A MASTERS THESIS PROJECT

REPORT
On

Estimation of Probable maximum precipitation and Probable

maximum flood of Gandhi Sagar catchment
At

INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR

Supervised by

Prashanth Reddy Hanmaiahgari

Associate Professor
Indian Institute of Technology Kharagpur

Submitted by

Abhishek Putatunda – 22CE61R01

M. Tech 2nd year, Hydraulic and Water Resources Engineering
Department of Civil Engineering

INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR

Abstract

As climate change is on the rise, and with an increase in climate hazards, improvement of
detection and monitoring of flood events has become a priority. Floods are increasing in
frequency and magnitude, which is continuing to rise due to climate change. Our objective of
the research is: prepare the PMP and PMF for the generation of flood inundation map under
specific conditions. We have considered the Gandhi Sagar reservoir as our study area. This
report includes PMP calculations using gridded rainfall data and available point data from
rain gauge stations in the GS catchment area. The Exponential (EXP), Gamma (GAM),
Generalised Extreme Value (GEV), Generalised Logistic (GLO), Generalised Normal
(GNO), Generalised Pareto (GPA), Gumbel’s Extreme Value (Type 1 distribution) (GUM),
Hershfield method, Kappa (KAP), Log-Pearson (Type 3 distribution), Log-Normal (Type 2
distribution), Log-Normal (Type 3 distribution), Normal (NOR), Wakeby (WAK), Uniform
ditribution and Weibull (WEI) methods are used to generate PMP (probable maximum
precipitation)for different return periods for the GS and RPS basins.The L-moment method is
used to calculate probability distribution parameters such as mean, standard deviation and
frequency factor used in the above-mentioned methods. To estimate the PMP values, (i)
gridded rainfall (daily) data over the upper Chambal catchment from 1901- 2021 has been
collected from IMD, (ii) The daily rainfall data measured during 1968- 2006 at rain gauge
stations in the upper Chambal catchment has also been collected from govt agencies of
Madhya Pradesh and Rajasthan, (iii) The daily rainfall data during measured during 1990-
2017 by Gandhisagar Dam Authority (15 rain gauge stations) and RPS Dam Authority (one
rain gauge station) has been obtained. (iv) daily rainfall data at the RAPS plant during 1964-
2022 has been collected from NPCIL, and (v) In addition, the PMP atlas of the Ganga basin
prepared by the CWC and IMD has been used to arriveat the probable maximum rainfall in
thecatchment areas of the RPS and GS dams. The atlas provides a source of readily available
design storm rainfalls and PMP values for different durations for any area in the six major
river basins (Ganga, Brahmaputra, Cauvery, Godavari, Mahanadi, and Narmada and Tapi) to
design engineers. The upper Chambal basin area upto Rawatbhata is referred to catchment
no. 404 in the PMP Atlas for which the Rainfall data of 26 stations in the upper Chambal
catchment (catchment 404) have been used. Then we compared the calculated PMP values to
those available in PMP Atlas. To generate the PMF we deployed SRTM DEM of our study
area. We developed a hydrologic model in HEC HMS to generate the PMF in our study area.
Then storm hydrograph is generated for different return periods for Gandhi Sagar catchment.

2
1. Introduction

Flood is an outcome of heavy rainfall and the accumulation of excessive water in every living
area1. Floods may occur due to the overflow of water from the reservoirs or due to torrential
downpours of rain in places where the drainage systems are not adequately maintained.
Influenced by climate change, the frequency and intensity of meteorological disasters mainly
floods have increased significantly in recent years. Floods can take place naturally due to
heavy rainfall or cloud bursts in hilly regions. In man-made causes, it may be due to improper
operation of dams or due to sudden breaches of dams. “1979 Machchhu dam failure on
Machchhu River”2 is an example of such type of failure in which the earthen embankment of
this dam failed and the result was a severe flood in Morbi town (Gujrat). Failure of concrete
dams is rare to the rarest event. However, to minimize the risk of flood in the downstream
area, dam break failure studies are required for concrete dams also. In the present study, a
nuclear power plant is associated, hence an in-depth study is required with possible rainfall of
even 1000 to 10000 years return period.

In this study,the Exponential (EXP), Fisher – Tippett (Frechet Type 2 distribution), Gamma
(GAM), Generalised Extreme Value (GEV), Generalised Logistic (GLO), Generalised
Normal (GNO), Generalised Pareto (GPA), Gumbel’s Extreme Value (Type 1 distribution)
(GUM), Hershfield method, Kappa (KAP), Log-Pearson (Type 3 distribution), Log-Normal
(Type 2 distribution), Log-Normal (Type 3 distribution), Normal (NOR), Wakeby (WAK)
and Weibull (WEI) methodsare used to generate probable maximum precipitation in the study
region. Here,the study region refers to the catchment areas of the RPS and GS dams. The L-
moment method is being used to check the homogeneity of rainfall data as the main
advantage of this method, in comparison to conventional moments, is the existence of all
orders under only a finite mean assumption. L-moments are measures of the shape of a
probability distribution. It may be used for summarizing data drawn from both univariate and
multivariate probability distributions.

To estimate the PMP values, (i) gridded rainfall (daily) data over the upper Chambal
catchment from 1901- 2021 from IMD, (ii) The daily rainfall data measured during 1968-
2006at rain gauge stations in the upper Chambal catchment from govt agencies of Madhya
Pradesh and Rajasthan, (iii), The daily rainfall data during measured during 1990-2017 by
GandhisagarDam Authority (15 rain gauge stations) and RPS Dam Authority (one rain gauge

1
www.who.int/health-topics
2
https://en.wikipedia.org/wiki/1979_Machchhu_dam_failure

3
station), (iv) daily rainfall data at the RAPS plant during 1964-2022 from NPCIL, and (v) the
PMP atlases of six river basins prepared by the CWC and IMDhave been used to calculate the
probable maximum rainfall in the study area. The PMP atlas provides a source of readily
available design storm rainfalls and PMP values for different durations for any area of the
river basin to design engineers. The whole Ganga basin area covered in the PMP atlas is
around 10,00,000 km2, which is subdivided into 21 catchments, numbered 323 to 325 and
401 to 418 by IMD (ganga-final-report-volume-i.pdf).The rainfall data of 26 stations have
been used in PMP atlasfor the upper Chambal catchment (catchment 404) in this study.

1.1 Study area

The location of the Chambal Basin is presented in Figs 1 and 2. It covers Madhya Pradesh,
Rajasthan and Uttar Pradesh (For more details Refer to Report 1). The Chambal Basin is
further divided into 4 parts viz. i) Upper Chambal ii)Up to confluence with Kalisindh and
Parwati iii)Banas iv) Lower Chambal. The Upper Chambal basin is located up to the Rana
Pratap Sagar (RPS) Damwhich is the second dam ina series of 4 Dams on the Chambal River.
Gandhi Sagar Dam is the firstand major storage reservoir in the series and is located 52 km
upstream of RPS Dam (Fig. 2).

