8/10/2015

Hydrological Frequency Analysis

Frequency Analysis

 First step in designing variety of engineering projects

related to water resources problems
 Design of reservoirs, bridges, highways, drainage systems,
soil conservations works, hydro-electric projects etc.

 In case of soil conservation, we need to determine

probability of occurrence of particular extreme rainfall. This
information is determined through frequency analysis of
point rainfall data.

 Deals with the chance of occurrence of an event

equal to or greater than a specified magnitude

1
 8/10/2015

Frequency Analysis
 The objective of frequency analysis of hydrologic data is to
relate the magnitude of extreme events to their frequency of
occurrence through the use of probability distributions

 The hydrologic data analysed are assumed to be

independent and identically distributed, and hydrologic
system producing them is considered to be stochastic,
space-independent, and time-independent

 Independent: data is not related with adjacent observation, i.e.,

randomness

 Identically distributed: data coming from same population and having

same statistical properties (homogeneity). All data reflect same type of
hydrologic processes

Frequency Analysis

 Suppose, P is the probability of occurrence of an

event (rainfall) whose magnitude is equal to or in
excess of a specified magnitude X

 The recurrence interval (return period) is related to P

as follows:
1
T  (1)
P

 This represents the average interval between the

occurrence of a rainfall of magnitude equal to or
greater than X

2
 8/10/2015

Frequency Analysis
 If return period of rainfall of 20 cm in 24 hour is 10
years at a certain station A

 It implies that on an average rainfall magnitudes equal to or

greater than 20 cm in 24 hour occur once in 10 year period,
i.e., in long period say 100 years, 10 such events can be
expected

 However, this does not mean that such rainfall

events will be separated by 10-year intervals

 It is much more likely that two or more events may occur

within one year or month

Frequency Analysis

 In the design of structures generally annual data is

used for frequency analysis

 However, analysis is accurate only if length of record

is sufficient

 In the case of soil conservation structures, at least 10 years

data are essential

 Furthermore, estimating the frequencies of expected events

greater than twice the length of the record should be
avoided

3
 8/10/2015

Pre-requisites

 Before performing frequency analysis, data

should be checked for

 Adequacy of length of the record

 Randomness of data

Test of Adequacy of Length of Record

 To check the adequacy of length of record

Mockus developed the following equation:
Ymin  4.3 t10 log Q2 2  6 (2)

Where
Ymin = Minimum acceptable years of record
t10 = Student “t” value at 10% significance level and
(y- 6) degree of freedom
y = Number of records in the sample
Q2 = Ratio of 100 year maximum rainfall to 2 year
maximum rainfall and is defined as below:

4
 8/10/2015

Test of Adequacy of Length of Record

ZT  100
Q2  (3)
ZT  2

ZT  Z 1  CV . CT  (4)

 Where,
ZT= Annual rainfall for T years recurrence interval
Z = Mean annual precipitation of the sample
CV= Coefficient of variation of sample (ratio of standard deviation
to mean)
CT= Frequency factor, and is determined as follows
 2.45
 0.577  lnln T  ln T  1 (5)
3.1416

Example

The annual precipitation for station A is given below.

Check the adequacy of length of record using the
Mockus model

Year Annual rain (mm) Year Annual rain (mm) Year Annual rain (mm)
1966 1084 1973 1254 1980 1697
1967 1047 1974 1166 1981 1849
1968 1255 1975 888 1982 1880
1969 1555 1976 1219 1983 1407
1970 1128 1977 1213 1984 1380
1971 1398 1978 1188 1985 1556
1972 1283 1979 1468 1986 1250

5
 8/10/2015

Solution
 Here, y = 21
Ymin  4.3 t10 log Q2 2  6
Where
Ymin = Minimum acceptable years of record (?)
t10 = 1.34 (from table), for 10% significance level and
(21-6 = 15) Degrees of Freedom
Z = 1341.2 mm
SD = 255.7 mm

SD 255.7
CV  
 0.1907
1341.2
Z

Solution
 Using (5) for T =2,
 2.45
C2  0.577  lnln 2  ln 1   0.164
3.1416
 Similarly, for T = 100,
 2.45
C100  0.577  lnln 100  ln 99  3.137
3.1416
 Using (4), for T = 2,

Z2  1341.21  0.1907 0.164  1299.2652

 Similarly, for T = 100,

Z100  1341.21  0.1907 3.137   2143.5406

6
 8/10/2015

Solution
 Using (3)
ZT  100 2143.5406
Q2    1.6498
ZT  2 1299.2652
 Putting values in (2)

