THE USE O F STOCHASTIC MODELS IN A HYDRO-AGRICULTURAL

D E V E L O P M E N T P R O J E C T I N LEBANON"

by
J.H. V i s s e r

ABSTRACT

Stochastic modelling techniques were employed i n order to

p r o v i d e l o n g s e q u e n c e s o f m o n t h l y s t r e a m f l o w and w a t e r d e m a n d n e e d e d
for i r r i g a t i o n s c h e m e d e s i g n (the h i s t o r i c f l o w r e c o r d s b e i n g t o o
s h o r t t o s e r v e t h i s purpose).

The w a t e r r e q u i r e m e n t c a l c u l a t i o n s (with the Blaney C r i d d l e

f o r m u l a ) were b a s e d o n g e n e r a t e d s e r i e s o f m o n t h l y r a i n f a l l and
m o n t h l y m e a n temperature. T h e g e n e r a t i o n "in phase" o f t h e s e v a r i a b l e s
w i t h s t r e a m f l o w , e n s u r e d t h a t a dry y e a r w a s c h a r a c t e r i s e d by h i g h
d e m a n d and l o w f l o w , T h i s s t r a t e g y o f Ilin phase" g e n e r a t i o n w a s
preferred to t h e more usual treatment o f assuming a fixed annual
c y c l e o f d e m a n d s and a l l o w e d f o r a b e t t e r a s s e s s m e n t o f t h e d e s i g n
p a r a m e t e r s and a b e t t e r e c o n o m i c e v a l u a t i o n .

The Ilin p h a s e t 1 s e r i e s r e q u i r e d t h e g e n e r a t i o n o f

- m o n t h l y r a i n f a l l (simple m o d e l d u e t o a b s e n c e o f
persistance)

- monthly mean temperature (mixed model with auto regression

and l i n e a r r e g r e s s i o n on m o n t h l y r a i n f a l l )

- annual streamflow (with linear regression on annual

rainfall )

- mr oe ng tr he ls ys i so tn r e(for
amflow, related t o the annual flow, using auto
stations having a historic record o f more
t h a n 10 y e a r s )

- mr oe ng tr he ls ys i os tn r eoafm fa l od we ,s era es lo an ta el id z te do
the annual flow, using auto
v a r i a b l e and a n d l i n e a r
r e g r e s s i o n o n t h e s a m e v a r i a b l e o f a n o t h e r (better) station.
(for s t a t i o n s h a v i n g l e s s t h a n 1 0 y e a r s o f record).

U n d e r t a k e n j o i n t l y b y t h e G o v e r n m e n t o f L e b a n o n and t h e F o o d a n d
A g r i c u l t u r e O r g a n 2 s a t i o n o f t h e U n i t e d Nations.
436

RESUME

Des m é t h o d e s s t o c h a s t i q u e s ont é t é u t i l i s é e s pour pouvoir

disposer d e s é r i e s l o n g u e s dtapport et de demande mensuels, nscessai-
r e s pour I r é t u d e d f u n projet d ' h r i g a t i o n . CLes s é r i e s h i s t o r i q u e s
d'apport étant t r o p courtes pour être utilisées).

Les besoins en eau ( c a l c u l é s a v e c l a f o r m u l e de Blaney-

-Griddle) ont été b a s é s s u r des s é r i e s g é n é r é e s de la p l u i e et de l a
t e m p é r a t u r e mensuelles. La g é n é r a t i o n "en phase" de ces variables
a v e c les a p p o r t s a fait que l'année s è c h e s e caractérise par une
demande élevée et des apports faibles.

Cette s t r a t é g i e de g é n é r a t i o n Ifen phase" a été p r é f é r g e par

r a p p o r t à l a méthode p l u s h a b i t u e l l e d l u n cycle annuel f i x e de l a
demande, et a p e r m i s u n e m e i l l e u r e & v a l u a t i o n de l a g e s t i o n diun
projet d'irrigation a ï n s i quiune m e i l l e u r e évaluation économique.

Les s é r i e s "en phase" o n t n é c e s s i t é la génération d e

- pluie m e n s u e l l e (modèle simple vu l'absence d e p e r s i s t a n c e )

- tete m de
p é r a t u r e m e n s u e l l e (modèle m i x t e d e r é g r e s s i o n sérielle
r é g r e s s i o n l i n é a i r e par r a p p o r t ?i la p l u i e mensuelle

- apport a n n u e l (régression s i m p i e par r a p p o r t

annuelle)
à i a pluie

' - apport mensuel, par r a p p o r t à i f a p p o r t annuel, a v e c

r é g r e s s i o n s é r i e l l e (pour les s t a t i o n s ayant a u m o i n s 1 0
a n n é e s d'observations)

- rapport m e n s u e l , par
égression sérielle
rapport à i f a p p o r t annuel, a v e c
diune variable désaisonnalisée et de
r é g r e s s i o n l i n é a i r e p a r r a p p o r t à la même y a r i a b l e drune
autre (meilleure] s t a t i o n ( p o u r les s t a t i o n s ayant m o i n s de
10 annges d'observations).
 43 7

1 - INTRODUCTION

1.1. One of the objectives of the UNDP/FAO Project LEBANON 13,

concerning the hydro-agricultural development of North Lebanon, was to study
an irrigation scheme of about 7 O00 ha in the KOURA-ZGHARTA region. For the
water supply of this scheme a dam has to be constructed on the Aasfour river.
The reservoir inflow can be provided by the Aasfour discharges together with
part of the streamflow of an adjacent river.

To assess the performance of the design reservoir long series of

streamflow are needed to be routed through such a reservoir, Such long series
of historic records were missing and the presence of outliers (very wet and
several consecutive dry years), made it virtually impossible to establish
with any confidence a return period for these outliers.

1.2. The hydrologic information available in the Project area was

based mainly on the following data :

---- two streamflow series with

thirteen streamflow series
several rainfall series of
some temperature series of
14 years of record
with 3 to 5 years of record
about 30 years of record
about 15 years of record.

A good correlation exists between annual rainfall and streamflow

but low values are found of the correlation coefficient between monthly rain-
fall and streamflow. This can be explained by the fact that the response of
the catchments to rainfall has a delay factor of one to two months due to the
presence of snow and / or springs. It was thus impossible to apply the conven-
tional technique of extending the shorter streamflow records by correlating
them with the longer rainfall records, Unfortunately monthly streamflow data
are needed for reservoir analysis studies,

As the conventional techniques were unable to provide these data,

the use of hydrological modelling techniques became necessary.

1.3. Two main categories of mathematical models can be used in

principle for this kind of problems : Deterministic and stochastic. The first
category permits to extend the length of the historic streamflow series to
the same length as the (longer) historic rainfall record. The stochastic mo-
delling however permits to generate synthetic events of any length adequate
for certain design purposes.

A mixt use of a stochastic input into a deterministic model could

be useful in principle but unfortunately no such valuable generating models
for daily rainfall existed.
 43 8

1.4. The statistics, such as the e x p e c t e d f r e q u e n c y o f failure

of the design system, depend largely on the variation of the streamflow, i.e.
on the values of the variance of monthly and annual flow. The stochastic model-
ling techniques can improve the estimate of the variances of the shorter
records by using the i n f o p t i o n available in the longer series.

It was for all these reasons that the Lebanon 13 Project decided
to apply stochastic modelling techniques.

1.5. In order to apply generated series of streamflow in the

reservoir simulation studies it was necessary to generate also long series of
rainfall and temperature in order to calculate long series of water demand. To
avoid generating series of streamflow, rainfall and temperature that were un-
correlated, a method of "in phase" generation was adopted, This phasing will
ensure, for example, that during a dry year the values of streamflow and rain-
fall are both low, together with high temperature values resulting in high de-
mand for the year.

1.6. For the calculation of crop water needs the Blaney-Criddle

formula was used in view of the insufficiency of data for the Penman method.
However from the point of view of the methodology, t h e r e 2s no o b j e c t i o n t o
replace the Blaney-Criddle formula by that of Penman or another. The methodo-
logy for adjusting water resources and water demand as applied to North Lebanon
is explained schematically in the attached flow chart.

2 - CHOICE OF TYPE OF STOCHASTIC MODELS

2.1. The stochastic models, described hereafter, for the gene-

ration of long time series for different variables which are mutually in phase,
were proposed by Mr. J. Bernier, Chief of the Statistics Group at the Labora-
toire National d'Hydraulique, Chatou (France) and consultant to the North
Lebanon Project for stochastic hydrology. The models chosen were in response to
the availability of data and other local conditions as well as to the objecti-
ves of the study in particularly to provide input data for the simulation
studies, which explains the use of monthly values of the different variables.

2.2. The elements on which this choice was based were :

a - -
caracteristics of available data :
streamflow : the presence of 2 series of 14 years of record
 439

- temperature : the presence of some series of about 15 years of

record and a significant value for the corre-
lation between monthly rainfall and temperature
during spring and autumn.
- rainfall : several series of 30 years of record and a good
correlation between rainfall and streamflow on
annual basis but a bad one an monthly basis
(due to snowfall and / or karsticity)
- perennial flow of the rivers.

(the temperature and rainfall series together permit the use of

Blaney-Criddle's formula for the calculation of crop water needs)

b - The requirements imposed by the methodology used for adjusting water resour-
ces and water demand :

-- the series of streamflow and demand have to be in phase

the "being in phase" of streamflow and demand requires automati-
cally the'"in phase" generation of the series of rainfall, tem-
perature and streamflow.

2.3. The short streamflow series (3 to 5 years of record) can

only be used after "deseasonalisation" of the variable (3) resulting in series
of 36 to 60 months of record in which the different characteristics of the par-
ticular months have been neutralised,

2.4. The following transformations of the variables were necessa-

ry in order to be able to use the normal distribution :

-
- the logarithm of the discharges instead of the discharges
the square root of the rainfall instead of the rainfall (the
temperature is taken without any transformation).

2.5. The particular features described above led to the applica-

tion of the following generating models :

1) A simple model for monthly rainfall due to the absence of

persistance in the monthly rainfall series. This model uses
only mean and variance of the historic record together with
generated random numbers (eq.1).

2) A mixed model for monthly temperature using autoregression

plus regression on another variable (monthly rainfall) and
a random number generator (eq.2).
44 O

3) A simple regression model for annual streamflow using re-

gression o,n annual rainfall plus a random number generator
(for stations with at least 10 years of record) (eq.3).

4) An autoregression model for monthly streamflow using the

relation monthly / annual streamflow to give monthly stream-
flow plus a random number generator (eq.4).

5) A mixed autoregression model for monthly streamflow of a sta-

tionary ("deseasonalised") variable with regression on the
same variable of another station plus random number generator
(stations having less than 10 years of record) (eq.7).

3 - DESCRIPTION OF MODELS USED

3.1. Generation of monthly rainfall

Hypothesis : normal distribution of (Pt = monthly rainfall).

The statistical analysis of the historic series of monthly rain-

fall gives the values of mt = the mean of the fit of the month t (12 values)
and the s2t = the variance of the fit
of the month t (also 12 values),

Generating model :

fitemt -+ st . u (eq.1)

( 6 = o for Pt = 0)
where : u is the random normal deviate with zero mean and unit variance
-u = o, s2(u) = 1

The square root of the rainfall depends thus on mt, st and U.

3.2. Generation of mean monthly temperatures

Hypothesis : normal distribution of Tt (Tt = mean monthly tempe-

rature of month t).

The generator model uses a correlation of temperature with

 441

rainfall and is as follows :

Tt,i = mean monthly temperature of month t (t runs from 1 to 12 and i

represents the ith month after the start generation, i = 1,2 ....>
Ut = the mean of the mean monthly temperatures of month t, in period
of record (12 values)

K i = square root of the rainfall of month t,i

mt = the mean of the square root of the rainfall of month t, in period of

record (12 \talues)

U = random nonual deviate with 3= O and s2(u) = 1

vt = the variance of the residuals of T in the month t, in period of

record (12 values)

al,t
and d 2 , t = partial regression coefficients (for method of calculation
see standard works, e.g. Ven Te Chow "Applied Hydrology", page 8:6û)

For the first value of Tt,l, ill the following fonuula is used :

Tt-1 = M t-1 + G u'

where M and v are known and a single value for the random normal deviate u'
is sufficient to define the value of Ttml.

Once the coefficients of the generator model for mean monthly tem-
perature are known for each month, a sequence can be generated which will be in
phase, at the monthly level, with rainfall.

3.3. Generation of monthly streamflows for stations having at

least 10 years of record

The poor degree of correlation found in the study area between

monthly streamflows and monthly rainfall, due to the effect of snow and karsti-
city of the basin, necessitated first a correlation of the annual streamflows
(14 years of record) and the annual rainfall (the same station as used in para
3.1. above). This was achieved through the following equation based on the
hypothesis of a normal distribution of log Q :
 log ~a = M +e( q-
m> +K u íeq.3)

M = the mean of the logarithm of the annual flow Qa (14 values of Qa)

m = the mean of a (Pa = annual rainfall for the rainfall station 32 years)

u = random normal deviate

v = variance of the residuals of log Qa

In order to adjust to the long series of rainfall (32 years)

the following correction to M was prepared. This type of correction is only
necessary when records of the period of M are not typical of the period of m.

In fact, since the rainfall record (32 years) is longer than

the streamflow record (14 years) and the two are correlated, the estimate M
of the population mean of log Q and the estimate v of the variance of the
residuals can be improved, leadtng to revised estimates M;! and v2 as follows :

2
s2 (log Qa> = 2
s1 (log Qa) - r
2

a
(eq. 3b)

where the suffixes 1 and 2 refer to the 14 and 32 year records respectively.
2
The value of 62 (log Qa) so obtained can be used to derive an
improved estimate v from the relation :
2 2
v2 = (1 r s2 (log Q,) (eq.3~)

All the coefficients of equation 3 being known, a long series

of logarithms of annual flow Q, can be generated and will be in phase with
the annual rainfall.

The next step is the generation of the variate Zt,i = log Qt,i - log Qa (eq.4)

(eq.4)
 443

where :
-Zt = mean of (log Q
rt =
t,i - log Qa for period of recorii (12 values, t

correlation coefficient of Zt on ZtWl (12 values)

= 1,2,... 12)

s(Zt> = standard deviation of Zt (12 values)

vt = variance of the residuals -.

= n 2
s (2,) . (1 C.
2
rt)

u = random normal deviate with u

-= n-2
O, s2(u) = 1.

The application of equation (4) then gives a generated series

of Q t a i following.the transformation

Qt,i
-- Qa e ('t,i' íeq.4a)

for each month. The annual totals of these Q t a i should be equal to the values
of Qa generated with equation (3) but this will not usually be so, in
which case the following correction must be made for the generated Q
. tai
QL,i
E a'

Q'a
-
'Qt,i
(eq. 4b)

I
where :
Qt,i = revised value of Q t a i (monthly generated flows)
Qa = annual flows generated by eq. (3)
QIa
= annual sum of monthly flows Qt,i generated by eq. (4)
The result of these processes is a generated series of monthly
streamflows which are in phase annually with rainfall which is in turn in
phase monthly with the mean monthly temperatures.

3.4. Generation of monthly streamflows for stations having less

than 10 years of record

For short series equations (3) and (4) cannot be used because
a short series does not permit a sufficiently precise calculation, month by
month, of the mean, variance and correlation coefficient. To avoid this diffi-
culty the observed series of the monthly streamflow Qt i can be "deseasonalized"
to produce a new series Y giving, in the case of 4 y e a h of observations, 48
values of Y. To "deseasonalize" one of two transforms was used :
444

Y = log Q
t,i
- a cos (
-
12
(eq.5)

where (eq.5) : the origin was taken as the month of November (t = 1)

e =
the phase determined in such a way as to place the mean
maximum flow in the required month, which also determines
the month of minimum flow which will be six months later.

a = the amplitude of the variate (the logarithm of the monthly

flows) chosen in such a way as to give the best fit to
the observed maximum and minimum monthly values.

The choice of the transform depends on the shapc of the hydro-

graph. In the Lebanon project equation (5) was found to give less good re-
sults and therefore equation (6) was used.

Each of these equations gives a series of Y which combines all

months. From this series the mean (i)can be calculated and also the variance
(<)and-the serial correlation coefficient (ry) giving only -e value for
each of Y, sy and ry for the series of Yi. Only the value of Y will be used
directly in the equation which follows (7). SY and ry are used indirectly in
the coefficient Oc
i,t'
In order to use the information contained in a lcng series
the same "deseasonalization" is applied to this long series, giving a series
X. The next step is to calculate the correlation coefficients between Yi and
Xi, Yi+l and Xi, Yi+l and Xi-1 and even Y.+1 and Xi- .
The highest coefficient
is taken to indicate the interval (O, 1, 2
or 3 montEs) between the Y and X to
be used. The generator model for Yi for the chosen station, making use 01 the
monthly flows of another station, can then be written in the form :

where :

-YY = as given in eq. (5) or (6)

= mean of Y (during period of record)
Xi = deseasonalized generated variate 05 flow at another
station having a long record
 44 5

-X =I mean of X (during period of record)

u = r a n d a normal deviate
v = variance of the residuals of Y

When all the coefficients of equation (7) are known and data have
been generated for the "long" record station as described in para. 3.3, the model
can be used to generate the long series of Yi and so, using the transforms in
equations (5) and (6) above, long series of monthly streamflows Q
t,i'
The resulting values of Qt,+ are related to the monthly generated
flows of a station having a long record which are themselves in phase annually
with the annual rainfall. The annual rainfalls are the sum of the monthly rain-
falls which themselves are in phase with the mean monthly temperatures.

4 - CONCLUSIONS

4.1. The checks used on the generated time series are the sta-

(winter -
tistical moments (mean, variance and coefficient of variation) and periodicity
summer). It was found that the generated time series were in general
of good quality but that the value for the coefficient of variation (i.c. the
variance) was too high. This did not matter in the Lebanon case because the
reservoir simulation studies done with these time series kept us on the safe
side, but for the sake of completeness some kind of correction should be intro-
duced in future.

4.2. All the programmes have been written in Fortran IV for the
IñM 1130 computer in such a way that they can be used separately or in series.
In this latter case the calculation time is about one hour per run. To obtain
as output in one run the results of system simulation (reservoir size, irrigable
area, failures and effects on other water users) the following input data are
needed :
(a) a historic record of rainfall (monthly)

(b) a historic record of mean temperature (monthly)

(c) a "long" (12-14 years in North Lebanon) historic record of streamflow

(monthly

(d) a "short" (3 to 5 years) historic record of streamflow (monthly)

446

the duration of sunshine as a percentage p of the maximum possible

the monthly crop coefficient K (by crop)

the maximum usable soil moisture storage (by ero;)

the coefficient of growth of the plant (by crop) e

the phasing of irrigation development of the whole area, and the
subdivision of the area by crop (in %)

the rate of changedver from the present olive groves to the new crops

coefficient of irrigation efficiency

geometric characteristics of the reservoir

P (e) and (f) are necessary for the application of the Blaneydriddle formula.
 Il m

I II I

P 0 0
I I
m

1. II ..... i ace -. 3.7.4 or t a

I FLOW CHART FOR DATA GENERATION A N D SYSTEM ANALYSIS

North Ltibonon irriqaion scheme

1-1 i o'indirti. Bencator
Jaouory 1972 lidopied from Prqe~i Droir~ng AE-2513
 RELATIVE IMPORTANCE OF DECISION VARIABLES
IN FLOOD FREQUENCY ANALYSIS

Wallis, J.R.
IBM, Thomas J . Katson Research Center , Worktown Heights , N .Y,, USA

Matalas, N. C.
U.S. Geological Survey, Washington, D. C., USA

ABSTRACT

Monte Carlo simulations were used to assess flood and overde-

sign losses that result from differing choices o f assumed frequen-
cy distribution, plotting position, criterion o f best fit and
length of record. Probabilities o f best fit for an assumed world
distribution, given a real world distribution, are given.

RESUMEN

El método de simulación de Monte Carlo se utiliza para eva-

luar los daños producidos por máximas crecidas en funci’on de las
leyes de distribución de frecuencias, de las estaciones utilizadas
y de la calidad y extensión de las series hidrológicas. De la m u e s
tra puede obtenerse el valor minimo teórico de los daños estimados.
450

Introduction

In many cases, the design of multipurpose water resource

systems includes flood control as one of the purposes. While the
design process may specify the sizing of flood control structures
as a function of the T-year flood, the design process must cope
with the uncertainty as to the magnitude of the T-year flood.
Given that floods are a random phenomenon, the magnitude of the
T-year flood depends upon the underlying probability distribution
of flood events and the values of the distribution's parameters.

Among the objectives of flood frequency analysis is that of

determining the magnitude of the T-year flood, referred to as the
design flood. While the underlying distribution of floods is
unknown, an estimate of the design flood can be provided. A dis-
tribution may be assumed or chosen i n accordance with some criter-
ion of best fit to observed flood sequences. Given an observed
flood sequence of length n, and an assumed o r chosen distribution
estimate of the distribution's parameter values, the design flood
can be derived. These estimates are subject t o sampling errors,
the magnitudes of which depend upon n, and on uncertainties that
are only partially a function of n. To reduce sampling errors,
longer flood sequences are needed. To acquire longer sequences
through direct observation might necessitate the delays in the
design of the water resource system. Delays would be economically
feasible if over the period of data collection no benefits were
foregone. In those cases where benefits would be foregone effec-
tively longer sequences might b e obtained through regional analyses.

However, even with a very large but finite flood sequence,

uncertainty would still exist i n the estimate of the T-year flood.
The Uncertainty arises because the assumed or chosen distribution
used to estimate the T-year flood does not necessarily have t o
be the correct real world distribution, and in fact if a criter-
ion of best fit is used, the chosen distribution might vary given
another flood sequence of equal length.

By using an estimate of the T-year flood as the design flood,

either of two types of losses is likely to b e incurred. The first
type refers to overdesign costs of flood control structures which
would be incurred if the estimated T-year flood exceeded the true
value of the design flood. The second type refers to the down-
stream damages which would be incurred from underdesign if the
true value of the design flood exceeded the estimate of the T-year
flood. In the design process, what is of concern is not how well
.a particular distribution fits an observed flood sequence par se,
b u t to what extent the t w o types of design losses are affected by
the choice of a particular distribution. To the designer, the
 451

criterion of best fit refers to choosing a distribution to minimize

d e s i g n losses.

To gain some insight as to the magnitudes and sensitivities

of t h e d e s i g n l o s s e s t o u n c e r t a i n t f e s i n t h e c h o i c e o f a f l o o d
f r e q u e n c y d i s t r i b u t i o n a n d e s t i m a t e s o f t h e distribution's p a r a m -
eter values, several computer-based experiments, employing Monte
Carlo techniques, are currently being performed. I n this paper
the nature o f these experiments i s briefly discussed, and some
e x p e r i m e n t a l r e s u l t s as t o t h e p r o b a b i l i t i e s of f i t t i n g of
observed flood sequences with particular distributions are
presented.

Monte Carlo Experiments

T w o s e t s o f d i s t r i b u t i o n f u n c t i o n s a r e considered. The first

set, referred to as the real world set, consists o f several dis-
t r i b u t i o n s , any o n e o f w h i c h m a y b e t h e u n d e r l y i n g d i s t r i b u t i o n
o f floods. From the real world set, a distribution function is
c h o s e n w h e r e t h e distribution's p a r a m e t e r s v a l u e s a r e r e l a t e d t o
t h e m e a n , u, t h e s t a n d a r d d e v i a t i o n , o, a n d t h e c o e f f i c i e n t o f
s k e w n e s s , y. F o r this distribution, 18,000 flood sequences o f
length n are generated.

T h e s e c o n d s e t o f d i s t r i b u t i o n functions, r e f e r r e d t o a s t h e
i m a g i n e d or a s s u m e d s e t , c o n t a i n s s e v e r a l d i s t r i b u t i o n s , a n y o n e
of which may be fitted to observed flood sequences. Each element
of the imagined set is fitted to each of the generated sequences,
and o n the basis of various methods for defining plotting positions
and measuring goodness of fit, the particular distribution of best
f i t i s d e t e r m i n e d f o r e a c h g e n e r a t e d sequence. F r o m e a c h of t h e s e
d i s t r i b u t i o n s , t h e f l o o d h a v i n g a n e x c e e d a n c e p r o b a b i l i t y o f 1/T
i s determined. These floods are estimates o f the real world flood
o f e x c e e d a n c e p r o b a b i l i t y 1/T.

Initially, overdesign and underdesign linear loss functions

in terms o f the difference between the real world T-year flood
a n d i t s e s t i m a t e a r e assumed. Nonlinear loss functions will b e
c o n s i d e r e d i n s u b s e q u e n t experiments. G i v e n t h e 18,000 v a l u e s of
t h e d i f f e r e n c e s b e t w e e n t h e r e a l w o r l d T-year f l o o d a n d i t s
estimates, the probabilities of incurring overdesign and under-
design losses and the expected values o f the losses are estimated.
S i m i l a r l y , t h e s e v a l u e s a r e e s t i m a t e d f o r e a c h o f t h e o t h e r dis-
t r i b u t i o n s b e l o n g i n g t o t h e r e a l w o r l d set.

Three flood control design objectives are considered:

1 ) m i n i m i z i n g t h e e x p e c t e d o v e r d e s i g n l o s s e s , 2) m i n i m i z i n g t h e
e x p e c t e d u n d e r d e s i g n losses, a n d 3 ) m i n i m i z i n g a w e i g h t e d s u m
o f t h e e x p e c t e d o v e r d e s i g n and e x p e c t e d u n d e r d e s i g n losses. For
452

t h e t h i r d o b j e c t i v e , t h e w e i g h t s , s a y CI a n d ß, w h e r e a > O, ß > O,
a n d o: + ß = 1, a r e varied. Among the methods for defìning plotting
p o s i t i o n s a n d m e a s u r i n g g o o d n e s s o f fit, t h e p a r t i c u l a r m e t h o d a n d
m e a s u r e b y w h i c h t h e v a r i o u s d e s i g n o b j e c t i v e s a r e m e t w e r e deter-
mined. The sensitivities of the design losses t o less than optimal
choices of the plotting position method and measure of goodness
o f f i t w e r e assessed.

T h e e x p e r i m e n t s w e r e c a r r ì è d o u t f o r e v e r y f e a s i b l e p o i n t in
the following experimental hyperspace:

p = 2600
u = $00
y = O, 1/4, 1/2;*, 1, 1 . 1 4 K 2
n = 10, 30, 50, 70, 9 0

The results of these experiments are conditional o n the distribu-

t i o n s b e l o n g i n g t o t h e r e a l w o r l d set. Subsequent experiments
will consider prior information on the real world distribution
function and regional estimates of the distribution's parameter
values.

P r o b a b i l i t i e s o f B e s t Fit:

Some preliminary results of these experiments are presented

n a m e l y , t h e p r o b a b i l i t i e s o f b e s t fit. Both the real world and
imagined world sets consisted of three elements
distribution, the log-normal dìstribution, and the Type I extremai
-- the normal

( G u m b e l ) distribution. F o r each distribution belonging- to the real

woerfd s a t ) i8000 s e q u e n c e s were g e n e r a t e d f o e~a c h f e a s i b l e
point in the experimental hyperspace. Floods for each sequence
o f l e n g t h n w e r e r a n k e d i n o r d e r o f m a g n i t u d e f r o m t h e largest,
h a v i n g r a n k m = 1, t o s m a l l e s t , h a v i n g r a n k m = n. The flood o f
r a n k m w a s a s s i g n e d an e x c e e d a n c e p r o b a b i l i t y , P[m,nl , b y b o t h
the "Weibul method," defined as

P W [m,nl = m/(n+l) (1)

a n d by t h e " H a z e n m e t h o d , " d e f i n e d a s

P,[m,nl = (2m-i) / 2 n (2)

( S e e Chow: 1964).

F o r a given element of the real world set and a given element

o f t h e i m a g i n e d w o r l d s e t , t w o s e t s o f d i f f e r e n c e s of f l o o d magni-
t u d e s w e r e f o r m e d f o r e a c h g e n e r a t e d sequence. T h e f i r s t s e t con-
sisted of the differences between the observed, that is generated,
 453

floods and the corresponding imagined world floods having exceed-

a n c e p r o b a b i l i t i e s d e f i n e d by P w ( m , n J . Similarly, the second set
of differences was based on exceedance probabilities defined by
P H (m,n).

Two measures of goodness of fit were considered

squares o f the differences in flood magnitudes and the s u m o € the
-- the sum of

absolute differences in flood magnitudes. F o r each sequence based

o n an e l e m e n t o f t h e r e a l w o r l d s e t a n d r e l a t i v e t o e a c h m e t h o d o f
assigning exceedance probabilities, the element o f the imagined
s e t w h i c h p r o v i d e d t h e b e s t f i t t o t h e s e q u e n c e of f l o o d s w a s
determined, where best fit was defined by each o f two criteria --
m i n i m u m sum o f s q u a r e s o f d i f f e r e n c e s i n f l o o d m a g n i t u d e s a n d
m i n i m u m s u m o f a b s o l u t e d i f f e r e n c e s i n f l o o d magnitudes.

The probability of the event that a sequence, having a par-

t i c u l a r e l e m e n t o f t h e r e a l w o r l d s e t as i t s u n d e r l y i n g d i s t r i b u -
t i o n , i s b e s t f i t t e d by a p a r t i c u l a r e l e m e n t o f i m a g i n e d w o r l d
set, relative to a particular method o f assigning exceedance
p r o b a b i l i t i e s a n d a p a r t i c u l a r c r i t e r i o n of b e s t f i t , w a s e s t i -
m a t e d b y N/18,000, w h e r e N d e n o t e s t h e n u m b e r o f t i m e s t h e e v e n t
o c c u r r e d a n d 18,000, t h e t o t a l n u m b e r o f t i m e s t h e e v e n t c o u l d
h a v e occurred. T h e p r o b a b i l i t i e s f o r e a c h o f t h e 36 p o i n t s i n
the event space relative to each feasible point in the experi-
m e n t a l h y p e r s p a c e w e r e determined.

Remarks

F o r t h e e x p e r i m e n t a l h y p e r s p a c e , t h e e s t i m a t e s of t h e p r o b a b i l -
i t i e s of b e s t f i t m u l t i p l i e d b y 1000 o v e r t h e e v e n t s p a c e a r e g i v e n
i n Lahles 1 t h r o u g h 9, w h e r e M S S d e n o t e s m i n i m u m s u m of s q u a r e s ,
M S A D d e n o t e s m i n i m u m s u m of a b s o l u t e d i f f e r e n c e s , N d e n o t e s t h e
n o r m a l d i s t r i b u t i o n , G d e n o t e s t h e GuInbel d i s t r i b u t i o n , a n d L d e n o t e s
t h e l o g - n o r m a l distribution. These estimates were based o n the follow-
ing additional experimental operating rule --
of the coefficient o f skewness, y, f o r a generated sequence w a s equal
if the computed value

t o o r l e s s t h a n 0.007, t h e n t h e s e q u e n c e w a s c o n s i d e r e d t o h a v e b e e n
d r a w n f r o m a n o r m a l distribution. While these probabilities give
s o m e i n d i c a t i o n as t o t h e p o w e r f o r i d e n t i f y i n g t h e r e a l w o r l d f l o o d
distribution from an observed sequence, they d o not give any indica-
tion of the optimum strategy to use f o r choosing a design distribu-
tion. I n t e r p r e t a t i o n of t h e s e e x p e r i m e n t a l r e s u l t s i n t e r m s o f
overdesign-underdesign strategies will b e the subject o f subsequent
papers.

Reference

C h o w , V e n T. (1964). Handbook o f Applied Hydrology, McGraw-Hill.

 454

Table 1. -- Real World is normal

Weibull Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G
~~ ~~

10 527 468 5 602 314 84 523 289 188 556 234 211
30 531 432 36 608 341 51 530 432 38 579 364 58
50 528 455 17 614 360 26 526 465 9 489 387 24
70 526 468 6 613 377 io 526 473 2 595 397 8
90 532 466 2 620 374 5 529 470 1 603 394 4

Table 2. -- Real World is Gumbel with skew = 1.14

Weibuil Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G
10 216 766 i8 287 523 190 212 416 371 245 358 398
30 40 646 314 78 624 299 40 658 303 57 590 352
50 7 612 380 27 642 331 8 717 275 19 629 352
70 2 596 402 10 625 364 2 756 242 7 641 352
90 0 591 409 6 622 372 O 767 233 3 635 361

Table 3 . -- Real World is log-normal with skew = i/4

Weibull Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G

10 441 551 8 521 370 109 436 323 241 474 261 265
30 342 573 85 433 469 99 341 570 89 395 481 124
50 279 659 62 382 550 68 279 683 38 347 583 70
70 234 727 38 339 621 41 232 750 18 313 649 $8
90 205 771 24 313 661 26 204 788 8 290 688 23
 455

Table 4. -- Real World is log-normal with skew = 112

Weibull Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G
10 359 630 11 442 421 i37 356 352 292 393 290 317
30 197 643 161 281 554 165 196 641 163 240 556 204
50 116 721 i63 202 655 143 116 780 i04 169 678 153
70 71 789 i40 147 735 117 70 859 71 124 758 118
90 46 837 i17 118 789 93 46 901 53 97 812 91

Table 5. -- Real World is log-normal with skew = v1/2

Weibull Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G

10 302 684 i4 382 461 157 298 375 327 333 315 353
30 i14 662 224 i86 597 218 i14 662 224 i50 582 268
50 48 688 264 106 672 222 48 775 177 83 677 241
70 22 712 266 64 719 217 21 832 i47 49 731 221
90 11 737 253 43 756 201 11 865 125 32 765 203

Table 6. -- Real World is log-normal with skew = 1

Weibull Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G

10 235 750 i5 309 508 183 231 406 362 260 350 390
30 50 658 292 96 631 273 51 663 286 72 594 334
50 13 625 363 39 649 312 13 736 251 27 640 333
70 5 607 388 20 651 330 4 767 229 i4 653 334
90 1 592 406 io 657 333 1 787 212 6 667 327
 Table 7 .--Real World is log-normal with skew = 1.14
Weibull Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G

10 209 775 17 281 528 192 205 418 376 234 364 403
30 34 651 314 70 638 292 34 660 306 52 594 354
50 7 606 386 24 638 338 7 721 272 i7 633 351
70 2 586 412 9 635 356 2 755 243 6 641 353
90 O 582 418 4 632 364 0 777 223 3 647 351

Table 8. -- Real World is log-normal with skew =

Weibull Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G

10 169 814 16 231 565 204 166 438 396 193 387 421
30 17 655 328 38 658 304 16 667 317 27 615 358
50 2 620 378 11 657 332 2 727 271 6 648 346
70 0 618 381 3 658 339 O 772 228 2 671 327
90 O 629 371 1 664 335 o 804 196 1 685 314

Table 9. -- R e d World is log-normal with skew = 2

Weibull Hazen
MSS MSAD MSS MSAD
n N L G N L G N L G N L G

10 113 865 21 163 614 223 112 473 415 135 427 438
30 4 696 300 11 710 279 4 710 286 8 679 313
50 0 723 277 1 751 247 O 798 202 1 754 245
70 O 770 230 0 795 205 0 860 i40 O 807 193
90 o 810 190 O 826 174 O 903 97 0 842 158
 SHOT NOISE MODELS FOR SYNTHETIC GENERATION
OF MULTISITE DAILY STREAMFLOK DATA

bY
G , WEISS

Department o f Mathematics Imperi'al College

University o f London

AB STRA CT

While multisite models for generating synthetic streamflow d a -

ta on a monthly basis have been successfully used, adequate daily
models are lacking, In particular, existing models based o n Gau-
ssian processes are unsuitable in reproducing the recessions which
are clearly observable in daily data. The models currently being d e
veloped a t Imperial College, London, under contract for the Water
Resources Board, England, are based on "Shot Noise" or filtered P o i
sson processes. These processes consist o f a series o f Poisson
events, each o f which generates a pulse of random height and some
fixed recession shape, In the simplest o f these models the pulses
consist of jumps which are exponentially distributed in magnitude,
and which decay exponentially with a fixed decay rate. This is a
continuous time first order autoregressive (Markovian) process, and
its instantaneous values have a Gamma distribution. This model can
be fitted to streamflow data so as to preserve the observed means,
standard deviations, serial and cross correlations o f daily data.
Using a more complicated model consisting o f two shot noise proce-
sses, monthly statistics can be preserved in addition to the daily
statistics. This model gave satisfactory results with data from S O -
me East Anglia sites,

--
RESUME

Alors qu'on a réussi a

construire des modgles capables de €OU:
nir artific2ellement des sérri'es ae dé:its moyens mensuels en plu-
sieurs sites, :n manque encore de modeles satisfaisants pour les.va
leurs journalieres. I1 faut souligner en particulier que les modeles
stochastiques actuels, basés sur des processus gaussiens, ne sont
pas capables de reproduire les décrues qui sont faciles 2 mettre en
évidence dans les relev6s.journaliers. Les modèles qu'on est en
train de mettre au point a l'Imperia1 College (Londres), pour le
compte du Water Resources Board (Angleterre) sont basés s u r le
"shot noise" o u processys de Poisson filtr'es. Ces processus se com-
posent d'une série d'évenements obéissant a une loi de Poisson, dont
chacun produit une impulsion d'amplitude aléatoire dêcroissant s u i -
vant une forme déterminée. Dans les plus simples de ces modèles,
les impulsions consistente en des sauts dont les amplitudes aléatoA
res sont distribuées de facon exponentielle et sont affectées, une
fois produites, d'une décroissa?ce exponentielle dans le temps, l a
constante de temps étant fixée a l'avance. Ceci constitue un p r o c e 5
sus (Markovien) autorégressif de premier ordre continu dans le temps,
dont les valeurs insta2tannées sont distribuées suivant une loi
Gamma. Le modèle peut etre ajusté aux données disponibles concernant
l'écoulement de facon a respecter les moyennes, les écarts-types et
les corrélations croisées des débits journaliers. En utilisant un
modèle plus compliqué, formé de deux processus ltshotn o i ~ e ' ~i,l est
possible2 en plus des caractéristiques statistiques des valeurs
journalieres , de conserver celles des valeurs mensuelles. Ce-modèle
a donné des résultats satisfaisants pour un ensemble de rivieres
dans l'Est de l'Angleterre.
Introduction

The present work is aimed at supplying multisite daily synthetic stream-

flow data for the British Water Resources Board. The Water Resources Board
is currently developing regional simulation programs to help in the planning
and operation of water supply, and sensitivity tests have shown that at the
level of detail used in these programs, monthly data are insufficient and
daily data are indeed required.

Generation of synthetic monthly data in Hydrology was apparently first

attempted in the Harvard Water Program [I], and has been developed as a useful
tool since ([2], [3J,[ 41). Basically, in the methods previously used, the
data or a transformed series obtained from the data are assumed to follow a
multivariate Gaussian 1st order autoregressive process. Some sophisticated
transformations and some higher order regression models have also been used
([5], [6])*
Autoregressive models based on the Gaussian distribution have also been
applied to daily data [7J, [SJ. Such models, however, are inadequate in the
following sense : one of the prominent features of daily streamflows is the
presence of peaks and recessions clearly observed in the data. Yet no model
based on the Gaussian distribution can reproduce these recessions, no matter
what transformation or what order of autoregression is used.

In this paper a class of models which reproduce recessions is introduced.

A simple model from this class, of the same degree of complexity as the
Gaussian 1st order autoregressive process is developed, and its implementation
for data generation described. The theoretical aspects of reproducing recess-
ions are discussed. Finally, the application of the model to some British
streamflows is illustrated.

Filtered Poisson Processes

Let N(t) be a P&isson process, let Y be a random variable, and let
w(t,y) be some function. Let the sequence ...,7-1, TO, 71, O. be the times
of events of the process N(t), and let ...,
y-1, yo, YI, ...
be mutually
independent random values having the same distribution as Y, and all of which
are independefit of N(t). A filtered Poipson process X(t) is defined by :

For further details of these processes refer to CS].

A physical interpretation in hydrological terms can be given to the
filtered Poisson process. The events at random times T ~ given
, by the Poisson
process can be thought of as beginnings of rain storms. The random value ym
associated with T; could correspond to the amount of water in the rainstorm.
Finally, T and y,,, will produce a response in the flow given by w(t-.cm,ym)
and thus w?t,y) represents the system transfer function.
 459

The foregoing interpretation is only approximate, since rainstorms are

not independent as the Ym'S are required to be in the definition. One can
however imagine that an independent series of climatic events exists initially,
and that w(t,y) is the transfer function which transforms such climatic events
into streamflow.

The Shot Noise Process

A particular linear filtered Poisson process was adopted for modelling

streamflow, which is referred to as the shot noise process. In a filtered
Poisson process, let N(t) be a Poisson process with rate 3 .
Let Y be a
random variable with an exponential distribution and with mean 8, and let
w(t,y) = ye-bt, for t > O. The shot noise process is defined as :

jump height, and b -

- -
The process has three parameters :9 the event rate, 8 the average
the decay rate. !he process has the following properties
(found by applying theorems in [9J) :

X(t) has a Gamma (Pearson type 2) distribution with parameters (Q, q/b),
and thus X(t) is nonnegative and positively skewed, and has probability density
function :

The moments of X(t) are given by :

(s > o)
From eq. (2) the process at time t+s, X t+s), can be written as :

The two terms in (5) are independent. The first represents the effect
of events previous to t, and is equal to e-bs X(t). The second includes the
events in ( t , C+s) and is the innovation term. Denoting the innovation by
~ ~ ( t + sone
) has
460

Thus, the shot noise process is in fact a 1st order autoregressive process
in continuous time. However, it differs from the Gaussian 1st order autoregre-
ssive process in that ES(t+s), instead of being Gaussian, has a skewed distri-
bution with a positive probability of being exactly zero. This arises when
no events occur in (t, t+s).

Some Aspects of Modelling Recessions

When modelling a stochastic process in hydrology one often makes the in-
exact but not unrealistic assumption of a linear system. This amounts to
assuming that the process X(t) is of the form :

where dY(t) is a completely uncorrelated and independent process, which

describes all the randomness in X(t), and h(t) is the system transfer function.

A usual choice for dY(t) is a Gaussian white noise. The characteristic

feature of the shot noise process is that dY(t) is chosen as zero almost every-
where, except for a series of spikes. These spikes occur at random time
instants determined by a Poisson process and each spike has some random mass.
(Note that in eqns. (1) and (2) summation over the spikes replaces the integral
in (7)).

The choice of the transfer function h(t) determines the autocorrelation

of X(t). In particular h(t) = e-bt gives a 1st order autoregressive process,
and corresponds to a single linear reservoir. In addition h(t) must determine
the shape of recessions in X(t).

However, if dY(t) is chosen as Gaussian white noise, no recessions will

appear in X(t). The absence of recessions may be explained intuitively by the
fact that Gaussian white noise is changing by minute quantities very quickly ;
hence the recession shape of h(t) appears in minute form and is immediately
swamped by the next change in dY (t) .Thus linear Guassian processes cannot
reproduce recessions, irrespective of the form of h(t), and the same will be
true even if a non-linear transformation is used pointwise on X(t).

The ability of the shot noise process to reproduce recessions prompted

its use in the model1ing;of daily streamflows.

Averaged Sampling of the Shot Noise Process

Natural streamflow and the stochastic shot noise process are continuous
time processes. Recorded daily streamflows and the synthetic data to be
produced are on the other hand discrete time processes. The usual approach
in the modelling of monthly data is to consider the data as a discrete sample
of the process, i.e. the values of the continuous process at discrete time
points. However, discrete sampling is inaccurate, since the data are actually
obtained by averaging the flows over the period between the discrete sampling
 461

time points. The difference between the two approaches, of discrete sampling
or of average sampling, is negligible for serial correlations of up to 0.5 P=
which are typical for monthly data. It is however substantial for
which is typical for daily data.
0-8 e=
It is assumed here that streamflow follows a continuous time shot noise
process X(t), and that the observed data (and the generated data), are averages
of this process over a period of T = 1 day. The data X,, X2, ...
are thus
defined as :

The moments of X. are slightly different from those of X(t), and are given
by : J

E(X.1 = SQ
J
Var (x.)
3
-Q2
= 9
b
2 [b-(l-e-b))
(9)
b2

(s 1)

In addition, the averaging changes the shape of the recessions. Whereas

in the process X(t) recessions start from a vertical rising limb, for the
averaged values X the transfer function is of the shape,
j
f;
1
(1-e
-bt)
<
(o t < 1)

that is the rise is gradual over O \< t < 1.

Fitting the Shot Noise Model to Daily Streamflow

In fitting the shot noise model an approach similar to that employed by

Matalas [3J is used. Values of 9 , 8, b are calculated which preserve the
values of 1.1, o2 and p(l) obsezved in the data. Thus the sample mean, variance
and first serial correlation, p, 82 and g(1) are calculated from the hietor-
ica: data. These are substituted in equations (9), which are solved for 4,
8, b. The estimated decay rate 6 is solved for from the thigd equation by
numerical methods, and the other two equations yield 8 and 9 .

An alternative approach would be to estimate b directly from observed

recessions or from the unit hydrograph of the basin, and to estimate 3 by
46 2

observing times of peaks in the data. This latter approach was attempted for
the British streamflow data at our disposal. However, preservation of observed
p, o2 and (1) using this method did not ensue. A similar approach may
however prove useful for different data, for instance, streams in semi-arid
regions, where data may consist of short records of frequent observations.

Synthetic Generation of Shot Noise Data

Let 3, 8, b be the parameters estimated from historical data. The algo-

rithm for generating synthetic data is as follows :

Denoting by Xt, t = 1, 2, ...,the averaged shot noise to be generated,

and by X(t), t = O, 1, 2, ...,
the values of the continuous process, one
obtains from (5) and (IO) :

where the first term in eqns. 11 and 12 is the contribution from events
preceding t , and the second is the contribution of events in (t, t+l).

Starting with an initial value for X(O), XI and X(1) are generated.
X(1) is then used to generate X and X(2) and so on. Assuming XI, ...,Xt
and X(t) have been generated, tge following steps lead to Xt+l, X(t+l).

1) The first terms of (11,121 are calculated, from X(t). X(t) can then be
discarded.
2) Time of last event preceding (t,t+l) need not be remembered. Event times
are initiated by putting m = O, .cm = O.
3) The next event .cm+l is generated as zm+l = T ~ + I ,where I is a random
number generated from an exponential distribution with mean (1/3 1.
4) If T ~ >+ 1 ~all events in (t,t+l) have been exhausted and so generation
of Xt+l and X(t+l) is complete.
5) For T ~ + I< 1, ya+-, is generated as a random number from an exponential
distribution with mean 8.
6) The contribution of ym+ 1 to (11, 12) is calculated as :
1 (l-e-b(l-Tm+l 1 ) and
; e-b(l-.cm+l), and added to the values

of Xt+l and X(t+l) respectively.

7) m is set to m+l and steps 3 to 7 are repeated.
Thus the generation requires only random numbers from exponential distri-
butions, which are easy to create.
 463

Multisite Shot Noise Processes

The shot noise process already defined can be easily extended to a

multisite process. Let Xl(t), ...,
XM(t), be continuous shot noise processes
at M sites, with parameters Sk,Qk, bk, k = 1, ...,
M.
A multisite process incorporating all of the parameters will be defined
by assuming that some of the events occur simultaneously at several sites,
and give rise to correlated jumps yk at these sites.

For two of the sites, k and 1, let the events which occur simultaneously
be at rate Jkl(with 3 k 1 < 3 k, 3 kl < 3 i), and let the jumps associated
with a simultaneous event, y , yl, have a correlation coefficient ckl. Then
the correlation between \(ty and X,(t) is :

c + I
kl
&. . 2

In this expression the first term shows the effect of the different decay
rates on the cross correlation, (i.e. the effect of the two different recession
shapes), and the other two terms arise from the correlation between the two
series of events and jumps.

By (13) Ski and Ckl can be chosen for each pair k,l, to preserve the
observed cross correlation in the multisite data.

The Double Shot Noise Process

Some difficulties arose in fitting the shot noise process to average

daily flows from some English streams. While the model did preserve the mean,
standard deviation and lag one serial correlation coefficient of the daily
data, when the synthetic data was averaged over months, the synthetic monthly
data had much smaller standard deviations and lag one serial correlation
coefficients than those observed in the historic data. Moreover, in the syn-
thetic data the recessions decayed too fast towards zero, and too many rises
and recessions were generated.

Inspection of the historic daily ctreamflow hydrographs showed that the

streams modelled have a pronounced base flow component which is not reproduced
by the shot noise process.

Therefore a more sophisticated model was proposed, which assumes X(t)

to be the sum of two independent shot noise processes, Xq(t) with parameters
+ A ,QI, bl and X2(t) with parameters $ 2 , 82 and b2 (cf equation 2). In
these two process $1, QI, bl are assumed to be larger than $2, 82, b2, so
that Xl(t) has more recessions, higher jumps and a faster decay rate than
X2(t). In physical terms, Xl(t) may be thought of as representing a surface
runoff mechanism and X2(t) as representing a baseflow mechanism.
 464

In fitting the model, the six parameters can be calculated so as to

preserve the observed mean, standard deviation and lag one serial correlation
of the observed daily flows, and the standard deviation and lag one serial
correlation of the observed averaged monthly flows.

An Application of the Double Shot Noise Model

Data from the river Nene in East Anglia and some of its tributaries, and
of one tributary of the neighbouring Great Ouse was used to generate synthetic
data. The Nene flows through East Anglia, and discharges into the Wash. It
has a drainage area of 1630 km2, it receives an average annual rainfall of
623 mm, and has an annual runoff of 157 mm.
The historic data consists of 11 years of average daily streamflows
concurrent at 8 sites. !Che double shot noise model was fitted to the data so
as to preserve at each site the overall mean, the standard deviation of the
daily and of the monthly series, and the lag one serial correlation coefficient
of the daily and the monthly series, and so as to preserve the daily cross
correlations between the sites. Seasonality was accounted for through estima-
ting the parameters separately for each calendar month.

Twelve series of synthetic data, each of them equal in length to the

historic record, were generated. Table 1 summarises some of the results from
the historic and generated data. The table includes quantities calculated for
the River Nene at Orton, close to the outflow point, and for the calendar month
of January. Flows are listed in m3/s.

The table gives a comparison between properties of the historic data,

properties of the theoretical model calculated analytically, and properties
of the synthetic data. Column 1 refers to the historic data. Columns 2-4
refer to the theoretical model. Column 2 contains quantities calculated for
the double shot noise model, and the decomposition of these quantities into
the process modelling the surface runoff mechanism (fast process) and the
process modelling the baseflow mechanism (slow process) are given in columns
4 and 3 respectively. Columns 5-7 refer to the synthetic data. Column 5
contains values which are averages of all the twelve synthetic series while
columns 6 and 7 list the lowest and highest values obtained for each quantity
out of the twelve series.

The different rows of the table list the values of several quantities
of interest. The quantities which are starred, are those used in fitting
the model. For some of those quantities the model preserves the historical
value exactly, while others were only preserved approximately, due to numerical
difficulties. e12 is the cross correlation between Nene at Orton and Great
Ouse at Thornborough Mill.

The rest of the quantities listed were not used in fitting the model,
and success in preserving them can serve as a measure of the adequacy of the
model. Of these quantities which were not fitted, the lag two and lag three
daily serial correlation coefficients are extremely well preserved. On the
other hand the skewness of the data was overestimated by the model. It is
very encouraging that the model seems to yield reasonable values of high and
low flows.
 465
h
2
3
e
(d
r,
..
c
O
ci
k
O
c>
(d
al
!2
al
z
O
ci
W
al
o
O
k
pi
al
4- *rl
M Ca
Y
u3 O
?
f
?
O ci
O
-3
al
rl
P
r
2
a
pr\
w
“o.
Co
?
r
P;
-7
9
O
O
eO
.rl
T
I
z.
ci
rl
pi
Co pi
“o. u3
c;.
nl
4
o, O
I
r
al
d
P
(d
H
 466

Conclusions

The shot noise model has been developed as a physically more realistic
model of daily streamflow data than has heretofore been proposed, and in
particular models recession effects which are a prominent feature of daily
streamflow data.

In its basic form, the shot noise process in its conception, statistical
properties and with the associated method of fitting and method of data gener-
ation is as simple and easy to handle as the Gaussian 1st order autoregressive
model.

The use of the double shot noise model for some English streans gave
satisfactory results, and illustrates the adaptibility of this class of models.

It is felt that these mdels, with their emphasis on events and recessions,
could with further research provide a link between deterministic and stochastic
hydrology. Thus studies by deterministic methods of the instantaneous unit
hydrograph and of the mechanism of base flow etc. could provide some of the
parameters needed for a stochastic model based on shot noise processes.

Acknowledgements

This work is financed by the Water Resources Board of England and Wales,
and is being carried out at Imperial College of Science and Technology in
London under the supervision of Professor D.R. Cox of the Department of
Mathematics and Mr. T. OIDonnell and Mr. P.E. O'Connel1 of the Hydrology
Section, Department of Civil Engineering, to whom I am deeply indebted for
ideas and help.
 467

References

1. Thomas, H.A. and Fiering, M.B. (19621, Mathematical synthesis of stream-

flow sequences for the analysis of river basins by simulation, Ch. 12 in
'Design of Water Resources Systems', by Maass, A., et al, London, Macmillan,
pp. 459-493.
2. Fiering, M.B. (19641, Multivariate techniques for synthetic hydrology,
Proc. Am. Soc. Civ. Engrs., J. Hydraul. Div., Vol. 90, HY5, pp. 43-60.

3. Matalas, N.C. (19671, Mathematical assessment of synthetic hydrology,

Water Resources Research, Vol. 3, pp. 937-945.

4. Young, G.K. and Pisano, W.C. (19681, Operational hydrology using residuals,
Proc. Am. Soc. Civ. Engrs., J. Hydraul. Div. Vol. 94, HY4, ppe 909-923.

5. Beard, L.R. (1965), Use of interrelated records to simulate streamflows,

Proc. Am. Soc. Cive Engrs., J. Hydraul. Div., Vol. 91, Hy5, pp. 13-22.

6. Moreau, D.H. and Fyatt, E.E. (19701, Weekly and monthly flows in synthetic
hydrology, Water Resources Research, Vol. 6, pp. 53-61.

7. Quimpo, R.G. (19681, Stochastic analysis of daily river flows, Proc. Am.
Soc. Civ. Engrs., J. Hydraul. Div., Vol. 94, HYI, pp. 43-58.

8. Payne, K., Newman, W.R. and Kerri, K.D. (19691, Daily streamflow simulation,
Proc. Am. Soc. Cive Engrs., J. Hydraul. Div., Vol. 95, HY4, pp. 1163-1180.

9. Parzen, E. (19641, Stochastic processes, San Francisco, Holden-Day,

pp- 144-159.
 FLOOD CONTROL DESIGN WITH LIMITED DATA -
A COMPARISON
OF THE CLASSICAL AND BAYESIAN APPROACHES

Eric F. Wood

Department o f Civil Engineering

MASSACHUSETTS INSTITUTE O F TECHNOLOGY

ABSTRACT

Water Resource planners usually design flood control structures

by choosing an extreme value model. The model's parameters are esti-
mated from the available streamflow data and design decisions are m a
de by finding the discharge related to a particular level of risk.
In most design problems the data on extreme events is severely limi-
t e d , making parameter estimation difficult. Two different parameter
estimation approaches are investigated -classical and Bayesian
which are applied to a flood design problem f o r a nortkeastern U.S.
-
river. The classical approach uses the maximum likelihood cri'terion
for parameter estimation. The Bayesian approach i s performed for 05-
jective prior information based upon observations f r o m other rivers
and for subjective prior information derived by considering the
effect o f river basin development upon flood discharges. The results
indicate that the classical and Bayesian approaches lead to diffe-
rent design discharges for the same level o f risk.

Normalmente los Ingenieros Hidráulicos diseñan las estructuras

para e l control de crecidas eligiendo u n modelo de valores extremos;
los parámetros del mismo se estiman mediante los datos de aforo y
se diseña mediante la determinación de la crecida asociada con u n
cierto nivel de riesgo. En la mayoria de los casos, la información
disponible sobre los valores extremos es muy escasa, haciendo muy
difícil l a estimación de los parámetros. Se investigan dos métodos
de estimación de pardmetros -e l clásico y el Bayesiano - que s e
aplican a un problema de control de crecidas para un rio d e l n o r d e s
te de los Estados Unidos. E l método cl'lsico utiliza e l criterio de
verosimilitud máxima para la estimación de parámetros. El enfoque
Bayesiano se desarrolla con informaci8n priori" objetiva basada
en observaciones realizadas e n otros r í o s , y con información "a
priori" subjetiva obtenida a l considerar e l efecto del desarrollo
de la cuenca sobre los caudales de crecida. Los resultados indican
que los métodos clásico y Bayesiano llevan a diferentes valores de
caudales de crecida para un mismo nivel de riesgo,
470

INTRODUCTION

In water resources planning the hydrologist's main function is analysis

that lead to engineering decisions. The decision variable is not a hydrologic
variable but a general engineering variable like the height of a dike or the
size of a spillway. The separation of hydrologic analysis and economic
analysis can not occur if efficient designs are to be obtained. In the
decision process the design variables are related to the estimation of hydro-
logic variables through a loss function which reflects the different economic
implications of the project.

If one accepts this role for the hydrologist within the planning process
then a number of important qualitative implications follow.* It is useful to
describe complex phenomena such as rainfall or runoff processes by the applica-
tion of probability theory only from the point of view that it produces a more
economical design. If streamflows can be treated as random variables thep it
is consistent to treat the unknown parameters of the distributions of stream-
flows as random variables. To treat the parameters of distribution of random
variables as random variables is not "permitted" within the framework of
classical statistics.

The extension of this argument is that it is useful and professionally

sound to treat any uncertain factor as a random variable if it leads to better
decisions. This includes variables such as the quality of workmanship in con-
struction, cost and benefit adjustments due to inflation as well as the more
traditional hydrologic variables.

The designer should consider three types of uncertainty in his analysis -

uncertainty of a probabilistic nature (i.e. frequency of occurrence), statisti-
cal uncertainty due to the limited number of observations from which parameters
are to be estimated, and professional uncertainty arising from incomplete infor-
mation concerning the underlying process and its probabilistic representation
(Cornell, 1972). The methodology of the decision process must be able to con-
sider these forms of uncertainty as well as be able to utilize professional
judgement obtained from related experience with similar projects.

Bayesian analysis within the framework of statistical decision theory

(ñaiffa, 1968) prescribes a methodology for making decisions under uncertainty.
Decision theory allows the decision maker to consider together both the uncer-
tainty of the modelled process, the quantifying of the decision outcomes and
the preferences for these outcomes. Bayesian analysis is a probabilistic frame-
work by which the uncertainty in any design variable and the knowledge about
that variable can be considered. This paper is concerned with the latter aspect
of the proposed methodology. The application of Bayesian analysis in water
resources planning is gaining acceptance as decision makers recognize the
*
Parallel arguments have been used previously in support of a statistical
decision approach to structural reliability analysis by Cornell (1972).
 471

inherent advantages that combining information sources and treating uncertain

parameters as random variables leads to better designs. In recent years many
researchers have made significant contributions to this area. These include
the work of Bernier (1967), Shane and Gaver (1970), Davis et al (1972a)
Bogardi and Szidarovszky (1972) amongst others.*

Probabilistic Model Formulation

One area of particular concern to water resource planners is the analysis

of extreme events, mainly floods. This problem is especially applicable to the
issues raised earlier since data is often scarce, with the consequences due to
an inadequate design often severe.**

The issues we wish to focus upon in this paper is a comparison of the

classical approach and the Bayesian approach to flood analysis and design when
direct observation of extreme events are either scarce, non-existant, or non-
stationary. Substantial urbanization of a river basin introduces non-
stationarity effects into the direct observations, thus decreasing their
information content.

The first step in any analysis, classical or Bayesian, is the construction

(or assumption) of an underlying probabilistic model which represents the
physical process. Consider the hypothetical streamflow trace presented in
Figure 1. The flows of interest are those flows greater than Qo and it is
assumed that flows larger than Qo can be described by a Poisson process (the
time between events are exponentially distributed) with an average annual
arrival rate and the probability distribution of the flows of interest (flows
greater than Qo) can be represented by the exponential distribution

where

This is a fairly general form since the upper tails of many distributions
may be represented as exponential. The proposed model has been used for
extreme flows by Shane and Lynn (1964) and Todorovic and Zelenhasic (1970) and
for rainfall events by (Davis et al (1972b)) and Grayman and Eagleson (1971).

*The numerous papers at the International Symposium on Uncertainties in

Hydrologic and Water Resource Systems, December 11-14, 1972, Tucson, Arizona,
U.S.A. is proof of the growing interest in this field.
**The decision makers may also consider besides economic consequences social
and professional consequences due to failure of a flood control structure.
These may be loss of life, disruption of community services and the loss of
professional prestige. On the other hand, over design commits resources that
could be used on other projects.
472

The probability that, in any single occurrence, a discharge z (z

A A
= q - QO)
exceeds the discharge z is Pz, where

Pz = 1 - FZ(z) = e
-a2

the process of these occurrences is Poisson with an average arrival

rate VP and the probability that in time t n exceedances of level z will
Z
occur is
n -UP t
P[N = n 1 = (vP,) e
(3)
n !

No exceedances of z in time t is just P [nz = O] = F ( 2 ) ; the cumulative

Z FZ(z) gives
P from (2) into
probability function of z. Substituting

F
Z
(z) = [ fvte
-az
z>o
z<o

The probability the z = O is equal to the probability that a peak discharge q

is less than Q,. IIf z is such that the probability of exceeding z is small
and the arrival rate of such events is small then FZ (z) can be approximated by

FZ (z) 1 - vte -a2

(5)
The probabilistic model of the underlying physical process serves both the
classical and Bayesian analyst but in slightly different ways.

Assume that there is no uncertainty in the model itself but only in its
parameters CY and V .
The classical analyst then obtains point estimators,

V and a , (usually by the maximum likelihood criterion) from the observed

streamflow record. His probabilistic model is

The Bayesian analyst, meanwhile, obtains probability density distributions on

the unknown parameters, v and a , from combining all sources of information.

The Bayesian approach to the use of probabilistic methods recognizes that

the subjective information of the analysis is inseparable from the objective
aspects. Subjective infomation is incorporated into the analysis through a
prior probability distribution which reflects the information content. This
prior information is combined with objective information
tions - -
direct data observa-
to provide the analyst with a posterior distribution. This reflects
all of his information. If the prior information is vague and the sample infor-
mation is very good then the posterior distribution of the information will be
 4'13

negligibly affected by the prior. The opposite also holds. The prior informa-
tion may be looked upon as that information an analyst wovld use if he had no
observable data. In the design for floods a number of sources of information,
are available. These include such sources as regression equations based on the
physical characteristics of the basin (Benson, 1962) and analytical derivation
of extreme flow dynamics (Eagleson, 1972) as well as engineering experience and
expertise. To ignore these sources is to throw away potentially significant
information which could lead to better designs. The use of diffuse or non-
information prior is in most cases wrong since it side steps this important
aspect of Bayesian theory.

The prior information and the direct observations are combined through
Bayes theorem

f" (a) = t(aJUamp1e) f'(a) (7)

where f" (a) is the posterior probability distribution of parameter a

&(alSampie) is the likelihood function of a given the observed

samples.

f'(a) is the prior probability distribution of parameter a.

The posterior distribution, f'l(a), can be found analytically if the prior dis-
tribution is a natural conjugate. To obtain the posterior distribution from a
prior which is not a natural conjugate usually requires the application of
numerical methods. The gamma-1 probability density function is the natural con-
jugate for both parameters Y and a. The Bayesian distribution of z, FZ (21, is
obtained from:
-FZ (2) = 1
all Y all a
FZ (zIv,a) f"(v) f"(a) dvda (8)

where FZ (zIv,a) = Fz (z) of equation (5)

By assuming the posterior distribution on parameter v to be gamma

parameters u" , SI' and parameter a to be gamma -
1 with
1 with parameters Y", E"
-
(these aze obtained with natural conjugate priors) permits analytical evaluation
of 1 -
FZ (2).

v a
474

- Fz
1 (2) = Ut
L 1 +E
J
""+1

where cI=vII+1
E''
3=-
u"+l
S"
This is the probabilistic model for the Bayesian analysis. It is interesting
to note that the form is completely different from classical analysis model
(equation 6).

Design Model Formulation

The motivation for developing the probabilistic models of extreme events

is to apply them in making decisions. Suppose we are interested in the damage
associated with the exceedance flow z which is larger than the flood pro-
tection flow level r . The total cost is comprised of a damage cost C,(z)
and a protection cost C (r). For our example, let's assume that the damage
cost can be expressed:!b

C,(z) = C1 (z-r) (10)

while the cost of protection can be expressed by:

C,(r) = K+ Co r (11)

If the expected criterion is used to evaluate different protection levels then

the expected cost, E[c], of protecting for a flow r is
m

z=r
For the classical model f(z) is:

from differentiating equation (6). Thus the annual exp,ected damage E[CZ] from
flooding when flood protection r is provided is

a
This assumes that r is large enough that the upper tail approximation for
FZ(z) is valid.
 475

In the Bayesian framework equation (12) applies but with the Bayesian
density function f(z) which is obtained from equation (9) as:

The annual expected damages due to flooding is

J
again assuming that the upper tail approximation of Fz (z) is valid.

Example Application

The analytical formulations developed here are applied to a river in the

Northeastern region of the United States. The mean of fhe maximum yearly flood
is about 5800 cfs. Exceedance events were considered to be flows greater than
10,500 cfs which is somewhere around the 10 year recurrence intervals. Only
three flows in the 37 years of record (1929 through 1965) exceeded this base
flow.

Bayesian Parameter Estimation

Estimation for a
Prior information on a , the event magnitude distribution, was obtained
from a regression on 36 other Northeastern United States basins. The regression
related exceedance flows to physical characteristics found within any drainage
basin. The following regression was obtained:
.153 o 2.87 A .81 .74 .54 .65
Qm- St
where*
is mean exceedance flow, in cubic feet per second
%
A
is
is
orographic factor
drainage basin area, in square miles
S is main channel slope, in feet per mile
T is average January, degrees below freezing, in degrees Fahrenheit
is percent of surface storage area plus .5 percent
St
Since partial duration series and annual flood series are virtually identi-
cal above a frequency of about the 10 year flood (Langbein, 1949) the use of the
annual series for the prior was considered to be adequate for this example.
Research is presently being conducted by the author to study the problems of
appropriate prior information.
*The physical characteristics are from Benson (1962) and the streamflow data
from the U.S. Geological Water Supply Papers. (1301-A, 1721-A, 1901-A).
 From the regression an estimate of the mean exceedance flood, Q and an
estimate of the variance of the mean flood were found to be: P9

= 734 cfs
QP
5 2
V[Q ] = 4.9 x 10 (cfs)
P
Due to the assumed distribution of the magnitude of exceedance events, the mean
exceedance flood can be related to the event magnitude distribution parameter
by Q = l/a. If Q is assumed to be distributed as an inverted gamma -1 dis-
tribetion with pargrneters v' and R' then a is distributed gamma -1 with para-
meters v', $' (biffa and Cchlaifer. 1961); that is

with

V[Qp] = v'>1
(v')2 (VI-i)
This gave parameters v' = 2, R' = 1468. Thus the prior distribution on O! ,
f' (a) is
-1468a
fgYl(a) = e a2 (i468I3
r (3)
The posterior of , f"(a) can now be evaluated by equation (7). Thus
5 -33668a
f" (a) = Ka e (20)
YI
where

n
The posterior of a is gamma -1 with parameters E' = + zi ; VI' = VI+ n

Estimation for v .
The estimation of prior information on u, the average arrival rate,
involves, as a first step, the estimation of the first two central moments of
the distribution of the arrival rate of a peak flow that will exceed the base
flow Qo. In our example, this base flow was 10,500 cfs. There are some
probabilistic or statistical methods one may use to approach this problem or
the engineer may have said simply "based upon my experience in the area, my
best estimate of v is .1 and there is a 50-50 chance that v could be plus or
minus .O25 of .l'I. The implication of that statement is that the standard
deviation is about ,033. If that is accepted for our example and if a g a m
distribution for the prior of V is assumed then
- 1

f'
YI
(v) - e
-0'V
(s'v)
r (u'+i)
U'
s' (21)

with u'
s'
-
=
8
92

The posterior distribution of , fy1" (u) is just

3 .-37 v8 .-92~
f II (y)
Y1
= v (22)

v" e- 1 2 9 ~
yhich is g a m -
=
1 distributed with parameters u"
v= .O85 events per year.
- 11, si' = 129, and mean

Substituting these into the Bayesian design model of Equation (9) yields

1 - -FZ (z) 5 .085t

Thus the Bayesian model, 1 --

F2 (z), of equation (23) is shown in
Figure 2. The effect of considering diffuse prior information (no observations
in no years of data) is also shown in Figure 2.

Classical Proceedures
Application of the classical estimation proceedures is straight forward.
Estimators for both the average arrival rate, v , and the parameter of the
event magnitude distr&bution, a can be obtained by applying the maximum likeli-
hood criterion. For v , the estimator for v, the likelihood function is:

and the maximum likelihood criterion; 2 = O yields

a,
j= 0 = -
3
37
= .O81
tr
Similarly for a;
-5
a = n = 9.3 x 10

zi
i=l
478

A Kolmogorov-Smirnos test on the models using the derived estimators

passed the .10 significance level with ease.

Thus the classical estimator model for our example is

1 - FZ(z) .O811 e
-9.3 1 0 - ~ ~

which is the probability of observing a peak flow z in the next interval of

time. The classical model, 1 -
FZ (z) represented by equation (25), is compared
to the Bayesian model in Figure 2.

Design Application

Using cost coefficients for Equations (10) and (11) as being:

4
c1 = $10 Icfs

K = $25 x lo4 equivalent annual cost over a

= $102/cfs proposed 50-year project life.

for the classical design proceedures in Equation (16) utilizing the Bayesian
design model and in Equation (14) for the classical design model.

The expected annual cost of providing protection against the 100 year
flood is presented in Table I.

100 yr flood Expected Flood Equivalent Annual

discharge Damages ($1 Protection Cost-50 Yr Life($)
Bayesian Model 17500 7 105 20 105
Classical Model 22500 10.76 lo5 25 105

- ComparisonsandofClassical
~~

Table I Costs and Damages for Bayesian

Models

For each model the flood which had the lowest expected total cost also was the
100 year flood.
Discussion
Incorporating Non-Stationarity Effects

The problems of flood analysis when non-stationarity has been introduced

into the streamflow records due to increased development of the drainage basin
have not been completely solved. Recent studies by Bras (1972) have shorn
 479

increases in flood peaks of developed catchments of between 30% to 115% depend-

ing upon the particular size and shape of the storm. Basin development tends
to remove natural stream storage areas as well as decrease impervious areas and
holding ability of natural ground cover. The effects of changing these charac-
teristics can best be investigated by a deterministic catchment runoff model
that utilizes a stochastic rainfall generator (Harley, Wood and Schaake, 1973).

The Bayesian analyst has a number of options open to him which include a
rainfall analysis, a runoff analysis from the catchment analysis, and other
approaches. He can either utilize the streamflow data or ignore it, applying
his engineering judgment in many ways.

The classical analysis has few, if any, options open to him. The strict
application of his theory permits him only to consider the historical record
which will not apply to the developed basin. If the amount of development is
small, then the historical record may still contain valuable information but if
extensive modifications have taken place and the classical analyst still uses
his historical record then he must be able to defend it.

Conclusions
The role of analysis is to aid decision making. The two approaches pre-
sented here lead to quite different design decisions.

The classical approach restricts the analyst to the observable hydrologic

data to which other information sources can not be added. Furthermore, it is
not possible to include within the analysis other uncertain parameters which
may affect the design. Instead some other artifical mechanism is used such as
adding a factor of safety to the design variable, designing for the largest
possible event or using some other method which can not be related to a mean-
ingful economic (or social) preference criterion.

Too often too much weight is given to a few observable data points and too
little weight to other available information. Lhe Bayesian analysis is a meth-
odology which enables the combination of information sources as well as allows
the explicit evaluation of the effect of all sources of uncertainty upon the
decision variables. The application of the Bayesian approach will lead to
better design than will a classical analysis which is restricted to a few obser-
vations and whose conclusions are difficult to interpret.

Acknowledgments

The work was supported by the Office of Water Resourc Research, Office
of the Interior, United States Government under Grant No. 14-31-0001-9021.

References

1. Benson (1962). "Factors Influencing the Occurrence of Floods in a Humid

Region of Diverse Terrain" U.S. Geological Survey Water Supply Paper 1580-B,
Washington, D.C.
480

2. Bernier (1967). "Les Methods Bayesiennes En Hydrologie Statistique"

Proceedinvs of the Int. Hydro Symp., September 1967, Colorado State
University, Fort Collins, Colorado, USA.

3. Bogardi and Szidarovszky (1972)."The Margin of Safety for Compensating

Losses due to Uncertainties in Hydrological Statistics" Proceedings of
the Int. Symp on Uncertainties in Hydrologic and Water Resource Systems.
Dec. 1972, University of Arizona, Tucson, Arizona, USA.

4. Bras (1972)."Effects of Urbanization on Runoff Characteristics of Small

Basins in Puerto Rico" Unpublished Bachelor of Science thesis, Department
of Civil Engineering, Massachusetts Institute of Technology, Cambridge,
Massachusetts, U.S.A.

5. Cornell (1972). "Bayesian Statistical Decision Theory and Reliability-

Based Design" Structural Safety and Reliability, (A. Freudenthal, ed.),
Pergamon Press, New York.

6. Davis, Kisiel, and Duckstein (1972a)."Bayesian Decision Theory Applied to

Design in Hydrology" Water Resources Research, Vol. 8 No. 1.

7. Davis, Duckstein and Kisiel (1972b)."Uncertainty in the Return Period of

Maximum Events : A Bayesian Approach" Proceedings of the Int. Symp. on
Uncertainties in Hydrologic and Water Resource Systems. December 1972,
University of Arizona, Tucson, Arizona, U.S.A.

8. Eagleson (1972). "Dynamics of Flood Frequency" Water Resource Research

Vol. 8, No. 4.

9. Grayman and Eagleson (1971). "Evaluation of Radar and Raingage Systems for
Forcasting" Ralph M. Parsons for Water Resources and Hydrodynamics T.R. No.
138, Department of Civil Engineering, M.I.T., Cambridge, Mass. U.S.A.

10. Harley, Wood, and Schaake (1973). "The Application of Hydrologic Models to
Urban Planning" Presented at 54th Annual Meeting, American Geophysical
Union, Washington, D.C., April, 1973.

11. Langbein a949). "Annual Floods and the Partial Duration Flood Series"
American Geophysical Union Transaction, V. 30, pp. 879-881.

12. Raiffa (1968). Decision Analysis, Addison-Wesley, Reading, Mass., U.S.A.

13, Raiffa and Schlaijer (1961). Applied Statistical Decision Theory, M.I.T.
Press, Cambridge, Mass., U.S.A.

14. Shane and Gaver (1970). "Statistical Decision Theory Techniques for the
Revision of Mean Flow Regression Estimates" Water Resources Research,
Vol. 6, No. 6.
 481

16. Todorovic and Felenhasic (1970). "A Stochastic Model for Flood Analysis"
Water Resources Research, Vol. 6, No. 6.

17. United States Department of the Interior. "Surface Water of North Atlantic
Slope Basins, throu$-11950", U.S.G.S. Water Supply Paper f301, Washington,
D.C., 1957.

18. . "Surface Water of North Atlantic

Slope Basins, 1950-60", U.S.G.S. Water Supply Paper 1721, Washington, D.C.,
1969.

19. . "Surface Water Supply of the U.S.

1961-65Jater Supply Paper 1901, Washington,
D.C., 1969.
482

4
Q

TIME

O. DISCHARGE Q I EXCEEDANCE DISCHARGE 2 = Q-Qo

Qo

Figure 1: Typical Discharge Record Showing Exceedance Events and

Showing the Probability Density Functions for both Discharges
and Exceedance Events.
 483
O
O
O
ln
N
O
O
O
O
01
O
O
O
Lrl
ri
O
o
O
O
i
O
O
O
U1
O
rn
O 2
O O
 T H E USE O F M A T H E M A T I C A L ( D E T E R M I N I S T I C ) M O D E L S

General Report
bY
J. E . N a s h
U n i v e r s i t y College, Galway. I r e l a n d .

Authors and Titles; At the time of writing four papers have been received. These

axe:

(1) "A Rainfall-Runoff Model Based on the Watershed Stream Network" by

J.W. Deleur and N.T. Lee, of the School of Civil Engineering, Purdue
University, and the Department of Agricultural l3conomics of the

university of Illinois, respectively.

(2) "Monthly Streamflow EstLnation from Limited Data" by C.T. Haan,

of the Agricultural Engineering Department, University of Kentucky.

(3) "Obtaining of Deficient Information by Solving Inverse Problems

from Mathematical runoff models", by V.I. Koren and L.S. Kutchent,

of the Hydrometeorological Centre of the U.S.S.R.

(4) "The. Mathematical Model of Water Balance for Data Scarce Areas" by

Nabil Rofail, Water Resources Department, Desert Institute of Cairo.

Introduction: In view of the relatively small number of papers it had been

suggested to me by the OrgGisers that I should include some introductory comment

of my own,on the subject of catchment modelling. However, while the number of

Papers is indeed small, they are all interesting and two of them are of a
 486

mathematical nature and will require time to elucidate. One of these is particular3

interesting,in that it presents a practical tool developed and used in the

Soviet Union but, as far as I am aware, not generally known in the West. This

paper also happens, understandably, to be very difficult to follow, and on

these two accounts, I propose to devote a somewhat disproportionate part of the

time available to its consideration. I feel sure that the other authors will

not consider this in any way a slight and, as I am sure that this distinguished

audience would prefer me to devote any time available to the consideration

of khis interesting technique, I shall keep my own general comments on the

subject of modelling as brief as possible.

Hydrological Modelling: The variety of titles among the papers we are considering

reflects the widely different senses in which the term modelling is understood.

Nevertheless, there is a strong common link between them. This would seem to be,

that a natural phenomenon such as the conversion of rainfall into discharge, or

the movement of water in a porous medium, is represented by an hypothesis or

model, expressed as a series of operations which are,performed on one

function of time (the input) to convert it to another function of time (the

output), or as a single mathematical relationship, a partial or ordinary

differential equation which must be solved in terms of boundary CQnditions.

The relationship between the three elements may be represented by

Where the relationship is of a causal nature this may be indicated by

6 f fe.c E-
y ?)
 487

I make this distinction because the mathematical solution of such problems is

often against the direction ,of the arrow, which from the mathematical point of

view may,therefore,be considered irrelevant. I shall use the terms cause and

effect for emphasis, only when the direction of the arrow is physically relevant

Either diagram represents a relationship between three quantities one only of

which may be unknown in any realistic problem.

The solution sought may be the output, the input, or a description of,

(or parameters of) the operation itself (e.g. a unit hydrograph or the coefficients

of a differential equation).

Generally speakingrit is true in the hydrological context that the operations,

viewed in the direction from cause to effect are stablelin the sense that bounded

causes produce bounded effects and small variations in the causes produce smaller

variation in the effects. Precisely because of this fact, the inverse operation

discussed by Koren and Kutchment, of the discovery of the cause of an observed

effect, or the discovery of an operation itself, tends to be unstable and small

variations oierrors in the observed output produce larger variations or errors

in the computed cause,or the computed values of the parameters of the operation.

Por this reason the solution of the inverse problem is usually very much more

difficult than the solution of the direct problem.

The Direct Problem: This has at least a logical simplicity and the difficulties

which arise usually involve questions of convergence of finite difference

solutions of differential equations. The paper by Nabill Rofail describes the

solution of one such problem which we shall discuss in some d e a i l later.

488

Among the inverse problems it is useful to distinguish three types

(a) The input is unknown

(b) The values of the parameters of the operation are unknowr

(c) The form and parameter values of the operation are unknowr

In the particular case of a lumped linear system where the input, the

operation and the output may be represented by

L

where x(t) is the input, y(t) is the output and h(t) the impulse response, the

three classes collapse to one. For such systems the form of the operation may

be described uniquely by the impulse response of the system and,theoretically at

least,this may be found without prior specification of its form. Therefore the

second and third classes merge. Furthermore, because of the symmetry in h and

x in the two equations, a symmetry which becomes more obvious when the relationshi1

ià expressed in terms of Lapace transforms through the Faltung theorm,

Ys) = X(sj HU) (3)

the problems of discovering h and x are mathematically the same, so that all three

distinctions vanish. in the nonlinear problem, however, or when recognition

of the system implies discovery of the coefficients of a partial differential

equation (a distributed system) the distinctions remain valid.

The Lumped,Linear Model: For functions which are not simple expressions, it is

usually easiest to deal with these models in discrete form. Eqs. 1 and 2 are

replaced by €Y] = IhJ i4 (41

and ius = PI €h3 c=>

 489

where [x] and [y] are vectors of the input and output respectively, sampled

at equal time invervals. [hJ is a rectangular matrix formed from the

ordina tes of the impulse response as

h, O 0 :

[h] =
.

Similarly Lx] in eq. 5 is a rectangular matrix formed from the input

ordinates in the same way.

The direct. problem of finding fy3 is trivial. The inverse problems of

finding 1.3 of eq. 4 or {XI from eq. 5 are mathematically the same.
Because of the,great stability of the operation in the direct direction,

solution of the inverse problems tends to be very unstable and therefore

difficult. Traditional means usually involve a least squares solution;(Snyder)

or the imposition of constraints on the impulse response e.g. harmonic analysis

with a limited number of terms (O’D~nnell). A new method of obtaining a least

squares solution under a constraint is described in the paper by Koren and

Kutchment.

Distributed Linear Models: If the input is distributed in space in a

constant manner the direct problem is essentially the same as that of the

lumped linear system (Kraijenhoff van de Leur, Venetis). If however the

input is arbitrarily distributed in space,the differential equation must be

solved numerically for the direct problem, usually by reducing the problem

to linear difference equations, which are solved at the node points of an xt

plane. An example is provided in the paper by Nabil Rofail.

 490

Threatment of the inverse problem to discover the input or the coefficients

in a known linear system expressed as a partial differential equation,are rare.

An attempt to discover the characteristics of an aquifer is described in ref. 6 of

the paper by Koren and Kutchment, and the paper itself gives two examples of such ai

attempt to determine values of the conveyances and cross sectional areas as

functions of space and time in an open channel.

The General Nonlinear Problem: The direct problems are again relatively

straightforward - the nonlinearity complicates the numerical solution of distribute(

systems (partial differential equations) but .the lumped parameters systems are

scarcely affected.

The Inverse Problem involves, generally, postulation of the form of the

operation (i.e. a conceptual model) and estimation of the parameters by

successive approximations. The first approximations are inserted in the model

and the output computed. This is compared with the observed output and a

single expression of the observed errors (the objective function) is systematically

reduced by subsequent trial and error adjustments of the parameters values.

Examples are provided in the papers by Deleur and Lee and Haan. The major'

difficulties with this method are that the set of solutions obtained may not

be unique and, particularly when two or more parameters represent similar

operations,the optimised values are subject to very .high sampling variance.

It would be interesting to speculate whether such problems could be made amenable

to direct least squares approximations, as so often used in the corresponding

linear case. Theorétically, this would seem possible, but it may be, as seems

to be generally assumed, that the complexity. of the equation representing the

 491

dependence of the objective function On the parameters might render its formulation

difficulty. I feel however that this possibility ought to be explored.

Having thus classified the papers according to the nature of the problem discussed

we corne to a consideration of the papers themselves in some detail. These I

would like to take in the order of their classification above.

492

A Mathematical Model of Water Balance for Data Scarce Areas

by Rofail

The title of this paper is somewhat misleading. The subject matter is in

fact the numerical solution of the linearised equations of motion of groundwater

in an unconfined aquifer. It is therefore a case of a direct solution of a

distributed linear system. Neglecting any vertical component of velocity, the

horizontal components parallel to the x and y axis in a homogeneous aquifer are

proportional to the gradients of the piezometric head (h+z). The constant of

proportionality (k) is known as the coefficient of permeability, (authors eqs.

1 and 2).

The continuity equation (aut4ors eq.3) expresses the fact that the rate

of rise of the surface of saturation , at a’point in the aquifer, is

proportional to the rate of percolation down to the aquifer at this point, plus

the net rate of flow towards the point (the negative of the divergence). The

constant of proportionality is known as the specific yield and is given the

symbol/h in authors eq. 3. These two equations are combined in the authors eq.4

by replacing the velocity terms by the corresponding gradients of the piezometri

head, yielding an equation in the head only.

3Yhtz) k ukæl a h ,
031
3% b>c
 493

This equation contains first and second order derivatives of the depth h and

the elevation of the aquifer bed z, and is non-linear due to the occurrence of

- By further assuming that the gradient of h is small relative to that of z,

ah
terms such as (&become small relative to s.
9
s and are dropped,thus

linearibing the equation (authors'eq. 5)

9 8 - h $:jL- 11 WZ. - ah az-

arc-E
-h--,bbLh
u
-hg?, ah .
DY
N = o
k--
a y k
- (Id
This assumption is attributed to"Boussenzq" and is stated to be that the

powers of derivatives of the first order are of a lesser order of magnitude

than the derivatives themselves. This would, of course, be an acceptable assumption

but it does not seem to be that which is in fact made by the author in obtaining

his eq. 5 from eq. 4. Perhaps the authors would like to comment on this.

To emphasise the linearity of eq. 5 in terms of the partial derivatives of h,

the partial derivatives of z (assumed to be known) are written as 7,, and fy

and the second order derivatives (also assumed to be known as r,, and fiv

in eq. 10. (Note that in the text the quantities 7,, and Vy are incorrectly

stated to be the partial derivatives of h; this in only a printing errer).

In order to advance the solution from time n'to t h e n+l the author

replaces the equation of motion with two distinct finite difference approximations.
 494

The first is applied to the first half of the time interval and the second to

the second half. For ease of comparison I reproduce here the equivalences as

used in the two successive steps.

j and k refer to the node point location in the x and y directions and n

to the number of the time interval. approximation

1st half 2nd half
derivative step. -step

These approximations have a certain symmetry, which when the finite differenc

approximations are added for the two half time steps,make the results consistent

with the original equation up to the second order. They have the additional merit

that the finite difference equation for the first half step in time, can be
n 42
written implicity in terms of linear combination of (h,-l ., h, and hJ+l)
with the coefficients all known (author's eq. il). Similarly the equation

applied to the second half step yields an implicit equation linear in

h, and hrti
(LI rt' ail the coefficients again being known (authors eq. 12).

These implicit equations can be solved as a linear set when the boundary condition8

are provided to yield h for all node points at a single time. Repetition extends

the solution through time.

 495

"A rainfall runoff model based on the watershed and stream network"
ßy Delleur and Lee

The problem discussed, i.e. that of recognising a lumpedpon-linear model,

is in the inverse category and the method used is that of postulating the form

of the system and optimisation of the parameters.

The authors begin by pointing out that even with forty years of record the

errors in estimating the parameters of a stochastic model of annual flows

may be quite high. They mention also that stochastic linear models of the

rainfall-runoff process,likewise,require long term series of both the rainfall

and runoff for their calibration. They conclude, therefore, that for regions

with inadequate data one may have to resort to deterministic models either of

the "black box" or of the "physical" type. This distinction is made according

to whether the model form attempts to mirror the physical processes or is merely

a linear regression. The authors point to the obvious deficiency of black

box models,that they cannot be transferred from one location to another because

there is an absense of a one to one relationship between the parameters of the

model and the parameters the watershed. They conclude, therefore, that a

physical model requiring only a small number of identifiable parameters or a model

based on data which can be obtained in a relatively short time is required.

I am not sure what the authors mean by ''a stochastic linear model of the

rainfall runoff process" nor am I sure that these several models are

alternatives for the same piirpose. It seems to me that if one requires a

stochastic model in order to generate a time series having the properties of

the observed sample it can scarcely be rubstituted for by a deterministic model

(whether of a black box nature or otherwise) relating rainfall to discharge.

 496

Before such a model could be used to produce a synthetic discharge record the

stochastic properties of the rainfall input would have to be computed and a

synthetic rainfall record fed into the deterministic model. Thus it would seem

that the problem had merely been transferred rather than solved by the substitution

of the deterministic model. If of course a much longer ,rainfall record was

available this might be useful. The authors intention is to provide a deterministi

model with a small number of parameters preferably identifiable from the catchment

characteristics. I don't .think there would be any argument about the usefulness

of such a model even if it would not substitute for a stochastic model for a

different purpose.

The authors suggest that the availability of modern techniques of photography

and general remote sensing technology make it possible to observe relevant

catchment characteristics on a large scale and therefore to include these in the

deterministic model. In the authors'model use,is made of the following catchment

characteristics obtained by aerial photography - the plan form of the stream

network, the topography, and the soil type. The model attempts to relate the

areal mean of the rainfall to the discharge at the gauging site, both as functions

of time. The structure of the model depends largely on the concept of a contributi

area which is the area contributing at any given instant to the flow at the

gauging station i.e. the function of time representing that portion of the

catchment from which runoff is currently passing the gauging station at the time

in question. Obviously the contributing area is a functi,on of the catchment

wetness and the model for this quantity, described by the authors'eq. 1 is clearly

dependent upon the antecedent rainfall.

 497

The contribut?.ng area at t i m e i b t is A ( f d k ) and the total catchment area is A,.

I assume that in eq.11 the second negative sign from the right in the numerator

should in fact be positive, and with this interpretation I understand the

assumption of eq. 11 to be that the contributing area expressed as a proportion

of the total catchment area varies with the Nth power of a wetness index, which

is obtained by the sumation up to the time under consideration of the proportion

of the net rainfall in each previous time interval, weighted in an exponential

manner according to remoteness in time. Later it is stated that this equation

is subject to a constraint of continuity, that is, that the total effective

rainfall is equal to the total discharge. It is not explained how khis condition

is fulfilled but if eq. 1 1 is taken as a statement of proportionality rather than

of equality, giving, therefore, the relative v a l u a a t ail times of the contributing

areas, the constant of proportionality may be chosen to satisfy the requirement

of continuity. This is my interpretation of the authors'intention. The quantity

B would seem to be a constant loss rate existing throughout the storm. I think,

perhaps, the authors might like to clarify these few points and explain what

happens if the rainfall intensity is less han B.

The next step is to distribute the total contributing area A(7At) along the

channel. of the catchment. The contributing area per unit length of channel

At> is obtained under the following assumptions.

1. Constant velocity at a given time throughout the catchment.

2. Uniform distribution of drainage density.

3. Uniform distribution throughout the catchment of first order streams.

498

Under these assumptions the total contributing area at any given time

is distributed according to the distance (or time of flow) from the gauging site,

in th? same way as tli? number of channels is distributed according to distance

from the gauging site. Thus it is possible,on an examination of the stream

system,to define the function al2 6 k) in space and time.

The input of each reach, in each time element, is obtained by multiplying

the appropriate contributing area by the net rainfall intensity (i.e. the

total rainfall intensity minus the loss rate) and this input is routed through

the channel system by a 1inear.method (Dooge and Harley) to give the output

at the gauging station as a function of time.

Because the routing is linear the output due to the input on a given reach

could be represented by a convolution integral (but the kernel may vary from

reach to reach). The total output as a function of time is obtained as the

spatial integral of this convolution integral. The kernels themselves vary

with stream slope, a reference discharge, and a roughness parameter.

To reduce the complexity it is proposed to replace the actual stream network

by a "folded up" one in which (I believe) all stream elements lying the same

distance from the gauging site would be assumed equal to one another in the

properties of length, roughness, slope and reference discharge. They would

also agree, of course, in the depth of contributing area. It is mentioned later

that the roughness and slope parameters are obtained by actual observation but

it is not clear to me how this can be done in the idealised or "folded up" model.
 49 9

The parameters of the model are:

A catchment area
D and N numerical parameters in eq. 1.1

B the constant loss rate

CZ the roughness coefficient (one parameter only)

QR the reference discharge (one parameter)

SL the main charnel slope (one parameter)

Subsequently in the work, the parameter B was set to zero and the area

and slope parameters were obtained by physical measurement (I think the authors

might like to explain this a little further) the remaining parameters D,N, CZ

and QR were obtained by optimisation. Details of the method are not given,

nor are we told what sampling variance of the optimum values was obtained.

We are told,however,that the model was insensitive to D (when D exceeded 0.5)

and a fixed value of D = 0.8 was chosen. QR was found to vary only slightly

between 1.1 and 1.4 cubic meters per second for rainfall values ranging from

2.5. to 14 milimeters. Thus only N and CZ are left as free parameters.

The model was applied to 13 basins in the eastern haî€ of the United

'States and when.the optimised values had been obtained, relations were sought

between these and the catchment characteristics, so that these relations could

be used to provide estima'tes of the parameters values for use subsequently

on wigauged catchments.

N was found to vary with the ratio of runoff to rainfall volumes for the
storm according to authors eq. 8.
# = esp (0.464- R,)/O.24L
500

This ratio itself was found to vary between storms in accordance with the

daily temperature, an index of soil permeability, rainfall volume, and maximum

in t ens ity e

CZ was significantly related to basin area, stream slope and the base

flow value at the time of occurence of the storm.

Using these relations between the model and the catchment to estimate

the parameters of the former for insertion in the model which was subsequently

fed with'the observed rainfall, good results were obtained, hydrograph peaks

being reproduced with an error of the order of 20% in magnitude, and 10% in

timing .
 501

Monthly streamflow estimation from limited data

by Haan.

The purpose of the exercise described in this paper is very similar to that

in the paper by Deleur and Lee. A four parameter model is used to compute

monthly runoff from daily rainfall and the parameters are related to catchment

characteristics by regression equations. The model is of the physical rather

than the "black box" type according to the distinction of Deleur and Lee

and according to the classification I have suggested, the problem is inverse

non-linear lumped.

The structure of the model is not described but we are told that there

are four parameters.

fmax- maximum infiltration rate (cm-hr)

'max
-- maximum daily seepage loss (cm)
c -- "the water holding capacity of that part of the soil, from
which the evapo-transpiration rate is less than the potential

rate, unless this portion of the soil is saturated".

F
s
-- fraction of seepage that becomes runoff.
The input to the model is a series of daily rainfall values and average

monthly values of potential evaporation (evapo-transpirat'ion).

The optimisation is obtained by comparing computed and observed values

of monthly discharges and summing the squares of the errors to obtain the

objective function. The search is carried out along the axis of each parameter

in turn.
502

The model was.app1ied to 27 catchments in Kentucky and South Carolina and

a four percent average error was found in the prediction of the annual discharge.

This of course is not a very efficient test of the model. Details of the results

obtained are not provided - in particular the estimates of the sampling variances
of the optimum parameter values are not provided.

To provide for ungauged catchments,regressions were sought for the optimum

values of the parameters obtained from 17 catchments on certain characteristics

of these catchments. The independent variables were 12 in number (see table 1

of the paper) leaving, it would seem, only 5 degrees of freedöm, though perhaps

even this is an overestimate as covarience terms appear in the regression equations

It would be interesting to learn how significant the coefficients in these

equations appear to be.

Having obtained the regression equation the model was applied to six

catchments not used in obtaining the regressions. The model parameters were

obtained from the regressions and the runoff simulated. Percentage errors for

the total runoff for the whole period varied from 1.8 to 11.8. These figures

do not indicate how the model performed over shorter periods, for example of

one year, one month, peak flows, etc.,etc..

On a single catchment the effects of different methods of parameter

estimation are explored. Firstly, the parameters are estimated from the

regression equations asd a percentage error (in the total flow?) of 8.64%

observed. Optimisation of the parameters in the first year of record surprisingly

increased this figure to 10.13% and when 2 or 3 years of records were so used

figures of 2.19 and 9.38 were found. The last optimisation was a rather curious

one. The parameters were first obtained through optimisation in the first years
 503

record and the remaining 21 years Of Output simulated. Next the worst two of

these years, from the point Of View Of agreement between computed and observed

outputs, were noted and the parameters optimised.again,independently in the

records of these two years. The final parameters were taken as weighted

averages of the two sets weighted according to the sum of deviations of

observed and simulated flows. When the simulation for the full period of record

using these final values of the model parameters was made the observed

error in the total discharge was only 0.56%.

504

"Obtaining of Deficient Information by solving inverse problems for Mathematics

Runoff Models" by Koren and Kutchment.

The authÒrs'definition of an inverse problem is in agreement with %hat which

1 have been using. They explain the difficulty of obtaining solutions to such

problems and mention the lack of uniqueness of the solutions obtained by

postulating the form of the operation and adjusting the coefficient or parameters

by trial and error. Instead they propose the application of an algorithm due to

Tikonev which restores the proper posing of the problem and limits the possible

variation of the solution in accordance with "a priori" information on the

so lu t ion.

Unfortunately, I am not familiar with the sources quoted and my interpretatio

of the method derives solely fromthe present paper. I would hope the authors

would forgive me if I misinterpret their intention and I would hope that, if

at all possible, time should be provided to allow them to correct me and explain

s e z r a l points of difficulty which I stil1,only very imperfectly,understand.

The method involves the algebraic minimisation of an objective funtion and

is akin to Lagrange's method of undertermined multipliers.

Consider a function F(h) where h is a vector .hl,h2...and suppose

we wish to minimise F(h) subject to a constraint on h, e.g.T(h) = O.

Lagrange's method states that the conditional minimum of F(h) occurs at the

same h ao the unconditional minimum of G(h,a) where

 505

G(h,oc) a F(h) + - 9th) (is1

A formal algebraic proof is possible but scarcely necessary. Along the

constraint cp(i-i) =O the functions G(h,*) and F(h) are identical and therefore

their (conditional) minima agree. But the unconditional minimum of G(h,a)

obtained by differentiating G(h,ch) with respect to h a n d u and simultaneously

equating the derivatives to zero, implies q(h) =O or the general (unconditional)

and conditional minima of G(h,a) agree. Consequently the unconditional minimum

of C(h,5) gives the value of h which corresponds the conditional minimum of F(h).

An optimum value for a i s also found. The method used by the authors would seem

to be an adaptation rather than the straightforward use of this method. A

series of values of the vector h which minimise G(h,@) for a series of values

of agradually increasing from zero toward the optimum (SC are found by
opt
differentiating. The first of these vectors h (corresponding to = O) corresponds
to the unconditional minimum of F(h). The last (corresponding toca*
opt
corresponds to the conditional minimum (i.e. to the constraint fully implemented)

and the intermediate solutions correspond to the partial implementation of the

constraint.

in this way the investigator is enabled to seek about in the vicinity of the

optimum h for one which provides a reasonable compromise between satisfying the

constraints and minimising the function.

in the three examples quoted in the paper the physical problem is reduced,

in one case after the application of much ingenuity, to the solution of a set of

linear algebraic eqs.

4
Q =AZ (I 6)
Where Q and h are vectors and A a rectangular matrix. Assuming redundancy among

the equations a least squares solution could be found by minimising

F(h) 5 /I P h - all" (I7J
506

As shown by Snyder, the solution of this equation is

A*Ah = A*Q (12)

or h = (A*A)-l A*Q (1 9)

In the inverse problem, in the hydrological context (e.g. h is the impulse

response or the input to a linear system) this equation is often badly conditioned

and the h obtained may be seriously distorted by small errors..in Q or A.

In particular, h may fail to conform to some physical requirements (e.g. unit

area or smoothness of the impulse response). The authors method is to minimise

i.e., to find h from (A*A+ a E ) h = A* Q (where i? is the unit matrix)

with preselected values of e(presumab1y increasing from zero.

Obviously the smaller the value of q the more /lh112 is permitted to increase

and therefore increasing q corresponds to increasing the permissible fluctuation

in h. Presumably a selection is then made between the several h, bearing in

mind that the nearer oi is to zero the nearer h is to the least squares solution

of the equation.

It is clear that the constraint need not be precisely stated. It is sufficiei

that the coefficient of arepresents some quantity which increases with the

undesirable property of h. If the constraint can be precisely stated,e.g.rh = 1,

the Lagrangian method would probably be the better.

Of the three examples quoted by the authors, the first involves finding the

effective rainfall input given the impulse response and the discharge. The

problem as we have seen is identical (even in the constraint) to that of finding

the unit hydrograph given the input and output. The authors mentioned various
 507

constraints includkigrh = 1 which would yield

9 =Ikh-QI\ + aZh

and a smoothness constraint /h/ yielding

+ ec
=Ih-Q/I hl I)
Straightforward application of Lagrange s method would of course yield h =O
which would be useless. The solutions for lesser values of &would permit h # O
2
while restraining /h// .

The second problem discussed by the authors is that of discovering the

coefficients of the Saint-Venant eqs. for flow-in open channels

- ;;= $+$*&(e)
.II +i&($) (d

as +Fk= 0 )
This problem arises in, for example, flood routing, where it is impracticable

to measure the conveyance and areal relations K(z,x) and F(z,x) for each

cross section. Instead, smoothed values of these functions may be obtained

from observations made on the discharges and waterlevels as functions of space

and time, Q=Q(x,t) and z=z(x,t), during the passage of a particular flood.

ûnce.these relations have been established they may be used directly in subsequent

routing operations.

The continuity equation integrated with respect to x, provides

Q(x, t >-Q(o, t 1 = &F (i,,t 1%

where 7) is a dummy variable along the length x. In ?finite difference form,

this equation applied to a reach of channel from K=O to K=i, becomes

f$
I< :
(j+l, k 1 + F (j+l, k+l 1-F (j,k 1-F (
j,k+l)]
It.
~ ~ Q : q ( j + l , o+) Q(j+i,i) - Q(j,il-Ki,od

where j and k refer to time and space, respectively.

508

The authors state that this eq. maybe arranged as

- ? +
AF =Q

one such equation existing for each discrete time and the vectors running, as it

were, along the channel.

It would seem that 2);has been omitted, but even allowing for this, I cannc
express eq. 24 in this form. Nor does it seem to me that eq. 25 is redundant.

It would certainly be interesting to have this point cleared up, but I think we

can all accept that the finite difference equation can somehow be reduced to a

set of linear equations between the changes in time in F(x) and in distance iii

Q(>r). Such a set of equations would apply for one instant only and would take

the form of

The authors apply the algorithm already expiained,with the constraint that

Ik-FOIr, where F is an initial estimate of F is kept small,

0 - 0)

Q@,F,) = //AF-Q11 +aIIP-gI (27

is minimised for chosena's by solving

(A*A+aE )F=A*Q+ a EP C=
With regard to choosinga the authors mention the "method of discrepancy" whit

I don't quite understand, nor can I see how the smooth variation of P with time

can be insured, as P(x) seems to De found independently for each time step.'

Perhaps in calculating F the values of F obtained .in the previous time step may bc

used in the algorithm for F and thus, by constraining //F-Fo1\2 the change in F
O
is distributed regularly over all x
's.

Having found, P(x,t), thus, from the continuity equation, and having z(x,t)
 509

already, F(x,z) can be found.

The dynamic equation is similarly used to find K(x,t) and hence K(x,z).

Details are not given by the authors but the computation would seem to be

quite independent of the Computation of P(x,t,).

in their thikd and final e x m p l e the authors deal with the same equations

but Pith different boundary conditions. This time they assume 'that the

discharges are known as functions of time, only at the beginning and the end

of the channel reach, i.e. Q(o,t) and Q(L,t).are known. They assume also that

z(x,t) is known. These conditions are more parsimonious than in the former

case.

The difference between the discharges at the ends of the channel reach

is related to the rate afincrease in storage in the channel by the continuity

The right hand side is a single known quantity for sach t h e step and may

thus be expressed as a vector in time. The left hand side, because of the

integration with respect to x, is also a vector in th(unknown).

Assume that P(x,t) can be expressed as a smooth function of space and

time by a Chebishev polynomial.

Ex;>anded, this would be an ordinary polynominal in x and z and in xz with

constant coefficients depending on Aks.

510

To evaluate the coefficients, P(x,z) could be inserted in eq.29 but as this

is a difference equation Li P, the solution would be underdetermined at least

to the extent of the arbitrary constant. Instead of using F, the authors use

the top width B(x,z) and.expand this in x and z as

I€ there are m by n t e m in the expansion of B(x,z),there will be m by n

unknown coefficients A and,therefore,at least this number of equations in

ks
the form of eq.29 must be found. This can be done by taking sufficient time

intervals in the rising and falling hydrograph and evaluating the right hand

side of eq.29 accordingly to yield xl,x2,X3,...............

For every k and s the quantity

(321

is known for every.x and t and the integral with respect to x can therefore be

found. Hence the coefficients of every A in eq.29 after substitution

k.9
can be written down yielding.

%st s
k
'
= Xt and this can be arranged in matrix form, if (33)

necessary, and I think it would be necessary, with Akiswritten in vector form

A,l, Ak2, Ak3, ............. Thus the linear equation in the unknown Aks would
be obtained as
(Sf 1
0 Q = 9x
where $ is the matrix of the coefficients Cks in eq. 33 and the solution for
9
8 the vector of unknown A,s found by application of the algorithm.

(Q* F B I 0 = 9*x @a
In this case, the constraint imposed is 191 small. The choice afa(and

therefore of e) seems to be made at the value ofawhere a further change

inawould produce only a m i n h u m change in 8 expressed by eq.36.

 511

Y
(c= 2 p(QP+,)
J- I
-
Once F(x,z) has been determined, and remembering that we have already

z(x,t), K(x,z) can be found from the dynamic equation simplified to

and, thus, all the parameters of the equation are available. I cannot quite
follow the authors explanation of this part of the project. There may

be some subtleties here which I am missing,but I can see no particular difficulty

in solving it along the lines I have indicated.

This is an extremely interesting though difficult paper and I hope the

authors will be available to correct iiiy very inadequate exposition of it and

perhaps resolve some of the difficulties which I have mentioned and others

which may be tròuhling other colleagues.

References

(1) Snyder, W., Tennessee Valley Authority (1961)

"Matrix operations in hydrograph computations"

(2) O'Donnell T., (1960) 'Instantaneous unit hydrograph derivation by harmonic

analysis'IASH (Helsinki) Pub No 51

(3) Kraijenhof van de Leur, D.A., "A study of non-steady ground water flow with
special reference dto a reservoir coefficient"

De Ingenieur, 70(19 1 (193 81.

(4) Venetis C. "Estimating infiltration and/or the parameters of unconfined

aquifers from ground water level observations" Jour Hyd. 12(1971)

 DONNEES INADEQUATES ET MODELES MATHEMATIQUES
DE LA POLLUTION EN RIVIERE

Par J.BERNIER

Laboratoire National d'Hydraulique

CHATOU - France

ABSTRACT

The value o f I'inadequate1l information must be judged

relatively t o the mathematical tools used and the practical problem
t o be solved, Concerning river pollution where t h e problem is to
design projects as sewage treatment plants for instance, the
limitation o f the usual Information collected in situ is shown,
These limitations do n o appear w i t h the standard math.ematjca1 model.
Taking in account of more realistic stochastic model allows us to
measure the value of t h i s information and to design experiment for
collecting acceptable data,

RESUME

La valeur de l'information inadequate doit être jugee en

fonction des problèmes à rbsoudre et des outils mathématiques utili-
sês pour cette résolution, En matière de pollution en rivière où il
s'agit de dêfinir les caractdristiques des moyens de lutte comme
celles des stations dlspuration par exemple, on montre les limita-
tions de l'information usuellement recueillie in situ, Ces limita-
tions n'apparaissent pas avec les modèles mathgrnatiques standard,
La prise e n compte de modèles stochastiques plus réalistes permet
de mesurer la valeur d e cette information et de definir les condi-
tions de collecte de données acceptables,
51 4

I - INTRODUCTION

En matière d'aménagement des ressources en eau, l'insuffisance

des données est souvent le premier écueil auquel on se heurte. Cependant
pour mieux apprécier la validité et la précision des réponses aux questions
posées à l'hydrologue ou l'ingénieur, il faut noter que cette insuffisance
de données n'est en fait que le reflet de l'inadéquation des inhthodes uti-
lisées pour résoudre les problèmes. Ces méthodes doivent etre adaptées >
la nature de l'information disponible ou 2 recueillir. Certes dans bien
des cas les données disponibles doivent être coiilplétées mais l'organisa-
tion de la collecte des données complémentaires, le choix des conditions
opératoires de mesures ne peuvent valablement être &finis qu'en fonction
des méthodes iiiobilisant l'inforination recueillie. Dans ce contexte, certai-
nes iitéthodes usuelles sont particulièreiiient inadéquates. On peut en trouver
des exemples dans le domaine de la pollution en rivière notamment dans
l'étude du inouverrient et des réactions auxquels sont soumises des matières
polluantes en riviere 2 l'aval d'un point de rejet en vue d'apprécier la
capacité d'autoépuration de la rivière compte tenu de ce rejet. Nous trai-
terons ici du seul problème de la pollution biochiinique'caractbris6e par
le bilan d'oxygène.

II - LA METHODE USUELLE

La méthode classique utilise le modèle de Streeter et Phelps

décrivant le bilan dynamique d'oxygène sous la forme de deux équations
différentielles (voir la liste de notations en fin de note).

J
 515

Mises sous forme intégrale et en faisant apparaître l'abscisse

longitudinale x prise le long de la rivière et liée au teiiips d'écouìe-
2
nient t et à la vitesse moyenne du courant u pa; t U
, les équa- =-
tions donnent :
X
-K3 u
L(x) = Lo e

O Ù d (x)
(x) =ao
e

est le déficit en oxygène :3 = Cs

~

-
~
- K

C.
(
-X
2 u -e e
'J
- Kg U)

Ainsi peut-on mesurer l'incidence d'un rejet (spécifiant les

concentrations initiales d'oxygène Co et de demande biologique en
oxygène L )sur l'autoépuration à l'aval.

La mise en oeuvre courante de ce iiiodèle demande l'estimation

préalable des coefficienb de reoxygénation K2 et de biodégradation K1 et K3.
On utilise généraleinent des formules empiriques ou quelques observations
recueillies dans le tronçon de rivière étudié. Les multiples formules
empiriques disponibles établies en laboratoire ou sur des rivières de
caractéristiques biochimiques particulieres présentent des résultats
extre'mement dispersés difficilement extrapolables hors des limites du
doniaine où elles ont été établies. I1 reste l'inforniation recueillie
dans chaque cas d'espèce. A la limite Kg et K2 peuvent ëtre calculés
par l'intermédiaire des formules (2) en fonction d'un seul couple d'ob-
servations amont (Lo, Co) et d'un seul couple aval ( L(x), C(x)). Une
telle procédure, trop souvent utilisée pratiquement, est justifiée dans
le contexte du modèle déterministe strict décrit par (2) mais ce carac-
tère déterministe est extrêmement fallacieux c o m e nous allons le voir.
Par ailleurs il importe de se soucier de la cohérence spatiale des mesures,
le décalage des époques d'observations a l'amont et à l'aval devraient
tenir compte du temps d'écoulement t . En pratique cette condition iinpé-
rative est rarement respectée et la non prise en compte de la menie masse
d'eau 2 l'amont et à l'aval entrarne une dispersion notable des observations
et des estimations erronées de KI, K2 et Ka.
516

III - INSUFFISANCES DU MODELE DETERMINISTE DE STREETER ET PHELPS

Le modèle (i), (2) schéiuatise l'hydraulique de l'écoulement :

celui-ci est supposé permanent, uniforme ; de plus on néglige la disper-
sion de polluants dissous ou en suspension dans l'eau imputable au phéno-
mène de diffusion turbulente et aux fluctuations de vitesse dans chaque
section. Cependant l'effet de cette dispersion est surtout notable en
régime transitoire et devient négligeable en régime de pollution permanent
ou lorsque cette pollution (caractérisée par L et C I présente une évo-
lution lente, d m s le cas de rivière à coefficient de dispersion faible
(cf. [i] et [2] )
.

L'inadéquation en nature du modèle de Streeter et Phelps provient

surtout des hypothèses caractérisant les phénomènes biochimiques :

- prise en compte de la seule phase carbonée des réactions de

dégradation des matières organiques (non prise en compte de
la phase azotée) ;

- non prise en compte des effets de la photosynthèse de la respi-

ration des boues de fonds, de la sédimentation des matières
polluantes etc ...
Certaines tentatives [a] ont été faites pour inventorier et
modeliser
. .
plus completement les phénomènes mais la multiplicité des para-
mètres et les difficultés pratiques d'estimation de ces paraniètres rendent
illusoire l'apparente précision qui semblerait résulter d'un inventaire
exhaustif des mécanismes biologiques et physico-chimiques.

Par ailleurs si les concentrations en oxygène dissous peuvent

être mesurées in situ avec une précision acceptable, il n'en est pas de
même de la demande biologique en oxygène qui n'est pas estimée directement
mais seulement par 1 l intermédiaire d'un test chiriiique, la DB05, effectué
au laboratoire sur des échantillons prélevés en rivière. Des essais ont
montré que l'imprécision de ce test est notable et que la chaîne complexe
des conditions opératoires, depuis le prélèvement en riviere jusqu'a
1 'analyse chimique en laboratoire introduit des erreurs systématiques et
aléatoires importantes. On ne peut donc considérer cette DBO comme une
5
mesure exacte de la consommation d'oxygène in situ mais comme un index
représentatif de cette consommation en plus ou moins bonne comélation
statistique avec elle.
 517

IV - UN MODELE STOCHASTIQUE

I1 est classique en statistique de prendre en compte globalement

l'ensemble des phénomènes négligés dans un modèle déterministe schématique
sous foriiie determes d'erreurs aléatoires. Cette conception permet d'intro-
duire la souplesse nécessaire à une bonne adéquation du modèle aux données
d'observations en nature. Les équations (1) sont remplacées pour un sys-
teme différentiel stochastique :

Les paramètres pl et p2 représentent les irioyennes, c'est-à-dire la

2
part systématique. des erreurs dues aux phénoiiiènes négligési o12 et u 2
représentent les variances des termes d'erreurs. E, et .C2 sont alors
les erreurs centrées réduites (de moyenne nulle et de variances égales à 1).

Dans ce contexte aléatoire, les paramètres K1, K2, K3 peuvent

être interpréter coime les coefficients de la régression statistique des
variations de concentrations en fonction des grandeurs L e t d et ils
ont une signification statistique plutôt que physique. On peut alors rem-
placer l'index DB05 par tout autre index qui soit en corrélation avec
la demande en oxygène (par exemple : demande chiiiiique en oxygène, carbone
organique total, etc... indexes qui sont plus aisément mesurables que la
DB05). I1 est possible d'intégrer le système ( 3 ) (cf. [41 1. En adinettant
des conditions initiales fixées
LO
et g o au point origine x = O, on
peut montrer que L(t) e t 3 (t) sont des variables aléatoires quasi-
gaussiennes dont les espérances niathématiques et variances sont :

- K3 t
E (L) = -
P2 + (Lo - -
K2
P2
)
Kg
e (4)

E ( d ) = - tPi
- P 2 pci +- 1' (Lo - -P2
)I e
-Kpt
K1 K2 Kg - K3-K2 K2

K1 P2 - L o ) - K3t
+-í- e (5)
K3-K2 K3
 518

- 2 K3t
v22 ( i -e 1
Var (LI = K3

2 2 -2 K,t
Var ($1 = ( u1 +
2 u2
2
t ui r, (i -e 1
(K~-K~) K3-K2 K2

(i -e- 2 K t
3 ) K1 ( u u r+-
2
=2
Kg K3-K2 1 2 K3-K2

(i -e 1 (7)
Kg + K2

formules dans lesquelles r est le coefficient de corrélation entre les

erreurs Cl et E,.

v- LE MODELE STOCHASTIQUE AVEC ERREURS DE MESURES SUR L et d

Connie nous l'avons déjà souligné, 1'inadéquation du iiiodèle aux

données in situ est liée également aux erreurs de mesures importantes sur
ces données. Pour analyser plus complèteinent le problèiiie, il importe de
préciser les conditions d'observations. Nous ne traiterons ici que de la
méthode usuelle OU l'on observe deux points amont x et aval distants
de x = - xcI.
Compte tenu des foririules (4) à (7) le modèle intégré peut être
écrit :

en regroupant sous forine des constantes a et b les termes indépendants

des conditions initiales dans les espérances inathématiques et sous forme
des constantes p1, p 2 , p 3 les coefficients de Lo et do.Les variables
aléatoires d'écart EL et I
!& ont L
alors des variances u
et u 2 données
par les formules (6) et (7) et qui sont donc indépendantes de ces conditions
 519

initiales. En fait ce que l'on observe n'est pas directement L ou 3

m i s deux grandeurs X et Y telles que :

X = L + 7)
1

Y = a+ 7)
2

où Il e; 7)2 sont les erreurs de mesures aléatoires de variances respec-

tives VL et Va2 . L'estiiriation in situ des parainetres du modèle stochas-
tique et notanunent des vitesses de reoxygénation et de biodégradation
K2 '
K1 et K3 , demande la connaissance préalable des variances d'erreurs de
mesures. De façon précise on supposera qu'ont été obtenus n ensembles
de grandeurs observables (Xoy X1, Yo, Y 1) aux deux points amont et aval
du tronçon de rivière considérée. Les paramètres statistiques de ces cou-
_ - - - 2 2 2 2
ples (moyennes Xo, XI' Yo¶ YI , variances S x0, S x13 S yo, S y1 , et
covariances S&xl3 Sx0y1, etc ...)permettent l'estimation des coefficients
du inodele au moyen des relations :

P, = sxlxo
2 2
SX, - VL

(10)

- -
a = X1 - p 3 Xo (12)
 520

VI - LE PROBLEME DES ERREURS D'ECHANTILLONNAGE

Dans la plupartdes cas pratiques l'estimation des paramètres du

modèle d'oxygène ne peut être effectuée que sur un nombre n de répéti-
tions d'observations assez liiriité. I1 importe alors de iiiesurer la précision
de ces estimations par leurs variances d'échantillonnage. Nous ne pouvons
ici développer l'ensemble des formules, nous renvoyons le lecteur à [2]
pour un aperçu sur le problème. En expriniant les divers paramètres comme
2
des fonctions des variances et covariances S2 S yl, Sxoxl, etc ...
XO
les formules précédentes peuvent permettre le calcul approché de ces va-
riances d'échantillonnages à partir de celles des variances et covariances
estimées (cf. [51
). En ce qui concerne notaniiiient les paramètres pi qui
déterminent les vitesses des réactions biochiiriiques , on aura des formules
de la foriiie :

VI1 - VALEUR DE L'INFORMATION RECUEILLIE EN NATURE

Pour la suite de la discussion il est conimode d'appeler variances

2
d'erreurs d'adéquation du inodele les paramètres u L , mg2 car leur valeur
est liée à l'importance de l'explication des variations d'oxygène dissous
par les paramètres C, DB05 ... pris en compte. L'interprétation des
variances d'échantillonnages, perinet de préciser le noiiibre d'observations n
nécessaires à l'obtention d'une précision d'estiniation donnée ; on pourra
observer généralement la loi générale de l'augmentation de n en fonction :

- des valeurs croissantes de l'erreur d'adéquation .

- des valeurs croissantes de l'erreur de mesure.

Quels que soient les paramètres de pollution pris en compte, quelles que
soient les procédures opératoires de mesures, il restera toujours des
erreurs d'adéquation et de mesure irréductibles. Dans de telles circons-
tances, on ne peut utiliser les procédures classiques d'estimation des
modèles d'oxygène supposés déterministes et qui n'utilisent que des infor-
mations trop partielles. réduites trop souvent 2 un unique ensemble des
4 valeurs Xo, Yo, XI, Y1. I1 est absolument indispensable de faire des
mesures répétitives en nombre n suffisant. Le modèle stochastique est
 521

alors un guide précieux pour la planification de la collecte de cette

information et des procédures opératoires de mesures.

Placé devant un problème de décision, le gestionnaire de la qua-

lité de l'eau d'une rivière aura donc a sa disposition des observations
cohérentes recueillies in situ ; mais ce n'est pas la seule source d'in-
formations disponible. Le gestionnaire dispose également de données plus
ou moins qualitatives sur les vitesses de réactions, de forinules semi-
empiriques diverses [6 1 dont la dispersion des résultats est telle qu'elle
ne peut. donner que des ordres de grandeurs assez grossiers. Une telle
information est cependant précieuse si elle permet de réduire le nornbre
d'observations in situ. La disposition d'un modèle stochastique permet
l'utilisation des méthodes bayésiennes [7] , [a] dont le but est l'incor-
poration des inforiiiations de diverses origines dans un modèle quantifié.
Davis, Kisiel et Duckstein [E] ont montré tout l'intérêt de ces techniques
appliquées ?i l'étude des risques,associées aux décisions en matière de
gestion des ressources en eau et au calcul de la valeur économique de
l'information hydrologique qui tend à réduire ces risques. Ia prise en
compte d'un modèle stochastique est la première étape de l'approche déci-
sionnelle dans les problèmes de qualité de l'eau.
522

N O T A T I O N S

c : concentration en oxygène dissous, moyenne par section de rivière

concentration en oxygène dissous 2 la saturation
Cs -C déficit en oxygène
DB05 : demande biologique en oxygène mesurée au laboratoire sur 5 jours
à la température de 2OoC sur un échantillon supposé représentatif
de la rivière
K1 : coefficient de consoinmation d'oxygène dans le modèle de Streeter -
Phelps
coefficient de réoxygénation
K2 :
K3 : coefficient de dégradation de la D B O restante L
L : deniande biologique en oxygène restante
x : abscisse longitudinale de la rivière
u : vitesse moyenne de l'écoulement.
 523

B I B L I O G R A P H I E

[i ] W.E. DOBBINS : BOD and Oxygen relationships in streams

Froc. ASCE Sanit Div. S A 3 - 1964

[2] J. BERNIER - P. LENCIONI : Utilisation d'un modèle stochastique pour

organiser 13 collecte in situ des données de qualité
de l'eau d'une rivière - 15ème Congrès de 1'A.I.R.H.
Ictainboul 1973.

[3] D. LEFORT : Modèles mathérriatiques de pollution en riviere -

La Houille Blanche - nuiiiéro spécial 8/1971

[4] D.R. COX - H.D. MILLER : The theory of stochastic processes

Methuen - 1965
[5 ] T.W. ANDERSON : An introduction to multivariate statistical analysis
Wiley - 1958

[6]
M. NEGULESCU - V. ROJANSKI : Recent Research to determine reaeration
coefficient - Water Research - Vol 3 no 3 - 1969
[71 J. BERNIER : Les méthodes bayésiennes en hydrologie statistique.
Froc. Intern. Hydrology Symp. Colorado State Univer-
sity - Fort Collins - 1967
[8 1 D.R. DAVIS - C.C. KISIEL - L. DUCKSTEIN : Bayesian Decision Theory
applied to design in Hydrology - Water Resources
Research - Vol 8 no 1 - 1972.
 "REGIONAL GROUNDWATER RECHARGE ESTIMATES VIA METEOROLOGICAL DATA"

SAMUEL P.COOK AND SAMUEL G,MBURU

ABSTRACT

In arid regions the planning of agricultural development

requires estimates of the availability o f groundwater, Adequate
detailed data bases are unlikely to exist in most areas o f interest
in developing countries, We have attempted to compute the
groundwater recharge potential for East Africa using primarily the
available meteorological data, The procedure has been to generate
a synthetic year by averaging the meteorological data for each
month at each meteorological site over a period of years, Then from
the monthly precipitation, the estimated evapotranspiration and the
estimated run off is subtracted, This computation proceeds according
to an assumed soil moisture storage and transport model whose
throughput constitutes the potential groundwater recharge, Contour
lines o f this cuantity are plotted and these can serve as a guide
to rural planners for optimizing the selection of sites for new
development,

RESUME

Aux regions arides quand on fait un plan du development agri-

cole il faut evaluer l'eau souterraine desponible, I 1 est tres peu
probably qu'on trouve les donnees de base assez detaillees dans la
plupart des regions en observation aux pays que son en train de se
developper. On a essayé de computer le potentiel de la recharge des
?aux souterraines pour les pays en Afrique de l'Est en utilisont,
?rincipalement, les données meteorologique disponibles, On a produit
sne ann&e des donnges synthetìques en faìsant l a moyenne les
ionnêes meteorologiqueo pour chaque mois a chaque h s t a l l a t l o n meteo-
nologique pendant une periode des annêes, P u i s on a soustrait de l a
>recipitation mensuelle, l'evaluation de ltevapotranspirat2on et de
L'ecoulement total, Cette computation continue suivant un modele
;uppose de l a capacité de l'eau et du transport de l'eau dans le sol
lui evalue le potentiel de la recharge de l'eau souterraine. Les
:ourbes de niveau de cette quantite sont tracées et les organisateurs
lu developpement rural peuvent s'en regler pour optimiser la
;election des situations pour des projets neufs.
526

At the East African Agriculture and Forestry Research

Organization we are interested in estimating regional groundwater
recharge rates to provide basia information for agricultural
planning. Our approach is based on a water balance calculation. The
gmundwater recharge is the residual which reinains after subtracting
from the precipitation the losses due to evaporation and surface and
subsurface run-off. The total precipitation can be computed reasonably
well from the rainfall records. The remaining terms can be estimated.

T. Woodhead, 1. Dagg and D.A. Rijks (1, 2, 3, 4) have

studied extensively the computation of the Penman potential
evapotranspiration from the data sources in Eaet Africa. The actual
evaporative losses may be estimated with the use of a physioal model
of the soil moisture storage and transport system. Por a regional
computation in a predominently arid region the net surface and
subsurface run-off may be taken a8 aero to yield an upper bound on
the potential groundwater recharge. The accuraoy of the final result
will depend on the ohoice o f the soil moisture storage and transport
model. This must represent the average regional response to the
stimuli of rain, wind and sun.
Water Balanoe Oalculation 0.f Groundwater Beoharge

GWFli Potential Groundwater Resharge

Pa Precipitation
Eo: Penman Potential Evapotranspiration
E : Actual Evapotranspiration
Q: Soil Moisture Storage
AQ; Change in S o i l Moisture Storage
RO: Rn-off
GWR = P -E -
-.U) R O
In order to compute the actual evapotranspiration from the
Penman potential evapotranepiration the dynamics of a soil moisture
model m e invoked. Many such models arc possible and future studies
may improve this phase of the work. For the present computation a
nbucket model" has been chosen. This model allemes that moisture
stored in the soil root sone is freely available for transpiration
up to the Penman potential Eo demand. If the root zone soil moisture
is exhausted no further evapotranspiration takes place regardless of
the Penman Eo demand. Furthermore, lhe root abne has a finite
capacity, Qo, for moisture storage. In the event the monthly
precipitation minus the monthly Penman Eo exceeds Qo, downward
percolation o f the excess moisture takes place and constitutes the
potential groundwater recharge.
 527

"Bucket Model" of Soil Moisture Stcrage ana Transport

1. Root zone has a maximum moisture storage capacity, Qo.

2. If Q S Q o , groundwater recharge i? cil.
3. If monthly moisture input plus storage exceeds Qo, the
excess is potential groundwater rechzrge.

W e have made use of meteorological data collected by

T. Woodhead, M. Dagg, and D.A. Rijks (iq 2, 3, 4). This data is
based on 80 stations in Kenya, 50 in Tanzaniz and 17 in Uganda, at
ozly a few stations were oomplete records o f precipitation, wind run,
and insclation available. Several methods were devised by Wlcndhead to
fill in the blanks. In order to eompute the Penman potential
evapotranspiration, Eo, according to the method of McCullooh (5) the
following inputs are needed:

R: insclation in cakories/cm 2/day

n/N: =?io 3f observed to maximum possible number of daily
smshine hours.

Ta: zverage screened ambient temperature 'C.

o
Td: mean deily temperature of den point C.

U: win& run in miles per day at 2 meter elevation.

Eased cn fifteen stations a linear regression between R and

n/N has been derived (6, 7). This regressior, was used to derive one
of thene quantities when the other was available from the meteorological
recnrds. When neither R nor n/N were recorded use was made of a,
rehtionship esteblished (1, 2) between the monthly mean of daily
sucshim duraticn and the tot21 cloud amount. Estimates of total
cloud smount are made at most civil airfields. When records of wind
run were not available use was made of e relation established
betaeen Beaufort scale assessments of wind velocity and wind run (3).

After these procedures had been applied to complete the

recr-ds they were processed to construct a synthetic year. Por each
site sverages over the length of the record were made for each given
mocth of the year. Monthly values of precipitation and Penman
potential evapotranspiration were obtained. A t each site an estimate
was also made of the model parameter, Qo, the maximum m i l moisture
storage capacity in the root zone.

The soil moisture storage and transport system is conceived

o f E? a dynamic system whose output is the downward percolating groundwater
recharge which is determined jointly by the system driving funation
3n3 the system physical parameters. The present output depends on the
whcle past behavior of the input. If the syetem ie 3inesr, the system
oiityut is the convolution of the system unit impulse rnspcnse and
tke system c?ri.ving function. If the system is non-licesr, given the
528

input we can compute the output ueing the system parameters.

For our bucket model system we proceeded as follows.

If the Penman Eo exceeded the precipitation for several months we
tentatively assumed that the soil moisture was totally depleted at the
end of the dry season. Starting at that month the bookeeping was
begun arid carried out each month for the duration of the synthetic
year. On the other hand, if the precipitation exceeded the Penman
Eo for most o f the year the assumption was made that the soil was
saturated with moisture at the end of the wettest period. Then the
oookeeping was completed for the synthetic year. One of these
assumptions always led to a consiatent sat of bookeeping entries.
The monthly groundwater recharge was summed at each site over all
mrnths of the synthetic year. These totals were then noted on a
m a p and contour lines of equal groundwater recharge were interpolated.

After obtaining the meteorological data the single model

payameter 20 determines the reault. Therefore, the sensitivity of
the cont Ars to a variation of Qo ia of interest. It is evident that
If the soil moisture is not completely depleted at some time during
7-e gear, a change in Qo alone will not affect the groundwater recharge.
For this situation, only the moisture in permanent storage will be
changed. On the other hand, if the sril moiature is completely
exhausted at one time during the year, a change in Qo will change
the throughput by an equal and opposite amount. Therefore, in arid
regions a low soil moisture storage capacity favors groundwater
recharge. Low storage capacity implies that a short intense rain
will rapidly fill up the soil moisture reservoir and the excess will
quickly percolate downward beyond the root zone to the subsurface
storage aquifer. In arid regions vegetative cover is generally less,
which re8'uces the losses due to tranepiration. Deep rooted
vegetation wnuld negata this advantage.

In order to investigate the eensitivity of the potential

groundwater reoharge contours to a change in the value of QG, the
ac.mputations were carried out for two choices of this parameter at
each site.

Oritiaue of the Method

le hare attempted to obtain same idea of the regional

potential groundwater recharge ratea using the available data, mainly
meteorological records. By the use of a simple aoil moisture
storage and transport model we estimate the actual evapotranspiration
from the Penman potential evapotranspiration. The residue from the
precipitat irn after subtracting the evapotranspiration and the
increment tc soil moisture storage we identify as the deep percelation
or potential groundwater recharge. The model we used for the soil
mt-isture is a one parameter model and we have studied the effect on
tSe throughpiit of the choice of this parameter.
 529

Several improvements are poserible in the treatment. A study

co-ild be carried out to improve the accurary of the model. It is easy
tri devise multi-psrameter models. These could be compared to field
me-rsurements. Another refinement would incorporate sdditional input
data. Neutron soil moisture probes are now fairly widely available.
Fer future work such data should be incorporated in the computation.
The use of this additional input can lighten the burden of the model
in the determination of the actual evapotranspiration.

The neutron moisture probe data could be used in the

following way which modifies the present soil moisture model. Soil
misture profiles could be taken at monthly or weekly intervals to a
depth of five meters. The total soil moisture to a fixed depth will
??e totaled fnr each measurement. During the interval between
measurements the precipitation and the Penman potential
zvspotranspiration will be totaled. If during that interval there
i? soil moist*ire storage exceeding the wilting point in the roet zone
than the a m e l evapctranspiration will be taken as the Penman Eo. If
w t , the actual evap6transpiration will be taken as zero.
With this accounting procedure we could compute the sum of
the deep percolation and the difference between surface and subsurface
run on and runoff. If these last categories are in approximate
halance, then we have the deep percolation or potential groundwater
recharge as the residual, From measurements at a network of sites the
regional maps msy be constructed.

The authors wish to thank the Director o f EBBFRO for

aermission to present this paper at the Symposium on the Design 00
Tater Resources Projects with Inadequate Data, Madrid, June 1973.

REFERENCES
I.. Woodhead, T. (1966). Empirical relations between cloud
amount, insolation and sunshine duration in East Africa:
i, E. Afr. Agric. For. J., 2,pp211.

2. Woodhead, T. (1967) Empirical relations between cloud

amount, insolation and sunshine duration jr? East Africa:
II, E. Afr. Agric. For. J., 2,pp474.
3. Woodhead, T. (1970) mapping potential evaporaticn for
tropical East Africa; the accuracy of Penmen estimates irem
indirect assessments of radiation and wind speed, Proc.
Reuding Symposium World WgletgFi,B&lanCe
. I._ I - I. A.C.H.

Dagq, M. and Woodhead, T. and Rijks, D.A., Evapcmtirr

jn Enst Africa, nul. I.A.S.H., XV, 1, pp61.
530

.
..
c McCulloch, J .S .O. (1965), Tables for the rapid i:omputE+,i.on
of the Penman estimate of evaporation, E. Aï'?. AgTjc. For.
J., 22, pp.286.
6. Woodhead, T. (1968), Studies of Potential Evaporaticn in
Kenya, Government nf K e n p , Nairobi.

7. Woodhead, T. (196R), Studies of Potertial Evapuration i.c

Tarizania, Dar es S a h a m .
 A RAINFALL-RUNOFF MODEL BASED O N T H E W A T E R H E D STREAM N E T W O R K
by J.W. Delleur and M.T. Lee*
School o f Civil E n g i n e e r i n g
Purdue University '

West Lafayette, Indiana 47907, USA

A B S TRAC T
P h y s i c a l m o d e l s o f t h e rainfall-runoff p r o c e s s a r e better
suited t h a n either s t o c h a s t i c or black box m o d e l s f o r a r e a s w i t h
limited data, T h e m o d e l parameters m u s t have a p h y s i c a l s i g n i f i -
cance, be convenient t o obtain and t h e i r n u m b e r should be s m a l l ,
T h e framework of a m o d e l meeting t h e s e o b j e c t i v e s is proposed a n d
is based primarily o n t h e g e o m o r p h o l o g i c c h a r a c t e r i s t i c s o f t h e
stream network o b t a i n a b l e from m a p s or f r o m a e r i a l photographs.
T h e r e is a n a l y t i c a l a n d experimental evidence t h a t h y d r o g r a p h s a r e
dominated by direct r u n o f f from very short overland flow p a t h s
from precipitation o n transient, n e a r channel wetlands. T h i s
wetland a r e a is dynamic i n t h e s e n s e that it varies i n t e r m s o f t h e
history o f t h e excess o f the p r e c i p i t a t i o n o v e r t h e "B" h o r i z o n
permeability, The distribution o f t h e dynamic c o n t r i b u t i n g a r e a
along the m a i n stream is obtained under t h e a s s u m p t i o n s that t h e
velocity o f flow along t h e stream n e t w o r k is uniform, that t h e
d r a i n a g e density is a constant within a g i v e n watershed and that t h e
first o r d e r s t r e a m s are uniformly distributed i n t h e basin, T h e
r u n o f f from t h e d y n a m i c contributing a r e a is t h e n r o u t e d t h r o u g h t h e
synthesized stream network t o obtain t h e direct r u n o f f a t t h e b a s i n
outlet,

RESUME

Les m o d è l e s physiques des t r a n s f e r t s pluies-débits s'adaptent

m i e u x q u e les m o d è l e s stochatiques o u q u e l e s "boîtes noires" a u x
r é g i o n s o ù les d o n n d e s sont l i m i t é e s , Les paramêtres de ces m o d e l e s
doivent être p o u r v u s d'un s e n s physique, f a c i l e s 2 o b t e n i r e t l e u r
nombre doit être petit, Le c a d ~ ed u m o d è l e p r o p o s é s e c o n f o r m e 2 c e s
o b j e c t i f s et est basé principalement s u r l e s c a r a c t é r i s t i q u e s g ê o -
morphologiques d u r l s e a a fluvial q u e l'on peut o b t e n i r de c a r t e s ou
de photographies a é r i e n n e s , I1 a dtb d é m o n t r ê a n a l y t i q u e m e n t et
expérimentalement q u e l e s hydrogrammes sont e n g e n e r a l d o m i n b s p a r
l e ruisellement s u r de petits p a r c o u r s s i t u ê s dans l e s z o n e s m o u i -
liges près des c o u r s d'eau, Ces zones m o u i l l ê e s s o n d y n a m i q u e s d a n s
l e s e n s qu'elles v a r i e n t pendant la p l u i e et a v e c l a s a i s o n , U n m o -
dèle mathêmat2que d e ces z o n e s est f o r m u l e e n f o n c t i o n d e I f h i s t o i r e
de l a p r é c i p i t a t i o n excédant la permsabilité de l'horizon iiB't. La
distribution de ces z o n e s l e l o n g du r d s e a u f l u v i a l est o b t e n u e en
supposant que l a vitesse de l'écoulement est uniforme d a n s l e r é s e a u
fluvial, que l a densitd de drainage est c o n s t a n t e pour l e b a s s i n
donne, et q u e l e s cours d'eau du p r e m i e r o r d r e sont u n i f o r m d m e n t
d i s t r l b u h d a n s le bassi'n, Les r u i s s e l l e m e n t s d e s z o n e s d y n a m i q u e s
sont achemint% a u t r a v e r s d'une s y n t h è s e du r d s e a u f l u v i a l p o u r o b t e -
nir i'ëcouïement à l'exutoire.

* Current Addres,s: D e p t , o f A g r i c u l t u r a l Economics, Univ. o f

Illinois, Urbana, 1llinoi.sy 61801 y USA,
 532

Stochastic models for the generation Of river flow sequences require long
historical time series for the appropriate calibration ofthe parameters. In
many parts of the world, actual streamflow records are not sufflciently long to
attempt to deflne an elementary model such as a flrst order Markov process for
annual streamflow series. According to Rodriguez-Xturbe [i] for serles shorter
than 40 years the error in estimating the annual man might run from 2% to 20%.
for the variance from 15% to 60%. and for the rank one serial correlation it
might be as high as 200%. It may be fìatile to attempt to develop generating mo-
dels which preserve parameters, the estimation of which carries such an uncer-
tainty. The formulation of mnthly models requires shorter records, but with a
record of i5 years, the error in the rank one serial correlation coefficient is
still of the order of 40%. Btochastic linear models of the rainfall-runoff pro-
cess likewise require long time serles of both the rainfall and runoff for their
calibration. It would, therefore, appear that for regions with inadequate data,
one may have to resort to deterministic models. At this point, the choice may
be between a %lack box" type of model and a physical model. Black box models
ceanot be transferred from one location to another as the meaning of the peu-8-
meters in terma of their representation of the components of the hydrologic
cycle is usually undeflned. The proper choice appears to be a physical model
which requires a small number of easily identifiable parameters, or a model
based on data w-hich can be obtained in a relatively short time, perhaps by new
techniques.

These physical models could conceivably be formulate&, by making use of in-

formation that is becoming available through remote sensing from airCrart and
from satellites. By means of these new techniques, large areas CBP be observed
and analyzed in a short time, end require a small amount of observations on the
ground. The recent developments in remote seneing technology thue seem to point
to a new direction for hydrologic investigations in -(LB with inadequate data.

Remote sensing from aircraft or from satellite is best applied to observing

or monitoring fairly large area and thus lende itself to the hydrologic studies
of complete watersheds. Images taken at different times cm show changes In the
watershed, such as variations ia the land w e . The potential of remote sensing
for water resources stubies has been discussed by Kiefer and Scherz [2], but the
principal application of remote sensing to hydrology has been through aerial
photography.

The eye can see light from about .4 to .7 microns, but photographscan sense
from about 0.3 to 1.0 microns, thus extending the range to lower and higher wave
lengths. Color end color infrared photography have been used with great success
in forestry .and in agricultural crop identification. [3] Themai scanning op-
erates in the heat emission part o? the energy spectrum in the w a n length riPr
3 to 20 mincrons. Pluhowski [4] shoved that with Infra-red Imagery in the 8 to
14 micron range, it is possible to discern thermal contrasts of '1 or .C'2 Thin
technique enables the hyarolo,5lSt to detect areas of &K>inid water dlacharge mad
to identify circulation patternr in large vater bodies.
 533

More advanced techniques include mUitiSpeCtra1 scanning and side looking

radar. Multispectral scanners produce as many as 20 separate images in wave
lengths ranging from the reflected infrared region to the ultraviolet region.
These images may then be analyzed by means of computer data processing programs
which classi* the surface materials. This classification is accomplished by
separating materials in a known area according to their spectral response char-
acteristics and then applying these criteria to unknown areas. [3] The side
looking airborne radar can operate through dense cloud covers. It has been used
in mapping southeastern Pan- and northwestern Columbia, which could not be
mapped by conventional aerial photography becauee of the cloud cover. As an ex-
ample, Weaver [5] cites that the meandering pattern of the hiira river was re-
vealed by this technique.

Black and white, color and color infrared photography combined can be used
to delineate water bodies, rivers and streams, the drainage structure of water-
sheds, to give indications on the underlying geology and on the soil types of
the region. Waltz and Myers [6]have shown that there exists a significant cor-
relation between the optical density measured from an aerial film and soil water
content measured by neutron probes and also between ground water temperatures as
measured through infrared thermal scanner and the soil water content of fallow
or bare soil. Zachary et.al. [7] has applied multispectral. remote sensing to
soil survey research in Indiana. These techniques may also be used for enalysie
of water quality and for monitoring water pollution. [a] A general review of
the application of remote sensing in the management of earth resources has been
prepared by Colwell [9].

It appears that at present, black and white, color and infrared aerial
color photography, can be used to obtain the basic information regarding stream
networks, water bodies, main geologic and soil features needed in hydrologic in-
vestigations. It also appears that in the near future, more dependable informa-
tion on soil water will become available through remote sensing. The remote
sensing techniques thus appear to be of particular interest in areas with inade-
quate data, as a substantial area can be mapped in a relatively short time with
a minimum of ground observation.

MODEL FRAMEWORK

It is the purpose of this paper to explore the feasibility of developing

rainfall-ninofi models based primarily on information that can be obtained from
remote sensing aerial photography and to establish a framework for such models.
The simpler observations obtainable f r m the aerial photography being the plan
form of the stream network. the topography and the soil type. the proposed model
is based on these three types of information and particularly on the stream net-
work s

Geomorphologists have developed parameters which describe the topology, the

structure, the planform and the relief of stream networks. [lo] Some of these
parameters can be used ae indices of the hydrologic behavior of the basins since
scmral characterietics of the hydrograph depend upon the efficiency of the
534

drainage networks. For example, the bifurcation ratio (ratio of number of streem
segments of one order to number of stream segments of next higher order) is an
important control over the peakedness of the runoff hydrograph. Another geomor-
phologic parameter which affects the m o f f pattern is the drainage density ( s m -
mation of stream lengths divided by basin area) which is approximately one halr
of the reciprocal of the overland flow length. A high drainage density indi-
cates a rapid removal of the surface runoff, a decrease in the lag time and an
increase in the peak of the hydrograph.

A model based on the stream network also lends itself to the application or
the dynamic source area concept rather than the application of classical Horton
infiltration theory for the purpose of estimating the runoff-producing-rainfall.
Freeze [li] has shown theoretically that on concave slopes with lower permeabil-
ities and on all convex slopes, hydrographs are dominated by direct runoff with
a very short overland flow path from precipitation on transient, near channel
wetlands which form the variable response area. D u M e and Black [12] reported
that the area contributing to the overland flow ie dynamic in the sense that it
varies seasonally and throughout a storm. Nutter and Hewlett 1131 have depicted
the growth of the source area during a storm from areas adjacent to the lover
order streems and gradual4 expanding to the main stream in one direction and to
efflmeral stream in the other. It seems logical to assume that the response
area will depend on the soil type adjacent to the stream and on the antecedent
rainfall.

In view of the complexity that would result from estimating and routing the
runoff in each tributary, it is proposed to synthesize the stream network by
folding it along the main stream in a manner similar to that m e a by Lkwge [i41
with the time-area diagram. Several routing procedures could be used, the %om-
plete linear routing" method of Dooge and Harley [i51 was used because of its
superior accuracy among other linearand emperical methods. It is in the appli-
cation of the routing procedure that the slope of the main stream plays a major
role. For full details the reader is directed to ref. 17.

iVRMULàTION OF THE MDDEL

The model of the contributing area A(iAt) is expressed by the relationship

i-1
- B]At - B]At
A(iAt) = Ao i;,
I
[R(kAt)
T
-
+ [R(iAt)
(1)

i k=O
[R(kAt) B]At

where A(iAt) is the contributing (response) a n a at time iAt

1
R(kAt) is the rainfall intensity at time kAt
B is the "B" horizon permeability
D is the fraction of the antecedent rainfall contributing to the
response area
N is a parameter
T is the total nuniber of sampling points of the runoff -&-ogreph
 535

k is an index to count the time of antecedent rainfall excess,

k * i
i is an index indicating the current time

Equation (1) is subject to the conetraint that the continuity equation must
be satisfied, namely the volume of rainfall excess muet be equal to the volume
of direct runoff:
T
1
i=o
QO(lAt)At - T
1
i=o
A(iAt) [R(iAt) - B]At

where Q,(iAt) is the direct runoff at the outlet at time Ata

The synthesis of the stream network is based, in part, on the observation

made,by Leopold [i61 that there is no definite tendency îor the flow velocity to
have a great change along the length of the stream for a given retuni period or
fkequency. It may thus be assumed that locations having equal distances meas-
ured along the stream network to the outlet, have the same runoff travel time to
the outlet. If it may be further assumed that the drainage density is approxi-
mately uniform within a watersheä, then the total stream length upstream of a
particular point on a stream ia proportional to the tributary drainage area at
that point. Thus the distribution of the travel times is proportional to the
distribution of the drainage areas along the stream reaches, and only the latter
need to be considered. Fig. 1 shows the method of estimation of the distribu-
tion of the drainage a r e m s(JAL) along the main stream reaches for a idealized
watershe d.

The vol^ of runoff may be obtained by adding the runoffs from each of the
elementary contributing areas. Calling a(jAL, IAt) the dynamic response area
at stream reach jAL and at time iAt, the continuity equation may be written
T T S
1
is0
Qo(iAt)At 1 1
i=O,j=o
a(JAL, iAt) [R(iAt) - BIAL At (3)

where S is the total nimiber of stream reaches.

Assuming further that the first order atreams or the atream Bources are plai-
formly distributed over the watershed, then, at a given time iAts the ratio 0%
the dynamic response area a(jAL, iAt) at stream reach jAL to the tributaPy
drainage area at the same stream reach is equal to the ratio of the total pe-
sponse area A(iAt) to the total watershed area Ao- Thus

The continuity equation (3) thua becows

536

The direct runoffs from the individual stream reaches are then routed
through the stream network by means of a linear routing procedure. li- 2
shows schematically the routing procedure for a stream reach. The input,
X(JAL, kAt), in reach j at time kAt is the direct runoff given by

X(JAL, kAt)
AO
-
ao(jAL) A(kAt) [R(kAt) BI - (5)
and the routed outflow from reach,j, at time iAt is Y(JAL, iAt), given by the
convolution integral shown in pig. 2 which Is approximated by the convolution
sum
i
Y(jAL, iAt) 2:
k=O
Hu(JAL, (i-k)At) X(jAL, kAt)At (6)

where H (JAL, (i-k)At) is the kernel function or instantsneous unit hydrograph

of the b e a r routing procedure used. The runoff at the outlet, Bo, is obtained
by summation over the stream reaches

2C (iAt)
S
= 1
i
1
j=O k=O
H,(jAL, (i-k)At) * Ia0(jAL) A(kAt) [R(kAt)-BI
AO
- At ' (7)

where A(kAt) is given by equation (1). The kernel fbctions for 10 of the most
common linear routing models have been listed by Toebes and Chang [la]. In this
particuiar study the linear channel routing kernel function used is based on a
linearization of the Saint Venant equations developed by R w g e and Harley [15].
This kernel function has three parameters: the stream slope, a reference dis-
charge, and a roughness parameter.

IMPLEMENTATION OF THE MODEL

The watersheds selected for testing the model are located in the state of
Indiana, near the center of the esstem half of the United States. Thirteen ba-
sins were used with areas ranging from 8 to 400 square kilometers. The drainage
maps for these watersheds were prepared from aerial photographs at the scale of
1:20,000 by the staff of the Airphoto Interpretation Laboratory, School of Civil
Engineering at -due University. The maps used were at the scale of 1:63,360
(one inch equals one mile). The longitude and latitude of all stream junctions
and stream sources within the basins were digitized and stored on punched cards
by means of an automatic digitizer. The details of assembly and of the storage
of the hydrologic and geomorphologic data on magnetic tapes have been reported
by Lee, Blank and Delleur [is]. Fig. 3 presents a CALCOMP restitution of the
drainage network of a watershed from the data stored on magnetic tape. Also
ahawn on Fig. 3 are the etream link m d the drainage eue8 distributions along
 537

the main stream. The rainfall imposed on the dynamic contributing areas is then
used as the input into the linear routing procedure for each main stream link
and then summed over all the stream links. Fig. 4 (right) shows the outflow hy-
drograph obtained by the complete linear routing method using the parameter val-
ue shown (QI3 = reference discharge in d/Sec, CZ = roughness coefficient in
d 2 / s e c , SL = main stream slope). With the exception of the slope, the parame-
ter velues were obtained by an optimization procedure which minimized the differ-
ence between observed and calculated peak discharges and observed and calculated
timesto the peak discharge. The parameters so obtained were correlated with
climatological and geomorphological characteristics of the watersheds with the
following results.

For the watersheds used in this study it was found that the outflow hydro-
graphs were not sensitive to the choice of D for D > 0.5. A value of D = 0.8
was used. The value of B was taken as zero as the soils were generally imper-
vious because of their clayey type and high permanent water table in the contri-
buting areas adjacent to the streams. The value of N was found to be related t6
the m o P f ratio, R (ratio of measured runoff to measured rainfall):
r'

=
0.464 -
0.242
Rr
for D = 0.8, B =O (8)

The runoff ratio was in turn related to the storm characteristics, the tempera-
ture and an average soil permeability index of the basin by a regression equa-
tion of the type
a 8 6 ~
(9)
Rr = Tmin I' 'max

where T+n is the minimum daily temperature when the storm occurs, Sf =: soil
permeability index determined by assigning soil permeability VaEues to major
soil types occurring in the basin, and calculating the weighted average for each
basin, Pt is the rainfa-ll. volume and P- is the maximum rainfall intensity. The
independent variables in the right hand side of Eq. (9) are listed from left to
right in order of decrearkig significance. Al1 the exponente were negative and
=
less than one (a = -0.42, f3 -0.15, y =
-0.18, 6 = -0.25). The multiple corre-
lation coefficient was 0.91.

The roughness coefficient C, was significantly correlated to the basin area,

the stream slope, and the b m e flow per unit area by a regression equation of
the t y p

where Bf is the base flow per unit area when the storm occurs, is the drain- 4
age area and So is the slope of the main stream. The independent variables are
listed in order of decreasing aignificance. The exponents p and v were positive
(1.0 and 1.4 respectively) but A was negative (-0.21)* The multiple correlation
coefficient was 0.64. The value of the reference discharge varied between nar-
row limits, 1.1to 1.4 cubic meter per second for a t o m wlumas ranging from 2.5
538

to i4 mm. Making use o? equations 8 t h r o u a 10, the model regenerated well the
shapes of the hydrographs; the peak discharges were in general, reproduced with-
in 20% and the times to peak within 10% of the observed values. It should be
remembered that equations 8, 9 and 10 are unlikely to be vali8 outside of the
geographical arca ?or which they were obtained. They indicate, however, the
type of variables which influence the model parameters and their corresponding
sensitivity.

DISCUSSION AND CoNCurSIONS

The framework has been developed for a model which makes it possible to es-
timate the runoff from rainfall and from data obtainable from aerial photography
and from remote sensing. As presented, aerial photograph is needed for the de-
termination of the stream network, the main stream slope and the watershed a r e a n
In addition infrared color photography and/or ground observations are needed to
obtain the soil permeability index, the soil types for the estimation o? the ''B"
horizon permeability and the base flow.

with the rapid progress of the remote sensing technology, it is expected

that, in the near future, the amount of field work necessary may be greatly re-
duced and lidted to calibration areas to obtain the spectral response charac-
teristics needed for the interpretation of the remote sensing scanning.

The rainfall-runoff process in a watershed was simulated by three basic

components: a dynamic contributing area model, a contributing area distribution
curve which integrates the contributing areas along the stream network, and a
linear routing technique. In the proposed dynamic contribution area model the
exponent N, which quantifies the rate of expansion of the response area, Was
found to be the dominant parameter, and was found to be correlated to the -off
ratio. The "B" horizon infiltration and the weight of the antecedent rainfd1 D
were not the primary parameters. In the linear routing, the roughness Parameter
was found to be correlated to geomorphologic parameters and to the baseflow Wr
unit area. The reference discharge did not change significantly from Storm to
storm or from watershed to watershed.
ACKNOh'LEIKMENT
The Work presented herein was supported by Lhe Office o? Water Resources R e
search, U.S. Department of the Interior under grant OWRR-B-008-IliD, by the Pur-
due Research Foundation under grant XR 5869 and by PiPrdue University. m-
thors wish to exprese their thanks to the sponsors,
REFERENCES

1. Rodriguez-Iturbe, I. (1969) Estimation of Statistical Parmeters for An-

nual River Flows, Water Resources Research, 5, pp. 1418-1421-
2. Keifer, R. W. and J. A. Sherz (1971) Aerial photograph for Water reSomceS
studies, Jour. o? Surveying and Mapping Div. , Am. Soc. civil Enva. vole 97,
No- SU29 PP. 321-333.
 539

3. Laboratory for Agricultural Remote Sensing, Purdue University, Lafayette,

Indiana (i968 and 1971) Remote Multispectral sensing in Agriculture, Vol.
3 (Annual Rept., 1968) Res. Bull. No. 844, Agr. Exp. Station, also Vol. 4
(Annual Rept., 1971) Res. Bull. NO. 873, Agr. Exp. Station.
4. Pluhowski, E. J. (1972) Hydrologic interpretations based on infrared ima-
gery of Long Island, New York, Geel. Surv. Water Supply Paper 2009, U.S.
Govt. Print. Off.
5. Weaver, K. F. (1969) Remote sensing: new eyes to see the world, Natl. Geo-
graphic, Vol. 135, NO. 1, pp. 47-73.
6. Waltz, F. A. and Y. I. m e r s (1970) Remote sensing of hydrologic resources
in the Great Plains, Rept. #I, Remote Sensing Inst., Univ. of South Dakota.
(Available from IVTIS, NO. PB 195 451)
7. Zachary, A. L., J. E. Cipra, R. J. Diderickson, S. J. Kristof, and M. F.
Baumgardner (1972) Application of multispectral remote sensing to soil
survey research in Indiana, Lab. for Application of Remote Sensing, Purdue
Univ., Lafayette, Indiana, Print 110972. '

8. Scherz, J. P. (1971) Monitoring water pollution by remote sensing, Jour. a i

'
Survy. and Map. Div., Am. Soc. Civil Engr., Vol. 97, No. SU2, pp. 307-320.
9. Colwell, R. N. (1973) Remote sensing in the management of earth resources,
American Scientist, Vol. 61, NO. 2.
10. Strahler, A. N. (1964) Quantitative geomorphology of drainage b&ns and
channel networks, in Handbook of Applied Hydrology, V. T. Chow, Ed., McGraw-
Hill Book Co., pp. 4-40, pp. 44-74.
11. Freeze, R. A. (1972) Role of subsurface flow in generating surface runoff,
2, Upstream Source Areas, Water Resources Res., 8. pp. 1272-1283.
12. Dunne, T., Bnd Black, R. D. (1970) Partial area contributions to storm run-
off in a s-1 New Englua watershed, Water Resources Res., 6, pp. 1296-1311.
13. Nutter, W. J., and Hewlett (1971) Stream flow production from permeable up-
land basin, paper presented to the Third Internatl. Seminar for Hydrology
Professors, Furdue Univ., Lafayette, Ind., USA, July 1971.
14. Dooge, J. C. I. (1959) A general theory of the unit hydrograph, Jour. o?
Geophys. Res., Vol. 64, NO. 2, pp. 241-256.
15 Dooge, J. C. I. and B. M. Harley (1967) Linear routing in uniform c h m e l a ,
Proc. inti. wdroïogy Symp., Sept. 1967, Fort Collins, Colorado, USA, 1, pp.
57-63.
16. kopold, L. E. (1953) Downstream change of velocity in rivers, Am. Jour. of
Science. 25. PP. 606-624.
17. Lee, M.-T. t&d-J. W. Delleur (1972) A program for estimating muoff from
Indiana Watersheds, Part III, Analysis of geomorphologic data snd a d y n d c
contributing area model for runoff estimation, Purdue Univ. Water Resources
Res. Center, Lafayette, Ind. Tech. Rept. No. 24.
18. Toebes, G. H. and T. P. Chang (1972) Simulation model for the Upper Wabash
surface water system, Purdue Univ. Water Resources Res. Center, Lafayette,
Ind. Tech. Rept. NO. 27.
19. Lee, M. T., D. ~ l a n k ,J. W. Delleur (1972) A program for eetimting m o r i
from Indiana watersheds , Part II , Assembly of hydroloaic euid geomorphologic
data for small watershed0 in Indiana, Purdue Unio. Water Resource9 Ra#. Ccn-
ter, Lafayette, Indiens, Tech. Rept. No. 23.
 540

STREAM REACH
a

U
O
O

2 4 6 8
STREAM REACH i
FIGURE I DRAINAGE AREA DISTRIBUTION ALON
THE STREAM REACHES
 541
INPUT 5 q (0,t)

OUTPUT = q (L,t)

INPUT -1 S Y S T E M 1 OUTPUT -
INPUT OUTPUT

DE LTA
FUNCTION

-
t t

q(0.t) = Xf0,t)

2 PHYSICAL D ? A G R A M OF U P S T R E A M INFLOW
INSTANTANEOUS UNIT H Y D R O G R A P H
FOR SINGLE STREAM REACH
542
m (D U N
 543
I
Q
O. a
œ
CJ
O
œ
n
>
r
âLL
=IA
o- 0
"w
zI-o-z3
I-
o
W
rr
n
4
O
(u
d
O
8
ï
v)
f d-
8
N
 MONTHLY STREAMFLOW ESTIMATION FROM LIMITED DATA

C.T. Haan

--
ABSTRACT

A four parameter, monthly water yield model has been develop-

ed and tested that makes it possible to estimate monthly streamflow
volumes from daily rainfall information. The four parameters of the
model can be determined from a s little as two years of observed
monthly streamflow data. This makes it possible to install
temporary, short term stream gaging stations to collect two or
three years of monthly streamflow data and from this data determine
the four parameters o f the water yield model. The model uses a
self-optimizing procedure so that the user is not necessarily in-
volved in the parameter estimation process. A study of 24 water-
sheds has also shown that the four model parameters can be related
to soil, geomorphic, and geologic characteristics o f the basin.
In this way the parameters for an ungaged basin can be estimated
without requiring any data from that particular basin. Once the
model parameters are determined, long traces o f monthlv streamflow
data cán be simulated us ng only daiiy rainfa 1 as a model input.

RESUME

__
On a développé et mis a u point un mode e à quatre paramètres,
pour l'évaluation du rendement mensuel en eau, qui rend possible
une estimation du volume de l'écoulement mensuel à partir de
l'information que constituent les pluies journalières. Les quatre
-
--r a-m...5- t -r e- -s di1
nn - -- - - - - -- - - -- -
- - m...o d-3-1-e- n ~ i i.v e--n-t P-+ P P d 8 t P v r,,n, -.
tions aussi restreintes que les résultats de deux ans d'observation
--- -
i n.L s > n a-etiv ~ t .
r-.- i..
n.I~"L.n..m- r n a -
des débits. Cela permet d'installer des stations de jaugeage des
débits, temporaires, à court terme, afin de réunir des informations
portant sur deux o u trois ans, s u r les débits mensuels, e t , 2
partir de ces informations deduire les quatre paramètres du modèle
de rendement d'eau. Le modèle fait usage d'un procédé qui évalue
automatiquement les informations et les dispose de l a meilleure
facon possible (self-optimizing procedure) de sorte que l'usager
n'a pas nécessairement à se préoccuper de l'estimation des para-
mètres. Une étude de 24 bassins (watersheds) a aussi montré que
les quatre paramètres du modèle peuvent être reliés aux caracteris-
tiques géomorphiques et géologiques, ainsi qu'a celles du sol, pour
un bassin (basin) donné. De cette facon, les paramètres pour un
bassin non jauge peuvent être évalués sans que l'on ait besoin de
faire appel a des renseignements sur ce bassin en question. Une
fois que l e s paramètres du modèle sont déterminés, on peut produire
artificieilement un tracé important (long traces) des débits
mensuels en se servant seulement, pour alimenter le modèle, des
précipitations journaïières .
546

The design of surface water supply systems requires an

estimate of streamflow volume characteristics. Ideally, the
designer would like to have available long streamflow records.
Generally, for the design of water supply reservoirs, a long
record of monthly runoff volumes would be sufficient. Unfortunate-
ly these streamflow records are not available for a vast majority
of the streams draining watersheds of up to 1500 square kilometers.
Thus, methods of estimating monthly runoff volumes for these un-
gaged basins are needed.
Watershed models that simulate continuous streamflow records
offer promise in this area. Several comprehensive watershed
models are presently available [1,2,3]. These models may be
categorized as parametric models in that they contain several
parameters that must be estimated before they can be applied to a
particular watershed. The parameters must be determined manually
and are based on a comparison of observed streamflows with
simulated streamflows.
Realizing the ability of computers to find parameter sets that
satisfy certain objectives and the differences that occur when
different people determine the values for model parameters, Liou
141 attempted to develop a self-optimizing procedure for the
Stanford Watershed Model [i]. Further work on the Stanford Model
was done by Ross [SI in an attempt to relate the optimum model
parameters to watershed characteristics.
Haan [6] has developed a relatively simple runoff model that
enables one to estimate monthly streamflow from daily precipitation
and estimated average daily potential evapotranspiration. The
model contains four parameters that must be determined for each
basin. These parameters are :
f
Smax
= maximum infiltration rate (cm/hr).
Cmax
= maximum daily seepage loss (cm).
= water holding capacity of that part of
the soil from which the evapotranspiration
rate is less than the potential rate
unless this portion of the soil is
saturated (cm).

'Associate Professor, Agricultural Engineering Department,

University of Kentucky , Lexington, Kentucky, 40506 , USA.
 Fs = fraction of the seepage that becomes
runoff (-1.
The optimum values for these parameters are defined to be
those that minimize the sum of squares between the observed and
simulated monthly runoff volumes. Thus, to get the optimum
parameter values , some observed streamflow data is required.
Work with the model has shown that two years of monthly streamflow
data are usually sufficient to obtain a satisfactory set of
optimum parameter values.
The model has the capability of determining the optimum
parameter values for a particular basin when provided with daily
rainfall, average daily potential evapotranspiration by months,
some observed monthly streamflowc for the basin, and a set of
initial estimates for the parameter values.
The optimization procedure that is presently used is a simple
univariate technique. The value of the objective function
n
c (Vo - Vs. ) is computed using the initial estimates for the
i=l i 1
parameter value. in this expression, n is the number of monaths of
flow and Vo is the observed volume and Vs. the simulated volume of
streamflow $or the ith month. Next the vatue of one of the para-
meters is chanued bv a fixed amount and the obiective function is
recomputed. &e vaiue of this parameter contiAues to be changed
as long as the objective function is improving (getting smaller).
The other three parameters are adjusted in the same manner one at
a time. Since these parameters are not independent, the entire
process is then repeated one or two times. The result of this
iterative process is taken as the optimum set of parameters.
This model has been tested on 24 watersheds in Kentucky
ranging in area from 1.74 to 1225 square kilometers and on 3 water-
sheds in South Carolina ranging in area from O .ll to 2.27 square
kilometers. The results of these evaluations are given in 161 ,
[71 and [8]. Defining the average prediction error (%) as 100
times the absolute value of the difference between the observed
and simulated average annual streamflow divided by the observed
average annual streamflow, the average prediction error for these
27 watersheds is 4.0 percent. For these watersheds, the average
annual runoff varied from 18.7 to 48.6 cm.

For streams on which there are n o records available, at least

two procedures can be used to estimate the optimum parameter values.
548

If sufficient time is available, a temporary stream gaging station

can be established and operated for two or more years. This
station would only have to provide information on the monthly
flows. The data from this short-term gaging program could then
be used in the optimization scheme described earlier.
Jarboe and Haan [7] have used a second technique for estimat-
ing the four model parameters for ungaged basins. This method uses
streamflow information from gaged basins in the vicinity of the
ungaged basin of interest. The optimum model parameters for the
gaged basins are determined and related to measureable character-
istics of the gaged basins. These relationships are then used
to estimate the model parameters for the ungaged basin.
The basin characteristics used by Jarboe and Haan [7] are
shown in Table 1. The four model parameters were related to these
factors using multiple linear regression. Twenty-three watersheds
were included in the study. Six of the watersheds were selected
at random and treated as ungaged basins. The remaining 17 basins
were used in developing the following prediction equations for the
model parameters :

fmx = 11.83 - 11.51 Smax - 0.0147 SdSb - 0.030 A Hg

- 0.334 PkFc + 0.692 VrPR
(1)

nax
= 0.073 + 0.0031 Wc + 0.00075 Iw L - 0.0021 P H
a g
+ 0.00011 FcL - 0.0057 V H
r g (2)

C = 7.69 + 0.739 IwSb + 0.011 S H + 0.0243 FcIw (3)

d 9
Fs = 0.325 + 0.0068 L + 0.444 PkSb + 0.00027 PsSd (4)

- 0.018 WcPk

These equations should not be used on watersheds (1) greater

than 100 square kilometers in area, (2) on urban watersheds, or
(3) on watersheds that differ greatly in their hydrologic char-
acteristics from the watersheds used to derive the equations.
 549

Table 1. Watershed characteristics used by Jarboe and

Haan [7] to estimate the water yield model
parameters.

Geomorphic Factors
A Basin area (km')
percent of basin under forest cover (%)
Percent of basin in lakes and ponds (%)
Slope of the main stream (%)
% Length of the main stream (km)

Soil Factors
W Average available soil water capacity (cm)
HC U. S. Departnuint of Agriculture, Soil
Conservation Service hydrologic soil
group converted to a numerical index
from 1 to 4 (-1
Sd Average soil depth (cm)
P Average soil permeability (cm/hr)
:p
Average permeability of upper soil horizon
(cm/hr)

Geologic Factors
vr "Rock" volume = mean basin elevation above
basin outlet times the basin area (krn3)
,
I Water availability index (an index ranging
from 1 to 4 depending on the ability of the
material underlying the basin to yield water
to wells) (-)
5 50

Table 2 presents a s-ry of the results of using the above

4 equations to estimate the parameters of the model on the 6 basins
that were taken as unqaged. The simulated runoff values were
obtained by estimating the model parameters from equations 1
through 4 and then using these estimated parameters in the water
yield model to simulate monthly streamflows. The six watersheds
listed in table 2, although actually gaged, were considered as
ungaged and not used in developing equations 1 through 4. These
results indicate that reasonably good estimates of runoff volumes
can be made on watersheds for which no streamflow records are
avai lab le.

EXAMPLE APPLICATION

The South Fork of the Little Barren River in Kentucky was

used to illustrate the application of this model under various
conditions. Streamflow records have been maintained by the U.S.
Geological Survey for this watershed for the period October of
1948 through September of 1970. A U.S. Weather Bureau rain gage
at Edmonton, Kentucky, about 6 1/2 kilometers from the watershed,
was used to provide the needed precipitation input. Some of the
watershed physical characteristics are given in table 3. This
watershed was not selected because of the ability of the model to
simulate its monthly flow, but because of the long gaging record
for the stream that could be used to check the simulated results.
Table 4 summarizes the various simulations made on the South
Fork of the Little Barren River watershed. Methods a through d
are examples of how the model might be used to simulate streamflow
from a previously ungaqed area. Method a required no streamflow
records in that the parameters were estimated from equation 1
through 4. Methods b, c, and d illustrate how the model can be
used if it is possible to initiate a stream gaging program and
collect 1, 2, or 3 years of data respectively on monthly runoff
volumes. In method e the entire 22 years were used in a procedure
described by Haan 161 to obtain the parameter values. In table 4
the percent error is as previously defined, the correlation
coefficient is the simple correlation between the observed and
simulated monthly flows for the entire 22 year period of record,
and the slope is the slope of a simple regression line relating
the observed and simulated flows.
 551

Table 2. Comparison of observed and simulated average

annual runoff for six Kentucky watersheds when
the mode1 parameters are estimated by equations
(1-4).

Observed Simulated
Average Average
Annual Annual Percent Watershed
Watershed Runoff Runoff Error Area
2
Helton Branch 43.59 crn 44.63 cm 2.4 2.20 km
McGills Creek 41.50 46.41 11.8 5.54
Perry Creek 34.16 33.55 1.8 4.45
Stillwater Creek 48.59 42.98 11.5 62.16
Little Plum Creek 46.74 48.01 2.7 13.33
N. F. Nolin River 39.90 43.46 8.9 94.28

Table 3. Physical characteristics of South Fork of the

Little Barren River watershed, Kentucky.

Area 47.4 km2

Forest cover 62 %
Lakes and ponds 0.11 %
Slope of main stream 0.32 %
Length of main stream 15.96 km
Available soil water capacity 17.68 cm
Index of USDA hydrologic soil group 2.30
Average soil depth 84.84 cm
Average soil permeability 3.02 cm/hr
Average permeability of upper soil horizon 3.35 cm/hr
II Rock 'I volume 2.25 km3
Water availability index 2.0
552

The mean annual runoff for the South Fork of the Little
Barren River is 50.17 cm. Thus an error of 1 percent represents
an average annual error in the simulated runoff of O .5 cm. When
the model parameters were calculated from equations 1 through 4,
the error in the average annual runoff was 4.3 cm. Figure 1 shows
a portion of the simulated and observed streamflows for the water-
sheds. The simulated monthly runoff shown in this figure were
obtained using parameters calculated from equations 1 through 4
in Haan's [6] wateryield model. Again it is cautioned that these
equations may not produce reliable parameter estimates for regions
hydrologically different than Kentucky. The technique of deriving
parameter prediction equations should, however, be valid else-
where.
Methods b, c, and d of table 4 illustrate how a few years of
streamflow data can be used to estimate model parameters which in
turn can be used to simulate long traces of monthly flows. The
variable nature of streamflow from year to year is apparent in
the runoff records from this watershed. As an example the first
three years of the 22 year record produced the highest, third
highest, and sixth highest annual runoff. The average annual run-
off for the first three years was 75.79 cm as compared to 50.17 cm
for the entire period of record. It was these three wet years
that were used in determining the model parameters indicated in
table 4 under methods b, c, and d. This indicates that even
though the years used in obtaining the model parameters may not be
representative, reasonable estimates of streamflow can still be
obtained.
Table 4 also indicates that the accuracy of the simulation
depends on the years used in determining the model parameters.
The fact that using two years of flow data to obtain the
parameter values produced better simulated results for the entire
22 year period than did the parameters obtained from three years
of data is not unusual; however, in general the more years used
to obtain the parameters, the better will be the simulated
results.
Method e consisted of (1) optimizing the model on the first
year of record, (2) simulating the entire 22. years of flow with
these parameters, (3) reoptimizing the model on the 2 years from
the entire 22 year record that produced the poorest fit, and
(4) finally determining the final parameters as a weighted
average of the resulting two optimum sets of parameters where the
weighting factors are the sum of the deviations of observed flows
from simulated flows. The parameters obtained in this manner
 553

Table 4. Methods used to optimize parameters on S.F.L. Barren

River and summary of simulation results.

Method Des criD tion

(a) Parameters calculated from equations 1-4.