Fig. 1: Location of Chambal Basin in the Fig. 2: Location of Upper Chambal Basin
Country

4
1.2 Data used

1.2.1 Digital Elevation Model (DEM)

SRTM DEM with a spatial resolution of 30 m is used for catchment and river network
delineation of the study region.

1.2.2 Rainfall Data

IMD has gridded rainfall data of 1-day precipitation events with a resolution of (0.250 x
0.250)3. The collected dataset from IMD is converted into a usable dataset and then gridded
rainfall data (Fig. 3 and Fig. 4) of our study region is extracted from the whole dataset so that
we can estimate the maximum precipitation depth of our study region. Other than IMD,
rainfall data from the Gandhi Sagar Dam Authority (Fig. 5), the Rajasthan Government, and
MP Government are also used in the present study.

(a) (b)
Fig. 3: a) Grid generation for upper Chambal Basin b) Rain gauges maintained by Gandhi
Sagar Dam Authority

3
https://mausam.imd.gov.in/

5
1.2.3 Survey of India Toposheets

Topographic maps with a scale of 1:50000 downloaded4 from georeferenced are used in the
present study. These topographic sheets will be sed to generate DEMs, which will further be
used to calculate PMF of our catchment.

2. Methodology

Finding probable maximum precipitation for 1-Day for various return periods in the study
area is the main objective of the present report. These results will be further used to find
Probable maximum flood both at GS and RPS dams.

The process flow chart of the study is as follows:

Step 1
• Collection of Topographic maps

Step 2
• Locating the raingauge stations in our study area

Step 3
• Calculation of Probable maximum precipitation

Step 4
• Georeferencing and digitization of topographic maps

Step 5
• Development of SRTM digital elevation model

Step 6
• Development of storm hydrograph using SNYDER's unit hydrograph

Step 7
• Catchment delineation using HEC HMS

Step 8
• Development of hydrologic model in HEC HMS

Step 9
• Importing of hourly rainfall distribution data

Step 10
• Development of storm hydrograph using HEC HMS model

2.1 Rain gauge stations in the study area

The following agencies collect the rainfall data in the Upper Chambal basin.

I. Gandhi Sagar Dam Authority or Water Resources Department of MP

II. Central Water Commission
III. District Administration (Rajasthan & Madhya Pradesh)
IV. Rana Pratap Sagar Dam Authority (They maintain only one rain gauge at the Dam)

Most of the locations are the same, yet there is some variation in the location and number of
rain gauges of various agencies. Gandhi Sagar Dam Authority mostlyusedits data for the gate
operation. We have worked on two separate catchments of Rana Pratap Sagar and Gandhi
4
https://onlinemaps.surveyofindia.gov.in/FreeMapSpecification.aspx

6
Sagar dam. The areas of these two catchments are calculatedusing QGIS. We delineated the
whole catchment to divide it into several sub-catchments and estimated the area of these two
catchments separately.

(a) (b)

Fig 4: (a) Gandhi Sagar Basin and (b) Locating of rain gauge stations on GS catchment using
Google Earth Pro 7.3

7
Table 1 Locating of Rain gauge stations in our study area using Google Earth Pro

Sl. No. Gandhi Sagar Catchment Rana Pratap Sagar

catchment

1 Bhanpura Rawatbhatta

2 Ujjain

3 Indore

4 Pat

5 Gandhisagar

6 Dhar

7 Ratlam

8 Choumahla

9 Rampura

10 Dewas

11 Nagda

12 Badnagar

13 Neemuch

14 Mandasaur

15 Mahidpur

2.2 Collection of Gridded Rainfall Data

Rainfall gridded data is then downloaded from IMD5. These rainfall data are then converted
into MS Excel files so that we can further use them for our calculationpurposes.The gridded
data of the 1-day rainfall series is then extracted for the Upper Chambal basin and it can also
be imported to QGIS 3.30.3 to visualise the rainfall distribution over a particular area. The
annual maximum daily rainfall is calculated for the preparation of the PMP of our study area.

5
https://mausam.imd.gov.in/

8
3 Calculation of probable maximum precipitation

PMP is being estimated using several methods such as the Hershfield method, Gumbel’s
extreme value (Type 1 distribution) method, Fisher – Tippett (FrechetType 2 distribution)
method, Log-Pearson (Type 3 distribution) method, Log-Normal (Type 2 distribution), Log-
Normal (Type 3 distribution), and Weibull method. We generated the PMP for the 1-day
annual maximum rainfall series for both gridded rainfall data and available rain gauge station
data.

The PMP atlas prepared jointly by CWC, IMD and RMSI Pvt Ltd are also used to find the
PMP values of 1-day, 2-day and 3-day PMP for various return periods in Gandhi Sagar dams
catchment areas.

3.1 Hershfield method

Hershfield’s statistical procedure can be used to estimate the maximum probable rainfall for a
given catchment. Based on the general frequency formula, 𝒙𝑻 = 𝑥̄𝑁 + 𝐾𝑆𝑁 (Chow et al.
1988), PMP can be calculated. Hershfield (1965) calculated the frequency factor as the
maximum observed value at a station, minus the mean of the annual series, divided by the
standard deviation,

𝑥𝑚𝑎𝑥 − 𝑥̄𝑁 −1
𝐾𝑚 =
𝑆𝑁−1

where 𝒙𝑻 is the rainfall for the return period 𝑇, 𝐾 is a variable frequency factor dependent on
the probability distribution that fits the extreme value series, 𝑥𝑚𝑎𝑥 is the highest recorded
value from the series of N annual maximum rainfalls, x̄ 𝑁 is the mean and 𝑆𝑁 is the standard
deviation of the rainfall series, 𝐾𝑚 is the frequency factor which is the largest of all the
calculated 𝐾𝑚 values for all stations in a given area, x̄ 𝑁−1 and 𝑆𝑁−1 are the mean and standard
deviation of the annual maximum rainfall series, excluding the maximum observed value
respectively. Next, 𝐾𝑚 values are to be plotted against their corresponding mean of annual
maximum rainfall series. 𝐾𝑚 occurs randomly even though the means and standard
deviations at nearby stations might be very nearly equal. The largest 𝐾𝑚 value is to be
computed based on the observed rainfall series. Enveloping 𝐾𝑚 as a function of the mean
serves a transposition purpose. By the combination of the data, each station is permitted to
reach an extreme possibility of 𝐾𝑚 . In other words, each station with a corresponding mean is

9
assumed to have the same value of 𝐾𝑚 . This assumption does not imply that the standard
deviations are equal (Method for Estimating Probable Maximum Rainfall; David M.
Hershfield, 1965).

Finally, the PMP is calculated as,

3.2 Gumbel’s method

Extreme value Type I (EV-I) distribution, also known as Gumbel distribution, is a limiting
probability distribution which is used to model the maximum or minimum values from a
sample of independent, identically distributed random variables, as the size of the sample
increases. Gumbel (1941) introduced Gumbel's distribution. It is one of the most widely used
probability distribution functions for extreme values in hydrologic and meteorological studies
for the prediction of flood peaks, and maximum rainfalls.

