CIVL3155 HYDROLOGY AND FREE SURFACE FLOWS

Hydrological Risk
September 2023

Intended learning outcomes

After successfully completing this module you should be able to:

• Define the annual exceedance probability.

• Define hydrologic risk.

• Estimate empirical average recurrence intervals and exceedance probabilities for an annual
flood series

• Fit a probability distribution to the observed annual flood series

• Apply the best-fit probability distribution to estimate the peak discharge of a required design
average recurrence interval

Pre-reading material
The pre-reading material for this tutorial is Chapter 10 of Mays (2012) “Ground and Surface Water
Hydrology”, John Wiley & Sons, USA, 617 pp. You can access this text via The University of
Queensland Library (library call number GB661.2.M388 2012).

Alternatively students can access the material from chapter 13 from the textbook by Brutsaert (2005)
with particular emphasis on sections 13.1-13.4. Reference: Brutsaert, Wilfried (2005). Hydrology -
An Introduction. Cambridge University Press. An electronic copy of the text is available from The
University of Queensland’s Library.

Students that are interested in gaining additional background information on groundwater properties
and well hydraulics encouraged to explore the range of “hydrology” texts that are available both
online and in The University of Queensland’s Library. Most modern texts cover this material in
similar ways.

Page 1 of 21
1. Consider the data in table below for annual rainfall at College Station (Texas) (data sourced from
Chow et al (1988) Applied Hydrology. McGraw-Hill. Singapore, 572 pp.).

Table 1 – Summary of annual rainfall for College Station

Year 1910 1920 1930 1940 1950 1960 1970
0 -- 1237.0 1137.9 1252.2 792.5 1168.4 861.1
1 1013.5 1120.1 863.6 1122.7 685.8 1125.2 805.2
2 787.4 1087.1 1158.2 1059.2 939.8 960.1 800.1
3 1074.4 1229.4 947.4 782.3 1188.7 751.8 1513.8
4 1069.3 868.7 1110.0 1361.4 683.3 891.5 1282.7
5 1043.9 823.0 1061.7 876.3 645.2 1262.4 980.4
6 729.0 1178.6 1051.6 1277.6 584.2 929.6 1102.4
7 426.7 988.1 792.5 1112.5 1435.1 825.5 729.0
8 866.1 947.4 894.1 548.6 1102.4 1567.2 812.8
9 1432.6 1285.2 891.5 1196.3 1049.0 1204.0 1315.7

Use relative frequencies to estimate the probability that the annual rainfall in any year will be i) less
than 890mm? [1 mark] ii) larger than 1140mm? [1 mark], and iii) in the range 890 – 1140 mm [1
mark]

2. Consider the data in table above for annual rainfall at College Station (Texas) (data sourced from
Chow et al (1988) Applied Hydrology. McGraw-Hill. Singapore, 572 pp.). Estimate the probability
that there will be 2 successive years with rainfall less than 890mm? State any assumptions used.
Compare your estimate with the actual relative frequency of this occurrence in the data. [1 mark]

3. Fit the Standardised Normal Distribution to the last 10-years of data in table 1 above and use it to
calculate corresponding probabilities for question 1 above. Compare the results with your frequency
estimates from question 1. [3 marks]

4. Roughly 110 independent storms occur across the year for a tropical city in Queensland. Their
average duration is 5.3 hours. In a year (8760 h) the average interarrival time will be (8760 – 110 ×
5.3) / 110 = 74.3 h. Now the Exponential distribution, f(x) = λ e-λx, is useful for describing the
distribution of interarrival times, x (h). Fit the exponential distribution (using results above) by the
Method of Moments (in notes: λ ≈ 1/average; P(X ≤ xi) = F(xi) = ∫f(x)dx = 1 - exp(-λx)) and estimate

Page 2 of 21
the probability that at least 4 days (96h) elapse between storms. [3 marks]

5. Using the same information and distribution from the above question estimate the probability that
exactly 12 h elapses between storms. [3 marks]

6. Using the same information and distribution from the above questions estimate the probability that
the separation time between any 2 storms is no more than 12 h. [3 marks]

7. Consider a series of Annual Maximum Floods of length N. Determine the AEP (probability of
exceedance) of the 2nd highest ranking flood and the smallest flood using (a) the relative frequency
method (i.e., rank/N) [1 mark], (b) Gringorten's Gumbel formula [1 mark] and (c) the Weibull
formula [1 marks] for N = 10, N = 50, N = 100. Comment on which is the most appropriate estimate
and differences between the formulae.

