Hydrology in Mountainous Regions.

H - Artificial Reservoirs; Water and Slopes

(ft-ooeedings of two Lausanne Symposia, August 1990). IAHS Publ. no. 194,1990.

Reservoir sedimentation and flushing

W.R.WHITE
Head of River Engineering, Hydraulics Research,
Wallingford, UK

ABSTRACT This paper is concerned with the engineering

assessment of the deposition and re-erosion of sediments
in reservoirs. This deposition and re-erosion can be
assessed by methods which vary in their complexity.
Historically, simple empirical methods have been used
which require only a limited amount of data and analysis.
More recently, however, computer models have been
developed which can simulate the mechanics of sediment
movement within the reservoir as flows and levels change.
These are used to predict sedimentation patterns and also
to look at the possibility of the strategic flushing of
sediments through the dam in order to extend the useful
life of the scheme.

An assessment of catchment sediment yield is a

prerequisite to any consideration of sediment problems in
reservoirs and this has been shown to be difficult and
often imprecise. Sediment yields vary with climate, the
geology of the catchment and land-use practices.

INTRODUCTION

Reservoir sedimentation and the consequent loss of valuable water

storage is becoming increasingly important in tropical countries.
High sediment yields are natural in the tropics and are balanced by
the high rates of erosion and soil production. When this balance is
disturbed by man's activities then the sediment yield is dramatically
increased at the expense of soil renewal. The steady rise in soil
erosion in tropical countries due to increased cultivation has
endangered reservoir projects and caused doubts about the viability of
existing and future schemes. The impoundment of water for potable and
irrigation supplies, hydro-power, and flood control is a necessary
step towards improved national incomes. Untimely sedimentation may
reduce the benefits and, if it is ignored, remedial measures may
become prohibitively expensive.

ASSESSMENT OF RESERVOIR SEDIMENTATION

The assessment of the deposition of sediments can be made using simple

empirical desk calculation techniques or more sophisticated numerical
modelling. In the case of existing reservoirs these estimates can be
improved by measuring sedimentation during the early life of the

129
W. R. White 130

reservoir and using these base-line data to help with predictions of

future deposition.

Survey and computation of reservoir sedimentation

Figure 1 shows sediment yields derived from recent reservoir surveys

carried out by Hydraulics Research in Indonesia, the Philippines and
East Africa. The order of magnitude difference in annual sediment
yield is due to the large differences in soil erosion between the
sites in the humid, steeply sloping drainage basins of south east Asia
and those in the drier, less steep areas of central Kenya. In
addition sediment yields fluctuate annually due to variations in
rainfall intensity and distribution, and over longer time periods due
to changes in land use and runoff. The important point to note is
that all the sediment yields measured by means of these reservoir
surveys (after allowing for errors in survey and computation) are
several times higher than the figures used during the design of the
dams. In the absence of reliable sediment transport data and sediment
yield estimates from tropical areas, designers have had to use data
from temperate regions.

Reservoir Country Drainage Annual Survey

Area Sediment Period
Yield
km 2 tonnes/km 2

Karangkates Indonesia 2050 4200 1973-77

2050 1550 1977-82
Selorejo 238 4660 1970 - 77
238 1370 1977-82
Tokol 180 4200 1975-77
Kindaruma Kenya 10000 460 1968-81
Kamburu 9520 430 1968-74
9520 770 1974-78
» " 9520 280 1978-81
Magat Philippines 4123 3800 1982-84

FIG. 1 Sediment yields based on recent HR reservoir

surveys.