Description: The EV-I distribution for maximum/minimum values is the limiting model as n
approaches infinity for the distribution of the maximum/minimum of 𝑛 independent values
from an initial distribution whose right/left tail is unbounded, that is, the initial cumulative
distribution approaches unity (zero) with increasing/decreasing values of the random variable
at least as fast as the exponential distribution approaches infinity.

The normal, lognormal, exponential, and gamma distributions all meet the requirement for
the maximum values, whereas only the normal distribution satisfies the conditions for
minimum values. The probability density function of the EV-I distribution is as follows:

1
𝑓𝑥 (𝑥) = 𝑒𝑥𝑝 [∓(𝑥 − 𝛽)/𝛼 − 𝑒𝑥𝑝 (∓(𝑥 − 𝛽)/𝛼)]
𝛼

where, −∞<𝑥<∞; −∞<𝛽<∞; α > 0. The (–)ve sign implies maximum value, and the (+)ve sign
implies minimum value. The CDF of the EV-I is as follows:

𝐹𝑥 (𝑥) = 𝑒𝑥𝑝 [∓𝑒𝑥𝑝 (∓(𝑥 − 𝛽)/𝛼)]

where −∞< 𝑥<∞; −∞<𝛽<∞; α > 0. The –ve sign implies the maximum value, and the (+)ve
sign implies the minimum value. The parameters α and β are scale and location parameters
with 𝛽 being the mode of the distribution.

The mean, variance, and skewness coefficient are as follows:

𝐸 (𝑋) = 𝛽 ∓ 0.5772𝛼
10
 𝑉𝑎𝑟 (𝑋) = 1.645𝛼 2
𝛾 = ∓ 1.1396
For a hydrologic extreme event X, following Gumbel’s distribution, the exceedance
probability of 𝑋 = 𝑥0 is given by,
(𝑥−𝛽)
−
𝑃 (𝑋 ≥ 𝑥0 ) = 1 − 𝑒 −𝑒 𝛼

Let us simplify the equation by introducing a dimensionless variable 𝑦 known as Gumbel’s

reduced variate, given by, 𝑦 = (𝑥 − 𝛽)/𝛼, where 𝛼 and 𝛽 are the scale and location
parameters of Gumbel’s distribution, respectively.

−𝑦 1
⇒ 𝑃 (𝑋 ≥ 𝑥0 ) = 1 − 𝑒 −𝑒 =𝑇

𝑇
⇒𝑦 = −𝑙𝑛 (𝑙𝑛 (𝑇−1))
For Gumbel’s distribution, standard deviation and mean are given by,
𝑆𝑥 = 1.2825𝛼

𝑥 = 𝛽 + 0.5772𝛼
Now for a particular return period 𝑇, let us designate 𝑦 as 𝑦𝑇 and 𝑥 as 𝑥 𝑇 , then
(𝑥𝑇 − 𝑥)
𝑦𝑇 = [1.2825 ] + 0.5772
𝑆𝑥

(𝑦𝑇 − 0.5772)
or, 𝑥 𝑇 = 𝑥 + [ ] 𝑆𝑥
1.2825

or, 𝑥 𝑇 = 𝑥 + 𝐾 𝑆𝑥

(𝑦𝑇 − 0.5772)
where the frequency factor, 𝐾 = [ ]
1.2825

Hence, the above equation is modified to take care of the finite sample size N as shown
below

𝑥𝑇 = 𝑥 + 𝐾 𝑆𝑥
𝐾 = (𝑦𝑇 − 𝑦𝑛 )/𝑆𝑛
where𝑦𝑇 = −[𝑙𝑛 (𝑙𝑛 (𝑇/ 𝑇 − 1))] is reduced variate for return period 𝑇 and 𝑦𝑛 is reduced
mean, a function of 𝑇 and sample size 𝑁; as 𝑁 →∞,𝑦𝑛 → 0.5772. 𝑆𝑛 is the reduced standard
deviation, a function of 𝑇 and sample size 𝑁; as 𝑁 → ∞, 𝑆𝑛 → 1.2825.

11
3.3 Log-Pearson distribution

This distribution is extensively used for the estimation of rainfall distribution, and flood
discharge calculations. In this method, the variate is first transformed into logarithmic form
and the transformed variate is then analysed. If 𝑋 is the variate of a random hydrological
series, then the series of 𝑍 variates,

𝑧 = 𝑙𝑜𝑔𝑥

𝑧𝑇 = 𝑧̄ + 𝐾𝑧 𝜎𝑧

𝐾𝑧 is a frequency factor, which is a function of the return period 𝑇 and skewness

coefficient𝐶𝑠 .

𝜎𝑧 is the standard deviation of the 𝑍 variate series, is given as,

𝛴(𝑧−𝑧̄ )2 𝑁Σ(𝑧−z̄ )3
𝜎𝑧 = √ and the coefficient of skew is determined as 𝐶𝑠 =(𝑁−1)(𝑁−2)σ 3
(𝑁−1) 𝑧

𝑁 is the sample size and z̄ is the mean of 𝑍 values

After finding 𝑧𝑇 the corresponding 𝑥𝑇 value is calculated by,

𝑥 𝑇 = 𝑎𝑛𝑡𝑖𝑙𝑜𝑔 (𝑧𝑇 )

The value of the frequency factor corresponding to the Skewness coefficient is obtained from
the Log-Pearson Type 3 distribution table.

3.4 Log-Normal distribution

The Log-Pearson distribution is reduced to the Log-normal distribution if the value of the
skewness coefficient is taken as zero. Hence for Log-normal distribution, Cs = 0.

3.5 Pearson 3 distribution

The Pearson III distribution belongs to the Gamma distribution family, being a generalized
form of the two-parameter Gamma distribution, with shifted x, and a particular case of the
four-parameter gamma distribution.

(x − γ) α−1 x−γ 1 x−γ

f(x) = · exp (− ) = β ·dgamma ( , α)
β α ·Γ(α) β β

12
 x−γ
𝑋 1 𝑋 x−γ x−γ Γ ( α, )
β
F(x) = 1 − ∫𝑌 f(x)dx = 1 − β·Γ(α) (∫𝑌 ( β )α−1 . exp (− ) dx =
β Γ(α)

x−γ x−γ
F(x) = 1 – pgamma ( , α) = 1 – pchisq (2. , 2.α)
β β

x(p) = F −1 (x) = γ + β·Γ −1 (1 − p; α) = γ + β·qgamma (1 − p, α)

where α, β, γ are the shape, the scale and the position parameters and x can take any values of
range γ < x < ∞ if β > 0 or −∞ < x < γ if β < 0 and α > 0; µ, σ represent the mean (expected
value) and standard deviation. If β < 0 (negative skewness) then the first argument of the
−1 −1
inverse of the distribution function Gamma, Γ (1 − p; α) becomes Γ (p; α). The
expressions of the inverse function x(p) = F −1 (x), using the mean and standard deviation, are
valid only for the method of ordinary moments.

3.6 Kappa distribution

The distribution represents a particular case of the four-parameter Kappa distribution and a
generalized form of the two-parameter Kappa distribution by adding a location parameter
(shifted x), being presented for the first time in 2009 by Park et al.