8. Consider the Standard Normal distribution for an Annual Series with N observations illustrated
below:

f(z )
0.4

0.3

0.2

0.1

0
-3 -2 -1 0 1 A 2 3

A
What is the AEP of a flow event with a z value corresponding to Point A if
−
 f (z)dz = 0.92 ? What
will be the rank of this storm event to the nearest integer if N=35? [2 marks]

Page 3 of 21
9. What is the probability that at least one flood of AEP = 0.02 (ARI = 50 years, T = 50 years) will
occur during the 30 year design life of a flood control project? [1 mark]

10. In an N year sequence, risk is defined as the probability of occurrence of a particular event in at
least one year (in an N year sequence). Reliability is equal to (1 - risk). Consider a small, temporary
(3 year design life) flood mitigation dam designed to contain the 20 year flood event, what is the
reliability of the designed dam (i.e., what is the probability it won’t be overtopped in the 3 years)?
Also estimate the risk that it will be overtopped at least once in the design life. [1 mark]

Page 4 of 21
11. Consider the annual series given in the table below for a small stream (the Jason St gauge). Use
the data to estimate the AEP for a flood of 60 m3 s-1 using a plotting position probability estimate [1
mark]. Also fit a Gumbel distribution to the data as well as a Log-Normal distribution to derive an
AEP estimate for a flood flow of 60 m3 s-1 [1 mark].

Table. Jason St gauge annual maximum series.

Year Flow Year Flow

(m3/sec) (m3/sec)

1972 33.4 1990 27.9

1974 133.4 1991 15.3

1975 6.7 1992 45.8

1976 26.6 1993 1.5

1976 11.5 1994 58.6

1978 13.7 1995 44.8

1979 20.1 1996 32

1980 24.1 1997 13.4

1981 20.9 1998 13.8

1981 34.8 1999 23.1

1983 21.5

1984 34.4

1985 23.2

1985 8.3

1986 18.3

1988 44.6

1989 66.4

Page 5 of 21
12. Given the annual flood series statistics in the table below, calculate and tabulate flows for the 1
in 2, 5, 10, 25, 50 and 100 years floods assuming Extreme Value Distribution (EVD) and Log-Pearson
Type III distribution (LPT3), respectively [2 marks].

Mean of floods (m3/s): 1430

Standard deviation of floods (m3/s): 1660
Mean of logarithms of floods: 2.818
Standard deviation of logarithms of floods: 0.641
Skewness coefficient of logarithms of -0.458
floods:

13. If a structure can withstand a flood with a peak discharge no greater than 1,950 m 3 s-1 and has
been designed to have an economic life span of 100 years. Determine the risk of failure assuming the
Extreme Value Distribution when the mean and standard deviation of the annual flood series are 410
m3 s-1 and 280 m3 s-1, respectively [2 marks]. The Extreme Value Type 1 (EV1) distribution (also
known as the Gumbel Distribution) is described by:

   X − a  
F ( X ) = exp  − exp  −   
   b  

where F(X) is the probability of an annual maximum Q ≤ X and a and b are fitting parameters defined
by:

Q 6
a = Q − 0.5772b b=


where μQ and σQ are the population mean and standard deviation respectively.