However, not all the variation in sediment yield can be explained

as a change in rainfall and runoff. The accuracy of the result
depends upon the detail of previous basin surveys. The volume of
deposited sediment is generally obtained as the difference between the
total reservoir capacity at the time of survey and the capacity at
some time in the past (usually pre-impoundment). In this way the
result is obtained by taking the difference of two, large,
approximately equal numbers. A small error in the calculated
reservoir capacity may give rise to an error up to two orders of
magnitude different from the calculated volume of deposited sediment.
Research at Hydraulics Research has focused on examining suitable
methods for surveying and computing deposited volumes. The aim has
been to provide information on the most appropriate method for the
accuracy of result required.
 131 Reservoir sedimentation and flushing

Survey techniques There are numerous ways of surveying a

reservoir to determine the elevation of the bed and hence the total
storage and sediment volume; and the method and equipment chosen
should reflect the nature of the reservoir project. The method of
data processing and volume calculation should be decided first since
this will constrain the choice of equipment. In many cases estimation
of design capacity is based upon small-scale mapping, whereas later
range line surveys give more detailed information. Combination of
these two types of data, particularly when pre-impoundment data are
sparse, can lead to poor estimates of sediment volume.
Traditionally survey data are collected along pre-determined range
lines which can be re-surveyed so that the changes of sectional shape
can be noted directly. Depth measurement is obtained by using an
echo-sounder (sonar) with a chart or digital readout. From our
experience, more discrepancies in cross-section data can be attributed
to errors in the initial triangulation than any other source.

Computational techniques The literature contains details of a

variety of techniques for computing the volume of deposited sediment
but gives little guidance on their relative merits. Reservoir
managers with limited financial and skill resources need to know how
the choice of method will affect the accuracy of the results. Studies
undertaken by HR have gone some way in identifying suitable methods.
These are categorised into three groups :
® End-area methods (based on range line surveys)
® Contour methods (based on contour information)
• Combined methods (based on range line and pre-impoundment contour
information)

Terrain models (see Figure 2) The use of Digital Terrain

Models (DTMs) has increased dramatically over the last few years.
They are used to model ground surfaces and can be used in processing
reservoir survey data and in calculating the volume of deposited
sediment.

FIG. 2 Three-dimensional representation of DTM, Manjirenji

reservoir, Zimbabwe.
W. R. White 132

Use of DTMs necessitates investment in expensive surveying

equipment with integrated data collection in a form which is readily-
transferable to a computer for processing. The high cost of DTMs
means that their use for reservoir surveys in the developing world
will for some time be limited to specialist survey companies. The
advantages of DTMs are the improved accuracy, speed of computation,
and reduction of preparatory work (setting up range line beacons).
The accuracy of the results depends heavily on the adequacy of the
surveys, in particular the pre-impoundment survey.

EMPIRICAL METHODS OF PREDICTING SEDIMENTATION

Until recently reservoir sedimentation could only be assessed using

simple, empirical methods. To estimate the volume of deposited
material the notion of trapping efficiency was introduced. The
trapping efficiency of a reservoir is defined as a ratio of the
quantity of deposited sediment to the total sediment inflow.
Gottschalk (1948), Churchill (1948) and Brune (1953) provided
simple graphical means to determine trapping efficiency and these have
been used extensively. Since, however, the trapping efficiency must
depend upon the sediment size, the flow through the reservoir, the
distribution of flows into the reservoir and the way that the
reservoir is operated, it follows that such estimates of trapping
efficiency can only provide approximate values which may, on
occasions, be seriously in error.
After analysing several reservoirs in the USA, Churchill came to
the conclusion that along with the retention time, the transit
velocity, ie, the velocity with which the water flows in the
reservoir, governs the trap efficiency. If the water held in the
reservoir is moving fairly rapidly in the reservoir, very little
sedimentation will occur because the turbulence associated with the
higher velocity hinders settling, even though the retention time may
be high. He introduced a parameter known as sedimentation index which
is the ratio of the period of retention to the mean transit velocity.
The trap efficiency of the reservoirs was found to increase with
increase in the sedimentation index.
If a large percentage of the sediment in the stream is moving in
the form of density currents, then the concept of mean transit
velocity introduced by Churchill is questionable. In such a case, the
velocity of density currents may be very different from the mean
transit velocity of the flow.
Brune analysed the records of 44 different reservoirs in the USA
(41 of which were normal ponded reservoirs) and found that the
capacity to inflow ratio gives better correlation with trap efficiency
than the capacity to watershed area ratio. Figure 3 shows Brune's
plot of the trap efficiency against capacity to inflow ratio.
Pitt and Thompson (1984) developed the methodology proposed by
Brune and used the concept of "reservoir half life" (ie, the time
taken to fill 50 percent of the reservoir storage with sediment).
Figure 4, based on Brune*s curve, shows that reservoir half life
varies roughly in proportion to reservoir size and the inverse of the
sediment concentration. Assuming an average sediment inflow
concentration in the range 1000-4000ppm, most reservoirs with storages
in excess of 50% of the mean annual inflow will have half lives
 133 Reservoir sedimentation and flushing