α x − γ α − α+1
f(x) = β (α + ( ) ) α
β

x−γ 𝑥−𝛾 α 1
F(x) = 1 − . (( ) + α) − α
β β

1
α·(1 − p) α
x(p) = γ + β· (1 − (1 − p) α)α

where, α, β, γ are the shape, the scale and the position parameters.

3.7 Generalized Exponential Distribution

The Generalized Exponential Distribution is an alternative to the two-parameter Gamma and

Weibull distributions. It was introduced by Gupta and Kundu in 1999.

f(x) = α·β· exp(−β·x) · (1 − exp(−β·x)) α−1

F(x) = 1 − (1 − exp(−β·x)) α

13
 1
− 𝑙𝑛 (1 − (1 − 𝑝)𝛼 )
x(p) = β

where α, β > 0 are the shape and the scale parameters; x > 0.

3.8 Normal distribution

Normal distribution, also known as Gaussian distribution or bell curve, is a continuous

probability distribution.

A continuous variable (X) that can take any value in the real line with a symmetrical (with
respect to its mean) bell-shaped distribution of probability. Normal distribution is the most
frequently used continuous probability distribution function. When the mean is zero and the
variance is 1, the distribution is called a standard normal distribution.

The pdf of the normal distribution is given by,

−(x−μ)2
1
f X (x; μ, σ2) = 𝑒 2σ2 dx, − ∞ < x < ∞
√ 2πσ2

The CDF of the Normal Distribution is given by,

−(x−μ) 2
𝑥 1
FX (x; μ, σ ) =
2
∫−∞ 2πσ2 𝑒 2σ2 dx, − ∞ < x < ∞
√

where, μ and σ2, where μ is the mean and σ2 is the variance.

3.9 Uniform distribution

Uniform distribution is the simplest and symmetric continuous probability distribution

function. It is defined over a range (known as support) such that its occurrence is equally
possible (equiprobable) over any subinterval of the same length within the support. Thus, the
interpretation could be as follows, 𝑋 is equiprobable over any subinterval of the same length
within its support, α and β, where α and β are the minimum and maximum limit of the
support respectively.

The probability density function for the uniform distribution is as follows,

14
 1
fX (x) = β − α, α ≤ x ≤ β

The cumulative density function for the continuous uniform distribution is as follows,

x−α
FX (x) = β − α, α ≤ x ≤ β

The mean, variance, and coefficient of skewness of the uniform distribution are,

E (X) = (β + α)/ 2

Var (X) = (β − α) 2/ 12

γ=0

3.10 Wakeby distribution

The inverse form of the five-parameter Wakeby (WAK) distribution is expressed as,

α γ
x(F) = ξ + β {1 – (1 − 𝐹)β } - δ {1-(1-F) - δ}

where, ξ, α, β, γ, and δ are the parameters of the Wakeby distribution

3.11 L-moments

L-moments and L-moment ratios are quantities useful in the summarization and estimation of
probability distributions (Hosking, 1990) describes the theory and applications of L-moments
Here we give an expanded discussion of some of the theory. Moments, such as mean,
variance, skewness and kurtosis, are traditionally used to describe features of a univariate
distribution. Hosking (1990) introduced an alternative approach using L-moments, which are
defined as certain linear combinations of order statistics. The main L-moments’ advantage, in
comparison to conventional moments, is their existence of all orders under only a finite mean
assumption. L-moments, being measures of the shape of a probability distribution, may be
used for summarizing data drawn from both univariate and multivariate probability
distributions.The L-moments-based estimates are obtained in a similar way as in the moment
method, which means the population L-moments are equated to their corresponding sample
quantities. Hosking (1990) gives parameter estimators of some common univariate
distributions and highlights L-moments because they sometimes provide better estimates than

15
the maximum likelihood method (particularly for small samples and heavy-tailed
distributions). L-moments are also employed in hypothesis testing, particularly in RFA which
yields reliable estimates of high quantiles of extreme events using data from sites, which have
similar probability distributions.

ASPECTS:
The following are the aspects:

1. PWMs and L-moments

2. Data screening and missing value corrections
3. Test of regional homogeneity
4. Frequency distributions used
5. The goodness of fit measures and
6. Development of relationship between mean annual peak flood and catchment area

Probability weight moments:

Population univariate L-moments: A population L-moment is defined to be a certain linear
combination of order statistics (the letter L just emphasizes that the L-moment is a linear
combination) which exists for any random variable with finite mean. Hosking (1990) defined
the population L-moment of the rth order as a linear combination of the expectations of the
order statistics 𝑋1 : 𝑛 ≤ 𝑋2 : 𝑛 ≤ … ≤ 𝑋𝑛 : 𝑛 of a random sample of size n drawn from a
univariate distribution of a random variable X with cumulative distribution function F,

Probability-weighted moments are defined by Greenwood et al. (1979) as,

1
𝑀𝑖, 𝑗, 𝑘 = ∫ 𝑥 (𝐹 )𝑖 (𝐹 )𝑗 (1 − 𝐹 )𝑘 𝑑𝐹
0

𝑥
where, 𝐹 = 𝐹(𝑥) = ∫−𝑥 𝑓 (𝑥 ) 𝑑𝑥 is the cumulative density function and 𝑥(𝐹 ) is the inverse
of it; i, j, and k are the real numbers.

The particularly useful special cases of the PWMs 𝛼k and 𝛽 jare,

1
𝛼𝑘 = 𝑀1,0,𝑘 = ∫ 𝑥(𝐹 ) (1 − 𝐹 )𝑘 𝑑𝐹
0
1
𝛽𝑗 = 𝑀1,𝑗,0 = ∫ 𝑥(𝐹 ) (𝐹 )𝑗 𝑑𝐹
0

16
These equations are in contrast with the definition of the ordinary conventional moments,
which may be written as,

E(Xr) = ∫{𝑥(𝐹 )}𝑟 𝑑𝐹

PWMs involve successively higher powers of non-exceedance probability (F) or exceedance

probability (1-F) and may be regarded as integrals of x(F) weighted by the polynomials Fr or
(1-F)r. As the quantile function x(F) is weighted by the probability F or (1-F), these are
named as probability weighted moments.

However, PWMs are difficult to interpret as measures of scale and shape of a probability
distribution. This information is carried out in certain linear combinations of the PWMs.
These linear combinations arise naturally from integrals of x(F) weighted not by polynomials
Fr or (1-F)r but by a set of orthogonal polynomials (Hosking and Wallis, 1997)

Hosking (1990) defined L-moments as a linear combination of probability-weighted

moments. In general, in terms of 𝛼k and 𝛽 j, L-moments are defined as,

λ r+1 = (-1)’ Σ p*r,k𝛼k = Σ p*r,k𝛽 k

Where,prr,kis an orthogonal polynomial (shifted Legendre polynomial) expressed as

(−1)^(r−k)(r+k)!
p*r,k= (-1)r-krCkr+k Ck = (𝑘!)^2 (𝑟−𝑘)!