Page 6 of 21
14. The table below shows the classified monthly precipitation data measured at a station. The data
are divided into nine categories. The exponential distribution is often used to approximate rainfall
data such that P(X ≤ xi) = F(xi) = ∫f(x)dx = 1 - exp(-λx)). If λ = 0.105, use the chi-square test with 5%
level of significance to evaluate whether the exponential distribution provides an adequate fit to the
observed data [Example sourced from Karamouz et al., 2013]. [3 marks]

Precipitation
depth class
Interval [mm] Actual frequency fa
1 0≤x<5 38
2 5 ≤ x < 10 26
3 10 ≤ x < 15 12
4 15 ≤ x < 20 10
5 20 ≤ x < 25 8
6 25 ≤ x < 30 3
7 30 ≤ x < 35 2
8 35 ≤ x < 40 1
9 40 ≤ x 0
Sum 100

15. Find the average occurrence of a 10-year flood (return period of the flood is 10 years) in a 100-
year period? What is the probability that exactly this number of 10-year flood will occur in a 100-
year period?

Page 7 of 21
Solutions

1. Consider the data in table below for annual rainfall at College Station (Texas) (data sourced from
Chow et al (1988) Applied Hydrology. McGraw-Hill. Singapore, 572 pp.).

Table 1 – Summary of annual rainfall for College Station

Year 1910 1920 1930 1940 1950 1960 1970
0 -- 1237.0 1137.9 1252.2 792.5 1168.4 861.1
1 1013.5 1120.1 863.6 1122.7 685.8 1125.2 805.2
2 787.4 1087.1 1158.2 1059.2 939.8 960.1 800.1
3 1074.4 1229.4 947.4 782.3 1188.7 751.8 1513.8
4 1069.3 868.7 1110.0 1361.4 683.3 891.5 1282.7
5 1043.9 823.0 1061.7 876.3 645.2 1262.4 980.4
6 729.0 1178.6 1051.6 1277.6 584.2 929.6 1102.4
7 426.7 988.1 792.5 1112.5 1435.1 825.5 729.0
8 866.1 947.4 894.1 548.6 1102.4 1567.2 812.8
9 1432.6 1285.2 891.5 1196.3 1049.0 1204.0 1315.7

Use relative frequencies to estimate the probability that the annual rainfall in any year will be i) less
than 890mm? ii) larger than 1140mm?, and iii) in the range 890 – 1140 mm?

SOLUTION:
Total observations = 69

i) < 890mm? The number of observations satisfying this = 23. Therefore the probability P(<890 mm)
= 23/69 = 0.33

ii) > 1140mm? The number of observations satisfying this = 19; Therefore the probability P>1140
mm) = 19/69 = 0.28

iii) in the range 890 – 1140 mm? The probability P (890 – 1140 mm) = (1- 0.28) – 0.33 = 0.39

Page 8 of 21
2. Consider the data in table above for annual rainfall at College Station (Texas) (data sourced from
Chow et al (1988) Applied Hydrology. McGraw-Hill. Singapore, 572 pp.). Estimate the probability
that there will be 2 successive years with rainfall less than 890mm? State any assumptions used.
Compare your estimate with the actual relative frequency of this occurrence in the data.

SOLUTION:
Assuming that observations are independent the probability P = 0.33 x 0.33 = 0.11. Comparing this
with the actual relative frequency of this occurrence in the data shows there are 9 such pairs (out of
possible 68 pairs) and therefore the probability P 9/68 = 0.13 (reasonably close but not an exact
match).

3. Fit the Standardised Normal Distribution to the last 10-years of data in table 1 above and use it to
calculate corresponding probabilities for question 1 above. Compare the results with your frequency
estimates from question 1.

SOLUTION:
Using method of moments we have: Average = 1020.32; s = 270.00 and can then calculate Z the
standard normal variate: Z = (x-Average)/s

Thus, we want P(-0.48 < Z < 0.44) = F(0.44) – F(-0.48) = 0.67 – (1 – 0.6844) ≈ 0.35

Values for F(0.44) and F(-0.48) are read from the tables of cumulative probability of the standard
normal distribution (recall that F(-0.48) = 1 - F(0.48)).

Compare with your frequency estimates –not too bad?