íb 1 Parameters determined by optimization on first year of
data.
(Cl Parameters determined by optimization on first two years
of data.
(dl Parameters determined by optimization on first three
years of data.
(e) Parameters optimized by Jarboe 181.

Percent Correlation C
fmax 'ma,
Method Error Coefficient Slope cm/hr cm/day cm FS

(a) 8.64 0.91 0.93 3.58 0.21 13.33 0.49

(b) 10.13 0.87 0.96 3.45 0.25 16.38 0.54
(c) 2.19 O .92 0.91 3.58 0.20 . . 12.57 0.54
(d) 9.38 O .92 0.89 5.61 0.22 11.81 0.69
(e) O .56 o .91 0.92 3.30 0.22 12.45 0.58

when used with the watershed model were able to simulate the 22
years of record with an average annual error of only 0.56 percent
or 0.28 cm. Obviously this technique cannot be used on a data
scarce watershed. It is included here only to provide an
indication of the ability of the model to simulate monthly stream-
flows.
This model like most parametric hydrologic models, is
in a constant state of change as improvements are incorporated to
make the model easier to use, to reduce computer processing time
and to increase the accuracy of the simulations.
554

SUMMARY

Two procedures for using a four parameter water yield

model for simulating traces of monthly streamflaw from watersheds
with either no or very limited streamflow information are
presented. The two procedures are (1) to relate the model
parameters to watershed physical characteristics using stream-
flow data from watersheds located near the watershed of interest
or (2) to establish a short term gaging program on the stream
draining the watershed and use these streamflow records to
determine the model parameters. Once the model parameters are
determined, long streamflow traces can be generated using either
measured or synthetic daily rainfall. These two procedures were
illustrated on a watershed in Kentucky and demonstrated that
reasonably accurate estimates of monthly streamflow can be
obtained.

Acknowledgements: The work in which this report is based was

supported in part by the Kentucky Division of Water and in part
by the Kentucky Agricultural Experiment Station as a contribution
to Southern Regional Project S-53 "Factors Affecting Water Yields
from Small Watersheds and Shallow Ground Aquifers". The paper
is published with the approval of the Director of the Kentucky
Agricultural Experiment Station.
 555

BIBLIOGRAPHY

1. Crawford, N. H. and R. K. Linsley. (1966). Digital simul-

ation in hydrology: Stanford watershed model IV. Technical
Report 39 , Stanford University , Department of Civil
Engineering, Stanford, California.
2. Holtan, H. N. and N. C. Lopez. (1971). USDAHL-70 Model of
Watershed Hydrology. Technical Bulletin No. 1435, Agricultural
Research Service, U. S. Department of Agriculture, Washington,
D.C. 84 pp.
3. Tennessee Valley Authority. (1972). Upper Bear Creek
Experimental Project: A continuous daily streamflow model.
Division of Water Control Planning, Hydraulic Data Branch,
Knoxville, Tennessee. 99 pp.
4. Liou, E. Y. (1970). OPSET: Program for computerized
selection of watershed parameter values for the Stanford
Watershed Model. University of Kentucky Water Resources
Research Report 34, Lexington, Kentucky.
5. Ross, G. A. (1970). The Stanford Watershed Model: The
correlation of parameter values selected by a computerized
procedure with measureable physical characteristics of the
watershed. University of Kentucky Water Resources Institute
Research Report 35 , Lexington, Kentucky.
6. Haan, C. T. (1972). A water yield model for small watersheds.
Water Resources Research 8 (No. 1) , pp 58-69.
7. Jarboe, J. E. and C. T. Haan. (1972). Calibration of a four-
parameter water yield model to small ungaged watersheds in
Kentucky. Paper No. 73-207 for presentation at the 1973
Annual Meeting of the American Society of Agricultural
Engineers, Lexington, Kentucky, June 17-20, 1973.
8. Jarboe, J. E. (1972). Calibration of a four-parameter water
yield model for use on small, ungaged watersheds in Kentucky.
Unpublished M. S. Thesis in Civil Engineering Library , University
of Kentucky, Lexington, Kentucky.
556

20[
-- PART OF
S.F.L. B A R R E N R.
RUNOFF RECORD

- OBSERVED R O
-O

- SIMULATED R O

TIME

Figure 1. Example of monthly streamflaw simulation

results using calculated model parameters.
 OBTAINING DEFICIENT INFORMATION BY SOLVING
INVERSE PROilLEMS FOR MATHEMATICAL RUNOFF MODELS

V.I. Koren and L.S. Kutchment?:

AB ST RACT

Possibilities are considered for increase o f deficient

information for extending observation series by solving the
"inverse problemt1for mathematical runoff models. The results
of applying the theory of "improperly posed problems'' are
presented. Examples are given for representing hydrological,
geometrical and hydraulic characteristics of the basin by
lumped and distributed parameter runoff models.

RESUME

Les auteurs examinent les possibilités de la resolution

du problème inverse appliquée aux modèles mathématiques d'eco!
lement, en vue de compléter les lacunes des séries d'observa-
tions et d'étendre la période couverte par ces séries. Ils ex-
posent les résultats qui ont été obtenus par l'application de
l a théorie des problèmes posés incorrectement. Ils citent des
exemples de détermination des caractéristiques hydrologiques,
topographiques et hydrauliques a l'aide de modèles d'ecoulement
globaux ou matriciels.

k: Hidrometeorological Centre of the USSR.

 558

Mathematical modelling of hydrological processes is increas-

ingly used to provide for missing information and to extend hydrolo-
gical time series. Mathematical models are predominantly used for
the solution of the so-called 'direct problem', consisting of deriv-
ation of unknown hydrological variables by solving respective differ-
ential equations with known coefficients and known initial and bound-
ary conditions, In a large number of cases it is necessary to solve
the 'inverse problem' namely to find the coefficients and establish
the initial and boundary conditions using observed values of the
hydrological variables included in the equations. This approach has
as yet gained relatively rare use due to the fact that the solution
of the 'inverse problem' is more difficult than that of theldirect
problem'. The solution of the 'inverse problem' may be circumvented
by multiple solutions of the 'direct problem' for example by the
methods of trial and error and subsequent optimization. Thia may
lead however to a non-unique or inferior solution. The principal
difficulty in the solution of the inverse problem consists in the
fact that it may be incorrectly posed and thus leads to the non-
existence of some or any initial conditions or leadsto a solution
in which a small change of initial conditions (data) due for example
to observational errors, results in major changes in the results.
This has caused in the past a reluctance toward the use of this
method, since the solution being of very low accuracy and high un-
certainty casts doubt on its physical significance.

A number of studies were made in recent years (particularly

by A.N. Tikhonov and his school) aiming at the correct posing of
the problem by establishing the necessary conditions for it.
A.N. Tikhonov has shown that it is possible to u13e a priori inform-
ation on the solution to ensure a continuous dependance of the
solution of an incorrectly posed problem on its initial conditions
and to derive special algorithm:: which prevent bringing out the solution
outside the limits of its uniqueness and of the existence of its initial
conditions. In particular it made possible to solve with sufficient
stability such classical incorrectly-posed problems as the integral
equation of the first type, algebraic systems with improper initial
conditions, the Cauchy problem c?f the Laplace equation and others.
The theory of the 'inverse problem' has thus stimulated the formu-
lation of algorithms used in many scientific and technical fields.
The method was particularly useful in geophysics, where it permitted
the solving, for example, of problems of determination of rock charac-
teristics not accessible for direct measurement as well as restora-
tion of missing information, to cite only the most important points.
The use of this method in hydrology appears also as most promising.
Examples of such studies, used in hydrological practice, are given
below. They illustrate also the principles and possibilities of the
theory of incorrectly posed problems.

1. Determination of the input functions of the models

with lump parameters

Let us suppose that the process of transforming an input h(t)

in the catohment (effective rainfall or an inflow) into an output
Q(t) can be described by the Duhamel integral:
 559

where P(t) is some known function of influence. Then having the observa-
tions on Q(t) and knowing the function P( t) (by historic observations or

h(t). Thus an improperly posed problem is solved -

from physiographic and hydraulic data) it is possible using (1) to derive
consisting of an in-
tegral equation of the first type. It is possible to solve this problem
on the basis of A.N. Tikhonov's algorithm. Integral (1) is replaced by
a summation according to the method of rectangles and a smoothed functional
curve is constructed:

3
where Q = a vector, designating the Ordinates of the given hydrograph Q(t);
-b
h = a vector of the unknown ordinates h A = a matrix with elements
P ; d =a positive Constant. Finding5'the minimum of this functional
mk&'it possible to receive a sequence of stable solutions %,which
converge to the accurate solution providing there are no errors in the
given data. However since there are always errors in these, changing
the parameter &(called parameter of regularization) we select such
solution which corresponds best to the a priori information about the
function h(t). For exam e good results are obtained with the aid of
the condition Th(t)dt=$(t)dt.
o

Other kinds of a priori information, allowing the narrowing

of the interval of unknown solutions, may be a suggestion on the smooth-
ness of the solution, the non-negativeness of the ordinates, the closeness
to some known function and so on. Naturally, the narrower the interval
of the solution, the higher accuraoy will be obtained. Results in using
functional (2) to determine the input functions of the runoff models,
described by the Duhamel integral, are presented in greater detail
in (3)' where examples of constnicting effective rainfall, hydropower
station releases and snowmelt intensity are treated. Another approach
to the solution of the inverse problems for models described by the
Duhamel integral (linear models with lump parameters) are indicated in
(6).

2. Determination of geometric and hydraulic charactexistics

of river channels using observations of flow

To describe unstea9flow in a river channel Saint Venant

equations may be used:

(3)
560

where 2 (x,t) = stage at point x at time t, Q(x,t) = discharge, K(x,z)

forces of resistance1 g= acceleration of gravity. Because of great
variability of geometry and roughness of the river channels the
functions F(x,z) and K(x,z) determined by the observations in
separate points are not quite representative for the whole river reach,
even with large frequency of observations. Thus a problem of determin-
ing the averaged relations P(x,z) or B(x,z) = aF/aZ and K(x,z) by ob-
servations of flow (the determination of coefficients of the system (3))
is of great significance for the establishment of the most characteris-
tic geometry and hydraulic properties of the river channel as well as
for ensuring sufficient accuracy of the calculationa. It can be shown
that this problem is improperly posed and for its solution it is
necessary to derive special calculating algorithms. We shall discuss
below two of the approaches tried by us in solving this problem.

(A) The discharges and Water levels are known in a rather large
number of sites.
Integration of the continuity equation (3) with respect to x,
leads to:

Finite differences are substituted for the derivatives and

instead of an integral it is possible to construct for every time moment j
the following system of equationst

In order to solve this system it is necessary to have Q(x,t) F(x,o) and

F(o,t). As the problem is improperly posed the solution of the
system (5) is unstable. For its regularization the solution of
A.N. Tikhonov's functional is with introducing initial approximation.
As a result for every time suchFarefoanä which correspond to the
minimum of the functional.
 561

where 2 is the given initial approximation, d = positive parameters,

thus a golution is found which not only secures the minimum of square
deviation of the right part of the system (5) from the left part, but
at the same time it is least deviated from the initial approximation.
The condition of functional extreme gives:

To select the quantitydmethod of discrepancy has been used.

The idea of this methori COnSiStS in conforming the accuracy of the
problem's solution to the accuracy of observed data.
It is supposed that the error0 of the given information form-
ing discrepancy of the system (5) are known and an d is found which
secures this discrepancy . '
8 It is possible to prove that if the
functional (6) is used the parameterdsecuring the given discrepancy
is unique. The initial pproximation can be made in a rather crude
mannerbarticularly for 9 = O), however giving a good initial approxi-
mation contributes toam-aocurate optimum d . Use of the initial
approximations provides great possibilities for improvement of the
solution by introduction of a priori information. Such a priori
information can be an empirical relationship between geometrical and
hydraulic characteristics, observed in separate sites, and different
theoretical formulas (for example, we have used the equation of the
typical form of river Otrinnel derived from the principle of minimum
dissipation of energy).
The values of F (x,t) found according to equation (4) have
been used for determining the characteristics of the resistant forces.
For this purpose the momentum equation has transcribed:

Derivatives with respect to t have been replaced by forward directed

finite differences and the integrals have been replaced by sums de-
rived by the method of rectangles. The resulting algebraical systems
have been solved for all time intervals with the help of the same
algorithm as the system (5) (without the initial approximation).
Aa for determining F(x,t) and K(x,t) the discrepancy has been
taken equal to 5 per cent of the average module from the left integral
equation's part.
This method has been tested on data obtained by special
observations of unsteady movement in the m e r c a river and it has
given satiafactory results (a comparison ha8 been made between the
relations F(x,z) and K(x,z) which have been derived by different
floods by measurements in separate sites) (see figure 1).
562

(B) The stages are known in a rather great number of sites and
the discharges only in the first and the last site.
Let us integrate the continuity equation with respect to
time (in the interval (Ti, Ti+l)) and to distance (in the interval
(0, LI):

to solve it in the form:

where ‘y-the Chebishev polinomials. Let us put (10) in (9):

No terms with zero polinomial are in the left part of the equation (li),
because in this case the integral would be equal to zero. The equation
(li) is therefore not sufficient for the full determination of the
function (10). However it can be used for determining the function
B(x,z), which can be presented:

To find the coeffiaients we shall construct a system of equations

(their number must not b e h w than Y=(n +l)m , and change the limita
of the integration with respect to time in (dl
so as to embrace the
whole amplitude of variation of discharges and of stages on the rising
as well as on the falling, part of the hydrograph. Let us write this
system in the matrix forms

Bere is the matrix of +th order, its elements are equal

 563

id- vector of the unknown coefficients $(B. x - right part with elements:

Since the system (13) is unstable, ita solution is possible

with A.N. Tikhonov's functional. AS a result the following system is
found :

where%'* - matrix transformed with relation t o p , E - the unit matrix.

The parameter of regularization d has been determined from the
conditions of minimum of the function

where A.PP,,),
values JO(.
A (dp) -
j-th elements for the two successive
+or determining (ni + i) coefficients entering in (10)
we shall replace in (9) discharge with the product of a cross section
area and the velocity of the current U(x,t) and shall make the proper
integration with respect to time and to distance. Putting in the
resulting equation the relation (10) we shall find:

Here C - matrix of (nI + I) x N-th order with elements

The rest of symbols are the same. lystem (17) is solved by analogy
with system (13). Having determined 3 and it is possible, asing
relations (10) and (12) to find function B(x,z). This approach has
been tested on the data of special observations in the Svir river.
In figure 2 functions B(x,z) for some sites, calculated by relation (10)
are shown: furthermore widths were determined aocording to topographic
data. For controlling the results of these calculations the discharges
in-the intermediate sites have been determined with the help of equation:
564

These discharges have been found as very close to those observed.

The coefficients received from the different floods have turned out
%
to be quite simila and this fact indicates their sufficient stability.
Let vs see now a scheme for determining the hydraulic'characteristics
of river channels. We use the dynamic St. Venant equation, assuming
that the ineLtial terms are equal to zero

Putting (18) into (19) and integrating with respect to distance in

the interval (0,ï) we gett

às earlier we shall find the solution in the form:

Putting (21) into (20) and using Tikhonov's functional by analogy with
the previous one we construct the system of equations for determining
the coefficients Dks:

&=Z(t)
-b
where D -- Z(t),
vector of unknown coefficients, 3 - vector with elements
- matrix of (n2+1).(m2+l) N-th order with elements

The function B(t, t) hae been calculated according to relation

(12) including the earlier determined coefficients Ilks.
The found
functions have been oompared with the functions determined by the method
of optimization. It was found that a strong smoothing is observed.
This can be eliminated by taking logarithms in equation (19).
 565

References
566

I50

SO e

e
/Iff 8
a 0
E
e a *'
96

0
 567

x r acm .x=
3.9cm
I l
I
I
I
I
I
I

I I I
15.0 250 2.50 i50 250 350

i
/

I I I
200 390 4uu
 THE MATHEMATICAL MODEL OF
WATER BALANCE FOR DATA-SCARCE AREAS
by
Nabil Rofail

ABS TRACT

The mathematical model of water balance for data-scarce areas

is designed. The solution of this problem is considered of general
type of boundary conditions.

The equations of motion and mass conservation lead to the l i n

ear parabolic partial differential equation. The equation is solved
by implicit scheme and the alternating direction procedure is appli
ed for computation. The numerical procedure has second order accu-
-
rracy and is unconditionally stable. As it contains no iterative --
routines, it is also exceptionally economical in computing time and
memory requiments. Therefore the procedure is recommended for areas
of inadequate hydrological data.

RESUME

L'auteur décrit un modèle mathématique de bilan hydrologique

destiné aux régions pour lesquelles on dispose de peu de données. -
La solution du probleme est envisagée pour des conditions aux limi-
tes générales.

Les équations du mouvement et de la conservation de masse c o n

duisent à une équation aux différentielles partielles linéaire para
bolique. Cette équation constitue un système implicite qu'on résoud
par approximations successives. Le mode de calcul numérique est in-
conditionnellement stable et permet une précision du second ordre.
Comme il ne contient pas de procédé itératif, il est aussi exception
nellement économique en temps de calcul e; en dimension de mémoire.
I1 est donc recommande pour les régions ou les données hydrologiques
sont insuffisantes.

~~ ~ ~

* Water Resources Dept. Desert Institute, Mataria, Cairo, Egypt.

570

Introduction

The system of equations of motion and equations of mass con-

servation for the ground water flow leads to linear pzrabolic di-

fferential equations. In the present work the impervious bed has

been considered of any configuration and the recharge or dischar-

ge from the aquifer has been introduced. Moreover the effect of

boundnries are considered either of a river or of a continuation

of the aquifer of different parameters. The present aaterial deals

with the solution of the balance equation that can be easily app-

lied for areas of inadequate hydrologicd. data.

Formulation of muation

The aquifer is considered homogenous and the effect of the

impervious bed is &%Ven. According to Dupuit assumption, the equ-

ation of motion cam be written as followsj (see Fig. i ) ,

v = - k 2 i h - t ~ )= - k a h - k dz (2)
"Y "Y
The equation of mass conservation can be considered as follows;
 571

!he dischuge or the recharge to the aquifer is considered as a

:unction of time to the exposed area i.e. N =f (x,y,t) equat-

tons (1),(2) and (3) proYide the following equation system,
- kh QQ(lZz' - k d(h+z) a h
9 X ax
- kh - -
Q'íh+z) -k a(h+z) ah -N=a (4)
9Y * ay al/
?or simplifying the solution of equation (4), Boussensq acsump-

:ion (the powers of derivatives of one order, SLTB of lower ma-

pitude than the derivatives themselves )is applied, therefore

the system will be,

kherefore equation (5) can be written in the alternate farm,

512

\A consistent imdicit difference scheme

The three implicit difference scheme has been applied for equ-

ations (6) and (7), considering the grid spacing of A 5 and time

intervai A t , (see fig. 2 >.Thus leads to;

AS
and (g), such that;

Equation (10) is found by Taylor's series expansion, to be term

by term consistent with equation (5). The von Neuinann method is

used for examining the stability of the finite deference scheme.

It has been found out that the amplification factor 5 1, that

1 m A S
the original component e will not increase with time.

Therefore the scheme can be considered absolutely consistent to

second order accuracy, i.e. to O (A s z , A tz), and absolutely

 573

unconditionally stable. Thus it is not unexpected as the applied

scheme is implicit.

Alternating Direction Algorithon

Equations (8) and (9) can be written respectiviely in the

following form,

@here, A,B,C and D are coefficients of known values,i.e.,

574

Each of equations of equation5 (11) a d (12) forms a tridiago1

al vector system that may be solved using the aìternating directior

algorithm (e.g. Richtmyer and Mooton,1967 1, by introducing auxili-

ary variable E and P in the x sweep as follows:

Introducing (11) in (131, that provides recurrence relations;

5 = - CJ (AJ El,,+B,)-I , ,$= Cq-A, $+,)(A,%+,+%)
Method of Computation:

Thensgion of interest in the x-y plane in which the numurical

solution of equation (5) is carried out, is divided by a mesh of g~

id lines. The distance between grid lines need not be the same, ths

is considered one of the reasons for recommending this metnod of cc

putption for areas of inadequate hydrological data? Equation (8) is

solved for the x sweep for the time level (9) to the time level

(n + j$ 2, and equation (9) for y-sweep for the time level (n +

to the time level (n 4 1 )

.The recurrence can be initiated from
boundpry conditions of any two - point type; two velocities (con-
tinuiation of the aquifer ), two depths (bounded by rivers ), one

velocity and one depth.

If the boundary is a continuiation of the aquifer, this could

be illustrated as follows;
 575

.s the boundary point is situated at JJ A x, the equation (14) can

expressed as follows,

r the, case where the boundary is bounded by a river, therefore;

The comput.afbncan be applied for the x-sweep by determining the

zfficients P and P from one bounary to the other b o u n d u y and bet-

3n them for all grid points. Thus the new values of the potentials

the time level (n +

-
% )can be determined from the recurrence re-

;ion (13). These new values of the time level (n + M )are used
? the computation of the coefficients of y-sweep and the values of
;ential at the end of time level (n + 1 )have been found out (see
IW chart diagram ). This method is known as the multi - sweep method.
576

Conclusions

A program in AIGOL - 60 has been executed on ICL - 1900 machine

for the water balance equation. The program was tested for different

boundaries (e.g. river, dyke, .*) of different parameters.

As the method of solution is based mainly on the three implicit

difference scheme that there is no conditions for choosing the dis-

tance between the grid points and the time interval A t. U s e this
procedure contains no iterative routines and it has been found out
that this method is exceptionally ecomomical ln computing time and
memory requifements and the solution is considered of high accuracy

Thus this method of computation is recommended for areas of in-

adequate hydrological data.
 577

BOR n = 1 STEP 1 UNTIL nn

i
COU'UTATION OF SwIEE;p IN X - D I W T I Ù N

I
[COMPUTATION i)F h AT n +
.y
$1
1 h[- = J.Km h

,COMPUTATION OF SWBE2 IN Y-DIHECTION

FOR J = 1 STEP 1 UNTIL JJ
c
[COMPUTATIUN FOR E & F FOB kk - 1 to i ]
\'
1
ICOMPUTAIION OF h AT n+l
I

FLOW CRART DIAGRAM

 570

Symbo1e

h :heigtit of water table above the impervious bed.

u,v: the flow components. in x and y directions respectiv;aly.
k :Coefficient of permeability.
:Coefficient of specific yield.
ïV t intensi- of infiltration to the ground water table.
A S :the grid spacing.
A t :the grid spacing on t-axis.
JJ :the number of grid points on x-axis.
kk :the number of grid pointe on y-axis
nn :the number of grid points on t-sis.
3 :any grid point on the x-axis.
k :any grid point on the y-axis.
n :any grid point on the t-axis.

AOKNOWLEDGWT

This work is sponsored in part by the Water Resources Depart-

ment of the Desert Institute, Cairo, Egypt, to which the author is

gratefuì .
 579

Literature

1. Abbott M.B. (1967 ).Difference methods, Lecture note, Inter-

national course of Hydraulic Engineering, Delft, Holland.

2. Mitchell A.R. (1969 )

.Computational methods in partial diff-
erential equations, John Wiley.

3. Nabil Rofail (1972 ).The numerical computation of parabolic

equation using inplicit difference scheme and alternating
direction methods, gth Conference on statistics and computat-
ional science, Cairo, Egypt. pp. 572~-593.

4. Richtmyer R.D. and Mooton K.P. (1967 )

.Difference mothods for
ijlitial value problems, Interscience.

5. Uri Shamir, (1967). The use Of computers in ground water hydrology,

hydro dynamics Laboratory, Beport NQS. 105, Yasaachusetts.
580

3ig.l Diagramatic representation of unconfined aquifer

r:a YR U J+1 J J-1

k-1 n+l
t
cr>
4
k
t
tn
0

4 k-1 n
'COS- AS 3
a
Fig.2 The three level Scheme
 DATA ACQUISITION A N D METHODOLOGY FOR A SIMULATION MODEL
O F THE LLOBREGAT DELTA ( B A R C E L O N A , SPAIN)

Francisco VilarÓ Rigo1 y Emilio Custodio Gimena

A B S TRACT

T h e Llobregat Delta (Barcelona) is a 80 sqKm., area supplying u p to

150 million cubic meters per year o f water for industrial, urban and
agricultural uses, in order of decreasing importance. The construction
o f a exploitation simulation model has been necessary in order t o study
carefully the new problems concequence of a increasing rate o f
abstraction, the conversion of extense irrigation lands in industrial
areas, the dredging o f a new harbor and the forcoming river regulation
with dams. Historical data were initialy scarce. In one hand they were
restricted t o the rainfall and main river discharge knowledge and in
the other hand t o some disperse ground water level data and file o f
well drillers logs without interpretation. After the classification o f
the existing data, some specific studies were iniciated in order t o
complete the knowledge o f the system and finally the model w a s
constructed, followed with an important stage of value adjustment,
specially those related t o intermediate aquitard properties. The
ajusted model has been used in several stages t o forecast t h e response
t o preestablished possible future situations.
Key words: scarce data, model, adjustment, exploration.
I

RESUMEN

El delta del Llobregat (Barcelona) constituye una zona de 80 km2,

que llega a proporcionar hasta 150 millones de m 3 anuales de agua p a r a

tancia. Ha sido necesaria l a construcci6n de un modelo de simulacibn

de la explotaci6n a fin de estudiar con detalle los problemas apareci-
-
usos industriales, urbanos y agrícolas, p o r orden decreciente de i m p o r

dos a causa de la cada vez más intensa explotación, transformación de

áreas agrícolas extensas en industriales, apagado de un nuevo puerto y
próxima regulación d e l río mediante embalses. t o s datos histbricos
existentes inicialmente eran escasos. Por un lado se reducían al cono-
--
cimiento de la pluviometrla y del caudal d e l rio principal y por otro
lado a algunos datos esporadicos de niveles del agua y un archivo de
perfiles de pozos sin elaborar. Se han realizado estudios detallados
orientados a complementar el conocimiento del sistema y finalmente se
--
ha construido el modelo con una importante fase de ajuste de los valo-
r e s estimados, en especial a los referentes a l acuitardo intermedio.
El modelo ajustado ha sido utilizado en varias fases de previsión de
--
respuesta ante determinadas situaciones futuras posibles.
Palabras clave: datos escasos, modelo, ajuste, explotación.

( ) Comisaría de Aguas del Pirineo Oriental y Curso Internacional de

Hidrología Subterránea. Barcelona.
-
582

1.- LOCATION A N D B A C K G R O U N D

T h e B a j o Llobregat is a n a r e a s p r e a d i n g f r o m B a r c e l o n a
E a s t w a r d s and t h e G a r r a f L i m e s t o n e Massive W e s t w a r d s and SW
(Fig. 1). It is l a r g l y o c c u p i e d b y t h e valley of t h e L l o b r e g a t
r i v e r and its delta, w h o s e a l l u v i a l f o r m a t i o n s o c c u p y a r o u n d
80 Km2., o f w h i c h s l i g h t l y o v e r 50 Km2. c o r r e s p o n d t o t h e d e l t a
itself.

T h e proximity t o t h e important u r b a n n u c l e u s of Barcelona,

t h e f e r t i l i t y o f t h e land, t h e easy a v a i l a b i l i t y o f w a t e r and
t h e e x i s t e n c e o f a b i g market f o r its products, h a v e g i v e n r i s e
t o and important a g r i c u l t u r a l and i n d u s t r i a l development. Today,
t h e i n d u s t r y i s t e n d i n g t o t a k e t h e p l a c e o f f a r m i n g and both
a r e r e j e c t e d b y t h e e x p a n d i n g u r b a n area o f t h e t o w n o f Barce-
l o n a . On t h e o t h e r hand,the c u r r e n t e x p a n s i o n o f t h e B a r c e l o n a
h a r b o u r , n e e d s new s e r v i c e a r e a s t o b e p r e p a r e d , w h i c h t h e B a j o
Llobregat easily offers.

T h e p r o b l e m s o f i m p o r t a n t w a t e r extractions, o f increasing
interest in t h e s a n d s and g r a v e l s o f t h e d e l t a and valley f o r
construction, the additional communication lines, the prolifer-
a t i o n of d i s c h a r g e s and t i p p i n g o f a l l c l a s s e s , etc., c r e a t e a
h a r m f u l and a p p r e h e n s i v e climate, and l e a d s t o t h e d e s t r u c t i o n
o f t h e a q u i f e r s b y e m p t y i n g and c o n t a m i n a t i o n and it may p r o d u c e
a d e e p s e a intrusion. I t s r a t i o n a l a d m i n i s t r a t i o n r e q u i r e s a
g o o d k n o w l e d g e o f t h e c h a r a c t e r i s t i c s and h y d r a u l i c o p e r a t i o n
o f t h e a q u i f e r s in t h e area.

I n 1909 a d e t a i l e d s t u d y w a s made o n t h e g r o u n d w a t e r
hydrology of t h e d e l t a (61, b u t t h e s y s t e m a t i c and detailed
s t u d i e s s t a r t e d in 1964, w h i c h i s t h e i n i c i a l point of a s e r i e s
o f ‘ m p o r t a n t w o r k s a n d r e p o r t s w h i c h are partly listed in t h e
references. They h a v e mostly b e e n prepared b y p e r s o n n e l of t h e
G e n e r a l H y d r a u l i c W o r k s Board, t h r o u g h t h e E a s t P y r e n e e s W a t e r
C o m m i t t e e a n d the D e l e g a t i o n in B a r c e l o n a o f t h e P u b l i c W o r k s
G e o l o g i c a l Service.

T h e c o m p l i c a t e d f a c t o r s r a i s e d t h e need t o have a s i m u l a t i o n
m o d e l o f t h e a q u i f e r s y s t e m s available. T h e P u b l i c W o r k s
G e o l o g i c a l S e r v i c e built a R-C (resistors and c a p a c i t i e s ) m o d e l
in 1970 as a first a p p r o x i m a t i o n , and almost s i m u l t a n e o u s l y ,
t h e East P y r e n e e s N a t e r Board and t h e P u b l i c Works G e o l o g i c a l
Service prepared a digital mathematical model o f the explotation,
c a p a b l e of f u r t h e r d e t a i l s a n d m o r e f l e x i b l e u s e (i). T h e m a i n
problem w h e n b u i l d i n g s u c h m o d e l s l i e s in t h e s c a n t y h i s t o r i c a l
d a t a available, s i n c e t h e s y s t e m a t i c c o n t r o l s t u d i e s o f t h e a r e a
w e r e initiated s i s t e m a t i c a l y a f t e r 1966.
 583

2.- AQUIFERS I N T H E AREA

F i g u r e 2, shows t h e general f e a t u r e s of t h e a q u i f e r s in
t h e area, by means of three cross-sections. I n t h e Llobregat
valley, t h e r e is a single a q u i f e r of c o a r s e g r a v e l which d i v i d e s
up in t h e delta entrance, into t w o s u p e r p o s e d ones, s e p a r a t e d
by a silt-clayey intercalation, which i n c r e a s e s in t h i c k n e s s
t o w a r d s t h e sea. T h u s an upper a q u i f e r , which is mostly a w a t e r
t a b l e one,and a deep confined a q u i f e r with a weakly s e m i -
pervious r o o f are separated. T h e silt intercalation n a r r o w s
and becomes sandy towards the delta margins, and finally
disappears, thus a l l o w i n g both a q u i f e r s t o l i e directly a b o v e
one another, and in easy h y d r a u l i c r e l a t i o n ( 8 ) ( 9 ) (14).

T h e a q u i f e r of t h e valley and t h e deep a q u i f e r of t h e d e l t a

present a r e a s of high transmissivity where t h e r e a r e important
pumpings, whereas t h e upper a q u i f e r o f t h e delta support o n l y
r e d u c e d explotation.

Both t h e d e l t a and t h e valley a p e m a r g i n a t e d by m a t e r i a l s

which may b e considered a s impervious.

3.- EXTRACTIONS A N D HYDRAULIC OPERATION

In f i g u r e I, t h e m a i n extractions and hydraulic c o n d i t i o n s

o f t h e m o d e l a r e a were shown. I n t h e delta, t h e t w o l a r g e s t
pumping c e n t r e s a r e f o u n d in Prat d e Llobregat and t h e F r e e Port,
and t h e y g r a v i t a t e on t h e d e e p a q u i f e r ; i n t h e valley t h e y l i e
alongside a l o w e r end (Cornella-Sant J o a n D'Espi) and n e i g h b o u r -
hood of Sant Feliu de Llobregat. Other extraction n u c l e i a r e
f o u n d along t h e S U edge of t h e delta, b e s i d e s o t h e r i s o l a t e d
points, served from both aquifers. T h e u p p e r a q u i f e r of t h e delta
r e c e i v e s an excellent r e c h a r g e from irrigation r e t u r n f l o w a n d
w a s t e water discharge, and it is drained by t h e sea, t h e f i n a l
stretch of t h e river, t h e d r a i n s o f t h e airport and t h e m a r g i n a l
pumping areas. T h e aquifer of t h e valley r e c e i v e s i t s m a i n
r e c h a r g e t h r o u g h r i v e r water i n f i l t r a t i o n and from t h e i r r i g a t i o n
canals, but t h e permeability of t h e beds impedes t h e m a i n t e n a n c e
o f a direct hydraulic connection, and a non-saturated m e d i u n
exists between w a t e r t a b l e a n d t h e bottom of t h e s u r f a c e water.
T h e deep aquifer o f t h e delta r e c e i v e s t h e r e c h a r g e direct f r o m
t h e valley or from t h e upper a q u i f e r in t h e m a r g i n a l a r e a s o r by
v e r t i c a l infiltration t h r o u g h t h e silt lens. T h e s e r e l a t i o n s
and a c t i o n s can be seen i n t h e double piezometric s u r f a c e of
f i g u r e 3, and are reflected in some detailed s t u d i e s b a s e d o n
balance criteria. ( 9 ) (10) (141, h y d r a u l i c c o m p u t a t i o n s ( 1 8 ) ( 1 9 )
a n d g e o h y d r o c h e m i c a l evaluation (2) ( 3 ) ( 5 ) (8).

4.- MOTIVATION O F T H E MODEL

Delta g r o u n d w a t e r explotation f o r i n d u s t r i a l u s e s h a s b e e n
increasing at a rapid pace during the last t e n years, at t h e
5 84

same time as normal extractions f o r the Barcelona supply have

b e e n d r o p p i n g a s a r e s u l t of t h e direct u t i l i z a t i o n o f t h e
r i v e r water, a f t e r a s u i t a b l e treatment. T o t a l e x t r a c t i o n
h o w e v e r h a s g r a d u a l l y i n c r e a s e d and it w i l l r i s e p o s s i b l y in
t h e i m m e d i a t e f u t u r e w h e n it i s n e c e s s a r y t o r e a c t i v a t e t h e
u r b a n supply w e l l s t o meet g r o w i n g âemand. T h e t o t a l c a p a c i t y
o f water stored in t h e a q u i f e r s y s t e m a n d easily m o b i l i z a b l e ,
i s b e t w e e n 100 and 150 m i l l i o n m3., a s m a l l f i g u r e c o m p a r e d
w i t h t h e a n n u a l e x t r a c t i o n w h i c h n o n e x c e e d s 140 m i l l i o n m3.,
and c a n r e a c h 200 w i t h t h e present e x i s t i n g p u m p i n g capacity.
This m e a n s t h a t in t h e a b s e n c e o f r e c h a r g e , in a f e w m o n t h s ,
c e r t a i n p a r t s o f t h e a q u i f e r dry u p o r a r e left w i t h a n i n s u f -
f i c i e n t s a t u r a t e d t h i c k n e s s t o m a i n t a i n v e l 1 discharges. T h e
river infiltration does not increase when the extractions rise,
a s a r e s u l t of its d i s c o n n e c t i o n w i t h t h e w a t e r t a b l e in t h e m a i n
r e c h a r g e a r e a ( c o r r e s p o n d s t o t h e valley), a n d t h e r e is n o o t h e r
i m p o r t a n t r e c h a r g e s o u r c e e x c e p t t h e sea, t h i s i n d u c i n g a
s t e a d i l y a d v a n c i n g s e a w a t e r intrusion. ( 5 ) (20).

T h e s t u d y of t h e effect o f n e w e x t r a c t i o n s o r o f different
n a t u r a l o r a r t i f i c i a l h y d r o l o g i c a l , r i v e r situations, a s a
r e s u l t o f i t s dam r e g u l a t i o n o r w a t e r t r a n s p o r t a t i o n t o o t h e r
a r e a s and a l s o t h e c o n v e r s i o n o f f a r m i n g a r e a s i n t o i n d u s t r i a l
z o n e s , is c o m p l e x . P o r t h i s r e a s o n it w a s d e c i d e d t o built a
s i m u l a t i o n m o d e l t o a n a l y s e t h e e x p l o t a t i o n o f t h e g r o u n d waters,
w h i c h would a l s o h e l p t o a s s e s s t h e r i v e r recharge, t h e s e a
i n t r u s i o n (by indirect e v a l u a t i o n ) and t h e interferences.

T h e different o b j e c t i v e s a n d v a r i a b l e s t o b e e s t i m a t e d m a y
b e s u m m e d u p a s f o l l o w s : (4)

Study of t h e e f f e c t s o f t h e e x p l o t a t i o n in c e r t a i n places,
w i t h o r without d i s a p p e a r a n c e o f s o m e o f t h e present p u m p i n g s .
S t u d y o f t h e a r t i f i c i a l r e c h a r g e e f f e c t s b y s p r e a d i n g and
by wells, and a n a l y s i s o f t h e i r t e c h n i c a l , e c o n o m i c and
legal feasibility.
S t u d y o f t h e e f f e c t s and s u i t a b i l i t y of a r e c h a r g e l i t o r a l
barrier to reduce sea intrusion, in the upper aquifer, in
t h e d e e p one, o r i n both, im s e l e c t e d areas.
Study o f t h e L l o b r e g a t r i v e r r e g u l a t i o n e f f e c t s and/or
d e r i v a t i o n o f l a r g e r d i s c h a r g e s f o r supply.
Study of t h e s u p p r e s s i o n e f f e c t s o f i r r i g a t e d a r e a s o r t h e
m o d u l a t i o n of t h e i r r i g a t i o n discharges.
S t u d y o f t h e g e o t e c h n i c a l p r o b l e m s derived f r o m a b a n d o n m e n t
of t h e m a i n c u r r e n t pumpings.
Study o f t h e o p e r a t i o n of t h e a q u i f e r s a s l o c a l r e s e r v o i r s
f o r t h e most a d e q u a t e s e r v i c e o f demand.
 585

The study of these possibilities includes:

1) Determination o f the external balance,

2) Determination of the internal balance.
3) Estimation of the fresh water discharges into the sea and
river.
4) Estimation of the sea water encroachment areas and their
possible evolution.
5) Estimation of the deficits which may turn up in the
different zones.

5.- MODEL NETWORK

The shape of the piezometric surface, the distribution of

the pumpings, the plant o f the aquifer system and present
knowledge, advised an assymetric network, digital mathematical
model, similar to the one established by the California Water
Resources Department (17) as being the best suited. In accordance
with the already known estimation principles (12) were made the
necessary adaptations for its programming and handling on the
double memory IBM 1630 computer at the Computation Office of the
Public Works Ministry in Madrid, and a series o f special
modifications in the boundary conditions. Its constructive and
network details have been published on several occasions ( 1 ) (4)
(15). In the delta, the two aquifers are simulated by means o f
a double network of polygons (fig. 4) connected by a vertical
conductor branch. The sea condition is established as a constant
level directly for the upper aquifer and by means of a resistent
element (the aquitard) for the deep aquifer. The condition of the
draining river is imposed as om another constant level, and the
river condition in recharge area is established giving a recharge-
discharge set of figures by polygon in terms of the river discharge.

6.- RESOLUTION OF INSUFFICIENCY OF DATA. ADJUSTMENT.

At the time when the need for the model came about, the
number of available data were small, especially regarding the
length of the observation period.

The number of data figures needed is very varied and com-

prises those referring to the geometric form of the aquifers
and their hydraulic parameters, up to those referring to the
temporary and spacial distribution of the extractions, passing
by the infiltration of the rainwater, the river and the irrigations
( 6 ) ( 7 ) and they should have a sufficient precision and represen-
tative nature in accordance with the model network. A set of data
should be available in each node and branch of the projected
model.
586

In this case, the geological structure was well known,

owing to a high number of bore-holes (fig. 3) and wells with
filed lithological log, but not S O the hydraulic characteris-
tics of the different formations. These values were fractionary
and corresponded to some precise data of tests in piezometers
and wells made very ofer under difficult conditions, and some
few prolonged pumping tests, with complicated interpretation
due to the notable piezometric fluctuations that are produced,
wich sometimes exceed a metre throughout the day.

With the available data, a plan of isotransmissivities was

completed and a distribution of the seepage coefficient of the
aquitard (intermediary silts) was estimated, based on a few
granulometric tests and geohydrochemical considerations, which
only gave the approximate magnitude.

Clearly a model built under these circunstances,with a

poorly known connection with the river, and for which there was
only a few partial semi-quantitive figures, mostly obtained by
statistical analysis of the river discharges, supply extractions
and levels in valley ( 1 3 1 , is a long way from reproducing the
reality, An adjustment process is necessary, based on comparison
of its response to actions taken from the historic series and
comparison with the effects observed in the aquifer. These
actions are the extractions and recharges and the effects are
the piezometric levels.

The adjustment process requires a sufficiently long and

complete set of historic data, in order to complete, correct and
suit the imprecise data, or the estimated or non-existent data.
This adjustment process allows some data to be corrected if
other can be taken as sufficiently precise. Otherwise, no
sole situation is reached, or there is no satisfactory solution
nor one which responds to the real conditions of the prototype
or real system. The set of historic data should be for each
polygon, and this is difficult even in well known areas, and
with a good systematic of measurements. In the case of the Llo-
bregat delta, it was decided to take as "exact" data, despite
certain uncertainties in their determination:

a) The extractions by pumping anã the recharges by wells and

drains, using as contrast criterion: for agricultural uses,
the irrigated surface anã calculated water needs; for indus-
trial uses, the type of production, number of workers and
real production in those cases where it was known; and for
supply uses, the urbanistic level and population served.

b) The infiltrations of the rainwater, the irrigation water

and runoff of the nearby areas obtained from water balances
in the soil, and therefore of theoretic type. Nevertheless,
as this is a mild climate area, flat and with classic
irrigation crops, a small error is expected.
 507

c) The water losses of the canals by infiltration based on the

inlet and outlet measurements and the irrigation quantities.
In winter, these irrigation quantities are almost non-existent,
wich permits an acceptable estimation.

d) The piezometric surfaces and hydrograms of available ground

water. Most hydrograms have been obtained with eight water
level recorders, plus daily measurements on another six
piezometers, plus monthly readings on a few more points. The
piezometric surfaces correspond to intense and periodical
measurement campaigns of one or two days duration, but these
may have errors due to variations in the measurement hour,
or introduction of some dynamic data or tridimensional flow
areas; nevertheless they are sufficiently valid.

e) River discharges, obtained with certain guarantee at the

upper valley inlet, in Martorell.

f) Geometric dimensions of the modelled units.

The data to be adjusted are:

1. on the model in itself, based on already mentioned

previous values, and with a pre-established variation
margin, taken from existing knowledge.

- Transmissivities of the surface and deep aquifer, with

reduced variations.

- Vertical permeability of the aquitard (intermediary

silts) for which the previous data could be notably
erroneous.

- Porosity of the water table aquifer, only admitting

slight variations in accordance with the lithology.

- Coefficient of elastic storage of the captive aquifers

within a logical margin according to the existing
structure and figures on the interpretation of pumping
tests and the response to sea tide in some water level
recorders of ground waters.

2. on the actions impossed on the aquifer in the adjustment

period, not directly known.

- River recharge, estimated previously from balances,

simplified analysis of the piezometric surfaces of the
valley and a statistical correlation betwen discharges
of the river-supply extractions and levels in the
valley (13).

- Discharge to the river in the final stretch, estimated

by partial balances and sumplified analysis of the
piezometric surfaces. This is a relatively small value.
588

- Discharge t o t h e s e a a n d s e a w a t e r e n c r o a c h m e n t values,
a c c o r d i n g t o the a q u i f e r and t h e c o a s t l i n e a r e a con-
sidered. M e a s u r e d very roughly due t o e s t i m a t i o n dif-
ficulties, e x c e p t i n g t h e c e n t r a l c o a s t a l s t r e t c h o f
the w a t e r t a b l e aquifer.

T h e d i s t r i b u t i o n o f t h e r e c h a r g e b e t w e e n t h e u p p e r and deep
a q u i f e r o f t h e d e l t a i s a r e s u l t of t h e a d j u s t m e n t , and a l s o i s
t h e w a t e r d i s c h a r g e c i r c u l a t i n g t h r o u g h t h e aquitard.

T h e c h i e f d i f f i c u l t i e s r e g a r d i n g t h e a d j u s t m e n t are derived
f r o m i n s u f f i c i e n t d a t a on l e v e l s a n d a n e e d t o account o n t h e
s e a s o n a l v a r i a t i o n s , o w i n g t o t h e great i m p o r t a n c e o f e x t r a c t i o n s
i n r e l a t i o n w i t h t h e q u i c k l y m o b i l i z a b l e g r o u n d s t o r a g e volume
o f w a t e r . T h e first p i e z o m e t r i c s u r f a c e u s e a b l e is at t h e b e g i n -
n i n g o f 1966, and a l t h o u g h a n o t h e r s i x c o m p l e t e o n e s and one
partial one were available, their distribution was neither regular
i n time, n o r c o v e r e d each of t h e q u a r t e r l y p e r i o d s i n t o w h i c h
t h e y e a r w a s t o b e d i v i d e d up. It w a s t h e r e f o r e decided t o u s e
the a v a i l a b l e p i e z o m e t r i c s u r f a c e s , w i t h m i n o r c o r r e c t i o n s t o
a d o p t t h e m t o t h e f i n a l m o m e n t of e a c h q u a r t e r l y interval,
forming new interpolated piezometric surfaces, based on the data
o f t h e c o n t i n u o u s p i e z o m e t r i c m e a s u r e m e n t s in s o m e p o i n t s , already
d i s c u s s e d , t r y i n g t o m a i n t a i n t h e flow s h a p e character.

T o complete the quarter figures, the water balance estimations

w e r e m a d e in the s o i l , and t h e e x t r a c t i o n s w e r e c a l c u l a t e d from
t h e i n v e n t o r y a c c o r d i n g t o t h e a n n u a l rate o f u s e and moment t h e
w e l l s went i n t o o p e r a t i o n i n s o m e c a s e s , or b a s e d o n t h e demand
c u r v e s given by some users.

In f i g u r e 5, a s a m p l e o f t h e r e s u l t of t h e f i n a l adjustment
p r o c e s s c a n b e seen, t a k i n g t h e 4 y e a r s o f f i g u r e s , d i s t r i b u t e d
i n t o 16 quarterly t e r m s . T h i s f i n a l a d j u s t m e n t need 1 3 s t a g e s
w i t h t h e d e f i n i t i v e network. S o m e p r i o r t r i a l s w e r e m a d e w i t h a
s i m p l i f i e d n e t w o r k , t o know t h e m a g n i t u d e and c o n v e r g e n c e r a t e s
of the estimation process. This adjustment stage incorporated
the m o d i f i c a t i o n s s u g g e s t e d by t h e p r e v i o u s one, m a i n l y m o d i f y i n g
t h e h y d r a u l i c c h a r a c t e r i s t i c s o f t h e u n i t s a n d t h e r e c h a r g e o f the
river. Before making a modification, the results obtained were
c a r e f u l l y s t u d i e d , t a k i n g i n t o account p r e v i o u s r e s u l t s w i t h early
s t a g e s , and in o r d e r t o b e i n a c c o r d a n c e w i t h t h e p h y s i c a l c h a r a c -
teristics of the system.

An i m p o r t a n t r e s u l t of t h e a d j u s t m e n t p r o c e s s is not only t h e
c o r r e c t i o n o f t h e i m p r e c i s e data, b u t o b t a i n i n g o t h e r n e c e s s a r y
d a t a f o r t h e m o d e l e x p l o t a t i o n p r o c e s s , p r e v i o u s l y unknown. Such
is the relation Q r (river discharge) versus I R , thus allowing
t h e ( r i v e r i n f i l t r a t i o n ) c o m p u t a t i o n o f I R ( n o t m e a s u r a b l e ) with
a v a i l a b l e data on QR. T h e a d j u s t m e n t o b t a i n e d s h o w s t h e r e i s t h i s
relation with a sufficient statistical degree of significance.
 589

Figure 6 shows the inicial map of transmissivities of the

deep aquifer and the valley gravels and the one obtained after
the adjustment. The differences are not important and in many
cases, the variations are not merely a correction of an erroneous
value, but the adaptation of a precise value (test in bore hole
or well) or of a regional value (pumping test or analysis of
piezometric oscillations) to the dimensions and forms of each
polygon.

7.- UTILIZATION OF THE MODEL

The model has been built for use under different conditions
as those prevailing during theadjustmen process. This creates
various problems. For example, the validity of the model for
other distributions of the pumping or recharge-disoharge, or
those corresponding to piezometric surfaces notably different.
Also, one must consider the validity of the Q r- 1, (river
discharge - river recharge) relation, under different circumstances
of the river system or after conditioning works in the bed. The
adjustment period is rather short, but sufficient to insure
credible results under conditions similar to the adjustment
ones and in time periods not much greater. I f one attempt to
simulate 20 years or under pumping conditions with other centres
of extraction, noticeably different as those existing now, the
results would possibly only be semi-quantitative.

One of the recent processes of using the model arose to study

the possibility of temporarily increasing the groundwater
extractions for supply, in the event of a succession of dominatly
dry year combined with a delay in the first service of the new
surface water regulation works of the Llobregatriver (ll), taking
into account the normal pumping increase forecastsfor other pur-
poses. The injuries and needs of redistribution and conditioning
of the pumpings under various foreseen hypothesis have been
assessed, and the sea water encroachment and the later return to
a "normal" situation after these regulation works have been
finished. Some of the possible extraction situations have not
been made as they produce excessive drops which prevent the
pumping capacity of the wells to be maintained.

The use of the model permits the aquifer system of the Bajo
Llobregat to be handled as a regulating reservoir, analysing
the guarantees of the different ground water demands in different
natural or man-made hydrological situations, and a knowledge of
the rate and location of the progressive salinization process or
the effectiveness of the measures adopted to reduce it. These
eventualities were analysed by seven different hypothesis
following the adjustment process ( 1 ) (15), including the analysis
of the possible artificial recharge. The model, in its explotation
phase, works with s i x monthly intervals instead of the quarterly
intervals of the adjustment.
590

8.- C O N C L U S I O N

The c a r e f u l 1 m o d e l l i n g o f an a q u i f e r p e r m i t s a very u s e f u l
w o r k t o o l t o b e o b t a i n e d , e v e n t h o u g h t h e i n i t i a i d a t a is
i n c o m p l e t e or n o n - e x i s t e n t in c e r t a i n a s p e c t s , p r o v i d e d a n o t h e r
s e r i e s o f s u f f i c i e n t l y p r e c i s e d a t a , or with k n o w n e r r o r is
a v a i l a b l e , a n d w h i c h is s u c h t h a t it p e r m i t s a n adjustment
p r o c e s s with a s u f f i c i e n t n u m b e r o f steps.

9. - REFERENCES

1. C u e n a , J. and C u s t o d i o , E. (1971) - C o n s t r u c t i o n and adjustment

o f a t w o l a y e r m a t h e m a t i c a l m o d e l o f the L l o b r e g a t Delta.-
I n t e r n a t i o n a l S y m p o s i u m on M a t h e m a t i c a l M o d e l s in Hydrology.-
I n t e r n a t i o n a l A s s o c i a t i o n o f S c i e n t i f i c Hydrology.- Varsovia.

2. C u s t o d i o , E. (1967) - Etudes geohydrochimiques dans le delta

du Llobregat, Barcelone (Espagne) - Geochimie, Precipitations,
H u m i d i t é du Sol, H y d r o m e t r i e . A s s e m b l é e G é n é r a l e d e Berne,
1967. A s s o c i a t i o n I n t e r n a t i o n a l d ’ H y d r o l o g i e Scientifique.-
Gentbrugge - pp. 134/155.

3. C u s t o d i o , E. (1969) - Ground water entries in the Llobregat

r i v e r delta.- H y d r o l o g i c a l R e a s e a r c h D o c u m e n t s No. 6 , B a r c e -
l o n a W a t e r R E s e a r c h , A p p l i c a t i o n s and S t u d i e s Centre. Speech
in P a m p l o n a 1967 - pp 205/237.

4. C u s t o d i o , E., C u e n a , J . a n d BayÓ, A. (1971) -

Problem,
e x e c u t i o n and u s e o f a t w o l a y e r m a t h e m a t i c a l m o d e l f o r t h e
L l o b r e g a t d e l t a a q u i f e r s (Barcelona).- First S p a n i s h
tuguese - A m e r i c a n C o n g r e s s on E c o n o m i c Geology.- Madrid-
-
Por-

Lisbon. s e p 1971. S e c t i o n 3. pp 171/198.

5. C u s t o d i o , E., Bayo, A . and P e l a e z , M.D. (1971) -Geochemistry

a n d w a t e r e n t r i e s f o r s t u d y o f the m o v e m e n t o f g r o u n d w a t e r
in t h e L l o b r e g a t r i v e r d e l t a ( B a r c e l o n a ) - First S p a n i s h -
P o r t u g u e s e - A m e r i c a n C o n g r e s s on E c o n o m i c Geology.- Madrid-
Lisbon. s e p . 1971. S e c t i o n 6 pp 51/80.

6. C u s t o d i o , E. and L ó p e z - G a r c i a , L. (1972) - C o n s t r u c t i o n and

u t i l i z a t i o n p r o c e s s o f a model. C h a p t e r 5 B a s i c T h e o r y on
A n a l o g i c a l and D i g i t a l M o d e l s of Aquifers. I n f o r m a t i o n s and
S u t i d e s , B u l l e t i n , 37, G e o l o g i c q l S e r v i c e o f P u b l i c Works.
Madrid.

7. C u s t o d i o , E. (1973) -
Basic data for building an aquifer
s i m u l t a t i o n model. C h a p t e r 16.1. on S u b t e r r a n e a n Hydrology
O m e g a Editorial. B a r c e l o n a (at press).
 591

8. Custodio, E. and others (1973) -

Compiling of works made
during the period 1966/1972 in the Bajo Llobregat. Water
Board of the East Pyrenees and Public Works Geological
Service. Barcelona (in preparation).

9. Llamas, M.R. and Molist, J. (1967) -

Hydrology of the Besos
and Llobregat River deltas.- Hydrological Investigation
Documents NQ 2 - Water Research, Applications and Studies
Centre. Barcelona. Speech in Barcelona (1966).

10. Llamas, M.R. and VilarÓ, F. (1967) -

Die Rolle der Grund-
wasserspeicher bei der Wasservorsorgung von Barcelona.-
Das Gas-und-Wasserfach, Wasser Abwasser, Vol. 34. No. 1 5 ,
August 1967. pp. 945/953.

11. Martin-Arnaiz, M. (1972) -

Report on the explotation
possibilities of the Llobregat river delta aquifers. General
Board of Hydraulic Works. East Pyrenees Water Board. Barce-
lona (prior report).

12. Mc Neal, R.M. (1958)

network -
-
An asymetrical finite difference
Quarterly of Applied Mathematics. Vol. XI. No. 3
1958.

13. Montalbán, F. (1969) -Factorial analysis of the oscillations

o f the deep aquifer of the Llobregat river. Hydrological
Investigation Documents No. 6. Water Research, Applications,
and Studies Centre. Barcelona, Pamplona speech (1967).

14. Ministry of Public Works (1965).- Study of the Total Hydraulic

resourcs of the East Pyrenees -
Second Report East Pyrenees
Water Board and Public Works Geological Service. Barcelona.

15. Ministry of Public Works -

Report on the construction and
application of a mathematical simulation model of the Llo-
bregat delta aquifers.- Study of the Total Hydraulic Resources
of the East Pyrenees. Central Area. Report CE-111.- East
Pyrenees Water Board and Public Works Geological Service.
Barcelona.

16. Santa Maria, L. and Marin A. (1909) -

Hydrological studies
on the Llobregat river basin.- Bulletin of the Commission
of the Geological Map of Spain LX 2nd Series.

17. Tyson, H.N. and Weber, E.M. (1964).- Ground water management
for the nations future computer simulation of ground-water
basins - proceedings of the ASCE, Journal of the Hydraulics
Division. New York. Jyly 1964.

18. VilarÓ, F. (1967) -Balance of the present use of the Bajo

Llobregat. Hydrological Investigation Papers No. 2. Water
Investigations, Applications and Studies Centre. Barcelona
Speech in Barcelona, (1966) -
pp 155/169.
592

19. VilarÓ, F. and Martin Arnbiz, M. ( 1 9 6 8 )

of t h e B a j o Llobregat - -
Hydric Balance
Hydric B a l a n c e S e m i n a r F.A.O. - -
Geology and Mining I n s t i t u t e of Spain. Madrid.

20. VilarÓ, F. Custodio, E., -

and B r u i n g t o n , A.E. ( 1 9 7 0 )
W a t e r i n t r u s i o n and w a t e r pollution i n t h e P i r i n e o Oriental
Sea

(Spain) -
Memphis, Tennence.
ASCE N a t i o n a l Water Resources Engineering Meeting,
- M e e t i n g Preprint 1122.
Fi g. 1 .- Plano general de situaci& y de extracciones.
General location and pumping map.
594
O ' ' . I ' U
O O 0 0 O O
Yi
- s
sariaw soiiaui
'"" S ~ J I ~ Usoiiaw
I -
r
 595

O
- Escal o - S c a l e
1 2 3 L SKm.

L o g u n a s pantanosas natural es
Limite d e los z o n a s p e r m e a b l e s
Limite del o c u i f e r o p r o f u n d o
-2- Isopieza del acuifero profundo I ml.
..2-.- Isapieza del acuifero suprficial [ m 1.
e S o n d e o piezomctrico

Natural m a r s h y lagoons
-.-.-
B o u n d a r y of t h e p e r m e a b l e o r e a s
B o u n d a r y of t h e d e e p a q u i f e r
A- Isopiestic line O f the drepoquifer(ml
,-2--- Isopiestic line of t h e upperoquiferIm)
O b s e r v a t i o n b o r e - hole

Fi g. 3 - Superficies PiezornCtricas en Abril de 1.967

situaci& de los sandeos.
(s& Custodio) y

Piezometric surfaces in April 1.967 (after Custodio) and 1 0 C b

tion of the boreholes.
596
I
 597
/* :
Y--
I
I in
- a
.e
I u
I o 4
I C
in
c
598

Tranrmirividad del aeuifero del

valle y profundo del delta e n m2/dia
Transmissivity of valley a n d delta
d e e p aquifers in s q m l d a y
--i-d- - - Dato inicial Preliminary figure
1000 Valor ajustadocon Value m s t c h c d with the
el modelo model

Fig. 6.- Valores de la tranrrtirividaà del acuftero del valle y prohrado &l dal-
ta del Llobregat.
Values of the valley aad delta upper aquifers OP the Llobregat delta.
 and in Etudes et rapports d ‘hydrataptie 16

gn of
r-resources projects
inadequate data
Proceedings of the Madrid Syinposiurn
June 1973

Elaboration des projets

d’utilisation des ressources en eau
sans données suffisantes
Actes du colloque de Madrid
Volume 2 Juin 1973

-
Unesco W M O - IAHS
Unesco - O M M - AISH
Studies and reports in hydrology/Etudes et rapports d’hydrologie 16
TITLES IN T H I S SERIES / DANS CETTE COLLECTION

1. The use of analog and digital computers in hydrology: Proceedings of the Tucson Symposium.
June 1966 / L'utilisation des calculatrices analogiques et des ordinateurs en hydrologie: Actes du
colloque de Tucson, juin 1966. Vol. 1 & 2. Co-edition IAHS-Unesco / Coédition AISU-Unesco.
2. Water in the unsaturated zone: Proceedings of the Wageningen Symposium, June I967 / L'eau dans
la zone non saturée: Actes du symposium de Wageningen, juin 1967. Edited by / Edité par P. E.
Rijtema & H . Wassink. Vol. 1 & 2. Co-edition IAHS-Unesco / Coédition AISH-Unesco.
3. Floods and their computation: Proceedings of the Leningrad Symposium, August 1967 / Les crues
et leur évaluation: Actes du colloque de Leningrad, août 1967. Vol. 1 & 2. Co-edition IARS-Unesco-
W M O / Coédition AISH-Unesco-OMM.
4. Representative and experimental basins: A n international guide for research and practice. Edited
by C. Toebes and Y.Ouryvaev. Published by Unesco.
4. Les bassins représentatifs et expérimentaux: Guide international des pratiques en matière de re-
cherche. Publié sous la direction de C. Toebes et V. Ou-vaey. Publié par l'Unesco.
5. 'Discharge of selected rivers of the world / Débit de certain cours d'eau du monde. Published by
Unesco / Publié par l'Unesco.
Vol. I : General and régime characteristics of stations selected 1 Caractéristiques générales et
caractéristiques du régime des stations choisies.
Vol. II: Monthly and annual discharges recorded at various selected stations (from start of obser.
vations up to 1964) / Débits mensuels et annuels enregistrés en diverses stations sélectionnées
(de l'origine des observations à l'année 1964).
'Vol. III: Mean monthly and extreme dlscharges (1%5-1969) / Débits mensuels moyens et débits
extrêmes (19651969).
6. List of International Hydrological Decade Stations of the world / Liste des stations de la Décennie
hydrologique internationale existant dans le m m d e . Published by Unesco 1 Publié par l'Unesco.
7. Ground-water studies: A n international guide for practice. Edited by R. Brown, I. Ineson. V. KO-
noplyantsev and V. Kovalevski. (Will also appear in French, Russian gnd Spanish / Paraitrg
également en espagnol, en français et en russe.)
8. Land subsidence: Proceedings of the Tokyo Symposium, September 1969 / Affaisement du sol:
Actes du colloque de Tokyo, septembre 1969. 'Vol. 1 & 2. Co-edition IAHS-Unesco / Coédition
AISH-Unesco.
9. Hydrology of deltas: Proceedings of the Bucharest Symposium, May 1969 / Hydrolaße des deltas:
Actes du colloque de Bucarest, mai 1969. Vol. 1 & 2. Co-edition IAHS-Unesco / Coédirion AISH-
Unesco.
10. Status and trends of research in hydrology / Bilan et tendances de la recherche en hydrologic.
Published by Unesco 1 Publié par l'Unesco.
11. World water balance: Proceedings of the Reading Symposium, July 1970 / Bilan hydrique mondial:
Actes du colloque de Reading, juillet 1970. Vol. 1-3. Co-edition ZAHS-Unesco-WhfO 1 Coédirion
AISH-Unesco-OMM.
12. Results of research on representative and experimental basins: Proceedings of the Wel1inp;ton
Symposium, December 1970 / Résultats de recherches sur les bassins représentatifs et ex érimen-
taux: Actes du cowoque de Wellington, décembre 1970. 'Vol. 1 & 2. Coedition IAHS-Jnesco /
Coédition AISH-Unesco.
13. Hydrometry: Proceedings of the Koblenz Symposium, September 1970 / Hydrométrie: Actes du
colloque de Coblence, septembre 1970. Co-edition ZAHS-Unesco-WMO / Coédition AISH-Unesco-
OMM.
14. Hydrologic information systems. Co-edition Unesco-WMO.
1s. Mathematical models in hydrology: Proceedings of the Warsaw Symposium, July 1971 / Les mc-
deles mathématiques en hydrologie Actes du colloque de Varsovie, juillet 1971. Vol. 1-3. Co-
edition IAHS-Unesco-WMO / Coédit on AISH-Unesco-OMM.
16. Design of water resources projects with inadequate data: Proceedings of the Madrid sym ,
June 1973 / Elaboration des projets d'utilisation des ressources en eau sans données sufp:z:
Actes du colloque de Madrid, juin 1973. Vol. 1-3.Co-edition Unesco-WMO-IAHS/ Coéditiori Unesco-
OMM-AISH.
Design of
water resources projects
with inadequate data
Proceedings of the Madrid Symposium
June 1973

Elaboration des projets

d’utilisation des ressources en eau
sans données suffisantes
Actes du colloque de Madrid
Juin 1973

A contribution to the Lnternationat Hydrological Decade

Une contribution a la Ecennie hydrologique internationale
Con reshmenesen csuañol

Volume 2

Unesco - W M O IAHS 1974 -

Uiiesco - O M M - AISH
Published jointly by
the United Nations Educational,Scientific
and Cultural Organization,
7, Place de Fontenoy, 75700 Paris, >
3)

World Meteorological Organization,

41 av. Giuseppe-Motta,Geneva,and
the International Association of Hydrological Sciences (President:J.-A.Rodier),
19,rue Eugène-Carrière,75018 Paris

Publié conjointement par

l’organisationdes Nations Unies pour
l‘éducation,la science et la culture.
7,place de Fontenoy,75700 Paris.
l’organisationmétéorologique mondiale,
41, av.Giuseppe-Motta,Genève,et
l’Association internationale des sciences hydrologiques (président: 3.-A. Rodier),
19,rue Eugène-Carrière.75018 Paris

Impreso por el Centro de Estudios Hidrográficos,Madrid

The selection and presentation of material and the’opinions expressed in this publication
are the responsibility of the authors concerned and do not necessarily reflect the
views of the publishers.
The designations employed and the presentation of the material do not imply the
expression of any opinion whatsoever on the part of the publishers concerning the legal
status of any country or territory,or of its authorities,or concerning the frontiers
of any country or territory.

Le choix et la présentation du contenu de cet ouvrage et les opinions qui s‘y

expriment n’engagentque la responsabilité des auteurs et ne correspondent pas
nécessairement aux vues des éditeurs.
Les dénominations employées et la présentation des divers éléments n’impliquent
de la part des éditeurs aucune prise de position à l’égard du statut juridique de l’un
quelconque des pays et territoires en cause,de son régime politique ou du tracé
de ses frontières.

ISBN 92-3-001137-1
0 Unescc-WMO-IAHS-1974
Printed in Spain
 PREFACE

The International Hydrological Decade (IHD)1965-74 was launched by

the General Conference of Unesco at its thirteenth session to promote
international co-operationin research and studies and the training of spe-
cialists and technicians in scientific hydrology. Its purpose is to enable
all countries to make a fuller assessment of their water resources and a
more rational use of them as man’s demands for water constantly increase
in face of developments in population, industry and agriculture. In 1974
National Committees for the Decade had been formed in 108 of Unesco’s
131 Member States to carry out national activities within the programme
of the Decade. The implementation of the programme is supervised by a
Co-ordinatingCouncil, composed of 30 Member States selected by thc Ge-
neral Conference of Unesco, which studies proposals for developments
of the programme, recommends projects of interest to all or a large
number of countries,assists in the development of national and regional
projects and co-ordinates international co-operation.
Promotion of collaboration in developing hydrological research techni-
ques, diffusing hydrological data and planning hydrological installations
is a major feature of the programme of the IHD which encompasses all
aspects of hydrological studies and research. Hydrological investigations
are encouraged at the national, regional and international level to streng-
then and to improve the u6e of natural resources from a local and a global
perspective. The programme provides a means for countries well advanced
in hydrological research to exchange scientific views and for developing
countries to benefit from this exchange of information in elaborating re-
search projects and in implementing recent developments in the planning
of hydrological installations.
As part of Unesco’s contribution to the achievement of the objectives
of the IHD the General Conference authorized the Director-General to
collect, exchange and disseminate information concerning research on
scientific hydrology and to facilitate contacts between research workers
in this field.To this end Unesco initiated two series of publications: Studies
and Reports in Hydrology and Technical Papers in Hydrology.
The Studies and Reports in Hydrology series, in which the present
volume is published, is aimed at recording data collected and the main
results of hydrological studies undertaken within the framework of the
Decade, as well as providing information on research techniques. Also
included in the series are proceedings of symposia. Thus, the series com-
prises the compilation of data, discussions of hydrological research techni-
ques and findings,and guidance material for future scientific investigations.
It is hopped that the volumes will furnish material of both practical and
theoretical interest to hydrologists and governments participating in the
IHD and respond to the needs of technicians and scientists concerned
with problems of water in all countries.
A number of these volumes have been published jointly with the In-
ternational Association of Hydrological Sciences and the World Meteoro-
logical Organization which have co-operated with Unesco in the imple-
mentation of several important projects of the IHD.
 PRÉFACE

La Conférence générale de l’Unesco,à sa treizième session, a décidé

de lancer,pour la période s’étendantde 1965 à 1974,la Décennie hydrologique
internationale (DHI),entreprise mondiale visant a faire progresser la con-
naissance en matière d’hydrologie scientifique par un développement de
la coopération internationale et par la formation de spécialistes et de
techniciens.Au moment où l’expansion démographique et le développement
industriel et agricole provoquent un accroissement constant des besoins
en eau, la DHI permet à tous les pays de mieux évaluer leurs ressources
hydrauliques et de les exploiter plus rationnellement.
I1 existe actuellement dans i08 des 131 Etats membres de l’Unesco un
comité national qui, pour tout ce qui a tratit au programme de la Décen-
nie, impulse les activités nationales et assure la participation de son pays
aux entreprises régionales et internationales. L’exécution du programme
de la DHI se fait sous la direction d’un Conseil de coordination composé
de 30 Etats membres désignés par la Conférence générale de l’Unesco;ce
conseil étudie les propositions concernant le programme, recommande
l’adoptionde projets intéressant l’ensemble des pays ou un grand nombre
d’entre eux, aide à la mise sur pied de projets nationaux et régionaux, et
coordonne la coopération à l’échelon international.
Le programme de la DHI qui porte sur tous les aspects des études et
des recherches hydrologiques, vise essentiellement à développer la col-
laboration dans la mise au point des techniques de recherches, dans la
diffusion des données hydrologiques, dans l’organisation des installations
hydrologiques. I1 encourage les enquêtes nationales, régionales et interna-
tionales tendant à accroître et à améliorer l’utilisation des resources na-
turelles,dans une perspective locale et générale. I1 permet aux pays avancés
en matière de recherches hydrologiques d’échanger des informations; aux
pays en voie de développement, il offre la possibilité de profiter de ces
échanges pour élaborer leurs projets de recherches et pour planifier leurs
installations hydrologiques en tirant parti des acquisitions les plus récentes
de l’hydrologie scientifique.
Pour permettre a l’Unesco de contribuer au succès de la DHI,la Con-
férence générale a autorisé le Directeur générale à rassembler, à échanger
et à diffuser des informations sur les recherches d’hydrologie scientifique
et à faciliter les contacts entre les chercheurs dans ce domaine. A cette
fin,l’Unesco fait paraître deux nouvelles collections de publications: (< Etu-
des et rapports d’hydrologie»et «Notes techniques d’hydrologie».
La collection N Etudes et rapports d’hydrologie»,dans laquelle est publié
le présent ouvrage, a pour objet de présenter les données recueillies et les
principaux résultats des études effectuées dans le cadre de la Décennie
et de fournir des informations sur les techniques de recherche. O n y trouve
aussi les Actes de colloques réunis sur ce sujet. Cette collection publie
donc des données, des techniques et des résultats de recherches ainsi
qu’une documentation pour les travaux scientifiques futurs.
O n espère que ces volumes apporteront aux hydrologues et aux gou-
vernements qui participent à la DHI des matériaux d’un intérêt tant pra-
tique que théorique, et qu’elle répondra aux besoins des techniciens et
des h o m m e s de science de tous pays qui s’occupent des problèmes de l’eau.
Certains de ces ouvrages sont publiés en coopération avec l’Association
internationale des sciences hydrologiques ou I’Organisatioii mMorologique
mondiale dans le cadre de projets réalisés conjointement par ces orga-
nisations et l’Unesco.
Design d water resources projecis with inadequate data: P-dings d the Madrid aympoaium,
June 1973 / Elaboration des projeta d'utilisation dei resswTas en eau aona d o n k a auffluntes:
Actea du wlloque de Madrid. juin 1973

Volume II Contents Table des matidres

Foreword/Avant-propos

TOPIC II.1A .METHODS FOR STUDIES LN DATA-SCARCEAREAS AND

THE INFLUENCE ON DESIGN OF THE INADEQUACY OF
DATA FOR PURPOSES OF PLANIFICATION OF WATER
RESOURCES (EXCLUDINGFLOOD AND LOW FLOWS).
METHODOLOGY FOR ASSESSING HYDROLOGICAL CHA-
RACTERISTICS IN DATASCARCE AREAS.
POINT II.1A .METHODES D'ETUDES UTILISEES DANS LES REGIONS OU
LES DONNEES SûNT INSUFFISANTESET INFLUENCE SUR
LE CALCUL DU PROJET DU MANQUE DE DONNEESPOUR
L'ELABORATION DES PROJETS DE L'UTILISATION DES
RESSOURCES EN EAU (A L'EXCLUSION DES CRUES ET
DES DEBITS DE BASSES EAUX).
METHODOLOGIE POUR L'EVALUATION DES CARACTE-
RISTIQUES HYDROLOGIQUES DANS LES REGIONS OU
LES DONNEES SûNT RARES.

BASSO,EDUARDO. (UNDPMIMO)GENERAL REPORT 1

ABIODUM,ADIGUN ADE. (NIGERIA)

Water resources projects in Nigeria and the hydrological data employed in
their planning and development ................................ 21

BASSO, E., ARRIAGADA, A., NEIRA, H.,PEREZ DELGADO, M . (COSTA

RICA)
An example of regional co-operation for improving the hydrologicaland
meteorologicalinformation ................................... 35

CUBAS GRANADO,FRANCISCO.(SPAIN)
Existing methodology for estimating free water surface evaporation .... 59

CUSTODIO,EMILIO.(SPAIN)
Geohydrological studies in smallareas without systematic data ........ 77

DALINSKY,JOSEPH S. (ISRAEL)
Methods of analysing deficient discharge data in arid and semi-arid zones
for the design of surface water utilization ....................... 95

D'OLIVEIRA,EMILIO EUGENIO.MIMOSO,JOAO JOSE.(PORTUGAL)

Mapai river hydrological study (Limpopo'sriver) ................... 141
D'OLIVEIRA,EMILIO EUGENIO.MIMOSO,JOAO JOSE.(PORTUGAL)
Application of Coutagne's and Turc's formulas to southern Mozambique
rivers ................................................... 121
HERAS,R. (SPAIN)
Report hydrological programa of the Center for Hydrographic Studiesfor
the investigation of hydraulic resources with insufficient data ......... 155
KARAUSHEV,A.V.,BOGOLIUBOVA,I.V.(U.S.S.R.)
Computation of reservoin wdLnrntition ......................... 199

KLIGUE,R.K.,MECHDI EL SACHOB (U.S.S.R.)

CilnilritionofrunoffinIraq .................................. 207

KUZMIN,P.P.,
VERSHININ,A.P.(U.S.S.R.)
Determination of evaporation in caw of the abmnce or inadequacy of
data ..................................................... 217

PENTA,A.,ROSSI,F.(ITALY)
Objective criteria to daclare a aerier of data sufficient for technical pur-
poses .................................................... 221

QUINTELA GOIS,CARLOS.(PORTUGAL)
Objective criteria used in hydrology with inadequate data ............ 24 1

SMITH,ROBERT L.(U.S.A.)
Utilizing climatic data to appraise potentiai water yields ............. 253

STANESCU,SILVIU.(COLOMBIA)
Determination of hydrological characteristics in points without direct
hydrometricdata ........................................... 265

TEMEZ,J.R.(SPAIN)
New models of frequency law of runoff starting from precipitations .... 287

TRENDEL, R.,DER MEGREDITCHIAN,G.,RULLIERE,MARIE CLAIRE.

(FRANCE)
Traitement opérationnel des données pluviométriquesentachées d'erreurs
ouinsuffisantes ............................................ 301

TOPIC II.1B .METHODS FOR STUDIES IN DATA SCARCE AREAS AND

THE INFLUENCE ON DESIGN OF THE INADEQUACY OF
DATA FOR PURPOSES OF PLANIFICATION OF WATER
RESOURCES (EXCLUDING FLOOD AND LOW FLOWS).
INFLUENCEOF INADEQUACY OF HYDROLOGICAL DATA
ON PROJECT DESIGN AND FORMULATION.
POINT II.1B- METHODES D'ETUDES UTILISEES DANS LES REGIONSOU
LES DONNEES SONT INSUFFISANTESET INFLUENCESUR
LE CALCUL DU PROJET DU MANQUE DE DONNEESPOUR
L'ELABORATION DES PROJETS DE L'UTILISATION DES
RESSOURCES EN EAU (A L'EXCLUSION DES CRUES ET
DES DEBITS DE BASSES EAUX).INFLUENCE DU MANQUE
DE DONNEES HYDROLOGIQUES SUR LE CALCUL DU
PROJET ET SA FORMULATION.

BEARD,L.R.(U.S.A.)
GENERAL REPORT 315
BANERJI,S.,LAL,V.B.(INDIA)
Design of water resources projects with inadequate data in India. General
&Particular Case Studies ................................... 323
JAMB,IVAN C.(U.S.A.)
Data requirements for the optimization of reservoir dengn and operating
dedetermination .......................................... 335

REID,GEORGE W.(U.S.A.)
The design of water quality management projecta with inadequate data 349

SABHERWAL,R.K.(INDIA)
Designing projects for the development of ground water resources in the
alluvial plains of northern India on the basis of inadequate data ....... 365

SEXTON,J.R.,
JAMIESON,D.G.(U.K.)
Improved techniques for water resource systems design ......... 383

WEBER,J.,KISIEL,CHESTER C.,DUCKSTEIN,LUCIEN (U.S.A.)

Maximum information obtainable from inadequate design data: from
multivariate to Bayesian methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401

TOPIC 11.2- CURRENT PRACTICES FOR ASSESSING DESIGN FLOODS

AND DESIGN LOW FLOWS,INCLUDING THE USE OF
SYNTHETICUNIT HYDROGRAPH,WITH PARTICULAR
EMPHASIS ON MAXIMALISATION AND MINIMALISATION.
POINT 11.2 - PRATIQUES COURANTES POUR L'EVALUATION DES
CRUES ET DES DEBITS D'ETIAGES PRIS EN COMPTE DANS
LE PROJET,COMPRENANT L'EMPLOI D'HYDROGRAMMES
UNITAIRES DE SYNTHESE,AVEC ETUDE PARTICULIERE
DE LA MAXIMALISATIONET DE LA MINIMALISATION.

ROCHE,MARCEL.(FRANCE)GENERAL REPORT 419

BATLLE GIRONA,MODESTO.(SPAIN)
Estimation of floods by means of their silt loads ................. 439

BERAN,M.A.(U.K.)
Estimation of design floods and the problem of equating the probability
ofrainfailandrunoff ........................................ 459

DAVIS,DONALD R.,DUCKSTEIN,L.,KISIEL,CHESTER C.,FOGEL,MAR-

TIN M. (U.S.A.)
A decision-theoretic approach to uncertainty in the return period of
maximum flow volumes using rainfall data ....................... 473

HALL,M.J.(U.K.)
Synthetic unit hydrograph technique for the design of flood alleviation
works in urban areas ......................................
485

HELLIWELL,P.R., CHEN,T.Y.(U.K.)
A dimensionless unitgaph for Hong Kong ........................ 501

HERAS,R.,LARA,A.(SPAIN)
Study of maximum floods in small basins of torrential type .......... 517
HERBST, P.H., VAN BIWON, S., OLIVIER, J.P.J.,HALL, J.M. (SOUTH
AFRICA)
Flood estimation by determination of regional parameten from limited
data .................................................... 541

JARASWATHANA,DAMRONG.,PINKAYAN,SUBIN.(THAILAND)
Practices of design flood frequency for small watersheds in Thailand ... 553

KINOSITA,TAKEO.,HASHIMOTO,TAKESHI.(JAPAN)
Design discharge derived from design rainfall .................. 551

LEESE,MORVEN N.(U.K.)
The use of censored data in estimating t-yearfloods ......... 563

POGGI PEREIRA,PAULO.(BRAZIL)
Assessment of design floods in Brazil ........................ 517

RENDON-HERRERO,OSWALD.(U.S.A.)
A method for the prediction of washload in certain small watersheds .... 589

RODIER,J.A.(FRANCE)
Méthodes utilisées pour l'évaluation des débita de m e des petits c o m
d'eau en régions tropicales .................................... 603

SOKOLOV,A.A. (U.S.S.R.)
Methods for the estimation of maximum dischargea of snow melt and
rainfall water with inadequate observational data .................. 615

VLADIMIROV,A.M.,CHEBOTAREV,
A.I.(U.S.S.R.)
Computation of probabiustic valuea of low flow for ungauged riven . 625

WON,TAE SANG.(U.S.A.)
A study on maximum flood discharge formulas .................... 635

TOPIC III - RELATION BETWEEN PROJECT ECONOMICS AND HYDROLO-

GICAL DATA
POINT 111 - RELATION ENTRE LES DONNEES ECONOMIQUES DU PRO-
JETET LES DONNEES HYDROLOGIQUES

BURAS,NATHAN. (ISRAEL)
The cost-effectivenessof water resources systems considering inadequate
hydrologiddata ........................................... 649

FILOTTI,A.,FRANK,G.,PARVULESCU,C.(ROMANIA)
Optimization of water resources development projects in case of inade-
quate hydrologic data ....................................
661

POBEDIMSKY,A. (ECE)
Relation between project economics and hydrologicai data ........... 683
INTRODUCTION

The Symposium on the Development of Water Resources Projects with

Inadequate Data was held in Madrid from 4 to 8 June 1973 for the purpose
of focusing on the methodology for hydrologic studies for water resources
projects with inadequate data and on current practices for the assessment
of design parameters.
The Symposium was opened at the Palacio de Exposiciones on the
morning of 4 June by Miniester of Public Workes of Spain Addresses were
then given by Dr.Dumitrescu on behalf of the Director General of Unesco,
Professor Nevmec on behalf of the Secretary-Generalof WMO,Dr.Rodier
as President of IAHS and by Dr. Briones, on behalf of the Spanish Na-
tional Committee for the IHD.
The Symposium was attended by 480 participants from 77 countries.
The technical programme, detalled in the Table of Contents, included
consideration of 3 major areas:

1. Methodology for hydrological studies with inadequate data,

2. Current practices in different countries,
3. Relation between project economics and hydrological data.

Each area was further sub-divided into topics for each of which the
individually contributed papers were abstracted into a general report,orally
presented by an invited expert, and followed by discussion.
Since the individual papers were not presented at the Symposium orally
by the authors, thery are reproduced here in the orden in which
they were reported in each general report under each topic.
Contents
Table des matières
Volume II
Foreword/Avant-propos................................

TOPIC II.1A - METHODS FOR STUDIES IN DATA-SCARCEAREAS AND

THE INFLUENCE ON DESIGN OF THE INADEQUACY OF
DATA FOR PURPOSES OF PLANIFICATION OF WATER
RESOURCES (EXCLUDINGFLOOD AND LOW FLOWS).
METHODOLOGY FOR ASSESSING HYDROLOGICAL CHA-
RACTERISTICS IN DATA-SCARCEAREAS.
POINT II.1A - METHODES D’ETUDESUTILISEES DANS LES REGIONS OU
LES DONNEES SONT INSUFFISANTES ET INFLUENCE SUR
LE CALCUL DU PROJET DU MANQUE DE DONNEES POUR
L’ELABORATION DES PROJETS DE L’UTILISATION DES
RESSOURCES EN EAU (A L’EXCLUSION DES CRUES ET
DES DEBITS DE BASSES EAUX).
METHODOLOGIE POUR L’EVALUATION DES CARACTE-
RISTIQUES HYDROLOGIQUES DANS LES REGIONS OU
LES DONNEES SONT RARES.

BASSO,EDUARDO. (UNDP/WMO)GENERAL REPORT

ABIODUM,ADIGUN ADE. (NIGERIA)

Water resources projects in Nigeria and the hydrological data employed in
their planning and development ................................

BASSO, E., ARRIAGADA, A., NEIRA, H., PEREZ DELGADO, M. (COSTA

RICA)
An example of regional co-operationfor improving the hydrological and
meteorological information ...................................

CUBAS GRANADO,FRANCISCO.(SPAIN)
Existing methodology for estimating free water surface evaporation ....

CUSTODIO,EMILIO.(SPAIN)
Geohydrologicalstudies in small areas without systematic data ........

DALINSKY,JOSEPH S. (ISRAEL)
Methods of analysing deficient discharge data in arid and semi-aridzones
for the design of surface water utilization .......................
II

D’OLIVEIRA,EMILIO EUGENIO.MIMOSO,JOAO JOSE.(PORTUGAL)

Mapai river hydrological study (Limpopo’sriver) ...................

D’OLIVEIRA,EMILIO EUGENIO.MIMOSO,JOAO JOSE.(PORTUGAL)

Application of Coutagne’sand Turc’s formulas to southern Mozambique
rivers ....................................................

HERAS,R.(SPAIN)
Report hydrological programs of the Center for Hydrographic Studies for
the investigation of hydraulic resources with insufficient data .........

KARAUSHEV,A.V., BOGOLIUBOVA,I.V.(U.S.S.R.)
Computation of reservoirs sedimentation .......................

KLIGUE,R.K.,MECHDI EL SACHOB (U.S.S.R.)

Calculation of runoff in Iraq ..................................

KUZMIN,P.P.,VERSHININ,A.P. (U.S.S.R.)
Determination of evaporation in case of the absence or inadequacy of
data .....................................................

PENTA,A.,ROSSI,F.(ITALY)
Objective criteria to declare a series of data sufficient for technical pur-
poses ....................................................

QUINTELA GOIS,CARLOS.(PORTUGAL)
Objective criteria used in hydrology with inadequate data ............

SMITH,ROBERT L.(U.S.A.)
Utilizing climatic data to appraise potential water yields .............

STANESCU,SILVIU.(COLOMBIA)
Determination of hydrological characteristics in points without direct
hydrometric data ...........................................

TEMEZ,J.R.(SPAIN)
New models of frequency law of runoff starting from precipitations ....

TRENDEL, R., DER MEGREDITCHIAN, G., RULLIERE, MARIE CLAIRE.

(FRANCE)
Traitement opérationnel des données pluviornetriques entachées d’erreurs
ou insuffisantes ............................................
TOPIC II.1B - METHODS FOR STUDIES IN DATA SCARCE AREAS AND
THE INFLUENCE ON DESIGN OF THE INADEQUACY OF
DATA FOR PURPOSES OF PLANIFICATION OF WATER
RESOURCES (EXCLUDING FLOOD AND LOW FLOWS).
INFLUENCE OF INADEQUACY OF HYDROLOGICAL DATA
ON PROJECT DESIGN AND FORMULATION.
POINT II.1B - METHODES D’ETUDESUTILISEES DANS LES REGIONS OU
LES DONNEES SONT INSUFFISANTESET INFLUENCE SUR
LE CALCUL DU PROJET DU MANQUE DE DONNEESPOUR
L’ELABORATIONDES PROJETS DE L’UTILISATION DES
RESSOURCES EN EAU (A L’EXCLUSION DES CRUES ET
DES DEBITS DE BASSES EAUX).INFLUENCE DU MANQUE
DE DONNEES HYDROLOGIQUES SUR LE CALCUL DU
PROJETET SA FORMULATION.

BEARD,L.R.(U.S.A.)
GENERAL REPORT
BANERJI,S.,LAL,V.B.(INDIA)
Design of water resources projects with inadequate data in India.General
&ParticularCase Studies ...................................

JAMES,IVAN C.(U.S.A.)
Data requirements for the optimization of reservoir design and operating
rule determination ..........................................

REID,GEORGE W.(U.S.A.)
The design of water quality management projects with inadequate data .

SABHERWAL,R.K.(INDIA)
Designing projects for the development of ground water resources in the
alluvial plains of northern India on the basis of inadequate data .......

SEXTON,J.R.,
JAMIESON,D.G.(U.K.)
Improved techniques for water resource systemsdesign ..............

WEBER,J., KISIEL,CHESTER C.,DUCKSTEIN,LUCIEN (U.S.A.)

Maximum information obtainable from inadequate design data: from
multivariate to Bayesian methods ..............................

TOPIC 11.2- CURRENT PRACTICES FOR ASSESSING DESIGN FLOODS

AND DESIGN LOW FLOWS,INCLUDING THE USE OF
SYNTHETIC UNIT HYDROGRAPH,WITH PARTICULAR
EMPHASIS ON MAXIMALISATION AND MINIMALISATION.
IV

POINT 11.2- PRATIQUES COUFUNI"'I'S POUR L'EVALUATION DES

CRUES ET DES DEBITS D'ETIAGES PRIS EN COMPTE DANS
LE PROJET,COMPRENANT L'EMPLOI D'HYDROGRAMMES
UNITAIRES DE SYNTHESE,AVEC ETUDE PARTICULIERE
DE LA MAXIMALISATION ET DE LA MINIMALISATION.

ROCHE,MARCEL.(FRANCE)GENERAL REPORT

BATLLE GIRONA,MODESTO.(SPAIN)
Estimation of floods by means of their silt loads ..............

BERAN,M.A.(U.K.)
Estimation of design floods and the problem of equating the probability
of rainfall and runoff ........................................

DAVIS,DONALD R.,DUCKSTEIN,L.,KISIEL,CHESTER C.,FOGEL,MAR-

TIN M.(U.S.A.)
A decision-theoretic approach to uncertainty in the return period of
maximum flow volumes using rainfall data .......................

HALL,M.J.(U.K.)
Synthetic unit hydrograph technique for the design of flood alleviation
works in urban areas ........................................

HELLIWELL,P.R.,CHEN, T.Y.(U.K.)
A dimensionlessunitgraph for Hong Kong ........................

HERAS,R.,LARA,A.(SPAIN)
Study of maximum floodsin small basins of torrential type ..........

HERBST,P.H.,VAN BILJON,S., OLIVIER,J.P.J.,HALL, J.M.(SOUTH

AFRICA)
Flood estimation by determination of regional parameters from limited
data .....................................................

JARASWATHANA,DAMRONG.,PINKAYAN,SUBIN.(THAILAND)
Practicesof design flood frequency for small watersheds in Thailand ...

KINOSITA,TAKEO.,HASHIMOTO,TAKESHI.(JAPAN)
Design discharge derived from design rainfall ......................

LEESE,MORVEN N.(U.K.)
The use of censored data in estimating t-yearfloods ................
POGGI PEREIRA,PAULO.(BRAZIL)
Assessment of design floods in Brazil .............................

RENDON-HERRERO,OSWALD.(U.S.A.)
A method for the prediction of washload in certain small watersheds ...

RODIER,J.A.(FRANCE)
Méthodes utilisées pour l’évaluation des débits de crue des petits cours
d’eau eri régions tropicales ....................................

SOKOLOV,A.A.(U.S.S.R.)
Methods for the estimation of maximum discharges of snow melt and
rainfall water with inadequate observational data ..................

VLADIMIROV,A.M.,CHEBOTAREV,A.I.(U.S.S.R.)
Computation of probabilistic values of low flow for ungauged rivers ....

WON,TAE SANG.(U.S.A.)
A study on maximum flood discharge formulas ....................

TOPIC III - RELATION BETWEEN PROJECT ECONOMICS AND HYDROLO-

GICAL DATA
POINT III - RELATION ENTRE LES DONNEES ECONOMIQUES DU PRO-
JET ET LES DONNEES HYDROLOGIQUES

BURAS,NATHAN. (ISRAEL)
T h e cost-effectiveness of water resources systems considering inadequate
hydrological data ...........................................

FILOTTI,A.,FRANK,G.,PARVULESCU,C.(ROMANIA)
Optimization of water resources development projects in case of inade-
quate hydrologic data ....................................

POBEDIMSKY,A. (ECE)
Relation between project economics and hydrological data ...........
 Foreword

While the need for hydrological and meteorological data of many types
for the design of water resources projects is obvious, it is often found,
especially in many developing countries,that such data are either lacking
or inadequate.
Recognizing the existence of this problem, the Co-ordinating Counci.1 of
the IHD appointed a group of experts (third session, Paris, June 1967) to
study the problem of design of water resources projects with inadequate
data.
Similarly,the Commission for Hydrology of WMO (third session,Geneva,
September 1968) established a Working Group on Hydrological Design
Data for Water Resources Projects to prepare guidance material on this
subject for the WMO Guide to Hydrological Practices and to maintain
liaison with the IHD group of experts appointed by the Co-ordinating
Council.
As a means of taking stock of the work carried out by the hydrological
community in coping with project design with scarce data, Unesco and
WMO jointly convened a symposium on this subject. The Symposium was
organized with the co-operation of the IAHS and the Spanish National
Committee for the IHD and was held in Madrid from 4 to 8 June 1973 at
the invitation of the Government of Spain.
The Madrid Symposium concentrated on the methodology of hydro-
logical studies for water resources projects with inadequate data and on
current practices for the assessment of design parameters.
The Minister of Public Works of Spain opened the Symposium at the
Palacio de Exposiciones on the morning of 4 June. Addresses were given
by Dr.Dumitrescu on behalf of the Director-Generalof Unesco, Professor
Nemec on benalf of the Secretary-Generalof WMO,Dr.Rodier as President
of IAHS and by Dr.Briones, on behalf of the Spanish National Committee
for the IHD.
The Symposium was atteneded by 480 participants from 77 countries.
The technical programme, detailed in the Table of Contents, included
consideration of 3 major areas:

1. Methodology for hydrological studies with inadequate data;

2. Current practices in different countries;
3. Relation between project economics and hydrological data.

Each area was further sub-divided into topics for each of which the
indivi,dually contributed papers were abstracted into a general report,
orally presented by an invited expert, and followed by discussion.
 This volume of proceedings was compiled by the Spanish National Com-
mittee for the IHD; it includes all the general reports and individual
papers presented at the Symposium, as well as the discussions. It is issued
as a joint Unesco/WMO/IAHS pub,licationin the spirit in which the three
Organizations have collaborated during the IHD.
Since the individual authors did not present their papers orally at the
Symposium, the papers are reproduced here in the order in which they
are discussed in the general report for each topic.
Unesco, WMO and IAHS wish to record their thanks to the Spanish
National Committee for the IHD for the many contributions of its members
towards the organization of the Symposium, and for the Committee’s as-
sistance in the publication of these proceedings.
 AVANT-PROPOS

I1 est évident que,pour élaborer des projets d’utilisation des ressources

en eau il est nécessaire de disposer de données hydrologiques et météoro-
logiques de types très divers; or il apparaît que ces données sont souvent
inexistantes ou insuffisantes, notamment dans beaucoup de pays en voie
de développement.
Conscient de ce problème, le Conseil de coordination de la DHI a créé,
lors de sa troisième session (Paris,juin 1967) un groupe d’experts chargé
d’étudier les moyens d’elaborer des projets d’utilisation des ressources
en eau sans disposer de données suffisantes.
D e son côté, la Commission d’hydrologie de l’OMM a constitué à sa
troisième session (Genève, septembre 1968) un groupe de travail sur les
données hydrologiques nécessaires à l’élaboration des projets d’arnénage-
ment des ressources hydrauliques; ce groupe de travail a été chargé de
formuler des recommandations destinées à figurer dans le Guide OMM des
pratiques hydrologiques, et d’assurer la liaison avec le groupe d’experts
de la DHI créé par le Conseil de coordination.
Afin de faire le point des travaux accomplis par la communité hydro-
logique en ce qui concerne l’élaboration de projets pour lesquels on ne
dispose pas de données suffisantes,l’Unesco et l’OMMont décidé de réunir
conjointement un colloque consacré à cette question. Ce colloque,organisé
avec la collaboration de I’AISH et du Comité national espagnol pour la
DHI, s’est tenu à Madrid en juin 1973, à l’invitation du gouvernement es-
pagnol.
Le colloque de Madrid a traité en particulier de la méthodologie des
études hydrologiques sans données suffisantes et des pratiques courantes
utilisées pour l’évaluation des paramètres de calcul.
Le colloque a été ouvert par le ministre espagnol des travaux publics,
le matin du 4 juin,dans le cadre du Palais des expositions. Des allocutions
furent prononcées par M. Dumitriscu, au n o m du Directeur général de
l’Unesco,par le professeur Nemec, au n o m du Secrétaire général de l’OMM,
par M.Rodier, président de l’AISH,et par M. Briones, au n o m du Comité
national espagnol pour la DHI.
480 participants, venant de 77 pays, participèrent au colloque.
Le programme technique, dont le contenu détaillé figure dans la table
des matières,portait sur trois domaines principaux:

1. Méthodologie des études hydrologiques sans données suffisantes;

2. Les pratiques courantes utilisées dans différents pays;
3. Relation entre les données économiques du projet et les données
hydrologiques.
 Chacun de ces domaines était subdivisé en thèmes, et sur chaque thème
un rapport général synthétisant les communications individuellles était pré-
senté par un expert, puis suivi d’une discussion.
Les Actes du colloque, établis par le Comité national espagnol pour
la DHI, comprennent l’ensemble des communications individuelles et des
rapports généraux, ainsi que le compte rendu des débats auxquels ils ont
donné lieu. Ils constituent une publication conjointe de l’Unesco, de l’OMM
et de I’AISH, reflétant l’esprit dans lequel les trois organisations ont col-
laboré pendant la DHI.
C o m m e les communications individuelles n’ont pas été présentées ora-
lement par leurs auteurs, elles sont reproduites dans l’ordre où elles sont
apparues dans le rapport les concernant.
Unesco, l’OMM et I’AISH tiennent à remercier le Comité national es-
pagnol pour la DHI du concours qu’il a apporté à l’organisation du colloque
et à la publication de ses Actes.
 METHODOLOGY FOR ASSESSING HYDROLOGICAL CHARACTERISTICS
IN DATA SCARCE AREAS

General Report
bY
Eduardo Basso*

INTRODUCTION
Three of the eight papers reviewed in this Report describe in general the f o r m
of assessing hydrological characteristics in data-scarce areas (Nigeria, A n g o -
la and the Central A m e r i c a n Isthmus). T h e other five papers deal with the ap-
Plication of certain particular methods, covering estimation of runoff, evapo-
ration, sedimentation and other hydrological parameters. Therefore, the re-
vision will be m a d e in this order.
T h e procedure to be followed in this summarizing report consists of present-
ing s u m m a r i e s of the papers followed by a discussion of the m a i n subjects and
by s o m e general c o m m e n t s on the whole subject.

REVIEW OF THE PAPERS

Okavango Basin in Angola. - T h e paper by Mr. Quintelag** presents the
studies m a d e in Angola and in particular those of the Okavango Basin, which
is one of the big international rivers of the South of Angola. T h e basin is shown
in Figure 1 of the paper. Its drainage area is about 150 O00 K m 2 and m o s t of
the rainfall occurs f r o m October to April. Altitudes range f r o m 1000 to 1800
meters. F o r the rainfall studies, 28 stations could provide 20 years of records
after completing s o m e shortages by correlation. With these annual values, a n
isohyetical m a p w a s drawn taking into account altitudes and s o m e climatical
factors. F r o m there, the m e a n annual rainfall w a s computed and analysed by
applying the Foster-Hazen method. The result is shown in Figure 2 of the
paper, f r o m which a m e a n annual precipitation of 9 5 0 mm w a s debermined.
19 flow measuring stations operate in the basin, but only records for 7 years
w e r e available. A s the m e a n rainfall of these seven years is near theaverage
the author concludes that the m e a n annual flow can be estimated by averaging
the flows of those seven years for every station. O n e station operatdby the
South African Services had longer records (25 years) and for it the Foster-
H a z e n method w a s used (Figure 3 of the paper). Finally, Figure 4 s h o w s a
curve indicating the variation of the specific annual flow within the drainage
area.

* Project Manager, Central A m e r i c a n Hydrometeorological Project, (UNDP/

WMO). Managua, Nicaragua.
** See list of references at the end of this Report,
2

Niveria's Case. Abiodun's p a p e r g deals first with Nigeria's water policy and
with the institutional arrangements in relation 90 water resources studies.
It later presents s o m e examples of utilization of hydrological data in existing
projects.
T h e Kainji multipurpose s c h e m e is located o n River Niger (Figure 1 of the
paper). Although construction was started in 1964, n o water levels w e r e ob-
served prior to 1959. T h e precipitation network w a s also insufficient until in
1953 w h e n n e w stations w e r e installed allowing a seven year record (1953-59)
f r o m which the rainfall over the catchment area w a s calculated for this period.
A relatively long record at Jebba, upstream of the d a m , could not be used
because of lack of adequate datum information. A correlation between monthly
rainfall and runnoff at Jebba w a s obtained using the newly observated discharges
at Jebba for seven years and w e -ed for ccmp&ing the discharge f r o m the basin
between N i a m e y (upstream of the d a m site at Kianzi) and Jebba Cbserved and
computed flows for two years are shown in Figure 2 of the paper.

T h e &ect of lage up to one m o n t h were considered in the correlation (equation

1 of the paper). Finally, the discharge at the damsite w a s obtained substracting
the eetimated runoff on two areas, estimating runoff coefficients of O. 1 and
O. 2 (equation 2 of the paper).
T h e paper refers t h e m to the problems produced by the lack of hydrogeologi-
cal data in Midwestern Nigeria.

in Western Nigeria m a n y long and reliable evaporation and rainfall data are
available, but river discharges are very scarce. According to the author, the
standard practice is to base the water s c h e m e design in a conservative f o r m
using a monthly evaporation of 127 mm and computing runoff with the formula

in which Q, is the catchment annual runoff, A the basin drainage area, Rpsn
the basin rainfall value corresponding to correspondjng the probability of -Ur-
taace.in 5 û y ~ a r eCo;'Coefficient of runoff for the basin, estimated at 4%. Abig
d u n indicates that variations of this formula are widely used in western Nige-
ria. and the uee of a f o r m of it was not used for computing the flood for the
spillway for Asejire Project because of the advice of a foreign consultant. Ins
tead, a runoff of 490 l/sec w a s used. basedan similar occurrences in other
West African dtreams.

T h e paper refers also briefly to the L a k e Chad basin studies which count with
cooperation f r o m the United States Geological Survey, FAO and UNESCO. In
this case FAO'e efforts have been directed to the harmonization and evalua-
tion of the data. infra-red aerial photography has been used in connection with
theee taska.
T h e paper concludes with a n appraisal of the studies used and with a brief des-
cription of the future activities in the field of water resources investigations in
Nigeria. Here, the use of n e w techniques such as rsmote sensing is r e c o m -
mended.

T h e paper by Basso. Arriagada, Neira and P é r e z13.-

T h e Central A m e r i c a n Hydrometeorological Proiec
describes the activities
of the Central A m e r i c a n Hydrometeorological Project, a co-operative effort
 3

between the countries of the Central A m e r i c a n Istbmis and the United Nations
Development P r o g r a m m e acting as executive agency the World Meteorological
Organization. T h e m a i n objectives of the Project are: (i) Installation of a basic
network of meteorological and hydrological station; (ii) Collection, processing
and publication of the data; (iii) training of personnel by m e a n s of courses,
fellowships or through technical publication and manuals; and (iv) T h e institu-
tional strengthening of the meteorological and hydrological services of the
area.
At the beginning of the Project (1966) the conditions in the area varied widely
f r o m country to country. In the average the few river discharge measuring
stations had short and sometimes unreliable data, the meteorological network
w a s poorly distributed and bore no connection with the hydrological network, a
defect that has been also reported in other of the papers under review; ra-
diation, evaporation and rainfall intensity information w a s completely insuf-
ficient, and --except for one or two countries-- no sediment or water quality
measurements were m a d e at all. A few capable technicians were available,
but extensive training w a s a pressing need.

T h e paper describes in s o m e detail the steps taken by the Project, as result

of which the present situation is quite satisfactory for developing conditions.

Of particular interest for the subject of this meeting is the description of s o m e

methods proposed by the project for assessing hydrological characteristics
with insufficient data.

T h e use of the sediment rating curve, Figure 7 of the paper. has been used for
computing sediment transportation. T h e remarks on the variation of the coe&
ficient n of the equation C S A Qn with annual precipitation (G:sediment dis-
charge, Q: Discharge; A, n coefficients) are of interest in aMlyZing scarce se-
diment information. Figure 8 of the paper shows the results of s o m e m e a s u r e -
ments m a d e by the Project, indicating the effect of rainfall and vegetation cover
in the sediment yield. T h e effect of the destruction of the vegetable cover by a
volcanic eruption should be noticed as a quite particular case.

Flood and rainfall envelopes (Figure 9 and 10 of the paper) have been used as
a first estimate of m a x i m u m discharges and precipitation studies. Studies of
regionalized flood frequency analysis are n o w under way.

Other achievements of the Project include etudies for determining evapotrans

piration and water requirements for irrigation, studies o n runoff forecasting
-
groundwater studies using a regional analog computer, etc.

T h e report refers also to the problem of network implementation in areas with

access problems, and the use of prefabricated elements Éhould be noted
(Figures 3 and 4 of the Report). Figures 5 and 6 shows the change in areal
coverage as result of the action of the project. T h e successful use of m o d e r n
mechanical methods for processing meteorological and hydrological information
should encourage other developing countries in the use of these methods.

T h e report concludes with a r e m a r k o n the importance of adequate institu-

tional support for these activities, which<imitia1Ly requi res the creation of
concern of the Governments on the importance of meteorology and hydrology.
 4

Est i m a t i ng Water Yiel ds

S m i t h ' s 9 paper presents an interesting example of estimating water yields

using only precipitation and temperature measurements.
T h e basic water balance equation applied to a catchment area m a y by expressed
as:
P R + E + AS
(1)
P precipitation; R total basin outflow, E evapotranspiration and AS
change in storage. F o r a long period AS b e c o m e s negligible, and making
s o m e transformations in equation (1) it is possible to rewrite it as:

Thus, in the long term, the runoff oefficient C is governed by climatic consider-
ations. ln 1967 Guisti and López$ proposed that the m e a n stream discharge
could be determined as a function of the m e a n annual precipitation and the basin
climatic index, BCI, defined as:

where P : average monthly precipitation in centimeters and T: average monthly

temperature in degrees centigrade. A relation between C and BCI based in 250
catchments in the United States and Puerto Rico is s h o w n in Figure 1 of the paper.

T h e use of regional relations between BCI and P as those s h o w n in Figure 2 of

the paper allows to derive C only f r o m precipitation data. T h e basic C versus
BCI relationship was tested with satisfactory results as those shown in Figure 3
of the paper.

T h e basic relationships can also be used to appraise the effect of changes of the
precipitation, If subscript 1 represents natural conditions and 2 represented
augmented conditions (in the case of a n increase in rainfall) then the gain in ru-
noff can be written as:

Where, PM L P2/Pi

Jntable 1 the author c o m p a r e s the results of using this method with the results
of ueing hydrologic simulation as reported by several investigators with good
agreement..
Using a reasonable amount of judgment it is possible to determine flow charac-
teristics other than the mean. Figure 4 shows a comparison of calculated and
observed annual runoff distributions for the Marias de Cygnes River, Kansas,
USA. However, the limitation of this method, as clearly indicated in the text
of the paper, should be considered before using it. This also applies to the
estimation of monthly yields allocating t h e m in proportion to their contribution
to the BCI (a t w o m o n t h running average should be used due to tag problems).
 5

T h e use of the basic relation can also be extended with the help of certain flow
and miscellaneous field measurements.
T h e paper closes showing the application of the method for appraising the po-
tential yield characteristics of coastal aquifers in southern Puerto Rico and
presenting one example of the adjustments required w h e n the natural conditions
have been changed by man's activities.
Application of Coutagne's and Turc's F o r m u l a s

T h e paper by D'Oliveira and M i p ~ s o 6 Japplies Coutagne's and T u r c formulas

for the southern M o z a m b i q u e rivers.

Coutagne's general rule states:

D-H-KH2
= H - E;H: M e a n rainfall height; K! Coutagne's constant
D: Runoff deficit
Also C = KH where C: Runoff Coefficient ; &
H
T h e m o s t probable value of K is obtained by equating to zero the first derivative
-
of E(C KH)2, which results in:

Turc's general rule can be expressed as:

P. H

/.z-g-
Where =
L: Turc's constant A t 25T t O. 05T3,
P: Evaporation plus percolation looses (runoff deficit), H: Precipitation,
A: Constant; T: M e a n temperature (In degrees centigrade)
T u r c applied his rule for 254 basins, using A 300, finding that in 53% of the
cases the difference between the real and computed D was lees than 40 mm;in
43% of the cases this difference w a s less than O. 1 of measured D and in 65% the
difference w a s less than O. 2 measured D.
T h e application of both formulas to seven basins w a s divided nto two groups; the
L i m p o p o River group (Rainfall 450-650 nun;temperatures 18' C-20' C) and the
Incomati, Sabie, Umbeluzi and Usoto G r o u p (Rainfall 800 mm;temperatures
higher than 20').
Detailed results are presented, which can be sumarized as follows:
 6

C ontanne's T u r c relation
K =A 300
P e r cent of D -Dcalc
greater than O. 1 Dcalc
L i m p o p o area
Elephants River o. O00055
Beit Bridge O. 000031
Trigo de Morais O. 000047
All group o. 000050 64%
Incomati, Sabie. Umbeluzi and Usoto area
Incomati River O. 0001 50
Sabie River O. 000131
Umbeluzi River O. 000145
Usuto River O. 000162
All group O. 000140 16%
-the use of Turc's relation with A 300 produces poor results, the authors
present a nomograph (Figure 2 of the paper) to compute the value of A. Using
these n e w values of the constants the difference between calculated and measured
D is reduced to acceptables levels.
Estimati on of Lvapotranspiration

T h e paper by K u z m i n and Vershininu deals with the determination of evapora-

tion in case of the absence or inadequacy of data.

Since methods for direct evaporation m e a s u r e m e n t s are still being developed,

computations are the m a i n source of information. These can be divided into
three groups: (i) methods based in the physical analysis of the process, (ii)
methods combining the physical analysis with semi-empirical constants deter -
mined f r o m actual evaporation in representative regions and (iii) purely statisti-
cal methods.

T h e first group includes methods using heat balance equation. water balance
equation and turbulent diffusion. I n the USSR equation (1) which has been
deduced f r o m the simplified equation of the heat balance of the land surface with
the account of B o w e n ratio is widely applied.

where: E: Evapotranspiration, R is the measured value of the radiation balance

of the surface, B is the heat income into the soil, L is the latent heat of evapor-
ation, C p is the heat capacity under constant pressure, H is the atmospheric
pressure, t and e are respectively the differences in temperature and water
pressure measured at two levels above the ground.

Equation (1) should rather belong to the second group than to the first one, since
it does not represent all physical factors that affect the phenomena.
Full water balance is not applied in the practice but in the case of deep water
table. in this case, the fobwing equation is used in the USSR for estimating-
evapotranspiration f r o m non-irrigated fields;
 7

E = X +(W1 ..W 2 ) (2)

X: Precipitation; W1 and W 2 are the moieture storage in soil at the beginning
and at the end of the design period. S o m e conditions for using this relation
are indicated by the author. Another partial solution of water balance equa-
-
tion is the estimation of m e a n annual s u m s of evapotranspiration as the dif
ference between precipitation and runoff. After indicating the possibilities of
methods based on turbulent diffusion problems in deriving a universal equation
are also stated. Therefore, the convenience of equations using non-specialized
observations is evident. O n e of these for regions
- of natural moistening is due
to Budyko:
-Ro XL
;h Ro ( c m year-1 ) (3)
4
equation (3) includes only one observational parameter, X: long t e r m average
precipitation (cm/year) R o is the average annual radiation balan e of the undey
lying surface which can be obtained f r o m the m a p of Referenced. L is the
latent heat of evaporation, A method for distributing the m e a n annual s u m s
estimated f r o m equation (3) is explained by the ,authors.

Monthly evapotranspiration f r o m irrigated fields are estimated with the help

of simplified heat balance equations. T h e standard error is about 15% w h e n
special observations are available or about 30% with standard observations.
T h e use of em+ical relations similar to that of Blaney and Criddle can be
used only if the empirical coefficients are tested and corrected for each point
of their application.

T h e m o s t simple equations allowing the estimation of evaporation f r o m water,

s n o w and ice surfaces by m e a n s of standard observational data, are the
following binomial and m o n o m i a l equations:

E (a t ab&) (es - e2)

and E = A U 2 (es - e2)
being: E; evaporation -/day; U, wind speed at the height z above the sur-
face in misec; es and e2 are the m a x i m u m water vapor pressure estimated
f r o m surface temperature and water pressure at 2 meters in mb; A, a and b
are experimental constants. F o r estimating evaporation f r o m s n o w the values
of a -0.18 ab=O. 098 z = l O m should be used in equation (8). F o r lake evaporation
a=O. 14 b=O. 72 and 9 m should be used in equation (8). Other cases are also
discussed in the paper.

RESERVOIR SEDIMENTATION
T h e paper by Karaushev and Bogeliubevag presents a method for estimating
reservoir sedimentation based o n the equation of sediment balance a s applied to
the whole reservoir or its parts.

T h e inflow of sediments is computed by observational data or by indirect

methods. T h e outflow of sediments is computed based in hydraulic and sedi-
m e n t characteristics. T h e determination of sedimentation during one year is
reduced to estimating tha portion of the sediment inflow that is accumulated in
the reservoir.
8

Equation (1) of the paper shows the computation of sedimentation for any si ze
fraction in a design interval:
m -6
Paj = is1
-
Z Pi in j Qter j A tj 10 si terjils (1)

P aj is the amount of sediments of all size fractions trapped by the reservoir

during k tj; Pi in j is the inflow of sediment for each i-th fraction; Q ter j is
the m e a n water outflow (m3/s) and Si ter j is the m e a n turbidity (concentration)
for the time 4 tj and for the i-th fraction of size. Equation (2) to (9) are used
for computing si ter j and are based in hydrod namic considerations and the
reservoir characteristics such as length and depth. T h e amount of bed load in
the reservoir is computed by equation (10):

P a bed j = l o m 3 (R bed in j - R bed ter j) A tj (1 0)

P a bed j is the weight of bed load in the reservoir (Tons), R bed in j and R bed
ter ' indicate bed load discharge at the initial and terminal discharge sites
(Kgjsec), A tj is the time interval, Bed load, R bed, is computed with Shamovls
equation (equations 1 1 to 14 the text).

T h e annual accumulation of all sediment fractions for the first year of reser-
voir operation is obtained by adding the suspended and bed sediments as indicated
in equation (15) of the paper. T h e value Pai (tons) so obtained is transformed
into volumetric units Wai:

duced volume W -
Y s is the specific weight of the sediment (T/m3). After the first
W a is used for the computations of next year.

F o r the computation of the chronological variations of sedimentation the S h a m o v

method is recommended:

W h e r e W a t is the sediment volume in t years; W a l is the sedimentation volume

during the first year, computed as explained before, W a ext is the extreme
volume of sediments in the reservoir, approximately computed by:

W h e r e W is the initial volume of the reservoir, Ur is the area of river cross

section w h e n discharge is close to maximum and up is the m a x i m u m cross
section area of the upper pool near the dam.

Surface Water Utilization in Arid and S e m i Arid Zones. -

T h e paper by Dalinskyly
shows *e experience of Tahal-Water Planning for Israel Ltd. in various methods
of analyzing stream flows. F o r planning of utilization the following information
is required: (a) the average volume of annual flows (P.ave),. representing the
stream water resources potential; the average annual feasible utilizable flows
is a portion of this value; (b) the stream's flow regime including flood frequency;
(c) the stream variability within a season, a year, or f r o m one year to another.
 9

F o r determining the annual flood return periods the author proposes the use of
the well k n o w n T =
formula; for longer return periods the estimates of
m
order of magnitude of annual flows for longer return periods can be obtained
by extrapolation on probability paper.
T h e next section deals with the well k n o w n flow-durati on curves,

T h e concept of Ilhorizontal cut" of the stream hydrograph is useful in the case

of a diversion of a stream, as indicated in sketchs 2 and 3 of the paper. T h e
"horizontal cut1'can be expressed mathematically as:

Qd = Qi when mix
... (2)
Qd (ad)m a x when Qi 3 Qd) max

Where: Q: atreamflow discharge

Qd diverted discharge
(Qd) max: m a x i m u m diverted discharge

F o r a period of n years, a series of n annual diverted volumes can be obtained

and the average diverte&annual flow (va)can be calculated for each value of
=
(Qd) max. T h e funtion U d f (ad) m a x has the f o r m indicat& in sketch 4.
Three zones can be distinguished in this curve; in zone I U d is
'mmax
relatively large and almost constant; in zone II the derivative decrease quickly
a s (Qd) m a x increases; in zone III the derivative trends to eeru,w h e n (Qd)max+ Q

M o s t of diversions will be economically justified in_zone i, and unfeasible in

Z o n e IU. F o r m u l a (3) can be used for calculating U d f r o m the flow-duration
curve.
Adjustments for baseflows or m i n i m u m diverted discharges can be m a d e easily
changing the origin o f coordinates.

W h e n there are limitations to the diversion of baseflow discharges, a "double

cutttis required as indicated in sketch 5. This case will arise w h e n baseflow
is undesirable due to high salinity or other reasons for diversion. A m a x i m u m
desirabledischarge is determined generally by sedimentation conditions. T h e
value of U A , average diverted flow can be computed as:

where ÜBis established by m e a n s of a horizontal cut and E

by m e a n s of a
vertical cut. Funtions Ug and Uc can be easily calculatfi by computer, An
example is shown in Fig. 2, A p p A. Using equation (5) UG can be calculated
f r o m the flow-duration curve without use of a computer.

Another uae of the vertical cut is presented for planning of diversiomwith li-
mitations of maximum discharges due to sedimentation:

Qd
Qd
m Q
=O for
for
Q
Q Il
4 Qdmax
> Qdmax
T h e resulting discharge curve is combined with the sediment concentration
flow discharge curve s h o w n in Figure 4, App. A for computing the sediment
transport as detailed in App. B. of the paper.
10

Next section deals with the determination of annual storable flows as a .function
of reservoir capacity. Assuming that losses during the rainy season can be
neglected, the following equation applies:

-UR: is the n years' averageannual amount of water stored in the reservoir

(Net average capacity'RN)
UR)^ is the amount of water stored in the i th year;
UR)i Ui w h e n Ui 4 R N
Iud. i: (RN)i when Ui P (RN)i
Ui: is the annual streamflow
(Rn)i represents the net reservoir capacity in the ith year
Limitations of these relations arc indicated in the text (Neglecting losses, etc. )

Relations between K a n d the reservoir e f f i c i e n c y h s function of average

Uave
net reservoir capacity are shown schematically in sketch 6. (The meaning of
the three zones is the s a m e as in sketch 4). This analysis is important for
preliminary estimates and/or feasibility calculations. Recent investigations
reported in the paper prove that these relations can be approximately estimated
on a regional basis using as a parameter the dimensionless standard deviatio-

F o r m u l a e for computing RN and RN. based in the decrease of the capacity of

the reservoir due to sedimentation are given in other section of the paper. T h e
paper concludes recommending hydrological investigations to find, on a regional
basis,parameters allowing to represent the m a i n functions discus sed in the
article.

ASSESSING m R O L O C I C A L CHARACTERISTICS IN DATA -SCARCE AREAS

Estimating flow regime w h e n no data are available. -
T h e first problem with
which the hydrologist has to deal in data-scarce areas consists of obtaining
the hydrological characteristics of the region. First of all, the average dis-
charge has to be estimated. F o r this, a wide variation of methods can be used
depending in the availability of information

T h e worst case consists in a complete lack of information H e r e the estimates

should be based in observations in similar gauged zones. T h e estimations m a d e
for Asejire Project s e e m to belong to this case. However, even in this extreme
case, the scarce available information should not be neglected. Topography,
altitude, shape, orientation, geology and ve@aHecover of the basin are easily
obtained and should be always used.

T h e m o s t elementary equation is the surface relation:

A
Q = A-b Q b
where: Q :Flow at the site under study; Qb : F l o w at a base station; A: Surface
of the basin under study and Ab: Surface of the basin of the base station.

This equation, ii obviaidya wrypoor representation of the +noniena, and sbculd be used
 11

only for gross estimates.

W h e n precipitation data is available,thism d h o d can be improved by introducing the

precipitation data for the basin under study.(P)and for the basin of the -base
station (Pb). T h e relation in this case is:
A P
Q a--
A b p b Q b (b)
If the yield of the basin is defined as K e 2,
equation (b) becomes:
PA

Q - K b P A (4
A variation of this method, used sucessfully in Chile and Central America" J
consists 8f analyzing the variation of K with the basin conditions, topography,
elevation, vegetation, geology, orientation, etc.. .Figure 1 shows a n example
of this method.

Abiodun uses this method in his paper. N o explanations, however, are given
for the criteria in selecting K = O. 04 and for the use of a 1:50 years precipi-
tation. This s e e m s a n exaggerately pesimistic estimation and should result
in underestimation of the water resources. However, since as the paper explains
that efforts are been m a d e for enlarging the scope of the hydrological investigat-
ions in Nigeria, it is hoped that soon it will be possible to revise these c o m p u -
tations with m o r e accurata methods.

W h e n the Economic Comission for Latin A m e r i c a decided to m a k e a prelimi-

nary survey of the water resources in the Central A m e r i c a n Isthmus, the UNbP/
@Mo project prepared the m a p s of curves of runoff deficit s h o w n in Figure 2,
wnich allowed first estimates for ungauged areas. T h e trace of the curves
should take into account the already mentioned physical factors.

A further improvement consists of the introduction of climatic factors. such as

temperature. Examples of these methods are those of Khosla, Langbein,
Coutagne, T u r c and the one propeed by Smith in his paper. Application of
these methods with universal constants produce, sometimes, large errors, so
they should be limited to regional use, previously determining their constants
in gauged areas of similar characteristics.
in D'Olivieria's case, the use of Coutagne's rule with the original constants
would hata introduced very large errors in the estimates. T h e s a m e occures
when A = 300 Le used in Turc's formula.

As Smith shows in his pa er a good relation between precipitation and t e m p e -

rature (or Flimatic indexf can be found in a Fegionalized basis. T h e reporter
has added entra1 A m e r i c a n values to Smiths relations for Puerto Rico and
Kansas with good results (Figure 3). However, the Constants used in thesb
rnethds elaould be verified-ona regional basis.
A check has been m a d e to all these methods using t h e m to estimate the m e a n
annual discharge (in mm)of eleven CentralAmerican streams of quite different
conditions. T h e following methods have been used: Equations a. b and c,
Coutagne, T u r c and Smith, and the results are s u m m a r i z e d in table i.
12

Table L - Comparison of the use of several methods for estimating m e a n annual

runoff (mm)of Central A m e r i c a n streams.
Country Drainage Estimates of m e a n annual runoff Obserx
and Basin ed
Station sq Km. Eq.a Eq, b Eq,c Coutagne Turc Smith runoff

Guatemala
Candelaria 849. 5 470 480 550 1550 1830 1190 440

Honduras
R e . Pimienta 883.8 720 720 750 540 1250 320 740
El Salvador
Bande ras 432.8 89 0 840 800 7 50 1500 660 500
San M a r c o s 180. O 560 590 620 81 O 1500 1000 590

Nicaragua
Dar $0 91 5 160 160 200 50 670 50 110
Tamarindo 165 280 250 480 340 1040 230 500
Costa Rica
Cachi 904.1 2280 2000 2160 21 O0 2250 1700 2000
El H u m o 135 26 50 5700 6000 (6 500) (5240) (6100) 6270
Palmar 486 3 1590 2300 2350 2270 2420 1800 1970
Panamá
David 1392 1970 2150 2200 2880 2620 2150 2650
Majk 321 8 1370 1520 1550 880 1670 800 1560

Average
error ‘já 35 21 15 40 39 60
( ) Extrapolations.
Equation (c) gives the best results, followed by the simple areal relation corrected
for taking into account the change in precipitation (Equation b). However, these
are the results for a particular area, Central America, and there is no assurance
that similar results should apply to other regions of the world. T h e best advice
could be to try several of these methods and check as soon as possible the re-
sults with m e a s u r e m e n t s at the site under study.

-
Extending short or incomplete records, T h e m o s t c o m m o n l y used method for
extending short or incomplete records is to correlate the records of the station
with the records of a station with longer records. T h e correlation can be done
with m e a n annual, m e a n monthly, m e a n daily or instantaneous discharges: the
quality of the correlation decreasing in this order. F o r daily or instantaneous
dischargestlag effects have to be taken into account. in larger basins --as in
the case reported by Abiodun-- lag effects apply also to monthly discharges.

T h e quality of the correlation can b e determined easily by m e a n s of simple

statistical tests. This quality depende on the physical and meteorological
characteristics of the basins being compared. in general, correiations between
two stations nearly located over the s a m e river give good results. The fol1 w-
ing results should be expected w h e n the basins of Figure 4 are c o m p a r e d l d :
 13

Basins c o m p a r e d Quality of correlation

1 and 2 G o o d : Basins of similar form,

size and orientat i on.
2 and 3 Fair : T h e orientation of the
valley is different.
3 and 4 Poor: Different altitude and
orientation.
2 and 5 Fair : S a m e f o r m but different
a It it ude.
4 and 6 P o o r : Different characteristics
5 and 6 P o o r : Different orientation and
altitude.

W h e n no hydrometric information is available, correlation can be tried with

longer precipitation series.
A long time average can be obtained assuming a constant yield of the basin, or:

In this case, long (n year) precipitation records are available, Pt and Qt are
the rainfall and discharge averages over the t years for which discharge data
=
are available. This method, after checking that P n P w a s used by Quintela
in his paper. However, a n interesting verification in that case would had been
correlating the seven year's records with the South African station

Studies m a d e in Chile and in the Central A m e r i c a n Isthmus s h o w that the re-

sults of correlation studies are far m o r e reliable than the methods explained
in the preceeding section. However, extreme caution has to be exercised w h e n
records are too short, carefully avoiding to be too influenced by s o m e statisti-
cal indicators. In this case comparison with other methods is a n useful auxiliary
tool. Complete verification of the base information should be the starting point
of any extension of hydrological records.

Estimating evaporation and evapotranspiration. -

T h e estimation of evaporation
and evapotranspiration has several important implications in hydrological
studies, such as computations of reservoir evaporation, water balances and of
requirements for agriculture.

Direct m e a s u r e m e n t s are difficult; the U. S. Weather Bureau type A pan, the

m Q s t frequently used instrument in developing countri es, is not always correctly
read and the relation f r o m pan to lake evaporation remains in doubt . The
development of a simple formula for computing potential evaporation, is there-
fore of great importance.

K u z m i n and Vershinin give an excellent s u m m a r y of formulas used in the USSR..

F o r data-scarce areas, however, formulas based in the physical interpretation
of tFe fenomena are quite difficult to apply. Equations (2) and (3) are certainly
promising and it would be interesting to have m o r e details o n t h e m Binomial
formulas are widely used. Equation (9) lightly different coefficients has
been used in Chile and in Central A m e r iwit&
ca with unsatisfactory results.
14

Equation (9), as reported by K u e m i n and Vershinin, l

y
y been compared with
Blaney -Griddle, P e n m a n , Hargreaves -Christiansen and M e y e r formulas
for five locations in the Central A m e r i c a n I sthmus with the following results:
Table II

Evaporation computed with several formulas

Station Madden San José Chorrera Guija G ua t emala

Panamá Costa El El Guatemala Average
Formula Rica Salvador Salvador

USSR, Binomial
lake evaporation 79 8 488 1293 1133 765 89 5

Blaney-Criddle 2062 1749 2041 1910 1647 1882

Penman 1328 1077 1345 1530 1350 1326

Hargreaves-
Christiansen 1400 1 O00 1340 1280 1070 1218

Meyer 1394 1227 2147 1577 1628 1594

Potential
Evaporation
(Measured in
p a n x O. 77) 1020 1145 1760 1460 1050 1287

in average, the best agreement ie reached P e n m a n formula. However, as the

Hargreaves-Chrintiansen equation w a s a plied using only temperature, humidity
and precipitation (wind was estimated7 it provides a simple alternative, Blaney-
Criddle and M e y e r (a simple binomial formula) give excessive values. T h e USSR
binomial formula gives very low values, which is probably due to the obvious
differences in climate with respect to the conditions f r o m which the formula w a s
de rived.

Sediment studies. -T h r e e of the papers s h o w examples of sediment determinat-

ions, which quite frequently have to be m a d e with insufficient information,

T h e first problem refers to estimating sediment yields f r o m streams without

sediment measurements. Figure 8 of the paper on the Central A m e r i c a n H y d r o
meteorological Project shows the wide variation, within a regiun, of sediment
yields. Thus, determining sediment transportation loads without field m e a s u r
menta is quite unreliable. Hydrologic and hydraulic information f r o m the
-
gauging stations "somehow improves these estimates. However, very simple s q
diment m e a s u r e m e n t s allow a relatively acc rat estimate of suspended sediment
loads. A good correlation has been found18)1 between the concentration of a
s a m p l e taken with a bottle by U n d of unskilled obskrvers and the m e a n concen-
tration obtained with conventional sampling.

T h e sediment rating curves (two examples presented in the papers) allow to

compute the total suspended load in a quite simple form, T h e points of the curve
relating the solid and liquid discharges have a large dispersion (due to errors
 15

in measurements, differences in the raising and decreasing stages of a flood,

variations in the availability of sediment supply, etc. ), but its m e a n trend has
been found to be relatively stable, which allows estimates of suspended loads
with series of observations as short as one year.

Determination of bed load presents just the opposite problem. Direct m e a s u r e -

ments are difficult and provide in m o s t of the cases non-meaningful results.
U s e of well-known formulas is thus encouraged, in spite of the fact that they
give enormous differences. Therefore, the use of several methods is suggested
including, if possible, methods, such as modified Einstein, which use the avail-
able suspended sediment measurements. It is also quite useful to observe the
'!critical discharge", .i e. discharge at which the bed m o v e m e n t starts, which
in s o m e cases can be determined by detection. of stone noise by the stream
gaugers.
Karaushev and Bogeliuva's paper deals with the important problem of predict-
ing the chronology of the filling of a d a m , and esents a n e w interesting ap-
proach to the problem also studied by B r o w n l v . However, the m a i n difficulty
as seen by this reporter, is the estimation of "in situ" specific weight of the
settled suspended sediment. T h e problem here is that w h e n the sediment is
very fine its settlement is very slow and subject to relatively complicate laws.
For this the formula of Reference 14/ can be used:
yT : -k k( T
T-l Log T - 1 )
k is a constant depending on the size and mechanical distribution of the material;
YT the specific weight after T years and y1 the s ecific weight of the sediment
("in situ") after one year of settling. R e f e r e n c e l g g i v e s values of k and Y
but to the reporteis knowledge, no check of these values have been m a d e for1,
m o s t of the world. These "in situ" determinations are difficult, since it is
practically impossible to obtain indisturbed samples of submerged clay or lime.
T h e use of y R a y diffusion probes can be usefull, but they require careful
laboratory calibrations and expensive equipment.

Water Resources Studies. -

Dalinsky's paper present a n interesting and simple
method for preliminary studies of water resources projects, and should be con-
sidered a preliminary approach to those exposed in other sections of this Sym-
posium.

CONCLUSIONS
Assessing hydrological characteristics in data-scarce areas is indeed a difficult
p r o b l e m T h e difficulties in the studies increase inversely with the amount of
information. available, not because of the intrins ic mathematical and operational
problems, but because extremely good judgement is required. Unfortunately
good hydrological judgement depends o n the knowledge of the meteorological,
physical and hydrological characteristics of the region under study.

Several excellent examples have been s h o w n of what can be done with scarce
information, but the possibilities of big mistakes appeared also evident. These
can be avoided either with excellent judgement or with the help of a few, but
adequate data. T h e s e data do not need to be long t e r m series or sophisticated
measurements, thus can be collected at a relatively low cost. This cost re-
presents only a small fraction of the eventual overexpenditures or losses f r o m
poorly designed schemes.
16

T h e ideal, obviously, would be to undertake in each data-scarce area a c o m -

prehensive meteorological and hydrological survey, as the U DP/WMO projects
in several parts of the world. Evaluation of these projectslg allows to s h o w
several concrete examples where a small investment in these basic surveys has
resulted in economic benefits several times larger than the expenditures in m e -
teorology and hydrology.
REFERENCES
Quintela Gois, C. - S o m e Criteria Used in Hydrologic Studies with Ina-
dequate Data. S y m p o s i u m on the Design of Water Resources Projects
with Inadequate Data. Madrid 1973.

Abiodun, A. A. -Water Resources Projects in Nigeria and the Hydrologi-

cal Data Employed in their Planning and Development. S y m p o s i u m on the
Design of Water Resources Projects with Inadequate Data. Madrid 1973.

-
Basso, E., Arriagada, A., Neira H. and P é r e z Delgado, M. A n Example
of Co-operation for Improving the Hydrological and Meteorological Inform-
ation. S y m p o s i u m on the Design of Water Resources Projects with Ina-
dequate Data. Madrid 1973.
-
Smith, R. Utilizing Climatic Data to appraise Potential Water Yields.
Simposium on the Design of Water Resources Projects with Inadequate
Data. Madrid 1973.

Giusti, E. V. and López M. A. -

Climate and streamflow of Puerto Rico,
Caribbean Journal of Science, Vol. 7, pp 87-93, 1967.

D'Oliveira Martens, E. E. and M i m o s o Loureira, J. J. - Application of

Coutagne's and Turc formulas to the Southern M o z a m b i q u e rivers. Sym-
posium on the Design of Water Resources Projects with-Inadequate Data.
Madrid 1973.

-
Kuzmin, P.P. and Vershinin, A. P. Determination of Evaporation in
case of the Absence or Inadequacy of Data. S y m p o s i u m on the Design
of Water Resources Projects-with Inadequate Daia. Madrid 1973.

Materialy Mezhduvedomstvennogo Sovetchchania PO probleme Izuchenia

i Obosnovania Metodov Rasheta Isparenia s vodnoi Poverkhnosti i Suchi.
(Materials of Interagency Meetings on the P r o b l e m of Study and Substan-
tiation of Methods for the Computation of Evaporation f r o m Water and
L a n d Surfaces). Edited by CGI, Valdai 1966.

Karaushev, A. V. and Bogeliulova L V. - Computation of Reservoir Sedi-

mentation. S y m p o s i u m on the Design of Water Resources Projects with
Inadequate Data. Madrid 1973.

-
Dalinsky, J S. Methods of Analysing Defficient Discharge Data in Arid
and Semi-arid zones for the Design of Surface Water Utilization S y m p o -
s i u m on the Design of Water Resources Projects with Inadequate Data.
Madrid 1973.

-
Central A m e r i c a n Hydrometeorological Pro.iect. Manual de Instruccio-
nes: Estudios Hidrológicos (Manual of Instructions: Hydrological
- Studies)
Publicación No, 70, San José, Costa Rica 1972.
 17

Central A m e r i c a n Hydrometeorological Proiect. -

Medida de la Evapora-
ción (Measurement of Evaporation) Publicación No. 19, San José, Costa
Rica,. 19 68.

Brown, C. B. -
Discussion of "Sedimentation in Reservoirs" by B. J.
Witzig" transactions ASCE Vol. 109, 1944, pp 1080-1086.
Office of Indian Affairs, Bureau of Reclamation, Tennessee VaBey Authority
-
Corps of Engineers, Geological Survey, Department of Agriculture and
Iowa institute of Hydraulic Research. A Study of Methods Used in M e a s u r -
m e n t and Analysis of Sediment Loads in Streams. Report 9 "Density of
Sediments Deposited in Reservoirs", St. PBul District Sub-Office, Corps
of Engineers, Hydraulic Laboratory University of Iowa, Iowa City, Iowa
194.
Central A m e r i c a n Hydrometeorological Project. -Estimación Preliminar
del Balance de Aguas en el Istmo Centroamericano (Preliminary estimat-
ion of the water Balance in the Central A m e r i c a n Isthmus) Pubiicación
No. 18, San José, Costa Rica 1968.

-
World Meteorological Organization. Twenty Years of WMO Assistance.
W M O - N o . 338, Geneva, Switzerland 1972.
-
Central A m e r i c a n Hydrometeorological Project. Deficiendas de Agua
en Centro A m é r i c a B P a n a m á (Water Defficiencies in Central A m e r i c a
and P a n a d ) Repodprepared by G. Hargreaves as a consultant to the
Central A m e r i c a n Hydrometeorological Project. Publication No. 88,
Managua, Nicaragua, 1973.

Central A m e r i c a n Hydrometeorological Project. -

E m p l e o de la Muestra
Puntual para la Determinación del Sedimento en Suspensión (Use of the
Puricbial Sample for the determination of Suspended Sediment) Publica -
ciÓn No. 1, San José, Costa Rica, 1967.

-
Central A m e r i c a n Hydrometeorological Project. M a n u e l de Instruccio-
nes: Hidrometría (Manual of Instructions: Hydrometry) Publicación No.
47, Segunda Edición, San José, Costa Rica, 1972.
18

Figure 1. -
Method for estimating
hydrologic yield of
u ngauged areas
( F r o m Referencelu)

Figure 4. -
Basins used for
checking results of
correlation
( F r o m Reference ,l#
19
 20

300

200
x

tra1 A m e r i c a

Kansas

L 1 I I
5u 1 O0 200 500

P M E A N ANNUAL P R E C I P I T A T I O N CENTIMETERS

Figure 3. - Central A m e r i c a n Values plotted into Smith's BCI f(P)

 WATER RESOURCES PROJECTS IN NIGERIA AND
THE HYDROLOGICAL DATA EMPLOYED IN THEIR
PLANNING AND DEVELOPMENT

Adigun Ade Abioduna

ABSTRACT

The need for adequate water supply to meet the demands of

Nigeria's growing population is well known. However, the technical
adviser is seriously handicapped in his planning efforts by the
lack of sufficient information. As a result, different kinds of
data and different levels of efficiency have been employed by the
various agencies which have planned the existing major water related
projects un Nigeria. This investigation shows and intensity of
rainfalls and the attendant floods, small scale project modelling,
projections based on hydrologic data from other but climatologically
similar places, provision of missing data by statistical correlation,
and intensive surveys over short periods to obtain rapid and exten-
sive information. These schemes have been reviewed and the hydrologic
information employed in designing them has been appraised. This study
also shows that Nigeria must intensify her efforts to provide exten-
sive basic data on her surface and groundwater resources if costly
mistakes are to be avoided in the future. A case is also made for the
use of new techniques such as Remote Sensing for rapid identification
and appraisal of these resources.

RES UME

On sait quels sont les besoins du Nigbria pour un approvision-

nement en eau capable de satisfaire les demandes de s a population
croissante. Or il se trouve que l e conseiller technique y est sérieu-
sement handicapé, dans son effort de planification, par l'insuffi-
sance de l'information. Les diverses agences qui sont chargées, a u
Nigeria, des grands projets d'aménagement des eaux, doivent utiliser
des données disparates ayant des niveaux d'efficacité différents.
L'analyse des problèmes montre que l'gtude de ces projets doit faire
appel -3 l'information locale sur la fréquence et l'intensité des
pluies, et les crues qui en sont la conséquence (petits aménagements),-
aux évaluations tirées des donnés hydrologiques recueillies dans de
régions climatiques semblables, -à l'utilisation des corrélations
pour boucher les lacunes,- a l'observation intensive sur de courtes
périodes pour étendre rapidement l'observation. Des efforts ont été
faits dans ce sens, mais il reste que le Nigeria doit les intensifier
pour rassembler une masse importante de données de bases sur les
ressources en eaux de surfaces et en eaux souterraines, afin d'éviter
dans l'avenir de coûteuses erreurs. On ne néglige pas non plus
l'utilisation des techniques nouvelles, telles que la détection 'a
distance, pour améliorer lainventaire de ces ressources.

* Lecturer, Dept. of Agric. Engineering, University of Ife, Ile-Ife,

Nigeria.
22

1. IBTRODUCTIOH
The developat of water reaiources miithin the p a t deeade, in Nigeria,
has concentrated moetly on the prorieion of adequate pipe-borne water for
domeetic and institutional supplies. The trend, however, ia changiiig, and
it is now realieed that water resource8 developent, a8 a ipa$Or economlo
revolutionary tool, ihould Qiphaeise ita total harnesaing, control and
utilization to provide in addition to watar supply, such other benefits 88
hydro-power, irrigation water, flood control, water transportation into and
from the hinterled, fish and wild-life, recreation and pollution abatarsient.
The awaxenees of these needa has provoked riome dee Unking and has,
in part, precipitated the putting together of the Färat ]81962-68) and the
Second (1970-74) National Developent Plans. The objective of the latter,
according to the National Economic Counail, being
"the achievement and mainteamce of the highest poeaible rate
of increase in the standard of living and the creation of the
neceieary conditions to this end, inoluding public support
and awareness of both the potaiti&le that exist and the sac-
rificee that will be required."

The implementation of the variow schemes coatained in t h e m pl- have

experienced sime hardship especially where technical man-power and information
were needed. In many inatances, the technPlogist is often called upon to
make far reaahing professional deaisiona, and quite often, he ia seriouelg
handicapped in hie $Lanning efforts by the lack of scientific information.
This problem of planning dthout facta waa amply stated by Andu (1) about
bore hole drilling (for water) in Yeatern Nigeria:

"1 have emphasised the handicap due to ecantinesa of hydrolo-

gical data; and eince the gigantia Five Year Developnent
Programme cannot be held up becaune of this, the practice
now is to confine drilling to areas with favourable geolo-
gical formations. Time f a c b r has made any exploratory

hole w e n in a geologically favourable area is chancy-

test drilling virtually impoaaible. The location of a bore

and there ia no sufficient guarantee that water of adequate

quantity ehall be etruck. It ie not uncomon to drill far
deeper than expected where the exhibited geological patterme
indieate otherwise...."
 23

In order to achieve thd goals spelled out in the National D w e l o p e n t

Plane, expertise are often imported to analyae our local data or to use
their "ingenuityn to generate needed scientific information on which our
planning and development programmes could rely. More often than not, histo-
rical data are either scanty, unreliable or absent, and the synthesized
data can only be 88 reliable as the historical but scanty data available.
For many foreign experts, handling the problwie of the tropics is a new
educational experience and most of these techniml consultants, who are
often from temperate climates can only draw on their bowledge and ewe-
rience of their own temperate environment and adapt them to plan for the
needs of the tropical zones.

Most of the existing water resources sehemes have been handled in the
nanner enunciated above, and sane of these techniques can in some caees be
referred to as "educated guesen work by the experts. Hence, this paper
examinea, in closer details, a few of the existing water resources projecte
in Higeria with a view to high-lighting the kinds of hydrologic data, analysis,
and the different levels of efficiency that have characterized their planning
and developeat. Such an evaluation should offes some guide-lines for
systematic planning in the future.

2. HYDROLOGICAL DATA COLLECTION

The hydrological data needed to effeat adequate study of water re-
sou~.cesinolude data on precipitation, evaporation, stream-flou and ground-
water. In Nigeria, the sole responsibility for collecting rainfall data
reste on the Federal Meteorological Service (IPPS). Although M!CS maintains
over lux) rain gauging stations throughout the country, utilizing the
recorded data Rom these statione for water resources planning would require
further interpretation and analyeis. This is so because these stations
uere not established in relation to river basins. Furthermore, those co-
llecting the rainfall data such as the local school teachers and looal post
office personnel owe no allegiance to the WI
TS since the latter never rewards
them in any way or form for their services. Hence, the accuracy and relia-
bility of data collected under the aforeaentioned condition are often in
grave doubt.

The measurement of evaporation data acroas the nation is also done by

the employing some 68 clans A evaporation pans in an area almost 590,000
-
square kilometres. In additiop, there are three lysimeter statione in Nigeria
two at Ibadan and one in Zaria. Although reservoirs are being built on 8
continuing baais, and evaporation acrose the land v m i w between 102 to 204
eentimetrea a year, the impact of evaporation on the yields of these reservoirs
is probably still not fully realised.
24

Stream flow data are collected by auch ageucies as the IpLand Waterways
Department (IWD) and the Ministries of Work. The former maintains over 100
gauging stations along the major rivers of Higeria for the expresseu purpose
of recardiiig stage heights which are used to determine navigable waterways.
The Ministries of Work on the other hand are m o m interested in potential
areas for the location of highway bridges, hence, most of their hydrological
stations are non-self recording. The unavailability of diacharge measure-
ments or proper rating curves that could be used to interprete the recorded
stage heights has rendered most of the date available unworkable. In the
Northern States, where there were ZIO reel hydrological net-rrodf until after
1960, most of the rivers are non-perennial and shifting; the latter situatinn
makes it mandatory to provide more than one ratirig curve per Station per
seaBon thus rendering most of the available record difficult to interprete.

The Geological Survey of Nigeria (GSN) is totally responsible for co-

llecting data on the groundwater resourceB of the nation. Most of GSl's
efforts have been concentrated in the Northern States where there is abun-
dant supply of grounawater and very little surface water supply. The GSü
in collaboration with the United States Geological Survey, haa carried out
some investigation in the Chad Basin complex, and estimates have been made
of the life of the aquifers in the basin as a result of groundwater mining.
However, no attempt has been made to quantify the annual natural groundwater
recharge or the contribution of the groundwater to river discharge. Informa-
tion is also not available on the groundwater flou conditions.

Although the agencies cited above collect vast quantities of data

annually, the fact ia that until very recently, the data collected have been
piecemeal and the hgdrologicd records were never checked nor analyred. In
many cases, the records have no duplicate8 and hence distribution is often
-
impomsible. This state of affairs is often due to two major factors lack
of funds and lack of badly needed technical mawpower. This dearth of
adequate hydrological information has not however precluded the planning and
the actual developent of a -ber of major water schemes in Nigeria euch as
the Kainji Dam and Lake Project on River Niger.
 25

30 HYDBOLOCICAL DATA USED IN EXISTING PROJECTS

Sequential generation of hydrological data has been a tool the hydro-
logist haa often used to create synthetic records, in the absence of very
long hiebrical records, that could be used in hia water reaourcea planning
efforts. Since the generated set of data is only as good as the historiaal
set employed in such a synthesis, the historical set should not be too ahort.
In the absenoe of such a hiatorical information, the planning anã derelopent

niques of hydrological data derivation as -

of the existing water resources schemes in Nigeria have relied on such tech-
local information, eduaateä
gueas method, projection based on hydcological data from other but climatolo-
gically similar places, provision of missing data by correlation and intenaive
surveya over short periods. A few significant schemes are examined below.
A. KAINJI W BND D q
The moat wide-ranging water reeourcee project undertaken to date in
Nigeria is the Kainji Dam and k k e Saheie on River Niger (Fig.1.). The scheme
was conceived to provide hydro-eleckic pwer, flood control, regulated water
for navigation, and fishery benefits. Although actìml construction started
at the Kainji site in 1964, water levels were never observed there prior to
1959. The two nearest statitma where
Y ologic data were observed on the
Niger prior to 1959 vere Jebba (Nigeria and Niamey (Niger), both of which
sandwich the Kainji site and are 906 anä 1630 kilometres respectively a m y
from the Atlantic mouth of Xivex Niger.
The pre-construction density of rainfall net work within the catchment
area of the Keinji project was too low to serve as the baais for any reliable
hydrological interpretation. Conaequently, new rainfall uging stations
were established for the project and a seva year record K955f959) was
obtained b y the consulting f i m (2). Through the application of the Thieasen
method, the total amount of rainfall on the catchment area was calculated
for the seven year period.

Although records of water levels at Jebba were available for the years
1915-24 and 1947-64, the ahiftirig positions of the gauges during those years
&e it impossible to oorrelate the datum points of all the gauges used.
Consequently, a decieim was made to correlate the rainfall with the nia-off
within the Niamey-Jebba catchment, using the newly observϊ atage discharges
at Jebba for seven years, and to employ this correlation curve with the cal-
culated rainfall data to establish a 1939-59 discharge reaord for Jebba.

Owing to the relative insignificant average value of the inflou between

-
Jebba and KainJi, the obaei-red and the generated discharge
were aaaumed to be the same for linin31 which ie upstream
were analyzed accordingly. In establishing a satisfactory
-
data for Jebba
of Jebba and
correlation between
the rainfall and runoff data, two steps were taken:
 26

Because of the wide variation in both the rainfall and the discharge
data betueen gauging stations, only monthly totale were uaed in the
d Y S i 8 . This approach was found to produce smoother oorrelation
curves than thore obtained from daily or 5 4 a y records.

The Jebba-Niamey catchment area i8 extensive, and the runoff contri-

bution to the Niger flow from the Dahomey catchment area takes a
longer time to reach Jebba than the o t h r catchent6 downstream.
Hence, a sequence of lag time wa8 introduced into the data analysis
to yield an expression heeeby derived as
where
QA,P+= CU~,.O + CPRe&-Ah),$ + C3Rr.n (1)
BA,,.,,= Runoff of the month of August from the Jebba-Biamey
catchment srea;
R 4 . ~U Bainiail of the month of July on the Dahomey catchment
area;

R($-k),s= Bainfall of second-half of July plus that of first-

half of August on the Sokoto basin;

e,,.
= Rainfall of the month of August for the rest cf the
Jebba-Niamey catchent area; and

'
C,, C2, C are nuioff coefficients far Dahomey, Sokoto and the
rest of Jebba-Niamey catchment area respectively.

The introduction of the coefficients of runoff in the above equation

became necessary as a result of the wide variation in the geographioal
nature of the catchment area.
Through the step enumerated above, the m 4 f f data from debba-iainey
catchment area were deduced for the period 193959, and th6Se w8re added to
the Niamey observed record. The net result is w.2, the hydrograph of the
Niger discharge at Jebba. This figure was in turn used to develop the maas
inflow curve info b k e I[a;Lnji. The most important atreaai between gaiaji and
Jebba is River Oïi with an estimted aiktchment correlation ooefficient of
0.2 The remainder of the drainage basin had an estimated runoff coefficient
of 0
.
1 These coefficients were used by the oonaultants in the equation
wwe =
Q,,,,, QJebh - -e,, - O.'F?, (2)
Q E river discharge in eubic metres/ulrit of time
P = rainfall in cubia metres/unit of time

to arrive at the mass inflow curve for Lake Kainji.

 27

The daily discharges used in the hydrologfcal analysis are very

interrelated and the peaks are interdependent. S i m e these daily discharges
exhibited a tendency towards persistence in succeasive stream flows, the
Goodrich dietributiona were used in the frequency calculations; the latter
were of the exponential type and were similar to the exponential G a b e l
distributions.

B. WATER SUPPLY I I MIDWESTERN NIGERIA

Water resources activities in the lid-West are centred mostly on

water supply. The latter is tapped, in general, from the various aquifere

--
which underlie 9% of the State. The Benin sand aquifer has the greatest
potential about 333 metres thick extending laterally to an appreciable
distance but the hydrological studies from which the aquifer chacterie-
tics could be obtained are etill in the planning stage. m y of the ;aquifers,
such as the Benin sand (3) and the Coastal Plain aquifers oan be described
only in the moat general tarma because of the lack of recorded data. There
i8 also no information on the hundreds of veils t h t tap water daily from
these aquifers.

C. UTER SUPPLY I I WESTERN NIGWIA

The Western Nigeria Water Corporation is entirely responsible for the
planning and the developinent of Water supply in the State. The Corporation
obtains the necessary evaporation and rainfall data, m y of which are very
long and reliable, from the Federal Baeteerological Service in Lagoa. However,
because of the scantiness of data on river discharges, the standard praotice
in those parts of the West, where surface uater has been developed, is to
base the rater scheme design on the following hydrological assumptions in
addition to a very liberal monthly evaporation of 127 mm:

(i) A conservative runoff coefficient of 4s

(ii) A once-in-50 years recurrence probability in rainfall with

where P = Percentage probability W rainfall being equal to or

lesa than a given talue;
m = rank of the year; and
n = number of years of record

The catchment annual ruaoff, Q, which is based on these assumptiom

can be computed from the expreseion

where A = Basing drainage mea:

28

= Baein rainfall value correspnâing to the probability of

orne-in-% yeare oocurenae.
Co = Coefficient of m f f for the basin.

]Equation (3) or a forin of it hae been widely applied on the numerous eurfaae
water apply eohetmee in the West. Eowvwr, bey%$ of the vast arai- area
of 7,500 sq. kilometres that is governed by the,project, a form of the equation
(3) shown above was not employed to predict the maximm probable flood.
Instead, Professor M. Parde of the University of Grenoble in France, a
speciaìiet in the field of flood studies, advised the consultants that a
runoff in the order of 490 litres per second per square kilometre is known
to hava occured within West African strema of similar importance ae the
Oehun river on which the achenie is estsb1ished.b Sapply Ibadan aith water.
Hence t h b value was used in caltuleting the project's spillway
deeign flood of 3680 cmbic metres per second.

The developaenf of groundwater resonroes in the West has encountered

a number of difficulties. When the existing wells were been developed, the
areal extent of the bed-rock formation was not fully known, and the lithogra-
phic characterietics of the water bearing formations, in many cases, were
etill to be studied. Absence of perfonaance data on the wells has only
aggravated the situation, and in most cases, local informtion on dug w e U ,
was often obtained from the inhabitanto.

The Geological Survey of Nigeria ( W N ) in collaboration with the United

States Geological Survey haa undertaken aome imestigativ6 work which had led
them to edict a 30 year yielding life for the Lake Chad Baain middle eone
aquifer K g . 3 ) at a withdrawal rate of 5000 gph with wells placed at 16 kilo-
metres apart. Huiidred8 of bore holes have been drilled but co-ordinated day-
to-day performance data on these bore-holes are lacking. Most of the informa-
tion that can be readily obtained on the basin's aquifere are available only
in special reports. It is also irapossible to undertake a meaninghil study
of the basin's aquifere within Bigeria along since four countries share the
b a i n area. The FAQarpd' -@are assisting the Lake Chad Basin Commiedon
a regional organization of the countries that have territorial claims over
-
parts of the basin, to

(i) Compile all the available date in the baain;

(ii) m e l o p an a n a l m e compu4er model that would miwilate au. the

activities that affect the quantity of water in the basin; and

(iii) Define the various aquifers in the Chad bydro-geological basin and
erdeavour to arrim at a synthesie or composite picture covering
the correlations between the atmospheric, emrface and groundwater
ae well as between individual aquifera.
 29

From hydrological stand point, the Kainji project has been more inten-
aively studied than any other water scheme in Nigeria. A number of h o u n
standard methods were used to develop some reasonable results such as the
correlation established between runoff and rainfall of the Jebba-Niamey
catchment area. Since the project design flows were in p r t derived from the
discharge data obtaineä at Niamey, the accuracy of the rating curve used to
determine the Niamey discharges shonld have been verified. The importance
of this project also warranted a longer hydrological record than the 20 year
reconstituted record used. This could have been sequentially generated.

In order to emure enough water supply, coneervative estimates of rain-

fall, runoff and liberal estimater of evaporation have been the 8taRdard
practice in the West. But such educated guesaes have not prevented water
shortages resulting froa both drought and under-design for the needs of the
communities served. It appears that these educated guess teahniques haye
never taken into consideration that moat of these watershetie would be opened
up in the near future ae a result of extensive and medhanised farming practices.
The occurence of %aximam possible rain-stOmR vouìd also yield higher peak
f l o w than most of the existing achemea, except the Aaejire project, have
been designed to handle. And the eubsequent floafiing resulting from the
faracing practices or the maximum rainstorm would not only wipe aut these
water schemes but would also endangaz lives and property.

Groundwater develogaent based on inadequate or scanty data has been

found to be both unecoaoioicnl and frustrating. Typical examples include
bore hole failures at Agbor and Warri owing to the mollapse of cui-,
and poor yield as a result of drilling for water in granite aone at Ijebu-
Ode. The problame of g r o d w a t e r developaent in the Hid-Vest are eubatan-
tial, and only a through analysis can fore-stall more problema in the
future. And the efforts of Mo within the chad Baain has aided and aacele-
rated not only the haraonizatim of the existing data, but aiso the evaluation
of data concerning evaporation and temperature meaeurenent by infra-red
aerie1 photography.
5. HPDBOLOOIcbL RESS(XIRCE
Planning without facts ehould na longer plague the water resource8
program of the nation, m r e especially, as we shift our emphiusis from
single giirpose water supply to multi-purpose schemes such aa the Kainji and
Kano river schemes. The latter, also a victim of hydrological data scarcity
ia enviaaged to p r i d e barrefita in such areas ae irrigation, hydro-pouer and
flood control. In order to enmare systematic planning in the future, the
newly created Water Resoaaraoa btitute.8haeld eatablieh a Hydrological
resource^ Centre whoee primary h c t i o n uill firet be the aollecting and
compiling of the existing piecemeal data and nates that are scattered all over
30

the nation. This responsibility should be a continuow one and the inforaur-
tion so collected ahould be published annually and made available to the
public on eale. The centre should also standardice, nation wide, the inetru-
mentation and hyàrological data collecting and recording procedures.

The information available to the centre can be upgraded both in quality

and in quantity through the application of Remote Sensing Technique (RST).
The latter ia currently available through participation in the United Stateet
urth ~esourcea~echnologySatellites program (EE~TSP). The data ana imegery
obtaiasd though such a p m g ~ a mcan be utilised in sweral ways including
the rapid identification and appraisal of our water resource^. The immediate
hydrological investigation required in Nigeria, uuder the EELTSP, includes:

The overall delienation of the aquifers of the Chad Basin and the
pattern of the groundwater movsment in the basin. Such information
can be integrated into the basin'e existing analogne model;
the monitoring of changes in reservoira' levels resulting from
evaporation and changes in rater courses resulting R o m erosibn and
siltation;

%he identification and quantification of the groundwater resources

of Southern Nigeria including the location of the position and
evaluation of the extent of salbwater intrusion along the coastal
aquifers. Data obtained from ERTSP would generate more awareness
of the problem in surface water developeat and would provide
needed information on which conjunctive ground and surface water
programa oould be baed.

6. CoáIcLusIOB
Within the past decade, a =ber of water resources schemes have been
developed, ard in genSral, these schemes hare been plaaned with very limited
hydrological data tbat were often extended through the applicaticm of
statistical techiqueo to provide rational design parameters. In others,
"educated guess" technique wan subatitnted. The net resuit of auch methode
haa been the failure of many water supply schemes to meet demands eapeoially
aurfng the dry seaeon ana the location of unproductive bore-hoiea which had
to b abaadoned.

The future of io~iiy of these a c h e s osnnot be accurately preàicted at

this point. bwever, there is the urgent need to upgrade the scanty data,
through a contirmoire qioiritoring proc688, on which t h e m schemee were built.
such a step WOUM provide a sound baeie for a ture schemer, and would e m m e
the propat execution of any modification on existing sc@nms when warranted.
 31

The scarcity of reliable data i s mostly due to the acute shortage of

hydrologists and middle-level techniciana in this diecipline. Herne, it
will be necessary for the Water Resources IMtitUte in collaboration d t h
some of the eriating Universities to develop and execute achenes uhereby
a large number of such teohnologiate and techniciana might be trained
locally to meet the urgent needs of the nation.

In order to eneure systematic planning in the futare, the power to

collect, compile and hairnonise all the hydrological data in the country
should be vested in a Hydrological Resource Centre. The hydrological
information available to such a centre could include data and imagery
obtained through the use of Remote Sensing Technique. The information
obtained will enable the nation to accelerate the pace of ita natural
resources developneat.
32

BIBLIOGWHY

1. Andu, J. A. (1965). Exploitation and Developent of Groundvater in

Western Nigeriay Ministry of Works and Trsmeport, Ibadan, Nigaria.

2. IJEDECO (1961). Niger Dame Project, Vol. 2, Hydrology and Beilervoir

Operation, Report auhitted to the Fedaral Ooverment of Nigeria,
Lagoa
3. Tahal (Water Planning) Ltd. (1965). Master Plan for Urban and Rural
Water Supply, Report submitted to the Hid-West Ministry of Works and
l!rsnsport, Benin, Nigeria.

4. Tahal Consulting Ebgineers Ltd. (1969). Akungba-Shapureka-Ido&

Water Supply Sahane, Plauning Report suinuitte8 to the Western Nigeria
Water Corporation, Ibadan.

5. -
Tahal and Motor Columbus Ltd. (1961). Ibadan Water Supply hejire
Daia, Final Design Report submitted to the Western Nigeria Hinietry of
Works and Tranpport.

6. Miller, B. E., R. E. Johnaton, J. A. Oloni, and J. A. Umma (1968).

Groundwater ñyärology of the Chad Baein in B o m and Dikwa Emiratem,
N.E. Mgeria, with special Eaphasis on the Flow Life of the Artmian
System. USGS Water Supply Paper 1757-1, U.S. Govt. Printing Oîîïce,
Washington, D.C.

7. üHESC0 (1970). Study of Water Resources in the Chad Basin, Report

on the resulta of the hoject, Conclueionis and Recamendatione, PSriS,
fiFrance.

8
. HEDECO (1970). Feasibility Study-gan0 River Project: Report eubdtted
to the %o State Winistriea of Agriculture sad Iaatursl Besotarcea, and
Worka and Sumeye, b o , Isigeria.

9. Federal Ministry of Infonuation, Iagos (1970). Second Hationel

Developent Plan, Fed. Govt. Printer, Lagos, Hgeria.
 33
. .

3G1 Ma2 of W Africa showing R Niger and its tributaries

---_ -~~drograph
derived S m m rom

I
I L
-
4dL

derived
.ved.
- N+ogTh
w*
From &sei
-.rat5

FIG 2 Hydrogragh of R Niger at bbba (1955-571 (Reference 2 )

 A N EXAMPLE OF REGIONAL CO-OPERATION F O R IMPROVING
THE HYDROLOGICAL A N D METEOROLOGICAL INFORMATION

Eduardo Basso*
Andrés A r r i a g a d a n n
Hcsrnando Neira**
Manuel PBrez Delgado***

ABSTRACT

T & e Centra8 A ~ s ~ i c aHny d r o m e t e o r o l o g i c a l P r o j e c t i n i t i a t e d

in S e p t e m b e r 1961 r a p r e s e n t s & c o - o p e r a t i v e e f f o r t a m o n g t h e c o u n t r i e s
of t h e C e n t r a l Amsricea Isthmus ( C o s t a Rica. E l S a l v a d o r , G u a t e m a l a ,
H o n d u r a s , H i c e r a g u a a n d Panam%) and t h e U n i t e d N a t i o n s D e v e l o p m e n t
P r o g r a m m e , acting aar e x e c u t i v o a g e n c y t h e World M e t e o r o l o g i c a l
O r g a n i z a t i o a . Its o b j e c t i v e s a r e the following: ( i ) i n s t a l l a t i o n o f a
b a s i c n e t W 6 P k of m e t e o r o l o g i c a l and h y d r o l o g i c a l s t a t i o n s , ( i i )
c o l l e c t i o n , p r e c e e s i n g and p u b l i c a t i o n of t h e d a t a , ( i i i ) t r a i n i n g
o f p e r s o n n e l by neans o f course.%, f e l l o w s h i p s or t h r o u g h t e c h n i c a l
p u b l i c a t i o n s end m a n u a l s and ( i v ) t h e i n s t i t u t i o n a l s t r e n g h t e n i n g of
t h e m e t e o r o l o g i c a l and h y d p a l o g i c a l s e r v i c e s in t h e a r e a . I m p o r t a n t
P r o j e c t a c t i v i t i e s h a v e b e e n t h e P a e t i n g of n e w e q u i p m e n t used in
d e v e l o p e d c o u n t r i e s in o r d e r t o s t u d y t h e i r a p p l i c a t i o n t o t h e
c h a r a c t e r i s t i c s and t r o p i c a l c l i m a t e of t h e a r e a , and t h e d e v e l o p m e n t
and a p p l i c a t i o n of m e t h o d s f o r m e t e o r o l o g y , h y d r o l o g y and s e d i m e n t
s t u d i e s w i t h l i m i t e d i n f o r m a t i o n . It i s c o n c l u d e d t h a t t h e i r use in
other a r e a s w i t h s i m i l a r c o n d i t i o n s c a n b e u s e f u l and t h a t r e g i o n a l
cooperation can b e one effective means for coping with inadequate
d a t a t h r o u g h t h e p o o l i n g of i n d i v i d u a l c o u n t r i e s e f f o r t s .

RESUMEW
E l Proyecto Hidrometeorológico Centroamericano iniciado en -
S e t i e m b r e d e 1967 c o n s t i t u y e un e s f u e r z o c o o p e r a t i v o e n t r e l o s p a í -
s e s d e l I s t m o C e n t r o a m e r i c a n o ( C o s t a R i c a , El S a l v a d o r , G u a t e m a l a ,
Honduras, N i c a r a g u a y P a n a m á ) y e l P r o g r a m a d e l a s N a c i o n e s U n i d a s
p a r a e l D e s a r r o l l o , a c t u a n d o c o m o a g e n c i a e j e c u t o r a la O r g a n i z a c i ó n
M e t e o r o l ö g i c a Mundial. S u s o b j e t i v o e p r i n c i p a l e s l o s c o n s t i t u y e n : -
(i) la instalación de una red básica de estaciones meteorológicas e
h i d r o l b g i c a s , (ii) l a r e c o l e c c i ö n , proceso y p u b l i c a c i ó n d e l o s da-
tos, (iii) e l adiestramiento del personal ya sea con becas y cursos
o mediante publicaciones y manuales tgcnicos, y (iv) e l robusteci--
miento institucioaal d e los servicios meteorolbgiccs e hidrológicos
en e l área. A c t i v i d a d e s i m p o r t a n t e s d e l P r o y e c t o h a n s i d o l a p r u e b a
d e n u e v o s equipos u t i l i z a d o s an p a í s e s d e s a r r o l l a d o s p a r a e s t u d i a r
s u a d a p t a c i ó n a l a e c a r a c t e r í s t i c a s y c l i m a t r o p i c a l d e l area, y e l
desarrollo y aplicacids de métodos para la ejecución de estudios m o
teorológicos, hidrol6gicas y de sedimentación con información limi-
tada. S e c o n c l u y e e s t i m a n d o q u e s u u s o e n o t r a s á r e a s C O Q c o n d i c i o -
n e s s i m i l a r e s p u e d e s e r d e utilidad.

* P r o j e c t Manager, C e n t r a l A m e r i c a n H y d r o m e t e o r o l o g i c a l P r o j e c t
** H y d r o l o g i s t Expert., C e n t r a l A m e r i c a n H y d r o m e t e o r o l o g i c a l P r o j e c t
*** H y d r o m e t e o r o l o g i c a l Expert, C e n t r a l A m e r i c a n H y d r o m e t s o r o l o g i c a l
P r o j ect
36

INTRODUCTION
T h e Central A m e r i c a n Hydrometeorological Project is a joint effort betweenthe
Governments of the Central A m e r i c a n Isthmus ( Costa Rica, El Salvador, G u a t e
mala, Honduras, Nicaragua and P a n a m á ) and the United Nations Development
P r o g r a m m e . T h e World Meteorological Organization acts as Executing A g e n c y
T h e Project started in September 1967, at a cost of 9. 2 millions dollars (3.3
millions UNDP and 5.9 millions Governments), which m a k e s this Project one
of the largest in this field. in M a r c h 1973 a second phase of the Project w a s
started devoted mostly to the Coordination and Consolidation of the activities in
Meteorology and Hydrology. This second phase has a duration of three years
and the global contribution of UNDP adds to 1.3 millions dollars.

P R O J E C T OBJECTIVES

T h e m a i n objectives of the first phase of the Project, already completed w e r e

the following:
Installation of 290 hydrometric stations in the six countries.
Installation of 830 climatological stations (60 main, 240 secondary and 530
pluviomet ric).
c) Institutional strenghtening of the Meteorological and or Hydrological Services
and the collection, preparation and publication of the data in both the n e w
and old stations.
d) Training of the personnel, by m e a n s of fellowships, seminars, courses,
publications and on-the-job training.

At the end of the project the n u m b e r of stations constructed surpassed by far the
goals; m o r e than 350 hydrological and m o r e than 950 of all kinds
of meteorological stations w e r e completed. T h e achievements in the activities
of data processing and publication as in personnel training w e r e m o s t remarkable.
In s o m e countries, meaningful results were obtained in the important task of
institutional building. In others the present condition is not yet adequate for the
needs of their development, but it is expected that during the Second P h a s e it
will be possible to complete the necessary arrangements for this. F o r Co-ordi-
nating at a regional level the activities in Meteorology and Water Resources In-
vestigations, a Regional Committee w a s created. This Cornmiittee, formed by
the presidents of the National Coordinating C o m m i t t e e s of the six countries, has
proved to be a n excellent arrangement and can be considered a good example of
regional Co-or dination.

PRE-PROJECT CONDITIONS
T h e conditions before the beginning of the project varied widely f r o m country to
country. However, in s o m e countries it was practically inexistent. Although
about 180 hydrometric stations w e r e in operation in 1966, only a few provided
reliable data. E v e n these had a very short period of records, normally less.
than five years. In s o m e countries the stations consisted only of a staff gauge,
without bridge or cable for flood measurements, in other cases limnigraplis w e r e
installed without a device for checking the river levels, Sediment m e a s u r e m e n t s
w e r e m a d e in only one country and water quality determinations were made. How
ever, the m a i n deficiencies arised f r o m the methods of collecting and processing
 31

the data. T h e O. 6 depth method of velocity m e a s u r e m e n t w a s used in s o m e cases

introducing errors in the stream gauging data. T h e discharge rating curves were,
generally extrapolated graphically, and no checks w e r e m a d e for the consistency
of the resulting information. Even though m o s t of these defects were recognized
the counterpart pcrsonnel lacked the m e a n s and the influence for improving the
situation.

T h e situation in Meteorology w a s similar. E v e n considering that s o m e countries

had fairly well organized cervices, the network w a s absolutely insufficient for
the needs of the region. Several Services had only one m a i n meteorological
station, in the principal airport. T h e n u m b e r of secondary stations in good w o r k
ing standard were less than 20. T h e rainfall observation network comprised only
the.main inhabited areas, and even there, only a f e w recording instruments w e r e
available. S o m e countries showed a complete lack of rain recorders and others
of evaporation stations. T h e processing of the data w a s quite rudimentary, and
their publication with a few exceptions, inexistent. Only five meteorologist with
university degree were available, and all five worked in one country. Practical-
ly, no co-ordination between meteorological and hydrological services existed.

TRAINING
T h e activities of personnel training at all levels w e r e considered fundamental
and received a preferential treatment f r o m the project.

Training in the Region. Training in the region w a s done with courses --in-
- -
cluding cour s e s by correspondence -, seminar s, on-the job training, confer ence s
and publications. Without including on the job training, approximately 500 people
received formal or informal courses. This does not include the personnel
trained by other W M O projects, such as the Chair of Meteorology at the Univer-
sity of Costa Rica or the Mobile Center for Training of Meteorological Personnel.
Practically all the graduates of these courses are engaged in activities connected
with the Project.