L-moments in terms of probability as given below,

λ1 = 𝛼0 = 𝛽0

λ2 = 𝛼0 – 2𝛼1 = 2 𝛽1 - 𝛽0

λ3 =𝛼0- 6 𝛼1 + 6 𝛼2 = 6𝛽2− 𝛽1+ 𝛽0

λ4 = 𝛼0- 12 𝛼1 + 30 𝛼2 - 20 𝛼3 = 20𝛽3− 30𝛽2+ 12𝛽1+ 𝛽0

λ1 is a measure of location

λ2 is a measure of scale

where λr are L-moments, r=1,2,3…

λ1 is first L-moment

λ2 is second L-moment

17
λ3 is the third L-moment respectively

𝛽0 ,𝛽1 , 𝛽2 , and 𝛽3 areparameter estimates by Hosking and Wallis (1997)

The procedure based on PWMs and L-moments are equivalent. However, L-moments are
more convenient, as these are directly interpretable as measures of the scale and shape of
probability distributions. Subscript, 1, the mean, is a measure of location, and 2 is a measure
of scale or dispersion of random variable. It is often convenient to standardise the higher
moments so that they are independent of units of measurement.

τr = λr / λ1 for r = 3, 4

Analogous to conventional moment ratios, such as the coefficient of skewness τ3 is the L-

skewness and reflects the degree of symmetry of a sample. Similarly,τ4 is a measure of
peakedness and is referred to as L-kurtosis. These are defined as

L-coefficient of variation (L-CV), (τ) = λ2 / λ1

L-coefficient of skewness, L-skewness (τ3) = λ3 / λ2

L-coefficient of kurtosis, L-kurtosis (τ4) = λ4/ λ2

Symmetric distributions have τ3 = 0 and its values lie between -1 and +1.

The calculation of PMP for the1-day storm event using Hershfield method, Gumbel’s method,
Log-PearsonType 3 method, Weibull’s method, Log-Normal Type 2 method and Log-Normal
Type 3, Wakeby distribution, Kappa distribution, Normal distribution, Uniform distribution,
GEV distribution, Exponential distribution, Gamma distribution, GLO distribution, GNO
distribution, GPA distribution are given in the results and discussion.

3.12 PMP Atlas

The PMP Atlas is presented in two volumes namely ‘ganga-final-report-volume-i’6 and

‘ganga-final-report-volume-ii’7. The volume-i gives basin information, details of storms
analyzed, generalized PMP estimates and statistical PMP along with the procedures. The
volume-ii consists of maps and tables supporting the main body of the report. Importantly
volume-ii gives location maps of storm centers of 1-day, 2-day and 3-day durations with
PMP rainfall estimates with various return periods.

6
https://cwc.gov.in/sites/default/files/ganga-final-report-volume-i.pdf
7
https://cwc.gov.in/sites/default/files/ganga-final-report-volume-ii.pdf

18
Fig. 5: Location of Upper Chambal Basin (Catchment number 404) in Ganga Basin (Source:
PMP Atlas Volume ii)

Based onthe procedure given in Volume I of PMP Atlas for the Ganga basin, various Tables
and corresponding figures are presented in Volume II. Refer to Table 1-1 on page 30 (PMP
Atlas), rain gauge stations used for PMP calculation of the Upper Chambal basin are
presented in Figure 11 (PMP Atlas).

The PMP estimation procedure involves the formation of the Depth Area Duration curve and
their areal and temporal distribution over the basin. PMP Atlas used the physical method
approach for PMP evaluation. The main assumption in the physical method is that PMP will
result from a storm event in which there is an optimum combination of the available moisture
in the atmosphere and the efficiency of the rainstorm mechanism. Factors, which influence
the rainstorm event efficiency, include horizontal mass convergence, topography, induced
lifting, vertical motion and the rate of condensation (ganga-final-report-volume-i). The
procedure of estimating PMP involves the use of observed rainfalls for various major events
over the basin. The storm rain depths are enveloped by smoothing over a range of areas and
durations. This method is quite accurate for PMP calculation. The procedure starts with the
collection of daily rainfall data from various rain gauge stations in and surrounding
catchment (for our study that is catchment 404 of PMP Atlas) under consideration.

19
 Fig. 6: Rain gauges used in the PMP Atlas

In the case of PMP estimation, rainfall analysis using the DAD envelope approach and
moisture maximization is applied. These envelope rainfall depths are considered in-situ
Standard Project Storm (SPS) values. The SPS values are then multiplied by the MMF of the
storm contributing to the envelope curve for the respective area and duration to get the PMP
estimate. For example, the MMF for the rainstorm of July 2005 at Sagar was calculated as
1.12. In the case of grid point PMP estimation, a grid system is constructed at a 1-degree
resolution of latitudes and longitudes for non-orographic areas. In the case of orographic
areas, the grid system of 0.25-0.5 degrees is constructed (ganga-final-report-volume-i). The
rainfall maximization is done by using the DAD analysis to estimate the enveloping rainfall
depths at each grid point for the range of the areas and durations (1-3 days). Since a rainfall
transposition is assumed in our study, the SPS values are estimated by multiplying the grid

20
rainfall depths with the Transposition Adjustment Factors, TAF, (which considered the
impact of location adjustment and barrier adjustments).As an example, TAF for 16-17 Aug
1969 was estimated as 0.86 for catchment No. 404 corresponding to a 340mm grid rain depth
and an area of 20,000 sq. km. (ganga-final-report-volume-i).

The in-situ Standard Project Storm (SPS) are obtained from corresponding DAD envelope
curves of the catchment (Catchment-404 of PMP Atlas) for 1-day, 2-day and 3-day durations
separately. These SPS (in situ) values are multiplied with the MMF to obtain in situ PMP
values. The PMP for different durations over a basin is derived by maximizing the highest
rainfalls obtained from major historical rainstorms that have occurred over the region. This
maximization is done by simply multiplying the highest rainfall values by the moisture
maximization factor (MMF). The MMF for a rainstorm in a place is the ratio of precipitable
water corresponding to the maximum persisting dew point temperature on record at the
original location of the rainstorm corresponding to the maximum persisting dew point
temperature of the rainstorm. The objective of maximization is to determine the physical
upper limit of rainfall, which occurs if the moisture available to the storm is at maximum.
The most important factor in moisture maximization is the estimation of moisture or
precipitable water available in the atmosphere. The MMF is thus determined based on 12-
hour or 24-hour persisting storm dew point temperature and the maximum ever-recorded
persisting dew point temperature for the area under study (ganga-final-report-volume-i).

4 Calculation of probable maximum precipitation

4.1 Calculation of PMP of 1-Day storm event

Probable maximum precipitation (PMP) is estimated using IMD gridded rainfall data,
available rain gauge station data for the Gandhi Sagar catchments, using Hershfield’s method,
Gumbel’s method, Log-PearsonType 3 method, Weibull’s method, Log-Normal Type 2
method and Log-Normal Type 3, Wakeby distribution, Kappa distribution, Normal
distribution, Uniform distribution, GEV distribution, Exponential distribution, Gamma
distribution, GLO distribution, GNO distribution, GPA distribution for different return
periods such as 25 years, 100 years, 1000 years, and 10,000 years, the maximum value of
calculated PMP of the 1-day storm event and the corresponding maximum value of frequency
factor is tabulated below. The gridded data is available from the years 1901 to 2021 and

21
downloaded from IMD. In these methods, the mean, standard deviation and frequency factor
are determined using the L-moments method.