  4.3 x 1.34 x log1.6498  6  7.5693

2
Ymin

Ymin = 8 years

 Hence, available length of record (21 years) is adequate

Test for Randomness or Persistence Test

 Hydrologic data must be checked for independence

 Independence means that the outcome of hydrologic

variable (rainfall amount) at a given time does not
depend on the value of variable at previous time

 The following two tests can be used for checking the

independence of the hydrologic data

 Lag one serial correlation test

 Turning point test

7
 8/10/2015

Lag One Serial Correlation Test

1 N 1
  X  X  X i 1  X 
r1 
N  1 i 1 i
(6)
 X i  X 
1 N 2
N i 1
Where N is the number of data points, and X is the hydrologic
variable

 For data to be random lag one serial correlation

should be within:


1
 1.96
N  2
N 1 (7)
N  1 2
3

Example

The annual rainfall for station A is given below. Check

the randomness of data using lag-one serial
correlation coefficient.

Year Annual rainfall Year Annual rainfall Year Annual rainfall

(mm) (mm) (mm)
1966 1084 1973 1254 1980 1697
1967 1047 1974 1166 1981 1849
1968 1255 1975 888 1982 1880
1969 1555 1976 1219 1983 1407
1970 1128 1977 1213 1984 1380
1971 1398 1978 1188 1985 1556
1972 1283 1979 1468 1986 1250

8
 8/10/2015

Year Flow Mean

x x_bar xi - x_bar xi+1 - x_bar Col D * Col E Col D * Col D Numerator Denominator r1
1966 1084 1341.19 -257.19 -294.19 75662.99 66146.94 35938.146 62265.30 0.577
1967 1047 -294.19 -86.19 25356.42 86548.04

1968 1255 -86.19 213.81 -18428.34 7428.80

N 1
  X i  X  X i 1  X 
1969 1555 213.81 -213.19 -45582.15 45714.51
1
1970 1128 -213.19 56.81 -12111.25 45450.18 r1 
N  1 i 1

 X i  X 
1971 1398
1 N 2
56.81 -58.19 -3305.77 3227.32

1972 1283 -58.19 -87.19 5073.66 3386.13

N i 1
1973 1254 -87.19 -175.19 15274.94 7602.18

1974 1166 -175.19 -453.19 79394.66 30691.70

1975 888 -453.19 -122.19 55375.56 205381.61

1976 1219 -122.19 -128.19 15663.66 14930.51


1
 1.96
N  2
1977 1213 -128.19 -153.19 19637.56 16432.80
N 1 N  1 2
3
1978 1188 -153.19 126.81 -19426.01 23467.32

1979 1468 126.81 355.81 45120.04 16080.66

Range is calcutaed
1980 1697 355.81 507.81 180683.46 126600.42

as -0.466 to 0.366;
1981 1849 507.81 538.81 273612.61 257870.51

1982 1880 538.81 65.81 35458.80 290315.70

1983 1407 65.81 38.81 2554.04 4330.89 Since r1 calculated is

1984 1380 38.81 214.81 8336.66 1506.18 outside the range,
1985 1556 214.81 -91.19 -19588.58 46143.13 data are not random
1986 1250 -91.19 8315.70

718762.92 1307571.24

Turning Point Test

 If
Xi-1 < Xi >Xi+1 (8a)
Or
Xi-1 > Xi <Xi+1 (8b)

 Then Xi is assigned a score of 1 otherwise 0

 This way the Total score, SN, for a data series is

found

9
 8/10/2015

Turning Point Test

 The Normal Standard deviate, u, is then calculated as

follows:   2
 SN   N  2 
u   3 
 16 N  29  (9)
 90 

Where, SN is the score number, and N is the number of data points

 If ‘u’ lies within  1.96 , then series is random

Example

 For the given annual rainfall data, check the

randomness of data using turning point test.

Year Annual rain (mm) Year Annual rain (mm) Year Annual rain (mm)
1966 1084 1973 1254 1980 1697
1967 1047 1974 1166 1981 1849
1968 1255 1975 888 1982 1880
1969 1555 1976 1219 1983 1407
1970 1128 1977 1213 1984 1380
1971 1398 1978 1188 1985 1556
1972 1283 1979 1468 1986 1250

10
 8/10/2015

Solution
Year Rainfall (mm) Score Year Rainfall (mm) Score
 Score Number,
1966 1084 1977 1213 0
1967 1047 1 1978 1188 1 SN = 10
1968 1255 0 1979 1468 0
 2 
1969 1555 1 1980 1697 0
10  21  2 
1970 1128 1 1981 1849 0 u   3   1.46
1971 1398 1 1982 1880 1  16 x 21  29  
1972 1283 0 1983 1407 0  90 
1973 1254 0 1984 1380 1
1974 1166 0 1985 1556 1  Since calculated u lies
1975 888 1 1986 1250 between  1.96 , the
1976 1219 1
data series is random