Fellowships. 37 fellowships with a total of 324 m e n - m o n t h were m a d e avail-

able to the Project, for the preparation of n e w personnel or for the improve-
m e n t of the training of the existing ones. T h e s e fellowships were a fundameotal
completement for the local training which w a s devoted mainly to a large n u m b e r
of low level technicians. M o s t of the fellows completed successfully their
studies, and s o m e obtained higher degrees in well k n o w n Universities. T h e im-
portance given to the practical training m u s t be noted; the course for preparing
technicians in meteorological instruments--Buenos Aires-- m u s t be specially
remarked. Unfortunately, s o m e of the fellows ieft their jobs with the G o v e r n -
m e n t s o m e t i m e after the completion of their studies, which m e a n s that their
and U N D P ' s effort w a s wasted. However, the percentage of fellows in this
case w a s relatively low, 13%. in addition, the Project co-operated actively
to obtain fellowships f r o m national and multilateral sources. in such a way,
64 m o r e fellowships w e r e obtained, without including the ones used for the
courses already mentioned. As a consequence, the total of trained personnel
has been significantly higher than the quantity that should have resulted only
f r o m the fellowships assigned to the Project. E v e n so, the shortage of capable
personnel can still be noticed, specially in the Meteorological Services.
38

Publications. T h e Project considered that one the m o s t effective f o r m s of

training in a dispersed regional project w a s the intensive use of technical
publications. in general, the reaction to these publications w e r e encouraging.
and resulted in a large d e m a n d of them, both f r o m the countries of the area
and f r o m outside. Their m a i n merit aves to the fact that bibliography in
Spanish was m a d e a.vailableto all counterpart levels. W h e n this paper was p r g
pared, about 100 technical publications and 60 reports had been released, .To
m a k e the diffusion of Project activities m o r e available a by-mounthly newsletter
w a s edited, O v e r 500 copies of each issue w e r e printed, making possible for
all m e m b e r s of the Committee to k n o w the activities of the others. T h e editorial
activity of the Project stimulated also the publications of the counterpart, in-
creasing their technical reports and data publication. B y far the m o s t important
publication of the Project is the "Manual of Instructions". which has been plan-
ned in four volumes. T h e first one deals with "Hydrometry" and comprises
three chapters: (1) Field measurements and installations, (2) Data processing
and (3) Sediments. Standards are set for the installations, field m e a s u r e m e n t
and methods for processiqg the information. T h e second volume is devoted to
IIHydrological Studies" comprising: (1) Verification and Correction of Hydrolo
gical Records, (2) Extension of Hydrological Records, (3) Duration and Variation
-
Studies, (4)Hydrometeorological Studies, (5) Floods. (6) Draughts, (7) Hydrolo-
gical Forecasts (8) Hydrological Studies for P o w e r Developments (9) Agricuitu-
ral Hydrology (10) Economic Aspects in Hydrology (11) U s e of Mechanical Data
Processing. T h e third volume refers to "Meteorological Observations" and has
been edited only in a preliminary form. T h e last volume "Ground Water H y d r o -
logy" will be prepared -hthe future. T h e M a n u a l is a i m e d to the m e d i u m level
technicians and includes several numerican examples, with information of the
area. Special emphasis has been given to the specific problems arising f r o m
the lack of long and reliable records. (1) (2).

EQUIPMENT
. T h e equipment component, formed the m a j o r part of UNDP's contribution,
adding to a total of about 1,9 million dollars.

Meteorological Equipment. T h e m a i n meteorological stations (type A) w e r e

in general provided with universal wind recorder, m e r c u r y barometer, m i c r o -
barograph, psychrometer, maximum and minimum thermometers. thermohy-
drograph, set of geothermometers, Robitzch actinograph, Campbell-Stokers he-
liograph, Piche and tank evaporimeter, tank level anemometer, water thermo-
meter, raingauge and rain recorder (Figure 1).
T h e ordinary climatological stations were provided with psychrometer, m a x i m u m
and minimum thermometers, raingauge, rain recorder and Piche evaporimeter.
in m o s t of the station of this type evaporation tank, pan level anemometer, and
thermohydrograph w e r e installed and in several, anemograph, heliograph, acti-
nograph and or soil thermometers w e r e also included. (Figure 2). S o m e stations
included also agrometeorological instruments. such a s soil-moisture gauges,
d e w recorders, lysimeters, extrasoil thermometers. etc. Several pre-Project
stations w e r e reinstalled and or completed with new instruments. T h e m e t e o r o
logical equipment included also six standard barometers, which were included
in the principal station of each country, replacement parts for the period of the
project and for s o m e time after its completion and equipment for inspection and
maintenance. Also included w e r e equipment for part of a regional laboratory
for calibration of meteorological equipment. T h e equipment provided was, in
 39

general, of good quality, and adequate for the needs of the Project. As far a s
possible equipment of complicated operation or maintenance w a s avoided, pre-
ferring simple and sturdy ones suitable for tropical conditions. In s o m e cases
'defects w e r e detected, but they w e r e satisfactorily corrected by the manufactu-
rers, by introducing several changes in the design of the instruments.

Hydrological Equipment. T h e stations installed included the total or part

of the following elements provided by UNDP: limnigraph --of the float type or
bubble gauge type (manometric)--damping pipe, housings, connections and pack
ings of the limnigraph, sets of staff gauges, cables and accessories (Cable c a q
for the cableway instailation plus a reasonable quantity of spare parts. It is im-
portant to note the fact that the standardized prefabrication of the construction
elements, especially the cableway towers, which w e r e designed for 3, 6 and 9
meters height, allowed a simplification of the construction of the stations, making
easier its transportation and mounting;a fact that w a s of fundamental importance
to reach isolated and difficult zones (Figure 3). T h e equipment w a s designed in
order to ensure a m a x i m u m of safety during construction and operation, provi-
ding the cable cars with safety brakes, the towers with stairways protected with
safety rings, etc.. . In addition the publication of standards of construction and
operation aimed to ensure this objective. A s a consequence of this, and rever-
sing the pre-project conditions, serious accidents happened neither during the
constructionnor during the operation of the stations. (Figure 4) T h e equipment
included a current-meter calibrating tan!!, which w a s installed at the Universi-
dad Centroamericana in Managua. Probably due to the careful supervision of
the design and construction of the building, the installations remained undamaged
by the earthquake of D e c e m b e r 1972. in order to gain experience with m o d e r n
instruments for level recordings, three digital limnigraphs w e r e qperated expe-
rimentally for s o m e years. T h e results of their operation was, in general, LUISA
tisfactory, because of extreme humidity and lack 'of adequate maintenance. This lias
proved that the selection of mechanical equipment w a s a wise one, and that the
gradual introduction of digital equipment should wait for m o r e development to
solve the observed defects and to allow training of specialized personnel.

Equipment for flow and Sediment Measurement. T h e Hydrological services

w e r e provided with flow meters, counterweights, winches and cranes, etc. to
ensure the adequate operation of the hydrometric network. Selecting the type of
current meters w a s also subject of detailed studies, and it w a s decided to use
both the axial and the Price current meter. after a consideration of their relative
merits. T h e m a n u a l of Instructions (1) of the Project contains instructions re-
garding the criteria to be used in selecting one or other instrument. in addition,
m e a s u r e m e n t s m a d e in Costa Rica and E l Salvador proved that the difference
between the measurements m a d e with both kinds of current m e t e r s is very small.
T h e Project started and intensive p r o g r a m m e of sediment sampling, for which
the acquisition of standardized D49 and DH-48 samplers has been fundamental
for the successful achievement of this goal. In addition the Project provided
construction,laboratory, navigation and transportation equipment.

Data Processing Equipment. This comprises fundamentally t w o groups: e-

lectronic calculators and peripherical comnutation equipment. T h e first group
cornprises conventional and programmable calculators, which have been used
preferentialy in the computation of streamflow measurements, hydrograms and
discharge rating curves. T h e second group includes mainly card perforators
for imput to conventional electronic computers. This equipment will be used as
a base for the future data processing centers planned in the second stage of the
Project.

DESIGN AND CONSTRUCTION OF THE NETWORK

T h e design of the climatological network w a s based upon the following crite
ria:
a. The first priority for the m a i n mateorological stations w a s given to the irri-
plementation of the basic synoptic network, which had been planned before the
Project. T h e remaining m a i n stations were located in intermediate points,
trying to obtain a relative uniform density and a good representation of the dif-
ferent climates of the area. Preference w a s given to installation in the main
airports.
b. W h e n possible, a n ordinary station w a s installed in each m a y o r agricultural
area. In isolated valleys with characteristic microclimates, a n effort w a s m a d e
to asign a station to each of them.
c. T o obtain the necessary interrelation between the hydrological and the m e -
teorological network, at least an ordinary st,ationw a s assigned to each m a y o r ba-
sin or sub-basin with co-ordinated operation of the meteorological and h y d r d o -
gical networks.
d. in scarcely populated areas the m a i n consideration w a s the availability of
observers.
e. Finally, the availability of air, land or water access w a s a limiting factor
in s o m e jungle, mountainous or isolated areas. T h e pluviometric network w a s
planned following the recommendations of the Guide for Hidrometeorological
Practices of W O , with the limitations imposed iy the lack of observers and
the inaccessibility of s o m e regions. Regarding the design of the hydrological
stations, these w e r e located in the following places:

i N e a r the m o u t h of the principal rivers and or their m a i n tributaries. ii. in

each main lake. iii. At the outlet of each main lake. iv. W h e r e d a m s of m a -
jor hydraulic works w e r e planned. v. At the entrance of a river to a m a y o r
valley. vi. At the crossing of a m a j o r river of a n international boundary. in
addition, s o m e stations were located in urban and m i n o r basins, based mainly
on utility criteria or, in s o m e cases, for use as representative basins. it
was planned to m a k e sediment m e a s u r e m e n t s in part of the network mainly at
the stations listed under i and iv. T h e complete plan w a s co-ordinated at a
regional level and approved by the Regional Committees (3). T h e impact of the
Project in the meteorological network coverage can b e appreciated in Figure 5
which s h o w s the situation before and after of the Project. A similar compari-
son has been m a d e for the hydrological network in Figure 6.
Sediment Measurements. (4)O n e of the subjects of m a i n interest for the
Project w a s m e a s u r e m e n t of the sediment loads of the rivers, since --with the
exception of Costa Rica-- practically n o information was available at the begin.
ning of the Project. At present., systematic samplings are m a d e in 136 of the
gauging stations in the area. Samplings are m a d e in accordance with the usual
techniques and are later analized in the laboratories established in the six
 41

countries to derive the sediment load. W h e n access problems limit the n u m b e r

of measurements, s o m e local observers take a point daily sample it w a s found
that in m o s t of the cases the concentration of this sample correlate well with the
average of the compusite samples. T h e use of the s e d b e n t rating curve, rela-
ting the solid discharge (G)with the liquid discharge (Q)has been used for c o m -
pleting the records. Figure 7 shows one typical sediment rating curve and
Figure 8 s u m m a r i z e s s o m e of the first results obtained by the project.
T h e bed load is computed using several of the usual formulas, and several ex-
amples have been published in arder ta explain the procedure to the counterpart
technicians (12). An interesting result concerning the sediment rating curve is
that the coefficient of the equation G A = an, varies between 1.4 a d 4. O. T h e
lower values of n (1.4 to 2. O) are associated with rivers crossing arid areas,
and the .value of n in general increases as the rainfall also increases. A theory
for explaining this has been developed by the project (20) and will be the object
of further publications.
'
Water Quality. Although this objective w a s not originally though of, the
Project has started a minimum p r o g r a m m e of m e a s u r e m e n t s of water quality.
At present systematic samplings are m a d e in only 48 stations of Costa Rica,
El Salvador and Guatemala, but it is expected that in the future this p r o g r a m m e
will be expanded.

STUDIES A N D A P P L I E D RESEARCH
M o s t of the problems in hydrology and meteorology in Central A m e r i c a
arise f r o m the lack of appropriate information, therefore this subject falls
directly in the m a i n t h e m e of this Seminar. Although in the area of the Project
a few, very few, meteorological stations existed with information up to the begin
ning of the century, this fact did not help m u c h in the evaluation of water re-
sources and m u c h less for the feasibility studies. T h e network established b y
the Project provides the necessary coverage so in m o s t of the cases the problem
is n o w o f "insufficient data" and not of complete "lack of Information1I.In Central
A m e r i c a it is n o w possible to undertake the study of the potential resources of a
basin or for estimating the m a x i m u m design flow, even considering that the
stations giving a n adequate coverage have only two or three years record. With
this information, a m o d e l of the weather responsible for the major floods can be
prepared. Based in this weather m o d e l it is possible to develop a n isohyetical
m o d e l which CaA%e transposed in time to the m o s t intensive storms, knowing
only data at a few rainfall stations and very iimited hydrological information; the
m a x i m u m historical gauge levels par example. T h e above mentioned m e t h o d is
now being used for the design flood of a large hydroelectrical d a m in southern
Costa Rica, and will be published in a future report of the Project. Data on
wind, present weather, meteorological phenomena, temperature and humidit y
obtained at two possible sites for the n e w airport for Tegucigalpa for short
periods of observation, have established the need of further information for a
meaninful decision. in this case the lack of information on cloud cover and vi-
sibility m a d e impossible a decision a s in the previous case. Therefore, it can
be seen that the problems of evaluation with insufficient data differ substantially
f r o m one case to another, and it is impossible to propose fixed solution methods.
T h e first case shows h o w the action of the Central A m e r i c a n Hydrometeorological
Project has m a d e possible the evaluation of water resources with insufficient in-
formation by m e a n s of a closed and co-ordinated w o r k between the meteorologist
42

and the hydrologist. in the Symposium, it would be important to recognize this

fact. Special enphasis has to be placed in the fact that in the area of evaluation
of natural resources with limited information, the problems will be solved best
wtth a close collaboration between hydrologists and meteorologists, since it is
impossible to separate the aerial and terrestial phase of the hydrologic cycle.
T h e lack of hydrological information can be compensated with meteorological
information and viceversa. "Elastic relations" which allow to extrapolate the
few observed data, based on s o m e knowledge of the mechanics of the p h e n o m e -
na, should be used as far as possible. T h e fact that the hydrometric data are
based on pluviometric information m u s t not be forgotten, since it provides the
m o s t effective tool for the evaluation of water resources. T h e scope of this

- -
paper m a k e s impossible to detail all the studies of the Project. A list of s o m e
of t h e m of which m o s t w e r e published is the following: Studies for determining
water requirements for irrigation (5) (6) (7) (8). Studies o n runoff forecasting
(9) (10). (Already being used for forecasting the operation of several reservoirs

bed load, for several projects. - -

in the area). Effect of the eruptions of the Irazú Volcano on the sediment dis-
charge of the Reventazón River (11) (12). Sediment computations, specially
Assistance for the computation of design flood
for several projects. - Development of methods for estimating floods in the
area. Figure 9 shows s o m e flood envelopes for all the Central A m e r i c a n area.
Figure 1 O shows s o m e rainfall envelopes for the area (1 3) (14).-Groundwater

-
studies with the analog computer w e r e m a d e for the Project at El Salvador (1 5).

-
Water balance studies (16) (17) (18). Figure 11 shows schematically the re-
sults of a preliminary study for all the Central A m e r i c a n area. Effect of the
temperature on the sediment load (19) Figure 12 s u m m a r i z e s the result of this
study.

STUDIES WITH INADEQUATE DATA

T h e inadequacies of data arise f r o m (i) incorrect data and (ii) short or insuf
ficient records. Although coping with this is one of the tasks of the Second
P h a s e of the Project, efforts for correcting and extending the available data have
been m a d e up to now. T h e M a n u a l ofInstructions of the Project (2) details the
techniques suggested for this.

Double m a s s curves are used for a first check of the quality of the data.
W h e n errors are found in the hydrological records they are generally due to in-
correct extrapolation of the stage-discharge curve. Jn this case several methods
for determining this curve are proposed, s o m e based in hydraulic relations and
other in the hydrological balance of the basis.

T h e filling or extension of these records is m a d e either using simple or mgl

tiple corelation andfor estimating the runoff based in the meteorological data
and basic characteristics.
U p to now, the checking and extension of meteorological and hydrological
records has been m a d e following specific needs, but it is planned to undertake
thin task in a co-ordinated and comprehensive f o r m for all Central A m e r i c a n
Isthmus during the Second P a h s e of the Project.

DATA PROCESSING A N D PUBLICATION

O n e of the main Project activities has been to ensure the prompt and adequate
 43

processing of the information. This h a s been achieved, both in meteorology and

hydrology, by m e a n s of m o d e r n systems based in the use of electronic computers.

Meteorology. T h e data collected at the stations are directly written in the

computer entrance forms, except where, due to limitations of the observer, this
has to be done in the central office of the meteorological services. T h e detail of
the f o r m s and instructions for filling t h e m are indicated in Publication No 84 of
the Project (20). T h e results of reading the graphs of the recording instruments
are also filed on the form. At this stage, the adjustment of the graphs by cornpa
rison with the direct reading instruments has to be made. Finally, before
punching these data on IBM cards, the consistency of the data is checked. This
system allowed the publication of the first meteorological yearbook (21) using
services of a rented computer. In the future, this system will be changed for
one that.wil1 requiere a m i n i m u m of services of commercial firms. Plans for
mechanizing the reading of bands and for preparing s o m e secondary processing
are also being taken into account. E a c h country will prepare its part of the
yearbook on uniform format, so that the preparation of a regional yearbook wili
consist of joining the national parts only.

Hydrology. T h e action of the Project has allowed the standarization of data

processing, following the usual recommendations in this kind of w o r k T h e r e -
fore it is n o w possible to ensure the reliability of m o s t of the records that are
published. At the s a m e time the deficiencies of the previous data are n o w evi-
dent. Therefore, the revision of these old data constitutes a fundamental activity
of the second phase of the Project. T h e Project has proposed a complete nie-
chanized processing, as indicated in the instructions (2) (22), but for the lack of
computing facilities this objective could be achieved only partially. In practice,
the computation of stream gauging is m a d e mechanically, either by m e a n s of
p r o g r a m m a b l e calculators or by conventional computers. T h e use of small pro-
g r a m m a b l e calculâtors or mini-computers will be extended to the second phase
of the Project. T h e translation of the graphs of the limnigraphs has been m a d e
up to n o w by manually, but the rest of the process f r o m there on is m o r e or less
mechanized up to the tables for publication. Mechanization of all this process
is contemplated in the second phase of the Pr'oject. T h e rest of the processes,
.
.i e. :rating curves, sediment computations, duration curves, etc.. , is m a d e
manually or with the use of the f e w p r o g r a m m a b l e calculators provided up to
date, but the trend is towards to a complete mechanizations of these computations.
T h e Project has published four regional yearbooks (23). of which the last three
have been prepared with the help of electronic computers. T h e s e publications
have received excellent c o m m e n t s by the users of the information. T h e yearbooks
contain in addition to streamflow records, lake levels, sediment discharges,
water quality, duration curves and flood envelopes.

OUTLOOKFORTHEFUTURE
T h e impact of the Project on the meteorological and hydrological activities
in the Central A m e r i c a n Isthmus h a s been impressive not only in the a m o u n t of
available information, but in the increase of the public concern with the importance
of these. T h e second phase of the project is aimed m a i d y to completing the
-
*

institutional strenghtening necessary to ensure the continuity of the activities re

quired for providing the basic information needed for the social and economical
 44

development of the Central A m e r i c a n Isthmus. T h e Project will have at that

time prepared the local Services for providing all necessary information for p r 2
ject design W h e r e this information is insufficient, tools will be available for
mbking a reasonable good estimate which will avoid delaying the implementation
of,the Project. W h e n the information is inexistent, criteria for obtaining a m i n i
mum set of data will be well k n o w n to the local technicians. Finally, the m e t e 2
rological and hydrological services will be in a good position for influencing na-
tional polices on natural resources, ensuring a rational and efficient use of them.

REFERENCES
1. -
PHCA. M a n u a l de Instrucciones;Hidrometría, Publicación No49
2. -
PHCA. Manual de Instrucciones; Estaciones Meteorológicas, Publication
No70.
3. -
PHCA. P r o g r a m a Regional de Instalaciones; Publication No20
4. -
PHCA. Medida de Sedimento s en Algunos Ríos del Istmo Centroamericano
Publication N"79.

5. PHCA. El cálculo de los requerimientos de agua en Costa Rica. Publica-

tion No39.

6. Hargreaves, G. Requerimientos de Irrigación y Balance de Agua; Proyec-

to propuesto Arenal, Costa Rica, Publication No87 del PHCA.

7. Hargreaves, G. Necesidades y Requerimientos para Irrigación; C o m a y a g u a

y Vecindades, Hondruas. Publication No86 del PHCA.
8. Hargreaves, G. Deficiencias de A g u a en Centroamérica y Panamá. Publica-
tion No88 del PHCA.

9. -
PHCA. Previsiones de Escorrentía. Publication N"46.
1 o. PHCA. Pronósticos Hidrológicos para la Operación de Plantas Hidroeléctr-
'=&tas del Seminario de Managua) Publication N091.

11. Basso, E. Sediment m e a s u r e m e n t s in several rivers of the Central A m e r i -

c a n Isthmus. Fall meeting of the A m e r i c a n Geophysical Unnion, S a n F r a n -
cisco 1971.

12. -
PHCA. Medidas de Se dimento en algunos rfos del istmo Centroamericano,
Publication No79.

13. Basso, E. S o m e Methods for Estimation of Floods with Limited Information

in O n e Tropical Area. Second international Hydrology S y m p o s i u m Fort
Collins, Colorado 1972.

14. -No
PHCA. Envolvente de Precipitaciones en el Istmo Centroamericano, Publi-
cation 81.
 45

15. PHCA. Factibilidad del Riego con pozos en el Proyecto Usulután El Salva
dor, Publication No25.
16. PHCA. Estimación Preliminar del Balance de A g u a s en el Istmo Centroa-
mericano; Publication No18.

17. Alghren, L. ; Basso, E; Jovel R. Preliminary Evaluation of the Water

Balance in the Central A m e r i c a n Isthmus. S y m p o s i u m o n the Water Balance
in North America; Banff 1970.
18. PHCA. Estimación preliminar gel Balance de A g d a s del L a g o de Managua.
Publication No7 5.

19. PHCA. Efecto de la Temperatura en el Transporte de Sedimentos. Publi-

cation No6 1.

20. PHCA. Curva de Descarga de Sedimentos. Publication No8.

46

Figure 1
M a i n Met e or olog ica 1 Stat i on.
 47

A_....

Figure 2
O r dina ry Meteor olog ical Station.
48

Figure 3
Typical Hydrometric Installation.
 49

Figure 4
Prefabricated Cableway Tower.
50

Figure 5
Meteorological coverage, before and after
the Project.
 51

Figure 6
Hydrological c o v e r a g e after the PFoject.
52

Figure 7
Sediment rating curve.
 53

AVERAGE ANNUAL PREC/P/TAT/ONM M

Figure 8
Results of the Sediment m e a s u r e m e n t s
54

Figure 9
Flood Envelopes.
 55

Figure 10
M a x i m u m Rainfall Envelopes.
56

Figure 1 1
W a t e r Balance in Central A m e r i c a
 57

StZE OF PARTlCLES M M

Figure 12
Effect of T e m p e r a t u r e in sediment transportation.
 METHODOLOGY EXISTING FOR ESTIMATING
FREE SURFACE WATER EVAPORATION

by
Francisco Cubas Granado

ABS T R A C T

The purpose of t h i s paper is t o recount the metodology

f o r e s t i m a t i n g f r e e s u r f a c e w a t e r e v a p o r a t i o n and p a r t i c u l a r y
in t h e c a s e o f a r e s e r v o i r w h e n s t u d i n g t h e r e g u l a t i o n c u r v e s
theceof o r t h e regulation-exploitation system, for estatistics
and e m p i r i c a l methods.

RESUMEN

E l objetivo de este artículo es recopilar los distintos

métodos para estimar la evaporación en lámina libre y particu-
l a r m e n t e e n e l c a s o d e u n e m b a l s e e n f u n c i ó n d e la r e g u l a c i ó n
que efectue y del sistema regulación-explotación utilizando m 5
todos empíricos y estadlsticos.
 60

Al1 water returning to the atmospherd due solely to evaporation

procosoen is an important element in the hydrologic cyole. Moreover,
it io a limitin;? factor for the effecient utilisation o 9 free surface
m t o r (reservoira, lnken, rivers, etc.).

In vie# of its big influonce in the water cycle, wvapqration has

beon the subject of innumerable surveys which, because of the diversity
o f sndc pursued in enmh one thereof, have not given rise to R h o m o p m a o u r
theory that could be accepted unanirnoualy. The sole purpose of this
pronent papar io to recount the methoùology exinting for estimating
I'rso rJurfaoo MttCr evaporo.tion. l~nrticularlyin the oose of a reßervoir,
vhen studgin,? the re$Tlation curve8 thereof or the re~lation-exploitation
riyntern, formulan are required which nay enable the m a p o m t i o n ocourring
to b e cotimnted ox evaluated when it occurs on J. large sc¿ile or must
lie taken into account for r.orkinL; out these calculations. Suah formulas,
oithcr empirical or on n phgBica.1 basis, may often mitipte the lack of
"in nitu" data.

1.2. %ctoro affectinp: the atmomhere's evaporating Dower

The a.tmoephore's evaporating power in tho avapoxation rate, ex-

prossed in millimetres of water, for tho period determined (mm. of
uistor per day, for example).

The atmoaphere'o evaporating povier faotors are: the hypometrio

deficit, temperaatureof the water, temepraturo of the air, insolation,
fiperd and turbulence of the wind, barornotrio pressure, the quality Of

Lhc irntar and u1ti tude.

In fact, moct of these po.ra,metere are corelatecl to each other and

th@ nractica.1 formulae used Tor eva.luntinf:evaporation gfily use8 those
:mmmnotera which a m e the moot important or easiest to measure.

1.3. b'actorr, afrectinn wa.-aorcr.tion

of Tree viater surfacea.
¡?ret?wa.tc,r r:iirfn.cr rv&porntion, ,Tiven that th'e atmosphere's
:>oi:e:r ir: cox:t:.i.nt, dencnds on ita ri.rea..l+and
~wl)n~*it~tj.n,~~ dapth. The I?I~.SS
 61

OF imter acto as P. regqlator so that if it is not R large area and

in aliallow, t!ir ternpcrn.tnre of thp whole eanilg follovrs the law of
thermal variation as np1)lied to itr, surfo.ce.

Iprne imter nurfri.ce evaporation in all the less in hot seaaona m a

p e n t r r in cold venther, the bi,:ger in area and in 'depth the water
napne ie.

It can bo considered that, ba.nically, free surface water evaporation

dc:)enda on:

-- ' i h enorgy availabla from uo1n.r rr.diation.

The perceptible heat traiiomitted throunh the air.
- 'l'ho airari cnpncity to trianoport imter vapour.

Tlir diff~rentmethods propozcd for estimating waporation may be

poiipod into tiro ccitr:;orieo:

EL)Empirioal methods giving rise to Yormulae baaed, mostly, on

BEIlton,a law with modi í.'icationo to Che f m t o r s affeating evaporntion.

b) VathodB with n rational basis of physicaï theories that may ha

aummrd up ant
- Ilcthotln baned on vater evaluation, consisting in performing
EL water input and output LaLance with evaporation being
calculated as an unknown in the balnnoe equation.

- ilethocin h
ard on the evnluation of energy where the bnlence
a
The csl-
mnde io a,n energy enterinc and leaving brzlanoe.
culation of evaporation io aimilar to the foregoing g~oup.

- Methodo bnaad on the rnnßri trnnuaort theory vihere cvaporation

in evaluated from the wind cpeed and the vapour preoouro
,ymdient between lhe surface i!ntPr and the supwinounbent
lnyerc of air.

The irater or cner.3 bo.l:i.nce methods are theoretically suitable for u88
in cn1culntin:y evaporation in 1 a . k and
~ ~ reservoirs. Nevertheless, it is
62

difficult to a.pnly tham in prnotice because of the error committed

in msaaurin(: some terms in the balance.

Nore rocent renenrch hno drmonstratod that over relatively long

prriodo, > t Icaat one month, the potential ovapo-transp'iration is
conota,nt Pnd only depends on climatic factors. This has led the
researchers to Geek empiric formulae depending on these factors.

In pmrcr!.l, empiric rormnlae Iinvc. been nought after by oo-relating

t i on wit li tti e Po 1 lowin,n met oo ro logica1 fa0tcw i
eva n o r-,

- The temperature of tho air

-- Incident SOIR.T ro.dia,tion
Air humidity
- A combination of tho foresing

Howover, many of these îormulno hnvo to be checked in practice

b(?îore using them on uurfnces or areas which are not thoee where
they were first obtained. Their contrast is obtained by oelibrating them
by actual measurements of evapora.tion on the basis of ihatruments al-
ren4dy exintin,? (rafte, tanice, evmorimeters, ato.). Xn fact, the only
prccodure fox direot meanurement of evaporation lies in solving a water
bn lance.

Let uö romnmber that:

. The hycrornetric deficit or atmosphere saturation deficit,
obtained ao n difference between saturating vapour tenaioh Fe lo the
irater n u r f m e tompera,turrT and the actual vapour tension pa in the
ri.mbisnt air, ir, the main ;,pa.rameterof the atmoophere avnpomting QOWBF.

- 'The hyqrometric condition or r3eiFee E of thc air

rcTrm5.n:: to t h o viater nurfoce temperature T is the quotient b e t m e n the
 63

the tencione 1%. 8.nd Pe (f. = Fan/fe) and represents the relative
humidity o r -tha air.

- The psiahrnmetric diffarence, obta.ined an a difference

b<rt,i.ic.cn the temperature 1 of the dry thermometer 8nd the temperature
T' of tha kict thermorneter, in noticeably proportional to the evapo-
cttion.

- 'Tho water aoturatiiic vapour tension incrmsea with temp-

eraturo anù decreasea with altitude.

- Xvaporntion incransea when kmrometric pressure drops.

- J:vitpûra.tion ( l e c r ~ a s ewhen
~ tho concentration of water
dia-olvad in salt increa.ae.a.

- Tho moot common are:

- hapoximetorß (Piche; Livin,lnton; Bellnni)
- Expcrimentnl basins or evaporation tanka (ClR8S A;
Colorado t,ype; ,%ld Xvapcrimatero)
- Clasairare.

whocta charac-terioticsand opcratinE nyntem can be consultad in tha

rxin ting bibliogwphy.

'Phn ovtiporation mennurot1 in thane a.pparntuo may be oonoidered REI

nn npproximntion of the atmosphere's evaporating power. Thene measurements
rnatile empiric fornulas to be establishad whose ooefficiento have to be
Inter nda.)ted to each onoe studied.

Tho common characterintic of all thhsne menouremants is that of

ronxen anting mscroolirnati c fact ors isthich, therefore, disp ens e with the
n:iturc) of the wirfacer: on which they are taken. In fact, the majority of
precn.utionn tzikcn in nsteorolor$cnl stations for taking these mansurenento
xrc intondcrl ::o thnt thcy cAn ;IC takcn as iiidnpendantly a6 poucible from
1oc:i.ï pocu i.inritiao or I'roiii t h nitin:: of the inntruaenta. IIonevar,

~ * vi):o~T.1 .tio ;I i:; :i.i~ cs: rrit i :,1 lg ni c 1.0c Ii .li:& t ic p h rnox e no n. .L'!i c r c)fo re,
64

it in not p017::ii~i.eto expect a,nythin!yrno:t*e L h m an a.pproririo,tioii

froin
this typo OP rrtimnte, a.ü IJC? have u.lreedy mid.

The rooults obta.ir.eci in the evapora.tion tanka must be multiplied

by tho trin!c coefficient, which will be peculiar to ench type, in order
to oritir!ia.-tc!
the actual evawration.

ïri addition, t h e pa.rn.meterva.luec wc h3.w to consider are those

exia Lin:: in the air-water surface intoqhase which are generally impoccible
to rdiwniire w i t 1 iro Iin,voto observo them at the inost ncceosibla points where
i.1; il: niipponcd that their values Co-rrlíite well with those which would
h:i.vr beon obt:iiiiccl in tho micl intarphase.

s
0
n
-
.
1
3
2 forrnula

In 1802, I):i.lton deduced that , just likc! other parameters, evaporation

on a. free m t e r surface is proportioria.1 to the hygrometric deficit. IIence,
hi2 cvo.por:i
tion forrnula:

E = II
(Pc? - Fa) = o( (Fe - Fa.), depending on the hygrometric
deficit

or, what ir: tho ao.rne8

= d :pe (i-€), drprndin,yon tho saturatinC;

vapour tenci ion and hjr$romet ric
de I;rea,
 65

In thc Pirat expression of Uaiton’s forinuls, II represents the total

:)resnurc! (,pmplus water vapour) above the evaporatine ourface. H’s
inl.‘lirenceonly intervenes ac a corrective term in evaporation problem0
riml m;ry b r tiiccanxlsd in R. rirst approximation.

‘i’hccocPPioicnt 4 in cha.racteri::tic of the meterological station under

oo iin irl erat ion.

‘J!hi:i formula civon very vRria.ble evo.poration valuos Prom one p l m e to

i’i.notJior, uhioh limite iti; une.

- i3 i:: the water evaporated in o. month of n days.

- P o (in mrn. of r.lg.) ir: tho nvera,p saturz,tin,swater vapour
tension nt temnoraturs T. This in obtained from hygrometric
tnbles.
- Fn. (in min. of JIg) in the actual a.veragemonthly water vapour
ttrnoion of‘ tho fiar nt the tima of tho T readings. Thin ia
obtained by multiplying Fe by the hypometrio deGree.
- I3 (in min. o.€ !!go) in the Lx1,roniotricaxerage monthly pressure.
- T (in QG) ia the a,vora.gemonthly tomporature.

in irhiolri:
66

vhore:
-E (in mrn.) io tho evn7oration in 24 houm.
- U, the :i.riiìnesn. 'Phi:: is calculated by tha equation D=lO@-humidity
at u ntmocphereo.
-V (in miles/hour) io the :i,vera,yewincl opeed over 24 hours.
- 'I (in QB), .the ::.vcrnp temperature over 24 hours.

3) Tiorton's eqiiation:

i - p + r Y - i
W-h
with P i o for n inrater ahcet irith 2. crflall ßurfaoe.

4) Rohwer' a equation:

E=O.7'(1 (1./16s-0.01:36 n1.Y. ( F d a )

 67

D = averaee barometric pressure.

C numerioal coefficient
1% = prenent vapour pressure
Be sotimating vapour pressure
h m lblative humidity of tho air
P = fraction of time during which the wind is turbulent.
t = nurnber cf days.
Ta = average tempwature cf the air
Tw = average temporature of surfaoe water.
N P monthly wind speed average
'y wind factor,

&3.6. ûbnarvation

The dií'ïiculty in applying these formulae lies in that the

mnjoiity of the vnrisbles appear as nn average vsluo and it is possible
io2 their valueci not to represent their total VQlUe well.

'1.1. Critime on the mothods

Tho methods based on the e n e r m balance enter more into the

field of research than in that of praoti-1 usage. Yhey can be usied for
contrant or chock purposes. They can give evaporation Values over periods
or tima which are nufficiently long.

Tho methodo based on the water balonce or OR ,th0 mass twnspcrt

ihoory 1i.m likawins mur@ suitable for uoinc 81 ohecke than fa2 graatioal
utla~a. ThePie are, however, more recomendable for invsetigating svapomticn
over ohort periorln of timo (a few hours).

'Pheee nothaaa also always hitve to be chaoked &na contrasted, as

hnppene with tho ampirio mothods, on the basis of direot eveporntion mean-
iiromonto, bccnuee without these contTaStB and the modifications resulting
thorofrom, oounte-active resulto may be given.
 68

In view oi the faet thRt in the imter or energy balance

rn(>thods it is difficult to measure the terms appearing for evaporation
ctiidy on free inter eurfaccs, the use of methods based on the mass
transport theory is more otxongly recomnendad.

3.2. :mos t r m m o r t methods

In masE tmnsport methods, the value of evaporation is entimated

ïroin the wind opeod aiid the vapour pressure ,?radient between the surface
water cind the laysra of air, bu w i n g a formula of the typez

R = ri fl (u) f 2 (ao - 1%)

iilioro I I ir. the pronortionality constant commonly known as the 'ImaSB

tmmiport coePí'iciontlt, 21 and f2 nre knovm functions of the wind apeed
anti vapour prop nure {yadisnt res? cct ively.

The í'o.re{:oin,y formu1t-c ir: rrritten, in its more u m a 1 forrnt

iihioh enn.bleii E to be calculnted if i:e know the value of H beforehand.

3.7.1. Cp.lcula.tin,s I!

N'E: value in obtained in two waynr

1) Dy catirnntiny: the evaporation by other methods and dividing

i I; by ths product U.(Fe-Fa)

2) Dy obtaining n linear c.clyreocionarqmtion b e t w n n the chango

[)I' tlio r:tnto OP the wa.tcrl\lI md.thr product U(l%-%), in the following
i.!ayg
A 11 fl.IT. (F'c?-~) 5C
iiherr bhc coii:;to,rit C giver. the avcrn,Te loss from filtration in the free
1I:l:LPr ::iii..l'n.cc.
 69

Thr ficrr t pro.-,edurefor i”ind,inf;the mass transport coefficient

mtriiirec n. qrncise water balance crhich forces exact measuring o € the input
and output o r the surface wa.t,orunùer study to be ca,rried out. This may
not; bc prn.ctica.lin till CRDBB.

The nocond. procedure in cntisfmtory if the losses not due to

rva.pora,tiondo not vary to any ,great extent or are not the most part of
tho m t n r loet from thc surface in question.

It ia interesting to look into the aeroAynarnic formula, here,

that cnnbloo T7 to bc calouliited R B n function of the shape (perimeter) and
six0 (nroa) of the free irater nurface, i+moncFtother variables.

3.%.2. The rnntooroloaianl .station rrcruirod Eor viorking out the

a ero dynnmic met hod

The fo11owin:y rnateriii.1 iu necesnsry for the typical station:

- A Ch:::: A raft on the edge of the water surface.

- ‘l‘woCUD type wind B u c e o
- %io limnigrnphs
- A water tempersture recordar.
- A hy,irothermograph and a pyrmometro.
- Several plqvioTra.phs cet out all around the perimeter o f the
ourfnco in question.

J.?. 3. kiorkinrt out the evanore,tion esuation

Ilvnpomtion ia, with rajytrd to i’aorkinc out the formula mentioned

in 3.2.1., ri. diffunion proceso in which the water vapour is transported
froin the tinter nurface to thn Fi.tmocnherc.

Vrrtichl trn.nni)orta.tlonof tho vo.nour depends on thn effective-

naon of the tur.hiilent mixture in the lor:er layers of air, r.nd the main
inl.‘liisncetharcin in tho wind npcred and ro1ir;hnaon O P the surfa.ce. A t u r b
ii1an.t coefficinnt ban be found which varies rrith the vind cpeeù Por each
 70

tlotcrininotl niirfnco, ir1 thc caso whcrrby the narodynarnio characteristics of

tlic 1::tter romain conntant.

The tramsport o f the vapour takes place under the vapour

~)recsure{:radient cet up between thc vapour saturation pressure' at the
curlno<?w t e r temperature r m d the vapour pressure in the ambient air. The
wind acting on the isacterourface is continuously ctirring up the wn,ter
vapour nnd maintninu the ,yradient.

Unücr theno conditions, the m o u n t of vapour transported is

,:ivrn Uys

L¶oreovor, the wind flow on the m t e r nurfaoe produces a

clittin,? forcas

In thcrie formulae;

-p ir, tho density of the air.

- Kw io the turbulent transport coefficient for the inaso of
water.
- KM io the turbulent transport coeffioiant for the active
force.
- q io thc specifio humidity.
- 2 io tho vertical ooordinnte
- U in the wind speed.

8) The coefiiciento Kri and KI3 osre the stime. Thin equality
hna ,icm the subjrat of rrinny numrcyo and it ia gcnemily aooirptQd ns
va 1id .
 71

b) It iu accci>tcri thnt the laminar sublnyer ia of a negligible

thicknsnn nntl thnt tho turbulent boundary layer extondB below the
wrfaca water.

c) The riincl cpeed profile is given by the law:

m
u(z) = aZ

tihere t h conntc?.nto2 and fi depend on the etability nnd roghness of the

::iirface.

Combini'iig the equations given for E ana T,we havez

From $ho wind profile i:: found:

irliioh a.lloi~nus to irrite:

dz. E. zm-1

Integrating for

E = <
u2

Throou,yh a,nalap;ytrith the fluid flow through & u n i f o m tube,

i t c m be shotrn thnt:

O,:! Q.2 0.4, 1,8

U
(2m+l) (di)

nliich civer: rice to tho followinc expreesion for E:

 72

where:
- -
Tho a.vern.gospeed value U is equa.1 to the product
(K2. IJ2)
- K1 and 1 2 aro numerical constants.
- V ic the kinetic viscosity.
- X i n riven by X e, with A and P being the area and the
perimeter of the surface under study.

139 expresein,y the aaecific humidities as a function of the

vapour t en0 iam, the evnpora.tion equation becomes t

if 6 is tho thickncco of the turbulent layer, it is easy to show that:

K l = m + l ; K 2 = 1 .
m 4 3

and thon:

Tho ciluationn (1) ; m l (7’) aro the formuin proposed by thio

ii,o-rotlyrinmicrnet hod.

In order to facililate the calculation of evaporation in small

.ux’hcrs, lhe Col loirinl: hypoihcocs can Uc rnnde:
 73

1
a) Give the wind profilo exponent the valus m=s
b) Take a, value of 6 metren a6 the thickness of the turbulent
boundary layerr

Thon, tho mass tranoport coeffioient ie given by2

o1 2
N = (2.62 x 10-4) p
(A)

3.2.5. Conclusiono

Thena fortnulno ara vary useful when, st al1 times, tho area
nnd perimeter of the \rantersurface under survey are known. They are very
important, then, for applying to tho study of the evaporation change that
would OCOUT in n re~ervoirin every situation thereof.

In order to oalculate the value of N, th4 values for U, Fe and Fa

obtained at R meteorological station near the zone under survey can be
UCoa.

The inclusion of the perimeter and area in the formula for

cnlcutnting I compenses the variability OS this mass transport cooffioient.

The application of this formula to very irregular ehapad water

ririrfiloao may lead to an excaesiva ca.lculation of evaporation because of
the effect of tho perimeter in the formula. In such caees, to mitigate
.thio exoesn, we c ~ maker
n

u.Thorthw&t W I o 3 tzmanOo formula

‘Phis expraeoeo evaporation by

(in inches/hour)
 74

EJh ere:

- F1 ana B'2 are the vapour pressures (in inchee of Hg) at

nltitudeo hl and hp.
-
- U1 and U2 are the winü speeds (in m/h) at the said altitudes.
iri thc nvertrn,Te temperature (inop) of the air between altitudee
hl n.nd h2

3.4. PItnmnn's formula

'Chio haCj tho expresoion:

E I 0.4 (1 4 0.17 U) (Fe - Fa)

whore E io civen in mm/day and the wind veloaity U, at 2 m. height, in

mile5 per hour.
 75

UT i3LIOGRAPlIY

t l ; d o d o nen uso y DU empleo para cdlculo de la eva,potranspira~idn~~,

by Paustino Lonnno Cnrcfa.- February 1964.- Publication no. 23 of the
C.E.11. OP tiio Xiriistry of Public Blorks.

"L'hytl xo loeie d R 1 '

ingenieur", by C. ii6rn6ni&$as. -Publishad by Eyro 1l es.

'llI:indbaok of applied i!:;ciro1.ogyW, by Van Te Chon.-Published by Elei.c-Graw Hill.

"Netodos prdcticoo P R M . el estudio hidrologico completo de una cuenca",

by R. fieran.- Published by the C.F.H. of the Ministry of Publio Works.
 GEOHYDROLOGICAL STUDIES IN
SMALL AREAS WITHOUT SYSTEMATIC DATA

Emilio Custodio Gimenan

ABSTRACT

Frequently are needed studies t o profit ground water resources by

means o f wells or galleries in areas with non existing data on river
and spring flows and on recharge, but in which injuries may be imposed
on pre-existing water uses. One begins looking for available data in
several kinds of files and inquiring local people. Moreover, the size
of the existing water concessions and their specific use allows the
appraisal o f mean and base discharge. The pluviometry is obtained though
the closest stations, and some corrections on judgement. The key problem
is the effective ground water recharge calculation, beeing solved
through the consideration of three independent points o f view:
1 ) modified hydrometeorological balance
2 ) ground water flow calculation based on existing or estimated data
3 ) salt balance, specially chloride, based on water -table chemical
analysis and rain water composition
Generally is possible to get coherent results. As an illustration,
three cases are presented:
a) Montroig Area (Tarragona). It is a coastal plain
b) Riera de Carme Basin (Barcelona). It is a limestone formation
c) Famara Massive (Lanzarote, Canary Islands). It is a basaltic
formation in an arid clima
Key words: scarce data, ground water, chemical balance, perameter
estimation, subterranean flow, case histories.

RESUMEN

Con frecuencia deben realizarse estudios para aprovechamiento de -

aguas subterráneas mediante pozos o galerías en zonas en las que exis-
ten datos sobre caudales de ríos y fuentes, ni sobre la recarga, pero
en las que se esperan afecciones a usos y a establecidos. Se procede a
la búsqueda de los posibles datos en l o s archivos y al interrogatorio
de los habitantes. P o r otro lado la importancia de las concesiones --
existentes y su destino permite apreciar los caudales y los caudales -
de base. La pluviometría se interpola a partir de las estaciones más -
próximas efectuando correciones estimativas. El problema clave es e l -
cálculo de la recarga eficaz a los acuíferos y se ataca bajo tres pun-
tos de vista:
1 ) balance hidrometeorológico modificado
2) cálculo del flujo de agua subterránea a partir de datos disponibles
o estimativos
3) balance en sales, en especial en cloruros, a partir de los análisis
del agua freática y de la composición del agua de lluvia.
En general se obtienen resultados coherentes. A título de ilustra-
ción se comentan tres casos prácticos:
1) área de Montroig (Tarragona). Es un llano costero
b) cuenca de la Riera de Carme (Barcelona). Es un macizo calcareo
c) macizo de Famara (Lanzarote, Islas Canarias). Es un macizo basálti-
C O en clima arido.
Palabras clave: datos escasos, agua subterránea, balance químico, e s t i
mación de parámetros, flujo su,bterráneo, casos reales.
9; Comisaria de Aguas del Pirineo Oriental y Curso Internacional de Hi-
drología Subterránea. Barcelona.
78

1. -
INTRODUCTION

Frequently, geohydrological studies are made in small basins

where problems of water use exist or are foreseen. To solve
these problems, data is required which has not usually been
compiled o r taken down. Usually, there are only a few
pluviometers in the area, and are of dubious reliability; there
are have no measurements of the water courses as they are small
o r ephernerous, and the springs or sources have not been
controlled. On the contrary, the exploitations established may
be o f a the same order of magnitude of the total available water
resources.

It is not possible to give general working norms, since

there is a very wide range o f climatic, geological structural
conditions etc. After setting out some general rules, three
cases will consequently b e discussed, showing notable differ-
ences in conditions, discussing the form of operation and the
guarantee of the estimations made.

The main objectives of the work to be carried out may be

summarized as follows:

a) Knowledge of the groundwater flow pattern, including

recharge, circulation and discharge. The identification
of the main aquifers is one of the stages to be covered.

b) Obtain a reasonable hydraulic balance, if possible

coherent with the results of various independent
estimation processes.

c) Analysis of the existing and projected water up-taking

ernphasing the possible interferences between them and
with the water courses and springs, and also, if possible,
obtaining the foreseen user's extraction programme.

It is important to remember that one is obliged to cany ont

these studies should be during certain months along o r at the
most within a year; consequently the only data available to
compute the components of the mean hydrological cycle are
those existing at the time. The hydrological data taken during
the study are not mean values, but depend on the climatic
conditions during the study and past actions, and they should
consequently be corrected to obtain a mean or pre-established
situation.

One needs to solve the problem by various channels a s

independently as possible. In principle, they may be included
in any o f the following three large groups:

a) Hydrometeorological methods.

b) Geohydrochemical methods.

c) Hydrodynamic methods.
 79

Details o f these methods will not b e discussed in this

paper a s their general lines are well known. F o r further
details, the reader may consult the two volume text:
"Hidrologia Subterránea" coordinated by M. R. Llamas and E.
Custodio, at present being printed by Ediciones Omega, Barce-
lona.

The investigation and special methods are expressly

excluded, since this is not the right place to discuss them,
but the studies to solve the real problems raised and which
require a prompt answer and an order o f magnitude o f their
confidence. More delicate works can later be set up to find
or affirm the basic estimations and hypothesis.

2.- DATA COMPILATION AND SYNTHESIS

T h e data compilation and synthesis work is necessary in

any hydrological study, but in small basins with insufficient
data, it assumes peculiar features, since it is frequently
necessary to test all possibilities in various aspects.

First, it should b e defined the sort of is necessary data,

to later define the search places where data can b e found and
finally establish the methodology of compilation and elaboration.

When discussing the three factors defined in the introduction,

will be specified what data is necessary, if they already exist.
The places where the data can be found vary from one country to
another, and from one place to another and a list of them would
prove very tedious. The official centres officially in charge o f
filing and compiling certain types of data, and their publications
should be permanently in mind. The consultation and help of local
experts may prove essential, and also the water-well companies;
furthermore one should not forget the local people as well,
without whose collaboration many important aspects may pass
unnoticed, and even essential data o r also some time the main
profited sources and wells.

Except perhaps for very little developed areas, with abundant

water resources, the local people have a noticeable, often
unconcious, knowledge of the local hydrology, o f a qualitative
nature, but which may be quantized and built up with adequate
surveys. This method of obtaining data not only saves a lot o f
work and time, but perhaps is the only way of obtaining historic
knowledge and erroneous conclusions, by building a logical
structure on not well foundes basis.

A good knowledge of the local idiosyncrasy and people

customs is needed for these tasks, and they should not be given to
under-qualified people who raise suspicions and are not capable
of handling, screening and correcting the information received.

Generally speaking, the local inhabitants are not very

willing in principle to reveal their knowledge, out of fear
80

it may prejudice them. The interviewer should be prepared

to “waste time” in winning over their confidence and present
the survey without them noticing it, making notes discreetly.

One should try to get the information to flow out on its

own, just channelling it and loocking for the interesting
details.

It is generally difficult to pass judgement on the data

obtained in this way and it requires a great critical sense,
a good knowledge o f the area and a continuous contrasting.

The collaboration of the local inhabitants is most

important in locating springs, bore-holes, wells, etc., and
t o establish the most important characteristics of them. On
the other hand, the local corporations and Town Councils are
usually important sources of information.

3.- -
OBTAINING THE OBJECTIVES

T o obtain the objectives listed in the introduction, it

is frequently necessary to set up a general water balance in
homogeneous part ia1 areas , bearing in mind the limitations
and inevitable errors they contain. One should not only see
an equation between mean values in the word ”balance”, but
also the possible variations in the different values
intervening and their interconnection (SC). This is specially
important when the ground-water reservoir capacities available
are small in relation with the water volumes t o be exploited
annually, giving rise to accentuated seasonal effects.

This raises the problem that in one of these small areas

where data is scarce and not very reliable, an elaboration
and definition is necessary, with a depth not common in the
case o f large basins.

One o f the greatest unknown factors is usually the

infiltration and recharge to the aquifers , which should b e
estimated using the best methods available.

4.- - -
HYDROMETEOROLOGICAL METHODS

The hydrometerological methods to establish water balance

and define deep infiltration are the classical ones, except
in arid or semi-arid areas, where a daily computation ( 1 0 )
must be made to avoid excessive errors in monthly data handling.

(*) The autor is conscious of the limitations of the water

balance but feels it is a very useful tool if the person
handling it is aware of its restrictions and errors, and
the variability o f the involved magnitudes.
 81

The pluviometry must be obtained through the usually

scant stations available, which generally do not cover the
mountainous parts where the pluviometry is usually greater
than in the lowlands.

A first measure is t o correlate the different stations

and complete the series, trying to obtain a definition of
areas with the same rainfall (quantity, distribution and
intensity), making estimated altimetric and topographic
corrections.

Next, the graphs of accumulated deviations of the

pluviometry should be drawn, and these will be the basis of
the study on the springs and water courses discharge and
water-level in the wells. With these relations hips, the
conditions observed during the study will be changed into
mean conditions or those conditions of particular interest
in order to ascertain extent the pluviometric variations
influence the ground waters.

When estimating the surface runoff, the knowledge of

the local people may provide interesting data helping the
morphological appreciations made. Frequently, local people
can tell the heights and frequences of water in the river
beds under various circumstances, and thus draw an initial
scheme of the system. When there are permanent waters is
frequent the presence of manufacturing or irrigation
installations which use them, and in this case they are usually
dimensioned for the base discharge or some figure slightly
higher. A knowledge of this discharge and the user's remarks
are of great importance, as it permits the characteristics
of the surface hydrology to be reconstructed approximately,
based on one or various river flow measurement campaigns in
selected points. The absence of noticeable surface uses by
means of simple derivations, may be a clear sign of temporary
discharges.

Rarely are there homogeneous and well defined crops in

these basins, and frequently there is forest, brush, bare
rock areas and great slopes, and consequently the classic
evapotranspiration estimations are not applicable. Added to
this is the rare availability of the necessary meteorological
data, excepting some thermometric station. Thorntwaite's
method (9) may give an initial idea of the potential value.
Successive balances based on an estimated field capacity,
enable the real evapotranspiration to be calculated by
difference. In low pluviosity areas, with a high evapotrans-
piration capacity, the errors may be very important and a
value of the infiltration plus surface runoff below 10 or 20
per cent of the annual mean pluviometry, may only give a mere
indicative figure. In this case, other balance methods should
be established.
82

5. GEOCHEMICAL METHODS

In studies where there is insufficient data, the chernical

characteristics of the ground water may be extremely useful,
if they are correctly interpreted.0ne of the chief advantages
of the geochemical methods lies in the low variability of the
chemical composition of the ground waters, averaging the
annual and seasonal variations, and in the low cost of an
overall indicative analysis if there are sufficient points for
the sampling. The interpretation however, is a delicate affair
and should be made by an experienced person with sufficient
knowledge of the local hydrogeology and geology.

On the one hand, the geohydrochemical methods may help

to establish the paternof the groundwater .flow, comparing the
analysis of various points of water, using graphs (mainly
those of logarithmic vertical columns or Schoeller's;
triangular with three fields, or Piper's; and Stiff's modified
polygonals) (8) and ionic indexes, helped if necessary by dis-
persion diagrams (correlation between two chemical charac-
teristics (8).

From another point of view, the chemical composition of

the ground water may further information on the recharge. For
this, it should b e admitted that the aquifer does not notably
modify the salt contents of the infiltered water. To remove
the possible influence of solution and modifying phenomena
such as ionic exchange, redox reaction aggressiveness to
carbonates, precipitation etc., the chloride ion is taken as
reference, which can only be changed by an addition of a new
chloride ion by the aquifer. In alluvial aquifers, limestone,
dolomite, etc., no important additions are expected if for
from the sea.

In this case, all the chloride of the ground water would

come from rain, and therefore we can state: ( 2 ) ( 4 ) ( 5 ) (8):

(P - E) . Ca I . Cs
Where:
lr
P annual mean pluviometry
E = annual mean surface runoff
I = annual mean deep infiltration
Ca mean concentration in rain water chloride
C, = concentration in ground water chloride

It can be easily deduced that:

I (F - E)
cs
 83

Some care is required when applying the method. One is

that the activities on ground surface should not modify the
chloride contribution. The method therefore has a dubious
application in intensive crop areas with irrigation or in
areas with disposal and infiltration of important amounts of
direc.t residual waters or though tippers, etc., and also in
immediate coastal areas with direct sea influence.

The Cs value may easily be obtained from the ground

water analysis provided this is stable. If not, the method
should not be applied. The value of Ca is not normally known,
as it is rare to find systematic analysis of the rainwater;
during the study some analysis of this rainwater may be made,
but before taking a content as the mean value, various deter-
minations must b e compared, since this content varies each
season according to the origin of the clouds and even within
a same rainfall. During initial estimation attempts, it may
be considered that in areas several dozen km. away from the
sea, the chloride content is generally less than 1 0 ppm, and
that in areas some hundred kms from the sea, it is less than
1 ppm ( 8 ) .
In areas near the coast, the variations and contents may
be higher. Close to populated areas, in particular if these
are industrial, high values may also occur.

The soluble salts brought along by the infiltrated water,

may not only come from the rain, but also from the atmospheric
dust and this is another reason for doubt. In principle, the
rainwater collector units should also collect the atmospheric
dust, but at a sufficient height to avoid the local particle
movement at low level.

Experience shows that the method is good in arid areas,

where the rain concentration due to evaporation is high and
the surface runoff is scarce. The result i s more problematic
in the more humid areas where the infiltration is an important
fraction of the pluviometry and where the surface runoff is
notable and errors in estimation greatly influence the P - E
value. However, in these areas it is possible to apply the
salt balance in order to separate the components of the
hydrogram of a gauging station, if a sufficient chemical
analysis series is available during a rainfall and the later
period, but the author has no direct experience in such cases
( 1 2 ) (15).

6. HYDRODYNAMIC METHODS

The hydrodynamic methods try to determine the infiltration

based on the hydraulic characteristics of the aquifer and the
piezometric surface. The most correct way of making the balance
is using a simulation model, but this is generally a detailed
study phase and requires a notable amount of data (13).
 The methods given herein refer to simple situations within
the study area with a well defined piezometric surface and
with a pattern and slope which scarcely varies throughout the
year, so that an almost stationary situation can be imagined,
with a well differentiated recharge and drainage area. The
method means that the flow of water per unit of transversal
width is equal to the mean recharge upwards. The application
means having the mean transmissivity o f the aquifer in the
analysis area obtained by means o f some pumping tests and
bore-holes and that the piezometric surface has been observed
in a sufficient number o f points to precisely know the mean
gradients. The estimation is a mere application o f Darcy's law.

q T. i.

where q = discharge per unit width

T = transmissivity
i = piezometric gradient

Darcy's law is generally valid in most normal circumstances.

7. EXAMPLES

To illustrate the above, three examples have been chosen,

corresponding to studies in areas of less than 100 Km2, one in
a semi-humid area, another in a semi-dry area and the other in
a sub-desert climate. To better locate the data, the example
has been broken down into multiple paragraphs:

7.1 Montroig Area

Location.- S.W. of Tarragona, in the Baix Camp (fig. 1)

Physiographic characteristics.- Flat coastal strip, 4 km

wide, bordered by the mountain range. The water divide line
is 12 km from the sea ( 1 ) ( 3 ) (5).

Geological characteristics.- Plain of detritic materials

resting on clay formations. Mountain range materials are of
low permeability (1) (11).

Water exploitation.- Traditional use for irrigation. Near

the coast, pumping of 10 m3/year approximately, for supply of
the Vandellos Nuclear Station. New extractions for Tarragona
are ready to start in a short time (3) (5).

Basic problem.- Find out the resources and sea intrusion

process when new well will start pumping.

Existing data.- Scarce, reduced and partial hydro-

meteorological data. Mean pluviometry 400-500 mm in the plain,
higher in the mountain range (1). Almost non-existant hydrological
data; there are no permanent water courses. The existing ones
 85

are short-lived dry creeks. Contributions are estimated be

means of local surveys. Almost non-existant hydrogeological data
prior to the studies for the Vandellós Nuclear Station; later,
values of the transmissivity o f the piezometric gradients in a
reduced area were available. The aquifers drain directly into
the sea ( 1 ) (5) (11).

Hydrometeorological balance.- Of dubious value owing to

incertitude of data and low infiltration (2).

Geohydrochemical balance.- Good application conditions in

non-irrigable land areas. There is no direct data on the
chloride content of the rain water, but this can be obtained
by comparison with similar areas with data (2).

Hydrodynamic balance.- Ideal conditions for application,

but the interpretation of the pumping tests becomes complex (5).

Results.- The mean recharge obtained by each of the three

methods was as follows, in thousands of m 3 per year, per km
of coastline$: ( 2 ) .

a) Hydrometeorological method ....... 600

b) Geohydrochemical method .......... 900
c) Hydrodynamic method .............. 1.100

Method a) foresees a progressive marine intrusion; method

b) foresees a critical situation and method c) a certain
residual flow to the sea, which would stabilize the salt water-
fresh water interfacies in a new position.

Check-ups.- Two lines of piezometers were installed to

control the sea intrusion and three water table elevation
recorders to control the levels were installed. After over
three years exploitation, the interphase movement is more in
accordance with result c) than with the other two. The
hydrometeorological method is excessively pessimistic. The
geohydrochemical method is acceptable in a first approximation,
and the hydrodynamic method is the closest to reality (5).

7.2. Riera de Carme Basin

Location.- S. of the town of Igualada (Barcelona) (fig. 2).

Physiographic characteristics.- It spreads over 100 k m 2

betwen 2 5 0 and 900 m of altitude. The length of the small river
is 25 km.

Geological characteristics.- The materials are clay, silt

and limestone Eocene formations, resting on clay and chalk of
the Keuper formation. The tectonic alteration is important.
There are important travertine and calctuff formations (6).

* In reality, a maximum and a minimum value was calculated.

 86

Water exploitation.- The discharge of the Riera de Carme

is notably regular. The main ground regulating reservoir are
the alveoline limestones, discharging relatively important
springs. The main spring discharges in Capellades, outside the
basin. There is a n intense industrial use.

Basic problem.- Some wells have been chilled from which

100 l/sec., are going to be pumped continuously. Find out
whether it is possible to obtain this discharge in dry seasons
and what type of injuries will be produced to springs and water
courses. The tests have been made i n an extraordinarily humid
season, and it is therefore essential to obtain the probable
situation under other conditions.

Existing data.- Various peripherical pluviometric stations,

and only one interior o n e , at present out of service. Spacial
distribution of the pluviometry relatively regular, around
600 mm/year.

The hydrological data is very scarce. With only one gauging

station operating since 3 years ago. The runoff characteristics
have been reconstructed, based o n a inventory, survey of the
canals, seried gauging and comparison with other,basins. The
normal basic discharge o f the river is 400 l/sec., which should
rise to 500 ì/sec., if the ground discharge to a nearby basin
is considered. The hydrogeological data are almost non-existant,
except a prolonged pumping test lasting for two months, and '

various tests on bore-holes (6). Various springs have been

regularly gauged and the data has been apparently satisfactorily
completed, by means of local surveys on the field and in
factories.

Hydrometeorological balance.- Not very reliable since most

o f the basin has high slopes, with wood or brush, and with
sometimes very permeable materials.

Geohydrochemical balance.-In the main springs area, ground

water has 15 to 24 ppm in Cl-. The scarce rain water samples
show chloride content betwen 5 and 10 ppm. The accuracy is
very low. Direct surface runoff is not known with an adequate
degree o f confidence. Data is only indicative. Possibly the
chemical conditions for hydrogram components separation by
means o f salt balance discharges are optimum, but has not been
made as the influence of the industrial discharges is not very
well known. .

Hydrodynamic balance.- The important variability of

transmissivity conditions, of the main fractured aquifer and
its complex arrangement, make estimations difficult. The best
way is by a study of discharge recession curves in selected
points.

Results.- Estimation of the total infiltration in millions

of m3/year, including the groundwater discharge outside the
basin (6):

a) Hydrometeorological method ........ 10

b) Geohydrochemical method ........... 18
 a7

c) Hydrodynamic method ........ ?

d) Separation of hydrogram
components .................
15

Check-ups.- No direct check-ups are made, but they will be

obtained after completion of the study wiht the 2-month pumping
test, and related observations.

7.3. Famara M,assive

Location.- N. of the Island of Lanzarote, Canary Islands

(fig. -3).

Physiographic characteristics.- Massive of over 600 m. in

altitude, which forms a notable cliff over the W. coastline.
Spreads over 80 km2. Very scanty vegetation, almost sub-desert
climat e.

Geological charact.eristics.- Tahular hsalts, o f more than

1.000 m. thickness, buried cinder cones, very continuous and very
little permeable subhorizontal clay-like levels (almagre].

Water exploitation.- Reserve area of ground waters for

supplying the capital. Collections by means of galleries which
penetrate d e e p into the massive, with horizontal drills and at
present also some vertical ones, to increase the drainage. Discharge
obtained 2 to 3 l/sec., at present temporarily increased to 20
l/sec. ( 4 )

Basic problem.- Get to know the reserves of the massive and

determine the exploitable discharges, their rate and recession
curves, Assess the possible resources.

Hydrometeorological data.- Sufficient rainfall network, except

in the highest areas, where most o f the low infiltration mus-t be
produced. Compiling of data and detailed elaboration by the
Hydrographic Study Centre I n Madrid c"].
Mean rainfallbelow 200 &year.

Hydrological data.- Absence of surface runoff except in

strong storms. There are no direct available.

Hydrogeological data.- Almost non-existent, except in the

galleries where there is a data record and several deep exploration
bore-holes. There are various small springs and oozes, of very
fine or inappreciable discharge, inventoried by the Public Works
Geological Service, and with some chemical analysis.

Hydrometeorological balance.- Of dubious Worth, due to the

highly arid climate and because many suppositions have had to be
made. A daily balance calculates a mean infiltration o f 3 mmLyear.
There is possibly no recharge if the daily rainfall does not exceed
20 to 40 mm., which only occurs a few times in a period of several
years.

By reason of the Scientific Study of Water Resources of the Canary

Islands, made by the General Board of Hydraulic Florks of the
Spanish Government, and UNESCO.
88

Hydrogeochemical balance.- Chloride content of the

infiltration water obtained from the analysis of the small water
oozes on the almagres at high altitude (around 300 to 700 ppm.)
and surface wells in Haría (900 ppm). The water of the galleries
is more salty possibly due to basalt contributions by the high
holding time. There are no direct data on chloride content in the
rainwater, but it can be estimated from the data of the island of
Gran Canaria (4).

Hydrodynamic balance.- Made with precautions from the

freatic surface obtained by a careful study of the data on the
bore-holes and galleries, and the hydraulic characteristics of
the basalts obtained by various procedures (4). The hydraulic
gradients at times exceed 10% per cent in a plane at O elevation.

Results.- The rechar e obtained by each of the three

5
methods in thousands of m /year, are (4):

-
Mean - -
Max. Min.

Famara Heights (19 Km2) .... 225 28 5 190

Famara Lows (28 Km2) .... 140 224 84
Marginal plains (25 Km2) .., 50 75 25
Total (72 Km2) ,. .. 415 584 299

In this case, the best method would appear to be the

geohydrochemical one. There is no direct verification, but
additional information is obtained by means of isotopes and
ambient radioisotopes, apart from a study of the salinity o f the
soil and dust, in elaboration,

Co'plementary,-Since the explotation is mainly of

reserves, it has been computed by hidrodynamic study methods o f
the recession curves of the gallery discharges, that the water
yield should vary between 0,03 and 0,05. This figures, jointly
with the other data allow to estimate exploitable reserves in the
gallery area, betwen 20 to 60 million m 3 , The overall transmissi-
vity is 100 m'/day, with a thickness between 200 and 500 m.

8. CONCLUSIONS

In areas with small infiltration in relation to the

pluviometry, the geohydrochemical method applied with
precautions, is a very useful tool which can improve the
hydrometeorological balance method. In more humid areas, the
results are not so clear. The hydrodynamic balance is the best
method but in some cases it needs appropriate conditions for
application, and in any case, it requires numerous reconnoisance
tests to determine the hidrodynamic characteristics of the aquifer.

9. REFERENCES
 89

1. Custodio, E., Molist J., and Martin Arnaiz, M. (1968).

First Report on the works for supply of the Vandellós
Nuclear Station. Geoteclinics Geokgists Consultants.
Barcelona.
2. Custodio, E. (1969). Report on the present state of
the possibilities of the Montroig collections t o - supply
water to the Vandellós Nuclear Station (internal report).
3. Custodio, E., Bayo, A., and Orti, F., (1971). Geological,
Hydrogeological and geochemical characteristics of the
coastal aquifers between Cambrils and L'Ametlla de Mar
(Tarrag ona)i.- -
Fi rs t Span i s h Portuguese -Ameri can Congr es s
on Economic Geology, Madrid-Lisbon. Section 3. pp 1471170.
4. Custodio, E., and Saénz de Oiza, J., (1972) Geohydrological
study on the Famara Massive, Lanzarote. General Board of
Ilydraulic Works. Las Palmas -Barcelona. 2 U 4 pp.
5, Custodio, E., and others (1973). Final Report on the
works to supply the Vandellós Nuclear Station. Geotechnics
Geologist Consultants, Barcelona (being elaborated).
6. Custodio, E., and others (1973). Geohydrological study of
the Carme Basin. Barcelona. East Pyrenees Water Board
and Public Works Geological Service. Barcelona.
7. Custodio, E. (1973). Hydraulics of water collections.
Section 9 of Subterranean liydrology. Vol, 1.- Omega
Editorial. Uarcelona (at printers).
8.
Sub terruiieanz liydrology , Vol, 2,
(at printers),
-
Custodio, E. (1973). Geo1iydrocliemistry.- Section 10 of
Omega Editorial, Barcelona

9. Martin Arnaiz, M. (1973). Components of the Hydrological

Cycle, Section 6 of SUBterranean Iiydrology, Vol. 1.- Omega
Editorial. Barcelona (at printers).
10. Mero F. (1969). Approach to daily hydro-meteorological water
balance computations for surface and groundwater basins.-
Proceedings ITC - UNESCO Seminar on Integrated Surveys for
River Basin Development.- Delft. pp 89/116.
11. Orti , F. (1970). Notes o n the hydrogeological prospecting
made for supply of the Vandellós Nuclear Station (Tarragona)
Geological Research Institute of the Provincial Delegation.
Vol. XXIV, pp 75/88 Barcelona.
12. Pinder, G.F. and Jones, J.F. (1909) Determination of the
ground water component of peak discharge from the chemistry
of total runoff.- Water Resources Research, Vol 5. N o 2
April 1969 pp 438/445.
13. Public Works Geological Service (1972). Basic Tiieory on
analogical digital models of aquifers. (See especially
chapter 5, Process of construction and use of a model
by E. Custodio and L. Lopez-Garcia), Information and
Studies, Bulletin No. 37 -
Madrid, October 1972. 178 pp.
Public Works Geological Service,
 90

14. VilarÓ, F., Custodio, E., aiid Bruington, A.E., (1970)

Sea Water intrusion and water pollution in the Pirineo
Oriental (Spain). ASCE National Water Kesources Engineering
Meeting, Memphis, Tennesse. Meeting Preprint 112.
15. Visocky, A.P. (1970). Estimating the groundwater con-
tribution to storm runoff by the electrical Conductance
method.- Ground water, Vo. 8. No. 2, March-April 1970.
pp 5/10.
 91

Fig.1 - S i t u d c i ó n del Area d e M o n t r o i g

Locat i on map of Fdontroi g Area
92

Fig. 2 - Sitiiacibri de la Cuenca del Carme

Locat i on of Carme Has i n
 93

Q
Al c g r a n z a
 METHODS OF ANALYSING DEFICIENT DISCHARGE DATA
IN ARID AND SEMI-ARID ZONES FOR THE DESIGN OF SURFACE WATER UTILIZATION

bY
Joseph S. Dalinsky

-
TAHAL Water Planning for Israel Ltd., Tel Aviv, Israel
Technion, Israel Institute of Technology, Haifa, Israel

ABSTRACT
-____
This paper surveys various methods of analysing stream flows:
frequency of annual volumes, discharge-volume relationship with
horizontal, vertical and double hydrograph cutting, and calcula-
tion of the storage volumes available as a function of reservoir
capacity.

Application of these methods, which have been successfully

applied by Tahal-Water Planning for Israel Ltd. over the past '

ten years, can generate data for the design of surface water
utilization schemes when flow records are available for only
a few years.

The understanding and application of the general design

aspects, even if only qualitative, enables the planning engineer
to reduce his basic hydrological requirements to less than 10
years duration.

It is proposed that applied hydrological research be d i

rected towards evaluation o f a number of important hydrologi-
cal design parameters on a regional basis to enable nondimen-
sional curves to b e established.

RESUME

L'auteur examine différentes méthodes pour l'analyse du

débit des riviére: fréquence des volumes annuels-relation entre
ces volumes e t les débits dérivés pour l'utilisation par tronca
t u r e ' d e s hydrogrammes, cette troncature pouvant etre verticale,
horizontale, ou les deux à la fois- calcul des volumes stockés
disponibles en fonction de la capacité du réservoir.

Ces méthodes ont été utilisées avec S U C C ~ Spar Tahal-Wa-

ter Planning, pour Israël Ltd, au cours des dix derni2res années.
Leur application permet de fournir des donnérs pour l'aménage-
ment des eaux, lorsqu'on ne dispose de données di'éculement que
pour u n ' p e t i t nombre d'années.

Une mise en oeuvre intelligenye des aspects généraux d'un

projet, même sous une forme purement qualitative, permet à l'i2
génieur de planification de se contenter, pour les données hy-
drologiques de base, de moins de 10 ans d'observation.

L'auteur propose que la recherche hydrologique appliquée

soit orientée vers l'estimation des parametres hydrologiques -
important à la réalisation des projets. Une telle étude doit
être menée sur une base régionale et déboucher sur l'établisse-
ment d'abaques adimensionnels.
96

INTRODUCTION

The need for water in the arid and semi-arid zones is in most cases
greater than the water resource potential, since there are generally

large areas of good soils that cannot be cultivated as a result of the

scarcity of water for irrigation. Hence, the planning of surface water

utilization is directed toward maximum exploitation of the lfmited re-

sources at reasonable cost.

The annual yield of surface water resources varies considerably as

a result of the extremely non-uniform climatic conditions that prevail in

arid and semi-arid zones. In many cases, the available source cannot by

itself provide an adequate supply and other solution& must be found,

Possible solutioqs are as follows:

(1) To recharge surface water to suitable groundwater aquifers which

would serve as long-term reservoirs. In this case the input

will be the varying annual volumes of surface water, in addition

to the natural replenishment, while the output will be an ap-

proximately constant annual rate.

(2) In the case of supply from surface reservoirs fed from intercep-
tion of stream flows, this source can be integrated with some

other certain or steady but limited source. In this case, in

years of adequate flow from the less reliable or variable

source, the yield of the steady source is retained €or use in

those years in which the yield of the variable source is in-

adequate or non-existent.
 97

In such cases, planning should be based on the average flow which

can be diverted and recharged, or stored in the surface reservoir under

dif fereniconditions.

The techniques proposed in the following are aimed at calculating

these average values as a function of planning parameters such as maximum

diverted flow or maximum net capacity of the surface reservoir.

Acquaintance with streamflow regimes is best acquired by study of

hydrographs. The more detailed the information, the more exact the ùnow-

ledge it provides. Data of hourly or daily flow rates are important,

although detailed time dependence of discharges (instantaneous discharges)

are the subject of greatest interest; where daily discharges do not under-

go rapid changes (e.g. in rivers, springs, and baseflows), monthly data

may be sufficient.

Hydrograph study provides valuable information on the flow regime

and can lead to diversified techniques of analysis.

The following information on streamflows is essential for the plan-

ning of utilization:

- The average volume of annual flows (Uave 1, which represent the

stream water resources potential; the average annual feasible

utilizable flows is a portion of this value.

- The stream's flow regime: Is the stream perennial, intermittent,

or ephemeral? Does it have a significant base flow or only dis-

continuous floods? What is the duration of flow or floods, the

yearly number of floods, and the interval between successive

floods?
98

- Streamflow variability, which comprises variability within a season

or a year and variability from one year to another.

Hydrograph analysis can provide the required information s&h as:

monthly and annual flows and their frequencies; flow-duration curves;

annual average flows in relation to diverted discharges - represented by

horizontal and vertical hydrograph cuts; and annual average flows in rela-

tion to possible reservoir capacities.

In many cases the planning has to be done while insufficient hydro-

logical data are available. The techniques presented in this paper enable

the limited data available to be used efficiently and therefore reduce to

a minimum the period of records needed for planning purposes - very often

to less than 10 years, if the period for which records are available can

be taken as representative of climatic conditions.

A. ANNUAL FLOOD BETURN PERIODS

For practical purposes of surface water utilization planning, annual

flood return periods can be computed by using the established formula:

T E -n + l ... (1)
m
where: T is the return period (in years);

n is the number of annual flow data;

m is the serial number of annual flow data arranged in descending

order, by size.

By using the above formula, return periods approximately equal to

the period for which data are available can be reasonably evaluated. The

estimates of annual flows (order of magnitude) for longer return periode

can be obtained by extrapolation on probability paper by using the points

 99

which were calculated according to formula (1) as plotting points; though,

for practical purposes, rare annual flows are of little importance, if any,

since in most cases, a project based on rare flows will not be economic,

B. FLOW-DURATION CURVES

Flow-duration curves express the average duration of occurring die-

charges equal to or greater than given values (Qi -> Q 1); or dischargea

equal to or smaller than given values (Qi 5 Q21e Schematic representation

of flow-duration curv is given in Sketch 1.

Sketch 1: Flow-Duration Curve - Schematic Representation

The duration can be expressed as the average number of days per year

on which the said discharges occur, or as the total number of days in n ,

years (see illustration of flow-duration curves ‘for the Qishon stream in

Fig. 1 in App. A), or as the relative duration (which is similar to

relative frequency ).

The computed discharges can be hourly or daily averages, or averages

for any period - in accordance with the aims of the analysis and the
nature of the data, In general, average daily discharges will be used

when the daily discharge fluctuations are not appreciable, or wnen the

representative changes are daily changes. For streams characterized by

 100

a flood flow regime - where the flows are of short duration and there is
no significant baseflow - average hourly discharge or averages for even

shorter periods can be chosen. It is customary to express the relative

duration in percentages (p).

The area delimited by Lhe curve Q = f(p), when /Qdp or Z(Q x Ap), is

equal to the average discharge of the stream.

The relations can be easily and economically established when calcu-

lations are made by computer, in many cases as by-product of the computer

analysis of streamflow data. For planning purposes, the direct use of

flow-duration relations or curves is not convenient and their use is limi-

ted to assisting the computation of data needed for drawing curves which
represent the horizontal and vertical hydrograph cuts (as shown in the

following sections of this paper). It should be stressed that: (1) The

flow-duration relations can represent the streamflow character; (2) These

relations can be achieved with satisfactory accuracy in the zone of the

practical importance (the zone where relatively small or medium size dis-

charges occur) using data of a relatively short period (few years, mostly

less than 10 years).

C. AVERAGE ANIUAL FLOW IN RELATION TO MAXIMUM DIVERTED DISC-GE -

HORIZONTAL CUT

When stream diversion i8 considered, whether by gravity flow or

pumping, the dependence of annual diverted flows on the maximum diverted

discharge is computed using the historical data. The curve representing

the dependence of the average annual diverted flows on the maximum diver-

ted discharges can be considered as the stream's "visiting card". The

meaning of the diversion, from the hydraulic aspect, is that all the
 101

discharges which are equal to or smaller than a certain magnitude,,,,),Q(

are being diverted (see Skbtch 2). When the discharge (Q,) exceeds,,,),Q(
the diffeTence [(Q,) - (Qd)max I will overspill, while the diverted dis-

charge, Q,, will be approximately constant, at the magnitude of about

(Qd)maxg

Q
I

Qd i Diversion
discharge

Sketch 2: Schematic Layout of Diverdion Sketch 3: Hydrograph Horiaontal Cut

From a hydrological point of view this means a "horizontal cut"'of

the streamflow hydrographs (see Sketch 3).

The "horizontai cut" can be expressed mathematically as:

where:

is the streamflow discharge;

Qi
is the diverted discharge;
Qd
is the maximum diverted discharge.
(Qä'rnax
 102

For every maximum diverted discharge a certain volume can be diverted

every year; for a period of n years - a series of n annual diverted volmes

can be obtained, out of which the average annual diverted flow (Ud)can be
calculated for each value of (Qd)
max'

The function cd = f (Qdlmx has the shape illustrated schematically in

Sketch 4.
When Qd -+ œ, *U
then where U e represents the stream poten-
d ave' ave
tia1 (average annual flows).

The curve representing the dependence of Üd on (Qd )

max
can be divided

into three m i n zones according to the tangent slopes:

Zone I: AÜd/A(Qd)mx is relatively

large and almost constant. The

I
diversion will be most justified A I
economically - I
H

Zone II:
-
AUd/A(Qd)max quickly de-

creases as (Q ) increases.
d max
This is a transition zone.between 5
v
CI

Zones I and III.

Zone III:
-
AUd/A(Qd),, diminishes

as (Qd)mx increases, and tends

Uave
towards zero when (Q ) * m, In
d max
this zone, diversions are usually

not worthwhile o
-
Sketch 4: Ud - f [(Qd)max]
 103

-
Note:

(i) The above relations can be easily obtained from the computer
at small cost. They can also be calculated without a computer,

in some cases easily (depending on the data).

(ii) If the flow-duration curve is used, expressed by relative

durations (p), and the average nupiber of flow days per year
-
(ta) are given, then üd
- = f [(Qd),,] can be calculated by

the formula:

.. . (3)

where:
-t is the average number of flow days per year;

p is the relative duration (expressed as a fraction) of

discharges equal to or greater than the appropriate Q.

(iii) Althmgh the historical streamflow data will not be repeated

in the future, the calculated diverted volumes represent

fairi,y satisfactorily the amounts and distribution of the

expected volumes for practical purposes of planning.

In planning streamflow utilization by diversion of flows up to a

certain maximum diversion discharge, the above relations, based on a

simple "horizontal cut" of hydrographs, are widely used (see illustration

of horizontal, vertical and double cuts of hydrographs in App. A, Fig. 2).

The maximum diverted discharge is determined according to direct or in-

direct economic considerations (for instance - the limitation of the

diverted discharges in order to minimize the inflow of sedimentary

materials). When limitations exist with regard to small discharges,

this method cannot be applied (see Section D).

1 o4

When water rignts refer to baseflows and/or discharges up to a minimm

diverted discharge, Qo, (when Q > Q only QO is utilized), the computation

-
- >. In this case, new scales will Óe used: Qd* (Qd - Qo) and
(Q0; U
O
- -
can be made by translating the pivot of the axes to a new starting point

:U = Üd - Üo. If the point (Qo; U )lies outside of Zone I, or at its

O

edge - the best flows are already utilized, .even though this does not mean

a priori that the proposed scheme will not De feasible.

D. AVELUGE ANHUAI., FLOW IN RELATION TO DIVERTED DISCURGES - DOUBLE CUT

When there are limitations to the diversion of baseflow discharges,

or any definite discharges less than a certain magnitude, the relation of

average annual flows to diverted discharges cannot be computed by means of

a simple "horizontal cut", In these casesp a "double cut", i.e. a vertical

and horizontal cut, is required (see Sketch 5).

Double Cut Horizontal Cut Vertical Cut

Sketch 5: Schematic Representation of Double Cut

Such a case will arise when baseflow is of undesirable quality for

diversion purposes - generally too saline; conversely, during flood flows

tne water is of good quality and can be diverted up to a certain maximum

value, (Qd)max*
In this case the double cut - shown in Sketch 5-A - ie
 105

used, When from a certain discharge onwards tiie sediment concentration is

undesirable for artificial recharge or from the aspect of reservoir capa-

city losses, and it is decided not to divert those discharges - a vertical

cut for Q = (Qd)mx is used (see Sketch 5-C).

Since during planning it is still unknown which Q and which (Q )

d max
will be selected, different combinations have to be examined. Tiiis can be

done by establishing a series of curves which describe the relations be-

tween annual average flows to maximum diverted discharges for different

values of Q [Qo = constant]. This analysis has the disadvantage of being

related to discontinuous values of Qo and the need for repeating the calcu-

lations for each value of Q (when the storage capacity of a reservoir is

O

involved, many sets of curves'will be required). Such work is superfluous

and can be limited to the calculation of only two curves - the horizontal

cut curve and the vertical cut curve - if the following equation is used:
-UA = -UB - -uc
... (4)
where:
- =
UA fl Qo; (Qd)mxl, represent the double cut';

-UB = f2 ],,),Q( is established by means of a horizontal cut;

o

-U = f [Q ], is established by means of a vertical cut.

c 3 0

It is possible to calculate ÜA for any desired combination of (Qd)max

and Qo by the use of the two curves (see Sketch 5-B and C). Each of the

functions ÜB and Û can easily be calculated by computer. These functions,

C
as calculated for the Qishon stream in Israel, are illustrated in App.&
'pig. 2.
 1Ob

Function ÜC
-
can easily be calculated from U*, without use of a compu-

ter, on the basis of a flow-duration curve, when the duration indicates

the average number of days per year of any given discharge or discharges

exceeding the given value. In this case the relation between the two will

be expressed by:

-
where:
ta and p as in equation (3)

more (see Sketch 3); t* =

-
t* is the average number of days per year of a discharge of Qo or
ta x (P)~,.

E. THE USE OF THE VERTICAZ. CUT

In addition to the contribution of the vertical cut curve for

simplifying the double cut technique and its use for planning of di-

versions with constraints of maximum discharges owing to sedimentation

IQ0 (QdImxi Qd Q for Q 5 (Qd1-i Qd = 0 for Q ’ (Qd)maxla the

curves are used for calculating the average annual sediment concentra-

tion or load.

The average annual sediment load can be calculated using simul-

taneously the vertical cut curve and the curve describing the relation

of the sediment concentrations to flow discharges (mostly log-log rela-

tions). The computation is carried out as demonstrated in Appendix B.

The average annual volume of sediment load, (Us)ave, of the stream

is calculated as:
m
... (6)
 107

Accordingly, the average annual sediment concentration, ave(Cv), is

calculated as:
(Uslave
i-
.. (7)
ave('v) 'ave
where :

(AUVIj (ÜVli
-
- (Uv)i-l,
indicates the contribution of the discharges

-
within the limits of Qi-i to Q,, to the average annual flow;
(Uv )i indicates the average annual flow from the vertical cut
curve [for Q 2 Qil.
-
[(Cv)avel, is the mean volumetric sediment concentration of the dis-
charges within the limits of Qi-i to Q,.

j indicates the intervals of the discharges (AQ), chosen for

the calculation.

It should be noted that since the significant reduction in reservoir

storage capacity resulting from sedimentation in arid and semi-arid zones

is mostly due to rare high rate floods, there is a need for data of a

relatively long period. In such cases, it is therefore recommended to

use probability analysis in the evaluation of the frequencies. However,

regional analysis supported by analysis of historical flood water marks

for rough estimates of the "maximum historical floods" enables relatively

short period data to be used for evaluating the expected average annual

sediment load order of magnitude (in this case - channel sections with
a stable bed should be chosen; otherwise large mistakes may occur as a

result of marked changes in the channel bed).

F. ANNUAL STORABLE FLOW AS A FUNCTION OF RESERVOIR CAPACITY

The quantities of reservoir-stored streamflows which can be

utilized depend on flows, net reservoir capacity, reservoir operation,

and seepage and evaporation losses.

When the reservoir is to be emptied every year (e.g. there is

a rainy season in which the flows are stored and a dry season when the

stored water is used), it is possible to estimate the annual stored

108

quantities according to annual flows and net reservoir capacity, as long

as losses, at least during the rainy season, are small. When the expected

losses are large, the possible losses must be known or estimated before

calculations can be made; however,this information is often not available.

When losses during the rainy season can be disregarded, the reservoir

can every year store quantities smaller than or equal to its net capacity:

iL
UR = -
n
i-1

in which

(URIi = ui when Ui 2 RN
(UR)i = (%li when Ui -'(%)i
Here:

Ui
indicates the annual streamflow in the ith year - when the
reservoir is on the channe1,and annual diverted flow - when

the reservoir is off the channel;

(%)i representing net reservoir capacity in the ith year;

(UR )i is the amount of water stored in the ith year;

- is the n years' average annual amount of water stored in the
UR
reservoir (whose average net capacity is
-
%)
n is the number of annual data (calculated by the use of either

observed, historical, or of reconstructed synthetic data).

The quantity ÜR is always smaller than Uave, when Uave designates

the average possible annual inflows into the reservoir, such as average

diverted streamflow or average streamflow.

 109

Annual net reservoir capacity is defined as freel,,annualcapacity

up to maximum operational height (e.g. up to the spillway crest). When

the reservoir is operated in consideration of a planned dead storage, net

operational capacity is constant until the dead storage is replete with

sediment; i.e. (%Ii =

-
RN = constant. Storage losses are dependent on

two major factors: the volume of annually deposited sediment, and the

volume of water remaining in the reservoir at the end of each year (the

"remaining volume" is generally constant, owing to the reluctance to pump

mud, except in reservoirs which store water for more than one year; the

case of such reservoirs 'is not dealt with in this article).

The following should be noted:

(a) The computations as described give approximate solutions.

If losses (by seepage and/or evaporation) are relatively

large, at least monthly water balances are required in order

to calculate the stored inflows with reasonable accuracy.

(b) The amount,,ofwater which can be annually utilized also depends

in each case on the operational regime of the reservoir and on

the losses (there is a difference between the utilized and the

stored amounts, since losses occur while the stored water is

being utilized).

(c) In arid and semi-arid zones - in many cases, due to the limited
potential of the stream and the considerable seepage and evapora-

tion losses - surface water utilization is based on artificial

recharge of aquifers. In such cases the reservoirs are used

for regulating and silting purposes; therefore (UR)i can exceed

the net reservoir storage capacity, as it is a product of a

 number of floods which entered the reservoir after it was

emptied or partly emptied (after every flood the stored water

is transferred to spreading grounds for artificial recharge).

Principally, the calculations, the character, and the analysis

of the relations between FR and -

RN are the same as dealt with

in this Section.

The average annual volume of stored water (uR)and the reservoir

efficiency (ÜR/uaVe) as functions of average net reservoir capacity (i$)
are shown schematically in Sketch 6. (The meaning of the different zones
is as explained in Section C for Sketch 4)
d

Zond Zonelzone (UR/Uave>

A Zonalzone ZoneI

2-

Sketch 6:
Rd -
Schematic Representation of U R = f (s)and (ÜR/UaVe) = F (s)
The curves, which summarize the aforementioned influences, illus-

trate the contribution of average net reservoir capacity (Q). Here too,

as in the analysis of the relations illustrated in Sketch 4, different

zones of the curves can be discerned, characterized by the magnitude of

the slopes of the tangents to the curves (AÜR/A% or A(ÛR/Uave)/A$).

These slopes represent the marginal additions of the average annual

quantity of storable water for the addition of a unit of net reservoir

capacity .
 111

It is characteristic that as
-
% increases, AÜR/AQ decreases. For
high values of RN the value AÜR/A$ is small, as it represents rare flood

flows; its reliability is therefore limited.

An analysis of this kind is of great importance for preliminary

estimates and/or feasibility calculations, since it makes it possible

to find easily the approximate economic solution.

Recent investigations made by the Surface Water Utilization Depart-

ment of Tahal - Water Planning for Israel Ltd. prove that the relationship
- -
between UR and % can be approximately estimated on a regional basis using

as a parameter the dimensionless standard deviation (the ratio u /U

u ave’
where U is the standard deviation, and U is the average annual flow).
U ave
It was found that the ratio ouIU is, in many cases, of regional
ave
character. (The above-mentioned investigations have not yet been con-

cluded and hence cannot yet be summarized).

The function described above is illustrated in App. A, Fig. 3.

G. THE COMPUTATION OF RN AND RN

Annual net reservoir capacity can be computed from the equation:

(%li = - ... (9)

where :

($>i is the net reservoir capacity at the end of the

ith year;

(%)i-1 is the net reservoir capacity at the end of the

(i-i) th year;

(Rs)i is the volume of the sediment trapped in the

reservoir during the ith year.

112

The volume of the sediment deposits trapped in the reservoir during

a certain year can be calculated from the equation:

(CS)i
(RsIi = (EVIix Uix(T.E.Ii = - YS
x Ui x (T.Ea)i ... (10)
where :

is the average concentration of the transported sediment

during the ith year, by volume;

(Cs)i is the average concentration of the transported sediment

th
during the i year, by weight (e.g. in p.p.m);

is the average specific weight of the trapped sediment

YS
(generally approximately constant, depending on sediment

qualities and reservoir operation);

represents the reservoir inflow in the ith year;

ui

(T.E.Ii is the trap efficiency in the ith year - the portion of the
sediment which remains in the reservoir (if there is any

overspill, part of the sediment leaves the reservoir with

the overspill).

For a design period of n years, especially when the value of n is

high (tens of Years), the total loss in reservoir storage capacity re-

sulting from sedimentation can be calculated as:

Therefore, the average annual loss in reservoir storage capacity

will be

-
Rs = - (Rs)i = Uave x (T.E.1 x [ave(Cv)l ."a (11)
n
i=l
 11 3

where :

is the average concentration, by volume,of the sediment

ave('v)
deposited by the transported water at the reservoir

location;

(T.E.) ,the average trap efficiency.

The average net reservoir capacity for a period of n years will be

estimated as:

... (12)

where :

is the initial reservoir capacity.

Ro

It should be noted that since it is impossible to predict the future

annual flows, there is no other practical possibility of evaluating the

net reservoir storage capacity. For practical purposes, the use of aver-

age net storage capacity (%)

is sufficient.

H. RECOMMENDED HYDROLOGICAL INVESTIGATIONS

Hydrological investigations directed towards finding parameters

which enable non-dimensional curves to be established which represent

the main functions discussed in this article, are recommended, especially

on a regional basis.

The reconstruction of such a regional synthetic curve, even though

not "scientifically accurate", will be of great assistance in planning

surface water utilization schemes, and especially in planning the first

stage of such schemes.

 114

BIBLIOGRAPHY

This article is based on the experience gained in Tahal - Water

Planning for Israel Ltd., in the last 20 years, &.on the Technical
Reports published by Tahal in Hebr’ew, as also on the foliowing works.

1. Kuiper, E., Water Resources Development, Buttexworths,

London, 1965

2. Linsley, Ray K. and J. B. Franzini, Water Resources

Engineering, McGraw-Hill Book Co., London, 1964

3. Searcy, J. K., Flow-Duration Curves, Manual of Hydrology,

Geological Survey Water Supply Paper 1542-A, Washington D.C.,
1959
 li 5

APPENDIX 13: CALCULATION OF AVERAGE ANNUAL SEDIMENT. VOLUME

TRANSPORTED BY LOWER QISHON FLOWS

1.
-
Streamflow hydrographs were used for preparing a %cirtical cut curve"
representing the average annual values of flow (Uc) contributed by
discharges up to any value of QI as explained in Sections D and E,
and illustrated in Fig. 2 of App. A.

2. Simultaneous data of sediment concentrations and instantaneous dis-

charges, supplied by the Israel Jydrological Service, drawn on a
log-log paper enable the construction of a correlation line between
the average sediment concentration and the instantaneous discharges.
In order to be on the safe side, the line was removed toward the
higher concentrations (of each discharge) - see Fig. 4 of App. A.
3.. The calculations are shown in detail in the following table.
 11 6.

APP.
Sneet 2

CALCULATION OF AVERAGE ANNUAL SEDIMENT LOAD

EXAMPLE: LOWER QISHON STREAM (ISRAEL)

-uc -cs
L CU. mfs ec

5
MCMIY r

O
3,O
PPm

300
5ave x AÜc
tonlyear

900
1 3.0 400
3.0 600 1 ,800
3 6.0 800
2. o 1 ,O00 2,000
5 8. O 1 ,200
2.4 1 ,450 3 ,480
10 10.4 1,700
1.0 2,000 2,000

1 --
15

.20
Total
LEGElID:
II..

12.0
13. O
O. 6

1.0
13.0
2,300

2,700

-
2,500

4,000
1,500

4,000

15,680

-U2Q == diccnarge
average annual flow volume related to Q calculated by
vertical cut of hydrographc (from Fig. 2 of App. A)
I -
AUc = the interval of Uc contributed by discharge interval
-Cs = average sediment concentration, by weight (from Fig. 4
-
(Cs)ave =
-
of App. A) , high values
average CS for discharge interval

4. The result obtained from the calculations shown in the above table,
is that average annual sediment load transported by the Qishon stream-
ilows amounts to about 16,000 ton. Assuming an average trap efficiency
of 90 percent and sediment deposits specific weight of 1.5 ton per cu.m -
the average annual value of sediment trapped and deposited in the planned
reservoir will be about 10,000 cu.m per year (16~000x0'9 9,600
1.5
i0,OOO cu.m per year).
 117

Appendìx A

---Ueriium
Rainy yerre - average:
reinfall years
t* > 40 daye
- averaec:
I
20 1 t* 40 daye
Dry year# - average: t* < 20 daye
.,...u. Average for 1940141 to 1964165

t* - dischargee exceeding Q
Number OP daye with discharges cx-
ceeding 1.5 m3Jeec

-
FIG. 1: FLOW-DURATION CURVES FOR QISHON
STREAM (ISUAEL)

FIG. 2: HORIZûNTAL, VERTICAL AND DOUBLE

CUTS OF HYDROGRAPHS OF QISHON
STREAM (ISRAEL)
118

Amendix A

FLG. 3: T H E DEPENDENCE OF T H E AVERAGE STORABLE

FLOWS AND T H E STORAGE EFFICIENCY ON
T H E NET AVERAGE RESERVOIR CAPACITY A T
T H E UPPER QISHON (UPSTREAM T H E HYDRO-
METRIC STATION T O WHICH T H E DATA OF
FIG. 1 AND 2 REFER), I S M L
 11 9
Appendix A
 APPLICATION OF COUTAGNE'S A N D T U R C F O R M U L A S
TO THE SOUTHERN MOZAMBIQUE RIVERS

E m i l i o E u g é n i o D'Oliveira M e r t e n s
J o ä o José M i m o s o L o u r e i r o

ABSTRACT

C h e c k i n g o f Coutagne's and T u r c f o r m u l a s , w a s p u r p o s e d t o
obtain values, though approximated, for t h e annual mean runoff
o f t h e s e v e r a l r i v e r s at s o u t h e r n M o z a m b i q u e w h e r e f e w g a u g i n g
s t a t i o n s exist. T h e r e f o r e , m e a s u r e d r a i n f a l l and t e m p e r a t u r e
v a l u e s w e r e c o l l e c t e d f r o m t h e m e t e o r o l o g i c a l and g a u g i n g stations,
a s w e l l a s t h e r u n o f f values o b s e r v e d i n t h e l o c a t i o n s .
We c o n c l u d e f r o m t h e r e s u l t s o b t a i n e d t h a t t h e a p p l i c a t i o n
o f t h e s e r u l e s h a s g i v e n us, w i t h r e l a t i v e g u a r a n t e e , t h e a n n u a l
m e a n r u n o f f v a l u e s , w i t h d e v i a t i o n s i n f e r i o r t o 10% w h i c h c a n b e
considered as satisfactory.

RESUME

Les f o r m u l e s de C o u t a g n e et T u r c o n t é t é u t i l i s é e s p o u r
o b t e n i r d e s valeurs, même a p p r o x i m a t i v e s , d e l'écoulement m o y e n
a n n u e l p o u r l e s d i f f é r e n t s f l e u v e s d e l a r é g i o n sud de M o z a m b i q u e
d a n s l a q u e l l e on n e d i s p o s e q u e d'un n o m b r e t r è s l i m i t é d e
stations de jaugeage.
Les c a l c u l s o n t S t 6 e f f e c t u é s à p a r t i r d e s v a l e u r s d e s
p r é c i p i t a t i o n s et d e s t e m p é r a t u r e s m e s u r é e s a u x s t a t i o n s m é t é o r o -
l o g i q u e s et p l u v i o m é t r i q u e s , a i n s i q u e d e s v a l e u r s d e s é c o u l e m e n t s
observées à différentes stations.
L e s r é s u l t a t s o b t e n u s m o n t r e n t q u e l'application d e c e s
d e u x f o r m u l e s donne, avec u n e p r é c i s i o n r e l a t i v e , d e s v a l e u r s d e
l'écoulement m o y e n annuel. Les é c a r t s sont i n f é r i e u r s à lo%, c e
qui peut être considéré comme satisfaisant.
 The hidrological phenomenons o- greater interest, relating to the hidro-
logical studies of a catchment area under consideration, are namely:-

Rainfall
Air temperature
Relative humidity
Evaporation
Hidrometical records
Flow discharges of streams and runoff
Sediment discharges

The main purpose of a certain hidrological study, consists on the deter-

mination for each one of the observed actions, of the variability principles
thereof at distinguished intervals, analogy principles of the phenomenon itself
from site to site and of the correlation principles amongst the several pheno-
menons.
One of the basilar elements necessary for the planning of an economical
development program is the knowledge of the value and distribution of its hidrg
logical resources.
In Mozambique, registration of the hydric resources has been facing
great difficulties not only in what refers to the extension of the territory but
also, and essentially, by lack of observations of the hidrological phenomenons,
namely the runoff and flow discharges of rivers and water-sources.
From a report presented by Dr. L. Turc on the 3rd.iiidrological Ehgeneer-
ing Congress organized by the 'Societé iiidrologique de France', which took
place in Argel, in 1954, we were suggested to follow the idea of verifying the
possibility in the application of Coutagne's and Rirc general rules, related to
the Southern Mozambique water-sources.

2 - COUTAGNE'S AND TURC GENERAL RULES

These general rules allow us to estimate, by simple calculation, the
value of a catchment area's runoff deficit, provided that rainfall and tempe-
rature are known.
 123

/2.

Runoff deficit - is the difference between mean rainfall height

pertinent to a certain site in the water-source and the corresponding height
to the flow discharge estimated at the referred site.

2.1 - COUTAGNE'S GENERAL RULE

Being :
H = Mean rainfall height
E = Hficient rainfall height, that is, the height which transform
itself theoretically, in the whole, to runoff.
D = Runoff deficit = H -E
C = Runoff coefficient =2
H
K = Coutape's constant

general rule is

D =H - KH2 being E = KH2

and since C = E and D=H-E
H

C,D-H or C = KH
H

Now as:

(C - KH)2 = (c 1 - KH + (c 2 - KH 2)2 + (c 3 - KH 3) 2 + ....**

to minimize this sum, it will do equalizing zero to the first derivative:

(C 1 - KH 1) H 1 + (C 2 - KH 2) H 2 + .......= O
wherefore:

CH-KH2 = O K = -H2
CH

From the above determination it is given the most probable value for K.
124

/3.

2.2 - TURC'S GENERAL RULE

L = Turc's constant
P = Evaporation plus lost by percolation
T =Mean temperature
A = Constant

H
'dFtJ2
L2

being
L = A + 25 T + 0,05 T 3
wherefore

The author still precises that applying his general rule in 254
Catchment areas, considering A = 300, distributed towards every climate in the
world, it has been r e c k m e d that values of 0 observed and calculated from the
referred general rule, came out as to the undermentioned results:-

in 53% the cal. D - me86 D < 40 mni

or
in 43% the cal. D - meas. D < 0,l meas. D
or rather

i,n 65% the cal. D - meas.D < 0,2 meas. D

 125

/4.

3 - REPORT ûF THE CONSIDERED LOCATIONS

Described hereyder are the considered locations at the Limpopo's

(incl. Elephant's River), Incomati, Umbeluzi, Sabie and Usuto Rivers (D. 1)

3.1 - ELEPHANT'S RIVER

Location: Maçuço - Mozambique

2
Catchment area: 66.600 Km
Mean rainfall height: 636 mm
Mean temperature: 18,9OC.
Observation years : 1944/45 to 1970/71.

3.2 - LIMPOPO'S RIVER

Location:Beit bridge -
R.A.S.
2
Catchment area: i88.000 Km
Mean rainfall height: 481 mm
Mean temperature : 2OoC.
Observation years : 1955/56 to 1963/64

3.3 - LIMPOPO'S AND ELEPHANTS RIVERS

Location: Vila Trigo de Morais - Mozambique

2
Catchment area: 340.000 Km
Mean rainfall height: 541 mm
Mean temperature: 20,2OC
Observation years: 1951/52 to 1969/70

3.4 - INCOMATI'S RIVER

Location: Ressano Garcia -2 Mozambique

Catchment area: 21.600 Km
Mean rainfall height: 832,2 mm
Mean temperature: 18,8OC
Observation years : 1955/56 to 1969/70

./.
 126

3.5 - SABIfl'S RIVER

Location: Machatuine - Mozambique
Catchment area: 6.200 Km2
Mean rainfall height: 766,4 mm
Mean temperature: 20,7OC
Observation years: 1955/56 to 1969/70

3.6 - UMBELUZI s RIVER

1

Location: Goba - Mozambique

Catchment area: 3.100 Km 2
Mean rainfall height: 820,3 mm
Mean temperature: 21,6OC
Observation years : 1955/56 to 1970/7i

3.7 - MAPUTO'S RIVER

Location: Sipofaneni - Swaziland

2
Catchment area: 12.903 K m
Mean rainfall height: 83i,7 mm.
Mean temperature : 22OC
Observation years : 1958/59 to 1964/65

3.8 - Therefore we get two distinguished groups in regarding to pluviosity

and temperature:-
- Limpopofs River Groue - with mean rainfalls between 450 and 650 mm
and temperatures from 18OC to 2OoC;
- Incomati's, Sabie, Umbeluziaid Usuto Group with mean rainfall
values of 800 m. and temperatures higher than 2OoC.

4 - APPLICATION OF COUTAGW'S GENERAL RULE

We have tried Coutagne's general rule for each one of the above
groups and locations therein.
 127

4.1 - LIMPOPO'S RIVER CATCHMENT AREA

2
The 340.000 Km of the Limpopo's River Catchment Area relating to
2
Vila Trigo de Morais' gauging station, include the 66.600 Km of Maçuço's
2
gauging station at Elephants' River and the 188.000 Km pertinent to Beit
Bridge location.

4.1.1 - For
the 27 observation years at Elephants' River we have reached to
the following type of Coutagne's rule:-

D =H - 0,000055 H (1)

The most probable values for the measured runoff deficits (612,9 mm)
and calculated ones (613,3 mm) differ in 4 mm to a mean deviation of _+ 8,7 mm
and a mean observation error of 7,4 mm.
Extension of this rule for the available 66 rainfall observation
years would be plainly acceptable in view of the fact that for a period of 34
years of which we own the closest possible runoff estimatives, difference is
kept for the measured and calculated deficit.

4.1.2 - For the 9 observation years in Limpopo's River area at Beit Bridge we
reached to the results hereunder, to Coutagne's rule:

D = H - OJ00O031 H2 (II)

recording the most probable values of the measured runoff deficits (472,7 mm)
and calculated ones (473,7 mm) being the mean deviation and each observation
error of _+ 6,2 mm and 3,7 mm respectively.

4.1.3 - Finally for the i9 observation years in Vila Trigo de Morais, situated
after the confluence with Limpopo's and Elephants Rivers, Coutagne's rule
presents us the following result:-

D =H - 0,000047 H2 (III)

Measured and calculated runoff deficits have a similar probablest

value, but the mean deviation is increased in ,+ 10,l mm and mean observation
error amounts to f 7,6 mm.

./.
 128

/7.
4.1.4 - It is left to determinate now an available Coutagne's general rule
to the entire group of 55 observations, as the principal elements taken to
its calculation - -
runoff coefficient (C) and mean rainfall (H) are not
dependent values on those of the referring catchment areas, same being consi-
dered to the runoff deficit.
Ordering the values of the observed mean rainfall, the undermentioned
rule is calculated (Q1):-

D =H - 0,000050 H2 (IV)

Through the same comparative system, it was obtained to the measured

runoff deficits, the value of 560,3 mm and for the calculated ones through
the same rule (IV) 56O,6 mm being 0,3 mm the difference thereof.
Mean deviation of the measured and calculated values amounts to
-+ 9,l mm with a mean observation error for each one of 7,2 mm.
Seing that the values of medium rainfalls, relating to the observa-
tion periods, are respectively of 636, 481 and 541 mm with a short difference
from the medium normal rainfall, we may conclude that Coutagne's rule of
which coefficient is equal to 0,000050, can be applied to every catchment
area of which medium rainfall is comprehended between 450 and 650 mm. However
it is necessary to point out that its application, in view of the great ex-
tensions that they comprehend, cannot be considered as absolutely precise,
except for mean values. Application of this rule every year may lead us to
mistake, since it calculates a regular correlation amongst runoff and rainfall
which is not precised in the practice because of the powerful stream of Lim-
popo's River, namely before the confluence with Elephants' River.

4.2 - CATCHMENT AREA'S GROUP OF INCOMATIIS, SABIE, UMBELUZI AND USUTO RIVERS

2
The 43,803 Km of this group comprehend all the rivers which drain
off in Lourenqo Marques' Bay and are located in an area, mean altitudes of
which excede the 800 m. and mean rainfall is estimated between 750 mm and
850 mm.
All the above catchment areas are neighbouring.

4.2.1 - For the 15 observation years of the Incomati's River at Ressano

Garcia rainfall station, we conclude from Coutagne's general rule the next:
D = H 0,000150 H2 (V)
-
./.
 129

The mean deviation value amounts to 19,7 mm and the mean observation
error to i5 mm for the measured and calculated runoff deficits of 737,3 and
739,O mm respectively.

4.2.2 - For the Machatuinels rainfall station of Sabie's River and Goba's
rainfall station of Umbeluzi's River, respectively with 15 and 16 observation
years pertinent to an identical period, Coutagne's general rule figures like:

D =H - 0,000131 H22 (VI)

D =H - 0,000145 H
To the first location, measured and calculated runoff deficits are
similar (684,7 mm) and mean deviation amounts to _+ 20,4 mm.
At Umbeluzi's River, mean deviation amounts to the decreasing value of
-+ 181 mm and difference between the measured and calculated deficits amounts
to 718,s and 719,O.

4.3.3 - For the 7 observation years at the Sipofaneni's rainfall station in

Swaziland, the main confluent of Maputo's River, Coutagne's general rule is
as follows:-
D =H - 0,000162 (VI1)
Measured values (717,3 mm) and calculated ones (717,O) differ from
O,3 mm and mean deviation amounts to &17,3.

-
4.3.4 Similary to what has been done in Limpopo, it was arranged the group
of 53 observation years (Q2) in regard to mean rainfall, which alters from
500 mm to 1.540 mm and the exposed Coukagne's rule is:-

D =H - 0,000140 H2 (VIII)

Difference from the results determined by measuring and obtained from

this ru1.e (VIII) is of 1,2 mm with a mean deviation of f 19,3 mm and _+ 15,8 mm
for the medium error of each observation.
To this catchment area's group mean rainfall wherein exceeding 800 mm
and mean small deviations qualifying same as of minor torrentiality, and,
consequently, higher stream regularity, application of Coutagne's general rule
more than granting us accurate values for the annual runoff deficits yet allow

./.
130
/9.

us its application year after year.

5. - APPLICATION OF TURC'S GENERAL RUlE

Application of this rule, such as formed by Turc, that is, considering

A=300, could not be used but running the risk of forming gross estimate errors
seing that in Mozambique, the catchment areas in study have great extensions,
usually.

5.1 - We have tried to both of the groups the application of the general rule
formed by Turc, and have found the next following results:-

5.1.1 - IJMPOW'S CATCHMENT AREA

Difference calc. D - meas. D L 20 mm - 53%

II II L 40 mm - 76%
II II A 40 ~UW - 24%
II II L 0,Ol m.D - 14%
II 11
< 0,05 m.D - 22%
II II L 0,l m. D - 36%
II II
1 0,l m. D - 64%
II II Ls 0,2 m. D - O

5.1.2 - INCOMATI'S, SABIE, UMBELUZI AND USUTO GROUP-(Q2)

Difference calc. D - meas. D c 20 m - 3%

11 Il < 40mm - 66%
II II > 40 ìüiìì - 34%
11 II < 0,Ol m.D - 15%
II II < 0,05 m.D - 50%
11 II < 0,l m.D - 84%
II 11 > 0,l m.D - 16%
11 Il > 0,2 m.D - 10%

5.2 - Percentages differ from the obtained values by Turc for his Group of
254 catchment areas and considering the 64% (Limpopols Group) of events
superior to 0,l of measured D, we are not abled to consider the rule as
applicable.
 131
/lo.

Therefore it was tried to find a rectifying solution of the constants

in function of the mean rainfall and annual mean temperature for every catch-
ment area. Thus we have traced the graphic shown in D.2, consequence of
succeeding considerations on the measured values.

5.3 - Upon the above application of Turc's general rule, established the
constant A from the graphic, we reached to the following results:

5.3.1 - LIMPOPO'S CATCHMENT AREA (Q3)

Difference calc. D - meas. D < 20 mm 88%

II 11
< 40 mm 95%
II II 40mm 5%
11 II d 0,Ol m.D 33%
II II < 0,05 m.D io%
11 II < 0,l m. D 10%
II II > o,1 m. D O
11 11 > 0,2 m. D O

5.3.2 - INCOMATI'S, SABIE, UMBELUZI AND USUTO CATCHMENT AREAS (a)

Difference calc. D - meas. D < 20 mm - 43%
II II < 40 iiìüi - 87%
II II > 40 ïìüü - 13%
II 11 ¿ 0,Ol m.D - 25%
II II < 0,05 m.D - 8%
II 11 < 0,l m.D - 96%
Il 11
> 0,l m.D - 4%
11 II > 0,2 m.D - O

5.4 - From the whole of the 108 compared values,

we conclude that in 87% of
the cases the difference among the calculated and observed values does not
exceed 40 mm and in 96% the difference does not exceed 0,l from the measured
runoff deficit. Nevertheless, the most relevant results are that in 2% of the
cases the deviation does not exceed 0,Ol meas.D and 53% does not amount to
O,O5 of the observed runoff deficit.

./*
132
/li.

6 - CONCLUSIONS
In the water-sources situated at the South of Save's River, there
might be applied the Coutagne's and Turc general rules on the following way:

COüTAGNE'S GENERAL RULE:

Catchment areas with annual medium rainfalls coaiprehend between 450

and 700 nun.

D =H - 0,000050 H2
for mean rainfall superior to 700 nun.

D =H - o,000140 H2
TURC'S GENERAL RULE

Determining constant A from the graphic

Utility in the application of these rules becomes evident in view

of the non-existence of gauging stations along the multiple water-sources in
the area under consideration, meteorologic and rainfall stations taking their
place instead.
But, seing that for the inventorying of the hidrological resources
is matter of extreme necessity the knowledge, though approximated, of the
annual mean rainfall, and yet because it has become evident through the appli-
cation of the aforesaid rules that precise values amount to less than i%, we
may say that have succeed with the reaching of our main purposes.
133
D1
 ln
U
L

850 C
-.mU
- W
u
O
P
a
800 .E-
-I
VI
O
.-
a

750

A (TURC2

700 35 O

650
400

\
\
\
\
6 O0 \
\
\ '
\ i
\ C5 O
v

55 o

500

500

A B A C O P A R A A DEIERMINAFAO
DA C O N 5 f A N T E DE T U R C
450
D E T E R M I N A Ç ~ ODO C O E F I C I E N T E DE C O U T A G N E
Rios Limpopo e Elefantes
 1
DEIERMIN4ÇÁO DO COEFICIENTE DE COUTPGNE
Rios Limpopo e Elefantes
 2
DETERMINPÇÃO DO COEFICIENTE O € COüTAGNE
Rios Incomati, Sabié, ümùeltízi e Maputo

OEFIC E N I :EPICI1 d r E S C O A U E N 1
-
- A
e E SCGAIAI HZ C H lî*

-__ C - OC A L C U L A OM- o
-
Qt
u 259d81 - -. 25,99 452 -21 441
-Qm 392.599 @+5Q - 502 - -6 - 36
AQ6 3 1 W P - .
. XL§Q__ .I29 +-L3 269
Q,lQ 93L776 5Z,60 513 -12 144
All- 34kUlQO- - .- - 6430 524 -
-18 324
Q,05 3 á LO M .. in,oo - 16i) i18 . 324
n,u- 381924 67.98 547 -18 324
9,08 384.400 .- 49,6Q_ 571 + 4 - IL
4.05 419,881 32.01 610 i26 -636
-0,l3 412.164 83,46 560 -24 57t
0,16 448.900 107.20 562 -46 211C
0,11 A 9 8 329 74,47 603 -10 -100
0,08 462.40Q 1>4,40 622 + 6 36
0,lQ 463,761 48,LO 6Q7 -1O 101
0,07 467,856 . 47.88 630 +19 36 1
0,08 485,809 55 I 76 640 +LI 171
Li1 504.100 78.10 634 -6 36
0.10 132.9QO 73,OO 655 -1 1
o,11 547.600 . 81.40 662 - 2 4
0.11 549.061 81.51 656 -9 bl
O ,O8 599-076 - 61,92 708 i17 289
O, 14 605.284 ~ . 108,92 567 -2 7 729
0,13 606 .a41 -~ 101.27 676 -19 36i
0.09 624.00 71110 713 +10 100
0,OI 628-849 55,SI 737 + 32 1074
O, 08 649.636 64,4t? 739 +2 3 520
o ,oa 622.400 65,60 749 *23 529
0,13 685 584 . 107,64 721 -12 144
0,15 687.241 124,35 704 -29 841
Od3 714.025 . 109.85 738 -8 6 1"

o,09. 734.449 77,13 780 +?5 625

0.11 736 164 94.38 762 + 7 49
0,12 17O.8134 105 ,ih 777 + 2 4
O ,O9 774.40Q 79120 799 +2 7 729
0,13 784.996 115,lß 771 - 6 36
Q,16
0,12
792.100 -
792.100
142.40
106,80
744
785
-36
+ 5
1296
2c>
0,11 793.881 9h,01 792 +12 __144
0,l.l 195.664 96,l' 797 +11 121
0,12 802.816 107,~J 79 1 + 7 49
OJ 5 846.400 -. 138.00 780 -22 484
-e*-
-- 81.9_41
.
-
1 7 c , 7 7

-
813
-- +
.41
-
-
6
- -
38
2
DETERMIN4C$O DO COEFICIENTE D E COUT4GNE
 3
Q U A D R O COMPARATIVO DOS RESULTADOS OBTIDOS
PELA FORMULA DE TURC E DA CONSTANTE TI,
RADA DOÁBACO

ZONA Bios Limpopo e Elefantes mor

140

4
Q U A D R O COMPARATIVO D O S RESULTADOS OBTIDOS
PELA FORMULA D E T U R C E DACONSTANTE TI-
RADA DO ÁBACO

ZONA Incomati, Sabi6, Umùelilzi e Naputo

 MAPA1 H I DROLOGI C A L S T U D Y (L I M P O P O 'S RI V E R )

E M I L I O E U G E N I O D'OLIVEIRA MERTENS

J O Ã O JOSE MIMOSO LOUREIRO

ABSTRACT

Lack o f o b s e r v a t i o n s in t h e f l o w d i s c h a r g e s a n d r u n o f f ,
t a k e n at the f u t u r e l o c a t i o n of mapai's d a m , h a v e c o m p e l l e d
us t o the e s s a y i n g o f d i v e r s e d m e t h o d o l o g y v i e w i n g i t s
obtention.
It w a s s e l e c t e d t h e m e t h o d o f t h e s p e c i f i c a 1 r u n o f f
t e c h n i c w h i c h h a s c o n d u c t e d us t o most c o n s i s t a n t and
s i g n i f i c a n t r e s u l t s in c o n j u n c t i o n w i t h t h o s e o b s e r v e d a n d
c a l c u l a t e d f o r o t h e r l o c a t i o n s at t h e c a t c h m e n t area.

RESUME

L'abscence des o b s e r v a t i o n s r e l a t i v e s a u x d é b i t s et
écoulements measurables au futur lieu du barrage du mapai,
n o u s a f o r c é d'essayer d i v e r s e m é t h o d o l o g i e p o u r e n o b t e n i r .
I 1 a ;te c h o i s i e l a method: de l a t e c h n i q u e d é s d é b i t s
specifiques que nous a conduit a des resultats tres concor-
dants et significatives en conjunction avec ceux observés
e t c a l c u l é s p o u r les a u t r e s l i e u d u b a s s i n versant.
142

1 - CATCHMENT AREA
1.1 - Site, area, relief and hydrography:

The hidrographic basin of the Limpopo River has its major part in the
territories of South Africa, Rhodesia and Botswana, its area of 412 O00 km2
being devided in the following manner (Drawing 1):

South African Republic .................... 193 500 km2

Rhodesia .................................. 66 O00 km2
Botswana .................................. 73 O00 km2
Mozambique ................................ 79 500 km2

Rounded off, the catchment area is situated beyween 220 and 260
South and 269 and 350 East, its highest altitude being 2.300 metres near the
city of Lydenburg.
In National territory, situated between parallels 210 and 250 South and
meridians 310 and 359 East, the basin has to the North, that of the River Save,
to the South, that of the River Incomati and to the East, that of the River
Govuro, and a coastal strip where a few closed catchment areas are found from
which the water-sources accumulate in lakes.
In Mozambique there is no noticeable irregularity, this occurring only
in the limiting zone to the south of the Limpopo, in a reduced area with eleva-
tions of 400 metres.
In its total length the average height is of 840 metres, its being 977,
964 and 950 metres, respectively in Beitbridge, Mapai and Trigo de Morais.
The average slopes of the course of the water are:

Upper stream ....................................... 2, 50 d

lan
Central stream ..................................... 1,80 m/km
Lower stream ....................................... 0,Og m/km

The Limpopo is one of the most important rivers of South Africa and
Mozambique and, as happens with the Incomati River, it is contained in the
lower part of the great drainage area, which includes more than half of the
Transvaal and a considerable part of South Rhodesia.

./.
 143

The Limpopo is a strange river, very changeable and capricious, perhaps

due to the influence of the dissimilarity of its hydrographical basin; its vol5
me of water is extremely variable as in dry weather it is very reduced and during
the rainy season, reaches heights of 7 metres which flood large areas of ground
in the central and lower courses. The Limpopo River, when it enters our terri-
tory, has already a definite bed, where it has three large tributaires: on the
right bank, the Elephants River, and on the left bank, the Nuanetzi and the
Changane; it is to these that it owes its permanent volume of water for the
flow from those joining it, its principal supplier being the Elephants River, a
water-source which crosses a region of high rains, its hydrographic basin having
a somewhat impermeable geological configuration.
It belongs to the hydrographical system of the African Continent and it
is of the torrential rate of permanent volume.
The course of the water which takes the name of Limpopo River, is formed
by the junction of the Marico and Crocodile Rivers which have their sources at
an altitude of 1.500 metres to the west of the city of Pretoria.
The principal tributaries of the right bank, all with their sources in
the Transvaal, from the source to the mouth of the Limpopo River are as follow:
River Matablas, Pongola, Palala, Sand, Pafuri (flowing in close to Pafuri,
already in Portuguese territory) and the Elephants River, the largest and most
important which joins it within Mozambique after some 110 kms. On its left bank,
the Limpopo receives large courses of water all with their sources in Rhodesia,
the principal ones being:
River Notwani, Macloutsie, Tuli, Umtzingwane, Bubye, Nuanetzi (which has
already flown about 50 kms in Mozambique) and the River Changane.

1.2 - Geological Aspect, Soils and Vegetation:

In the Limpopo basin, formations are found which belong to different
systems, such as Karroo, Waterberg, Primitive System.
The basin in South African and Rhodesian territory seems to be consti-
tuted of basaltic lava, Serie Ecca, siliceous detrital rocks (sandstone) of brown
red and purple colours, formations of conglomerates, graphite and gneiss.
In Mozambique, the basin is mainly constituted of sedimentary formations.
In a narrow area near the border, volcanic rocks are found, in the upper
course of the Limpopo River and Elephants River formations of the Cretaceous Era,
in the rest, Quaternary formations with alluvium, sandstone, calcarium and sand
deposits.
./.
144

The vegetation in foreign territory is mainly constituted of bush and

grass, of great density in the highlands, and mixed bush and grass plains. In
Mozambique, the vegetation is of the bushy type and plains with some trees,
level grass plains and large stretches of grassy land.
The predominant soils in our territory are: sandy in the coastal area,
salty in the river vales, soils of mananga in the lower Changane and conglome-
rates.

1.3 - Climate:
In respect to the area situated in Mozambique, it appears that the
average annual temperatures are practically the same in almost all the basin,
being 240 C with the exception of the north eastern side, where it goes as low
as 220 C.
On the coastal and north-eastern areas, the average maximum daily tem-
peratures are 300 and 320. C and in the central area 34Q.C.
The average temperature in the hottest month is 280 C. and the lowest
260.c., the annual variation of these averages being between 60 and 9Q.C.
The average temperature in the coldest month is 20% in the central
area, and 18% in thê rest, while the coastal area has an average minimum in
the coldest month of 12%.
The annual average relative humidity in the central area is 65%, increa2
ing to the north and south to reach the highest rate of 75%.
According to the classification of Koppen, the climate of the basin is
in general the dryness of steppes with a dry season in winter, dryness of the
desert in the area of Pafuri, dryness of the steppes in the south of the basin,
and in the coastal area, tropical raininess of a savanna.
The predominant winds in the months of September to February are those
from the East and, during the other months, almost entirely with predominance
from the West.
In the whole basin, one finds that it is situated between the isothermics
of 240 and 170 with the average temperatures of 200, 2003 and 2002 respectively
for the areas of Beitbridge, Mapai and Trigo de Morais.

1.4 - Hydrological Occupation:

Both the South African Republic and Rhodesia have a network of udometric
and hydrometric stations which, for the African Continent, can be considered
dense: one pluviometer for 200 km2 and one hydrometric station for 4 o00 h 2 .
 145

There are readings from 31 Rhodesian udometric stations and from 90

South African posts, the majority of which with more than 30 years of existence.
The more significant hydrometric stations in Rhodesia and South African
not only for the area they cover and their locality, but for the extension of
their records, are:

Rhodesia :

- River Tuli - 4 144 km2

- Unzimgwane River - 2 533 km2
- Bubye River - 8 029 lan2

South Africa:

A3 MO7 - Eerste Poor - Groot Marico Rivier - 8 588 h 2

A2 M25 - Hardekool Bulti - Crocodile River - 21 i09 lan2
A5 MO2 - Vischgat - Palala River - 2 341 h 2
A5 MO3 - Oxenham Ranch - Limpopo River - 97 850 km2
A 7 MO4 - Beitbridge - Limpopo River -180 O00 h 2
A7 MO3 - Zamenkomst - Sand River - 6 900 km2
A 9 MO1 - Schuinshoogte - Luvuhu River - 912 lan2
- Liverpool - Olifants Rivier - 4 2 352 km2
- (354) - Olifants Rivier - 27 928 km2
- Manorvlei - Letaba River - 668 km2
- Letaba Ranch - LetkaRiver - 4 716 h 2
- Driehoek - Blyde River - 2 199 km2

In Mozambique, there are 33 udometric stations, some with a significant

period of regular readings, and 12 stations for the measurement of water volu-
me, some equipped with linmographs, the most significant of those with regular
readings being those of Maçuço (66 O00 km2) and Tiobine (68 450 km2) on the
Elephants River; Vila Trigo de Morais (340 O00 km2), Pafuri (235 930 km2),
Mapai (246 O00 km2), Mohambe (342 780 h2), João Belo (407 970 km2) on the
Limpopo River, and Chibuto (43 200 km2) on the Changane River.
As regards evaporation, there are 6 U.S. Class A Tina evaporemeters in
Rhodesia, 20 Standard Symons in South Africa, and 5 U.S. Class A Tina evapore-
meters and 7 Piche atmometers in Mozambique.

./.
146

Also in the Portuguese part of the basin are 3 lysimetric stations, 3 cli
matological stations, 2 agronomic-climatological posts and 4 climatological posts.

2 - RAINS
2.1 - Introduction:

From the analysis of the normal isohyetal map, it is noted that the basin
is within the ishoyetal extremes of 400 and 1.500 mm., the monthly distribution
of rainfall being divided in a deficient way throughout the year,with a concentrg
tion of about 85% of the total during the months from October to March inclusive.
To determine the average annual rainfall in the Limpopo basin as far as
Mapai, three methods were used:- of the rainy districts; of the area of influes
ce and the isohyetal figures.

2.2 - The method of the rainy districts

The South African Republic is divided into restricted zones by rains of
average equality, certain rainy districts under the same principle also dividing
the areas of Rhodesia.
In accordance with the udometric records for the period 1914/15 to 1963/
/64 (50 years), the average annual rainfall figures were determined in respect
to the basin, taking into account the fall in each district.
From the period of 50 years, the average annual rainfall arrived at was
579 mm, its having been 582, 566 and 560 mm respectively during the past 10, 15
and 25 years.

2.3 - Method of Area of Influence

In accordance with the records existing for the period 1954/55 to 1963/
/64 (10 years) and using the 71 udometric posts, the average annual rainfall was
determined, the figure obtained being 514 mm.

2.4 - Isohyetal Method

With the normal figures - 30 years - recorded at the udometric posts,.
isohyetal curves were interpolated, the areas between the adjacent isohyetal
figures being thereafter measured.
The average figure for the rainfall in the basin thus obtained was 504m.
 147

2.5 - Correlation of the methods:

For the common period of 1954/55 to 1963/64 and 1955/56 to 1963/64 of
which the average annual rainfall figures are available, the correlation between
the two methods was determined, the following equations having been obtained:

1954/55 to 1963/64 -x = 0,94y + 99

1955/56 to 1963/64 -x = 1,02y + 6 2

which give a lineal relation between them, the rates of the correlation being
very significant (0,98 and 0,96) the figures arrived at showing only small dif-
ferences .

2.6 - Resumé:
ANNUAL RAINFfiLL

Averages :

Period of 50 years ........... 579 mm

Period of 25 years ........... 560 mm
Period of 15 years ........... 566 mm
Period of 10 years ........... 582 mm
Rainiest year ................ 971 mm (1924/25)
Rainiest year w/lOO year occ-
rence ........................ 1 101 mm
Driest year .................. 355 mm (1963/64)
Driest year w/lOO year OCCUT-
rence ........................ 290 mm

3 - RUNOFF
3.1 - Elementary Principles:
In view of there not being any measurements of the volume of the Limpopo
River in Portuguese territory, it was necessary to resort to comparative studies,
taking as a basis the specific flowage in the various hydrometric stations,
existing upstream in the region of Mapai.

./*
148

3.2 - Details of the Study

Barrows, in his book "Water Power Ehgeneering" stated that "it is often
possible to consider, without serious error, that the specific volume of a river
is similar to the successive contours along the same river". The same author
also advises that, whenever possible, comparisons and corrections should be esta
blished, not only in respect to the rainfall, but also to the altitude, slopes,
constitution and the rock formations of the soil.
Based on the measurements made at the Hydrometric stations of the Repu-
blic of South Africa and Rhodesia which cover 195 841 km2, 79,6% of the respective
basin in the area of the barrage of Mapai, we took into account the specific run-
off year by year and we calculated the specific runoff of the locality under
study.
Since 1963, efforts have been made to estimate the average annual runoff
always using the methods of the specific volume but adopting various criteria.
The average annual figure obtained for the period of 12 years was 3 095
million cubic metres, which corresponds to the specific runoff of 12 580 m3
k2 -1 (chart attached).
Thus we have:
Study in 1963 (6 years) 13 700 m3 (k2)-1
specific runoff
Study in 1965 (7 years)specific runoff 13 658 m3 (k2)-1
Present study (12 years) specific runoff 12 583 m3 (k2)-1

If we observe the sequence of the years in which measurements existed

for the 3 studies realised, it will be noted that the period of 12 years includes
an excessively dry year (1963-1964) and one high runoff (1966-67) while there
were no measurements taken in 1965-66 at the fundamental station of study (Beit-
bridge) due to the hydrometric station having been under water (floods in Februa-
ry, 1966) and as a result of which, the figure now obtained is necessarily defec-
tive.
The 1970 study covering a period of 15 years, places the specific average
runoff as 13 O00 m 3 (k2)-l.
For the 246 O00 km2 of the river basin, the figures obtained for the
average annual runoff in the region of Mapai are respectively as follow:

1963 study .............. 3 370 million m3

1965 study .............. 3 360 million m3
1970 study .............. 3 200 million m3
Present study ........... 3 095 million m3

./.
 149

Any of these figures fall within the admissible limits based on the rate
of the runoff observed at the stations of Beitbridge and Vila Trigo de Morais,
the average of which is respectively 0,015 and 0,026.
Thus for a figure of 3 O00 million m3 and for the average rainfall of
579 nun., runoff coefficient is 0,021, which is within the observed limits.
Using the method of Coutagne only for the average annual figures, for
a rainfall of 579 mm as the average over 50 years, we arrive at a runoff of
2 969 106 m 3 and for 582 nun as the average for the past 10 years, the figure
of 2 999,7 106 m3 .
Observing and trying all these ways and means, we shall adopt chart
attached hereto for the annual runoff because, as they arise from direct mea-
surements, they fall within all the estimated figures.

3.3 - Monthly Distribution

The monthly distribution is based on a hydrometric station in the Repu-
blic of South Africa (Beitbridge), which already has years of sufficient read-
ings, its area being much like that of Mapai.
Thus we haïe:

October ................ 0,5%

November ................ 0,6%
December ................ 4,6%
January ................25,%
February ................33,8%
March ................ i8,i%
April ................ 8,s
May ................ 4,3%
June ................ i,%
July ................ 1,1%
August ................ o, 9%
September -
................ 0,3% 10%

4 - FLOODS
The River Limpopo is typically torrential and as such, not only dries
during consecutive months as it is susceptible to exceptional floods.
Many records of the volume of floods have been compiled in the Republic
of South Africa since 1915, always based on specific volumes and they obtained

./.
150

measured details which extended to the area of Mapai, gave us the figures of
16 925 m3/s (1933) and 12 792 m3/s (1966).
Thus, using various formulas, one can estimate the volume of floods for
return periods of 100 and 200 years.

100 years 200 years

Water affairs formula ............ 13 243 14 963 m3/s

Mimoso Loureiro formula ........ 14 304 18 375 II
Fuller formula ................. 19 763 21 284 11

(Period 34 years and specific

volumes measured)
Larivaille formula ............... - 15 375 I'

The estimate is thus very difficult and depends a great deal on the
type of barrage to be adopted.
The hydrograph of a maximum flood was also determined, based on the
formula of Giandotti in the calculation of the times of concentration of the
peak and of the swell of the flood, and on the hydrographs of the floods record-
ed at the border (Pafuri) during the years 1955, 1958, 1959, 1966 and 1967.
It was verified that, in the flood of 1966, there was agreement in the
calculated and observed times, because the calculation placed the figure at
153 hours and from observation, at 150 hours for the time of concentration,
there being, however, a difference in the time of the swell of 600 with 724
hours (sketch attached).

5 - Evaporation and Solid flows

Using the details measured in Rhodesia, South Africa and Mozambique,

we can place the resulting evaporation at M a p i as 1 344 mm. As to solid volu-
mes, the figures are few and vary greatly; as for the rest, definitely, confix
med by the complex composition of the hydrographic basin, as for the same volu -
me one obtain 21,8 kg/s and 97,l kg/s, without any meaning.
 151
0
N
o
n
N
1
I
I
W
s
u
o
a
r
U
2
n
-I
4
3
z
z
4
 Relation of hydrological programs of the Center of Hydrographic Studies
for complete studies of hydraulic resources with insufficient data

Dr. Rafael H E R A S

1 LISTA Lista las ochenta columnas de las fichas.

The programme lists the eighty columns of the cards.

2 LIS 65 Lista las ochenta columnas de las fichas, poniendo en la

cabecera de la pagina las columnas numeradas y saltan-
do página cada 65 fichas.

It lists eighty columns of the card, putting on the heading

of the page the numered colum and skipping a page after
every 65 cards.

3 LIFLA Lista fichas con flags.

It lists carde with flags.

156

4 LIMO1 Lista cinta de papel del limni'grafo.

It lists ribbon of the limnigraphe.

5 LICAN Listado de longitudinales.

Longitudinal listing.

6 LIGUI Listado de transversales.

Transversal listing.

7 LGMET Listado de datos geográficos.

Listing of geographical data.

8 LPMET Lista la cuenca, n ú m e r o de estación, aflo y datos de prg

cipitaciones mensuales.

It lists the basin, n u m b e r of station, year and monthly

rainfall data.

9 LFRNT Lista cinta de papel perforada en código FERRANTI.

It lists perforated ribbon in FERRANTI code.

 157

10 LI-C-CA Lista cabeceras de los canales de aforos.

It lists the headings of the channels of valuation.

11 L-NI-24H Lista precipitaciones máximas en 24 horas.

It lists the m a x i m u m rainfall in 24 hours.

12 LTVP Lista tablas de vertidos probables,

It lists tables of probable downpour.

13 L T A C Lista las tablas de alturas-caudales.

It lists tables of altitudes-flows.

14 TAVAL Lista los valores de las curvas alturas-caudales y está

preparado para obtener valores que no figuren en las ta
blas, interpolando linealmente entre los dos puntos más
próximos.

It lists the values of altitudes-flows lines and it is

prepared to obtain values not appearing in the tables,
by linear interpolation between the closest points.
158

15 TAB-1 Este programa lista, de las cinco fichas de que consta

la información de cada pozo, lo siguiente: número de pg
zo, fechas de muestreo, número de muestras (laborato-
rio), latitud, longitud, nivel de agua/l, columna de agua,
horas de bombeo, caudal en Ils., procedencia, número
de laboratorio, fecha del análisis, temperatura del aire,
temperatura del agua e indicador.

This programme lists the five cards which contains the

information of each well, as the following:number of
well, dates of sampling, number of samples (laboratory),
latitude, longitude, level of water/l, column of water,
hours of pumping, flow in Ils., origin, laboratory number,
date of analysis, temperature of the air, temperature of
water and indicator.

16 TAB-2 Lista: número de pozo, fecha de muestreo, calcio, mag-

nesio, manganeso, sodio, potasio, cloruro, sulfato, fluo
ruro, silice, fosfato y carbonato.

It lists: number of well, date of sampling, calcium,

manganese, sodium, potassium, chloride, sulphate,
fluoride, silica, phosphate and carbonate.

17 TAB-3 Lista: número de pozo, fecha de muestreo, bicarbonato,

nitrito, nitrato, amoniaco, boro, hierro, pH, resistivi-
dad, gravedad especifica, sólidos disueltos, litio, estrog
cio, ni’quel.

It lists: number of well, date of sampling, bicarbonate,

nitrite, nitrate, boron, ammonia, iron, pH, resistivity,
gravity/m, solid in dissolution, lithium, strontium, -
nickel.

18 TAB-4 Lista: nimero de pozo, fecha de muestreo, cobalto, iodo,

bromo, molibdeno, zinc, plomo, cromo, cobre, vanadio,
m e r curio, arsénico.
 159

It lists: number of well, date of sampling, cobalt, iodine,

bromine, molibdenum, zinc, plumbum, chromium, cop-
per, vanadium, mercury and arsenic.

19 TAB-5 Lista: número de pozo, fecha de muestreo, pH, resisti-

vidad, gas carbónico libre, oxigeno disuelto, dureza, dg
reza (sin carbonatos), alcalinidad, T.A., HC03.

It lists: number of well, date of sampling, pH, resistivity,

free carbonic gas, oxygen disolved, hardness (without
carbonates), alkalinity, T.A., HC03.

20 TNFAG Tablas de interpolación polinómica de cuarto grado.

Tables of polynomical interpolation of the fourth grade.

21 TANG Tablas de senos, cosenos y tangentes.

Tables of sine, cosine, tangents.

22 TNUA Mete en disco tablas de números aleatorios.

It puts in the disk tables of fortuitous numbers.

23 T-V-P Mete en disco tablas de datos hidrológicos.

It puts in the disk tables of hydrological data.

160

24 T-EVA Mete en disco tablas de evaporaciones.

It puts in the disk tables of evaporation.

25 EN-PRT D a d a una serie de estaciones, almacena e n disco dichas

estaciones.

It stores in the disk stations which are previously given.

26 SAPRI Obtiene un listado de las estaciones almacenadas e n disco.

It obtains a listing of stations stored in the disk.

27 PDPRC Perfora datos con anos consecutivos para los diversos

programas de regulaciones.

It perforates data with consecutive years for the different

programs of regulations.

28 PC 128 D a d a una serie de caudales diarios con formato 12 F 6.2,

los perfora con formato 9 F 8.3, para ser utilizados por
el programa I~TCDAP".

Given a series of daily flows with format 12 F 6. 2, it

perforates t h e m with format 9 F 8. 3, to be used by the
p r o g r a m I'TDFAF~~.
 161

29 PRD-I Ordena un m á x i m o de 1.000 datos, de m a y o r a m e n o r ,

con formato 12 F 6.2.

It puts in order a m a x i m u m of 1.000 data, in descendent

order, with format 12 F 6. 2.

30 PRD-2 Igual que el anterior, pero con formato 9 F 8.3.

T h e s a m e as above, but with format 9 F 8.3.

31 PRD-3 Ordena n ú m e r o s en c o m a fija, de m a y o r a menor.

It puts in order n u m b e r s with fixed point, in descendent

order.

32 DUPLP Duplica las ochenta columnas de las fichas.

It duplicates the eighty columns of the cards.

33 DUP-Mp Duplica las ochenta columnas, modificando la columna

que se desee.

It duplicates the eighty columns, modifying the c o l u m n

that is wished.

34 SUMNU S u m a o resta un n ú m e r o a una serie de datos mensuales.

It adds or deducts a n u m b e r f r o m a series of monthly data.

162

35 SUMRE Dadas dos series de datos mensuales, las suma o las

resta.

it adds or deducts two series of monthly data which are

given.

36 SUMAR S u m a series de datos hidrológicos.

It adds series of hydrological data.

37 MULTI Multiplica series de datos por un número fijo y obtiene

el listado, así c o m o las fichas perforadas, con estos nue
vos valores.

It multiplies series of data by a fixed number and it

obtains the listing, in the same way as the perforated
cards, with this new values.

38 MULTA Multiplica los 12 números de la primera fila por cada

uno de los datos, los pone en orden decreciente y los lis
ta en 12 columnas.

It multiplies the 12 numbers of the first row by each one

of the data, it puts them decreasing order and it lists -
them in 12 columns.

39 M U L 12 Multiplica los datos mensuales por el número que ocupa

el lugar correspondiente a ese m e s en la primera ficha,
lista y perfora.

It multiplies the monthly data by the number that occupies

the corresponding place of that month in the first card, it
lists and Derforates.
 163

40 MULAN Multiplica los datos mensuales de cada ano por el & m e -

ro que ocupa el lugar correspondiente a ese ano en las
primeras fichas que lee.

It multiplies the monthly data of each year by the n u m b e r

that occupies the corresponding place to that year in the
first cards that is read.

41 MRS 12 Multiplica, s u m a o resta dos o una series de datos sin

limitación.

It multiplies, adds or deducts two or one series of data

without limitation.

42 M A TEN Eleva una matriz a la potencia enésima.

It elevates a matrix to the n power.

43 SECUA Resuelve un sistema de ecuaciones (40 c o m o máximo).

It solves a system of equations (40 as m a x i m u m ) .

44 DIS -KM D a d a la situación geográfica de una serie de estaciones

por su latitud y longitud, este programa selecciona los
grupos de estaciones a c o m p a r a r con el criterio de que
las distancias entre las estaciones sean menores de una
cantidad fija.

Given the geographical position of a series Of stations by

their latitude and longitude, this p r o g r a m m e chooses the
group of stations to be compared with the criterium that
the distance a m o n g the stations be smaller than a fixed
quantity.
164

45 LISDA Lista series de datos mensuales (sin limitación de exten

sión), imprime cabecera y calcula la s u m a anual y las -
medias mensuales.

It lists series of monthly data (without limitation in its

scope), prints heading and computes the yearly s u m and
the monthly averages.

46 MAXLL Dados los valores de precipitación total mensual, dias

de lluvia y los valores máximos en 24 horas, obtiene los
máximos en 24 horas y los dilas de lluvia a escala anual.

Given the values of total monthly rainfall, days of rain

and the m a x i m u m values in 24 hours, it obtains the-
m a x i m u m in 24 hours and the days of rain in a yearly
scale.

47 INT-ES Dados los caudales medios mensuales y anuales, la apoy

tación media de los años precedentes y el caudal máximo
(medios diarios e instantáneo y la fecha), obtiene cauda-
les medios anual y mensual, caudal y aportación mensual.
Deduce la aportación y caudal anual, caudal medio y apor
tación media de la serie anual.

Given the monthly and annual average flows, the average

afford of the preceding years and the maximlim flow (daily
and instantaneous averages and the date), it obtains year
ly and monthly average flows, monthly flow and afford. It
deducts yearly afford and flow, average flow and the aver
age afford of the yearly series.

48 INT-EM Dado el volumen embalsado, aportación de salida y media

precedente, obtiene entradas y salidas, reserva, aporta-
ción del año, aportación de entrada y salida, media de e;
trada y salida (aportación y caudal).
 165

Given the stored volume, afford of exit and the preceding

average, it obtains entries and exits, reserve, yearly -
afford, afford of entry and exit, average of entry and -
exit (afford and flow).

49 INT -CA Realiza la misma función, pero sin el caudal máximo

instantáneo.

It performs the s a m e function, but without the m a x i m u m

instantaneous flow.

50 ADfDB-1 Dadas las series de datos hidrológicos de un conjunto de

estaciones a escala anual, mensual, etc., forma "esta-
ción tipo" (media aritmética de las estaciones de cada
grupo) y a continuación compara cada una de las estacio
nes con su "estación tipo", acumulando las series y dag
do, además de las sumas acumuladas, la relación entre
las acumulaciones de cada estación con las acumulacio-
nes de la "estación tipo".

Given the series of hydrological data of an assembly of

stations at yearly scale, monthly scale, etc., it forms
''typestation!' (arithmetical mean of the stations in each
group) and afterwards compares each stations with its
'hype station", accumulating the series and giving besides
the accumulated sums, the relation among the accumula-
tion of each station with those of "type station".

51 ADfDB-2 Este programa es análogo al anterior, pero no utiliza -

estación tipo, haciendo todas las comparaciones posibles
en cada grupo.

This programme is analogous to the preceding one, but

it does not use type station, performing all the possible
comparisons in each group.
166

52 A-AC-96 Acumula datos hidrológicos mensuales (de 1 en 1 hasta

96 en 96 meses). Imprime 96 cuadros.

It accumulates hydrological monthly data (from 1 to 1 up

to 96 in 96 months). It prints 96 charts.

53 AC-96-P Este programa es análogo al anterior, pero perfora los

resultados en ficha.

This programme is analogous to the preceding one, but

it perforates the results in card.

54 AM-AC 1 Acumula aportaciones mensuales y ordena de menor a

mayor .

It accumulates monthly affords putting in ascendent order.

55 A M - C 96 Dada una serie de datos hidrológicos mensuales, acumu-

la de 1 a 96 meses consecutivos, y los ordena de menor
a mayor.

Given a series of hydrological monthly data, it accum;I-

lates from 1 to 96 consecutive months, and putting them
in ascendent order.

56 AC-T~D Acumula todos los valores de una serie.

It accumulates all the values of a series.

 167

57 AD~BP Dibuja en el Plotter los diagramas de las acumulaciones

dobles.

It designs in the Plotter the diagrams of the double --

accumula tion.

58 ALAOA Dadas las aportaciones diarias, las lista, ordena de m g

nor a m a y o r y las acumula.

Given the daily affords, it lists, puts in ascendent order

and accumulates them.

59 TCDAP D a d a una serie de caudales diarios, la transforma en -

aportaciones diarias.

Given a series of daily flow, this p r o g r a m m e transforms

t h e m in daily affords.

60 TCAP~ D a d a una serie de caudales mensuales, la transforma en

aportaciones mensuales.

Given a series of monthly flows, this series is transform

ed in monthly affords.

61 TAPPC D a d a una serie de aportaciones mensuales, la transforma

e n caudales mensuales.

Given a series of monthly affords, it transforms t h e m in

monthly flows.
168

62 TANAD Dada una serie de aportaciones naturales diarias, la -

transforma en aportaciones derivables diarias.

Given a serles of daily natural affords, the programme

transforms them into daily derivable affords.

63 C-C-D-ES Dada una serie de caudales diarios, obtiene los caudales

derivados diarios en m3/s., medias y aportaciones men_
suales, caudales clasificados, aportación y caudal total
del año.

Given a series of daily flows, the programme obtains the

daily derivable flows in CU. m/s., averages and monthly
affords,classified flows, afford and total flow of the year.

64 C -C-D-CA Realiza la misma función que el programa anterior, pero

con caudales mensuales.

It performs the same function as the preceding programme,

but with monthly flows.

65 C-C-D-EM Dados los volúmenes y salidas diarias de un embalse, o b

tiene las reservas diarias (Hm3), caudales diarios (sali-
das en m3/s. ), media mensual, salida y entrada mensual,
evaporacion y un resumen anual (caudales medios, salida,
entrada, evaporación).

Given the volumes and the daily exits of a reservoir, the

programme obtains the daily reserves (CU. Hm),daily
flows (exits in CU. m/s), monthly average, monthly entry
and exit, evaporation and an annual summary (averages
flows, entry, exit, evaporation).
 169

66 C-C-D-E1 Dados los datos de alturas de escala diaria y los valores

de las curvas alturas-caudales de una estación, calcula
los caudales diarios de dicha estación. Los caudales que
no figuran en las tablas se calculan interpolando lineal-
mente entre los dos m á s próximos.

Además del caudal diario, este programa obtiene los cag

dales máximos y mínimos, los caudales medios mensua-
les y las aportaciones mensuales.

Given the daily scale heights and the values of the height-
flow charts of a station, the programme computes the --
daily flows of said station. The flows that are not appear-
ing in the tables, are calculated by linear interpolation-
between the closest points.

In addition to the daily flow, this programme obtains the

m a x i m u m and minimum flows, the monthly average flows
and the monthly affords.

67 C-C-D-E2 Dados los datos de alturas de escala diaria y los valores

de las curvas alturas-caudales de una estación, calcula
los niveles diarios en metros, los caudales diarios en
m3/s. , medias mensuales, máxima instantánea, aporta
ci& mensual en Hm3,caudales clasificados y un resu--
m e n de los datos del año (aportación y caudal total y es-
pecifico y caudales caracteristicos).

Given the daily scale heights data and the values of height-
flow lines of a station, the programme computes the daily
levels in meters, the daily flows in CU. m/s., monthly -
averages, instantaneous maximum, monthly afford in --
CU. Hm, classified flows and an annual summary data --
(afford and total flow, specific flow and caracteristic --
flows).

68 C-C-D-ES Dados los datos de alturas de escala diaria y los valores

de alturas-caudales de una estación, calcula los caudales
diarios y el caudal medio anual.

Los datos son los obtenidos en limnigrafo.

170

Given the daily scale heights data and the values of height-
flow lines of a station, the p r o g r a m m e computes the daily
flows and the yearly average flow.

Data are obtained by the limnigraphe.

69 CURGA A partir de unos puntos base tabula una tabla de gastos.

Lista y dibuja los diagramas de las curvas de gastos.

F r o m a basic point, it tabulates a tables of expenses.

It lists and designes the diagram of the expense lines.

ME Y ME D a d a s las aportaciones mensuales de una serie de esta-

ciones de una cuenca, calcula e i m p r i m e las medias m e c
suales de cada estación y la media de las medias de todas.

Given the monthly affords of a series of stations of a basin,

it computes and prints the monthly average of each station
and the average of all means.

71 MEDIP D a d o s los datos diarios de años de una estación, los

i m p r i m e y calcula las s u m a s y medias mensuales de c a
da ailo y las medias de las medias (mediorum).

Given the daily data of c years of a station, the p r o g r a m m e

prints t h e m and computes the sums and monthly averages
of each year and the average of the n_ m e a n s (mediorum).

72 MET, M E Dada una serie de datos mensuales, calcula ias medias

para cualquier periodo.

It calculates the average for any period of a series of

monthly data, which are given.
 171

73 CICLJD Calcula las medias acumuladas en periodos de n años y

sus relaciones con la media del periodo total.

It calculates the accumulated averages in periods of E

years and their relations with the average of the total
period.

74 MED -AN Dada una serie de datos anuales, calculala media para
cualquier pedodo.

Given a series of yearly data, the programme calculates

the average for any period.

75 A-ESP Dadas unas series pluviométricas reales, obtiene una s e

rie real, media de las anteriores, a partir de la cual ob-
tiene otra de precipitaciones efectivas, de la que se con-
siguen las aportaciones especificas y los caudales en una
cuenca.

Given an actual pluviometrical series, the programme

obtains an actual series, average of the preceding, from
which it obtains another series of effective rainfalls, -
from which it gets specific affords and the flows in a basin.

76 -
INF 1 Dadas las precipitaciones mensuales de una cueBcay los
coeficientes de capacidad de infiltración mensual en mm.,
de humedad inicial del suelo y de superficie en Has., ob-
tiene las aportaciones especificas, infiltración y evapora
cion.

Given the monthly rainfall of a basin and the coefficients

of monthly infiltration capacity in mm.,initial humidity
from the earth and from surface in Has., it gets the --
specific affords, infiltration and evaporation.
172

77 INF -2 Realiza la m i s m a función que el p r o g r a m a anterior, pe-

ro a nivel diario.

It performs the s a m e function as the preceding one, but

in a daily level.

78 EVAP-1 D a d o s los volúmenes, superficies y reserva, halla las

evaporaciones.

T h e p r o g r a m m e computes the evaporations, knowing the

volume, area and stock.

79 EVAPT D a d o el n ú m e r o de estaciones, superficie de la cuenca

(kmz), aportación media (media de una serie) y precip'
tación media, obtiene la aportación en Hm3,coeficiente
de escorrentia y déficit de escorrentía.

It obtains the afford inCU. H m . , runoff coefficient and

deficit of runoff, knowing the n u m b e r of stations, area
of the basin (sq. Km. ), average of the afford (average
of a serie) and the average rainfall.

80 REST-l Lista: n ú m e r o total de muestras, media aritmética, va-

lor m á x i m o , valor minimo, desviación tipica, tempera-
tura del agua.

T h e p r o g r a m m e lists: total n u m b e r of samples, arithmeg

cal m e a n , m a x i m u m and m i n i m u m value, standard desvia
tion. temperatura of the water.

81 REST-2 Lista los m i s m o s parámetros para calcio, magnesio, mall

ganeso, sodio, potasio, cloruro, sulfato, fluoruro, silice,
fosfato y carbonato.
 173

It lists the s a m e parameters for calcium, magnesium,

manganese, sodium, potassium, chloride, sulphate,
fluoride, silica, phosphate and carbonate.
-

82 REST-3 Lista los m i s m o s parámetros para bicarbonato, nitrito,

nitrato, amoniaco, hierro, resistividad y sólidos disuel
tos.

It lists the s a m e parameters for bicarbonate, nitrite,

nitrate, a m m o n i a , iron, resistivity, solid in dissolution.

83 REST-4 Lista los m i s m o s parámetros para pH, resistividad, gas

carbónico libre, oxígeno disuelto, dureza, dureza (sin -
carbonatos), alcalinidad, T.A. , HC03.

It lists the s a m e parameters for pH, resistivity, free

carbonic gas, oxygen dissolved, hardness, hardness -
(without carbonates), alkalinity, T.A., HC03.

84 D-FLSI D a d o el perimetro y la superficie por encima de cada co-

ta de una cuenca, obtiene el rectángulo equivalent e, coe-
ficiente de Gravelius, indice de pendiente, pendiente m e -
dia y altitud media de la cuenca.

Given the perimeter and the area above each elevation of

a basin, the p r o g r a m m e obtains the equivalent rectangle,
Gravelius factor, pendant index, average of the pendant
and average altitude of the basin.

85 GE~NE A partir de coordenadas de puntos fijos dados, de medi-

ciones, direcciones y distancias para una conexión de pu9
tos aislados o múltiples, el p r o g r a m a obtiene las coorde-
nadas de los nuevos puntos.
 F r o m coordinates of fixed given points, of measuring,
directions and distances to a connection of isolated or
multiples points, the p r o g r a m m e gets the coordinates
of the n e w points.

86 DIS-2P D a d a s las coordenadas de dos puntos, calcula su dis-

tancia.

Given the coordinates of two points, it calculates their

distance.

87 HELME Convierte coordenadas instrumentales en terrestres

(Helmert ).

It converts from instrumentais coordinates to terrestrial

(Helmert).

88 REPPU E n función de las coordenadas de los puntos calcula azi-

mutes y distancias para replanteo.

In function of coordinates of the points, it calculates azi-

muths and distances to be reconsidered.

89 ARCES Calcula la superficie en planta que abarca cada curva en

un embalse.

It calculates the area which comprises each line in a

reservoir.
 175

90 EAKIN Dada la superficie del tramo máximo de un embalse, su

longitud y la superficie del perfil de sondeo, obtiene el
volumen de esas superficies.

Given the area of the m a x i m u m stretch of a reservoir,

their longitude and the area of the profile of sounding,
it obtains the volumes of those surfaces.

91 RAPEM Cubica un embalse según la fórmula RAPEM.

It cubes a reservoir according to formula R A P E M .

92 C~R-AM Dada una serie mensual de datos hidrológicos, calcula

la correlación ortogonal, los momentos de la serie reg
pecto al origen y el coeficiente de correlación.

Given a monthly series of hydrological data, it calculates

the ortogonal correlation, the moments of the series with
regards to the origin and the correlation factor.

93 CPR-LM Calcula la correlación lineal mensual, obteniendo el cog

ficiente de correlación, las medias, varianzas, disper--
si&, coeficiente angular y ordenada en el origen de la
recta de regresión.

The programme calculates the monthly linear correlation,

obtaining the correlation factor, the averages, variances,
dispersion, grade angle coefficient and ordinate at the -
origin of the regression straight.

94 CPR-AN Este programa se diferencia del anterior Únicamente en

que, partiendo de los datos mensuales, utiliza solamente
176

los anuales, efectuando la correlación entre ellos con -

arreglo al esquema ya reseñado.

This programme differs from the previous one only in

that starting from the monthly data, it uses only the -
annual ones and performing the correlation among them,
in accordante with the indicated scheme.

95 C~R-DM Con los datos hidrológicos mensuales de tres estaciones,

z, x e y, realiza la correlación doble de x e y con z, ob-
teniendo la ecuación del plano de correlación.

With the monthly hydrological data of the three stations,

z, c and y, it performs the double correlation of x and y
with z, obtaining the equation of the plane of correlation.

96 CPR-PA El esquema es igual al de los anteriores que realizan cg

rrelaciones, dando éste la ecuación de la parábola de re
gresión de y sobre x.

The scheme is identical to the previous one which performs

correlations, , the equation of the parabola of regression
of y on x is given by the scheme.

97 C ~ R - ~ R Dadas dos series mensuales de datos hidrológicos, calcg

la la correlación ortogonal obteniendo la recta cuya suma
de cuadrados de distancias a los puntos es minima.

Given two monthly series of hydrological data, the pro-

g r a m m e calculates the ortogonal correlation, obtaining
the straight whose s u m of squares to the points in min-
imum.
 98 CfbR-A-p D a d a una serie mensual de datos hidrológicos, realiza
la correlación anual ortogonal.

Given a monthly series of hydrological data, it performs

the yearly ortogonal correlation.

99 CPR-48 Realiza la correlación ortogonal para los m e s e s de estia

je (4)y el resto de los m e s e s (8).

It performs the ortogonal correlation for the s u m m e r

months (4) and to the balance of the months (8).

100 CPR-LP Dadas dos series de datos hidrológicos, obtiene la recta

de correlación ortogonal entre sus logaritmos.

Given two series of hydrological data, it gets the straight

of ortogonal correlation a m o n g their logarithms.

10 1 COR-O-P El esquema es igual al del p r o p a m a "CqR-qR" y, ade-

m á s , dibuja la nube de puntos en el Plotter.

T h e s c h e m e is identical to " C O R - O R " p r o g r a m m e , in

addition it draws the clouds of points in the Plotter.

102 cpycfb Completa y corrige una serie de datos hidrológicos dan-

do las ecuaciones de las rectas de regresión (y=ai x bi) -
entre las dos series (se puede hacer con una ecuación o
con dos simultáneamente).

It completes and corrects a series of hydrological data,

giving the equation of the regression straights ( x = a ix - bi)
between the two series (it can be done with a n equation or
two simultaneously).
178

io3 ~p-DRR Completa los datos hidrológicos según la recta de regre-

sión.

It completes the hydrological data according to the reg-

sion straight.

104 IN-D-M1 Inventa datos mensuales de una estación con datos anua-
les, a partir de los datos mensuales de otras dos esta--
ciones.

It creates monthly data of a station with yearly data, star

ting f r o m the monthly data of other two .stations.

105 IN-D-M2 Inventa datos mensuales de una estación con datos anua-
les, a partir de los datos mensuales de otra según la 'Óf
mula B (I) = [A (I) * ( S U M B (I) / SUMA (I) 1.

It creates monthly data of a station with yearly data,

starting from monthly data of other station, according
to formula: B (I) = I A (I) * ( S U M B (I) / S U M A (I) 1 .

106 COQUI Calcula 20 correlaciones ortogonales, entre elementos

quhnicos, pintando por Plotter los puntos y la recta de
regresión y dos paralelas a una distancia igual a la dis-
persion.

T h e p r o g r a m m e calculates 20 ortogonal correlations,

a m o n g chemical elements, drawing in Plotter the points
and the regression straight and two parallel lines to a
distance identical to the dispersion.
 179

107 BERKA Dibuja el diagrama de Berkaloff-Scholler por el Plotter,

uno por cada pozo, en una escala logaritmica. Pinta los
puntos de los siguientes elementos: CA, MG, ALC, CL,
SO4, HCOQ + CO3,NO3. y une con segmentos dichos -
puntos.

It designs the diagram of Berkaloff-Scholler by means of

Plotter, one diagram to each well, in logarithmical scale.
It draws the points of the following elements: CA, MG, -
ALC, CL, SO4, H C 0 3 t COQ,NO3, and it joins the - -
mentioned points with the segments.

108 STIF Dibuja el diagrama de Stif.

The programme draws the Stif diagram.

109 PIPER Dibuja el diagrama de Piper. Consta de dos triángulos

equivalentes;en la base del primero se marca en 70el v a
lor C A , en mgl/l, y en los otros dos lados, M G y NA+K,
y mediante paralelas a 10s lados opuestos obtenemos un
punto.

D e la misma forma, en el segundo triángulo pintan en el

lado base, CL, y en los otros dos CO3 t HCH03 y so4 t
t NOP en 70y mediante paralelas obtenemos otro punto.
Esta operación se repite para cada pozo y para las ocho
zonas.

It draws the Piper diagram. It is composed of two equiva

lent triangles; in the base of the first one, it marks in 70
the value CA, in mgl/l, and in the other two sides, M G
and NA t K, and by means of parallel lines to the opposite
sides obtaining a point.

In the same way, in the second triangle, it marks in the

sidebase, C L and in the other two sides, Co3 + HCHO
and SO4 + NOP in 7 0 and by parallel lines obtaining another
point. This execution is repeated for each well and for the
eight bands.
180

i10 G ~ K A los valores de una serie de datos hidrológicos, le ajug

ta una ley de Goodrich y contrasta la bondad del ajuste -
mediante el test de Kolmogoroff.

Given a series of values of hydrological data, it fits with

Goodrich’s law and contrasts the perfection of the fitting
by Kolmogoroff’ test.

111 GPK~L A partir de una serie de datos hidrológicos anuales, se

ajusta una ley de Goodrich y se contrasta la bondad del
ajuste mediante el test de Kolmogoroff.

Los datos de entrada son series mensuales. El programa

obtiene las series anuales mensuales, la media, los m o -
mentos respecto al origen de orden 2 y 3, los momentos
centrales de segundo (varianza)y de tercer orden, as( -
c o m o los parámetros de la ley de distribución de Goodrich.

Starting from a series of hydrological annual data, Good-

rich’s law is adjusted and the perfection is contrasted by
Kolmogoroff’test.

The entry data are monthly series. The programme --

obtains, annual series, monthly series, the average, the
moments with regard to the origin of order 2 and 3, the
central moments of second (variance)and third order, -
the programme obtains also the parameters of the Good-
rich distribution law.

112 GPKAF Dada una serie de datos hidrológicos, ajusta la ley de dig
tribución de Goodrich.

Given a series of hydrological data, the Goodrich distri-

bution law is adjusted by the programme.
 181

113 GPKPL Realiza la misma función que el programa IIGQKOLII y

dibuja las curvas.

It performs the same function that the "GOKOL"

programme and draws the curves.

114 GPK 70 Dada una serie de datos hidrológicos, ajusta una ley de
Goodrich y contrasta la bondad del ajuste mediante el
test de Kolmogoroff.

Given a series of hydrological data, it fits with Goodrich's

law and contrasts the perfection of the fitting by Kolmogo
s test.
roff '

115 GUMB 1 Ajusta una ley de Gumbel a una serie de datos hidrológi-
cos. El programa nos da los diversos valores que resul-
tan de la ley de Gumbel, ajustada para tiempos de recu-
rrencia de 5, 10, 25, 50, 100, 500 y 1000 anos ylista --
los datos originales clasificados de menor a mayor, asig
nándoles a cada uno la frecuencia 2n-1 / 2 N, donde n es
el número de orden y N el total de datos.
Datos de entrada (12 F. 6.2). Anuales.
This programme fits the Gumbel's law to a series of --
hydrological data. The programme gives us several values
according to the Gumbel's adjusted law, for time of --
recurrences 5, 10, 25, 50, 100, 500, 1000 years and it
lists the original data classified in a crecent order assign-
ing to each one a frecuency equal to 2n-1 / 2 N, where n is
the number of order and N the total number of data.
Entry data (12 F 6. 2). Yearly.

116 GUMB 2 Realiza la misma función que el programa "GUMB l", pg

ro los datos de entrada son (9 F 8.3).

It performs the same function as "GUMB l", but the entry

data are (9 F 8.3).
182

117 GUMB 3 Realiza la misma función que el programa "GUMB 1".

pero los datos de entrada son mensuales.

It performs the same function as the "GUMB l", but the

entry data are monthly data.

118 GUMB 4 Ajusta una ley de Gumbel a una serie de datos mensua-
les. Obtiene mensuales y máximos anuales.

This programme fits the Gumbel's law to a series of

monthly data. It obtains also monthly and m a x i m u m
annual series.

119 GUMB-P Realiza la misma función que el programa "GUMB 1"

y además dibuja la nube en el Plotter.

It performs the same function as the programme "GUMB 1"

and draws the clouds of points in the Plotter.

120 C-C-D Dados los datos diarios de aportaciones naturales, se ob

tienen datos diarios de aportaciones derivadas para dis-
tintos caudales de derivación, con los que se obtienen ds
tos mensuales de aportaciones naturales y derivadas. Con
estas parejas de datos se ajustan unas curvas, que sirven
para obtener datos mensuales de aportaciones derivadas
cuando sólo se tengan datos mensuales de aportaciones -
naturales.

Given the daily data of natural affords, the programme -

obtains daily data of derived affords for different flows -
of derivation by which are obtained monthly data of natural
and derived affords; with these pairs of data, s o m e lines
are adjusted, which are used to obtain monthly data of
derived affords, when only monthly data of natural affords
are had.
 183

1.21 GAMMA D a d a una selección de 10 valores equidistantes de la fun_

ciÓn g a m m a de X (con 16 cifras significativas) y sus seis
primeras diferencias en el intervalo (1, 2), obtiene por
interpelación cualquier g a m m a de X.

Given a selection of 10 equidistant values with the func-

tion g a m m a of X (with 16 significative digits) and their
six first differences in the interval (1. 2), the p r o g r a m m e
obtains by interpolation any g a m m a of X.

122 NUMRE D a d a una selección de 10 valores equidistantes de n y

sus seis primeras diferencias en el intervalo (-0.75,
4.25), obtiene la función inversa de la función de G o o d -
rich.

Given a selection of 10 equidistant values of n and their

six first differences in the interval (-0.75, .4.25), the
p r o g r a m m e obtains the inverse function of the Goodrich
function.

123 A M - C 4P Ajusta una ley de frecuencias parabólicas a los diez m e -

nores valores de una serie de aportaciones clasificadas,
sacando el valor de la aportación correspondiente a una
garantía dada.

T h e p r o g r a m m e fits a law of parabolic frecuency to the

ten least values of a series of classified affords, obtain-
ing the value of the afford corresponding to a given - -
guarantee.

124 CD SOO Calcula las aportaciones acumuladas correspondientes a

una garantía dada, obteniendo la curva de seguridad.

It evaluates the accumulated affords corresponding to a

given guarantee, obtaining the safety lines.
184

125 CD SO1 Calcula las aportaciones acumuladas durante un afio,

obteniendo la curva de seguridad.

It calculates the accumulated affords during a year,

obtaining the safety line.

126 CD 502 Calcula las aportaciones acumuladas, después las clasi-

fica considerando los siguientes periodos:

Oct Nov Set

Oct + 1 Nov + 1 Set + 1

____-----------
Ott + k N O +~k Set + k

y obtiene después las aportaciones acumuladas corres -

pondientes a una garantia determinada. Posteriormente,
calcula las demandas acumuladas para los mismos peri2
dos y seguidamente las superficies evaporantes corres-
pondientes a un volumen cualquiera vi, tomando como sg
perficie evaporante en un m e s la media aritmética de las
correspondientes al estado inicial y final, y aplicándolo
a la evaporación unitaria mensual.

A partir de estas pérdidas obtiene, para la curva de ga-

rantia dada, las pérdidas totales para cada periodo.

Finalmente, calcula por iteración la curva de seguridad

por meses, según la ecuación

siendo:

Ei (G) = volumen embalsado al principio del m e s (i)

para las curvas de garantia G.

Dik = demanda real acumulada desde el principio

del m e s (i) durante k meses sucesivos.

Pik = pérdidas del embalse acumuladas desde el

principio del m e s (i) durante k meses suce-
sivos.

Aik (G) = aportación del embalse acumulada desde el

principio del m e s (i) durante k meses suce-
sivos, que tiene una probabilidad G de ser
superada.
 185

It evaluates the accumulated affords, and afterwards it

classifies them considering the following periods:

Oct Nov Set

_ _ _ _1 _ _Nov
Oct t
- - -1 -
_ - -1 - -Set
+ +

Ott t k N O t~k Set + k

computing afterwards the accumulated affords corresponcj

ing to a determined guarantee. Lately, it evaluates the -
accumulated demands for the same periods and thereafter
the evaporating areas corresponding to a given volume vi,
taking as the evaporating areas in one month, the arithm-
etical mean corresponding to the initial and final state -
and applying this to the unitary monthly evaporation.

Starting from this losses, the programme obtains for the

given line of guarantee, the total losses to each period.

Finally, it evaluates iteratively the safety line by months,

according to the equation

where:

Ei (G) = stored volume at the beginning of the month (i)

for the guarantee line G.

Dik = actual accumulated demand since the begin-

ning of the month (i) during k consecutive-
months.

k
i
' losses in the reservoir accumulated, since
the beginning of the month (i) during k conse-
cutive months.

Aik (G) = accumulated afford in the reservoir since the

beginning of the month (i) during k consecutive
months, that has a G probability of being over
passed.
186

127 CDSSE Calcula las curvas de seguridad de un embalse, para -

cualquier nivel de garantía de suministro de una d e m a n
da dada en función de la serie histórica de aportaciones
de hasta 60 anos de duración.

It evaluates the safety lines of a reservoir, to any level

of giiarantee of supply of a given d e m a n d in function of
the teorica1 series of affords up to 60 years of duration.

128 REGCV A partir de la serie de aportaciones mensuales en un pun

to, calcula las capacidades de embalse estricto mediante
el método de las diferencias acumuladas, por la expre-
-
sión c = q Aki.

El m i s m o p r o g r a m a distingue dos casos:

a) Regulación a caudal constante

Mediante el programa obtenemos el principio y la
duración del período de vaciado del embalse, del
intervalo de m e s e s sucesivos que da el m á x i m o VE
-
lor positivo a la s u m a de las diferencias ni q Aki,
siendo qi el caudal m i n i m o continuo garantizado -
durante el periodo considerado. Calcula también el
volumen medio regulado en % de la aportación m e -
dia y e n Hm3,el caudal regulado y las capacidades
de embalse estrictas para asegurar estos caudales
en tanto por ciento de la aportación media y en H m 3 .

b) Regulación con caudal variable

E n este caso el caudal para hacer el cálculo de la -
regulación es variable en cada uno de los meses. -
El programa nos da el principio y la duración del -
periodo de vaciado del embalse, volumen medio re
gulado en 70de A m y en H m 3 , capacidades de e m b F 2
se estricto en '7 de A m y en H m 3 y, además, el nu-
m e r o de Has. regables con los volúmenes medios -
regulables.

Starting f r o m the series of monthly affords in a point, the

p r o g r a m m e calculates the capacities of strict reservoir
according to the method of the accumulated differences, -
by the expression C z ni q Aki. -
T h e s a m e p r o g r a m m e distinguishes two cases :
 187

a) Regulation at a constant flow

B y m e a n s of the p r o g r a m m e w e obtain the beginning
and the duration of the period of emptying the resec
voir, the interval of following months which gives
the m a x i m u m positive value of the s u m od differen_
-
ces ni qi Aki, being qi the m i n i m u m continuos -
flow guaranteed during the period under consider-
ation. It calculates also the average regulated --
volume in 70of the average afford and in CU. Hm,
the regulated flow and the strict capacities of --
reservoir to assure these flows in percentage of
the average afford and in CU. Hm.

b) Regulation with variable flow

In this case the flow to perform the calculation of
the regulation is variable in each month. T h e pro-
g r a m m e gives us the beginning and the duration of
the emptying period of the reservoir, averages -
regulated volume in 70of A m and in CU. Hm, c a p a
cities of the strict reservoir in 70of A m and in CU.
Hm in addition, the n u m b e r of H a s irrigables with
the average regulable volumes.