Km vs Mean Rainfall
8

Km
2

0
0 50 100 150

Mean Rainfall

(a) (b)

Fig 7 : a) Boxplot of 1-Day annual maximum rainfall b) Plot of Km against the mean of annual
maximum rainfall series for Gandhi Sagar stations

Fig 8: Plot of precipitation depth against the non-exceedance probability of annual maximum rainfall
series for various distributions

22
 a. Gandhi Sagar b. Nimach

c. Bhanpura d. Mandasaur

e. Chaumala f. Badnagar

g. Ratlam h. Rampura

Fig 9 : L - moments diagram for various rain gauge stations for 1-Day maximum rainfall of
Gandhi sagar catchment

23
 i. Nagada j. Pat

k. Dewas l. Dhar

m. Ujjain n. Indore

o. Mahidpur

Fig 10 : L - moments diagram for various rain gauge stations for 1-Day maximum rainfall of
Gandhi sagar catchment

24
 Table 2: Probable maximum precipitation depth for 1-D storm using various distributions

Name : Gandhi Sagar Dam Catchment area (sq. km.): 22553

PMP PMP PMP

(using IMD (using MPWRD (using Gandhi
gridded rainfall rain gauge data) Sagar Dam
data) Authority rain
gauge data)
Hershfield’s method (1-day storm) 469.19 459.89 353.51
Return period = 25 years 191.87 201.15 240.93

Return period = 100 years 241.07 246.42 304.8

Gumbel’s method
Return period = 1000 years 322.06 320.94 409.93

Return period = 10,000

years 402.9 395.33 515

Return period = 25 years 184.77 249.8

197.82
Return period = 100 years 232.19 321.33
Log-Pearson-Type3 257.17
Return period = 1000 years 308.10 436.6
method 367
Return period = 10,000
years 419.42 549.26
496.17
Return period = 25 years 184.77 249.2
188.97
Max Return period = 100 years 223.61 318.06
Weibull’s method 230.28
precipitation Return period = 1000 years 282.59 426.93
depth, PMP 293.18
Return period = 10000 years 337.7 532.86
(mm) 352.07
Return period = 25 years 198.76 185.54 187.86
Log-normal Type-2 Return period = 100 years 253.86 234.66 231.81
method Return period = 1000 years 356.12 321.54 306.37
Return period = 10,000
years 470.73 417.96 385.44
Return period = 25 years 194.47 218.9 246.86
Log-normal Type-3 Return period = 100 years 273.74 323.46
248.93
method
Return period = 1000 years 345.46 359.64 466.93

Return period = 10,000

years 452.77 444.54 636.6

Return period = 25 years 207.57 203.12 247.06

Return period = 100 years 275.4 267.32 324

Exponential
Return period = 1000 years 388.06 373.96 467.2
Distribution
method Return period = 10,000
years 500.72 480.59 636.66

25
 Return period = 25 years 192.36 187.94 245
GLO Return period = 100 years 262.79 352.66
262.51
Distributions
Return period = 1000 years 430.09 484.61 653.2
method
Return period = 10,000
years 697.72 1018.6 1256

Return period = 25 years 188.21 247.13

191.55
Return period = 100 years 228.73 304.53
GPA distributions 220.27
Return period = 1000 years 313.13 403.06
method 247.56
Return period = 10,000
years 510.92 516.73
261.61
Return period = 25 years
172.21 219.46
174.92
Return period = 100 years 195.84 251.66
Normal distribution 199.88
Return period = 1000 years 227.19 294.4
Max method 233.01
Return period = 10,000
precipitation
years 252.99 329.66
depth, PMP 260.27
Return period = 25 years 174.92 164.27 238.7
(mm)
Return period = 100 years 199.88 168.43 246.06
Uniform distribution
Return period = 1000 years 233.01 169.68 248.27
method
Return period = 10,000
years 260.27 169.81 248.49
Return period = 25 years 222.16 239.26
232.26
Gamma distribution Return period = 100 years 281.32 296.26
method 286.38
Return period = 1000 years 381.36 383.73
384.71
Return period = 10,000
years 431.28 466.53
450.68
Return period = 25 years 250.02 262.26
254.36
Kappa distribution Return period = 100 years 286.12 330.46
method 296.81
Return period = 1000 years 402.16 453.4
352.96
Return period = 10,000
years 550.13 631.86
454.64
Return period = 25 years 202.16 230.66
220.23
Wakeby distribution Return period = 100 years 232.66 280.46
240.21
method
Return period = 1000 years 304.24 372.33
303.21
Return period = 10,000
years 386.22 495.8
351.66
Return period = 25 years 242.23 249
198.28
GNO distribution Return period = 100 years 270.02 334.2
method 286.96
Return period = 1000 years 354.66 503.73
398.60

26
 Return period = 10,000
years 498.99 719.33
550.03
Return period = 25 years 242.22 247.46
241.26
Return period = 100 years 302.96 339.2
GEV distribution 330.01
Return period = 1000 years 502.12 551.86
method 524.23
Return period = 10,000
years 619.46 904.46
719.03

Goodness of fit is observed for different distributions for various stations of Gandhi Sagar
catchment.

Fig 11 : Measure of goodness of fit for various stations of Gandhi sagar catchment

27
4.2 Calculation of PMP of 1-Day storm event using PMP Atlas

Probable maximum rainfall is estimated using PMP Atlas for Gandhi Sagar and Rana Pratap
Sagar catchment (catchment No. 404 of ganga-final-report-volume-ii). The estimated value of
maximum rainfall depth of 1-day storm, 2-day storm and 3-day storm events are tabulated
below.

Rainfall events that can affect the catchmenthave been analysed using the DAD envelope
method. The DAD curve is reproduced here (Fig. 21) from the PMP atlas.

DAD Curve for 1-Day storm

400
350
300
PRECIPITATION (MM)

250
200
150
100
50
0
0 10 20 30 40 50 60

CATCHMENT AREA (1000 SQ. KM.)

Fig 12: Envelope DAD Curve (1-day) for Catchment-404

Table 3: PMP values for 1-Day storm obtained using DAD curve

Name Catchment area (sq. Precipitation depth from 1-Day DAD

Km.) envelope (mm)

Gandhi Sagar dam 22553 251

These PMP values are determined after multiplying Standard Projected Storm values with
MMF. Next, probable maximum precipitation depth of 1-day, 2-day and 3-day annual
maximum rainfall series of our study area is obtained for different return periods such as25
years, 100 years, 1000 years and 10,000 years using isohyetal contour maps provided in PMP
Atlas volume ii.