Methods for Frequency Analysis

 There are two methods for performing frequency
analysis

 Empirical or Graphical method

 Frequency factor method

 Empirical Method

 The exceedence probability of the event is obtained by the

use of empirical formula, known as plotting position formula

 Several plotting position formulae are developed and some

of them are given below:

11
 8/10/2015

Plotting position formulae

Method P (probability)
m
California N (10)
m  0 .5
Hazen N
(11)
m
Weibull N 1 (12)
m  0.3
Chegodayev (13)
N  0.4
m  0.44
Blom (14)
N  0.12
 Where m is rank assigned to the data after arranging them in
descending order of magnitude

 Thus, the maximum value is assigned m =1, the second largest value (m
=2), and the lowest value m =N, N being the number of records

 Weibull formula is the most commonly used plotting position formula

Empirical Method
 Having calculated P and T for all the 20
Rainfall Frequency Curve

events in the series, the variation of 17

Rainfall (cm)

rainfall magnitude is plotted against

14
11
the corresponding T on semi-log or 8

log-log paper 5
1 10 100
Return period, T (years)

 The rainfall magnitude for any recurrence interval can be determined

by extrapolating the plot between magnitude and recurrence interval

 Empirical procedures can give good results for small extrapolations

but the errors increase with the amount of extrapolation

 For more accurate results, analytical methods using frequency factor

are used

12
 8/10/2015

Example
 For a station A, the recorded annual 24-h maximum rainfall is
given below

 Estimate the 24-h maximum rainfall with return periods of 13 and

50 years.
 What would be the probability of a rainfall of magnitude equal to
or exceeding 10 cm occurring in 24 h at station A?
Year Rainfall (cm) Year Rainfall (cm) Year Rainfall (cm)
1950 13.0 1957 12.5 1964 8.5
1951 12.0 1958 11.2 1965 7.5
1952 7.6 1959 8.9 1966 6.0
1953 14.3 1960 8.9 1967 8.4
1954 16.0 1961 7.8 1968 10.8
1955 9.6 1962 9.0 1969 10.6
1956 8.0 1963 10.2 1970 8.3
1971 9.5

Solution
 Selecting Weibull Formula for Plotting Position
m Rainfall, cm P =m/n+1 T=1/P m Rainfall, cm P =m/n+1 T=1/P
1 16 0.043 23.00 12 9 0.522 1.92
2 14.3 0.087 11.50 13 8.9 0.565 1.77
3 13 0.130 7.67 14 8.9 0.609 1.64
4 12.5 0.174 5.75 15 8.5 0.652 1.53
5 12 0.217 4.60 16 8.4 0.696 1.44
6 11.2 0.261 3.83 17 8.3 0.739 1.35
7 10.8 0.304 3.29 18 8 0.783 1.28
8 10.6 0.348 2.88 19 7.8 0.826 1.21
9 10.2 0.391 2.56 20 7.6 0.870 1.15
10 9.6 0.435 2.30 21 7.5 0.913 1.10
11 9.5 0.478 2.09 22 6 0.957 1.05

13
 8/10/2015

Solution
Rainfall Frequency Curve
20
17

Rainfall (cm)
14
11
8
5
1 10 100
Return period, T (years)
 After interpolating and extrapolating the above graph, we can
determine rainfall magnitude for 13 and 50-year return period
respectively

13 year RI Rainfall = 14.55 cm

50 year RI Rainfall = 18.00 cm

 For Rainfall = 10 cm, T =2.4 years and P = 0.417

Intensity Duration Frequency Relationship

 Many studies require rainfall intensities of different
durations and return periods.
 watershed management, runoff disposal and

erosion control,
 The relationship between intensity (i, cm/hr),
duration (D, hr) and return period (T, years) can be
expressed as follows:

Where K, x, a and n are constants for a given catchment.

14
 8/10/2015

IDF Curve

Values of constants in IDF eqn – (Source: CSWCRTI- Dehradun)

City K x a n
Bhopal 6.93 0.189 0.50 0.878
Nagpur 11.45 0.156 1.25 1.032
Chandigar 5.82 0.160 0.40 0.750
h
Bellary 6.16 0.694 0.50 0.972
Raipur 4.68 0.139 0.15 0.928

15
 8/10/2015

Example
 Compute 10 year, 1 h design rainfall intensity
for Bhopal and Nagpur
Solution:
For Bhopal

6.9310 
0.189
10.708
i    7.50 cm/hr
1  0.50 0.878
1.427

For Nagpur

11.4510 
0.156
16.398
i    7.10 cm/hr
1  1.251.032
2.309

16