129 REG25 Realiza la m i s m a función que el programa "REGCV",

pero la entrada de datos está calculada para que regule
estaciones durante 25 horas.

T h e p r o g r a m m e performs the s a m e function as the pro-

g r a m m e " R E G V F " , but the entry data is evaluated to
regulate stations during 25 hours.

130 REGVA D a d a una serie histórica de aportaciones mensuales, unos

consumos, una serie de precipitaciones mensuales sobre
el cultivo, unas capacidades de embalse m á x i m o muerto
y dando distintos porcentajes del consumo, calcula las -
variaciones de volumen embalsado, los dé€icits y verti-
dos, después de abastecer los regadios con unos consu-
m o s determinados, a los que descuenta la precipitación
sobre el cultivo. El programa tiene en cuenta la evapora
ciÓn mensual del embalse.
188

Given an historical series of monthly affords, some -

consumptions, a series of monthly rainfall over the crop,
a capacities of m a x i m u m dead reservoir and giving dif-
ferent porcentages of the consumption, this programme
calculates the variations of volume of the reservoir, the
deficits and emptying, them to supply the irrigated land
with a determined consumptions, deducting the rainfall
on the cultivation. It has in consideration also the month-
ly evaporation of the reservoir.

131 REG-RA Estudia la regulación para riegos y abastecimientos de

forma análoga al REGVA.

This programme studies the regulation for irrigation

and supply in the s a m e way to the R E G V A programme.

132 REG-K2 Estudia la regulación conjunta de un sistema de embal-

ses considerando evaporación, para lo que utiliza los -
siguientes datos:

a) Las series de aportaciones en uno, dos tres o c u ~

tro embalses de los que se trata de efectuar una -
regulación conjunta.

b) Los consumos mensuales en Hm3 , suponiendo que

se consume anualmente el 100% de la aportación -
media de cada embalse.

c) La evaporación mensual en cms.

d) Las caracteristicas de los embalses, ecuaciones

de las curvas alturas -volúmenes, superficies-voli
menes, capacidad total y volumen de embalse --
muer to.

El programa realiza entre las aportaciones de cada e m -

balse sorteos equiprobables de 5 en 5 años, hasta 1000
-
años, y a las series de 50, 100, 150 1000 les aplica el
proceso de regulación conjunta, en hipótesis de consumo
de diversos % de la aportación media, dando como res-
tado el 70de fallos en cada serie de anos para cada uno -
de los embalses considerados.
 189

It studies the compound regulation of a system of reser-

voirs considering evaporation, usind the following data:

a) The series of affords in one, two, three or four

reservoirs of which it treats to realize a compound
regulation.

b) The monthly consumption in CU. Hm, supposing -

that one hundred per cent of the average afford is
used yearly in each reservoir.

c) Monthly evaporation in cms.

d) The characteristics of the reservoirs, equations

of the height-volume lines, area-volumes, total
capacity and volume of the dead reservoir.

The programme performs among the affords of each

reservoirs equi-probable casting lots every 5 years
until 1000 years, and to the series.50,100, 150, 1000
the programme uses the compound regulation process,
in the hypothesis of different 70of consumption of the
average afford, given as a result the percentage of -
failures in each series of years for each reservoir under
consideration.

133 REG-SU Este programa estudia la regulación sucesiva de una se-

rie de embalses, sin limitación de número, utilizando las
curvas de regulación del programa anterior y en la hipó-
tesis de que la capacidad de embalse se utiliza para reg5
lar la aportación de la cuenca propia y los caudales no re
gulados aguas arriba, obteniéndose como resultado los -
volúmenes regulados por cuencas parciales y totales.

This programme studies the sequential regulation of a

series of reservoir, without limitation of number, using
the regulation lines of the preceding programme and -
under the hypothesis that the capacity of reservoir is -
used to regulate the afford of its own basin and the non
regulated upstream flows, obtaining as a result the regu
lated volumes by partial and total basins.
190

134 REG-KI Igual que el REG-KI sin embalse muerto.

T h e s a m e as REG-KI, but without dead reservoir.

135 REG-K3 Realiza la m i s m a función que el programa "REG-KI",

EAM calculando directamente varias hipótesis.
RYPJU
RELLO
RESE
EBBE

It performs the s a m e function as the p r o g r a m m e "REG-

K2", evaluating directly several hypothesis.

136 CMAR Dados los valores de abastecimiento, riegos en valor

absoluto y 70,calcula los consumos mensuales.

Given the values of supply, irrigations in absolute value

and percentage, it calculates the monthly consumption.

137 EAM Estudia la explotación de hasta seis embalses, interco-

nectados entre ellos por una red principal de canales de
conducción, que se simula mediante una malla de 24 n u
dos.

Utiliza una serie de aportaciones generadas por sorteo

aleatorio teniendo en cuenta o no la autocorrelación de
las aportaciones anuales.

Los Órdenes de desembalse se establecen en función de

los vertidos probables de cada uno de los embalses y de
la d e m a n d a a satisfacer. Se tiene en cuenta las pérdidas
por evaporación en los embalses y la capacidad de las -
conducciones.
 191

EAM It studies the development of up to six reservoirs, inter

connected among them by a principal net of channels of
conduction, which are simulated by a mesh of 24 knots.

The programme uses a series of generated affords by

fortuitous casting lots having present or no the self-cor
relation of yearly affords.

The orders of emptying are established in relation to the

probable emptying of each reservoir and of the demand
to satisfy. Having present the losses by evaporation in
the reservoir and the capacity of the conductions.

138 RYPJU Este modelo simula la explotación y la producción ener-

gética de un conjunto de aprovechamientos.

Se aplica a un sistema de aprovechamientos (embalses y

saltos hidroeléctricos) situados sobre dos rios en forma
de Y, al cual se pueden añadir caudales regulados en --
otras cuencas o detraer caudales regulados por el sist:
ma.

A partir de una serie de aportaciones generada por sor-

teo aleatorio y teniendo en cuenta la autocorrelación, se
establecen los Órdenes de desembalse en función de las
demandas y de los vertidos probables en cada embalse.

El modelo calcula las producciones en todos los saltos

y la garantia de suministro de la demanda prevista.

This model pretends the exploitation and energetic pro-

duction of an assembly of utilizations.

It is applied to a system of utilizations (reservoirs and

hydroelectric waterfall) located on two rivers in the form
of Y, to which it could be added regulated flows in other
basins or take away flows regulated by a system.

Starting from a series of affords, generated by fortuitous

casting lots and having present the self-correlation,the
order of emptying in relation to the demand and the pro-
bable emptying in each reservoir is established.

The model calculates the productions in all waterfalls

and the guarantee of supply of the calculated request.
192

139 RELLO Este modelo simula la explotación coordinada de los re-

cursos superficiales y subterráneos.

Supone la existencia de un embalse subterráneo del que

se puede extraer un caudal uniforme prefijado, en fun-
ción de los estados de los embalses del sistema.

Utiliza una serie de aportaciones generadas por sorteo

aleatorio, teniendo en cuenta la autocorrelación de las
aportaciones anuales y las curvas de seguridad de un -
embalse equivalente a la suma de los embalses del sis-
tema, determinadas mediante el programa CDSSE.

L a explotación se simula teniendo en cuenta las pérdidas

por evaporación en los embalses y obtiene la garanda de
suministro de la demanda de abastecimiento junto con -
los valores de la extracción anual media del acuifero y
del periodo de máxima duración de la extracción máxi-
m a prevista.

This model pretends the coordinated exploitation of su-

perficial and underground resources.

It assumes the existence of an underground reservoir

from which a uniform flow can be extracted fixed in ad-
vance, depending on the state of the system of the resec
voir.

It uses a series of generated affords by fortuitous casting

lots, having present the self-correlation of the yearly
affords and the safety lines of a reservoir equivalent to
the s u m of the system, which is determined by the pro-
g r a m m e SAFLI.

The exploitation is simulated, having present the losses

by evaporation in the reservoir, the guarantee of supply
of the demand, together with the values of the annual -
average extraction and of the period of m a x i m u m calcu-
lated extraction.

140 RESE Este modelo simula un sistema de explotación con varios

embalses situados sobre una misma corriente, uno de -
los cuales puede ser el origen de un aprovechamiento -
hidroeléctrico.

La explotación se establece a partir de una serie de apor

taciones generadas por sorteo aleatorio y teniendo en -
 193

cuenta la autocorrelación de las aportaciones anuales,

en función de las curvas de seguridad de un embalse tg
tal equivalente para atender a una demanda de usos cog
suntivos. Al mismo tiempo que calcula la garantía de s u
ministro de la demanda prevista obtiene las produccio-
nes históricas en todos los saltos, distinguiendo la enec
gi’a de puntas y la energía producible en horas llenas en
el periodo critico (nov. a feb. )de máxima demanda ener
gética. Ordena los valores de la energía optenidos y as{
puede suministrar los valores de la energia de distinta
calidad garantizada en el per<odo a que se extiende la si
mulación.

This model pretends a system of exploitation with several

reservoirs located above the same current, one of which
could be the origin of a hydroelectric use.

The exploitation is established starting from a series of

generated affords by fortuitous casting lots and having-
present the self-correlationof the annual affords, in -
function of safety lines of a total equivalent reservoir to
attend to a demand of utilizations, At the same time evs
hates simultaneously the guarantee of supply of the cal-
culated demand, it obtains the historical productions in
all waterfalls, distinguishing sharp end energy and the
energy producible in full time in the critical period (nov.
-
to feb. )of m a x i m u m demand of energy. It orders the
values of the obtained energy and in this way it can sup-
ply the values of the energy of different quality in the-
simulation period.

141 EBBE Este modelo realiza el estudio de regulación de una cuen

ca a partir de una serie de aportaciones generadas por -
sorteo aleatorio teniendo en cuenta la autocorrelación de
las aportaciones anuales y en función de los vertidos pro
bables de los embalses previstos calcula los valores de
los trasvases posibles al mismo tiempo que la garantia
de suministro de las demandas previstas y las producci2
nes y consumos energéticos en los aprovechamientos hi-
droeléctricos de la cuenca.

This model performs the study of regulation of a basin,

starting from a series of generated affords by fortuitous
casting lots, having present the self correlation of the
annual affords arid in function of the probable emptying
194

of the calculated reservoirs. It calculates the values of

the possible overflows at the same time that it calculates
the guarantee of supply of the calculated demands and the
energetic productions and consumptions in the hydroelec
trical utilizations of the basin.

142 LAMI 1 Estudia la laminación de un embalse, supuesto un nivel

inicial determinado y pudiendo utilizar uno o varios si-
temas de desagües, en función de las caracteristicas -
del embalse y de la crecida.

It studies the lamination of a reservoir, supposing an

initial level determined in advance and being possible
the use of one or several drainage systems, in function
of the characteristics of the reservoir and the flood.

143 HIDR 1 Calcula el hidrograma para diversas hipótesis de inten-

sidad horaria de precipitación, coeficiente de escorren-
tia y duración de la tormenta.

It evaluates the hydrogram for differenthypothesis of

hourly intensity of rainfall, runoff coefficient and the
duration of the storm.

144 HIDR 2 Calcula el hidrograma con intensidad y coeficiente de

escorrenti’a corriente.

It computes the hydrogram with normal intensity and

usual runoff coefficient.

145 ABC Realiza el estudio económico (análisis, beneficio y cos -

to), expresándolo en forme de corrientes monetarias a s
 195

tualizadas en función de la tasa de descuento y obtiene

la ratio
Beneficio - Gastos
costos

It performs the economic study (analysis, profit and

price) expressing it in actual monetary currency in func-
tion of the standard rate of deduction and it obtains the
ratio
Profit- Expenses
Prices

146 ABC 10 Programa ABC para estudio económico de varias cen-

trales.

It programmes ABC for an economical study of several

centrals.

147 A B C TV Programa ABC para estudio económico, con tasa varia-

ble.

It programmes ABC for an economical study, with varia

ble standard rate.

148 BNZ Dado el número de muestra, tiempo de lectura y número

de desintegración, obtiene la concentración de Tritio pa-
ra muestras de agua.

Given the number of the sample, lecture time and n u m -

ber of disintegration, it obatains the concentration of -
Tritium for sample of water.
196

149 PARAM Calcula los siguientes indices : R M G / R C A , R N A / R K ,

R N A / R C A , A N A / R M G , (RCA-RMG), RALC, BR/CL,
R C L / R H C O 3 , R H C O 3 / R C L , RS04/RCL, (RCA+ RHCOQ)/
/ (RCH RS04), R H C 0 3 / R H C 0 3 t RS04 t RCL,R A L C / R C L ,
+

R C A / R A L C , SAR, ICB, ID, FI, Tipo de agua. Además,

lista los elementos en meq/l. y en 70.

It calculates the following indexes: R M G / R C A , RNA/RK,

R N A / R C A , A N A / R M G , (RCA-RMG), RALC, BR/CL,
R C L / R H C 0 3 , R H C 0 3 / R C L , RSOq/RCL, (RCA t RHCOS)/
/ (RCHt RS04), R H C 0 3 / R H C 0 3 + RSO4 + RCL,R A L C /
/ RCL, R C A / R A L C , S A R , ICB,ID,FI, Type of water.
Besides, it lists the elements in meq/l. and in %.

150 HISTO Calcula los histogramas de las siguientes relaciones,

clasificándolos en clases y valores fuera de clase:
RCL/RS04,R C L / R H C O 3 , R A L C / R C L , RNAIRCA, RCL,
RSO4, R H C 0 3 , RN03,RALC, Res. seco, T.D.S.,D u -
reza total. El histograma lo dibuja por impresora.

It calculates the hystogram of the following relations,

classifying them in classes and values out of class:
RCL/RS04,R C L / R H C O 3 , R A L C / R C L , R N A / R C A , RCL,
RSO4, RHC03,RN03,R A L C , Res. (dry), T.D.S.,total
hardness. The hystogram is designed by the printer.

151 TUBEC Dado un muestrario de tubedas de diferentes diámetros,

con sus precios y caracteristicas hidráulicas,, determi-
na para una configuración topológica y topografica de la
red y para diversas hipótesis, la combinación de distri-
bución de tubos m á s económica que permita el suminis-
tro solicitado con la minima pérdida de carga.

Given a sample book of pipes of differents diameters,

with their prices and hydraulical characteristics, it
determines for a topological and topographical form of
the system and for several hypothesis, the combination
of distribution of pipes more economical, that allow the
solicited supply with the minimum loss of loading.
 197

152 CANAL Definido un canal por sus secciones y pendientes en difg

rentes tramos, as: como por distintos tipos de cornpuer_
tas, el programa determina la evolución de los caudales
transportados en el curso del tiempo, as: c o m o los cala
dos alcanzados en los distintos tramos del canal.

Permite estudiar las maniobras de apertura y cierre de

compuertas m á s convenientes para la explotación del cg
na1.

Defined a canal by its sections and pendants in different

stretchs, as well as by different types of floodgates,-
this programme determines the evolution of the flows
carried in the course of time, as well as soakage reach-
ed in the different stretchs of the canal.

It allows also to study the process of opening and closing

of floodgates more convenient to exploitation to the canal.

153 SER-EL Depura los datos suministrados por las empresas hidro-
eléctricas relativos a la producción mensual de energi'a
de los diferentes saltos de cada una, recogidos en tarje
tas perforadas. L a depuración se hace verificando la -
concordancia de los datos geográficos, número de horas
de utilización de los controles en función de la potencia
instalada y producción.

Una vez corregidos todos los errores detectados, pre-

para unos cuadros resúmenes estadisticos de producción
de energTa, clasificados por diferentes conceptos :

Producciones globales por cuencas hidrográficas en ca-

da mes.

Producciones globales UNESA, IN1 y otr os en cada mes.

Producciones anuales clasificadas por:

Empresa o concesionario.

Centrales por magnitud de su producción.

Centrales y cuencas por magnitud de su produc-

cion.

Centrales y rios por magnitud de su producción.

Centrales y provincias por magnitud de su produc-

ción.
198

It purifies the supplied data by hydroelectrical companies

relatives to monthly productions of energy of the diffe--
rents waterfalls of each one, collected in perforated cards.
The cleansing is done verifying the harmony of geographi-
cal data, number of hours of utilization of controls in -
function of installed power and production.

Once corrected all the detected errors, it prepares a

summary of statistical charts of production of energy,
classified by different ideas :

Total productions for hydrographical basin in each month.

Total productions UNESA, INI, and others in each month.

Annual production classified by:

Company or dealer.

Centrals by magnitude of production.

Centrals and basins by magnitude of production.

Centrals and rivers by magnitude of production.

Centrals and provinces by magnitude of production.

 COMPUTATION OF RESERVOIRS SEDIMENTATION

A.V. Karaushev, I.V. Bogoliubova

State Hydrologic a 1 Institute
Leningrad, USSR

--
ABSTRACT

Methods for the computation of sedimentation by suspended

sediments and bed load of the projected reservoirs are given,
or the first year Of the reservoir operation computation is
made according to the balance of sediments computed by the
difference between the transport capacity and the hydraulic
parameters o f the current at the upper pool (transient region)
and at the dan of the reservoir. The subsequent attenuation of
the process as well as the total duration o f sedimentation is
evaluated by empirical relations obtained from the observational
data on reservoirs under operation.

RESUME

Les auteurs exposent des méthodes pour le calcul de

l'envasement des barrages par les matériaux transportés en
charriage o u en suspensión. On fait, pour la première année
d'exploitation, le bilan des matériaux déposés dans la retenue
par différence entre ce qui entre à l'amont (station de mesure
dar,s la z o ~ edu remous) et ce qui sort par le barrage; ces
mesures s o r t reliées aux paramètres hydrauliques du cours d'eau.
u', extrapole les résultats dans le futur en utilisant des
relations empiriques obtenues pour d'autres réservoirs en cours
!1 exploit a t ion.
200

The construction of reservoirs in mountain areas and at the

foothills on rivers with a considerable sediment concentration
inevitably faces with the necessitg to remove or to impede sedi-
ments transported by the river to keep the projected capacity
of the reservoir. The present paper gives methods accepted in
the USSR providing the evaluation of possible sedimentation rate
for the whole reservoir or for its individual parts during the
first year of its operation and for subsequent years.
Methods for the computation of re semoirs sedimentation are
based on the equation of sediments balance applied to the whole
reservoir or its parts, to the gross composition of the transport-
ed sediments or its particular fractions. The use of this equation
makes it possible to compute the difference between sediments in-
flow and its discharge out of the reservoir i.e. sediments
accumulation. The inflow of sediments is computed by observational
out of the reservoir
data or by indirect methods. The discharge of sediments is estim-
ated by equations of the transporting capacitg of the current at
the specified values of water discharge Q, mean depth Hm, mean
current velocity Vm, and granulometaAc sediment composition.
The computation of sedimentation during one year is reduced
by the determination of that portion of sediment discharge
which is accumulated in the reservoir. When starting computation
it is essential to establish design values of annual water
discharge, of suspended sediments and bed load, as well as typi-
cal chronological graphs of these values for the inflow site of
the reservoir. It is recommended to divide the hydrograph of the
typical year into 3 or 4 design time intervals& and to compute
$
sediments accumulation according to the values o Q, Vm Hm, etc.,
averaged for every time interval. The computation of seAimenta-
tion rate is made by individual fractions i, in this case it is
sufficient to subdivide all the transported fractions into 3 or
5 categories. Then sediments are summarized according to all
categories of the fractions.
The computation of sedimentation by suspended fractions for
a design interval A tj is niade by equation:

where: Pa j is the amount of sediments of all the fractions

(tons) in the reservoir or in the design area dur A t ;
*i in J is inflow of sediments of the i-th fractio3tonsg
d u r m g time ~t through the initial (upper) discharge site of
the reservoir $ or its part)determined by the chronological
graph or by computation d as made for the upstream area; Qter
is mean water discharge ( sec) for time a t through the
terminal (downstream) discharge site of the Jeservoir (at the
dam) or the design area; Si t is mean particular turbidity
for time A t . of the i-th fractfoi at the terminal discharge site
of the resehoir (certain area) (g/m3); ~ t is j time interval
(sec).
 202

Turbidity of the i-th fraction at the terminal discharge

site Si ter j is computed by equation of A.T. Karaushev:
- G"AL
(2)
where: Si 4" J Ois particular turbidity at the initial discharge
site mean or time interval B ta; S? is the turbidity
corresponding to a particular thins$of%idg capacity of the current
computed by equation (6) given below; e is the base of natural
l2gari t h ;
G is dimensionless value determined by equation

where: ui is fall velocity of fraction i under consideration;

kg is a parameter having a dimensionality of velocity and which
is computed by equation
LL; ri
The value of r
of sediments whichi=
(4)
is the value of hydroneclnanic param?.t;er
be obtained by graphs according to the
Chezygs coefficient C and ratio of i*
(Fig. 1 ).In equation (3)
AL indicates the length of the reservoir (or it8 part) given
In relative units:
- =---- A L
AL
H m (5)
where: A L is the length of the reservoir (or design area (m);
IL, is mean depth of the reservoir (or some area), (m) for time
iIltel?ral A t * o
A particulAr transporting aapacj. of the current :S tr
(for the i-th fraction of
-
i8 C O uted with the %se of
data on bed load composition. The value of 8tr j is computed
with the use of hydraulic elements of the current mean for time
interval A tj related to the whole reservoir or its d e s i m area:

Here a indicates a correcting factor estimated by the ratio of

actual and computed turbidity at the initial discharge site:

a =-SS4.
- r
s cokvlvip (7)
2 02

d is composition (in per cent) of the i-th weighted

f&&&o& in the roiling portion of bed load.
The value 0% droil i is determined by the ratio

-- IC0 ,
bed i
(8)
roil i - t CL

where A is the portion (per cent) of the i-th fraction in

bed load &&&sition; r is gross portion (in per cent) of the
weighted fraction in bed load composition. In this case sediments
wPth fall velocity corresponding to the condition u 4 1- are
regarded as weighted fractions, and Vi indicates maxmum value
of the vertical component of the puls%on velocitg. The latter
salue is computed by a special equation according to mean velocity
of the current and Chew's coefficient. Gross turbidity of roiling
(Sroil) is obtained by equation:

where: N is characteristic dimensionless number depending on

Chew's eoefficient C; is the ratio of velocity at the bottom
!
go mean veloci%y; the est of the symbols are given in previous
equations, *
When comguting.Si tr for the first time interval the composi-
tion of bed load xn the reservoir is accepted according to the
averaged conposition of river alluvium in the cñannel and in the
f lood-plain; for subsequent intervals it is essential to consider
the composition of sediments obtained by computations.
If sedimentation is computed for certaih areas, then Si 8'
estimated by equation (2) for the first (upper) area 1s use as
j
Si in j for the second area domstream, etc.
The computation of reservoir sedimentation by bed load is made
according to the same design intemals a8 by suspended sediments.
For an approximate evaluation it i8 possible to be confined to
the computation for flood periods when the major portion of coarse
fractions flows into the reservoir.
The amount of bed load in the reservoir is determined by the
difference:

(10)
where: Pa bed is the weight of bed load in the reservoir (tons);
n t.J is time interval (sec); Rbed in and Rbed ter jindicate
bed load discharge at the initial and termi.mil discharge sites
(kg/sec) nean for design time intervalat.. For bed load discharge
computation it is reasonable to recommed $he equationsof G.I.
Shamov, V a N e Gomharov, I.V. Egiaearov, K.I. Rossinski, et al. The
equation of G.I. Shamov is the most simple one providing a suffici-
 203

ent coincidence of computation results with the data of measure-

ments at a wide range of fractions dimensions:

where: Rbed is bed load discharge (kg/sec); B is current width

(m); Hm is mean depth (m); a, is mean diameter of mobile particles
of bed load (m); V m is mean velocity of the current (m/sec); vsed
is mean velocitg of the current (m/sec) when fractionswith Q
in diameter stop moving; K is coefficient consider- non-homo-
geneitg of bed load composition.
The value of is computed by equation:
d =c~oíz.i.ul!
WL
WL
i=/ L '
(12)

where: d, and d. res ectively indicate percentage and mean

diameter o) a certain ?i-th) fraction; summation is made accord-
ing to all the mobile fractions, the number of these fractions
is-indicated as m.
Separation of immobile (coarse) fraction is made by
equation: 3

- 9,o12 'yl -
4,t - L-i H, (13)
The value of Vsed is obtained by equation:
fis
=
3,) d, H ~
(14)
All the computations of bed load transport are made accord-
ing to hydraulic elements mean for time interval A tj.
Annual accumulation of all the sediment fractions for the
first year of reservoir operation is determined by equation:

where: Pai is gross sediment weight for the 1st year (tons);
and P a bed j respeCtiVely indicate the W i g h t of suspend-
88 g%hde.ents and bed load for the 1st year (tons)j j is the number
of design internali n is the number of intervals during a year.
If computation 1s made according to certain areas, then
summation is made for all the areas to obtain gross sedimentation
of the reservoir. The obtained value for the whole of the reservoir
is transformed into volumetric units:
204

where: Wai is the volume of sediments during the 1st ear (m3);
P is the weight of sediments for the 1 t year (tons5; fs is
&e volumetric weight of sediments (t/m )3 .After the computation
being done the initial volume of the reservoir W is corrected.
W, W -
The obtained volume of the reservoir at the end of the 1st year
= Wa is used for the computation of sedimentation for the
next year. $he computation of sedimentation for subsequent years
may be performed in the same way as for the 1st year or by
extrapolation equations considering time attenuation of sedimenta-
tion.
It is recommended to use the equation of G.I. Shamov for the
computation of chronological variations of sedimentation:

where: Wa t is the volume of Sedimentation (m3) in t years;

M'ad is sedimentation volume during the 1st year (m3) computed
by the methods described above; Wa ext is the extreme volume
of sediments in reservoir (m3) approximately computed by
equat
- ion:

where: W is the initial volume of the reservoir (m3); Wtis

area of river cross section (m2) when dischar e is close to
maximm;Ldp is the maximum cross section area fm2) of the upper
pool near the dam.
The method is applicable to the reservoir as a whole.
One year should be accepted as a design time interval. In
case of correct initial values the method provides variations
of reservoir sedimentation close to actual.

R E F E R E N C E S
1. Bogoliubova I.V. Resultam polevykh issledovaniy i rascheta
stoka vlekomykh nanosov r. Mzymty (The results of field
investigations and bed load discharge computation for the
Mzymta river) Trans. of the State Hydrological Inst.,
1968, ~01.156, PP* 3943.
.
2. Karauschev A .V Rechnaya gidravlika (River hydraulic 8).
Leningrad, ñydrometeorological Publishing House, 1969,
PP. 303-778.
.
36 Razumikhina K.V Primenenie formuly transportiruyushchei
sposobnosti dlia rascheta godovogo stoka vzveshennykh
nanosov (Application of transporting capacity equation
for the computation of annual discharge of suspended
sediments). Trans. of the State Hydrological Inst., 1969,
vol. 175, ppe 137-154.
4. Ukasania PO raschetu eailenia vodokhranilishch pri strof.t;el-
nom proektirovanii (Instructions for the computation
of reservoirs sedimentation for engineering projects).
Leningrad, Qdrometeorological Publishing House, 1968,
54 PP.
5. Shamov G.I. Recbnye nanoqy (River sediments). Leningrad,
Hydrometeorological Publishing House, 1959, pp. 2-282.
 205

Figure 1. - Computation of reservoirs sedimentation

 C A L C U L A T I O N O F R U N O F F IN I R A Q
R.K. KLIGUE, M E C H D I E L S A C H O B

ABSTRACT

T h e r e g i v e n a g e n e r a l c h a r a c t e r i s t i c o f r u n o f f in
I r a q and i t s d i s t r i b u t i o n w i t h i n t h e a r e a o f t h e country.
The a u t h o t s a n a l y s e c o r r e l a t i o n o f r u n o f f with e l e v a t i o n
o f the t e r r i t o r y , r i v e r b a s i n a r e a and o t h e r f a c t o r s w i t h
t h e aim o f using t h e s e r e l a t i o n s h i p s f o r r e g i o n s w i t h o u t
r u n o f f data. H y d r o l o g i c t i m e s e r i e s a n a l y s i s o f r u n o f f
and a n a l y s i s o f r u n o f f f l u c t u a t i o n s t h r o u g h t h e t e r r i t o r y
a r e cited.

RES UME

On d o n n e l a c a r a c t é r i s t i q u e g é n é r a l e d'écoulement
f l u v i a l en Irak et s a d i s t r i b u t i o n p a r l a t e r r i t o i r e . S o n t
a n a l i s d e s l e s r e l a t i o n s e n t r e é c o u l e m e n t f l u v i a l , l'altitude
de l i e u , l'aire de s u r f a c e r é c e p t r i c e e t d'autres facteurs.
L'objectif d e c e t t e a n a l y s e d'est u t i l i s a t i o n de c e s
r e l a t i o n s p o u r l e s r é g i o n s a v e c l'absence d e s d o n n é e s
d'écoulement. O n fait l'analyse d e s s é r i e s h y d r o l o g i q u e s
et d e s v a r i a t i o n s de d é b i t s p a r l a t e r r i t o i r e .
208

The investigation of river flows in Iraq is of great

importance keeping in view the constantly increasing water
balance stress in the country that arouses the necessity to
design special rnultipurpose projects as well as projects
aimed at more complete use of water resources by construct-
ing irrigation systems, overyear storage reservoirs, by im-
proving crop management under irrigation and by wider use
of groundwaters.
The main rivemof Iraq - the Tigris and Euphrates -
cross the oountry by their middle and lower reaches. Con-
fluencing they form a river called Shatt-al-Arab flowing
into the Perbian Gulf. The main tributaries of the Tigris
within Iraq are the Greater Zab, Lesser Zab, Adhaim and
Diyala. The Euphrates river have no tributaries on the ter-
ritory of the country. The arid regions are characterized
by the existence of "wadi".
Water resources of Iraq are mainly determined by the
flow of the Tigris and Euphrates rivers making about 77.7 km3

a year - about 22.2 km3 flows into the sea and 55.5 km3
(71.4%) is used for irrigation, municipal and industrial
water supply and power generation. A considerable part of
flow is lost due to evaporation, transpiration and filtra-
tion.
The mean annual flow of the Euphrates on entering the
territory of Iraq is 928 cumecs decreasing downstream (Na
\I
-
siriya) to 454 cumecs. Thus, the rate of flow changes along
the river length from 3.52 to 1.57 1/s. k
m
'
.
The Tigris river on the territory of Iraq has several
 209

big tributaries, which increase its flow from 587 cumecs

at Tusan to 1534 cumecs at Salman-Pak. Downstream the flow
decreases due to intensive withdrawal for irrigation and
makes 49.6 cumecs.at Qalat-Saleh. The mean annual rate of
flow in the Tigris basin changes from 12.7 l/s.km 2 in the
upper part to 0.26 l/s.km 2 (Qalat-Saleh) in the lower
part .
The coefficient of annual flow variation (C,) for the
Tigris and Euphrates changes from 0.26 to 0.31 decreasing
with the altitude of watershed (Haam, m). It can be ex -
pressed by a correlation:

The flow of the Tigris and Euphrates is distributed

very uneven through a yew' - the greater part of it falls on
the flood period (April-May), making about eo"/4 in the upper
reaches and 5% - in the lower reaches.
The beginning of flood period is closely dependent on
the mean altitude of watershed m) and may occur in
January-April. The mean duration of flood period in the
piedmont regions is 45 days, in the middle mountain regions -
90 days and in high mountain - 135 days. The more is the
water availability through the year, the longer is the flood
period. Por the year with mean water availability the
duration of the flood period (I, days) may be calculated
by the equation:

T = Hmean - 20.6
14.4
210

In Icaq the maximum flow of rivers occurs due t o snow

melting and rain. The rates of maximum daily flows of the
Tigris river and its tributaries vary from 220 l/s.km2 (the
Khaair) and 135 l/s.km 2 (the Greater Zab) in the mountain
2
regions to 1.6 l/s.hm2 (the Tigris-Amara) and 0.71 l/s.hm
(the Qalat-Saleh). in the south of the Mesopotamian lowland.
The rates of maximum daily flow of the Euphrates river vary
from 13.3 l/s.km2 (Hit) to 8.5 l/s.km 2 (downstream the dam
Hindiya). The rates of maximum monthly flow of the Tigris ri-
2 (the Greater
ver and its tributaries vary from 48 l/a.km
2
Zab river at Eski-Kalek) in mountain regions to 0.6 l/s,km
(the Tigris river at Qalat-Saleh) in the south lowlands. For
the Euphrates river the rate of maximum monthly flow fluc -
tuate from 8.6 l/s.km 2 (Hit) to 4.0 l/s.km 2 (Nasiriya). As
I rule there is observed an increase in the maximum rates of
flow throughout the mountain regions a
In the mountain regions of Iraq,above 2000 m (the Tigris,
Greater Zab and Euphzlates river basins) the rate of maximum
daily river flow increase from 8.5 to 136 l/s.km2, making
the mean about 35 l/e.km 2 for every 100 m. For this region
a certain dependence between maximum rates of flow (%ax)
and mean annual rates (Mme= an. )is traced.
 211

In the middle mountain region with the altitudes 1000-

2000 m (the Lesser Zab and Diyala basins) the rates of maxi-
2
mum daily flow fluctuate within the limits 48-112 l/s.km ,
increasing averagely by 38 l/c,km2 on every 100 m of alti-
tude* The dependence of maximum and mean annual flows for
this region is expressed by the correlation

- 45.8
%ax äaiïy 0.138

In piedmont regions, lower than 1000 m (AdhaFm and

Khazir river basins) the rates of maximum daily flow being
inversely proportional to the altitude of watershed increase
from 48 to 220 l/s.km2. The phenomenon is observed in the
case of the maximum monthly flow. This can be explained by
the fact that the Khazir river has the lesser, compared to
the Adhaim basin, area of watershed but receives greater
rainfall. The dependence between the maximum and mean annual
flows is expressed by the correlation.

%ax monthly = 2.5 %ea* an.

The coefficients of maximum daily flow variation for

the rivers of Iraq fluctuate within the limits from 0.14
(the Tigris-Amara) to 0.74 (the Adhaim-Injana). The coeffi-
cients of maximum monthly flow are limited by 0.18-0.29. There
observed the inverse proportion between coefficients of va-
riation and the altitude of watershed. The correlation of
coefficients of maximum monthly flow variations (Cv max 1 and
the coefficients of annual flow variations (Cv an. )can be
put as: v' max monthly =(2*1 v' -
an. ) 0.21
212

For the rivers of Iraq it is characteristic the increase

of maximum flow with the decrease of the watershed area
(F b2),
Barnax daily = 62 - 0.2 F + -9
Minimum flow usually occurs in autumn (September-Octo-
ber) and is caused by the groundwater depletion by the end
of the hot and dry period. Small rivers in piedmont regions
dried up as early as the beginning of summer and till the
winter rains they have no flow.
One of the most important factors of the Iraq rivers
regime is the low-water period, when the river flow is cha-
racterized by stable low levels and discharges and when the
rivers under the condition of great reduction of surface flow
or its complete cessation are recharged through groundwaters.
Low-water period usually occur in summer or autumn (oftener
from June to December). Its beginning (t days from the begin-
ning of the year) has a certain dependence on altitude,

days = 64.6 (jy,ean-lOOO)OSI75 -K

+

where, K - coefficient of water availability of the designed

period. It can vary from + 20 (for high-water period) to -20
(for low-water period). At the mean level it approaches zero.
As a rule the greater is the altitude of a watersheii, the
shorter is the period of low water. In piedmont zones it
makes averagely 171 days, in middle mountain zones - 159 days
and in high mountain - 153 days. The latter is mainly due
to more favourable conditions of humidification in the moun-
tains where the rainfall reaches 1000 mm than in piedmont re-
gions where the rainfall makes 200-300 m.
 213

The rates of minimum flows of Iraq rivers have a wide

range of variations from 0.04 (daily) and O.Oí'(monthly) for
the river Adhaim (Injana) to 5,56 (daily) and 5.71 (monthly)
l/s.km2 for the Greater Zab river (Bekhma). Usually the
rates of minimum daily and monthly flows increase with the
altitude and this relationship can be expressed ass

'L$in = 5.8 IO-.10 3 J

On the territory of Iraq there observed quite a defi-

nite reduction of the minimum flow rates with the increase
of a watershed area. For example, on the Tigris river at
2
Al-Fatha the minimum monthly flow equals 2.95 l/s.km , the
2
watershed area being 1076b0 km and downstream the Kut bar-
rage the flow -
1.35 l/s.km 2 and the watershed area -
177540 b2.For the Greater Zab and Tigris river basins
(high mountain zone) the relationship will be similar,

%in monthly = 6.35 - 3.14 IÕ7F

%in daily = 6.14 - 3.08

The decrease of the minimum flow rates with the in -
crease of a watershed area can be explained mainly by great
withdrawals of water for irrigation and due to considerable
evaporation.
The mean annual flow of Iraq being studied better than
the minimum one their interrelation may arouse certain in-
terest
'min monthly = '
O
s
4 'mean an.

%in daily = O*** mean an. 0.15

214

These relationships do not reflect the conditions

in the lower reaches where the regime of the rivers is great-
ly distorted under the influence of anthropogenic factors.
The variations in the flow of Iraq rivers (mean, mini-
mum and maximum) occur principally at the same time. For the
investigation of the minimum flow variation of the Tigris
river (at Mosul, 1920-1970), the Euphrates river (at Hit,
1932-1970) and the Greater Zab river (1938-1970) they used
the method of the differential integral curves permitting
to bring out the succession of low-water and high-water
groups of years in the considered period. The duration of
low-water periods with minimum flow (mean coefficient of wa-
ter availability 0.78) can change within 1-18 years (the Tig-
ris river at Mosul) and high-water periods (mean coefficient
of water availability 1.24) - within 1-11 years. The coeffi-
cients of variation (C,) for minimum monthly and daily river
flows fluctuate from 0.18 (monthly, the Tigris river at Al-
Fatha) and 0.17 (daily, the Euphrates river at Hit) to 1.27
(monthly,the Aähaim river at Injana) and 2.03 (daily, the
Adhaim river at Injana). The value of the coefficient of va-
riation is inversely proportional to the altitude which is
explainable by a considerable aridity of lowland territories
in Iraq.
The values of Cv for minimum flow, definitely correlate
to Cv for mean annual flow. For the watersheds where the
withdrawal of water for irrigation is low this relationship
can be expressed by the following empirical equation,

v' min daily = 1 0 ~v'~ mean

5 an. - 0.16
 215

In Iraq because of the drying up many rivers have no

flow for a considerable period (the Adhaim, Al-Wend, Galal-
Bedrah, Wadi-river, etc.). The watershed areas of drying-up
2
rivers may reach 13000 km (the Aàhaim river) and the dura-
tion of a drainless period exceed 250 days. In Iraq, especial-
are
ly, in its south-western part there a numerous strema of
temporal nature, "wadi", which have flow only several days
a year. The length of some of them reach many tens of kilo-
meters. This phenomena is a result of the extreme aridiky of
the region where they are met (the rainfall is less than
100 mm and evaporation - over 2500 mm).
It should be noted that the given relationships of &if-
ferent flow characteristics on the territory of Iraq despite
their approximate nature and the necessity of further preci-
sion, permit to give duly evaluation of a number of flow pa-
rameters for the insufficiently studied regions of the consi-
dered territory.
 D E T E R M I N A T I O N O F E V A P O R A T I O N IN CASE O F T H E
ABSENCE O R I N A D E Q U A C Y O F DATA

P.P. Kuzmin, A.P. V e r s h i n i n

State Hydrological Institute
L e n i n g r a d , USSR

ABSTRACT

The possibilities f o r the determination of evaporation

f r o m w a t e r s u r f a c e and l a n d are g i v e n in c a s e o f t h e a b s e n c e
o f d a t a o f d i r e c t e v a p o r a t i o n m e a s u r e m e n t s . T h e a n a l y s i s and
classificatiqn o f methods for the computation of evaporation
are p r e s e n t e d . P r a c t i c a l r e c o m m e n d a t i o n f o r t h e d e t e r m i n a t i o n
o f e v a p o r a t i o n by m e a n s of s t a n d a r d o b s e r v a t i o n a l d a t a f r o m
hydrometeorological stations are given.

L e s a u t e u r s e x a m i n e n t l e s p o s s i b i l i t é s d'évaluation d e
l'évaporation des s u r f a c e s d'eau l i b r e lorsqu'il n'existe p a s
d'observation directe. I l s a n a l y s e n t l e s d i f f é r e n t e s m é t h o d e s
u t i l i s é e s et e n p r o p o s e n t u n e c l a s s i f i c a t i o n . I l s font d e s
r e c o m m a n d a t i o n s p o u r l'évaluation d e l'évaporation 'a p a r t i r
des o b s e r v a t i o n s s t a n d a r d s e f f e c t u é e s d a n s les s t a t i o n s
hydrométéorologiques.
218

The determination of evaporation under natural conditions

is of great importance for the ewtimation of the present and
future water resources, Por water resources management and
for the solution of various theoretical problems in the field
of hydrology and meteorology. Methods of direct evaporation
measurements under natural conditions are still being developed,
therefore computations are the main source of information.
The existing computation methods might be subdivided into
three groups. The first group comprises the methods based on
the physical analysis of the evaporation process. The second
group (combined ox complex methods) includes methods based on
physical principles combined with semi-empirical constants which
can be determined with the help of accurate measurements of
actual evaporation in representative regions.
Methods based on the statistical analysis using only
empirical relations, where empirical constants and coefficients
are h i g w variable and depend on meteorological conditions,
make the third group.
Besides, according to the basic data (factors) included i-nto
the design schemes, it should be noted that computation methods
may be complex and simple as well as difficult and easy to be
applied in practice. In this respect the most simple and
practicable methods are those of the third and some of the second
groups, while the methods of the first group which are based
on the physical analysis, are most inconvenient in practice.
The first group includes the well-known methods of estima-
tion of evaporation from heat balance eqqtion, water balance
equation and turbulent diffusion method /ll/; the accurate
solution of these equations cannot be obtained because it is
impossible to estimate with sufficient degree of accuracy some
individual components of the above equations.
The estimation of the turbulent heat exchange between the
underlying surface (water or land) and the atmosphere is one
of the difficultiee of the solution of heat balance equation.
This component can be estimated approximately without conside-
ration of temperature stratification and horizontal gradients
of turbulent heat exchange (advection).
In particular, in the course of estimating evaporation from
the reservoir surface it is difficult to determine time
variation8 of the heat accumulated by the reservoir (heat
content) as well as heat income and losses due to all kind8
of water inflow and outflow (both surface and subsurface).
Therefore this method is applied only in research studies.
Heat b8Lance equation of the land surface is more complicated
than that of the water surface /U/#
In the 'USSR, however, 8 method of estimating evapotranspira-
tion has been developed and is widely applied,from thefollowing
equation /14/:
ß-B
(1)
 219

which is deduced of the heat

balance of the land with the
account of Bowen ratio:

Here: E is evapotranspiration, R is the measured value of the

radiation balance of the surface, B is heat income into %he
soil, II is the atmospheric pressure, P is turbulent heat
exchange with the atmosphere, C is heat capacity under
constat pressure, L is the latgnt heat of evaporation;
at and a i are respectively the differences in temperature
and water vapour pressure measured at two levels above the
ground.
Equation (l), naturally, would rather belong to the second
group of methods than to the first one; it does not include
horizontal gradients of turbulent heat exchange (advection)
and temperature stratification. It can be applied for homo-
geneous areas large enough to ensure wind r u over homo-
geneous top cover over plain area at the distance of 300-400 m.
The relative standard error of 10-day and monthly evapotranspi-
ration sums estimated from equation (1) for the re ions of
natural moistening and for irrigated fields makesf 1%.
Equation (1) cannot be recommended for the estimation of
evapotranspiration in very dry regions (semi-deserts, deserts)
Full water balance equation is not applied in practice
.
since it is both difficult to determine water exchange with
the bed of reservoir (the difference between underground
water inflow and outflow in a reservoir) while estimating
evaporation from the water surface and to determine water
exchange between the upper layer of the aeration zone and the
underlying ground (upward and downward streams of moisture in
the ground) while estimating evapotranspiration from the land
surface in a river basin.
In case of deep water table (no lesa than 3-5 m) the
simplified water balance equation is used in the USSR to
estimate evapotranspiration from non-irrigated agricultural
fields; according to this equation evaporation is estimated
from precipitation (x) and the change of moisture storage in
the upper soil layer:
E =K+(L4pW.) (2)
where8 W and W are moisture storage in soil at the
beginnid and a? the end of the design period.
Equation (2) can be used only under the condition that all
precipitation is absorbed by the soil and no surface runoff
is formed and besides that the depth of rainfall water per-
colation should not exceed the depth up to which soil moisture
content was measured and moisture content was determined. Such
conditions usually exist during the vegetation period. The
220

depth of the upper layer of soil in which soil moisture storage

should be determined is 1 m in wet areas and up to 3 m in
arid zones. In case of reliable estimation of precipitation
and moisture storage the standard error of the estimation of
monthly sums of evapotranspiration by this method makes
approximately l5-2m. Another example of a partial solution
of water balance equation is the estimation of mean annual
sum8 of evapotranspiration as the difference between precipi-
tation and runoff /3/.
Theore tical and experimental development of the turbulent
diffusion method which is also known as the gradient or
aerodynamic method, has not yet reached the stage which would
allow its wide application in practice /1,2,9,12,15,16/. However,
this method :is promising. Being universal, this method is
based on gradient measurements of wind speed and air humidity
and provides estimation of evaporation from any land or sea
surface irrespective of the state and character of the latter.
Accurate enough (universal) solutions of the equations
of the first group require a consideration of a great number
of factors and special observations to be made. The develop-
ment of simplified semi-empirical and empirical design schemes
will provide possibilities for the estimation of evaporation
without the data of specialized observations. At present it
seems possible that standard observational data from hydro-
meteorological stations are enough far the estimation of
long term average annual and monthly evaporation sums for
a given territory and monthly evaporation sums for individual
different surfaces -
years, as well as for the estimation of evaporation from
snow cover, swamps, forests, irrigated
and non-irrigated agricultural fields /1,4,5,10,17/.
'

Most convenient are the methods of computation of mean

annual evapotranspiration sums for the regions of natural
moistening based on the equation developed by Y.I. Puàyko /IO/.

The right part of equation /3/ includes only one parameter

taken from standard observational data, i.e. long term average
precipitation X (cm year'l) which reflects the natural moisture
-
content of a region. Another parameter, reflecting the heat
regime and the character of the underlying surface averag
annual adiation balance of moistened surface Ro (kcal cm -2
year -$ is $&en from the map prepared by N
L is the latent heat of evaporation (kcal,
.
$*).
Efimova /IO/.
The standard
error of evaporation estimated from equation (3) makes about
17% /6/0
Mean annual sums estimated from equation /3/ can be
easily distributed by months because long term mean monthly
 221

evaporation sums given as percenta e of the annual sum, change

regularly according to geobotanic ?soil-climatic) zones. This
method is called the method of percentage ratios /7/. The
percentage ratios by months are given in tables, developed
by experimental or design ways. Table I, for Instance, presents
monthly evapotranspiration values as percentage of annual sums
for the main geobotanic zones of the European territory of the
USSR . Table I

C onifm o u s
forests O 0,5 2 6 17 25 22 15 8 4 O,5 C
Mixed and
decideous forests,
fore st-steppe s 0,5 I 3 9 18 20 ia i3 9 5 3 05

Steppes 1 I 3 1 1 1 9 2 0 1 6 12 8 5 3 I

M.I. Budyko suggested a combined method for the computation

of monthly evapotranspiration sums with the use of the main
elements of heat and water balances /1/. This method can be
applied in practice since it is based on the use of the standard
observational data, i.e. precipitation (x), runoff
temperature and humiditg.
c
y), air
In this case it is assumed that, when soil moisture content
is less than its water holding capacitg, monthly evapotranspira-
tion (E )is proportional to th8 monthly sum of potential
evapora8on (E )and to the average monthly storage of
productive moisgure in 1-metre layer of soil WI + W2
2
(WI and W2 are moisture storage at the beginning and at the
end of the month), that is:
222

where: W is critical storage of productive moisture in

soil lay& 1 m deep at which and above which
to Eo. Equations (4) and (5) are applied to
the year. WI at the beginning of the first warn month (Fa
spring) is estimated approximately, and later it is assumed
to be equal to W estimated for the end of each previous
month from equatzon:

or from equation:

where: y indicated runoff.

Eo and W O are taken from graphs and tables included in
publication /IO/. The value of E depends on the conventional
humkdkty deficit of the air whicf: is determined as the
difference between maximum water vapour pressure estimated
from mean monthly air temperature, and vapour pressure of the
air at the altitude of 2 metres.
This method was developed in two variants and is applied
for the estimation of long term average monthly evapotranspi-
ration sums of individual months of certqin years /4/.
-
Aver ed areal evapotranspiration values (areas to 1000
3000 km2y are determined from equations (3) and (4) (7).
-
Design schemes for the determination of evapotranspiration
from different kinds of surfaces include parameters which are
seldom measured at observational stations. For instance, in
the scheme developed by V.V. Romanov /13/ evapotranspiration
from a swamp is assumed to be proportional to the radiation
balance of the swamp surface (Q =dR ); in the scheme
developed by S.F. Fedorov /18/ evapotranspiration from forests
is proportional to the potential evaporation and the proportion
coefficient is presented as the function of the radiation
index of dryness (R/LX).
Monthly sums of evapotranspiration from irrigated fields are
estimated with the help of simplified heat balance equation /i/
 223

using special observational data, the standard error bei%

l5%, or with the help of modified formulae of the complex
method /ïg/ u s i q standard observational data, the standard
error being 3%. To estimate evapotranspiration from irrigated
agricultural fields empirical design schemes similar to that
of Blaney and Criddle /li/ can be used if only empirical
coefficients are tested and corrected for each point of their
application,
Most simple design schemes allowing estimation of evapora-
tion from water, snow and ice surfaces by means of standard
observational data, are the following binomial and monomial
equations:

and

(9)
where E is evaporation in mm/day, U, is the wind speed at
the height 2, above the surface in m/sec; es and e2 axe t h
maximum water vapour pressure estimated from the surface
temperature and water vapour pressure at the height of 2 m
in mb; A, a and b are coefficients estimated from experiments.
Substituting a
assuming z =
= 0.18 ab= 0.098 in equation (8) and
IO m one can obtain the formula for the estima-
tion of evaporation from the snow surface /8,11/,
When a L 0.14, b = 0.72 and z = 2 m, equation /8/ can
be used for the estimation of evaporation from lake (reservoir)
surface. In this case in equation (8) parameters uz , e and
e 2 shouid be substituted by correspondent values measure8 at
different points above the reservoir and averaged for a month
with respect to the whole water area of the reservoir,
In case of the absence of such obsemrational data one can
use the data from land meteorological stations situated in the
same climatic zone. T9e transition of the obeained above-land
coefficients &z/ efand 4; to the corresponding above
reservoir values h o u d be carried out with respect to the
transformation of the air flux affected by the underlying SUT-
face, the topography of the environment, the rate of wind
protection of the reservoir and the average length of wind
run above the reservoir /l7/.
In conclusion it should be mentioned that the present paper
224

deals with the methods which can be used in practice and produce
relatively reliable estimates of evaporation on the basis of
standard observational data from meteorological stations.
Therefore more complicated methods of the first group
which cause difficulties being applied in practice, and
numerous empirical design schemes which produce unreliable
results, are not treated here. Penman and Turc methods are
not mentioned since they are known well enough. It should be
mentioned as well that the above classification of methods
is conventional.
All methods are closely interrelated, and their develop-
ment, particularly the improvement of methods of computation
in case of inadequate data, depends greatly on further
experimental and theore tical research on the evaporation
problem.
R E F E R E N C E S
1. Qudyko Y.I., 1956. Teplovoi balans zemnoi poverkhnosti
(Heat balance of the Earth's surface). Hydrometeo-
rological Publishing House, Leningrad.
2, Buãyko M.I., 1948, Isparenie v estestwenoykh usloviakh
(Evaporation under natural conditions). Hydrometeorolo-
gical Publishing House, Leningrad.
3. Wodnye resuray i wodny balans territorii Sovetskogo Sojuza
(Water resources and water budget of the USSR area).
Hydrometeorological Publishing House, Leningrad, 1967.
4. Zubenok L.I., 1968, Ob opredelenii sumaiarnogo isparenia za
otdelnye godg (On estimation of evapotranspiration
during particular years). Trans. of GGO, vol. 233,
Leningrad.
5. Konstantinov A.R., Astakhova N.I., Levenko B.A., 1971,
Metoày rascheta isparenia s selskokhoziaystvennykh
polei (Methods for the computation of evaporation
from agricultural fields), Hydrometeorological
Publishing House, Leningrad.
6. Kuzmin P.P., 1966. Teoreticheskaya skhema otsenki oshi-
bok rascheta isparenia s poverkhnosti sushi (Theoreti-
cal scheme of evaluation of estimation errors of
evaporation from land). Materials of Interagency
meeting on the problem of study and substantiation
of methods of evaporation computations from water
and land. Ed. GGI, Valdai.
7. Kuzmin P.P., Zubenok L.I. Konstantinov A.R., Astakhova N.I.,
. .
Vinogradov V .V , 1968 Vnutrigidivie rasprede lenie sum-
suschi na territorii SSSII (Annual distribution of
evapotranspiration from land over the USSR territon),
Trans. of GGI, vol. 151.
 225

8. Kuzmin P.P., 1953. K metodike issledovania i zascheta

isparenia s poverkhnosti snezhnogo pokrova. (On
methodology of research and computation of evapora-
tion from snow pack surface) Trana. of G I , vol.
41 (95).
9. Leichtmap D.L., l9W. Profil vetra i obmen v prizemnom
sloe atmosfery (Wind profile and exchange in the
lowest atmosphere). Izv. AN SSSR, ser. geofis.,
No.1.
.
IO Materialy mezhduvedomstvennogo sovetchchania PO probleme
izuchenia i obosnovania metodov rascheta isparenia s
vodnoi poverkhnosti i suchi. (Materials of Inter-
agency meeting on the problem of study and sub-
stantiation of methods for the computation of
evaporation from water and land surfaces). Ed.
by GGI, Valdai, 1966.
11. Measurement and estimation of evaporation and evapo-
transpiration. Technical Note No. 83, WO-N0.201. TP.
105, 1966, Geneva.
12. Monin A.S., Obukhov A.M., 1954. Osnovnye xakonomernosti
turbulentno o peremeshivania v prizemnom sloe
atmospgery $Principal laws of turbulent mixing in
.
the lowest atmosphere) Trans. of Geophysical Inst.,
AN SSR, vol. 24 (151).
13. Romanov V.V., 1962. Isparenie s bolot Xvropeiakoi terri-
torii SSSR (Evaporation from swamps from the USSR
European territory). Hydromet. Publ. House, Leningrad;
14. Rukovodstvo P O gradientnym nabliudeniam i opredeleniu sos-
tavliajushchikh teplovogo balansa (Guide on gradient
observations and determination of heat balance compo-
nents) * Hydromet. Publ. House, Leningrad, 1962.
15. Rusin N.P., 1959, Gradientny metod opredelenia isparenia
s sushi i ego ispolzovanoe na seti stantsiy (Gradient
method of estimation of evaporation from land and its
use on the network of stations). Trans. of III-rd
All-ünion Hydrological Congress, vol. III, Hydromet.
Publ. House , Leningrad.
16. Tbornthwaite C.W. and Holtzman B., 1942. Measurements
of evaporation from land and water surfaces. U.S.
Dept. Agr. Technical Bul. 817.
17. Ukazania P O raschetu isparenia s poverkhnosti vodoemov
(Instructions for the computation of evaporation from
reservoir surface). Hydromet. Publ. House, Leningrad,
1969
18. Fedorov S.F., 1969. O reaultatakh issledovania digrologiches-
koi roli lesa. (On the research results of hydrological
role of forest). Trans. of GGI, vol. 176.
19. Kharchenko S.I., 1968. Gidrologia oroshaemykh zemel
(Hydroìogy of irrigated areas). Hydromet. Publ.
House, Leningrad.
 O B J E C T I V E C R I T E R I A T O D E C L A R E A SERIES OF
DATA S U F F I C I E N T FOR T E C H N I C A L P U R P O S E S
by
P e n t a A., R o s s i F.

ABSTRACT

It i s supposed: that f o r t e c h n i c a l p u r p o s e s it i s
n e c e s s a r y t o e s t i m a t e t h e v a l u e s xo that an h y d r o l o g i c a l
v a r i a b l e x may a s s u m e w i t h a g i v e n p r o b a b i l i t y 6; t h a t x c a n
b e m e a s u r e d d i r e c t l y and t h a t its n v a l u e s h a v e b e e n r e c o r d e d .
T h e s e r i e s of t h e n v a l u e s o f x i s ' d e f i n e d s u f f i c i e n t
i f it c o n s e n t s t o e s t i m a t e x o w i t h a r e l i a b i l i t y a d e q u a t e f o r
technical purposes.
By r e f e r r i n g t o t h e u s u a l s t a t i s t i c a l m e t h o d o l o g i e s ,
the authors present objective criteria to recognize whether
t h e s e r i e s o f n values i s s u f f i c i e n t . The a u t h o r s f u r n i s h s o m e
d i a g r a m s that i n d i c a t e w h i c h m i n i m u m v a l u e s o f n a r e n e c e s s a r y
for t h e s e r i e s t o b e c o n s i d e r e d s u f f i c i e n t .
From t h e d i a g r a m s it is e v i d e n t t h a t f o r t h e s a m e v a l u e s
o f n t h e s e r i e s s u f f i c i e n c y is s t r i c t l y l i n k e d t o t h e v a r i a b i -
l i t y o f x.
Particularly, the authors considere the normal, t h e log-
n o r m a l and t h e d o u b l e e x p o n e n t i a l ( G u m b e l ) d i s t r i b u t i o n s , t h e
m.ost a p p l i e d l a w s o f hydrology.

--
RESUME

On s u p p o s e qu'à l'égard d u p r o b l è m e t e c h n i q u e i l f a u t
e s t i m e r l e s v a l e u r s X Q qu'une v a r i a b l e h y d r o l o g i q u e x p e u t
a s s u m e r a v e c l a p r o b a b i l i t é 0, que x peut ê t r e m e s u r é e et q u e
n v a l e u r s de x ont été e n r e g i s t r é e s .
La s é r i e d e s n v a l e u r s d e x est d é f i n i e s u f f i s a n t e s i
par e l l e on peut e s t i m e r xa a v e c u n e c o n f i a n c e a d é q u a t e a u b u t
d u technicien.
E n se r a p p o r t a n t a u x m é t h o d o l o g i e s s t a t i s t i q u e s u s u e l l e s
on donne des criteriums objectifs pour reconnaitre si la série
d e s n v a l e u r s est s u f f i s a n t e .
O n d o n n e d e s d i a g r a m m e s p a r l e s q u e l l e s on i n d i q u e l e s
valeurs minima du nombre n qui son necessaires afin que l a série
soit suffisante.
D'après l e s d i a g r a m m e s i l a p p a r a i t é v i d e n t que, n a y a n t
l a m ê m e v a l e u r , l a s u f f i s a n c e d e l a s é r i e dépend d e l a v a r i a b i -
l i t é d e x.
En particulier, les auteurs considèrent la loi normale,
la l o i log-normale e la l o i d e G u m b e l , q u i sont p l u s f r é q u e m m e n t
e m p l o y é e s en hydrologie.
228

Symbols and definitions

1: Let us indicate by I
- , x a generic hydrological variable3
-E, a x and y, respectively the mean, the standard deviation
and the coefficient of variation of the x population;
- @(XI, the distribution function of x ;

@ e
-xQ, , the value of x corresponding to the cumulated probability

Moreover, let us also indicate by :

-x , with 1 4 i s n , the n values of x registered, in each

--
single year,iduring the observation period i
x and ax, respectively the estimates of [and a ;
- Pix) , the estimate of the distribution function (Dix);
-- xppa,, the estimate of
y(xP,@),
xa ;
the sampling coefficient of variation of 5
E@

2: If x is normally distributed,the best estimate xpsio, of x

(D
is obtained [1 1by t
X pn(D = Z + u@ (1)

where u is the value of the variable u, that in equation:

@
1 2

@(U) E - 1
2
du (2)

corresponds to the fixed value of @ .

The sampling coefficient of variation of xp could be obtained
approximately [2] by I -@

or whenever n is sufficiently high, the equation (3) becomes :

1 +u$/*
(3')
i i n
 229

3: If x is distributed according to the log-normal function,

having established that y = log x , we indicate by :
- Yi 9 the value of y corresponding to the generic value xi ;

the n
- I
y and s
Y'
values of Yi.
respectively the mean and the standard .deviation of

Therefore, the best estimate x of x is obtained [a] by

P=UJ 0
log xp' E y. + U@ 8 (4)
=a> Y
where the value of uUJ is deduced by means of equation (2) while the values
of y and s are deduced respectively by the equations :
' n

-
Y"
13.1
log xi

n
and
r n

s = n-1
Y
The sampling coefficient of variation of xPIUJ could be obtained
approximately [
21 by :
2
1 + u0/2

- 1 (7)

or, whenever y* is sufficiently mall, the equation (7) becomes :

I 2

4: If x is distributed according to the double exponentlal,namely

@umbel function, an almost correct and efficient estimate xP=@ of x UJ is
obtained [4] by :
(8)
xppUJ= + K UJ

where Ka, is the value of the variable K that'in the equation :

K --
= 6
x
(0,5772 + In In -1
a,
(9)
corresponde to the fixed vaïue of <D .
230

The sampling coefficient of variation of x could be obtained

approximately [ 2 1 by :

5: When the variable x is measured in k gaging stations, lying

in a detertnined zone, there exists an hydroloRica1 similitude between the k
stations if the parameters, or some parameters a l , z2 , ..... of the x
diatribution assume the same value or if they vary from one to another with a
known regression relation in function of a certain number of parameters y1 '
y2 ..... [5].
The inter-station correlation is the correlation which exists, in
such cases, among the values of x registered in them contemporaneously (e.g.
in the same year if maximum and minimum annual values are considered).
Therefore, the information that can be derived from the k stations,
considered all together, in regard to the x distribution parameters a l , a2,..
..... is the same as the information furnished by a number k of independent
stations. Such a number, known as the equivalent number, depegds both on k
and on the mean interstation correlation coefficient F , thus becoming so
smaller than k, the higher the value of F is.
So, e.g. if in the k stations the mean 6 of x would assume the
same value, the information that the complex of the data registered in the k
etatiomwould furnish in regard to would.be equal 16) to the information
furnished by an equivalent number of independent stationsequal to :
k
keE l+F(k-1)

Basic Risk. Uncertainty and Effective Risk

6: Normally, for design purposes, by referring to a given hydrolog&
cal variable x,we indicate by :
xd , the value of x that Is assumed as the basis for the design
(deBiRn Value) ;
N , the desinn duration.
Particularly, in a flood problem, we select a value O f so that
there exists a probability of failure W that xd will be exceededxdat least
once in N years.
Consêquently, xd coincides with the value x D of x whlch corre
sponda to a value of D of the cumulated probability furnished by :
 231

E.g. when N = 25 years and W=0,025, @ is equal to 0,999.

Likewise, in a drought problem, we select a value of
xd so that
there exists a probability of failure W that xd will not be exceeded at least
once in N years.
Consequently, instead of using equation (121, we must apply the
following equation :

E.g. when N
@ = l -
=
(1 - w)
25 years and W
1 /N

= 0,025, @ is equal to 0,001 .

The basic risk is defined [71 as the risk that would be encountered
if, by knowing the probability distribution of x,we would assume xd E X @ .Such
risk is measured by means of the probability of failure w.
In reality, however, the distribution of x is not known. Consequent-
ly, having fixed the basic risk W and having calculated @ by means of eqpa-
tions (12) or (131, with the use of a series of n values of x,only an esti-
of x could be had, aiid, therefore, to assume x = x Q an error
mate x
equal toP =?x
encountered ?;treater
-x D ) would be made. In reality the effective risk that is
than the basic risk due to the uncertainty with which
the value of x could be estimated.
0

Sufficiency of a Single Series of Data

7: Once the basic risk W has been determined, to judge whether a
single series of data is sufficient for technical purpose4,i.t is necessary to
take into account the uncertainty with which x @ could be estimated.
Generally, by considering also the observation periods which are usu-
ally available, a series of at least 3 0 + 4 0 data is defined IIlonP and it is

-
implicitly retained sufficient; a series with less than 30+40 data is defined
Vshort1I and is considered insufficient.
In reality, however, such criterion might be erroneous. In fact, if
the uncertainty, with which x0 could be estimate, is measured by means of
y{xp } , from eq. (31, or eq. (7) or eq. (101, we recognize Immediately that
the s a d uncertainty, beside n ,depends also on :
i ) the variability of the hydrological magnitude x being considered,
which can be measured by y ;
ii) the probability Q of the design value xd , which is a function
of the basic risk W and the design duration N.
In particular, let us consider e.g. the annual rainfall depth x u h
distributed generally according to the log-normal function [ 81, with a coefficient
of variation y, which varies from 0,l to 0,9 as we progressively move from
sub-humid zones to semi-arid and arid zones, the mean annual rainfall changes
from vaïues of circa i 500 mm to values of circa 50 mm [9J .
 232

As it could be noticed from the diagram (a) of fig. 1, if it were nec

essary to estimate the median value x 5o of x , a long series could be retained
sufficient from a technical point of v h w for each of the possible values of y x ,
since in no case y{xp, would be greater than 15%.
However, wheh we fix the duration N equal to 25 years and the basic
risk equal to 2,5%, by applying eq. (12) or eq. (13) we notice that we must refer
to values of @ equal to 0,999 or 0,001. In this case, from the diagram (b)
of fig. 1 it ie evident that a long series of data would be sufficient from a
technical point of view only if y were rather low.
In fact, even for values of y, greater than 0,5, Y { X ~ , ~ ) could
s b e greater than 20%.
On the other hand, from the same diagrams (a) and (b) of fig. 1 ,it
can be derived that a short series of data, which is certainly insufficient
for values greater than y , could be sufficient if y, would assume too
small values.
Analagoue considerations could be made if x follows the double
exponential distribution by examining the diagrams (a) and (b) of fig. 2 in
which are represented the function of y{xp, 1 as n and
Y,
for @ P 0,368
(corresponding to the distribution mode) -and fi? CD= 0,999.

the Data Registered in Other Stations

9: The regions where regular hydrological measurements have been
taken for a short period of observation, have often arid or semi-arid climate,
therefore, it becomes practically impossible to estimate from a single series
of data the values that, with a given probability, those magnitudes might assume.
It becomes therefore necessary to recognize if it is possible to improve the
estimate of xrp in a given station by using the data obtained in others. As
it is known, to render this possible, it is necessary that the different stations
considered b e hydrologically similar (see pgr.5). For this to happen, it is
necessary that the values taken by x in the &d stations depend on common
meteorological and hydrological factors. Consequently, this implies that there
exists an inter-station correlation.
It is udeful to point out that from this point of view it is very
important to consider either one of the hydrological magnitude. In fact, the
mean inter-station correlation coefficient P is amaller when the daily or
weekly rainfall is considered, while it is greater when we take into account
annual rainfall [101 . In the case of annual rainfall, in a research conducted
from the information furnished by 1141 pluviometers installed in the Western
D.S. and in the South-West California [lOl, Caffey has shown that the mean
inter-station correlation coefficient F situated in a zone meteorologically
homogeneous varies from Q,30 to 0,SO. In a recent research on pluviometers
installed in Basilicata and in Southern Italy, we have found f n 0 , 5 t 0,6 and
 233

in a research on the Morocco pluviometers, being conducted at the time of this

report, r = 0,90 which is still higher.

10: By referring to the mean value 6 of x , in the diagram of.

fig.3, ne have repreeented equation ( 1 1 ) which formulates the law according which
the equivalent number ke of independent stations, defined in pgr.5, varies as
a function of r' and the number k of statione installed in the zone.
,

As it can be observed from fig.3, for each value of F , ke increasel

at each increase in k ,tending asymptotically toward a maximum value kernax= F
Consequently, the maximum increase of information that is obtained in
regard to 5 by applying the hydrological similitude criteria is inversely
proportional to F . E.g. when F = 0,5 , the information, at the most, could be
doubled; for still greater values of ? , which are often encountered in hydrology,
the advantage obtained could be almost negligible.
On the other hand, no real benefit is obtained by increasing the number
of k stations above a certain limit strictly connected to F . To prove this,
we have represented in the diaáram of fig.4 the law with which
k
-
ke varies as
a function of k for different values of r. As it can be noti$eyxif we are
satisfied with the 90% of the maximum information that can be obtained, by
accepting that -k
ke
= 0,9, this objective could be reached with only 9
stations, for F = with only 4 stations for r 5 0,7.

CONCLUSIONS

11: In eome countries, systematic, reliable, homogenous measurements

have been taken for only few years and in few stations. Moreover, to render the
problem more severe, such regions have an arid, or semi-arid climate. Therefore,
due to the extreme variability of the hydrological magnitudes, with the same
number of data, the uncertainty with which the probability distribution of them
could be estimated, is greater.
Consequently, in the said regions it is particularly important to
utilize all the information that the few available data could furnish, by applying
either correct statistical methods to interpret each single series of data and/or
by defining objectively some hydrological similitude criteria that would consent
the interpretation on how the magnitude varies from one station to another.
Particularly, for a reference magnitude x , by applying the hydrolog-
ical similitude criteria, it is possible :
a) to obtain a reliable estimate of x o even for points where no
direct measurements of x were even taken ;
2 34

b) to improve the estimate of x in points where only few data

are available. Q,
The advantages obtained in regard to point b) could be noticeably
limited by the inter-station correlation located within an hydrologically
homogeneous zone.
In any case, only when all the information available has been uti-
lized, it is possible to establish whether the data available are sufficient
or not to be used in practical applications.

12: If the available data in the region should be insufficient, a

supplementary research program would be necessary. Even in this case, it is
absolutely necessary to take into account the information furnished by all the
data available so that the research program is carried out in an adequate manner.
On the other hand, we must be well aware of the results of a short
research program.
In fact, if an appropriate localization of the stations is made, it
is useful :
i) to individualize and improve the delimitation of the region in
hydrologically homogeneous zones ;
ii) to determine the regression law of a variable x as function of
some parameters which characterize the point or the basin (e.g. the regression
relation of the mean rainfall depth vs the level of the point or the regres-
sion relation of the mean annual runoff vs mean annual rainfall).
On the other hand, when both aims have been attained, the research
program could be useful also to estimate the probability distribution of x
in different points (or basins) only if in the region there are one or more
gaging stations functioning for a long time.
 235

1. KENDALL M.G. and STUART A., (1967). The Advanced Theory of Statistics.
London, Griffin, Vol. 2, 2nd Ed., p.54.

2. ROSSI F., (1 972) Distribuzione di campionatura di alcune grandezze stg

tistiche. Fac. di Ing. delltllniv. di Napoli, 1st. di Costr. Idr.,
Quad. no 5 .
3. AITCUISQM J. and BROWN J.A.C., (1957). The Lognormal Distribution.
Cambridge University heas.

4. LOWRaY N.D. and NASH J.E., (1970).

-
Methods of Fitting Double Exponential
Distribution. Journal of Hydrology, 10, pp. 259 275.

5. VIPARELLI C., (1 965). Idrologia applicata all 1 Ingegneria. Parte II, Fond.
Politecnica del Mezzogiorno d'Italia, Napoli.

6. MATALAS N.C. and BENSON M.A., (1961). Effect of Interstation Correlation

on Regression Analysis. Journal of Geophysical Research, Vo1.66, nolo.

7. YEVJEVICH V., (1972). Probability and Statistics in Hydrology. Water

Resources Publ., Fort, Collins, Colorado.

8. MARKQVIC R.D., (1965). Probability Functions of Best Fit Distributions of

Annual Precipitation and Runoff. Hydrology Papers no 8, Fort Collins, Co-
lorado.

9. GARCIA-AGREDA R., RASULO G., and VIPARELLI R., (1 973) Pluviometric Zones
and the Criteria to Define their Boundaries for Regions with Scarce Data.
Simposio sobre proyectos de recursos hidraùlicos con datos insuficientes,
Madrid.

10. CAFFEY J.E., (1965). Inter-station Correlations in Annual Precipitation

and in Annual Effective Precipitation. Hydrology Papers no 6. Fort Collins,
Colorado.
 20 40 60 *O n
OBJECTIVE CRITERIA TO DECLARE A SERIES OF DAPA SUFFICIENT I'OR
TECHNICAL PURPOSES
OBJECTIVE CRITERIA TO DECLARE A SERIES OF DATA SUFTICICNT FOX
TECHNICAL PURPOSES
 10

Ke

4 I

2
6
-
/
- O,50
I
O,70
I

O Y
Fig. 3
10 20
?=%O

30
1 1
I
I
I
I
K. 50

OBJECTIVE CRITERIA TO DECLARE A SERIES OF DATA SUFFICIENT FOR

TECHNICAL PURPOSES
l,oo - 1,oo-
/

0,8O

0,60

0,4O

0,20

10 20 30 40 K 50
Fig. 4

OBJECTIVE CRITERIA TO DECLARE A SERIES O F DATA SUFFICIENT FOR

TECHNICAL PURPOSES
 SOME CRITERIA USED IN HYDROLOGIC STUDIES
WITH INADEQUATE DATA

C a r l o s Q u i n t e l a Góis

ABSTRACT

I n t e r r i t o r i e s w h e r e t h e h y d r o l o g i c n e t w o r k s are
s t i l l s c a r c e , i t i s n e c e s s a r y t o adopt s i m p l i f i e d d e s i g n i n g
c r i t e r i a w h i c h m i g h t lead t o s u f f i c i e n t l y r e l i a b l e r e s u l t s .
In t h i s paper t h o s e w h i c h a r e n o r m a l l y used for t h e h y d r o -
logic characterization of t h e drainage basins under these
c o n d i t i o n s are p r e s e n t e d and e x a m p l e o f t h e i r a p p l i c a t i o n
t o O k a v a n g o B a s i n in A n g o l a i s g i v e n .

RESUME

Dans l e s t e r r i t o i r e s o ù l e s r é s e a u x h y d r o l o g i q u e s
s o n t e n c o r e insuffisants, i l f a u t r e c o u r i r a d e s p r o c é d é s
de calcul simplifiés qui puissent conduire a
des résultats
d i g n e s de confiance. L'auteur e x p o s e d e s m é t h o d e s n o r m a -
l e m e n t u t i l i s é e s p o u r évaluer. d a n s d e t e l l e s c o n d i t i o n s
les caractéristiques hydrologiques des bassins, e n prenant
p o u r e x a m p l e d u b a s s i n d u C u b a n g o , e n Angola.

* Civil Engineer - M e m b e r of t h e W o r k i n g Group

o f t h e O v e r s e a s M i n i s t r y ( P o r t u g a l ) f o r t h e I.H.D.
242

1. Introduction

In the framework of the hydraulic policy which has b en followed for he

last years in the Portuguese Overseas Provinces, the study o f the general plans
for the development of the water resources plays a very important role. In the
- -
two main provinces of Africa Angola and Mozambique this action led to the fact
that the main drainage basins are already covered by such studies; that enables
an adequate hydroelectric and hydro-agricultural overall planning to be made,

Hydrological studies are obviously the fundamental basis of such general plans
because they. determine the hydrologic characterization of the basin and from the-
re the preliminary design of the several schemes and estimate of their potentia -
lities. In this field international cooperation which was achieved with the other
territories of Southern Africa, as a result of established agreements, is also of
a great importance and it gives an idea of the value that water has got for the
common development on that part of the world.

The inhospitable characteristics of these areas together with the communica-

tion difficulties and low human occupation result usually in very scarce and
recent hydrologic networks so that on carrying out hydrologic studies one faces
the difficulty of applying the classic methods or those used for more developed
areas.

Therefore it is necessary to adopt approximative methods and special crite -

ria enabling to arrive at sufficiently correct and reliable results for the ai -
med purposes.

In this paper the methods which have been followed for carrying out the abo-
ve mentioned hydrologic studies are presented and the approximate criteria that
have been adopted as a result of inadequacy of data are pointed out; at the end
a practical example is given for the case of a drainage basin in Angola. Only
the aspects of rainfall and run-off in average terms are stressed because they
are o f most interest for the hydrologic studies of general plans.

2. Rainfall

Among the hydrologic data, rainfall is commonly measured for a longer pe-
riod, even in developing territories. Although networks do not cover satisfacto-
 243

rily the areas to be studied, they enable the characterization of the phenomenon
with enough accuracy to be achiewed.

Usually the daily precipitation data measured in raingauges normally loca-

ted at villages or townships are available. Record periods of twenty years ormore,
at least in some of the stations, are frequent and the use of correlation techni-
ques enables to obtain monthly rainfall all over the stations of the network, On
the other hand, uniform rainfall regime of the African subtropical regions with a
long period of four months without precipitation is well known, which makes it ea-
sier to fulfil some failures in the records.

The study of that regime is usually done by taking the annual weighed pre-
cipitations obtained from the isohyet maps drawn for the basin. The isohyet me -
thod is considered to be the most adequate when dealing with incomplete informa-
tion, because local surveys, topography, etc. may help to introduce corrections
or indicate the best drawing of the curves of equal precipitation so that a pat-
tern, as close as possible with reality, can be obtained. Once the basins have
usually a drainage area of tens of thousands of square kilometers, the used sca-
le for drawing isohyet maps is normally 1:l O00 000.

After those maps are obtained, some characteristic sections are chosen and
the weighed values are determined. These are the bases for the study of the rain-
fall regime and periods of about 20 years permit the application of stochastic me-
thods. Among these, the method of Hazen-Foster has been considered to be the most
adequate to interpretate the phenomenon. After graphical and analytical confirma-
tion of its applicability, it is possible to obtain the mean annual value and tho-
se corresponding to characteristic return periods. The probability relating to each
one of the years of the period can be obtained as well.

This analysis gives a first idea of the natural sequence of the years and
principally the occurence of dry periods and their degree of drought so that fur-
ther studies for comparison with the run-off can be done.

The study of rainfall is usually completed with a short analysis of dry and
wet seasons and mainly of the frequency with which longer dry seasons may occur.

3. Run-off

As far as flow measurements are concerned, data is always very scarce and only
few flow stations in Portuguese Africa have records available for more than 5 to
10 years. Besides, it has been verified that the study of general plans normally
shows the need and lead to the best choice and establishment of the hydrometric
networks.

Stochastic methods cannot be applied safely with such short periods and
therefore the first approximative criterium to be used is trying to characterize
the available flow record period by relating it with the similar period of the
rainfall studies. Hence it is possible, as a first approximation, to consider the
same probability of occurence for the annual flow and rainfall of a certain year.

From this it is often possible to chose certain years which can be consi-
dered as average or with a given degree of dryness. Therefore a critical period
corresponding to an unfavourable sequence of years can be chosen in order to fix
the storage capacity of interannual reservoirs and to obtain a complete regula-
tion of the flows. This sequence is normally formed by an average year followed
by two or more dry years with fixed characteristics. Undoubtedly this is an appro-
ximate approach, but experience has shown that for studies at the level of gene-
ral plans this analysis is quite acceptable and safe because the pessimism in
the reasoning compensates the.uncertainties resulting from the inadequacy of da-
ta.

Sometimes, as an exception, there exists in the basin a measuring section

with a longer period of records and for which stochastic methods can be applied.
Two ways can then be followed, (1) correlation analysis with other stations of
the basin, trying to obtain more data for those which have shorter records or
(2) characterization of the shorter period by relating it with the longer one
of that station in a similar way as mentioned in the previous paragraph for the
rainfall.

The first method is not always easy to apply, because the rivers might
show a change of regime along their course as a result of the phisiography and
correlations are no more valid.

The second one is more reliable and on applying it, it is possible to ar-
rive at safe and easily interpretable results. Normally one can obtain not only
the annual flow but also the monthly ones of the average and dry )ears of the cho-
sen critical period and therefore carry out more reliable regulation studies.
 245

The study of rainfall/run-off relations has not been, as far as our expe-
rience is concerned, successful for large drainage basins as a method of enlar -
ging the available flow record period. This is probably the result of the speci-
-
al type of the rainfall regime of those regions short, heavy and localized storms-
together with high temperatures and evaporation rates which affect the usual me-
chanism of transforming rainfall into run-off. Besides, this method would only
lead to global annual values and its distribution along the year is not possible
to obtain.

4. Application example

4.1 - General characterization of the problem

The Okavango is one of the three big international rivers of the

South of Angola. It springs on the central plateau of the territory and flows
more or less North-South down to the border with Southwest Africa where it
shifts eastwards, forming the border, crossing Kaprivi Strip and spreads in-
to a wide swampy area (Figure 1).

Its drainage basin in Angola is about 150 O00 km2 from which 61 O00
km2 belong to its main tributary Cuito.

The northern part of the basin is the most rainy one and there the
altitudes reach 1 800 m, decreasing gradually southwards to 1 O00 m.Here the
climaté is semi-arid. Rainfall occur in the wet season from October to April;
the other months are dry.

From the geological standpoint, the northwest part of the basin is

formed by igneous and metamorphic rocks; sedimentary formations occur in the
rest of the basin.

The hydrographic pattern is characteristical as well, the tributa -

ries being normally parallel to each other and flowing North-South. The sha-
pe of the beds is ruled by the local geological and topographical conditions.

As far as the vegetation is concerned, it changes from the more or

less dense forest in the North into the savana in the South.

The problem was to carry out the general plan for the development of
246

the water resources and obviously the first step was the hydrological study.

In the following chapters a summary will be presented of the analy-

sis made for the study of the rainfall and run-off, according to the methods
and criteria mentioned above in this paper, once the available data was ina-
dequate.

4.2 - Rainfall studies

For the rainfall studies, the records of 28 stations for the period
1943/1970 were available. However, only from 1950/51 onwards, the number of
stations with complete records was sufficient and therefore the basical study
period considered was 20 years, from 1950 to 1970. Some shortage of monthly
records necessary for the evaluation of the annual values was easily overcome
by correlation with more complete and reliable stations.

With these annual values, the isohyet maps were drawn on a scale
1:l O00 O00 introducting the influence of altitude and other known climatical
factors and avoiding a cold interpretation of the plotted values.

After chosing some characteristic sections, the weighed annual rain-

fall was determined and analysed by applying the Foster-Hazen method. Figure
2 shows a diagram with the sequence of annual precipitation and the correspon-
ding probability graph for a section of the main course of the river where the
international border starts.

From the joint study of these graphs, some conclusions can be drawn.
First of all, the applied stochastic method can be considered adequate to in-
terpretate the phenomenon and therefore it is possible to determine a mean an-
nual precipitation of 950 mm as well as precipitations corresponding to cer-
tain return periods. One can note the occurence of a sequence of four dry
years which might be considered as the basis of the critical period for regu-
lation purposes.

4.3 - Run-off studies

The basin has 19 flow measuring stations and the records started to
be obtained early in 1963. Before that date, there were some random measure-
 247

ments made with floating device> but their reliability was doubtful. The net-
work is nowadays equipped with automatic level gaugings and flows are measu -
red with current meters suspended from steel cables crossing the river from
one bank to the other.

It was then possible to have flow records for a period of 7 years

consisting of maximum and minimum flows, average daily flows, and consequent-
ly monthly and annual values.

For such a short period stochastic methods are not applicable with
reliability; nevertheless the analyses made for the rainfall showed that such
period has average characteristics and therefore the mean annual flow can be
estimated by averaging the flows of those seven years for every station.

The same criteria cannot be applied to determine the dry year flow,
because in this seven years period (1963/1970) any of the years of the cri-
tical period obtained from the rainfall study is not included.

Fortunately, there is a station in the international strech measu-

red by the South African Services which has got records for a longer period
(25 years) from 1945 on, although some of its valueshave been obtained by cor-
relation. It was then possible for this station to apply the Foster-Hazen rne-
thod which showed a rather well interpretation of the phenomenon.

Figure 3 shows in the same way as for the rainfall the diagram of
annual flow sequence and the probability graph.

The former indicates a notorious resemblance with the one of the

rainfall, being characteristical the four dry year period 1966/1970. It skws
as well that the period 1963/1970 is an average one and that 1966/67 can re-
present the dry year of the critical period.

In order to obtain the annual flows in any section of the river,tk

curves showing the variation of the specific annual flow with the drainage ba-
sin for the average and dry year, were drawn (Figure 4); these curves show
bi uniform pattern and thus one can consider them sufficiently reliable for
obtainment of the desired values.

The regulation of flows can be studied by considering the sequence

248

of an average year followed by four dry years as determined above.

5. Conclusions

Some criteria normally utilized for hydrological studies of the general

plans for the development of the water resources of rivers in semi-arid areas of
Portuguese African territories were presented and an example of their application
given. The obtained results are obviously approximate but they can be considered
sufficiently safe for the purpose and moreover when decisions would be taken for
the design of specific projects further data will be available and then a more re-
liable analysis can be made.

.........................
249
W
I
-I
rnm
6 O0 SEQUENCE G R A P H

LOO

200

O00

BOO

600

LOO

200

YEAR

FOSTER-HAZEN ADJUSTMENT
Pmm
500

LOO

300

200

100

O00

900

-
0
a%
0
<n
0 - N Y i
-
0 O
N
O 0 0 0
- < m u > c m
0 O O
m
y>
m
y> m m
m O b m 6
m a m r n m
PROEABILIT Y

FiGQRE 2 - STUDY OF ANNUAL RAINFALL

 io6,' SEQUENCE G R A P H 251
10 O00
LL
I
L
O
z 9000
I
3

< 8000
z
z
7000

6 O00

5000

.4oco

3 WO

2 O00

1 O00

FOSTER- H A Z E N ADJUSTMENT
lo6 m'
1.4 o00
IL
Y
O
z 3 O00
3

2
U 2 ooa
3

z
U
1000

3 000

9 o O0

a O00 i

7 O00

6000
1
5 O00

4000

3000

2 O00

1 O00
-ri yl
0 0 0 - h i - 0
- 0
N
0 0 0
m - m w p i
0 0 0
m
o
01
~n
m
In "01
m m o i m m
0 . m m m m
PROBABILITY
FIGURE 3 - STUDY OF ANNUAL RUNOFF
252
.
3
2
a 6
W
U
W
a II:
a
:; W l-
I z
I
U
W
I-
a
I
U W
3 I
n u
a
U
n
.
3 Z
a
3
O
A
3 I
L
O
J
a
3
Z
z
3
0 Q
2
LL
u
u
=l
U
a
ul
z
W
W
3
3
I-
W
m
I
a
3
Y
a
K
0
I
U
2 W
a
3
0
LL
 UTILIZING CLIMATIC DATA TO APPRAISE POTENTIAL WATER YIELDS
Robert L. Smith"

A B S TRA CT
Precipitation and temperature measurements often represent the
only significant hydrologic data available in developing areas. Initial
assessments of potential surface and ground water supplies must build
on this limited climatic base. Early in the planning studies there is
need f o r an accurate estimate of mean annual streamflow, and of the
probable variance in annual flows. These determinations can be made
utilizing an empirical function relating the mean annual runoff
coefficient to the aforementioned climatic parameters. The relationships
have been tested in a wide range o f environments, and their general
utility can be extended appreciably with limited surface and subsurface
observations. Applicability o f the recommended relationships is
demonstrated by selected case studies involving a variety of problems.
Included are examples illustrating the calculation of: (a) mean yields
for ungaged areas, (b) the probability distribution of annual flows for
ungaged areas, (c) daily flow duration curves, (d) potential yield of
selected groundwater areas, and (e) the potential impact o f precipita-
tion augmentation on surface water supplies.

RESUMEN
A menudo las medidas de precipitación y temperatura son los Ú n i
cos datos hidrolbgicos disponibles para áreas en desarrollo. Los esti-
mados iniciales sobre abastecimientos potenciales de aguas superficia-
les y subterráneas deben partir de esta limitada base climática. Muy -
pronto en e l curso de la planificación se hace necesario un estimado -
preciso del caudal promedio anual y de l a variación probable en flujos
anuales. Estas determinaciones pueden hacerse mediante la utilización
de una función empírica relacionando el coeficiente de escorrentía me-
dia anual con los antes mencionados parbmetros climáticos. Este tipo -
de relación ha sido puesto a prueba en una amplia serie de medio am--
bientes y su utilidad general puede extenderse apreciablemente con li-
mitadas observaciones sobre y bajo tierra. El éxito con que se han --
aplicado l a s relaciones recomendadas se demuestra por medio de casos -
escogidos que cubren una variedad de problemas. Se incluyen ejemplos
que ilustran el cálculo de: (a) rendimientos promedios para áreas Ca--
-
rentes de medidas, (b) la distribución probabilística de caudales a n u a
les en áreas carentes de medidas, (c) curvas diarias de caudal-dura---
ción, (d) rendimiento potencial de áreas de agua subterránea escogidas
y , (e) e l impacto potencial de l a incrementación de precipitación s o - -
bre abastecimientos de agua superficial.

-
JI Deane Ackers Professor of Civil Engineering, University of Kansas,
Lawrence, Kansas, USA.
254

The water resources planner is often required to appraise the water yield
characteristics of streams for which flow data is unavailable. In these situ-
ations the initial appraisal has to be based on climatic factors supplemented
by prior experience in similar terrains. This paper presents an empirical
relationship designed to further this appraisal, and which the author has found
useful on a number of occasions.

The basic water balance equation applied to a catchment area may be

expressed as
P = R + E + AS (1)
where all t e m o represent units of depth over the catchment area, and P = pre-
cipitation, R = basin outflow, E = evapotranspiration and AS = change in
storage.

For the condition of an extended time interval the AS term becomes

negligible. In this case, and after dividing all terms by P, the equation may
be rewritten as
-
RIP = c = 1 E/P (2)
Thus, in the long term the runoff coefficient C is governed by climatic con-
siderations.

Geographers and agricultural scientists have long utilized climatic

parameters in appraisinz water balance questions relating to management of soil
moisture. In 1967 Guisti and Lopez [i] proposed that the mean stream discharge
could be determined as a €unction of (a) mean annual precipitation and (b) the
basin climatic index. BCI. The latter is based on the work of Thornthwaite i21

Jan *'"I

where P is the average monthly precipitation in centimeters and T is the aver-

age monthly temperature in degrees Centigrade.

If the hypothesis presented by Guisti and Lopez has merit, it should be

possible to develop a relationship between BCI and the deviations from the mean
line d r a m on a sc.atter diagram of average precipitation versus average runoff.
Their initial efforts to develop such a relationship were limited to examina-
tion of relatively short term data in Puerto Rico. Smith 131 subsequently ex-
tended this approach by examining data from approximately 250 ca.tchments in the
United States and Puerto Rico. The resulting empirical relationship between
the coefficient C in equation (2) and the BCI is graphed in Figure 1. It dif-
fers appreciably from t.he curve initially presented by Guisti and Lopez. The
available data provided firm definition of the relationship for BCI values
ranging froiri 35 to 150. Currently, extension beyond these limits is most ten-
tati.ve and i.s Eased on the following. The lower end was extended to the obvious
terminal al: the origin. Extension of the upper end of the curve was based un
concurrer.tappraisal of the nature of the 13CI vs P relationship in high rainfall
zrens, and on recognition that the change in C with X I should be such that the
i.iic.rment.alpercent of precipitation which becomes runoff is constantljr
 255

increasing bur never exceeds unity. One word of caution. Data utilized in
developing the re1 onship was obtained €rom catchments for which the sub-sur-
face outflow was negligible. Thus the ruhoff calculated by Figure 1 represents
tot91 runoff and cannot be directly equated to streamflow in those instances
wtie're a significant percentage of the yield'is discharged as sub-surface flow.

Utilization of the relationship is enhanced by conversion of existing

climatic data into a basic P vs Bdi relationship for the area in question.
Worldwide the relatioriship between BCI' and P varies markedly. Regionally It
preciably with topographie considerations. However, for a given
he relationship between BCï and P is weJ.1 defined. Figure 2 illus-
trates a typ'ical relationship for a basin in the State of Kansas in the central
United States Qhere elevation changes are negqigible, and similar relations fcr
Puerto Rico where elevation is a significant factor. Note that the slope of the
relationship also varies slightly with location. Figures 1 and 2 caq be used
conjunctively to develop the mean annual rainfall-runoff relationship for the
catchment. Experience has shown that actual data will scatter about the curve
so determined because AS is seldom negligible on an annual basis. The individ-
ual curves tend to approach a 45" asymptote as evapotranspiration tends E O
become fully satisfied and thereby constant. For example, in Puerto Rico the
evapotranspiration demand is satisfied at all elevations when the rainfall
exceeds SOO centimeters, but the magnitude of this consumptive loss is a func-
tion of elevation.

The basic C vs BCI relationship has been tested in several ways with
satisfactosy results. Figure 3 will s e m e LO illustrate. Figure 3(a) presents
a coiqparisbii of calculated versus observed discharge for thirty streams in
Puerto Rico [4]. The calculated values were determined via conjunctive use of
the appropriate curve from Figure 2 and Figure 1. Since the qbserved records
wexe relatively short, many no longer than three years in length, the applicable
BCI was based on the average precipitation during the period of observed stream-
flow. BCI values for these streams range from 49 to 178. Figure 3(b) presents
the mean annual precipitation versus mean aqnual runoff relationship €or the
State of Kansas. The solid curve thereon was based on observed data from 122
basins [5]. The dashed cuí-ve was calculated using the Kansas curve of Figure 2
and Lhe basic coefficient chart of Figure 1. Basin BCI values for the ctndition
of mean precipitation range from 25 to 70.

The basic relationships can also be utilized to appraise possible stream

response under several yeats of above or below n o m a l precipitation. For exam-
ple, in recent years appretiable attention has been directed to the potential
application of weather modification tachniqves in improving water supply con-
dtionc. Although the bulk of the research effoxt has been directed toward
seeding techniques and understanding the mechanisms of cloud physics, several
investigators in the 1Jnited States, via the use of hydrologic simulation tech-
niques, have attempted LO explore how streams would respond to a given increase
in precipitation. The relationships in Figures 1 and 2 can be utilized to
estimate the percent gain in runoff thaL will oc €or a given increase in
average precipi tatjon. Let the subscript I represent natural conaitioi-s, sub-
script 2 represent augmented conditions, and the symbol PM equal P2/P,. Then
256

(PM1 c*-cl
Percent gain in runoff = 100 -
Ri
=
PC-PC
100 2 2 1 1 3 1 0 0
plcl cl
(41

Table 1 summarizes the impact of precipitation augmentation on water yield

as determined by hydrologic simulation and as reported by Linsley and Crawford
[6], Crawford [7], Lumb [8] and Smith [3]. The first three authors utilized
the Stanford Watershed Model and the latter utilized the Kansas Watershed Model.
In aggregate, these investigators conducted simulations on 14 separate water-
sheds, 13 in the United States and one in New South Wales. The last two col-
umns provide a comparison of the average increase in yield as determined by
simulation and as estimated by use of Figure 1. The calculations assumed that
the slope of the BCI vs P relationship was equivalent to the typical Kansas
curve. This approximation introduces some error because the slope of this
relationship does vary slightly from watershed to watershed, Nonetheless, the
calculated and simulated values are -most comparable. Examination of the com-
puter simulations again reveals that year to year increases scatter about the
mean value listed in the table.
-
Table 1 - Comparative evaluation of the impact of precipitation aiigmentatiun
- on mean yield.
,ength lbserved
-
of Period % Gain i Runoff
'eriod ainfall - -
Runoff Computer
'ears cm/year iainfall - PM #irnulatiori :alculated
One HundredlTen Mile Creek, 17 88.4 .205 1.05 16 17
Kansas - 17
17
88.4
88.4
.205
,205
1.10
1.20
33
74
34
70
Stranger Creek, Kansas 11 - 20 90.5 .200 1.10 35 32
Doniphan Creek, Kansas i/
Black Vermi lion River,
- 8
14
86.4
77.5
.261
.125
1.10
1.05
30
21
31
22
Kansas A 14 77.5 .125 1.10 41 40
14 77.5 .125 1.20 94 87
Salt Creek, Kansas 11- 8 58.0 ,066 1.05 23 26
8 58.0 .O66 1.10 49 52
8 58.0 .O66 1.20 107 117
S. Fk. Solomon River, Ks 11 - 20 53.1 .O57 1.10 41 53
Beaver Creek, Kansas
Cottonwood Creek, Calif. 21 -
21
2
45.6
40.9
.O16
,080
1.10
1.15
62
82
-51
78
Wollombi Brook, 3l
New South Wales -31 5 107.7 .141 1.10 35 40
Beargrass Creek, Ky T~ 5 110.6 .403 1.10 20 25
Arroyo Seco, Calif. -
41
5 68.2 .386 1.10 19 24
LaBrea Creek, Calif. -4/ 18 28.6 .O84 1.10 41 44
Dry Creek, California -- 22 130.0 .472 1.10 18 20
Saxtons River, Vermont 41
--
i/ From data presented by
16
rn 111.6
i70)
.499 1.10 - 19 20

2/ From data presented by Linsley and Crawford (1962)

--
3/ From data presented by Crawford (1965)
4/ From data presented by Lumb (1969)
-
5/ Not calculated
 257

Earlier reference was made to the fact that a plotting .of annual precipita-
tion-runoff values for a given basin will scatter about the mean annual rela-
tionship one develops with Figure 1 and the basin applicable Figure 2. Also,
it was noted that year to year percentage gains in flow from precipitation
augmentation, and as determined by computer simulation, would scatter about the
average gain observed for the entire period of record. This scattering is due
to the well established phenomenon of hydrologic persistence and reflects short-
term storage changes. Question arises, therefore, as to whether the relation-
ship can be used to determine flow characteristics other than the mean. The
answer is yes but a reasonable amount of judgment is required. Determination
of the distribution of annual flows will serve to illustrate.

Available climatic data can be utilized to develop the probability

distribution of basinwide annual precipitation, and the basin average curve
for Figure 2. When working with a basin whose geologic structure is not con-
ducive to the development of significant baseflow components, i.e., a basin
with minimum persistence characteristi.cs, an estimate of the prcbability dis-
tribution of annual flows can be developed by direct application of the pre-
cipitation probability function to Figures 2 and 1. Figure 4 provides a
comparison of calculated and observed annual runoff distributions for the
Marias des Cygnes River, Kansas, USA. This basin has little natural storage
and experiences a wide range.in annual precipitation, from less than 50 an to
more than 150 a.

Experience has shown that the foregoing approach is generally applicable

to the above average years. However, where lag or persistence is expected to
be a significant factor the lower portion of the distribution function should
be handled differently. In this case, replotting of the precipitation proba-
bility function using a two year running average will provide a more appropriate
solution. The effect, of course, is to convert the naturally skewed distribu-
tion which results from direct application of the basic coefficient relation-
ship to a more normal distribution so often encountered in the annual flow
relationship. Exercise of the judgment option inherent in the alternative
approaches outlined above requires that the planner be cognizant of the nature
of typical distribution functions in basins of similar geologic character.

For areas where freeze is of minor concern mean monthly yields can be
estimated by allocating monthly values in proportion to their contribution to
the BCI as defined in equation (3). However, this calculation should be made
using the average two month running total due, again, to the problem of lag.
Extension of this concept as a means of developing a stochastic generator of
monthly yield needed for preliminary appraisal of storage-yield relations is
currently underway. That is, monthly BCI values based on two month running
averages are being utilized to determine the regression, correlation, and
standard deviation parameters required for stochastic generation of long term
monthly yield 191.

Utility of the basic relationships can be extended to the determination

of additional flow characteristics with the acquisition of certain short-term
258

and miscellaneous field measurements. For example, experience has shown that a
daily flow duration curve obtained from a short-term record acquired over a pe-
riod of two to three years can be adjusted to a long-term appraisal if the ordi-
nates of the short-term record are expressed as a dimensionless ratio to the
average flow observed during the short record period. Subsequent mul.tiplication
of these ratios by the long-term mean as determined from Figures 1 and 2 will
provide a reasonable approximation of the long-term flow duration curve,

The relations described herein have also proven useful in appraising the
potential yield characteristics of coastal aquifers in southern Puerto Rico [IO'.
Historic groundwater use from these aquifers far exceeds the possib1.e direct
recharge assuming all the locally generated flow, as determined from Figure 1,
is di.verted to the groundwater aquifer. In this case the principle recharge
mechanism, excluding the recirculation effect of well irrigation, is infi.l.tra-
tion of surface water as it flows across the alluvial plain. Figures 1 and 2
were utilized to determine the mean surface inflow from the mountainous central
core at Lne point where the water entered the coastal plain. Following the
analysis of various short-term flow duration records which were available, this
mean yield was converted to a daily flow duration curve as described above.
Local stream seepage measurements , available from the U. S. Geological Survey,
were coupled with other similar information from prior studies to develop a
channel infiltration rate as a function of channel width and slope. Applica-
tion of the potential loss capacity of each channel to its flow duration curve
allowed subdivision of the surface flow into two components; the portion which
was infiltrated into the subsurface and the portion which escaped tcj th* sea.
An areal mass balance was then performed to determine the magnitude of trie
subsurface discharge to the sea (precipitation on the plain plus streamflow
from the mountains minus the sum of direct local runoff plus evapotranspiration
plus surface flow escaping to the sea). The recharge due to infiltration and
the subsurface discharge to the sea, both as determined above, were then incor-
porated in a subsequent mass balance of the subsurtace aquifer in which ïwliarge
was equated to ilet pumping (gross p-mpirig minus recirculation or return flow)
plu: subsurface discharge t u the sea. The calculations were repeated '01 con-
dition? other than the mean, e.g., the vondition of protracted drouth. Results
of these calculations provided a satisfactory explanation of the respons< in
aquiftr water levels that has been expeiienced during both noimal and subnormd
climatic condi t I .,ns.

One additional word of caution beforc losjxg this discussion. The

estimates obtained from Figures 1 and 2 as2uine natural catchments, and wtural
climatic ,oriditLons. Whenever man's activities have materially altered he
naturai environment, e.g., by applicatio? of irrigation water or the I U I ~ ruc-
tion of appreciable impervious areas, adjustments must be niade. The fc,l 3ing
will serve L'J illustrate.

Thr lower 6500 hectares of Cherry Creek, Colorado is partially iirbanized.

Approxinial eli' 1 5 percent of the area is impervious surface for which hie run..ff
coefficient appioxiniates 0.9, and an addit,(nai 10 perrent is in urban awns
which are heavily irrigated (an average of ~uciut t.7 cm per year). 'Hie m e m
a n m a l rainfall appri,.hates 38 m . n d ';c tesponding BCI Is 32. The ad-jii.;. I:
 259

BCI for the irrigated area approximates 87. An estimate of annual yield, with
and without consideration of the man induced changes, is summarized below.
Percent Moisture Applied Weighted Runoff

Natural Conditions
Area
100
(a.)
38
-C
.O25
cm.
.97
Modified Conditions
Natural 55 38 .O25 .52
Impervious 15 38 .goo 5.12
Irrigated 30 104 .320 -10.00
15.64
The observed mean annual discharge from this 6500 hectare portion of the basin
for the four calendar years 1966-69 was approximately 16 centimeters.

The foregoing example is of interest on two counts. First, it illustrates

the procedure required for adjusting the appraisal of mean yield to accommodate
significant man induced changes. Second, it provides a relatively unique exam-
ple of the impact of urbanization on flow response.. In many areas the effect of
urbanization is to reduce the opportunity for recharge and thus diminish baseflow
contributions. The reverse is true for the example cited above. .Here, the im-
pact of lawn irrigation in a relatively dry climate has created a substantial
baseflow contribution to the stream and a significant increase in overall yield.

In summary, precipitation and temperature measurements often represent the

only significant hydrologic data available to the water resources planner. AE
empirical function relating the mean runoff coefficient to the aforementioned
parameters hac been developed. The relationship has been tested in a wide
range of environments, and has proven most useful in undertaking preliminary
assessments of water availability. The general utility of the relationship can
be extended appreciably with limited field data and the application of basic
hydrologic concepts. Continued exploration of the utilization of climatic data
in the preliminary appraisal of water yield characteristics must be encouraged.

Acknowledgements

A portion of the material described herein was deve.loped during conduct of

a research project sponsored by the Kansas Water Resources Research Institute
and the Office of Water Resources, U. â. Department of the Interior. The
author is also indebted to Black & Veatch, Consulting Engineers, Kansas City,
Missouri; R. A. Domenech & Associates, Hato Rey, Puerto Rico; and the Puerto
Rico Aqueduct and Sewer Authority for permission to cite information developed
by these several offices in their analysis of water availability in Puerto Rico.

References Cited

1. Guisti, E.V. and Lopez, M.A., (1967). Climate and streamflow of Puerto
Rico, Carribbean Journal of Science, Vol. 7, pp 87-93.
260

2. Thornthwalte, C. W., (1931). The climates of North America according to a

qew classification, Geographic Review, Vol. 21, pp 633-55.

3. Smith, RL., (1970). Water utilization aspects of weather modification in

Kansas, Contribution No. 46, Kansas Water Resources Research Institute,
Lawrence, Kansas.

4. Black & Veatch - R. A. Domenech & Assoc., (1971). Water Resources of

Puerto Rico, phase 2, surface water appraisal, Puerto Rico Aqueduct
and Sewer Authority, San Juan, Puerto Rico.

5. Furness, L.W., (1959). Kansas streamflow characteristics, part 1, flow

duration, Kansas Water Resources Board Technical Report No. 1, Topeka,
Kansas.

6. Linsley, B.K. and Crawford, N.H., (1963). Estimate of the hydrologic

results of rainfall augmentation, Journal of Applied Meteorology,
Vol. 2, NO. 3, p p 426-427.

7. Crawford, N.H., (1965). Hydrologic consequences of weather modification:

case studies, Human Dimensions of the Atmosphere, University of Chicago
Press, Chicago, Illinois, pp 41-57.

8. Lumb, A.M., (1969). Hydrologic effects of rainfall augmentation, Tech.

Report 116, Dept. of Civil Engineering, Stanford University, Palo
Alto, California.

9. Thomas, KA., Jr. and Fiering, M., (1962). Mathematical synthesis of

streamflow sequences for the analysis of river basins by simulation,
Design of Water Resource Systems, Chapter 12, Harvard Press, Cambridge,
Mass achuse t ts.

10. Black & Veatch - R. A. Domenech & Assoc., (1970). Water Resources of
Puerto Rico, phase 1, ground water appraisal, Puerto Rico Aqueduct
and Sewer Authority, San Juan, Puerto Rico.
 261

o -7

O .6

o -5
z
9
t-
U
k
0
o
En. 0.4
z
U
w
2

kI* 0.3
COORDINATES
O COEFFICIENT
z
3
K
z
Q
w
E 0.2

110 .430
0.1 I20 .470
140 .535
I GO ,583
I RO .624
20 o .655
o .o I I I
O 40 80 120 I60 200
BASIN CLIMATIC INDEX

Figure 1 - Basic climatic index related to the ratio of mean runoff

divided by mean precipitation
262

300
PUERTO RICO CURVES
ME& BASIN ELEVATION
200 - 1500 METERS
1000 METERS
x
Li1
SOO METERS
O
z
o100
i=a
-
-
z
J TYPICAL
o
5
cn
a 50-
m
CURVE
-

20
20 50 100 200 500
MEAN ANNUAL PRECIPITATION - CENTIMETERS
Figure 2 - Selected examples of the relationship between precipitation
and basin climatic index
 263
OBSERVED MEAN D f S C H A R G E IN CMS

‘5
w
1.0
E 40.
0
w2 0.5
J 2 o.
3
o
-1 cn
a u:
O IO. E
W
o.2 _-
T.
I-
z
w
MEAFJ ANNUAL RAINFAL.LA 4. u
RUNOFF RELATION FOR I
STATE OF KANSAS, USA. IA.
L
O
2. z
3
u
J
Q
I 3
BASED ON 122 DATA POINTS ( F U R N E S ) z
z
<
CALCULATED USING STANDARD COEFFI- z
CIENT C H A R T a
w
1.4 2

).2
I I I -
40 80 100
60
M E A N A N N U A L PRECIPITATION - I20
CENTIMETERS
Figure 3 - Sone canparative rec4ults obtaiiied with the basic BCI vs C
relationship
264

IO

O
in
ó
J
a
3
U
W
2.
k
4
ai
2
0 1.0
a:
a
k-
a
W
3 0.5
9
-J
\
4
æ
-++H-OBSERVED ANNUAL RAINFALL \
O000 O B S E R V E D ANNUAL S T R E A M F L O W
z
z CURVE FITTED TO RAINFALL DATA
\
a O ooo\
*I*
---- CALCULATED STREAMFLOVJ C U R V E
0)
\
IJJ
3
o \
-J O \
3
J 0.1
rr(
53
z
z \
a \
O L
0.05

__
.O5 .I5 .30 .50 .70 .85 .95
PROBABILITY OF EXCEEDANCE
Figure 4 - An example plot of calculated versus observed probability
distribution of annual flows
 DETERMINATION OF HYDROLOGICAL CHARACTERISTICS
IN POINTS WITHOUT DIRECT HYDROMETRIC DATA

S i 1v iu S t an e s c u Jc

ABSTRACT

In Colombia, hydrological information is very scarce.

Consequently no direct hydrometric data are available for most
of the sites of projected hydrotechnical works and exploitation
of water. Therefore one must generally apply methods of genera-
lization, transfer of direct information from observed points
to points of interest, and indirect estimation of the hydrolo-
gical characteristics. In relation with this, there are several
proceedings of indirect determination of mean, maximum and
minimum runoff, as well as o f other characteristics of the
hydrological regime which are applied to the concrete conditions
of Colombia. The examples, which are included to illustrate the
application of the methods pointed out, are selected from complex
hydrological studies, elaborated or in the process of elaboration,
within the frame of the activities of interpretation and hydrolo-
gical calculations worked out in the Colombian Service of Meteoro
logy and Hydrology.
-

RESUMEN

En Colombia, la información hidrológica es muy escasa. Co-

mo consecuencia, la mayoría de l o s sitios de proyectos de obras-
hidrotécnicas y aprovechamientos de agua no disponen de datos hi-
drométricos directos. De tal manera, se deben aplicar ampliamente
métodos de generalización, de transferencia de información direc-
ta desde puntos en que se dispone de observaciones y mediciones -
hacia puntos de interés práctico, así como métodos de calculo in-
directo de las características hidrológicas. Relacionado con esto,
se presentan varios procedimientos de determinación indirecta de
caudales medios, máximos y minimos, a s í como de otras caracteris-
ticas del régimen hidrológico, aplicados a las condiciones concre
tas de Colombia. Los ejemplos que se incluyen para ilustrar la -
-
aplicación de los métodos indicados son seleccionados de estudios
hidrológicos complejos, elaborados o en curso de elaboración den-
tro del marco de labores de interpretación y cálculos hidrológi--
cos desarrollados en el Servicio Colombiano de Meteorología e Hi-
drología.

* Hydrology Expert - World Meteorological Organization and United

Nations Development - Program. Bogotá, Colombia
lombiano de Meteorología e Hidrología.
-
Servicio Co--
266

Introduction and generali ties

In Colombia, owing to the scarcity of direct hydrometric
information, in most of tile specific cases, the sites of hydro-
technical works, water uses or other works which come in direct
or indirect contact with the rivers, d o not coincide with the
sites of the hydrometric stations. In such cases, the necessary
hydrological parameters must be determined by methods of indirect
estimation. The indirect nydrologic estimations are not however
considered as final values, but only used in an approximate way
or for guidance. The magnitude of these estimations is verified
by means of hydrological field activities, which are generally
grouped into two categories:
a) Temporary hydrometric stations.
b) Expeditional hydrological activities.
T h e temporary hydrometric stations with intensive and complex
programs of observations and measurements have been widely used
in Colombia, and in particular for dam and reservoir projects
for hydroenergetic purposes, drinking and industrial water uses,
for the most important towns in the country, irrigation and
drainage districts of national importance, and, to a lesser
extent, for navegation, construction arid protection of bridges,
and also for preservation of hydrographic basins.
The expeditional hydrological activities have also encountered
a very wide field of application, especially in aqueduct projects
for medium and small towns, road bridges and sewers, land use
programming, reforestation programmes,.prefeasibility studies for
work projects which come in direct or indirect contact with the
water currents, and also when planning the u s e of hydric resources.
Regarding this latter aspect, the expeditional hydrological
activities not only act as a means of verifying indirect Iiydro-
logical estimations, but tiicy moreover constitute almost UE only
really acceptable and reliable way of assessing the scope of the
hydrological parameters in over half the national territory (in
other words 60ú-000 km2) , where the stationary hydrological
activities are almost completely missing.
T h e use o f temporary hydrometric stations is very similar t o
the use of what are normally called secondary stations. As this
is generally a known method, no details on this question will b e
given. Only the most important features of the problem will b e
presented. More emphasis however will b e made when describing
the most usual methods of the expeditional hydrological
activities, since i n Colombia, these vast fields of application
are not only found in the past, but also i n the present and the
future.
 267

Verification of hydrological estimations by m e a n s of

temp o r a ry hy d rometric stations,
T h e hydrometric activity of one or several temporary
stations with complex and intensive observations and measurements
programmes, verifies and completes the hydrological estimations,
riot only by means of direct data it provides during its operation,
but also through the possibility of spreading its series o f direct
data, by means of correlatioris with data of other reference
stations, which have been working for over 15 years, in zones
of similar hydrological characteristics.
In most cases, preference is to install one o f the temporary
hydrometric stations exactly in or near the sites of the projected
works. The stretch o f river corresponding to the work site does
not always however meet satisfactory conditions to install a
hydrometric station. In such cases, various hydrometric stations
must be installed in the hydrographic basin where the site of the
works is found, and later deduct the hydrological parameters
by interpolation, balance of discharges or relation w i t h physio-
graphic or morphometric characteristics of the basin,
In many cases, although a temporary hydrometric station can
be installed in the very site of the projected works, it proves
preferible to install several more stations in the corresponding
hydrographic basin, in order to have possibilities of
controlling the activity developed in t h e station of the works
site, complete the eventual gaps in observations and measurements,
avoid errors and confirm the results,
T h e working duration of the temporary hydrometric stations
is variable, pursuant to the specific conditions. W h e n the
temporary station is right in the river as .the reference station,
and the areas o f its hydrographic basins differ by less than lo%,
the correlation is generally established in o n e year alone.
\then the two stations are in the same river, b u t the areas of
their basins differ by over lo%, the time needed to establish
a reliable correlation takes various years, In short, w h e n the
temporary station is not i n the same river as the reference one,
the operation of the first one does not end when an acceptable
mathematical correlation is obtained, but as soon a s this is
physically verified, during a period which contains sufficient
humid and dry average years, in other words, rather representative
for the average nultiannual situation, Otherwise the correlation
obtained, although good from a mathematical point of view, m a y
only express a temporary situation, which would lead t o great errors
Besides the above cases, situations have occasionally been
found where the reference station was too f a r from the site of
the projected works. Hence a direct correlation was practically
impossible to establish. T h e problem could however be solved
with various intermediary stations, which facilitated the transfer
of data, by means of chain correlations.
268

The decision to suspend the running of a temporary hydro-

metric station has always constituted a great difficulty.
Generally, only in a very few cases can the duration o f operation
o f these stations be really considered sufficient. Therefore,
even after the works execution has commenced, it is considered
preferible to continue running temporary stations n e a r these
work sites, and these stations sometimes remain in operation
e v e n after the corresponding exploitation has started. T h e
supplementary data supplied by these stations prove highly useful
to complete and confirm the hydrological estimations, and also
as guidance for eventual improvements in both the works and in
the water uses programmes,
Verification of indirect hydrological estimations by
cxpeditional methods
The verification of hydrological estimations by means of
temporary hydrometric stations with intensive and complex
programmes of observations and measurements, constitutes a superior
method, from a qualitative point o f view, in relation with the
verification of estimations by means of expeditional hydrological
activities. It is not always however possible to install and
operate stations in the sites of the projected works and water
uses, or in their hydrographic basins. In scarcely populated
regions, it is difficult t o operate hydrometric stations, but
the estimation of the main hydrological parameters of these
areas is essential to plan the uses of hydric resources o n a
long term basis and also for prefeasibility studies. This
situation is due to certain specific conditions, In Colombia,
in more than half the national territory, the land communication
lines are completely or partly missing, o r are in a deficient
state, such that in rainy periods, penetration is rarely possible.
In these areas moreover, it is very difficult to find
satisfactorily qualified people to act as hydrometric observers,
and the installation of limnigraphs, in areas which have n o
watchkeepers, generally proves a hazard and failure.
T h e lack of hydrometric networks, and the difficulty of
organizing stationary hydrological activities in almost half
the national territory, constitute conditions which favour the
wide use of expeditional hydrological activities. Although
these cannot give well defined determinations of the hydrological
system characteristics, as the Stationary systematic hydrometry
cannot be substituted, they constitute essential work to verify
hydrological estimations or to obtain approximate o r guide
indications in isolated regions of difficult access, where the
installation and operation of hydrometric stations fail t o
encounter satisfactory conditions.
T h e hydrological measurements in campaigns are frequently
applied i n Colombia to determine the following factors:
 269

a) Maximum discharges, duration of floods and

time o f wave propagation;
b) Minimum flows and duration of low waters;
c) Sediment charges;
d) Water temperatures;
e) Physical, chemical and biological
characteristics of the water;
f) Overall hydrological characteristics of
the currents.
üetermination of maximum discharges and flood characteristics
b y means of hyd rological expeditions
In most of the concrete situations (except the case of
flood sweeping), the aim is to determine the maximum runoff,
pursuant to information on maximum historic levels and maximum
floods known in the region, which supposes the almost total
absence o f evident traces of maximum waters in the beds of the
currents.
Thus, the information obtained on the field, acquires
decisive importance. It can be classified into two categories:
a) Information supplied by the river bank dwellers,
b) Microphysiographic analysis in the largest beds
of the rivers and o n their banks.
The information solicited from the river-bank dwellers
refers to the following factors:
a) The maximum level of the greatest flood known,
b) The year and eventually the date when the flood
came about,
c) The time the flood waters took to reach their
maximum level,
d) T h e time the waters took in dropping to their
normal levels,
e) Eventual artificial influences o n the maximum
runoff system,
Pursuant to the specific possibilities, the information
on maximum waters is solicited from river-bank dwellers who
have lived on the premises for over 30 years and the questions
are put to various people, in order to have a chance to compare
replies,
T h e microphysiographic analysis in the largest beds of
the currents and on their banks, consider the possibility
of finding certain traces about the maximum water levels,
These analysis generally refer to the following factors:
a) Geomorphological aspect,
b) Alluvial material and grounds;
c) Vegetation and organic vegetable material
270

The geomorphological or morphohydrographic aspect o f the bed

constitutes the first sign o n the possibility o f the river waters
overflowing. The indicative details are micromorphological
aspects of very recent age, traces of erosion processes, alluvial
formations scarcely fixed by the vegetation etc, Sometimes
one can even determine lines o f separation between the lowest parts
of the banks, characterized by very recent morphogenetic
processes, and the upper parts of same, relatively fixed and o f
more advanced evolution. A l l the information resulting from
the detailed analysis of the bed micromorphology, cannot lead
to an exact determination of the maximum level of the waters, but
it does offer a first and very useful general guidance, on the
extension of the maximum flood and its possible lines o f demark-
ation on the banks or in the largest bed.
T h e micromorphological analysis is completed with observ-
ations on the alluvial materials and the soil. The fine sediments,
coming from the smaller bed, found on the banks, are a sure sign
of flooding. The secondary soil, discontinuous on surface, and
those which are scarcely at the beginning of the formation
processes, also indicate the overflow of the waters, Finally,
the mineralogical analysis of the fine sepry recent sediments
o f the largest bed, may indicate the presence of materials which
are not of that place but come from upstream in the section under
study, which indicates the flooding of the larger bed. The
vegetation can also indicate the overflow o f the waters, O n the
one hand, the discontinuity of the vegetable formations indicates
the approximate limit of the flood. On the other hand, the
detailled inventary of the vegetable species of the area can
constitute a highly important piece of information, because all
vegetable material which is different, in the larger bed, may
have been brought by floods from upstream. The detailed
laboratory analysis of the vegetable content of the sample
sediments aiid soils may lead to decisive results, when pollen
particles are found in them which do not belong to the vegetable
species of the area under study, but to others from upstream zones.
During the .land activities, the river-bank dwellers'
information is always completed w i t h microphysiographic analysis
made in the largest beds o f .the currents and on the banks o f same.
Without these analysis, the riversiders' information cannot be
verified, and can consequently i d to very great mistakes, T h e
errors arise from subjective reasons which make the riversiders
hide the truth or merely offer information o n unknown events.
The microphysiographic information, although unable to fix the
maximum level of the waters, indicates essential approximations,
as general guidance and verification factors.
Besides finding information on the characteristics of the
maximum runoff (iiiformation frorii river-side dwellers and micro-
physiographic information), the following main operations are
carried out in each section studied:
 271

Survey of three cross-sectional profiles, spaced at

equal distances or more, of the river width, and
continuing for n o less than 1 m , above the maximum
historic level of the waters. During the survey, the
maximum levels are markeù on the profiles, and any
lithological sign of soil or vegetable removed from
the place;
Survey of the longitudinal profile of the current,
with a length equal to or at least 5 times the width
of the river.
Execution of at least one gaging (if the natural
conditions so permit)
Approximate drawing of the river span, including the
largest bed, the marking of the cross-sectional profiles;
indications on the types and sizes of lithological
materials and vegetation of the largest and smallest bed,
the lines defining the maximum flood, certain reference
elements, etc.);
Sampling of sediments of the smaller bed and of alluvial
material, and eventually soils from the larger bed and banks,
along the cross-sectional profiles made;
Inventary of the vegetable species of the area and eompiling
of vegetable remains differing to the local species,
The litliological samples of soil or vegetables are suitably
packed, and all the necessary references are marked on the
packages to establish the site from which they have been taken.
Afterwards, pursuant to possibilities, the samples are analysed
in the laboratory.
The above activities are made in various representative
sections of the hydrographic basin studied, in order to have
sufficient data available to permit a comparison of values,
an analysis of the territorial variation of same and generalization
of runoff maximum. During the field work, the information from
different sections are permanently compared, bearing in mind the
territorial continuation of the processes, the variation of
the magnitudes, the periods and dates on which the events have
come about, etc,
Once the field and laboratory activities have beencompleted,
the following factors are determineo during office work:
a) Maximum discharges of homogeneous probability (generally 1%).
b) Main flood characteristics;
c) Eventually, time of wave propagation.
T h e discharges corresponding to the maximum historic levels
are calculated by hydraulic methods. T h e measurements made
during the expeditions help to determine the hydraulic formula
factors which contain the rugosity coefficient. These values
are not used directly when estimating the maximum discharges,
but merely offer comparison criteria. Once the maximum historic
27 2

discharges corresponding to a certain frequency have been

calculated (for example 3% if they have been produced in
30 yars), the values should be increased, in accordance with
the coefficients, which permit one to pass from larger
frequencies to rare occurrences. Thus a homogeneization of data
is made (the 1 % probability is convenient), essential for
comparisons and generalization. In order to change values
of various probabilities into 1% probability values, it is
preferible to use coefficients established with base on the
direct hydrometric data available in the same zones or in
regions which are hydrologically similar. If these completely
fail, coefficients will then be used estimated with base on the
theoretic frequency curves, considered adequate for the region
under survey,
A final verification of the 1% maximum probability discharges
is made through generalizations of various forms. The most
comfipn are the type: Qmax = f (A); lg qmax = f(1gA); qmax =
f ( ); etc, where Qmas = maximum discharge, in m3/s; A -
area
of R e basin in km2; qmax = maximum yield, in l/s/km2, or mm;
H m = average elevation of the basin, in m ; n = a subunit exponent,
specific for the natural conditions of a given zone; f = a
different function for each zone.
The chief flood characteristics (swelling time and total
duration of same) are also verified by comparison of data and
generalizations. These latter are determined by reason of
various morphometric and physiographic factors of the hydro-
graphic basins (length of currents, gradients of same, etc,)
Likewise, the time of wave propagation is also verified and
defined. The generalizations are generally determined in
relation with the lengths and gradients of the currents,
The final verification of the results is made b y comparison
with the direct hydrometric data available in the region. Thus,
it is not acceptable that the 1% probability maximum discharges
estimate? by expeditional methods, be less than the discharges
measured in hydrometric stations, during short intervals,
Determination of minimum discharges and duration of low waters
by expeditional methods.

In most of the concrete situations, the minimum low water

characteristics are determined during the expeditions made to
find the maximum runoff, subject to the condition that these be
made during low waters.
T h e activities developed on the field have three categories:
a) Compiling of information froin the riversiders.
b) Hydrological and topographic work in the bed of the current.
c) Observations on the lithology and freatic layers of the region,
The reports from the riversiders refer to the following aspects
 273

a) Eventual interruption of the runoff,

b) Minimum historic levels;
c) Year and month when the runoff w a s interrupted o r when
the minimum level came about,
d) Low waters and duration of same,
e) Eventual artificial influences on the minimum runoff system.
Preferibly the information is requested from various people
w h o have lived near the river, for ovcr 30 years. The most
marked sections €or analysis are those which pertain to spans
of current where ancient floodgate openings are found, a n d also
the sections near to irrigation land. T h e existence o f
derivations, upstream from the section under survey, must be
considered, to avoid considering the minimum discharges in
influenced state as minimums in natural state.
Tile hydrological aiid topographic work in the bed refer to
the following:
a) Execution of measurements;
b) Topographic survey of loiigi tudinal prof iles,
T h e measurements are generally made by wading. After making
the measurements, the wet section is drawn and on this, the line
of the surface of the water corrcsponding to the lowest water
(in accordance with the information on minimum historic levels).
T h e topographic survey of longitudinal profiles is m a d e on
the water surface, and spreads for at least three times the
width o f the lower bed. These operations are made in various
sectioiis representing the basin or area under survey, which
are generally assimilated in the main confluences, T h e inform-
ation is permanently compared, bearing in mind the territorial
continuity of the hydrological phenomena.
Throughout the hydrological expeditions , the lithology of
the region is continually observed. If geological maps are
available, they are taken to the field, to have prior indications
o n the areas where there are permeable rocks Any discontinuity
in the runoff during low waters should be explained either a s
a result o f human activities or due to lithological influences,
Research o n the depth of the freatic layers, in existing wells,
is also made, and also on possible contacts o f these layers with
the flows, which could constitute an important additional inform-
ation for estimating the minimum runoff characteristics.
T h e estimations, interpretation, verification and generalization
of data are made at a later stage, at the office. T h e minimum
discharges are estimated by means of hydraulic methods, For the
factor containing the rugosity coefficient, the values are used
which result from the measurements made, The discharges of
diverse statistical probabilities are transformed into 9 7 %
probability discharges (three times in 100 years) to obtain liomo-
geneous values which can be compared. The coefficients used t o
274

change values of greater frequency into values o f lesser

probability shauld be determined based on direct hydrometric
data o f the zone or regions with similar hydrological system,
If these fail, determinate coefficients may be used based o n
the theoretic frequency curves, considered adequate for the
region studied.
In the event of intermittent runoff flows once in 30 years,
all the minimum discharges w i t h probabilities above 959. may
be considered the same o r zero.
T h e 97% probability minimum discharges are firstly analysed
in relation with the areas o f basins and by means of balances of
discharges. T h e yields are analyzed by means of generalization
relations, which may be o f type qmin = f (iim) for mountainous
areas and qmin = f (B%) o r qmin = f (Ud) for flat areas, In
these relations qmin = minimurn yield, in l/s/km2, or m m ; Iim =
average elevation of the basin, in m ; B% = forestal covering
coefficient o f the basin, in % ; Ud = drainage density or density
o f the hydrographic network in km/km2; and f = a different function
in each zone.
If maps are available with monthly mean isohyets, the minimum
yields may be compared w i t h the monthly mean precipitations of
the driest month, by means of relations of type qmin = f(Pm),
where Pm = mean precipitation of the driest month, in the basin
correspoiiding to each section, in mm.
T h e duration of the low waters is analysed from the point of
view of territorial continuation o f the hydrological phenomena
aiid moreover, in relation with the distribution of the precipit-
ations within the year, When special meteorological maps are
available, indicating the average duration of the drought periods,
this data can be used t o verify the maximum low water durations,
by means of relations o f type T e = f(Ts), where T e = time or
duration o f the drougnt period, in days; and f = a different
function in each zone.
Determination of sediment charges by expeditional methods
T h e hydrological campaigns to determine sediment charges
are made during high waters periods and after floods of certain
importance have occurred, T h e expeditions organized during high
waters periods try to determine sediment charges in suspension,
whereas t h e others refer to haulage volumes.
The main activities to determine sediment charges in suspension
are a s follows:
 275

a) Sampling o f waters with suspensions;

b) Execution of measurements;
c) Geomorphological observation of the land.
The water samples are taken during tlie execution of
measurements, in tlie same points where tlie speeds o f the
water are measured. T h e measurements and water sampling
are made in various characteristic sections of the hydro-
graphic basin or study zone, where bridges or other facilities
are available to execute the gagirigs.
iluring these runs, permanent geomorphological observations
are made on the existence of erosion processes, land degradation,
lithological conditions ardvegetation, and also their relation
with washing of the soils, etc. A l l these factors help to
explain the abrupt changes in the territorial variation of the
sediment concentration. To make our work easier and as general
guidance, it is convenient to take t o the field the lithological
or geological, general geomorphological and special geomorphological
maps (gradient, fragmentation of the relief, erosion, etc) if
they exist.
Once they have been estimated, the sediment charges in
suspension are analysed, bearing in mind the territorial
continuity o f the hydrological processes, Any discontinuity
ihould be explained by tlie different contribution of any o f
the affluents, or by evident morpholithological changes in the
basin, which determine changes in the erosion and the transport
of sediments. The concentration o f sediments in suspension m a y
moreover be analysed in function o f the territorial variation o f
tlie corresponding runoff, within the hydrographic basin o r area
studied.
In order t o establish the magnitude of the averages of sedi-
ment concentration in suspension, various hydrological campaigns
are made, until relations between discharges and sediment
charges, i n various characteristic sections can be determined.
Thus, the verification and generalization of concentrations or
sediment charges in suspension is made through the flow and run-
off magnitudes,
Finally, the verification of the magnitude of the average
values o f eoncentretion of suspensions is made, by morpholitho-
logical zones, in function of the variation of the gradients,
coefficients o f covering with forestal vegetation, etc.
T o find voluniEs of dragged sediments, certain activities are
carried out, during low water periods, and the following are the
most important among these:
a) Set up marks and fixed reference points.
b) Topographic surveys of alluvial accumulations in f l o w beds;
c) Set up and recover traps for sediments and measure accumul-
.
a t i on s
 276

These operations are performed in various characteristic

sections , generally downstream of important confluences regard-
ing the drags contribution. To obtain a general idea on tlie
size of the sediment charges dragged along, various campaigns
are made. In the first, the marks and fixed reference points
are set up on the banks, in the larger lied, and sometimes even
iii tile smaller bed of the currents, and also the sediment traps
i n the smaller and larger river beds. In later campaigns,
topographic surveys are made of the alluvial accumulations;
the sediments accumulated in the traps are removed; the marks
are repaired and also the reference points that have been
damaged during floods, and the traps are again set up for bottom
sediments.
l h e dragged sediment charges are analysed in relation with
the magnitude of the discharges and suspension charges, and also
in function of the morpliolitliolpgical local conditions (litho-
logical complexes, erosion and gradient processes, etc.)
Finally, coeffients may be established which, for each zone
of specific morpnolithological conditions, indicate the magnitude
of tlie proportion that the sediment charges dragged along
represent , i n relation with the suspension charges.
Determination of water temperatures by expeditional methods
The hydrological expeditions which determine the water
temperatures, refer to the following operations:
a) Measurement of air temperatures.
b) Measurement of water temperatures,
c) Observations on the land lithology.
The air temperatures are measured in order to have values
available to determine correlations between these and the water
temperatures. Once the correlations have been established,
characteristic values and the variation in space and time of
the water temperatures can be determined, based on the values
o f t h e former, Naturally, in such cases, maps with isotherms
of the air in tlie surveyance regions are available,
‘She water temperatures are measured in various characteristic
sections, parallel with those of the air temperatures. The
variation in their values is analysed bearing in mind the territ-
orial continuation of the hydrological processes, Aiiy jump in
the water temperatures, throughout a flow, should be explained
either by confluences with different temperature flows, or by
imp0 r t ant prouiid water contributions,
Observations oii the lithology o f the region are made to
detect possible substantial ground water contributions, and related
t o this, explain the sharp changes in temperature experienced by
the waters throughout the flows,
 277

in order to compile representative data not only from

the territorial variation point of view, but also regarding
tiic temporary variation, expeditions are made tliroughout all
the seasons of the year.
Determination of physical, chemical and biological character-
istics of the water by means of expeditional methods
T h e campaigns to determine the quality of the waters are
organized during low water periods, when the physical , chemical
and biological characteristics of t h e f l o w waters are most
stable.
Iii cases o f waters whose quality is unchanged by human
activities, the characteristic sections for expeditional w o r k
are the confluences. To the contrary , the conf luences w i t h
drainage and sewerage are also taken into account,
The main activities carried out during the campaigns arc
a s follows:
a) Compiling o f water samples;
b) Execution o f measurements;
c) Analysis of water samples , and eventually, preservation
and packing o f same;
d) Geological observations.
The water samples are analysed on the field, if mobile
laboratories are available (the most suitable]. When there
are n o possibilities of making complete analysis on the field,
the samples are preserved, and at least the analysis o f the easily
changeable characteristics are made, anù which can be only
determined i n fresh tests. ‘Iiie samples sent to the laboratory
are suitably packed, and all tiic indications regarding the site,
and date of collection are noted o n the packets.
T h e measurements are made to find the discharges t o w h i c h
the characteristics measured correspond, anù also the amounts
of waters available for dilution o f chemical coiicentrations,
of vital use, especially in cases of pollution tipping.
The geological observations are similar to those made during
expeditions to find the miiiimum runoff characteristics. In the
case of physical, chemical and biological qualities of the f l o w
waters, any sharp change should be explained either by artificial
influence (tipping of pollutions), o r by natural influence, due
to confluences w i t h flows of different biological, chemical and
physical characteristics, o r due t o an abundant food o f ground
waters from different lithological zones,
As soon as thc physical, cliemral and biological character-
istics o f the water have been determined, they are analysed,
bearing in mind the territorial continuity of the hydrological
278

processes, and they are verified, according to litliokgical

zones, in relation with the discharge and runoff magnitudes.
Determination o f whole hydrological characteristics of the
Flows by means of observations and measuremeiits on campaigns
I n practice, the caso very frequently turns up of there
being no direct hydrometric data available in certain Iiydro-
graphic basins, or estimations of various hydrological
characteristics must be verified.
In such situations, complex expeditional hydrological
activities are developed, based essentially on tiie following
p r i nc ipk s :
a) iixccution of simultaneous measurements hydrologically,
in various sections;
b) Periodicity of campaigns, in accordance wi tli the hydro-
logical method phases;
c) Installation of recorder apparatus, with long duration,
autonomous operation;
Ci) General Observations o11 the genetic characteristics of
the hydrological sys tem.
The measurements may refer to most of the hydrological
characteristics (levels, discharges, sediment charges, temperature
and physical, chemical arid biological characteristics of the
waters, etc,) and they are made in various representative
sections of the basins under survey, and also in nearby
hydrometric stations, located iii areas with similar hydric
system.
T h e principle o f Iiydrological simultaneity should be strictly
respected during the measurements, in order to compare and
correlate the results. From an operational point of view, this
supposes a need to execute w o r k by means of various teams of
hydrologists working parallel, in accordance wi tli strictly
established programs regarding the sections and measurement hours,
The periodicity o f the campaigns, iri functinn of the hydro-
logical system, is irnposcd a s a compulsory condition, in order
to establish the variation ranges of the characteristics measured
and the correct correlations between the data of n r i o u s sections
arid those of the nearby liydrometric stations.
"lie installation of recording apparatus, of long duration
autonomous operation, is convenient when the periodicity and
frequency o f the campaigns cannot be assured on a satisfactory
level, and also when oiie is trying to complete the correlations
between the data of the sections studied and the reference
hydrometric stations. tiowever, in most specific cases, the land
difficulties prevent execution o f works for installation of
recorder apparatus (lack of roads and labour; the maintenance and
 279

periodic inspection of the installations and apparatus cannot

be assured, etc.)
The measurement o f tlie hydrological characteristics i s
organized i n accordance w i t h the indications given in the '
above paragraphs.
T h e data analysis is made bearing in mind the territorial
continuity of the hydrological processes and the correlations
between various sections and the reference hydrometric stations,
arid also in terms of the local physiographic conditions
influencing the variation of the hydrological system factors,
Naturally, the most important thing is to properly determine
tile correlations so as to extend the series of data of the
sections studied, in terms o f the long series o f data available
in the reference hydrometric stations. I t is therefore
convenient for the expeditioiiai hydrological activities t o b e
developed in each zone, at least during two complete years.
The verification, analysis and interpretation of the data
is made before suspending the field activities. Pursuant to
tlie results, the initial programmes can be changed and t h e
work intensified, to define the processes which have not yet
been satisfactorily determined. The total suspension of the
expeditionai hydrological activities in the study area can only
be macle after conclusive results have been obtained, or,
exceptionally, w h e n the sure conclusion is reached that the
methods used are sufficient to deterriiine or verify the hydro-
logical characteristics w h i c h must be known.
bibliography
1. Uiaconu, C., Lazarescu D, (1965). Iiidrologie, Bucuresti.
2. Irjorld Meteorological Organization (1970). Hydrometeorological
Practices Guide, OMM-No. 168 TP. 82, Geneva.
3. Roche, M. (1963). llydrologie de suÏface, Paris,
4. Stanescu, S. (1969). Chief prcsent day problems of the
national network organization of hydrological stations
in Colombia, Aperiodic Publication 1 , SCMII, Bogota,
5. Stanescu, S. (1971). Expeditional Iiydrological Activities.
Aperiodic Publication 22, SCMiI, Bogota.
6. Vircol, Al. (1960). Calculul debitelor maxime folosind
cercetarile expeditionare, Studii de liidrologie 1, Bucuresti.
7. World Meteorological Organization (1972). Casebook o n
Hydrological Network Design Practice, WMO- No. 324 , Geneva.
280

E STAC I O N M E T E ORO L O G IC A I N G E N I O M A N U EL I TA
A
PROMEDIO 1901 -1970

A
1100

Ah0

ESTACION HIDROMETRICA C A U C A -JUANCHITO

PROMEDIO 1934-1970

260

24 O
1935 1940 1945 1950 1955 1960 1%5 1970

COMPARACION DE PROMEDIOS MULTIANUALES SUCESIVOS (GLISANTES) DE PRECIPITA -

CION CON EL PROMEDIO DEL PERIODO 1901-1970 EN LA ESTACION METEOROLOGICA
INGENIO MANUELITA (A) Y DE C A U D A L C O N E L PROMEDIO DEL PERIODO 1934.1970
EN LA ESTACION HIDROMETRICA C A U C A - JUANCHITO (8)

FIGURA I

I GRAFICO PARA CURVA DE FRECUENCIA I

COMPARACION DE C U R V A S DE FRECUENCIA DE C A U D A L E S MEDIOS A N U A L E S

DEL PERIODO 1934-1970 Y i951 -1970 EN L A ESTACION HIDROMETRICA
CAUCA JUANCHITO -
FIGURA 2
281
o
a
Y
3
3
L
.
l
I
L
3
-3
L
282
ae oec
n
I I
L J
I I
I I
I I
I I
I* I
I *I
I I
I I
L. .J
'il!
O
E l
0 I
I
9i%'- x x O
o
O
N
0 -
\
283
284
I
m
r(
a 4
a a
3 a
2 3
I
L 2
IL
0 0 0
I
 285
\
b \
\
\
\
a
-4
I
l
\
I

\
U
.
n
i
-
v>
c' l
l
 NEW MODELS O F FREQUENCY LAW OF RUNOFF
STARTING FROM PRECIPITATIONS

J.R. TEMEZ
Professor
1.T.O.P.College-Madrid
ABSTRACT

Two interesting applications of a new hydrometeorological

method are developed with scientific rigour.
From the frequency law of annual precipitations, we can
deduce symply the law of runoff, and with precission, proved
in all tne esperimental verifications. To do it, we must know
or estimate the potential evapotranspiration on the basin (ETP)
and the minimum effective precipitacion (Po).
An analogous reasoning, yet simplier, allow us to convert
the frequency law of maximum precipitations in the law of
volumes of superficial runoff in floods. The only necessary
datum is the minimum effective rainfall P L , analogous to the
Po *

We can simplify the calculations with special paper o f

double scale. In them, the statistic function of runoffs is
the same as precipitations if we read each one in the corres-
pondent scale.
The profit o f this methodology, evident in bassins without
data of discharge is also important when we know the registers
of flow because it makes easy models of adjustment more reaso-
nable than the classic one o f Galton, Goodrich and so o n , and
in this way we avoid nonsensical extrapolations in the intervals
of large and small values.

- -
RE S U ME N
Dos interesantes aplicaciones de un nuevo método hidrometeo
rológico se desarrollan con rigor científico.
A partir de la ley de fre'cuencia de las precipitaciones --
anuales, se deduce la de aportaciones de manera sencilla, y con
precisión, demostrada en todas las comprobaciones experimentales.
Para elio solamente se necesita conocer o estimar la evapotrans-
piración potencial en la cuenca (ETP) y l a lluvia mínima eficaz
(Po).
Un razonamiento análogo, aún más simple, permite convertir
la ley de frecuencia de máximas precipitaciones en la de volume-
nes de escorrentía superficial en avenidas. El Gnico dato necesa
rio es la lluvia mínima eficaz P L , análoga a ia Po.
Los cálculos se simplifican con papeles especiales de doble
escala. En ellos, la función estadística de aportaciones es l a -
misma de precipitaciones con tal de leer cada una en la escala -
correspondiente.
La utilidad de esta metodologia, evidente en cuencas sin d a
tos de aforo, es también importante cuando existen registros fo-
ronómicos, pues facilita modelos de ajuste más racionales que --
los clásicos de Galton, Goodrich, etc., y se evitan así absurdas
extrapolaciones en los intervalos de grandes y pequenos valores.
2 88

1. MOTIVATION

The precipitations, phenomenon of general type, are adjusted

correctly to the also general classical frequency laws: Gumbel (maximum
rainfalls), Gauss (annual precipitations*), and so on, altering the mean value
and the dispersion from some places to others.

These precipitations are transformed partially in runoffs, but is

fundamentaly a deterministic process, peculiar of each basin in relation to
their edafogeological and climatical characteristics. Therefore the regime of
runoffs can not be defined with the statistic functions, at present in use, which
ignore the concrete parameters of each basin, significatives of the hydrological
cycle. These functions, by virtue of the liberty that their indetermined coef-
ficients give them, are able to adjust to the experimental points only in the -
interval of the intermediate values, but they point their inadequate conception
out to represent the hydrological phenomenon in the band of small and large
values where the disadjustments are significative and many times nonsensical,
such as what happens when the runoff of low frequency are negatives and the
high ones exceed notably the cipher of the precipitations. Their inadequateness
is manifested more clearly in basins of irregular regime than in the regular
ones, since these last are easy to adjust any curve to the reduced range of
variation of the registers not indicating the erroneous extrapolations which are
done about the extreme values.

O n the other hand, in the basins without registers of flow, w e

must define correctly from the precipitations, not only the mean flow, but also
the other parameters which are characteristical of their hydrological regime
like functions of frequency of runoffs floods and so on.

These considerations have moved the author to develop a new hydro

meteorological method of precision and scientifica1 base, though of
simple application.

The article shows two interesting applications of this metodology,

which will be the object of an exhaustive treatment in a later publication.

*
The author will propose in a later publication a modified Gauss law.
 289

2. TYPE OF FREQUENCY L A W OF THE A N N U A L

RUNOFFS PROPOSED BY THE AUTHOR

When w e treat of regulating high percentages of the mean flow of

a river, the decisive fact is the frequency law of their annual runoffs, lossing
importance in the study, the precise knowledge of the monthly variations, more
sensible to the geological characteristics of the basin, on the other hand deci-
sives when the volumes of water to regulate are low.

Therefore it has a big interest to determine the said law starting

from the precipitations of the basins without registers of flow, or by the
convenient adjustment to the experimental points if there are registers of
flows. The method that is exposed later on, solves the probleme in both cases.

The balance of water in a period like the water year,

permits to establish the relation

A = P - E where A = annual total runoff

P = annual precipitation
E = annual and actual evaporation

This equation applicable to the individual values of one or several

years, suggest us also about the relative configuration of the frequency of run-
offs and rainialls laws.

For very big values of P,the actual evaporation will be identified

with the potential one, practically constant from s o m e years to others (in the
humid years is smaller than in the dry ones). On the other hand, in the extre-
mely dry years all the precipitation will evaporate from its own basin E = P,and
A = O. In intermediate conditions E and A will increase with p.

The qualitative sight of the precipitations and runoffs laws is

represented by the figure 1. W e can calculate the frequencies of A from the
frequencies of P if w e know the values of 6 (P)that in this way it has the
significance of a middling evaporation to that precipitation.

That relation 6 = 6 (P)according to the considerations done previo-

ly, will be represented in the figure 2 and different from some basins to others
in relation to the capacity of retention of water on their soil R and the regime
of temporal distribution of their climatical variables.

The method proposed consists in establishing the family of curves

i
ETP = P$
(-E T P)
290

among which w e must select the most suitable in each concrete case.

-
W e can make that selection in relation to R , but the curve,
according to what was said before, is also conditioned E by the temporal
distribution of the precipitation and the evapotranspiration, which is variable from
a meteorological zones to another. Therefore w e think more practical to
represent the influence as a whole with all these variables by their inmediate
effectPo (figure 2), minimum effective precipitation from which frequency
corresponds a null runoff.

Having present the definition of 6 , the figure 2 is transformed in

the figure 3.

The calculations of the frequency laws of precipitation and runoffs

in many basins with registers of flows were made, and in all of them w e found
out that the values, correspondent to a same frequency, are combined realy
by a relation of the type schematized in the figure 3 arid has the following
expression:

A = O para P < PO y para P > PO

that permits to transforme the frequency law of precipitations in the frequency

law of runoffs.

The figure 4 contrasts the results obtained by this process and

the usual ones at the basin of the Guadalmellato river. In front to the good
adjustment of the author’s law, Galton gives nonsensical high values, higher
including to the precipitations, in the interval of high frequencies, while Good
rich in the interval of small values, decisive in the studies of regulations of a
river, recomends negatives ciphers including for frequencies higher to O, 10.
The functions of Goodrich, and specially Galton, ignoring the physical sense
of the hydrological phenomenon, are not capable of simultaneously adapting
to the total range of values.

The adjustment of the author is repeated in the figure 5 with

double scale of ordinates: the normal and their transformation according to
formula (2). In this way the law of frequency of runoffs must be the same as
that precipitations provided that to read each one in the correspondent scale;
effectivelly w e can verify that the experimental points as much as the preci-
pitations as the runoffs are confounded and are distributed in the grapht round
about to an only curve.
 291

Guadalmellato is an example of the many empirical cornprobations

carried out, which demostrate the big precision of the method proposed here.

The procedure of calculation consists in the following steps:

1) Calculation of the frequency law of precipitations,

2) Determination of the potential evapotranspiration value of the

basin (ETP)deductible from the evaporimetrical measures (or
charts) existing in the zone.

3) Valoration of the minimum ei'fectiveprecipitation Po. The data of

flows, if they exist, must orientate the computations to choose
theparameter Po more convenient. O n the contrary the valoration
of Po must be guided by the values obtained in other basins gauged
of that zone and based in the capacity of retention of water in their
soil. Po is equal to the capacity of retention of the soil plus a
function of the climate which sign can be positive or negative
depending on the cases. The author prepares an orientative table
of values of Po,though a hydrologist can estimate sufficiently
exact that cipher with so dear physical sense, based only in their
experience and in a superficial knowledge of the characteristics
of the basin.

In humid climates where the precipitations of low frequency are

much higher than PO,is enough a gross approximation of this
last parameter.

4) Once that data are known, the runoff A with frecuency F is obtain-
ed in relation to the precipitation P of that same frequency by the
formula (1) or their equivalent (2).

It must not be forgotten that the knowledge of the frequency law

determines automatically the value of the mean runoff. Inversely w e can
choose the parameter Po with the condition to proporcionate a mean runoff
equal to the previous valuation by other procedure.

In any case, if there are registers of flow the values of ETP and
Po will be needed to get the best adjustment with the experimental points.

To programme the method to the use of computers will yet be

easier the calculation.
292

3. CORRECTIVE R U N O F F

Let us imagine that after a serie of years of intermediate

characteristics, a year so dry is produced that all their precipitation is
evaporated and not producing any runoff. In spite of that circunstance, the
flows of the river will be not necessarily null, since they can feed f r o m the
underground reserve (or superficial) of the basin decreasing in agreeable to
the curve of exhaustion of the base flow.

T h e corrective runoff, or variation of the reserve from the end

of a water year to the end of the following one, will be an analogous function
to the represented one in the figures 6 and 7:with high values of theprecipi-
tation, the reserve will increase and discharge of the river, diminish in this
quantity; on the contrary with low values it will decrease, to feed the super-
ficial flows; in m e a n raining years the reserve will not change.

For smaller frequency than the correspont one to the frequency

of m i n i m u m effective precipitation F (PO),the total runoffs are identified with
the corrective ones, responding to t e r m s fundamentally different so far mention
ed. Neither the potential evapotranspiration, ETP, nor the Po has n o w any
incidence in the phenomenon, as neither the value of P;the law is determined
by the curve of exhaustion of the reserves of the basin, as well as by the
frecuencies of the initial state of said reserves and of PO.

T h e classical functions of frequency are not either capable of

being adapted simultaneously to the interval of usual values, prevailing the
direct runoff of the year, and to the different interval of small values corres-
pondent to the variation of the reserve. These functions treat t h e m indiscrimi
nately with an intermediate dull adjustment in which ignoring this real duplicity
to get out of orbit the dry runoffs, which are precisely the decisive ones in
the regulation process of a river.

T h e article will not extend in the detail of this corrective runoff

which in several cases is necessary to have present, while in the others, on
the contrary, it has little importance, like what happens in impermeable basins
with little variation of their reserves of a year to the next one and of which
m e a n value and pluviometrical regularity are at least moderated, in this w a y
the probability of precipitations close to the m i n i m u m effective one PO is
extremely small not participating on the computations. T h e substractive term,
that according to the graph of the figure 6, m u s t be applied also to the zone of
strong precipitations, represents a percentage very small of the total runoff
in these dates which is not worthwhile considering.
 293

4. MAXIMUM A C T U A L E V A P O T R A N S P I R A T I O N
ON A DRY CLIMATE

A s it has been exposed previously, the precipitations increase

the availabilities of water for the evaporation and this one will increase up to
the potential evapotranspiration, or more exactely to the potential evapotrang
piration of the humid years which is about O, 9 times their mean value. The
previous affirmation is evident to humid climates, but not so to the dry ones.

In climates like the mediterranean one, there is a season of the

year (Summer), when their potential evapotranspirations are m a x i m u m
while the precipitations are practically null as much in dry years as in the
plenty ones. It exists in these dates a permanent deficit of precipitation; the
actual total evaporation of the year does not reach ever to the value of the
potential one and its m a x i m u m value will be the potential evapotranspiration
of the period of precipitations (ETP) p increasing in the capacity of retention
of water on the soil (R)evaporating in posterior dates.

Is very important that in these cases the ETPwhich intervenes in

the formulas be replaced by the m a x i m u m actual evapotranspiration ETP*
where (ETP)*= (ETP)pt R.

It could be said that the ETP of the formula will in any case be
the least of the following values:

1) potential total evapotranspiration of the year

2) potential evapotranspiration in the period of rains increased

in the retention of the soil

5. FREQUENCY L A W OF VOLUMES OF M A X I M U M FLOODS

The relation between the total rainfall P’and the volume of

-
superficial runoff A’ is of the type schematized in the figure 3 for P A, but
now, treating of a phenomenon of short duration and strong concentration of
rainfall, exist the following differences:

. The evaporation in so short time and in an atmosphere of big

relative humidity is worthless and not altering the process.
2 94

The essential element is the quantity of water that can be retain

ed in the soil, characterized by a minimum actual rain Pó,
similar in idea to the Po of the annual runoffs but with ciphers
much smaller.

According to the documentation of the Soil Conservation Service

of EE.UU. and verified with several studies of the author, the relation is:

If P’ G P O , A’= O andif P’ >Pó

universal law in relative values to Pó, which is their only indetermined

parameter (figure 8).

The previous one, suggests the creation of a new special paper

(figure 9) with scale of frequency according to Gumbel and double scale of
ordinates: the normal one and their transformed as:

x = It (it JTx)
deducted from the equation (4).

If w e draw the frequency law of m a x i m u m rainfalls on the mention

ed paper, that same straight will define the frequencies of the volumes of flood
A ’ reading them in the correspondent scale. The method can not be simpler.
The figure 9 shows an example of application to the basin of the Cheliff at
Algerie.

The hydrologists defenders of the analytical and non-graphical

adjustment, only have to transform the law of m a x i m u m precipitations accord-
ing to formula (3) or their equivalent (4).

There upon these volumes are related only to surface runoff and
to obtain the total ones is necessary to increase them in the correspondent
groundwater runoff,worthless however in the interval of the high values, the
most interesting to the calculations.

In summary, once the frequency law of m a x i m u m rainfalls is

defined, the process of calculation of the volumes of flood only need the
estimation of the value of the minimum effective rainfall PA,
 295

T h e book "Design of Small D a m s " of the Bureau of Reclamation

take in the information of the Soil Conservation Service and facilitates a tables
which suggests values of the Pó, principally in relation to the nature and thick-
ness of the soil, although modified by the type of cultivation; this book explains
that each concrete case will depend naturally on the humidity of the soil in the
initiation of the rainfall and the values of the formula are considered to an
intermediate conditions in the dates of presentation of the floods. T h e author
in keeping with the theory of the Soil Conservation Service but precisely by
that remarkable influence of the initial humidity of the soil, the Pó of the
formula has to change also in relation to the climate and in this manner, other
things being equal, it will be higher in a dry one than in other humid one, where
there is a big probability that at the beginning of the rainfall, the soil would be
in proximate conditions to the saturation of the water.

If data of flows could exist, the experimental points of the volume

of superficial runoff in the m a x i m u m flood of each year of register, could
advice the Pó to choose to obtain the best adjustment.

T h e conditioning factors of Pó and of Po are basically the s a m e

and it m u s t not be forgotten, since any information that w e can orientate in the
estimation of one (for example the tables of the Soil Conservation, Service) can
be used in the determination of the other.

In the order of magnitude it can be said that Po is aproximately

fifteen times greater than Pó ; a study directed in establishing with greater
precision this relation would be interesting.

6. CORRELATIONS P R E C I P I T A T I O N S - R U N O F F S

T h e relations:

2
(P - Po)2 A' = (P'- Pó)
A = Pt ETP 2 Po - Y P't 4 P ó

are rooted in the essence of the hydrological cycle and have a big physical
signification. Besides obtaining its specific end in the transformation Of
frequency law, it also m a k e s clear the types of correlation m o r e adequate
to the individual values of these variables.
296

7. LIMTS OF THE METHOD

This method, a s any other hydrometeorological one, is not

strictly applicable to a singular basin with appreciable captures of water from
other zones or leakages towards them, since their flows are conditioned also
by precipitations outside the said basin.

It is conceived for regimes fundamentally rainy and it has not

been studied for any possible adaptation to the snowy ones.
 297

Fig.I . ESQUEMA DE SITUACION RELATIVA DE LAS LEYES DE FRECUENCIA DE "P"Y "A':

RELATIVE SITUATION SCHEME Op FREQUENCY L A W S OF "P" AND "A",
i

4_
ET P

I
P
ETP

Fig.2 . ESQUEMA DE VARIACION CE 6.

VARIATION SCHEME OF 6.

P = Precipitacidn anual de frecuencia F. P = Annwl pr.cipitol)an at treginncy F.

A = Aportacion especifica onud dr lo misma frecunieia A = Annwl rprcific totalrunoff Or the SQIY fp.p-y
F F.
5 :Diferencio entre P y A
de ia misma frecuencia F. 6 :DiffWlnCe beîueon P and A of iha s a w m y F.
ETP: Evapotranspiracidn pciencid. ETP x Patentid ewpotranspira tion.
R,= Precipitacidn a cuya frecuencia F (Po) correspon- PO= Precipitation ta which frequrnoy F( Po) Oorrewponh
de uno aportación nula. o null iotalrunoff.
 -
-7-
l
!
-f
!
i i
I
4I

0 51

Fig. 4 r 5 - AJUSTE EN LA C U E N C A DEL GUADALMELLATO

ADJUSTEMENT AT GUADALMELLATO BASIN.

---
--- Ley da Goodrich.
Ley d. ßolton
---
--- Qoodrich's low
Galton's br
O Puntos experimentales de atoro. o Annual runoff miperiiaentol peints.
Puntos rperimntoler de lluvia.
 299

AA

P A

Fip. 6 y 7 - ESQUEMA DE APORTACION CORRECTIVA DE L A LEY DE FRECUENCIA DEBIDA A L A VARIACION

DE L& RESERVAS.
CORRECTIVE CONTRIBUTICW SCHEME OF FREOUENCY L A W W I N G TO VARIATION OF RESERVES

P =Procipitacibn anual de frecuencio F , P =Annual precipitation of frequency F

A = Apartacidn especifica M U O I de la misma frocww ' A =AnnuOJ aQocific t#alrUnolf of the s- tre-
cia F. qurncy F

E T P = Evapotranipiroci6n potwcial, 1 ETP- Patrntlal rwpotronspiration

=
PO Procì~itoci~0 cuyo frccunicia F(Po) corres- , PO = Proclpitotion to which frequency F(Pe) carres-
pande una apwtación nula. , pondi a null totalrunoff.

A A = Aportación carroctiva. A A = Corroctlve totalrunoff

&,= Máxima valor de la aportaci6n correctivo , A O = Maximum valu# at correctivo totalrunoff

 30G

-
PA
A'

a - R E L A C I O NENTRE P' Y A' DE UNA MISMA FRECUENCIA.

RELATION BETWEEN P' AND A' OF THE S A M E FREQUENCY.

6 T- .- . ~- . -- - . ~~ -

-
P'
?A
Bassin of the Cheliff river (ALGERIE 1
~

Pk = 35 IlMn.
5 1
Experimental pointa P'

4 4

3 1I
I P'with frequency F
I- -- - -- -

ol , i c

FIP.9 . GRAFICO ESPECIAL CON DOBLE ESCALA. EN EL UHA MISMA RECTA REPRESENTA LAS LEYES DE
FRECLÆNCIA DE P' Y A'.
SPECIAL GRAPHIC WITH DOUBLE SCALE.ON WHICH THE SAME STRAIGHT LINE REPRESENTS THE
FREQUENCY L A W S OF P' AND A'

P' = Precipitación toto1 de un

aguacero de frecuen- P'vTotal precipitation of o rainfall with frequency F.
cia F
A' = Volumen de escorrentia auperficio( de b misma A', surface runoff volume of the rame frequency F
frecuencia F en mdrimaa avenidar. in maximum floods.
Ph= Precipitación de un opuacero do frecuencia
F (PO )o lo que correrpondo una escorreniio
-
PA= Precipitotlon of o rainfall with frequency F( Po 1 to
which corresponds o null rurtoce rurloff.
superficiai nulo
 TRAITEMENT OPERATIONNEL D E S DONNES PLUVIOMETRIQUES
ENTACHEES D'ERREURS O U INSUFFISANTES

R. Trendel - Der Megreditchian - Mme Rulliere

ABSTRACT

The Bureau o f Water of the National Meteorology, at present

applies a method that allow us, under certain hypothesis, to obtain
the equation of lineal multiple regression, which permits to
calculate the theoretical values of the monthly rains. The
application of this formula is possible by the existance o f base
data, corresponding to each season an index actual value/theore-
tical value that is useful to correct.
In this way, we can calculate the values o f theoretical rain,
that allow to correct and complete the series, and also the rainy
periods in season without data or with inadequate data. It is
carried out an analysis of correlation to establish the degree o f
guaranty of this method and to choose the parameter to use in the
different possible hypothesis and, particulary, the iterative
method of Van Isacker.

- -
RESUME

Le Bureau d e l'Eau de La Météorologie Nationale applique

actuellement une méthode opérationnelle, découlant sous certaines
hypothèses de l'équation d e regression linéaire multiple, permet-
tant de calculer des valeurs dites "théoriques" des pluies men-
suelles.
L'application de cette formule est rendue possible grâce à
l'existence d'un important fichier de normales. A chaque estation
correspondant un indice (valeur réelle/valeur théoriqye), il est
alors aisé de repérer les valeurs qui divergent trop a l'intérieur
d'une m ê m e zÔne d'homogénéité.
On calcule ainsi les valeurs d e la pluie "théorique"
permettant de combler les données manquantes et de pallier les
erreurs les pius grossières.
De même, pour les précipitations, on calcule les valeurs
mensuelles et s'il y a lieu, celles des episodes pluvieux, pour
les poster fermés ou insuffisants.
On effectue une analyse de corrélation pour étayer le degré
de validité de la méthode opérationnelle et effectuer le choix des
paramètres à utiliser. On examine les possibilités offertes dans
ce domaine par la méthode des composantes principales, en parti-
culier sous la forme itérative d e Van Isacker.
Différents critères sont également testés pour déceler les
valeurs douteuses, éventuellement entachées d'erreurs.
Certaines indications sont fournies sur la répartition
rationnelle du réseau pluviométrique.
 302

I - DETECTION AUTOMATIQUE DES ERREURS :

1 - Pl~~e_theoriq~e_'encuelle
Le B u r e a u d e L'Eau d e l a M é t é o r o l o g i e N a t i o n a l e a m i s a u p o i n t
u n e m é t h o d e p e r m e t t a n t la c r i t i q u e a u t o m a t i q u e d o s d o n n é e s p l u v i o -
m é t r i q u e s . E l l e est a p p l i q u é e d a n s le d o m a i n e r e l a t i v e m e n t p e u
é t e n d u d'un d é p a r t e m e n t f r a n ç a i s ; e l l e u t i l i s e u n e f o r m u l e empi-
rique permettant de calculer 1a"pluie théorique" mensuelle pour
u n e s t a t i o n d o n n é e , en f o n c t i o n d e la s o m m e p o n d é r é e d e s v a l e u r s
r é e l l e s d e la p l u i e m e n s u e l l e aux a u t r e s s t a t i o n s d u département.
c o n s i d é r é , les c o e f f i c i e n t s d e p o n d é r a t i o n é t a n t le r a p p o r t du
seuil d e référence d e cette station 5 celui des autres stations.
La f o r m u l e p r o p o s é e est d e la f o r m e :

O.~
0
Pth - p l u i e t h é o r i q u e m e n s u e l l e d e la s t a t i o n à é t u d i e r
m - seuil d e référence d e cette station
ms - s e u i l d e r é f é r e n c e d e La s t a t i o n s
Pr(s)-pluie r é e l l e m e n s u e l l e d e la s t a t i o n s
n - nombre de stationssans données manquantes utilisées
pour.l'interpolation
La f o r m u l e p r é c i t é e d é c o u l e d e la f o r m u l e u t i l i s é e e n a n a l y s e
objective pour l'interpolation d e Gandine.
Les seuils de référence des stations pluviométriques ont ete
o b t e n u s e n p a r t a n t d e s n o r m a l e s m e n s u e l l e s p u b l i é e s p a r la M é t é o -
r o l o g i e F r a n c a i s e C.13
______________---_--
Indice d'homogénéité
O n a p p e l l e r a i n d i c e d ' h o m o g é n é i t é le r a p p o r t " p l u i e r é e l l e
mensuelle/pluie théorique mensuelle" calculé pour une station
donnée. I L est bien evident q u e ce rapport s e r a nul pour une sta-
t i o n n e p o s s é d a n t a u c u n e d o n n é e p e n d a n t le m o i s é t u d i é , e t qu'il
s e r a i n f é r i e u r 5 l a v a l e u r r é e l l e si la s t a t i o n p o s s è d e d e s d o n -
nees manquantes.
L o r s q u e l'on p o r t e l e s v a l e u r s d e s i n d i c e s d e t o u t e s l e s s t a -
t i o n s du département sur u n e carte, i l apparait d e s zÔnes d'homo-
généite bien délimitees 5 l'intérieur desquelles ces valeurs sont
t r è s v o i s i n e s ; c e qui veut d i r e que, d a n s c e s z Ô n e s , Les p l u i e s
sont fortement corrélées. Les stations ne s'inscrivant pas dans
La r é p a r t i t i o n s p a t i a l e d e s i n d i c e s s u r l e d é p a r t e m e n t s o n t j u g e e s
d o u t e u s e s ; t e l l e e s t la b a s e d e la c r i t i q u e p r o p o s é e .

2 - _______ _______ -_----

Pluie mensuelle estimée
~ ~

Pour I estimation de T a pluie mensuelle des stations possbUdi~t

u n e s é r i e i n c o m p ! & t e , n o u s u t i l i s o n s la s o m m e p o n d é r é e d e s i n d j c c s
d ' h o m o g é n é i t e r e l a t i f s a u x t r o i s s t a t i o n s c o m p l è t e s lec. p l u s ~ r ' o -
c h e s m u l t i p l i c e p a r la p l u i e t h e o r i q u e d e l a s t a t i o n e n q u e s t i o n .
.../
 303

La f o r m u l e est d e La f o r m e :

Ob
'est - pluie mensuelle estimée d e , l a station à étudier
KS - facteur de pondération relatif à la station s
CS - indice d'homogénéité d e la s t a t i o n s
Pth - pluie théorique mensuelle d e ta station à étudier
Le f a c t e u r d e p o n d é r a t i o n u t i l i s é ici e s t f o n c t i o n de l'inverse
des distances entre stations.

3 - ------------ a n o m a l i e-
D é c a l a g e s e t --------- s
Pour rechercher des anomalies o u décalages éventuels, souvent
dus à des erreurs d e transcription, nous appliquons le principe
suivant :
O n c o n s i d e r e qu'une v a l e u r j o u r n a l i & r e (nous la noterons flk)
e s t d é c a l é e o u a n o m a l e si :
a) O, a l o r s q u e la v a l e u r j o u r n a l i è r e p o u r c h a c u n e d e s
trois stations les p l u s proches est supérieure
o u é g a l e à 1 mm.
b) vk supérieur à 3 mm. avec une valeur j o u r n l i è r e nulle
a u x t r o i s s t a t i o n s les p l u s p r o c h e s .
Dans Les deux cas, i l faut que les c o n d i t i o n s s u i v a n t e s soient
vérifiées pour u n jour donné;
n o m b r e d e s t a t i o n s of = O 1
nombre total de stations
G?T
nombre de stations où >,¶ 3 mm.
< y
1
nombre total de stations
4 - ------
Cumuls
La r é p a r t i t i o n d e s c u m u l s s'effectue à l'aide d u m ê m e p r i n c i p e
d e p o n d é r a t i o n q u e d a n s l a f o r m u l e (21, a p p l i q u é e a u x p l u i e s j o u r -
nalières des trois stations les plus proches possédant des donnees
n o n manquantes et n o n cumulées les j o u r s considérés.
La formule utilisée est la s u i v a n t e :
 O0
Cjlest(j) - pluie journalière estimée au jour j
étudiée
a la station

Fil-(s,j> - pluie réelle journalière L a s t a t i o n s Le j o u r j

n, - p l u i e c u m u l é e d e la s t a t i o n é t u d i é e
"1 - ler jour des données cumulées
"2 - d e r n i e r j o u r du. c u m u l ("1 \< j < n2)

D a n s la f o r m u l e (I), n o u s u t i l i s o n s u n s e u i l d e r é f é r e n c e établi
a p a r t i r d e s n o r m a l e s é t a b l i e s p a r A n g o t e n 1913 p o u r l a p e r i o d e
1850-1900. Pour l e s s t a t i o n s n'existant p a s à c e t t e é p o q u e , c e
s e u i l e s t o b t e n u p a r l a m é t h o d e d u t r a c é d e s isohyètes. P o u r a v o i r
d e s valeurs aussi précises que possible, nous corrigeons reguli&-
r e m e n t c e s e u i l d e r e f e r e n c e a u f u r et 3 m e s u r e d u d é v e l o p p e m e n t
du fichier.
Pour cela, nous calculons les moyennes mensuelles des données
d u f i c h i e r , e n t e n a n t c o m p t e s'il y a l i e u , d e s o b s e r v a t i o n s m a n -
q u a n t e s . C e s m o y e n n e s c o n s i d é r é e s c o m m e p l u i e s r é e l l e s d a n s la
f o r m u l e (1) p e r m e t t e n t d e c a l c u l e r les i n d i c e s d ' h o m o g é n é i t e qui
d e v r a i e n t ê t r e v o i s i n s d e 1 . Si l e s c o e f f i c i e n t s a p p a r t i e n n e n t à
l ' i n t e r v a l l e (0,90 ; 1 , î O ) l a n o r m a l e e s t a c c e p t é e , s i n o n e l l e
est modifiée.

II - REMPLACEMENT DES DONNEES MANQUANTES OU ABERRANTES PAR DES

V A L E U R S C ALC ULEES.
1Pre m-
------é t-----
hode :
E n c a s d e d o n n e e s m a n q u a n t e s , l a f o r m u l e (2) p e r m e t d e c a l c u l e r
l a p l u i e e s t i m é e . Si l a d i f f é r e n c e " p l u i e e s t i m é e - p l u i e r é e l l e "
e s t n é g a t i v e , l e s d o n n é e s m a n q u a n t e s d e l a s t a t i o n n e s o n t p a s re-
cherchées car, dans la plupart d e s cas, e l l e s correspondent a des
traces o u 5 des cumuls oubliés.
Pour calculer ces données, nous utilisons toujours le même
p r i n c i p e d e p o n d é r a t i o n , c e qui c o n d u i t A la f o r m u l e :

- pluie
Test(j) estimée du jour j d e la station étudiée
V%(s,j) - p l u i e r é e l l e j o u r n a l i è r e d e l a s t a t i o n s le j o u r j
Rm -- p l u i e r é e l l e m e n s u e l l e d e la s t a t i o n à é t u d i e r
n o m b r e d e p k r i o d e s d i s t i n c t e s d e d o n n e e s manquantes
"1 1 - l e r j o u r d e d o n n e e s m a n q u a n t e s d e l a lierne p b r i o d e
n2( - dernier jour de cette période(nll<j<n21;l<I<m)
 305

--------
REMARQUE
Cette méthode ne donne pas toujours des résultats acceptables.
a) L o r s q u e la s t a t i o n d o n t o n veut c a l c u l e r la p l u i e e s t i m e e s e
trouve à la frontière séparant deux zônes d e répartition spa-
t i a l e d i f f é r e n t e s d e s i n d i c e s d'homogénéité, la p o n d é r a t i o n
utilisée relative aux trois stations les plus proches, traduit
u n e distribution particuliere d e s p o i d s qui peut s'éloigner d e
la réalité.
b) A l o r s q u e l e s i n d i c e s d ' h o m o g é n é i t é c a r a c t é r i s e n t l e s p r é c i p i -
t a t i o n s m e n s u e l l e s a u x s t a t i o n s , i l s e n t r e n t d a n s le c a l c u l d e
la p l u i e j o u r n a l i è r e e s t i m é e .
P o u r t o u t e s c e s r a i s o n s , i l a é t é n é c e s s a i r e d e r é d u i r e l'échel-
le d u t e m p s . L a c r i t i q u e a u t o m a t i q u e d e s d o n n é e s p l u v i o m é t r i q u e s
est maintenant appliquée aux épisodes pluvieux. L'efficacit6 d e ce
p r o c é d é a d é j à é t e v é r i f i é p a r l'étude d e c e r t a i n s m o i s n e p r é s e n -
t a n t qu'une s e u l e p é r i o d e p l u v i e u s e .
Pour pallier ces difficultés, nous avons mis au point une deuxiè-
m e m é t h o d e p e r m e t t a n t d ' o r i e n t e r Le c h o i x d u m é t é o r o l o g i s t e .
------------
2ème methode :
E l l e e s t b a s é e s u r la r e c h e r c h e d'un r a p p o r t d e p r o p o r t i o n n a l i t é
m o y e n e n t r e la s o m m e d e s p r é c i p i t a t i o n s c o r r e s p o n d a n t a u x p é r i o d e s
d e s d o n n é e s m a n q u a n t e s e t la d i f f é r e n c e d u t o t a l m e n s u e l c o m p l e t
avec cette somme.
I N o u s u t i l i s o n s ici l e s t r o i s s t a t i o n s l e s p l u s p r o c h e s s a n s
données manquantes.
A p p e l o n s Pr(l), Pr(2), Pr(3) l e s t o t a u x m e n s u e l s r e s p e c t i f s d e
la p r e m i è r e , s e c o n d e et tro.isikme s t a t i o n s .
D I , D2, D3 l a s o m m e d e s p r é c i p i t a t i o n s t o m b é e s r e s p e c t i v e m e n t
à ces trois stations durant les périodes considérées.
P/,(s), l a d i f f é r e n c e P r ( s ) -
Ds (s v a r i a n t d e 1 3 3).