Page 9 of 21
4. Roughly 110 independent storms occur across the year for a tropical city in Queensland. Their
average duration is 5.3 hours. In a year (8760 h) the average interarrival time will be (8760 – 110 ×
5.3) / 110 = 74.3 h. Now the Exponential distribution, f(x) = λ e-λx, is useful for describing the
distribution of interarrival times, x (h). Fit the exponential distribution (using results above) by the
Method of Moments (in notes: Statistical Hydrology i.e., λ ≈ 1/average;
P(X ≤ xi) = F(xi) = ∫f(x)dx = 1 - exp(-λx)) and estimate the probability that at least 4 days (96h) elapse
between storms.

SOLUTION:
λ ≈ 1/ave = 1/74.3 = 0.0135

The probability that at least 4 days (96h) elapse between storms is given by:

P(X ≥ xi) = 1 - F(xi) = 1 - ∫f(x)dx = 1 - 1- exp(-λx)

1-F(96) = exp(-0.0135 x 96) = 1 - 1 - 0.27 = 0.27

5. Using the same information and distribution from the above question estimate the probability that
exactly 12 h elapses between storms.

SOLUTION:
The probability that exactly 12 h elapses between storms is 0 (this is inferred by the definition of
continuous pdf – i.e., the probability of any exact value is zero. This is why we pose questions in
terms of greater than, less than or between two values)

6. Using the same information and distribution from the above questions estimate the probability that
the separation time between any 2 storms is no more than 12 h.

SOLUTION:
The probability that that the separation time between any 2 storms is no more than 12 h is given by:

P(X ≤ xi) = F(xi) = ∫f(x)dx = 1 - exp(-λx)

F(12) = 1- exp(-0.0135 x 12) = 1 - 0.85 = 0.15

Page 10 of 21
7. Consider a series of Annual Maximum Floods of length N. Determine the AEP (probability of
exceedance) of the 2nd highest ranking flood and the smallest flood using (a) the relative frequency
method (i.e., rank/N), (b) Gringorten's Gumbel formula and (c) the Weibull formula for N = 10, N =
50, N = 100. Comment on which is the most appropriate estimate and differences between the
formulae.

SOLUTION:

2nd highest flood: rank is 2

N r/N Weibull: r/(N+1) Gringorten:

(r-0.44) / (N+0.12)

10 0.20 0.18 0.15

50 0.04 0.04 0.03

100 0.02 0.02 0.02

Bottom ranked flood: rank is N

N r/N Weibull: r/(N+1) Gringorten:

(r-0.44) / (N+0.12)

10 1 0.91 0.94

50 1 0.98 0.99

100 1 0.99 0.99

Results show a general trend of converging towards the same result as N increases. As noted in
lectures the relative frequency method is not appropriate when considering the smallest flood on
record as it has a probability of 1 (i.e., it will always be equalled or exceeded in any given year)

Page 11 of 21
8. Consider the Standard Normal distribution for an Annual Series with N observations illustrated
below:

f(z )
0.4

0.3

0.2

0.1

0
-3 -2 -1 0 1 A 2 3

A
What is the AEP of a flow event with a z value corresponding to Point A if  f (z)dz = 0.92 ? What
−

will be the rank of this storm event to the nearest integer if N=35?

SOLUTION:
A
The AEP of a flow event with a z value corresponding to Point A if  f (z)dz = 0.92
−
is given by:

P(X≥xi) = 1 - 0.92 = 0.08.

To estimate what the rank of this storm event is (to the nearest integer) if N=35 we need to apply one
of the plotting position (ranking) methods. Using Gringorten we have:

AEP = 0.08 = (rank – 0.44)/(N+0.12).

Rearrange to find rank (roughly): rank = (0.08*35.12)+0.44 = 3 (rounded to nearest integer).

9. What is the probability that at least one flood of AEP = 0.02 (ARI = 50 years, T = 50 years) will
occur during the 30 year design life of a flood control project?

SOLUTION:
P = 1- (1-1/50)30 → P = 1 - 0.545 → P = 0.455 (= risk of 50y event during design)

Page 12 of 21
10. In an N year sequence, risk is defined as the probability of occurrence of a particular event in at
least one year (in an N year sequence). Reliability is equal to (1 - risk). Consider a small, temporary
(3 year design life) flood mitigation dam designed to contain the 20 year flood event, what is the
reliability of the designed dam (i.e., what is the probability it won’t be overtopped in the 3 years)?
Also estimate the risk that it will be overtopped at least once in the design life.