100 4-

JL
S ^
••_ i~

+
>^^
80 +
^ S "**
/ ^ < * ^
/ / - Envelopes
60

>'A/
JU 7Z ^

/
+/ ^ D e <>ign curve for normal ponded
ervoirs

/ y /

20 / /' /
/ /
7~
/ ; / • /
10"J 10"' Iff'
Ratio of capacity to annual inflow

FIG. 3 Trap efficiency of reservoirs after Brune (1953).

measured in hundreds of years. For storages in the range 5-50% of the

mean annual inflow, half lives will be measured in decades whilst for
the smaller reservoirs the half life may be less than 20 years.
Kabell (1984) presented a similar methodology based on
observations of sedimentation in reservoirs in Zimbabwe. Figure 5
shows how the reservoir half life depends on the initial storage rates
and the concentration of fine sediments entering the reservoir.
Kabell used the curves given in Figure 5 and suggested that all
reservoirs with an initial storage ratio of less than 0.10 are
uneconomic. Furthermore, in areas of poor land use practices, a
minimum economic initial storage ratio of 0.25 is likely to apply.

001 0-1 1-0

Ratio of reservoir volume to annual inflow

Fig. 4 Reservoir half life as a function of the average

sediment concentration of the inflow, reservoir volume and
mean annual flow after Pitt & Thompson (1984).
W. R. White 134

Time (Years)

FIG. 5 The influence of the initial storage ratio and the

mean sediment concentration of the reservoir inflow on
reservoir half life after Kabell (1978).

HR has recently carried out work to provide a method of estimating

reservoir sedimentation which was simple to apply but took account of
the size and shape of the reservoir, the nature of the sediment and
the hydraulics of the flow.
Rather than consider trapping efficiency to be a yearly average an
instantaneous trapping efficiency was defined. By dimensional
analysis it was concluded that the instantaneous trapping efficiency
was a function of three non-dimensional variables
i - bdw
n —q~
b
7 -
z
2 T
7
Z
_d
3 -T
where
b = mean width
d = mean depth
L = length
w = fall velocity
Q = discharge
By using a numerical reservoir sedimentation, see next section,
the variation of trapping efficiency with these variables was
determined.
The instantaneous trapping efficiency can then be determined for
different times during the year and integrated to give the annual
trapping efficiency.

NUMERICAL MODELS FOR PREDICTING SEDIMENTATION

Recently, with the availability of computers, it has been possible to

develop numerical models of reservoir sedimentation. These models
 135 Reservoir sedimentation and flushing

calculate the water flow and sediment movement throughout the

reservoir and can provide a reliable and detailed estimate of the
impact of sedimentation.
Reservoir sedimentation results from a complex interaction of a
number of physical phenomena. While water flow can be satisfactorily-
described our understanding of the movement, settlement and
consolidation of sediment is not complete. Despite this, it is
certainly possible to improve on how these processes are described in
existing numerical models.
Most numerical reservoir models are time-stepping models. For
given initial conditions, the equations are used to predict what
happens in the reservoir over a short time-step. The process is then
repeated a number of times over the required time period. The
time-step used depends upon the size and nature of the reservoir but
is typically of the order of a day. Once the geometry of the
reservoir and incoming river, and the nature of the flow and sediment
are specified, three modelling stages can be carried out : firstly of
the reservoir's storage, secondly of its flow and thirdly of the
sediment transport within it.