Nous calculons : 3
K=ZK s L (5)
s=' P;(q
O U K s est u n f a c t e u r d e p o n d é r a t i o n , f o n c t i o n d e l ' i n v e r s e d e l a dis-
t a n c e s é p a r a n t la s t a t i o n é t u d i é e d e l a s t a t i o n s.
C o n n a i s s a n t P:, le t o t a l m e n s u e l i n c o m p l e t d e la s t a t i o n é t u d i é e ,
e t La v a l e u r d e K d ' a p r è s l a f o r m u l e (5), n o u s p o u v o n s e c r i r e :

(6)

o b D r e p r é s e n t e la s o m m e d e s p r é c i p i t a t i o n s c o r r e s p o n d a n t e s a u x
j o u r s d e s d o n n é e s m a n q u a n t e s p o u r la s t a t i o n é t u d i é e .
Pour calculer les quantités journalibres m a n q u a n t e s , i l suffit
d e r e p r e n d r e la f o r m u l e (4) u t i l i s é e p o u r la p r e m i è r e m é t h o d e e n
remp1,açant p a r D.
306
52
I
.
c o
4
L

O
e.
I I
N N N
n n n
-
.e
W
c
I-
Ii
.I-
U
W
i1
4
0
I
Y .
0
I-
C
2
I-
 307
-
<D
m
T
a
m
>
o
z
cn
v-
a
O
v-
-a
23
Y
Y
a
w
=-
æ
O.
-i1
-
O
W
a
w
a
m 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 9 O0000
e
Q ~ O O O O O O O O O O O O O O O O O O O O O O O O O O0000
*
s - ~ o o o o o o o o o o o o o o o o o o o o o o o o oo o o o a
t
Q:000000000000000O0000Q0000 00000
t
yI toe, 00.0o o oooo 0-0 o o o oo a a o o o oo.oee
.I
* ~ 0 0 O 0 O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0
t
n ~ o o o o o o o o o o o o o o o o o o o ~ ~ o o0
o0o0 0 0
t
~:ooooooooooooooooooooooooo 00000
1 - 1 1 I
 308

III - CALCUL A PARTIR DES STATIONS EXISTANTES DES VALEURS MENSUELLES

ET DES EPISODES PLUVIEUX POUR LES POSTES FERMES OU INSUFFISANTS
Les développements précédents nous permettent d e :
1 - c a l c u l e r le s e u i l d e r é f é r e n c e d'une s t a t i o n d o n t les d o n n é e s
n ' e x i s t e n t q u e p o u r u n e p é r i o d e t r è s c o u r t e (2 o u 3 a n s )
sous r é s e r v e c e p e n d a n t qu'elles n e s o i e n t p a s erronées.
2 - d é t e r m i n e r d ' a p r è s l a f o r m u l e ( 1 ) La p l u i e t h é o r i q u e d ' u n e
s t a t i o n B l'aide d e s o n s e u i l d e r é f é r e n c e et d e s d o n n é e s
disponibles pour les autres stations du département
3 - c a l c u l e r l a p l u i e e s t i m é e d ' a p r è s l a f o r m u l e (21, ( L e m é t é o -
r o l o g i s t e p o u v a n t é v e n t u e l l e m e n t l'obtenir B p a r t i r d e la car-
t e des indices d'homogénéité) '

4 - rechercher les p l u i e s journalières en utilisant les formules

( 1 ) (2) e t ( 4 ) d a n s Le c a d r e d e s é p i s o d e s p l u v i e u x .

L e s i m p é r a t i f s d e l ' é l a b o r a t i o n d'un f i c h i e r v a l a b l e d e p l u v i o -
m é t r i e a v a i e n t d é t e r m i n é l'adoption d a n s la p r a t i q u e o p é r a t i o n n e l i e
d e la m é t h o d e s i m p l e d e c r i t i q u e d e s d o n n é e s , q u e n o u s v e n o n s d'ex-
poser.
P a r a l l è l e m e n t B c e l a , L e B u r e a u d e L'Eau a p o u r s u i v i d e s r e c h e r -
c h e s t h é o r i q u e s a f i n d'elucid.er l e d e g r é d e v a l i d i t é d e la m é t h o d e
a d o p t é e et l e s a m é l i o r a t i o n s qu'il c o n v e n a i t d e l u i a p p o r t e r .

IV - 'LE P R O B L E M E D E S D O N N E E S M A N Q U A N T E S C8,3,43
Quatre methodes ont été vérifiées sur u n fichier donnant les
hauteurs des pluies journalieres en 15 stations d e s Côtes du Nord
p o u r l e s m o i s d e J a n v i e r d e 1'1 a n n é e s c o n s é c u t i v e s ( d e 1961 2 1971)
1 - Analyse en composantes principales
_ _ _ _ _ _ _ _ _ _ - I _ - - - - - - - - - - - - - - - - - - - - -

Le p r i n c i p e d e La m é t h o d e e s t le s u i v a n t :
O n p a s s e d e s d o n n é e s initia1es:r;ii v a l e u r d e l a p l u i e le j o u r
i B la s t a t i o n j a u x d o n n é e s centrees'réduites =;i 3c.i , où
m
% SJ
x.j:A
i:4
Xij , Sj '2 2
ni

b-4
(lu- .
l e s v a l e u r s m a n q u a n t e s é t a n t r e m p l a c é e s p a r la m o y e n n e m e n s u e l l e
d e la s t a t i o n c o n c e r n é e . O n c a l c u l e l e s v a l e u r s p r o p r e s A; et Les
vecteurs p r o p r e s c i de la matrice d e corrélation. Les composantes
p r i n c i p a l e s s o n t a l o r s d é t e r m i n é e s p a r la t r a n s f o r m a t i o n l i n é a i r e
d e s d o n n é e s i n i t i a l e s 3 l'aide d e la m a t r i c e d e s v e c t e u r s p r o p r e s .
O n e f f e c t u e u n e r e c o n s t i t u t i o n a p p r o c h é e d u fichi'er initial e n
n e c o n s e r v a n t q u e l e s p r e m i b r e s c o m p o s a n t e s p r i n c i p a l e s . Les va-
leurs manquantes sont alors remplacées par les valeurs ainsi re-
constitu6es.
L ' e f f i c a c i t é d e l a m é t h o d e e s t m e s u r é e B l'aide d u c o e f f i c i e n t
 30 9

ry
où x e s t ta v r a i e v a l e u r , ta v a l e u r r e c o n s t i t u é e , g x L'écart
m o y e n q u a d r a t i q u e p o u r La s t a t i o n c o n c e r n e e , N L e n o m b r e d e s " t r o u s "
supplémentaires introduits dans le fichier d e façon aléatoire afin
d e t e s t e r l a m e t h o d e , La s o m m a t i o n é t a n t e t e n d u e à t o u t e s l e s va-
leurs d e x correspondant aux trous supplémentaires.
U n e é t u d e e x p é r i m e n t a l e a m o n t r e q u e le n o m b r e o p t i m u m d e c o m -
posantes principales retenues pour reconstituer le fichier initial
é t a i t n = 4.

2 - ---_--_-__-
A n a l y s e d e s --___-
c o r r e ----------
spondances
La m é t h o d e est a n a l o g u e 3 la p r e m i è r e m a i s , a u Lieu d e p a s s e r
aux variables centrées réduites =Li -3C.L o n u t i l i s e La t r a n s -
6J
formation classique en analyse des correspondances
(8)

I *
o ù xi.=- L.r..est la te j o u r i p o u r L'ensem-
n jz4 Y 1
ble du département, x . j = - g x~j e s t La m o y e n n e m e n s u e l l e à la
1 m ir4
s t a t i o n j et x..=-P n j:d
=.j=L 2
rn i r 4
~~.
Les valeurs propres d e la matrice d e corrélation d e s nouvelles
V a r i a b l e s t i j s o n t t r è s v o i s i n e s : 1 , 2 4 . 10-2 <i;< 5,53. I O - * .
Les itérations successives n'appoztent aucune amélioration p a r
r apport à l'approximation de départ tij = O pour les données man-
quantes. Cela revient à prendre

3 - __-_______-______--_-
Analyse de regression I

Les d o n n é e s m a n q u a n t e s s o n t r e c o n s t i t u é e s A l'aide d'une e q u a -

t i o n d e r e g r e s s i o n où les p r é d i c t e u r s s o n t Les h a u t e u r s d e p l u i e
aux s t a t i o n s environnantes. Le coefficient 6% p r é s e n t e u n f a i b l e
m i n i m u m p o u r u n n o m b r e d e p r é d i c t e u r s é g a l à 4, 5, 6 o u 7.

Les résultats obtenus par ces trois méthodes sont tres voisins
e n m o y e n n e , c e qui p r o v i e n t s a n s d o u t e d e la f o r t e c o r r é l a t i o n en-
t r e Les s t a t i o n s é t u d i é e s p o u r u n e p é r i o d e d e l'année c o r r e s p o n -
d a n t A u z e r é p a r t i t i o n r é g u l i e r e d e s p r é c i p i t a t i o n s . La d e u x i 8 m e
m é t h o d e xiJ=%. x =.i p r h s e n t e a l o r s l ' a v a n t a g e d e La s i m p l i c i t é .
x..
4 - -- ...............................
Méthode du maximum de probabilité

Soit X = [x,,---,xbJ u n v e c t e u r a l é a t o i r e g a u s s i e n ,
X % F I (o,~,,) O; EX = o ,Vdl(s EXXI-
a ) O n s u p p o s e Vx, c o n n u e e t r é g u l i è r e . La d e n s i t é d e p r o b a b i l i t é
de x e s t d o n c d e la f o r m e
 310

I -4
Pour maximiser {
(
)o o n m i n i m i s e La f o r m e q u a d r a t i q u e xV X .
La c o n d i t i o n d e m i n i m u m est ainsi
X'V'1dX = O
si t'on p o s e X I , =Z,--- ~n l e s v a l e u r s c o n n u e s e t "&+d j - - - -1
7t, l e s v a l e u r s i n c o n n u e s ( m a n q u a n t e s ) l a c o n d i t i o n d e m i n i m u m
s'-écrira

.-
O n o b t i e n t n-k é q u a t i o n s à n-k inconnues pour determiner les
valeurs i n c o n n u e s xp + 4 > ----
=,.
b) O n n e c o n n a i t p a s V x x . O n c a l c u l e a l o r s l e s c o v a r i a n c e s , c o u p l e
p a r c o u p l e , e n e f f e c t u a n t la s o m m a t i o n p o u r les v a l e u r s s i m u l t n n e -
m e n t non-manquantes pour chaque couple donné. On obtient ainsi une
m a t r i c e V x x , qui n'est p a s n é c e s s a i r e m e n t d e f i n i e p o s i t i v e é t a n t
d o n n é q u e l e s i n d i c e s d e s o m m a t i o n =ont é t e n d u s à d e s e n s e m b l e s
differents. On diagonalise ensuite V x x , o n range Les valeurs.pro-
p r e s p a r o r d r e d e d é c r o i s s a n c e et o n n e c o n s e r v e q u e c e l l e s d'entre
e l l e s qui s o n t s u p é r i e u r e s 5 u n seuil p o s i t i f d o n n e . U n s e u i l o p t i -
m a l s e m b l e e x i s t e r q u i e s t d'.autant p l u s g r a n d q u e l a t a i l l e d e
l'échantillon est petite.
O n a p r e s e n t é d a n s le t a b l e a u 1 l e s v a l e u r s d u c o e f f i c i e n t ' 6
(c-f 7) e n f o n c t i o n d u p o u r c e n t a g e d e t r o u s d a n s l e f i c h i e r p o u r
les trois premikres méthodes retenues.
L e t a b l e a u Z p e r m e t d e c o m p a r e r , p o u r l a q u a t r i è m e m é t h o d e , Le
g a i n o b t e n u e n r e m p l a ç a n t l a . d o n n e e m a n q u a n t e , n o n p a s p a r La v a -
l e u r m o y e n n e m a i s , p a r l a v a l e u r r e c o n s t i t u e e à l'aide d e c e t t e
m é t h o d e , l e s crit&re_c d e q u a l i t é é t a n t respec.tivement,E(r-;E)',
E ( ~ c - 2 ) ~et -4-t*Cr,~).
v - RECHERCHE DES ZONES HOMOGENES DE PLUVIOMETRIE PAR U N E METHODE DE
VISUALISATION DE MATRICE D'INTERDISTANCE

S o i t N p o i n t s d a n s l ' e s p a c e R P . O n c o n s t r u i t La m a t r i c e s y m e -
trique NxN dont les termes sont les distances euclidiennes e n t r e
points, mesurés d a n s l'espace RP ;
O n r e c h e r c h e u n e r e p r é s e n t a t i o n p l a n e d e s N p o i n t s X I , X z , ...,
X N d e R P à l'aide d e N p o i n t s i m a g e s Y I , Y z , - Y p d e Rz, d e f a ç o n
à c e q u e l a d i s t a n c e e n t r e d e u x p o i n t s i m a g e s Yi e t Yj s o i t La
p l u s p r o c h e p o s s i b l e d e l a d i s t a n c e e n t r e X i e t Xj. ,

P o u r c e l a , o n p a r t d'une c o n f i g u r a t i o n a r b i t r a i r e d e s Y i et o n .
d e p l a c e c e s p o i n t s d e f a ç o n a m i n i m i s e r u n c r i t e r e d e t y p e Xa
e n t r e d i s t a n c e s r e e l l e s et d i s t a n c e s images. C e t t e m e t h o d e n e n e c e s
s i t e p a s d e d i a g o n a l i s a t i o n d e la m a t r i c e d e d i s t a n c e s e t p e u t ê t r e
a i s e m e n t m i s e e n o e u v r e . L a p r 8 c i s i o n d e La v i s u a l i s a t i o n d i m i n u e
q u a n d le n o m b r e d e p o i n t s a u g m e n t e . I l s e m b l e t o u t e f o i s p o s s i b l e
d e t r a i t e r d e s m a t r i c e s 1 5 O x 1 5 C a v e c u n e p r é c i s i o n d e reppesen'
tation satisfaisante.
 31 1

Pourcentage de trous 15,19 19,6 24,3 29,4 32,9

d a n s le f i c h i e r

Méthode des composantes o,33 0,34 0,26 0,35 0,51

principales

Analyse des O ,33 0,36 0,20 O ,33

correspondances

Méthode de 0,30 O ,30 0,21 0,32 0,46

regression

Tableau I - Valeurs des 6% en fonction des pourcentages de

trous d a n s le fichier

JANVIER JUILLET

tations E(X -%y E(*- ;y 1 - r2 :(x -ry E(X-%)=

1 o1 381 2795 0,14 319 1542 0,21

161 335 1454 O ,23 177 Y65 0,18

471 755 2864 0,26 179 992 0,18

1131 343 1318 0,26 1 o1 1126 0,09

1211 797 1844 0,43 191 1027 0,18

1271 359 649 0,55 97 257 O ,38

1581 5 79 3453 0,16 409 1311 0,31

1681 253 1632 0,15 145 1262 0,11

1711 43 4 21 59 0,20 220 2285 0,10

2101 391 1236 0,32 160 822 0,19

21 51 327 1477 0,23 3 85 1 3 2'ï' 0,29

2231 41 6 2533 0,16 2 88 1413 o ,20

2281 390 2791 0,14 186 1448 0,13

2 62.1 761 2418 O ,31 161 1511 0,1Î

3621 51 2 1699 0,30 221 85 1 0,26

-
-
T'ablcau 2
 312

A p a r t i r d e 24 s t a t i o n s n o t é e s A,B,C,-, X d a n s la région Sud-

O u e s t d e l a F r a n c e ( F i g 2), o n a c o n s t r u i t 1 2 m a t r i c e s d e d i s t a n c e s
(une p a r mois) B partir d e s hauteurs d e pluie j o u r n a l i 6 r e s rele-
vées dans chaque station. C e s douze matrices ont été visualisées;
o n t r o u v e r a e n e x e m p l e ( F i g 3 ) l e s d e u x v i s u a l i s a t i o n s D é c e m b r e et
Mai. I l a e t é p o s s i b l e d e t r o u v e r q u a t r e g'roupes d e s t a t i o n s voi-
s i n e s qui s e r e t r o u v e n t s u r c h a q u e v i s u a l i s a t i o n .
C e s g r o u p e s o n t é t é r e p o r t e s s u r la c a r t e et c o r r e s p o n d e n t A
des zbnes pluviométriques homogènes en ce sens que deux stations
d'un m ê m e g r o u p e c o n t " p r o c h e s " v i s 3 v i s d e l a p l u v i o m é t r i e .
C e s t h 6 m e s d e r e c h e r c h e o n t é t é d é v e l o p p é s s o u s l'egide d u
B u r e a u d e L'Eau e t r e n d u s o p é r a t i o n n e l s a v e c L a p a r t i c i p a t i o n
a c t i v e d e M r s . B . R A M B A L D E L L 1 , J.F. R O Y E R , J.C. B A R E S C U T , J . Z I R P H I L E .

Notons en conclusion.que ces recherches s e poursuivent au

B u r e a u d e L'Eau e t q u e d ' a u t r e s m e t h o d e s d ' a n a l y s e d e s d o n n é e s
sont également étudiées dans le but d e parvenir 3 une meilleure
connaissance du champ d e pluviométrie.

REFERENCES BIBLIOGRAPHIQUES

CI1 A N G O T A. ( 1 9 1 1 - 1 9 1 4 ) A n n a l e s d u B u r e a u C e n t r a l M e t e o r o l o g i q u e
d e France
C21 B U C K S.F. ( 1 9 6 0 ) A m e t h o d o f e s t i m a t i o n o f m i s s i n g v a l u e s i n
multivariate data suitable for u s e with a n electronic computer.
J o u r n a l of t h e R o y a l S t a t ' i s t i c a l S o c i e t y , S e r i e s 8.22
pp. 3 0 2 - 3 0 6

C31 A F I F I A.A., E l a s h o f f R.H. (1966). M i s s i n g o b s e r v a t i o n s i n

multivariate statistics. Journal of the American Statistical
A s s o c i a t i o n . 61 p p . 5 9 5 - 6 0 4 .
C 4 1 K E L E J A N H.H. (1969). M i s s i n g o b s e r v a t i o n s i n m u l t i v a r i a t e
regression : efficiency of a first order methods. American
Statistical Association Journal. 6 5 p p 1609-1616.

C51 S A M M O N J.U. ( 1 9 6 9 ) . A n o n l i n e a r m a p p i n g f o r d a t a s t r u c t u r e
analysis. IEEE Transactions o n computers
pp. 401-409.
-
Vol C -
18, N' 5
 31 3

ARNE I

II

Fig. 1 -

I
'
5
ndiccs d'homogéneitE
314

12 B I S

Fig.2 -

Fig.3 - Exemple d e visualisation des matrices d’interdis-

tances.
 I N F L U E N C E OF I N A D E Q U A C Y OF H Y D R O L O G I C A L DATA
O N P R O J E C T DESIGX AND I'ORMULATION

GENERAL REPORT

by
Leo R. Beard (1)

N A T U R E OF DATA I N F L U E N C E O N PROJECT DESIGN

In evaluating the effect of data inadequacy on water resources project

design, it is important to recognize that a moderate error in project size that
might result is not necessarily accompanied by a proportional over-all loss in
project net benefits. As a matter of fact, the difference between benefits
derived from almost any water resources project and the costs of that project
changes very little over a relatively large range of project size in the vicinity
of optimum project size. However, it is in this range that added uncertainty
in design reduces project net benefits on the average, beca:ice net benefits
decrease in both directions from the optimum, and, even thoiigh increased expected
cost due to uncertainty is usually a smaìl fraction of the total project cost, it
can be large enough to justify care and extra cost in obtaining data for more
reliable design.

W h e n project design level is quite different from tconomic optimum

(and this can occur because of financial constraints, political constraints, and
other factors) , then the net project benefits change vei *J rapidly with errors in
design magnitude, but these errors tend largely to carit-el in the expectation
computation. Hence, in general but not always, errors in determining over-all
project size have far less than a proportional effect on project net benefits,
provided that the project operation can be modified as necessary to m a k e effective
use of the project facilities under conditions different from those anticipated
during design.

O n the other hand, rather minor inadequacies in data can have a n unex-
pectedly large effect on the over-all project size. In flood control design, for
example, errors due to data inadequacies can cause differences as great as a
factor of 2 or 3 in estimating extreme flood sizes corresponding to specified
exceedence probabilities. In the case of drought regulation (water supply) , a
change in magnitude or duration of a prolonged drought can result in differences as
great as a factor of 2 or 3 in the amount of supplementary supply (usually storage)

(l)TechnicalDirector, Center for Research in Water Resources, The University of

Texas, Austin, Texas, USA.
316

that must be provided for. Here again, though, project design magnitude
does not necessarily respond linearly to changes in flood or drought magnitude,
because cost and benefit considerations have a strong dampening or stabilizing
influence.

INFLUENCE OF DATA INADEQUACY ON PROJECT SIZE

Considering then, that there is no simple relationship between data

inadequacy and project net benefits, it is safe to say that evaluation of the
effects of data inadequacies on design requires a detailed study of the
inadequacies and all of the interrelated factors that influence project design.
Such detailed studies have been demonstrated in rather simplified applications
in work cited by Mr. lames and used as a basis of the studies described by
Mr. James.

In his paper, “Data Requirements for the Optimization of Reservoir

Design and Operating Rule Determination ,” Mr. James develops the theory
and some practical demonstrations for determining the optimum length of
stream gaging stations where their value for reservoir design and operation
alone is considered. In effect, the question to be answered is, h o w soon
should gaging records be started if a project will be constructed at some
distant time in the future. His basic solution is first for a known future
construction time, and then he considers uncertainty in the time of construction.
Benefits of gaging records are a function of increased efficiency of design and
operation.

Although m u c h simplification of the design and operation problems is

assumed, the concepts developed by Mr. James are of fundamental importance.
It is interesting to note that optimum record periods are in the order of 25 to
50 years, but there is insufficient information in the paper to determine whether
the basis of these results is real or largely assumed. Perhaps the author could
elaborate on this.

Stream gaging records are of value for many things other than project
design. It would be helpful if the author could express some opinions on whether
other benefits exceed these or are rather minor. It would seem off-hand that our
great heritage of hydrologic data could not have been justified many years ago on
such grounds alone, and yet w e k n o w Chat the body of data that n o w exists is
invaluable.
 317

I N F L U E N C E OF DATA I N A D E Q U A C Y ON METHODOLOGY

In addition to affecting the size of a project, data inadequacies can

greatly influen..e the methodology used in planning and designing a project.
Professor Reid o in his paper, "Tiie Design of Water Quality Management
Projects with Inadequate Data, " points out that mathematical models must
be built with availability of data in mind, that there is never as much data as
needed, and that the only defense against inadequate data is judgment. H e
describes a number of water quality models very briefly in the form of mathe-
matical equations, but does not attempt to describe their purpose or application
or to delineate the need for data in each case. Perhaps he could elaborate on
this. H e expresses some thoughts on the cost of waiting for more data before
designing a project, and points out that an important element is the zost of
postponing the stream of net benefits from the project.

Dr. Reid suggests 8 quality parameters that are commonly measured in

the U.S. with adequate reliability and accuracy and at reasonable cost. It
would be helpful to discuss these in relation to the models described, with
particular attention to data gaps that would exist if only these paranieters are
measured.

Dr. Reid also suggests a time scale for a progressive pollution abate-
ment program, showing abatement of lake eutrophication by 1980, reuse by
1990 and recycling by 2000. This is apparently for the United States, but
would be of interest to other countries. It would help if some of the abbrevia-
tions used would be explained, if distinction between reuse and recycle is
explained, and if the basis for or origin of the table were stated.

M E T H O D S USABLE WITH INADEQUATE D A T A

T w o other papers prepared for this session describe specific methodology

that should be used when data inadequacies exist.

In the paper, "Designing Projects for the Development of Ground Water

Resources in the Alluvial Plains of Northern India on the Basis of inadequate
Data, " Sarherwal describes generalized ground-water yield criteria,developed
for guidance in developing ground-water supplies in the Punjab until such time
a s systematic data on ground-water reservoirs becomes available. The develop-
ment of high-yield crops has occasioned a marked increase in ground-water
exploitation a s a n assured supply for critical irrigation needs. In order to
further increase the use of ground water effectively, studies based on such
criteria are essential.
318

The criteria described are based on approximating the pertinent

components of the hydrologic cycle. Of primary concern are those components
associated with replenishment of the ground-water supplies. Formulas are
given for the amount of rainfall that contributes to deep percolation, seepage
from lined and unlined canals, recharge from water courses, and return
seepage from irrigated fields. It w a s found that horizontal movement of ground
water is very small compared to vertical recharge and could therefore be
ignored in this set of approximate criteria. Water withdrawal criteria consist
of generalized values for evaporation from water-logged areas and draft from
various types of wells.

Planning of n e w wells is based on a water balance study using these

generalized criteria and a safety factor dependent on the region. An example
of criteria application is given for the Bist D o a b Tract.

Mr .Sarherwal supports his paper with an abundance of background

material indicating the importance of this subject to the economy, to the
ecology and to social conditions in India. It would appear that s o m e elaboration
on the role of surface water development in conjunction with ground water
management would also be very useful in such an outstanding paper.

In their paper, "Design of Water Resources Projects with Inadequate

-
Data in India General and Particular Case Studies," Banerji and La1
describe a variety of methods used in India for the estimation of design
floods and for monthly and seasonal runoff quantities.

Rough approximations of seasonal runoff from monsoon rainfall are

obtained with Strange's Table of runoff ratios for apparently arbitrary
categories of good, average and bad catchments. A less arbitrary method
of estimating runoff from rainfall uses Khosla's Formuld, which simply
substracts monthly evaporation and transpiration loss from monthly precipi-
tation. The loss is a universal, unique function of average monthly temperature.
A modified formula is given for calculating annual loss from annual temperature
in order to compute annual runoff from annual rainfall.

A third technique for obtaining runoff from rainfall is the correlation

of short runoff records with rainfall on a n annual or monsoon-season basis
and then estimating runoff for all the years of rainfall reconl. The fourth
technique uses the standard unit-hydrograph method for relating runoff to
rainfall.
 319

Methods of estimating peak flows include empirical formulas

relating maximum observed floods to size of catchment area , envelope
curves of maximum floods and regional flood frequency analysis. Criteria
are given for obtaining probable maximum precipitation and unit-hydrographs
for ungaged catchments. An interesting exponential recession technique
in lieu of the unit-hydrograph technique is described.

The authors do not discuss the degree of development or of flood

protection that is needed for various types of structures, so it is difficult
to visualize h o w their criteria would be applied for a great variety of
structures such as culverts, levees, d a m s and spillways, where there is
a great range in the degree of safety needed. Also, they d o not indicate the
degree of adequacy of the methods and whether further development of
methodology or increased amounts of data would substantially improve the
reliability of project design. It appears that they are in an excellent
position to render judgment in this matter, and perhaps they would d o so
in their discussion at this session.

MINIMUM DATA R E Q U I R E M E N T S

There is always the question as to the minimum data required for any
design, and, of course, this varies with the type of project and level of
development. In a paper, "Minor water Resource Projects Formulation
on Micro Hydrological Data for Standardization and Quicker Execution
in Developing Areas: Guidelines, " received through written communication,
Mr. Sikka discusses the problems of data needs in reldtion to development
of projects of moderate size. H e supplies a list of miriimum data requirements,
which should be of value to countries outside of India a s well as to India.
These include topographic and soil mapping, m a n y types of hydrologic data
and data on irrigation efficiency. Special emphasis is placed on the fact
that past drought periods can be exceeded in the future and that this should
be taken into account in design.
320

Mr. Sikka discusses environmental impacts and the needs for indices
of environmental conditions and for value weights that can be related to
economic efficiency benefits and costs. H e lists water and air quality,
wilderness and scientific areas, esthetic features and wildlife habitats
as environmental elements of principal concern. H e stresses the conservation
of water resources through more efficientapplication of irrigation water. H e
also discusses the conjunctive development of surface and ground waters and
related data needs.

Mr. Sikka brings up the question of the adequacy of basin-wide studies

of surface waters in conjunction with ground-water aquifers that extend beyond
the boundaries of river basins. This is a rather common circumstance, and
it is apparent that the scope of surface water studies must be extended where
ground water is a substantial element and where horizontal movement of ground
waters across the river basin boundaries is significant. This emphasizes the
importance of obtaining surface and ground-water data extending far beyond
river basin boundaries in some studies.

INFORMATION C O N T E N T OF DATA

M a n y of the effects of data inadequaciesdiscussed thus far in this

general report are direct influences that are relatively easy to understand.
There are some subtle effects that are little understood and yet can have
major impact on project design.

Weber, Kisiel and Duckstein in their paper, "Maximum Infonnpon

Obtainable from inadequate Design Data: From Multivariate to Bayesian
Methods," discuss some theoretical aspects of a subiect that is critical in
the use of inadequate data and has considerable impact even where substantial
data exists in many applications. They examine the effect of possible
inapplicability of theoretical assumptions underlying techniques such as linear
regression, discriminant functions,canonical correlation, principal component
analysis, factor analysis and cluster analysis. In many cases, departure of
data from underlying assumptions such a s linearity or normality will cause
erroneous results. More markedly and more generally, confidence estimates
will be in error.
The authors discuss the complexity that is introduced into Bayesian
analysis by uncertainties in the basic assumptions and cite some degree of
success in applications to discriminant analysis.
 321

Remarks relative to the interpretation of the results of principal

component analysis are interesting. Attempts to identify the physical
significance of the components or to use the components in subsequent
regression analysis bring up serious questions. The general reporter feels
also that such attempts would constitute a misapplication of the technique
(as the authors m a y feel also).

This paper does not attempt to answer the problems but simply identifies
them. It should be of great value if it occasions attempts by the authors or
others to find answers to these problems. It would be useful if the authors would
comment on the effects that inapplicability of assumptions m a y have on the
stability of maximum-likelihood solutions. The general reporter has witnessed
cases where highly erratic results were obtained through use of maximum-likelihood
parameters that were apparently sensitive to the form of a distribution function
and where the data were not known for sure to fit the assumed distribution.

G E O G R A P H I C CONS IDERATIONS

None of the papers in this session discuss the differences in the

various geographic regions that affect the adequacy of data. It is known that
many rivers are very stable and that a relatively small amount of data can be
adequate for fairly reliable hydrologic determinations. O n the other hand,
there is almost never sufficient data for evaluating the runoff potential of
some highly erratic streams where flows some years m a y be ln0 to 1000 times
as great as flows in other years.

Also, there is a difference in the nafure of data transfer potential in

various geographic regions. In regions where genera! storms or general snow-
melt floods predominate to produce high conelations among hydrologic events
within the region, data at a long-record site m a y be used to effectively extend
data at a short-record site. In this manner, short records can be m a d e to serve
for long records to a large extent. It should be noted, however, that this permits
estimates whose reliability is limited to that obtainable with the longest records
of the region.

O n the other hand, where great hydrologic heterogeniety exists, such

as where small-area storms predominate, information might be transferred if
the rainfall-runoff process can be modeled accurately. In this case, there is
a virtually unlimited amount of information in a region that might be assembled
to yield estimates far more reliable than those obtainable from the longest
records. At present, the technology does not exist for effectively assembling
such data, but the potential certainly is there.
322

SUMMARY

In summary, it is not at all obvious how data inadequacies can affect

design without making a detailed. study and without a thorough understanding
of the factors Involved. Errors due to data inadequacies can accidentally
improve a design, but the expectation is that better data will produce better
designs, as long as sound policies and technology are employed.

M a n y important contributions are contained in the papers for this

session, and the authors are to be congratulated on their efforts. They have
studied the need for and value of basic data and the impacts of data deficiencies
on techniques and on design adequacy, and have defined n e w problem areas
where special considerations are required in the use of small data samples.
 DATA IN INDIA -
DESIGN O F WATER R E S O U R C E S P R O J E C T W I T H I N A D E Q U A T E
GENERAL Q PARTICULAR CASE STUDIES

S.Banerji" E V.B. Lal* - INDIA

I

A B S TRACT

I n d i a has r i c h e x p e r i e n c e i n s u c c e s s f u l c o n s t r u c t i o n o f w a t e r
r e s o u r c e s p r o j e c t s w i t h i n a d e q u a t e data. W h i l e r a i n f a l l d a t a o f
c o n s i d e r a b l e l e n g t h are a v a i l a B l e i n or a r o u n d t h e c a t c h m e n t , r u n o f f
o b s e r v a t i o n s are u s u a l l y a v a i l a b l e for 1 0 y e a r s or l e s s , C o m m o n l y
s o m e g a u g e s i t e some d i s t a n c e a w a y from the dam s i t e may b e a v a i l a b l e
Data o n s o i l m o i s t u r e , infiltration, and e v a p o t r a n s p i r a t i o n a r e
almost n o n - e x i s t e n t ,
T h e paper, based on a s t u d y o f s e v e r a l i m p o r t a n t r e p o r t s r e l a -
t i n g t o m a n y p r o j e c t s s i t u a t e d in different c l i m a t o l o g i c a l , t o p o -
g r a p h i c a l and g e o l o g i c a l r e g i m e s , d e s c r i b e s t h e p r a c t i c e s f o l l o w e d
in: ( i ) t r a n s f e r r i n g r a i n f a l l d a t a from a h y d r o l o g i c a l l y s i m i l a r
r e g i o n t o the r e s e r v o i r c a t c h m e n t by s h o r t term c o r r e l a t i o n , Ciil
e s t a b l i s h i n g c o r r e s p o n d e n c e b e t w e e n r a i n f a l l and r u n o f f by a p p l y i n g
a r e g i o n a l e m p i r i c a l formula, or by first deriving a r e g r e s s i o n
equation for r a i n f a l l vs, r u n o f f f o r the s m a l l period for w h i c h s i m u l -
t a n e o u s r e c o r d s o f b o t h p a r a m e t e r s a r e a v a i l a b l e and t h e n a p p l y i n g
it t o l o n g e r r a i n f a l l r e c o r d s f o r g e t t i n g t h e discharge s e r i e s (iii)
t r a n f e r r i n g gauge discharge r e l a t i o n s h i p o f a distant site t o the
dam site t o w o r k out t i m e d i s t r i b u t i o n o f i n f l o w s , and t h e peak flow.
For t h e latter, a m e t h o d e v o l v e d f o r e s t i m a t i n g m a x i m u m f l o o d in the
Narmada and M a h a n a d i rivers, p r i n c i p a l l y based on a s s e s s i n g the
c o n t r i b u t i o n due t o different z o n e s o f c a t c h m e n t e a c h e x t e n d i n g t o 1
day's f l o w time, has b e e n d e s c r ì b e d for t h e B e n e f i t o f m o n s o o n r e g i o n s .

RESUME

L e s I n d e s ont une grande e x p é r i e n c e d a n s l a r é a l i s a t i o n d'amlna-

g e m e n t s des e a u x h p a r t i r de d o n n é e s i n s u f f i s a n t e s , L'observation
directe des d é b i t s ne porte en g é n é r a l q u e s u r d e s p é r i o d e s de m o i n s
de 10 ans, t a n d i s qu'on d i s p o s e s o u v e n t de d o n n é e s de p r é c i p i t a t i o n s
s u r de l o n g u e s p é r i o d e s . D'autre p a r t , il est c o u r a n t de d i s p o s e r
d'une é c h e l l e l i m n i m é t r i q u e 3 q u e l q u e d i s t a n c e d u s î t e du b a r r a g e p r o -
jeté, a l o r s q u e l e s d o n n é e s s u r l'humidité d u s o l , 1 i i n f P l t r a t i o n et
ì'êvaporation sont p r e s q u e ìnexfstantes.
L'étude p r é s e n t é e est basée sur p l u s i e u r s r a p p o r t s i m p o r t a n t s
r e l a t i f s à d e s p r o j e t s s i t u é s d a n s des r é g i o n s q u i p r é s e n t e n t d e s
c o n d i t i o n s v a r i é e s en c l i m a t o l o g i e , t o p o g r a p h i e et g e o l o g 2 e . L e s
a u t e u r s décrivent l e s m é t h o d e s e m p l o y d e s p o u r (31 t r a n s f é r e r l e s
d o n n é e s de p r é c i p i t a t i o n s d'une r é g i o n h y d r o l o g i q u e m e n t a n a l o g u e a u
bassin q u i a l i m e n t e l e r&servol'r, Ci'il & t q b l i r u q e c o r r e s p o n d a n c e
entre l e s p r e c i p i t a t i o n s et l'écoulement p a r l'application d'une for-
m u l e e m p i r i q u e r é g i o n a l e ou d'une I q u a t i o y de [email protected], c a l c u l @ e s
s u r l a p é r i o d e d'observation c o m m a n e des p l u z e s et des d é b i t s et u t i -
l i s é e s a v e c des d o n n é e s de p r g c i p i t a t i o n s d e l o n g u e d u r é e p o u r o b t e n i r
une e x t e n s i o n des debits, (iii) d é t e r m i n e r l a r e l a t i o n hauteur-débit
a u droit d u barrage 1 partir d'une r e l a t i o n é t a b l i e a u n e s t a t i o n s i T
t u é e 3 q u e l q u e d i s t a n c e pour e s t i m e r la d i s t r ì b u t i o n des d&bìts et l e s
p o i n t e s d e c r u e s , L e s a u t e u r s exposent a t i t r e d'exemple une m é t h o d e
utilisée p o u r é v a l u e r l a c r u e m a x i m a l e des r i v i b e s N a r m a d a et M a h a -
n a d i ; c e t t e m l t h o d e , i n t é r e s s a n t e p o u r l e s r é g i m e s de m o u s s o n , t i e n t
c o m p t e de l a c o n t r i b u t i o n des d i f f é r e n t e s p a r t i e s du b a s s i n , l e
découpage correspondant a un isochronisme journalier.
~

* S c i e n t i s t s , S e c r e t a r i a t o f 1.H.D' National Committee

 324

1.0 introduction

India h a a rich experience in 'successful' construction of water msources

projects with inadequate hydrologioal data. Since 1951, when tha f i m t Five Year
Plan commenced, 537 major and mriàium projeots, each having a reservoir etorage of
over 6167 hectare-mtree i.e. 50,000 aumfset, ham been taken up and about 300 have
been completed (1). However, since most of the gauge ami discharge sites f a
regular observation on Indien riveru have been eet up only after independenoe in
1947, runoff observatioas or even gauge-readings, if at all available at the sito
of a proposed dam were of very short duration, sw, lesa than 10 g b m . 'phr redseeing
factor in this situati= has been tknt for most aress of the oouutry long records
of rainfall, of 50 years or more ere gen9ra;lI.y available. Aluso, ooemonlJ aoma gauge
site would be evaileble on the ccgcerned river som distema away from tkie dai site.
Data on soil moisture, infiltration end evapotranpipiratiai axe praotiedïy non-
existent.
1.1 hia mesent paper is based m e tatu* of neveral intportant reports (vide
appendix 1) mlat- to meny projects of variona siaes rituated in different
ûlbtOlOgiû& t o p m p h i o a l a d @ û l ~ i C 8 l =gimes Of the C m t r j . W pY.æOtiCeS
described relake to the following t h m e oetegoriee of problem:

(i) 'Jkarisferring rainfall date from an adjacent, lydrologieally nimilar

regiai to tlie reservoir o a t c h n t e
(U)Eatciblisw oormepandeme between rainfall and runoff
(iii)Tramaferring g aw discharga relatioriehip of a distant site to dam site
M œ t of t h e e relate to estimation of runoff vdrmies. Eltimation of p.nk
flore has been discussed separately.
2.0 Indien Pcaotice
2.1 It meg be etated hem that 88 there is l e g e di.arsity not o d 7 in the
size and the region of l m & A o n but also in the nature of data available for
differed projeote, lhm are no 8~tau&d8 or mallar8 W d - 8 to -rocmm
problem of dafa maralty. Then, he8 ban no partioular Pmferenoe for either
the M i t i@rograph or t b etatietiad Mqoenny distribution In &te- tìm
'dosigm flod8, and, bpending upoe the gravity ni wzumxmnnoe of a likely failure
of t b s t m û t m , attsmpts hem b e n meâe, wimromr m-88- a d poisiblev
to arrive at the design flood by =me of both t h s e a p p r o d m s and a for others
of r e g i d applloatian.
2.2 1Lo 8 general guideline it h m been pmsuribed that inajar and odium groJects
should be deeigmd for Probable Msximtri Flood (m) "that w o d d reriult from the
 325

co&ination of critical msteorslsgicd and hYhOle@;iO f oonditians

ca.sidend physi.~c~llgpossible in the mgid"' Ln oases of S P J Q ~ wojects
with mery larga catchm;.nte whem applicaticm Of imit Wai.oS=Qhis b d V h a b b l e
a 1OOO-yeer flood e s t h t e ia attempted by f n q w n c y onalg.eie from a dkchprgs
aeries at site c m t m t e d frem &ta of rainfall or other i n f o m t i m that Ippg b~
amilable. In t b C P B ~of germanent ba-8 tu
less thpn 6167 ha. m the design is to be based on the Stanbra Pr@d**t
Flood (SPF) that would m s d t from Ithe mest severe combination of ~ t e ~ ~ l W i c ~
and hg&ologic omditime, considersd reasonably characteristi0 ef liha =gim
excluding e x t m m b r a m oo?abineticars', er a 1oO-year f l m d w h i a * r IS
For smaller projects design flood ae;y be estimated by approximate end empirioel
methods applicable locally. (2~3)-
3 . 0 Transferring Rainfall Data from Adjacent Catchment

In Cases where rainfall data for considerable periods are not

available for the catchment upto the damsite two tendencies are discerni-
ble. If it has been possible to construct a discharge series at the dam
site by some technique from discharge data available elswhere no
outstanding necessity has been felt for precîpitation data for estimating
the flood peaks or periodic inflow volumes e.g. Tehri Dam. Otherwise
attempts are made to work out precipitation figures for the catchment
under consideration. If the number of raîngauge of raingauge station
in the project catchment is samall, the statìons in the adjacent region
considered hydrologically similar are utilised for constructing
Thiessen's Polygon for working out weighted average figures o f rainfall
in different years within the catchment e.g. Hasdeo (Bango) Project. It
is also possible to have a few years'data at some specially set up sta-
tions within the catchment and to correlate them with the observations
at some stations with longer records, lying outside the project catch-
ment but still within a hydrometeorologically similar region. The
short-term correlation thus established is then applied to the longer
records of outside stations and the series completed for the project
catchment,
4.0 E s t a b l i e u C o r m s D o n m y 1& run& f
Bainfali reoords are @rerally available for projeet-catohwnte in
the form of 24 hour ralnfail amounte o b e e m d at a fixed hour for moat staticne
and 88 continuous recorde for selected etations with self-moording raingaugda.
For estimating runoff data the follmlng methode am generally follmedt-
a. Regional correlations, like Strange's Table; b.Khosla's F o d a
C. Regression equations defining correlatia betweem short-term
raipfall runoff data; à. whograph application.
While PrrtboC a,b,o, yield estiiiatee of runoff volume, hydrograph application
ia good for eetiaieta of flood volume as well aa flood peaksl.
4-1 l & 2&! 2wC-OXUl a S W Q E ' S TABLE
It gime permntage of runoff from 10~18ocn-1
rainfalls for different
lnàian catchmanta, which were rathbr subjeotively divided into three oatagories~
'good' 'average' and 'bad'. Thus far a total monsoon rainfall of 1000 e a good
catchment will yield 37.@ runoff, an average oatchnient, 2& ami a bad eatohRIent
18.776. Inspite of the faat tìmt theee tablea am ncm very old and o m yield only
 326

rough estimates, they ere often applied in the projects for assessrnt of runoff
volUries, e.g. Chambal Valley Development Scheme, where such calculations baw also
been checked againat the observed data of a few years. €&o (3) has also used Strangego
table while working out dischargea for Nagarajuriaegar aiid Srisailm projects.
4.2 s
'a&-xs Fornu4
Khoela (4) working 'on t b rational concept that runoff is the residual of
rainfall after deduction of evaporation and transpiratian loss' aesuiped that
'temperature can be taken to be a complete maaure of all the factore which are
responsible €or the loes of rainfall to runoff'. The formula hos no mgional l i d -
ta5cms of applicability.
1
-
His empirical formula is Ra Pm -Lp wbm Rm, ,P Lm a m rssp8otiln3ly the
runoff, rainfall and 'loss' figures for a given month in mu. Lm is taken as equal to
5 'Ern,ilkre Tm is the man monthly tempereture in centigrades and is more than
4.5"C. For Tm<4.50C lose is estimated from the following tabler
T,(OC)

L~ (RUI)
21
4.5 - - '7
'1
18 15
- -
12'
12.5
18'
io
4.2.1 If instead of a month, a year i6 teken es the period of Consideration, t b
relationship become
-
RA 5 PA X (45TA+ 800)
where Ra & P 828 the carresponding yearly values of runoff a n d rainfall, Ta is t b
mean annual %mparature for that year, aid X ia a oonstant. The value of x can be
determinad e w n from a year's observed values of RA, PA & T In area8 'wbm
rainfall is very low or it comes in oooasional oloud bursts4'the period far R-P-L
relationship is suggested to be 10-day or l 5-wrather than a month or a year.
4.2.2 The Mrakud D m Project employs Khmh fornula. F i g u ~ sof annual =Off
and l a s thus worked out ham also been compared with the figume arrived at hy
a different method, namely, the tramfer d g a w discharge figume et 8 distant
site to the projeot site.
4.3 Short-tem Reai.8roion sauti-
A B8gmriionequat i o n t âeïirlopd Q O ~ h t b gthe reinfall-runoff volirso
for the ehorf period for whioh t h e n data may be available er spalally colleoted
and them this equation ie applied to thi, long-tern r~lnftdldate to gst tbe
cormrpcmding runoff eitimatee.
4.3.1 ~n t b m d e o (Boago) Project, regresdon equetiœœ have been workad out
between (a) anuual rainfall and ennui runoff# (b) 110n8008 r8infal -0QIL
nmoff and (c) mamoon raâaîeìl ami mnud runoff on tb basie of tis observad data
of rainfall and runoff for 10 y e w e between 1959-1969. Ths leet OIY 8~ k Jutified
in view of the fact that th6 monsoon rainfall o o o u r r ~duri- JUM to Ootokr
1. nietric form is talsen from Bìmnt Siagì4 'PunaePentalr of Irrigrtion
Engiasering', 1967, ROW- mó.*
 327

-
aucounte for ab ut 9% of tlm anatm1 rainfall. Out of thee thee equstioriey the
momoon rainfdl-monaoon nuioff equatian gave the highemt QOrIdaticeiS coefficient
(0.869) o d thia was u. ad to derive the a m o a l runoff from the ennuel raiafoll f m s .

4.4 &hr-Dh Applicationr The tuchnique is lairly well haai. Later diseusaions
will ahow how the design storm is selected and its tiiir, distribution obtaineà for
applylag the mcipltetion figues to t h unit hydrograph.
5.0 -Disc- Belet ioliahip of a Distant site to the Barn Site
We piok up t b m e o- studies Vix the Eirakuà Dam, tts ThiLTi Deia a d the
NagarJunaaagar to illustrate hou this is being done.
5-1 In the Hhalcud Dam project (1947) the &am was prapomd to be looeted at a
site near SamboLpur w h r 8 gauge recad8 existed sinee 1921, but there were no
COrreaPpopdine gauge dieche@ curves. Ebwever, at Earaj, a site som 230 miles down-
stream froaSembalpur, gauge diaßhaqp recards existed sinue 1868. The gaugs madia@
at Sambdpur were corzhlated with the gauge readiq at NaraJ, d i n g due allowance
for the tinia.1- and similar epuea, discbrge c m 8 were prepared for S d e l p u r
and checked 8gainst the dally discharge obrervatians o M d eince J m , 1946.
5-2 The Tebri Dam Project (1969j e n w b w s construction of a dam e o r ~ ~thb s river
BhagFratM near Tehrii tha catohnmt arsa upto t b dpn eite ia 7511 8q.h.inaluding
2328 8 q . b of constantly snou bound axea. Daily rimr g&ugoa andwe8-U~ âiaaharge
observatiais at tkm damaite .ere available only fra May, 1964. This Catchment is
a part of t b Ganga cstcharnt in which, at Bairele, near Haricbrar about 105 km.
damst- Of 'pehri, deilr aid dia- dot8 S v d l a b h € h a i l l 9 O l o The
catcent up%o Raiwala t 23000 8q.b. inoludbg 8450 8q.h- is anow-
bound.The Raiwala data have been u t i l i s e d to compute runoff at Tehri
in 10-day periods of the year. For this purpose the runoffs for
different 10-day periods, in the period o f actual observation of
discharges at Tehri, have been compared with the corresponding 10-day
runoffs of Raiwala, assuming a one-day time lag for the flow to reach
Raiwala from Tehri. The percentages o f Tehri discharges to correspon-
ding Raiwala discharges have been plotted against the relevant 10-day
periods for the period of observation, 1964-66. These percentages vary
for the same 10-day period from year to year due to variation in
precipitation, temperature, humidity, vegetation, soil moisture etc.
and for individual catchments of the tributaries of the river Ganga
above Raiwala. The required factors have been worked out as below:
AvoroRs of runoff et T e m .
r i Ave- of runoff at Raiwtle
~ e i n gr vaime for &ifferet 10- periodi, the -off f i g m e at Beirala
have ben canverted to f-a for Tehri for 30 (fra 1936 to 1966).
5.2.1 'Ffiib Rairela Qata h a w albo heen wed, ia ocajimotian with the ih&-tea
rseard at TekrirL, for estimating the flood peds et 'psbri, &a- Baiwda 88 t h
etaticus, tfie peroeritege d tias e pcirticular noOb hae been equalla8 or
e m e e b d le plotted agaInet the flood 021 a sed-log paper to giw e lozig-tsra data
c m for t b inder itattian. Bor t b short-term for which data ara available both -
for % h i arrd Baida, rimilar o m s u. plot- far both th6 et&ticma. Tbs IOW-
term c m for the project 8t&iOn (%-i) L then coiiltriiabd from the abow three
o m s , and ths flooda of various frequencies &PB o b t a b d from this OPM. It is,
328

h m e m r , only apIQ of m w w adopted in the projeot for eetimting tha flood peek.
5.3 Ra0 (5,3) applies a different apprcmh to determim peak flooda at a section,
when discharge data are available for a diffemnt site al- the river. W
principle applied is simple: thedischage observed at a dametream aite ieequal to
the discharge at an upstream site plue the dischar@ contributed by me interniedia*
c a t c b n t mincis the oharial' trough' oapaieitg between the two sites. W 'trough'
capacity oan be computed ideally with the help of croes eectime of the river at
close internals, or otherwise in the absence of thie inforretian, by taking the
average width of the river flow at one end, the difference in the depth of f l w on
the day of Peak flood and 24 hours before its occurmncc and the length of the river
reach into woount. Th? inflow from ths intermediate c a t c h n t q be worked out from
the rainfall recorda using strange*s table. In this way the flood aeries at the
upstream point ia Constructd for a number of years and subjected to frequsnoy
analysis far estimeting tb design flood of a given recurrence interval.
6.0 E s t h a u = of P e d F l o a
N o d l y , peak floods are estimated by several mithoda before adopting a
design flood. Such niethods range f r a empirical fofiaulae directly giving peak flows
from a oatchniant of given area to the elaborate etarm-trailspoeition and aiaximisation
mthode. "hey m y broadly be clamifiad into tuo oategoriest
(a) Non-mteorological =th& (b) Meteorologici1 methods
In the non-mteorologieal oategory we may include t b following: Empirioal
formulae, Enveloping Curves, Regional Flood Frequency analysis. In the meteorological
category we inelude æ?thoQe that proceed frcm a t o m analysis. They msy or mqr not we
unit hydrograph.
6.1 .O Non-ktoorolouiad Cateaory
6.1 .1 Enipiriaal F a
(a) Ths noet popul formulae link the peak flood with the a m a of the basin,
like Diokm's, Q= C A Y 4 , for the Central and Northern India, the Byve's,
g CABB, for the south India, the &lis Q- O0O ."or
the Bombay aegica, wherm Q &vea the peak ra k- faa shaped c a t o h m m b ki
f disohaya in cusecs, A, tis a m a
in sq. miles and C is a coneteat differing fron loeation to looation. &oaueß Of
their simplicity euch regional formulae still hi wide use for getting a f i m t
appraaimatiar of the likely flood.
(b) Quite often, if high flood mark6 ara avrilable with raferrtaae to old trees
or =oient atructplae, or OWE from the m o r y of th looal inhabitruita, elow-ama
method is employed as (UI aid to gueoo tkm ordar of t b dieoharm. No reliable idea
c m obviously be bad of t b Seetion prevail* at fhs t- of flood fim, end t h s m
is diffiaulty in estabbliehin& ttie bed slope, which i8 teken OB equal to tkm SudaCe
S l o p establiaha0 from mrka at different points. Kutter's or uamih@' Coeffioient
of rugosity ie e i t b r (usumd, or dete-d by t b eubmtitution of i e a e m d äata
for a few flood8 In th oonoerned formula.
%Sidea t h faot t h t eu& formulae are ueeful only for limited regiolipl
applioation, present ri& eoope far eubjectim fagtom in choosing the valm
of tïæ constant. Also, it is not paisible to have any idea of tbie probable frequencg
 329

of the flood so estimated, ao that a p a r t l u l m value of C may give a flood which

may be too high for designing a minor work, m e y a culvert and too loa for &signing
a spillwey.
6.1.2 Envelope curvesi
Working on t b assumption that basins of similar hgdmlogical characteristics
should produce the sana mexlmum floods psr unit of catchment ama, Kanwar Saia and
Karpov plotted data of mm~imumfloods in Indian rimiers againet the drainage amss
producing thoee floods on a log-log paper, and gave two envelope curves one enveloping
data of South Indian Basins ond the other enveloping data far northern and oentral
Indian basina. !be likely maximum flood from a catchment of given area is then
expected to be indicated by these curves. Besides tìm basic inadequaoy that these
curves relate flood potential only with the drainage area, they do not provide for the
occurrence of floods of higher magnitude than those on reoord.
6.1.3 Resi onal Flood Freauew-
Data of all the stations (points) in a statietioally homogeneoua region are
combimd to produce a flood-frequenoy oume that is a s s w d to be valid for the entire
region and can thye be applied to determine flood of a retuni period for an imgauged
catchment In the region. The simplified procedure recommended by tìm Central Water &
Power Commissian (2) is as follows:-
All stations in the regim with flow recorde of 10 years or more are eelected
and for each etation a frequency curva go- upto a 100-year flood Is constructedby
the Gumbel's Ethod, with a confidence-band of 9% reliability. All points a m tested
for homogeneity as f ollors:
The ratio of 10-year flood to man annual flood is determined for each point!
this ratio awreged for all points is taiœn to give the mean 'lo-year ratio' for the
ama. The m t u r n period corresponding to the ran 'lO-year ratio' t i m e the maen
annual flood Is detemlned from the frequency o u m e of each station and plotted
against the number of years of record far that etatim oq a sed-log test graph. If
tiæ pointa far ail tim etations lie between the 9% confidence limits, they are
oonsidered homogeneous.
The frequency curves of different stations in a homogsneoua region a m regarded
as different estimates of the regimal curve, and tlaey are averaged as follows:
For eaoh statim, flood ratioe (flood of a return period T over the niean
m u a l flood) a m computed for a number of arbitrarily selected values of T. The rean
of the flood ratioe for all stations for a particular period T is taken to represent
the flood ratio for the r e g i d curve. The resulting mans for different vaìws of T
are plotted a t b extreiiie value probability paper and the best fit line through them
gives tïm mquized regional frequency curve.
The application of thie curve t o an w u g e d catchment rsquims M setimate of
the E=BIIILup1 flood for the wtchnmnt. This is d o m from another e m which gives
the plot of man u m u a l floods at different statims agai-t tbe corrsepmding
drainage amm. From this C- the value Of ttn? likely flood W d t
the are of the naxg mgaugad oatobment can be mado
6.200 &teorOlaasop1~
 330

records of all the precipitation stations in the region of, and around, the project
catchment, which may rather subjectively be ree;ardeà 88 hyäromteorologicdly homo-
geneous, are studied to sglect storms of high rainfall covering an area more or
less equal to or larger than the project oatchment. Far this purpose it mu b
neceesary to carry out the umtaai Deptharea-Duration ( D U > analysis of selected major
st >rp~s6nd from there maximum one-day, maximum two-deg, maximum three-day precipi-
tati- are worked out. These aeleoted s t o m , are then transposed to the project
oatcìnrent adjusting the precipitation axis d s o to an orientation that will give the
maximun runoff producing effect, if such directional change of storm axis is within
20° from the original axis. The storm is then maximised for the moisture content by
applying a moisture-adJustiPent factor (maf) defimd as the ratio of the max. preci-
pitable water over the catchmnt, W piex, and the precipitable water of the storm,
P
W This factor can be worlred out from consideration of the repremntatiw dew point
op the storm, and the mucimm dew point over the catchment and then finding out the
corresponding precipitable waters from the 'Pressure Vs Precipitable Water' diagram
between the pressure range 1000 mb to 300 mb. Alternatively, in the absence of
sufficient data, a multiplying factor lying between 2C$ to 5s ia assumed.
Havhg thue determined the design storm, the tina-distribution of the rainfall
has to be obtained. From DAD analysis maximtua rainfall depths for durations of 6,12,
18,24,36,48 etc. hours are obtained for each of the atorpis and expressed a8 percent-
of the total rainfall, From a study of these pementages suitable distribution for
the desiga storm is arrived at. Alternativelyif a limited number of self-moording
r a i n g e s are available the ti- distributim cum be obtained from the continuous
records. If no self-recerding gauges are available time distribution based on the
experience of storma elsewhere in comparable area is adopted. Effective rainfall
for difr'emnt time incremnte is estimated by any of tb usual rays, vis,(a) tha
calculation of infiltration loes by finding the total surfme f l w s from actual
flood-hydrographe and c o m m n g them with corresponding rainfall vol~aese.g.
Tenughat krojeat or (b) by simply assuming a runoff factor and applying it to the
design stozm values, e.g. Fíasdeo (Bsngo) Project. These effective rainfall values
are then arranged in the oritieal sequence which may be a m m or less sgmnietric
arrangement of valiies with the greatest value in the middle, or m y be determined
by arranging the rainfall increaients against the ordinates of the design d t
hydrograph ln such a way that the longest odinate faces tke largest effectiw
rainfall and the next largest ordinate faces t b next largeet rainfall increment
and so on, and then reversing this arraageeient to give t b oritical seqrrenœ. It is
then applied to the design unit hydrograph, which can ba derivad by eriy of the
wual nieans,actual obsemtiolls ar synthetic.
6.2.1.1 A recent report (6) suggeete e new Psthod to qatimete t 3 design flood
peak (50-yem r e c u m m e ) from small oatch~mnta(25 Km to 500 KID1. It takes into
account selected besin characteristios (length and weighted .Ban B l o P of the b m i d
88 representative of the b e e h response to tiie storm intaet and the atora P-PieterS
like areal to p o N t rainfall ratio. The procedure hae been evolved from an -lysis
of short-term diaeharge data (5 to 10 yeare) for 60 drahmege basi- Of different
slopes and s i m soattered all over India. It Gen be briefly summed up a8 Pollairs:
The weighiiù m m slope of the main stream, defined 88 given belw,
worked out'
-CL 2 -
= ( Li/+ SB
+L21 2 + .....1
 331

where Lc is the length of the mPin stream ln d l e s fra th@ maeuremnt site to a
point on the main stream near the centre of grevity (CG) of the catchment area,
and S1,S2 etc. are the slopes of the stream in the remhee of lengths L,,L2 etc.
into which the length Lc is divided. Lengths axe mesured from the topoaheet,ln=l mile,
From the value of s, the peak rate of flow Qt, in a tc-hour m i t graph in cuuecs can
beestimated by t h following formiilaer-

(i) Qtc
(ii) Qtc -
I 16000 A%2'3,
320 A6, if
if
s>
s 4 0.0028
0.0028
0-9
t, is the duration of t h rainfall excess given by 255/(Qtc/A)
where A is the area of the catchment in sq. miles. For estimeting the design rainfall,
a 'design storm by6tograph' table hae been p m g r e d giving point-rainfall volume (m)
of 5O-gear return period for durations varying from 15 minutes to 24 hours, and
these are then conmrted to arsal rainfall volume by applying a m a 1 to point rainfall
ratios that have been worked out earlier by analysing data of 12 àense networks. To this
areal rainfall a uniform loss rate is applied which is determimd from the empirical
relatioioohips which have been deriwd for different types of soil. This rill .determine
tb rainfall exoess in t, hours, and the Qtc value multiplied by this excess would
give the design flood peak.
Bowever, these formulae need to be tested further by the field events.
6.2.2 Meteorol w c a l catB P O N i without usina t b U.G

Banerji m d Mantan (7,û) h a m adopted a new approach for eStiIU8ting volume

and peak of runoff from data of atoras. They have studied flood in the N a m a d a and
the ~ h a n a ä ioatohmants. Studring hydrographe of a number of floods (including i a
floods), t h y find out the ti- base of the hydrographs after the bese flow has been
eeparated; it was 6 days in each of these oaseu. The basin is then subdivided into
zone8 of 1-day travel time each, by using the following equation (9)
Tc = L1 .i5 /7.700 ii0.38
where Tc is the time of ooncentration, calculated for all the m i n tributarie8 to tirs
points of outflow in hours, L is tìm length of the remoteat point in the zone t o tb
outlet point in feet ani II ia the diffemzuu in elevatia betwen the waterslmd outlet
and the moet distant ridge h feet.
It io assumed that the contribution to the flood-vol- from each ~ O P Y I is
depeudent on the average a m a 1 depth in e w h didelm, ths scdl moiet- cmdition
and the retenticm orpmity; the proentsge contribution of flood vol- from elch
zone is th- made independent of the %d catchment characteristioB. It has b e n
further aaaued that t b infiltration or retention deoleame from upatream to dom-
8tnam zoma so that if K i8 the Storage f a c t a (considered a8 the f m t i O n Of
ecierage preci itatian & p a a p p a r ~aa runoff )in the zone nearest to the point
of outflcnr,! l is tb starogb faotor f m the nth diviaion m y from the outflm point.
The total daily runoff *Fit at the outlet can then be sriPiPed up as

Shew An i8 t h area and Pn-, ia t b average precipitation recorded in the nth

division. The value of K is seleoted, by trial and error, from past records of
discharge for which simultaneous precipitation data are aleo available. A graph
is then Plotted between values of K obtained for different perioda and corns-
ponding antecedent c a t c h n t rainfall.
For working out peak runoff rates (7) the method does not need a unit
hydrograph, but utilises all the sante the es8ential underlying principle that the
ordinates of two hydrographs for the 8amB basin and similar tine bese are proportional
to their respective volumes of runoff. The correspondence may be effected between
tkie biggest storm on record of which only rainfall data are available and any or
all available hydrographs if the oharacteristios of s t o m are meteorologically
similar to tke outflow point.
Heferences

1. India, Irrigation and Power Projecte (Five Year Plans) 1970, Govwrzuœnt of India,
&inistry of Irrigation and Power.
2. Eetimation of Design nood, bcoimPsndeà Procedures 1972, Govt. G f India,
Central Water 4 Power Commission.
3. b o , G.ii~. 1569 Modern Trenda in Hydrologic Computations, New Celhi
Central Water and P m e r Commission.
4. Khoela, A d . 1949 Analysis and Appraisal of Data for t h Appraisal of water
Resources. Central Board of Irrigatia Jour. pp 410-422.
5. Ha0 G.A.H. 1967. Computation technique for Probable Maximum Flood Discharge
at place in the river while gauge dischare data is available for anotber
pïaoe with special referena to dam on Krishna river, India. Proc. Int. Sgmp.
Floods and their Computation, Aug. 1967, Leningrad, u-s-sj.~*
6. 1973 Flood Estimation Directorate, Central Water & Power Commission, New Delhi,
&sign Office Report No. 1/1973.
7. Bansrji, S d h t o n , D.C. (1967). On estimating peak discharges correeponding
to heaviest redorded a t o m in a oatchment. Ind. Jour. W t . and Ceoph. V01.17
Spl. N0.M 297-306.
6. Banerji, S.Manton, D.C. (1967) Determination of t h e distribution
of rainfall floods in large catchments using hydrometeorological
data. Unesco I n t . Symp. on Floods and their Computation, Lenin-
grad.
 333
--
V A
v
8
z
n
Q)
d
 I
n
4
n
4
4
n
4
-
4
rl
 D A T A R E Q U I R E M E N T S F O R T H E O P T I M I Z A T I O N OF
RESERVOIR DESIGN A N D OPERATING RULE DETERMINATION

J a m e s , I v a n C., Ir
U.S. G e o l o g i c a l S u r v e y , W a s h i n g t o n , D.C., USA

ABSTRACT

A p p r o a c h e s t o t h e d e s i g n of m u l t i p u r p o s e r e s e r v o i r s h a v e u s u a l l y
a s s u m e d a g i v e n s e t o f o p e r a t i n g rules. Conversely, studies o f oper-
ating r u l e s h a v e o f t e n t a k e n r e s e r v o i r s i z e as fixed. In only the
f o r m e r c a s e h a v e e s t i m a t e s of t h e o p t i m a l d a t a r e q u i r e m e n t s b e e n made.
This paper gives the estimates of and compares the optimal length of
data sequences for reservoir design where operating rules are fixed,
for operating rule determination where reservoir design is fixed, and
for the combined determination of operating rules and reservoir size
for a multipurpose reservoir where the benefit is a piecewise-linear
f u n c t i o n of s t o r a g e a n d release. A strategy is developed for the
e c o n o m i c a l l y e f f i c i e n t d e s i g n o f t h e c o m b i n e d p r o g r a m of a d d i t i o n a l
d a t a c o l l e c t i o n and p r o j e c t d e f e r m e n t . T h e s h a p e of t h e b e n e f i t s
f o r e g o n e v e r s u s t i m e f u n c t i o n is s u c h t h a t p r o j e c t d e f e r m e n t i s
usually optimal only where very short hydrologic records exist, and
t h e e f f e c t o f an u n c e r t a i n p r o j e c t i n c e p t i o n d a t e i s t o i n c r e a s e t h e
o p t i m a l - l e n g t h o f t h e d a t a sequence.

RESUMEN

M é t o d o s p a r a e l d i s e ñ o d e un e m b a l s e m u l t i p r o p ó s i t o u s u a l m e n t e
h a n a s u m i d o u n j u e g o f i j a d o d e r e g l a s d e operación. Reciprocamente,
l o s e s t u d i o s s o b r e l a s r e g l a s d e o p e r a c i ó n f r e c u e n t e m e n t e h a n ini-
c i a d o c o n u n t a m a ñ o f i j a d o d e embalse. Estimaciones de los reque-
rimientos Óptimos de datos s e han hecho solamente en e l caso ante-
rior. Este artículo presenta las estimaciones del largo Óptimo de
series de datos para e l diseño de embalses con reglas fijadas de
operación, para la determinacìón de reglas de operación cuando e l
e m b a l s e s e fija, y p a r a l a d e t e r m i n a c i ó n j u n t a d e r e g l a s d e o p e r a c i ó n
y t a m a n o p a r a un e m b a l s e m u l t i p r o p ó s i t o e n q u e l o s b e n e f i c i o s s o n u n a
f u n c i ó n c o n t ì n u a p o r a r c o s d e a b a s t e c i m i e n t o y descarga. Una estra-
tegia se desarrolla para el diseño eficiente economicamente de un
programa junto de aplazamiento del proyecto y recopilación de datos
adicionales. La f o r m a d e l a f u n c i ó n d e b e n e f i c i o s r e n u n c i a d o s c o n t r a
tiempo es tal que el aplazamiento del proyecto usualmente sea Óptimo
c u a n d o e x i s t e n s o l a m e n t e r e g i s t r o s h i d r o l ó g i c o s m u y cortos. E l efec-
to de una fecha incierta del comienzo del proyecto es crecer e l largo
Ó p t i m o d e l a s e r i e d e datos.
336

Introduction

Fundamental t o any development process is an information base

for use in making planning, design, and operational decisions.
This input has measurable costs and benefits as do the other inputs
such as planning resources, capital, and site values for alternate
uses. Economic efficiency requires that the balance between the
inputs of a development process be such that the marginal returns
on all inputs are equal. These marginal returns should be equal
to their marginal costs where budget constraints are not active,
o r equal to their shadow costs when they are active. Viewed as
another input, information may be conceptually handled as any
other input. A s Weiner [i] succinctly states:

"Information is only one of many development inputs;

development, i n turn, is but a transformation process
adopted in order to reach certain objectives. Infor-
mation is, thus, purely an instrumental objective and
not a final purpose in itself, a basic fact we some-
times tend to forget. "

Information requirements for project development include such

diverse factors as hydrology, future prices and extent of markets,
climatology, topography, soil classification, demand level, geol-
ogy, demography, and political trends. In reviewing this list of
information requirements, it can be seen that hydrologic data,
particularly those of a stochastic nature such as streamflows
have distinguishing characteristics that require a different treat-
ment than other information inputs such as economic, demographic,
political, and physiographic data. Climatic and hydrologic phe-
nomena require rather long data sequences t o develop suitable
representations of their generating mechanisms. In contrast,
programs for the collection of physiographic information such as
topographic, soil, and geologic data can be deferred until shortly
before these inputs are needed in the planning process. Models
for projecting economic, demographic, and political trends into
the project life horizon heavily weight the latest inputs, thus
these data are usually collected only shortly before their use.

Planning the hydrologic data collection program requires the

longest lead time and is usually accomplished when a high degree
o f uncertainty exists about other project information inputs.
Efforts devoted to hydrologic network design cannot rely on highly
formal tools i n the absence of these other inputs. An alternative
is the development of heuristic rules based on generalizations
from results of pilot studies of optimal data record length for
specific planning and design situations.

The relationship between the timing o f investments in data

collection and the time stream of benefits gained through the use
 337

of t h e s e d a t a i n the design a n d o'peration o f w a t e r developmecii

projects is an important consideration w h e n t h e s e time s t r e a m s of
costs and benefits are discounted t o a common point in time.
Discounted marginal costs o f the f i r s t y e a r s o f stream gaging are
m u c h h i g h e r than the costs o f gaging j u s t b e f o r e construction.
D e s i g n and construction produce large sunk costs for w h i c h t h e r e
i s little recovery from incorrect decisions unless t h e d e s i g n s
h a v e incorporated a high d e g r e e o f o p t i o n flexibility s u c h as
through staged construction o r o t h e r o f t e n costly m e t h o d s of main-
taining decision liquidity.

T h e problem for the d e s i g n e r of a w a t e r development p r o j e c t

i s t o m a x i m i z e net project b e n e f i t s s u b j e c t to exogenously
supplied constraints. The d e c i s i o n variables pertaining t o the
w a t e r supply design of a r e s e r v o i r are usually the s i z e o f storage,
r e l e a s e target, and operating rules f o r determining s p e c i f i c
releases. Herein, the decision variables are divided i n t o those
physically immutable design values'such as sizing, and t h o s e
operating rule variables w h i c h could conceivably b e c h a n g e d t o
r e f l e c t t h e results of new information.

Whereas design sizing requires h i s t o r i c a l data, t h e use of

information for defining operating rules may be another matter.
Additional data collected as normal requirements o f project oper-
ation may b e used t o update operating policies. On a t h e o r e t i c a l
basis, sequential decision theory provides a methodology f o r a
f l e x i b l e and continually updated operating policy. I n practice,
h o w e v e r , political and i n s t i t u t i o n a l constraints m a k e changes in
operating policy a difficult and expensive process.

T h e determination of trade-offs b e t w e e n capital and o p e r a t i n g

expenditures is a straightforward process w h e n marginal b e n e f i t s
are known. Planning decisions, by t h e i r very nature are f u r t h e r
removed from the t i m e stream of benefits than are design decisions.
L i t t l e is known about the m i x o f planning process i n p u t r e s o u r c e s
w h i c h achieve an optimal d e s i g n from the viewpoint o f e c o n o n i c
efficiency. A large effort cannot b e expended t o d e t e r m i n e t h e
o p t i m a l length of record a t every p o s s i b l e site: rather, s i m p l e
g u i d e l i n e s are needed that a n s w e r s u c h q u e s t i o n s as:

a) H o w m u c h streamflow d a t a i s o p t i m a l at a site €or t h e

expected design decisions and economic p a r a m e t e r s ?

b) W h a t is the m o s t efficient o p e r a t i o n o f a g a g i n g
s t a t i o n w h e n there are uncertainties i n d e c i s i o n s a n d
parameters?

c ) I € the gaging station h a s a l r e a d y . b e e n operated l o n g e r

t h a n the optimal length, w h a t factors would j u s t i f y
à3scontinuing o r retaining t h e s t a t i o n ?
338

An approach to the determination o f optimal lengths of gaging

for the design sizing and operating rule determination o f irriga-
tion reservoirs is presented herein. This system was chosen for
several reasons. In the western United States irrigation reser-
voirs receive the majority of their inflow in the spring months
from rainfall and snow-melt runoff. Usually only small quantities
of natural flow are available during the peak demand months o f
July and August. The essence of such a system can be captured by
a model in which inflows must be stored for satisfying demands in
later seasons or future years. This permits the use of annual
flows and operating rules which are nonseasonal in nature. Bene-
fits from irrigation projects are also usually more easily measured
than for other types of water supply.

The value of data in a particular situation is often limited

by constraints in the decision space. F o r example, if arable land
were limited because of available streamflow, reservoir design
would become sensitive to the irrigation requirements, not mea-
sures of the available streamflow. Systems which have a high
degree of operational flexibility may have low sensitivity to the
availability of data since operating policies can be changed to
consider streamflow data collected after project construction.

T h e Value of Data for Reservoir Design and Operation

The seminal work on the determination of optimal data lengths

for project design is that of Dawdy, Kubik, and Close [2]. This
work has been extended by considering the effect of discounting
and benefits foregone by Moss [31, Herfindahl 141, and Tschannerl
151.

F o r the most part, these studies have dealt with the situation
where operating rules were given and design siting w a s the primary
decision variable. Hydrologic data also have value for the deter-
mination of optimal operating rules.

Operating Rule Determination

The transform of state variables such as storage and inflow

into release values is accomplished through operating rules.
Perhaps the simplest and most often used operating rule for analyt-
ical purposes is the Z rule, where the release R is defined by:

Minimum { S + I, T 1 for S + I
Maximum { T, S + I -Vm
Vm
for S + I > V m

where S is the carry-over storage, I is the inflow, T is the

target, and Vm the maximum conservation storage. This rule would
S e optimal if losses were linear with the deficits below the target
value and no benefits o r losses accrued from releases greater
 339

t h a n the target value. Though these may b e r a t h e r r e s t r i c t i v e

conditions, the Z. rule h a s proved t o b e a useful c o n c e p t u a l tool.
I n recognition of nonlinear loss functions and s e a s o n a l differ-
ences i n expected inflows, Maass et. al. -161 describe modifica-
t i o n s t o the basic release rule called t h e pack r u l e and t h e
hedging rule. Young 171 and H a l l and Howells [ 8 ] used r e g r e s s i o n
analysis o n optimal deterministic releases derived from d y n a m i c
programming to define release rules. R u s s e l l 191 u s e s d y n a m i c
programming t o determine the form o f an optimal operating policy
following a development similar t o t h a t o f G e s s f o r d and K a r l i n 1101.
Both investigators assumed s e r i a l independence i n inflow sequen-
ces. G a b l i n g e r and Loucks [ill, Loucks and F a l k s o n 1121, and
L o u c k s (131 consider design and o p e r a t i o n of s t o r a g e f a c i l i t i e s
w h e r e f l o w s can be described b y M a r k o v transition processes.

T h e Z rule for release determination is not o p t i m a l w h e n

marginal losses increase with the amount o f the r e l e a s e d e f i c i t
or w h e r e significant benefits accrue from competing r e s e r v o i r
purpo'ces such as recreation o r w a t e r power. O n e m e t h o d f o r devel-
oping operating rules is to s e l e c t from a s e t of o p e r a t i n g rules
w h i c h are defined as a function of state variables such as s t o r a g e
and t a r g e t draft. Parameters values of these r e l e a s e r u l e s c a n
t h e n b e optimized in conjunction with reservoir d e s i g n parameters.
T h e selection of a particular functional form o f t h e r e l e a s e rule
w i l l depend upon the objective function, and the n a t u r e o f t h e
assumed mechanism which generates the inflows. For other than
extremely simple problems, this determination is non-analytic.

T h e operating rule chosen f o r t h i s study was:

R = Maximum'{O.O, M i n i m u m (aA - ßB+T, Si+J)}

where
A = Maximum (0.0, S+I-T'Vol
B = Maximum-{O.O, T+V -S -I}
O i
si e: Minimum {V ;
rn
si-1 +I)
a n d CL and ß are r e l e a s e rule parameters, R is t h e r e l e a s e t o bene-
f i c i a l uses, and ifo is a minimum t a r g e t conservation storage. In
o t h e r words, t h e relea'se is t h e t a r g e t m o d i f i e d b y a f r a c t i o n a
of t h e end-of-season contents a b o v e minimum p o o l l e v e l V o o r b y a
f r a c t i o n 6 of the end-of-season contents b e l o w m i n i m u m p o o l level.
T h i s r e l e a s e rule i s m o r e s e n s i t i v e t o p r o j e c t e c o n o m i c s t h a n t h e
2 rule, y e t adds only the t w o p a r a m e t e r s a a n d $ t o t h e optirniza-
tion problem.

T h e previously defined parameters a, ß, T, a n d V , t h e a c t i v e

s t o r a g e volume above Vo, were determined by t h e m e t h o a of g e n e r a l
340

f u n c t i o n m i n i m i z a t i o n d e s c r i b e d b y B e r m a n 1141, u s i n g a r e s e r v o i r
simulation model which returned the negative of discounted net
p r o j e c t b e n e f i t s f o r any s e t o f t h e f o u r p a r a m e t e r s f r o m t h e o p t i -
m i z a t i o n routine.

The selection of a project benefit function is not a n easy

task. The marginal economic response of crops to marginal water
a p p l i c a t i o n s v a r y widely. F o r this analysis a linear benefit
f u n c t i o n w a s used w i t h t h e f o l l o w i n g coefficients.

C1 .O016 $/m3/ s h o r t t e r m l o s s f o r end-of-season

storage below V
0
C2 .O004 $/m3/ s h o r t t e r m b e n e f i t f o r end-of-season
storage above V
O
c3 .o12 $/m3/ short term loss for releases below
target
3
C4 .O04 $/m short term benefit for releases above
target
C6 .O057 $/m3/yr long t e r m b e n e f i t f o r t a r g e t r e l e a s e s

Assumed marginal capital cost3 for providing reservoir storage C

r a n g e d f r o m .O017 t o .O04 $/m /yr. All annual benefits and
5
c o s t s w e r e d i s c o u n t e d at 6 % t o t h e s t a r t i n g p o i n t o f p r o j e c t
benefits.

Analysis of Data Requirements for Irrigation Projects

F i v e g a g i n g s t a t i o n s w e r e s e l e c t e d w h i c h h a d b e e n used i n
t h e d e s i g n o f e x i s t i n g i r r i g a t i o n p r o j e c t reservoirs. The his-
torical data were extended with generated operational hydrology
t o a l e n g t h o f 200 years. The reservoirs were sized and operat-
i n g r u l e s d e t e r m i n e d f o r a l l 1 0 0 , 50, 33, 20, 15, a n d 1 0 - y e a r
s e q u e n c e s o f t h e 200-year s e q u e n c e , a n d o n 8 and 5 - y e a r s e g m e n t s .
f o r s o m e o f t h e projects. F o r each set o f determined parameters,
t h e r e s e r v o i r w a s s i m u l a t e d f o r t h e 200-year p e r i o d a n d d i s c o u n t e d
v a l u e s o f t h e o b j e c t i v e computed. These values were then averaged
a c r o s s a l l e q u a l - l e n g t h s e g m e n t s o f record. Average objective
function values were then plotted against segment length and a
c u r v e s m o o t h e d i n as s h o w n f o r a n e x a m p l e i n f i g u r e 1. T o the
values o f this curve are added the present value of the cost o f
g a g i n g , a n d a t o t a l c o s t c u r v e results. The optimal length of
r e c o r d o c c u r s a t t h e m i n i m u m o f t h e t o t a l c o s t curve.

Operating rules were held constant at their optimal values,

a n d t h e a n a l y s i s r e p e a t e d f o r r e s e r v o i r s i z i n g alone. Also,
reservoir size was fixed and the analysis repeated f o r operating
r u l e s alone. F o r clarity, only the objective function values
 341

and not the total cost curves are shown in the published figure
for these analyses.

These latter analyses were done to attain comparability with

previous studies on the worth of data where operating rules were
fixed and only reservoir sizìng allowed to vary.

Discussion of Results

The present value of the cost of a gaging station record

which is assumed to cost $2000 per year o f operation is given in
table 1 assuming an interest rate of 6%. Thus, in a 50-year
gaging record, the first 10 years has a marginal cost of $ 5 8 0 , 0 0 0 -
$310,000 = $270,000 as compared to the original $20,000 investment
for that record. The effect of discounting the gaging costs is
to make the marginal cost of gaging much higher than might ordi-
narily be perceived.

The results from these studies are tabulated in table 2.

For the economic parameters used, several results are apparent.
The optimal data length for both design sizing and operating-
rule determination is considerably longer than the optimal data
length for design sizing alone, but only slightly longer than
the optimal data length for operating rule determination. There
is a strong correlation between optimal data length and size of
stream, as is seen in figure 2.

I f a gaging station is already in existence and h a s already

achieved its planned "optimal" length of record but the project
has not been yet built, then the decision problem must b e viewed
from a different economic viewpoint. For this, the present value
of the marginal benefits of an additional year of record must
equal or exceed the cost of that marginal year of gaging. The
right-hand column of table 2 is the point on the marginal benefit
curve that equals t h e marginal cost of gaging. F o r example, if
the decision maker designing a reservoir on the Smoky Hill River
near Arnold, Kansas was one year away from construction, then he
should continue to gage i f the total available record is less
than 55 years, even though it may exceed the apriori optimal
length of 30 years.

In previous work by James, Bower, and Matalas í151, it h a s

been suggested that the total variability of meeting a water
quality target was more strongly influenced by economic and
political uncertainties than hydrologic uncertainty. This w a s
based upon a multivariate sensitivity analysis o f output measures.
Such a n analysis is similar t o measurement of the type A error,
which is an apparent error i n design caused by incorrect economic
parameters evaluated at those incorrect parameters. The type B
 P
rlC
a m -rl
C&ci
5.4 m a, O O rn W
oc,-.+ P O O rn
n ax m rl rl
LiOalP
al A A
v) a C Li
E ao o
C O 3 w
4 4
ueoooooo ci C P
m ~ r l m ~ ~ e w om o P C h
ci ne-Julwcnmw ci T
I .FI rl rn rn CT ul
E m m N C N rl rl
al -rl o V
E
ci
P
C
am
rn .rl m
;
C
P
m
br
a,
rl
2
rl Li
a
iu al P h
O ci C d ul O b IC
V -4 C * cy N
u 0
5
a,
a,
rl
a,
m
Li
Li rn al
ci a
m Li. O
O
a, iu. a
E I C P
4 00000000 rn m C
c, 00000000 S’ 4 rn
~ W 0 0 0 0 0 0 ci C c i a, IC -J O PI
a, . œ œ œ - œ . - e. m a 4 -J w m
s L D ~ W O O W W N C.
N I C ~ F I W W a,~ ~
-FI Li 3
ma,&
ci
rlmul0m-J rll ma
W . œ œ œ n o
O C rlrlm a
O k
a, .rl O U
Ici U a,
4 0 a, m
ma, \
*z
cio
d
a
a,&
e
m
rl d< W rn
N
E U E * mal
m
a, .rl U P I* O
mid ci @Li N N
a 4 a * a
Lid O . 4s
pia V
00000000 I m
rlNmeulWbQ) I -rl
a
rl
(Y O
al m a c
d a, E m m a
nid rl id &I d ci
A ci O m D
E a E rl LiC
€4 O O ala x@ E-
e Liu
al I-.
9 %
alo
i l -
O h .:;“a‘
* n “ d rlrl
u+>
om
X
rlrl n d o CLi
om
sa, mal rifi
E c m - a XLI Url
u v
O mcn mpi 4
.rl rl E h xL i L i
ci mLi aLi XLi
m il6 X d o m a m
ci ma, Lia, Bal dal
VI E C V C
 343

error, o r efficiency loss, is the pertinent measure t o compare

against efficiency losses from inadequate gaging records. The
type B error is the loss measured with the true economic param-
eters of a system whose desfgn was optimìzed with the incorrect
parameters o

The type A and type B errors resulting from 40% errors i n

each of the cost parameters used are shown in table 3.

Table 3. Type A and type B errors resulting from 40% error

in cost coefficients for a reservoir on the Smoky
Hill River near Arnold, Kansas

Cost Coefficient Type A error T y p e B error

(% of net project benefits)

c1 recreation 3.1 < .1

c2 recreation < .1 < .1
c3 short run deficit .6 < .1
c4 short run release 22 2.5
c5 res e rvoi r 10 < .1
construction
C6 long run release 60 11

F o r comparison, on this same project the efficiency loSS for

incorrect record lengths of 5, 10, and 1 5 years is 0.4, 1.4, and
3.7% resPectivelY. Project feasibility is usually quite sensitive
to type A errors and t o the interest rate used for discounting
project costs and benefits to a common time name. Type A - a n d B
errors were not estimated for errors i n interest rate, however,
on this project a change i n interest rate from 6% to 5% would
change the optimal record length from 30 t o about 35 years.

This analysis has taken the manager's viewpoint o f a project

gaging network, whose objectives are overall economic efficiency.
Budget constraints are not active. T h e network manager seldom has
precise information on the date o f project inception. H i s problem
is then t o minimize the losses by starting gaging at a point
years in the future such that the expected loss
m
C P(t)L(t--r) is a minimum, where P(t) is the probability t h a t
ta0
the project will be started t years i n the future, and L(t-T) is
the t o t a l loss function as sho& i n the example i n figure 1.
Water resources development projects also must consider other
criteria than economic efficiency. Considerations of regional
income distribution and the degree of risk aversion in the parties
receiving the economic benefits w i l l temper gaging network
decisions.
344

If the analysis presented herein is t o b e used t o determine

ghe optimal sìarting time f o r a gaging station a t a previously
yngaged site, s o m e estimates of the f l o w characteristics w i l l
bave to b e made. Estimates o f flow characteristics prior to gag-
ing c a n b e made for sites in the Unfted States using regional
gelationships presented in Thomas and Benson n61. Additional
gaging information can then b e incorporated into the prior ecti-
wgtes using Bayesian analysis.

cgnclusions

Optimal record lengths for determining reservoir sizing and

ekerating rule parameters can b e determined by computing expected
project benefits from designs resulting from varying lengths of
design data. Optimal record lengths f o r the combined process of
reservoir sizing and determining operating rule parameters are
significantly longer than for reservoir sizing alone under a
fixed optimal operating rule.

T h e optimal length of record increases with s i z e of stream,

ganging from 7 t o 4 7 years for the five str ams used i n this study,
9
w h i c h range i n discharge from .25 to 27.1 m /sec.

Econcmic efficiency requires that the marginal worth of col-

ìecting additional streamflow data b e greater than the marginal
costs when these two values are discounted to a common point in
%ime. Hence the decision problem of discountinuing an existing
gage is different than the decision problem for the optimal
gtarting time for the gage because the discount factors applied
g p the initial year are larger than those applied t o the final
year.

References

Wiener, Aaron, (1972). T h e role o f water in development,

New York, McGraw-Hill, Inc.

(1970). T h e value
Dawdy, D.R., Kubik, H.E., and Close, E.R.
of streamflow data for project design -a pilot study,
Water Resources Research, v. 6, no. 4, pp. 1045-1050.

MOSS, M.E. (1970). Optimum operating procedure f o r a river

gaging station established t o provide data f o r design of a
w a t e r supply project. Water Resources Research, v. 6, no. 4,
pp. 1051-1061.

Herfindahl, Orris C., (1969). Natural resources information

f o r economic development, Baltimore!, Johns Hopkins Press.

Tschannerl, G., (1971). Designing reservoirs with short

streamflow recoräs, Water Resources Research, v. 7, no. 4,
pp. 827-833.
 345

6 Maass, Arthur, et. al. (1962). The design of water resource

s y ~ t e m s ,Cambridge, Harvard University Press.

7. aung, G.K. , (1967). Findiiig reservoir 01 crating ruler. Jour.

f the Hydrau1ic.s Div., A m Soc. of Cìvi I Enqr.. v . 9 3 . no.
HY6, pp. 297-

8. Hall, Warren A., and Howell, David T., (1963 . The uptimiza-
tion of single purpose reservoir design with the appii ition
of dynamic programming to synthetic hydrology samples, Jour.
o f Hydrology, v. 1, pp. 3 5 5 - 3 6 3 .

9 Russell, C. Bradley, (19121. A n optimal policy for operating

d multipurpose reservoir, Operations Research, v :O, no. 6,
pp. 1181-1189.

lu. Gessford, John, and Karlin, Samuel, Optimal policy for hydro-
el-ectric operdiicns, pp. 179-200 in Arrow, Kenneth, J.. Karlin,
Samuel, and Scarf, Herbrrt (1958). Studies in the mathematical
theory of inventory an..;productions, Stanford, Stanford Univer-
sity Press.

1 . bablinger, Moshe, and Loucks, Daniel , (1970). Markov models

for flow regulation, J O U I . of the Hydraulics Div., Am. SOC.
Civil Engrs., v. 96, no. HY1, pp. 165-181,

IL Loucks, D.P., and Falkson, L.M., (1970). A comparison of Some

dynamic, linear and policy iteration methods f o r reservoir
operation, Water Resources Bulletin, v. 6, no. 3, pp..385-399.

13 Loucks, D.P., (1970). Some comments on linear decision rules

and chance constraints, Water Resources Research, v. 6, no. 3,
pp. 668-671.

14 Berman, Gerald, (1969). Lattice approximations t o the minima

J f functions of several variables, Jour. of the Assn. for
omputing Machinery, v. 16, n,,. 2, pp. 286-294.

15. James, I.C., 'II, Bower, B.B., and Matalas, N.C., (1969 ,
Relative importance of variables i n water resources planning,
Water Resources Research, v. 5, no. 4, pp. 1165-1173.

16. Thomas, D.C., and Benson M.A., (1970). Generalization O f

streamflow characteristics from drainage basin characteristicsr
1v.S. Geol. Survey Water Supply Paper 1975, 5 5 P.
346
U
 347
O 10 20 30 40 50
OPTIMAL LENGTH OF RECORD
 T H E DESIGN OF W A T E R Q U A L I T Y M A N A G E M E N T P R O J E C T S
W I T H I N A D E Q U A T E DATA

G e o r g e W. Reid
Regents Professor
University o f Oklahoma

ABSTRACT

One o f t h e i n c r e a s i n g l y i m p o r t a n t e l e m e n t s i n t h e d e s i g n o f
w a t e r r e s o u r c e p r o j e c t s is, o f c o u r s e , t h e m a n a g e m e n t o f q u a l i t y
and a t e c h n o l o g y t h a t w a s a l m o s t p u r e l y h y d r o l o g i c a l and h y d r a u l i c
is now being expanded t o i n c l u d e w h a t m i g h t be classed a s t h e
e n v i r o n m e n t a l and e d o l o g i c a l impact a r e a s and s y s t e m s , S o , it is n o
l o n g e r s u f f i c i e n t t o u n d e r s t a n d t h e i n t e r r e l a t i o n s h i p s , f l o w s and
t r a n s p o r t s , but t o t h i s must b e a d d e d t h e i m p a c t s on ttbe lisJrng and
n o n l i v i n g w a t e r , and p e r i p h e r r a l e n v i r o n m e n t s ; w i t h a n e e d to
develop ecological models or more specifically, water quality m o d e l s ,
U n f o r t u n a t e l y , t h e r e i s r a r e l y a d e q u a t e d a t a t o properly d e s c r i b e
these interrelationships, The methodology used for hydrological
studies involving inadequate data such a s the transfer o f okserved
p o i n t s t o p o i n t s o f i n t e r e s t ; s h o r t t e r m interise s t u d i e s ; & , Y u s e o f
s i m u l a t i o n t e c h n i q u e s , can and are being u s e d i n quality m a n a g e m e n t
m o d e l i n g , P e r h a p s m o r e b a s i c is a n u n d e r s t a n d i n g of d a t a r e q u i r e m e n t s ,
u s i n g t h e s y s t e m approach, t h e s e q u e n c e o f e v e n t s a r - i l ) p r o b l e m
f o r m u l a t i o n , ( 2 ) symbol1 m o d e l i n g , ( 3 ) d a t a c o l l e c t l o n , (4) a n a l y s i s
and ( 5 ) design. (See F i g u r e 1 ) F r e q u e n t l y , t h e o r d e r is , h a n g e d ,
p a r t i c u l a r l y t h e entire p r o c e s s w i l l s t a r t w i t h a v a i l a b l e data.
The complexities, o f c o u r s e , a r i s e due t o t h e fa.t t h a t t h e
rocesses associated with water quality management: hydraulic,
I

hydrologi,al, chemical, b i o l o g i c a l and e c o l o g i c a l -- are extremely

and i m p e r f e c t l y understood. So, t h a t is a c o m p l e x r e a l i t y , w i t h a
g r e a t many v a r i a b l e s o n w h i c h t h e r e i s a v a i l a b l e veri p o o r m e a s u r e s
and w h i c h t h e m s e l v e s i n t e r r e l a t e in w a y s very i n a d e q u a t e l y
understood -- must b e m e a s u r e d a n d a p p r o p r i a t e l y r e l a t e d t o be u s e f u l ,
C e r t a i n l y , one r e c o g n i z e s t h e s u p e r i o r i t y o f a n e x p l i - i t q u a n t i -
f i a b l e d a t a and m o d e l s o v e r i n t u i t i v e m o d e l s a n d hurlChes. T h e
a l t e r n a t i v e s t o s u c h a m o d e l , b a s e d on p a r t i a l k n o w l e d g e , is a m e n t a l
m o d e l , b a s e d o n t h e m i x t u r e o f i n c o m p l e t e i n f o r m a t i o n and i n t u i t i o n
similar to those controlling most political decisions. A mathematical
m c d e l deals w i t h the s a m e i n c o m p l e t e i n f o r m a t i o n a v a i l a b l e t o a n
1 1 * u i t i v e model, but t h r o u g h o r g a n i z a t i o n o f i n f o r m a t i o n f r o m m a n y
iifferent s o u r c e s into a c l o s e d l o o p at l a s t a n a l y s e s i s p e r m i t t e d
and d a t a n e e d s s t u d i e d ,
350

RESUMEN

Uno d e los elementos a l t a m e n t e i m p o r t a n t e s en e l diseño d e p r o

y e c t o s d e r e c u r s o s de a g u a es, d e s d e luego, e l m a n e j o d e l a calidad
y t e c n o l o g i a que f u e c a s i puramente hidrolögica e h i d r á u l i c a y está
s i e n d o a h o r a expandida para i n c l u i r lo q u e d e b e d e ser c l a s i f i c a d o -
como á r e a s y s i s t e m a s d e impacto a m b i e n t a l e s y ecológicos. Asi que -
y a n o s e r á s u f i c i e n t e entender l a s interrelaciones, f l u j o s y trans--
portes, p u e s a éstos d e b e n d e ser a g r e g a d o s l o s i m p a c t o s en l a s ---
a g u a s c o n y s i n p r e s e n c i a d e f o r m a s d e vida y los a m b i e n t e s perifi--
cos; con la necesidad d e d e s a r r o l l a r m o d e l o s e c o l ó g i c o s o m á s e s p e c l
f i c a m e n t e m o d e l o s d e calidad de agua. Desafortunadamente, r a r a v e z -
existen datos a d e c u a d o s p a r a d e s c r i b i r p r o p i a m e n t e e s t a s interrela--
ciones. La m e t o d o l o g í a u s a d a para e s t u d i o s h i d r o l ó g i c o s incluye in--
f o r m a c i ó n inadecuada, t a l e s c o m o el c a m b i o d e puntos o b s e r v a d o s a -
puntos d e interés; e s t u d i o s i n t e n s o s d e c o r t o plazo; o u s o d e técni-
cas d e simulación, p u e d e n y h a n sido u s a d a s en m o d e l o s d e m a n e j o d e
calidad. En e l m o d e l a d o existe s i e m p r e una c i e r t a incompatibilidad -
e n t r e puntos de s u s t a n c i a y g e n e r a l i d a d ; r e q u e r i m i e n t o s d e informa--
ciÓn y l a r e p r e s e n t a t i b i l i d a d d e l m u n d o real. E l o b j e t i v o d e s d e lue-
go, e s p r o v e e r por m e d i o d e u n a a b s t r a c c i ó n i d e a l i z a d a u n comporta--
m i e n t o a p r o x i m a d o el c u a l es s i e m p r e u n c o m p r o m i s o entre simplicidad
y realidad. En a ñ o s r e c i e n t e s u n a g r a n cantidad d e m o d e l o s h a n sido
d e s a r r o l l a d o s , pero, d e s a f o r t u n a d a m e n t e parece h a b e r u n a l t o g r a d o -
d e polarización. En u n extremo, hay u n elegantisimo y s o f i s t i c a d o m o
delo b a s a d o en t é c n i c a s e c o n o m é t r i c a s r e q u i r i e n d o u n a l t o g r a d o de
e s p e c i f i c a c i ó n d e información, que en l a r e a l i d a d n o existe, Por --
-
otro l a d o d e l espectro, l o s s e n a r i o s d e p e n d e n c a s i m u y poco d e info;
m a c i ó n , m á s sobre conceptos. La necesidad b á s i c a es para m o d e l o s en
aìgfin l u g a r entre los d o s extremos q u e están c o n s t r u i d o s u s a n d o in--
f o r m a c i ó n existente y q u e p u e d a n s e r r e s p o n s a b l e s a l a s n e c e s i d a d e s
d e l a s a g e n c i a s d e a c c i ó n , Es en esta r e a l i d a d en l a c u a l e l a u t o r -
h a d e s a r r o l l a d o una s e r i e de m o d e l o s d e c a l i d a d d e agua. Los proyec-
t o s siendo m o d e l a d o s son g e n e r a l m e n t e d e u n a n a t u r a l e z a t a l q u e la
r e a l i z a c i ó n f i n a l o c u r r i r á b a s t a n t e d e s p u é s d e l a partida d e l o s di-
-
s e ñ a d o r e s y p o r t a l los p r o c e d i m i e n t o s d e evaluación d i r e c t a s o n im-
posibles, n e c e s i t á n d o s e d e a l g u n a f o r m a d e e v a l u a c i ó n o integridad
interna. E l problema es que usando c u a n t a i n f o r m a c i ó n esté disponi--
-
ble, para 50 a 1 0 0 años a l a f e c h a y haciéndolo d e m a n e r a que n o s e a
t a n e l e g a n t e que s e c o n v i e r t a en u n m o d e l o dogmático. E l autor h a -
d e s a r r o l l a d o una s e r i e d e m o d e l o s r e s p o n d i e n d o a l desafio. La esen--
cia de la m e t o d o l o g i a e s r e c o n o c e r l a c o m p l e j i d a d d e u n problema y -
t r a z a r una c o m b i n a c i ó n d e t é c n i c a s d e i n v e s t i g a c i ó n d e o p e r a c i o n e s ,
t é c n i c a s deterministicas, a s i c o m o m’etodos empîricos, f e n o m o l ó g i c o s
y analiticos. Modelos p a r a s i s t e m a s d e rios r e s p o n d e n a la p o l u c i ó n
o r g a n i z a d a de cuatro maneras: bioquimica, biodegradable, sedimentos
nutricionales, incluyendo modelos adicionales para flujos urbanos y
poluciiin dispersa, a s í c o m o f l u j o s rurales. T o d o s los m o d e l o s u s a r o n
i n f o r m a c i ó n existente y ésto los sitG‘a para m o d e l a d o p r o n o s t i c a b l e
de n i v e i de l a s c u e n c a s s i e n d o c o m p u t a r i z a d o s y s i s t e m a t i z a d o s y es-
-
t á n s i e n d o u s a d o s en p r o b l e m a s específicos en e l S u r o e s t e d e los Es-
t a d o s Unidos.
 3 51

Problem Formulation: To arrive at a water resource project design, the number

of variables is enormous, and they are mostly nonlinear. The structure of
the system is more hierarchical than functional, and many of the parameters
and variables are unquantified at present, certainly those associated with
ecology. Nonetheless, to some degree, a merging of disciplines and the
increased use of the system approach has been taking place in the study o€
urban systems, and it is not just a matter of collecting data and figuring
out what one has.
Lf one looks at the type of models being postulated €or the design of
water quality systems today, it will be seen (Figure 2) that they fall
within a spectrum ranging from erudite mathematical models at one end of
the spectrum to scenarios at the other. In the first case, the mathe-
matical models may be rigorously developed in a mathematical sense, but
are all too often of little use in describing a real complex system in
inadequate data. On the other hand, the scenario model - little data,
numerous ideas --may accurately depict the significant elements of the
real system, but it is of little use to the engineer-planner because he
cannot manipulate it. or quantify it.

The target one should try to hit is a reasonable and useable balance
hetween the poles of intuition and selecting hard data. One would like
to be able to use the mathematical rigor of the physical scientist and.
at the same time, give equal weight to the heuristic insight of the social
scientist. The result would be a useable model for a system design. So,
perhaps, or certainly, for planning purposes, one is dealing with the lowest
level of quantification that allows good estimates and the lowest level of
complexity which gives a reasonable picture of the real world system with
the lope of expounding in both directions.

The dpplication of mathematical modeling techniques to water quality

management can significantly aid the decisionaakers to arrive at better
decisions. Thus, modeling provides relevant facts and alternatives, the
decision-maker chooses the strategy. Operational modele are still prim-
itive, primarily becaiise of the probilistic or random nature of the
physical processes involved in waste diffusion. One is sometimes inclined
to be skeptical of the value of increasing model sophistication which
often seems to have progressed much further than our understanding of the
rmrnplex real world situation; all models currently proposed in the literature
have enormous data requirements which far exceed those data usually available,
and which, for the most part, must be derived from actual measurement.
Many parameters in the more sophisticated models are simply not known in
actual situations.

The water quality management design problem require:

1. The cause and effect relationship between pollution from any
source and the present deteriorated quality of water in the estuary.
2. Forecasting variation of water quality due to the natural and
man-made causes.
3. Methods of optimal management, including treatment and flow
regulation to control the quality in the estuary for muaicipal, industrial,
agricultural, fisheries, recreation and wild life propagation.
4. Chemical, biological, hydrological, hydraulic, at the same t h e ,
same place, and same accuracy.
 352

Models In modeling there is always a certain incompatibility and repre-

sentativeness of the real world. The aim, of course, is to provide through
an idealized abstraction an approximate behavior of the system which always
is a compromise between simplicity and reality. Water quality models can be
used to simulate, describe and predict, and programming leading to optimi-
zation of design. Programing which leads to policy requires an explicit
set of objectives, or an objective function to maximize benefits or minimize
costs. Simulation does not require explicit results. So, simulations are
misunderstood, if one expects to use the numerical projections and values.

Using numbers is wrong if it leaves the impression that design projections

are in any way predictions of the future. It is helpful, E as a prediction
but to get one to realize how short-sighted --
how present-oriented- images
of the future ordinarily are, but extrapolltion of present trends is a time-
honored way of looking into the future. Most people intuitively and
correctly reject extrapolations --
the point is that it provides indications
of the system's behavioral tendencies and as an analysis of current trends,
of their influence on-each other, and of their possible outcomes.

Models may be classified usefully by areal extent into national, regional

and local . At the highest, or national level, data is necessary for broad
planning purposes, such as to determine an overall level of water pollution,
to determine the total investment necessary for pollution abatement, to
determine national policies and to project the problems into the future.
At the second highest level, the regional level, all of the above information
is necessary, plus the particular information needs for the region. The
third, local level, consists usually of checking the operation of waste
treatment plants to insure compliance with regulations and statutes. Thus,
due to the different requirements and objectives, a data program which may
be optimal at one level, is usually far from optimal at some other level.
Unless a clear objective has been set, there is no guarantee that all
critical bits and bytes of information are collected, and that the gathering
of useless data is minimized. Similar calssification classification can be
made with relation to time.

Hypothetical attempts to describe the intricate relationships between

nutrients, phytoplankton, zooplankton, fish, detritus, bacteria and man-
induced waste loads. There has resulted a great variety of models. One of
the first developed, classical Streeter-Phelps equation, describes adequately
the deoxygenation and reoxygenation in the river. The familiar form of the
oxygen sag equation is:

where: D --- oxygen dificit at time t

oxygen deficit at time zero
Do
Lo -
BOD at time zero
t = time (distance) in days
deoxygenation coefficient
kl =
k2 = reoxygenation coefficient

This equation has been expanded to provide for evection and diffusion; algae
growth, beuthal deposits, etc., into, inreality, impossible data requirements.
The basic need is for models somewhere between two poles that are built using
existing data and as such can be responsive to the needs of the action agencies.
Tt is in this realm in which the author has developed a series of water quality
models. The projects being modeled generally are of such a nature that the
ultimate realization will occur long after the departure of the designers, and
as such direct validation procedures are impossible, necessitating some form
 353

of internal validation or internal integrity. The problem is one of using

what information is available for a 50-100 year future, and doing it in
such a fashion that it is not so elegant that it becomes a classroom make-
believe world, The essential thread in the author's methodology is that of
recognizing the complexity of a problem and drawing on a combination of OR
techniques, deterministic techniques, as well as imperical, phenomological,
and analytical methods. River system models respond to organized pollution
L,I modes.

There are suggested six categories of stream responses: biodegradable,

nutritional, bacterial, solids, persistant of slowly degradable chemicals
and thermai. The response of a given stream to these categories can be
formulated; or the reverse. given an instream criteria (RQS), allowable
effluent quality can be calculated. The specific criteria now can be
grouped under response headings; for nutritional, one might select N, P,
NIP, or AGP, etc. If primary treatment is established as a lower con-
straint on the effluent, the solids criteria can be deleted; and further,
if a public health constraint on toxic and bacterial levels can be exercised,
fosir rather than six responses can now be used leaving a four-by-four matrix
'o be examined.

TABLE I

Municipal Industrial Agricultural Recreational

Biodegradable Controlled by D. O. levels
Nutritional Controlled by N and P levels
Thermal Controlled by Temperature increases
Persistent
Chemical Controlled by Salt, CCE's or ABS, etc.

So, a response/use matrix, changing with time will set goals; based on a
matrix such as the one in Table I "d alternative socio-operated projections.
A linking technical basin model can be built and operated to provide the optimal
use of water resources, and of necessary treatments; or in pianning for
ruture population increases and the concomitant increased use of water, it
is possible to build mathematical models depicting the optimum treatments and
stream flows necessary to meet ths RQS. The one-to-one input-output relation-
ships f-r he four c-tegorius vf waste discharges follows with the Low Flow
Augmentation FA), associated with each treatment level (mi) , will be QL,
QN, Qp and .,Q This is a terminal flow in M O . T'Li is a fraction where i
refers to BOD, N and P.

BIODEGRADABLE MODEL (L)

yY + (i-Y) PE or P A (P)
Q,
CS - RQsDO

(i) where:
Y = Fraction of total population in SMA's
=E Efficiency term, Point LoadIUniform Load
PE = Population Equivalent Ln millions
P = Percentage discharge to river, expressed
 Cs =
Variable -
as a fraction, Decision
(1-TL)
DO saturation level 6 given temperature

A = 942,900 relates to stream character-

k2 4 * istics

where n is essentially the number of reoxygenized volumes, Ir the reaeration

2
constant, L the reach, V the velocity
stream Ltself is subject to management.
-- these valués will change as the

ACCELERATED RITROPHICATION MODEL

Z'P (1-%-1.44 (1-5) (TLL3250) (3)

Qp = 2-P (1-TLp) - .27 ( 1 - 5 ) (TLL 1080) (4)

F, ROS

where:
Qp or QN = Nutritional Dilution Required, MGD

Z = Relative portion impounded and

effected by ,SQR level

T$ or %
P
-
= Population, millions
Phosphorus or Nitrogen removal level

F or Fp
N
RN
- =
expressed as a decimal
BOD/, Ratio divided by optimum
combining ratio
BOD removal level expressed as
a decimal
RQS, or RQS, = Acceptable level, RQS determined
by RQSAGp

THERMAI. MODEL (T)

Qr =
ATw C - AQ
AT + C
Q
= Thermal Dilution Required, MGD

A Lw - Allowable temperature difference between added flow and RQSt (t-RQSt)

- To)
A TQ = Allowable temperature change (RQST
( Ratio of K/Vx when K - Geometric mean for Bowmen's ratio and V =
 355

subsidance velocity
A Q = Waste Flow, MGD

CONSERVED OR PERSISTENT SHEMICAL MODEL (C)

These models, though’cast in terms of dilution requirements, can be

altered, given a diluted level to provide permissible loadings. The
models (2) thru (6) are based on organized (sewered) pollution. Models
for storm drainings or dispersed pollution have also developed such as:

DISPERSED POLLUTION MODEI. (D)

Y2 = 4.8 + 0.0827X2 + 0.489X8 (7)

where Y is BOD
3
Y =
5
-
0.188 1nX + .310 Inxl0
2.36

where Y is ON and Y6 is PO, in

5
Y6 = 2.90 + .OOOOSX1 .OOOlX, -.0137X8 - - .741Xll
and Xi = population
X2 =. population density
X
3
= number of households
X8 = comercial establishments
Xl0 = streets
Xll = environmental index

Models (2-9) can be used to relate waste inputs to stream responses under
varying municipal stream characteristics and against varying goals (RQS).
Many technical models are available to project flows (Q), and other stream
characteristics it2, L, V, et.. but a final model is needed for evaluation of
the effects of the rural upstream watershed programa on downstream runoff to
complete the set. Such a model was developed for the Congress in 1969. ls2

For details of model. development see, THE OUTLOOK FOR WATER, Wollman and
Bonem, The John Hopkins Press, Baltimore & London, 1971, Appendix C., p. 203.

This was a special consultative report to the Secretary of the Interior,

October, 1967.
 356

UPSTREAM USE MODEL (U)

Y=-16+XX
13 - u7x2 (10)
Where:
Y =i percentage of nonna1 runoff
X1 -i percentage of normal precipitation
X2 * percentage of watersheds controlled by hydraulic structures
Xj = annual above one inch precipitation

In these equation, the simple Phelps equation (1) has been reduced to:

dL = % dO = f dO (11)
n
That is to say, the load equals the capacity. Distribution factors are
added, load is put in terms of people, PE's, etc. This is useable. On the
other hand and by way of contrast, O'Connor uses a one dimensional, differ-
entia1 equation, first involving:

- kdL - kn L" + ka (Cs -C) -

Expanding this to three dimension, (x, y, z,) would require:

ao E 2 E a20
- i P
at -xa: ++ i-
EZa2 o -
ax ax
- kic , etc. (13)

Also the evaluation of E's, U, Ki, etc. in terms of velocity, solar energy,
depth, turbidity, etc. 3

The effectivehess of models is, of course, acceptance. Actually, very few

models have been used. Limitations of applying them to "real" systems are
rooted in many factors, most related to data inadequacies; the acquisition
of proper data, adjustment of non-homogenity, or inconsistency, to M I Ua ~
few.

SYSTEMS ANALYSIS AND WATER QUALITY, Thoman, Environmental Science Service,

New York, 1972.
Every model, or system, is always embedded in a larger system in space
or time, so one is limited to selection of a free body cut and exogenously
determined parameters. Finally, serious factors, mostlv associated with
social values cannot, at present, be quantified.

An efficient use of models thus, argues for different models to answer

different question. For example, one for sediments, one for social costs,
etc. The systems process is iterative and continues while the models are
refined and until satisfactory results are obtained.

The flow of information for all the mested models eventually leads to the
decision process. Forward and feedback information flows take place between
models until the alternative selection and information developed is accepted
for decision-making. As illustrated, there is no attempt to "hang" all
models together. More important, different levels of data, can be used in
each mode, providing homogenity in each model.

DATA

The data must support the models. Some of the questions for which answers
are needed are, goals, include,:
1. What significant parameters of water quality should be measured,
for an alert system, for treatment plant control, for a quality forecasting
system, for a river management system?
2. What should be the periodicity or time interval in collecting
specific data?
3. What are the cross correlations of these parameters?
4. Are there any synergisLic relationships between the parameters?
5. What is being accomplished to develop instrumentation that can
gage quantitatively those essential parameters, such as BOD, that are not
being measured automatically at the present time?

So, there are all sorts of data, much of it redundant. One needs a model
to discover needs, costs, etc. The process is shown graphically in
Figure 3.

Data has a cost, collection and deferral of decisions.

The quantity of information collected should be increased so long as the

present value of the investment opportunity (or cost savings if this is the
use to which the information is put) is increased by more than the cost of
the information.

The expected value of a decision will be low with little data available, but
will rise with more data available. With little data available. the solution
often would be overstated (resulting in unused capacity) or understated
(resulting in lost opportunity), thus reducing the expected present value of
tht opportunity. For small enough quantities of data, the expected value
will be negative.

The conclusion thatthe decision take place when the cost of getting one more vear
of information is equal to the resulting increase in expected present value.
The cost of getting one more year of data is made up of two elements, the
outlay during the during the year to get the data, k, and interest on the
expected present value of the opportunity one would experience if a year of
waiting is not included. That is, if V(t) is the basic function, one should
not wait until its rate of increase, V’(t), is equal to [rV(t) + LI, where
r is the rate of discount (the rate of return on investment).

Several conclusions are evident. First, it never will pay to wait for
”complete” information. Second, an extremely important element of the
problem is the cost coming from postponement of the stream of net revenues
from the decision. This factor means it does not pay to accumulate data
until the increment in expected value is equal to the annual cost of the
data.

Experience in the United States has resulted in the common utilization of

only eight water quality parameters that are thought to satisfy the re-
quirements of reliability, accuracy, and low maintenance. These parameters
are dissolved oxygen, pH, turbidity, conductivity, temperature, OñP, solar
radiation intensity and chlorides. Time sequence is important. Parameters
needed today may not be the correct ones torrorrow.

T A B L E II

TIME SCHEDULE FOR WATER POLLUTION ABATEMENT

-
Time
Secondaw
Treatment
- -
BOD
Eff
-N&P
Eff Eff
TDS Thermal
-Ef f
1960 X
1970 X X
1980 X X X
1990 X X X X
2000 X X X X X
Criteria Fish KJlls Eutrophi- Reuse Recycle
Water Treatueur. cation
Problems

Figure 4 suggest a water pollution abatement time scale; that is, the
standard will be upgraded with time, and the resource must be used
within these constraints.

One is still concerned with the frequency with which data should be
collected, the optimum locations of collection, the provisions for data
storage and the resources for analysis of the data. The use of a short-
term survey approach or establishment of a minimal number of permanent
stateions. An analysis of historical data will yield insight into those
parameters which require continuous analysis because of significant fluc-
tuations and help to identify those locations which best identify changing
conditions in the receiving water.

In contrast to the monitoring of a simgle point over a long period, studies

can be concenrrated over shorter times but more intensive. There is a
questi01 ,f manual collection versus continuous, automatic recording. All
parameters of interest can be determined on a continuous basis and the results
 359

transmitted to a central storage facility, while water quality parametere

that can be economically and dependently measured in the field are still
somewhat limited.

CONCLUSIONS

Briefly, models to illucidate design parameters should be built with

available data in mind. By a process of separating and nesting, sub-
models can overcome inconsistencies. If goals are precisely stated as
to function, various parameters can be represented by what ia available.
The author has developed a series of models using very general data,
leaving a latitude of alternative data.items to define a parameter. Data
has a cost, collection and opportunity or decision errors also cost. If
inadequacies continue, short-term intensive studies are justified, either
now or backward, for example, point reviews can be used. Manual systems
can be replaced by automatic monitors; all eight suggested parameters
handled by electrodes. Generally speaking, however, automatic monitors
tend to provide more data than are needed, because noone dares to turn
these expensive machines off or set the sampling interval to such a time
interval that meaningful deviations can be recorded.

One never has adequate data, nor can one afford to wait for it. So, models
must be made using every device available, recognizing that the final
result will still involve uncertainty and risks, and require judgement
the only defense against inadequate data.
-
360

SELECTED REFERENCES

1. Biswas, Asit K. PROCEEDINGS, INTERNATIONAL SYMPOSIUM ON MODELLING

TECHNIQUES IN WATER RESOURCES SYSTEMS. Volumes 1 and 2. Ottawa,
Canada: Environment Canada, 1972.

2. Herfindahl, Oris C. NATURAL RESOURCE INFORMATION FOR ECONOMIC

DEVELOPMENT. Baltimore: John Hopkins Press, 1969.

3. Krenkel, Peter A. (ed.). PROCEEDINGS OF THE SPECLAZTY CONFERENCE

ON AUTOMATIC WATER QUALITY MANAÇEMENT IN EUROPE, No. 28. Vanderbilt
University, 1971.

4. Mancy, Khalil H. (ed.). INSTRUMENTAL ANALYSIS FOR WATER POLLUTION

CONTROL. Ann Arbor, Mich.: Ann Arbor Science Publishers, Inc., 1971.

5. Public Health Service. U. S. Department of Health, Education and

Welfare. SYMPOSIUM ON ENVIRONMENTAL MEASUREMENTS, VALID DATA AND
LOGICAL INTERPRETATION. Cincinnati, Ohio: Public Health Service, 1964.

6. Public Health Service. U. S. Department of Health, Education and

Welfare. SYPPOSILJM ON STREAMFLOW REGULATION FOR QUALITY CONTROL.
Cincinnati, Ohio: Public Health Service, 1965.

7. Thomas, William A. (ed.). INDICATORS OF ENVIRONMENTAL QUALITY.

New York: Plenum Press, 1972.
 361

PROBLEM
FORMULATION
DESIRED DATA
- L

V
DATA
COLLECTION
ADEQUATE DATA
c

I
t

ANALYSIS
1-SIMUUTION
2-PROGRAMING
V C

L
DESIGN
CRITERIA -
Figure 1 .
362
 363
L
zal
c
C
al
E
al
0)
O
C
3
Qo
2O
*
0"
07
E3
.o)-
UI
I
8-
.
c
O
.-
N:
LaJ
a u
c .-
O
L
C
ua;g'
I L I
 DESIGNING PROJECTS FOR THE DEVELOPMENT O F GROUND WATER
R E S O U R C E S I N T H E A L L U V I A L P L A I N S O F N O R T H E R N I N D I A ON
THE BASIS O F INADEQUATE DATA,

BY

B. K, S A B H E R W A L

ABSTRACT

U t i l i z a t i o n o f ground w a t e r p o t e n t i a l t o d e v e l o p i r r i g a t e d
a g r i c u l t u r e in the a l l u v i a l p l a i n s o f N o r t h e r n I n d i a t h r o u g h
"Push button" w a t e r w e l l s h a s p l a y e d a v i t a l r o l e t o b r i n g a b o u t
t h e G r e e n R e v o l u t i o n for m e e t i n g country's f o o d d e f i c i t . But t h e
p o s i t i v e d e v e l o p m e n t on t h e food f r o n t i s only a p h a s e . C o n t i n u i n g
p o p u l a t i o n g r o w t h and t h e r e s u l t a n t i n c r e a s e i n demand f o r f o o d ,
f i b r e and o t h e r s e r v i c e s o b t a i n i n g f r o m w a t e r u s e a r e a d d i n g t o t h e
water requirements thereby underlining the urgency to hasten
execution of projects capable o f delivering assured water supply to
m e e t the d e m a n d s o f h i g h y i e l d i n g v a r i e t i e s (-HYV] c r o p s , T h i s c a n
be a c h i e v e d by i n s t a l l i n g m o r e w a t e r w e l l s i n t h e a l l u v i a l p l a i n s
o f I n d i a r i c h i n ground w a t e r p o t e n t i a l . G r o u n d w a t e r r e s o u r c e
t h o u g h it g e t s r e p l e n i s h e d a n n u a l l y , i s not an i n e x h a u s t i b l e r e s o u r c e ,
E c o l o g i c a l r e s p o n s i b i l i t y m a k e s it i n c u m b e n t on t h e p l a n n e r s o f
ground w a t e r d e v e l o p m e n t p r o j e c t s t h a t t h i s p r e c i o u s r e s o u r c e , I S
not e x h a u s t e d due t o over e x p l o i t a t i o n , S u r f a c e w a t e r s a r e t a n g i b l e
a n d t h e i r p o t e n t i a l c a n be p r e d i c t e d u p t o r e a s o n a b l e c e r t a i n i t y on
t h e b a s i s o f long t e r m o b s e r v a t i o n s o f f l o w i n c h a n n e l s . A s s e s s m e n t
o f g r o u n d w a t e r p o t e n t i a l o n t h e o t h e r hand i s q u i t e c o m p l i c a t e d .
The d i f f i c u l t y a r i s e s o n a c c o u n t o f t h e f a c t that g r o u n d w a t e r
r e l a t e s t o t h a t i n v i s i b l e part o f h y d r o l o g i c cycle w h i c h o c c u r s
b e n e a t h t h e l a n d s u r f a c e . E v a l u a t i o n o f ground w a t e r r e s o u r c e t o a
high d e g r e e o f a c c u r a c y i s a m u l t i d i s c i p l i n e s t u d y i n v o l v i n g ,
c o l l e c t i o n , a n a l y s i s and s y n t h e s i s o f h y d r o l o g i c a l , g e o l o g i c a l ,
meteorological, geophysical, hydrochemical data, computing quantums
o f r e c h a r g e , d i s c h a r g e and b a l a n c e of g r o u n d w a t e r i n a b a s i n o r a
s u b - b a s i n and c o r r e l a t i n g t h e r e s u l t s w i t h t h e c h a n g e s i n g r o u n d
w a t e r l e v e l s and i t s r e g i m e , A c o m p r e h e n s i v e s t u d y o f t h i s t y p e i s
t i m e c o n s u m i n g a n d c o s t l y , I n view of t h e l a t e s t d e v e l o p m e n t s i n
g r o u n d w a t e r hydrology the a v a i l a b l e h y d r o l o g i c a l a n d g e o l o g i c a l
d a t a is not a d e q u a t e e n o u g h f o r a c o m p r e h e n s i v e and p r e c i s e
a s s e s s m e n t o f g r o u n d w a t e r p o t e n t i a l t h o u g h e x p l o i t a t i o n of g r o u n d
w a t e r i n I n d i a c o m m e n c e d q u i t e s o m e t i m e back. O n t h e o t h e r h a n d
p r e p a r a t i o n and e x e c u t i o n of p l a n s and s c h e m e s f o r t h e e x p l o i t a t i o n
o f g r o u n d w a t e r c a n n o t be held o v e r t i l l t h e c o m p l e t i o n of s u c h a
study which may take four to five years, ît has therefore become
necessary to adopt some reasonably accurate methodology t o evaluate
t h e g r o u n d w a t e r p o t e n t i a l w i t h t h e h e l p o f t h e a v a i l a b l e d a t a and
p l a n g r o u n d w a t e r e x p l o i t a t i o n p r o j e c t s o n i t s b a s i s t h o u g h at t h e
s a m e t i m e k e e p i n g m a r g i n for s u b s e q u e n t a d j u s t m e n t s w h e n b e t t e r
data becomes available, Appraisal techniques and adopted criteria
for an approximate evaluation of ground water balance i n water table
a q u i f e r s are d e s c r i b e d with p a r t i c u l a r r e f e r e n c e t o t h e B i s t D o a b
T r a c t o f the S t a t e o f P u n j a b - I n d i a w h i c h h a s a n a r e a o f 9000 sq.
k i l o m e t e r s and w h e r e 80% o f a n n u a l r a i n f a l l o c c u r s i n t h e m o n t h s o f
J u l y t o September. S i g n i f i c a n t p a r t o f t h e a s s e s s m e n t s t u d y i s t h e
r e c h a r g e t o g r o u n d w a t e r f r o m t h e a n n u a l f l o w o f a b o u t 1.25 M.A.F.
of; s u r f a c e water t h o r o u g h a net w o r k o f u n l i n e d a n d l i n e d i r r i g a t i o n
c a n a l s a n d i t s u l t i m a t e s p i l l a g e i n t h e c r o p p e d fields. O n t h e
 366

discharge side is the d r a w a l by a p p r o x i m a t e l y , 0.1 m i l l i o n e x i s t i n g

s h a l l o w and deep water water w e l l s w h i c h a r e e i t h e r electrically o r
d i e s e l driven. T h e e l e c t r i c a l l y driven w e l l s have unmetered e l e c t r i c
s u p p l y , the t a r r i f being o n t h e basis o f horse power o f t h e e l e c t r i c
motor. F o r b o t h t h e t u p e s o f water w e l l s l o g books r e c o r d i n g t h e
n u m b e r o f hours a t u b e w e l l o p e r a t e s are not being m a i n t a i n e d by t h e
private o w n e r s , This aspect further a d d s t o t h e problem o f working
out a c c u r a t e drawals from and return s e e p a g e t o ground water in
t u b e w e l l irrigated f i e l d s , I n t h e a b s e n c e o f a d e q u a t e data to
c o r r e c t l y evaluate ground water p o t e n t i a l a n d p r e s s i n g necessity to
exploit the p o t e n t i a l f o r food production s t a t i s t ì c a l or e m p i r i c a l
m e t h o d s have been adopted t o work out ground water balance and t h e n
apply a r e a s o n a b l e safety f a c t o r to t a k e c a r e o f short c o m i n g s in t h e
approach. In t h e project a r e a s water t a b l e f l u c t u a t i o n s are a l s o
being observed more f r e q u e n t l y t o c l o s e l y w a t c h t h e effect o f
additional draft,

RESUME

L'utilization du p o t e n t i e l des e a u x s o u t e r r a i n e s pour l e

développement de 1"agriculture i r r i g u é e s dans l e s p l a i n e s a l l u v i a l e s
de l'Inde du Nord a u m o y e n d e s p u i t s d'eau d u button-préssoir a
joué un r ô l e v i t a l p o u r a c c o m p l e r la "Revolution Verte'' afin de
s a t i s f a i r e l e s b e s o i n s d e f i c i t a i r e s d e s a l i m e n t s du pays. Mais l e
développement positif sur le front d e n o u r r i t u r e n'est qu'une phase.
La continuation de la c r o i s s a n c e de l a p o p u l a t i o n et l'augmentation
r e s u l t a n t e du besoin de nourriture, tissus, et d e s a u t r e s s e r v i c e s
utilizant l'eau n e c e s s i t e n t l e s besoins d e l'eau supplementaires,
ainsi soulignant l'urgence de l'ëxecution d e s p r o j e t s capables de
l'alimentation fourniture assuT+e d'eau p o u r s u b v e n i r la demande des
r é c o l t e de haute z-eqdement. O n peut s a t i s f a i r e c e t t e demande en
installant p l u s d e p u l t s d'eau d a n s l e s p l a i n e s a l l u v i a l e s d e l'Inde
du Nord, r i c h e s en p o t e n t i e l d e s e a u x souterraines. D e s r e s s o u r c e s
des e a u x souterraines quoiqu'elles s e r e m p l i s s e n t c h a q u e année, n'est
p a s u n e r e s s o u r c e ingpuissable. La r e s p o n s a b i l i t é écologique le r e n d
o b l i g a t o i r e a u x planificateurs des p r o j e t s d e s e a u x souterraines d e
voir q u e cette r e s s o u r c e prlcieuse n e s e p p u i s s e pas, en raison de
sur-éxplóitation. Des e a u x de s u r f a c e sont t a n g i b l e s et o n peut p r é d i r e
l e u r p o t e n t i e l jusqu'une c e r t i t u d e r a i s o n n a b l e s u r la base des o b s e r -
v a t i o n s à long t e r m e de l'écoulament d e s e a u x d a n s l e s canaux. L'esti-
m a t i o n du p o t e n t i e l d e s e a u x s o u t e r r a i n e s p a r c o n t r e est bien c o m p l i -
quee. La difficulté s'8le've en r a i s o n d u fait q u e l'eau souterraine
s e r a p p o r t e à cette p a r t i e invisible d u c y c l e h y d r o l o g i q u e qui s e
f a i t au-dessous d e la s u r f a c e de la terre. La n a t u r e héterogene d e s
f o r m a t i o n s g é o l o g i q u e s à t r a v e r s l e s q u e l l e s l'eau s o u t e r r a s n e c i r c u l e
r a j o u t e à la c o m p l z x i t ê du problame. Y a l o T i s a t i o n des r e s s o u r c e s d e s
e a u x souterraines a u n e haute d e g r 6 d'exactitude est u n e étude de d i s -
c i p l i n e s m u l t i p l e s comprenant recuîl, a n a l y s e et s y n t h e s e des d o n n é e s
hydrologiques, géologiques, m é t e o r o l o g i q u e s , g é o p h y s i q u e s et hydro-
-chemiques, calculant l e s q u a n t a de r é c h a r g e , d@scñarge et l e bi'lan
l'eau s o u t e r r a i n e d a n s un bassin o u sous-bassin et m e t t a n t en c o r r é l a -
tion l e s r é s u l t a t s avec des c h a n g e m e n t s d a n s l e s n i v e a u x d'eau s o u t e -
r r a i n e et s o n régime. Une é t u d e d e t a i l l é e de c e t t e t y p e d e m a n d e s p l u s
de t e m p s et est coûteause. En v u e des p l u s d e r n i e r s d ê v e l o p p m e n t s d a n s
l'hydrologie de l'eau s o u t e r r a i n e la d o n n é e s h y d r o l o g i q u e s et g é o l o g i -
q u e s d i s p o n i b l e s n e sont p a s a s s e z pour u n e e s t i m a t i o n c o m p l J t e et
 367

exacte du p o t e n t i e l d e s e a u x s o u t e r r a i n e s , bien que l'exploitation

des e a u x souterraines commence i l y a quelque t e m p s d a n s l e passé.
D'un a u t r e cote, la p r é p a r a t i o n et l'dxecution d e s p l a n s o u s c h e m e s
pour l'exploitation des eaux s o u t e r r a i n e s n e peut pas &tre a r r ê t é e s
jucqu'a la complétion d'une t e l l e é t u d e q u ì p u i s s e p r e n d r e , 4 o u 5
ans.Donc, i1 est devenu n é c e s s a i r e d'adopter une m é t h o d o l o g i e
r a i s o n n a b l e exacte p o u r e s t i m e r le p o t e n t i e l des e a u x s o u t e r r a i n e s
avec l'aide des donnêes dìsponsible et p l a n i f i e r des p r o j e t s
d'exploitation d e s eaux souterraines, a u même t e m p s en r e t e n a n t u n e
marge pour l e s m o d i f i c a t ì o n s s u b s é q u e n t e s quand p + u s d e d o n n é e s
seront disponlbles. L e s t e c h n ì q u e s d'estìmatïon, p o u r u n e v a l o r i s a -
tion approximative de balance d'eau s o u t e r r a i n e decrite a v e c une
r é f e r e n c e particulière à BIST D O A B t r a c t e Etat de P u n j a b en I n d e q u i
a un t e r r a i n de 9000 kilometres-carres et o u 80% de pluie a n n u e l l e
arrive a u x mois de Juillet.Septembre. L a p a r t i e s i g n ì f i c a t i v e d'6tude
estimative concerne la r é c h a r g e 2 l'eau s o u t e r r a i n e de l'écoulement
a n n u a l d'environ 1.25 M.A.F. (million a c r e p i e d s ) d'eau de s u r f a c e
par un r é s e a u de c a n a u x d'irrigation a l i g n e s et n o n alignes, et son
utilization ultîme dans les c h a m p s c u l t i v é s , A c b t k de d é c h a r g e m e n t
1.0 million des existants p u i t s d'eaux q u i s o n t o p é r é e s s o i t p a r
electricit6 soit par essence. Des puits m e c h a n i z e s par e l e c t r i c i t 6
assurent une alimentation d'eau s a n s c o m p t e u r d'electrictricite, le
t a r r i f étant basé s u r le C.V. d e s m o t e u r s e l é c t r l q u e s , Pour l e s d e u x
t y p e s de p u i t s d'eaux, des c a r n e t s à r é g l e c o n c e r n a n t le n o m b r e des
heures qu'un puit S opére, n e sont pas t e n u s par l e s p r o p r i é t a i r e s
prives. Cet aspect ajoute encore a u problème de c a l c u l s d e s p u i s e -
m e n t s exacts de l'eau s o u t e r r a i n e dans l e s c h a m p s i r r i g u é s a u m o y e n
des puits à moteurs é l e c t r i q u e s , Dans l'absence de d o n n é e s de v a l o -
r i s e r c o r r e c t e m e n t le p o t e n t i e l d'eau s o u t e r r a i n e & l a n é c e s s i t é
pressante d'exploiter le p o t e n t i e l pour l a p r o d u c t i o n d e n o u r r i t u r e
l e s m e t h o d e s empiriques et d e s t a t i s t i q u e s ont Btd a d o p t é e s p o u r
r e t r o u v e r la balance d'eau-souterraine et d'appliquer u n f a c t e u r
r a i s o n a b l e de sÛréte de bein r e n d r e c o m p t e d e s f a u t e s d a n s l a m a i n è r e
d'aborder, Dans l e s r e g i o n s s o u s o b s e r v a t i o n o n é t u d i e a u s s i t r e s
souvent le n i v e a u de variabilité d'eau p o u r r e m a r q u e r de p r e s
l'effect d'eau puisée en s u p p l e m e n t ,
 368

1. UTROWCTICN
1.1 India i6 the seventh largeet country in tbe
World. ït's area is 328 million hectarest 3-28 million bqtlue
iiiïorn9ters) with a population of 547 million (1971)
Agricultural out put accounts for half of thr country's Gross
23: tioneï droductí GPW.
1.2 In the year 1947 wheu the country was divided,
the major irL-i;jation syskais únd %cod piav~ucLi? ?reas were
lost to rtkistm resuïtiag ILI a deficit of 4 dillions t m m s
of food grains. India had t b r e f c m to iwort ,ILL t;.c's.us
from the major wheat producing countries of the world till
the advent of Green devolution recently brought about by
the'incrrased utilization of country(s surface water resources
for irriyqtion from 93745 priïlion ~1 m (76 million acre ft.)
in IL361 to 222000 milîion CU m (186 miliion acYe a.) at
Grosent ?nd tof ground water lli000 Pilllion CU m (I30 million
ecre ft.) üse í~fhigh yielding variety (W) seePS of
cercnls h k e wheat ,rice,wiize,Jawar and B j r a hes Ris0
hastened to a great extent the tremendous increase III
'food cut put. Cevelopmgnt Of rtwarf varieties of wheat made
Possible following the introduction of valuable genetic
material from bxtco 141 1962 has alone increased the production
of t h b important cereal from neerly 12 to 23 million tonnes
within a period of about five yeam.
1.3 Eilt the maximm production per unit of any
Particular variety of m d seed is the result of a set of
cultivation practices proper doses o9 ioputs prophylactic
and curative measures to check the atta& OP insects, pests
and disewes end above all adequate irrigation at proper time.
2. INDIA-PH!EICAL AND OTHE=TI FJ3AlWRES.
2.1 main land can be
Physio raphially I n d k a i s
divided into six divisfons comprishg of i-
i) the Himslayan mountains
ii) the indo-Gangetio Plains
iii) the Central Hiagi Unda
the D e c m Plateau
the Eastern Coastal Belt
vi) the Western Coastal Belt
2.2 The H i m l a moimtains are of comparatively
recent origin. The Deccan E t e a u end the CentraR Hi@ Lande
are composed of ancient rocks. The Plains are hilt up
of layers of sends, clays of m o l o loally very recent &te.
The metern anditestem Coastal befts comprise of deltaic
a n d sedimentary marine deposits.
2.3 About 7of of the country's a m a is under lain by
hard rock with a thin soil cotrer at top derived f ra l%o
wealtherinn of rocks. IO 1i, mainly the Indo-Gan etic Phkis and
the two deltaic Eastern and Westem Coastal d t s which are
made up of alluvial solls and sedfmentwy deposits varying in
thickness from a few hundred feet ln the coastal belts to
thousands of feet in the Plains.
2.4 Ailuviai soils are suitable for agricultum
and respond well to artificial irrigation. Being generally
permeable in character and having laysrs of coarser deposits
also provide under ground storage for seepage water. NO
wonder the Indo-GangetLe Plains, 8nd the tu0 coestal belts
though accomt for on1 l./3 of the oauitry'e laad rnam ?ut
suwort atmut of tL caintrps pornlation.
2.6 tu
The m a o r snow fed riveris of country naaiely
the triaitaries of the U&s, the cianges and the Bwhaii Putra
flow rlugyishly through the indo-Gangetic Blain. The main rivers
 369

flowing to the coastal belts are the X a r k d a and the Tapti on

the western side C m d the Maha Nadi, the Godavari, the Krishana
rind the Cavery on the eastern sir%. All these rivers outfall
into sea. The rfvers also provide irririation supklies to the
vast net work of m a l systems part of which was constructed
about a century back. host of the old canals are designed as
Il m n of the riveril schemes and are unlined. The unlined cana.ls
act es additicnal souY'ce of recharge to ground water besides
seepage from rivers, streams and rainfall.
2.6 In dia s clima.te ranges from con t in en ta.1 to
oceanic, from extrems of heat to extrerns o f cold, Prom high
a.ridity a.nd negligible rainfall to excessive humidity and
torrential rainfall. Sauth destemi monsoons in summer accounts
for m r e than 85% of the precipitation a n d that too in a
short span of about 4 months. The great diversity in weather
conditions and uncertainity of rainfall results in the prevalence
of draught condition in about one third of the country.
3. G R O W D :,! AT-g2 1
7s
3.1 In the face of variability and tirircliability of
rsinfall and also lack of' adequate storage support for some Of
the major canal irrigation schemes, tappin:? of zround water
resource through iiells and tubewells for intensive agriciiltu re
has pla ed a yitzl role in ushering the (Green Revolution
Y
particu a.rly in these parts of the country where low or badly
distributed rainfall is quickly lost thrm$l evaporation Ixit
where g-ound watnr potential is available stored in alluvial
deposits
3. 3 Cultive.tion of high yielding varieties and
intmsive cropping dernad water at the right time and of the
r e w i r e d quantity. These pre-requisits have made the
cultivators in areas with copious ground watcr supplies take
to the instcllatïon of their own', push aittontt irrigation
systems. The water scarcity during the yesr 1365-67 which
created draught conditions almost all over the country acted
as catalyist to boost up exploitation of ground water poteiitfal
t h m u & diesel or electrically operated tubwells 100 feet
to 200fbet dealfor the protection of Crops.
3.3 The Govemrcent also rose to the occassion and
undertook to provide large saale loan finance to the cultivstors.
for the installation of tubewells on their farms in areas where
the ground watcr potentialities were promising. The result is
that at present an investment of about Rs.2000 crores hac Filreacbr
bean Lade in the field of ground water exploitation in the
country. Most of this investment has taken place in private
sector.
3.4 The following table, indicates the progress Of
insxellation of tubwells in the a3untry:-
(In thousands I

Mo.of private tubewells 3

- 250 1965
100
-
A969
279
1971
-.-
470
(anticipated)

rio.of diesel pumps 66 471 837 1150

No.of electric pump sets 19 513 1080 1620
Total 88 1084 2196 3240
370

3.5 The spectacular development of ground water

utilization in the country has been influenced by a npmber of
factors namely the rzcognition by farmers of the importmt role
played by ground water in sustaining modern agricultural
techniques, incrceaed availability of institutional credit for
financing the ground water exploitation programme, rapid
electrification Of rural areas, local availebility of technical
h o w how to drill well8 with machines, a n d indiyenously
manufactured pumps, motors and other equipmat for the
construction of wells and above all large scale village road
.
deve lopment p rogr amme
3.6 Heavy investments in gound water exploitati.
schemes and the involvement of the Government back institufional
credit far the purpose has made it incunibent to plan a n d execute
this programme of utmost natio'lal imoortance duly sumorted by
proper assessment of ground water potential.
4. HYDROLOCEIC CYCB.
4.1 All the waters in existance e n be located by
what is ïmìwn as 1) hydrologic cyclen or 11 Nater Cycle". This
C cle involves total earth system comprising of the atmosphere,
d e hydro-sphere Snd the lithosphere. The activities of the
n ilater Cycle" are vast extending from an average depth of about
half a mile in the lithosphere to abcut 10 miles in the
a tmsphere .
4.2 Hydrologic cycle is greatly influenced by the
geologic history of a particular area. If the geology consists
of alluvial fmnatlons, water w i l l occur in the openings
between granular XcFosits; EUT; if the area f O r ~ t i O n Sare rocky,
the ground water Will be found in decomposed parts of ro&B,
freotures or in tabular openings in soluable rocks or opening
in lava formed by flow or gas expansion during solidification.
Guide lines to evaluate ground water potential in alluvial
formatias have only been discussed in this p3Per.
4.3 Ground water originates from surface water and
gets renewed or recharged with the down vard percolation Of
precipitation, flow in stream, canale, return flow f r a
irri,ated fields etc. Propm assessment of this valuable
resource f o m d in Permeable geQlOgac formations a n d in motim
through the voids or pore spaces in an area requires working
out its total storage and quantities whir& are annually pumped
out or replenished into the ground water reservoir. bality Of
grOUnd water .leo requires to be known. Comprehensive studies
a n d explorati-ns -re necessary to evaluate the potential to a
hfgh degree of accuracy.
5. APPRbACH TO W O R K OUT G R W d D WATER BULLANCE;
ON TI% &.SIS OF INADECrJATE DATA.
5.1 hthodology for the precise eva2natioB of ground
vater potential is quite complicated. The difficulty arises 691
account of the fact that ground water relates to the.t invisible
 371

pert of hydrologic cycle which occurs 'beneath the land surface.

9etero~~eneoUs nature of the ,(.f?ologicalformat ions through which
ground water moves e-dds to the coarplexity of the problem.
5.2 It ha.s been observed by pump tests that in Punjab
which is the Northern-Western part of the Indo-Gangetic Plain
alluvial materials constitute an extensive hetrogeneous and
.
a.nisotropic unconfined aquifers Discharge from tubwells as
deep a$ 300' results in- draw-down of water tsóle over larye
prea. ?nd is sustained by dawrltering o f surface watr,r recharne,
such condit iow jrevail through out the top aauffers ~f slluaiurn.
5.3 Planning and designing of ground vmte? development
through small and medium sized tubewells (1.10 to 200 feet
deep) in the ground water &sins and sub-basins o0 the indo-
Gmjetic plain cím therefore be ,compared to re:.err)ir prObLem.
'

This approach cal-1s for drawing iipm the fresh water table
a,quifem upto the Safe Yield which should not. exceed ^che long
term mean -annual supply o r recherge involving wet and dry years.
In view of lack of complete data the genersl. i"om CJf bhe squrtie.cn
of hydrologic equilibrium in thcs project areas has been simplified
cmdsuitably adjusted to arrive at worh.ble ,g:rourid;iater hiance.
In areas having ground water quality problem Safe Yield cannot
be equated to mean annual recharger
5.4 Installation af tubewci-11s upto 300 feet for
irriz8tion is being practised in India since 19.34-36. In
edditìon, tubwells vere a.lso install.ed Por municlpal,rai%ays
Tnd indust,ia1 use. Geological Survey of India, state Zrri%ticn
&partmats and Central Gróund -.:a.terBoard have bom iminte.in1ng
oeological, hydrological, geochemical and other ground water
data of a rudimentary character. Irrigation Departnlsnts have
also meintahed record o f water table fluctuations keduced to
mean sea level ( 1%.Lr) from a net work of observa.tion wells.
kmccipitatScn,racord is kept by Indian bieteorological Department.
Ifit no are?wlse systematic investi.zations and exploration to
a-ssess ground water potential were conàucted. In the absence
of adequate ùata to evaluate ground water potlontkal on the basis
of lat$st deVelOPI~ent6in g r m d water hydrology and pressing
necessity to exploit ground water potential statistica.1,
analytical and empiriel wtbods were resorted to arrive at
preliminary quantitative evaluatîon of ground water balances in
the pro,ject areas
5.5 Ground water balance te Plan schemes was
computed on the collp,ction and malysis of the following basic
data in project areast-
1. Village -wise iocatim5 and other details of existing
tu heiaelle
2. Colleetion of reliable litholo- of tubewer-ils.
3. Iso-pstch featums E@ revealed by litho-lons and
g e o l o g i d correlation of strata upto the available
depths to broadly understand t.he geometry of aquifers.
372

4. ,Sample observatiais of pumping rates o? existing tubewells

5. dimple surveys to assess pumping hours for surn'lier and
.
by u s h g simple devices(0rifice or ri,qht Angled '6-notch)
winter crops to work out present ground water draft
in the Gi'oject .-;.rea.
6. Locatim of raingaue stations m d annual ra.infel1 datn.
for the last 20 to 30 years. '

7. ::eighted mean average annual rainfall fill;ures for

different blocks of the area by Theisson method.
8. Locations of existing pugln?JdischPrp sites on stresms,
drains and CO! lectfon of run off data monthwise.
il. Available ground water qiality data to demrcate fresh
and inferior ground water zones.
10. Location of existing c m e l irriyation s,Ftem m d data
about their len$.hs ,. sectims,( lined/unlined) d e s i l e d
dischzcrges, actual flow time and areawise.
11 0 Yeriod-wise flow in enals Rt the point OP entry into
and exit from the project area.
12. Locations of existkg water table observation wells
and past data of tiatcr table fluctustbas for pre
and post monsoon periods
13. hater table depth data to delineate high water table
14
.
areas
Cropping pet tern, cropkning cslander a n d water
requirements for summer a n d winter crops.
.
15 :,orkingout zvera- value of specific yield of the
formations, either by pump tests or empirically.
6. TJXH1vICk.L CRITl3RI.A FOF GROTD ?dUKE PLZ./L".JJCI:
C0PîJT;RTIDN.
Technical criteria adopted to work out ground
water balance is given as under:-
6.1 Rechars oroni R&infall
Unconfined %quifers get recharwd from local
rainfall. Based on the sQiilles ccnducted in the Ganges &sin
for period 1937-78 to 19Sû-51 a relationship vas evolved to
WO& out net p e n e t m t i m of rain water to water teble in
alluvial areas
2/5
Rp t 2*O(R-15)'
'Ihere R average annual rainfall Zn inches.
=
Rp annial rainfall penetration to water
table in inches.
ïhis relationship applies to areas having
annual rainfall in excess of 15".
6.2 beemge from Canals
seepage loss values from unlined canals based on
 373

experimental data are given below. These figures are inclusive

of evaporation losses which form only a small proportion:-
i) In the :tete of Uttar Brzdech-ïndia it is about 8 Cs.
(GU ft/:iec.jfor ordínary clay loam to about 16 cs.for
sandy loam Per million 5;q.feet of wcttad parameter.
.
i,verage k i n g 10 CS ./k.qt
li> In the Wz1iarashtr.a State-India these sire calculated at
15 Cs./M sat. for discharse upto 250 CS. anil lO.&s!./M%ft
for hiisher discliarzes.
iïi) In FunJab and Haryana state -India =lue, of ..eepa<e loss
in Cs./ìvI e f t p is zxpressed by the formla :-
(a) K8 4x4 -.O625 for unlined chmnels
d .O56
KE ix; for lined channels.
where K is the se.epa.gein cusecs pfx million square Feet of
wetter area anci Q is. the discharge in channels in cusecs
(CU ft,/riec.j
6.3 Recharge from Mater CouI’ses.
,Accordingto the findinqs in the Crop Planning
Committee Report of 1954 for Phakra canal meas in ïuiijab, out
of the total quantity of water that enters a canal at head,
nearly 45$ is lost during transit In canal, diatriaitaries a d
water courses. This means that for every 109 cft of water
received in the field 182 cft of water is releaspd at the c m 3 1
hend. Seepage loss in water courses also is tczken E S 20% of the
flow.
6.4 Retuin s e e m g e from irrigated fields.
PPrt of irrigation water spread on field
percolates below the root zone of crops m d is added to the
.ground water body. !:;hat percento9 of the water spread for
irrigation crops percolatcs riown depenas on character of soal,
types of crcp 2nd quant.ity of watering and its frequency at,?<
AccOr6iang to Crop planning Committee 1iepqJ39~
Rhakra C2na.l .&rea referred to above out of 100 cs water ;,te-;,,.
reaching the field 70 CS. is utilized by the crcp plant” & $
i
;.,

the remaining 30 CS. goes under mound.

6.5 Recharge from Hilly Stream and
S t o m water Drains.
It is difficult to exact ìy analyse.!
caused because neither the c:uantum of flashy disch9?ges no.? the
sections of the hilly Streams(Choes) a
parameter has been evûlved on the basi
Last b i n s stream in Bist Doab T r m t o
approximation to the seapage precipita
through the numerous choe beds is taken asss?
6.6 s u b SUPface Flow
It has been observed th&-
 374

alopes of the $round water table in the $lains the quantity of

subterrainem flow g ~ t t h g$.h and out of project are- is
negligible 2s conipared to vertical recharge from rainfall,
c m a l P B ~ P U ~ L return
, flow from irrigated fields etc. Hen-
this item has been omitted from computatiuns on both sides
of hydraulic equation amialy due to'the non-availability of
adequate data.
6r? Ground wqtrr loss due to non bene3tçkl
etrapo-tranepiration in water logged areas.
EvagotreEspiration losses ars related to depth
of water table froin th@ ground surface 81.14vegetatkq cover.
Airing July to October period when recharge to grow-d water
reservoir is maximum Qe to rainfall and high supplies in rivers
and canals etc. water table rises towards ground surface. In
the riverain md water Jogged tracts, the depth of Found water
veries between zero to 6 feet below lanu surface. in 1963,
'J.S. B.RI conducted experiments/ obervctiuns for salvaging
ground water t6i?g evawqsteü from ground LI' transpired non
beqeficially by vegetation In the central part of san-his Valley,
Ceil'tral Coloredo (U.S+) t Graphs were plotted, correlating E.T.
loss to ground water depth E.T.loss th negligible if water table
is lowered to 12.5k
Persistance of higher water table 5r water
logged amas indicates that recharge to ground waster is equivalent
to the E.T.loss or may be $ven morg. Pending detailed studies,
it would be reasonably p o d planning to draft grouad water
within the limits of water est9rnated to be lost through evapo-
transpiration.
608 Draft froor the e x h t h g state (deep) and
Arivate (shallow )tubewells.
The State tubewells (1.5 to 2r0 CS. cawcity)
ere planaad to operate at 22 hours a day for 240 days in a y6arr
H e n œ d m a l draft from state Tubewells varies f r m 660 acre ft.
to 880 acre ft. Shallow tubeiwslls (0.2 to 0.8 cs.cepacity) for
a b u t 800 to i000 hours per year. Draft f r a these wells cm b
taken as 15 acre ft, to acre e,per well per year. por wells
driven by animal power dralpt Is taken W 5 acre ft. while for
drinking sugply wells in villages the draft 1 acre ft. has been
adopted per mimam.
6s9 SpaqIpg q$ s h a l l o ~tubewells, '
Closdl spaaed tubewells m u s e mutual hydlaulic
interfpence due to ovQr f a p p h g 09 their ocrnes of depression.
In thi&y populated &we&9 of the IndolGangetic Plph@ Qr
a b r y Basin ? m m ho&&hgs are very small < abcut 10 to 15 acres
~

.
per head ) The t u W e l l s aro of 0.2 to 0.3 cubic feet/ $eco
dfaeharge a n d 100'-190' deep. These wells do not run more than
10 to 15% of t h e time in a year. After tests results minimm
epacinp of such wells has been kept at a b u t 600 feet 6
 375

6.10 Grotmd WatEr Ealmce

Computing the item of ,annual recharge and
discharge as por criteria discussed above if a balance is
struck 5 first estimation of the balance of T2cbyze potcntntlal
in a Project crea beccmes laown for planning further explcit-ticq.
A reasotinble factor of safety can be adopted to plen exploitetim
of the comnuted ground water ImiBncs which tcakes care of the
saps in the available data or the appraisal approech. The fa.ctor
will depend on area conditicns.
7. GHOTjND b~dkiiBLLA.?JiCGIN B145TDCAB 'I'R.'ICT
7.1 Bict Doab is a triangu1a.r part of the Punjab
StF.te (India )enclosed the rivers Sutlej and the Beas on
9
t?:c ciäes and a i v a l u hi1 s (lover iïj.ml8yan ranyes) on thc
third side. Three districts of the state ncmely Jullundur ,
doshiarmr and Kapurthala are located in th;c tract .
7.2 The area of the trect is 9900 Fq. Kilometers
mostly mugrising of alluvial plpin except the 8 miles wide
belt of Shivalik Hills on the XGrth-eestorri ,?id&. depth of
alluvium in the plain as revealed by seismic surveys is thousends
feet.
7.3 Gmeral water table is a b u t 20 to
%c feet in the
plains. Avcrzge ?.nnual rainfall in hilly region 1s 1200 mms
tJhils in the plains it varies between 914 mm tc 635 mm. 80% of
the rainfpll occurs during the monsoon period.
7 04 The soils are fertile and the ??SB 112s Copious
gromnd water supplies. There are abcut 0.2 million irriqatiw
tubwells clnà dugwells in the area. Quality OP g r m n d water is
qood for cultivation.
7.5 The tract is also irri,gated tl-iraigh Bist m a b
eanal which draws its supplies fYom the barrage on tho river
L u t k j at Rupar. A b u t 1.25 M.A.F. of water is used annually
for cultiva.i;ion. The net work of canal system Measures 7,54.28
Lilcirieters out of which 34.5i) K m . is lined. m c t h c r feature
of tile; area is ;iutiierous hilly utrea=( clioec) which descend
from the :,hivz~likhills and fiord only &i.ing monsoon period
w ith flashy uischw .;es
7.6 The recharge and discharge computntions ta
work out the ground Kater bzlance in the m e a dis+;rictwiseare
talxilated 88 per stateuats i & 11. .A safety factcr 0.60 has
been cùopted to the computed figures to arrive a.t the exploitable
grourd water stential. The water ttible f1uctuatir:ns end the
rainfall hi the area 8i-e being closely observed to watch the
strezses m d strains covered by ground water exb>lcitat*on on
the shallow unconfined aquifers under water table conCAtions
8. ca4crusCáu
8.1, Demographic trends indicete that the Lndia's
populatun is likely to inn,ease to 700 millions by the end of
the present dea&. On the k c i s of the piiojecte8growth rate
 376

c f 1.45 per tbausand per anliuam during 1981-85 the population

w m l d rise to 300.millions in the year ZOO0 A.D. i.e. about
S5$ increase over the 1371 population. Keeping in view the
expected improvement in the standard of living of the people
during the intervening period, the food and fibre requirements
will increase by a b w t 100% of 1371 production. Suck an
enormous increase in the production is possible through intensive
agriculture and bringing additional areas under irrigation by
o p t i m a utilization of the water resources (Surface and qround )
8.2 Tnis will eventually result in intensified drawels
of ground waters from shallow as well as deep aquifers. Ground
water resource though it gets replenished annually, is not an
inexhaustible resource. Ecological respansibility m k e s it
incumbent on the planners of ground water development projects
that this precious resource: is not exhausted due to over
exploitatbn arid is so utilized that it also remakis ava.ilable
ïor the yencrstion to come. Therefore aremise potential of
ground wa.ter anà its safe yield both from shallower and deep
aquifers neads to te assessed as accurately as possible to
prepa1.e realistic exploitat ion p l m s and schemes. This aspect
has ben duly recognized and separate state level orgpnizations
cmpï-ising of hydrologists, hydrometeorologist, geologist,
s
agronomist, geoph scist and drilling engineers have been set up
to carry out deta led investigationa. These detailed studies
will however take time. However to maIntaln the continuity of
N grow more food 1) compaign exploitation of ground water
recharge &lance may be planned on the %sis of Safe Yield worked
out with the approximations and applicatmn of safety factors
suited to each project area.

mmmv CES
1. Report of the Irrigation Commissian, 1972, Volume-I,
Ministry of Irrigation and irower, New Delhi.
2. Krishnm, M.S.,m Geology of India and E m m a
3. T o l m , C.J.,l) GroundWater " .
4, Ehattacharya, R.P. 1) Ground Water supplice, depletion of
water table a n d penetration of rain water to ground water
table in Western Uttar Pradesh (India )IIo
5. U.S.G.S. Water-supply paper, 1608-G Anplycis of Aquîfcr
Tests in the Punjab Region of West Pakistan
6. tJ.s.';.S. :uater supkly papc'r, 1608-G '1 Ground Water
Hydrciogy of the Punjab, West Pakistan 1~1th'=mphasiS
of ?rcblems caused by Canal Irriggtion II .
 377

7.

8.
9.

10 o

11b"

12

33.
14

15o.
37 8
STATEMENT I

ITEMS
-
-
ISWlhRpI

1.410

.7ss
b.30

-
-
D.310

0.oy
0.27

0.06

&IO<

o .ai

ODs'

o .41
o .Ia!
 379

A DRAFT

1
41
II 9

6SC

7 S2

96

2s

. Il

CROSS DRAF T(1+2+: 0.309

o.132
380

S E\SM\CL \ NE’S
R E F LE CTION h
REFRACTI ON 8-
TEST WELL LOCATION 8
CONTOUR INTERVAL 0 2 K M
DATUM M.sL
hLLUV\kL P U \ H 5 O
?LRYLbRY (SM\\uhL\KS)
 381

M A P 15
 I M P R O V E D T E C H N I Q U E S FOR WATER R E S O U R C E S Y S T E M S DESIGN

J R SEXTON
D G JAMIESON

WATER R E S O U R C E S BOARD, R E A D I N G , E N G L A N D

ABSTRACT

F l o w d a t a inadequacy can t a k e d i f f e r e n t forms. One e x t r e m e is

t h e complete l a c k o f a n y i n f o r m a t i o n but t h e m o r e u s u a l c a s e is
insufficient l e n g t h o f r e c o r d since v e r y l o n g s e q u e n c e s o f f l o w
data are r e q u i r e d t o evaluate t h e y i e l d and r e l i a b i l i t y of w a t e r -
- r e s o u r c e s y s t e m s w i t h c o n f i d e n c e . Using t r a d i t i o n a l c o n c e p t s o f
failure and r e l i a b i l i t y , all w a t e r - r e s o u r c e s y s t e m s are being
designed on inadequate data with o n l y t h e d e g r e e o f i n a d e q u a y
varying between s c h e m e s , T h e use o f s i m u l a t i o n a s a d e s i g n t e c h n i q u e
has necessitated a more r i g o r o u s d e f i n i t i o n o f r e l i a b i l i t y w h i c h
a c c e p t s the lack o f data yet m a i n t a i n s a m e a n s o f c o m p a r i n g t h e
r e l i a b i l i t y of d i f f e r e n t s c h e m e s both in t e r m s o f frequency and
m a g n i t u d e o f f a i l u r e , A new d e f i n i t i o n o f r e s e r v o i r r e l i a b i l i t y
h a s been used for t h e h y d r o l o g i c a l d e s i g n o f t h e Wash Estuary
Storage, a proposed series o f p u m p e d - s t o r a g e r e s e r v o i r s in s o u t h - e a s t
England.

RESUMEN

La i n s u f i c i e n c i a de d a t o s de f l u j o p u e d e t o m a r f o r m a s d i s t i n -
t a s . Ocurre e l caso e x t r e m o de l a f a l t a t o t a l de i n f o r m a c i ó n , p e r o
lo m á s usual e s l a duración i n s u f i c i e n t e de r e g i s t r o p u e s t o q u e s e
n e c e s i t a n c a n t i d a d e s i n o r d e n a d a s de d a t o s de f l u j o p a r a q u e s e e v a -
I Ú e n c o n f i a n z a la e f i c a c i a de s i s t e m a s de r e c u r s o s h i d r á u l i c o s . Em-
p l e a n d o c o n c e p t o s t r a d i c i o n a l e s d e l fracaso y de l a e f i c a c i a , t o d a s
l a s i n s t a l a c i o n e s de r e c u r s o s de agua s e h a n c o n c e b i d o c o n d a t o s d e
flujo inadecuados, c o n g r a d o de i n s u f i c i e n c i a c o m o s o l a v a r i a c i ó n
e n t r e ellas. El uso de s i m u l a c i b n c o m o m o d o de d i s e ñ a r s i s t e m a s c o m -
p l e j o s de r e c u r s o s de a g u a exige d e f i n i c i ó n m á s r i g u r o s o de e f i c a c i a
q u e m i e n t r a s a c e p t a la falta de d a t o s de flujo m a n t i e n e s i n e m b a r g o
u n medio de c o m p a r a r la e f i c a c i a de un p r o y e q t o c o n otro y e n t é r m i -
nos de s u f r e c u e n c i a d e ella y en g r a d o de s u fracaso. Un c o n c e p t o
de esos -la f r e c u e n c i a de p o T c e n t a j e c u m u l a t i v o - s e ha e m p l e a d o en
e l disefio h i d r o l ó g i c o Ifel d e p ó s i t o d e l e s t u a r i o d e l Washff, s e r i e de
depÖs*itos de r e s e r v a a bomba en e l s u d e s t e de I n g l a t e r r a ,
384

INTRODUCTION
The analysis and study of water resource systems can be conven-
iently subdivided into three stages, planning, design and operational.
Each stage has its own specific flow data requirements and what maJr be
adequate for one stage could well be inadequate for another. At the
planning stage, a large number of possible combin&t&ons of sources are
evaluated but not in detail: the requirement for hydrological data is
minbal, since the yields of individual sources need only be determined
approximately. The most promising combinations of sources are sub-
sequently examined in considerably more detail at the design stage.
This stage is concerned with aspects such as frequency, probability and
reliability all of which make considerable demands in terms of data
quantity and quality. The requirement is for long period of flow
records which may have a time increment of a day or more. In the oper-
ational staze, the data requirement emphasis changes from long-term
flow records to shorter but more detailed flow records perhaps even on
an hourly basis.
This paper is concerned with the relationship between the assess-
ment of reliability, the definition of failure and flow data inadequacy
at the design stage. Flow data can be inadequate in many ways: it may
be that there is no data or Rot enough data, or the wrong data has been
collected. Data can be of inadequate quality or have too coarse a time
increment between successive values. To sunmiarise, inadequate data is
an occupational hazard to all those involved in the hydrological design
of water-resource systems. However, with traditional concepts of
reliability and what constitutes a failure, the problem of flow data
inadequacy will remain for a very long time.

In the planning of water resources for England and Wales, many

diverse types of sources such as pumped-storage reservoirs, multi-
purpose reservoirs, rivers, aquifers and estuarial storage are being
considered. &ch proposed source is n? longer considered in isolation
hut as part of a much larger water-resource system. In these circum-
stances the individual yield of the proposed source loses importance
since it is the yield of the system as a whole that requires evalua-
tion. The increase in the scale of the problem caused by consideration
of a water-resource system as a whole has outdated many of the tradi-
tional techniques for analysing the performance of a resernoir: some of
the implicit assumptions have been made invalie by the complexity of .
modern water-resource systems, other assumptions have never been valid.

m H O D OF ANALYSIS
owing to the complexity of the water-resource systems currently
envisaged and the lack of theoretical techniques cspable of analysing
such systems, simulation is considered to be the only viable method of
analysis. A simulation model of a proposed water-resource system can be
constructed by joining appropriate component models of particular types
 385

of reservoirs iri an ofder corresponding to the physical system.

Examples are given in Fi$ures 1 and 2 of component models for a pumped-
s t o r q e reservoir and a pumped aquifer. It should be appreciated that
not all the links indicated in these models need be included since in
the specific application some can be set to zero.

The simulation is structured in a general form with physical con-

straints such as the capacity of the reservoir, maximum pumping capacity,
minimum residual flows in rivers etc treated as input vaxiables. The
model can then be used to find the frequency with which the system fails
to meet the specified demands and the sensitivity to changes in any of
these or other input variables in terms of frequency of failing to meet
specified iieiads. The relative importarice of each data ita! xc thus
be determined and the effect of data inadequxy can be qusntirird i?i
terms of confidence limits on the resulting reliable yield. LÅoFeovcr,
since the w a ~in which a water-resource system is managed will dfect
the reliability of the system, different operating rules can be compared
and evaluated.

Since the design of the system is concerned with rare events,

large amounts of historic or synthetic flow data have to be routed
through the models. Consequently the component models have to be rela-
tively simple to keep camputing costs down and therefore they are essen-
tially accounting procedures with lags and attenuation built in.

Given adequate data it is possible to include both conservative and

degradable water quality parameters in the model. The build up of
pollutants in various parts of the system can be monitored in the seme
w a ~as the quantity of water and the performance of the system can be
depicted as histograms of both quantity and quality of water (Figure 3).
In this WEIJthe interactions between water quality and quantity can be
investigated.
BPPLICBTION
The techniques described [i) are being used in the hydrological
evaluation of the Wash Storage, a pumped-reservoir scheme in the estusry
of the Great Ouse, a river in south-east England (Figure 4). The pre-
liminary estimate for the total capital cost of the scñeme is
2140 O00 O00 at 1971 prices. The work outlined here forms a small part
of the e2 900 O00 feasibility study though much of it will have applica-
tion even if estuary storage is rejected. A schematic diagram of the
whole system is given in F i g u e 5 with symbols defined in Table 1. The
complexity of the complete system has necessitated the division into
three interlinked subsystem namely, the Welland and Nene, the Great
Ouse and the Wash Storage. The first two subsystems define the poten-
tial input to the third.
3 86

The Welland and Nene subsystem which comprise8 the right hand portion
of Figure 5 is a model of a pumped-storage reservoir, %pingham (now under
construction), in conjunction with a confined aquifer, the Lincolnshire
Limestone. Water will be pumped into lbpingham from both the River
Welland and River Bene when the flows axe in excess of specified minimum
values. Rnpingham can be used for a variety of purposes including meet-
ing direct-supply requirements as well as regulating the lower Welland to
enable it to support downstream abstraction. Some of the water from
h p i n g h m will be returned to the Nene aa effluent, upstream of the
intake pumps for Ehpingham. Water from Rnpinghem will also be used to
maintain the flow in the River Glen. The Lincolnshire Limestone is used
mainly for direct-supply in conjunction with abstractions from the Welland
but any spillage from the aquifer helps to maintain the flow in the Glen.
The possibility of artificially recharging the aquifer from the lower
Velland has been included.
The Great Ouse subsystem comprises the left hand and upper centre
portions of Figure 5. The model is a simulation of an existing pumped-
storage reservoir, Grafham Water, in association with an unconfined
aquifer, the Great Ouse Chalk. Grafham Water is replenished by pumping
water from two points on the Bedford Ouse, a tributary of the Great Ouse.
Agairi, there is an element of recirculation since some of the water
supplied direct to a demand centre is returned as effluent upstream of
the reservoir's intake pumps. The Great Ouse Chalk aquifer has been
modelled as six interlinked unconfined aquifers. In a scheme shortly to
be promoted all the sub-aquifers are to be used for direct-supply and
river regulation. Obviously pumping water from an unconfined aquifer
will affect the natural outflow from the aquifer to the tributary.
Poreover, if the aquifer is drawn down, the possibility of seepage
through the bed of the tributary exists. Both these effects have been
incorporated in the model.
The lower centre portion of Figure 5 is a schematic representation
of the proposed first two stages of the Wash Storage which comprises the
third subsystem. Water could be pumped from both the Great Ouse and the
lower Nene. The possibility of having sea-water recirculation schemes
on both the Great Ouse and lower Nene has been included. This enables the
low-flow constraint at the tidal limit of each river to be zero.
Sgnthetic flow data generation techniques [27 have been used for
this invastigation. Currently the historic flow record on the River
Nene has been used as the master series and all other subsidiary flow
sequences have been obtained by regreseion on the logarithmic values of
flow. hproved multisite daily data generation techniques are being
developed under contract o] and will be used when available. Prior to
being used as the master series, the Nene record was corrected for all
upstream abstractions and effluent returns to obtain the 'natural' flow
series.
 387

INADEQUACY OF FLOiV DATA

A simulation model such as that used in the hydro1oglc.d design
of the T.3h Stor2.p rcyui.:>e., a coneiderable amount of information as
input data. It is inevitable that some of this data will be inade-
quate in one form or another. The usual case is where some inîorma-
tion is available but in insufficient quantity to estimate input
paxmeters reliably, m d for some parts of the system there is a
complete absence o€ data. To amplify these problems specific ex-
amples which have been encountered in the bdrological desi,m of the
Wash Storage axe given together with the way in which they have been
partly overcone.

INADESUACY DUE TO H U DATA

in modelling an unconfined aquifer such as the Great Ouse Chalk
it is evident that when punping the aquifer for either water-supply
or river regulation, the natural outflow from the aquifer to the
river will decrease. However, pumping the aquifer will have no
effect on the run-off from the non-aquifer portion of the catchment.
It is the combination of these two flow components that is measured
by the downstream gauging station. In short, if the aquifer is to
be developed by pumping, it is neaessaxy to have two inputs, the
recharge to the aquifer and the run-off from the remaining portion
of the catchment when only one measurement of the combined effect is
available. No details on the natural recharge of the aquifer were
known.

The aasumption was made that the downstream flow comprised two
flow recimes, a slow response from the aquifer itself and a fast
response from the remainder of the catchment. Having separated oyt
the base flow component, the overall 'proportion of base flow to
surface flow for the period of historic record was ascertained. The
surface flow component alone was cross-correilated with the correspond-
ing historic flow data for the master station on the River Nene. The
cross-correlation was performed on the logarithmic flow values which
gives weighting to the low flows and avoids the difficulty caused by
zero flows. This'relationship was then used to generate the surface
flow component direct. The base flow component could not be treated
in a similar manner since this was a measure of the output from the
aquifer rather than the input.

It was assumed that the temporal distribution of the surface

flow component was indicative of the periods when natural recharge
occurred. Therefore the surface flow component was scaled by the
overall ratio of base flow to surface flor and used as input to the
 388

recharge process. This data stream was attenuated by an exponential

delay function to simulate porous-media flow prior to adding the
percolate to the water already in storage. The delay induced by this
process was made equal to the observed mean delay between rainfall
ind the resulting maximum well levels.

The aquifer above the threshold constraint defining when channel

loss occurred, was modelled as a single linear storage. Consequently
the natural outflow from the aquifer to the river is proportional to
the mount of water in storage, the storage coefficient being derived
from the base flow recession. In this W¿QJ the effect of pumping the
aquifer was to reduce the amount of water in storage thereby reducing
the natural outflow from the aquifer without interfering with the
surface flow component.
INADEQUACY DUE TO INJCOIJPLEZ'E M T A
Although a historic flow record was available close to the pro-
posed abstraction point on the Ely Ouse, it would have been of little
use for the hydrological design of the Wash Storage even if it had
'been an accurate flow record. The river acts as a source of supply to
both industrial and agricultural consumers as well ag a disposal
system for treated effluents. No detailed records have been kept of
abstractions or returns and consequently the record can not be adjust-
ed to obtain natural flows. Ideally it would have been fax simpler to
have used the natural flow record at this station and account for the
net changes as time progressed rather than to have to construct a
simulation model of the entire river basin. In this specific case,
however, development of the chalk aquifer necessitated a simulation of
the entire basin. Fortunately better quality flow records existed on
ail of the important tributaries which were all upstream of the.major
industrial and agricultural demands.

