Vol.

6 Issue

Generation of Rainfall Intensity-Duration-Frequency curves for the Barak River Basin

Violina Basumatary, Briti Sundar Sil

National Institute of Technology, Silchar, Department of Civil Engineering, NIT Road, District-Cachar, Silchar, Assam
788010, India, e-mail: violinabasumatary2725@gmail.com, britisil@gmail.com

Abstract. Analysis of a design storm, explained as the expected rainfall intensity for a given storm duration and return
period is carried out to establish Rainfall Intensity-Duration-Frequency (IDF) relationships. The IDF relationships
are essential for the designing of hydraulic structures for future planning and management. The intent was to deter-
mine IDF relationship for the Barak River Basin in India. This region receives heavy rainfall but lacks potential in
harnessing the available water resources. Rain gauge stations are not so frequent in this region. Under this condition,
the engineers are bound to use IDF relationships. Using the 24-hour annual maximum rainfall and disaggregated
hourly maximum rainfall as a measure of rainfall intensity for each station, IDF curves and isopluvial maps of rain-
fall intensities of 30-minute to 24-hour durations were generated. Analysis was done using the best fit amongst the
Gumbel, Log Pearson Type III and Lognormal probability distribution functions. The goodness-of-fit tests indicated
the Log Pearson Type III distribution was a suitable distribution for the area. After developing IDF curves for each
station, an average IDF curve was developed using the Thiessen Polygon method in ArcGIS. The IDF relationship
developed can be used as valuable information for the planning and management of hydraulic structures in the region.

Keywords: Barak River, design storm, IDF curve, isopluvial map, probability distribution, Thiessen Polygon

Submitted 2 May 2017, revised 17 July 2017, accepted 25 October 2017

1. Introduction precipitation in the form of rainfall is life-supporting, it can

also cause serious problems. An abundance or shortfall of
Rainfall is a fundamental element in the hydrologic rainfall, if not controlled, has an impact in all environmental
cycle. To design structures influenced by rainfall or con- events, and extreme events have serious consequences for
cerned with its accumulation and conveyance, engineers human society. The evaluation of extreme rainfall is critical
are required to be able to assess rainfall. Assessment of in hydrologic studies. Interpretation of their return periods
rainfall is usually done using Intensity-Duration-Frequency and design values are of great significance for the design
curves (IDF curves) for various water resource-related and management of water resource systems and reservoirs,
schemes. An IDF curve is a graphical representation of and planning for weather related urgencies, etc. depends
the probability that rainfall with a particular intensity and on knowledge of the frequency of these extreme events.
duration will occur and the probable time interval between IDF curves give useful information about when there will
storms with similar characteristics. IDF curves are a potent be flooding in an area, and when a particular rainfall rate or
representation of the extreme rainfall that is expected in a certain volume of flow will recur in the future. The IDF
a region of interest, which reflects the average intensity of curves are usually derived by carrying out annual maximum
rainfall at every return period for all durations of rainfall. analysis of historical precipitation data, assuming there is no
The determination of the probable frequency of extreme change in climate. IDF curves are useful in determining the
rainfall events of different intensities and durations is the peak runoff in a catchment, using the rational method, and
basic step in designing a flood control structure, in order to for designing channels and other water ways (Isikwue et al.
provide an economic size of the structure. On most occasions 2012). The design rainfall intensity for a specified storm
designing a structure that can withstand the greatest rainfall duration and frequency is generally estimated from a set of
ever to have occurred is not practical. Having periodic fail- statistically derived IDF curves suitable for a given region.
ures is more economical than designing for a very intense IDF relationships are suitable for estimating design
storm. However, where life is concerned, the designed storms (DS) to obtain peak discharge and hydrograph
structure should be able to administer a greater runoff than shape in any hydraulic design. The DS are applied abun-
has ever been recorded. To achieve this, data providing dantly to design flood control structures (Ewea et al. 2016).
information about return periods of extreme rainfall events Although many studies have been done to develop the IDF
of various intensities and durations are vital. Although relationships in the Brahmaputra River Basin of Assam
2 V. Basumatary, B. Sundar Sil

