Frequency Analysis

Lecture 9

Dr. Shahid Ali

National University
of Computer & Emerging Sciences DEPARTMENT OF CIVIL ENGINEERING
 Frequency Analysis

• The purpose of frequency analysis is to estimate frequency of occurrence of

floods, droughts or storms of maximum rainfall. Hydrologic parameters from
available record of data are predicted.
• Using these parameters the data is extrapolated for future forecasting.
• The accuracy of prediction depends on accuracy and correctness of available
records.

Frequency Analysis is based on statistical methods.

 The following factors can affect the frequency analysis

 Non-cyclic nature of random events like floods and droughts,

 Analysis based on limited data, and

 Data being non-homogeneous

Return Period

It is an average value of time of occurrence of hydrologic outcome. It is an

average and not the exact period of becoming equal to or exceedance from a
certain value of a hydrologic quantity.

For example a flood of say 50,000 cumecs that has been exceeded on the average
once in 20 years has a probability of exceedance from this value in any one year
equal to 1/20. This does not mean that every 20 years a flood of this magnitude
will occur.

Return period is also called recurrence interval.

Probability
It is measure of chance for an event to exceed or equal a certain level.

 For an event ‘A’, ‘probability of A’ denoted by ‘P(A)’ is the ratio of the

number of possible outcomes favorable to A (denoted by ‘NA’) to the total
number of possible outcomes for the hydrologic experiment / phenomenon
(denoted by ‘N’).

Mathematically:

P(A) = NA / N
 Relationship between Probability and Return Period
Relationship of probability of occurrence of a flood ‘p’ in any year with its
return period ‘T’ is given as:

p=1/T

The probability that the flood will not occur in any year is given as:

p = 1- p = 1 – 1 / T

The probability that the flood will not occur for ‘n’ successive years is given as:
n

 p   11 / T n
 
 
 Risk

The probability that the flood of given magnitude (having return period ‘T’)
will occur at least once in ‘n’ successive years is called risk (denoted by ‘R’).

n

 p   1  11 / T n
R = 1-  
 
Random Variable

If outcomes of a hydrologic experiment or phenomenon are quantitative measurements

(numerical values) and the outcomes depend on chance, then the resulting measurement
variable is called a random variable. It is also called a chance variable, a stochastic variable
or simply a variate.

Probability Distribution

A function f(x) of hydrologic quantity is a probability function or a probability distribution

of the discrete random variable ‘X’ if for each possible outcome ‘x’:

• f(x) > 0
• Sf(x) = 1
• P(X = x) = f(x)

A graph of all possible values of random variable along with their respective probabilities
shows a probability distribution or a probability function
Histogram

A histogram consists of a set of adjacent rectangles whose bases are marked off by class
boundaries along the x-axis and whose heights are proportional to the frequencies
associated with respective classes.
Example 1

Table 1, below, gives precipitation depth in ‘cm’ for first eight months at certain location.
Construct the frequency distribution diagram.

Table 1 Precipitation data

Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Jan 7.5 8.2 6.2 8.9 7.8 5.4 9.4 9.9 10.9 10.8 7.4
Feb 9.7 11.6 12.6 5 10.2 9.2 12 9.9 7.3 7.3 8.4
Mar 10.3 10.1 10 11.1 6.5 12.5 7.8 6.5 8.7 9.3 12.4
Apr 10.6 9.1 9.7 9.3 6.2 10.3 6.6 7.4 8.6 7.7 9.4
May 7.7 12.8 8.7 5.5 8.6 9.6 11.9 10.4 7.8 7.6 12.1
Jun 4.6 14 8.1 11.4 10.6 11.6 10.4 8.1 4.6 6.6 12.8
Jul 6.8 7.1 6.6 8.8 8.8 10.7 10.8 6 7.9 7.3 9.3
Aug 9.3 8.9 10.1 3.9 6 6.9 9 8.8 9.4 11.4 10.9

Solution
We note that the largest value is 14.0 and the smallest value is 3.9 so that the Range
is 14.0 - 3.9 = 10.1 cm.