INADXQUACY DUX TO INSUFFICIENT DATA

Traditionally the criterion for assessing the reliability of a
reservoir system has been the mean recurrence interval between failures.
This concept of return period has generally been defined quantitatively
in one of two ways, namely, a once in T year event where T is typically
50 or 100 y e a m or in terms of probability where it is said that there
is a 100 per cent chance of failure occurring in any one year. Assum-
T
ing that reservoir failures axe rare events and that the time between
failures has an exponential distribution, these two definitions are
 389

equivalent and the probability of m failures within n years is given

by :
-0

phn> =-e Mrn

n!
consequently there is a 37 per cent chance of there not being a once
in T year event in any T year period of record.

Even in the recent past attempts have been made to isolate low
flow events with return periods of 50 or 100 years from a short
period of historic flow data. The usual lengths of these records
typically range from 20 to 50 years. These lengths of record axe
totally inadequate for isolating such rare events and consequently
very little codidence can be placed in the results obtained. For
exmple, to be 9% certain that an estimate of return period is
within 2 10 years of a 50 yeas return period would require 2000
years of data and to be 9@ certain that the estimate was within
f 5 years would require no less than 11,000 years of data. biore-
over, even to be 9% certain that the return period was in the
r,mge of 50 years to 100 years would require 1600 years of data.
These data requirements show the absurdity of tho present reli-
ability criterion. It infers that all water-resource systems are
designed on inadequnte data with only the degree of inadequacy
varying between schemes.
Even if a once in T yeas low flow sequence could be isolated,
there is no guarantee that this would produce a once in T year
failure rate in a reservoir system designed to withstand such an
event. Shortkves in water supply are not independent events due
to the effect of storage. If a reservoir has failed one year and
has not recovered it is more likely to fail in the follovnng ye:=
than if it had been full at the start of the year. Consequentlg
reservoir failures come in groups rather than completely random
sequences and an event less severe than a once in T year flow
sequence closely following on a similar loa-flow sequence could
ceuse the system to fail. The occurrence pattern of these extreme
low-flow events is therefore as important as their severity and
individual events should not be taken from the historic record and
used in isolation when designing a reservoir system. Unfortunate-
ly the historic flow record provides just one realisation of the
occurrence pattern at a given point and the probability of the
historic sequence being repeated in the future i3 infinitesimal.
Consequently even if the whole historic record were used and even
if it contained what were considered to be extreme events there is
no guarantee that this would enable a realistic prediction of the
reservoir's reliability to be made.
 390

The difficulty of determining 'rare1 events from 'short' data

cannot be overcome. Recently the use of synthetic data generation has
alleviated some of the problems. The historic flow &ta is used to
estimate the parent population by modelling statistical chaxacteris-
tics and many synthetic samples can be generated each of which is
equally as likely to occur in the future as the historic record was to
have occurred in the past. In this way vaxious occurrence patterns
may be obtained and long perio&of synthetic data can be ?%gzìxdd as
producing a larger sample from the infinite population of possible
flows than the historic record affords. With the larger sample there
is a. correspondingly increased chance of the record containing (rare1
flow events providing a 'true' model has been used. However, the
synthetic data can only be as representative of the parent population
as the historic data. If an untypical historic record has been used
or there is insufficient data for the reliable estimation of node1
parameters then little confidence can be placed on the generated
sequences and in particulm on inferences about extremes within the
data.

in looking for a suitable design criterion we must accept the

lack of data and use a criterion that can be estimated with more
confidence from the same &ta. Rather than defining failure as a
reservoir or aquifer becoming empty, an event which would under-
standably be accepted only raxely, the introduction OP rationing of
water supplies can be used as the definition of failure. This would
o c c w when only a certain amount of water remained in store and would
obviously be tolerated more frequently. In practice a reservoir
would not be used at normal demand until it was empty. Instead a
level of storage would be reached below,which the supply would be
rationed. If rationing could be accepted, say, every twenty year3
then this would be a more frequent event and one has a correspond-
ingly increased confidence in the design.

Another shortcoming of the return period criterion is that it

gives no indication of the magnitude of the shortage. For example
in figure 6 the reservoir failed only once in the first case whereas
in the second case it failed twice. Therefore although the first case
is clearly the more severe condition the concept of return period
indicates the second is worse as it has twice as many shortages.
This is only.to illustrate a point but in practice reservoir failures
do group together which poses the problem of deciding whether such
a series should be consideyed as a single failure or a number of
individual failures. Therefore return period is not ideally suited
to describe the pattern in which reservoir shortages occur. An alter-
native criterion i~ required which must be a measure of both the
frequency and magnitude of failures. It must be flexible enough to
allow for a variable definition of failure (as the introduction of
rationing is somewhat subjective) and it must be simple to calculate.
 3 91

The concept of cumulative percentage frequency (CPF) of a

.specified failure level being reached meets these requirements.
CPF measures the percentage of time that the reservoir is at or
below a spec.ified storage. It will not however differentiate
between sw one &y of failure every year or a one hundred day
failure every hundred years. F i b w e 6 shoirs that the first case
would have a CPF of and the second 100 (x + y 1 which
V V
'correctly assigns the less severe shortage to the latter.
The CPF value for any reservoir state can easily be obteined
from the storage histopans already referred to (Figure 3). By
rerunning the model with different demands a graph showing the
CPF of vmious storage levels for different demands can be con-
structed (Fibwe 7). Given a reservoir level at which rationing
would be introduced the relationship between quantity of water
and reliability can be obtained. In this way the effect of dif-
ferent policies on reliability can be easily determined in t o m s
of yield and the definition of failure does not need to be pre-
judged.
CONCLUSION
Hydrological design criteria are based on rare events and
there will always be some degree of inadequacy in flow data.
Synthetic flow àata is only a partial solution because the
techniques are dependent upon the assumption that the historic
sample is representative of the infinite popultxtion of flows.
Even then, the historic data will only contain limited infoma-
tion on long-term periodicities and persistencics which are
important when examining rare events.'
Where no flow information is available there seems to be
little alternative to improvisation. This inay take the form of
transposition of data, scaling flow data or estiinating data
indirectly as in the case illustrated. Any improvisation should
always be treated with suspicion and attempts made to verify it
if possible, Failing this, simulation can be used, at a cost,
to ascertain the sensitivity of the system's performance to this
input. If the outcome is insensitive to that specific input there
is little cause for concern. If on the other hand the outcome is
sensitive to that input,at least it shows where the àata collec-
tion effort should be concentrated.

Another way of improving the confidence in the prediction of

reliability, given a limited amount of flow data, is to choose a
better design criterion by changing the definition of failure.
392

Hence the proposal is made that the introduction of rationing should

be used as the definition of failure since this would be tolerated
more frequently than the complete emptyiw of the reservoir. Nore-
over, by changing the concept of reliability to one which is both a
measuce of frequency and magnitude of failure rather than just the
frequency of failure enables two schemes to be compared objectively
even :?hen based upon a small amount of flow data. Thus the cam-
bination of synthetic flow data generation, introduction Qf ration-
ing au the definition of failure and cumulativa percentage frequency
as a masure of reliability helps to overcorce the problem of
inadequate flow data.

ACiO-f?LEIlc~
The authors thank their Director, Sir Norman Bowtree, for
permission to publish this paper in which the views expressed are
those of the authors and not necessarily those of the Hater Resources
Board.

1. Jamieson, D.G., Radford, P.J. and Sexton, J.R. (1973).

The Hydrological design of water-resource systems.
Water Resources Boosd. (To be published)

2. Bloomer, R.J.G.B. and Sexton, J.R. (1972). The

generation of synthetic river flow data. Water Resouroes
B o a d publication No. 15.
3. Weiss, G. (1973). Shot noise models for synthetic
generation of multisite àaily streamflow data. Symposium
on Desi,v of Water Resouxces Project with Inadequate Data,
Uadrid.
 3 93

TâBLE 1
LIST Q SYNBOLS ASSOCIATZD WITH FIGURE 5
D h a n d centre
E Effluent retumi
R Naturd recharge

AR Artificial recharge
S Seepage or spill-
I Natural inflow
L "ranslational delay
P Precipitation
V Evaporation

t b P
tc- AtiPinimum-flow constraint
 394

a
SECOND RIVER FLOW
- -A
II minimum flow constraint

FIGURE 1 Component Model of a Pumped-Storage Reservoir

FLOW
II minimum flow constraint

FIGURE 2 Component Model of a Pumped Aquifer

 3 95

NUMBER OF DAYS RESERVOIR AT THAT STORAGE

w W
O
O
t
O O i
I I
1I
I
4 BOTTOM WATER

O
2
v)
W
<
v)
-I
rn
4 3
f 1
Lp

O
6
æ
O
uI
P)
4 a
n
B3 n
o c
E 0
n
2
4 3
:
: o!
-s
n
al
I
o)
o)
3
<r,
P
-
11296
4 FULL
39t;
 397
w
398
Q
v>
m
*
o
c
al
3
U
e
L
al
m
Q
c
c
al
t
al
I \
\
D e
> .-i
al
.-
r3 .-
ir
&
O
f e
Y E
u3
r a l
0 -
U
a
u
Q>
e
O
Y
-
al
f
YI
a
a
E
LL
- 30VüOlS üIOhM3S3ü
3 99
I-
 MAXIMUM I N F O R M A T I O N O B T A I N A B L E FROM I N A D E Q U A T E DESIGN DATA:
FROM M U L T ï V A R I A T E TO BAYESIAN M E T H O D S
J e a n W e b e r 1 , Chester C.Kisie12 and L u c i e n D u c k s t e i n 2

ABSTRACT

An o v e r n i e w is given o f s o m e t h e o r e t i c a l and e m p i r i c a l i s s u e s
involved in d e s i g n i n g water pesource p r o j e c t s in t h e face o f i n a d e -
quate data. The primary f o c u s i s o n m u l t i v a r i a t e a n a l y s i s o f s a m p l e s
w h o s e p r o p e r t i e s are n o t c o n s i s t e n t w i t h t h e a s s u m p t i o n s o f t h e
analysis, The multivariate models discussed include multiple linear
r e g r e s s i o n , d i s c r i m i n a n t functions, c a n o n i c a l c o r r e l a t i o n , p r i n c i p a l
c o m p o n e n t s , a n d f a c t o r and c l u s t e r analysis. Each o f t h e s e m o d e l s
is d i s c u s s e d in t e r m s of i t s a s s u m p t i o n s , d a t a r e q u i r e m e n t s a n d
a p p l i c a t i o n s in h y d r o l o g i c r e s e a r c h . T h e B a y e s i a n a p p r o a c h to p a r a -
m e t e r e s t i m a t i o n and decision m a k i n g i s i n t r o d u c e d f o r t h e p u r p o s e
o f c o n s i d e r i n g b o t h t h e uncertainty d u e t o i n a d e q u a t e d a t a a n d
e c o n o m i c losses.

RESUME
Les a u t e u r s exposent q u e J q u e s c c F s i d é r a t i o n s g é n é r a l e s , t h é o -
r i q u e s et e m p i r i q u e s , sur 1'élaboratio.i d e s p r o j e t s d'am€nagement
des e a u x q u a n d o n s e t r o u v e e n p r d s e n c e de d o n n g e s i n s u f f i s a n t e s .
Ils m e t t e n t l'accent s u r l e s p r o b l è m e s q u e pose l'analyse m u l t i v a -
r i d e l o r s q u e l e s é c h a n t i l l o n s q u i l u i s o n t s o u m i s ne r é p o n d e n t p a s
a u x h y p o t h e s e s d e base de c e t t e analyse. L e s m o d è l e s m u l t i v a r i é s
dont i l est q u e s t i o n comprennent: l e s régressio?s linéaires, l'ana-
l y s e d i s c r i m i n a t o i r e (variable d e p e n d a n t e discrete), la c o r r é l a t i o n
canonique, l e s c o m p o s a n t e s p r i n c i p a l e s , l'analyse f a c t o r i e l l e et
l'analyse groypde. Chacun de c e s m o d e l e s est e x a m i n é s o u s l'angle
de s e s h y p o t h e s e s d e base, d e s d o n n é e s qu'exige s a m i s e e n o e u v r e
et de s e s a p p l i c a t i o n s en r e c h e r c h e h y d r o l o g i q u e . L'approche bayé-
s i e n n e d e liestirnation d a s p a r a m è t r e s et d e l a décìsion, p e r m e t
d'introduire a l a f o i s l'incertitude d u e à l'insuffisance d e s d o n n é e s
et s e s c o n s é q u e n c e s économiques.

lprofessor, Department o f M a n a g e m e n t , U n i v e r s i t y o f Arizona, T u c s o n ,

A r i z o n a 85721.
2Professors, Department of S y s t e m s and I n d u s t r i a l E n g i n e e r i n g a n d
Department o f Hydrology and W a t e r R e s o u r c e s , U n i v e r s i t y o f A r i z o n a ,
Tucson, A r i z o n a 85721.
 402

1 .O Introduction-
This DaDer considers L.e problem o .-recastinq o hvdroloaic variables for
water r e s o k e projects when the data-are inadequati, thät is, when there is a
mismatch between data and model. This mismatch is considered in terms of multi-
variate methods of data analysis. Mismatch implies a discrepancy between model
structure and structure suggested by the data and/or data inadequacy in relation
to model requirements. Several types of data inadequacies are considered in the
context of models frequently used in hydrologic research. The discussion is from
two related points of view; it considers limitations of a model in terms of the
assumptions on which it is based and sensitivity of the predictions of a model to
data inadequacies of various types. These considerations are inextricably re-
lated since the more restrictive the assumptions of a model are, the more likely
it is that data obtained are inadequate for estimating the parameters of the model.
Uncertain input information for the design of water resource systems is the
result of the inability of hydrologists to model large basins in substantial de-
tail as projected by Freeze (1972) and a result of the "curse" of small samples in
developing space-time series models and probability density models of flow, pre-
cipitation, temperature and evapotranspiration. Problems of extending data at a
design site and to ungaged sites are of long standing concern. An awareness-of
the implications of assumptions in mu1 tivariate statistical methods applied to
these problems is important to subsequent steps of coping with the consequent
assumptions and offering alternatives and decision strategies.
1.1 Model Building and Its Assumptions
When data such as streamflow are obtained, it is almost always for the ulti-
mate purpose of designing or operating a structure (bridge opening, dam, drainage
structure); one intermediate step consists of predicting or forecasting future
events (floods or droughts) using a model. The sequence of events in accumulation
of knowledge for predictions can be characterized as follows: some knowledge is
obtained by observations, a preliminary theory or hypothesis (for example, log
normal probability density function (pdf) of flow) is formulated on the basis of
these observations, additional data are obtained perhaps more systematically, the
theory or hy othesis is revised and/or refined (for example, log Pearson type III
pdf of flews!,
additional data are obtained, and so forth. As this interaction
between theory and data proceeds, the theory becomes more reproducible and per-
haps less general and the data required for its verification or modification also
become increasingly accurate, so that the design process may be started without
having to use large safety factors to compensate for uncertainty.
At some point, after accumulation of sufficient supporting data, a theory or
hypothesis is generally accepted and, unless subsequent theory and/or obser-
vations strongly indicate otherwise, the theory is used for prediction of a design
quantity such as the 50-year flood Q(50). By this time the theory is frequently
referred to as a model. As a theory becomes generally accepted, even tentatively,
the purpose of obtaining data gradually shifts; data are used less as a basis for
reformulating theory and more as a basis for estimating the parameters of a model
whose form has been determined, at least in most respects. Unfortunately. it is
frequently tempting to accept a theory and corresponding model prematuwly,
 403

especially if the urgency of making predictions or forecasts is compelling (for

example, in a decision to be made at once on the building of flood control works,
a water supply reservoir, or hydroelectric power dam).
Premature acceptance of a model can have very serious consequences, particu-
larly since the model is likely to be idealized to the point of being unrealistic
or to hold only under very restricted conditions, such as time invariance of a
watershed (Foge1 et al., 1971). Any model is an oversimplification of reality.
This is inevitable, because the purpose of theory is to simplify reality, which
is enormously complicated, by abstracting from it those elements that explain a
large proportion of the observed phenomena. Watershed models certainly are in
'

this cateogry. Acceptance of a model thus involves a compromise between realism

(e.g., a distributed model involving all details of the water cycle) and simpli-
city, represented by a lumped model. In economics and behavioral science, the
predictions of a theory or model are said to be appropriate, ceteris paribus,
that is, other things being equal (not varying). Under controlled laboratory
conditions extraneous variation can be minimized; in the real world, it gener-
ally cannot. Thus , i n using a rainfall-runoff model for forecasting streamflow,
for example, it is important to know how robust the model is to its assumptions,
including the ceteris paribus assumption which, for example, precludes urbani- .
zation. That is, it is important to know whether minor perturbations in the con-
ditions under which a model is applied have relatively small or relatively large
effects on its predictions. Clearly, if the predictions of a watershed model are
sensitive to changes in a variable, this variable should be included in the model.
Unfortunately, a model developed for one set of conditions is frequently used
under quite different conditions without consideration of the inadequacies of the
data obtained under those conditions. A linear rainfall-runoff model that has
been shown to yield a correct design of a culvert draining 50 km2 cannot be extra-
polated to a 500 km2 watershed. Here the predictions of the model may be quite
erroneous.
One of the major difficulties in choosing and using models for decision-
making in hydrology or other engineering design is the mismatch between available
models and available data. The implications of the mismatch are not clearly
understood. Most of the analytic derivations of the properties of deterministic
and statistical models (for example, sampling distributions of estimators , tests
of significance, standard errors of estimates and predictions, and so forth) are
based on assumptions that are almost always violated in applications. These
assumptions include linearity in the parameters of the system and of estimation
equations, normality of population distributions, independent random sampling,
large samples (for applicability of asymptotic results) and a variety of ass-
umptions concerning the covariance or correlation structure of multivariate ob-
servations and/or their errors. In most applications at least one of the assum-
ptions of the model is violated; the relevant concern should thus not be with
the properties of the model when its assumptions are satisfied, but should be
with the properties of the model when its assumptions are violated in various ways
and to various extents. The research in this area is not nearly sufficient to
provide practical guidelines (Dhrymes et al., 1972).
Specific assumptions and their violations are discussed in the next sections
for several models. In addition to inadequacy with respect to these assumptions
data may be inadequate in several more general respects including sample size,
missing observations, measurement errors , secondary variables.
404

These types of data inadequacy occur, for example, in the method of regional-
ization used by the U.S. Geological Survey (USGS). Assume that a stream f l w
characteristic, say the 50-year flood Q(50), is needed at an ungaged site for
design purposes. Regionalization may be used to calculate Q(50) in those cases
for which a data collection network for the region has been in operation for a
certain time. One method of regionalization, used by the USGS, relies on re-
gression analysis (Thomas and Benson, 1970). First, the desired streamflow
characteristic Q(50) is regressed upon basin characteristics, such as basin area,
precipitation, channel slope, elevation, forest cover and soil index. Then,
given the basin characteristics of the ungaged design site, the regression equat-
ion is used to predict Q(50) for that site.
For most regional data collection networks, many sites have record length of
the order of 20 years or less; small sample bias thus is quite substantial, so
that distributions other than the normal distributions should be used to compute
the logarithm of the flow Q(50); see Metler (1972). The basic reason for using a
regional regression is that data are missing in space at the design location and
large scale physically based models of combinations of river basins are non-exis-
tent to help in augmenting the data. Although calibration curves of flow veysus
gage height are periodically recalculated, there are difficulties associated with
sediment flows and with recording of flow data. All the variables entering the
regression equation constitute by definition secondary data (primary data is the
flow itself).
2.0 Multivariate Models and Data Inadequacies
The follwing sections concern multivariate models frequently used in
hydrologic research and focus on model limi tations arising from inadequate data.
The multivariate models discussed include multiple regression, discriminant
functions, canonical correlation, principal components ,:and' factor and cluster
analyses. Each of these multivariate models, with its assumptions, data require-
ments and applications, is discussed in the following sections. Primary refere-
nces on these models are Anderson (1958),'Christ, C1966), Dem ster (1969) Dhr mes
(1970 and 1972), Harmon (1967), Johnston (19631, benta (1971!, Morrison (19671,
Press (1972), Tryon (1970), Zellner (1971).
2.1 Multivariate Linear Regression
The purpose of multivariate linear regression analysis is to obtain an
equation -%-for predicting the value of a dependent variable 1 as a linear
function of a vector of k independent variables &=[Xi ,..., Xk]. The criterion
for obtaining the vector k=[bi, ..., &] is that the sum of squared errors
-
(y-@) '(Y-Xi) be minimized.
Linëarregression and analysis of variance, which can be viewed as a special
case of linear regression, are the only multivariate models for which there has
been considerable investigation of the effects of violation of the assumptions.
The assumptions. of the multivariate linear regression model ~ = X B + E can be stated
as follows: E sN(0,~)where c=$I is the covariance matrix o f t h e multinomial
&
(N) population, Ts non-stocFastTc, and has rank k<n.
 405

The number of observations is n, so 1is n x 1, X is n x k, is k x 1 and E is

n x 1. Thus the ässumptions are linearity, normality of errors, serial indëpen-
dence of errors, nonstochastic independent variables, and an observation matrix
of full rank with the nunber of independent variables less than the number of
observations.
The assumptions required for estimating a model are generally much weaker
than the assumptions required for inferences concerning the estimates. In
particular, normality assumptions are usually required for inference but not
for estimation (say of the 6's). This is the case for multivariate regression.
Since estimates are of little use without knowledge of their distributional
properties, the assumptions stated for mu1 tivariate linear regression and for
models discussed subsequently are those required for standard tests of signi-
ficance and confidence intervals.
The effects of violating the various assumptions of multivariate linear re-
gression are summarized in Table 1; also included in the table are procedures
for detecting violations of assumptions and proposed al ternatives for remedying
detected violations.
2.2 Canoni cal Corre1 ati on
The purpose of canonical correlation analysis is to study linear relation-
ships between two sets of variables l'=[Yi ,...,Y,] and L'=[XI ,...,Xq]. Peck
(1972) uses the analysis to determine whether 12 meteorological parameters (like
vorticity, vertical velocity, wind speed at different elevations and from various
directions, temperature differences, etc.) are sufficient to predict variations
in orographic winter precipitation patterns without the need for storm typing.
Nimnannit (1969) uses the technique to relate spring runoff at a set of stations
in the target region (where clouds are seeded) to runoff at a set of stations in
the control region (no seeding); the urpose is to assess the effectiveness of
weather modification. Torranin (19725 investigates the potential of the method
for (1) forecast of monthly precipitation of three large areas of the U.S. west
coast and (2) forecast of seasonal snowmelt ruiroff for three gaging stations in
the Flathead River Basin in Montana. The applications in hydrology have been
very few in nunber.
The analysis obtains vectors and such that the correlation between &'Y-
and b'& given by
.
r=
---
a'YX'b

------
Ja'Y 'Y a b 'X 'Xb
is maximized, subject to the normalizing conditions C'L'Y~ = 1 = d'&'Xb.
Subsequent vectors are obtained such that for each successive vector %e canonical
correlation is maximized subject to normalizing conditions and the condition of
independence with respect to previous vectors.
The number of canonical correlations between r=[Yl,. ..,Y,] and X'=[Xi,. ..,Xq]
is min !p.q), although in practice usually only the first few canonic3 correlates
are of interest. For the special case when either or is scalar, that is, con-
sists of only one element, canonical correlation is equivalent to multiple corre-
lation. Except in this special case, canonical eorretatim analysis is not useful
406

for prediction but is of value only to aid in formulating a modez; this is in

contrast to hydrologic uses mentioned above.
Under the assumption that Y' and X' are jointly normally distributed, the
joint significance of sets of cänonicar correlations can be tested using a like-
lihood ratio statistic. Unfortunately, the exact distribution of this statistic
is complicated. An approximate large sample distribution has been obtained, but
its convergence properties have not been studied. Thus , inferences concerning
canonical correlations can appropriately be made only on the basis of large
samples from a multivariate normal population. No information is available con-
cerning the nature and extent of the effects of violations of the assumption of
normality on the distribution of canonical dorrelations. Canonical correlation
analysis thus appears of limited use for building models for eventual use in
design with limited or inadequate data, in contrast with the hydroiogic uses
mentioned above.
2.3 -
Discriminant Analysis
The purpose of discriminant analysis differs from the purpose of multi-
variate linear regression analysis only with respect to the type of prediction
required for the dependent variable; in regression analysis the dependent variable
is continuous and its value is to be predicted, while in discriminant analysis
the dependent variable is discrete and its classification is to be predicted, for
example, classification of watersheds.
For the case of a dichotomous dependent variable, discriminant analysis can
be computed as a special case of multiple regression analysis by using a d u m y
variable having values zero and one for the dependent variable and point biserial
or biserial correlations between the dependent (dummy) and independent variables.
The regression coefficients obtained by this type of analysis are proportional
to the coefficients obtained by discriminant analysis.
The follwing discussion concerns discriminant function analysis for a
dichotomous dependent variable; the discussion can readily be extended to a
dependent variable having more than two categories.
Suppose that the independent variables &=[Xi ,...,Xk] are jointly normally
distributed in each of two populations with mean vectors and g and connnon
covariance matrix c of full rank k. If the prior probabilities of each population
(pop) are equal ana the costs of misclassification are equal, then the probability
of misclassification is minimized by using the following rule for classification
of an observation g
+ g ) 'in~pop 2 if K'L<$(g1+u)'L
classify in pop 1 if ~ ' ~ + ( ~ 1classify
The function 6 = r " ( ~ -g)is the discriminant function.
EI,
Unfortunately,
uand cy
ibis classification rule involves the population parameters,
which generally are u n k n w n in practice. The elements of the dis-
criminant function can be estimated on the basis of random samples of n1 and n2
observations from the two populationsand classification can be based on the esti-
mated discriminant function d=-'(x -Q)where is the sample covariance matrix.
It can be shown that this vector ddferentiates the populations by maximizing the
variance between samples relative to the pooled variance within samples.
Although the discriminant function based on sample estimates of the para-
meters can be justified intuitively, its small sample distributional properties
are extremely complicated and its convergence properties have not been investi-
&
gated. The affects of nonnormality of are not known. Thus discrinimant function
x
analysis is appropriate only when is normally distributed for each population
and, in addition, its application to small samples is appropriate only if the popu-
lation mean vectors p~ and u and the comnon population covariance matrix are
known.
2.4 Principal Components
The purpose of principal component analysis is to reduce the dimensionality
of K=[Xl ,...,Xk] on the basis of dependence among the variables. For example.
Fiering (1964), in his work on extending the single-site streamflow synthesis
model to the multi-site case, applied the technique to a river basin with p gag-
ing sites each site having an n-year record of annual flaws. Craddock (1965)
applied the principal components method to monthly temperature series from 1680
to 1963 for Central England. Other applications include increases in sediment
discharge from 31 watersheds after two major floods in northern California
(Anderson, 1970), sediment network design in California to insure accuracy of
predicted sediment yield (Wallis and Anderson, 1965), establishment of the uni-
formity of a hydrological region in Northland, New Zealand (Blake et al., 19-70),
derivation of a water yield model from monthly runoff data (Snyder, 19631,
identification of watershed factors from annual precipitation and runoff data of
watersheds in Coshocton , Ohio and Riesel, Texas (Diaz et al., 1968), and short-
range forecasts of river stage or discharge on the river Kolyma, U.S.S.R.
(Nechaeva and Mukhin, 1968).
Mathematically, principal components analysis transforms the X's to a set
of variables which are pairwise uncorrelated and of which the first has maximum
possible variance, the second has maximum possible variance subject to the con-
dition of being uncorrelated with the first, and so.forth. Principal components
are estimated on the basis of a random sample of n observations as follows. The
first principal component of & is denoted by Li=X a~ and g, is obtained such that
Z'1L1=g'lX1h1is maximized subject to the normaTizing constraint g'1&1=1.
n e secona principal component Q=X- a is then obtained by determining 7uch
that g&= am2X-'X9is maximiaed s d j e c t to the normalizing constraint ti23=1
and the independence constraint &'I3 = O. This procedure is repeated until the
k principal components have been obtained.
Large sample distributional prooerties of principal components have been
obtained assuming that has a multivariate normal distribution with a covariance
structure such that the covariance matrix c has k distinct characteristic roots.
The effects of nonnormality and the covergence properties of the large sample
distributions have not been investigated. Small sample distributional properties
of principal components are not known; this again limits the use of this technique
for the problems considered here.
In many cases determination of the number of principal components needed to
account for a reasonably large proportion of the variance in X is a matter of
judgment on the part of the investigator. Even if the investTgator is willing to
make this decision on judgmental rather than statistical grounds and he concludes
that a relatively small number of principal components seem to account for a
reasonably large proportion of the variance in X, there is still the problem of
interpreting the principal components in terms of the original variables.
 40 8

hfortunately, pi.incipal components are not a h a p interpretable and this hae

been a deterrent to the extensive use of principal components in developing
models.
The use of principal components as independent variables in regression
analysis has been suggested for the purpose of reducing the dimensionality of 5
and thus avoiding problems with degrees of freedom and for the purpose of
circumventing the problems resulting from multicollinearity in X. For hydrologic
applications see earlier references in this section as well as Singh's (1970a)
application for predicting infiltration in an aspen-grassland watershed in
southwestern Alberta, Canada. Al though principal components have been used as
independent variables in regression analysis by numerous investigators, there is
no generally accepted procedure for determining the number of principal com-
ponents to be included in such analyses, nor is there agreement concerning whether
x
it is acceptable to include one of more of the original variables in addition
to principal components.
In spite of these shortcomings and limitations, a principal component
analysis could possibly lead to a better use of insufficient or correlated data
for hydrologic prediction. For example, it offers an alternative to the method *

of regionalization described elsewhere in this paper.

2.5 Factor Analysis
The purpose of factor analysis is to account for the covariance structure
of a set of observable random variables in terms of a minimal number of un-
observable or latent random variables referred to as factors. Among hydrologic
applications have been those that sought decision rules that resulted in reduced
inventory and survey costs for specific areas and problems, as in the study of
the chemistry of groundwater quality (Dawdy and Feth, 1967), in the design of a
hydrologic condition survey in the TVA system (TVA, 1965), in parameter screen-
ing for watershed analysis (Shelton and Sewell, 1969), in predicting reservoir
losses in cavernous terrain (Knisel, 1970) and in reducing a set of edaphic
variables for a soil (Singh, 1970b).
Factor analysis estimates the coefficients to be us& in expressing each
response 'variable as a linear combination of a small number of unobservable
common-factor variables and a (latent) specific variable. The common factors
generate the covariances among the observable variables (responses) and each
specific term contributes only to the variance o f the particular associated
response variable. The coefficients of the common factors, estimated by factor
analysis, are not required to be orthogonal and their matrix is unique only up
to multiplication by an orthogonal matrix. The observations are assumed to be
a random sample from a multivariate normal population of full rank and the nunher
of common factors is assumed to be known; both of these requirements limit the
use of the technique in hydrology. The factor analysis model can be written as
-
X = n l + c where X is pxl, is pxm, 1 is mxl and is pxl, There are thus p
response variames L'=[Xi, ...,X,], m common factor variables r=[Y, ...,],Y and p
specific-factor variables E'=[E~ ,...,E 1. The matrix gives the factor loadings
where aij is the loading of the ith regponse variable on the j e common factor
variable. The conunon-factor variables l'=[Y1 ,...,Y,] are independently
distributed N(0, 1). The specific-factor variables E'=[E~ ,. are independently
 409

distributed N(0,q~~).Factor analysis estimates the elements of the loading

matrix A. Maximuil: likelihood estimates can be obtained assuming that is
multivariate normal with covariance matrix L=@&' of full rank p.
Note that principal components can be viewed as a particular solution of
the problem of factoring the covariance matrix. The-principal components
solution ignores variance associated with a specific response variable and re-
quires the factors (components) to be orthogonal and of decreasing importance in
accounting for (common) variance in the response variables.
Assuming normality, the adequacy of the m-factor model can be tested for
large samples using a likelihood ratio test of the null hypothesis E=&+$'
against the alternative hypothesis that is any symmetric positive definite
matrix. In most applications the number of common factors is not known and
successively larger numbers of factors are extracted until the goodness of fit
hypothesis is accepted or the computing routine fails to converge. Successive
tests used in this procedure clearly are not independent and the statistical
properties of the result are unknown.
Factor analysis has been used since the beginning of the twentieth century
to study the covariance structure of multivari te observations. Many variations
3
of the model discussed above have been propose and many estimation procedures
have been developed. Unfortunately, factor analysis, in any of its forms, may
be very difficult to interpret in practice. Part of the difficulty arises from
the fact that, regardless of the method of extimation used, the factor solution
is unique only up to a rotation of the axes. Various criteria, notably those
involving simple structure, have been suggested for obtaining the rotation most
readily interpreted; in practice, considerable subjectivity may be involved in
applying these criteria, even if their appropriateness is not in question.
Another difficulty in factor analysis arises from the fact that evaluation of
factor scores for use in subsequent analyses is not uniquely defined; several
intuitively appealing approaches have been suggested, but there are no apparent
criteria for choosing among them. Thus there are serious problems involved in
interpreting the results of factor analysis and using them in subsequent analyses.
In addition, relatively little is known about the sampling properties of
the estimates obtained in factor analysis (see Matalas and Rieher (1967) for
hydrologic discussions of this issue). The test for appropriateness of structure
assumes normality and large samples; unfortunately, the alternative hypothesis
for this test may not be the most interesting alternative in many applications.
There is considerable evidence that factor analysis can give meaningless results
if its assumptions are ignored.
As an example of the last point, consider Rice's (1970) use of variables
describing the physiography of experimental basins on the San Dimas Experimental
Forest in southern California. His goal was to identify variables which would
be useful in flood prediction. He notes "that the hydrologist might be better
rewarded if he turns his efforts toward developing physiographic variables
which better portray hydrologic processes rather than relying on a mathematical
artifact such Bs factor analysis to appraise the utility of various expressions
of basin physiography." This point is borne out in a non-hydrologic study by
Armstrong (1967); he finds that, while factor analysis "explains" a large pro-
portion of the variances, it fails to identify the known factors in the model!
41 O

2.6 Cluster Analysis

The purpose of cluster analysis is to group multivariate observations accor-
ding to various cri teria based on their degrees of homogeneity and .heterogeneity.
In hydrology, cluster analysis can be used to classify watersheds, flaw regimes,
and climates. Bogardi et al. (1972) have used it to group statistical pro-
perties of monthly water levels in Lake Balaton (Hungary). Hydrologic appli-
cations are very few. Cluster analysis can help to identify variables for in-
cl us i on in mu1 ti vari ate regression.
The other multivariate methods discussed above assume that the variables
belong to particular populations and that these populations have specific
(usually normal) distributions. In cluster analysis the variables are not assumed
to have even the minimal structure of belonging to particular populations and
the purpose is to establish appropriate populations as a basis for structuring
the variables.
Techniques of cluster analysis have been developed, almost exclusively, not
only for computer application but also on the basis of computer analysis. Al-
though mathematical rigor is minimal and statistical inference is almost non-
existent for cluster analysis, very useful results have been obtained in appli-
cations. Because of its (lack of) assumptions concerning population structures
and distributions, cluster analysis is applicable to a wide variety of hydrologic
and other problems;. its results can be useful if they are recognized as tentative
and i f even tentative conclusions are based only on results from large samples.
There are several questions or decisions that must be considered in any
cluster analysis: the number of clusters must be determined, the cluster
boundaries must be established, the method for handling correlated variables must
be specified, the technique for examining similarities must be chosen, and so
forth. Several approaches have been proposed for each of these aspects of cluster
analysis. Which criteria or rules of thunh are most appropriate depends on the
problem. Regardless of the techniques and criteria chosen for cluster analysis,
the investigator usually examines successive computer printouts and uses his
judgment to al ter apparently poorly selected cri teria and techniques. Compared
with the other multivariate analyses discussed, cluster analysis is more of an
art and less of a science, but so is engineering design under uncertainty assoc-
iated with insufficient data.
2.7 Bayesian Inference
The preceding discussion of multivariate models is entirely from the point
of view of classical sampling theory. Several of these models have been analyz-
ed from the Bayesian point of view and these results are summarized in the follo-
wing discussion. Bayesian inference incorporates, with sample infomation, the
investigator's prior information concerning the sampling distributions of the
parameters to obtain point or interval estimates. More general, however, is
Bayesian decision theory that incorporates both prior information and a loss
function with sample information in order to obtain parameter estimates or to
determine the optimal decision. Bayesian analysis is intuitively appealing; in
many applications the investigator has considerable prior data or experience as
a basis. for prior parameter distributions and in most applications he has at
 411

least a general idea of the?(economic) loss function associated with inaccurate

estimation. Unfortunately, the results for many multivariate Bayesian methods
are complicated and at best are applicable only for large samples. However,
since many classical results also have this limitation, Bayesian methods may be
p:i ferable because of their flexibility in incorporating prior distributions and
loss functions in the estimation of parameters or i n determining optimal decisions.
Also, human beings are better at estimating prior distributions than at esti-
niatlng posterior distributions (Ferre11 , 1972).
There has been considerable application of Bayesian methods in mu1 tivariate
linear regression analysis. Bayesian point estimates of the regression coeffi-
cients can be obtained with or without incorporating loss functions and Bayesian
interval estimators (credibility intervals) can be formulated.
As discussed in section 2.5, the maximum likelihood factor analysis solution
is unique only up to a rotation of the axes; the use of subjective information
in a Bayesian analysis is an intuitively appealing basis for eliminating this
ambiguity. Unfortunately, the technical difficulties involved in obtaining
numerical solutions have thus far precluded use of this approach.
The Bayesian approach has also been considered for canonical correlation
analysis; unfortunately, even for the simplest assumptions with respect to both
the prior distributions of the parameters and the sampling distributions of t h e
data, the Bayesian results for canonical correlation analysis are so complicated
that their applicability is extremely limited.
The most notable success of Bayesian methods in multivariate analysis thus
far has been for discriminant functions. As summarized above, a number o f methods
based on the sampling theory viewpoint have been proposed for discriminant
analysis, but these results are unsatisfactory for use with small samples. The
Bayesian approach provides a useful and simple al ternative.
Consider the case of classification into one of two mutually exclusive pop-
ulations. Denote the populations by Pi and P , the vector of observations by
x'=[xl, ...,xk]. the density functions by fl(Kf and fp(&) and the prior probabi-
Tities by p1 and p2 where p +p2=1. The costs of misclassification are C(211) if
an observation from P1 is classified in P2 and C(112) if an observation from P2
is classified in Pl. The problAem is to determine a classification rule of the
following form: partition K into regions R1 and R2 such that if ZERI, the obser-
vation is classified in Pi and if x ~ R 2 ,the observation is classified in P2.
The expected cost of misclassTfication is given by

and the corresponding ,classification rule is

f+X) C(l M P 2 f+x) C(112)PZ
R1:q-g 'copl R2: < cop1
This classification rule involves the densities f1(&) and f2(&) which may not be
known. Assume that fi(&) and f2(&) are multivariate normal with mean vectors PJ
and c.
and common covariance matrix Then the above rule can be written
PIP,
c(1 c(1 12)P2
R, :L'i.-+(IL,+q) 'L>log, R2:~'6-+(~1+4)'&<loge
41 2

where 6 = E- 1( p l y ) and x'6 is the discriminant function. If p1=p2=% and

C(112)3(2ll),'-this reducësto the rule, given in the discussion of the sampling
theory appLoack to discriminant analysis. As noted in that discussion, sample
estimates x x and 2 may be used to obtain from the sample data without
knowledge d'i c.
-2 and
Bayesian aiscriminant analysis can easily be extended to other cases; for
example, fl(x) and f2(&) may have some form other than the normal distribution
or there may be more than two populations into which an observation may be
classified. Finally, for the sake of completeness, we should mention the use of
Bayesian decision theory in design to imbed uncertainty in parameters resulting
from inadequate samples into a loss function (Davis et al., 1972; Davis et al.,
1973).
3.0 Sumnary and Conclusions
In this overview, we have critiqued the current status of multivariate
methods of data analysis because of their central position in making estimates
and predictions of both hydrologic and econometric (e.g., cost) inputs to design '
of water resource systems. The design implications of many of the assumptions
-
in these methods remain to be evaluated a task of importance to many profes-
sional disciplines. In contrast to the current tactic of using multivariate
models without considering the assumptions involved, we believe that use of
Bayesian decision analysis, while not the final answer, may be a viable alter-
native for anticipatinq poor design. Bayesian analysis offers flexibility in
incorporating prior (subjective) knowledge about probabil i ty distributions on
design parameters and it encourages the design engineer to invoke his general
ideas of loss functions associated with inaccurate estimation in many specific
design problems. The focus is on the consequences for a specific use and not on
a precise design estimate. The latter has a subtle linkage of probability and
utility, depending on one's value structure, but Bayesian decision analysis
encourages specific consideration of each in an open manner. With the continuing
emphasis on environmental impact evaluation, such an approach seems timely and
necessary in the face of small samples of hydrologic and other environmental
data.
4 .O References
Anderson, H. W. 1970. Principal components analysis of watershed variables
affecting suspended sediment discharge after a major flood. Int'l. Assoc.
for Hydrologic Sciences. Publ. 96, pp. 404-416.
Anderson, T. W. 1958. An Introduction to Multivariate Statistical Analysis,
New York: John Wiley ti Sons, Inc.
Armstrong, J. S. 1967. Derivation of theory by means of factor analysis or
Tom Swift and his electric factor analysis machine. The American Statis-
tician. 21(5), pp. 17-21.
 41 3

Blake, G. J., A. D. Cook and D. H. Greenall. 1970. The use of.principa1 com-
ponent factor analysis to establish the uniformity of a hydrological region
in Northland, New Zealand. Int'l. Assoc. tiydrol. Sci. (IAHS) Pulb. 96.
Bogardi, I., L. Duckstein, and C . C. Kisiel. 1972.. Distribution of dynamic
water level in a shallow lake, paper prepared for Fall Annual Meeting, AGU,
San Francisco, Calif., December.
Christ, C. F. 1966. Econometric Models and Methods. New York: John Wiley
and Sons, Inc.
Craddock, J. M. 1965. A meteorological application of principal component
analysis. The Statistician, Vol. 15, p. 143.
Davis, D., C. Kisiel and L. Duckstein. 1972. Bayesian decision theory applied
to design in hydrology, Water Resour. Res., Vol. 8, No. 1, pp. 33-41.
Davis, D. R., L. Duckstein, C. Kisiel, and M. Fogel. 1973. A decision-
theoretic approach to uncertainty in the return period of maximum flaw
volumes using rainfall data, paper to be presented at Symposium on Design
of Water Resource Projects with Inadequate Data, UNESCO, Madrid, Spain,
June.
Dawdy, D. R. and J. H. Feth. 1967. Application of factor analysis in the study
of groundwater quality, Mojave River Valley, California. Water Resour
Res., Vol. 3, No. 2, pp. 505-510.
Dempster, A. P. 1969. Elements of Continuous Multivariate Analysis. Reading,
Mass : Addison-Wes 1 ey .
Dhrymes, P. J. et al. 1972. Criteria for evaluation of econometric models.
Annals of Economic and Social Measurement, Vol. 1, No. 3, pp. 291-324.
Dhrymes, P. 1970. Econometrics: Statistical Foundations and Applications,
New York: Harper & Row.
Diaz, G., J. I. Sewell and C. H. Shelton. 1968. An application of principal
component analysis and factor anal sis in the study of water yield. Water
2
Resources Res. , Vol. 4 NO. 2, Pp. 99-306.
Ferrell, R. 1972. Subjective inputs and uncertainty in water resources
decisions, Proceedings, Int. Symp. on Uncertainties in Hydrologic and Water
Resource Systems, Univ. of Arizona, Tucson, Ariz., Decenber.
Fiering, M. 1964. Multivariate technique for synthetic hydrology. J. Hydraulics
Div., Proc. Amer. Soc. Civil Engrs. Vol. 90(HY5), pp. 43-60.
 41 4

Fogel, M, M., C. t. Kisiel and L. Duckstein. 1971. Space-time validation of a

rainfall model for summer-type precipitation, Mater Resour. Bull.,
Vol. 7, NO. 2, pp. 309-316.
Freeze, R. Allan. 1972. Role of subsurface flow in generating surface runoff
2. Upstream source areas, Mater Resour. Res., Vol. 8, No. 5, pp. 1272-1283.
Gray, Howard. 1972. Bayesian Decision Analysis of a Statistical Rainfall/Runoff
Relation, Tech. Report #14 of Reports on Natural Resource Systems, Univ.
of Arizona, Tucson, Arizona.
Harmon, H. H. 1967. Modern Factor Analys'is, 2nd Ed. Chicago: University of
Chi cago Press.
Johns ton, J. 1963. Econometri c Methods. New York: McGraw-Hi 1 1 , Inc.
Kmenta, Jan. 1971. Elements of Econometrics. New York: Macmillan.
Knisel, W. G. 1970. A factor analysis of reservoir losses. Water Resources
Res. Vol. 6, No. 2, pp. 491-498.
Matalas, N. C. and B. J. Rieher. 1967. Some comnents on the use of factor
analyses. Water Resources Res., Vol. 3, No. 1, pp. 213-223.
Metler, W. A. 1972. Bayes Risk Analysis of Regional Regression Estimates of
Floods. Master of Science Thesis, Dept. of Systems & Industrial Engineer-
ing, Univ. of Arizona, Tucson.
Morrison, D. F. 1967. Multivariate Statistical Methods. New York: McGraw-
Mill.
Nechaeva, N. S. and V. M. Mukhin. 1968. The use of statistical methods for
short-range forecasts. IAHS Publ. 81, pp. 405-416.
Nimannit, V. 1969. Multivariate analysis of hydrologic changes. Doctoral
dissertation, Dept. of Civil Engineering, Colorado State Univ., Fort Collins.
Peck, E. L. 1972. Relation or orographic winter precipitation patterns to
meteorological parameters. Proceedings, Int'l, Symp. on Distribution of
Precipitation in Mountainous Areas, World Meteorological Organization,
Gei 1 o, Norway.
Press, S. James. 1972. Applied Multivariate Analysis. New York: Holt,
Reinhart, and Winston.
Rice, R. M. 1970. Factor analyses for the 'iiterpretation of basin physiography.
IAHS Publ. NO. 96, pp. 253-268.
 415

Rodda, J. C., et al. 1969. Hydrologic Network Design, World Meteorological

Organizat ion/International Hydrological Decade, WO, Geneva, Sui tzerland.
Shelton, C. H. and J. I. Swell. 1969. Parameter screening for watershed
analysis. Trans. ASAE, Vol. 12, No. 4, pp. 533-539.
Singh, T. 1970a. A principal components regression model for predicting infil-
tration. Paper presented at 1970 National Fall Meeting of the American
Geophysical Union, San Francisco, Calif.
Singh, T. 1970b. A minimum entropy rotation of principal components for obtain-
ing simple structure in a hydrologic data matrix. Paper presented at the
1971 Fall Annual Meeting of the American Geophysical Union, San Francisco,
Calif.
Snyder, W. M. 1963. A water yield model derived from monthly runoff data. IAHS
Publ. 63, pp. 18-30.
TVA (Division of Water Control Planning). 1965. Design of a hydrologic
condition survey using factor analysis. TVA Research Paper No. 5.
Thomas, D. M. and M. A. Benson. 1970. Generalization of Streamflow Character-
istics from Drainage Basin Characteristics; U.S. Geological Survey Water
~

Supply Paper 1972, 55 pp.

Torranin, P. 1972. Application of Canonical Correlation in Hydrologic Pre-
dictions, Ph.D. dissertation, Dept. of Civil Engineering, Colorado State
Uni v. , Fort Col 1 ins.
Tryon, R. C. and D. E. Bailey. 1970. Cluster Analysis. New York: McGraw-Hill.
Wallis, J. R. and H. W. Anderson. 1965. An application of multivariate analysis
to sediment network design. Int'l. Assoc. Hydrol. Sci. (IAHS) Publ. 67.
pp. 357-378.
Zellner, A. 1971. An Introduction to Bayesian Inference in Econometrics.
New York: John Wiley & Sons, Inc. 480 pp.
41 6
u-
O
L
O
bal
n
u-
'I- s
lu
C U
O
.Y s
W O .
ru .r- u
:'Eo n E er
+*r E
WILL
S W <u
rumw
L'I- e,
FUU
.
lnl
-L
a o
a s .
m
:<cl-
m o a
a s
L O U
m -r
calm
o+ o>
L
s a
o L<t
.
Y Ia o
c,a
X %i3
Pi*
m .?r-
I55 g
U
.I-
c,
m
i m
S.C
o u
so
x G
mI
 417
WWEEv)
k.G
mvi L o w
E m o n u

-
m wv- E E
C L C O W
u O.- o L
I
c
CncnY:
L.r w c
2 l
u)-h
o w -1.- Sr-0
<c r: L L-I- h m m u
o mo+ h F L O W
E-r W L E- -.P.- L 3
00u L o a m v)L a
U u L c 0 E
W c 0.6-u
mu- L w
L0ic e E I c C>L CI
<cm *rml- .r'UE o
IC
w -0 w
C m w L L w.ru-
40
t
O uO) Y
+ E g w e uic-0 o
r w
.I- W a
VIa
VI
w
L O
cnL

0 m L
-cu,um
w L U O)%-
.- icv)
L S I - r -

rr%u- 0
w .P
.).we-* s
(np . .
r
e
w ~ u o c m w o n IO o 3
u w-I-icm > c> E m al.? L
m .I-
.I- ..)
Lu I
L L m s e al
na >4 c
o w O .
#-
r-
L eu L b
g z p
m e E m
c,
VI
E o 0'- L o,
-cicl-C> .n
v)
c -
r-

h
0
.r
L
m
li
W
V
ZX
u-c W
.- 3
I
X L P
 PRATIQUES COURANTES POUR L'EVALUATION DES CRUES E T
DES DEBITS D'ETIAGES PRIS EN COMPTE DANS LES PROJETS

Rapport General

P ar

M a r c e l ROCHE

Pour &borer les données hyärologiqueis néceseaires à la mise au

point diun projet d'aménagamcmt des eaux, on dispose finalement de trois
graisda types d'approche :

- u'utilieer que lee observatione concernaut les débite et re-

cueillie6 au site m b e de l'amhgement ou & p r o a t 6 ;
- utili- 6gelement les données climatiquee ãieponibles BUF le
baadn, notemnent les relevée de précipitatione i
.
.88 m v l r de formules r6gionalee1 ou de corr6lationa. Pour a-
trapoler des reeultate recueillis d dee statione hgdrom&riqUeE
et/ou plu~iométrigues tdtu&s BiUeure äaus ia regiou, à 1li.n-
ttieur OU d l'extbieur du basain fluvial.

Le premier typa groupe lee dthodee dites Wirscteetl,tandis qUe

lee d e w autree pmoèdemt de 1'Cvaluation indirecte. La m&hode "la plu8
aireate" consiste en l k d y s e statietique d'un Bchentillon, par exemple
420

de débits maximaux annuels t elle nécessite, pour que l'évaluation ait un

sens, une infoimation riche, des m e m e s précises de débite portant sur de
longues périodes.

des échan-
Le second type de méthodes ccmaiste en fait à l96tendrett
tillons existante de debits, portant BUT dee périodes courtee, en utilisant
les relations qu'on peut dkager entre lea dábits et les donnbe climatiques
disponibles sur une dur& beaucoup plus longue. C'est @ll'exteneion des don-
néedi, qui fait largement appel aux méthodes de r¿greesion et à l'anaïyse
multivariate, mais aussi dans certaine cas aux modèles dits conceptuels. On
peut rattacher à ce type lee m&hodes qui consistent à effectuer les trans-
formations ou à appliquer les régressions & u11 évènement climatique de pé-
riode de retour connue, ou considéré comme un maximum possible.

Le tmi&me type, enfin, rassemble les méthodes d'interpolation

ou d'extrapolation géographique. Ces méthodes vont de la simple analogie &o-
morphologique et climatique, aux raffinements de l'analyse factorielle, en
passant par les tablee et abaquee régionaux.
U paraft se dégager de cette &um&ration une id& simple de l'uti-
lisation dee m6thodee en face de l'information disponible ; il semble évident
a priori que les m6thodes de type I n'ont de sens qu'en cae d'information très
abondante, que o d l e s du type II correspondent A une information hydrologique
réduite, mais à une informetion cïimatoïogique consistante, et que le type
III groupe les mgthodes utilie8ss lorsqu'on ne dispose de pratiquement rien.
On pourrait en conclure que la m&hodologie de type I devrait être exclue
d'un symposium tel que celui-ci.

En fait, le jugement doit être plue nuancé, car on commence à ten-

ter des régionalisatione BUF lee lois de distribution, donc à rechercher par
là des estimations indirectes, a melanger dane lee études statistiques des
dondes de conaietances t A e différentes. C'est pourquoi nous rQerverom,
ciam notre expod, une piace à l * m & e statistique.
Chaque m&hode d'heluation fait appel à des outils de calcul qui
ne lui sont pae forchent propre. C'est aind qu'un modèle 6 structure détor-
ministe, ì'hydmgranmie unitaire par example, peut etre utilisé pour une ex-
tension des donnhe, pour u11 celcul de transformation d'una, averse de fr&
quence donnét, ou pour une extrapolation &graphique.
 421

ûn pourrait envisager, dans ce rapport, une présentation par

"outil de ~ e l c u l;~il
~ nous a paru plus cmfonne au eujet du eymposiuc
d'aùmter un ordre d'exposé qui se rattache, autant qu'il m i t possible,
à la quantité d'information disponible.
Les sujets proposés dans la question qui fait l'objet de ce rap-
port général se rappsrtent aux débits extr&nes, c'est-à-dire aux étiages
et aux crues. I1 faut bien reconnaître que le premier sujet n i a guère tenté
les ~ãpécialistes,puisque un 6eUl rapport, eur les quatorze que nous avona
examinés, traite des basses eau, Finalement, cela n'est pas tellement sur-
p r e m t ; les basses eaux sont liées de très pr&3 aux problèmes d'hydrogh-
logie ; or ce n'est que dans quelques cas trèer particuliers qu'il est possi-
ble de dégager des paraadtree morphologiques et climatiques simples qui
soient en relation directe avec ces pmblèmeo et qui permettent une t r a m p -
sition &,panhique suffisamment précise. Par contre, u11 auteur aurait pu
être tent6 par l'aspect " e x t e d o n des donnée^'^, qui dierpoee d'une méthodo-
logie courante sinon riche, du moina 88888 efficace ; cela ne s'est pa6
produit.

I1 était précisé enfin que les expos6s devaient se rapporter à

des llpratiquescourantes", ce qui semblait exclure les sujete de recherche
et certaim procédk de calcul non totalement dégagés de leur phase exp&i-
mentaie. Aussi n'Fneieterona noua par3 t m p LNF certaine aapecte qui no-
ont été soumie ; ceci ne veut pas dire que noua lee trouviane peu dignes
d'inter&

Dana notre prbentation des rapporta, nous parlerone d'abord des

étiages, puis des crues.

Le eeul rapport traitant de ce pmblème est celui de

MM. Vladimirov et Chebotarev [I3]- Lee autemrs définiasent avec m i n les
422

en
variables par lesquelles on caractÓriae lee basses eaux/G.R.S.S. :
- d&it journalier minimal de l'annw,
- d6bbit moyen meneuel minimal de l*annh.

Chacune de ce6 oarlablee est définie pour choune dee deux saisons
de basses eaux :celle d'hiver et celle d'&&-automne.

Les auteurs exposent ensuite ï'infïuence, sur ces débits de b M 6 e 6

eaux, des caractéristiques climatiques, en insistant Bur le A l e du gel, et
des carúct&istiques phyeiographiquuee, en notant le r81e particulièrement
important des lacs, des rnaraic, du sol, du sow-sol, du karst. 110 définie
sent enmite le cadre régionel dans lequel va s'exercer la méthadologie :
les bassins sont classés en

- petits bassins : jusqu'a 1 o00 1 500 h 2 e n zone de plaine hu-

mide, jusqu'à 2 000-2 500 h2dan6 les zones de montagnes oc les
zones de plaines peu humides, jutSqu'8 5 OOO-10 o00 km'
dans les
régiow arides ou pour les rivières soumises à un gel intense ;
- bassins inoyene : jusqu'8 75 O00 km2.

Pour lee petite bassins, on applique, dana un cadre strictement

régionali66 6 partir des caractériatiques phydographiquee et climatiques,
une formulo de la f o m e Q = a (A ;I. fIn oÙ Q eat le d B i t meneuel minimal
daps le texte) en
moyen (dit l~omuiì~l d/s,A ia eurface B.V. tan d, f un
correctif tenant compte d, la non coincidence hentueile du bassin topogra-
phique et du bassin souterrain ; a ot n srnt de6 paramdtres régionaux. Le
passage du débit % o d 1 * 6 dee débits de fr8quences donnbs er'effectue par
deux méthode8 diffbentes.

?our lee baemns moyenne, on utilise des cartes d'isopl&thes (iflolines)

pour la détexmination du débit d'étiage meneuel ; ces cartes sont établies
pour les valeur8 moyennes (normeles) et pour la f r b e n c e 80 % de dépasscirent.

Les étiages j o u m d i e r s sont dikìuits de6 étiages menauele par un

coefficient de.-p

Il s'agit donc d'un proc6dd8 de formule @kd8gionale1@

dont les parani&
tres sont d&erminb gar un catalogue, et non l i b à dee caractéristiques
 423

ghorphologiques et climatologiques meeurables. C'est pratiquement le seul

moyen de e'en eortir en matière de basaes eaux, pour los raieons que noua
avo- deja indiqubs. I1 est dommage que l'auteur n'ait pea indiqué quel-
ques valeure de cosfîiciente r6gioll~yx,ni montré sur un exemple la concor-
dance des réniltate du calcul avec des d o m b s observks.

Lorequ'on dit qu'on détermine dee crue8 en l'absence de données,

c'est faux. Sane donnéeel on ne caicüie r i a du tout. Simplement, on cherche ici
.4t r v e e r des d o ~ 8 e sexistantee ou à établir des relations entre le phé-
nomène qu'on &tudie et des données d'une autre nature. Toute L1infonnation,
et par suite +out le sérieux de6 e d h a t i o n s qui en dbdent, tiennent d a m
Ce8 dO~h6.

C'est pourquoi nous regrettons un peu qu'il ne se m i t pas trouvé

d'auteur pour exposer lea méthodee, pourtant de pratique bien courante, per-
mettant de reconetituer UTIcertain nombre d'6v&nementa marquanedu passé.
Noue verrou tout 6 l'heure qu'on commence a se prkoccuper d'insérer dane
des échantillons r4guliere de dbbite d m a w de crue5 annuelles, dee obser-
vations discontinues, parfoie t m n q u h et souvent entachées d'erreur6 beau-
coup plus grandee que celleede l'&chantillon régulier. Lorsque lee donnéee
sur les ornes sont r a m e et surtout portent SUT de5 périodes très courtee,
il devient extrhment important de dispaser du plus grand nombre de telles
obaervatiom.

chi peut aauvent trouver dana les archivea, dane la preme, ~ u r

de6 télégrammes adminietratifel dee indications parfoiPs trèB préci898
oertaines 5 m d e e crues. On peut également mener des enquatee 6ur place,
auprès dee riverains et em a'intbrsemant aux d&hhs& ou autres marques
laiss6es par ces orue~). 11 y a 1a toute une m6thodologie dont n o m no pou-
viam peu ne pas tout BU moina signaler l'existence.
424

des crues réellement observées e d traditionnellement un outil pour informa-

tion riche. Des études relativement récentes ont pourtant 6tk men688 afin
de voir s'il n'Était pas possible, en cas de données rares,

- soit de compléter l'échantillon de données rkgulierement recueil-

lies par des observations sporadiques et/ou t m n q u b s ,

- soit de faire de la transposition régionale sur les lois stutis-

tiques elles-mêmes

Bien entendu, lee deux ne s m t pae incompatibles.

La première question a été traitée, au moins partiellemmt, dens la

cornunication de M. Morven N. k e e e 183. L'auteur considère deux aspects de
la question. Dans le premier, il suppose qu'on dispose d'une série n o d e
d'observations (enregistrements par exemple) au cours de laquelle un certain
nombre de maximums annuels n'ont pas &é observés, par exemple par suite d'uno
déficience d'appareillage : on cornaft seuìernent , pour les crues concernhs,
le seuil inf&ieur du débit, seuil supposé constant eur la période. C'est le
cas pratique du stylet d'enregietreur qui sart des limites du tambour ou de
la table dbulante.

Le second aespect est celui de la station pour laquelle on dispose

d'une s k i e d'observations régULiBres de N années, courte mais complète,
des maximums m u e l a de cruea. On connaît par ailleurs, sur une p a o d e de
M autres annha, les n d8bits maximaux de crues ayant d & a d un seuil 5 ;
et on est certain que p e a t lee N-n anahs reetantes de cette période, aucun
débit n'a atteint ou d6paee6 xh. Ce eont là des hypotheses assez restrictives
qui correspondent au cas pratique des échelles de hautes eaux W l o i t é e s Pon-
dant u11 certein nombre d'années avant que les services se préoccupent des
basses et moyennes eaux ; ce peut (tre aussi le cas de certaines marques
relatives a des cmes historiques.

Lee deux problemes sont trait& par la méthode du maximum de vrai-

semblance, suivant la technique indicph p m Kendall aux paragraphes 32.15
et suivants de son ouvrage. L'auteur donne un exemple d'application a la ri-
vière Avon a la dation de Bath (U.K.). Le but de l'étude eat de déter-
miner le gain d'information apport& par les observations tronquées ou les don-
n k s que l'auteur qualifie d'historiques ; ce gain d'information est estimé
 425

ici par la r&ction apport& à 1'eITeur etandard d'estimation pour différent;

quant iles.

Cette 6tude est fort int&resaante, bien qu'on puisse ne pas être
pleinement d'accord avec touteei les hypathèees introduites par l'auteur. Elle
est susceptible d'importants prolongements vers d'autres formes d'information
tronquée ou sporadique, mais le traitement rieque alors de ne pas être aussi
simple.

En ce qui concerne la aeconde question, la traueposition des lois

statistiques a été mainteefois tenth, souvent avec succ8s. Nous nous permet-
tona de rappeler les 6tudes de U, Oktay h a n o g l u sur le5 mo~ulespluviom&
trique8 de IlAfrique de l'Ouest (Cahiers de l*O.R.S.T.O.M., hérie @droiogie>.
MM. Herbst, Van Biljon, Olivier et Hai1 noue présentent une c o d c a t i o n sur
la régionalisation des paramètres das lois de distribution des crues 151.
Les auteurs prennent comme m>dèle statistique la loi log-gaiinna
incomplète (log-Pearson III), tout en hoquant la possibilité d'effectuer
les mi3mee opératiom en Re basant m r une lo' de Qumbel. La méthode consiste
à
- calculer. '.es paP810dtime8 des lois pour toutes les longues séries
diaponibleB,

- apposer que chacun de ces paramatres d6pend deun certain nombre

de facteur@ gbmorpholagiques et climatiques (il cite la super-
ficie du bm&n, la pluie annuelle moyenne, la pente moyenne,
la longueur de la rivière, la pluie menquelle maximale médiane,
un facteur de f o m e mais n'utilise dans la suite du texte
que La surface du bassin A et la pluie annuelle moyenne R),
- appliquer pour chaque parametre de la loi une régression multiple
avec lea facteurs retanus ; les coefficients de la r6gression
sont calculés par lea moindres carrés et l'opération constitue
en fait un "lieeage gbgraphiquelt des valeurs de ces paramètres.

En r&äiité, lee auteum neiitilisent pas les param¿tres fLgurant

dana l'expresdon mathhatique de la Loi, &is la moyenne (des logarithmes
dee débite), ll&cart-type et 10 coefficient d1msiymétrie, ce dernier étant
du reste trait6 de fapon trèe diffhnte. Le p i n t le plus important au calcul
426

se rapporte à la variance d'estimation de tele paramètres et dee quanti-

qui en dkoulmt, toujours parr l'intermediaire de la loi log-ggnnaa.

Les auteurs donnant quelques résultats obtenue en Afrique du Sud.

Lo c a d c a t i o n de MM. Davis, Ducketein, Kisiel et Fogel /2]
traite du problhe de la distribution de la pkiode de retour correspondant
au déparssement d'un maxipnrm ou diun volume de crue donni, en partant des don-
nées sur les précipitations. La méthode se rapporte plut8t au type II,
d e les techniques de calcul expos6ee r e l h n t enti8rement de l'analyse
atatidique. La formule de transformation pluies-débits adoptée pour les
= -
volumes de cnie e& de la forme Q C (R A). Dana l'analyse de &bili-
té conduite par dmulation, les auteurs ne se préoccupent que de C et trou-
vent, c o m e il fallait s'y attendre, une énorme influence de la variance
d'estimation de ce paramètre sur la variance d'estimation de la période de
retour. I1 n'est pas pmuvé, c o m e semblent le supposer les auteurs, que
la variance de A n'ait qu'une influence négligeable.

2.2. - kJB-hodes_de rn-1; = *&e-d-og gee ~ o ~ . g

Dane Bon sena le plus littéral, l'extension Eee d o ~ deh d6bits
consiste en l'opération suivante.
- Une variate Y &tant définie mivant le phhomene i@rologique
qui intbesse et le probl8me qu'm a à résoudre (par emmpie
Y a d&it maximal instantané de l'année) ;
- on dispose d'un échantillon de n valeurs de Y obtenues par l'Ob-
semation directe des débits ;
- on dispoee d'un échantiiìon de N >
n valeurs de une 011 p h -
sieurs oaractgristiqueim cìimatoiogiques XI, U ...
Xk(pm 8xezaple :
averse d m a ì e annuelle, indice de pluies antk6dentee mtte
aveme, etock de neige sur le baasin au d h t de la fonte) N
contenant n i

(StJ - on cherehe à établir une r8grulsolon multiple (au simple) entre

11 et XI, X2 ...xk (ou seulement XI) :
 427

- on appïiqtm la régreseion trouvée aux Io-n valeur8 de XI

contenues dane la période commune de n années ;
...non
- on a ainsi une nouvelle e b i e de N valeurs de Y, plus longue que
la &rie originale de n valeurs, laquelle on peut appliquer
11analyse statistique.

OU B D N - on &ablit un modèle déterministe de transformation pluies-débits

(par exemple hydrogramme unitaire + fonction de ruissellement) ;
- on applique ce modele aux donn&s climatologiques XI ...,et on
procède comme pour lee données reconstitubs par régression.

On doit noter que, ni par une méthode ni par l'autre, on ne :ail;

de transposition ou interpolation gbgraphique. Les modeleis, qu'ils soient
régressionsll ou de structure déterministe, sont appliquh aux bassi116
mêmes pour lesquels ils ont et6 établis.

Le problème du gain d'information se pose dans les deux cas. I1

e ~ tbien évident qLe le nouvel Qchantilion de n valeurs de X observées
+ N-n valeurs de X calculés n'est pas équivalent, du point de vue quantité
d'information, à un échantillon de N valeurc de X observées, mais '1 un &chan-
tillon de NI valeurs, avec n < N u < N. Si on a procédé par régressions et
qu'on se soit m a n g é pour que ces régressions répondent à peu près aux con-
ditions suivantes :

- homsc8aasticit6,

- linéarité,

- distributions marginales n o d e s , en opérant b

ei changements
de vexiables ou dea anamorphoeesp le gain d'information, c'est-d-dire la wan-
tit6 NI-n, peut être facilement &du61 en comparant les variance8 des estima-
tiom.

Si on a utili& un modèle d&erministe, cette &elmtian n'est pas

imgdiate. Ii est nécetseaire de rechercher empiriquement la loi des écarts
rkiduels, ou la corrélation entre valeurs obeervbe et v e l e ~ ~calculées
W
428

en se baeent ~ u lee
1 n ann&s d'obmrvatioiiei conmnme0. il faut dire que bien
souvent on ne fait pas cette remherche et on se psisee de 1*6valuation des in-
tervallee de confiance.

L'avantage du modale déterministe, (notamment de ì'hydrolp.Emime uni-

taira, est de fournir la totalité de l'hydrogramme de a m e l donc simultané-
ment le débit maximal, le volume ruisselé et la fome. Taiidie que ces Biéments
(au moins débits maximaux et volumes) doivent &re étudiés séparbment par une
méthode strictement 1% régressionet'.

On peut élargir la notion d'extension des données en coneidérant non

plus In totalité de l@échantillon des Xi, mais des ensembles de valeurs de ces
données sélectionnée par des critères statistiques (averse de fr6quence donnée,
par exemple), ou bien cornidbée comme représentant dee situations partidi&-
rement défavorables eu égard au but poursuivi (précipitation maXimale probn-
ble, par exemple).

Cae& une attitude très répandue, qui correspond bien à la con-

ception moderne des crues de projet. Mais elle ne - J ~ B pas s- poser quelques
problèmes, ourtout quand on procède par hydrogramme unitaire. h effet, sur
lfenser;ibledes v:trLat?s Xi, une seule peut &tre introduite avec sa fréquence.
Supposons qu'on désigne par XI l'averse décennale et qu'on prenne pour X2 ...
X& des valeurs moyennes, ou médianes ; quelle probabilité peut-on attribuer
à la crue obtenue par application du modèle à l'ensemble des Xi ? On convient
souvent que la probabilité de la crue est la même que celle de l'averse.
C'est certainanent faux, mais dans quelle mesure ?

M. Beran, ciana la c o d c a t i o n qu'il noua soumet IlIltente de ré-

pondre a cette question. La e.ynthèse qu'il propose pour 1 ' W d r o g m m e e& tout
à fait classique. Le choix de l'averse est fait à partir de ia reìation
hauteur-dicrée-fréquence. La distribution des crues obtenues à partir de cette
averse est étudiée par une simulation effectub pour toutes les combinaisons
possibles d'un choix de

- 12 valeurs de 1~ durée de l@averBe,

- 3G 8 C h & ~de hyétogmrmee,
- 12 valeurs d'un indice d'humidité du bassin.
 429

La conclusion de l'autemr eat que, ai l'on utildm dee valeurs

m8dianes pour la durée de Le pluie, la r-ition de celle-ci au esin de
l'averse, le taux d'infiltration, la crue obtenue a une fr6quence voiaine
de celle qui a 6tB choidc pour la hauteur de precipitation. La forme àu
hyéto,p-amme ne semble pas jouer un grand Ale.

MM. Kitmaita et Haehimoto 171 exposW?i'au Japon on part de l'ana-

lyse statistique des prboipitations de deux jours pour lee petite bassins,
ou de trois J O W S pour les grands. Lee auteurs attachent une très grande
importance à la distribution de la ? h i e à l'intérieur de ces intervalles ;
l'étudc de cette ustribution sat rai'.? AU RB de temps horaire et ibdécri-
vent une méthode d'élaboration du hy6tograrame de projet qui met en jeu un
"facteur d'ag-an disse ment^^ et un hyétogramme dit lfreprésentatifl1qui n'est
autre qu'un h y é t o g r m e naturel observé lors d'une averse récente. Autrement
dit, on sélectionne une "formea*qu'on applique à la hauteur de pluie déter-
minée par l'analyse statistique.

Pour cette détermination, la période de retour choisie d b e n d en

fait du type c'e projet et des conditions économiques, sociales et politiques
dans lesquelles il est envisagé. La durée de cette période est de cinq à sent
ans pour un projet d'égout, de vi.?& ans pour un petit bassin urbdn, de
cent rms p o w un projet sur une gl<x..de rivière du Japor, et de d e w cents ans
s'il s'agit du calcul d'un évacuateur de crue pour un grand barrage.

Les modèles de transformation pr6cipitations-débits les plus c o u T m -

ment utilisés sont des modèles 1% réservoirs91 qui permettent d'introduire
aiebent des rhctions non linbires du bassin. On a &golement employé des
hydrograimes unitaires, mais seulement dans les bassins pour lcsquels on
peut penser que l'hypothèse de linéarité est acceptable.

Lee méthodes de traneposition &graphique, on dit a u s d interpola-

tion ou extrapolation gkgraphiquee, sant très nombreuses et vont de la siuple
comparaieon analogique de climat et de,gbmorphologie, H l'analyse factorielle
tra6 pousc8e des conditions de la genèse dee crues. De nombreux auteurs ont
Bt6 t e n t h par le eujet (ia moitié des rapporte da la queetion 11-21, mais
 430

se aont tous cantonnb daOs deux aspects particuliers du problème :la trans-
position de l'hydmgramme unitaire et l'utilisation de formules régionales
dbivées de la méthode rationnelle.
Noue rappelerone que, parmi d'autres, on peut considher la méthode
des courbes enveloppes comme une methode de tramposition géographique, quand
elle est assortie d'une "formule de r6gionalisation", c o m e c'eet le cas
pour les abaques de I'ranmu-Rodier (Cahier6 O.R.S.T.O.I., e w e hydrologie).
D'autres méthodee pourront 8tre conatmites H partir du catalogue des crues
exceptionnelles de l'U.N.E.S.C.O., lorsque celui-ci pourra enfin voir le jour.
De tels catalogues, lorsqu'ile comportent dee descriptions sufff-
a t e s dee cara& Bristiquee climatiques et gbmrphologiquee du baaseiin (sans
toutefoia trop compliquer lee C ~ O S O E ) constitueraient
~ par leur sede exis-
tence un outil de tout prender choix pour l'bvaluation dee crues en l'absence
de &m&s insuffisantee. Cseat m8me à vrai dire la seule choae qui actuelle-
ment fasse vrdnent ahfaut.
Rappelons enfin que la transposition des loi8 de distributions des
crues pourrait atre traitée BOU cette rubrique. Nous avons pr8î8i.h en parler
6 propos de l'analyse etatietique,davantage pour uno queetion de m&hodologie
que dans un souci de préeentation logique.

L'hydr-ogranme unitaire paraît $tre encore, malgré ses dé-

tracteurs, un inetment de choix pour l'evaluation des crues mar lee petits
baseins. Noue n'ailone pas ici emtemer une foie de plue une äi~cuseionam
la d6finition de ce derni- terme. Noue a m n e d6jà parlé de e m utilisation
à un mame batwin, il s'agit maintenant ae voir comment on peut trawposer
lee résuitate.
Cette traaepdition est eseentielle dans la m6thoùologie
ConcermELllt lee cruce dee petit8 bamuins. ûn 88 b u t e bien qu'il n'est pas
poesible d'entretenir des r b a u x de longue d u d e sur la totalité des petits
bassiris d'un pays. I1 n'est m8me pas toujours possible, pour chaque petit
projet, de mettre en oeuvre sur le bassin correspondant des observations
d'une densité suffisante, pendant une dur& suffieante p u r l'application
 431

de l'bydrogrme unitaire au bassin lul-m&ue.

Pour lee petits bassins, l'inauffisance, et m b e l'absen-

ce totale de donnbs au lieu d'utilisation, est donc la &gte. La préparation
des données hydrologiques pour les projets consiste donc à échantillonner
un certain nombre de basains, dits reprbentatifs, correepondaat B un nombre
suffisant de conditione climatiques et morphologiques. Ce erant lee résultats
recueillis sur ces baasins qui permettent la mise en oeuvre des différentes
méthodes de trampsition.

H. ~ ~ ä i e r , sa communication [Il] , présente ia m h o -

dologie mise en oeuvre par 1'O.R.S.T.O.M. pour les paye tropicaux. Cette mé-
thodologie est bask sur les résultats obtenus par l'exploitation, pendant
des durées égales ou supérieures A trois ans, de plus de lo0 ensembles de
bassins représentatifs. Elle ee rapporte surtout aux zones mahéliemes et
tropicales, mais des résultats sont ¿galement disponible8 pour lea zones dé-
sertiques et pour lee zonee équatoriales.

Les parametree sélectionnés pour représenter la forme de

l%ydrograme eant :

- le temps dû b u e (Tb) OU durée du d,esûllment,

- le temps de montée (GI,

- le rapport K du d6bit de pointe au d.bit moyen de lib-
drogranune de ruissellement.
Le volume ruisselé est évalué à partir de la hauteur totale
de l'averee par l'intermédiaire d'un coefficient de ruissellement..$i

L'analyse des rewiltats disponibles a permis de lier les

paramètres %,l& et i à certaines caractéristiques gbamorphologiques du
bassin, soit :

- la Burface du bassin,

- une clame de relief (R) d8terminb A partir d'un indice

de pente,
- une clame de pennbbilit6 6valub 4 lleetime.
 Lee relations sont prkentbee SOUS f o m e d'abaques. Suivant
l'&umbation ci-deesus, ces abaques ne tiennent pas compte explicitement du
rdle pourtant importaut de la couverture végétale. C'est que, d a m les régions
étudiée^, cette couverture v¿&ale abend eaeentielleiacsnt de la zone clima-
tique dane laquelle se t m w e le bmsin. C o m e les jeux d'abaque eont établis
par Bones climatiques, loensemble tient compte simplicitemciait dea conditions
de végétation. Dane lee cas particuliers, il cmvient au epéciaìiste d'gvaluer
ie ltcoupde pouce" à donner pour tenir compte d'une anomalie de ce8 cmditione.

Le coefficient de f o m e K est &du6 mivant les zones

climatiques, la eurface du baesin, mu8 forme de tableau.
Les abaques fournissent des valeur6 moyennes de6 parmètres
c,iLL ALI^^..: L.1x1 ir; q 3 v c - 'iverse äécr-~qal-.

communication Is] consacrée e o u t à ilinfìuence

ans une
du degré d'urbanisation eur les c m e ~des petits bassine, M. Hall propose
une méthode de r6gionalisation des hyärogrmmes unitaires. I1 part d'un hydro-
g r m e unitaire dimension en utilisant comme paramètre d'khelle des
-- *
le temps 'r retctrd l i (la?). TL est d o r s exprimé en fonction du
llrapportde bessin" Z O L /p, où 2 e& en km, L est la longueur du cour6
d'eau principal, en km, et S la pente moyenne du cours principal en % .
W. Eelliwell et Chen, dans leur c o d c a t i o n [4] présen-
tent 6gaìsment une m6thode de traneposition r8gionale b a h a r un bydrogrme
BBPB dimmion. Leur p m b l h e e& absolument typique du cbantp dgapplication de
la m6thode de l'bydmgramme unitaire. Les rivières de la colonie de Hong Kong
sont très nombreueee pour un ai pctff territah (1 o00 k d de terres) et la
taille de leure baeeina est hidemuent td6 reduite. Il n'est pas concevable
d'utiliser dane ces conäitions la fermule classique du réseau hydrologique.
Lea hydrologues de Hong Kong ont donc s6lectionub quelques
baseins reprbentatife dont ilpl ont 6tudi6 en détail le comportement hydrolo-
gique. Le8 auteurs d b i v e n t lee mathodee d'analyse utiliekm qui m n t d'ail-
leiir8 t r h claesiques, sauf que le pa^ de tamps t d e caurt nécesmire (15 am
ou mine) a créé quelques difncuitb par euite de la ree&¿ de6 enregistre-
ments pluviographiquem exploitables. L'hydmgramme a a n ~dimension eet obtenu,
comme chez Hail, en multipliant les ordannies par le Lag, en divisant les
 433

abscisses par le h g et Bll r ~ e le~ tout

t un volume unité.
L'analyse a conduit, pour chaque baeain btudi6, à un hydro-
graimne unitaire moy'en, dont le tempe de retard (Lag) a 6th mie en relation
avec l'indice L L, /r8, 0ii E est la longueur de la rivière principale, Lc
la distance le long de cette rivière entre l'exutoire et le point le plus
près du centre du bsrasin, 9 la pente moyenne du cour8 principal. Le Ia[:
a été mio auSei en relation avec la siarface du baeain, et cette régression
donne du r e d e u m meilleure corrélation (0,92 contre O,%).

Touteo les formulee prbentkei par les auteurs sont d&riv&s

de LIméthode mt;.onnelle.
EIM. Jarmwathma et Pinkayan dkrivent dana leur rapport [6]
les abthode6 de d c u l dee crues utiliebea en Thai'lande pour les petits bas-
si-. Apre6 avoir rappelé la pauvreté dea donnke ditiponiblee dane ce pays,
ils adreseent quelques crítiquea à ia formuìe rationnelle cl as rip^^
Q P C i A et lui préfèrent la fozmule de Mc Math Q p: A C i @/A) "
'
qui in-
trodult la pente du tasdn.

La c o d c a t i o n de M. Pereira 9 damie entre autres qual-

que8 indíaatione mur i'utilieation de la methode rationnelle au Erbil, notam-
ment den valeurs du coefficient de miesellement. L'auteur y donne &alement
dea rsnseignsme&e Bur le@ tsmpe de retour dOpt68 dane ca, pp d - r a E le
type de l*amhgement et l'erivimmemnt, a r l'intensité dee pluies au &&fi,
BUF lea P.W.P.,sur l*eatimation des volumes N i m e l b à partir deer prkipita-
tio- (fonmile du Soi1 Conservation Service).
M. Sokalov 12 hoque l'emploi de l'hydrogrmme unitaire
en U.R.S.S., d e fait une place plue large & la méthode rationnelle, ainsi
qu'à des foruniLee empiriques de la forme
 434

où sax
est le débit maximal spécifique en m3/s.km 2, q un paramètre qui exprime
le débit spécifique extrême lorsque la gurface A du bassin tend vers zéro.
-
C est en fait égal à 1 ; n varie de 0,15 0,30 pour une crue de fonte de neige
-
à O,5 O,7 pour les crues dues à de violentes averses locales.
M. Won [ 141 expose les méthodes utilisées en République de
Corée. I1 propose une formule qui procède .?i la fois de l'hydrogramme global
(sinon utilitaire) et de la méthode rationnelle :

- qo = C Y A R/T
dans laquelle g, est le débit maximal, qo le débit avant la crue, C un coeffi-
cient de forme de l'hydrogramme, 9 le coefficient de ruissellement moyen, A la
surface du bassin, R la pluie totale, T la durée de la crue. I1 propose d'autres
formules concernant le temps de concentration, la courbe intensité-durée, la
durée critique de la pluie (t = <atc). I1 traite en fin le problème des ri-
n
vieres ayant des affluents importants. La documentation fournie est importante;
peut-être l'auteur fait-il un peu trop confiance à l'universalité de sa
formule.

3.- QUESTIONS DIVERSES.

Quelques auteurs ont traité des problèmes qui sont bien dans le
sujet du symposium, qui sont relatifs aux crues, mais qui se rapportent à des
questions un peu marginales par rapport aux préoccupations du plus grand nombre.
Nous avons déjà signalé la communication de M. Hall [33 a propos
de la transposition de l'hydrogramme unitaire, mais ce n'est pas là que réside le
principal intérêt de l'exposé. M. Hall fait état d'observations qui ont été
effectuées dans une zone en voie d'urbanisation (bassin de la Mole, près de
Crawley, U.K.). Après avoir exposé ses méthodes d'analyses, dont nous avons
déjà parlé, il montre comment la relation entre le temps et le retard (Lag) et
le rapport du bassin (L/ S) évolue avec le taux d'urbanisation. Ce résultat
ne fait du reste que traduire quantitativement le notion physique intuitive
de l'aggravation de la forme des crues avec l'imperméabilisation du terrain.
 435

M. Rendon Herrero a choisi, pour sa c o d c a t i o n bo] un eujet

bien particulier. Il s'agit du transport de sédiments en suspension €tudi6
à l'échelle de l'averse. L'auteur met d'abord l'accent sur l'importance des
apports latéraux de sediments (Washload), mitit par l'brosion en nappe (sheet),
soit par le ravinement (gully) par rapport aux dát&iaux du lit mis en jeu
durant le transport. Lee r6sultatcs sont interpr8t6e par des techniques ana-
logues & celles de l'hydrogranmie unitaire (s6dimentopame unitaire). Une
application est faite au bassin de Bixler Run (U.S.A.)

CONCLUSION

I1 eet certes Intbeesant de mettre au point des méthodes d'analyse

de plus en p l w bborées pour essayer d'amocher le moina mal possible les
caract6riatiques des crues et des basses eaux, lorsqu'on dispoae de donnbs
rarea ou peu précisoa. Xais il ne faut pas trop se faire d'illueion BUT la
portée réelle de cette tentative, ni oublier que toute la confiance qu'on
peut attribuer d une eetimation rénide dana la quantité d'information , c'est-
&-dire finalement dane la masee et la qualité des donnéee diqmniblee. La c o m -
titution de cette infomation n'est pai3 te&&, il est faia de dire qu'elle
ne pose plus de probl8mes.

Ea matière de cruet3 par exsmple, ce qui fait le plus défaut danri la

plupart des paye, e o u t lorsque les r i v i h s y epnt difficiles, torrentielles
et inetables, c'est une bonne connei~aancedes débits dee plus grandes crues
connues. L'organisation d'un service hydrologique efficace n'est pae une petite
affaire : elle demande une grande compbtence, un soin de tous les instants et
une certaine aportivité. Elle demande awai de l'argent et c'eet L4 que r b i d e
souvent la plus gnrnde difficult).
le
Notse conolwion aera donc que/meiUeur moyen de suppleer d la ca-
rence dee donube hyàrologiques est encore de s'attacher à la Buppremion, ou
tout au moins d la diminution de cette carence.
436

-
111 M.A. Bleuw (kglanä)
E.timation of d e d p floods and the problema of equating the
probability of rainfall and runoff -
R. DAVIS,bCKSTFZN, C. KIISIEL, N. Fo(zEL (U.S.A.)
-
A decision theoretic approach to uncertainty in the return
period of maximum n o w volumee using rainfall data -
131 M.J. HALL (U.K.)
Synthetic dthydrograph technique for the design of flood
alïepiatlon works in urban areas -
[4] P.R. EEUWIEL, T.H. CEW ( ~ o n g - ~ o z y )
A dimeriPiionlEss unitgraph for Hong-Kong -
[5] P.E. HERBBT, S. VAN BiLúûN, J.P.J OLMER, J.H. HAIL (South Africa)
Flood estimation bp determination of regional parameters from
-
r
limited data -
[GI D. J W A T H A N A , S. PINKAYAN (Thailand)

!l%arland -
Practice8 of design flood frequency for epiall Watereheds in

171 T. K0IWSITA, T. HAsHuIoTo (Japan)

- Design diecharge derived from design rainXall -
[8] W.N. LEEBE (U.K.)

-L91 The u m of asmoreci data in estimating T - y ~ a rfloode -

P.P. Pm!uzl?A (Brasil)
Amesment of deeign noode in bradl -
- o. RIBIDONmmmo
Eo] (U.S.A.)

wateraheäm -
A method for the prediction of W o a d in certain d l

FI] J.A. ROD= (fime)

-
Méthodes utilidee gour l'kaiuation dee dbits de crue des
petits cour6 d'eau en r8gione tmpicalee
 437

[la I A.A. SOICOIDI7 (U.S.S.R.)

Methods for the estimationa of meximum dischargea of snowmelt
and rainfall water vith inaàequate observational &ta -
I l q A.M. VLADMIROV, A.1. CHEBOTARGv (U.S.S.R.)
Computation of pmbabilietic values of lot flow for ungaugeü
rivers -
T.B. WON (Korea)
A study on maximum flood discharge f o n d e e -
 E S T I M A T I O N OF FLOODS B Y MEANS O F T H E I R S I L T L O A D S

Modesto Batlle G i r o n a
Dr. C i v i l E n g i n e e r

ABSTRACT

An empirical a n d e x p e r i m e n t a l f o r m u l a o f very s i m p l e
s t r u c t u r e is studied, to o b t a i n €he flows o f maximum f l o o d s in
r e l a t i o n to t h e sediment l o a d s that t h e floods produce, d e p e n -
ding only of the maximum s i z e o f a r i d i t i e s o f t h e channel.
T h i s f o r m u l a can be u s e f u l t o study also t h e b e h a v i o u r o f t h e
river bed, alluvial v o l u m e , a n d so on.

Se estudia una f ó r m u l a e m p í r i c a y e x p e r i m e n t a l de e s
t r u c t u r a muy simple, para o b t e n e r los caudales de m á x i m a s cre-
cidas e n f u n c i ó n de los a r r a s t r e s que éstas producen, depen-
diendo Ú n i c a m e n t e d e l t a m a ñ o m á x i m o de ’aridos d e l cauce. E s t a
f ó r m u l a puede s e r Ú t i l t a m b i é n p a r a e s t u d i a r e l c o m p o r t a m i e n t o
d e l l e c h o de los rios, v o l u m e n de acarreos, etc.
440

ESTIMATION OF FLOODS BY MEANS OF TIIliIR SILT LOADS

Based on the physical fact that every flood deposits

a mass of arids whose maximum cliametres are proportional to
the magnitude of the flood, by means of a reciprocal process
a n attempt was made to find a way of estimating tne discharge,
obscrving the silt loads produced,
Accordingly a very simple network formula has been
obtained. In a series o f 15 tests, the prevision of the
maximum floods that have occurred could be made (correspond-
ing to a return period between 100 and 500 years) WITH Ah'
LRKOK bELOW 13%. To do so, one merely has to know the maximum
size of the river-bed arids.
Besides being a new instrument to calculate floods,
this formula may, as indicated iii the "Summary" , open up an
interesting field of investigation regarding mobility of the
river beds, volume o f bed-loads, etc,
1. - FORMULA f'llOPOSEI3
1.1. WORK SCHEME, -
O n the one hand, the maximum silt load diametre
is function of the flood Jischargc.
On the other hand, the arids are moved by the
force o f the silt load, wnicii is proportional to the gradient
arid to tho draft.
Considering the above two factors, a formula was
sought which related the diametre of the deposited arid, with
the draft and gradient. The mathematical deduction o f this
relation is however inaccessible and a semi-empirical formula
was sought, verifying it and deducting the unknown values of
same, by neans o f experimentation.
Once a relation was obtained between the diametre
of the arids, tlie.gradiarit aiid the draft, this could be
deducted from the previous ones, thus defining the maximum
level obtained by the flood waters. Since the bed-section
is also known, the discharge of the flood which has borne
along the arid through this section, depositing it immediately
downstream, can moreover be obtained.
 441

Once the purpose of the study was specified,

a formula liad to b e proposed which would relate maximum
diametre-draft-gradient. The probing was systematized,
and the proposed formula was verified and as already
mentioned, the unknown coefficients of same wcre verified
with a series of 15 samples or tests. The margin of error
obtained was found and compared with other existing methods,
The return period of the flood-waters calculated with the
formula, was sought, defining an inferior limit, The possible
limitations o f the formula due to the petrography of the arids,
the morphology of the basins used or the non-existence of
certain sizes of arid, were studicd. Finally, the conclusions
drawn are summarized, All the documentation involvcd in the
tests, regarding diametres of arids and drafts observed, was
collected photograpliical ly .
1.2. - MATilEMATIC OBTENTION OF TiíE PROPOSGL) FOIMULA
Une tried to reach a formula, deducing it mathe-
matically from the silt load force equàtions, but the
influence on the larger arids cannot be defined quantitatively,
nor can the protector inter-action which the silt loads offer
between themselves, in the face of the dynamic thrust of the
current.
Various hypothesis were used, but the subsequent
elaboration did not crystallize into any practical formula.
Ori the other hand, adopting one or another hypothesis as base,
produced inadmissible differences of above 100%.
In view of the above, it was decided to employ a
semi-empirical formula, Its structure was obtained matliematically
but it has been verified and defined from the experiments made.
1.3. - UL.1)UCING A SEMI-EMPIRICAL FORMULA
To obtain the formula in question, various methods
were applied: a) .-
considering the dynamic thrust on the arid;
b).- balancing the silt load forces, The liermanek aiid
Manning formulae were likewise used in one case or the other
to determine the mean velocity,
When using the Manning formula, the possibility was
considered of the rugosity of the bed 'In" b e in g prop or t i ona 1
to one sixth the power of the arid diametre. ïliis is correct
in canals, but ir1 natural beds, the most accepted formulae of
river hydraulics (Ilermanek, Christen, Wiiikel, etc.) do without
the rugosity or adopt a constant value of same.
Equations were reached through different channels,
with identical structure:
Ila = -0b
.
u. 1
 442

but in which the exponents a and b varied in terms of

the velocity distribution law, adopted. The degree of
parabolic speed distribution i s normally one seventh,
and it was with this value that a arid b were calculated.
The values of "a" and "b" were also deduced in the hypo-
thesis of supposing a one ninth degree distribution.
ïiie results were:
Degree of the speed
distribution parabola 1/9 1/7
E X P O N E N T a b a b

Hermane k 1,28 O,78 1,21 0,71

Manning: n = Cte 1,11 0,78 1,OS 0,72
Manning: n = Cte 010 1 ~ 1 1 1#11 1#O5
Manning: 11
Manning: 11 -
= Lte
Cte 0 1/6
1,13
1,13
1,OO
1,33
1,O8
1,08
1,OO
1,33

Nevertheless, the empirical formula proposed and

00,s
. 11 -
verified with tests, was, as will be seen later on:

u.i
The formulae obtained in the previous table (where
the later experimental definition of B) is required,)
iiave no faithful tradition in the reality of the river
beds, as they give different values to those of the
said formula where "a" =
The most approximate

In this, m = 7.
1.4, - DLFINITION OF THE PROPOSED FORMULA
T h e formula we are trying to verify, w k e b y o n
e x p e r i m e n t a m obtaining the value of B, this value
should be practically constant, was:
1/2
ii = am
13. i
 44 3

II = Draft expressed in centimetres,

flm = blaximum diametre of the arid in centimetres,
i = Gradient
ki = Coefficient to be defined.
I n each test, knowing the draft 11, obtained directly
or from the discharge, in a sector, wliicli was termed "control",
tlie 0 o f tlie arids deposited in the area, if possible
immedyately downstream of same, was defined, calculating:
i/ 2
B = am
li. i
The series of 15 tests undertaken, permitted a verification
that coefficient B is almost constant; its value could be
calculated, and at the same time tlie exponent 1/2 of 0m was
found to be most suitable.
2.- RESULTS OBTAINED, -
2.1. - MET1iOL)OLOGY
First of all, a "control section" must be defined. Knowing
the maximum historic flood in a prudential period, the draft
corresporiding to H was calculated in it. This iì of the control
section, was in certain cases measured directly after some
important flood, through the traces of undergrowth and residues
that the current left in tlie river bed shrubbery T h e gradient
0.

i o f the span corresponding to the control section was sought,

and the maximum diametre 0 of the arids deposited downstream
in this section was meacurgd. With this information, the
following was calculated: 1/2
B = m'k
tl, i

Control Section:
The control section should be sited in stretches with as
uniform system as possible. Consider the influence o f bridges,
etc. The arids should clearly define 6,.
Gradients:
The "i" adopted, is that o f the stretch from 1 to 1,s kms.,
immediately upstream from the control section, It is obtained
from plan 1:50.000 of the Geographic and Cadastral Institute,
rounding off, if iiecessary, the excessive twists in the
longitudinal section.

-
Draft:
The "H" draft is measured from the lowest reading o f tlie
control section,
 444

izlaximiim diametre:
T h e arids probe area for defining Ø m will be chosen
downstream the "cnntrol section" and as ncar to this as
possible, so that the sizes observed are effectively the
largest that have passed through this section,
'To define the maximum diainetre, only those rounded
o r parallel-epipedical shape arids will be suitable , whose
smallest dimension is at least 2/3 (two thirds) the largest
orie, The 0 value will be the mean o f the largest two
dimensions of the arid.
The suitable arids for defining 0 should be found
with a minimum density of 1 per every pwo square metres.
Uy density, we understand the number of units in sight,
per river bed surface, I n some cases, this may even drop
to 1 per 4 square metres.
T h e choice o f arids o n which the 0 i s going to be
measured, demands a careful, critical judgement o n same,
considering the possibility of i t coming from the erosioned
sides and not upstream, that they may pertain to demolished
works, or under construction, etc, Certain geological know-
legge of the area will always prove most useful. Any kind
o f rock excepting slate is suitable.
It should be emphasized that in minimum densities ,
relation 2/3 o f dimensions, etc., a qualitative common sense
should always preside over ali rigorist criterion.
It is important to take photographs with scales which
will act as referenee so as to compare the field observations
at the office.
One must remember that the problem consists in a trans-
port through the control section of the arids, which will be
deposited immediately afterwards, and that they may even be
covered by other finer ones, which have settled when the flood
waters dropped.
Qu a 1 i ty :
T o have a p;rudeiit judgement of the suitability of the
tests made, they have been classified into GOOD (G), MEDIUM (M)
and FAIR (F) , in accordance with the guarantee deserved by
the definition o f the data obtained li, am and i.
pemarks :
1.- In low river-bed spans, with very slight gradient, it
is often difficult to define this on plans 1:50.000
and ori a smaller scale, the bottom oscillations are
excessive. In these cases, the river is also usually
 445

very wide, and the transversal section presents siiarp

relative off-levels. I n this case, the definition of 11
must be closely examined, to avoid falling into errors.
2.- In some cases, owing to the type of alluvial terrace
in wliicli the bed is fouiid, the prehistoric and millenary
arids cannot be differentiated from those corresponding
ti1 the latest historic floods.
2.2. KLSULTS OF TIIE TESTS.-
Below, the data i, 1i and id obtained in each test, anci
the resulting value of B are given:

Test Nomenclature and Site i 11 0 yual Coeff

;.i O ím) (CH) ity icient b
1 K. Llémana at crossroads 0,00822 2,81 18 M 1,84
Main road S,Martfn de
Llémaria
2 R. Llémana in Sta, Afra 0,00590 3,45 19 li 2,13
3 ‘ler in S.Julián de Kamis 0,00725 3,73 29 F 2,OO
4 Ter in S.Julián de Kamis 0,00725 5,63 60 I: 1,90
5 Ter in outlet of the Dar6 0,00100 7,93 25 F 2,OO
Uñar at crossroads N - I I
6
-
Madri d F r an c e li ig hw ay 0,00300 5,42 10 G 1,94
7 R.Uerneda at bridge Riu-
dellots,maximum flood 0,00237 5,lO 8 G 2,33
8 K.berneda at bridge Riu-
dellots, flood 11-X-70 0,00237 4,OO 5 G 2,05
9 Tordera in Sai1 Celoni 0,00835 2,52 16 G 1,90
10 Tordera in Tordera 0,00320 2,85 4 G 2,20
11 R. Rifer in San Celoni 0,00120 1,05 7.5 G 2 ,16
12 Corigost in La Carriga 0,00880 2,90 28 M 2,07
13 Cardoner in Manresa 0,00435 4,65 18 G 2,09
14 Llobregat in S.Vicente
de Castellet 0,00345 5,85 22 G 2,32
15 Llobregat in Martorell 0,00209 7,65 10 G 1,98
 446

2.3.- VERIFYING THE EXPONENT OF QIm.-

In principle, the value 1 was adopteù as exponent of OmD
proposing tiie formula: z
li
am
0,s
-r
ìiowever, the value 0,s was later esteemed and its worth
confirmed as the tests were made, owing to the scanty dispersion
presented by the values of B obtained.
The possibility of the dispersion of the resulting 13 being
l e s s with another exponent , is consequently likely.
dased on the values i, li and 0,,, obtained in the tests, the
of 0,,,: 0,4ü 0,45 -
b coefficient has been calculated
0,50 - -
for the following exponents
0,55 and 0,67.
A, The typical derivations obtained are:
for B = 0, 0140/i H - T = 0,195

e
E for U = ,0 i 11 - G = 0,175
o
0.55
for B = 0m H - G = 0,143
u 0.50 for U = 0, li - TT = 0,189
8 0.45 for B = 0, OaU7/i H - o = 0,484
J
0.40..

0.35
. 0.2
- Desviación típico
0.3 0.4 0.5
1 7
0.1
In the figure, the aforementioned results are represented,
and it can be seen how the minimum typical deviation corresponds
to the 0,s thus verifying that this value is the most suitable
as exponent of 0
,.

2.4 - CALCULATING THE COEFFICIENT U. -

Having obtained the value of U for each of the 15 tests madc,
what is proposed for the famula in question must be defined.
In order to consider the quality of each test in the determin-
ation of B, by means of a prodent mean, the ones qualified as good
(E) are assigned weight 3, the "mediumtv (M) , are given weight 2,
and the "fair" (F) , weight 1. Tnus we get:
B = 2,08lY2,08
 447

i3

Whereby the formula proposed will be:

-
The value of the coefficient L4 adopted is:
2,08

2,08 .i
li = maximum draft in centimetres (see 2.1)
Bm= maximum diametre in centimetres (see 2.1)
i = gradient (see 2.1)
2.5 SIGNIFICATION w r E OF TIE TEST SERIES.-
The "signification rate" of tlie test series made, or in other
words tlie quality of the whole ensemble of same and the mean ii =
2,08 was obtained. Accordingly, tiic likiliood of another value
of B, obtained as a result of a new series of tests, being
within specific limits, was calculated.
- (mean) aiid ';r (typical deviation) be the monthly
c h a r a c b g i g t i c s of the series of Values of B obtained; tlie number
o f representative tests is n = 15.
I n tiie interval of possible values of B included between:
cr
and Y + tp -.
if we choose a percentage of probability P (ex. p = 1%) where
tiie value o f the medn U of a new series of tests is outside these
limits, in the table of function of Student for this 1% and n-1
= 14 degrees of freedom, a value o f tp is obtained with which
the above mentioned "confidence interval" is defined. The

-=
probability p is the "signification level".
x 2,OG; q = 0,143

- = 0,03823
* ,

m
According to tlie Student t8t" function table, we get:
 448

sigriification Value of tp G
Confidence limits
level p for 1 4 degrees t p . m Be 1 ow Above
of freedom
X*tp.fl=$y
i ¿ - t p .Gm - o-
~~~~~ ~ __ -~ ~~~~~

0,l % 4,140 0,16 1,90 2,22

1,o % 2,977 1,11 1,95 2,17
2,o % 2,624 0,lO 1,96 2,16
5,o % 2,145 0,08 1,98 2,13

Thus, the probability that the mean of a new test series

is between 1,95 and 2,17 is 99%.
Let us set this conclusion out in terms more befitting
our problem.
The confidence interval between 1,95 and 2,17 admits a
possibility below 1 %, that the 13 obtained in a new series is
1,95; this represents a discrepancy of 0,11 in respect of the
mean of 2,OG of the experimented series.
o 1 1 *'Oo'
2,06
= 5I B 3 %
This difference of 0,11 means an error of
in the appreciation of the drafts. Let us see how much this is
?illen translated into flood water discharges:

and if the section is approximately rectangular T I1 , Whereby:

Qi = 0,34 li Ob7' ,i1l2,ki.b = 0,34 b.i 1/2 ,1i1,75

Thus an erro,,i2 draft of 5,3% multiplies the discharge

obtained by 1,053 = 1,095 which means a n error of 9,5%.
Therefore, the quality o f the series of tests made and
consequently that o f the B value adopted, can be defined as
follows:
The probability that in a new series o f tests, the value

-
of B defined for the formula proposed means an error in the
determination of discharges, below 9,5%, in respect of those
obtained with B 2,08, is 99%.
 449

2.6. - MAXIMUM ERROR A N D COMPARISON WITH OTIIER METIfODS. -

In practice, to determine the maximum historic flood,
applying the proposed formula, a certain number of tests
should be made in certain other control sections of the bed,
and finally obtain a mean, In this way, the inevitable errors
and discrepances will be compensated for, and which will take
place on defining the gradient (i) and in particular the
maximum diametre (e,) with which to enter into the formula.
The number o f verifications will depend on the exactness
to be obtained, and on common sense, in face of the data
defined in each "control section". O n an average, four tests
may prove sufficient.
Supposing that all the tests made correspond to a same
bed, we can find the maximum possible error by mixing
together the results of the whole series.

-
The five tests with highest B values are 2,33
2,20 -2,16 2,13; the mean is B 0 2 ~ 2 3 .
- 2,32 -
1,34 -
1,98 2,OS - -
Similarly, the 5 tests with lowest B values are 1,90
2,09; the mean is B = 1,99.
-
I n both series, those tests classed as "medium" (M) and
"fair" (F) have been omitted,
The difference between these means and the value B = 2,08
o f the formula is:
2,23 - 2,08 a 0,15
2,08 - 1,99 = 0,09
Considering the most unfavourable case where the tests
have given the 5 highest values o f 8, whereby their mean would
be B = 2,23 instead of U = 2,08, this means an error in draft
;.tW
appreciation of O 15 .lo0 = 7.2%.

which means a 13% error,

1 , 0 7 2 ~ ~1,129
~ ~ -
As seen i n 2.5, this indicates that the real discharge
has been multiplied by:
1,13

To conclude: THE MAXIMUM ERROR OBTAINED ~ViiliN ESTIMATING

'ï1ii.i IIISTORIC FLOOD DISCHARGE, ACCORDING T O VERIFICATIONS MADG
WITH THE PROPOSED FORMULA, IS 13%.
Other calculation methods:

Wit11 t h statistical method, the errors for return periods

 450

above 50 years, may be around 20 to 3090.

Specifically in the floods study of the Congost river
made at the liydrographic Confederation of tlie East Pyrenees ,
by the Measurerncnts Service, using the historia method, and
for a 50 year return period, a discharge of 160 m3./sec. is
obtained with the Gumbel method and 135 rn3./sec. with another
law of distribution, which means a difference of 20% between
iaoth methods. Applying the tational method to the same study,
biie gets a discharge of 305 m3./sec. for the same return period.
2.7. - RETURN PERIOD
The return period of the floods of the various tests
w a s also studied, to try and relate it with the discharges
foreseen with the formula, and at least obtain a lower limit
of samc.
T h e return period of the floods for which the formula
has been verified is:
Test 1 -----__----_-_- 400 years
Test 2 -------_-_---_- 400 years
Test -_----_----_--- 95 years
3
Test --------------- 95 years
4
Test 5 --_------_---_- 9 5 years
Test 7 -_------------- >70 years
Test 8 -__------_----- > 7 0 years
Test 9 ---------_----- 9 0 years
Test I l - - - - - - - - - - - - - -1000
- years
Test 1 2 - - - - - - - - - - - - - - -160 years
Test 1 3 - - - - - - - - - - - - - - -180 years
Test 1 4 - - - - - - - - - - - - - - -180 years
Test I S - - - - - - - - - - - - - - -180 years
In face of these figures, it would appear that the lower
limit of tlie return period is 100 years,
ilowever, it must be remembered that this conclusion is
based on a series of i3 tests.
A s indicated in 2.8, the influence of wear throughout time
is very scarce and does not change the return period of the
floodwaters,
I t can therefore be said that the fbod obtained generally
has a return period between 100 and 500 years.
 451

2.3. - STUDY OF TIIE POSSIBLE LIMITATIONS

Due to the petrography of the arids observed and their
possible erosion:
There is a possibility that the determination of the
maximum diametre d has only been made with certain rock
types, as the othefs were excessively worn by the erosion,
I f this has occurred, in other areas where there are no
arids of the most resistent types, false results of 0,
could be obtained,
T o approach this problem, the petrographic classific-
ation of the arids was made, based on the photos obtained
in the determination of the 0, of cach test, and which
defined 0, with the following symbolics:
A - Sandstone Co - Conglomerates Gn - gnes
B
C
- Basalts
- Limestone
G - Granite and
eruptive rock
P
Q -- Slate
Quarzite
The types of arid used in cach determination of the Ibrn
were:
‘rest 6 ------------ 2A 2G 3Q
+ +
Test 7
Test 8
-------------A + B + 2 C o + G + P +
------------ Q
2G 3B +
Test l o - - - - - - - - - - - -(2A + B + CO + 2G * 3 4
(A + B + G + 4Gn + 2A.
Test I l - - - - - - - - - - - -íG
+ 34
(3A + 4 4
Test 12- ----------- 2A + 2G + Cn + 2q
Test I S - - - - - - - - - - - -SA + 2C + 2G + 3Gn + Q
Test 1 4 - - - - - - - - - - - -2A + 4C
Test 1 5 - - - - - - - - - - - - SA + 14C + 2 4
Test l b - - - - - - - - - - - - (2A + 2C + Gn
(2C + P
Making a tctal used of: 26A + 13U + 32C + 3Co + 16G +
+ 9Gn + 22Q

In view o f the above results! evidently the rock type

does not influence the determination o f the diametre, since
all the classes appear as maximum arids (8 ) in considerable
number (except the slates which will be dipcussed below) and
it can therefore be supposed that the erosion, for the effects
o f the method adopted whep determining the 0, iii a way affects
any type of m t k .
452

This is because these arids are not from the bed downstream
and they consequently only suffer the effects of the river
erosion with high waters or normal flooding, which take place
intermittently arid not very frequently. On the other hand, the
large adjustment obtained in the value of B makes one suppose
the influence of the erosion in tlie various arids could not be
important.
It must however be emphasized tiiat the slate, as definers
of tlie 0 only appear twice, as a logical consequence of their
greater tensitivity to the environment, as is deduced from the
above table. For this reason, what lias been said in the above
two paragraphs cannot be applied to the slate arids.
We can consequently say that the definition of Om, is
independent of the arid petrography except in the case of slates,
which should not be used for the determination.
ihe to the region studied:
All the tests made have been in the provinces of Barcelona
and Gerona. Therefore extrapolation to another type of basin
could present doubts,
In the enclosed table , the chief geographic characteristics
o f the tested basins are indicated.
However, it should be stressed that the only thing that
can change the validness o f the proposed formula, is the arid
which defines 0, and according to the above paragraph, it has
been considered that the formula is valid for any type of
rock (except slate).
On the other hand, although not definite, it is most
significant that in a later test made in the Guaro river of
tlie basin in Southern Spain, near Vélez-Malaga, the value of
the U coefficient obtained is:
=
whereas that adopted is B - B 2,14
2,08 which reprecents a 3% error.
Although not conclusive, this result opeiis up a hopeful field,
awaiting an extension of the tests to other regions.
üue to the lack of arids:
Thecase may arise of there being no arids of a diametre
superior to a certain size, as they do not exist in the bed or
because they are retained by some dam or weir. The lack of such
arids does not produce any change since the floods correct the
gradient o f the river according to the existing sizes. In short
stretches where the local effect of a weir modifies tlie gradient,
this should be taken into account,
b
 453

The formula should be applied to river-beds whose

possible mobile bottom during the flood later permits the maximum
arids deposited by the flood peak to be discovered.
If a mobile bottom of considerable thickness is produced,
by means of burrows, the thickest deposits made by the flood
peak should be reached, aiid if the sediment thickness is very
important , the gradient adopted should be corrected. However ,
none of this kind have been experienced in the tests.
45
d
0 V
3- +
a
fl
3
x
+J
.rl
rl
O
a
U
t.4
o>
>
.FI
d
O
4
+J
VI
P,
E-
 455

3.- SUMMARY
To obtain the draft (H) that has been produced in the
maximum historic floods, knowing the size o f the arids (Om)
which may have been hauled along by this flood-water,
across a section termed "control", and the bed gradient (li)
the following formula is proposed:
o, 5

which should be experimentally checked when the coefficient

U is defined.
The above formula could not be obtained mathematically;
The most reached expressions of type:

the one nearest the proposed one is that where a

0,67.
- 1 and b =
To define U, a series of 1 5 tests was made, obtaining
B = 2,08.
There was a possibility of the proposed formula not being
correct, which would occur if the value of B obtained in the
tests w a s variable. However these values ali varied around 2,33
and 1,84. The typical deviation of the series was = 0,143,
The series also helped to contrast the favourability of the
exponent of 0 , since the 0,s produces the minimum typical
.deviat ion.
The "signification level" of the mean of the series of
tests made, corresponding to the "confidence interval" between
B = 1,95 and B = 2,17, is 99%. Expressed in other terms, it
means that the l i k g m d of another value of B, defined by a
new series of tests, having ari error in obtaining discharges,
less than 9,5% (regarding those obtained with B = 2,08) is 99%.
THE MAXIMUM ERROR IN DETERMINING DISCHARGES, ACCORDING T O
TIIE SERIES OF 15 TESTS, IS 13%.
The formula proposed is therefore:
O m a~5
li = 2,08.i
li - according
Draft of the maximum historic
to 2.1.
flood in crns. defined

0,- Maximum diametre of the arid in crns. defined according to 2.1.

456

i - Gradient of the bed, defined according to 2.1.

2,08 - Coefficient with dimensions L -l/Z ,
Tlie methodology to define these figures is indicated
in greater detail in 2.1.
The return periods of the floodwaters estimated with
the formula are generally between 100 and 500 years,
The control section is that in which i, II and the
silt loads Ibm which have crossed it, are defined.
In the river bed, whose discharge one wishes to define,
a series of tests will be made in accordance with the
precise ùegree of exactitude, and the guarantee that the data Bm
and i offer. An acceptable number may be four.
The formula is valid for any type of arid, except the
slates which should not be used to define firn.
Tlie formula was applied to beds whose mobile bottom
during the flood was sufficiently scarce to permit the
maximum arids deposited by the flood peak to be later dis-
covered. If this mobile bottom leaves the arids correspond-
ing to the maximum discharge hidden, by the smaller arids,
work may be done as indicated in 2.8, but in this study it
was not necessary to experiment with buried arids.
The 15 tests were made on the Catalan slope, The formula
also appears acceptable in other regions, but it has only
been verified with a test in Malaga.
This study does not pretend to have exhausted the
subject, but merely initiatesa new field of operations.
The series of tests can be expanded, The value of B can
be adjusted more in accordance with the variations of am
and i. The case of bed with far thicker mobile bottoms
may be studied , during the flood as indicated in 2.8, analys-
ing the buried arids. The observation of the arids need not
be restricted to the surface, By means of burrows the
diametres of the arids of the lower layers can be obtained,
thus extending the period studied. It may be used to measure
floods, previously photographing a panorama of the river bed
arids, with sufficient detail and after the flood water to
be studied, with another new photograph, establish the size
of the silt loads contributed by it. The dimensioning of the
protection rockfills of the river bed is defined with the
proposed formula arid for the slopes, the pertinent corrections
need merely be made, In terms of the draft of each flood,
the silt laden arids cari be foreseen and with the granulo-
metries of the bed, the sedimentation volume can be defined.
The longitudinal section can be studied in terms of the
 457

ADDITIONAL NOTE: Afìer the present doctoral thesis w a s

approved, the author continued making a series of tests
in various points in Spain, obtaining the following results:
1.

2.-
- Guadalquivir basin (Jacsi)
-Kiver Guadalbullón in Mengibar;
Ebro basin (Calatayud)
B - 2,lO

-River Jalón in Cetina; u 2,lO

3.- Ebro basin (Calatayud)
-River Jalón in Ateca; B - 2,lS

As can be seen, these values, added to the one obtained

in the basin in the South of Spain (Malaga) in the river
Velez Guaro, with B = 2,14, make solidly based hopes arise
that the formula is applicable for all types of basins.
A t the same time, it brings the number of tests made
up to 19, verifying t h e proposed formula,
458
ESTIMATION OF DESIGN FLOODS AND THE PROBLEM OF EQUATING THE PROBA-
BILITY O F RAINFALL AND RUNOFF

M.A. Beran
Floods Stuies Team, Institute o f Hydrology, Wallingford, Berkshire,
England.

ABSTRACT

Where data on river discharge are scarce it is a common engi-

neering design practise to concoct a design flood with the aid o f
rainfall depth-duration-frequency information and a catchment res-
ponse model. Two major waknesses o f this approach are (.a) the pro-
blem of the sensitivity of the design to legitimate changes in the
design assumptions and (b) the uncertainty of preserving the nomi-
nal rainfall return period in the design flood. A solution to the-
se problems is proposed which makes use of a computer simulation
investigating the sensitivity of flood magnitude to variations in
return period, storm duration, temporal rainfall intensity pattern,
infiltration l o s s rate, base flow and unit hydrograph shape. An e s
tension to the sensitivity analysis allows an estimate to be made
of any quantile of the distribution of flood magnitude based o n
sampling across all causative rainfall and antecedent conditions.

RESUME

El est courant, lorsque les données sur les débits sont insuf
fisantes, que l'ingénieuc élabore la crue de projet 'a partir de
l'information qu'il possede sur la distribution des pluies, en u t i
lisant un modèle de transformation pluies-débits. Les deux inconvk
nients majeurs de ce procédé concernent (a) la sensibilité de l'am$
nagement a la variation des paramètres du projet, (b) la conserva-
tion de la période de Tetour (ou de l a probabilité) lorsqu'on passe
de la ptuie de projet a la crue de projet. L'auteur propose une s o
lution a ces problèmes, en utilisant une simulation pour recher-
cher la sensibilité de la grandeur de la crue aux variations de la
période de retour, de la durée de l'averse, de la configuration du
hyétogramme, de la capacité d'infiltration, du débit de base, de
l a forme de l'hydrogramme unitaire. Une extension de cette analyse
de la sensibilité permet d'estimer n'importe quelle quantité de la
distribution des crues, en se basant sur un échantillonnage des
pluies et des conditions antécédentes.
 460

3. INTRODUCTION.
Modern engineercg pract<se requires the investigation of the likely
behaviour of a proposed construction under extreme conditions. Because scarcity
of data is the rule rather than the exception in the hydrological field it is
commonly necessary-toresort to indirect design procedures, for example the use
of rainfall probability information with a catchment response model, to derive
a design flood. One very common design procedure uses depth-duration-frequency
information to construct a design storm of some nomind return period which is
transformed to a design flood by means of a syntktic or observed unit hydro-
graph. Although conventionally the design flood is associated witb the same
return period as the storm which caused it it is clear that many sources of
variation which might influence the return period, such as those due to storm
duration, temporal and areal distribution of rainfall intensity, înfiltration
losses and base flow- are in fact ignored. The design engineer, in recognition
of this fact must make stringent ssumptions about these variables in order not
to inadvertantly reduce the return period of the flood.
It is relevant, therefore, to enquire into the effect of the ignored
variation by posing such questions as (a] what is the expected flood foìloning
the T-year return-period rainfall?; (b) what is the sensitivity of the design
flood to alterations to the assumed values of the variables?; (c) is there a
single combination of values that can be assumed fn order to approximate the
expected flood?
An approach to the solution of such problems is offered in this paper.
A simulation technique is described which samples the possible ways in which a
rainstorm of T-year return-period can cause floods, and derives their probability
distribution. (Section 2).
While this sensitivity aspect is relevant to current design practise,
perhaps of greater interest is the possibility of using simulation technique to
estimate the flood magnitude-frequency relationship*. The proposed method must
therefore be extended to consider floods resulting from all maximum
rainstorms and not merely from those of the numinal return-period (Section 3).

---I- -- _
--.
_ ~ l--ll_._.I_-I . _--"^.".----.-_-
* Nash ( I ) postulated a flood design procedure based on combining the incid-
ence of flood producing factors. Chow and Ramaseshan (2), Evans (3) and Dyck and
Kluge (4) present techniques conceptually not dissimilar to the one presented
here. Recent work by Eagleson (5) and Leclerc and Schaake (6) approach the same
problem in part analytically.
 461

2. THE SAMPLING PROCEDURE.

The procedure follows closely the steps used to estimate the design flood.

(a) Determine a nominal return period

(b) Choose a storm duration and calculate the total depth of

rainfall from the depth-duration-frequency relationship.

(c) Distribute the total rainfall within the duration to form

the gross rainfall hyetograph.
(d) Subtract from this an infiltration loss to form the net
rainfall hyetograph.

(e) Convolute the net rainfall hyetograph with the unit hydro-
graph to form the design inflow hydrograph.

(f) Process the inflow hydrograph and extract the particular

flood magnitude measure of interest.
In practical engineering application an arbitrary single choice is made at each
step (a) to (f); in the procedure described in this paper, however, the choice
is made from a selection of possible values, each one with a frequency proport-
ional to its probability of occurrence. Figure 1 illustrates the procedure as a
tree diagram on which the "single choice" method would be represented by a single
path.
As implied in figure 1 the continuous distributions of variables such as
rainfall duration are "discretized" so that each variable is made to assume only
one of a finite number of possible values to each of which a probability weight
is attached. Twelve values of duration, 36 temporal intensity patterns and 12
values of catchment wetness index ( C m .- and index of antecedent conditions
governing infiltration loss and base flow) are used. In a separate study to
provide rainfall information (Appendix 1 ) no dependences were noted between the
rainfall variables and this assumption was made throughout the simulation. This
means that the weights associated with each sampled variable value was itself
invariable; for example the weights associated with each of the 12 CWI values is
the same for 3 hour as for 48 hour duration storms. This particular consequence
might represent some departure from actuality as, in the United Kingdom, both
are seasonable variables.
However assuming independence and discretizing allowed considerable simpli-
fication in the programming and allowed the associated weights of each of the
12 x 12 x 36 combinations to be calculated from the product of the weights of
each of the contributing variables. This product weight is associated with the
flood magnitude in calculating statistics or assembling data into histograms.
To summarise, let pi be the weigM(or probability) of the ith duration,
Di ; let qj be the weight (or probability) of the jth hyetograph distribution,
462

Hj; let rK be the weight (or probability) of the kth CWI, CK; and let QijK
be the flood magnitude resulting from the combination of Di , Hj and Cu. Then
under the assumption of independence the weight or probability to be associated
with QijK is Wijk = pi qj rn and the expected flood magnitude is calculated from
E
B piqjrn Q i ~ n, while the mean flood magnitude following al1 storms of say
Fh'e fourth duration is calculated from E C W Qijh (Figures 2A and 2B).
j K 41~
Table 1 shows the results of the simulation for the IO -year return-period
at Burbage and Grendon. The contingent distributions show the effect of different
assumed values on the peak discharge. One noticeable result is that changes to
the rainfall variables have small effect on the average peak discharge showing
that the design flood would be insensitive to variations in hyetograph pattern
or storm duration. This is not to say that floods resulting from storms following
particular combinations of duration and hyetograph pattern cannot be found that
depart from the average, but as can be seen from the low standard deviations of
peaks contingent on chosen CWI values centrally chosen rainfall variables will
introduce little bias into the design flood. It has been found that this same
effect is even more marked when the measure of flooding being investigated
involves seme element of storage.

On the other hand, small changes in the CWI have a marked effect on the
resulting flood. It happens that a CWI value chosen to be near the median of the
distribution of CWI would have yielded a peak discharge only 5% in excess of the
expected flood.

Figure 3 shows some of the histograms of flood peaks following the 100-year
storm. These are noticeably negatively skewed and the modal value is typically
20% to 30% in excess of the mean. The inference from this is that a single choice
of each of the variables is likely to yield a flood that exceeds the average flood.
The sharpness of the histograms contingent upon CWI and the discrete sampling is
responsible for the spikey nature of the other histograms.

3. RAINFALL AND DISCHARGE DISTRIBUTIONS.

It had been noted in Section 2 and Figure 3 that the probability distribu-
tion of floods following rainfalls of fixed return period is negatively skewed.
One might anticipate from this that T-year return-period storms tend on average
to give rise to more floods with return period less than T-years than floods of
return period greater than T-years.

To test this and to derive the flood distribution the simulation was gener-
alised to sample the distribution of storm depths. Instead of sampling only
storms of depth and duration such as lie on a line of equal return period the
sampling is now conducted across all combinations of storm depth and duration.
The depth-duration-frequency is again used in order to calculate the probability
of occurrence of any combination (Figure 2C).

Figure 4 shows a comparison between the flood frequency relation as derived

 463

from the two simulations and from recorded flood peaks. In the case of Grendon
Underwood there is an apparent tendency for the simulated relation to under-
estimate the flood discharge based on the recorded peaks. although independent
evidence from regional analyses has suggested that the distribution as estimated
from the six only annual maxima would overestimate floods quite severely. However
the agreement with Burbage Brook, a small upland catchment in the Derbyshire
pennines with 43 years of data, is rather better. At small return periods the
generalised simulation produced lower flood values than the expected flood
following storms of that same return period.

4. CONCLUSIONS.

A technique has been described whereby the solution of several problems

pertinent to hydrological design in regions of inadequate data may be approached.
In particular, the sensitivity of the design flood to design assumptions can be
assessed. Experience with the technique suggests that the size of the flood is
determined more by the total depth of the rainfall than by its temporal distribu-
tion through the storm's duration. Correct choice of loss rate is in consequence
most important.

It appears that median values of duration, temporal distribution and loss

rate yield a design flood not far removed from the overall average flood follow-
ing the T-year storm. Because of the skewed nature of the flood distribution a
random choice of duration etc. would be more likely to yield a design flood
rather larger than the overall average.

The ability of the technique to reproduce tolerably well the flood magni-
tude frequency relation could be of very great value at a site where flow data
are scarce, whilst even at a well-endowed location the simulation result may be
used with profit to augment the flow record.

While attention has been concentrated on peak discharge as the measure of

flooding it should be emphasized that the technique is suited to more complex
design criteria. The hydrograph may be treated as an inflow and routed through
the scheme and so the actual design criteria of interest mqr be calculated.
Examples are:- (a) volume between inflow and outflow hydrographs for reservoir
freeboard design; (b) time to peak for a flood warning scheme; (c) volume over
a threshold level for a levee design.

The technique may also be adopted to use an entirely different catchment

response model such as that inherent in the rational formula, a multiple
regression equation or conceptual model although it can be expected that the
data requirements will be rather different from those of this investigation.
5. FUTURF RESEARCH.
The simulation appears promising as a tool for assessing the sensitivity of
design floods to variations in their causative factors and in estimating the
magnitude-frequency relationship for small return periods. However the technique
has not succeeded in reproducing the observed rapid growth in flood discharge
with increasing return period and it is here that further research is being
464

directed.
It is felt that the disparity between the definition of storms used to
determine the distribution of depth and duration (Appendix A -Introduction)
could be responsible for the "slow" growth and so long term autographic rainfall
records are to be. analysed to provide information on the distribution of the type
of storm used for the duration statistics.

Further investigation into dependencies between the variables could produce

results wkich would affect the aiscñarge distribution. For example seasonal sim-
ulation would reduce the coincidence of winter storm types with low summer CWI's
and vice versa.

The dependence of CWI on losses and base flow is essentially statistical

and this source of variability could be preserved fn the simulation by the addi-
tion of a random quantity to the values predicted from the best fit lines.
6. ACKNOWLEDGEMENTS.
Although few references have been cited the labours and opinions of others
have played no small part in the development of the procedure. Colleagues and
consultants of the United Kingam Floods Studies Team Dr. J.Y. Sutcliffe,
Professor J.E. Nash, Mr. M.J. Lowing, Mr. C. Cunnane, Mr. R.T. Clarke and
Mr. A.F. Jenkinson have all provided advice and encouragement. Mrs. J. Haworth
was responsible for the FORTFAN computer program and the numerical experiments
were run on the ICL 1906A of the Science Research Council's computing laboratory.

7. REFERENCES.

1. NASH,J.E. ; "Frequency of discharges from ungauged catchments".

Trans.A.G.U. , Vol. 37, No. 6, December 1956.

2. CHOW,V.T. and RAMASESHAN,S. ; "Sequential generation of rainfall and

runoff data". Proc. A.S.C.E., Journ. Hyd. Div. , Vol. 9 , HY4, July 1965.

3. EVANS,T. ; "River Eden flood relief studies". Feasibility report by

Sir M. Macdonald and Partners for Kent River Authority. Chapter 4,
September 1971.

4. DYCK,S. and KLUGE,C. ; "Investigations on the structure of frequency

distributions of floods". I.A.S.H. Warsaw, Vol. 3 , July 1971.

5. EAGLESON,P.S. ; "The dynamics of flood frequency". Trans. A.G.U.,

Water Resour. Res..Vol. 8, No. 4, November 1972.

6. LECLERC ,G. and SCHAAKE,J.C. ; "Derivation of hydrologic frequency

curves from rainfall". Water Resour. Res. (in print).
 46 5

TABLE 3

FLOOD DISCHARGE FOLLOUNG 3 O-YEAR RETURN PERIOD STORMS.

GFiENDON UNDERWOOD BURBAGE BROOK

MEAN STANDARD MAXIMUM MINIMUM MEAN STANDARD MAXIMUM MINIMUM

DEY DISCHARGE DEY DISCHARGE
m3/s m 3Is m3/ s *3/s 3
m /s
3
m /s
3
m /s m3/5
Overall 5.9 2.0 11.3 1.0 6.7 3.9 14.0 0.6
Constant duration storms.
1 hour 4.5 I .6 7.7 3.0 4.8 3.4 8.8 0.6
3 hour 5.7 2.0 10.0 3.2 6.4 3.8 13.5 0.9
6 hour 6.2 2.1 33.3 1.4 7.3 2.0 13.4 1.3
9 hour 6.2 2.0 11.3 3.6 7.3 3.9 14.0 3.1
12 hour 6.2 2.0 11.3 1.6 7.0 1.8 13.9 1.2
15 hour 6.o 1.9 11.1 3.5 6.8 1.7 13.7 1.1
18 hour 5.9 1.8 11.0 1.6 6.5 1.7 13.5 1.1
21 hour 5.7 . 1.7 10.9 3.6 6.3 1.6 13.4 1 .O
24 hour 5.0 1.5 9.8 1.4 5.8 1.5 12.5 0.9
30 hour 4.4 1.3 9.1 1.2 5.2 3.4 33.7 0.8
36 hour 4.0 1.2 8.4 3.1 4.7 Y .2 31.0 0.8
48 hour 4.1 1.2 8.9 1.2 5.3 3.4 12.3 0.9
Constant Quartile Type
I 5.8 2.0 13.3 1.0 6.2 1.8 12.i o.6
II 6.o 2.0 10.5 1 .O 6.7 1 .8 12.7 0.6
III 5.9 2.0 10.5 1.0 6.8 1.9 12.4 0.6
IV 5. a 2.o 11.1 1 .o 6.9 2.0 14.0 0.6
Constant CWI
i5 50" 1.6 0.2 1.9 1.0 1.3 0.2 1 .6 0.6
35 60 2.8 0.3 3.4 I. 8 2.0 0.3 2.7 1.2
55 70 3.9 O. 4 4.8 2.1 2.9 0.4 3.7 1.7
70 ao 4.7 0.5 5.7 2.3 3.7 0.5 4.7 2.2
80 go 5.2 0.6 6.4 2-5 4.4 o.6 5.6 2.4
90 100 5.6 0.7 6.9 2-7 5.1 o. 6 6.6 2-7
100 i10 6.1 O. 7 7.5 2.8 5.8 0-7 7.5 2.9
i10 120 6.6 o.8 8.0 3.O 6.5 0.8 8.4 3.2
120 130 7.1 o.8 8.6 3.2 7.3 0.9 9.4 3.5
130 140 7.7 0.9 9.3 3.4 8.1 1.0 10.4 3.9
i40 150 8.3 o. 9 10.1 3.8 9.0 3.1 1 3 -6 4.4
150 165 9.4 1.0 11.3 4.6 11.0 1.2 14.0 5.9
Recorded data
Graphical
fit 12.0
'-

I
8.6 --
MaX.
Likelihood1 3.1 3.O 8.2 1.6
* First figure refers to assumed CWI at Grendon, second to Burbage Brook.
 466

APPENDIX A - DATA REQUIREMENTS.

INTRODUCTION.

Statistical distributions were required for the three modes of rainfall

variability: depth, duration and temporal variability- for each catchment invest-
igated. In order not to predetermine any of the variability modes it was necess-
ary to define a storm in a manner unlike that of the customary rainfall depth-
duration-frequency diagram. The definition was expressed k terms of the condi-
tions for starting and ending a storm: a storm was considered to begin at the
onset of rain and to end when in the preceeding Y hours not more than X mms of
rain occurred. X and Y were chosen to represent the conditions under which a
flood hydrograph would return to near base flow and allowed sbrt spells of zero
rainfall to occur within a storm event.

Hourly analysis of catchment average rainfall was available from three

catchments; Grendon Underwood, Coalburn and Plynlimon (Wye). Sufficient records
were available to permit an investigation h t o statistical distribution of storm
durations and temporal patterns but not to conduct an investigation into storm
depth. For this element of the simulation, results of a depth-duration-frequency
analysis of the entire country were available from A.F. Jenkinson (Ref. Al).

The catchment response model is one currently under investigation by the

Floods Study Team. A relation between catchment wetness index (CWI) and total
storm losses, and CWI and base flow was used. A unit hydrograph based on recorded
unit hydrographs from the catchments were convoluted with the gross rainfall less
losses.

DETAILS OF THE SIMULATION DATA.

(a) Rainfall depth: The basic equation used to relate the T-year return period
rainfall of any duration (MT) to that of the five-year return period rainfall
( ~ 5 )is
M I M 5 = (T/5)'
where c is the "growth factor" and is related uniquely to M5 which is mapped
for the entire United Kingdom. Other necessary information required by the
simulation and provided in Ref. A1 concerns areal reduction factors to convert
point to areaì rainfall.

(b) Rainfall duration: This distribution is dependent upon the storm definition
and for the values X = 2 m s , Y = 5 hours used for both Grendon Underwood and
Burbage Brook simulation is given below
STORM DURATION 1 3 6 g 12 15 18 21 24 30 38 48
(HOURS)
RELATIVE FREQUENCY 5 12 26 20 13 30 8 1 1 2 1 1
(PER CENT)
 467

It was found that the distribution was very similar for hoth upland and lobiland
rainfall stations and varied slowly with changes to X and Y, longer storms
becoming commoner as the conditions for ending a storm were relaxed.

(c) Temporal distribution of storm rainfall: Several alternative schemes for

describing the hydrograph shape were investigated. The one chosen was due to
F. Huff (Ref. A2) in which four quartile types are recognised depending upon in
which of the four quarters of the storm duration tiïe largest rainfall fell. The
fine detail of the hyetograph is sampled by plotting all curves of the same
quartile type on a graph showing accumulating fraction of storm depth against
fraction of total storm duration. Composite storms can then be constructed by
connecting points which are exceeded by lo%, 202, 30% etc. of all storms. Sampl-
ing from these composite storms is analogous to sampling at regular intervals
from a distribution function in order to sample a variable in proportion to its
frequency of occurence,and were used by the sfmulation. The shapes of the campos-
ite storms were found to be insensitive to changes to the storm definition and
were nearly indistinguishable between upland and lowland catchments. The percent-
age frequency of the four quartile types were 12% type I, 32% type II, 35%
type III, 21%type IV.

(d) CWI distribution: CWI is calculated in nuns. from the soil moisture deficit
(SMD) as computed by the Meteorological Office (Ref. A3) and a five day anteced-
ent precipitation index (API5) using a daily decay constant of 0.5. The formuïa
usedwasCWI = 125-SMD+API5. It had been observed in a recent study (Ref. Ab)
that wet day rainfall and SMD were statistically independent and S O the end of
month values were adopted as representative of all cw? values. Oxford data nr&S
used to provide the distribution for Grendon Underwood and Buxton for Burbage
Brook. In the simulation a linear relation with CWI was used to calculate total
storm losses and the reciprocal of the temporal variation of CWI as the storm
progresses (assuming no evaporation to increase SMD) determined the loss rate
curve. An exponential relationship with CWI determined the base flow.

REFERENCES.

Al JENKINSON,A.F. ; "Meteorological office progress report. January 1972".

Report prepared for Floods Study Steering Committee.
A2 HUFF,F.A. ; "Time distribution of rainfall in heavy storms". Water Resour.Res.
Vol. 3, No. 4, fourth quarter 1967.
A3 GRINDLEY,J. ; "Estimation and mapping of evaporation". 1970 I.A.S.H.
symposium, Reading, I.A.S.H.
A4 BEM,M.A. and SUTCLIFFE,J.V. ; "An index of flood-producing rainfall based
on rainfall and soil moisture deficit". Journ. of Hydrology, Vo1.17, 1972
pp 229-236.
46b

FIGURE 1
 469

DuraîK

M 1 etc.
j

~ ~

FIGURE 2A FIGURE 2B
WïES
11 Consider case where two variables only aFFect discharge Q, for example storm duration
and CUI (Figure 24).

CWI in the second interval. C2, a discharge q(H) and a probability p(H)
ïoud
-
For each Combination of duration and CWI a value OF Q m d B probability of occurrence C U
be calculated. For exsmple combining the duration in the Fourth interval, Dy, with the
p(D,,)xp(ci) arc

'-1 Summing ail the probabilities in each discharge interval a discharge distribution [ F i m i
20) may be COOStmCted.
) This concept can be generalised to sample From Further variables.

I
Discharge den<
9
/ / / /

wm0.
a) Depth uld duration are plotted on the base plane ( P i w e 2c)
Durat ion
b) Each coibtiatim ia assoeiited nith a probability of o c c u r m c e u) g i v m by thr depth-
duration-frequency diagni.
c ) Contingent on each depth duration ccmbination B diatribution of diachugai like Pipure
OB C M be visudiaed on the vertical discharge arii.
d ) Integrating such densities above all points on the b u e m e 011 locu. or =qual retur,,
period yields the results of Section 2.
0) Integrating over the entire base plane Yields the d i h r i b u t i m of diichuge of Section 3.

FIGURE 2C
470

>-
u
æ
W
x
E
s
i=
0, 4
W
œ l
I
20- lI

18.
+
l

i
i

!
I

I
PEAK DISCHARGË- M"/S
Burbage Brook-Floods following 100-year Rainfalls
Distribution of all floods
Floods from storms of given duration
Floods from storms of given CWI ---
FIGURE 3
 471

Return period-years
I I I I I I
233 5 IO 20 50 100
Q/C M e a n annual
flood

2.2 Most likely peak following storms of given return period

\

2.0

1.5

/'Simulated flood peaks

1.0.
following storms of given

O.5 Peaks have been standardised by the arithmetic mean of

the recorded annual maxima 5.39mYs.
Plotting position corresponds to expected value of order statistic.
Graphical fit to plotted points so'recorded'line misses (1,l).

O 1 1 I I I I

O 1 2 s
Reduced
4
variate- y
5
FIGURE 4
 A DECISION - THEORETIC APPROACH TO UNCERTAINTY
IN THE RETURN PERIOD OF MAXIMUM FLOW VOLUMES
USING RAINFALL DATA

Donald R. Davis('), Lucien Duckstein(t:) Chester C. Kisiel('),

and Martin M. Fogel

ABSTRACT

The maximum seaaonal rUno#f YolUw Q #or an ungaged atream site is

derived using (1) an event-based rainfall mode1 for thunderstorma, and
(2) a linear rainfall-runoff model. Major emphasis is placed on effect
of uncertainty in parameters of rainfall inputs on the return period of
maximum runoff volumes in a season. The event-based rainfall model, de-
rived previously by the coauthors and others, has the following featu-
res: (1) the distribution of the number of events per season N is Pois-
son with mean m; (2) the d'1s t ribution of point rainfall amount R per
event is exponential with mean llu; (3) N and R are independent. More
explicitly, we obtain a correct distribution function for the return pe
riod T (x) under the uncertainty in m and u, and demonstrate the neces-
sity 0 P following this approach for a decision-theoretic analysis of a
water resource design problem. The approach enables us to design struc-
tures, relying only on rainfall data, on watersheds with ungaged
streams by taking into account uncertainty of design site parameters.
Also, we cari tailor the design to a specific problem rather than use a
pre-specified design flood, such as the magical lOO-year flood.

RESUME

Le volume d'écoulement maximum est calculé 2 un site non instru-

menté, en utilisant: (1) un modele de pluie d'orage construit par événe
ment; (2) un modèle pl,uie-débit linéaire. La maniere dont l'incertitudë
sur les paramètres du modèle de pluie affecte la période de récurrence
TQ(x) du volume $'écoulement maximum Q est analysée d'une manière quan-
titative, Le modele de pluie d'orage a les caractéristiques suivantes:
(1) se nombre d'événements par saison N suit une distribution de Pois-
son a moyenne m; (2) la quantité de pluie ponctuelle R par événement
suit une distribution exponentielle de moyenne l/u; (3) N et R sont des
variables aléatoires indépendantes, Nous obtenons la fonction dti distri-
bution de TQ(x) tenant compte de l'incertitude sur m et u et montrons
l'utilité de cette méthode pour une application correcte de la théorie
de la d6cision à un problème de planification de ressources en eau.
Nous pouvons ainsi de conceyoir des ouvrages sur des bassing déversants
sans données d'écoulement, a l'aide de données pluvincgtriques, tout en
tenant compte de l'incertitude sur les paramètres. Par ailleurs
pouvons spécialiser la conception & chaque cas d'e;tpèce au lieu'd:ItTli
ser une crue standard, telle la magique crue de pêriode de retour centë
naire.
1 Respectively, Assistant Professor and Professors, on joint appointment,
Departments of Hydrology and Water Reaources and Sistems and fndustrial
Engineering, University of Arizona, Tucson, Arizona 85321,
2 Department of Watershed Management, Same address as in.(l),
Professor,
 474

1.0 Introduction
Fle,ods or stream discharges are properly described by their durations and
volumes above a certain flow level and their instantaneous peak flows. Of
~I1cs.e three properties, this paper is concerned with the uncertainty in the
return period of maximum flow volumes which is a design parameter for flood pro-
tection and other structures. In particular, we consider the uncertainty due tc
inadequate data on small watersheds (up to 500 lon2).
Jt is well known that there is a good chsnce that a flow event Q.with a
large return period TR may be exceeded at least once in an R-year design period.
Typically, however, calculated risk diagrams (Gilman, 1964) do not consider the
uncertainty in the return periods of rainfall and flow events. TO a design
engineer, the uncertainty of inadequate rainfall or flow data cari result in either
overinvestment (overdesign) or underinvestment (economic losses) in the design of
flood retarding or retention structures or of water storage facilities (farm
ponds or water supply reservoirs for small towns or industries). The Bayesian
framework presented in this paper allows for an explicit consideration of hydro-
logic uncertainty as noted above and for a methodology to evaluate potential
losses associated with that uncertainty.
Approaches takea to arrive at estimates of the return period of hydrologie
f'low properties include:
(a) tipirical fitting of probability density functions to historical data;
in particular, the Soi1 Conservation Service (1965) fitted Pearson
Type III distributions ta flow volumes for various time periods in
Arizona. This approach disregards any available information in pre-
cipitation records or any knowledge about the rainfall-runoff prccess.
(h) Use of phenomenological relations such as a linear trensformation of
rainfall volume to flow volume as a basis for obtaining probability
density functions (pdf) of flow. The pdf of rainfall volume may be
denrribed empiri~ally (with its consequent uncertainty) or from a
procesc viewpoint .,herein individual rainfall events are modeled as
a stochastic process along the time axis (Duckstein et al. 1972).
(c) Use of detailer! dynemical flow equations to relate pdf of rainfall
psoperties to pdf of flow properties (Ragleson, 1972).
In this paper we use the second approach. Herein we build on previaus work
(Davis et al. 1972) where we evaluated the ucertainty in the return period of
point rainfall amounts from summer thunderstorms. We define an event-based
process in this case as a sequence of thunderstorms in tine. The return period
TR(k) of maximum point rainfall e (with k the rainfall smount or value of the
random variable 5) is derived by considering the following elements of the
event-based nrocess:
(a) l?hë number 1; of events per season is Poisson distributed with met-a m
(of number of events per season):

b) Rainfall events 53, R,,..., are independent identically distributed

random variables.
(cl The amo\‘Jit 5 of point rainfall per t\sent is exponentially distributed
with parsmeter u (equal to reciprocal of mean emount rainfall per event):
fR(klu) = ueBuk

(d) N and 5 are indepenaent.

 475

Then, the return period of k units of rain in a season, given the event-based
parameters m and u, is

Because m and u are uncertain due to small sample size, T is uncertain.

R
To encode the uncertainty, the posterior distribution of m and u represents
the likelihood of the values of m and u which produced the data. This posterior
is given by the conjugate distributions for the exponential and Poisson distri-
butions (de Groot, 1970, Chapt. 9). The distribution that is conjugate to both
of these is the gamma:
a a-1 -bx
b x e
gX(xla,b)
-. = r(a) (4)
For the Poisson distributution,
x m = , the parameter of the Poisson and estimated as m.
b =
n , the number of seasons.
a = &-I , the total number of rainfall events in n seasons.
For the exponential distribution,
x = u , the parameter of the exponential and estimated as Û.
a =
&-I , the total number of rainfall events in n seasons.
b =
h / û , the total amount of rainfall for the mn events.
The resulting F (x)s in each case are posterior distributions and represent
z
the likelihood that various values of m and u axe the values describing the rain-
fall process that we are observing, after getting the data. These posterior
distributions are used in a computer simulation to develop the posterior distri-
bution of TR(k). The mean of this distribution is the expected return period
E
[
T (k)Tl fo; a k-inch rainfall. Computer results' given by Davis, et al. (1972)
R
indicate that the return period of point rainfall is subject to considerable
uncertainty even with 20 years of data. The design and operational implications
are obvious for flood control, dry farming with irrigation, and water supply.
Next, we extend the procedure to uncertainty in return periods of seasonal flow
volumes on small watersheds.

2.0 Extension to Seasonal Flow Volumes

If m is the total number of runoff producing rainfall events in a summer
season, then the exact expected return period T (y) of the maximum seasonal
Q
runoff volume Q is, under our previous hypotheses,
T& (ylm,u) = [i-exp I-m + m F (ylu)~~-l
9 (5)
where F (ylu) is the distribution function of runoff per event CJ which we will
9
write F (y) for simplicity. Our approach is to obtain F (y) from the distribution
9 Q
function F (x) of rainfall
R
5 per event, using the linear rainfall-runoff relation-

where A are the initial abstractions depending on the watershed and is a co- c
efficient depending on the rainfall characteristics for a given watershed, in
particular, a time factor such as the maximum 15-minute intensity (Duckstein
--
et al. 1972).
 476

If we let

p = !-A for R > A

I

=o for 5 < A
then Equation (6) becomes
Fp(x) = 1 -exp (-u(x+A)) for x > O
= --
CP or y = -
cx; the distribution function of P is
(7)
and thit of €j(Feller, 1967, Chapt. 2) is
m

FQ(y) = I,"p($) fC(c) dc

because c
is a random variable as noted in previous work by the coauthors
(Duckstein et al. 1972). Since, physically, we cannot obtain more runoff than
c
rainfall, then O 5 5 1 , and a beta distribution for
Driate:
seems to be most appro- c
(9)
The uncertainty on a,b will not be considered in the present study. To sum up,
Equations (7), (8) and ( 9 ) may be combined to obtain

Equations (10) and (11) are now substituted into Equation (5) to obtain an ex-
plicit expression of Tg (ylm,u). Because we have the sufficient statistics,
fi and a, our knowledge of and m u
can be expressed as a pdf !Tiao and Box, 1973).
Hence, this encoded uncertainty results in a pdf on T (ylm,u).
Q
3. O Met hodolopy
To obtain the pdf of the return period on hand, T (ylm,u,n) is a problem of
transformation of random variables, where a closed form is beyond reach.
9
Thus, a simulation approach is used as follows: (a) consider a fixed

conjugate pars. %(mlfi,n)

-
yearly maximum flow volume Q = yo, and (b) sample values m,u are drawn from the
and gu(uli,n), respectively, as noted in our discussion
of Equation ( b ) , (c) these sample values are substituted into T (y,Im,u) to
obtain one value of the return period T 8
and (a) the process is repeated to
0'
obtain pdf of T for a fixed y (for example yo = Q
9 0.7 inch in Table 1). =
A similar procedure is then used to calculate the pdf of (T )
-
'
, which is
the probability of exceedance of y . The design parameter of
9
interest may be
O
either T (for sizing a small dam) or (T (estimating long-range replace-
B 9
ment costs of structures.)
Finally, to be considered in a later study is the pdf of maximum seasonal
flow Q that corresponds to a fixed return period. Such a pdf may be of interest
for flood plain insurance purposes and can be calculated by the same simulation
procedure as above.
 A DECISION -THEORETIC APPROACH TO UNCERTAINTY
IN THE RETURN PERIOD OS MAXIMUM FLOW VOLUMES
USING RAINFALL DATA

Donald R. Davis"), Lucien Duckstein (1)

( 2 j Chester C. Kisiel (1) ,
and Martin M. Foge1

ABSTRACT

The maximum seas.ona1 Trino$$ YolYge for an ungaged styeam site i s

derived using (1) an eyent-based rainfall podel $or thunderstorms, and
(2) a linear rainfall-runoff model. Major empñasis is placed on effect
of uncertainty in parameters o f rainfall inputs on th-e return period of
maximum runoff volumes in a season. The event-based rainfall model, de-
rived previously by the coauthors and others, has the following featu-
res: ( 1 ) the distribution of the number of events per season N is Pois-
son with mean m ; (2) the distribution of point rainfall amount R per
event is exponential with mean 1Lu; c 3 ) N and R are independent, More
explicitly, we obtain a correct distribution function for the return p e
riod T (x) under the uncertainty in m and u, and demonstrate the neces-
sity 09 following this approach for a decision-theoretic analysis of a
water resource design problem. The approach enables us to desigr, struc-
tures, relying only on rainfall data, on watersheds with ungaged
streams by taking into account uncertainty of design site parameters.
Also, we can tailor the design to a specific problem rather than use a
pre-specified design flood, such as the magical 100-year flood.

Le volume d'écoulement maxjmum est calculé i un site non instru-

menté, en utilisYnt: (1) un modele de pluie d'orage construit par 'evéne-
ment; (2) un modele pluie-débit linéaire. La maniere dont l'incertitude
sur les paramètres du modèle de pluie affecte la période de récurrence
TQ(x) du volume d'écoulement maximum Q est analysée d'une maniere quan-
titative. Le modèle de pluie d'orage a les caractéristiques suivantes:
(1) le nombre d'événements par saison N suit une distribution de Pois-
son à moyenne m; (2) la quantité de pluie ponctuelle R par événement
suit une distribution exponentielle de moyenne l/u; (3) N et R sont des
variables aléatoires indépendantes. Nous obtenons la fonction de distri
bution d e TQ(x) tenant compte de l'incertitude sur m et u et montrons
l'utilité de cette m'ethodf pour une application correcte de la théorie
de la décision à un probleme de planification de ressources en eau.
Nous pouvons ainsi de conceroir des ouvrages sur des bassing déversants
sans données d'écoulement, a l'aide de donn'e$s pluviométriques, tout e n
tenant compte de l'incertitude sur les parametres. Par ailleurs, nous
pouvons spécialiser la conception à chaque cas d'espèce a u lieu d'utili
ser une crue standard, telle la magique crue de période de retour cent:
naire.
'Respectively, Assistant Professor and Professors, on joint appointment,
Departments of Hydrology and Water Resources and Sìstems and Industrial
Engineering, University of Arizona, Tucson, Arizona 85721.

2Professor, Department of Watershed Management, Same address as in (1).

 474

1.0 Introduction
Floods or stream discharges are properly described by their durations and
volumes above a certain flow level and their instantaneous peak flows. Of
briese three properties, this paper is concerned with the uncertainty in the
return period of maximum flow volumes which is a design parameter for flood pro-
tection and other structures. In particular, we Consider the uncertainty due to
inadequate data on small watersheds (up to 500 h2).
It is well known that there is a good chance that a flow event 9 with a
large return period T may be exceeded at least once in an N-year design period.
R
Typically, however, calculated risk diagrams (Gilman, 1964) do not consider the
uncertainty in the return periods of rainfall and flow events. To a design
engineer, the uncertainty of inadequate rainfall or flow data can result in either
overinvestment (overdesign) or underinvestment (economic losses) in the design of
fiood retarding or retention structures or of water storage facilities (farm
ponds or water supply reservoirs for small towns or industries). The Bayesian
framework presented in this paper allows for an explicit consideration of hydro-
logic uncertainty as noted above and for a methodology to evaluate potential
losses associated with that uncertainty.
Approaches taken to arrive at estimates of the return period of hydrologic
flow properties include:
(a) Rupirical fitting of probability density functions to historical data;
in particular , the Soil Conservation Service (1965) fitted Pearson
Type III distributions to flow volumes for various time periods in
Arizona. This approach disregards any available information in pre-
cipitation records or any knowledge about the rainfall-runoff process.
(b) Use of phenomenological relations such as a linear transformation of
rainfall volume to flow volume as a basis for obtaining probability
density functions (pdf) of flow. The pdf of rainfall volume may be
described empirically (with its consequent uncertainty) or from a
process viewpoint wherein individual rainfall events are modeled as
a stochastic process along the time axis (Duckstein et al. 1972).
(c) Use of detailed dynamical flow equations to relate pdf of rainfall
properties to pdf of flow properties (Eagleson, 1972).
In this paper we use the second approach. Herein we build on previaus work
(Davis et al. 1972) where we evaluated the uncertainty in the return period of
point rainfall amounts from summer thunderstorms. We define an event-based
process in this case as a sequence of thunderstorms in time. The return period
TR (k) of maximum point rainfall (with k the rainfall amount or value of the
random variable FJ) is derived by considering the following elements of the
event-based process:
(a) The number N of events per season is Poisson distributed with mean m
(of number of events per season):

(b) Rainfall events R-1, G2,..., are independent identically distributed

random variables.
(c) The amount of point rainfall per event is exponentially distributed
with parameter u (equal to reciprocal of mean amount rainfall per event):
fR(klu) = ue-Uk

(a) $ and R are independent.

 477

4.0 Results
The results of the computer simulation are summarized in Tables 1 and 2 and
Figures 1 and 2. In these we consider the variance of c,
representative of con-
ditions on the watershed, and the variance in our knowledge about rainfall
parameters m and u.
Table 1 shows that u, the average rain per event, is much more important
than m, the average number of storms per season, as judged by the variance of
T ,(Var
9
?! ), for
9
different values of Var c. We also note the following
(a As Var c increases, EL?,? and Var ? decrease,
9
Thus by not randomizing
C the estimated return period of Q = 0.7 is much
higher. By varying C
the variable effects of rainfall intensity and watershed behavior on
the return period are anticipated;
(b) Var
9
Tincreases dramatically when Var $ = O for joint uncertainty in
m and u;
(c) The mean reciprocal return period (= exceedance probability p) and =
.-1
Var T, increase rapidly as Var
9
c
increases. This result is shown
because p is commonly used as the design parameter in hydrologic risk
analysis.
These patterns hold for all values of runoff volume used in the sensitivity analysis
=
(Q 0.5, 0.7 and 0.9 inches of runoff) as shown in Table 2.
As expected, the Var T decreases with doubling of available data (10
9
to 20 years used in the simulation) as summarized in Table 2. The E[id is only
slightly changed. A more general manifestation of the simulated process is evident
in Figure 2 where the posterior pdf (of return periods for 0.7-inch runoff) based
on 20 years of data has a much sharpy modal value than the posterior pdf based
on 10 years of data; note that mear, T is just to the right of the mode. While
9
not shown, the posterior pdf's become more peaked as Var increases. c
The effect of increasing runoff Qolume is to increase E[Td, Var T and
9
coefficient of variation CV(T ) as shown in Table 2. The latter result about
9 I

CV(TQ) also implies that a Var T increases more rapidly than E[Td. It is
9
intriguing to note the dramatic effect that the introduction of Var has on the
parameters.
The results in Table 2 for n = 10 years are shown in Figure 1, a plot on
Gumbel extreme value paper. As previously noted, as Var increases the smaller c
"Liil. From the tabulated results we note that so-called confidence limits for
ea& line would get wider as T increases because Var T increases with runoff
61 9
volume. These confidence limits are narrower for n = 20 years of data as is
evident from Table 2.
Of interest is the modest computer time (maximum of 25 seconds for 20 years
of data) per simulation run on the CDC-6400. Given the number of uncertain
parweters in this problem, it does not appear feasible to prepare charts and
graphs for routine design use unless more exhaustive computer studies are per-
formed.

4.1 Comments on Results A

In contrast to the classical empirical frequency approach in deriving T

9,
478

the event-based approach outlined here results in evaluation of uncertainty in

..
T from physically meaningful parameters like m and u. This is a much more
Q
efficient use of the available data on rainfall and runoff.
We have seen how the design would depend on the uncertainty in m and u and
on the interaction between*uncertainty in m and u and Var C. The end result, a
posterior distribution on T is of value to inference on hydrologic stochastic
9’
processes as discerned from limited data of value to the next important step
of invoking Bayesian decision theory for evaluating design decisions and for
judging if better designs are possible by waiting for-additional cata.
It would be desirable to express the moments (ErTJ and Var T ) of the
Q
posterior pdf in terms of m, u, Q and C, but this is intractable. -The next
approach for thinking about our results in simpler terms is to consider the
mean and variance of = 5 cp:
ECQI = EC~J CIFI
Var = - - -
E2[C] Var P + E2[P] Var C + (Var C) (Var P)

as given by Benjamin and Cornel1 (1970, p. 169). When C is not random, these
equations become E[g = CE[?] and Var 9 = C2 Var p. The variance of Q (and
thus its frequency of exceedance and its return period) is dramatically affected
by randomization of C. It is common in hydrologic design to choose a “frequency
factor” z (or standardized variate) in the relation 9 = Ere] + z (Var Q)1/2. ..
To contrast properly this classical approach to finding a design flow Q with the
method outlined in this paper wou1.d require a full-fledged decision theoretic
analysis for a specific design problem. The evaluation would have to be repeated
for each design use of the posterior pdf. Much work remains to be done in this
direction.

4.2 Relationship of results to Bayesian decision theory

Let the loss function for the design of a flood protection structure, say
a dike, be L(h,T) where h is the height of the dike and T is a design return
period such as T or an exceedance probability (T )-l. The result of our
a 9
investigation was to determine the posterior pdf f (t) as given in Figure 2.
T
Thus, we are now able to calculate Bayes risk, which corresponds to the optimum
design h*
+m
BR(h*) = min L(h,t) fT(t)dt (14)
..
h o
We can also calculate the worth of sample information to sharpen the estimate
of T (Davis et al. 1972) for each intended use of the data. Such studies are
9
left for the sequel. It is very important to emphasize that the worth of data
discerned by this methodology is based on the economic loss function associated
with a particular design use of the data; the results are not in terms of the
variance of the return estimate (return period in this case).

5.0 Conclusions
It is important to keep in mind when judging the results of the research
reported here that we are dealing with maximum flow volumes generated by a sequence
of thunderstorms during a season. Additional work is necessary to extend the
 479

approach to other runoff-producing precipitation events (including snow) during

the year. The use of the Gumbel distribution in this paper goes beyond its
classical use for the instantaneous rainfall and flood maxima during the year.
We thus have found the following points in our theoretical and simulation
arialy,is :
The approach enables us to design structur'es, relying only on rain-
fall data, on watersheds with ungaged streams by taking into account
uncertainty of the site parameters,
Using this approach we can tailor the design to a specific problem
rather than use a pre-specified design flood, such as the magical
100-year flood.
Simulation is an appropriate method for evaluating uncertainty in
estimates of physically-meaningful parameters arising in the event-
based approach.
Return period varies with record length rainfall, and watershed events,
etc. We have given an event-based approach to evaluate this variation.
The sensitivity analysis demonstrates the dramatic importance of un-
certainty in the average amount of rainfall per event and the importance
of considering variability in the rainfall and watershed parameter
called C in this paper.
The resats, if encoded in the posterior pdf of the return period
TI, allow the user to exercise inference or to find sensitivity of the
analysis to design decisions in the face of inadequate data. Bayesian
decision theory is the framework suggested for undertaking the decision
analysis.
The results have implications for design of a variety of hydraulic structures
in both urban and rural watersheds, in temperate and arid climates, and in
regions of the world confronted with inadequate hydrologic data. In the face
of changing watershed conditions, as reviewed by Fogel, et al. (1972), the
approach offered in this paper permits exercise of judgment on the effects of
lack of knowledge and of nonstationary meteorologic and hydrologic parameters
such as m, u and C. In o w j u m e n t , classical empirical frequency methods do
not provide such a clear basis for evaluation. Extension to nonlinear water-
shed models are possible as noted by Duckstein, et al. (1972) and Fogel, et al.
(1972).

6.O Acknowledgments
The work was supported in part by U.S. National Science Foundation Grant
GK-35791 and by a matching grant (Decision Analysis of Watershed Management
Alternatives) from the U.S. Office of Water Resources Research. !Che computer
programming skills demonstrated by Joel Friedman have contributed substantially
to the realization of the results.

7.0 References

BenJanin, J.R. and C.A. Cornell. Probability, Statistics and Decision for Civil
Engineers, McGraw-Hill Book Co., New York, 1970.

Davis, D.R., L. Duckstein, C.C. iíisiel, and M. Fogel. Uncertainty in the return
period of maximum events: A Bayesian a.pproach. Proceedings, International
Symposium on Uncertainties in Hydrologic and Water Resource Systems,
University of Arizona, Tucson, Arizona; 1972, pp. 853-862.
 480

Davis, D.R., C.C. kïsiel and L. Duckstein. Bayesian decision theory applied
to design in hydrology, Water Resources Research, Vol. 8, No. 1,
February 1972, pp. 33-41.

de Groot, M.H. Optimal Statistical Decisions. McGraw-Hill Book Co., New York,
1967.
Duckstein, L., M.M. Fogel and C.C. Kisiel. A stochastic model of runoff-
producing rainfall for summer type storms, Water Resources Research,
Vol. 8, No. 2, April 1972, pp. 410-421.

Eagleson, P.S. Dynamics of flood frequency, Water Resources Research, Vol. 8,

NO. 4, August 1972, pp. 878-898.

Feller, W. An Introduction to Probability Theory and its Applications, Vol. 2.

John Wiley, New York, 1967.

Fogel, M.M., L. Duckstein and C.C. Kisiel. Choosing hydrologic models for
management of changing watersheds, Proceedings, National Symposium on
Watersheds in Transition (American Water Resources Association) , Fort
Collins, Colorado, June 1972, pp. 118-123.

Gilman, C.S. Rainfall (Section 9). In "Handbook of Applied Hydrology,"

Edited by V.T. Chow, McGraw-Hill Book Co., New York, 1964, pp. 9-59.

Soil Conservation Service, Runoff volume-duration-probability analyses for

selected watersheds in Arizona. Central Technical Unit, Hydrology
Branch, SCS, U.S. Dept. of Agriculture, April 1965.

Tiao, G.C. and G.E.P. Box. Some comments on "Bayes" estimators, The American
Statistician, Vol. 27 (i), February 1973, pp. 12-14.
 481

Table 1: Sensitivity anaiysikon return period moments as

function of uncertain parameters for rural water-
shed with only 10 years of data.
~~ ~

Jncertain A Reciprocal
)arameters Var C Moments of return period T
- return period

T IQ cv( TQ)**
flean
~
?
(years Var
- ,..
nean variance

l & U O 41.82 538. .555 .O00299

.O005 39.00 442. .539 .o00318
.O05 26.33 153. .470 .o00328

* O5 6.30 2.35 .243 .o01809

)nly m .O005 35.36 8.30 .o81

.O05 23.41 3.89 .O84
O5 6.20 t 19 .O70

nly u .O005 37.61 383. .521

.O05 24.i4 110. .435

* O5 6.52 1.96 .215

* Conditions for the analysis: A = 0.4 inches, mean C 0.3 for beta =
distribution, Q = 0.7 inches on the average; rainfall is
distributed on basis of an exponential distribution for
amounts above 0.3 inches with an average of 14.0 storms/
season and an average of 0.39 incheslevent.

*+ Coefficient of variation of T
s-
 Table 2: Sensitivity analysis on return period moments
asa fun&ion of rainfall P, length of record n
ànd variance of C; both m and u are uncertain;
watershed is rural; conditions are as noted in
Table 1.

Average
runoff
n Var < I
Moments of return period 'f
volume Q (years of
data) &lean
(years 9
0.5 10 .o01
~

11.70
~-
I 58.07 .650

.O05 6.97 14.90 .553

O5 2.94 - 27 * 177

20 .O05 6.25 4.38 ,335

~~~

10 O 41.82 538 .555

.O005 39.00 441 .539
.O05 26.33 153 .470

O5 6.30 2.35 ,243

20 37.44 245 .418

24.36 68 .339
6.41 1.09 .163

10 O 271 48,085 .809

.O05 103 3 ,602 .582

O5 14.29 21.25 .323

20 .O05 95.95 1,721 I .432

 483

I .o

O.8

v,
w
I
o
z
- 0.6
w
-
z
3
-I
$ 0.4
LL
I
L
O
Z
3
0.2

O
2 5 IO 20 50 100 200

RETURN PERIOD, Y E A R S

F i g u r e 1: T h e effect of t h e v a r l a n c e o f C o n t h a Teturn
p e r i o d o f p u n o f f volume.
 484

.25 -

.20 -
>
.I5 -
w
3
o
w
CE
LL
.IO -

.O5 -

R E T U R N PERIOD, YEARS

Figura 2: Posterior p r o o a b i l i t r densiTy function o f return periods

f o 0.7-i’nch
~ runoff of r e c o r d length.
 SYNTHETIC UNIT HYDROGRAPH TECHIQUE FOR THE
DESIGN OF FLOOD ALLEVIATION WORKS IN URBAN AREAS

by
M.J. Hall
Lecturer in Civil Engineering, Imperial College
of Science and Technology, University of London
ABS TRACT

The development of rural land for urban, suburban or industrial

purposes can radically alter the flow regime of the catchment area WL
thin which such changes take place. The volume of surface runoff tends
to increase, the lag time of the flood hydrograph to decrease and the
peak rate of flow to increase. These ch-anges should be anti’cipated in
the design o f flood alleviation works for catchment areas undergoing
urbanisation, but in general, little quantitative information is a v a i
lable o n the magnitude o f the effect at different stages of urban de-
velopment. I f flow records are available from several catchment areas,
each o f which has reached a different stage of urban development, the
finiteperiod unit hydrographs derived from these data can be used as
an index to the influence of urbanisation. The application of a syn-
thetic unit hydrograph technique to flow records from both urban and
rural catchment areas within the headwaters of the River Mole near
Crawley, United Kingdom, has confirmed the feasibility of the
approach but has shown that more thought is necessary in choosing c a i
chment characteristics which reflect the character o f the urban deve-
lopment.

RESUME

L’utilisation des espaces ruraux pour le développement urbain et

industriel peut changer radicalement le régime hydrol2gique des bas-
sins concernés. Le volume du ruissellement tend a croitre, le temps
de réponse du bassin à décoitre et les pintes de crues s’amplifient.
Lors de l’élaboration des projets, ces modifications devraint être
prévues et on devrait chercher à atténuer l’effet des crues par des
travaux appropigs, mais o n ne dispose en général que d’une informa-
tion très succincte sur l’importance de cet effet aux différents sta-
des du développement urbain. Si on dispose de relevés de débits s u r
plusieurs bassins atteints à des degrés différents par le développe-
ment urbain, o n peut utiliser les hydrogrammes unitaires tirés de ces
données pour constituer des indices concernant l’influence de l’appli-
cation d‘u?e technique d’hydrogramme unitaire synthétique aux débits
observés, a l’issue de bassins urbains et ruraux, dans le bassin s u p 5
rieur de l a Mole, pres de Crawley (Royaume Uni), a confirmé les possi
bilités de cette méthode; elle a montré aussi que le choix des c a r a c y
téristiques du bassin reflétant l‘influence du d6velo;pement urbain
demandait une sérieuse réflexion.
486

I. INTRODUCTION

According to Toynbee [I], almost half the World's population had become
urban by 1969. This increase in the urban population has been accompanied
by an even more marked expansion in the area occupied by streets and buildings.
The development of rural land for urban, suburban and industrial purposes is
characterised by two important physical changes, both of which may have a pro-
found effect on the hydrological cycle of the area within which such urbanisation
takes place.

Firstly, the area covered by relatively impervious surfaces increases,

thereby increasing the proportion of storm rainfall which becomes surface run-
off. Owing to the concomitant decrease in soil moisture recharge, dry weather
flows are reduced.

Secondly, the natural surface water drainage system of the area is invariably
subjected to a variety of changes, ranging from realignment of channels to the
installation of stormwater sewerage. Since the flow velocities in the modified
drainage network are generally higher than those observed in the orlginal
natural channel system, both the time-to-peak and the length of the recession
of storm hydrographs tend to decrease as a catchment is urbanised. The increased
volume of runoff, and the shorter time within which that volume is discharged,
inevitably produce peak rates of runoff that are markedly higher than the flow
records from a catchment in its previous rural state would tend to indicate.

Although the effects of urbanisation on the flow regipie of a catchment

area have been appreciated qualitatively for over a decade [2], relatively
little information has been available on the magnitude of the changes brought
about by different forms of urban development. Of particular importance to
the engineer concerned with the design of flood alleviation works for urban
areas are
i) the frequency distribution of peak rates of flow ; and
ii) the shape of the flood hydrograph.

The changes in the magnitude of the parameters of the frequency distri-

bution of annual floods caused by urbanisati)onhavebeen studi,edby Carter [3],
Martens [4] and Anderson 151, each of whom approached the problem by means of
regional analysis. Much of the work on changes in the shape of flood hydro-
graphs has employed a similar treatment, with the finite-period unit hydro-
graph (TUH) being used as an index to catchment response. Flow records for
catchment areas in different stages of urban development within the same
hydrologically homogeneous region have been used to derive WH's of a pre-
determined duratiqn. Selected parameters of these !CUH's have then been expressed
in terms of pertinent catchment Characteristics using multiple linear regression
analysis. The relationships so obtained may then be employed to derive TUHts
for both ungauged catchments and gauged catchments in a more advanced state of
development. For example, Espey et al [6] used 5 hydrograph parameters : peak
rate of flow ; time-of-rise and base length of the hydrograph ; and hydrograph
widths at 50 and 75 per cent of the peak discharge. Since the catchment
characteristics selected by those Authors did not ipclude any parameter reflecting
changes in the surface water drainage system, an empirical coefficient (<I .O)
was required to gain agreement between predicted and observed times-of-rise
for urban catchment areas. The nature of this coefficient precluded its unmodi-
fied extrapolation to other areas 171. More recently, Rao et al [8] have shown
 487

that simple one and two-parameter linear conceptual models can be used to
advantage in characterising the changes iq catchment response caused by urban-
isation. However, a prerequisite to either approach is the availability of
hydrometric data for a sufficiently large number of urban and rural catchment
areas to effect a regional analysis. Where the number of flow records is
limited, techniques which employ as few hydrograph parameters as possible are
an obvious advantage.

In the following paper, a dimensionless unit hydrograph technique which

involves the use of only one parameter is outlined. The method of approach
is illustrated by means of data from an area in the south-east of England.
The paper begins in Section (2) with a brief description of the area and the
available hydrometric data, and continues in Section (3) with an outline of
the method by which TüH's were derived. The regionalisation of these TüH's
is discussed in Section (4). The paper concludes in Section (5) with a brief
diqcussion of the existing data inadequacies in Urban Hydrology.

2. DATA PRXPARATION

2.1 Data inventory

Between 1949 and 1969,the population of Crawley, a town situated some

30 miles to the south of London, increased from 5,000 to 68,000. The
developed area i? drained by the headwaters of the River Mole, a south-bank
tributary of the River Thames. !Che western side of the town drains to Ifield
Brook, whereas the centre and eastern sides are served by Crawters Brook and
Gatwick Stream respectively (see Figure 1). A major part of the urban area
lies on Weald Clay overlying Tunbridge Wells Sand, the latter outcropping to
the south of the area. The average annual rainfall in the Crawley region
(1916-1950) ranges from 750-850 rnrn.
There are 6 gauging stations within the area of interest, the details
of which are summarised in Table 1. Both the gauging statipns on the River
Mole and that on Gatwi.ck Stream are operated by the.Thames Conservancy ;
Crawley Urban District Council maintain the records at the remaining 3 sites.
For the purposes of the present study, data were available for all sites apart
from the River Mole at Gatwick Airport.
TABU 1 : Details of gauging stations within the Crawley region.

ref. river station catchment records

no. from

1 Mole Horley Weir 89.8 Nov., I961

- Mole Gatwick Airport 31-8 Nov., 1967
2 Gatwick Stream Tinsley Sewage Works 31.o Jul., 1952
3 Ifield Brook Ifield Mill 12.3 Dec., I958
4 Crawters Brook Hazelwick Roundabout 4.7 May, 1954
5 Crawters Brook Woolborough Road 2.2 Sew.. 1952
488

The positions of the 3 principal autographic raingauges located within

the headwaters of the River Mole are indicated on Figure 1 along with the
gauging stations. 2 of the 3 raingauges have been in operation since before
the first regular streamflQw measurements were taken at Tinsley Sewage Works
and Woolborough Road, and records from all 3 raingauges are available from
before 1961 when the two gauging stations on the River Mole were brought into
use.