SOLUTION:
Recall the critical event probability: P(Q≥ at least 1 in N years) = 1-(1-AEP)N
Given ARI = 20 years, AEP ≈ 1/20 = 0.05 and N = 3 years:

P = 1-(1-0.05)3 = 0.1427

Reliability = 1 - risk = 1 - P = 1 – 0.14 = 0.86

Risk = 1 - reliability = 1 - 0.86 = 0.14

11. Consider the annual series given in the table below for a small stream (the Jason St gauge). Use
the data to estimate the AEP for a flood of 60 m3 s-1 using a plotting position probability estimate.
Also fit a Gumbel distribution to the data as well as a Log-Normal distribution to derive an AEP
estimate for a flood flow of 60 m3 s-1.

Table. Jason St gauge annual maximum series.

Year Flow Year Flow

(m3/sec) (m3/sec)

1972 33.4 1990 27.9

1974 133.4 1991 15.3

1975 6.7 1992 45.8

1976 26.6 1993 1.5

1976 11.5 1994 58.6

1978 13.7 1995 44.8

1979 20.1 1996 32

1980 24.1 1997 13.4

1981 20.9 1998 13.8

Page 13 of 21
 1981 34.8 1999 23.1

1983 21.5

1984 34.4

1985 23.2

1985 8.3

1986 18.3

1988 44.6

1989 66.4

SOLUTION:
An event of 60 m3 s-1 is approximately the same as the third ranked event (60 m3 s-1). Applying the
simple plotting position method (P = r/N) gives AEP ≈ 0.11, ARI = 1/AEP → ARI ≈ 1/0.11 ≈ 9 years.
These results are also similar to those derived from the Weibull formula (AEP ≈ 0.11, ARI ≈ 9, see
table below). Recall that the Gumbel distribution is approximated by Gringorten’s formula [P = (r-
0.44) / (N+0.12)] which gives AEP ≈ 0.094, ARI ≈ 11 years.

To fit the Log-Normal distribution take the logarithms (base 10) of the annual maximum flows and
calculate the mean (1.35) and standard deviation (0.37) for these values as well as the standard normal
variate for the design flow of 60 m3 s-1 [z = (x-μ)/σ → z = (1.77 - 1.35)/ 0.37 → z = 1.15]. Then use
the look-up tables (or MS Excel function) to find F(z). In this case F(z) ≈ 0.875. The AEP is given by
AEP = 1 – F(z) → AEP = 1 - 0.875 → AEP = 0.125 and therefore ARI ≈ 8 years.

Table. Jason St gauge annual maximum series with probability estimates

Log10
Record Year Flow [m3 s-1] Rank r/N Weibull Gringorten Flows
1 1972 33.4 9 0.333 0.321 0.316 1.524
2 1974 133.4 1 0.037 0.036 0.021 2.125
3 1975 6.7 26 0.963 0.929 0.942 0.826
4 1976 26.6 12 0.444 0.429 0.426 1.425
5 1976 11.5 24 0.889 0.857 0.869 1.061
6 1978 13.7 22 0.815 0.786 0.795 1.137

Page 14 of 21
7 1979 20.1 18 0.667 0.643 0.647 1.303
8 1980 24.1 13 0.481 0.464 0.463 1.382
9 1981 20.9 17 0.630 0.607 0.611 1.320
10 1981 34.8 7 0.259 0.250 0.242 1.542
11 1983 21.5 16 0.593 0.571 0.574 1.332
12 1984 34.4 8 0.296 0.286 0.279 1.537
13 1985 23.2 14 0.519 0.500 0.500 1.365
14 1985 8.3 25 0.926 0.893 0.906 0.919
15 1986 18.3 19 0.704 0.679 0.684 1.262
16 1988 44.6 6 0.222 0.214 0.205 1.649
17 1989 66.4 2 0.074 0.071 0.058 1.822
18 1990 27.9 11 0.407 0.393 0.389 1.446
19 1991 15.3 20 0.741 0.714 0.721 1.185
20 1992 45.8 4 0.148 0.143 0.131 1.661
21 1993 1.5 27 1.000 0.964 0.979 0.176
22 1994 58.6 3 0.111 0.107 0.094 1.768
23 1995 44.8 5 0.185 0.179 0.168 1.651
24 1996 32 10 0.370 0.357 0.353 1.505
25 1997 13.4 23 0.852 0.821 0.832 1.127
26 1998 13.8 21 0.778 0.750 0.758 1.140
27 1999 23.1 15 0.556 0.536 0.537 1.364