 As the data are recorded to one decimal place, we may therefore locate the lower limit
of the first group at 3.5. Let us choose a class interval of 1.0. Then the class limits are
specified as 3.5 - 4.4, 4.5 - 5.4, 5.5 - 6.4,…. With this choice, the class boundaries are
3.45 - 4.45, 4.45 - 5.45, 5.45 - 6.45, … , which do not coincide with the given values.
Table 2, shows the required frequency distribution.
 Frequency Distribution of Rainfall Depths

Precipitation Depth Class Boundaries Frequency (No. of

Interval Occurrences)
3.5-4.4 3.45-4.45 1
4.5-5.4 4.45-5.45 4
5.5-6.4 5.45-6.45 5
6.5-7.4 6.45-7.45 13
7.5-8.4 7.45-8.45 12
8.5-9.4 8.45-9.45 19
9.5-10.4 9.45-10.45 13
10.5-11.4 10.45-11.45 10
11.5-12.4 11.45-12.45 6
12.5-13.4 12.45-13.45 4
13.5-14.4 13.45-14.45 1
Total 88
 Histogram for Frequency Distribution

20 19

18

16

14 13 13
12
12
10
10

8
6
6 5
4 4
4

2 1 1

0
3.45-4.45

4.45-5.45

5.45-6.45

6.45-7.45

7.45-8.45

8.45-9.45

9.45-10.45

10.45-11.45

11.45-12.45

12.45-13.45

13.45-14.45
Fig. shows the histogram for frequency distribution of monthly precipitation depths.
Along X-axis are class boundaries and along Y-axis is the frequency
Example 2

Find the probability that a 25 year flood will occur in particular 5 years period.

Solution

Here P(X≥25 Year Flood) is required to be determined in particular 5 year period. That
is:
‘n’ = 5
‘T’ = 25
From Equation 9.12
P(X≥25 Year Flood) = 1-[1-(1/25)]5
= 1-0.82
= 0.18 or 18%

A simplification of Equation yields

n = log (1-P(X≥x)n)/log[(T-1)/T]
Example 3

An embankment is designed for period of 20 years flood. Find, for even

chance, the life of embankment.

Solution

P(X≥20 Year Flood) = 50% = 0.50

T = 20 years
n=?
From Equation 9.14, we have
n = log (1-P(X≥x)n) / log[(T-1) / T]

= log (1-0.5) / log [(20-1) / 20]

= 13.5 years.
Statistical Methods of Analysis of Hydrologic Data
There are two main methods for frequency analysis – Computational and
Graphical.

Water resources projects often require frequency distribution of magnitudes. For

example in design of hydraulic structures flood of certain magnitude and of certain
return period is required. This needs predictions based on previous data. The data is
extrapolated by matching the data with statistical distribution. There are different
distributions of probability for hydrologic data, some of which are listed below:

 Normal Distribution
 Lognormal Distribution
 Exponential Distribution
 Gamma Distribution
 Pearson Type III Distribution
 Log Pearson Type III Distribution
 Gumbel’s Extreme Value Distribution
The purpose of all these distributions is to best extrapolate the hydrologic
data. Each distribution has certain advantages. Since a Normal Distribution is
defined by only two parameters - Mean and Standard Deviation, therefore it is
simple to use. Its details can be seen from any book on Statistics. Gumbel’s
Extreme Value Distribution, one of the commonly used distribution in
Hydrology is described below:
 Gumbel’s Extreme Value Theory
Gumbel assumed an exponentially distributed variable. He defined the cumulative probability
‘P’ as

P’ = Exp(e-y)

Therefore,
y = -ln [-ln[1-(1/T)],
1
𝑇=
1−𝑃
The event ‘QT’, having return period ‘T’ years is given by:
QT = Qav + б (0.78y-0.45)
Qav = Average of all values of ‘annual flood’, Qm
б = Standard deviation
Therefore,
б = SQRT [{n/(n-1)}{{∑Q²m/n}-Q²av}]
n=Number of years of record
The relationship between T and y is shown in figure
 1,200.00