2.2 Selection of storm events

The first stage in the analysis of the available data involved the
preparation of a short-list of suitable storm events for each of the £ive
catchment areas. The criteria used in choosing these events were somewhat
arbitrary, but in genera1,anattempt was made to confine the analysis to
hydrographs with well-defined peaks having both a smooth rising limb and a
smooth recession. Rainfall data for each of the selected storm events were
then abstracted. The raingauge at Broadfield was taken to be representatjve
of the rainfall patterns over the catchment areas draining to gauging statipns
3, 4 and 5 (see Table I>, and the arithmetic mean of the catches at Broadfield
and Gatwick Airport was taken for the areas commanded by gaugipg stations 1
and 2. The records from Tinsley Sewage Works were only used when no information
was availab1.e at either of the other gauges.

The above selection procedure produced 8 storm events at gauging statipn

3, 11 at gauging station 1, 12 at gauging station 4 and 16 each at gauging
stations 2 and 5, the majority of whi,chwere associated with rainfall totals
exceeding 12 mm.

2.3 Baseflow separation

The second stage in the analysis consisted of the separation of the base-
flow component from each of the recorded streamflow hydrographs, The procedure
adopted involved the plotting of the recession limb of each hydrograph on
semi-logarithmic graph paper with discharge on the logarirchmjc scale. A
straight line was then fitted by eye to the lower portion of the curve, the
point at which the recession departed from this straight line being taken to
mark the time at which surface runoff effectively ceased. The variatipn of
baseflow with time during the storm was then represented by a straight line
joining this point on the recession limb to the beginning of the rising limb
of the hydrograph.
The above method of baseflow separation, which is both straightforward
in use and less subjective than the majority of the alternatgve procedures
was applied to each of the 63 recorded hydrographs selected for analysis.
The ordinates of the resultant surface runoff hydrographs were then abstracted
at I-h intervals for all events at gauging stations 1-4 and at 30-min intervals
at gauging station 5. These data were subsequently transferred on to 80-
column punched cards along with the total recorded rainfalls witpin the same
time incrementso
 489

3. DERIVATION OF UNIT H!¿DROGRAPHS

There are two distinct methods of approach to determining the instant-
aneous unit hydrograph (IW) or finite-period unit hydrograph (TUH) of a
catchment area from rainfall and streamflow data [g]. The first of these
methods of approach involves the fitting of a linear conceptua1,modelto the
records of rainfall excess and surface runoff. The 4mpulse response functi.on
of the fitted mode1,is then taken to approximate the IUH of the catchment.
This indirect synthesis approach may be contrasted with the more direct methods
of analysis which operate on the rainfall excess and surface runoff data to
yield an IUH or TLTH wi,thout the need to postulate a model. The harmoni? method
for defining the TUH of a catchment [9], which was adopted for the purposes
of the present study, falls into the latter category.

3.1 The harmonic method of unit hydrograph derivation

In order to apply the harmonic method, the volumes of raiyfall excess

and the ordinates of both the surface runoff hydrograph and the TIM are defined
in terms of harmonic series. For example, if the equally-spaced ordinates of
the surface runoff hydrograph are given by yi, i = 1, 2, n, ....,

Yi.
= C[A~
+ P
j=l
-
cos j 2xi
n
+

j
sin j &]
n

If n is an odd number, p = (n-1)/2 and

B = 2
n
c y s i n k -;i
k
j k=1

The volumes of rainfall excess, xi, i = 1, 2, ...,

m, within the same
equal time increments may also be expressed as a harmonic series with the same
fundamental period and number of terms if (n-m) zeros are added to represent
the terms xi, i = m+l, m+2, ....,
n. This series will be identical in form
to equation (I), but with n harmonic coefficients a, b whose values can be
obtained by substituting rainfall excess volumes for surface runoff ordinates
in the equations (2). If the TITH is also assumed to have n equally-spaced
ordinates, Le. the same fundamental period as that of the rainfall excess
and surface runoff data, O'Donnell 191 has shown that the harmonic coefficients
a, ß, of the harmonic series which defines the ordinates of the TUH can be
calculated directly from the harmonic coefficients A, B, a, b usipg the linkage
equations

a
j
= -
n
a.A. + b.B
a. +b
but cio
-- -1
n
%
aO
~j eq. (3)

Pj -
- n
2 w
aj2+bj 2
 49 O

Substitution of the aj, ßj in a series expansion of the form of equation

(1) then gives the successive ordinates of the TUH, ui, i = 1, 2, ....,n
directly .
The application of any of the established methods of analysis, such as
the harmonic method, is liable to produce TUH's which are distorted by high-
frequency oscillations of varying amplitude. PhiiliFpee and Wiggert [IO] who
applied the harmonic method to data from 38 storms on 4 drainage basins in
thr vicinity of Detroit, encountered this problem but offered no explanation
as to its cause. More recent studies by Blank et al [Il], who used the Fourier
transform approach, which bears some relatipnship to the harmonic method, have
indicated that such oscillations can result from errors in the data and are
not necessarily caused by the inherent non-linearity of the rainfall-runoff
relationship. Blank et al [Il] also show that oscillatory TUH's can be avoided
by applying a low-pass digital filter to the rainfall excess and surface run-
off data prior to the derivation of the TUH.

One of the principal advantages of the harmonic method is its flexibility

in dealing with storm events which produce such oscillatory "UH's without the
need to use digital filters. This property of the method stems from the form
of the linkage equations (3) by which the aj, ßj of the harmonic series repre-
sentation of the TKH depend only on the harmonic coefficients of the rainfall
excess and surface runoff data for the same frequency. Individual harmonics
may therefore be omitted from the series representation of the TUH without
affecting the calculation of other üj, ßj. In particular, if the ordinates
of the TUH obtained by using all the aj, ßj exhibit high-frequency oscillations,
truncation of the series representation may help to eliminate these
oscillations. However, the amount of truncation applied should not be suffic-
ient to cause the hydrograph obtained by convolving the smoothed Tw with the
distribution of rainfall excess to depart significantly from the original
surface runoff hydrograph.

3.2 Appliiation to Crawley area data

A computer program was written to derive TUH's using the harmonic method
described ip Section (3.1) above. The computation began with the determination
of the distribution of rainfall excess using the @-index method. The total
volumes of both rainfall and runoff were calculated and their difference
averaged over the number of time intervals with non-zero rainfall. This
average t'lossttwas then subtracted from the recorded volumes of rainfall w5thin
each time interval, any negative differences being set to zero. The whole
procedure was repeated until the difference between the total volumes of
rainfall and runoff was less than 0.25 mm. Having obtained the distribution
of rainfall excess, the derivation of the TUH was carried out according to the
method outlined in Section (3.1) above. The surface runoff hydrograph was
then reconstituted by convolving the derived TUH with the distribution of
rainfall excess.

The data from all 63 storm events were processed using the full number
of harmonic coefficients in determining the ordinates of the TUH. The results
obtained were then plotted and compared. The majority of the derived TUH's
were found to exhibit high frequency oscillations of varying amplitude. The
storm events which gave rise to such behaviour were therefore re-processed
using fewer harmonic coefficients in determining the TUH ordinates.
 491

The choice of the most appropriate number of harmonic coefficients to

use for any given storm event is largely subjective. Truncating the harmonic
series representation of the TLTH may remove the high-frequency oscillations,
but the hydrograph obtained by convolving that TUK with the distribution of
rainfall excess should not depart markedly from the original surface runoff
hydrograph, particularly in regard to the magnitude and timing of the peak
flows. The amount of computer time that would have been involved in
systematically reducing the number of harmonic coefficients in the series
representation of the !
ì'
ü
ñ until the fit provided by the reconvolved surface
runoff hydrograph was no longer acceptable would have been excessive. A pilot
study using a restricted number of truncated,,series,eachhaving a predetermined
proportion of the full number of harmonics, was therefore carried out. For the
majority of the storm events, halving the number of harmonics successfully dampened
the high-frequency oscillations, and gave rise to a reconvolved hydrograph
whose maximum ordinate was generally within 5 per cent of the peak of the
original surface runoff hydrograph.

As a result of the above analysis, 8 TiJH's were obtained for gauging

station 2, 6 each for gauging stations 4 and 5, 4 for gauging station 1 and
3 for gauging station 3. The changes in flow regime which had occurred at
gauging stations 2, 4 and 5 during the period of record were immediately
obvious, and TiJH's for each of these sites were therefore grouped according
to the dates of occurrence of the storm events from which they were derived.
For convenience, these different groupings will be referred to by the letters
IfAt1 (for the earlier storms) and "BI1 (for the later storms). Of the 8
separate sets of TUH's, none consisted of less than 3 hydrographs. The TUH's
within each set were then plotted together using a common starting time, and
an "average" TiiH obtained by drawing in a smooth curve through the plotted
points, care being taken to ensure that the area under the curve was equivalent
to 25 mm over the catchment area.

Figure 2 shows the smoothed TITH'S obtained for Crawters Brook at Woolborough
Road, and is indicative of the change in flow regime which has taken place as
the town centre of Crawley has developed over a period of some 15-20 years.

4. REGIONALISATION OF UNIT KYDROGRAPHS

One of the simplest assumptions that can be made in regiqnaliTi9g a group

of unit hydrographs is that all TLTH'S of a common duration are reducible to
the same dimensionless shape. The scaling parameters that are required to
describe the dipensionless hydrograph (of which there are generally two) are
expressed in terms of catchment characteristics by means of a multiple linear
regression analyses. The appljcation of this approach,$ the present study
is complicated by the necessity to include independent variables which reflect
thg man-made changes within the catchment areas affected by the development of
Crawley.

Previous authors who have applied a dimensionless unit hydrograph approach

have differed widely ip their choice of scaling parameters, For example,
Commons u23 developed a l'basic hydrograph" with a time base of 100 arbitrary
units, a height of 60 arbitrary discharge units and an area of 1196.5 square
units. The scaling parameters required were peak rate of runoff and total
volume of runoff. In common with many similar pairings, these parameters
 49 2

are not entirely independent, and as Diskin [I31 has recently pointed out, a
choice of parameters which satisfy the constraint of unit area under the TLTH
is to be preferred.

A review of previously-published work on the hydrological consequences

of urbanisation showed that, of several possible time scaling parameters, the
lag time TL, defined as the time interval between the centroid of rainfall
excess and the centroid of surface runoff, has been found to show a consistent
variation with the length and slope of the main channel for both rural and
urban catchment areas 13-53. If, however, the constraint of unit area under
the TUH is observed, making the time scale of each TüH dimensionless by expressipg
the timing of all ordinates as a proportion of TL also determines the ordinate
scale. Hence, the dimensionless unit hydrograph can be specified in terms of
only one parameter, the functional form of the curve being

ut.TL = f (t/TL) eqe(4)

where ut is the ordinate of the actual TUH at time t.

The above method of producipg a dimensionless unit hydrograph was applied

to the data obtained from the 5 catchments in the Crawley area. The lag time
of the I-h TUH for each catchment was obtained by computing the tjme interval
between the origin of the TüH and its centroid and subtracting O.5h. Following
Carter [3) and Anderson [5], a double-logari4hmic plot of lag time against
basin ratio was prepared (see Figure 3). Basin ratio is defined by the quotient
,
@
L where L is the length of the main channel between the gauging station
and the watershed (km) and S the main channel slope. S is defined by the alti-
tude difference between points located 10 and 85 per cent of the main channel
length upstream from the gauging station divided by their distance apart [14].
Values of L and S were obtained from 1 : 25000 scale maps for all catchment
areas apart from that of gauging station 5 for which 1 : 500 longitudinal
sections were available.

In preparing Figure 3, the number of data was increased by the inclusion

of TUH's for 2 gauging stations on the River Wandle in the southern suburbs
of London to the north of Crawley (see Table 2). These hydrographs, which
were obtained by Nash [l5], relate to conditions before and after the execution
of channel improvement works. The data tabulated by Nash (loc. cit. Table 3,
p.323) were assumed to relate to a duration of 30 min. The reduction in lag
time shown by the Itpost-works" hydrographs was estimated by Nash from the
records for 3 adjacent catchment areas whose channels were considered to be
in an equivalent state to the improved conditions on the River Wandle. The
"post-works" hydrographs were therefore synthetic and not derived directly from
recorded data. Nevertheless, the data were considered to be useful in providing
an independent measure of the influence of channel improvement works without
a simultaneous growth in the impervious area within a catchment.

TABLE 2 : Details of gauging stations within the River Wandle catchment area
(from Nash [IS] 1
river station catchment
 493

Also plotted in Figure 3 are the lag timebasin ratio relationships

obtained by Anderson [5] for catchment areas in three different stages of
development. Drainage class N refers to natural (rural) areas. Drainage class
B includes areas in which the impervious cover ranges from 20 to 30 per cent,
the tributary streams are sewered but the main channels are retained in their
natural state. Drainage class U refers to fully developed urban areas having
more than 30 per cent impervious cover and all stream channels completely
sewered or improved and realigned. The majority of the gauging stations used
by Anderson in deriving these relationships were situated within the Washington
D.C. metropolitan area.

Examination of Figure 3 shows that data for gauging stations 2 and 3

exhibit markedly longer lag times than would be predicted from Anderson's class
N relationship. Since gauging station 3 is situated at the outfall of a mill
pond covering an area of some 8-10 ha, such behaviour is to be expected. The
presence of a lake of similar size within the headwaters of Gatwick Stream
catchment has a similar if not as pronounced an effect on lag time. In contact,
the data for gauging stations 1, 4A, 5A, 6A and 7A all show reasonable agree-
ment with the Anderson class N relationship, although bearing in mind the
channel improvements above gauging station 5 and the existing urban development
above gauging station 6, the broken line drawn above and parallel to Anderson's
equation is perhaps a better approximation to natural catchment conditions
within the Crawley area.

Figure 3 shows that at gauging station 5, an increase in the proportion

of impervious cover (as estimated from data supplied by Crawley Urban District
Council) from 5 to 26 per cent is associated with a reduction in lag time of
23 per cent, whereas at gauging station 4 an increase in impervious area from
18 to 27 per cent apparently causes a reduction in lag time of 72 per cent.
The latter anomaly is thought to result from extensive renewal of sewerage
within the catchment which took place concurrently with the increase in paved
area. The data from gauging stations 6 and 7 indicate that channel improvement
(not including installation of sewerage) can cause a 30-40 per cent reduction
in lag time. The results obtained at gauging stations 4 and 5 are therefore
not as inconsistent as they might at first appear.

Figure 3 also shows that the lag times for Anderson's developed (class B)
and fully developed (class U) catchments are markedly shorter than those
observed in the Crawley area. The lower broken line, drawn parallel to and
immediately above Anderson's class B relationship, is probably the best approx-
imation to the behaviour of developed catchments with approximately 30 per cent
impervious cover and improved channel systems within the Crawley area that the
available data will allow.

Having obtained the relationship between the chosen scaling parameter and
two readily-computed catchment characteristics, only the form of the dimension-
less curve is required to construct the I-h TUH for an ungauged catchment
within the Crawley area. Accordingly, the 8 observed TITH'S whose derivation
was described in Section (3) above were reduced to the form of equation (4)
using the appropriate observed values of lag time (see Figure 4). A single
dimensionless hydrograph was then fitted by eye to the plotted points, care
being taken to ensure that the area under the curve was unity.
 494

In practice, application of the method to produce a I-h TUH for an ungauged

catchment may be summarised as follows :

i) measure the length and slope of the main channel of the catchment area
from a 1 : 25000 Ordnance Survey map, and compute the basin ratio ;
ii) use Figure 3 to estimate the lag time of the catchment for a particular
stage of urbanisation ; and
iii) given the lag time, use the dimensionless unit hydrograph of Figure 4
to construct the I-h TLTH of the catchment.

5. CONCLUDING REMARKS
Since the procedure outlined above uses the lag time as the only scaling
parameter, there is an obvious analogy with the single linear reservoir model
whose storage constant is equivalent to the lag time as defined in the present
study. The major difference between the two approached lies in describing
the TUH by means of a series of plotted points, rather than an equation. The
peak of any TUH constructed from Figure 4 is therefore constrained to occur
at a specific proportion of the lag time rather than at a time equivalent to
the duration of the unit hydrograph.

According to Rao et al 181, the single linear reservoir model provides

an adequate description of the behaviour of both urban and rural catchments
less than approximately 13 km2 in area. Those Authors obtained an expression
for lag time in terms of volume and duration of rainfall excess and proportion
of impervious cover. The results of the present study (in particular, Figure 3)
tend to indicate that, for the area under study, proportion of impervious cover
provides a less than adequate description of the man-made changes within a
drainage basin. In the absence of additional parameters relating to changes
in the channel system, and perhaps distribution of impervious cover with respect
to the outfall of the catchment, the engineer concerned with the design of
flood alleviation works must rely on diagrams such as Figure 3 whose construct-
ion is unfortunately largely subjective and highly dependent on local knowledge
of the area.
The urbanisation of a catchment area provides one of the most dramatic
examples of man's interference with the hydrological cycle. Whereas the
expansion of any conurbation creates an increasing water demand for domestic,
industrial and recreational purposes, the very presence of the urban area
accelerates the processes by which locally stored ana precipitated water is
returned to the sea. Despite the major changes in the flow regime of a catch-
ment area which urbanisation can bring about, relatively little attention has
been given to the quantification of such changes when compared with other land
use changes,such as that of forest to grassland. Bearing in mind the large
sums which have been and are being devoted to flood protection schemes for
urban areas, the available information can justifiably be labelled as inadequate.
 495

ACKNOWLEDGEMENTS

The study described above was carried out on behalf of the Resources
Group for West Sussex County Council. The author wishes to thank Dr. T.M.
Prus-Chacinski, partner, C.H. Dobbie and Partners, for his encouragement
to prepare and permission to publish this paper. The assistance received
from the Chief Engineer, Thames Conservancy, Mr. E.J. Brettell, and the
Engineer and Surveyor, Crawley Urban District Council, Mr. H.J. Lumley, in
providing hydrometric data was also greatly appreciated.

REFERENCES

1. Toynbee, A. (1970). Cities on the move, Oxford Univ. Press, 257 pp.

2. Savini, J., Kammerer, J.C. (1961). Urban growth and the water regime,
U.S. Geol. Survey, Water-Supply Pap. 159l-A, 43 pp.

3. Carter, R.W. (1961). Magnitude and frequency of floods in suburban

areas, U.S. Geol. Survey, Prof. Pap. 424-B, pp. B9-SII.

4. Martens, L.A. (1968). Flood inundation and effects of urbanisation in

metropolitan Charlotte, North Carolina, U.S. Geol. Survey, Water-Supply
Pap. 1591-C, 60 pp.

5. Anderson, D.G. (1970). Effects of urban development on floods in

Northern Virginia, U.S. Geol. Survey, Water-Supply Pap. 2001-C, 22 pp.

6. Espey, W.H., Morgan, C.W., Masch, F.D. (1965). A study of some effects
of urbanisation on storm run-off from a small watershed, Centre for Res.
in Wat. Resour., Univ. of Texas, Teoh. Rept. KYD 07-65OI,.CRWR-2, IO9 pp.

7. Espey, W.H., Winslow, D.E., Morgan, C.W. (1969). Urban effects on the
unit hydrograph, in Moore, W.L., Morgan, C.W. (eds.), Effects of watershed
changes on streamflow, Proc. Wat. Resour. Symp. no. 2, Centre for Res.
in Wat. Resour., Univ. of Texas, Univ. of Texas Press, pp. 215-228.

8. Rao, R.A., Delleur, J.W., Sarma, B.S.P. (1972). Conceptual hydrologic

models for urbanising basins, Proc. Am. Soc. Civ. Engrs., J. Hydraul.
Div., 98 (KY71,pp. 1205-1220.

9. O'Donnell, T. (1966). Methods of computation in hydrograph analysis and

synthesis, Recent trends in hydrograph synthesis, Proc. Tech. Meeting
no. 21, T.N.O., The Hague, pp. 65-102.

IO. Philippee, J.T., Wiggert, J.M. (1969). Instantaneous unit hydrograph

response by harmonic analysis, Wat. Resour. Res. Centre, Virginia
Polytechnic Institute, Bull. 15, 36 pp.

11. Blank, D., Delleur, J.W., Giorgini, A. (1971). Oscillatory kernel

functions in linear hydrologic models, Wat. Resour. Res., 7, pp. 1102-1117.

12. Commons,, G,G. (1942). Flood hydrographs, Civ. Engrg. (New York), 12,
pp. 571-5720
496

13. Diskin, M.H. (1972). The role of lag in a quasi-linear analysis of the
surface runoff system, paper presented at the 2nd Internat. Hydrol. Symp.,
Fort Collins, Colorado.

14. Benson, M.A. .

(1959) Channel-slope factor in flood-frequency analysis,
Proc. Am. Soc. Civ. Engrs., J. Hydraul. Div., 85 (Kyk), pp. 1-9.

15. Nash, J.E. (1959). The effect of flood-elimination works on the flood
frequency of the River Wandle, Proc. Instn. Civ. Engrs., 13, pp. 317-338.
497
W
E
498
P
O
3 6 O
g" Sr
O
W
H/L '3ìVNIOäO HdVMCIOtKIAH IINíl
g o
 499

20

<.. 10
W
I
+
0
3 50

2.0

lo

0.5

0.2

0.1
0.5 1.0 ZD 10.0 2OD 40.0
BASIN RATIO, Z , K M

PLOT OF LAG TIME AGAINST BASIN RATIO FOR THE CRAW!.EY AWEA.
500
* o
v)
2 5
i 2
LC3
 A DIMENSIONLESS UNITGRAPH FOR HONG KONG

P. R: HELLIWELL
Department of Civil Engineering, University o f Southampton.

T.Y. CHEN
Royal Observatory, Hong Kong

ABSTRACT

The large number of individual catchments in Hong Kong makes it

impracticable to measure stream flows on all but a small proportion
of streams. Rainfall characteristics and topography are similar over
much o f the area.
Using data for several storms at ea-h of the seven stream gau-
ging stations, a mean dimensionless unitgraph was derived. Basin lag
was used in the conversion o f both time and discharge scales. For un-
gauged catchments basin lag can be estimated either as a simple func-
tion o f catchment size, shape and slope.
This work was based on records collerted in 1 9 6 4 and 1 9 6 5 , and
was one of the first studies made possible by the installation of a
network o f hydrometric stations in Hong Kong.

RESUME

Le grand nombre de bassins fluviaux du territoire de Hong Kong

fait qu'il n'est possible d'effectuer des mesures de débit que s u r un
faible pour centage d'entre eux. La pluviométrie et l a topographie
présentent des caractéristiques semblables sur la plus grande partie
du territoire.
E n s'appuyant sur les données recueillies à .ept stations de j a z
geage au cours d'un certain nombre d'averses, on a mis au point un h l
drogramme unitaire moyen sans dimension. Le temps de réponse du bas-
sin intervient dans les conversions à la fois pour l'échelle des temps
et pour celle des débits. Pour les bassins qui ne tont pas l'objet de
m sures des débits, le temps de réponse peut être estimé soit simple-
ment en fonction de la surface du bassin, soit en fonction de s a tai-
lle, de sa forme et de sa pente.
La présente étude est basée sur des observations recueillies e n
1 9 6 4 et 1965; ce fut une des premières qui aient été [,endues possibles
par l'installation d'un réseau hydrométrique dans le territoire de
Hong Kong.
502

Introduction

The British Crown Colony of Hong Kong is located on the coastline

of China, just inside the Tropic of Cancer at latitude 220N and longitude
114O. The land area of the Colony is approximately 1000km2, comprising
a section of the mainland, the islands of Hong Kong and Lantau, and a
large number of very small islands. The total area, including sea, is
approximately 2500km2.

It is an area of high relief, the highest point being over 1OOûm

above sea level. Drainage lines are short, usually less than lokm, giving
a large number of small steep catchment areas draining to the very long
coastline. In some areas, mast notably in the northwest, there is an area
of almost flat land between the hills and the sea shore, formed by silting
up .If shallow bays, and subsequent uplift of the land relative to sea level.
These areas are intensively cultivated, with vegetable crops replacing the
traditional rice cultivation where levels are high enough to be clear of
sea water intrusion, and fish ponds starting to replace brackish water rice
cultivation near sea level. Much of this flatter land, particularly near
the larger stream channels, is natural floodland, which makes stream
gauging at high flows very difficult.

The vegetation of the upland areas is of coarse grasses or mixed scrub-

land. Gathering wood for firewood and traditional seasonal burning of hill-
side vegetation tend to degrade the cover. Soils on the hills are coarse,
thin, and poor. Gullying, sometimes severe, occurs mainly in the west of
the Colony. In the East, the grass cover is complete, and sediment loads
are very low. The geology is predominantly granitic. Soil moisture and
groundwater storage are small.

Climate is seasonal. Winters are cool and dry, although periods af

li-ght rain do occur. Summers are warm (daily maximum temperature up eo 35W,
fvith very little diurnal variation) and wet. The mean summer half year rain-
fall at the Royal Observatory is 1850mm and the mean winter half year rain-
fall is 350mm. Observation-day rainfalls in excess of 250mm occur in most
years. Frosts can occur at levels above 600m, but snow does not fall.
Annual rainfall elsewhere in the Colony varies between 1250mm and 3000m.

Water supply has always been a major problem in Hong Kong. The small
size of catchment areas, the seasonal nature of rainfall, and occasional
severe droughts have presented a major challenge to the water engineers(l1.
Of necessity, reservoirs with small direct catchments have been built, and
water has been broughtin from much larger areas by systems of catchwater
channels or tunnels, intercepting many small streams which would otherwise
discharge to the sea. More recently, arms of the sea have been converted to
freshwater storage, at Plover Cove and at High Island.
 503

Measurement of Streamflow

Measurement of streamflow in all catchments in Hong Kong is obviously

impossible, The approach adopted has been to make measurements in a
relatively small number of basins spread through the Colony, and to transfer
data from these to other basins. This paper describes the method used to
generalise flood hydrograph data, and presents the resulting dimensionless
unitgraph.

Streamflow is measured by fixed structures. Sharp-edged and crump

weirs of various compound profiles, triangular and trapezoidal flumes,
Parshall flumes and broad-crested diversion weirs are all used. Data are
published annually(2).

Selection of Records for Study

Examination of records from streamflow stations and of autographic

rainfall records for sites in or near to gauged catchments showed that three
or more storms suitable for analysis were available at seven stations. There
were seven more catchments which had been or were being gauged, but no suit-
able records for this analysis were found among them, the commonest problem
being submergence of the measuring structure at high flow. Table 1 gives
notes on all streamflow stations, and Figure 1 is a map showing stations used
in this study. From the table it will be seen that all catchments are small.

Method of Analysis

The method used was that described in USBR Design of Small and
by Linsley, Kohler and Pa~lhus(~).

Studies of base flow had shown that depletion could be assumed to be

of the type qt = qokt. Base flow separation was achieved by plotting on log-
arithmetic graph paper. No attempt was made to separate interflow.

Duration of effective storm rainfall was found by applying the$-index

technique to the hyetograph, having found the total storm runoff by
integration of the storm runoff hydrograph. Storms with up to five unit
periods of excess precipitation were used.

Successive approximation procedures were used to find the unitgraph

ordinates.

In order that the period should be less than one third of the rise time
of the unitgraph (to avoid instability in the computations), it was necessary
to use a unit period of 15 minutes, except in the case of the smallest catch-
ment where 74 minutes was used.
504

With such short periods, the accuracy of timing of the chart records
of streamflow and rainfall is critical. In the cases of Hok Tau and
C’iung Mei, unshielded, tilting bucket rain gauges were sited on the stream
recorder house roof in order to ensure correct relative timing. Unfortunately,
the mechanism transferring the tipping bucket record to the chart proved
u.iieliable, and some records were lost. With the other stations, it was
hoped that timing errors would average out, but there is no evidence on this.
From experience, errors in long-period chart records using chart drives of
ZVknlday can be kept to less than five minutes by making corrections based
on check observations. On the other hand, standard daily autographic rainfall
recorders in the hands of all but the most careful observers can often show
fluctuations from correct time of ten to 15 minutes.

The measure of agreement between various unitgraphs for a given catchment

varied. Two examples, one showing consistent behaviour, and another showing
rather poor agreement, are shown in Figure 2. Average unitgraphs for each
catchment are shown in Figure 3. These were formed in the usual way by
averaging time to peak, magnitude of peak and total duration, sketching in
a mean shape, and adjusting the area under the curve.

The variation of the mean unitgraphs can be seen in Figure 3. A means

of unifying these was required. They were made dimensionless in terms of the
time to the centroid of the unitgraph and the volume of unit rainfall excess.

Time-axis values were divided by time to centroid of unitgraph and

discharge values were multiplied by time to centroid of unitgraph and divided
by the volume of unit depth of runoff over the catchment area.

The seven dimensionless unitgraphs and the mean dimensionless unitgraph

found from them are shown in Figure 4. The ordinates of the dimensionless
unitgraph are listed in Table 2.

Application to ungauged catchments

The size range of individual catchments included in the analysis

adequately covered the sizes of catchments found in Hong Kong. Similarly,
geographical distribution was quite good, only the eroded area in the west
being excluded. Catchment and stream slope variability was not so well
covered. The Tai P o Tau catchment included some lowland area, as did the
Kam Tin catchment. The other five were upland in type.

The time to the centroid of the unitgraph for the seven catchments is
the basin lag (i.e. time from centre of area of excess rain to centre of
area of hydrograph of excess runoff) plus half the unit period, (lag + 9).
 505

The basin lag for the mean unitgraph of each catchment was plotted
against catchment area and against-where L is the length of the main
s
stream projecked back to the catchment divide, as measured on 1:25000-
scale maps, Lc is the distance along the stream from the gauging station
to a point on the main stream nearest to the catchment centre of area,
and S is the stream slope as estimated by the difference in elevation of
the main stream at the catchment divide and the gauging station divided
by L). The correlation coefficients were 0.92 and 0.86 respectively. It
was significant that the Kan Tin value fell close to the regression line
when slope was included, and off the line where area alone was used.
However, in all work using the hydrograph, catchment area alone has been
used. Figure 5 6 show the relationship. Figure 6 also shows data from
Linsley et a d 4 j and Design of Small Dams(3).

The equations for estimation of catchment lag in ungauged basins are:

lag = 0.47 areaoss4 with lag in hours,area in km2

lag = 0.36 ( 3 ) 0 . 4 0 with lag in hours and lengths in km

si
To apply the unitgraph to any particular storm it is necessary to
estimate a @-index value, or loss rate. Studies of this for Hong Kong
conditions showed wide fluctuations between storms, ranging from 2.5 to
80m/h, with values commonly between 10 and 40mm/h. Judgement must be
used in selecting a suitable value. When reservoir spillway studies are
in hand, a very low value is appropriate. For drainage design, a value
nearer the mean would be used.

This dimensionless unitgraph has been used in conjunction with studies

of probable maximum precipitation over Hong Kong(5~6) carried out by the
staff of the Royal Observatory, to check capacity of existing reservoir
spillways and in the design of new dams in Hong Kong. Flood frequency
analysis has also been used but in the absence of long records this is
thought to be less reliablef7).

Conclusions

The dimensionless unitgraph derived by procedures developed in the

U.S.A. is useful as a design tool. Hydrographs from the seven catchment
areas, covering the range of sizes and types found in Hong Kong were unified
to an acceptable degree of accuracy using time to centroid of unitgraph in
converting scales into dimensionless values.
 506

Whether area alone or a more complex parameter should be used for

predicting basin lag is uncertain. Area alone appeared to be adequate
except in the case of catchments with extensive lowland area.

The volume of data available for this study was small both in terms
of length of records used and number of stations.

Acknowledgements

Thanks are due to the Director of Public Works, Hong Kong Government,
for permission to publish this paper; to Mr. J. Forth, who later extended
the work on floods to other methods of approach; and to Mr. Wong Shiu Ming,
present holder of the post of EngineerIHydrologist, for his valued ascist-
ance in checking the data in this paper and providing information.

References

1. Robertson, A.S. and La Touche, M.C.D., Assessing the Yield of Hong Kong's
Reservoirs, J. Institution Water Engineers, 23, (1969), 8, 507-519.

2. Hong Kong Rainfall and Runoff (Annually from 1965), Hong Kong, Water
Authority, Public Works Department.

3. United States Bureau of Reclamation. Design of Small Dams, (1960),

Washington, U.S. Govt. Printing Office.

4. Linsley, R.K. , Kohler and Paulhus, Hydrology for Engineers , (1958) ,

New York, McGraw-Hill.

5. Bell, G.J. and Chin, The Probable Maximum Rainfall in Hong Kong.
R.O. Tech. Mem. 10, (1968). Government Printer, Hong Kong.

6. Cheng, S. and Kwok, (1966) A Statistical Study of Heavy Rainfall in

Hong Kong. Tech. Note 24, Hong Kong, Royal Observatory.

7. Design Flood for Hong Kong, HS7, (1968), Water Authority, Public Works
Department, Hong Kong.
 Streamflow Stations in Hong Kong, to 1966

Station
-
Altitude
of Crest
Catchment
Area Readings Record
Name
I_
-m -@ Contro 1 Instrument Ob s erving P r o g r a m e Commenced Quality Remarks

Tai Lam Chung 15 16.2 Compound weir with ogee crest, Staff gauge. Frequent staff gauge readings Apr. 1948 Fair Discont inued
60ft. weir loft. low flow section. before Jan. 1950, thereafter May 1955
Locally-made float level recorder.
continuous recording with
daily observations.

Sham Tseng 30 2.0 30ft. compound weir with ogee crest, Continuous recording with
Jul. 1952 Fair Discontinued
Staff gauge. June 1956
3ft. low flow section. George Kent float level recorder. daily observations.

Tai Lam Chung 75 0.8 Compound V and rectangle sharp- Frequent staff gauge readings Jun. 1958 Good
Sloping brass staff gauge.
'A' crested suppressed weir. Munro vertical drum before June 1963, thereafter
continuous reading with daily first
and then weekly observations.

Tai Lam Chung 63 1.2 Compound V and rectangle sharp- Frequent staff gauge readings Jun. 1958 Poor
Sloping staff gauges.
'B' crested suppressed weir. before June 1959, thereafter daily
observations.

Shek P i Tau L 41.6 102ft. long with 4ft. wide broad- Daily observations before June 1964, May 1960 Poor
Sloping staff gauge.
crested weir. Munro vertical drum float level thereafter continuous recording
recorder. with bi-daily observations.

Ho Sheung Heung 5 16.9 40ft. long broad-crested weir. Daily observations before June 1964, May 1960 Poor
Sloping staff gauge.
Munro vertical drum float level thereafter continuous recording with
recorder. di-daily observations.

Tai P o Tau 9 15.2 Broad-cres ted weir. Oct. 1960 Fair Dis continued
Sloping staff gauge. Continuous recording from July 1961
Aug. 1963
Munro horizontal drum float eve1 to April 1963, daily observations
recorder. at other times.

Sha Tin 100 1.2 Compound sharp-crested rectangular Nov. 1960 Good Discontinued
Staff gauge. Continuous recording from Jan. 1961
weir without separating walls. 90° March 1963
Munro horizontal drum float eve 1 to Jan. 1963, daily observations at
V notch upstream for low flows. recorder. other times.

Hok Tau 85 6.0 Compound sharp-cres ted rectangular Daily observations before June 1961, Dec. 1960 Good
Sloping brass staff gauge.
weir, without separating walls. Munro horizontal drum float level thereafter continuous recording with
recorder before May 1964, thereafter daily first and then weekly observations.
Leupold & Stevens A-35 recorder.
Chung Mei 13 9.1 Compound crump weir with -90° V notches Continuous recording with daily first May 1962 Good
Sloping brass staff gauge.
upstream for low flows. Munro horizontal drum float level and then weekly observations.
recorders before April 1964, there-
after Leupold & Stevens 2A-35 recorder.
Siu Lek Yuen 74 2.1 Compound sharp-edged rectangular Continuous recording with weekly May 1964 Good
'Vertical brass staff gauge.
weir without separating walls. Munro horizontal drum float levei observations.
recorders before June 1964, thereafter
Leupold & Stevens A-35 recorder.
Tsak Yue Bu 41 1.6 Compound V and rectangle sharp- Jul. 1964 Good
Sloping brass staff gauge. Continuous recording with weekly
crested suppressed weir. Leupold & Stevens A-35 recorder. obscrvations.

Lo Shue Ling 3 10.8 Parshall flume, 15ft. throat. Continuous recording with weekly Jul. 1964 Poor
Staff gauge. Leupold & Stevens
:A-35 recorder. observations.

Kam Tin 3 11.7 Parshall flume, 25ft. throat. Jul. 1964 Fair
Staff gauge. Leupold & Stevens Continuous recording with weekly,
2A-35 recorder. observations.

Table 1
 509

TABLE 2

Ordinates of the 15-Minute Dimensionless Unitgraph

U
Time + (Lag+:) Discharge x -
lag+2
V

o. 20 0.05
0.30 o. 10
0.40 0.19
0.45 0.32
0.50 0.57
0.55 O. 71
0.65 1.00
O. 70 1.02
O. 80 1.01
0.95 O. 73
1.00 O .64
1 .O5 0.56
1.30 0.38
1.50 0.30
2.00 O. 18
2.20 O. 15
2.75 o .o9
3.40 0.05
3.90 O .O3
5.13 0.00
51 O
Q
Fig.1
 B
.-
L
o
N a
-e
.-
t
3
I
.-
-I
E
/
O
 O
cv
.+
-
t
3
I
.-
-I
E
- - t 2 O
A
--i--
-+
cc
;v
 S T U D Y ,OF MAXIMUM FLOODS 1.N SMALL BASïNS OF TORRENTIAL TYPE

Rafael HERAS
Dr. Civil Engineer

Angel LARA
Civil Engineer

ABSTRACT

The methodology of study is summarized for small b=

sins o f torrential character and it is applied to one of the gu-
llies of the Gran Canaria island, considering the geological and
geomorphological conditions of the basin and also the principal
physical characteristics of the same one. I n relation to all these
physical characteristics and o f a statistical complete study of in-
tensities, the hydrogram is established for different hypothesis
and the type of hydrogram is studied more unfavorable in relation
in relation to the duration-intensity-frequency curves of maximum
precipitations in 24 hours.

RESUMEN

S:e resume la metodologia de estudio para pequeñas

cuencas de carácter torrencial y se aplica a uno de los barrancos
d e l a isla de Sran Canaria, teniendo en cuenta las condiciones geo-
lógicas y geomorfológicas de la cuenca y también las principales c z
racteristicas físicas de la misma. En función de todas estas carac-
terísticas y de un estudio estadístico completo de intensidades, se
establecen los hidrogramas para distintas hipótesis y se estudia el
hidrograma tipo más desfavorable en función de las curvas duración-
intensidad-frecuencia de precipitaciones máximas en 24 horas.
 518

1. Generalities

This method has been applied to the Tirajana gully, which is

one of the most important in the south zone of the Gran Canaria island. The
high part of their channel is formed by a big number of gullies which have its
origin to an altitude of about 1,700 m., following the receiver basin a direction
sensibly north-west-southeast. The m a x i m u m longitude of the channel is 27 k m .
and the total area of the basin is 71,4 km2. Its location in the island is reflected
in the graph number 1.

2. Geology of the basin of the Tirajana gully

The region where the Tirajana gully is located is the southeast

of the Gran Canaria island.

For its location, it participates of the geological characteristic

of the half south of this island, appearing on the surface the most ancient
complex which have taken part in its formation, such as are the Ancient Basalt
of the Serie I, of basaltic alkaline-olivinical composition and formed by sub-
parallel running out with pyroclasts intercalated, the Trachysienite complex
with ignimbrites associated, of rhyolithical, panthelithical and trachyphenol-
ithical compositions, and the Phonolithical serie, composed in this zone by
running out, pius, end ignimbrites, frequently with laminar parting parallel
to the direction of the flow.

All these series are located principally in the middle zone of

the gully, existing also in form of little cropping out, principally phonolithic
and trachysienite, in the high part of the basin. O n the other hand, all the
mentioned series are practically impermeable in the process of infiltration
from the surface and, particularly, the series of Basalts I and Trachysienite
Complex, forming the majority of the substratum on which are seated the most
recent superficial formations.
 519

The series Pre-Roque Nublo and Roque Nublo, which have its
maximum power in the interior of the island appearing largely disseminated
in the Tirajana gully, specially in its middle and high zones.

The said series are composed lithologically by angular

fragments constituting xenolithical agglomerate with intercalations of
tephrithicallavas, basaltical running out and sediments. In these series, of
moderate permeability, a quick fall in the level of water is produced, being
therefore, its pondage coefficient very law.

The most modern basaltical serie that appear on the surface,

is the correspondent to the Basalts II, of basaltical olivinical composition and
constituted by aa and pahoehoe lavas, more permeable than the previous
formations. These lavas cover principally the northcast zone of the fully
disseminating also in smaller proportion in its middle zone.

Finally, this basin present a genuine characteristic which is

distinguished from the contiguous ones, since that a big part of its surface
occupied by sedimental formations, of which, the avalanches of various ages
constitute the principal cropping out of the zone of heading, while the low zone
of the basin is covered by deposits of recent alluviums, with bigger porosity
and higher permeability.

3. Physical Data

In order to know the characteristics of the basin to use them

fundamentally in the estimation of its velocity of propagation of maximum
flood, it has been calculated for the same one, the following characteristics:

, longitudinal section
. surface
. perimeter
. equivalent rectangle
, hypsometrical curve
. index of compactness
. index of slope
520

The values obtained have been the following:

compac tness equivalent slope

surface perimeter index rectangle index

17.4 k m 2 57.5 k m 1.90 L = 26. 10 O. 263

1 = 2.74

4. M a x i m u m floods

4. 1. General planning

The principal probleme presented is the absolute lack of

direct data of gauging with sufficient extension and guarantee, as m u c h in
the studied basin as in the rest of the island, therefore it is not possible to
study the flood from the direct data of m a x i m u m flows neither by comparison
with others basins, affinitive basins hydrologically. Therefore, using the
m a x i m u m available data, it has been performed the complete study of floods
by empirical and hydrometrical methods, constrasting each one of the
estimated parameters with data obtained by direct procedures in the Gran
Canaria gully.

4. 2. Empirical methods

In the formation of m a x i m u m floods intervenes multiple

causes, whose possibility of coincidence characterizes the risk. The surface
of the receiver basin is one of the causes a m o n g the principal ones, since
there exists a good correlation between the basin area and the m a x i m u m flood.

Using formulas that could tie directly the flows of floods with
the surface of the basin and others in which intervene others hydrological
parameters.

A m o n g the existing formulas it has been used those which are

in the joined chart; these formulas has been selected in relation to the hydro-
logical characteristics of the basin in the present study. In the mentioned
 521

chart it has been given, in the s a m e way, the values which are the result of
its application.

SANTI 423 (Tr = 500 años) KUICKLING 255 (Tr = 100 anos)

GREAGER 520 (Tr = 500 años) TURAZZA 820 (Tr= 500 anos)

FORTI 626 (Tr = 500 años) HERAS 780 (Tr = 500 anos)

ZAPATA 272 (Tr.=100 años) G.QUIJANO 292 (Tr = 100 anos)

The big dispersion of the results obtained of the s a m e ones

can be observed.

4. 5. Hydrometrical method

This method consists in trying to reproduce the meteorological

phenomenon and, in this case, w e will use the method of the isochronal curves,
to which it is necessary to discompose the surface of the basin in s o m e zones
(si, s2, ... sn) limited by lines (isochrones)in which the water fallen in one
of these ones delays in arriving to the point in wich w e estimate the flood,
sucesive times of value t, 2t, ..., being our case t half hour.

The velocity of the water if fixed by experimental and

empirical methods, in relation to the physical data, fundamentally of the
longitudinal section and index of slope, and other characteristics peculiar of
the basin (vegetation, geology and so on). In our case, w e have fixed as
velócity 6 km/hour in the low zone of the basin, up to an altitude of 600 m.,
above sea level, and 7 km/hour in the high part. Once fixed this one, the pointe
are obtained from which delays in arriving the water to the place studied a
s a m e time and with which, as contour line of a topographical elevation, w e
can draw the isochronical lines obtaining simultaneously the concentration
time, that in our case is of 4.3 hours.
522

If w e contrast this time with the one given by any of the

empii ical formulas existing (for example, Giandotti), w e obtain a difference,
by an excess of about 1 hour. This appreciable differenceis justified by the
quantity of sediment load which carry the floods in this type of gullies, and
produce a disminution in the m e a n velocity of propagation. The incidence of
the considered velocity in the flood peak is small, and so is only influenced
by the concentration time.

The isochrones once obtained, multiplying the area encircled

a m o n g the s a m e ones by the intensity of precipitation and the supposed runoff
coefficient, the flow is obtained in the studied point due to the precipitation in
each one of the zones.

They are, therefore, necessary the data of m a x i m u m

intensities of precipitations in the basin for a determined period of recurrence.
T o realize the statistical study of the intensity w e will use from among the
several laws of distribution of frequencies which are applied in hydrological
problems, Gumbel’s law, which is used principally for distributions of
m a x i m u m values. This law has been applied to the usable series of the interior
stations of the basin and to a series of stations of lap. All of them can be seen
in the graph number 1. The m a x i m u m annual values of precipitation in 24 hours
for several periods of recurrence are reflected in the charts numbers 1, for
the stations of the interior of the basin, and number 2, for the exterior ones.

In order to adjust the distribution of the values of m a x i m u m

precipitation in 24 hours and considering the probability of coincidence of said
values, the Gumbel’s law has been applied to the monthly data in all the
stations, for October, november, december, january, february and march,
resulting to be the months of October and november the most unfavourable in
relation to the floods, as it is deducted of the observation of the chart number 3.

Also, to contrast the distribution of m a x i m u m values in the

basin, the isomaximum curves has been designed with the values of m a x i m u m
precipitation in 24 hours to times of recurrence of 50, 100 and 500 years for
the m a x i m u m m a x i m o r u m annual values and for the m a x i m u m values of
October (which is the month of m a x i m u m intensity). These isomaximum curves
can be seen in the graphs numbers 5 up to 10 and have served like contrast of
the values obtained by Gumbel and also to adjust the m e a n intensity of
precipitation and its variation with the time.

With regard to the runoff coefficient, there is hardly no data

for m a x i m u m m a x i m o r u m flows, therefore considering the impermeability of
the middle and high zone and the greater permeability of the low zone, w e
estimate s o m e runoff coefficients of O. 85, O. 80 and O. 50, respectively, for
each one of the three considered zones. T o estimate these coefficients, which
could be reached in strong floods which would be produced after several days
 523

of considerable precipitation,it has been realize studies with all usable data
and considering the physical, geological and geomorphological characteristics
of the basin, detached in high, middle and low zones, it has been obtained
mean runoff coefficient of O. 78 that seems to be reasonably adjusted to the
characteristics of this basin.

The duration of the storm is an important factor in the

determination of the maximum flood, the maximum value of the peak flow is
used to obtain with durations of storm about the concentration time. In our
case, w e have supposed durations of storm of 1, 2, 3, 4, 5, 6 and 8 hours.
The precipitation for the several hypothesis has been estimated in relation to
the distribution of the maximum precipitation in 24 hours for smaller periods,
obtained from the short available data, which have been contrasted with direct
measures, obtaining the following values:

Duration of the storm (hours) 1 2 3 4 5 6 8

Precipitation in percentage of
the precipitation in 24 hours 35 43 57 69 75 ao 86

Although it is considered little representative the compiled

data of maximum intensities in several stations of the basin in study, the said
values are kept, putting us in security side. At the same time and in order to
procure greater aproximation to the actual phenomenon, we can consider three
stretch of different mean intensity coincident with the high, middle and low
zones, previously mentioned in the estimation of the runoff coefficients.

For the different hypothesis of duration of the flood, the

intensity, together, is distributed in the time in such a manner that in the
hydrograms of duration 1 and 2 hours it is considered all the unitary intensity
estimated and for 3 or more hours it has been supposed uniform intensity
during the two first hours and decreasing in a 20% each hour more of duration
until reaching a minimum of a 20% in the storm of 6 or more hours.

The isochrone curves used in the calculation of the hydrograms

as well as the different zones considered can be seen in the graph number 11,
and the hypothesis that the storm i s produced simultaneously in all the basin
has been made since that the hypothesis that began in the head waters and goes
displacing in the direction of the gully appears excessively unfavourable for
the climatological conditions of the basin. Once the runoff values, intensity
of precipitation and duration of the storm, are fixed, w e can obtain the flows
due to each zone and the accumumulated of these ones give the flows that would
reach the sea in each moment, supposing an infinite time of rain. Displacing
horizontally this curve in the time of duration of rain and calculating the curve
difference of the two, w e obtain the actual flows that reach in each moment.
 In relation to the study realized it has been considered the
hypothesis (i), applying in all of t h e m s o m e runoff coefficients of O. 80, O. 85
and O. 50 for each one of the different zones considered and the distribution of
intensities already cited for durations higher than two hours.

A s s u m m a r y of the hydrograms obtained, in the graphs

n u m e r s 12 up to 15 figure the correspondent to a duration of storm of 4 hours
and periods of recurrence of 100 and 500 years, the s a m e for the m a x i m u m -
m a x i m o r u m values of precipitation, as for the m a x i m u m of October. In the
chart n u m b e r 4 appears the distribution of intensities in space and time for
these hypothesis.

CONCLUSIONS

A s a result of the calculations realized by the different methods

and considering that the hydrograms obtained must be affected by a reducent
coefficient in relation to the hypothesis of calculation, in which it has been
considered s o m e m a x i m u m values of the runoff coefficient and s o m e m a x i m u m
intensities which must be reduced due to the non-coincidence of the distribution
in the space and time of m a x i m u m values in all the stations, w e obtain the
results that can be seen in the annexed chart.

T h e hydrogram type estimated figure in the graph n u m b e r 16.

T h e statistical study of m a x i m u m precipitation in 2 4 hours has been

realized in the period of 21 years, 1949-50 - 1969-70 and the data of
the usable stations has been contrasted and, generally, it appears to
have enough guarantee, but by the extension of the period used, resulted
as a risk to extrapolate for times of recurrence higher than 100 years.

T h e results obtained are conditioned by the empirical-theorical

methods used, due to the absolute lack of series of m a x i m u m flows,
although w e estimate that the m a x i m u m difference with the actual
values will not exceed 15%.

T h e study is only related to m a x i m u m values of flood and in it has not

been considered the effect of the solid flows.

(1)
of hydrograms with durations of storm of 1, 2, 3, 4, 5, 6 and 8 hours.
 525
rd
Fi
3
tJ
id
c
v
526
m
W h
r(
F:
ed
u
:
n
W
m
m
4
al
c
ree O
O W O
r
(
!x
O
H
!x
w
FI f
!z
H
W w
a,
c
ci
4-l cv
O w w
0)
c
O
'4
ci
a
4J
f
Wrn
o)
al
c
ci
c m
'4 W f
al
u
c
al
F
7u Wru
m
al
ci
lti
O
o .o o
H m o o
' 4 m
 527
W
W O
hl
UJ
wu.
rl
N
W f
[B
rl
a
O h
.rl
h w m
rl
a,
a
N
W m
1
rl
1
h
W 1
1
UJ
W O
1
m
w w
W
W h
W
W f
W N
O 0 0
m o o
r l m
52 8
-
-EE
m
m
>
-5:
529
 530

o-1

ISLA E

--T
6-3
 5 31

6-5

-
BARRANCO DE T I W A

ISOMAXIMAS-ANUAL
50 AfiOS
532

ISOMAXIMAS-ANUAL
100 AROS
 533

06ä

-
BIRRANCO DETIñàùANA

ISOMAXIMAS-ANUAL
soo AROS
534

0 2

\O"

G-8

ûARRANC0 DE TVLIWANA

ISOMAXIMAC-OCTUBR!
50 AROS
 535

0 2

OM

G-9

-
BARRANCO DE TIRUIW

-
ISOMAXIMAS OCTUBRE
100 AROS
536

G-10

BARRANCO DE TIRWANA

ISOM AXMAS-OCTUBRE
 537

oz

065

O78
538
s
' a.@..,
U
.
,
.
 539
L

+
! -
ir)
m
4
.N
d
N

r
T -
Nin
E
E
<v

a
a
O
r
m
W
ir)
O
O
FLOOD ESTIMATION BY DETERMINATION OF REGIONAL PARAMETERS FROM LIMITED DATA

P.H. EERBST, S. VAN BILJON, J.P.J. OLIVIER AND J.M. HALL

ABSTRACT

A regionalized study o f maximum annual flows of different short dura-

tions (including peaks) has been carried out. In view o f the limited
length o f record available at most of the gauging stations in the region,
an attempt has been made to develop a technique to strengthen the data
available at any particular point of interest, by using all available p e r
tinent flow data in the region. Having chosen the extrema1 dislribution
best suited to the region, the moments o f the sample (after adjustment)
are correlated with various catchment characteristics. This allows estima
tion of flood magnitude frequency curves at any site of interest within
the region, with associated confidence bands. Such frequency curves are
determined for various suitable time intervals which then allows the syn-
thesis of characteristic f l o w hydrographs, with a specific probability o f
occurrence attached to each, along Mith associated enyelopes correspon-
ding to specific confidence limits, Comparison with hydrographs derived
from rainfall input depths with specified probabplities, subtracting los?
s e s , and then using unitgraph methods, leads to the conclusion that a b e t
ter relation between probability of occurrence of a specific hydrograph,
and its magnitude, can usually be obtained by direct statistical methods,
than by more indirect deterministic techniques.

RESUMEN

Fia sido ejecutado un estudio regionalizado de gastos máximos anuales

de duraciones cortas y diferentes (.incluyendo valores máximosr. En vista
de la limitación de información disponible para la mayoria de las estacio
nec de aforo de la región, se ha intentado desarrollar un método que per-
mita reforzar dicha información para cualquier punto de interés usando t o
dos l o s registros existentes de la región. Habiendo elegido l a distribu-
ción extrema que mejor acomoda a la región, se han correlacionado los me-
mentos estadísticos de muestre0 (ajustando valores) con ias característi-
cas de diferentes hoyas. Esto permite l a estimación de curvas “magnitud-
frecuencia” de riadas para cualquier punto de inter’es dentro de la re-
gión, asociadas con bandas de confiabilidad. Tales c u r ~ a sde frecuencia
s e han determinado para convenientes intervalos de tiempo las cuales per-
miten la sfntesis de hidrógrafos de flujo caracteristicos, relacionados
con probabilidades específicas de ocurrencia, junto con envolventes que
corresponden a limites específicos de confiabilidad. Comparación con hi-
drógrafos derivados de precipitaciones de ocurrencia especifica, substra-
yendo pdrdidas y usando luego métodos de gráfico unitario, lleva a la
conclusión que una mejor relación entre probabilidad de ocurrencia de un
hidrógrafo determinado y su magnitud, puede obtenerse generalmente medi’an
t e métodos estadísticos directos e n vez de técnicas deterministicas indi-
rectas.
542

INTRDDUCTIOIY
There is no need to stress the importance of reliable flood magnitude frequency estimates in m t e r
resource development.
Whilst this problem is especially highlighted in developing regions, even so called developed
countries frequently suffer from a limitation of data on which to base reliable flood flow estimates.
An estimate of a flood with a specified recurrence interval (determined by specified design consideration)
should be accompanied by information concerning the reliability o f such an estimate, but this has not
usually been the case in the past.
This paper outlines a methodology by means of which all or most of the available flood flow information
in a region can be rationally analysed and assembled, by taking into account quantifiable parameters of
characteristics of the various catchments in the region and relating these to the moments of the frequency
distr, but ion assumed.
Methods are described by means of which a flood hydrograph with a specified recurrence interval can be

-
estimaied and a quantitative statement on the reliability of such an estimate can be made. lhe method was
drvtloped partly due to uncertainty about the validity of use of the method whereby rainfall intensity dura:
-
::on curves are applied to a transformation function (e.g. a I hour Unitgraph) due to the many assunptions
necessary in the latter approach.
It was felt that where some limited flow information does exist, an approach as outlined woulo provide better
estimates of flood frequencies and flood hydrographs, including also information on the reliability of such
estimates. Avoidance of any mention of the degree of uncertainty in any such estimate does not remove the
uncertainty, it only serves to diminish consideration of the fact that such uncertainty not only exists but
may be considerable.
FREOUtNCV DISIRIBUTIONS
In developing a methodology for flood frequency estimates (for annual extreme flows of various -hart dura=
tions) on a regional basis, it is essential to decide which frequency distribution should be assumEd to apply
throughout the rsgion. This is so, firstly for the reason that if a reasonable correlation between moments of
the oistribution assumed (whether the variables be transformed or not) and characteristics of the catchent can
be found, the same distribution must by force also be used for estimation purposes at some new site of interest
in the region.
Secondly it was considered that if a distribution is used which has a third parameter, this would provide
the necessary flexibility (adaptability) for the distribution to be ‘Iraiaxefi so as to fit that particular
region; the third parameter thus being a constant throughout the region (for every duration).
Furthermore, the possibility exists that such a third parameter may show some sensible variation if adjacent
regions are analysed in turn, thus promising the possibility of a “smoothing” thereof, providing there are no
gross geographic discontinuities. The coastal zone, consisting of rivers draining to the south eastern sea=
board of the Republic is considered suitable for such further analysis, a similar but less comprehensive study
having been carried out for those rivers mainly draining via the Orange, Limpopo and Komati river systems [i] .
The region chosen for use as a pilot study which this paper srnarizes, consisted of the north eastern part
of the zone mentioned and is shown on the locality map narked figure 1.
Data from wme of the gauging stations with a reasonable length of record in this region were used to com-
pare the log Gumbel and log Pearson Type III distributions. In the latter case the data were plotted on
specially made graph paper on which a distribution with a skew equal to that calculated from the sample concer-
ned, plots as a straight line. Camparison of the plots led to the conclusion that no particular superiority
of the one above the other was evident. The log Pearson Type III distribution was therefore chosen for the
reasons mentioned above. It should be stated however, that the techniques described in this paper could be
applied equally well to the Gunbel or log G m b e l distributions.
Moreover, the basic supposition that nature would be so kind as to ensure that the distribution of flou
extremes would follow some definite (simple) statistical distribution, should always be remembered for the
fallacy which it is. Ihis is especially true where two distinctly separate flood producing factors may per-
tain; and may operate either separately or conjunctively.
The authors feel, along with Harter 121 that there can be IM finality about the recommendations made by
the U.S. Water Resources b u n c i l 131 concerning the log Pearson Type III distribution, but for the various
 543

reasons stated, and the availability of the tables provided by Hartar, the exact form of the distribution
postulated is of lesser importance than is the proper utilization and assembly of all the available
flow data in the region, in a rational manner, so as to ohtain the best possible flood flow estimates
and concomitant reliability estimates.
The various durations of extreme flows in the region that was investigated in the pilot study sunniarized
herein were: peak flow, 1 day, 2 day, 4 day and 6 day average extreme flows. The logarithms (to base 10) of
these extremes were found to have a skew of 0,3 for peak flows, 0,4 for 1 day average extreme flows and 0,5
for 2 day, 4 day and 6 day average extreme flows. It would appear that there may be a relationship between
the skew and the duration and if this is also found to be the case in other regions this could cnnveivabìy
be used to obtain more stable estimates of the skew.
Special graph paper was developed for the skew values of 0,l (0,l) 1,O and 1,5. An example of this is
the paper on which the graph shown as figure 3 appears. The ordinate has both logarithmic and linear scales,
and the abscissa consists of both the emulative probability of exceedence (e.g. of a certain flow magnitude)
and a linear scale, the units of which are essentially in the number (and decimal fraction) of standard
deviations from the population mean p, corresponding to the probability of exceedence for the particular skew
-
value in question. This scale is identified as the K scale (K being analogous to Gumbel's reduced variate).
On the assumption then, that the logarithms of the annual extreme flows for the various durations are dis-
tributed according to a Pearson Type III distribution with the applicable regional skew values, the N year
flood can be obtained from the expresfion:
. .....................
X =X KS (i)
Here and S are the mean and ssandard deviation of the logarithms of the individual extreme annual flows.
X, is the logarithm of the N year extreiae flood magnitude, and K is the number of standard deviations from the
population m a n y that corresponds to the exceedence probability for the skew value in question (presented
in detail in Harters tables).
If the logarithms of the extreme flows are distributed according to the Log Pearson III distribution with

-
a skew ofï= 1 say, then if probability paper designed f o r ö = 1 is used, a straight line d r a w hereon for
specific values of X and S, will yield a flood magnitude frequency curve of XIversus .K As K is uniquely
related to the probability of exceedence, X, can be read and transformed to yield the flou value estimated to
be equalled or exceeded for any specified return period within the range.
ESTIMATION Of IHE M@KIIIS OF THE DISTRIBUTION
The problem therefore reduces to estimation of the values of i and S for a specific catchent. This is done
by correlation of all availab!e and pertinent measured flow data to catchment characteristics, so as to be able
to obtain best estimates for X and S. The variables investigated depend upon factors considered either as
possibly causal, or as possible contributary factors towards the occurrence of extreme flows. Although the
authors are aware of the possible application of factor analysis (or principal cuœponent analysis) here, it has
not been used during this study for various reasons [4'J .
The various independent variables considered were the following: area, mean annual rainfall, average
slope, river length, monthly rainfall with a tvo year recurrence interval (log normal distribution assumed) and
a shape factor. The data used in the present study are presented in Tables 1 and 2.
The regression nodels used were-all of the general form:
..
X a b log A+ c log R t ................. (2)
It may be noted that, as 1 is the mean of the logarithms of the extreme flous, the above formula using only
A and R is equivalent to the m d e l Q = 2 Ab RC where og,m,iS the geometric mean of the extreme flows at
a specific site. g.m.
In obtaining a further estArnate of i (1 )by simple or multiple linear regression, a value for the variance
of such a further estimate of X is always optained. This variance depends not only upon the degree o f variance
explained by the regression model, but also by the extent of the deviation of any of the independent variables

=
VAR (ie) Residual Variance[l + 7 - m)'
from its mean. In the case of equation 2 above the expression for the variance will be of the form:
1 + cZ2 (Log Ae - v)'
+ cj3 (Log Re

+ZcZj(Log A e - m > (Log R e - W ] ..................(3)

 544

The statistical theory applied here is very clearly set outlin text books on statistics [5,6] .
In short, for every regression equation used for estimation of X or S an accompanying equation is developed
e e
for the calculation of VAR (le) and VAR (Se).
The variables mentioned earlier were used in regression models to determine the regression equations
that would explaln the highest proportion of the variance of the dependent variables I and S, for the five
durations considered.
from some 150 regression models tested, the 10 equations that were selected as the best are presented in
lable 3. Values of ~22,C J ~ and C ~ Jare also presented for !se in an equation of the type represented by
equations 3 and 4, in order to calculate the variance of the X *s and the S 's. E.g. the equation for the
variance of a further estimate of X for peak flow, X by uS"e of equatiog 3 is as follows:
1 P t e 2
[n
VAR ($,e) E Residual Variance 1- + 0,1191 (Log Ae-Log A) + 11,6819 (Log R e - W I 2
+ 2~0,4545 (Log A e - W ) (Log Re- WR)]
= 0,0454 62
(this is for the brgenstond Dam site which has a catchment area of 528x10 m and a mean annual rainfall of
900 x 10-3m).
In this analysis it was hoped that the monthly extreme rainfall would be a more representative parameter
of the flood producing characteristic of rainfall than the mean annual rainfall. However, as both of these
parameters explained an approximately aqua1 amount of additional variance, it was considered advisable to
select the annual rainfall for use in the prediction equation. In view of the availability on magnetic tape,
on a large scale, of such monthly rainfall data, and the understandable hope that an extreme value rainfall
parameter would yield better results, this is a very disappointing result. It is however intended to invati.
gate this aspect further.
Hawing estimated the value of i and Se, a straight line flood magnitude-frequency curve can be drawn on
the graph paper with the apprcpriate skew, and X for any value of N within the range can then be read from
N
the graph. Ihis is done for the peak flow and for the various durations for which formulae have been developed,
thus allowing for the synthetization of a balanced hydrograph. Ihis is a hydrograph constructed symmetrically
around the peak. It can then be adjusted along the time axis (but with retention of the properties derived)
to its proper shape, either by means of information an unitgraph shapes [7] or by actual measurement of one
or two reasonably large floods at the site in question. Such measurements m u l d be arranged for at an early
stage of a feasibility study involving a specific site, if no data are available. It should be noted here that,
according to Nash [8,14 estimation of a unitgraph shape, even from only one good sized flood in a season will
yield more reliable results than that whlch can be obtained synthetically.
RELIABILITY OF ESTIMATES
In constructing the flood magnitude-frequency relationship we have:
XN=ie + KSe .............. (1)

If and Se are not independent, the problem of estimation of the covariance term arises, for which a
formula tuch as for VAR (1 and VAR (Se) has not been developed.
This problem was solves by use of an artificial population, distributed according to Pearson Type III,
=
w i t h y = O, Q 1. Skew, 8 was varied from O to 1,5 i.e. a different artificial population for each skew.
For each skeu an artificial population consisting of 10 O00 terns was prepared, by use of Harters tables.
Samples of size W ranging from 2 to 20 were then drawn. Every individual value drawn was, however transfor:
med by addition of unity so that in fact an approximation to a universe with O- = 1 andy= 1 was used.
for each sample size N, an adeguate number of samples were drawn to define-to a sufficient degree of accuracy,
the variance of i, s and cov h,s). for each sample drawn, the-values of x and s were calculated and then an
adequate number of such samples drawn so as to calculate var (x), var (s) and cov (x,s).for the smaller
sample sizes the number of sanples drawn were simply increased indefinitely until it was clear that the result
 545

:.J&converying. In this way curves were obtained showing how var (i), var (s) and cov (x,s) varies with saw=
ples size N, ranging from 2 to 20.
The results are presented in figure 2. Not all the curves developed are shown, but all the data obtained
was used in order to achieve an integrated matching set of curves.
lhe use of these curves, in order to solve the problem encnuntered with the existence of the covariance
term in equation 4 is eqlained as follows:
From the regression equations, values are obtained for Xe and VAR (ie) and also for Se and VAR (S 1.
(See Table 3).
PutX = k ; = k andS =ks-k
e _ i i e 2-2
(but x = s = 1)
Therefore
2
VAR (x,)=VA!? (k X)=kl var (S )=VAR fk2s)=k<: var (s)
1

By entering figure 2 with the calculated values of var (a)

and var (5) two values for the virtual sample sizes,
N1 and N to which level of reliability the estimates can be considered to be equivalent, are obtained.
2
These values may not be the same and it is suggested that the best value for N be calculated by use of
the equation:
2
r N i r N
2
N= 1 1 2 2 ...........,.. (5)
9
ri + 'r
2
where r and r are the multiple correlation coefficients for and S respectively, and N and N the corres=
ponding 1 equivaient lengths of record read from figrire 2. e e 1 2
This value of N can then be used to read the value of cov (ils) from tne applicable curve, and multiplied by
the factor k k for use in equation 4.
1 2
In order to calculate the confidence bands to be fitted to the line represented by the equation
XN=je Khequation 4 can be used to calculate VAR (X,) for the K-value corresponding to each XN.
By tanking the root o f VAR (X )and adding and subtracting this from the value of X throughout the range
N
involved, the upper and lower conhdence bands corresponding to one standard deviation can be obtained. This
can be interpreted that there is only a chance of 1 in 6 that the true value of the estimate sought would be
equal to or higher than the upper confidence band.
LIMIIED MEASURED DATA AVAILABLE
lhe case where some, but only limited observational data are available, is frequently encountered in practice.
Here the problem is how best to assemble the available data in the most rational manner, so as to obtain the
best possible estimates. Beard [q
has suggested formulae by means of which the reliability of data at a
station with a short period of record can be increased by correlation and adjustment with a nearby station with
a longer record, if the flows during the overlapping periods are intercorrelated.
Considered that after such adjustment (or without it if it cannot be done) the station data consists o f
xv,St and a number of years of observation, N' (whether real oc virtual in the case of adjustment by Eeard).
e h t h o u g h it may be possible to calculate values for VAR (1') and VAR (Sf) from the s a q l e itself, these
values are likely to be subjectJo such large sampliig fluctuafion that a bgtter approach would be'to read
the appropriate values of var (x), var (s) and cov (x,s) for the applicable N1 value from figure 2.
As before we have
 546

Iherefore

and COY (EL,S:) = k3k4 cov (;,SI

let 'i and S* represent the best pooled estimates of these parameters,
then we have:
Nie + N'ile
i* = ...............(6)
N +Nt

s =A
(N - l)Se + (N' - 1) Se
T ...........(7)
!if
.v N+fi'-Z
unbiased estimators are used)
Then we have:
xi .'1 . ..............................
KSI (8)
and VAR (xi) = VAR (2.1 Kz VAR (S') +

As before j* =kiandS':ks
. 2K CDV (i*,S*) ........(9)
5 6
Therefore VAR (i*) = k25 var (i)
VAR(S*) = k62 var (5)
and COV (i', )'S = k 5k6 COY (g,s)
As i = s = 1, k and k can be calculated. By putting N* = N + W 1 and entering figure 2 with this calculated
6
value for N*, "hues for var (i), var(s) and w v (x,s) can be rea$
These latter three values can then be used to calculate VAR (X,) for any specified K value thus making
possible the calculation of the confidence limits.
-
An example of a flood magnitude frequency curve (for peaks) is shown as Figure 3 (brynstond site),
along with the one Standard deviation confidence bands.
BALANCED HYDROGRAPHS AND DESIGN HYDROGRAPHS
Using the approach outlined above, an estimate of peak flow with a specified probability of exceedance and
concomitant upper and lower confidence limits corresponding to one standard deviation (or for any other confi.
dence level required) [Io3 may be calculated. lhe same can be done for each of the five durations, resulting
in a llbalanced hydrographll (i.e. symmetric about the peak) together with upper and louer envelopes wrrespon=
ding to the desired confidence level. In other words, for a confidence level corresponding to one standard de=
viation this implies that there is a 1 in 6 (or 16%) chance that the true hydrograph could be as big, or bigger
than the upper envelope.
lhe shape of this hydrograph can then be suitably adjusted along the tiw, axis, but preserving its derived
characteristics, so as to obtain an estimate of the hydrograph with the required probability of occurrence,
but incorporating the shape which is unique to the particular catchment.
CONCL US IOW
Results of work similar to that described here have in the past been used in the Department of Water Affairs
in a somewhat different manner [i] first for estimation of peak flou only and lately [il] also for e r t i m e
average flows over durations of somewhat longer periods. It is intended to extend the study to the whole of
the eastern and south eastern coastal zone of the Republic of South Africa.
The possible existence of medium term (e.g. from 3 to 7 years) non-stationarity of the various flood
magnitude-frequency curves due to such medium term variations in sea temperatures, will also be investigated
 547

,: series of recent flood disasters in the coastal zone mentioned makes such a study virtually imperative,
the controlling conditions may still not have reverted to normal). for smaller catchments a stratified
grouping approach will be followed to attempt to improve reliability.
More work is also intended to a t t m p t to determine the particular rainfall characteristic most closely
related to the flood producing attribute thereof.
lhe possibility of using an analogous approach to that described herein for estimation of other hydroloqical
parameters is not overlooked, e.g. low flow sequences with specified probabilities, mean annual runoff, etc.
This method also holds promise in rational hydrologic network plannirig [i31 or adaptation thereof.
Preliminary comparison of this method with older methods used by the Department, in some of which the
probability of the causative rainfall is put equal to the probability of the resulting runoff hydrograph,
s e m to inuicate that this method is preferable, not only from the point of view of accuracy of estimation but
also due to the frank admission and quantification of the reliability of estimation, and the extent of the
probable errors.
ACKNOWLEDGEME NT
The permission granted by the Secretary for Water Affairs to publish this paper is acknowledged.
The assistance provided by J. de Beer of the Computer Centre and by A.J. Muller, J. Botha, S. Fitchet and
L. Eskell in the preparation of the paper is greatly appreciated.
lhe guidance given by W.J.R. Alexander during the course of preparation of the paper is gratefully acknoum
ledged.
RE FERE NCES
1. Herbst, P.H. (1968). flood estimation for ungauged catchments, Technical Report No. 46, Department of
Water Affairs, Republic of South Africa.
2. Harter, H.L. (1969. A new table of percentage points of the Pearson Type III distribution, Technometrics,
II(I), pp.177-186.
3. U.S. Water Resources Council, (1967). A uniform technique for determining flood flow frequencies, Bull.
No. 15, Water Resources Council, Washington, D.C.

-
4. Haan, C.T. and Allen, D.M. (1972). Comparison of multiple regression and principal component regression
for predicting uater yields in Kentucky, Water Resour. Res., 8(6), pp. 1593 1596.
5. Ostle, B. (1963). Statistics in Research, Second Edition, Iowa State University Press, Ames, Chapter 8.
6. Ezekiel, M and Fox, K.A. (1959). Methods of correlation and regrassion#analysis, John Wiley & Sons, Inc.,
New York, pp.320-321.
7. Midgley, D.C., Pullen, R.A. and Pitman, W.V. (1969). Design flood determination in South Africa,
Report No. 4/69, Hydrological Research Unit, University of the Witwatersrand, Johannesburg.
8. Nash, J.E. and Shaw, B.L. (1%6).
-
Flood frequency as a function of catchment characteristics,
Symposium on River Flood Hydrology, Inst. of Civil Engineers, London, Session C 6, pp. 115 136.
9. Beard, L.R. (1962). Statistical methods in hydrology, U.S. Army Engineer District, Corps of Engineers,
Sacramento, California.
10. Mode, E.B. (1961). --
Elements of statistics, Prentice Hall, Inc., Nw Jersey.
11. van Blljon, S. (1972). flood volume frequency analysis Vaal Dam. Internal Report, Department of
Water Affairs, Republic of South Africa.
12. Thorne, R.B. (1966). River Engineering and Water Conservation Uorks, Buttervorths, London.
13- Herbst, P.H. and Shaw, E.M. (1969). Determining rain gauge densities in England from Limited data to

-
give a required precision for monthly areal rainfall estimates, Journal of the I.W.E.,
pp. 218 230.
23(4),
548
 549

GAUGE ûFFICIAL RAINFALL AREA SLOPE LENGIH NUMBER

SEOUENCE GAUGE AVERAGE MONTHLY Of Of MAIN OF Of
NUMBER ON NLMBER ANNUAL WITH CATCH- WATER MAIN YEARS
MAP OVER 2-YEAR MENI COURSE WATER OF
CAT CHMENT R E CUR R E NCE COURSE RECORD
INIERVAL
10+m 10-3n i06$ m/mxiû3 10% YEARS
1 XlMOl 888 204 5444 5.50 144.0 61
2 x2Mo1 1142 254 104 l3;49 26,6 19
3 XZMOZ 1074 231 176 9,47 32,2 19
4 XZM08 1163 289 181 35J1 25,7 23
5 Xx109 1070 264 280 18,40 28,2 12
6 X2M10 1187 289 127 19,42 15,8 22
7 X2Mll 849 183 401 16,37 26,6 15
8 X2M12 833 192 88 10,03 13,7 14
9 X2M13 907 208 1502 12,45 86,9 10
10 Xx114 1237 302 251 19,19 x),6 11
11 X2M15 936 213 1538 1256 76,4 12
17 XYOI 1492 368 174 33,03 17,4 23
13 XW6 985 226 761 17,37 53,6 12
14 w4Mo2 889 200 7122 5,05 232-6 18
15 W4M3 924 204 5843 7,41 162,5 21
16 m 4 929 210 448 3,91 24-9 10
17 wsMo5 920 221 751 3,65 5690 21
18 W W 947 213 176 14,95 1593 21
19 W5M7 868 191 536 2,68 46,7 16
20 WW8 862 203 119 4,79 27,4 18
21 w5Mo9 899 206 2805 12,29 9197 10
22 WW12 905 209 12769 8,02 184,3 12
23 w6MOI 1003 238 694 12,15 85,3 12
24 XlMo6 1233 279 585 18,Ol 49,l 12
25 m 3 921 205 218 4,73 25,7 13
26 WY10 953 213 2201 5,81 120,7 13
27 W13 1099 263 1155 10.93 88.5 10

Average of Log of
-
TABLE 2
Standard Standard Multiple
Prediction Equation Area Mean An. Error of Deviation: Correlation F
(Log A=Y ; log R.2) in Rainfall Estimate Dependent ‘22 ‘33 ‘23 Coefficient Value
10 10
dm2 10-3, Variable

P= 0,802Y+l,1372-3,880 2,790 2,991 0,207 0,511 0,119 11,681 0,454 0,92 67,5
SD =-O,119Y-O,9692+3,611 0,115 0,133 O,% 5,5
il= 0,798’f+0,7262-2,921 2,775 0,517 0,134 14,105 0,608 o,% 1%,8
S1=-0,092Y-0,4302+1,855 0,073 0,73 11,6
il2= 0,832Y+0,947Z-3,7% 2,775 0,530 0,134 14,105 0,608 0,97 187,8
S2=-0,08OY-O,4502+1,865 0,063 0,74 11,9
14’ 0,867V+1,1172-4,479 2,775 0,545 0,134 14,105 0,608 0,98 248,2
+O, 070Y-O,%32+1, 559 0,059 0,68 8,8
j6= 0,891V+1,2522-5,026 2,715 0,556 0,134 14,105 0,608 0,98 296,3
Sk=-0,081 V-O,3922+1,674 0,075 0.63 6.5
550

KM 20 K) O 20 40 60 80 100 KM
SCALE I I I SCALE

MAP SHOWING POSITIONS OF GAUGING STATIONS

FIGURE 1
551
E
a
P
u)
3
5 -
LI
œ
O
LL
 P R A C T I C E S O F DESIGN F L O O D F R E Q U E N C Y F O R S M A L L W A T E R S H E D S IN T H A I L L A N D *

Damrong J a r a s w a t h a n a
Director o f Hydrology Division
Royal Irrigation Department, Thailand

and

Subin Pinkayan
Associate Professor
A s i a n I n s t i t u t e o f Technology
Bangkok T h a i l a n d

ABSTRACT

B a s e d o n t h e fact t h a t a d e q u a t e h y d r o l o g i c d a t a do not e x i s t a n d
d e v e l o p m e n t o f w a t e r r e s o u r c e s p r o j e c t s c a n n o t be k e p t w a i t i n g u n t i l
d a t a a r e m a d e a v a i l a b l e . T h a i l a n d s h a r e s t h i s fact w i t h t h e o t h e r de-
v e l o p i n g c o u n t r i e s . T h e h y d r o l o g i c data c o n d i t i o n s in T h a i l a n d c a n b e
c a t e g o r i z e d a s follows. These are: (1) none o f any k i n d o f d a t a a v a i -
lable i n the catchment area; (2) some data available within neighbou-
r i n g a r e a s ; ( 3 ) some data with s h o r t p e r i o d o f r e c o r d ; and c4) c o n s l -
d e r a b l e d a t a a v a i l a b l e with low r e l i a b i l i t y and a c c u r a c y .
The purpose o f this paper is to present the general practices o f
hydrologic analyses in Thailand particularly o n design flood frequen-
cy in s m a l l watersheds. The m e t h o d w h i c h w a s t h e c o m m o n p r a c t i c e f o r
assessing design floods was based on the concept o f rational formula,
t h e unit d i s t r i b u t i o n g r a p h a n d t h e d e s i g n s t o r m o b t a i n e d by t h e c o n -
ventional procedures o f frequency analysis.

RESUME

Les d o n n é e s h y d r o l o g i q u e s sont i n s u f f i s a n t e s , m a i s l'aménagement

d e s e a u x n e peut a t t e n d r e . C'est u n e s i t u a t i o n que l a T h a ï l a n d e p a r t a
g e a v e c d'autres p a y s en voie de d é v e l o p p e m e n t . E n T h a ï l a n d e , o n peut
c l a s s e r c o m m e suit l a n a t u r e d e s d o n n é e s h y d r o l o g i q u e s : (1) i l n'y a
r i e n ; (2) o n d i s p o s e 4e quelque c h o s e d a n s d e s b a s s i n s v o i s i n s ; (3)
o n d i s p o s e de d o n n é e s s u r une c o u r t e p é r i o d e ; (4) o n a u n e g r a n d e
q u a n t i t é de d o n n é e s q u i n'inspirent pas c o n f i a n c e et s o n t p e u p r é c i -
ses.
L e b u t d e cette c o m m u n i c a t i o n est d e p r é s e n t e r l e s m é t h o d e s d'ana
lyse htdrologique habituellement utilisées e n Thailande, notamment
p o u r l'évaluation des c r u e s d e p r o j e t s u r l e s p e t i t s bassins. L e s mo-
d e s de c a l c u l l e s p l u s f r é q u e n t s sont b a s é s s u r l a m é t h o d e r a t i o n n e -
lle, l'hydrogramme u n i t a i r e et la r e c h e r c h e d e l'averse de p r o j e t p a r
l e s p r o c é d é s c l a s s i q u e s de l'analyse f r é q u e n t i e l l e .

* S u b m i t t e d for p r e s e n t a t i o n at t h e I n t e r n a t i o n a l S y m p s i u m o n t h e De-.
s i g n o f W a t e r R e s o u r c e s P r o j e c t w i t h I n a d e q u a t e Data, J u n e 4-9,
1973, Madrid, Spain.
554

INTRODUCTION

As far as the existing hydrological networks and its operation in Thailand

are concerned, it can be considered that the design of water resources pnojects
in Thailand are based on inadequate data. While the water resources develop-
ments cannot be kept waiting, it is the main function of hydrologist to modify
the conventional approaches to the hydrological assessment in planning of
water resources projects. Application and modification of conventional
approaches may have certain degrees of complication depending upon the avail-
ability and limitation of the information. Inadequacy of hydrological data
within the country may be catagorized as follows: (i) none of any kind of
data available in the catchment area; (2) some data available within neigh-
bouring areas; (3) some data with short and/or broken period of record; and
(4) considerable data available with low reliability and accuracy. Hydrologist
has to make a great attempt and utilized his own experiences in the country in
drawing up the basic fact, assumption and related knowledge to justify the
hydrological assessment of each case under study. With such attempts the
justification of a project could be made with only a fair degree of accuracy.
Most of the small storage reservoirs in the country were planned by applying
the modified Conventional approaches to the hydrological assessment, in which
the only available information is the rainfall data in the neighbouring areas.

HYDROLOGICAL DATA PROCUREMENT

It can be stated that hydrological observation has been initiated in

Thailand since 1831, when a staff gage had been installed at Ayuthya to observe
annual flood inundation within the Central Plain areas. The nature of flood
inundation had been a major factor reflecting the crop yield during the last
century. Such maximum water levels were indications of water condition in term
of good wateryear, too high flood or drought conditions which, in turn, would
help predicting the annual rice harvest.

Later in 1905 the streamflow measurement by surface float was introduced

to measure the discharge of the Chao Phraya river for the purpose of planning
of water conservation, diversion and irrigation control works. Development
at that time comprised of diversion schemes in the river valleys and tidal
irrigation in the delta area. Connected channels between rivers in the estuary
were also excavated to conserve water for irrigation and navigation from and
to Bangkok. Scientific approaches applied to planning and implementation of
irrigation and drainage schemes were carried out by the Royal Irrigation
Department since 1915. Development of water resources was gradually extending
towards headwater and slowly progressed.

Until 1952, when the storage work began, the modern methods and scientific
standards to be introduced in the hydrological investigation were recognized.
A network of comprehensive streamflow gaging stations was set up in major
tributaries where damsites for possible large reservoirs were found. Meanwhile,
numbers of stations were added consecutively in the accessible remote areas to
examine the runoff and flood yield from watersheds. At present there are over
230 rating stations operated by various government agencies. Among them there
are small number of stations that the drainage area is less than 100 square
kilometers. Many common problems of hydrological investigation and data pro-
curement still exist in the country hence they limit the expansion of the net-
 5 55

work particularly into the smaller watersheds. Among those common problems,
the limitation of financial support and lack of well-trained personnel are con-
sidered to be the main factors. Lack of popularity of work is another important
factor leading to have less fund allocated for hydrological investigation.

Besides the large multiple-purposes storage reservoir projects, the

surface reservoir of comparatively small storage volume called "tank" irrigation
projacis were commenced in 1951 in the Northeastern Region of Thailand.
Several small watercourses in the undulated topography were formed and appeared
to be good sites for storage tanks. The reservoirs range in capacity from
around 40,000 cubic meters up to 18 million cubic meters. Of course, the hydro-
logical investigation of such small basins has never been practiced in the
region as well as in the other parts of the country.

PRACTICES OF DESIGN FLOOD FOR SMALL WATERSHED

To present the general practices of design flood for small ungaged water-
shed, a case design by Royal Irrigation Department (1963) of the Sattaheep
Tank Project, Thailand, is described below.

It was the requirement of the Sattaheep Naval Station in 1963, to construct

a small reservoir of 2 million cubic meters capacity for domestic supply. The
proposed damsite has a drainage area of 10.9 square kilometers. None of any
kind of data is available except the rainfall data at Sattaheep, located about
8 kilometers west of the basin. Hence, the rainfall data at this station were
used in assessment of the design flood. Such adoption was based on the generally
practices that it was applicable where rainfall characteristics were similar.

Trials had been made by applying the empirical formula to determine the
=
maximum discharge. The rational formula, Q C i A, was found to be less
applicable as its coefficient. C, could not be determined correctly. The McMath
formula, Q ACi(S/A)1/5, was then introduced because it seems to be
more applicable as the formula involves the basin slope which is one of the
major factors governing the peak rate.

The frequency of design flood cannot directly be determined. In this case

it was assumed to be similar to that of one-day rainfall. From 24-year period
of daily rainfall record at Sattahepp, the maximum one-day rainfall amount
of 302.7 mm was observed on 6 October 1957. The computed frequency of occurrence
of this one-day storm rainfall is once in 40 years. From the conventional
frequency analysis, the 50-year frequency one-day rainfall amount of 320 mm
was adopted in the assessment of the inflow design flood. Such frequency was
assumed to be that of the design flood.

Careful inspection of the catchment had been carried out in order to

examine the basin characteristics and to estimate the concentration time. It
is apparent that the time of concentration of such small basin is very short
and usually is much less than one-day. The percentage of rainfall as a fraction
of one day was obtained from the graph of rainfall recorder. In this case
several storm events were examined and the envelope curve was used. The
rainfall amount falling within the time of concentration was calculated and
converted into rainfall intensity which is to be used in the McMath formula.
The basin coefficient, C, was estimated based on basin characteristics as
 556

inspected. The basin slope, S, was determined from the available topographic
map of the basin. The design peak discharge obtained in this case study was
43 cubic meters per second.

Other means of assessments were also made for comparison. The unit h y b o -
graph procedure was applied. The assumption of the base time of unit hydrograph
is important as it will result in varying peak rate. The storm runoff coeffi-
cient was carefully assumed and flood volume was computed. Peak flow rate was,
therefore, obtained by applying triangular distribution hydrograph to the flood
volume. The second comparison was made with the specific yields of flood
flows obtained from the actual streamflow measurements observed in larger water-
sheds by the Royal Irrigation Department (1965). The flood yield per unit area
computed from those stations were plotted against their respective drainage
areas. The possible maximum flood yield from smaller watersheds, in term of
cubic meter per second per square kilometeramay be read from the logarithmic
extrapolation of the envelope curve of specific yield. Such technique will be
one of the most reliable indirect approaches if the flood yields of small
streams are available with longer period of record. After several trials were
made, the design flood of 43 cubic meters per second were adopted in this study.
The assigned frequency was 50-year. The specific yield of flood flow was around
4 cubic meters per second per square kilometers, which is believed to be
adoptable in the area easily affected by tropical depression storms.

CONCLUSIONS

Several modifications of conventional approaches were used in planning and

design of water resources projects in small watersheds in Thailand. The results
obtained by such methods would be satisfied up to a certain degree. New con-
cepts and statistical techniques which give more reliability are needed to
design of small water resources projects in Thailand.

REFERENCES

1. Royal Irrigation Department (1963). Sattaheep Tank Project, Assessment of

Water for Storage, Hydrology No.137/63, Royal Irrigation Department, Bangkok,
Thailand.

2. Royal Irrigation Department (1965). Mean Annual Discharge vs. Drainage Area,
Envelope Curves of Maximum Recorded Peak Discharge, Specific Yield of Flood
Flow for Rivers in Thailand and Malaya, Hydrology No.186/65, Royal Irrigation
Department, Bangkok, Thailand.
 DESIGN DISCHARGE DERIVED FROM DESIGN RAINFALL

Takeo KINOSITA
Takeshi HASHIMOTO

ABSTRACT

A d e s i g n d i s c h a r g e f o r f l o o d c o n t r o l in J a p a n i s i n g e n e r a l
d e r i v e d f r o m a d e s i g n r a i n f a l l s i n c e d i s c h a r g e d a t a are not s u f f L
cient f o r designing. The p r o c e d u r e o f d e r i v a t i o n and i t s m e r i t s
a n d d e m e r i t s w i l l be e x p l a i n e d i n t h i s r e p o r t a c c o r d i n g t o fallo-
wing f o u r s t e p s . ( i ) A d e s i g n r a i n f a l l i n a c e r t a i n r e t u r n p e r i o d
is d e t e r m i n e d by a p r o b a b i l i t y process. ( 2 ) D e s i g n r a i n f a l l d i s -
t r i b u t i o n are o b t a i n e d by e n l a r g e m e n t o f r a i n f a l l d i s t r i b u t i o n s
o f recent r e p r e s e n t a t i v e s t o r m s . ( 3 ) A s i m u l a t i o n m o d e l f o r r u n o f f
i s d e c i d e d by r a i n f a l l s and r u n o f f s o f r e c e n t r e p r e s e n t a t i v e
storms. (4) A d e s i g n d i s c h a r g e is d e t e r m i n e d by t h e s i m u l a t i o n m o
del w i t h e n l a r g e d r a i n f a l l d i s t r i b u t i o n s .

RESUME

A u J a p o n , l e s d o n n é e s c o n c e r n a n t l e s d é b i t s ne s o n t p a s SUL
f i s a n t e s p o u r é v a l u e r l e s c r u e s d e projet; o n p r o c è d e d o n c g é n é r a
l e m e n t p a r l'intermédiaire d e l'averse de projet. Les a u t e u r s e x -
posent le p r o c é d é u t i l i s é , s e s m é r i t e s e t s e s i n c o n v é n i e n t s ; i l
s e d é c o m p o s e e n quatre é t a p e s . ( 1 ) O n détermin: p a r a n a l y s e fré-
q u e n t i e l l e une averse de p r o j e t c o r r e s p o n d a n t a u n e c e r t a i n e p ê -
r i o d e d e retour. (-2) C e t t e a v e r s e est d i s t r i b u é e d a n s le t e m p s e n
s'appuyant s u r d e s h y é t o g r a m m e s d'averses r é c e n t e s c o n s i d é r é e s
c o m m e r e p r é s e n t a t i v e s . ( 3 ) O n c h o i s i t u n m o d è l e de t r a n s f o T m a t i o n
p l u i e s - d é b i t s é l a b o r é à p a r t i r d'observations de p l u i e s et d e dé-
b i t s e f f e c t u é e s r é c e m m e n t a u c o u r s d'averses r e p r é s e n t a t i v e s . (4)
O n a p p l i q u e c e m o d e l e a u h y é t o g r a m m e de p r o j e t é l a b o r é e n (2).
 558

I. Introduction
Japan is located in the temperate and humid zone. A river in this coun-
try is comparatively small and its gradient is steep. Floods have occurred
very often since the prehistric age and been serious constraints against devel-
opment of the nation for a long time. Flood control is one of the major items
of water resource development works.
It is necessary to collect and analyze discharge data for design of flood
control projects. Authorized discharge gauging stations are 330 in 120 rivers
in Japan. There are many other non-authorized discharge gauging stations.
However, land developments and river improvement works have remarkably succeeded
and hydrological eituations of river badins are rapidly changing. This fact
induces that the discharge cannot be used for design purpose directly and used
only for verification of a runoff simulation model, and the rainfall which is
not affected by human activity is used for design purpose.
Mot only the peak discharge but also the flood hydrograph are necessary
for channel improvement, design of multipurpose reservoirs and soon. The pro-
cedure to obtain the design hydrograph will be discussed in this report.
2. Probability Analysis

Since discharge is originally derived from rainfall, the design discharge

for water resource system is determined by the design rainfall through the run-
off simulation model.
At the first step of this procedure, the total amount of the design rain-
fall within a certain period should be computed by means of the probability
analysis. The important assumption of this section is that the time series of
rainfall are produced by some stationary stochastic process.
The procedure of this analysis is divided into two.
(9Sampling from observed rainfall data.
cn) Frequency analysis.
The latter has been discussed by some hydrologists, 80 the authors intend
to focus their attention on the practical phase of the former. The series of
annual extreme values of rainfall within a certain dulation is selected from
the historical data. The duration in this paper is a period which is signifi-
cant to the design for the water resource system in the definite basin, and
cannot be so freely chosen. The rainfall within an adequate duration has the
closest relation to the magnitude of the flood discharge, and the rainfall with-
in a comparatively short or long duration has less relation to it. Therefore
the design duration must be appropriately Chosen according to the basin charac-
teristics, for instance the drainage area, the channel length, the slope and so
on.
The net work of daily rainfall observation covers all over Japan, and
daily rainfall data have been recorded for more than thirty years, at some sta-
tions a hundred years. On the other hand, the network of hourly rainfall ob-
servation is sparser than that of the daily rainfall. The hourly rainfall
data have been recorded for twenty years on an average. The credibility of
statistical estimations is dependent on the sample size, that is to say the
length of the series of observed data. Therefore the statistical analysis is
hardly applied to hourly rainfall data. Daily rainfall data are used for de-
termining the amount of design rainfall by means of the statistical analysis.
 5 59

Then, the design duration must be an integer multiple of a day. As noted

above, the design duration muet be selected ae a time in which the rainfall
has a close relation to the peak discharge. Since the time scale corresponds
to the space scale in natural phenomena, the duration for a smaller basin must
be a day, and that for a bigger basin must be three days in this country.
A daily rainfall in Japan is defined as a rainfall observed from nine
a.m. to nine a.m. the next morning. If a storm stretches over this boundary
of observation, a daily rainfall cannot represent a storm rainfall. 'Two
days'' seems a minimum design duration even for a small basin. The fact that
a big storm in Japan tends to continue more than a day requires this limitation
of the minimum design duration. The design duration for the statistical anal-
ysis is two days for a small basin and a medium basin, and three days for a big
basin.
The return period for design purpose is not determined by a mathematical
way, but by consideration of economical, political and social situations on the
basin. The sewerage system design claims for five to seven years as a return
period. For an urban basin, some period above twenty years is selected as a
return period. A big river basin in Japan requires almost a hundred years'
return period. For spillway design of a dam, about two hundred years' return
period is commonly used.

3. Enlargement of Observed Hyetographs

The amount of the probability rainfall for design purpose is determined
as shown in the above section. A careful attention should be paid to the
procedure for distributing the amount of the probability rainfall to the time
axis, because an hourly distribution of flood runoff, a hydrograph, is indis-
pensable for a flood control project, and a hydrograph is derived from an
houry distribution of rainfall, a hyetograph, through a runoff simulation
model. A hyetograph for design is deduced from that of a recent representa-
tive storm. A part of the hyetograph observed during the representative
storm is selected aiming at the time of occurence of the maximum amount of
rainfall, where the time is taken equally to the design duration.
Suppose N is the number of hours within the duration, Rpis the amount
of probability rainfall and Roi ( for i=1,2,... ,N is the observed rainfall
depth in the i-th hour. The enlargement factor R? is defined by the follow-
ing equation.

Then the enlargement factor is multiplied to each R to get the time series
of design hyetograph.

EX'-%, EF'sR, a. ,EF.Rou

This procedure is called enlargement of observed hyetograph. Several hyeto-
graphs are derived from several representative observed hyetographs in this
way.
If the amount of the representative rainfall is almost same as that of
the probability rainfall, This procedure is very successful. If not, the
enlarged hyetograph sometimes shows an unexpected pattern. In order to
avoid such an unexpected pattern, there must be some limitation for enlarge-
 560

ment. Several proposals were given for this limitation, but there's no
praiseful one. For this limitation is to be deduced not theoretically but
merely empirically. One of the proposals is presented in the following
paragraph.
A certain domain is set up including the basin concerned, From all
the rainfall gauging stations in the domain, maximum point rainfall values
are selected about various periods shorter than the duration. The enlarged
hyetograph is compared with these values. If the enlarged amount during
some period exceeds the m i m u m point rainfall value during the same period,
the enlarged hyetograph must be abandoned because of the rareness of occur-
rence. But this proposal raises another question. What region is appro-
priate as the domain? For instance, if we replace Japan with the world, the
selected value of maximum poin rainfall becomes greater at any period. In
spite of this question, this proposal seems reasonable. Because there must
exist a realistic upper bound on the amount of rainfall that can occur on the
basin within a certain period. An example is adduced. On the upper Kiso
River basin from 1951 to 1971, there were 29 representative s t o r m in which the
maximum rainfall amount d u r i g 4 8 hours was greater than 1 0 0 mm. The maximum
point rainfall values are made into Table 2. As a result of the comparison,
the exceedance is noted by symbol 'E* in Table I. If the exceedance has occur-
red at some domein, it also occurs at any narrower domain. And yet, in this
case, the exceedance is apt to occur in a shorter period than a longer period.
This fact suggests us that the hyetograph of a heavy storm is uniformer in its
time distribution than that of a common storm.

4. Runoff Simulation Model and Effective Rainfall Analyeis

In this section, a runoff simulation model is determined, and simultaneous-
ly an empirical rule is derived-on the separation of rainfall excess from obser-
ved rainfall.
Among the great number of rainfall-runoff convertion schemes, 'Storage Func-
tion Method' is commonly used in Japan. This method is expressed by the follow-
ing two equations.

where sris the storage in the basin, qL is the outflow from the basin, r excem-
sive rainfall, K and p are empirical constants dependent on the basin, and suf-
fix denotes delayed variable by a lag time Ta. These constants are necesea-
ry for the runoff simulation of the storage function, so they are previously
determined by sets of rainfall and runoff data of the recent representative
floods. As is seen in Fq. (2) this method contains a nonlinear process.
'Tank Model Method, aïso aseumes a nonlinear process, and is sometimes
used as a runoff simulation model. Unit hydrograph method has been improved
in this country, and is put to practical use today. But the use of unit hydro-
graph method ia restricted to the basine where the assumption of linearity is to
a certain degree appreciable.
 561

The separation of excessive rainfall from the observed one is an important

but an awfully suffering work. In Japan, the soil moisture of the basin general-
ly shows a violent variation from dry to wet according to the weather condition.
The soil moisture antecedent to the storm strongly governs the rising part of the
runoff hydrograph, sometimes even the crest. So the effective rainfall analysis
must be carried out carefully for the identification of model parameters. How-
ever, at the simulation for a design flood, a common value of the parameter repre-
senting the soil moisture in the basin is used.

5. Design Discharge
Finally, the design discharge is determined in this section. The enlarged
hyetographs of representative storms are used for the runoff simulation. A hy-
drograph corresponding to each design hyetograph is computed by the runoff simula-
tion model.
Table 1: Comparison of Enlarged Hyetographs with Maximum Point Rainfall Values

tom Upper Whole Chubu The

Ra E3 Japan
NO. Liso River Liso River District World

1
-1 -
mm

137.1 2.19
1
E
3 6
E
1 3 6 1 3 6 1 3 6

2 137.6 ?.i8 E E E E E E E E E
3 125.1 2.40
4 117.4 2.56
5 1 IO. 1 2.72
6 114.9 2.61
7 147.7 2.03
8 125.2 2.40
9 106.3 2.82
10 195.9 1.53
11 117.2 2.56 E E E E
12 107.6 2.79 E E
13 143.6 2.09
14 118.7 2-53
15 173.6 1.73
16 227.5 1.32
17 119.1 2.52
18 148.0 2.03
19 114.3 2.62 E
20 112.3 2.67
21 144.8 2.07
22 123-2 2.44
23 169.5 1.77 E E E
24 156.3 1.92 E E E E E E
25 118.1 2.54
26 136.6 2.20
27 148.1 2.03
28 263.9 1.14
- --
29
Symbols
225.6 1.33

EF : Enlagement Factor (Rf/Ro

Rp : Amount of Two Days' Probability Rainfall at 200 Years, Return Period
RD : Maximum 48 Hours' Rainfall Amount
E : ERLarged Amount exceeds Maximum Point Rainfall Value in this Domain
1*3 and 6 are periods in houds). The estimated value of R p is 300 mm.
 562

Table 2: Maximum Point ñainfall Values (mm )

I The World 350 600 1488900

Although areas of these hydrographs are almost same, peak discharges are
different each other. The reasom are (i) variety of hyetograph, (n)non-
linearity of runoff model, and (ai) loss factor Among these, (i) is the
most predominant. Owing to this fact and to M e it possible to obtain various
hydrographe, enlargement was applied to various representative storm hyetographs.
From the above hydrographs, one is selected as a design discharge. The
election of a design discharge itself brings up a brand new problem, but the
authors shall leave the discussion to another occasion.

6. Conclusion
The derivation of a design discharge is explained in thie re art. It is
composed of (1) design rainfall in a certain return period, (27 enlargement,
(3) a simulation model and (4) design discharge derivation. The procedure is
not fixed today. It will be improved everyday with development of hydrology.

Takeo KINOSITA, Takeshi HASHIMOTO : Public Works Research Institute,

Ministry of Construction,
Government of Japan.
 THE USE O F C E N S O R E D DATA I N E S T I M A T I N G T-YEAR F L O O D S

Morven N. Leese

Institute o f Hydrology, W a l l i n g f o r d , Berks, U.K.

AB ST RACT

T y p e s o f incomplete data t o be f o u n d i n c o n n e c t i o n w i t h
f l o o d s e r i e s are described, a n d it is s h o w n how s a m p l e s c o n t a i -
ning s u c h data may be used t o e s t i m a t e t h e p a r a m e t e r s o f a d i s -
t r i b u t i o n describing annual maximum flows. F o r m u l a e f o r t h e
s t a n d a r d e r r o r s o f t h e r e s u l t i n g e s t i m a t e s are a l s o given.
E x a m p l e s a r e t a k e n from a r i v e r for w h i c h c e n s o r e d d a t a exist.
P r e p a r a t o r y data s t a n d a r d i z a t i o n i s d e s c r i b e d , a n d t h e p a r a m e -
t e r s estimated using this data are c o m p a r e d u i t h t h e s e e s t i m a -
ted using t h e complete s a m p l e only. The m a r g i n a l value o f u s i n g
c e n s o r e d data in this context is a s s e s s e d by m e a n s of t h e subse-
quent r e d u c t i o n in the s t a n d a r d errors o f the e s t i m a t e s o f
T-year f l o o d s f o r various values o f T, and t h i s is r e l a t e d t o
the effort r e q u i r e d t o collect and s t a n d a r d i z e t h e data.

RESUME

L'auteur traite d'une c a t é g o r i e de d o n n é e s i n c o m p l è t e s

qu'on peut r e n c o n t r e r dans l'étude d'une s é r i e c h r o n c l q g i q u e COZ
cernant l e s crues. I1 montre comment des é c h a n t i l l o n s c o n t e n a n t
de t e l l e s données peuvent ê t r e u t i l i s é s p o u r e s t i m e r l e s param;-
t r e s d'une l o i de d i s t r i b u t i o n des m a x i m u m s annuels. I 1 d o n n e
également des formules p o u r c a l c u l e r les e r r e u r s t y p e s des e s t i -
m a t i o n s q u i e n résultent. I 1 p r e n d c o m m e e x e m p l e u n f l e u v e p o u r
l e q u e l existent de telles d o n n é e s t r o n q u é e s c'est-à-dire défi-
nies p a r u n s u i l a u q u e l e l l e s sont é g a l e s o u supérieures. I 1 in-
dique c o m m e n t o n peut p a r v e n i r à une n o r m a l i s a t i o n de c e s b o n n é e s
et c o m p a r e l e s v a l e u r s a i n s i e s t i m é e s p o u r l e s p a r a m e t r e s a ce-
l l e s qu'on puet o b t e n i r à p a r t i r du s e u l é c h a n t i l l o n des d o n n e e s
régulières. L'appréciation d u g a i n m a r g i n a l d'information dÛ a l a
prise e n c o m p t e des d o n n é e s t r o n q u é e s r e v i e n t à é v a l u e r la r é d u c -
t i o n de l'écart t y p e qui e n r é s u l t e p o u r l e s e s t i m a t i o n s d e s c r u e s
d e p é r i o d e d e r e t o u r T, p o u r d i f f é r e n t e s v a l e u r s de T. Ce g a i n
d'information est c o m p a r é 'a l'effort n é c e s s a i r e p o u r c o l l e c t e r e t
n o r m a l i s e r de t e l l e s d o n n é e s ,
5 64

1;JTFODUCTION

The design of hydraulic structures for a water resources project depends

in part on estimates of the floods which the structures may be required
to withstand during the project's economic life. The feasibility of the
project may be examined by comparing the cost of building each structure
to the requirements of its design flood with its anticipated benefits.
The latter may be realized in terms of reduced damage to the structure
itself as well as to surrounding property, and in more effective flood
plain use.

Precision in the estimation of a design flood conveys a monetary benefit

by mitigating the costs which arise from over or underdesign.
Nevertheless, the use of additional data to increase precision will have
a marginal cost which may be greater than its marginal benefit. In these
circumstances it is necessary to quantify the increase in precision, and
if possible to express this increase in financial terms. The design of
hydraulic structures thus involves both hydrologic and economic
considerations which require for their formulation: estimates of floods
with given return periods; values to be placed on the precision of the
estimates; cost and benefit curves for the structure.

The standard form of data for the estimation of floods consists of a series
of annual maxima derived from a continuous flow record. It is proposed to
show how data which is not of the standard form may still be used for this
purpose, and that the use of additional data of non-standard form'hcreases
the precision of estimation. It is not proposed to discuss in detail
the economic implications of the increase, but the order of magnitude of
the resulting cost-reduction is indicated by means of a simple example.
 565

ESTIMATION OF T-YEAR FLOOIX - STANDARD DATA

Jn order to estimate the flood with return period T, sey, (or 'T-year flood'),
it is first necessary to choose a probability distribution representative
of the annual maxima. The parameters of the distribution may then be
estimated from past records by one of many estimation procedures. The
Gumbel , or double-exponential, distribution is often used to represent
annual maxima because of a supposed validity on theoretical grounds and
although these grounds have been questioned (2), its extensive use justifies
its further study in this context. It has the following form:-
- (=)
F(r) = exp [-e -J , - o<x<w (1)

where F(x) is the probability of obtaining a value not exceeding x,

and a and u are parameters to be estimated.

The most common methods of estimating a and u are discussed by Lowery and
Nash (3). Of these, the method of maximum likelihood is considered here
because it allows the incorporation of non-standard data (as defined below)
into a normal sample. In addition to its convenience for present purposes,
the estimates obtained from the method have several desirable large-sample
properties, including unbiassedness. Before indicating how the method
may be modified to deal with non-standard data, the principal results for
a standard s-le are given.

!he likelihood function L and maximum likelihood equations for a sample of

N annual d m a x,, ~ 2 .,.. % (assumed independent) are as follows:-

i = 1
 566

i = N
-1/& E N + L P o

i = 1

where y
i’ the so-called ‘reduced variate’ a is given by:-

The flood of return period T, %, is then given by:-

Pr (x 5 %)= 1 - 1/T,

so that -
- e -% - u
e a
= 1 - 1/T.

m y T = %-u.
-9
an estimate oí’ 3 is then
a
A
3 =
A
Q + a YTa
A A

where <4; and u are estimates of and u and are obtained as the solutions
of (3)a and yT = -Ln Ï
i- Ln (1 - l/Tg
A

The large-sample standard error of

A A
3 is then obtained from the variance-
covariance matrix of u and u,
 567

the values of whoee elements are substituted into the following:-

(6)
h
V mw be approximated by replacing a by a. A pull derivation of the
above quantities may- be f o m d in Gumbel ( 1 ) and Kimball (4).

NON-STANDARD DATA

Those concerned with maximum flood estimation will be familiar with at

least two types of non-standard data found in connection with flood series:
missing peaks in continuous chart records and historic flood marks.

Hissing peaks occur when the flow is so high that the recording pen runs
off the eàge of the chart; whilst it should be possible to estimate a
missing peak discharge from h knowledge of the length of time the chart
limit is exceeded, the properties of such a method require further
investigation. The approach used here is to assume no more than that a
flood which has exceeded a chart limit has a peak discharge greater than
the flow corresponding to the flov at the chart limit.

Historic flood marks are usually to be found in waìls,bridges or on

specially coktructed flood stones. They indicate the levels of floods
which have risen above a fixed point during some historic period. In
certain circumstances, it mey be assumed that all such floods have been
marked, and that floods in the intervening years for which no marks exist
have failed to reach the fixed point.

These are the two types of data to be considered. They have this in common;
values are only specified if they lie on one side of a given threshold.
Samples which exhibit this property are known as censored samples, the
threshold being called the censoring point. If the threshold is fixed,
568

as it is in these caces, and the proportion of censored events is a random

variable, the censoring is type I; if the threshold is a random variable,
but the proportion of censored events is fixed, the censoring is type II.
A fui1 discussion of censoring is given by Kendall and Stuart(5), where
further references are given.

The incorporation of a . 'missing peak' or 'historic record' into a

standard sample is clearly a type I censoring problem. The general form of
the likelihood function Lc for a sample of n + k values of which n are below
the censoring point x
C
and are specified 8s x 1, x2 ... xn' and k ari? above
x and arë unkna~n,b BS follows:-
C

where f (x) is the appropriate probability distribution function. A similar

expressiori rr,ay be obtained for censoring above a censoring point, and maximum
likelihood equations m4y be obtained from either expression in the USUEL manner.

ESTIMATION OF T-YEAR FLOODS - NON-STANDARD DATA

Censoring above a threshold (Missing Peaks).

Suppose a sample consists of n annual maxima xl, 5,... xn (assumed

independent) and k missing peaks which are k n m to be above the chart
limit xc. The likelihood h c t i o n L and maximum likelihood equations are
C
then : -
 569

where yi = Xi -Y i Y, = X - u ; W =
-e J C
e .
o a

The variance-covariance matrix Y of ac

UC a the p u m e t e r s estimated
from equations ( 9 ) is then given by the inverse of Rc, whose elements ri
JC
are as follows:-

where c = eTC, and J and K are integrals which require to be evaluated

J 3 5
numerically. They are given by:

C
<*
W
; K = Je--*

C
Y
(11)

Censoring below a threshold (Historic Flood Marks).

Suppose that during a historic period of length M .yearsa r floods rose

above a threshold 5 and were marked, leaving I = M-r floods mmarked,and
assumed to be less than %. Suppose that in addition there are N years of
continuous record from recent times providing N completely specified values
(this is the usual situation). The likelihood fmction Lk end maximum
likelihood equations are then:-
 5 70

i=N+r i=N+r
- l/a L(N+r) - E y.1 +
i= 1
E YieYi
i= 1
+ leT%d = O;
(13)
i=N+r
- i/a r-(N+r) + C eyi
- i= 1
+

where yi = -
x.-u
1
a
; yh --%-u
a
; u=e

The variance-covariance matrix $for a and

A
B,the parameter estimates
estimated from equations (13) is then given by the inverse 'of #,
whose eisments r5 are 88 follows:-

rI2h = - E[a2LogI
aaau
- l/a2 {(N+M (0.4228) + M u(yh
i - -q1, 1

where J and K, are expressicm of the form of ( 1 1) with h 5 e* as the

lower limit of integration.

Equations (9) and (13) are thus the modified equations to be used for the
estimation of parameters f i m the na-standard floods data described. They
have been derived in the context of reliability theory, and are given in
(6). Similar expressions may be obtained for distributions other than the
Gumbel distributirm by substituting the appropriate p.d.f in (7) or its
equivalent fon censoring belar a threshold.
 571

An iterative technique for solving the equations (3) for a standard sample
may be found in Jenkinson (71, who gives a worked example. This technique
m y be used with slight adaptation for the solution of equations ( 9 ) and
(13); satisfactory results are obtained if the iteration matrix is left
unchanged, i.e given values appropriate to an uncensored sample. However,
care should be taken in the choice of initial values if the proportion
of censored values is high.

THE AVON AT BATH - AN APPLICATION OF THE EQUATIONS

The most satisfactory applications of these modified equations has been in

the extension of records to achieve significantly greater precision in the
resulting estimates. One such application was for data from the river
Avon at Bath, where a set of historic flood marks had been recorded during
a historic period prior to a fairly long continuous chart record. The
parametex estimated from the recent record alone were compared with these
estimated frcm a combined sample consisting of the historic floods and the
recent records. The data used is shown in t'able 1, 1940 being the date of
the beginning of the recent record.

It is necessary to perform a number of checks on historic data before

entering it into equations (13). for instance, the stage-discharge
relaticnship derived for the recent record may require adjustment before it
is applied to the historic flood marks, whose site may be at some distance
from the modern gauging station. The assumption that floods are marked if
(and mly if) they have risen above the threshold makes it necessary to
investigate the circumstances surrounding each flood mark. This mey be 8

time-consuming exercise, but it is one which could to a large extent be

carried out by local library or museum staff who have to hand contemporary
evidence such as old newspaper reports.
 572

TABLE 1. Annual Maximum Flooäa Used in the Estimation of

a and u in Gumbel's Extreme Value Distribution
(In Cumecs).

*
Water Water Water Flood
Flood Flood
Year Year Year

1865 206 1941 84 1955 128

i866 228 1942 149 1956 107
1874 12 1 1943 ' 73 1957 138
1875 218 1944 118 1958 169
1879 264 1945 128 1959 169
1882 362 1946 282 1960 352
1888 204 1947 98 196 i 12 1
375 1948 i04 1962 103
154 1949 1 q3 1963 277
239 1950 229 1964 110
302 195 1 136 1965 178
i86 1952 116 1966 172
255 1953 96 1967 31 1
148 1954 296 1968 125

!he following values were obtained for the data of table 1:-

N 0 32; M m 58; 1 = 48; r = 10; 5 -

and when these values, and the data of table 1, were substituted into
200,

equations (13), the estimates s h a m in table 2 were obtained.

A
Floods with various return periods were then estimated from the values of a,
and ;h obtained from the enlarged sample, by substitutitm in equation (41, and
their large-samgle stenâard errore were elso calculated from equation (6).
These are shown in teble 3, where values obtained from the original srmgle a m
ale0 shown.
 573

TAñLE 2. Estimates of P a r e t e m of Gumbel's Extreme Value

Distribution Using (1) Historic Flood Marks and Recent Data
and (2) Recent Data Alone. (In Cumecs).

Parameter a U Sample size

(1) Estimate: 47 128 29 recent values

Large-sample + 13 historic values
standard error: 2.5 27 + 48 censored values

(2) Estimate: 48 128 29 recent values

Large-sample
standard error:
I
27 29

TABLE 3. Estimates of Floods with Various Return Periods Using (1)

Historic Flood Marks and Recent Data and (2) Recent
Data Alone. (In Cumecs).

1 Reluni Period I 2.33 (Mean) 10 25 50 100 1000 I

(1) Estimated Flood: 155 234 278 311 344 , 453
Large-s ample
standard error: t7 -+12 +i6 219 2 2 3 i33

(2) Estimated ' Flood: 156 236 281 314 348 458
Large-sample
standard error: -+ I l -+20 -+26 231 -t36 +51

THE AVON AT BATH - THE VALUE OF ADDITIONAL DATA

It will be seen from table 3 that the sampling error in the 50year flood
estimate was reduced from 10% to 6% by the use of historic flood marks. Was
this reduction worthwhile in view of the effort required to standardize the
data? A number of approaches mw be taken ì.n answering questions of this
type. Ultimately they involve the formulation of expressions for the benefits
and c,osrs which arise from acquiring the data, and since these c m never be
fully known, the problem of evaluating the worth of stream flow data are far
from etrai@t-ionrard.
 574

Reasonable attempts have been made, however, by E~RS of simplifying assumptions.

une such attempt has been made by Wilson (81, whose method allows the
estimation of the reduction in cost of a small structure consequent upon an
increase in the precision of its desis flood estimate. His approach is now
applied to data from the A v m at Bath.

It is assumed that the total cost C mey be written as C = $+Z2 where C1 is

associated with construction costs and has the form:-

c1 = XTm> (15)
and C2 is associated with the probable future damage resulting from structural
failure and has the form:-
c2 = K2 p,
S
i 16)
where i is the optimum design flood with return period T, K 1 and K2 are
constants, and m and s are indices dependent an the particular structure.
For s m l l structures m and 8 may be hssumed eausl.

Wilson's formda depends partly on the fact that floods with r e t m periods
between 5 and 50 )ears may be represented by a power law of the following form:-
5= A$
(17)
where A is a constant. p is an index which Wilscm suggests may be estimated
as the ratio of the 50-
>
to the S-year flood. Writing n = lb - 8, the
following formula gives the reduction in cost (Ec) of a structure, given the
precision of the estimate of the design flood (Ex):-
_-
Ec = E 'E
2 x'
It should be noted that this formula applies only to small structures, with
design floods of moderate return periods (ie. between 5 and 50 years).

For the river Avon at Bath, p was found to be 0.2; taking FS= 0.75 88 a
typical value, an increase in precision from 10% to 6% may be seen to lead
to a decrease of 1% in the cost of a structure with a 50 years design flood.
While not a high percentage, this would represent in absolute terms a sum of
money considerably in excess of the cost of obtaining and standardizing
 575

the data. More importantly, a similar cost-reduction, by this analysis,

wouïd require a further 20 years of streamflow data from continuous records,
which might be impractical and would certainly be expensive.

Since the incorporation of the other type of nan-standard data considered,

Le, chart censoring resulting in missing peaks, entails no extra cost, and
bearing in mind that flood estimates are likely to be of interest in a
variety of contexts (not only one as in the above example), it may be
concluded that it is on the whole worthwhile to use additional data of the
types described.

Acknarle dgement

The author wishes to thank the following for their assistance:

Robin T. Clarke, who initially suggested this study and provided
valuable guidance during its progress; Con Cunnane, who made available
compu+.er programs,'adaptstions of which were used in this work> and
Dr Malcoiz D.Newson who collated and helped to standardize the historic data
used in the eyample. This paper is presented by permission of the Director,
Institute of Hydrology, Wallingford, Berkshire, U.K.

1. Gumbel, E.J. (1960) Statistics of Extremes Columbia University Press

Iiew York (1959).
2. Moran, P.A.P. (1959) !he Theory of Storage. Methuen and Co. London (1970).
3. Lowery, M.D. and Nash, J.E. (1970) A comparison of methods of fitting the
double exponential distribution. Journal of Hydrology IO, 259-275.
h. Kimball, B.F. (1949) An approximation t o the sampling variance of an
estimated maximum value of given frequency based on fit of doubly exponential
distribution of m&mum values. Ann. Math. Stat., 110-1 13.
5. Kendall, M.G. and Stuart N. (1961) The Advanced Theory of'statistics
Vol.11. Charles Griffin and Co., Ltd, London.
6. Harper, H.L. and Moore, A.H. (1968) Maximum-likelihood estimation, from
doubly censored samples, of the parameters of the first asymptotic distribution
of extreme values. h e r . Stat. Assoc. Jour. 63. 889-901.
7. Jenkinson, A.F. (1969) Estimation. of Maximum Floods. Chapter five of
W Technical Report 98, 193-227.
8. Wilson, K. C. (1972) Benefit-accuracy relationship for small structure
design floods. Weter Resources Research 8(2), 508-512.
 ASSESSMENT O F DESIGN F L O O D S IN BRAZIL

Paulo P o g g i P e r e i r a

ABSTRACT

The t e c h n i q u e s utilized by t h e D e p a r t a m e n t o N a c i o n a l de O b r a s de
S a n e a m i e n t o f o r c o m p u t i n g t h e c a r a c t e r i s t i c s o f flood t o be u s e d f o r
designing works against i n u n d a t i o n s a r e described. Very s e l d o m t r u s -
tworthy r i v e r flood discharge m e a s u r e m e n t s are obtained. In most ca-
s e s design flood discharges a r e e s t i m a t e d w i t h a basis o n t o p o g r a p h i c
d a t a which c a n be gathered q u i c k l y . U n t i l t h i r t y y e a r s ago t h e c o n t r i
buting basin area w a s multiplied by a s t a n d a r d unit discharge in or-
d e r to get t h e design flood discharge. L a t e r on, t h e r a t i o n a l m e t h o d
w a s addopted, m a i n l y f o r designing s m a l l c a n a l s . This s y s t e m w a s c o n -
siderably improved by t h e e x e c u t i o n o f . a s t a t i s t i c a l study o f heavy
r a i n s observed in t h e Country. The choice o f the h e i g t h o f s o m e d i k e s
w a s based o n the h i g h w a t e r levels a t t a i n e d d u r i n g ancient f l o o d s o b -
s e r v e d a n d s t i l l remembered by l o c a l people. It h a s been f o u n d n e c e s -
sary t o perform more elaborate and time-consuming h y d r o l o g i c a l o b s e r -
v a t i o n s and s t u d i e s f o r d e s i g n i n g dams. The u s e o f m a t h e m a t i c a l m o d e l s
i s s t i l l Incipient but promising. Design f l o o d s o f different s t a n d a r d
p e r i o d s o f i e c u r r e n c e are a d d o p t e d a c c o r d i n g t o t h e t y p e o f t h e w o r k ,
t h e s i z e o f t h e r i v e r and t h e u t i l i z a t i o n g i v e n t o t h e area t o be p r o
tected.

RES UME N

S o n d e s c r i t a s las t é c n i c a s e m p l e a d a s p o r e l D e p a r t a m e n t o N a c i o n a l
de Obras de S a n e a m i e n t o en l a d e t e r m i n a c i ó n de l a s c a r a c t e r i s t i c a s de
l a s c r e c i d a s a s e r c o n s i d e r a d a s e n e l p r o y e c t o de o b r a s c o n t r a i n u n d a
ciones. Raramente se c o n s i g u e n datos de m e d i c i o n e s f i d e d i g n a de l a s
d e s c a r g a s de c r e c i d a s de los c u r s o s de agua. E n la m a y o r i a de l o s ca-
sos e s t i m a n s e descargas de c r e c i d a s para e l proyecto, c o n b a s e e n da-
tos t o p o g r á f i c o s que pueden s e r o b t e n i d o s r á p i d a m e n t e . H a s t a t r e i n t a
a ñ o s atrás, e l m é t o d o u t i l i z a d o c o n s i s t i a en m u l t i p l i c a r e l á r e a de
l a c u e n c a h i d r o g r á f i c a c o n t r i b u y e n t e p o r u n a d e s c a r g a e s p e c i f i c a pa-
d r o n i z a d a p a r a o b t e n e r l a d e s c a r g a de c r e c i d a p a r a e l proyecto. De
a h í e n adelante, f u e adoptado e l m é t o d o r a c i o n a l , p r i n c i p a l m e n t e p a r a
p r o y e c t a r pequefios canales. E s t e s i s t e m a fue c o n s i d e r a b l e m e n t e m e j o r a
d o p o r l a e j e c u c i ó n de u n e s t u d i o e s t a d í s t i c o de l a s l l u v i a s i n t e n s a s
o b s e r v a d a s e n e l país. La a l t u r a de a l g u n o s d i q u e s fue e s c o g i d a e n b a
s e de los n i v e l e s de a g u a a l c a n z a d o s p o r a n t i g u a s crecidas, c u y o s v e s
t i g i o s p e r d u r a n t o d a v i a y s o n i n d i c a d o s p o r los m o r a d o r e s d e l l u g a r .
P a r a e l p r o y e c t o de r e p r e s a s h a s i d o n e c e s a r i o r e a l i z a r o b s e r v a c i o n e s
y estudios h i d r o l o g i c o s m á s p r e c i s o s y demorados. E l u s o de m o d e l o s
m a t e m á t i c o s e s aún incipiente, n o o b s t a n t e , promisor. T a m b i é n , a d o p -
t a n s e c r e c i d a s de p r o y e c t o c o n d i f e r e n t e s p e r i o d o s de r e c u r r e n c i a c o z
f o r m e e l t i p o de l a obra, e l c a u d a l d e l c u r s o de agua y l o s i n t e r e s e s
e n j u e g o de l a s c o m u n i d a d e s vecinas.
578

-1. INTRODUCTION
T h e D e p a r t a m e n t o N a c i o n a l de O b r a s de S a n e a m e n t o -
D.N.0.S.-
of the B r a z i l i a n M i n i s t r y of I n t e r i o r , h a s b e e n b u i l d i n g flood
c o n t r o l w o r k s f o r a l m o s t 4 0 years.
Such w o r k s i n c l u d e c h a n n e l i m p r o v e m e n t s , d r e d g i n g and l i n i n g
of canals, b u i l d i n g of l e v e e s , d a m s , c o n d u i t s and tunnels.
T h e first b a s i c s t e p i n the d e s i g n of these w o r k s i s the de-
t e r m i n a t i o n of the f e a t u r e s of the f l o o d s to b e c o n t r o l e d o r t a k e n
i n t o account. The m a i n m e t h o d s that h a v e b e e n u s e d for t h i s p u r p o -
s e are p r e s e n t e d in t h e f o l l o w i n g subtitles. It s h o u l d b e noted
h o w e v e r that n o t e v e r y m e t h o d r e p o r t e d i s s t i l l i n use.
T h e r e is a g e n e r a l i z e d l a c k of good r e l i a b l e h y d r o m e t r i c
o b s e r v a t i o n s and m e a s u r e m e n t s . As a c o n s e q u e n c e , i n d i r e c t h y d r o -
l o g i c m e t h o d s h a v e b e e n u s e d as a r u l e w i t h very f e w exceptions.

2. RATIONAL METHOD
T h e r a t i o n a l method i s the most w i d e l y a d o p t e d f o r d e s i g n i n g
c a n a l s and c o n d u i te.
it gives t h e d e s c h a r g e Q - -
t h r o u g h the e q u a t i o n Q
t h e elemerits of w h i c h are d e t e r m i n e d as f o l l o w s :
= CIA ,

T h e a r e a of t h e d r a i n a g e b a s i n
o r a e r i a l photographs.
- -A is o b t a i n e d from m a p s
W h e n n o n e is a v a i l a b l e , f i e l d s u r v e y s are
made.
The runoff coefficient - -
c d e p e n d s p r i m a r i l y on land use.
A s an e)rample the f o l l o w i n g table w a s c o p i e d f r o m (i), a r e c e n t
D.N.0.S.- O.A.S. p u b l i c a t i o n :
D o w n t o w n a r e a s , d e n s e l y b u i l t , w i t h p a v e d s t r e e t s and s i d e -
walks C = 0.70 t o 0.90
N e i g h b o r h o o d areas, less d e n s e l y b u i l t , w i t h p a v e d s t r e e t s
and s i d e w a l k s
Residential areas densely built, with paved streets C
,C 0.70

R e s i d e n t i a l a r e a s a v e r a g e l y i n h a b i t e d C = 0.55
S u b u r b a n r e s i d e n t i a l areas, s p a r s e l y built C = 0.35 to 0.55
to 0.65
-0.65

R e s i d e n t i a l a r e a s w i t h g a r d e n s and u n p a v e d s t r e e t s C = 0.30
V e g e t a t e d a r e a s , p a r k s w i t h gardens, u n p a v e d sport f i e l d s
c = 0.20
The v a l u e of the r u n o f f c o e f f i c i e n t for t h e d r a i n a g e b a s i n
is o b t a i n e d by a d d i n g the p r o d u c t s of t h e f r a c t i o n s of t o t a l drain-
a g e a r e a o c c u p i e d by e a c h land u s e , m u l t i p l i e d by the c o r r e s p o n d i n g
coefficient.
T h e d e t e r m i n a t i o n of the r a i n i n t e n s i t y
the f o l l o w i n g steps:
- I - is m a d e through
 57 9

a) A r e c u r r e n c e i n t e r v a l is c h o s e n , u s u a l l y obeyine, the
f o l l o w i n g c r i t e r i a (1 and 2):
R u r a l area U r b a n area
Small c a n a l (no levees) 5 years 10 years
Large c a n a l ( n o levees) 10 years 25 y e a r s
Small canal with levees 25 years 5 0 years
Large canal with levees 50 years 100 years
Small conduits for urban drainage 3 o r more y e a r s
b) T h e time o f c o n c e n t r a t i o n is c o m p u t e d by a d d i n g the t i m e
n eeded by the rainwater fallen on the r e m o t e s t part o f the watershed
e o r e a c h the canal or c o n d u i t , to the t r a v e l time n e c e s s a r y for the
w a t e r to f l o w to the point under study. The t r a v e l t i m e i s c o m p u t -
ed by d i v i d í n p the length of the c a n a l or conduit by the a v e r a p e
flow v e l o c i t y .
c) A total depth of r a i n f a l l is d e t e r m i n e d t a k i n g i n t o
account the chosen r e c u r r e n c e i n t e r v a l and a d u r a t i o n of r a i n e q u a l
t o the t i m e of concentration. (3) is r e s o r t e d t o f o r t h i s p u r p o s e .
T h e r a t i o r a i n f a l l d e p t h t r a i n d u r a t i o n gives r a i n f a l l i n t e n s i t y I.

3. I N T E N S E R A I N S IN B R A Z I L
In 1 9 5 7 D.N.O.S. edited O t t o Pfafstetter's " I n t e n s e R a i n s in
Brazil" p r e p a r e d mainly f o r a p p l i c a t i o n s o f the r a t i o n a l m e t h o d (3).
T h i s b o o k p r e s e n t s the r e s u l t s o f f r e q u e n c y a n a l y s i s of r a i n
fall v h l u e s recorded i n 9 8 s t a t i o n s of the B r a z i l i a n D e p a r t a m e n t o
N a c i o n a l le Meteorologia.
R a i n f a l l c o r r e s p o n d i n g t o s e v e r a l d u r a t i o n p e r i o d s of r a i n
(5, 15 and 30 m i n u t e s , 1, 2, 4, 8 , 14, 2 4 and 4 8 h o u r s , 1 , 2 , 3, 4
and 6 o b s e r v a t i o n days) w e r e a n a l y s e d s e p a r a t e l y f o r e a c h s t a t i o n .
-
R e c u r r e n c e i n t e r v a l s of the p r e c i p i t a t i o n s T -were carac-
t e r i z e d by the e q u a t i o n T = n / m , b e i n g n the t o t a l p e r i o d of o b s e r -
v a t i o n and m the n u m b e r of o r d e r o c c u p i e d by the r a i n f a l l i n a
s e r i e s w h e r e a l l observed i n t e n s e p r e c i p i t a t i o n s w e r e p l a c e d i n de-
c r e a s i n g order of magnitude.
T h i s b o o k p r e s e n t s d i a g r a m s , tables and f o r m u l a s that a l l o w
t h e d e t e r m i n a t i o n of design r a i n f a l l for the 9 8 s t u d i e d s t a t i o n s u p
t o 1 0 0 0 y e a r s of r e c u r r e n c e intervals. V a l u e s p e r t a i n i n g t o the
s t a t i o n n e a r e s t t o the p l a c e f o r w h e r e the d e s i g n i s b e i n g prepared
are u s u a l l y utilized. For checking representativeness, rainfall
f r e q u e n c y c u r v e s o f d a y l y p r e c i p i t a t i o n s of this s t a t i o n a r e s o m e -
times compared with similar curves prepared with data from a non r e
c o r d i n g r a i n g a g e i n s t a l e d at t h e a c t u a l p l a c e of t h e c o n t e m p l a t e d
works.

4. S T A N D A R D UNIT D I S C H A R G E S
A c c o r d i n g t o ( 4 ) the r a t i o n a l m e t h o d w a s a d d o p t e d w h e n D.N.QS.
b e g a n i t s a c t i v i t i e s many y e a r s a g o r e c l a i m i n g s w a m p s i n t h e n e i F h -
5 80

bn,irhood of Rio de Janeiro.

T h e r e a s o n s for this c h o i c e w e r e the a b s e n c e of d i s c h a r g e
~ ~ ~ . , ? i ~ ~ e m the e n t fs r,e q u e n t i n e x i s t e n c e of defined s t r e a m s i n the
s w a m - i s and because it w a s f e a r e d that the d r a i n a g e c a n a l s to b e c o n s
t r u c t c d w o u l d c h a n g e s o much the h y d r a u l i c c a r a c t e r i s t i c s of the
w a t e r s h e d s that the m e a s u r e m e n t s w o u l d not p r o v i d e a r e l i a b l e basis
f o r designs .
On the other h a n d , t h e r e w e r e n o r e c o r d i n g r a i n gage charts
f i o m ~ t . i c hr a i n f a l l i n t e n s i t i e s for d i f f e r e n t r a i n f a l l d u r a t i o n s
and r e c u r r e n c e i n t e r v a l s could b e deducted.
T h e r e w e r e only r a i n f a l l m e a s u r e m e n t s p e r f o r m e d with n o n r e -
r o r d i n g r a i n gages f o r a r e l a t i v e l y s h o r t period w h i c h showed a ma-
x i m u m p r e c i p i t a t i o n of 1 2 0 m m f o r l day (24 hours).
To get r a i n f a l l i n t e n s i t y , this observed d e p t h of precipit-
a t i o n w d s supposed t o b e uniformly d i s t r i b u t e d t h r o u g h the 2 4 hours
of observation.
S o , the r a i n i n t e n s i t y a d d o p t e d w a s a l w a y s the same, r e g a r d -
l e s s of the t i m e o f c o n c e n t r a t i o n of the v a r i o u s basins. As a con-
s e q u e n c e , the d i s c h a r g e b e c a m e d i r e c t l y p r o p o r t i o n a l to the drain-
a g e b a s i n area.
The r u n o f f c o e f i c i e n t a d d o p t e d €or r u r a l b a s i n s w a s 0.7,
o b v i o u s l y f o r c o m p e n s a t i n g the w e a k r a i n f a l l i n t e n s i t y used. For
t h e s e 3v a l u e s , the r a t i o n a l met9od e q u a t i o n gives a d i s c h a r g e of
1 0 0 m /s f L t a b a s i n of 1 0 0 k m .
A s a m a t t e r of f a c t this a p p l i c a t i o n of the r a t i o n a l method
w a s only a m e a n s of j u s t i f y i n g the s t a n d a r d unit d i s c h a r g e of 1 m 3 /
1s k m 2 that w a s an addopted r u l e of thumb. The b e h a v i o u r of the u~
l i n e d r u r a l c a n a l s d e s i g n e d a c c o r d i n g l y h a s b e e n good. Some ocasion
a l f l o o d i n g h a s o c c u r e d but w i t h o u t e x c e s s i v e r e s u l t i n g damage.
3
A n o t h e r s t a n d a r d unit d i s c h a r g e i s 0.5 m Is km2. It reeult-
e d f r o m a d e s i g n e s p e c i f i c a t i o n a s k i n g for p u m p i n g r a i n w a t e r out of
p o l d e r s w i t h i n a f e w d a y s for a v o i d i n g the b r e e d i n g of m o s q u i t o e s .
H e r e a g a i n r a i n i n t e n s i t y w a s n o t r e l a t e d t o the c o n c e n t r a t i o n time
of the d r a i n a g e basin.

5. D E S I G N F L O O D S F O R DAM S P I L L W A Y S - TRIANGULAR UNITGRAPH

T h e design OP d a m s p i l l w a y s i s usually b a s e d on flood h y d r o -

graphs. T h e t r i a n g u l a r u n i t g r a p h p r e s e n t e d in (5) h a s b e e n used
v e r y o f t e n b e c a u s e It p r e s e n t s the a d v a n t a g e of d o i n w i t h o u t h y d r o -
m e t r i c data.
It i s b e l i e v e d that p e a k d i s c h a r g e s o b t a i n e d b y this m e t h o d
a r e e x a g e r a t e d but flood v o l u m e s a r e correct. T h e r e f o r e t h i s method
i s c o n s i d e r e d good for c a s e s w h e r e the d a m r e s e r v o i r r e t a i n s much
of the flood volumes.
 581

T h e f o l l o w i n g e x a m p l e is b a s e d on r e c e n t d e s i g n computations
of a dam s p i l l w a y f o r N o r t h e a s t e r n Brazil.
a) The time of c o n c e n t r a t i o n w a s e s t i m a t e d by the e q u a t i o n
of the “ C a l i f o r n i a H i g h w a y s and P u b l i c W o r k s ” adapted f o r m e t r i c
units : 3
5 0.95 x (L /
TC
Tc = time o f c o n c e n t r a t i o n in h o u r s
I, = length of w a t e r c o u r s e i n k m m e a s u r e d from divide to
s p i l l w a y site.
H = d i f f e r e n c e in e l e v a t i o n i n m e t e r s b e t w e e n s p i l l w a y s i t e
and divide.
In our e x a m p l e L = 1 7 km, H = 400 m and
Tc = 0.95 ( l ì 3 / 4 0 0 ) ~ ’= ~2.5
~ ~h o u r s
b) The t i m e in hours from s t a r t to p e a k r a t e o f u n i t g r a p h
(T ) w a s computed as follows for e x c e s s r a i n s of 1 and 6 h o u r s

T
P
-
p e r i o d s (D)
D / 2 + 0.6 T c

-
For D
For D
=
=
1 hour, T
6 hour6,
P
T
P
1 1 2 + 0.6 (2.5)
= 6 1 2 + 0.6(2.5)
=
-
c) T h e time i n h o u r s from p e a k r a t e t o end o f u n i t g r a p h
2 hours
4.5 h o u r s

t r i a n g l e (T,) w a s computed as f o l l o w s :
Tr
For D
For D
= 1.667 T

=
=
P
1 hour, T r
6 hours, Tr
-- 1.667(2)
1.667(4.5)
=
3.3 h o u r s
= 7.5 h o u r s
d) P e a k r a t e s of u n i t g r a p h s f o r 1 mm e x c e a s r a i n f a l l of 1
and 6 h o u r s d u r a t i o n p e r i o d s w e r e c o m p u t e d as f o l l o w s :
I
A
‘p 1.8(T + Tr)
P
= p e a k r a t e in m 3 /e
4P
A

For D
= d r a i n a g e a r e a i n km2.
F o r D = 1 hour,

= 6 hours,
‘p
9
9 7
1.8(2 + 3.3)
œ
10.2 m 3 ~ s
97
--
I n the e x a m p l e A = 9 7 k m 2

4.5 m 3 / s
qp 1.8(4.5 + 7.5)
e) The e x c e s s r a i n f a l l s and c o r r e s p o n d i n g r u n o f f hydrographs
-unitgraphs - are r e p r e s e n t e d s c h e m a t i c a l y i n the a n n e x f i g u r e s t o -
gether w i t h l i s t s o f u n i t g r a p h d i s c h a r g e s c o r r e s p o n d i n g to the
m i d d l e of c o n s e c u t i v e one h o u r t i m e intervals.
6. PROBABLE MAXIMUM PRECIPITATIONS
The s p i l l w a y of the e x a m p l e w o u l d b e l o c a t e d u p t r e a m of a
large town and the f a i l u r e of i t s dam by flood o v e r t o p i n g w o u l d
c a u s e great p r o p e r t y d a m a g e and s e r i o u s l y j e o p a r d i z e human life i n
l a r g e numbers. T h e r e f o r e it v a s c o n s i d e r e d a p p r o p r i a t e to utilize
the m a x i m u m p r o b a b l e p r e c i p i t a t i o n f o r c o m p u t i n g the design flood.
T h e r e w e r e n o t e n o u g h s t o r m d a t a for e s t i m a t i n g d i r e c t l y
the v a l u e s of s u c h p r e c i p i t a t i o n . T h e indirect approximated method
p r o p o s e d i n (6) w a s used. It i s b a s e d on s u p p o s ing that m a x i m u m
p r o b a b l e p r e c i p i t a t i o n v a l u e s are i d e n t i c a l to those of a r e g i o n of
the United S t a t e s w h e r e r a i n f a l l s of 1 0 y e a r s r e c u r r e n c e i n t e r v a l
are the s a m e as those o b s e r v e d i n t h e w a t e r s h e d u n d e r study.
(3) w a s used f o r o b t a i n i n g 1 0 y e a r s r e c u r r e n c e i n t e r v a l p r g
c i p i t ô t i o n s from a s t a t i o n n e a r b y the s p i l l w a y s i t e and (7) p e r m i l
e d to l o c a t e the a r e a i n the United S t a t e s w i t h e q u i v a l e n t r a i n -
f a l l s and a l s o f u r n i s h e d the p r o b a b l e m a x i m u m 6-hour p r e c i p i t a t i o n
f o r a 10-square-mile area: 6 8 6 mm.
B y u s i n g c h a r t s f r o m (5) v a l u e s of p r o b a b l e m a x i m u m p r e c i -
p i t a t i o n s w e r e c o m p u t e d f o r the d r a i n a g e b a s i n u n d e r study, w h i c h
h a s an a r e a of 9 7 k m 2 = 37.5 square-mile, for t h e f o l l o w i n g listed
p e r i o d s of duration.
duraticm period computation rainfall
houss rnm
6 88% x 686 604
12 107% x 686 7 34
1 50% x 604 302
2 65% x 604 39 2
3 76% x 604 460
4 85% x 604 513
5 93% x 604 562
R a i n f a l l i n c r e m e n t s d i s p o s e d i n d e s c e n d i n g order of intensity
w e r e c a l c u l a t e d as f o l l o w s :
interval duration rainfall increments
hours mm
1 P1 = 302

P2
392 - 302 = 90

p3
460 - 392 68
513 - 460 = 53
P4
P5 = 562 - 513 = 49
P6 604 - 562 42
P12= 734 - 604 1130
 583

For obtaining the design precipitation, rainfall increments

w e r e t a b u l a t e d i n t h e f o l l o w i n g o r d e r a s s u g g e s t e d i n (5): P6, Pq;
P 3 , P1, P 2 , l 5 n a d P 1 2 ( s e e a n n e x table).

7. R U N O F F F C T I M A T I O N A N D C O M P U T A T I O N OF T H E D E S I G N F L O O D H Y D R C
GRAPH
T h e c o m p u t a t i o n o f the d e s i g n f l o o d h y d r o g r a p h o f t h e e x a m -
ple i s p r e s e n t e d i n t h e a n n e x t a b l e a n d w a s m a d e t h r o u g h the follo_w
ing steps:
a) R a i n f a l l i n c r e m e n t s o b t a i n e d a s of t h e p r e c e d i n g c u b -
title were added i n order to obtain accumulative precipitation.
b) Accumulative runoff or excess rainfall was estimated by
m e a n s o f the e q u a t i o n of t h e "Soil C o n s e r v a t i o n S e r v i ce" p r e s e n t e d
i n (5): 2
(P
R = P + 0.2
- S)
0.8 s
R = runoff in mm
P = accumulative precipitation in mm
S = maximum potential difference P
begining.
-
R at t i m e of rain's

S w a s e s t i m a t e d a s 100 mm.
c) Increments of runoff w e r e computed by subtracting the
accunulative runoff obtained for the preceding interval from the
accumulative runoff obtained for the interval under consideration.
d) I n c r e m e n t s of r u n o f f w e r e c o m p a r e d w i t h rainfal.1 i n c r e -
ments. T h e d i f f e r e n c e b e t w e e n t h e m s h o u l d a t t a i n at l e a s t l m m f o r
each interval hour. A s t h i s d i d n o t h a p p e n at t h e l a s t t a b u l a t e d
t i m e i n t e r v a l t h e i n c r e m e n t of r u n o f f f o r t h a t i n t e r v a l w a s r e c a l -
culated by subtracting 6 mm from Phe rainfall increment.
e) I n c r e m e n t s of r u n o f f f o r e a c h i n t e r v a l w e r e m u l t i p l i e d
by t h e u n i t g r a p h d i s c h a r g e s l i s t e d i n t h e a n n e x f i g u r e s a n d t h e p r o
d u c t s w e r e t a b u l a t e d i n the c o r r e s p o n d i n g t i m e i n t e r v a l s .
f) T h e a v e r a g e d i s c h a r g e o f t h e d e s i g n f l o o d i n e a c h t i m e
interval w a s obtained by adding the products resulting from the
p r e v i o u s s t e p f o r t h a t tf.me i n t e r v a l .
g) T h e b a s e f l o w w a s n o t t a k e n into a c c o u n t f o r b e i n g no-
minal.

a. STATISTICAL METHODS

F r e q u e n c y a n a l y s i s is a p p l i e d w h e n e v e r r e c o r d s t h a t a l l o w
i t s u s e a r e a v a i l a b l e , f o r r e a s o n s of b e t t e r p r e c i s i o n and r e l i a -
bility. Gumbel's a n d / o r Hazen's m e t h o d s a r e t h e m o s t f a v o r e d .
D.N.O.S. f i l e s k e e p r e p o r t s o f c l a s s i c a l h y d r o l o g i c a l s t u d i e s
mainly based o n frequency analysis of w a t e r level observations
and discharge measurements.
5 84

One of them is an o u t s t a n d i n g l y i n t e r e s t i n g example: the d e

t e r m i n a t i o n of the h e i g t h of l e v e e s for p r o t e c t i o n of the city of
P o r t o A l e g r e a g a i n s t f l o o d i n g s of the G u a i b a River.
I n that r e a c h the G u a i b a R i v e r f o r m s a n e s t u a r y and its
w a t e r levels are d e p e n d e n t n o t only o n the r i v e r d i s c h a r g e s as w e l l
as on the w a t e r l e v e l o c u r r i n g i n the lagoon w h e r e it f l o w s t o ,
w h i c h can b e s t r o n g l y i n f l u e n c e d by winds.
T h e r e w e r e l i t t l e k n o w l e d g e of the e l e m e n t s i n v o l v e d and
their effect.
On the o t h e r h a n d the G u a i b a R i v e r w a t e r l e v e l s had b e e n s y s
t e m a t i c a l y o b s e r v e d s i n c e 1 8 9 9 by m e a n s of a s t a f f gage installed
near d o w n t o w n P o r t o Alegre. The d a t a so o b t a i n e d w a s frequency an=
l y s e d and, a c c o r d i n g to Gumbel's method the b i g g e s t recorded f l o o d ,
w h i c h o c c u r e d in 1 9 4 1 , w a s found to h a v e a r e c u r r e n c e i n t e r v a l of
about 370 years.
L o c a l p e o p l e r e m e m b e r e d w h i c h p l a c e s had b e e n flooded and
w h i c h l e v e l s had b e e n a t t a i n e d b y the w a t e r i n d i f f e r e n t p l a c e s of
t h e t o w n d u r i n g the 1 9 4 1 flood. W i t h these i n f o r m a t i o n s it w a s
p o s s i b l e to d r a w a w a t e r - s u r f a c e p r o f i l e , w h i c h w a s confirmed l a t e r
by a h y d r a u l i c m o d e l of the estuary.
It w a s decided to set the crest of the l e v e e s 1.20 m a b o v e
that water-surface profile. N o d i s c h a r g e c o n s i d e r a t i o n s w e r e t a k e n
i n t o a c c o u n t a l t h o u g h d i s c h a r g e s w e r e e s t i m a t e d by m a k i n g u s e o f
the above m e n t i o n e d model.

9. MATHEMnTICAL MODELS
Up t o p r e s e n t time a l m o s t n o u s e h a s b e e n m a d e of m a t h e m a t i c
a l h y d r o l o g i c a l m o d e l s f o r d e t e r m i n a t i o n of d e s i g n flood c a r a c t e -
ristics.
R e c e n t l y , the S t r e a m f l o w S y n t h e s i s and R e s e r v o i r R e g u l a t i o n
(SSARR) M o d e l b e g a n b e i n g u s e d for f o r e c a s t i n g the b e h a v i o u r (flood
and l o w w a t e r l e v e l s as w e l l ) of the P a r a g u a y R i v e r and s o m e tribu-
taries. T h i s m o d e l w a s d e v e l l o p e d by t h e U.S. A r m y Corps of Engi-
n e e r s w h i c h a d d a p t e d i t for t h e P a r a g u a y R i v e r b a s i n a s part of the
a c t i v i t i e s of the "Project of the H y d r o l o g i c a l S t u d i e s of t h e U p p e r
P a r a g u a y R i v e r Basin" - a UNDP/UNESCO technically assisted project
f o r w h i c h D.N.O.S. is t h e r e s p o n s i b l e B r a z i l i a n counterpart agency.
T h e p o t e n t i a l i t y of S S A R R m o d e l for e v a l u a t i n g the caracte-
r i s t i c s o f d e s i g n f l o o d s of l a r g e r i v e r s is o b v i o u s and i t is ex-
p e c t e d be much used for this p u r p o s e i n the future.

10. CONCLUSION

D.N.O.S. h a s a l w a y s used a d d a p t e d f o r e i g n feekiikques f o r

a s s e s s i n g d e s i g n floods. O n the o t h e r h a n d , l o c a l d a t a h a s b e e n
used a s e x t e n s i v e l y a s possible. M e t h o d s that did n o t a l l o w easy
 585

'use of this d a t a h a v e not e n j o y e d preference. S u c h is the case of

e m p i r i c a l f o r m u l a s f o r r a i n f a l l i n t e n s i t y and flood d i s c h a r g e w h i c h
w e r e used only in a f e w instances.
E l a b o r a t e m e t h o d s h a v e not b e e n much addopted. The m a i n
r e a s o n for this may be the r a t h e r v a g u e e f f e c t of h i g h a c c u r a c y
a s s e s s m e n t of d e s i g n flood c a r a c t e r i s t i c s u p o n the e c o n o m i c s of
f l o o d c o n t r o l w o r k s in most cases, a fact that does n o t e n c o u r a g e
too many e f f o r t s f o r r e f i n i n g d e s i g n flood assessment.

REFERENCES
a

1. D.N.O.S. e O r g a n i z a q a o dos E s t a d o s A m e r i c a n o s (1972). R e l a t ó -

r i o d o Estudo p a r a C o n t r o l e da E r o s a o n o N o r o e s t e do E s t a d o do
P a r a n a , Rio de J a n e i r o , DNOS.
2. P o g g i P e r e i r a , P. (1967). C o n t r o l e d e cheias: c u s t o s e b e n e f i
cios, S A N E A M E N T O , Rio de J a n e i r o , DNOS.
3. P f a f s t e t t e r , O. (1957). Chuvas Intensas n o Brasil, Rio de Ja-
neiro, DNOS.
4. A r a u j a Goes, H. (1942). A B a i x a d a d e S e p e t i b a , R i o de Janeiro,
DNOS.
5; U.S. D e p a r t m e n t of the I n t e r i o r , B u r e a u of R e c l a m a t i o n (1960).
D e s i g n of S m a l l Dama, W a s h i n g t o n , U.S. G o v e r n m e n t P r i n t i n g
Office.
6. P f a f s t e t t e r , O. (1967). F l o o d s for S p i l l w a y D e s i g n , N e u v i e m e
C o n g r e s des G r a n d s B a r r a g e s , C o m i s s i o n I n t e r n a t i o n a l e d e s
G r a n d s Barrages.
7. U.S. W e a t h e r B u r e a u (1963). R a i n f a l l F r e q u e n c y A t l a s of t h e
U n i t e d S t a t e s for D u r a t i o n s f r o m 30 M i n u t e s t o 2 4 H o u r s and
R e t u r n P e r i o d s f r o m 1 to 1 0 0 Y e a r s , T e c h n i c a l P a p e r N Q 40 ,
W a s h i n g t o n , W e a t h e r Bureau, U.S. D e p a r t m e n t of Commerce.
586

FIGURES

c.ü 5 1 hour
L-

or r u n o f f

-- r

<(unitiraph)

-qP
= 10.2m 3 Is

T
- P
-hourly i n t e r v a l s
L- ---
- = 2 hours

0-1 1-2
Tr
2-3
3.3

3-4
4
hours

4-5
_ d i s c h a r g e s in m 3 / s 2.5 7.5 8.6 5.5 2.4

+D = 6 hours- --i

1 m m e x c e s s r a i n f a l l or runoff

c r u n o f f h y d r o g r a p h (unitgraph)

bT m4.5 hours+Tr= 7.5 hours-------g(

P

- h o u r l y i n t e r v a l s 0-1 1-2 2-3 3-4 4 - 5 5-6 6-7 7-8 8-9 9-10 10-ilil-12
- disch.in m3/s 0.5 1.5 2.5 3.5 4.5 3.9 3.3 2.7 2.1 1.5 0.9 0.3
 5 87
N
d ou VIU
m m N N
I 4- wrl
W
I: N
-
M
--l
-
N
N

I
-
d
d

I
-
1
O
W
Io
ci vi
a O
O c
c a
.IC1
G \ o )
rl 0 o c
€ n u
<a
rl
ld d
e
o)

(II
?
&I
U al
c U
.rl G
rl
al
E c
rl U
w W
 A METHOD FOR THE PREDICTION OF
W A S H L O A D IN CERTAIN SMALL WATERSHEDS

by
Oswald Rendon-Herrero

ABSTRACT

Present knowledge on the prediction of washload reveals that with

the exception of the universal soil-loss equation, and sediment-rating
techniques, a rational method does not exist that can accomplish this
task. A method is presented thar is analogous to Sherman's unit-hydro-
graph method o f hydrograph analysis. The ordinates of a sediment dis-
charge graph are divided by the excess runoff that mobilized it, prod:
cing a unit sediment discharge graph. When this is done for many storm
events, unit sediment discharge graphs are generated that vary conside-
rably in peak value and shape. The ordinates of the latter graphs are
then plotted logarithmically against their respective excess runoff,
yielding data points that can be fitted by straight lines. Predictions
pf sediment discharge or the generation of a sediment discharge graph
for a given excess runoff can be accomplished using the resulting
graph, Bixler Run Watershed, Pennsylvania, having a drainage area of
15 square miles, was selected as a data source. Granulometric tests
and otti~rrelated information disclosed that the suspended sediment in
Bixler Ruri is predominantly washload. Prediction of washload utilizing
the propose? method yielded errors that were considerably less than
that reported using available sediment transport formulae and techni-
ques.

RE C U ME N

Actualmente los conocimientos con respecto a la predicción de

"washload" son bastante limitados. Con las excepciones de la ecuación
universal d e pérdida de suelo y técnicas sedimentarias (sediment-ra-
ting) todavía no existe un método racional para resolver esta tarea.
El procedimiento presentado es análogo al método de Sherman (Unit hy-
drograph) o sea un análisis hidrográfico. Las ordenadas de la gráfica
de descarga sedimentaria divididas entre el volumen del derrame excesi
vo producen una gráfica unitaria de descarga sedimentaria. A l comple-
tarse este procedimiento para muchas lluvias, gráficas unitarias de
descargas sedimentarias son obtenidas y se notarán las diferencias de
los cambios d e valores máximos. Las ordenadas de estas Últimas gráfi-
cas son trazadas logaritmicamente versus sus respectivos volúmenes de
derrame excesivo rindiendo diferentes puntos de dato, los cuales pue-
den unirse con líneas rectas. Predicciones de descargas sedimentarias
dado cierto derrame excesivo pueden observarse en la gráfica obtenida.
El área seleccionada de 1 5 millas cuadradas, donde l o s datos fueron a d
quiridos queda situada en Bixler Run, Penns.ylyania. Pruebas granulomé-
tricas y otras informaciones relacionadas indican que la aescarga de
sedimentos en Bixler Run es casi todo "washload". El método presentado
rindió errores de magnitud mínima en comparación con los errores repor
tados por otras técnicas y fórmulas de transporte sedimentarias.
 590

INTRODUCTION
Relationships have been developed whereby the sediment transport of mater-
ials which are native to a channel can be computed with varying degrees of
accuracy. When the sediment transport is primarily composed of the lateral in-
flow of particulate matter eroded from the land surface (washload) in a basin,
the relationships derived are no longer velid(lg2). Heretofore, the leteral
inflow component of sediment discharge was predicted via the universal soil
loss equation(*), and sediment rating techniques. Predicted quantities using
these methods are subject to large errors. The universal soil loss equation
has the disadvantage of providing only annual predictions, The need for quan-
titative evaluation of washload is of paramount importance a t the present time.
A method is presented which is applicable to certain small watersheds and
which can enable the prediction of sediment discharge on a storm basis. By
"small" is meant those watersheds where the spati
is uniform over the watershed area. Some authors$3 ,w ribution of the rainfall
define a small water-
shed as being less than 161.0 or as much as 3219.0 square kilometers in area.
"Certain" refers to the sediment discharge graph's locus (sedimentgraph) depen-
dency on the soil type. For general stream conditions, fine-grained and
colloidal materials transported in suspension will yield. a sedimentgraph that
appreciably parallels the shape of its associated hydrograph; under similar
stream conditions, coarser particles in transport will not result in parallel-
shaped discharge graphs. The applicability of the series graph method depends
on the para!lel nature of the sedimentgraph and hydrograph for a given excess
runoff. Use ~f the adjective "series" is explained in the Analysis of Da
section of this paper. The series graph method is analogous to Sherman's F%>
unit hydrograph prc,zedure for the analysis of a direct discharge hydrograph.
The series graph method is demonstrated using Bixler Run Watershed, a
monitored drainage basin 38.9 square kilometers in area near Loysville, Penn-
sylvania (Figure 1). Granulometric measurements made of the bed, bank, and
suspended sediment, has established the sediment transport in Bixler Run as
being predominantly washload. Sediment sampling in the Bixier Run Watershed was
begun on February 1, 1954, using a U.S.D-43, and a DH-48 depth integrating
hand sampler(7).
The series graph method is used where the quantitative analysis of wash-
load is necessary for the prediction of sediment discharge and/or variation
with time. The prediction of total sediment discharge is required for example
where the rate of sedimentation can become problematic. This consideration is
particularly important in the allocation of storage volumes in new reservoirs.
WASHLOAD
Due to a series of rainfall-induced erosive processes, particulate matter
eventually reaches a stream course after being transported through a great
variety of distances in a drainage basin. Depending on such characteristics
as, for example, land slope and length, topography, and availability of trans-
portable surficial soils, various-sized particles can, given ample time, reach
the main waterways in a basin.
Depending upon the streamflow character, some of the eroded materials that
reach the stream course as lateral inflow combine with sediments native t o the
channel proper and continue to be transported downstream by the prevailing flow.
The lateral inflow of Sediment is known as washload. Sediment transport in the
 591

s t-ct'arn may be accomplished by four generally accepted modes depending primarily

upon particle diameter and stream transport capability. The transport modes are
known as contact, saltation, suspended, and solution load. The saltation load
in combination with the contact load is generally assumed to comprise the bed
load. The sum of the suspended, bed, and solution loads is called the total
load. Of particular note here is the fact that there is no sharp line of de-
marcation between m terials tifried a s bed load ot as suspended load. Some
authors (e.g., Graf ?I), Shen , Chow(3)) have indicated that in many instances
the washload may comprise from 90 to 95 percent of the total sediment load.
The scope of this paper is limited solely to washload. Bed load, and
suspended sediments mobilized from the bed, are not considered within the con-
text of this paper.
THEORY
Of the numerous sediment transport equations that have been presented,
none have been derived which account for the lateral inflow of water-soil mix-
tures (washload) originating from sheet and gully erosion of land surfaces in
a drainage basin. Shen(*) points out, "Finally, none of the equations for pre-
dicting suspended load account for the washload of the stream." Shen(') also
indicates that application of the available suspended sediment transport
equations to stream give rise to substantial error.
The existing sediment transport equations are based solely on the mobili-
zation of fine particulate concentrations (sediment suspensions) and coarse
layered masses of the bed, which are native to the stream channel. Of import-
ance here Is the fact that in most instances, the quantity of washload derived
from lateral i n f l m can be substantially greater than the suspended sediment
native to the bei. Several authors (e.g., Graf (11, Shen(') y Chow(3)) estimate
that the bed load contribution to the total sediment load is usually on the
order of five percent, and may in some cases be 'neglected from total load cal-
culati m s .
Given the flow condition and composition of materials native to the bed,
several relationships have been developed that .provide a general relationship
for the rate of sediment transport. It is not the intent of this paper to
present a development of the available sediment transport (bed load, suspended
load, or total load) equations, since their basis of derivation places them
outside of the realm of washload phenomena and, therefore, the scope of this
study. The reader is referred to Graf (1) y Shen(') , and Nordin and McQuivey(8)
for a general development and assessment of available sediment transport
formulae.
COMPILATION OF DATA: BIXLER RUN WATERSHED
Storm events were chosen according to accepted hydrograph analysis criteria
and which appreciably satisfied certain analogous sedimentgraph analysis con-
ditions. The storm events were primarily classified according to the degree to
which the locus of fhe sedimentgraphs were defined by sampling. I n many instances
sampling in the region of the crest of the sedimentgraph was not accomplished;
the Bixler Run project hydrologist therefore estimated the peak's shape from
the relative positions of the rise and recession sample points and from know-
ledge of previous sedimentgraphs where the peak was known. The latter class-
ifications yielded 63 storm events, which were grouped on the basis of runoff
derived during winter (October to March) and Sumner months (April to September).
592

UNIT GRAPH DEVELOPMENT (WATER AND SEDIMENT DISCHARGE)

Processing of the stage hydrograph and sedimentgraph for inidividual storm
s!i>entsinvolved as a first step the separation of base flow from the total dis-
charge. In the case of the stage hydrograph, baseflow was assumed to comprise
both groundwater flow and interflow. Base flow for the sedimentgraph was
assumed to be the sediment flow prior to the beginning of the rise of a sedi-
mrntgraph for a particular storm event. The base flow separation technique
W J S identical for both the stage hydrograph and sedimentgraph (see Figure 2).
Point A (or A') on Figure 2 is defined as the point where the rise of the dis-
charge graph begins and is determined by inspection. Line AB (or A'B') is a
tangential straight line projection, continuous with the base f l m curve pre-
ceding it, emanating from point A ,(or A') and bisecting a vertical line drawn
through the peak. Generally, the points B and B' were appreciably in phase
for most of the storm events considered in the analysis. On the average, where
such was not the case the hydrograph peak lagged the sedimentgraph peak by one
hour. The point C (or Cl) is determined by drawing tangents on the recession
and base flow portions of the curves; the bisector of the tangents intersects
at a point assumed to be at the termination of surface runoff C (or C').
Although the separation technique utilized in this analysi bitrary , the
important feature is the consistency of its use throughout $3fay3f the data
process ing.
The resulting direct flow discharge graph data was then processed by
computer to derive the unit hydrograph, unit sedimentgraph, excess rainfall,
and associated sediment mobilized.
Hyetogrqhs were constructed for the selected storms in order to determine
duration for thc derived unit graphs. This was donefor winter and Sumner
rainfall storms ordy.
GRANULOMETRIC MEASUREMENTS OF SUSPENDED,
CHANNEL-BED, AND CHANNEL-BANK MATERIALS
Grain-size analyses of suspended-load , channel-bed , and channel-bank
materials were conducted by the USGS District Office, Surface Water Quality
Branch, Harrisburg, Pennsylvania.
The compiled data serves as a basis for comparison of the materials trans-
ported during storm events and as a basis for reiative classification of the
prevailing transport mode (washload, bed load, etc.).
Results of 115 granulometric tests conducted on the bed, bank, and suspend-
ed sediment samples are plotted in Figure 3. Granulometric distributions ob-
tained from the tests generally plot as three distinct bands of points, with a
minor degree of overlapping. For clarity, only the arithmetic mean curves are
presented on Figure 3. These were determined by sumning the percents finer than
by weight at a given particle size diameter and material source (bed, bank, or
suspended), and obtaining an arithmetic average.
Figure 3 corroborates verbal communication between the project hydrolo-
gists of the USGS, Harrisburg, Pennsylvania, and this worker, to the effect
that materials encountered in the bed are primarily coarse-grained. In many
instances, bedrock is exposed at the surface. The suspended material, there-
fore, can only have had as its primary source of origin the watershed's land
slopes. The distinctness with which the individual mean curves plot on Figure
3 is also indicative of significantannoring of the bed. Armoring is the time-
 593

wise removal of fine particulate matter from the bed.

ANALYSIS OF DATA
Sed imcntgrriph Analysis
The original premise proposed in this study was that the unit hydrograph
- .

concept as applied to a direct runoff hydrograph was directly analogous in the

iinalyiiis of d sedim@negraph, A fami of a tirlit gedinientgraph idas indeed develop.
ed whose standard unit was 1.0 kilogram for a given duration, distributed over
the watershed area, analogous in unit-hydrograph analysis to 1.0 centimeter of
excess (effective) rainfall over the same area. The shape of the resulting
unit sedimentgraphs varied only slightly for different rainfall events of a
given duration, a s is anticipated in unit-hydrograph analysis. In order to
utilize such a unit sedimentgraph in generating a sedimentgraph for a particular
storm event, the total amount of sediment mobilized during the event would have
to be known or estimated. A relationship has been determined between total
sediment mobilized and excess runoff for single storm events. This is shown
on Figure 4, for runoff events resulting from winter and summer storms.
The latter approach would, therefore, entail the estimation of total
sediment mobilized on the basis of a known or predicted runoff excess a s an
initial step, followed by the selection, based on duration, of an appropriate
unit sedimentgraph. The latter can then yield a sedimentgraph by multiplying
the individual unit sedimentgraph ordinates by the total sediment mobilized.
A simpler approach, however, was adopted which has the advantage that consider-
ation of duration of runoff excess may be neglected altogether; the relation-
ship developed, as will be shown, is independent of duration.
Once the observed total discharge hydrographs and sedimentgraphs were
graphically convzrted to direct discharge graphs by deducting the base flow,
the following calculations were performed:
DDi = DTi - DBi (1)
Where DD. is direct water discharge in cubic meters per second (hereinafter
designated as crns)
DT is total water discharge in crns
i .
DB. is base water discharge in crns.
The subskript "i" refers to the time at which water discharge values (e.g.,
DTi) are measured on the hydrograph's abscissa. F o r this analysis the hydro-
graph's time base was divided into "n" two-hour increments.
Similarly,
SDi = STi - SBi (
2)
where SD, is direct sediment discharge in parts per million (hereinafter
designated as ppm)
ST. is total sediment discharge in ppm
SB: is base sediment discharge in ppm.
Thelmagnitude of the ppm units is equivalent to mg/A units (milligrams per
liter) as long a s the sediment concentration does not exceed 15,900 ppm (9).
Concentrations greater than 15,900 ppm have to be multiplied by a factor (9) in
order to convert ppm to mg/Q units. The units of.the direct sediment discharge
(SDi) are then converted from ppm units to kilograms per day, thusly,
 594

Where S i is direct sediment discharge in kilograms per day, 86.56 is a factor

for converting the sediment discharge to kilograms per day.
Excess runoff and the associated sediment mobilized are determined as
follows:

i-1

i=l
Where ER is excess runoff in centimeters per square kilometer of drainage basin,
A is the watershed area in square kilometers,
0.1157 is a factor for converting the remaining elements of Equation 4 to
centimeters per square kilometer,
ES is sediment mobilized in kilograms per square kilometer.
Individual unit sedimentgraph ordinates are determined thusly,

usoi = i'
Yhere USO. is the individual unit sedimentgraph ordinate in units of square
kilometer per day. Multiplying USOi by kilograms per squre kilo-
Eaters , yields kilograms per day.
Equation 6, as was previously pointed out, cannot directly be used in the
fashion of a unit hydrograph ordinate. The method, therefore, requires the
following operation,

i'
SGO.i = (7)
Where SGOi is an individual "series" graph ordinate in units of kilograms per
day per centimeter of excess.runoff per square kilometer. By
"series" is meant that in contrast to a unit sedimentgraph ordinate
which approximately superpose each other for a given duration, a
series of graphs are obtained which vary considerably in shape and
peak. For the purpose of discussion, Figure 5 will hereinafter be
referred to as a series graph.
The series graph lines were developed by plotting SGOi for a given excess
runoff. This entailed some judgment in the selection of coordinate points. The
latter procedure is analogous to selecting a mean unit hydrograph curve from a
number of curves, which in practice generally do not overlap for a given
duration. The judgment used was partly justified by the fact that the least
squares line fit of the selected coordinate points (p, p 2 2, etc.) have a
distinct tendency to plot approximately parallel to each other. This is in-
dicative of a prevailing trend.
Series graphs were constructed for winter including rainfall and snowmelt,
and for s m e r months. These are shown on Figure 5. The SGO. versus "ER"
coordinate points were plotted in time groups referenced to the peak discharge
(p). Thus p + 2 for example, refers to the direct discharge ordinate two hours
 595

after the peak; in all, the time increments considered were p,+ 2, p + 4, and
f i r the summer events only, p + 6. Generally p + 6 represents a negligible
discharge quantity, very frequently zero, and was therefore assumed to be
zero for winter rainfall and snowmelt events.
This writer is of the opinion that for this particular analysis a great
part of the data scatter on the series graph and Figure 4 , can be explained by
the mnnncr in which the sedimentgraphs were defined by sampling. The data
points are not shown since some overlapping exists as refers to p 5 n lines.
The sedimentgraphs selected for analysis did not have continuously defined loci.
A s a result, graphical interpolation and judgment by the USGS, based on exper-
ience and knowledge of sediment behavior, were incorporated in drawing the sedi-
mentgraphs between measured points. The observed sediment concentration points
were used a s guides. It may be possible, therefore, to considerably reduce the
scatter of points by adequately defining sedimentgraph loci for a given storm
event by more frequent sampling. Most of the sedimentgraphs considered herein
generally had from four to six, and at times as many as 10 observed sample
points defining th graphs; in many of the cases the USGS estimated the magni-
tude and location of the peak in its entirety. Consideration of scatter, at
least in this study, would suggest, that an attempt at explaining the variation
due to watershed soil types, vegetative cover, slope, etc., would be meaning-
less. This worker would, however, opine that the loci of well-defined
sedimentgraphs would lead to the development of series graphs prossessing
less scatter.
Exsmples of sedimentgraphs predicted on the basis of season and runoff
excess art. shown on Figure 6. Table I lists comparisons between predicted and
actual eroded sediment quantities in Bixler Run as shown in Figure 6. T o
illustrate the ;ange of applicability of the series graph method to Bixier Run,
variations in excess runoff for snowmelt or rainfall are included in Table I.
For the four storm events considered in Table I, the average error of estimate
for washload ranges from 16.1 to 16.5 percent as determined by the series graph
method and the ES versus ER graphical relationships, respectively. This is
based on comparisons with ac'tual conditions observed in the field. The errors
of estimate computed are all considerably below that reported for similar sus-
pended sediment load predictions, which may in some cases by greater than 100
percent.
TABLE I
Comparison of Predicted versus Computed Sedimentgraphs

Sediment Mobilized Percent Error

Date Excess (tons/sq. km.) Total Sediment
Runoff Source of Actual Predicted Bases (%)
centimetersf sq. Runoff BY BY BY BY
h. Series ES vs. Series ES vs.
Graph ER Graph ER
Method Curves Method Curves
10/19/68 .O35 W i n ter-Ra infa 11 19.6 27.9 27.7 29.6 29.1
03110/67 .343 Winter-Snowmelt 1523.0 1293.0 1330.0 17.8 12.6
10/04/62 .572 Winter-Rainfall 2710.0 2694.0 2555.0 0.50 5.7
05f07f 56 .O55 Summer-Rainfall 73.8 61.7 91.0 16.5 18.7
596

CONCLUS IONS
This study discloses two important findings for Bixler Run Watershed.
<I.That a relationship exists between excess runoff and the washload
that is mobilized by it over the watershed area.
b. And that the series graph can be used as a method to predict washload
and its variation with times.
Heretofore, it had been generally recognized that with the exception of
the universal soil-loss equation and sediment rating techniques a method did
not exist that could enable the prediction of washload in certain small water-
sheds. In adbition to this, there was n o method that could facilitate the
estimation of a sedimentgraph. It has been reported that predictions made
using the universal soil-loss equation or sediment rating techniques can result
in errors of estimate greater than 100 percent. Based on the analysis of hydro-
logic data for Bixler Run Watershed, the method presented in this study was
able to appreciably overcome the aforementioned limitation.
Granulometric measurements disclosed that practically all of the suspended
sediment in Bixler Run derives from the lateral inflow of washload from the
land surfaces; a negligible amount of suspended sediment is obtained from the
stream channel-bed.
Washload predictions for four storm events occurring during the winter
and summer months in Bixler Run averaged 16.3 percent error when compared to
actual results. The selected storms included rainfall a d snowmelt runoff
events.
A relationship (Figure 4) is presented between the total sediment mobilized
and excess runoff for single storm events during the period of record in Bixler
Run. Significmce tests revealed that the regression lines shown on Figure 4,
are highly signiijcant.
For Bixler Run Watershed the series graph and ES versus ER relationships
predicted washload with smaller erros than are generally reported in sediment
studies using existing methods.
REFERENCES CITED
1. Graf, W.H. (1971). Hydraulics of Sediment Transport, McGraw-Hill Book
Company, New York.
2. Shen, H.W. (1971). River Mechanics, Vol. i, Colorado State University,
Fort Collins, Colorado.
3. ,
Chow, V.T. (1964). - McGraw-Hill Book
Company, New York.
4. American Society of Civil Engineers, (1949). Hydrolorn Handbook, the
Committee on Hydrology of the Hydraulics Division, ASCE, Manual of Engin-
eering Practice, No. 28. Adopted January 17, 1949.
5. Wisler, C.O., and Brater, E.F., (1967). Hydrology, John Wiley and Sons,
Inc., New York.
6. Sherman, L.K. (1932). "Streamflow from Rainfall by Unit-Graph Method,"
Engineering News-Record, Vol. 108, p. 501-505, April 7, 1932.
7. United States Geologic Survey, "Station Analysis-Sediment (1955-1968 water
years), Susquehanna River Basin (1-5675), Bixler Run Near Loysviiie,
Pennsylvania," USGS, District Office, Harrisburg, Pennsylvania.
8. Nordin, C.F., and McQuivey, R.S. (1971). I'Suspended Load," River Mechanics,
Vol. 1, Chapter 12, Colorado State University, Fort Collins, Colorado.
9. United States Department of the Interior (1968). 'Water Resources Data for
Pennsylvania, Part 2, Water Quality Records," USGS.
597
598

; peak

HYDROGRAPH
1.50 \

1.25

1.00
I

1.75

?noon 6pm i2pm Gain 12nocn 6pm l2gin

6am 12noon
I_ -I.-.--I-

TIME IN HOURS

FIGURE 2 : TYPICAL STAGE HYDROGRAPH A J W SEDI?KNTGRAPW

(s-roiwiOF M A R C H 13,1963,BIXLER B U N WATERSHED )
5
599
600

3502
BIXLER R U N W A T E R S H E D

a
w
I-
w
I
O
4
Y
W
a
4
350.
u)
\
cn
z
a
a:
W
O
1
y.