Page 15 of 21
12. Given the annual flood series statistics in the table below, calculate and tabulate flows for the 1
in 2, 5, 10, 25, 50 and 100 years floods assuming Extreme Value Distribution (EVD) and Log-Pearson
Type III distribution (LPT3), respectively.

Mean of floods (m3/s): 1430

Standard deviation of floods (m3/s): 1660
Mean of logarithms of floods: 2.818
Standard deviation of logarithms of floods: 0.641
Skewness coefficient of logarithms of -0.458
floods:

SOLUTION:

Peak Discharge (m3/s)

Return Period, T (years) EVD LPT3

2 1157 736
5 2624 2323
10 3596 3994
25 4823 6817
50 5733 9410
100 6637 12391

Page 16 of 21
13. If a structure can withstand a flood with a peak discharge no greater than 1,950 m 3 s-1 and has
been designed to have an economic life span of 100 years. Determine the risk of failure assuming the
Extreme Value Type 1 Distribution when the mean and standard deviation of the annual flood series
are 410 m3 s-1 and 280 m3 s-1, respectively. The Extreme Value Type 1 (EV1) distribution (also known
as the Gumbel Distribution) is described by:

   X − a  
F ( X ) = exp  − exp  −   
   b  

where F(X) is the probability of an annual maximum Q ≤ X and a and b are fitting parameters defined
by:

Q 6
a = Q − 0.5772b b=


where μQ and σQ are the population mean and standard deviation respectively.

SOLUTION:
If we assume the mean and standard deviation of the annual flood series can be used to estimate the
population mean and standard deviation we have:

 Q 6 280  6
b= =  218.32 a = Q − 0.5772b = 410 − 0.5772b  284
 

   1950 − 284   
F (1950 ) = exp  − exp  −      0.999515
   218.32   

Therefore the probability that flow Q > 1950 in any given year can be estimated as:

P ( X  1950 ) = 1 − F (1950 )  0.000485

We also need to take into account the design life of the structure (i.e., 100 years) and estimate that
probability that Q > 1950 in this time period. This can be estimated as:

Page 17 of 21
PN ( X ) = 1 − (1 − P ( X ) )
N

where N is the number of years. Therefore in this case we have:

PN (1950 ) = 1 − (1 − P (1950 ) ) = 1 − (1 − 0.000485)

100 100
 0.047

Therefore the risk = 0.047 (or a 4.7% risk of failure during the 100 year economic life)

14. The table below shows the classified monthly precipitation data measured at a station. The data
are divided into nine categories. The exponential distribution is often used to approximate rainfall
data such that P(X ≤ xi) = F(xi) = ∫f(x)dx = 1 - exp(-λx)). If λ = 0.105, use the chi-square test with 5%
level of significance to evaluate whether the exponential distribution provides an adequate fit to the
observed data [Example sourced from Karamouz et al., 2013].

Precipitation
depth class
Interval [mm] Actual frequency fa
1 0≤x<5 38
2 5 ≤ x < 10 26
3 10 ≤ x < 15 12
4 15 ≤ x < 20 10
5 20 ≤ x < 25 8
6 25 ≤ x < 30 3
7 30 ≤ x < 35 2
8 35 ≤ x < 40 1
9 40 ≤ x 0
Sum 100

SOLUTION:
Use the fitted exponential distribution to estimate the frequencies for the upper bound of each of the
classes used for the observed data. That is for the first class:

P(X ≤ 5 mm) = 1 - exp(-0.105 × 5) = 0.41 → Therefore the expected frequency is 41.