1,000.00

800.00

600.00
T

400.00

200.00

-
- 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00
y

‘T’ Vs ‘y’ Plot

Log Pearson Type III Distribution

According to this N-year event can be calculated in a similar manner to the

normal distribution method, with Skew Coefficient ‘G’ given by:

 n²( X ³)  3n( X )( X ²)  2( X )³ 

G=    

 n ( n 1)( n 2) ³ x 
Where, X = log Q and бx = Standard Deviation of ‘n’ values of ‘X’

log Qr = Xav + K’ бx)

Where, ‘K’ is a factor depending upon probability ‘r’ and Skew Coefficient
‘G’
 Pearson Type III Distribution – ‘K’ Values
Skew Probability of Exceedance (%)
Coefficient 'G' 99 90 50 10 5 2 1
3.0 -0.667 -0.660 -0.396 1.180 2.003 3.152 4.051
2.5 -0.799 -0.771 -0.360 1.250 2.012 3.048 3.845
2.0 -0.990 -0.895 -0.307 1.303 1.996 2.912 3.605
1.5 -1.256 -1.018 -0.240 1.333 1.951 2.743 3.330
1.2 -1.449 -1.086 -0.195 1.340 1.910 2.626 3.149
1.0 -1.558 -1.128 -0.164 1.340 1.877 2.542 3.023
0.9 -1.660 -1.147 -0.148 1.339 1.859 2.498 2.957
0.8 -1.733 -1.166 -0.132 1.336 1.839 2.453 2.891
0.7 -1.806 -1.183 -0.116 1.333 1.819 2.407 2.824
0.6 -1.880 -1.200 -0.099 1.328 1.797 2.359 2.755
0.5 -1.955 -1.216 -0.083 1.323 1.774 2.311 2.686
0.4 -2.029 -1.231 -0.067 1.317 1.750 2.261 2.615
0.3 -2.104 -1.245 -0.050 1.309 1.726 2.211 2.544
0.2 -2.178 -1.258 -0.033 1.301 1.700 2.159 2.472
0.1 -2.253 -1.270 -0.017 1.292 1.673 2.107 2.400
0.0 -2.326 -1.282 0.000 1.282 1.645 2.054 2.326
-0.1 -2.400 -1.292 0.017 1.270 1.616 2.000 2.253
-0.2 -2.472 -1.301 0.033 1.258 1.586 1.945 2.178
-0.3 -2.544 -1.309 0.050 1.245 1.555 1.890 2.104
-0.4 -2.615 -1.317 0.067 1.231 1.524 1.834 2.029
-0.5 -2.686 -1.323 0.083 1.216 1.491 1.777 1.955
-0.6 -2.755 -1.328 0.099 1.200 1.458 1.720 1.880
-0.7 -2.824 -1.333 0.116 1.183 1.423 1.663 1.806
-0.8 -2.891 -1.336 0.132 1.166 1.389 1.606 1.733
-0.9 -2.957 -1.339 0.148 1.147 1.353 1.549 1.660
-1.0 -3.023 -1.340 0.164 1.128 1.317 1.492 1.588
-1.2 -3.149 -1.340 0.195 1.086 1.243 1.379 1.449
-1.5 -3.330 -1.333 0.240 1.018 1.131 1.217 1.256
-2.0 -3.605 -1.303 0.307 0.895 0.949 0.980 0.990
-2.5 -3.845 -1.250 0.360 0.771 0.790 0.798 0.799
-3.0 -4.051 -1.180 0.396 0.660 0.665 0.666 0.667
Example
The following Table shows Maximum Annual Discharge for Jhelum River. Estimate the flood
discharge for a return period of 100 years. Assume an appropriate distribution
 Year Qm (m³/s) Qm2 (m³/s) Year Qm (m³/s) Qm2 (m³/s)