28
a) 1-Day Annual maximum rainfall isohyets of Ganga b) 1-Day Annual maximum rainfall isohyets of Ganga river
river basin of 25 years return period(ganga-final-report- basin of 100 years return period(ganga-final-report-
volume-ii) volume-ii)

c) 1-Day Annual maximum rainfall isohyets of Ganga d) 1-Day Annual maximum rainfall isohyets of Ganga river
river basin of 1000 years return period(ganga-final-report- basin of 10,000 years return period(ganga-final-report-
volume-ii) volume-ii)

Fig 13: 1-Day Annual maximum rainfall isohyets of Ganga river basin of different return
periods (ganga-final-report-volume-ii)

Table 11: PMP values for 1-Day storm obtained using isohyet curves (Figs. 11)

Return period 25 years 100 years 1000 years 10,000 years

Storm
1-Day Annual Maximum 240 315 405 550
Rainfall (mm)

29
5 Georeferencing and digitization of topographic maps

The topographic maps are georeferenced using Q-GIS Georeferencer. This process is
executed using Open Raster feature and the output is prepared using transformation settings.
The georeferenced map is then imported to Q-GIS home screen. The coordinates are then
clearly visible on map. Then Raster map is then prepared for our research area on Q-GIS
3.30.3. The topographic maps are then digitized using Q-GIS and then a ShapeFile layer is
created on the topographic map.By using Add Polygon feature, the topographic map is
digitized for the formation of Digital Elevation Model of our study area. A virtual layer is
created on the topographic map.By using SRTM downloader, a plugin in GIS platform, a
Digital Elevation Model is prepared using Q-GIS and the DEM is exported on the
topographic map.

Fig. 14: Merged toposheets of study area

30
6 Development of SRTM digital elevation model
Digital Elevation Model is generated using Q-GIS, so that we can further import this data to
develop hydrologic model for the formation of storm hydrograph. We selected our study area
to generate the DEM of given catchment.

Fig 15: Development of SRTM digital elevation model using QGIS 3.30.3

7 Development of storm hydrograph using SNYDER's unit hydrograph

7.1 Relationships for estimating 1-h UH parameters for the study area
Table 12 Synthetic UH parameters

31
 Where, L = Length of the main stream (km), S = Equivalent stream slope (m/km), A= Area
of the catchment (km2).
7.2 Design loss rate
Generally, infiltration index and the initial losses values are derived from the available
rainfall-runoff records for the severe storms in the basin. Assuming the basin would be
saturated at the time of design storm the minimum infiltration rate and initial losses values
would be considered. The minimum infiltration rate and minimum initial loss, thus obtained,
are used to compute the effective rainfall of design storm. For this the initial losses must be
subtracted first from the rainfall increments and thereafter a uniform loss rate equal to the
minimum infiltration index is applied. However, in this study initial loss is neglected as a
conservative approach (as recommended in clauses 4.6 of AERB guide AERB/SG/S-6A).
7.3 Base flow for design flood
Base flow is the portion of stream flow that comes from the sum of deep subsurface flow and
delayed shallow subsurface flow. The recommended value of base flow 0.15 m3/s/km2 is used
in this study.
Table 13: Ordinates of flood hydrograph using SNYDER's unit hydrograph approach
Flowpath slope
Name Area (km2) Flowpath length (km) Flowpath slope (m/km) qpin cumec tpin hr
Subbasin-1 2176.9 160.95441 0.00179 1.79 0.131842 13.96029
Subbasin-4 1565.3 133.49428 0.0008 0.8 0.100981 18.62669
Subbasin-8 556.86 50.21539 0.00223 2.23 0.239553 7.318721
Subbasin-11 513.32 58.89528 0.00166 1.66 0.1969 9.047382
Subbasin-19 1155.7 101.21327 0.00117 1.17 0.13406 13.7107
Subbasin-15 1027.6 72.47414 0.00146 1.46 0.170348 10.58162
Subbasin-10 1283.4 99.06679 0.001 1 0.126446 14.60563
Subbasin-2 896.02 87.96239 0.00313 3.13 0.217721 8.115462
Subbasin-7 1728.3 162.05468 0.00229 2.29 0.14619 12.48467
Subbasin-17 837.95 66.7578 0.00137 1.37 0.171715 10.49055
Subbasin-21 1417.9 96.36921 0.00123 1.23 0.139911 13.09169
Subbasin-23 629.45 51.02359 0.00228 2.28 0.240195 7.297563
Subbasin-5 3430.4 181.21992 0.00092 0.92 0.094008 20.12511
Subbasin-24 149.22 23.44974 0.00371 3.71 0.414361 4.046563
Subbasin-13 2568.3 139.55975 0.00117 1.17 0.116714 15.92701
Subbasin-25 635.4 51.48654 0.00155 1.55 0.202574 8.77362
Subbasin-3 2036 104.42073 0.0013 1.3 0.138417 13.24456
Subbasin-6 511.81 53.55121 0.00304 3.04 0.266314 6.526783
Subbasin-16 177.62 34.19908 0.00281 2.81 0.312359 5.492877
Subbasin-14 4.7955 5.53315 0.00235 2.35 0.634374 2.553071
Subbasin-18 709.97 50.75043 0.00148 1.48 0.199814 8.90477
Subbasin-9 502.55 61.28373 0.00171 1.71 0.196046 9.090008
Subbasin-12 1479.9 99.87644 0.00243 2.43 0.184796 9.689896
Subbasin-20 768.24 57.00936 0.00153 1.53 0.192782 9.256551
Subbasin-22 579.14 51.13606 0.0037 3.7 0.295692 5.828459

32
 Flood
Hydrograph
1-h DRH ordinate ordinate
W50 W75 WR50 WR75 TB in hr Tm in hr UHordinateQp (cumec) (cumec)
6888.145851 7214.680851
17.5083 8.453968 5.455145 2.934221 51.13971 14.46029 287.0061
3793.565156 4028.360156
23.4118 11.16876 7.328843 3.886311 63.20084 19.12669 158.0652
3201.535166 3285.064166
9.133936 4.531303 2.81615 1.563837 31.82856 7.818721 133.3973
2425.740746 2502.738746
11.30952 5.560983 3.498972 1.922777 37.19093 9.547382 101.0725
3718.382442 3891.737442
17.19292 8.307963 5.355309 2.883089 50.46673 14.2107 154.9326
4201.185124 4355.325124
13.24308 6.469145 4.107644 2.239865 41.72448 11.08162 175.0494
3894.731691 4087.241691
18.32392 8.831062 5.713472 3.066321 52.86514 15.10563 162.2805
4681.984443 4816.387443
10.13623 5.006815 3.130434 1.729514 34.33768 8.615462 195.0827
6063.849726 6323.094726
15.64439 7.589403 4.865542 2.631562 47.11204 12.98467 252.6604
3453.323479 3579.015979
13.12824 6.415369 4.071451 2.221078 41.46048 10.99055 143.8885
4761.111846 4973.796846
16.41094 7.945455 5.107892 2.75617 48.7834 13.59169 198.3797
3628.575759 3722.993259
9.10733 4.518652 2.807814 1.559431 31.76097 7.797563 151.1907
7739.6537 8254.2137
25.31 12.03524 7.93304 4.19066 66.89555 20.62511 322.4856
1483.941446 1506.324446
5.027565 2.556838 1.535207 0.877831 20.59898 4.546563 61.83089
7194.13432 7579.37932
19.99483 9.601443 6.243262 3.33635 56.33644 16.42701 299.7556
3089.17898 3184.48898
10.96475 5.398402 3.390612 1.866059 36.36122 9.27362 128.7158
6763.603218 7069.003218
16.60404 8.035035 5.168969 2.787528 49.20105 13.74456 281.8168
3271.246302 3348.017802
8.138509 4.056892 2.50457 1.398701 29.26145 7.026783 136.3019
1331.55091 1358.19391
6.840338 3.434501 2.099166 1.182326 25.78085 5.992877 55.48129
73.01141147 73.73073647
3.160949 1.638861 0.958013 0.560393 14.6882 3.053071 3.042142
3404.685623 3511.181123
11.1299 5.47631 3.442514 1.893236 36.75955 9.40477 141.8619
2364.546255 2439.928755
11.3632 5.586282 3.515852 1.931604 37.31951 9.590008 98.52276
6563.496166 6785.481166
12.11898 5.941902 3.753596 2.055723 39.11257 10.1899 273.479
3554.463377 3669.699377
11.57299 5.685089 3.581818 1.966083 37.82038 9.756551 148.1026
4109.927794 4196.798794
7.261506 3.636922 2.230566 1.252662 26.92826 6.328459 171.247