state (Sharma et al. 2016a; Das et al. 2016), few studies of DS was conducted based on topographic and rainfall
have been conducted in the Barak River Basin. Generating characteristics to determine rainfall IDF relationships in
IDF relationships requires good quality and long-term his- Botswana for three homogeneous regions which were con-
torical data. A common problem in developing countries structed with the K-means Clustering algorithm (Alemaw,
is the non-availability or sparse network of rain gauge Chaoka 2016). Autorun analysis on a daily time scale was
stations, whose data are the ultimate basis for establish- used to generate wet and dry spell durations and rainfall
ing IDF relationships. Some interpolated rainfall data are intensities (Subyani, Al-Amri 2015). IDF curves were pro-
available on the Global Weather Data for SWAT site and posed using Kothyari and Garde’s empirical equation and
can be downloaded from the link (https://rda.ucar.edu/pub/ then IDF curves were derived by developing a modified
cfsr.html). This dataset is the result of close cooperation IDF equation so as to maintain compatibility with changes
between two organizations, the National Centers for Envi- in the rainfall pattern in Mumbai (Zope et al. 2016).
ronmental Prediction (NCEP) and the National Center for They proposed an IDF equation for Mumbai city using
Atmospheric Research (NCAR), who completed a global a single rain gauge station. Flood frequency analysis using
climate data reanalysis over the 36-years period from 1979 probability distribution functions were performed, as they
to 2014. are useful in urban development planning and floodplain
IDF curves are discussed in numerous hydrological management (Izinyon, Ajumuka 2013; Sharma et al.
books (Chow et al. 1988). A relationship for intensity 2016a). Various studies have been carried out in different
of rainfall, duration and frequency was developed for parts of Saudi Arabia (Elsebaie 2012; Al-anazi, El-Sebaie
Rwanda. The stations were divided into five homogeneous 2013; Subyani, Al-Amri 2015). The impact of climate
regions and quantile estimation was carried out using the change on rainfall IDF was studied in Alabama (Mirhos-
Generalised Logistic, Gamma and Pearson Type III and seini et al. 2013).
Generalised Extreme Value distributions (Wagesho, Claire The Barak River Basin has good hydropower poten-
2016). IDF curves were developed for shorter durations tial but it is rather under-developed and is lacking when
and various return periods in different regions (Logah et al. it comes to harnessing the potential of available water
2013; Rasel, Islam 2015; Palaka et al. 2016; Dar, Maqbool resources. Floods, drainage congestion and bank erosion
2016). IDF curves were developed for the Sinai Peninsula are major problems of the basin. Floods are caused by
with available methods of estimation of recurrence periods, banks spilling over due to inadequate carrying capacity
such as California, Hazen, Kimbell and Gumbel. A contour of the channels; this results from aggravation of the river
map of Sherman coefficients was developed using GIS beds, backing up effect of the main river on to its tributar-
tools. Rainfall intensities obtained using the Log Pearson ies and also due to excessive rainfall in the region. Due
Type III (LPIII) and Gumbel methods showed uniformity to poor drainage, many depressions remain waterlogged
with the results accomplished in studies carried out previ- long after monsoons. Hydrological analysis is a prereq-
ously on the area (Fathy et al. 2014). Isopluvial maps were uisite for the better utilization of the resources (Bora,
developed for the Tehri-Garhwal Himalayan region, India. Choudhury 2015). In this study, an attempt has been made
Isopluvial maps show the spatial variability and IDF rela- to develop an Intensity-Duration-Frequency model for the
tionships developed for the Bhagirathi-Bhillangana catch- Barak River Basin using annual maximum series (AMS).
ment (Sarkar et al. 2010). Spatial and temporal variation The IDF model can be used to estimate DS intensities.
of IDF parameters was developed in the Awash River The suggested expression allows the estimation of the
Basin, Ethiopia (Gebreslassie 2014). IDF curves were expected rainfall intensities for different durations and
developed using the Gumbel and LPIII distributions, with return periods in the basin area.
the Gumbel distribution showing higher results than the
LPIII distribution (Antigha, Ogarekpe 2013). The safety 2. Study area
factor was considered to be necessary while evaluating IDF
curves for different parts of a city because rainfall inten- The Barak River, one of the major rivers of south
sity is not uniform throughout a city (Sharma et al. 2016b). Assam, India, rises in the hill country of Manipur State,
A mathematical model was developed between peak flood where it is the biggest and most important of the hill coun-
discharge and return period to assess rainfall for different try rivers. The Barak Basin covers the States of Assam,
storm durations for the Bhadar catchment, Gujarat (Bhatt Manipur, Mizoram, Nagaland and Tripura. From Assam,
et al. 2014). A study was carried out to find the influence of it enters Bangladesh, where the Surma and Kushiyara
global warming on rainfall intensity in Toronto before and rivers begin, and is a part of the Surma-Meghna River
after 1980 (Carlier, Khattabi 2016). A regional analysis System. The Barak River is the single largest contributory
 Generation of Rainfall Intensity-Duration-Frequency curves for the Barak River Basin 3

the Indian Meteorological Department (IMD) one third

reduction formula, as given in equation (1). Chowdhury
et al. (2007) used the IMD one-third reduction formula to
estimate short duration rainfall for Slyhet city and found
the formula gave the best estimation for short duration
rainfall (Rashid et al. 2012).