Modelling reservoir storage

When flow enters a reservoir, its velocity drops dramatically and it

is no longer capable of transporting the coarser sediment fractions.
If the water level in the reservoir is near full supply level the
sediment will be deposited near the head of the reservoir. If the
reservoir is partially full then deposition will occur further into
the reservoir basin and at a lower elevation. It is thus important to
model the reservoir water level. The finer sediment will be carried
further into the reservoir where its deposition is controlled by the
opposing effects of particle weight and turbulence. To predict the
reservoir water level a storage sub-model is used. The sub-model uses
a continuity equation to relate the inflow of water into the reservoir
to any outflows plus the change in storage in the reservoir.

Modelling of flow

The water flow in the reservoir and the upstream river is determined
using a backwater calculation. The water level predicted in the
reservoir storage simulation, described above, is used as an initial
downstream boundary condition at the dam to enable the backwater
calculation to proceed upstream. This calculation provides water
depths, velocities and slopes at each cross section along the length
of the reservoir and up the incoming river.

Modelling sediment transport

In modelling sediment movement the primary concern is deposition. The

trapping efficiency and the location of the deposition depend on the
volume of water stored in the reservoir. Since, however, the water
level in the reservoir fluctuates and the inflowing discharge varies,
sediment that has previously been deposited may be subsequently
eroded. It is necessary, therefore, to be able to model both the
W. R. White 136

deposition and erosion of sediment. In performing such calculations,

which are of a volumetric nature, due allowance is made both for
initial density, and for subsequently increased density due to
compaction by overlying deposition. From the calculated velocities,
depths and slopes, the sediment concentrations at each section may be
calculated, but when modelling the sedimentation process it is
necessary to treat the sand and silt fractions separately. This is
because sand movement depends only upon local hydraulic conditions,
whereas silt movement is also influenced by preceding flow history.

Sand movement The transported sand sizes at each section are

calculating using one of the many established sediment transport
theories for non-cohesive materials, eg, Ackers and White (1973). The
movement is dependent upon the sediment diameter. For sediments which
do not contain too broad a range of different sizes a representative
sediment diameter, D 3 5 is often used. For widely graded sediments the
range of sediment sizes is divided into a number of classes each with
a representative diameter.

Silt movement The concentrations of the silt fractions entering

the reservoir depend on the drainage basin's sediment yield and total
annual runoff. The silt is convected with the flow but its
concentration reduces as some of the material settles out of
suspension onto the bed. The rate of settling is dependent upon the
fall velocity which in turn is dependent on concentration and flow
conditions. The calculation of the silt concentrations requires more
closely spaced sections than for flow. The resulting transport rates
of the silt fractions at each section are added to the sand transport
rates to obtain the total sediment transport rate. The change in bed
level at each section due to the variations in sediment transport rate
along the reach can then be determined.
At the head of a reservoir different sediment sizes are sorted
according to the ability of the flow to transport the material.
Studies at HR have shown that to represent conditions in this region
it is best to use a number of representative sand and silt sizes.
Predicting reservoir sedimentation using numerical modelling
offers a significant improvement over other methods based on simple
estimates of trapping efficiency. The method can readily take account
of :
• variable flows
» variable water levels
• sediment size
• reservoir geometry
• reservoir operating rules
and can provide :
• volume, location and compaction of sediment deposited over a
specified time period
« annual stage-storage curves
• longitudinal profile of the reservoir at any given time
• effectiveness of using sediment flushing to maintain storage
capacity
 137 Reservoir sedimentation and flushing