1,937 167.240 27,969 1,967 172.400 29,722

1,938 197.440 38,983 1,968 154.960 24,013
Maximum Annual Discharge at Jhelum River 1,939
1,940
186.080
146.080
34,626
21,339
1,969
1,970
187.160
98.760
35,029
9,754
1,941 137.000 18,769 1,971 98.440 9,690
1,942 168.880 28,520 1,972 178.760 31,955
1,943 178.400 31,827 1,973 180.400 32,544
1,944 138.720 19,243 1,974 121.680 14,806
1,945 171.480 29,405 1,975 157.200 24,712
1,946 105.880 11,211 1,976 212.520 45,165
1,947 116.960 13,680 1,977 145.920 21,293
1,948 209.240 43,781 1,978 167.880 28,184
1,949 201.400 40,562 1,979 137.800 18,989
1,950 237.720 56,511 1,980 160.520 25,767
1,951 151.600 22,983 1,981 180.920 32,732
1,952 148.400 22,023 1,982 136.560 18,649
1,953 120.760 14,583 1,983 185.120 34,269
1,954 167.040 27,902 1,984 139.640 19,499
1,955 135.960 18,485 1,985 106.480 11,338
1,956 187.080 34,999 1,986 174.040 30,290
1,957 221.040 48,859 1,987 214.760 46,122
1,958 202.040 40,820 1,988 215.880 46,604
1,959 296.600 87,972 1,989 163.600 26,765
1,960 122.480 15,001 1,990 212.080 44,978
1,961 127.920 16,364 1,991 235.560 55,489
1,962 97.360 9,479 1,992 209.120 43,731
1,963 175.280 30,723 1,993 189.000 35,721
1,964 197.240 38,904 1,994 185.640 34,462
1,965 216.920 47,054 1,995 204.080 41,649
1,966 172.480 29,749
Mean = 170 Total =
1,796,243.75
From Gumbel Extreme Value Distribution, given above
QT = Qav + б (0.78y-0.45)
Where,
y = -ln[-ln(1- 1/T )]
б=√[n/(n-1)]∑Qm2/n-Qav2
n = number of values
∑Qm2 = sum of squares of recorded discharge
Substituting values,
б= 39.67
For T = 100 years, y = 4.60
Q100 = 170 + 39.67 (0.78*4.60-0.45) =294.50 m³/s
Graphical Method
The extreme event can also be found by plotting. Different researchers have
proposed different formulae to relate return period with given variable. Some of
them are listed in table. Here m is the rank of the given variable. It has value of
m=1 for highest value of given variable, m=2 for next highest value and so on.
Also N is the number of the event.

Return Period Formulae

Sr. No. Formula Proposed by
1 Tr = (N+1)/m Weibull
2 Tr = N/m Californian
3 Tr = 2 N/(2m-1) Hazen

4 Tr = (N+0.12)/ (m-0.44) Gringorten

5 Tr = (N+0.2)/ (m-0.4) Cunnane

Example
The Maximum Annual Discharge for Jhelum River are given in Previous Example. Estimate
the flood discharge for a return period of 100 years by graphical method.

Tr=(N+0.12)/
Water Year Qm (m³/s) Arranged (m³/s) Rank 'm'
(m-0.44)
1,937 167.240 167.240 1 105.57
1,938 197.440 197.440 2 37.90
1,939 186.080 186.080 3 23.09
1,940 146.080 146.080 4 16.61
1,941 137.000 137.000 5 12.96
1,942 168.880 168.880 6 10.63
1,943 178.400 178.400 7 9.01
1,944 138.720 138.720 8 7.82
1,945 171.480 171.480 9 6.91
1,946 105.880 105.880 10 6.18
1,947 116.960 116.960 11 5.60
1,948 209.240 209.240 12 5.11
1,949 201.400 201.400 13 4.71
1,950 237.720 237.720 14 4.36
1,951 151.600 151.600 15 4.06
1,952 148.400 148.400 16 3.80
1,953 120.760 120.760 17 3.57
1,954 167.040 167.040 18 3.37
1,955 135.960 135.960 19 3.19
1,956 187.080 187.080 20 3.02
1,957 221.040 221.040 21 2.88
Solution

The calculations are done in spreadsheet below (Table 9.6). The first and second
column are water year and Annual maximum discharge obtained from stream flow
measurements. In third column values are arranged in descending order. A rank is
assigned to each discharge value starting with m=1 assigned to highest figure. The
fourth column shows calculation of return period Tr. Different formulae have been
proposed by different researchers. Here the used formula is Tr = (N+0.12)/(m-0.44)
proposed by Gringorten.