-z
e
cn
hl
..

c3
W
N
4 35.0
m
O
z
I-
z
Id
2
I

a
Ill
o)

3.5 w

( ;o098 0.0098 0.098

EXCESS R U N O F F ( ER ) , IN CENTIRIETERS / SQUARE I(Il.Ol\r;ETER

FIGURE 4:SEGIMENT Mû01LjZEü \E:;) VERSUS EXCESS RIINOFF (ER)

 6O1

I I I I 1 1 1 1 1 I I I I I I I I I
- EIXLER R U N WATERSHED
- legend:
- p peak value.
- p+n
II I,

--summer
n hours before (-1 or after (el peak.
roinf al I.
- -- ______winterrainfall.
winter snowmelt.

0.72C I I I I 1 1 1 1 11 1 I I I 1 1 1 1 1
I
0:00098 0.0098 0.098
602

ò A v a p-6 p-4 P-2 P P+2 pt4 pt6

TIME IN HOURS

FIGURE 6 : PREDICTED GRAPH OF SEDIMENT DISCHARGE VERSUS

TI M E
 METHODES UTILISEES pour 1'EVALUATION des DEBITS de CRUE
des PETITS COURS d'EAU en REGIONS TROPICALES

par J. A. RODIER

ABSTRACT

For most o f the tropical small streams studied by the author,

the floods result from surface runoff, field of application o f unit
graphs. Hydrometric networks are useless for the floods of these basins
(less than 500 km2). Two methods are described:

For area without cyclonic precipitations: the depth of the

storm of the frequency choosed for the project is computed assuming the
other characteristics equal to the more frequent values for the big
storms. The transformation of rainfall into discharge is made in two
steps: computation of runoff coefficient and flood volume, computation
o f characteristics of the hydrograph. This incorrect method gives good
results if used with judgement. Empirical graphs and rules have been
deduced from systematical researches on representative basins for c o m p ~
tation of the elements of the flood from physiographical data. A gene-
ral synthesis will permit a better characterization of the basins.

In area with cyclones: the precipitation depth are estimated

from the observations and high values of runoff coefficient are choosed
i n rel&tion with observations o r envelope curves are drawn from obser-
ved data 5.n the world.

RESUME

Pour l a plupart des petits cours d'eau tropicaux étudiés par

l'auteur, les crues résultent du ruissellement superficiel, domaine
d'application de l'hydrogramme unitaire.

Les réseaux hydrométri ues sont sans utilité pour les crues
9
de ces bassins (moins de 500 km ), Deux méthodes sont décrites:

Pour les régions non affectées par les cyclones: on détermine

l'averse de fréquence égale à celle de la crue du projet, les autres
caractéristiques étant les plus fréquentes pour les tres fortes aver-
ses. La transformation en débit est faite en deux temps: calcul du c o e
fficient de ruissellement et du volume de crue, calcul des caractéris-
tiques de l'hydrogramme. Cette méthode non rigoureuse fournit de bons
résultats si elle est employée avec discernement. Des diagrammes ou
des règles empiriques sont déduits de recherches systématiqyes sur b a c
sins représentatifs, pour calculer les éléments de la crue a partir
des données physiographiques. Une synthèse générale permettra de mieux
caractériser les bassins.

Dans les régions de cyclones: on détermine les averses d'après

les valeurs observées et on suppose des coefficients d e ruissellement
tres élevés en rapport avec ces observations, ou on établit directement
les courbes enveloppes a partir des données observées dans le monde.

Chef du Service Hydrologique de l'office de la Recherche Scientifique

et Technique Outre-Mer
Conseiller Scientifique à Electricité de France (DAFECOI.
 604

L'étude des ouvrages utilisant les eaux des petites rivières

tropicales ou méditerranéennes présente de très sérieuses difficultés
dès que l o o n aborde la dbtermination des conditions hydrologiques de
réalisation et d'exploitation des ouvrages, en particulier celle des
d6bits moyens annuels et surtout celle des ddbits de crues.

Les donnhes sur le régime hydrologique sont dans ce cas

inexistantes : la densit& des réseaux hydronktriques est faible, l e
nombre de stations amdnagées est nul ou dérisoire. LEh outre, les varia-
tions temporelles des débits sont si rapides que les donn&es de ces sta-
tions sont souvent difficiles à exploiter. Enfin, contrairement & ce qui
a lieu pour les grandes rivières, les crues de faible frequente ne lais-
sent aucun souvenir dans la mémoire des habitants les plus proches.

I1 est très sou-ntinipsible de procéder 2 une étude hydrologi-

que sérieuse sur le terrain pour un seul ouvrage car elle durerait long-
temps et son prix atteindrait ou dépasserait même celui de l'ouvrage lui-
même

Tout ce que l'on peut faire c'est organiser une telle étude à
l'occasion de l a r8alisation d'une série importante de tels ouvrages,
par exempie pour la construction de tous les ponts d'une longue voie
ferrée (chemin de €er transcamerounais), ou d'un grand axe routier, ou
lorsqu'o, ariiinage i l a fois 30 ou 50 petits barrages comme cela a 6t6 le
cas en I!AUî.Q-VOLTA il y a quelques années.

Autrement, on est conduit & utiliser les résultats de synthèses

& caractère g6ogr;pliique.

Nos hydrologues ont souvent rencontré ce problème en Afrique

Tropicale, en Am6rique d u Sud, dans les fles d u Pacifique et de l'Océan
Indien et ils ont mis au point différentes rnbthodes pour la détermination
des débits moyens annuels et des dLbits de crue, premiers 6léments que le:
ingénieurs demandent aux hydrologues.

Dans ce qui suit, nous ne traiterons que le probleme de l a dé-

termination des d&its de crue dans le cas de cours d'eau dont l e bassin
versant couvre une superficie inférieure 200 km2 et plus souvent infé-
rieure à 50 km2. Au-delà de ces surfaces, les m6thodes n e sont plus les
memes. Ellos correspondent souvent en effet l a limite d'emploi de l'hy-
drogramme unitaire et des modèles globaux.

Pour ces petits bassins, on considérera deux cas différents sui-

vant la genbse des crues exceptionnelles. Dans le premier cas, elles sont
dues a des orages convectifs avec prhcipitations intenses mais d'assez
courte durbej dans le second cas, il s'agit de précipitations cycloni-
ques ?ì iiitensité plus faible mais de plus longue durde, les derniers 616-
ments de l'hpisode pluvieux arrivant sur un sol pratiquement saturé.
 605

1. Cas de crues provoquées par des orages convectifs,

La mise au point de méthodes pratiques nous a demandé quinze

ans de recherches fondamentales. Nous passerons rapidement sur ces
recherches pour insister plus particulièrement sur la m6thodologie
proposGe aux ing6nieurs, cette méthodologie n'étant guère applicable
que pour des p6riodes de retour de 10 ou 20 ans. L'idée de base est
IQutilisatian de l'information pluviomhtrique existante et plus parti-
culièrement des sc'ries chronologiques de précipitations journalières
et la transformation des hauteurs de prGcipitations en débits de ruis-
sellemelit superficiel. Pour des averses de ce type et d'assez faible
frcquerice, en réginns tropicales et méditerranéennes, il se produit
g(!nbralernent du ruissellement superficiel, ce qui permet l'emploi de
la m6thode de l'hydrogramme unitaire.

1.1. Recherches fondamentales entreprises.

Les plus importantes ont ét6 les suivantes :

1.l.l.Etudes g&n&rales statistiques des pluies journalières. En

Afrique Occidentale OU elles ont été le plus poussées, elles ont port6
sur 1 O00 stations environ. L'étude simultanée pour un grand nombre de
statioils a conduit a des valeurs assez sûres des paramètres des lois de
distribuLion pour des p6riodes de retour de 10 à 20 ans. Eh particulier,
elle u cona.iit ?i abandonner, pour cette région du monde, la distribution
de GALTOW tro2 pessimiste,pour une distribution de PEARSON III. Sur le
plan pratique, on en a déduit une série acceptable de précipitations
journali&res de période de retour 10 ans ou 20 ans.

1.1.2.Etude de l'abattement (Inverse du rapport pour une méme frdquen-

ce, eiitre la hauteur de précipitations en un point et la hauteur de prcci-
pitations sur une surface donnée entourant ce point). Les études menées à
partir de données recueillies sur bassins représentatifs ont conduit a
des ordres de grandeur acceptables pour la pratique.

1.1.3.Etudes des courbes intensité-durée : ces études faites surtout

à partir des pluviographes des bassins représentatifs ont permis, pour
l'Afrique Occidentale, de donner des courbes-types.

1.1.4.Etude des relations pluies-débits : celles-ci ont été etudibes

averse par averse pendant plusieurs années sur une centaine de bassins
reprGsentatifs, qui ont également 6té utilisés pour les recherches vi-
sées aux points 1.1.2 et 1.1.3. La méthode des r6sidus a permis de dé-
teniiiner dans chaque cas la hauteur d'eau bcoulée HR ou le rapport KR
entre HH e t la hauteur de précipitation P eii fonction de P, des condi-
tions d'humidité du sol avant l'averse et de la durée de l'averse.

1.1.5.Etude de la forme des hydrogrammes. Sur les mgmes bassins re-

prdseiitatifs, on a pu appliquer la mgthode des liydrograrimes unitaires
et dbterminer la forme des hydrogranines-types. On en a retenu trois
 606

éléments caractéristiques : le temps de montée tm , la durée de ruis-

sellement tg et le rapport k entre le débit de pointe de l'hydrogramme
unitaire et le débit moyen pendant la durée du ruissemment.

1.2. bisthode de détermination des débits de pointes de crue et de leur

volume.

Le cas des fréquences décennales ou de fréquences voisines

est assez différent de celui de la crue maximale probable. Dans ce qui
suit nous traiterons le cas des crues de périodes de retour de 10 ans
ou 20 ans.

De façon générale, on a cherché à mettre au point des méthodes

simples qui puissent &tre utilisées sans ordinateur. Ces méthodes sont
probablement très différentes de celles qui sont élaborées actuellement
et qui seront vulgarisées dans quelques années, mais de nombreux pays en
voie de développement ne disposent pas, à l'heure présente, de moyens de
calculs suffisants et n'ont pas assez de personnel bien entraîné pour les
utiliser pour des fins hydrologiques.

C'est pourquoi, dans ce qui suit, on adoptera les principes sui-

vants, dont certains sont discutables, mais qui permettent aux ingénieurs
d'arriver à des résultats utilisables avec les moyens dont ils disposent.

1.2.1.Principes du calcul : Le point de départ est la série d'obser-

vations de précipitations journalières au poste le plus proche de l'ou-
vrage que l'on a & étudier ou un poste pluviométrique correspondant aux
mdmes conditions pluviométriques si la qualit6 des données du pluviomè-
tre le plus proche est insuffisante.

Des bassins de 50 km2 sont généralement assez homogènes, mais,

dans le cas de forte différence d'altitude, le poste pluviométrique choi-
si devra se trouver à peu près à l'altitude moyenne du bassin et non pas
au niveau de l'esutoire, ce qui rend le choix beaucoup plus difficile.
On étudie la distribution statistique des précipitations journalières ce
qui, en région tropicale, correspond à peu près à la distribution des
averses orageuses et on détermine l'averse correspondant ?I la fréquence
de la crue (période de retour 10 ans, 15 ans, 20 ans, etc...). On recher-
che, en Ltudiant lee 'enregistrements disponibles, quel est le schéma lo
plus courant des répartitions des intensités pour une averse donnée, on
examine également quelles sont les conditions moyennes d'humidité préa-
lables que rencontrent généralement les fortes crues. Enfin, on trans-
forme la hauteur de précipitation en un point par la hauteur de préci-
pitation moyenne sur une surface en la multipliant par un coefficient
d'abattement inférieur à 1.

Au moyen du modele de transformation des pluies en débits, on

transforme la pluie décennale en crue décennale en veillant bien à ce
que la distribution des intensités de l'averse, l'index représentant
l'humidité préalable, le mois de l'année lorsque celui-ci intervient,
correspondent aux conditions les plus fréquentes pour les fortes préci-
pitations. Sur le plan statistique, ceci est très contestable : la
 607

v8ritable solution consisterait 5 appliquer le modèle de transformation

pluie/débit à la totalité des averses observées au poste de référence,
sur 40 ans par exemple, et à étudier la distribution statistique de
l'échantillon de crues reconstituées sur 40 ans. Mais cette méthode
serait peu réaliste pour beaucoup de pays en voie de développement
parce qu'il est beaucoup plus difficile de mettre au point un modèle
valable pour toutes les averses qu'un modèle uniquement valable pour
les fortes averses et parce qu'ensuite la reconstitution des crues de
petits bassins pour 40 ans ne peut se faire qu'avec l'ordinateur.

La transformation pluie/débit se fait en deux temps :

lo - calcul du volume de crue par la détermination du facteur

1 .I .4.) ;
KR (voir

2* -à partir de ce volume, détermination du débit de pointe par la for-

me de l'hydrogramme (voir 1 .1 .5.).

Autant que possible, on a cherché à ramener ce calcul à des

opérations très simples dans un certain nombre de pays où le nombre de
bassins représentatifs était suffisant.

1.2.2.Pratique du calcul pour l'Afrique Occidentale :

Le5 études systématiques visées en 1.1.1. et 1.1.3. fournis-
sent des éléments pluviométriques permettant de déterminer l'averse de
fréquence cherchée avec son diagramme de distribution temporelle, pour
la majeure partie de l'Afrique Occidentale. Des études beaucoup plus
partielles effectuées dans d'autres r6gions du monde ont fourni les me-
mes données.

On réduit ces valeurs ponctuelles à des valeurs moyennes sur

une surface donnée en les multipliant par un facteur qui décroît de 1
pour une s v f a c e S inférieure à 25 km2, à 0,8 pour une surface comprise
entre 150 et 200 km2. Ces chiffres qui ne sont valables que pour les
orages convectifs des régions tropicales africaines, sont peut-$tre un
peu forts. Ils seront probablement diminués à la suite de recherches en
cours.

On dispose donc de la hauteur de précipitation .

,
P

Pour déterminer la valeur de Ks, on a établi des séries d'aba-

ques pour deux types de couvertures végetales naturelles (liées au cli-
mat), savane et savane boisée d'une part, steppe et savane à épineux
d'autre part. I1 n'a pas encore été possible d'établir d'abaques conve-
nables pour la forat tropicale.
 608

Les autres facteurs pris en considération pour la détermina-

tion de KH sont : la superficie du bassin, la perméabilité globale
du sol P et la pente R. A défaut d'index quantitatif pour R et surtout
P, on a établi deux classifications : R correspond à des plaines très
plates, RG à des pentes de montagne (pentes longitudinales supbrieures
à 5 $, pentes transversales supérieures à 20 $)o
Pl correspond à un sol rigoureusement imperméable, P5 à un
s o l très perméable (sable ou carapace latéritique très disloquée. Le
graphique 1 d m e un exemple de ces abaques pour des sols imperméables
(Pl - P2) et des pentes variables de R2 à R4. Ces abaques ont été &ta-
blies & partir des données des bassins représentatifs. Les valeurs de
KR correspondent des pluies de fréquence décennale (Pm vaciant de
&o à IO5 mm) dans ces régions, tombant dans des conditions d'humidité
du milieu de la saison des pluies.

Bien entendu, au cas OU des facteurs secondaires tels que le

réseau hydrographique, présenteraient des caractéristiques anormales,
par exemple lit marécageux, on devrait rectifier les valeurs de KR en
conséquence.

Le volume de ruissellement de la crue :

A ce volume il convient d'ajouter l e volume correspondant au

d6bit de base do1.t on peut avoir une idée sur le terrain, sans &tude
hydroiagique très difficile.

Pour la forme de l'hydrogramme,des abaques ont été également

mis au point;. On en trouvera un exemple au graphique 2 qui donne le
temps de base ou duróe du ruissellement en fonction de la surface du bas
sin et de l'index de pente pour les m8mes conditions de végétation que
le graphique n o 1.

La connaissance du temps de base TB permet de calculer le dé-

bit moyen de ruissellement :
'ruis seliemen t
M e

TB
M est obtenu en m 3/s.

Pour trouver le débit de pointe s, on utilise un coefficient

K (K P F) étudibi pour les mêmes rbgions sur bassins représentatifs.
 609

Pour la couverture végtitaïe steppe ou savane à épineux avec

des valeurs de KR pas trop &levées, on trouve des valeurs de K variant

supérieures à 50 -
entre 2,5 pour 25 km2 à 3,t pour 100 kmz. Si ces valeurs de KR sont
60 $ K varie entre 3 pour 2 km2 et 4,5 pour 50 km2.
On obtient QM en multipliant $1 par K et on ajoute le débit
de base.

Bien entendu, si le diagramme de répartition temporelle des

intensités et si la superficie du bassin sont tels que la crue n'est pas
unitaire, il existe des abaques complémentaires donnant le temps de base.

Dans ce qui précède, c'est volontairement que nous n'avons pas

Gtabli de formules pour repr6senter les courbes des graphiques 1 et 2
auxquelles nous voulons garder un caractère provisoire.

l.Z.3.Limitations de la méthode :

perficies inférieures & 50 -

Elle ne s'applique bien en Afrique tropicale que pour des su-
100 km2.
Comme nous venons de le dire, nos courbes sont provisoires et
on met ali point des modèles plus &labor& pour revoir les bases de nos
abaques quL nécessitent encore un sérieux effort d'homogénéisation des
données et des proc6dés de calculs.

Les problèmes de forêt tropicale exigent encore un effort im-

portant de recherches sur le terrairi.

%fin, il n'est pas très facile de classifier un bassin en caté-

gorie P2 ou P .Des recherches de physique du sol sont en cours pour arri-
ver ?i des règ3es simples permettant de le faire. C'est certainement 1&
le point le plus difficile.

Pour définir quantitativement des index R une bonne combinai-

son des facteurs géomorphologiques courants doit donner satisfaction.
Actuellement, cette méthode est très souvent employée en Afrique,
mais, dans bien des cas delicats, il serait plus prudent que les bassins
soient examinés auparavant par un hydrologue confirmé. Elle présente l'im-
mense avantage d'éviter toute véritable étude hydrologique sur le terrain.

1.3. Crue de période de retour supérieure à 20 ans.

C'est là un problème très difficile car la documentation plu-
viométrique est tout à fait insuffisante. Pour des périodes de retour
de l'ordre de 100 ans, une minutieuse étude critique des relevés de nom-
breux postes pluviométriques permet d'aboutir à un ordre de grandeur.
 61 O

En Afrique tropicale, les averses journalières centenaires

de caractère convectif sont peut-8tre de l'ordre de 200 ?i 3-400 mm en
24 heures, suivant les régions.
I1 reste ensuite à choisir une valeur de KR qui n'est plus
celle des abaques mais qui doit en tenir compte, car tous les bassins
pour de telles averses ne parviennent pas à la limite de O,&5 - O,9O.
Enfin, généralement, l'averse dure au moins 5 ou 6 heures et parfois
20 heures, elle n'est donc plus unitaire. On utilise donc les abaques
tels que ceux du graphique 2 pour établir les différents hydrogrammes
élehentaires qu'on ajoute après avoir découpé l'averse centenaire.

Enfin, s'il s'agit de la crue maximale probable, il ne reste

plus qu'8 appliquer la formule de FIERSHFIELD OU l'on ajoute à la valeur
moyenne de la précipitation journalière maximale annuelle 15 fois 1°é-
cart-type de la distribution de cette précipitation maximale. Mais il
faut d'abord partir d'une série de précipitations journalières de quali-
té suffisante pour en déduire une valeur correcte de l'écart-type.
D'autre part, si cette formule paraft excellente pour l'Afrique du Nord,
les régions soumises à des cyclones tropicaux, elle semble conduire à
des chiffres trop élevés pour les orages convectifs d'Afrique tropicale.
Bans ce cas également on revient à l'application de la méthode de l'hydro-
gramme unitaire pour des averses élémentaires successives, mais le choix
de la distribution temporelle des intensités est délicat. C e q u i arrive
souvent clest que d'un bout à l'autre de l'estimation, on arrive à de
telles cascadea de marges de sécurité qu'il est facile de fournir des
chiffres trop élovés.

2. Crues dues à des averses cycloniques :

2.1. Crues décennales :
L'averse décennale est plus difficile & définir que dans le
cas précédent, la distribution statistique est plus difficile à étudier
et les donnkes de base sont plus mauvaises (en cas de cyclone une bonne
partie des pluviomètres débordent), mais dans beaucoup de pays du monde,
on arrive 8 d6finir une valeur à peu près convenable de l'averse décen-
nale, on doit alors découper l'averse en averses élémentaires comme au
point 1.3. et on transforme ces averses en crues par la méthode des hy-
drogrammes unitaires. Très souvent, pour les dernières averses élémen-
taires rC, est voisin de 0,9O si la pente est notable, que la couverture
soit forestière ou non. Après calcul, il est bon de comparer le résultat
aux crues maximales connues dans le monde en utilisant des diagrammes
tels que le diagramme FWCOU-RODIER.
 611

2.2. Crue maximale probable.

Dans ce cas, on ne peut donner que des indications gén6rales.

Pour l'averse à prendre en considération, on pourra se référer, dans
les pays à fortes averses, aux valeurs maximales mondiales telles qu'el-
les sont données dans le Guide des Pratiques Hydrométéorologiques de
l'OMM ou aux résultats de la formule de HERSHFIELD, mais il sera encore
plus difficile que plus haut d'aboutir à une valeur convenable de l'écart-
type, ceci nécessitera une sérieuse étude critique des rares données
pluviométriques disponibles, en tenant compte du débordement 8ventuel
des pluviomètres. Le reste est plus facile car le coefficient de ruis-
sellement KR est de l'ordre de 0,gO.

Si la région est C O M U B pour avoir des averses exceptionnelle-

ment fortes, il est normal de prendre en considération des valeurs de
précipitations supérieures aux maximums mondiaux connus car, en pays de
cyclones tropicaux, la connaissance des averses de durée inférieure à
48 heures est très incomplète et les maximums mondiaux connus doivent
&tre considér6s comme piut8t provisoires.

ïb générai, dans les cas graves concernant les cyclones tropi-

caux, l'hydrologue arrive a la conclusion un peu décevante qu'il serait
prbférable que l'ingénieur prévoie son barrage de telle façon qu'il puis-
se être submergé par n'importe quelle crue.

Qua1 que soit le cas étuùié, un examen du terrain orienté vers

la recherche L;ss traces laissées par de fortes crues est n&cessaire.

I1 résulte de tout ce qui précède que les ñydrologues ont

encore de nombreuses recherches à faire pour aider efficacement les
constructeurs dans leur tâche.

Quelques références utiles pour les petits bassins de ces régions :

(1965). Guide des Pratiques Hydromét&orologiques, no 168

1. 0.M.E.i.
T.P. 82, Genève.
2. RODIER J., A W R G Y C. (1965). Estimation des débits de crues décen-
nales pour les bassins versants de superficie inférieure à 200 km2
en Afrique Occidentale, ORSTOM, Paris.
3. HERSHFIXLD D.M. (1963). Estimating the probable maximum precipita-
tion. Am. Soc. of Civil hhgineers Transactions, Vol. 128, Part I,
PP. 534-556.
 61 2

4. FRANCOU J., R O D B R J O (1967), Essai de classification des crues

maximales observées dans l e monde. Cahiers d'Hydrologie OHSTOM
vol. IV, n e 3, pp. 19-46. Paris.

5. BENSON M.A., (1968), Measurement of Peak Discharge by Indirect

Methods, OMM ne 225. TP. 1 1 g P Genève.
 613
N
E
Y
C
a
-O
-N
-
-0
I O
-0
-0
OI
W
-0
.o
h
O
-a
O
-YI
I O
-*
O
-m
-0
N
-0
-YI
-*
-m
-N
I
O O
I i ;
C
h YI - c
. -4.C
.*QL)c
-
2 5-

u)

!?a
Q
C
2o-
C
Q

v)
P
-f

+
15-

I o- 'Ø

'
4r

Ø
5- 'Ø
0

O- I

I 2 3 4 5 6 7 IO 20 30 40 50 60 7080 loo 2c
S en k m 2

Fig: 2
T e m p s d e b a s e e n f o n c t i o n d e R et de S
REGIMES SAHELIENS - SUBDESERTIQUES
 METHODS FOR THE ESTIMATION OF MAXIMUM DISCHARGES O F SNOW
MELT AND RAINFALL WATER WITH INADEQUATE OBSERVATIONAL DATA

Prof. A.A. Sokolov

State Hydrological Institute
Leningrad, USSR

ABS TRACT

The problem o f floods computation was accepted by the UNESCO

Co-ordinating Council for the IHD a s one of the most important
problems. For its solution the Working Group on Floods and their
ccmputation was established; it has realized several projects on
the THD programme essential for future research on floods, deve-
lopmen: and improvement o f methods for floods computation. The p c
per give;: an evaluation of the up-to-date state o f this problem
as means f L r its solution.

RESUME

Le Conseil de Coordination de l'UNESCO pour la DHI a estimé

que le problème du calcul des crues était tres important. Pour
l'examiner, i l a créé un groupe de travail sur les crues et leur
évaluation. Ce groupe a réalisé, dans le cadre du programme de la
DHI, un certain nombre de travaux très importants pour l'avenir
de l a recherche sur les crues, pour la mise en oeuvre et l'amélig
ration des méthodes de calcul qui le concernent. Dans la présente
communication, l'auteur fait le point de l a situation actuelle,
ainsi que sur les moyens de parvenir la solution du probleme.
 61 6

At present triere are two uifferent approac,ies to t.ie complita-

tlon of fiood discilarge of ungauged rivers cievelopec to a certain
dcbree quite independently. The first approach is based o n the
s~atisticalanalysis and generalization of field data on
flood runoff; the second approach is based on the genetic
analysis and synthesis of flood hy&ograph. As was: ctateu
by J. Nemec and M. Moudry (Czechoslovakia) at; the Leningrad
Symposium (1967) these two approaches ase gradually being
brought together and at; present they are used simultaneously
supplementing each other Lis,-/.
Humerous design schemes have been proposed to estimate maxi-
mum flqod runoff of ungauged and poorly gauged rivers.
According to the principles of approach and the scope of
flood computation these schemes may be divided into 2 main
groups:
3. Empirical or semi-empirical formulae for flood discharge
computation based on the account of some of its most important
factors (e.g., drainage area or maximum precipitation rate),
providing maximum water cìischarge only.
2. Methods considering flood genesis and providing the
possibility of plotting the whole hydrograph on the basis of
time inflow of snow melt and rainfall water and its trans-
formation into runoff as a result of losses by infiltration,
suriace retention, lag along the slopes and channel network.
A t k-esent, methods of maximum flood discharge computation
on the basis of the use of di€fereat design formulae are
widel applied. The most important formulae are as follows:
(a? formulae of extreme intensity, or the so-called
rational formulae, basea on the account of maximum or extreme
rainfall intensity during lag-time or flood flov concentration
i n its general form:

Qmax =/Cpa,d,A. (1)

where :g p is coefficient of dimensionality; ar is maximum
intensity of rain or snow melt during lag time ( c); d r i s
runoff coe ficient during this time interval; A is drainage
area in km 5.
Pormula (1) is usually applied for the computa-t;ionof
maximum runoff for relatively small basins (less than 200 km ) .
To determine design values of' dc curves of maximum precipita-
tion increase %, are plotted with the increase of time
interval Z' ? given as percentage of daily precipitation of the
same probability of exceedence (p )

The maximum mean precipitation rate Jcp for any time

interval is estimated by formula:
 617

Mean velocity and lag-time down the channel and slope6

are computed by simplified formulae of Chezy-Manning or
C he zy -Bazin.
The principal disadvantage of formula (1) consists in some
uncertainw and inaccuracy of the determination of lag time
c-
or flood concentration c, is interpreted by iiiJividiia1
scientists in different ways, therefore it causes iii-
accuracy of determination of principal parameters az and d,-
appearing in formula (1). Besides, formulae of type (1) do not
take into account the remaining flood elements (duration rise
and fall duration ratio) and do not provide the plotting of
%he vihole ïlood hydrograph essential for the determination of
maxima transformation in ponds and reservoirs;
(b) empirical or semi-empirical reduction formulae of the
general type:

where: Q,,,~
is maximum specific discharge, m'/sec per 1 k
m';
if
9.
A-O
is parameter comprising the extreme specific discharge
and C =1.O;cis addition to the drainage area
considering non-lineariky of dependence 4
~mp,aj@@~
within the range of small areas of the basin; n i s the
eqonent of reduction of maximum specific discharge3 with %he
and bheoretical data from n= 0.15 -
incrsase of basin area and varying according to experimental
0.30 for runoff maxima of
snow mel? water or caused by prolonged frontal rainfalls, to
=
n 0.5 -
3.7 for maxima caused by short heavy local storms.
Parameter +,,may be estimated according to the extreme rate of'
snow melt or rainfalls for minimum time interval, e.g. 1 hour,
or for snow melt water according to depth of runoff during a
flood.
In the first case when C = 1.0 formula (4) may be presented
as fo1lov;s:

where: %is coefficient of dimensionality; % d s maximum

hourly rato of snow melt or rainfall; d, is overland flow
coefficient
Since
.
does not exceed 10-15 m/hr for snow melt water
and 300-400 mm/hr for rainfall water (on the basis of compu-
tation of heat balance of snow melt), then the extreme value of
qe in fo mula (i+) if d o = 1.0 and k$= 0.23, may not exceed
3
2.8 -4.2 m /sec pe 1 lun2 for snow melt water and up to 84-
112 mj/sec per i bS for raidail water.
On the basis of these elementary considerations it is possible
to conclude %hat many empirical formulae of type (4) with
parameter 9. exceeding the mentioned limits have no physical
substantiation.
Due to some uncertainty of C value in formula (4) when
A -0, this Lormula is sometimes used as follows:
61 8

w is parameter (maxim specific discharge in

ì$/sec
Cp.6
ei'3:
if &ainage area B=2ûû by:
a is tiie exponent of reduction dcterrnined uy regional
dependences ep =fy/A/.
For practical computations on the basis of generalization
of empirical data a map of regional boundaries is prepared
with similar exponents of I2 .
The advantage of formula (6) consists of the fact that the
value of parameter c,, slightly depends on and
therefore the mapping of
possible e
cp,6 in the form of isolines is
In case of e o determination according to the depth of runoff
of snow melt water during flood hrnformula (4) may be present-
ed as follows:

where: & is t,ie coe€ficient considering a number cif otlier fac-

tors, in particular, duration and siiape of flood.
Formula (7) is used as the basis for the computation of
maxinum snow melt water äischarge for the whole USSII territory
L4L
Duriw recent years the reduct'on scheme is seldom used as
simple regional dependences =*[JI plotted by dependences
enveloping empirical points usually related to larger basin
areas.
In this case it is essential to take into account flood
runoff probability of oxceedence Its parameters are
differentiated according to climatic zones.
Along with the basin area which was previously accepted as
almost the only maximum runoff factor, numerous important
clinlatic factors are also taken into account, i.e. depth and
sate OP precipitation, snow melt rate, flood runoff depth;
and morphological factors as well, i.e. basin topography, lakes
and swamps areas, river network density, soils and subsoils
composing the basin mantle, etc.
7iith the increase of hydrological information and reliable
runoff data from small basins the reduction scheme has greatly
consolidated its positions since ibs basic parameters i.e.
reduction coefficient anCr maximum runoff of elementary (small)
basins have gainea a reliable substantiation. This particular0
concerns maximum runoff of snow melt water;
(c) the so-called volumetric formulae considering flood
shape and duration, besides maximum ordinate of flood, may be
given as follows:
 61 9

where: H is depth of precipitation or snow melt water;

& i s total or volumetric coefficient of runoff during flood;
Ttn ndicate general duration of flood or its rise phase;
4 and$ , are coefficients of flood shape, i.e. ratio of maximum
discharge to mean discharge.
Design formulae as (8) or (9) are preferable compared with
formulae of type (1) or (4) since all flood elements are co-or-
dinated but they ara applied only for simple one-peaked
floods for which general or volumetric coefficient of' runoff
is applicable.
Despite the existing numerous formulae the computation of
rainfall flood runoff for ungauged rivers is not reliable.
Every design scheme proposed is characterized by certain a6-
vantages and disadvantages. There exist no accepted design
schemes until now.
A general disadvantage of the majority of empirical design
formulae is in the cornputation of individual flood elements
without their co-ordination and without the account of genesis
anrid type of flood; the latter is essential for hydraulic engineer-
ing projects to make a correct estimation of the flood trans-
formation rate in ponds and reservoirs and the amount of
discharte through spillways. These disadvantages never occur
in genetx computation methods for floods of the 2nd group
based on the account of time variations of inflow of rain and
snow melt waters and their transformation into runoff hydrograph
as a result of non-simultaneous water lag from different basin
areas.
These methods are based on the plotting of runoff transit
curve showing the distribution of areas of simultaneous runoff
over time intervals.
The following methods may be mentioned for the computation
of floods:
(a) isochrone method based on the determination of ordinates
of the curve of unit areas distribution (transit curve) by
means of plotting the lines of equal transit (isochrones) on a
topographic map o€ river basin;
(b) unit hydrograph method, based on the determination of
transit curve by the ordinates of the observed unit flood
hydro .raphs caused by individual storms;
(cy method of mathematical floods simulation.
The method of isochrones is mainly applicable for floods
computation on small water courses with surface flow prevailing.
&/
Nhen the method of isochrones is applied it is essential to
use topographic map of the basin of sufficiently large scale
620

anu tile data on time variations of rain or snow melt water

Inflow, moreover, for small water courses it is necessary to
have data on such variations within a day and this causes certain
restrictions in the sphere of its application.
Unit hydragraph method is based, as it was mentioned, on the
plotting of curve of unit areas distribution according to the
ordinates of unit floods observed.
Unit hydrograph method is based on the lag theory expressed
by the so-called genetic formula of runoff:

where: Qdt indicates discharges at the outlet at the moment t ;

ht-r is effe tive precipitation per time unit A f at the
moment k-r; #c indicates ordinates of the curve of unit
areas distribution.
Tne unit ,iydrograplimetiiod is cnaracterized by its visuality
and.pa.ysica1 substantiation, it provides the plotting of design
flood nyúrograpii according to preciFitation; tiiis resulted in its
wide application in many countries of tile world despite some
draNoacks.
Its sppircasion is aiII1cuII; mainly ow- ‘GO m e inadequacy
of the mekhods used for averaging the observed individual floods,
separation cf multi-peaked floods and methods for infiltration
rate and flow coefficient determination, There is also some
uncertainty as to the applicabili- of the unit hydrograph method
to different drainage areas; also problematic is the relation-
ship between an individual storm duration and flood rise duration
or the time of peak shifting relative to rainfall maximum during
the flood.
All these factors limit the application of the unit hydrograph
method.
Lately the method of mathematical floods simulation has been
more and more widely used.
For instance, in one of the variants of trie method of mathe-
matical floods simulation applied in the USSR the analogy bet-
ween equation (10 ), describing flood formation resulting from
water lag and summation of individual discharges from different
parts of river basin, and the eq,uation describing the change of
current in the electric circuit, was used,
To apply this method practically, it is essential to develop
investigations connected with determining parameters set for the
specific electric analog computer to estimate flood runoff of
ungauged rivers.
The importance of research, computation and prediction of
floods for many countries of the world necessitates the inter-
national scientific co-operation on the problem of flood flow
computation.
 621

This co-operation is in particular exercised under the auspices

of UNESCO and WMO within the framework of the IIID programme.
For this purpose the Co-ordinating Council €or the IHD
established the Ciorking grou:, on floods and their compuation.
This Working group has studied and generalized, to some extent,
the international experience in the field of research and
computation of flood flow.
A great contribution was made by the International Symposium
on floods and their computation held in Leningrad at the ini-
tiative of UNESCO and with the participation of \YMO and IAkIS.
The proceedings of this Symposium were published, therefore
there is no need to cite and consider them here /18, 26/.
The review made by WMO on meteorological aspects in computing
flood flow is rather useful. ‘ïiiic review made up a technical
note on this problem /25/.
The results of processing of observation data on floods made
at the network of I H D stations also appear to be valuable. Ta-
king into account that in some large areas covered by the net-
work of IIID stations outstanding floods, may always occur. The
UNESCO IHD Working group on floods and their computation prepa-
red and published the technical note on collection and proces-
sing of data on floods / 2 7 / .
The Working group also developed the programme for the
World Catalogue of very large floods, according to which a
riiimber of countries were entrusted to collect, process and
publish the data on large floods.
Besides, the Working group considered it necessary to study
and generalize the international experience in the field of
floods covputation and is now preparing for publication the
Technical Note (Casebook), wiiich would reflect a great experience
in computation of flood flow, gained in many countries.
These pub lications together with the Proceedings of the Lenin
grad Symposium on floods will serve as a good basis to improve
methods of flood flow computation, which in its turn will
contribute to a more rational and economical selection of
parameters for hydraulic structures on rivers, their durability
and resistance to floods.

R E F E R E N C E S
Alexeev G.A. (1 966). Sklema raschetov maximainyh dozhdevyh
raskhodov vody PO formule predelnoy intensivnosti stoka,
(Computation of maximum rainfall discharge by means of
the formula of extreme runoff intensity). Transactions
of the GGI, vol. 134.
Befani A.N. (1958) Osnovy teorii protsessov stoka i puti
dalneishykh issledovaniy. (Theory of runoff processes and
directions of further research). Transactions of OGMI,
vol. 15.
 622

3. Velikanov M.A. (1931) GidromekhanichesQ analiz poverhnostno-

bo stoka. (Hyaromechanical analysis of the surface runoff).
Geophysics Nos. 1-2.
4. Voskresenski K.E. (1956) Gidrologicheskie raschety pri pro-
ektirovmii sooruzheniy na malyh rekah, ruchjah i vremennyk
vodotokakh (Method osn. i prakt) (Hydrological computa-
tions for structures on small rivers and temporary water
courses), Leningrad, GIMIZ.
5. Kalinin G.P., Milukov ?.I. (1958) Priblizhennyi raschet neus-
.
tanovivshegosya dvizhenia v o w h mass (Approximate
estimation of the unsteady water motion), Transactions of
T s W , vol. 66.
6. Kovzel A.G. (1951) Opyt projektirovania hydrografa vesennego
stoka dlya malogo vodosbora. (The designing of the hydro-
graph of spring runoff on small watersheds) .Tr.of GGI,v.31(85)
7. Kuzmin P.E. (1961) Protsess tayania snezhnogo pokrova.
(Snow cover melting) Hydrometeorological Publishing House.
8. Lvovich M.I. (L34-ö) Protsessy formirovania pavodkov.(Flood
formation), Transactions of GGI, vol. 10.
9. Moklyak V.I. (1965) Pormirovanie maximalnyh raskhodov ot
talyh vod i ih raskhoày (Formation of maximum snow-
melt discharges), Kiev.
10. l'rotoãyakonov M.M. (1966). Opredelenie maksimalnogo stoka
poverhnostnyh vod s malyh vodosborov. (Determination of
maximum surface runoff on small watersheds) Hydrometeos
raologicalPublishing House Leningrad.
11. Sokoiov A.A. (1963) Maximalnyi stok talyh vod elementarnyh
bassainov i priroda ego reduktsii. (Maximum snowmelt
runofi on elementary basins and the nature of its re-
duction). Transactions of GGI, vol. 107.
12. Sokolov A.A. (1966) Metodika rascheta maximalnyh raskhodov
talyh vod pri otsutstvii ili nedostatochnosti gidrometri-
cheskikh dannykh (Computation of maximum discharges of
snowmelt water in case of the absence or inadequacy of
hydrometric data) Transactions of GGI, vol. 134.
13. Sokolovsky D.L. (1937) Normy maximalnogo stoka vesennikh
pavodkov rek SSSH i metodika ikh rascheta. (Norms of
maximum spring flood runoff of the USSR rivers and the
technique of their computation). Hyarometeorological
Publishing House.
14. Sokolovse D.L. (1948) Metodika postroenia hydrografa liv-
nevogo stoka P O osaäkam (Plotting of rainfall runoff hydro-
graph on the basis of rainfall aata). Transactions of
GGI, vol. 14.
15. Sokolovsky D.L., Shiklomanov I.A. (1965). Haschety hydro-
grafov pavoàkov s primeneniem elektronnyh modeliruyshchih
UStroiStV. (Computation of flood hydrographs by means of
electronic modelling devices). Transactions of LGMI,
voi. 23.
16. Sokolovslry D.L. (1968). Hechnoi stok, (River flovi). 3rd
edition, Hydrometeorological Publishing House, Lenin-
grad.
 623

17. Stroitelnye normy i pravila. (19661,Chast II, rasàel II,

glava 7. Raschetnye maximalnye raskhody vody pri proek-
tirovanii gidrotehnichesmh sooruzheniy. (Norms and
instructions for civil engineering. Part II, section
II, chapter 7. Maximum design discharges for hydro-
-
technical structures). Normy proektivania (CH i II
II 4. 7-65). MOSCOW.
18. Mezhdynarodnyi simposium P O pavocikairi i ili raschetam.
(1969). (International symposium on floods ana their
computation) I and II, Leningrad, 15-22, August, 1967.
Hyàrome teorological Publishing House, Leningrad.
19. Ukasania po opredeleniu raschotnyh maximalnyh raskhodov
talyh vod pri otsutstvii ili nedostatochnosti gidro-
metricheskih nabludeniy. (1966) (Instructions for
the determination of maximum design snow melt dis-
charce in case of the absence or inadequacy of hyaro-
metric data). CH 356-66. Hydrometeorological Publishing
House, Leningrad.
20. Ukasania po opreaeleniu raschetnyh hydrologicheskih kharac-
teristik, CH 435-72 (1972). (Instructions for the esti-
mation of the hydrological design values CH 435-72).
Hydrometeorological Publishing House, Leningrad.
21. Ukasatel literatury P O pavodkam i ih raschetam (1967)
(Bibliography on floods and their computation)
Hydrometeorological YuDlishing House.
22. Chegodaev N.N. (1953). Haschet poverhnostnogo stoka s
m l y h vodosborov. (Estimation of surface runoff from
su.311 watersheds). Tranzheldorizdat.
23. Shiklomanov I.A. (1964). Kaschet transformatsii pavodkov
vodokhranilishchami i prudami pri pomoshchi electron-
nogo modeliruyushchego ustroistva (Computation of
flood transformation by ponds and reservoirs by means
of electronic modelling devices). Transactions of LGMI,
vol. 26.
24. Alexeev G.A. ana Sokolov A.A. General principles and
methods for the computation of flood discharges applied
in the USAR. Atti del convegno internazionale (Roma,
23-30 November 1969). Roma, ANDL! pp. 735-747.
25. Estimation of maximum floods. Technical Note No. 98.
LNO v No. 233, TP. 126, T;MO, Geneva (1969).
26. Floods and their computation. Proceedings of the Leningrad
Symposium. August, vol. 1 and 2. UNESCO/USH (1969).
27. Flood studies: an international scuiae for collection and
~ ~~-..

processing of data. Technica1"papers in hydrology No.8,

UNUSCO. Paris (1971).
28. Gray D-M. -Synthetic-Unit-Hydrographs for Small Watersheäs.
Proc. Am. Soc. Civ. Eng., vol. 87.
29. Linsley R.X., Kohler M.A. and Paulhus L.N. (1949). Applied
Hydrology McGrow Hill Book Company N.Y.
30. Morgan and Johnson (1962). Analysis of Unit-Graph Method.
Journal or Hydraulic Division, 88, NY-5.
31. Sherman L.R. (1932) Stream-Flow Rainfall by Unit- Graph
.
Method ¡in Mews Record.
32. Snyder F.F. f1938) Synthetic Unit-Graphs. Trans. Am.
Geophys. Union, vol. 19.
 COMPUTATION O F PROBABILISTIC VALUES
OF L O W FLOW F O R UNGAUGED RIVERS

Vladimirov A. M., Chebotarev A. I.

State Hydrological Institute

Leningrad, U.S.S.R.

ABSTRACT

The main characteristics of low flow (minimum daily, mon-

thly and seasonal flows) are investigated. The computation methods
are Sased on a combined use of geographical interpolation and pro-
bability analysis and considering the main factors affecting the
volume aiid regime of low flow. Principal characteristics of low
flow f o r mkdium-size rivers are determined by maps of flow isoli-
nes, those for small rivers are determined by regional empirical
correlation. Design flow is established by means of transition
coefficients. The principal computation methods discussed are deve
loped for U.S.S.R. rivers.

Les auteurs analysent les caractéristiques principales des

débits de basses eaux (journaliers? mensuels, minimum saisonnier).
Les méthodes $e calcul font appel a la fois à l'interpolation géo-
graphique et a l'analyse statistique, compte tenu des facteurs
principaux qui influencent l'abondance et le régime des débits de
basses eaux. Les principales caractéristiques des débits d'étiages
des rivières moyennes font l'objet d'une représentation cartogra-
phique; pour les petites rivières on les traduit par des relations
empiriques régionales. Les débits correspondant à la fréquence
choisie pour le projet sont établis en utilisant des coefficients
de transfert. Les auteurs présentent les principales méthodes de
calcul utilisées en URSS.
b26

Low flow is one of the principal phases of the hydrological river

regime. During the dry periods, when precipitation is usually at its
lowest, rivers have rather stable and relatively small discharges. Their
variations in the flow hydrograph tend t o approximate a horizontal line.
The lowest flow observed during a certain period is generally called the
minimum flow during that period.

Separation of low-flow period in river flow hydrographs

On rivers with distinctly expressed spring snowmelt floods and

autumn floods the period of low flow is observed during winter and summer-
autumn seasons. Its beginning in summer is determined by the end of
spring high-water period, i.e. when the intensive rate of decrease of
discharge tends t o become smaller. The summer period of low flow ends with
the arrival of the autumn floods or the appearance of ice in the river.
In the latter case the low-flow period is called the summer-autumn period.

The winter low-flow period begins at the appearance of ice events

in the river and continues until spring high-water period begins. In
case of no ice phenomena in the river the winter low-flow period is
assumed to be a period from the average data of air temperature falling
down contii:vously through O°C and below it up t o the beginning of the
spring high-wcter period.

The low-flow period includes also floods if the volume of each

of them does not exceed 10-15 per cent of the flow volume for preceding
and subsequent lol.:-flow periods, without taking into account the volumes
of floods already included. If the flow hydrograph has the form of a
saw-like curve (frequent floods of various magnitudes), the period of
low-flow includes floods with maximum discharges that are 3-5 times
greater than preceding daily minimum discharges (depending on the
volume of the flood peak). These criteria facilitate the plotting of
river low-flow periods, although they slightly overestimate the volume
of low flow.

In the U.S.S.R. low flow may be expressed as minimum daily,

minimum monthly (30-day) or minimum seasonal flow.

Seasonal flow is the average value of discharge (specific dis-

charge) for winter or summer-autumn seasons. The minimum monthly flow
is the average during the lowest calendar month in the given season.
On rivers with flood regime during winter or summer-autumn seasons when
the low-flow period is of less than two month duration o r is interrupted
by large floods, the smallest of the average discharges during a calendar
month may appear 1,5-2 times bigger than the minimum discharge. In such
 627

a case it is necessary to introduce a temporary correction taking into

account not a calendar month, but a 30-day period with the lowest flow.
If frequent and considerable floods make it difficult to find out the
30-day period of minimum flow, it may be reduced to 25-23 days in order
to exclude the influence of floods. This secures the genetic homogeneity
of the m i n i m u m flow of years with varying water volumes, which is impor-
tant in the determination of minimum flows for ungauged rivers.

Physiographic factors of low flow

The duration and volume of low flow depend on physiographic

factors which may be divided into two groups: (1) climatic conditions
and (2) factors of underlying surface.

lhe water resources of a certain basin depend on the climatic

conditions prevailing in that basin. Precipitation contributes to the
increase of ground water supply, while evaporation decreases its recharge
and supply. In winter low-air temperatures cause a considerable freez-
ing of soils and subsoils and contribute to the decrease of underground
flow into rivers. Climatic factors determine areal low-flow distribu-
tion in accordance with the low of geographical zonation.

1.i some cases and for small rivers particularly, low flow is
greatly intluenced by local (azonal) factors of the underlying surface,
i.e. the surface and underground flow contributions (lakes, swamps, soils
and subsoils, karst, etc.).

The influence of these factors may be so considerable and

exceeds the influence of the climatic conditions.

The most essential factor is the permeability of soils and sub-

soils. They serve as underground flow reservoirs, detaining water during
high-water periods and releasing it during low-water periods. The
capacity of underground storage is determined by the geological structure
of the area and its hydrogeological conditions. Loose and porous o r
crevassed deposits (sandstone, limestone, shingle, and the like) create
favourable Conditions for underground storage of water and far its sub-
sequent release during the low-flow periods i n rivers. Solid clay o r
monolithic crystal rocks (granite, gneiss) near the surface decrease the
regulating capacity of the storage and reduce the low flow. The influence
of karst-affected rocks on the regime and volume of low flow is determined
by their absorption capacity und the rate of water yield
it is, the less is their influence an the low flow.
-
the bigger
628

The contribution of underground water reservoirs t o the flow in

rivers is a factor of extreme importance in the study of low flows. In
this respect due consideration should be given t o the number water con-
tent and regime of aquifers contributions to river flow and the dynamics
of underground flow into rivers. These factors determine the contribu-
tion of underground aquifers to the flow in rivers.

The study of factors affecting the volume and regime of low flow
is a necessary prerequisite for the successful development of the com-
putation methods.

The analysis of the influence of main factors on conditions of

low flow formation necessitates the division of rivers into small and
middle-size rivers when developing computation methods since the process
of low flow formation is different for the two types of rivers. Large
rivers are not considered here.

Differentiation of small and middle-size rivers

The quantitative characteristics of a small river may be assumed

to be the value indicating the extent of aquifers discharge contribution
to total flow, i.e. the erosion depth of river channels. The determina-
tion of its îharacteristic is the ratio between the erosion channel depth
and the aquifeis depth feeding the river along its length up to the outlet.
Th.e quantitative estimation of the influence of main hydrogeological fac-
tors on low flow is difficult to make, while developing low flow compu-
tation methods involve additional characteristics: correlations between
the capacity of underground storage and drainage densities. In similar
regions there is a definite relationship between the volume of under-
ground storage, river channel erosion depths, watershed boundaries and
drainage areas. Therefore the value of river basin area provides an
integrated indicator of morphological and hydrological conditions of low
flow.

In this case, a criterion for the term "small river" may be the
largest (critical) area of the basin responsible for the complete drainage
of aquifers feeding the river and with the enlargement of which no varia-
tion of low flow modulus is observed. The value of the critical area is
established by graphs of relationship of minimum 30-day flow modulus with
river basin area for the physiographically similar regions.

For the U.S.S.R. rivers, the critical area of the basin ranges
from 1, O00 to 1,500 km2 in flat wet regions and in all mountain regions.
In semi h u m i d zones it rises to 2, 000-2,500 km2 due t o the lower depths
of uquifers drained by rivers. In semi arid areas rivers with 5,,000-
10, O00 km2 basin area are classified as small rivers.
 629

Computation of normal low flow of small rivers

In the U.S.S.R. computation practice for determining mean low

flow of small ungauged rivers, the following equation relating the dis-
charge to the river basin area is most widely used:

where Q is discharge (seasonal minimum) in m3/sec; A is river basin

area in km2; f is either a regional impermeable mean area or a
permeable contributing area outside the drainage basin. In the first
case, the parameter f has the sign minus (-), in the second case it has
the sign plus (+). Under usual conditions and permanent flow available
f = O. a, n are regional parameters characterizing conditions of low
flow formation.

The determination of the parameters of design equation (1) is made

for the regions selected on the base of a careful study of hydrogeological
conditions of the basins under study and on the analysis of principal
physiographic conditions. For instance, while dividing the territory of
the U.S.S.R. into regions the following were used: map of drainage of
water beating formations by rivers, hydrological descriptions of condi-
tions favouiing the formation of underground flows of regions, ground flow
map of the intcnsive water exchange zone, map of underground flow i n per-
centage of the total river flow and coefficients of underground flow in
percentage of precipitation. Also used are maps of annual river flow,
precipitation for warm and cold seasons, data on evaporation, air tempera-
ture for the winter season i n ice melt regions, topographic map of the
U.S.S.R., hydrological regionalization of the U.S.S.R., map of physiogra-
phic regionalization of the U.S.S.R. All these allowed to take into
account as much as possible all the characteristic features under which
low flow of selected regions is formed. The boundaries of regions with
similar low-flow conditions during winter and summer-autumn, were plotted
along the boundaries of sharp change of h y d 2 l o g i c a l conditions. For
instance, when in some river basins the change of hydrogeological and
other conditions take place, the change in the volume of river flow will
not be observed immediately, but gradually while the most notable change
will take place at the confluence of two rivers. In this case the
region boundary follows the watershed divide between these river catchments
across the point of their confluence.

Formula (1) may be used for the computation of flow of flat and
semi-mountainous rivers with the average accuracy of 152% (for 1 500
points on the U.S.S.R. rivers the deviation of computed minimum mean
long-term 30-day discharge was 17-2w of the actual flow for the summer-
autumn season, and for 750 points in winter it was 15%). Taking into
630

account the accuracy of determining the actual data, use of formula (1)
may be recommended for the computation of low flows of rivers of basin
areas not less than 20 km2 for humid zones not less than 50 km2 for
semi-arid zones, where the low flow volume is rather small and the in-
fluence of various local factors is most evident. In regions with very
complicated conditions of low-flow formation the area should be not less
that 100 km2.

A wide use of formula (i) in the designing practice (5, 6 ) paoved

it reliable.

In high mountain areas the altitude of the catchment may be of

a great significance as the factor reflecting the influence of vertical
zonation upon the conditions of low flow formation. Therefore, the low
flow modulus for regions similar in hydrogeology etc., is related t o
mean basin altitude.

Determination of low flow for middle-size rivers

The low flow volume of middle-size rivers, i.e. those with area
larger than the above stated critical area, but not more than 75 O00 km2,
is formea under principal influence of zonal factors. The flow modulus of
these rivers varies smoothly and in accordance with geographical zonation
(1-atitudinal or vertical) over the area. Therefore, low flow of middle-
size rivers can be determined by maps of flow isolines, made for a certain
characteristic of low flow. The flow modulus relates t o the catchment
centre, the interval between isolines is given in accordance with the map
scale and the value of flow variation over the area. In mountain regions
the average catchment altitude is taken into account; flow isolines may
not be closed, but end on the side of the mountain ridge without passing
over to the other side (due t o a great difference in wetness of slopes).
Maps are plotted both for the mean and for flows of various frequencies.
For instance, for the U.S.C.R. territory there are plotted maps of the mean
and S-frequency of minimum. 30-day winter and summer-autumn flows,
which allows t o determine the flow with the average accuracy of 10-20$.

Computation of low flow of different frequencies

For designing purposes the characteristics of low flows of dif-

ferent frequencies are of the highest importance. In the U.S.S.R. design
flow of 75-97$ frequency is mainly used. Necessary values may be deter-
mined with the help of three parameters: mean flow La), coefficient
variation of !Cv ' and skewness coefficient (CS).
 631

The second way is by the use of a transition coefficient from one

fixed frequency (e.g. 75 or 8%) t o another. This method has been lately
more and more widely used in the U.S.S.R., especially in its application
to low flow, since it is more accurate and simple than the method of three
parameters, and since available hydrometric data allow t o generalize for
almost the whole territory of the U.S.S.R.

The advantage of the transition coefficients method is proved

by the mere fact thót in this case the total mean square root error will
consist of the error of the flow of fixed frequency (u 1 and the error
P
'
of transition coefficient A , i.e.

When using three parameters, the same error will consist of three
terms: standard error (Qn), error of Cv (c($ and error of Cs ,( Cc,),
i.e.

Tt is evident that the error in the second instance will be

greater, and if we take into account unreliable methods for determining
coefficients C,, and Cs for any rivers, then the advantages of the method
of transition coefficients become quite obvious. It is the more so,
as in the range of frequencies under consideration (7597%) the curves of
low-flow frequencies are rather stable, gently sloping and quite reliable
in most cases. This stipulates a rather small (for the given frequency)
variability of transition coefficient for the area and season and, con-
sequently, its high reliability. Thus, for the U.S.S.R. rivers the value
of transition coefficient from the minimum 30-day discharge of 8% fre-
quency t o the discharge of 75% frequency varies from 1.03-1.06, and for
transition t o the discharge of 9056 frequency - from 0.83-0.91, i.e. the
value of coefficient x c h a n g e s only by 5-1056 and may be averaged for the
given frequency over a large area. Its value varies significantly only
for episodically drying or freezing rivers.

The flow of fixed frequency is established by formula (1) or by

maps of flow isolines, plotted for this frequency. Thus, for the U.S.S.R.
territory there are plotted maps of the minimum 30-day flow of 8056 fre-
quency (for winter and summer-autumn periods separately) and the maps of
flow of limiting season of 75% frequency. Also determined are the
parameters in formula (i) for the 30-day discharge of 8w and 7546 fre-
quency.
63 2

To determine flow of other frequencies, a table of transition

coefficients ;I has been prepared.

Computation of minimum daily flow is made by relationship with the

value of minimum 30-day flow (normal or fixed frequency flow for selected
regions) :

Q = K , Q
P p,30
where Op is minimum daily discharge of design frequency. Qp30 is minimum
30-day discharge of corresponding frequency, determined by maps of isolines
or by formula (1). k is the regional transition coefficients for the
given season.

For the U.S.S.R. territory the value of coefficient k, when

determining the minimum daily discharge of 8C$ frequency, varies from
0.59 to 0.90 in winter and from 0.45 t o 0.86 in summer-autumn seasons.
Its value depends on the degree of river flow depletion for the period
under study and on the volume of runoff during low-flow periods.

Determination of minimum daily flow of design frequency is made

by using ti:s above-mentioned coefficients A , since the frequency curve
of daily and 30-day discharges vary practically in the same manner.

The stated methods for the computation of probability values

of river low flow are given in Gosstroy Standards of the U.S.S.R.
/5,6/ and are widely used by designing organizations of the Soviet
Union.
 633

REFERENCES

1. Vladimirov, A. M.: Minimalny stok rek SSSR (Minimum flow

of the U.S.S.R. rivers) Hydrometeorological Publishing
House, Leningrad, 1970, p. 214.

2. Vladimirov, A. M.: Raschetnye minimalnye raskhody vody

(Design minimum discharges) Trans. of GGI, v. 188,
Hydrometeorological Publishing House, Leningrad,
1972, p. 244-272.

3. Kudelin, B. I. (ed.): Podzemny stok na territorii SSSR

(Underground flow in the U.S.S.R. territory), MGU
Publishing House, 1966, p. 303.

4. Popov, O. V.: Podzemnoe pitanie rek (Underground river

recharge) Hydrometeorological Publishing House,
Leningrad, 1968, p. 291.

5. Ukazania PO opredelenia raschetnykh minimalnykh raskhodov

vody rek pri stroitelnom prooktirovanii (Instructions
for determination of design minimum discharges of
rivers in engineering projects). CH 346-66, Hydrometeo-
x d o g i c a l Publishing House, Leningrad, 1966, p. 17.

6. Ukazania PO opredeleniu raschetnykh gidrologicheskikh

’ kharakteristik (Instructions for determination of
design hydrological characteristics), CH 435-72.
Hydrometeorological Publishing House, Leningrad, 1972,
p. 18.
 A STUDY ON MAXIMUM FLOOD DISCHARGE FORMULAS

Tae Sang Won, PhD.CE., Dr. En.${

ABSTRACT

This paper describes a new formula for the calculation o f

approaching velocity of rain water, and a number of new formulas
for the estimation of maximum flood discharge which have been deve-
loped by the author.

Many empirical formulas, which have limited application,

exist. However, in devising his formulas, the author derived theo-
retically the form of the basic maximum discharge formula for the
case of rivers with no tributaries, and determined stochastically
the value of the coefficients in his basic formulas using the re-
cords of observed measurements. Then the author derived theoretic2
lly many differe,it formulas for the case of rivers with tributa-
ries to fit in the actual localities of the site under considera-
tion, besides the basic formulas. So the author's formulas would
be widely applicable for rivers or sewer nets, and also for any
regions, countries, with different locality. The author could c o n
firm these facts through the numerical examples. The author's fo:
mulas may be used not only for estimating the design flood, but
also in flood routing. The author believes that his formulas would
be very helpful in the planning of water resources development p r o
jects -specially for those with inadequate data.
1

RESUME

L'auteur présente une nouvelle formule pour la vitesse de con

centration d'un bassin et en suggere d'autres pour le calcul du d g
bit de la crue maximale.

On trouve de nombreuses formules empiriques dans d e nombreux

manuels, mais ces formules sont d'une application limitée. L'auteur
parvient cependant à asseoir l a forme de sa formule sur des bases
théoriques, lorsqu'il s'agit de cours d'eau sans affluents; il pro
cede à l'évaluation des paramètres qu'elle contient par ajustement
statistique aux données d'observation disponibles. I1 generalise
ensuite à différents cas d e cours d'eau avec affluents. Les formu-
les proposées de'vraient pouvoir être appliquées n'importe o ù ,
aussi bien pour les cours d'eau naturels que pour les réseaux
d'assainissement; c'es ce que l'auteur peut confirmer par des
applications numériques. Les formules peuv:nt servir non seulement
au calcul des crues de projet, mais aussi a celui de la propagation
des crues. L'auteur pense que ses formules devraient rendre de
grands services dans la planification de l'aménagement des eaux,
spécialement lorsque les données disponibles sont insuffisantes.

fg Professor of Civil Engineering, Seoul National University, Seoul,

Korea.
636

I e XNTR001';TION
a
Charles F. Ruff defin& B m x i m m probable floodn as follows.
"The maximum probable flood does not mean the largest flood possible
but a flood so large that the chance of its being exceeded is no
greater than the hazards normal to all of man's activities." The
author will use here the term of 'maximum flood discharge" with the
same meaning of "maximum probable flood" as defined by Ruff.
It is very important to calculate maximum flood discharge cor-
rectly, and also it i5 a very difficult problem theoretically and
practically. It m y be impossible to establish a plan for flood con-
trol an8 water resources development or sewer nets projects without
reckoning correctly the maximum flood discharge or the design flood.
There are many methods for calculation of maximum flood dis-
charge, and we have to adopt the most suitable method in accordance
with the completeness of the data. However,the method of calculation
by the maximum flood discharge formulas,especially for the case of
those with inadequate data, is easy and simple for practicing engi-
nßers. There are many empirical formulas devised by many authors
such 88 Kuichling,Mead,KresnikeDickens,Metcalf and Eddy,Brix,Lauter-
burg,Possenti,Buerkli-Ziegler,Dr.Hisanaga,Kajiyama,and many others.
These old fosmulas have been devised empirioally and have limited
application. It will be clear that one may be unable to apply them
generallj.. Aliso it will not be strange to obtain results which may
be 10 or liio time8 of the correct values, according to selection OP
the coefficieqts in these formulais when these formulas are actually
applied to practical problems.
Generally speaking,the flood discharge depends upon the shape
of catchment,drainage area,amount of rainfall and the position at-
taoked by the heavy rainfall,pemneability,slope of the catchment,
shape of the water cours6,status of the surface,geological status,
etc. Strictly epeaking,such statua of catchment differs from others
from se68on to season,for every floo8,even in the same catchment as
well as in different draimge basins. In other words,flood disoharge
üepends also upon the inteneity of Fainfall which causes the flood,
duration of the rainfall and the position of the o a t e r of the lows,
or statua of the ground in case of heavy rainfall,vie.,dry ground or
saturated qround,etc.
As the maximum flood discharge depends upon many factors, as
stated above, it may be very difficult to express it in a formula.
However,if we can consider theoretioally correct value of approach-
ing velocity of rain water and intensity of rainfall, we may deduct
the maximum probable flood by getting the rainfall for a certain dis-
trictc The principle of derivation of the author's formulas belongs
to this process, and it may be said that this is an approach differ-
e.it from many scholars who had derived the old formulas.
~

* Ruff,Charles F.;nMaximuni probable floods in Pennsylvania Streamst'

.
Transactions ,American Society of Civil Engineers .Vol .i06,1g4l ,p 11 53
 637

In the first step, the author thought out a method to ascertain

correctly the approaching velocity of rain water. A t the same time,
the author found that the Rizha's (Germany) formula,the only complete
one for this purpose, could not be applicable to solve practical
problem8 as it gives too small values. In the second step, the author
studied the rainfall intensity curve comprehensively, and found out
theoretically when the maximum flood discharge may occur. In the
third step, the author has theoretically derived the maximum dia,
charge formulas for rivers with many tributaries by appljing the
general rules which he has determined by the first and second step.
In the fourth step, the author determined the discharge coefficient
in his formula from the actual records. The author was then able to
calculate the value of the discharge coefficient, with a great degree
of acouracy, of the rivers in Korea and Manchuria.
II. APPROACHING VELOCITY OF RAIN WATER
The approaohing velocity of rain water (U) is defined as the
mean velocity of rain,water approaching from the farthest point F in
a river basin to the point O where the maximum flood discharge is to
be ascertained,in other words, the mean velocity of flow between F
and O (Fig-1). There is only one formula to find such approaching
velocity so far expressed in equation, given by Rizha,Germany, and a
table given by Kraven,Germany.
1) RIZYA'S FORMULA
0"72 So'' (1)
where
a =Approaching velocity of rain water (km/hr)
s = H/L
H = Difference of elevation of height between O and F
L = Distance of OF (Length of water course)
2) KRAVEN'S TABLE
C
w (kln/hr)
above 0.01
12.6
0.01 -
10.8
0.005 below
7.56
0.005

Kraven had expressed only about approximate limite of Cd , the

author tried to formulate his table, to pass through the medium
points as follows.
(2)
3) THE AUTHOR'S FORMULA
The author succeasfully devised a new method to determine the
approaching velocity of rain water e3 ,theoretically which may be
applicable for rivers where the hydrographic surveying was completed.
Neglecting the principle and the process of derivation here, the re-
sult of the author's formula is illustrated as follows.
 638

The value of for rivers in Manchuria calculated using eq(3) is

shown in Table-i. As we see Table-1, the value of lies between 133
and í77, and we may be able to recognize that the Rizhass formula wil
not be of practical use because of the reason that the salue of iCt in
his formula is too smll,apparently,compared with the normal salues.
ide can also see from Table-1 that the value of K represents the
bottom slope OP the hydraulic siope of the river, and this fact coin-
cides with practice. Also it can be seen that the value of &, de-
ureases gradually according aa approaching the downstream of a river,
and this fact skok's that the bottom slope or the hydraulic slope of
a rivep generally decreases gradually as we approach the downstream.
III. RAPFALL INTENSITY CURVE
h'e can express the rainfall intensity by the following equation.
Table-I. The values of & and others fop rivers in Manchuria
Name of Rivers /c Range of S
Tumen 8. 277 4.82
(400)
-- 8,06 33,400 487.6 68.50. .LW
Whancheng R.
Rohe H.
207 6.72
177 3.54 -- 9.64
4.22
4,000 160.7
31,455 444.0
29,927 412.0
24.92 .l52
70.84 -159
72.64 .i76
149 8.37 -
Seasamorin R. 171 2.81 2.97

-- 3.16
Sealeog R. 3.48 510165 767.0 66.71 .O86
Tongleog 3. 207 1.70 10,318 33345 313.13 -093
The upatrean of the 150 1.83 2.30 178,699 1040.5 171.74 .O65
.

m i n Lacg R.
The middle of the Leog 158 1.62 -- 1.77
187,250 1199.0 157.49 0132
.
137 4-36 -
GhsnF: R. 159 3.90 5.514,958 163.0 30 42 .i86
Icwaslg R. 5.26 2,129 94.0 22.65 231
Van R. 149 7.07 -18,66 1,072 102.5 10.46 .lo2
4.062,361 178.0 13.26 e 074
134 1.43 - 1.96
Pa B. 150 3.68
Csnkai R. 515 51.0 10.10 .198
F:catchmtmt area
I = ß/(t + 1 (4) L:length of main water oourse
wher0
t = Bwatim
I = Average intensity of rainfall during duration t
=
OC,,^ Any-constant
Eq(4) represents a kind of hyperbola, and the constants a ar3.B can
be found by eq(5) by the principles of the method of the least s q u a m e
n(12t) - (ï)(ït)
d=
LI)' - n(I')
(Il(P2t) - (It)(12)
B= (I)~ - n(I')
E = nwnber of observations
Next let R be total. amount of rainfall aurin$ the buration t,
R = It = p t/(t + QL 1 (4)
 639

T ~ b l e - 2illustrates the values of the constants d and fi in eq(4)

for various regions. In this table, those for the regions of Korea
and Manchuria show the absolute maximum rainfall intensity curves
during those periods, Those for the regions marked with the asterisk
(*I were calculated by the author himself by the records of the re-
cording gauges.
IV. THE AUTHOR'S MAXIMUM FLOOD DISCHARGE FORMULAS
1) FUNDAMENTAL FORMULA FOR THE CASE OF A RIVER WITH NON-TRIBUTARY
(a) Retardation of Run-off
Table-2. The values of cc and P in eq(4)
Region d a Period taken Range of t
(min) (hour) P b the records (minutes)
Seoul ,K 59 0.938 7,860 131 1905 - 1920 5-60 min
1914 -
Inchon ,K 37.5 0.625 8,640 144 ditto 5-240
1914 - 1953
PymgYanR ,K 41 0.683 6,000 100 1920 ditto
Pusan ,K
Wonsan ,K
Taegu,K *
106.1
75
40.2
1.77 14,015 233.6
1.250 7,740 129.
0.67 8,711 145.2
1914
1929
-- 1920
10min-24hr
5-240 min
1953 10min-24hr
Chonj:i,K * 81.1 1.35 15,160 252.7 1918 -- 1954 ditto
Kwangju,K *
Mokpo,K *
*
90.4
101.8
1.51 10,866 181.2
1.70 11,398 190
1938
1916
- - 1954
1953
ditto
ditto
ChmgCheng,):
Sping,r? *
Tokyo,J
40.5
45.1
50
0.675 5,929 98.8
0.752 8,487 141.4
0.833 5,500 91.6
1937
i934 -
1891 -
1943 10min-48hr
1944
1911
ditto
5-60min
where a d/60 b P 3'1 /60 KæKorea M=Manchuria JsJapan
Prior to deecribing the flood discharge formula, the definition
of "retardationR must be understood. Now let O and F be the point
under coneiäeration and the farthest point of a catchment respective-
ly, 1 be the length of water course between O and F , CL) the approaoh-
ing velocity of rain water flowing from F to O, tc the time of con-
centration,i.s.,the time necessary for reaching O from F, tr the du-
ration OP rainfall,i.e.,the perlob between the beginning and enâing
of a reinfall (see Fig-i), then,
t, = l/Ld (79
T tr + tc * tr + l/W í 8)
where
T The time of the period between the beginning of a rain-
P
fall and the ending of the run-off due to the rainfall
at the point O
It may be better to use the author's formula for determining tc
When it becomes tr<t,, the run-off is subjecteä to retardation, and
in this case the rain water falling the whole catchment would not
reach the proposed point simultaneously. Because as the rain water
640

falling at F reached O, the rainfall would have ceased. In other

words, the rainfall causing the maximum flood discharge is the rain-
fall which falls in a part of the catchment.
(b) Fundamental Formula for the case of non-Retardation
The fundamental principles of the author's maximum flood dis-
charge formulas have already been described. The author adopted de-
ductive and inductive theories for the derivation.
Since it is unable to derive the rational equation of the flood
discharge hydrograph, the author expressed the peak of the flood
discharge hydrograph for the case of non-tributary and non-retar-
dation by the following equation.
q m - 9. a CfA R / T (9)
where
qm
qo
= Peak discharge
T = Duration of flood
t r = Duration of rainfall
-
in flood time
= Discharge of run-off in normal time
tr+ tc
tc = Approaching time or time of concentration
R = Total amount of rainfall during duration of tr
A = Catchment area
9 = Average
run-off factor
C A coefficient depending upon the shape of the flood
P
discharge hydrograph
Now let Fig-,? show a discharge hydrograph during a flood period.Then
the, peak disoharge q m - q O may be represented by eq(9)and the product
AR of eq(9) shows the total run-off during the period of flood
T. As this also represents the area of DMED of Fig-2 geometrically,
we may affirm that the authorls fundamental formula is reasonable
anal tically or graphically. Replacing the value of R of eq(6) into
ea(9 , 9 -
q m q o = C 9 AB / T =
C 9 A b t r /(tr + t, )(tr+ a (10)
We h o w through eq(1O) that the peak discharge is a function of tr,
anä it will take a limiting value to make the peak discharge maximum.
So differentiating eq(l0) with respect to tr
tr = (unit in hours) (11)
and.
T=t,+ t G = 6+t, (12)
Hence we know that the maximum flood discharge will occur when the
duration of rainfall t r satisfies eq(l1). Up-to-datesue have taken tr
generally without definite reason as follows: 5 or 10 minutes for de-
sign of sewers, 3 or 4 hours for small rivers flowing the vicinity of
a city, 24 h o m e or more for big rivers. However aceording to the
author's theory, the value of t r must satisfy eq(l1) to cause the
maximum flood discharge. Substituting eq(l1) into eq(lO),
 641

If we express in metric units,i.e. ,A(km2),R(~),t (hr),q (cms), (13)

becomes,
where
qm-q,= 0.2778C b 9 a
A / (t,+fic)( a + 1
(14)
a,b = any constants depending upon rainfall (see Table-2)
The value of t c can be found from eq(3) .(see Table-1)
(c) The value of the coefficient C
As stated above, the coefficient C in eq(9) depends upon the
8hape of the discharge hydrograph of the region .The relation be-
tween the kinds of the cuPve consisting the discharge hydrograph and
the value of C is illustrated deductively as fOllOW6.
Table-3. The value of C found by deduction
Kind of curve C kind of curve C
parabola 1.5 cosine curve 2.0
triangle(strai ht 2.0 probability 2.394
1ins? curve
Also the value of C can be calculated inductively from a dis-
charge hydrograph by using eq(91,whioh gives,
C 0 (qhi-Qo)T/ $ A R a ( Q m = Q , ) T / V (15)
nhere
V = The volume of run-off represented by the area DMED of Fig-2.
The value of C for the rivers in Manchuria found by the author using
eq(l5) are given in Table-4.
Table-4. The value of C for Manchrian rivers found by induction
Name of river Site of Duration of flood taken Value Value
measurement from the records of' T of .c
(hr ,day-hr ,day,month, r
Tongleog Ho Tidathergtse 15,3rd-21 ,4th,Aug,l9ii 30 1.664
n Sankankeu 12,10 .e 5,21 ,Sep,1939 257 1 .697
Whan Ho P eidakeng 15,24.-1 9 ,27, AUg,l94O 76 1.543
Main stream C hengsenkong 19,2nd- 6,6th,Sep,1939 83 2.090
of Leog Ho
ditto ditto 797th- 7,103 Se~r1939 72 1.966
Taitse Ho Whelongbo 12,31Jul-3,3~,Aug,1940 63 1.754
n n 9,4th-i7,6th,A~g,l940 56 1 0975
I n 16,6th-l9,8th, I 51 1.087
n n 17,2nd- 8 s 5thSSQp ,1939 63 2.137
n H 9,jth-l6,9th, " " 103 1.806
n 5,6th-l3,9th, Jul , .
80 2 204*
* show6 the value oalculated by estimation because of non-measure-
ment at the vicinity of the peak discharge.
(d) The fundamental formula for the case of retardation of flow
 642

The basic formula for the case Of non-retarclation ,mentioned

above, is applicable for the case of retardation of flow,too. But it
is necessary to multiply the coefficientp due to retardation, viz.,
q,-qo=yCpbf& A / ( t c + G L )(a+ Gc
1 (16)
p= f(tc/tr) (17)
It is clear that the value of the Coefficient /3 equals to 1 for the
case of non-retardation, but it beoomes less than 1 for the case of
retardation. The value of p varies inversely with that of t c / t, .
It is necessary to find out a general form of f(tc/tr) for
practical calculation. So the author tried to find out the general
form of the function f(tL/ty) atoohastically using some data o b
tained for rivers in Korea by some other methods. The author would
like to assume the general form of the function o f p as follows.
J ) = (1 + k 1 / ( tc/t,+ k 1
where
k = Any constant
Finding the value of k in above equation by the method of the least
squares, we get k =
4.802 .
Accordingly,
p= 5.802 / (tc /tr+ 4.802) (18)
2) 'THE MAXIYUM FLOOD DISCHARGE FORMULAS FOR THE CASE OF RIVERS
WITH TRIBCTARIES
(a) The maximum flood discharge st the confluence of a trlbutary
The author found that existence of tributaries affect greatly
the peak discharge of flood flow at the proposed site of the main
stream. Su the author derived many different formulas of maximum dis-
c h a r m for the case of rivers with tributaries, besides the basic
formula for the case of thoee with non-tributary. Therefore it would
be said that this is a great approach different from many scholars
who never considered the Influence of tributaries in their tradition-
al formulas.
KOW assume one of the simplest case as Fig-3. The discharge
hydrograph for this case may be illustrated as Fig-&. The value of q,
in ~ i g - 4shows the peak discharge of the triùutaryíI), and the value
of q 2 shows that of the main river (II) alone,excluding that of the
tributary(1) , also Qm shows that of the composed maximum discharge
to be occurred at the proposed site. The rational equation of the
curve ,i.e.,the true shape of the discharge hydrograph is unknown.
But the author would like to discuss about the shape of the curve in
the following. Let us consider two cases, one of them the simplest
case,i.e.,the case assumed that the discharge hydrograph consista of
an isosceles triangle, and the other the case assumed that it con-
sists of a parabolio ourve, to seek the effect of the nature of the
discharge hydrograph which influences on the peak disoharge Qn
 643

( i ) 'The case of an isosceles triangle

In this case, it evident from Fig-5,
TE
Um" qp+q,(2 --1 (19)
T,
(ii) The case of a parabola
Since it is evident as the nature of the parabola,at Fig-6,
q = 4qot/T - 4q,(t/TI2 í a)
we can get the following equation for Fig-?,

and by dQ/dt = O
Q,/TI + q2 /T2
(20)
to= 2 (q,/T,' + q,/T:)
Accordingly, substituting eq( 201 into eq( b) , we get

NUMERlCAL EXAMPLE
An ilìwtration is given here to compare the degree of accuracy
of the two cbses mentioned above.
=
Given T2 26 hr, T I = 20 hr,
Te/T,= 26/20 1.3 =
curve, &TA= (5000 + jOOOx1.3
Qm =
5000 + 3000x(2 1.3)
=
5000 cms, q = 3000 cms
/( 5000 + 3000x1.3x1.3 1
-
Next from eq(19), the straight line formula,
=
7100 cms
-
. Then since
from e q 8 1 ) , the case of assuming as parabolic
7865 cms

Hence,we h o w that there is not any remarkable difference on the re-

sults of calculation of the maximum discharge whether we assume the
discharge hydrograph as straight lines or a parabolic curve through
this numerical example. Aliso we can imagine that we shall obtain the
similar results with this numerical example even in the cases we
adopt Borne other ourves else than parabola for the discharge hydro-
graph,e.g.,cosine or probability curve. But adopting the oase as-
sumed as a parabolic curve is safer,easier to handle,& reasonable.
So the author would like to suggest those of the parabolic curve as
the general formula in this paper.
(b) The maximum flood discharge formula at the confluence for the
oaee of a river where n-1 tributaries flow into the confluence
(Fig-8 1
If we a s s m e the discharge hydrograph consists of a parabolic
curve, by the srne priciple with that in the previous paragraph, we

(0) The maximum flood discharge at the proposed site which is

located the downstream of a tributary
644

Now let O is the proposed site, 'O the confluence of the tribu-
tary in Fig-10, and t, is the necesssry tirne for reaching of rain
water from O' to O. If we assume the discharge hydrograph consists
of a parabolic curve.FiR-11 ,then
= - - -
Q Q;,+ qiez4q,(t-t,1 /TI 4 qi(t t ,l2 / TF+4 qr t/T+ 4 kt2/Ti (

(d) The maximum flood discharge at 8 proposed site where n-1 tribu-
taries join to the main river at its upstream side.(Fig-12)
if we assume the discharge hydrograph consists of 8 parabolic
curve, by the same principle with that in the previous paragraph, we

(e) The ma>;Smum flood discharge at a proposed site where m tribu-

t a i e s flow into this site and n-1 tributaries join to the
main river at its upetream side. (Fig-14)
This is the most general case. If we assume the discharge hydro-
graph consists of a parabolic curve, by the same principles, we get

ìGYi
V. CONCLUSION
The maximum flood discharge generally increase toward âown-
stream, a6 the result of increment of the drainage area. But as the
approaching time also increases approaching down stream,in other
words,ss the nearer approaching downstrem,the greater effect of re-
tardatlon. Accordingly the rate of increment of the peak discharge
decreases generally approaching downstream; and sometimes,i.e.,in
such oa8e18 where the approaching time remarkably increases compared
with the inorement of the drainage area, not only the rate but also
the actual absolute value of the peak discharge decreases at the d o m
stream than those of the upstream. These fsots are experienced some-
times in practice, In such cases,it was impossible to expreee this
fa& by the old formulas. However by the author's formulas, it is
 645

easy am2 theoretically sound to express this fact. Because as we see

the author's basic formulas-eq( 9)-(14), which represent the drainage
area A in the numerator and the factor of the approaching time t c in
the denominator. So it may also be said that the author's formulas
are very theoretical from the point of view of this fact.
As mentioned above,the author derived theoretically,i.e.,ration-
ally or stochastically many formulas of maximum flood discharge- the
basic formulas for the case of a river with non-tributary and many
other different formulas for the case of rivers with tributaries. Be-
cause the author found that the existence of tributaries affect great-
ly not only the peak discharge but also the entire shape of the dis-
charge hydrograph at the proposed site of the downstream. Consequent-
ly it may be posaible,by applying the author's formulas,to find the
real shape of the discharge hydrograph at the point under consider-
ation to be occurreti in some flood time.
Some scholars advocate that the actual shape of the flood dia-
charge hydrograph resembles to ~ig-16. On the other hand,some other
scholars insist that it should be resembled to Fig-17. But the author
should say that these theories both advocated by the traditional
scholars are those have not been touched to the core of the true theo-
ries. The real shape of the discharge hydrograph depends upon the lo-
cality Qf the point under consideration,in other words, it depends on
the relat4ve position of the proposed point and those of the conflu-
ences of th? tributaries on the mainstream under consideration. Conse-
quently it is resembled to ~ i g - 1 6in some cases, and also it takes a
shape resembleel to Fig-17 in some casessin accordance with the locali-
ty of the point under consideration. A s stated above,it would be able
to show the real shape of the discharge hydrgraph just fitted in the
locality of the proposed site by applying the author's formulas.
The author's formulas also would be applicable not only for the
purpose of reckoning of the design flood, but also for that of esti-
mation of the flood routing for some floods. In the case of flood
routing,it would be possible to obtain more correct results by taking
the real value fop tr instead of that calculated from eq(l1) in some
cases,i.e.,the real value of tr is greatly different from that calcu-
lated from eq(i1).
The author's formulas would be widely applicable for rivers or
sewer nets,& also for any regions,countries with different locality,
and it would be possible to obtain correct and accurate results by
selecting or assuming the values of the coefficieuits in his fQrUIUlaS
appropriately. Aceoräingly the author should like to suggest that the
author's formulas shall be applied in practioe in many regions and
also for many purpose8 as far as possible.
(p
646

F¡y -I Fi9 -2 Y
Fij-3
ot P
e e

t
t
-+
O

3
c,

O
Fìg -9
Fig- 8
Fig-7
647
 THE COST-EFFECTIVENESS OF WATER RESOURCES SYSTEMS
CONSIDERING INADEQUATE HYDROLOGICAL DATA
Nathan Buras, Ph.D.
The Lowdermilk Faculty of Agricultural Engineering
-
Technion Israel Institute of Technology, Haifa, Israel

Introduction.
The question of how much hydrological information is
necessary for the design of water resources systems has not
been answered satisfactorily as yet. Perhaps this question
does not admit of a unique answer, but rather of a range with-
in which the specific solution to a given situation may be
found .
In general, one can state intuitively that the cost of
a water resources project decreases with the amount of avail-
able hydrological data. For example, a longer hydrological
trace at a given reservoir site will yield improved estimates
of mean annual discharges and of extreme flows, so that the
dimensions of the dam and of the spillway may be reduced for
a given probability of failure during the same period of time.
On the other hand, additional hydrological data irivolve in-
creased cost, not only in terms of more gauging stations and
of the attendant manpower, but also in terms qf 'osts incurred
to the society by delaying the design and the covistruction of
the project until more data is collected and processed. Schem-
atically, one can show these two cost functions as two curves
intersecting in the data-cost space (Figure 1). However, of
practical importance are not the individual cost curves, but
the parabola which is the sum of the two functions. We shall
define, therefore, as adequate hydrological data the amount
of hydrological information corresponding to tho niinimum
ordinate of the total cost curve. This definition impli-s
that hydrological data in excess of this amount are as '.nade-
quate as those which are short of it: indeed, the effort put
in obtaining this additional information may increase the total
cost of the project. For this reason, we recommend the use of
the terms insufficient data for the information less than ade-
quate, and redundant data for the information in excess of the
point of adequacy.
The problem of adequate hydrological data is part of the
broader issue of planning water resource; s@erns. Within this
enlarged context, the hydrological data is but one of the
several planning variables, the others being socio-economic
considerations, organizational and i.nstitutiona1 structures,
political constraints, and so on. The role of the hydrological
data in a complex water resources system was investigated rela-
tive to the water quality in the Potomac estuary [I]. In this
analysis, four planning variables were considered: (a) hydro-
logical inputs; (b) models of the dissolved oxygen fluctuations
 650

in the estuary; (c) economic projections of the region serviced

by the water resources system; (d) water quality objectives in
the estuary. Under the specific conditions of the Potomac, it
was found that the performance of the planned water resources
system was most sensitive to the economic projections, and
least sensitive to the hydrological planning variable (10-
year and 50-year sequences of hydrological data).
Sufficiency of hydrological data.
It does not seem that there is today a generally accepted
method for the evaluation of the amount of hydrological data
with respect to their adequacy for planning water resources
systems. However, the problem was recognized for some time and
several approaches toward its solution were developed. One such
approach, based on the concept of information content of the
observed data [2], is oriented toward the determination of an
opt,imal.letwork of hydrological stations in a region. A some-
what similar approach is based on minimizing the sum of variances
of the estimates of the mean flows at gaging stations in a hydro-
logical network subject to a budgetary constraint [3]. All these
approaches attempt, in fact, to devise optimal strategies of
hydrc logical sampling.
However, when considering the L msequences of inadequate
h,droiogical data on the cost and effectiveness of water resources
Ftrur?l,!res anfi FroJects, it secas thnt the scope of the analysis
Iza> t o be broadened. This analysis takes into account not only
ali pcc:;ible sample results, but also computes the expected
worth 3r expected opportunity loss) of a strategy which assumes
that the best decisions (regarding the various components of a
water resouI-cessystem - to construrit or not to construct) are
dependent upon the information content of the observed sample.
+
This approach is called preposterior anal sis [4], because,
&hough carried out before the sample in ormation is obtained,
it attempis + a s s e s s e r i o r probabilities derived on a parti-
cular s a p l e rutcome.
A simple example will illustrate the preposterior analysis.
Suppose that the Development Authority of region Aleph is consi-
dering thF construction of a major dam. However, the Authority
wants i l s ~to appraise the advisability of obtaining additional
hydrological data thus delaying the planning and implementation
schedule by a few years. It is estimated that total costs involved
in obtaining the additional data, including costs generated by the
non-availability of water and water derivatives at the dam site
 651

6
during the additional time period, are 15 x 10 Monetary Units
(in short, 15 MMü). The contemplated structure needs an invest-
ment of 160 MMU, while the present value of the stream of net
benefits generated by it would add up to 200 W .
The Authority has two options:
al: build the dam
a2: do not build the dam
with the possible outcomes
el: the project is successful
9,: the project is a failure.
On the basis of past experience and with the help of a
firm of consulting engineers, the Authority reaches the con-
clusion that the prior probabilities of success or failure are
p(el) = 0.25
p(e2) = 0.75.
On the basis of the existing data the prior expected
opportunity losses (EOL) can be computed as follows:
Table 1.
Calculation of Prior Expected 0pportimj.ty Losses
a,: build the dam
Probability Opportunity Loss, Wej-giited Oppor-
Outcome p(e; - MMu tunity Loss, MNRT
el: success O O
û2: failure 150 120
EOL (u,) --120 MMU
m
a,: do not build the dam
Probability Opportunity Loss, Weighted Oppor-
Outcome p(e4 )- ndMu tunity LGSS, MMU
el: success 200 50
û2: failure O O
5u
EOL (a,) = 50 MMü
opt EOL = EOL (a,) = 50 MMü
652

Thus, with no additional information, the best decision

would b e not to build the dam. In this way, region Aleph would
forfeit only 50 NIMU, the expected opportunity loss.
Now the Development Authority turns to its Hydrological
Service asking its advice regarding the nature and usefulness
of the additional information which may b e obtained& the cost
of 15 MMU. The attitude of the Hydrological Service is that. by
and large the additional data would yield one of the following
three types of indications regarding the effectiveness of the
reservoir (in terms of streamflow regulation, hydropower gener-
ation, flood contral, etc.):
X1: increase in effectiveness
X2: no change
X3: decrease in effectiveness.
These variables could have been measured only when projects were
constructed, whether successful or not. Thus, the Hydrological
Service had in its records a set of joint probabilities P(X.1)gi)
as follows: J

Table 2.
Joint Probabilities P(X J W ; )
out co1iie
To taï
i' x1 x2 x3
0,: project successful 0.20 0.05 0.05
û2: project unsuccessful

To tal
- -
0.05
15
0.25
0.10
O.
0.55
7

0.60 1 .o0

Of course, the column totals represent the mar inal proba-

bilities of .the usefulness of the additional data: PTX,) = 0.25,
P(X,) = 0.15, P(X3) = 0.60.
The expected value of the information which may be obtained
by the additional hydrological data is reached by means of' a dia-
gram, as shown in Figure 2. The set of probabilities appearing
in the last branches of the decision tree are conditional probabi-
lities p(eiJxj),
 653

The amount of 28.9 MMU appearing at the node (a) in the

decision tree represents the expected opportunity loss if it is
decided to obtain additional hydrological information and if
optimal decisions would be made on the basis of the new data.
Comparingbhis amount with the 50 MMU obtained under Itno additional
data" policy (Table I), it appears that it is worth spending
50.0 -'.2&9 = 21.1 MMU in getting more hydrological information.
The difference between the outcomes of the two policies is called
the expected value of sample information. The expected net gain
of sample information is 21.1- 15 = 6.1 MMU, i.e., the expected
value of the sample information exceeds the costs incurred in
obtaining it. The Development Authority concludes, on the basis
of preposterior analysis, that it is worthwhile to get the addi-
tional hydrological information.
Cost-effectiveness.
Cost-effectiveness is, in fact, engineering economics
[5]. It is concerned with evaluation of a system worth, before
the decision is made to construct the system. Thus cost-effect-
iveness is future oriented, and because of it its mode of
expression is in terms of probabilities and expectations.
With reference to the situation represmted by Figure 1 ,
one can relate cost-effectiveness with the reciprocal of Cost,
i.e., l/(cost). In this way, the lower the cost of a project,
the higher would be its cost-effectiveness (which wculd also be
a measure of its worth).
Now, the cost-effectiveness of a system (as measured by
its worth) increases with the amount of information available
at the time when the system is designed. In other words, this
is a re-statement of the truism that the more we know about the
universe within which we design a system, the better the chances
to produce a good design. The increase in the cost-effectiveness
can then be observed with respect to two major aspects.
(a) Flexibility in plarmin . Because of hydrological,
economic,
gnicag-na the planner of a water
resources system, it is desirable to produce a flexible system.
In this context, byYlexibilitytlis understood one or more of
the following attributes:
654

(i) The possibility of increasing the capacity of the

system by adding additional components of the same kind (e.g.,
pwnp stations in a pipeline network).
(ii) The possibility of altering operating policies so
that the system may respond to a broader range of demands.
(iii) The possibility of modifying the system when the
nature of the demand changes, e.g., when there is a traasition
from irrigation to domestic and industrial uses of water.
(iv) The possibility of constructing the system in
stages, so as to respond to increases in the demand for water.
(b) Reversible vs. irreversible decisions. The design
of a system or of a component is a one-stage decision process:
the size and dimensions are established. If the system is im-
plemented, the design decision may have irreversible effects
upon the emironment, such as the transformation of a canyon
of unique scenic beauty into a man-made lake of doubtful
esthetic value. The decision to delay implementation is
reversible, since it keeps open the alternative to construct the
system. In addition, until the first decision is reversed,
additional information may affect several planning details, ar-d
also technologies may be improved in the interim.
As an example of the introduction of these two aspects
in the planning process, one can indicate thi. Israel Water
Scheme. The planning process was oriented toward increasing
the cost-effectiveness of the system, especially with respect
to flexibility in design and to the reversibility oî decisions [6].
The development of water resources progressed from local ground
water schemes, to regional groundwater projects, finally to the
construction of the major component related to surface water
resources - the National Water Carrier.
Planning and design with inadequate data.
The adequacy of hydrological data as defined by F i g n e 1
represents one aspect of the general problem of the consequences
of inadequate hydrological dataon the cost and effectiveness of
water resources structures and projects. Although this aspect
can be quantified, it is still somewhat mechanistic.
Another aspect would stress the linkage between data
and decisions in the planning process. Although this aspect
 655

also lends itself to'quantification, at least as far as the

data are concerned, it seems that it reflects also the quality
of the ensuing design (and operating) decisions.
It would be perhaps beyond the scope of this paper to
survey the state of the art in the planning and the design of
water resources systems with inadequate data. However, it
would be instructive to mention two of the more recent contri-
butions to this problem: one dealing primarily with surface
water, and another related to groundwater.
Wallis and Matalas [7] consider the problem of deter-
mining the capacity of a surface reservoir such that a given level
of demand be satisfied. Observed hydrological data were used to
generate synthetic flow sequences, using two different sequence-
generating mechanisms: (a) a well-known model based on the
Markovian process; (b) a model developed recently [8] which
assumes the process to have a finite memory length M and the
Hurst coefficient h; this is called the filtered type 2 process.
It seems that for streamflow regulation of up to 80$ of the mean
annual flow, the Markovian model may be quite useful for the
determination of the minimum necessary storage; for higher degrees
of streamflow regulation, the filtered type 2 model with h 7 2
should be used.
Maddock [ g ] used mixed integer programming methods for
evolving a planning and management model of a groimd water
development project. The model is oriented toward deterqining
three components of the overall system: (a) least cost operation
of existing wells; (b) least cost spatial and temporal schedule
for installing new weììs; (c) least cost transport system to.
convey the pumped water to a central demand point. The methodo-
logy developed is tested on a hypothetical sample problem in
which ground water development has to satisfy the demands for
water of a town. The concept of expected value of opportunity
loss (similar to the expected opportunity loss encountered in
the preposterior analysis) is used as a measure of how much
the errors inherent in estimating the model parameters will
affect the cost of the project in terms of overdevelopment or
underdevelopment. The results of the analysis indicate that the
reduction of uncertainty for the purpose of decreasing the
expected value of the opportunity loss should be a balanced
activity, i.e., beyond F- given point, further reduction of the
hydrological uncertainty will not improve the decision-making
process unless the economic uncertainty is alao diminished.
656

Concluding; remarks.
The consequences of inadequate hydrological data on the
cost and effectivepeas of water resources structures and pro-
jects were assumed to have a parabolic shape in the data-cost
space. The abscissa of the minimum point of thìs vertical
parabola defines the adequacy of data.
There are several methods for the evaluation of hydro-
logical data with respect to their adequacy for planning. One
such method using the preposterios analysis is presented in
some detail. This method enables the calculation of expected
opportunity loss generated by a program designed to obtain
additional hydrological data, as well as the expected value
of the sample information. If the expected net gain of sample
information is positive, it is an indication that the existing
hydrologic& data are insufficient.
Cost-effectiveness of projects is briefly discussed,
with some emphasis on its aspects regarding the flexibility in
planning and the irreversibilityof some design decisions. As
an example of these asepcts, the Israel Water Scheme illustrates
a planning process oriented toward increasing the cost-effectiveness
of the system.
Fi,nally,how to plan and design water resources systems
with less th&? adequate hydrological data was illustrated by
two examples. In the first example, synthetic hydrology was
used to determine the capacity of the reservoir, but the design
was sensitive to the type of model used to generate t h e synthe-
tic sequence. The second example related to a ground water
development project.
References.,
1. James, II, I.C., Bower, B.T. and latalas, N.C. (1969)
Relative importance of variables in water rescurces planning,
Water Resources Research, 5( 6), pp. 1165-1173.
2. Matalab, N.C. (1968) Optimum gaging station location,
Proceedings, IBM Scientific Computing Symposium, Water and
Air Resource Management, White Plains, N.Y., pp. 85-94.
3. Fiering, P.B. (1965) An optimization scheme for gaging,
Water Resources Research, 1(4), pp. 463-469.
 657

4. Hamburg, M. (1970) Statistical analysis for decision

making, N.ew York, Harcourt, Brace & World.
5. English, J.M. (1968) Cost-effectiveness, New York, Wiley.
6. Buras, N. (1971) Utilization of ground water resources in
Israel, Atti, Convegno Internationale sulle Acque Soter-
ranee, Palermo, pp. 674-680.
7. Wallis, J.R. and Matalas, N.C. (1972) Sensitivity of
reservoir design to the generating mechanism of inflows,
Water Resources Research, 8( 3), pp. 634-641.
8. Mataïas, N.C. and Wallis, J.R. (1971) Statistical pro-
perties of multivariate fractional noise processes, Water
-
Resources Research, 7(6), pp. 1460 1468.
9. Maddock, III, T. (1972) A ground-water planning model
basis for a data collection network, International Symposium
on Uncertainties in Hydrologic and Water Resources Systems,
Tucson, Arizonp, pp. 6.3-1 - 6.3-26.
 658

cost,
Monetary
Units

~~ ~ ~~~

Amount of hydrological data

Figure 1. The data-cost space.
 659

o.20/0
3
R. A

Figure 2.
Decision diagram for preposterior analysis, MMU.
 OPTIMIZATION OF WATER RESOURC ES DEVELOPMENT PROJECTS
Z N CASE OF INARE2UATE HYDROLOGIC VATA.
A. Filetti' ), G. Faank ' 1, C. Pahvuleb eu' ''I
Bucuhebti, Romania
----
I n t h o d u c t i o n .

The phoblemb which have t o be bolved b y hydhaulic

engineeh4 ahe 06 g h e d d i v e a i t q and deeibionb i n theih
dieldb o 6 actiuity o@en imply a conbidehable hen ponb abi-
lity, n o t only i n hebpect t o the economic conbequenceb,
but albo t o the bocial and ecologic eddectb 06 buch de-
cibionb. Being h e e d e d t o t h e mabtehing 06 cehtain natuh-
al phenomena, mobtly huled b y btochabtic l a m , t h e con-
ception as well a¿ t h e opehation o d wateh hebouhceb de-
velopment dthuctuheb depend o n the deghee o d knowledge a-
vailable on natuhal data, ebpecially on thobe helated t o
hydhologic euenth. The hydhologic hecohdb necebbahy t o
hydaaufic engineehs ahc not condined t o data &elated t o
liquid blow, though data 0 6 t h i b type ahe e b b e n a a l doh
theih acLluity, but a h 0 concehn bed-load p h o c e b b e b ,hiveh
and bank dynamics, wintea phenomena etc. Euidently,due t o
the aleatohq chahactes o b m o b t hydhologic occuhenceb,even
long ~recohdb 0 4 p a t hydhologic phenomena cannot oddeh an
abbolute baáety a¿ hegahdb avoiding ehhohb and deviation
@om t h e altehnaZiue which could be phoved ab optimal. lt
,i¿ neuehthelesb unanimoubly accepted that, ab the volume
and quality 06 in6ohmation o n tlecohded hydhologie events
incheas eh, i n conditionb o 6 i t l b cohhect intehphetation,
the phobability 06 cohhect ebtimation od dutute occuhencc
o b hydirologic phenomena inchease4 and, a2 t h e bame t i m e ,
the h h k b abbumed i n taking deCihion6 decaease.
~

'1 Vocto~-Engineeh,ChCed Engineeh o d the Rebeahch and Ve-

bign Inbtitute doh Wateir RebOuhCeb Engineehing.
I') Engineea, Team leadeh at the Inbtitute doh tlydsoetec-
thic SXudied and V e b i g n .
11') Voctak-Engineeh, Section leadeh at the Rebeakeh and
Debign Indtitute doh Wateh Reb ouhceb Engineehing .
66 2

It muAt be undeiraned t h a t the indluence 06 incom-

plete hydirologic data on t h e p o b b i b i e i t y 06 coirirectlg de-
teamining the technological, duncL¿onal and economic pa-
irameteu 06 wateir heb ouirceb development piroject¿ depend¿
i n gheat meabuhe t o the hydirologic chahacteh 06 the iriuez
b a i n , on t h e natuire 06 wateir U b e b , on t h e type 06 bthuC-
tuheb etc. Theaedohe,any opL¿mizaL¿on method mubt be con-
bidehed in the fight od t h e condition4 in which thib me-
thod i b apptied; the bpecidic 4itUafiOnb and the tenden-
cieb i n bolving t h e phoblemb m e n a o n c d in thib papeh must
be looked at only ab typical exampleb.
Undeh conditionb o6 incomplete hydhologic indoirmat-
ion, t h e methodb Ubually appfied ahe no longeir clbe~juland
it i8 necebbahy eitheh t o adapt thebe methodb t o the a-
vailable data oh t o adopt bimpleh phoceduheb which aire
conbibtent with thebe data. In belecting buch methodb,
6oÆlowing itemb mubt be taken into account:
- the methodb m u b t mahe integhal ube 04 the auaila-
ble volume 06 indohrnaL¿on.At the bame L¿me it m u b t
be kept i n mind ithat no phocebbing i d able t o cire-
ate quantitatively new i n ~ o h m a t i o n and thehedoh it
lb ubetebb t o thy genehafing in6ohrnat.ion not con-
tained in t h e basic data;
- bebideh hydhologic data irecoirded i n t h e aiueir ba-
din bubject t o analybib, it i4 pobbible t o take
advantage 06 additional indohmaL¿on irecoirded in
neighbouhing hiveir babinb. Indihect hydirology may
be ubed ebpecially i n ohdeir t o obtain quafitufive
indohmation hegairding the beabona1 oh annual d i b -
thibufion 06 dlow, the pobbibifity od occuirence 06
cehtain hydhologic phenomena i n vaaiou4 pehiodb od
the yeah, etc;
- t h e m e t h o d m u t not lead t o an ampfi6icaLLon od
e h h o u od the hydhologic basic data but,ab much ab
pobbible t o theiir attenuation.
A gheat d i v e u i t y 06 bituationb e x i b t b conceirning a-
vailable hydhologic data, covehing the whole dietd dirom
 663

total lack t o an acceptable volume od in6okmation. Theke-

dohe, tkying t o b e t up cefitain methodb o 6 gcneaat appbic-
ation i b haadly t o be hecommended. I n painciple, it
would be make cokhect t o talk about gkoupb ok typed 06
methodo having a common p ~ n c i p î e . Thebe have t o be adap-
ted t o concirete conditionb and objectiveb 06 each inveb-
tigated pkoject. The methods which can be appLied i n cabe
06 inadequate hydkological data may be ceabbidied ab dol-
LoWb :
- methodb baoed o n the genekation 06 bynthetic h y -
dkologic bequenceb, btakting ('ron1 given btatibtic-
al pazameteu o d hydkologic phenomena [ Monte Cak-
t o methods);
- methodo based on the genekation 06 hydhologic b e -
quenceb b y c o m e l a t i o n with kaindall data;
- method¿ based on the genehalization o 6 kebultb o d
watea he6 oukceo engineeking computation4 ;
- methods based o n t h e theoky od gameb.
Thib kepoht intend6 t o pkebent the wayd i n which
thebe method4 can be a p p a e d ' i n dome key pkoblemb 06
watek ire4 ouaced engineeking .

Dimenbioning 06 kivek blow hegulating

wokkh debigned dok rneefing ~ a t e k
demandb.

Une od the m o b t ubual paobtemb, cohkebponding t o an

incipient p h u e 06 watek keboukceb development, i¿ t o e d -
timate t h e capability 06 unhegulated UVC CM t o meet u b e ~
watek demand.
The bolution od thib type 06 pkoblemb i d based l e b b
on t h e detekmination 06 avekage @ow and dependendb gnea-
tty on tow watehb and o n chakactehibtic minimum valued 06
dibcchakgeb; thebe valued can pkebent a gkeat benditivity
t o t h e quantity 06 available in6okmatiav1, t o the methodb
06 dikect oh indikect detekmination od dtow valued and t o
the degkee t o which b a ~ i cdata have been extkapolated.
Theke6oke, the tack 06 adequate hydkotogic data and
664

pahticulahly 04 b a t i b dactohy hydhomethic hecohdb can be,

i n t h h c u e , t h e bouhce od impohtant ehhoh4. The w e 06
cohhelative methoda d6 hibky, as the helative valueb 06
low w a t e ~depend on t h e individual bupply 06 each wateh
c o u u e . On the otheh hand, the u4e o 6 bimulated 6low b e -
qU12MCe4, w u a l l y applied doh genehating mean monthly
d i b chahgeb, cannot be batib dactohy, bince daily valueb
o d minimum dibchahge can conóidehably did6eh dhOm mean
monthly valueb.
The dub-unitahy Ratio 06 minimum daily dibchahge t o
mean monthly dibchahge i4 w u a l l y bmalleh id the phoba-
bility o 6 exceeding t h e dibchahge i n c h e a e b and the buh-
dace od the hiuek basin d e c h e a e b .
The e h h o M which can be made in buch cabe4 can be
one 06 the dollowing:
- oueh-evalua-ting available low dlow,which can lead
t o a bmalleh phobability od being able t o meet
wateh demand o d u e ~ thib ; phobabifity m i g h t not
be acceptable becaube 06 t h e e x c e b b i v e l o b b e b due
t o dhequency and b e v e h i t y od wateh bhohtage. It
i4 Wohth mentionning that i n bome counthieb (as in
the U.S.S.R. ,Czechoblouahia, Romania and othehs 1
the phobabifity 04 being able to meet demand A
ebtablibhed b y btandahdb and A ,thehedohe, compul-
b ohy;
- undeheualuating available low dlow,which can lead
t o a da&e conclubion, that t h e hiVeh i4 not able
t o meet demand without blow hegutation and that
a stohage hCbehU0ih O h a diuehbion dhom otheh ex-
cedentahy h i u e u h a t o be budX.Thib would i m p l y
u4eleb4 expenbeb Oh, i n the bebt C a 4 e , U b e l Q b b im-
mobifibation o d capital.
A b wateh demand ghowb in compahibon t o available
wakeh hebOUhCQ6, t h e neCCbbahY deghee o d blow hegula-tion
i n c h e u e b and mutationb emehge concehning t h e bignidic-
ance 06 v a h i o u categohieh od hydhologic indohmation. In
thib benbe, the data helated t o t h e auehage inalow doh
 665

longeh L¿me pe&¿odb: monthb, beabonb, yeah4 oh even b e -

quenceb ad y e a u begin t o play an ebbential pmt i n de-
tehmining the pahametehb 06 btohage hequihed.
The indtuence 04 genehat hydhologic data on t h e va-
tue 06 thebe puhameteu i4 heuealed b y the geneaal phac-
-Lice 0 4 wateh hebouhceb engineehb a4 well a4 b y home
Apecial hebeahch phoghamb. Thib inbluenee i A made evid-
ent b y invehtigating :
- t h e genehat hetations between the chahacteaibtic
blow panameteu and t h e bpecidic deghee 06 deue-
topment 06 wateh hebouhceb;
- t h e ben&¿tiVitLj od hebUltb concehn.¿ng blow hegut-
d i o n at vahiou4 degheeb 06 apphoximation od t h e
inadequate hydhologic data.
Thub, a diut aspect 06 thib anatybib concehnb ihe
cohhelaXion b e m e e n the main puhametehb chahactehib.tiC
doh dtow dL¿thLbution: t h e vahiation coeddicient Cu, 2he
coeddicient 06 bkewnebb Cb and t h e coe6dicien.t o d sehial
cohhelation ir on one hand and the net volume od necebba-
hy ótohage hebehV0iJr.b and theih opehating policieh on
the otheh.
Vahiou4 sepohth conceaning genehatized J a b ultb on
the connection befween thebe p a t a m e t e u and the magnit-
ude 06 the ove&-annual component 0 4 the nequihed btohage
doh a given bade yield have been pubfibhed 111, 1 2 1 .
R e b u l a 06 buch hebeahch i b beAt made evident b y
diagrramb phebenting the c o m e l a t i o n between t h e coedtjic-
i e n a a, p and B, 06 which an example i b shown in diguhe
i. Following bymboeb have been uded :
- a, the blow aegulation coeddicient 04 the deve-
lopment, deiined ab t h e ha.t.io 06 t h e bade yield
od the dtohage hebehvoih t o the auehage dibchahge
06 t h e wateh C O U I L A ~;
- p id the phobabieity od meeLing wateh demand, de-
dined ab the &mit 06 t h e hatio 06 t h e numbeh 06
y e a u i n which no wateh bhohtage a p p e a u t o the
total numbeh 06 y e a u invebtigated;
- 0 c6 t h e btohage coeddicient, dedined ah the hu-
666

tio 06 the oveh-annual component 06 Atohage t o

the avesage yeaaly dibchahge 06 the hegulated ti-
Veh.
lt m U b t be k e p t in mind that the bame avehage dib-
chahge 2 could occuh i n vahioub beqUQnceb 06 bingutarr
dibchahgeb 06 t h e hivet; thib can be made evident b y an
a n a l y b d 06 hecohded beqUenCeb 06 dibchahgeb 06 hivehb
phebenting vahioub Valued 06 the p a h a m e t e u cv,Cb and 4.
Folîowing c o n c ~ u b i o n b may be dhawn dhom hebeahch deveto-
ped i n thib 6ield:
- the v a t i a d o n coed{i&ent Cv ha4 a dihect inblu-
ence on t h e volume 06 t h e btohage hebehvoih, the
Atohage coe6bicient B being the gheateh, the m o u
the vahiation coe6,jicien.t incheabeb ; genehaîly
the helaAive ghowthb 06 ß ahe gheateh, sometimes
even benbibly g h e d e k , than the gaowth 06 Cv, i6
the valuta 06 a a m high. On the con;titahy,6oh low
valued 06 a t h e gaowth 06 the necebbahy volume 06
btoaage in l e b b hapid than the ghowth 06 the va-
hiaiSon coe66icient [6iguhe 2 ) ;
- t h e coe{,jicient 06 bkewnebb c b ha an i n v e m e in-
ence ce on t h e volume 06 the btohage hebehvoih;
Zhu, i6 c b incheabeb the volume decheaseb coh-
hebpondingly. Genetally, pehcentual heduciSonb 06
B avre Amalleh than the pehcentual ghowthb 06 Ch;
- t h e coedbicient 06 betial cohhelation h has a di-
heet in6luence on t h e value 06 t h e btoaage; thub,
the incheabing 06 h leadb t o ghowthb 0 4 B, peh-
centual ghowthb 06 both parrametem being 06 the
bame ohdeh 06 magnitude.
Similah conbidehationb can a d o be made i n connect-
i o n t o the beabonal component l y e m l y component) 06 the
h e q d h e d btohage. Evidently, i n thib cabe, t h e coe66ic-
Leni2 06 vatiation, o 6 bhewnebb and 06 behiat cohhelat-
ion mubt be based on daily, decadal oh monthly avehage
dib chahgeb .
A beeond inteaesting aspect concehnb t h e inbluenee
06 t h e intehval taken in account 60k debign iday,decade,
 667

month etc) oh t h e Lime intehval doh which hydhologic da-

ta ahe ebtimated and wateh balance computations ahe un-
dehtaken on the hequihed btohage volume.
Vehy 06ten,waZeh balance id ebtablibhed on a month-
ly bad&, taking into account a bequence 06 mean monthly
dlowb COVehing a pehiod 06 ¿evehat qeah4, doh which he-
cohded oh indihectly detehmined hydhologic d d a ahe a-
vailable. Thib way 06 dealing with the phoblem imptieb
t h e assumpLion that the d h c h a h g e od the Jbiveh and t h e
demand o{ t h e u e h ahe baihey conbtant duhing a month.
Thib asbumption i b neveh abbolutely cohhect doh the di&-
chahge o d the hive&, nOh b o m e t i m e b {oh the Watch demand.
ZnvutigaLionb undehtaken i n thib 6ieÆd bhowed that
i n cehtain cabe6 the ,time pehiod taken into account hab
a gheat indluence on t h e hebultb obtained concehning t h e
volume 06 hequihed btOhUge.lt has been pobbible t o u t a -
btibh cohheîationb b e m e e n btohageb cohhebponding t o
,time pehiodb od a month oh a day ubed i n wateh balance
calculations. An example 06 buch an intehdependence i b
bhown i n biguhe 2.The genehat conclubion od the tedeahch
i& tha.t bhoht pehiodb, o6 a day, mubt be ubed 46 badie
a m e intehvaÆ only i6 t h e heqLUhed btohage,hebulted dhom
p h e a m i n a h y computation4 ubing monthly UVehUge valueb c4
Amall. foh gheateh 4tOhage volume4,the indluence 06 bhoht
Lime p e d o & i d negtigeable.
1 6 t h e invebtigation 06 lahge-bcale phoject.4 i4 un-
dehtaken, u6e od 6 yntheLic stheam- {low bequenceb had
btahted t o impobe itbet6 even in c a e 4 in which t h e vo-
lume 06 availabÆe hydhologic indohmation would have been
conbidehed adequate. Stheam-dlOW genehation methodb,
oh Monte Cahlo techniqueb, btah-t dohm c e h t d n comphehen-
hive hydhologic PaharneteM buch ab avehage dibchahge,
coeddicient 06 VahiatiOn, coe66icien.t 06 behiat cohhel-
ation etc, ebtimated on the b a b i b od a minimum 06 dihect
hecohdb oh b y genchalizing hebultb o d hydhologic inveb-
LigatiOnb i n bimilah ( M e u . Ahtidici& time behieb o 6
h u n d h e d and even thouband4 od hydhologic y e a u ahe g e -
nehated; thebe include a multitude 06 pobbible bequenceb
66 8

06 day, w e t and aveirage y e m b , which condeh a high he-

liability on the hebultb 06 watch balance and btohage
calculationb e

Such methodb, babed on bynthea2c btneam-6low b e -

quenceb, a m applied on a lahge ¿cale i n t h e U.S.S.R.
l4I,l5/, the U.S.A. 1 6 1 and otheh counthieb. ln Romania,
1 1 0 1 thib method .d being applied doh the btudy o6 the
development 06 lahge hive& babinh.
A hecent hebeahch 1 3 1 i b conceancd with the inilu-
ence 06 the length 06 the heal hecohd on t h e hequihed
stotrage,in 06 a seasonal oh y e u l y dlow hegulation.
The teseahch phogham covehed a gheat vahiety od bituat-
ionh, including di66ehent typeb 06 dlow dibthibufion
and bevehal VatUeb 06 berrial cohhelaZLon and btoirage
coeddicienfi.
The hebultb o 6 thih hebeahch Lead t o t h e conclub-
ion ( b e e diguhea 3 and 4 ) that t h e hequihed btohage in-
cheabeb with the length 06 the bequence. The inchease . &
dabteh doil higheh valued 06 the behial cohhelation and
stohage cae66icienh.
Though thib invehtigation w a ~concetned only with
yeahly (low rregulation, the h e b u l a obtained doh high
valueb o 6 the btohage coeddicient (maximum 1.0) phove
that the length o d the dihect hecohd and the bequence
06 w e t and day p e h i o d ~ have a bignidicant indluence al-
h o in t h e cabe 06 oveh-annual atohage.
A bthiking example in this b e n ~ ewas u p h e d e n t e d b y
the evolution od btohage hequihed 60s the wateh bupply
06 a big i n d u t h i a l plant i n Romania. Wateh balance cal-
culationb, 6ihbt undextaken i n the eahly 1960-ieb Wehe
based on dihect dlow hecohdb covehing only a pehiod od
15 yeah6, b t m f i n g @om 1947148. Extending thib bequence
b y cohhelation with hecohdb i n neighbouhing b a b i n b ge-
nehated, intek alia, an exthemely dhy hydhoîogic yeair
( / 9 4 2 / 4 3 ) .7 6 thib yeah wab included in the hecohd u e d
a~ b a b i b doh htonage calculationb, t h e hequihed btoaage
wab neahly the double ( 2 2 0 million cubic m e t e u ) 06 the
btoxage which would have deemed necebbahy i6 thib yeah
 669

had not been included into the hecohd ( 1 2 0 million cub-

ic m e t e u ) . A b a matteh 06 dact, the additionat hecohdb
06 t h e 6ottowing ten y e a u beem t o indihm t h e indihect
data obtained doh the yeah 1 9 4 2 1 4 3 ; it wad thehedohe de-
cided not to include t h i b yeah into the invebtigated b e -
quence when btohage caÆcutaL¿onb wehe again undehtaken
on the basi4 06 a .tongeh dcquence 06 dihect hecohdb.
In buch bituationb ,behideb the ub ual methodb 06 ex-
tending t h e hydhotogic time b e h i e b b y bimilahity with
otheh hiveh badinb, modehn techniques can be applied doh
detehmining the optimum decndionb. A botution might be
dound i6 the theohy 06 game4 i d applied; the hecommended
de&ion lead& i n t h i b c a d e t o minimum heghet, taking
i n t o account on one had t h e c o b t b ad t h e btohage h u e h -
voi& and on the otheh hand the pobbibte damageb. Thib i b
a cla6bica.î cade 06 a game againbt natuhe. Natuke’b se-
action i b not indluenced b y t h e phevioub d e c h i o n 4 con-
cehning the management o d wateh hebouhceb and i b eitheh
aleatohy oh huled b y an asbumed pkobabilibtic law 06
dib t&ibutio n .
06 c o u u e , t h e methodb based on the theohy 04 gameb
do not bolve t h e inadequacy 0 6 hyditotogic data. Ubing
them maheb howeveh the minimization 06 a d v e u e e66ec.tb
06 inadequate data po4bible. Theh~,joke, buch methods
bhould not be phebented in o p p o b i f i o n t o thobe based on
t h e genehation 06 new data oh on t h e genehatizaaSon o6
cehfain heb uttb od watek heb ouhceb engineehing calculat-
ionb. Both technique4 can be bimultaneoubly applied.
One 06 t h e advantageb o 6 the theoky 06 gameb i b the
pobbibifity 06 taking into account any inadequate data,
not only od hydhologic chairacte& (do& example, data &e-
dcilning t o t h e dUtUhQ development o{ watek u e ~ ) .
Noticing t h e ub e o 6 cohhelaLionb between hydirologic
p a h a m e t e u and Othe& chahacteJùbtic elemenX.4 06 the hi-
ve& basin: aveaage haindall, altitude, adtjoheAta,tion,
b t o p e etc. bome e n g i n e e u thied t o ebtablihh genehatized
helaZionbhipb be&een wateir heb ouhceb engineehing paha-
m e t e u , buch ah t h e kequiheb btohage, and geo-meteoholo-
g i c e l e m e n a . Such hebeahch has been cashied out in Po-
670

land. Methodb 06 t h i b type have, nevehthelebb, a limited

appficability, geneaalizationb being pobbible only at a
&regional beale. They may, evidently, give a view on the
b i z e 06 necebbahy W d e h hebOUhCeib development W O h k b and
can be 06 help doh the phefiminahy debign 06 ceht&n
bmall btohage damb. Theih ube, without a camparribon with
othes m o m elaboaate methodb i d howeveh n o t t o be hecorn-
mended i n t h e inwedtigation 06 impohtant btohage. hebeh-
VOihb.

Uimenbioning o d hydhoelecthic Noirkb.

The utilization 06 wateh poweh i d , evidently, hela-

t e d t o the cohhect knowledge 06 the natuhal potential
and o6 the conditionb 06 developing it. in t h i b c o n n e d -
ion, hydhologic data ahe necebbahy not only in ohdes t o
detekmine t h e genehal e66ect and edbiciency 06 poweh
plana2 but albo t o e b t a b l h h t h e development bcheme, the
enehgetic parrameteu and t h e charractexibtics 06 the
bthuctuhes. Fah p l a n a luith no btohage oh having a low
deghee 0 4 6low heguldtion a6 well c ~ 6 {oh t h e intake4 06
becondahy watch divehhionb it can be parrticulahly impoh-
t a n t t o ebtimate c ~ h h e c t l y t h e daily, and dometimed even
the momentarry blow hégime. Condthuction 06 irégime and
dlow dukation cuhveb , chahacte&ibtlc doh kelatively long
pehiod6,may be necebbahy. Such cuhueb alre uded in ohdeh
t o ebtablibh t h e deghee 0 6 utilization 06 the avehage
yeahly dh chahge and t o detekmine the inbtalled capacit-
ieb
The value4 06 the hydhologic p a h a m e t e u mentionned
i n the phevioub chapteh: auehage d.ib chahge, coe66icient
06 vahiation, coe6dicien.t 06 bkewnebb and coe6dicient 06
behial c o m e l a t i o n , ahe necebbahy i n ohdeh t o ebtablibh
t h e inbluenee 06 v a d o u b Atohage volumed o n the magnit-
ude and quality 06 electtic p o ~ e hphoduction and on the
enehgetic conditions 0 6 btohage hydhoelectaic plan&. O n
thib bahib, t h e opfimization o 6 the volume 06 the btoh-
age h e b e h v ~ i k b cb a h o podbible.
 FOR. mote advanced dlow conthol, t h e avehage multian-
nual dibchakge i6 t h e hydaologic etement whobe indluence
o n t h e economic eddiciency 06 hydhoelecthic plant6 i4
gheatest. In paht, thi6 i b due al60 t o the opehation 06
hydhoeÆecthic plana2 within 6taong poweh dydtemb ,genehal-
ly with inteknational l i n k d . I n buch 6Ybtem4, t h e e66ect
o 6 individual hydaologic 6ituationd i6 attenuated and the
whole hydhoelecthic poweh available can actually be uded
in the dybtem, even id only doh the dilling o6 the hebeh-
voim 06 pumped 6tohage powek p l a n a .
Re6eahch on the indluence 06 the length 06 t h e he-
coird on t h e value 06 the aveirage dibchahge hevealed that,
604 m o ~ teutopean wateh c o u h 6 e 6 , time behie6 covehing a
dequence 06 30 y e a m ahe u6ually batihbactohy. Conclub-
iond Reached at i n inve6tigating data concehning the a-
veaage d i d c h a q e doh Recoaded Lime behie6 06 vahiou6
length6 at the OIL6ova gauge on t h e Danube can be bahen
a6 an example. O n the basib od 133 yeah long hecoad6
(1838 - 19701 the avehage value 604 di66ehent 6 4 . ~ e b 06
a given Length, covehing 10 - 40 con6ecuL¿ve y e a u wehe
calculated and the highebt and Lowe62 value6 o6 thehe a-
vehage6 wehe examined. T h e irebutRb alre phedented in t h e
dollowing table:

10 124 6090 4520 12.5 -16.5 0.36

15 119 5910 4760 9.2 -12.0 0.28
20 114 5770 4850 6.7 -10.2 0.22
25 109 5660 5040 4.6 -6.8 0.17
30 104 5670 5200 4.1 -3.9 0.13
40 94 5650 5230 4.4 -3.3 0.11

Thehedohe, i6 t h e hydhotogic data ahe n o t adequate

doh the de6ign 06 hydhoelecthic developemntb, a6 a huÆe,
t h e b e data dhoutd be extended t o a 2 5 - 3 5 yeah6 long time
672

behie, b y analogy with othek watek couhdeb ok with kain-

ball. longeh extenhionb 06 hecohdb b y bimulation ate
veky kcvrely applied.

Debign and opehation 06 blood

conakol developmenth .

Flood canx7ca.t developmen;td ake kelated not only t o

economic bene6itb but aedo t o the becuhity 0 6 bocial
Lide i n wide axeab. At the same time, ab álood conthol
bthuctuheb ane debigned t o {ace hydhologic bituatioptb
exceeding extheme hibtohicaî hecohdb, the ube 06 cxthe-
mely accuhate data i b , in phinciple, necebbahy.
A ptoblem ahibing most dhequently i d helated t o the
debign 06 outle& o6 6low hegulating bthuctuheb. Othe&
impoatant phoblemb, buch ab the debign ,od bhidgeb oh 06
othes btkuctuheb chobbing hivetb, 06 dykeb and 06 blood
detenfion hehehvoitb axe aedo helated t o dlood hydhol-
OgY *
Except vehy hahe cab eb, debign hydhologic condit-
ionb have nevek been hecohded and have t o be e b t a b l h h e d
b y bpecLal computationb. In thih hebpect, ,two tendencieb
may be kemairhed:
-
t h e extkapolation 06 phobabifity dibtkibufion
cuhvcb 06 maximum aecokded ,$toodb ubing mathemat-
ical btatibticat methodb; thib extkapolation has
t o keach imposed debign paobabilitiea . Othe& me-
thodb, ubed i n o.theh hydhologic pkoblemb 6011 the
extension 06 tecohded fime bekieb ,buch ab bynthe-
fie blow genekation,ubing Monte-Cakto techniqueb.
M Q applied at a bah amalleh bcale bok 6lood con-
ak0l;
- the g e n e h a ~ o n06valueb 0 6 maximum dibchakgeb
and 0 6 ,$laod waveb b y hain6altlhun-od6 cohhelat-
ion; thib apphoach thiea t o compenbate the lack
06 hydhologic data b y uning hain{all oh Othe& me-
theohologic magnituded doh which, UA ually, longea
hecohdb ake available. l n theih dimplest 60hm,
 673

t h e b e methodb hephe4enZ ha.in6attlhun-od6 depen-

dencie6 which have been w e d doh a long time i n
hydhologic indihect btudie6 .Duhing t h e l u t yeah6
thebe methodb have been extended, 6ohmeh 6impte
dependencieb being developed i n t o phy6ioghaphic
haintjalllhun-066 mode&. Thebe mode& have been
ubed ebpecially in the U.S.A. and in Fhance.Theih
eidiciency wab pkoved e6pecially doh t h e htudy 06
dlood conthot phoblem6 .
The majoh did6iculty h a A e d b y applying genetic mo-
de& conbi6t.h i n a pheviou6 detehmknation 06 the pahame-
t e u 06 each model. FOh thi6 pukpo6c,aecohded dl00d6 ahe
houted thhough the model. Theiredom, u b e o 6 h&n6all/
hun-066 m o d e h implie6 the e x h t e n c e 06 a minimum o 6 hy-
dhologic hecohdb; it i b , howeveh, ju6t data on dlood6
which ahe dhequently mi6bing @om hecohd6.
l n the abbence 06 adequate hydhologic data, oveh6i-
zing m o s t d t o o d conthol 6thuctuhe6 i6 t o be hecommended,
even ifj it’6 c o n ~ e q u e n c e i6 a ubete66 i n v e ~ t m e n to 6 ca-
pital, i n ohdeir t o avoid t h e h i b h 06 ovehtopping and 06
eventual bucceeding 6ailuhe o d 6thuctuhQ6 181. Thi6 po-
&cy h juhtidied b y t h e exponential ghowth 06 the den6-
ity 06 economic and 6ocia.t objecaXve6 placed i n t h e a m a
in Which d t o o d COnthOt ~ 6 6 e ~ t06b t h e b e bthUCtUhC6 i6
he6ented. Eventual damage6 due t o exceedence 06 de6ign
pahameteh6 will Zhu6 be exponentially incheahed in com-
pahibon t o actual damage6 cohhehponding t o the 6ame
Stood.
Thub, adteh t h e hydirologic excedentahg peiriod 06
the la62 yeah6 and pahticu.tah.ty aóteh the 7970 and 7972
iloodh, the conclubion wa6 dhawn in Romania that 6lood
detention volume6 hebehved i n 6tohage lahel, which, a6 a
hule, Wehe phteViOU6ly 06 t h e ûhdeh 06 b - 1 2 % 06 t h e a-
vehage annual blow a m knbuddicient; incireahing the6e
volume6 i n butuhe up t o 20 8 06 the aveaage annual dlow
i4 hecommended. Except 4 ome i6 olated ea6 e6, the execut-
ion 06 hubrneuible dyke6 has been completely abandoned
bivice 1960.
674

In t h e debign 06 dyke6, the gkowth 06 blood levea

due t o eliminating the natukal deterifion o6 Blood in
the dtood plain and t o kiVeh bed dynamic¿, pahticulahlg
the haióing od t h e base 0 4 ehobion, obbehved on many wu-
tek couhbeb 06 t h e plain hegiOMb. At the tail watekd 06
cehtain btohage kebehvoihb located i n countkied with d e -
v e m climate, inadequate knowledge 06 wintek phenornena
may lead t o undehebtimafing t h e backwatek phoduced b y
t h e mas4 06 ice and t o undokbeen dlooding, .i6 no phe-
cauLion4 ate taken.
Finally, bok t h e cased whehe data on the genebis 06
6loodb ahe lacking, it i4 pobbible to apply methodb od
the theohy 06 gamed. Such methodb have been applied in
Romania 604 bevekal p k o j e c h 171.

Some aspecÁ2 o 6 multi-puhpob e

wutek ked ouhceb development.

One 06 t h e chahac$ehib.i.ics 06 the contempokahy wa-

t e k he6 Oukceb engineehing c o n b i b t b i n the multipuhpob e
and comphehen6ive development 06 h i W h b u i n b , within
an ebtablibhed development bcheme. V a G o u b pko jectb ake
phomoted a¿ bepahate deÜeÆopment btages 06 t h e genehat
bcheme, all phojecÁ2 being based on unitaky methodolog-
ieb and being conceived b o as t o meet the kequikemenÁ2
06 d l wateh ubeIL4 u well as 0 4 the COnthOl 06 deb-
thucfive e d d e c h 06 wateh. The btep b y step development
06 a hiveh basin i4 impohtant t o the phoblem dibcubbed
i n thib hepoht because 06 it'd pobitive conbequenceb he-
dlected i n the pobbibility 06 cokhecfing emou, 06 he-
ducing exaggehated h i b k b and even 06 attenuafing bail-
ukeb b y t h e way 06 conceiving Bututre p h O j U Á 2 developed
in t h e bame hiveh basin. Thió i4 helated t o the 6act
that t h e unitahg management 06 t h e watek kebouhceb o b
a hiveh basin cheateb a l e b b 4thLc.t dependence 06 each
phoject on t h e dibffkibufion 06 the dlow 06 t h e main hi-
veh among vahiou4 thibutahieb and heduceb the benbitiv-
ity t o e m o h b i n evaluating hydhaulic hCbOUhCe6 i n each
 675

~ p e c i b i cbite. Thib i.4 impohtant, ab global ebtimation

od the hCAOUhCeA 06 a kiveh b a i n i4 UbUal.& l e A b hub-
ject t o ehhohb than the e b a m a t i o n od the heAouhceb 06
the thibutahieb. On the otheh hand,the btep b y b t e p de-
velopment o 6 bthuctuheA doh wateir heb ouhced management
makes changed i n the p a h a m e t e a o d ultehioh phojectb
pobbible, A o ab t o COhJLect ~ 4 6 e c t b 06 Oveh oh undeh-
hizing the 4-thuctuheA built at phevioub development
A tag e b .
In the c a e 06 pkojecth doh which decibionb ahe
made undeh condifionb od inadequate basic data, inveb-
tigai2on 06 the p o b b i b i f i t i e b 06 butuhe sxtenbion o d
b h ’ i U C ~ U R e b d od gheat crteirebt.Such podbibilitied eli-
minate the necedbity 06 immobilizing capital doh the
development 06 initially o v e h ~ i z cd! bthuctuheb, deaigned
in thib way i n ohdeh not t o loo¿e the pobbibifitieb 06
an advantageou dite.
EhhohA committed due t o inadequate hydhologic
data ahC not fimited t o the design btage o 6 hy-
dhaufic AthUCtUheb. FOh the conception o d buch b t h U C t -
uheb t h e lack 06 dihecz hydhologic data can obten not
be avoided. 7 6 t h e implementation o d an adequate hydho-
meteOhOlOgic 6ohecabting nekwohk, including t h e meaAuh-
ing, thanhmibbion and data p h 0 C c h b i n g equipment d ad-
vibable i n Ohdeh t o Opehate a C Q h t L n hebChVOih, .¿A pu-
hely an economic phoblem. In t h i b benbe, the technical
oh economic analyAib 06 opehating conditionb 06 vahiOUA
lahge Acate p k o j e c h and the indluence 06 a good dohe-
ca6i2ng aybtem on theiir opehation Achedule4 lead inva-
tiably t o the conclubion that thebe Aybtemb ahc pahti-
culahly e 4 bicient.
Thib L¿ not only the cabe whehe hydhaufic Aybtemb
debigned 604 dtood canthot oh doh bade yield ahe ad-
dected b y t h e lack o d adequate indohmation oh dohe-
cabzb at Auch exteint, that theih opehation accohding
t o b chedule d phactically impohbible without imphov-
cng the indoirmational bybtem. The advantage i d evident
676

a d o when a betteh knowledge 06 t h e phobable pahame-

tehb o 6 butuhe hydhologic even22 leadb only t o opehat-
ional imphovemenX.6. Thib cb, doh inbtance, the cabe 06
hydkoelectk-ic p l a n a .
A n i l l w t h a t i v e example 06 an in6OhmatiOnal and
donecabX.ln9 netwohk concehnb khe lhon Gate¿ hydhoelec-
thiC plant o n the Danube. T h e wateh level 06 the bfoh-
age hcbehVOih at the dam vahiable, opehating bchedu-
&A phoviding doa an ab conbtant ah pobbible wateh le-
vel &the tail 06 the Atohage hebChVOih, Upbtheani 06
t h e IhOn Gates gohgeb. Thib policy cb due t o the con-
cenakation i n thib ahea o 6 bthuctuheb debigned doh t h e
photection 06 hipak-ian land and otheh development¿,
bthuctuheb which would be oventopped at higheh leve&.
Conbcquently, the hydhoelecthic plant bhould hegul&e
the wateh level & the dam bon an expected inblow, b o
UA to obtain t h e maximum u.ti.lizable head, t o avoid, U
much ah pobbible, the 0vehhpit.ling o 6 Waxeh and,& t h e
b a m e time, t o avoid the exceedence 06 the badety level
in the photected aheu. Vue t o a good dohecabtin9 net-
wohb on t h e Danube and on the main thibttahiebp up-
btlream o6 t h e phojecf, it w u pobbible,even duk-ing t h e
dihbt W O yeahb od opehaL¿ng, t o hegutate the daily
powea genetation i n buch mannet t h a t t h e deviaLion
6hOm the theohea2ca.t optimu did not exceed 1.2 %.
T h e utility 06 the exPendituaeh intended t o b e t
up and maintain an adequate 6onecasLing bybtem, pahti-
cutaaly within hiveh bain6 whehe dlood deten.tion he-
b eh
oiha ahe located has not any mohe t o be demonstha-
ted. Situation¿ may be met i n which an inadequate ope-
hation od OUtkkX.6 dhom Atohage hebehVOihb clln OvehpObe
nohmally buccebbive blood waved, leading t o an aggna-
vakion 06 the bituation which would have occuhed in
u n h e g u l d e d b&eamb.
 677

C o n c t u b i o nb.

The accukacy o6 the dolutionb 06 debign and opekat-

ion 06 hydkauîic wokkb d e p e n d not only on hydhologic
in6akmation, bu.2 albo on the degtee o 6 cokkect ebtimat-
i o n 05 a multitude 06 otheir 6 a c t o u , 60ir inbtance the e-
conomic and conjectukal condixXonb ,the wateir demand etc.
Thebe iactok obten aire much m o u uncektain than hydholo-
gic e v e n h . i n thib context, i n c a e 06 ceirtain ubeb 06
hydkautic bthuctukeb, Auch ab hydkoelecakic powea pko-
ducfion, watea tkanópokt, low dlow k e g u l a a o n boa watek
u e b , the hydhological in6oirmatian bhouLd be consideked
in the hame way a6 othen uncektain basic data, the qua-
îity 06 which h a a g t o b a î inbluenee o n the p04bibiti-
ty 06 o p ~ m i z i n g bOLUtiOn4. in buch cabeb, the necebbaky
accukacy 06 hydiroLogical data mubt be Looked at in cok-
helaZion with the accukacy o 6 othek e l e m e n h kelevant t o
the decibion. T h e m ake howevea alho othek typed 06 hy-
dkaulic b&uc.twreb,buch a6 thobe debigned 60k dlood con-
thol, 60ir which global analybib 06 accukacy 06 basic da-
t a i~ t e b b impoatant and hydkologic data have t o be ta-
hen bepaately into account.

R E F E R E N C E

1.- PLESHKOV, 1.F. Reguliirovanija kechnogo btoba -

GWdkometeoizdat, Uobkow, I96 1.
2.- DYCK,S.; SCHRAMM,M. Stochabtibche Methoden 6Ük die
Bemebb ung deb W a s b eupeichek-
haumeb. - Mitteilungen deb in-
cha@,
A Z i f U t e b {Ük W a b eir~ikt.6
Nk.28, BekÆLn, 1968.
3.- STEGARUZU, P. CokectiiLe d e debite zilnice i n
calcuîeÆe de gobpodairike a ape-
loir. - Hidkotehnica, Nk. 111972.
4.- SVANIDZE, G.G. û b n o v y k a chety keguîikovanija
aechnogo btoka metodom i\lonte -
Kaalo.- Tbiîibbi, L964.
67u

5. L’EZNlKOVSKll, A.A. Vodnoenehgetichebhie ha6 chety

metodom Monte-Kahlo - Enehgija,
idobbow, 1969.
6 . F l E R l N G , M.B. -
Stheamdtow Synthedib . MacMil-
lan, London-Melbouhne, 1967.
7. - VORVEA, A. ;F l LOTT i, A. PhObleme de gOdpOdahihe a ape-
loh CU aplicatic? la bazinul
Bahlu&. -1nbtitutul penthu Pla-
nuhi de Amenajahe b i Consthue-
tii Hidtotehnice ílPACHl.15 ani
d e activitate. Bucutebti, 7 9 6 8 ,
pp. 8 5- 96.
8.- FILOTTZ, A. D i b cubbiòn deb h a p p o h h conceh-
nant la photcefion de4 eaux en
ea6 de cûtabthûpheb. -GeWabbeh-
bchutz i m Katasttophen~all.Sym-
pobnum v o m 2 3 - 26 Ohtobeh i n
Flohenz. Födehafion Euhopäib cheh
Gewabbehbchutz, Vol. 15, Ziihich,
1969, pp- 9 2 - 96.
-
9. TEODORESCU, 1. Gobpodahihea Apeloh. - Ceheb,
FILOTTI ,A.; C H l R l A C V.%ucuhebti, 19 73.
10. -S?MON, A ; V l LAN. A GenehcZhea bihUhil0h hidhalogice
daha aUtOcOhe&ltie. - Studii de
Economia Apeloh, Vol. 1. lnbti-
tutul de Studii b i C e h c e t a ~
p e n a u Zmbunatatihi Funciahe
b i Gobpodahihea Apeloh, BuCu-
hebti, 1971, pp.311 - 366.
 679

ß , 1 ß ß
20 28

IO LO 1.0 10

C" cv C" CV
0.5 as 0.5 0.5

.
I5
-
[ r = 4 5 ;c,=2c,
ß
1 ß

ZD 20 2.o

10 ZO W

. .

0,s
CV CiI C" cv
0;s

, ß 'n ßr ia I I I

2.o 2.0 2A)

LO %O

C" cv CV cv

Fig.1.
n
6 80
 681

O /O0

pigc4. Average required etorage a8 a function

of record,for different degrees of
regulation8
c=0,5 I r - 0.3
 ItRELATIONS BETWEEN PROJECT ECONOMICS AND HYDROLOGICAL DATA"

by
A.Pobedimsky
Economic Commission for Europe

Introduction

In e::ce;j:;, water ilhich is very often considered as a source o;" wealth, m y

cause considerable damage or äisaster and impose a heavy burden on a country's
economy. sorn&jimez it may affect a g o u p of borderine countries (as for example,
those locate6 in the Danube and Rhine river basins).
The accelzrating rate of population eowth in the EC3 countries and the
economic proLpcss in Lechnologicrì changes, during recent years, causing the
depletion of natural resources have all rapidly increased the importance of
water resources development which, curing the last few decades, hac, become one
of the doninnting factors in the national economy of most stoLintries. All this
has obliged countries to improve their water resources nanagement so as to achieve
proper flexibility a d effectiveness corresponding to modern requirerncnts of national
economies.
Mater Phnagement, deals o-lher things with a verj importent component -
hydrological data which define the available water resources to me& national or
regional demnds.
The mtcliing of the bdznce or water resources md.neecls, as vel1 as the
planning and implementation of eppropriatc masures to provide the nececjsa.0 liater
cupply for a region, have become &ieslionr of hi& priority,
The belances oi' water resources and needs mentioned above, which serve to
elaborate the measures io be taken to avoid ne,r;etivec.onsequencesfor ths populaLion
and the regional economy, are now Peing used as a effective tool in m q v LC3
countries. The first internationa!. l4anue.l for the compilation of these balances,
now beiny: completed by the ZCg Conrmi-kLeéon idater Problems g d groups of national
experts, emphasizes that in regions with I.imited va-ler re::ouxces a high degree of
accuracy in their assessment is an essential condition ror e. rational economy.
The E.'mual enphasizes the importtince of the earliest possible organization of
hydrological studies in a river basin or 8 region where an intensive growth o i
uater needs is in prospect. ?he Phniisl defines in the following W P . ~the economic
impact of reliable hydrological daCC. on ua-ter resources dcvelopnent in particular
cnd on the national evonoq in genernlc ? h e more reliable the iiatter zupply, thi:
smaller i.iillbe the damage resultin, from cutr in periods of water shortagetf. By
cornparin: losses anU expenditure, it vili, in prlnciple, be poosi'ule to determino
the economic optimum.
Sufficient and accurate hydrological data promote effective vater management and
the prevention or didnuation of damage caused by such hydrological phcnomna as
severe floods, ice j m , rnuàflotis, intemive oedirnents, dangerous va.ter pollution etc.
684

On the other hand, the intensification of human activities in river basins

rnl :,Lcii, watersheds
including the increased anount of untreated effluents
uic .liarged FnLo the water couse, during recen.'cdecades, uraently calls for the
re1iai;l.e esse:;smnt of available wa-torresources. The ECS Nanual mentioned above
states that the consequences of h w n activities make advanced hydrolo$cal
research imperative. A relevant improvement of hydrological methods is
considered important.
5Je understand that this topic is of considerable interest to hydrological
services which must strike a balance between tho cost of gauging stations and
the probable futvre benefit that will result from the information to be obtained.
TakTng into account the special importance of hydrological data for water
resources management, various aspects of the development of hydrological networks
were discussed thoroughly at the ESE Seninar on Selected Water Problems in Southern
Europe convened in Zagreb, Yugoslavia, in October 1971. Certain conclusions
concerning the design of hydrological networks and their improvement were
reflected in the recommendations adopted by the ECE C o d t t e e on Water Problems.
Taking all this into account it is generally recognized that hydrological data
and well planned hydrological networks are prerequisites for efficient and sound
water resources planning.
The purpose of this paper is to appraise the possible economic effect of
insufficient hydrological data on the effectiveness of water planning and the
design of hydraulic engineering structures and their operation.
It is suggested to consider the following main aspects of the subject:
The economic consequences of a deficiency of hydrological data on water
planning, construction and operation,
"he impact of a deficiency of hydrological data on main water users.
The economic consesuences of a deficiency of hydroloaical data on water
P ~ construction
~ P and operation
The following questions could be raised in connexion with this aspect:
what is the extent of the economic impact of insufficient hydrological
records on a project and its subsequent operation?
in particular, what is the possible effect on investment in economic
development if hydrological data are not accurate enough and the records
are insufficient?
is the predominantly quantitative character of hydrological data sufficient
for modern intensive water resources development?
is it reasonable to postpone the initiating of water project planning and
the construction of water projects if the hydrological observations are
insufficient?
The available information on the experience and research in the ECE region
shows the following facts which could be emphasized in an attempt to answer the
above questions.
 685

Economic -acts Qf insufficient hydrological data and difficulties caused at

the key stages of water resources development:
Mater planning and d e s i m
Generally speaking, poor hydrological data and forecasts made on this basis
can lead to inappropriate proposals for investment in water engineering works and
the economic development of the region concerned. The importance of sufficient
data at the following staget of planning and designing can be pointed out.
The elaboration of schemes for niltipurpose development of water resources
i s greatly dependent on accurate hydrometrical data. The appropriate conclusions
according to which a scheme is to be designed cannot be derived from incorrect
hydrological data.
It should be underlined that possibilities of considerable miscalculations
exist in preinvestment studies as well as in further stages of planning and design
of engineering structures. They may be especially acute when, to estimate
characteristic river discharges, approximake methods and empirical fornulas
are used and the length of time of hydrological observation is relatively short
(considerably less than 30-40 years). Taking this into account, approximate
methods are being limited to the preliminary estimations for the elaboration of
schemes of development, but are not reconmiended for the design of water structures.
Appropriate attention to this matter has been given in the ECL i.ianual
mentioned above.
Different research during recent years has also analysed the importance of
hydrological data for large scale, long-term investment for general economic
development.
The conclusions of some of these studies may be summarised as follows:
It is considered that, if the period during which hydrological records have been
kept is not sufficiently long or the data are not sufficiently accurate, then the
investment in the economic development will be larger than is necessary, or the
possible production of the plant will be less, due to its smaller size. In both
cases the result will be a loss to the overall econon&.
Experience shows that various hydrologic parameters could have economic
importance for different stages and purposes of water resources development.
For example - flow variability and drought occurence for the design of storage
reservoirs; flood occurences for the design of spillways and other control works;
lesser importance is attached to the mean flow. However, as can be cited after
one a u t h o d

r/ D. Johanovic. Abstract. V h e Role of I-Qrdrologg and Hydromteorolow in the

Economic Development of Africat!Ma, No. 301, 1971.
2/ K.C. Wilson Cost-benefit approach to hydrometric network planningtt Water Res.
Research. October 1972.
686

Wariation of mean annual flow, depending upon hydrologic data, and length of
observation leads to overestimation or underestimation in determination of
sizes of water engineering structures, determination of regimes, capacities
of plants, irrigated areas" etc.
In fact the following implication of an underestimation of the mean annual
flow in the planning of water resources developnient might ba indicated:
(a) Smaller sizes of water storage capacity resulting in limited possibilities
of regulation of flows with subsequent adverse impact on:
(i) possibilities of self purification of water;
(ii) availabilities of water supply for drinking and industrial purnoses;
(Xi) potential of production by hydro-electrical plants;
(iv) quantity of water for irrigated areas;
(v) navigationaï capaciw of rivers;
(Vi) inadequacy of water storage size:: requires additional investments for
increasing dans, canals, etc. at a later stase.
On the other hand, over-estimtion of inem annui flow can lead to the following
implications:
(i) oversizing of iiriter engineering strictures, low efficiency of their
operation, larger investments in conparison with normal;
(ii) insufficiency of water for designated irrigation area3
(iii) energy production below planned target;
(iv) lesser dilution of effluents discharged and slower processes of self
purification of water.
Studies of the value of hydrological data are being carried out in several
countries. Comprehensive studies were conducted jointly by the United States Corps
of Engineers and the Geological Survey (1970) with the purpose of evaluaiing the
iiorth or" hydrological data for determination of the optimum water storage conserva.tion
capacity. Some American researchers have developad a mathematical relation betireen
the llworthllof data and the length of periods during which they were recorded.
Some research has been carried out to estimate the possible loss due to
inperfect information. The research, for instance considering optimal reservoir
design, analytically defines the opportunity loss as the difference between net
benefits associated with different hydrological data depending upon streamflow
record length. Reservoir designs are obtained by simulating flovs and selecting
that combination of storage capacity and target yield which gives the greatest
net benefits. Graphic functions obtained by this research shou that opportunity
losses decrease rapidly due to increasing the length of streamflow observation,
achiedng very small magnitude beyond thirty years of observation; this conforms
to many practical observations. Various types of reseaxch and observations show
that the cost of obtaining addition& hydrological data i.e. increasing the length
of observation is insignificant in relation tothe reduction of the esrected
 687

opportunity lox.
Some CanaCZan research developing generalized computer programmes to relate the
costs of operating and intensif'ying a hydrometric network to the resulting increases
in the accurecp oi: the three parameters mentioned above should be noted.
The available experience of differenct ECL countries confirms to an extent the
conclusions of the research mentioned.
As ior example in the USSR, it is considered:
- the economic benefits of the hydrometeorological service due to which the
hydrological and meteorological observations can be obtained are a high as
one billion roubles, i.e. 4-5 tines the amount that is spent on maintaining
this semice;
- the introduction of the use of hydrological forecasts in national planning
made it possible to raise by 10-15 per cent the efficiency of water
installations and to obtain correspondingly higher profits2/.
In the UnLted Kingdom and France, the relrvant benefits are estimated to exceed
the national hydrometeorological budgets at least 20 timed.
However, in many countries, especially in the developing ones, the povement
agencies responsible for h@rological observations do not have sufficient funds
available for the development of adequate nationwide hydrological network$. This
insufficiency of financial resources for the collection of hydrological data was
also emphasized at the ECE Seminar on certain uater problems, convened in Zagreb
in 1971.
The USSR experiences show that in the absence of hydrometeorological observation
data in an area selected for construction, a special hydrometeorological investigation
should be conducted with an expenditure of 2-3 thousand roubles for each million
roubles invested. This amount is considered to be an econom if, as a result,
sufficient hydrometeorological data proved to be available. Experience in other
countries confirms that the expense of acquired adãitional hydrological data
is much less than the losses involved in water engineering construction,the design
of which is based on inaccurate data. In this connexion as a positive experience,
a considerable extension of national and regional hydrological networks is taking
place. It should also be noted that in some countries the application of automatic
monitoring stations to control the quality and quantity of a river flow has
considerably extended during recent years. This modernization increases the
reliability of recorded data indispensable for accurate planning and design.

2/ E.J. Tolstikov, Vhe Benefits of Hydroirieteoroloefc Services in the UcsRtt

-
'dl40 V h e Economic Benefit of National Meteorologic Services'World Weather
Watch No. 27 i968
i/
-
Richard D.A. Kill IXydrological and Hydrometeorological data as essential
parameters for design or economic development projectsf! ld4l Eo. 301
 68 8

In spite of the sound persuasiveness of all the research mentioned above,

armther trend of research should be pointed out. Several studies and research
have been devoted to Solve the problem of whether it is desirable and profitable
to postpone the development of water resources or the construction of certain water
cngineering projects if the hydrological data for the river or particular site
investigated are insufficient. It is considered by some of the authors that, by
postponing the construction, the realization of the net benefits of that construction
is also being postponed.
The available results of research conclude that the risk of such postponement
must be carefully evaluated. Nore than that, some research clearly rejects
postponement as a protitable course of action. It is considered that only an
exceedingly loid discount rate makes the minimum total cost of postponement
e.g. for one year, equal to the cost without postponement.
Taking into account the definite discrepancy between these conclusions and
those previously mentioned we feel that further research should be continued,
emphasizing not only purely economic aspects, but also taking into account social
and technical aspects, including first of all the problem of safety of structures
and subsequently of the population. The lack of unaniixi~iy,even in a purely
economic approach, concerning the postponement of projects is to be noted.
Some authors concludeb/:
tlSincedemand for project outputs is presumably growing over time, the more a
project is deferred, the more quickly it is likely to be used, and the greater the
benefits generated per time period of project life will be.”
The conclusion seems to he very sensible.
P l d n , ? and desiminn of flood protection
This cspect deserves special attention, bearing in mind that economic losses
due to floods in the river basins of the world continue to be very high. Noreover,
the further extent of the economic development of regions being threatened by high
flows leads to further growth of economic damage from floods. Just one example, fron
the experience of the United States, shows that the frequency of floods, causing
major roperty damage of $50 million or more were increased aimst three times since
1 9 L & d Based on the currßnt status of flood control works and project conditions
of flood p l d m use and development the total annual floodcbage potential for the
nation is anticipated to increase from $1.7 billion in 1966 to $5.0 billion in 20Zoz/.
It is considered that, for elaboration of efficient national flood proLection
policies, the data on distribution of river flow durine a year, asml1 as the exact
characteristics of floods including maximum discharge, duration, tlme of flood end
volume of the flood flow are of considerable importance.
However, the determination of these characteristics becomes verydifficult if
hydrological observations are insufficient, as was mentioned above.
m e s \J. IIowe naenefit-Cost Analysis for Water System PlanningIl
Geophysical Union, bJashington D.C. 1971
- American
1/ IfTheNations Water Resources11 W.R.S. United States, 1968, 5-2-2.
 6 89

Experience shows that the frequency of peak floods, determined on the basis
of a short ceriod of observations c m be several times lower than the adequate
vdue, and that leads to considerable damage and to catastrophiee. The foll-owing
consequences of inaccurate flood forecasting should be pointed out as regards
water resources development and planning and designing of flood protection.
-
(a) Overestimation of flood discharge leads to unnecessary expenses in
weter engineering construction;
-
(b) Underestimation of flood discharge leads to full dektruction of the
sl.r.ucture, resulting in damage and possibly in loss of human lives in
the region.
Thus, proper estimation of possible flood peak, depending upon the quality
and accuracy of hydrological data is very important. At the samc tine the
experience of many countries still shows that: the continuing groi.rth of economic
damage to individuals and to the national economy in many regionu of the world,
caused by river floods, can be explained not oniy by the spontaneity of the flood
phenomena, but also by the lack of adeguate organization, insufficient hydrologic
observation and the necessary financial means, which very often are considerab1.y
less than the value of the damage caused. Talcing this into account the ECZ
C o h t i e e on Water Problems has recently initiated studies on the available
experience in rational methods of flood control planning in river basin
development with the purpose of extending this experience to al1 ECd countries.
In spite of the concept generally adopted that the development of sufficient
hydrological observation is very important for the planning of proper flood
protection and the organization of adequate operational measures to prevent
considerable damage, nevertheless the available infornation from some countricc
indicates:
- a lack of reliable data regarding rainfall intensities and corresponding
stream flow;
- insufficient studg of floods based on a detailed analysis of recorded flows
so that water management authorities in some river basins or their parts are
unable to arrive at more realistic estimates of expected floods;
- the insufficiency of the empirical and rather arbitarg methods used up to
now in some countries for estimating peak floods; inadequecy of such methods
has been proved and should no longer)%cceptable, in d e w of the magnitude
and importance of the projects undertaken.
It is also indicated in some countries that from the point of view of
safety and also from the economlc angle, the necessity for current study and
evaluation of the magnitude and frequency of the occurence of floods in
connexion with the economic development of Tiver basins has become essential.
The existence of dams in the vicinity of populated areas necessitates the closest
study of the anticipated probably floods, in order to provide adequate spillway
capacity for the safety of the dams, and the downstream areas.
690

In some countries flood forecasts are not yet of the desired accuracy, thus
increasing Lhe danger of economic damage or reducing the efficiency of water storages
down the river, while accurate flood forecasting can increase the overall potentialitie:
of a multi-purpose project.
The disoytrous floods which occurred in several regions of the world in the recent
decade, causing considerable loss of life an9 treinendous economic damage, pointed to
the need to extenä flood forecasting and warning syotems in many countries, especially
in their m.ìt vulnerable river basins. Relevant units established by national kjdro-
meteorological services or by central water and power agencies confirm their
effectivenes: It is considered that ths expenses for the maintenance of these unita
is only of mciest cost compared with the peins achieved by timely forece-sting. As
an example, considerable economic benefits are accrued in m q v countries when flood
forecast? nre usea to enable protective measures to be taken against the affects of
floods. In these cases, the economic benefits through iorecasts are caisulô.ted as
the differenw between profits from proteoted zoner or areas.
The use of flood forecasts in the IJSCR reduces the cost of damage äue (* floods
by 20-30 per 'cent. merience in the Uni'reä -:-catessonfirms that, the redudion of
flooe .iamagt aione would far outweigh the tow?. cost of [email protected] forecasting
serviqe, including the proposed network.:-'.8, A:$ participants were informed at the
Uniteu Nations inter-regional Seminar .,r, '?loo6 DamP-Ze Prevention Measures and
Management, convened in Saptenber 1969 i,. Tbidsi, ITSSR, the annual savings due
to the exlsting flood warning systems in the United States exceed $30 million a year.
Experience in India also confirms that the early expenditure on the maintenace of
the unit in the central water and power commission as well as the cost of the necessary
-
equipment is only a very moderate investment compared with the benefits which have
been brought by forecasting.
Importance of accurate data to show flow formation regardinp human activity at
watersheds of rivers
Attention should be drawn to the importance of strengthening research to analyse
the influence of human activity in watersheds on river flow. It is considered in some
countries that the euccessful design and operation of water engineering structures
could be possible if sufficiently accurate data to show the hydrological characteristic
of river basins under natural conditions of flow formation, with regard to the scale
and direction of changes caused by humans, were available. The importance of this
aspect has been emphasized among other publications by the ECE Manual mentioned in
the first part of this paper.

8/ M.A. Kohïer (United States National Weather Service) Wasebook on Hydrological

-
Network Design Practicet' WMO No. 324, 1972
s/ Sh.M. M a n s h a r d (Central Water and Power Codssion) tTlood Forecasting and
-
Flood Warning in India". Seventh Symposium The Civil and Hydraulic Engineering
Dept. Water Resources. May 1971.
 6 91

II. The inpact of a deficiency of i1ydrolopicai data on main water users

The impact on the following water users is considered:
(a) 1)onestic water supply
(b) mdustrial water supply
(c) Xydro-power generation and thermal power plant water supply
(d) irrigation
(e) 1knigation
(f) Fisheries
(g) in-terpollution control
Ex'üen3.ve research was carried out in i?ariy AC3 countries devoted to t'ne appreisal
of the impact of a deficient river flow aroused, among other reasons, by the lack of
hydrological obsorvations and insufficiently accurate forecasts. One such research
project has been carried out during -therecent decade 'by the central research institute
on water problems in the USSR (btinsl~)~. Analysing dieierent nethodx of estimation
for the appraisal of the impact of doficient water flow, the authors indicate that, in
sonle case:, estbation becomes difficillt by reason of too 5hoi-t a period ol 1iyarologica.l
obsemration:;. It is aïs0 pointed out that special estimation techniqucr to be applied
in cases of short available periods oi hydrological observations, which could be applied
for economic analysis, have not yet been created.
However available results of research carried out in the same corntry suggest
11 They atter-pt
some methoas for use under conditions of deTiCient hydrological dataJ.
to find a relztion between the extent of limitation of the water siipply and the
reduction of economic activities (e.g. the industrFal output) of a certain enterprize
in order to asress the economic damage caused by water shortages.
Domestic water supplx
It is generally adopted that domestic uater supply does not allow in-Lerruptions
i.e. the pyaranteed water supply for these consumers must be doze to 100 per cent.
Industrial water siipply
It is considered that the magnitude of economic damage connected. with stoppage
or partial reduction of water supply is considerably influenced by the quality of
hydrological prognosis.
Some researchers in countries with a centrally planned econoqv suggest to
subdivide the dmages, as followsw:

o/ 1.1. I-fechetov,V.N. Pluzhnikov, L.J. Popov IfBalanceof Water Resources and Needs
the method of optimal planning of water ressurces dsvelopmentl~Multiqurpose water
-
resources development and conservation. Minsk. 196û.
Ir/ V. Andreyanov IiInternal distribution of river flowl'. Gidrometizdat. USSR.
Centrcrl Research Institute of Water F'roblecis. USSR. Minsk. "Miilti-purpose
-
development and quality conservation of water resources" 1968.
692

(a) direct damages (expressed by direct cost of unproductive forms of enterprises

due to water stoppage or shortace);
(b) indirect -
measured by the loss of profit during the time of stoppage or
reduction of water supply.
The lack or insufficiency of research to define a relation between -Lhe extent
of the limitation of the water supply and the relevant economic damago to clifrerent
tLJhnologicalprocesses is being noted in some branches of production.
Generation of electric enerpy by hydro-electric plants and themai power plants
Proper flow predictions are important + o obtain optimum utilization of *dro-electi
and tharml. power plants. Timely 8n.d accurate ïlood forecasting in periods of lieais.
rains increase:.tlie overall potentiality of a multi-purpose project. Accurate
forecasting is especially important bearing in nind tho fact that hyciro-power
resources, depcnd on neteorological m u hydrological conditions vhich <an chanvre
considerably every year. As it is known, the mamitude of hydro-eaerD prnciuction
is to a great extent dependent upon natural river flows or upon the surface run-of2
from one square Idlonetrc:of the area of the basin. 'The unevenness of river floua
might essentially impact o@nùustrial value of hydropower resources and this in
its turn rzises special reriuirements as to the quzlity of hydrological provoses.
Fie le :k of a Tpecial n.ssessmeiït 02 !parentoc6 po-Lentiai1iy.kopower rcsoLLpces
in ElL coiinui'ie, mkes it c::pc::iaily di? 'icuLtto äe-teminethe guarmteed untcr
tlis.:i:argt- in L'no -ri.-mrr..
In 3mie :,:intriesi< UP.; cs'cabliohec d:zt the reciur:tion of Lischarge of fresh
water for thamal p o ~ e rplant by more ti;cii 10-15 per cent from tie 3evicnP.ted
quantity -p.uses reduction of the power generation in proportion relevant
approximatcly to the reduction of the water supply.
The relevant economic damage to the pouar productiun related to 1 m3 of water
which can nu; be supplied to the pouor plant could be as high as ten roubles per
1 2of wn-i;erl.2/ (USSR experience). zie limitation of power supply for differant
consumers, for example industrial consumers, due to poor hydrological forecasts
causes economic ,lamage,which could be estimated by the stme methodology as that
applied for industry due to limiting of vater supply.
Irrigation
Irrigation permits the rehction of water supply by 20-25 per cent compared with
the norms establishedu. However, further limiting of water very often causos
considerable economic damage, trhich is determined by the cost of production lost
due to insufficient irrigation.

merience in the USSR

&I. !'Power Reso&-ces of the USSR" Hydro-Power Resources tlNauka'f1967.
 693

Il,is emphasized that not onlx the entent of reduction o€ water supply is important
here but also .the seasonal period of this reduction which could be -<cry importmt
for vegetation. It is statea that the relation between an insufficient uater supply
and a,pi.cultuai production is difficult to e s t h t e because agriculbural yield
depends upon several factors taken as a whole.
River navipntion
Some researchers in the countries with centrally planned economies suggcst to
apply the following methods to determine economlc damage to navigation in cases or"
river flow reduction including unsatisfactory hydrological forecastsu:
1. The magnitude of norm.al navigational water release is being detenifineci for
ships of different mndels and curves of guaranteed river discharges established for
reaches of liater courses with limited water transport.
2. The extent of infringement (violation) of navigation interests (insufficient
loading of ships, etc.) is being determined for each internal river depth.
Pizhsries
It is admitted in some sources that damages to fisheries by the reduction of
river flosr have been estimated only subjec-Lively. Research which has been carried
out in the USSR concludes that former analyses of the inter-connexion between
ichthyolozic and hydrological conditions were predodnantly of a qualitative
character. Recent research points out that quantitative water requirements by
fisheries could be important for certain periods--/. 16
~ d r o l o d c a l .forecast Igaw-w&er pollukien eonkrol prom+mme4
S

The provision of satisfactory water quality in om rivers ha5 become one of t!p
most important aspects of the deteriorating husaan environment.
In that connexlon two aspects which have become very important may be improved
hydrological services:
better quantitative forecast of river flows;
better organization of water quality control.
it is known, there are many water quality characteristics. Dissolved oxygen (DO)
considered as one of the most important, which is affected by meteorological and
hydrological variables and by the discharge of organic waste along the watercourse.
First, to improve the methods of modern water quality control in river basins
requires systematic and accurate hydrologic forecasts, taking into account that
waste releases in niodern pollution abatement programes &pe sometimes scheduled
according 'Co predicted river flows. Any discharge of a considerable quantity of

rr/ Central Water Research institute. USSR. f'lkilti-purposewater resources

development and conservation'!. . Minsk. 1968.
19/ S.N. kits@ and M.F. Menkel "Genat.al scheme of multi-purpose development and
conservation of water resources of the USSR". Proceedings of Hydroproject
Vol. 12 "Energiaf1USSR. 1964..
69b

pollutant to the river in a period uith lower flow could create dangerous pollution
concentrzìions in the lower reaches of the -river. %u hydrological forecasting
has therefore aezome more iraportant fÒP watcr quality control in the river basin.
-
The other nspect is the importahce of better organization and modern
i n s t m n t a t i o n of water quality contrbl trg hyd.rolo~&cal sedces. The ECE Manual
mentioned above gives particular at4ention to water- qdity/data and research
simultaneously with hyàrological resáárch, reqtliring the extehsion and hprovement
of the system of observations on water”qua1ity in streams and reservoirs.
Information from the ECE countries, posticulariy diseussed at the ECE Sendnar
in Zagreb, 1971, points out khat considerable improvement in these functions is
being iqîemented in m m y countries,
As regards the assessment of economic danage caused by water quality
deterioration due to flow reductionsin rivers, including reasons caused by
untimely and inaccurate hydrological forecasts, there are as yet no methods
of assessment generally recognized or adopted. In some of the ECE countries
it is considered that the infringemant of sanitary water conditions in a river
basin, due to low flow, and the relevant damage d e e economi=.assessment.
Further comprehensive studies seem to be necessary in tlis field.
The same is said of the economic assessment of the impact of water quality
deterioration on water recreation. Correspondingly it 1s considered that it is
difficult to substantiate economically permissible standards of concentration of
polluters in a river basin.
However, there is no unanimity on this point. Research in some other ECE
countries show Lliat the increase in the present value of direct quantifiable
recreation benefits of inproved water quality, for example in the Delaware River
bacin could be as high as $300-350 niillion for the highest water quaïity clase
adoptadw.
Sediment control
In e:cicnding and further improving hydrological services and their forecasting,
sediment control shodd not be neglected.
Further water resources development and growth of human activity on the water-
sheds m e s quantitative and qualitative sediment data very important as a part of
hydrologic controls for different periods of the year. The iniportance of this
problem could be seen fromthe experience of q ECE countries.
It has been recognized by nany countries, that accumulation of sediments in
niagy cases ohortcns the effective economic life of water engineering structures such
as water-storagc ’,causes ii tensive wearing of pumps and turbines, requiring in all
cases new capiti investments. As Is known, the problem can be grcatly mitigated
by cidensive and complex control measures, including inproved forecasting.

12/ A.B. Iïneese and B.T. Bower. AnaJYSeS conducted by the University of Pennsylvenia.
Wanaging water quality’t. John Xopkins Press, Baltimore, 1968.
 695

Experience in the United States shows that the damage from sediments to water
management and national econow reaches more than $500 million a n n u a l a .
Data on ice and slush ice conditions
Inaccurate forecasting and warning ice and ice slush conditions cause
considerable damage to hydro-electric plants and the energy consumers,
causing reduction of industrial production in some of the regions of the middle
climatic and rmuntainous zones.
Conclueions
In conclusion we would U k e to emphasise the following:
1. Available experience and research being carried out in ECE countries clearly
demonstrate the considerable impact of insufficient hydrological data and
forecasting in all phases of planning of water resources development and use,
which in their turn could have an *act on economic development of certain
regions.
2. Available attempts in ECE countries to create and improve &.sting methods to
assess the value of hydrological data already have achieved certain positive effects,
for example the development OP mathematical relations between the present value of
the value of data and the length of their record, as well as development of some
methodology to assess the impact on main water users.
3. Available research clearly shows that the cost of additional hydrological data -
-
i.e. increasing the length of observation is imd@ficant in relation to the
expected losses due to insufficient data. However, insufficient funds available
in many countries, especially developing countries, are very often evaluated as
inadequate for proper development of hydrological networks as was revealed and
emphasized by different conferenoes.
4. In spite of considerable research being carried out in ECE and other countries,
as veli as various studies by specialized organizations, the following main
deficiencies in hydrological observations and research are noted in different
countries, which render a certain negative impact on water resources development
and effective use:
- the lack of reliable data on rainfall intensities and corresponding streamflowsj
practical usage of empirical and sometimes rather arbitary methods for estimating
peak floods; the lack of scientific studies on evaluation of the magnitudes
and frequency of the occurrence of floods; insufficient studies and research to
assess t h impact of human activity in watersheds on waterflow; lack of
special estimation techniques to be applied for economic analysis in cases
of available short periods of hydrolo,.Lcal observations; deficiencies of

u "Xnvironmental Problems".
to ECE. January 1970.
Monograph presented by the United States Government
 696

available methods of long-tem hydrologi.cal forecasts; insufficiency of

research to define a relationship between the extent of the limitation of
,he water supply and the relevant economic damage to different technological
processes; lack of generally recognized methods of assessment of economic
damage caused by water quality detexioration.
Strengthening of the research regarding the deficiencies listed above Is
considered as important and urgent.