FLUSHING SEDIMENTS THROUGH RESERVOIRS

Sediment flushing refers to the method of hydraulically clearing

accumulated sediments from a reservoir, usually by releasing flow
through low-level outlets at the dam.
As the low-level outlets are first opened in a reservoir with a
high retention level, the local concentration of flow entrains the
fine material deposits close to the outlet. This gives a false
impression to a lay observer of extensive desilting of the reservoir.
As soon as the local deposits are removed, this action stops. The
velocity of flow away from the outlet decreases very quickly so that,
in a relatively short distance, the velocity becomes too small even to
move the fine material.
Sediment flushing is not effective unless the reservoir is drawn
down to an extent that flow velocities over large areas of the
deposited material are sufficient to create high sediment transport
rates. Even then these hydraulic conditions have to be maintained for
a significant period of time in order to remove the high volumes of
sediments which will have been deposited over the previous months or
years.
In practice this means that reservoirs have to be drawn down to a
relatively low level for an extended period of time and for this
reason the trend in recent years has been towards larger and larger
low-level outlets. The proposed dam at Kalabagh on the River Indus,
for example, has low-level outlets with a capacity of more than
AOOOmVs.
Whether sediment flushing is a practicable, economic proposition
at particular reservoir sites depends on may factors:-
(a) The ratio of the annual run-off to the reservoir volume must be
high in order to provide the "extra" water required for flushing
(b) A climate with distinct wet and dry seasons is an advantage
because flushing can take place early in the wet season and the

10 I I
i r11"

R\^
10

£ 5
oi \ Spillway \ Spillway ;
I = 700m AD \ = 715mAD
\ Spillway
- = 708 AD -

10 i i i i
10 30 A0 50 54
Time (years)

FIG. 6 Decrease in storage volume with time, Kamativi,

Zimbabwe.
W. R. White 138

reservoir can be refilled to supply water through the following

dry season
(c) Flushing is more likely to be an economic proposition for water
supply schemes than for hydro-electric schemes because the need to
draw down the reservoir for extended periods can have a major
impact on power output
Hydraulics Research's numerical reservoir sedimentation model was
recently used (White and Bettess (1984)) to investigate whether the
net annual water requirement (1.1 x 10 6 m 3 ) of the Kamativi Mine in
Zimbabwe could be obtained by constructing a dam on the adjacent Gwai
river. A range of heights for the proposed dam were considered, but
all suffered from significant deposition.
The loss in storage against time for various proposed dam heights
is shown in Figure 6. The large ratio of annual flow (580 x 10 6 m 3 ) to
storage requirement and the fact that the workings of the mine would
tolerate an interuption to the water supply led to a study of using
sediment flushing through large low-level outlets to maintain the
storage required. This study identified under what conditions
flushing is a practical means of maintaining reservoir storage, and in
the case of Kamativi dam, showed that storage could be maintained for
a considerable period of time, Figure 7.

5-0
4-5
4-0
Sediment flushing
3-5 (extensive)

Sediment flushing
3-0
(moderate)

2-5

20

1-5

Annual volume water required X

10
10 20 30
Time (years)

FIG. 7 Effect of flushing on storage, Kamativi, Zimbabwe.

REFERENCES

Ackers,P. and White,W.R.(1973) Sediment transport : a new approach

and analysis. Proc ASCE, 99, HY 11, Nov, pp2041-2060.
Brune, G.M.(1953) Trapping efficiency of reservoirs. Trans AGU,
Vol 34, No 3.
Churchill, M.A.(1948) Discussion to Gottschalk (1948).
 139 Reservoir sedimentation and flushing

Gottschalk,L.C.(1948) Analysis and use of reservoir sedimentation

data. Proc F1ASC, USDA (Washington).
Kabell, T. C.(1978) Sediment storage requirements for reservoirs.
IAHS Publication No 144 (Proc IAHS Symposium, Harare).
Pitt,J. D. and Thompson, G. (1984) The impact of sedimentation on
reservoir life. IAHS Publication No 144 (Proc IAHS Symposium,
Harare).
White, W.R, and Bettess, R.(1984) The feasibility of flushing sediments
through reservoirs. IAHS Publication No 144 (Proc IAHS Symposium,
Harare).