Page 18 of 21
The chi-squared test statistic is then calculated χ2 = ∑((fa – fe)2 / fe)] by comparing the actual and
expected frequencies (see table below). Also note from the above table that for intervals/classes 6 to
9 the observed (actual) frequency is less than 5. Therefore all of these classes are grouped together
(one of the assumptions of the chi-squared test is that is valid only when the expected frequency in
any category ≥ 5). The degree of freedom is given by NC – NP – 1 → 6 – 1 –1 = 4 (NC = number of
intervals/classes/categories and NP = number of parameters used to fit the distribution → in this case
only 1, i.e., λ).

Considering a 5% level of significance and 4 degrees of freedom the relevant critical value can be
read from a standard chi-square table (χ20.05(4) = 9.488, read from upper tail area). Because the
estimated value of the test statistic is smaller than the critical value (i.e., 2.91 < 9.488) the null
hypothesis that the monthly precipitation is adequately fitted by the exponential distribution is
accepted.

Exponential
Cumulative
Precipitation Actual Probability Expected
depth class frequency at Upper frequencies
Interval [mm] fa Limit fe fa - fe (fa - fe)2/fe
1 0≤x<5 38 0.41 41 -3 0.22
2 5 ≤ x < 10 26 0.65 24 2 0.17
3 10 ≤ x < 15 12 0.79 14 -2 0.29
4 15 ≤ x < 20 10 0.88 9 1 0.11
5 20 ≤ x < 25 8 0.93 5 3 1.80
6 25 ≤ x < 30 3 0.96 3 -1 0.33
7 30 ≤ x < 35 2 0.97 1
8 35 ≤ x < 40 1 0.99 2
9 40 ≤ x 0 1.00 1
Sum 100 100 2.916338

Page 19 of 21
15. Find the average occurrence of a 10-year flood (return period of the flood is 10 years) in a 100-
year period? What is the probability that exactly this number of 10-year flood will occur in a 100-
year period?

SOLUTION:
Example sourced from Maity (2018). The probability of occurrence of 10-year flood in any year is
given by P = 1/10 = 0.1. Thus, the average number of occurrences in 100 years = E(X) = nP = 100
× 0.1 = 10. The probability of 10 occurrences of 10-year flood in 100 years can be evaluated using

binomial distribution:

Px ( x; n, P ) = nCx P x (1 − P )
n− x

Px (10;100, 0.1) = 100C10 ( 0.1) (1 − 0.1)

10 100 −10

Px (10;100, 0.1) = 100C10 ( 0.1) (1 − 0.1)

10 100 −10

Px (10;100, 0.1)  (1.7311013 )  (110−10 )  ( 7.618 10 −5 )

Px (10;100, 0.1)  0.132

Note that some calculators don’t have the capacity to calculate 100C10. In these cases, alternative
software such as spreadsheets or programming languages can be used to apply the combination
operator. For example, in MS Excel the combination operator is a built in function:

100
C10.= COMBIN(n,x) – COMBIN(100,10) ≈ 17310309456440

The Python “math” library (among others) also has a built-in function for the combination operator
(sometimes referred to as the binomial operator or binomial coefficient):

In [1]: from math import comb

In [2]: comb(100,10)
Out[3]: 17310309456440

Page 20 of 21
References
Brutsaert, Wilfried (2005). Hydrology - An Introduction. Cambridge University Press.
Karamouz M., Nazif S. and Falahi M. (2013) Hydrology and Hydroclimatology Principles and
Applications. CRC Press, Taylor and Fancis Group. 716 pp.
Maity R. 2018. Statistical Methods in Hydrology and Hydroclimatology, Springer Transactions in
Civil and Environmental Engineering, doi: https://doi.org/10.1007/978-981-10-8779-0_2
Mays (2012) Ground and surface water hydrology. Wiley. 617 pp.

Page 21 of 21