Computations for 100 year Flood

 Tr=(N+0.12)/
Water Year Qm (m³/s) Arranged (m³/s) Rank 'm'
(m-0.44)
1,958 202.040 202.040 22 2.74
1,959 296.600 296.600 23 2.62
1,960 122.480 122.480 24 2.51
1,961 127.920 127.920 25 2.41
1,962 97.360 97.360 26 2.31
1,963 175.280 175.280 27 2.23
1,964 197.240 197.240 28 2.15
1,965 216.920 216.920 29 2.07
1,966 172.480 172.480 30 2.00
1,967 172.400 172.400 31 1.93
1,968 154.960 154.960 32 1.87
1,969 187.160 187.160 33 1.82
1,970 98.760 98.760 34 1.76
1,971 98.440 98.440 35 1.71
1,972 178.760 178.760 36 1.66
1,973 180.400 180.400 37 1.62
1,974 121.680 121.680 38 1.57
1,975 157.200 157.200 39 1.53
1,976 212.520 212.520 40 1.49
1,977 145.920 145.920 41 1.46
1,978 167.880 167.880 42 1.42
1,979 137.800 137.800 43 1.39
1,980 160.520 160.520 44 1.36
1,981 180.920 180.920 45 1.33
1,982 136.560 136.560 46 1.30
1,983 185.120 185.120 47 1.27
1,984 139.640 139.640 48 1.24
1,985 106.480 106.480 49 1.22
1,986 174.040 174.040 50 1.19
1,987 214.760 214.760 51 1.17
1,988 215.880 215.880 52 1.15
1,989 163.600 163.600 53 1.12
1,990 212.080 212.080 54 1.10
1,991 235.560 235.560 55 1.08
1,992 209.120 209.120 56 1.06
1,993 189.000 189.000 57 1.05
1,994 185.640 185.640 58 1.03
1,995 204.080 204.080 59 1.01
The values of discharge are plotted on y-axis (linear scale) against computed return
period on x-axis (linear scale). A best fit line is drawn. The line is extrapolated to
find discharge against given return period and vice versa. We now try best-fit line

Plot (a)
Q is plotted against T with Linear Scale both along X-axis and Y-axis as shown in Fig.
Annual Maximum Flow (River Jhelum 1937-1995)
350

300

250
Discharge (m³/s)

200

150

100

50

-
- 20 40 60 80 100 120

Return Period Tr (Years)

Fig .Linear Axis Plotting

From plotted curve,

for Tr = 100 years, Q ~ 298 m³/s.
Plot (b)
Q s plotted on Linear Scale and T with Logarithmic Scale as shown in Fig.

Annual Maximum Flow at River Jhelum (1937-1995)

SEMI-LOG PLOT

Discharge Q (m³/s) - Linear Scale 350

300

250

200

150

100

50

-
1 10 100

Return Period Tr (years) - Logarithmic Scale

Fig. Q vs T plot with Q at Linear Scale and T at Logarithmic Scale

From Fig
or T=100 years, Q= 305 m³/sec.
This value of flow may be required for hydraulic design of a bridge across the river.
Plot (c)
Q is plotted on Linear Scale along Y-axis and Probability in % along X-axis

Plot (d)
Q is plotted on Log Scale along Y-axis and Probability in % along X-axis
Problem
Flood-Frequency computations for a river by using Gumbel’s Method, yielded the
following results. Estimate the flood magnitude in this river with a return period of
500 years.

Return Period T (years) Peak Flood (m³/s)

50 40,809

100 46,300
Problem

Annual maximum recorded floods in a river for the period 1951 to 1977, is given
below. Verify weather the Gumbel’s Extreme-Value Distribution fits the recorded
values. Estimate the flood discharge with recurrence interval of 100 years. Also,
estimate the flood magnitude in this river with a return period of 500 years.

Maximum Flood Maximum Maximum Flood

Year Year Year
(m³/s) Flood (m³/s) (m³/s)
1951 2947 1960 4798 1969 6599
1952 3521 1961 4290 1970 3700
1953 2399 1962 4652 1971 4175
1954 4124 1963 5050 1972 2988
1955 3496 1964 6900 1973 2709
1956 2947 1965 4366 1974 3873
1957 5060 1966 3380 1975 4593
1958 4903 1967 7826 1976 6761
1959 3757 1968 3320 1977 1971