33
8 Catchment delineation using HEC HMS
Streams and reaches are identified using HEC HMS to build our hydrologic model. The DEM
is delineated to form several sub basins.

(a) (b)
Fig 15: Gandhi sagar basin generated using HEC HMS
9 Development of hydrologic model using HEC HMS
The SRTM Digital elevation model is then imported into HEC HMS to develop our
hydrologic model. We used Basin model manager to create our well-defined catchment
boundary. After that, we established rain gauges inside every sub basin and we added Time
series data to each rain gauge inside those sub basins. To calculate basin loss we used SCS
Curve number method and to generate flood hydrograph, we used SCS unit hydrograph
method. The lag time and time of concentration of each sub basin is calculated using SCS
curve number method.

34
10 Importing of hourly rainfall distribution data
Hourly rainfall distribution over Gandhi Sagar catchment is given as follows. We created a
Meteorologic model to connect the rain gauges to each sub basin to ensure proper functioning
of our hydrologic model. A sink is created at the basin outlet, for which we generated the
flood hydrograph of our catchment.

120

100

80

60

40

20

0
0 5 10 15 20 25 30

Fig 16: 1-day time duration (hours vs. %precipitation)

11 Development of storm hydrograph using HEC HMS model

The following output was generated by performing hydrologic simulations for all four return
periods 25 years, 100 years, 1000 years, 10,000 years. The following graphs represent the
flood hydrographs and its peaks at various intervals.

Fig 17: Flood hydrograph of Gandhi Sagar basin generated using HEC HMS for a 1-Day
storm of various return periods.

35
Table 14: PMF values for a 1-Day storm obtained from PMP Atlas

Return period 25 years 100 years 1000 years 10,000 years

Storm

Peak of floid hydrograph 28,000 38,000 50,000 69,000

(cumec)

12 Results and discussion

Here in this report we have calculated probable maximum precipitation using various
distributions and tabulated. We have compared the calculated PMP values with those given in
PMP Atlas, Gumbel distribution performed satisfactorily well, which has given PMP values
as follows. PMP of 25 years, 100 years, 1000 years, 10000 years are calculated as 241 mm,
305 mm, 409 mm and 510 mm respectively.

Probable maximum floid is calculated using HEC HMS model, by performing catchment
delineation and sub basin formation. The peak of floid hydrograph is generated as 28000
cumecs, 38000 cumecs, 50000 cumecs and 69000 cumecs for 25 years, 100 years, 1000
years, 10000 years respectively.

13 Conclusions
This report includes PMP calculations using gridded rainfall data and available rain gauge
station data in the Gandhi Sagar catchment area. The Hershfield method, Gumbel’s method, Log-
PearsonType 3 method, Weibull’s method, Log-Normal Type 2 method and Log-Normal Type 3,
Wakeby distribution, Kappa distribution, Normal distribution, Uniform distribution, GEV
distribution, Exponential distribution, Gamma distribution, GLO distribution, GNO distribution, GPA
distribution are used to generate probable maximum precipitation in the study region. In
addition, the PMP atlas of the Ganga basin prepared by the CWC and IMD has been used to
calculate the probable maximum rainfall in the study area. The PMP values of 1-Day storm
computed using gridded rainfall data and available rain gauge stations are comparable to the
values obtained from PMP Atlas. The computed values are slightly smaller than those of
PMP Atlas. The calculated values of PMP using Gumbel distribution has given satisfactory
results, which are quite closer to the values given in PMP Atlas. The PMP values obtained
from PMP Atlas are more conservative which is representing higher/worst-case conditions.
The PMF of Gandhi Sagar catchment is produced using hydrologic model in HEC HMS,
which is quite satisfactory when compared to flood peaks of available Gandhi Sagar dam
flood reports.

36
12 References

1. D.M. Hershfield, Method for Estimating Probable Maximum Precipitation. American

Water Works Association. 1965;57(8):965-972.
2. Hosking, J. R. M. “L-Moments: Analysis and Estimation of Distributions Using Linear
Combinations of Order Statistics.” Journal of the Royal Statistical Society. Series B
(Methodological), vol. 52, no. 1, 1990, pp. 105–24.
3. Hosking, J., & Wallis, J. (1997). Regional Frequency Analysis: An Approach Based on
L-Moments. Cambridge: Cambridge University Press.
doi:10.1017/CBO9780511529443
4. Chow Ven Te David R. Maidment and Larry W. Mays. 1988. Applied Hydrology. New
York: McGraw-Hill.
5. Maity Rajib. 2019. Statistical methods in hydrology and hydroclimatology. Springer.
6. Gumbel E.J. The return period of flood flows. Ann. Math. Stat. 1941;12:163–190.
7. PMP Atlas for Ganga River Basin Including Yamuna, Final Report, Volume I: Main
Report, June 2015
8. PMP Atlas for Ganga River Basin Including Yamuna, Final Report, Volume II: Maps
and Tables, June 2015
9. CEurope (2023) URL: https://climate.ec.europa.eu/climate-change_en [Accessed in
Aug 2023]
10. PHED Rajasthan (2023)
https://phedwater.rajasthan.gov.in/content/raj/water/en/home.html [Accessed in Aug
2023]
11. Bhuvan Data (2023) https://bhuvan-app3.nrsc.gov.in/data/download/index.php
[Accessed in Aug 2023]
12. MPChambal (2023) http://mpwrd.gov.in/chambal-basin/ [Accessed in Aug 2023]
13. WHO (2023) www.who.int/health-topics [Accessed in Aug 2023]
14. Mausam (2023) https://mausam.imd.gov.in/ [Accessed in Aug 2023]
15. CARTO (2023) https://bhuvan-
app3.nrsc.gov.in/data/download/tools/document/CartoDEMReadme_v1_u1_23082011.
pdf [Accessed in Aug 2023]

37