( )
1/3
t
pt = p24 (1)
24

where: pt denotes the rainfall depth (millimetres) for t-hour

duration, p24 denotes the daily rainfall (millimetres), and
Fig. 1. Location map of the study area t denotes the duration (hours) of rainfall for which the rain-
fall depth is required. The rainfall events were analysed by
river to the northeast region of Bangladesh. The major breaking down the AMS to shorter durations of 30-minute,
tributaries of the Barak are all in India and are the Jiri, 1-hour, 2-hour, 3-hour, 6-hour, 12-hour and 24-hour.
the Dhaleshwari, the Singla, the Longai, the Madhura,
the Sonai, the Rukni and the Katakhal. The total area 3.2. IDF quantile estimation
of the Barak River Basin is 41 723 km2 which is nearly
1,38% of the total geographical area of India. In this study, In the present study, the return period (T) was com-
an area of nearly 25 000 km2 is considered, up to Badar- puted using the Weibull plotting position formula as given
purghat, India (92,58°E, 24,87°N) where it enters Bang- in equation (2),
ladesh. The topography of this region is characterized by
hills, hillocks, plains and low lying waterlogged areas. n + 1
T  = (2)
Generally, winter lasts from December to February and the m
period is rather dry. The climate of the Barak Basin is sub-
tropical, warm and humid. The annual rainfall ranges from where n is the total number of years of rainfall data and
2 500 mm to 4 000 mm. The periods from March-April m is the rank of the events. The probability of exceedance
and October-November are defined by low erratic rainfall of the rainfall values is the reciprocal of the return period
with occasional hailstorms. The period between May (Kumar, Bhardwaj 2015). An extreme event (XT) is decided
and September is defined by high rainfall with a risk based on the assumed probability distribution employing
of floods. The seasonal distribution of rainfall indicates the method of moments, method of L-moments or method
that while total annual rainfall is satisfactory, the dis- of maximum likelihood (Stedinger, Cohn 1986; Hosking,
tribution is uneven and most of the rainfall is confined Wallis 1993). Alternatively, with reference to frequency
to the period between June and August. Normally, the factors, it can be computed for the probability distribution
temperature ranges from a minimum of 12,2°C in January functions (Chow et al. 1988) as given in equation (3),
to 25,4°C in August and the mean maximum of 24,3°C
in January to 36,0°C in August. The map of the study area XT = X + KTσX (3)
is shown in Fig. 1.
where KT is the frequency factor which is a function of
3. Materials and methods return period and the parameter of the distribution, and
3.1. Data collection X and σX are mean and standard deviation of the variate.
For LPIII, the probability of exceedance corresponding
For the generation of IDF curves, daily precipitation to T-year recurrence interval was assigned to each variate
data at 23 stations in the study area were obtained from using the Weibull plotting position formula. The mean,
https://globalweather.tamu.edu/ for a period of 35 years standard deviation and coefficient of skewness for the log
(1979-2013). The stations considered for this study in transformed series were computed using equations (4),
the Barak Basin were located between 92,4°E to 94,3°E (5), and (6), respectively.
longitude and 23,5°N and 25,7°N latitude. From the data-
base, the annual maximum series (AMS) were extracted ∑log X (4)
log X =
and disaggregated to shorter duration rainfall series using n
4 V. Basumatary, B. Sundar Sil

∑(log X – log X )
√
2
of less than 24 hours, such as 0.5-hour, 1-hour, 2-hour,
σlog X = (5)
n – 1 3-hour, etc. are of huge significance for various drainage
designs and for water resource management, they were
n∑(log X – log X )
3

Cs = (6) chosen for the construction of the IDF curves for this study.
(n – 1) (n – 2)σ
 
3
log X The gist of this study is that frequency analysis of rainfall
intensity for various durations and return periods was used
The value of XT for a specified return period was com- to derive the spatial distribution of IDF for the region.
puted by equation (7), For the respective periods, the maximum of 0,5-hour,
1-hour, 2-hour, 3-hour, 6-hour, 12-hour and 24-hour rain-
log XT = log X + KT σlog X (7) fall are used for frequency modelling. The distributions,
namely the LPIII, LN and Gumbel were fitted to the rainfall
3.3. Goodness-of-fit (GOF) measures data. After model fitting and performance interpretation,
the best distribution was picked based on goodness-of-fit
For the selection of a particular distribution repre- tests, as discussed.
senting a rainfall series, for the dependable projection of
a storm event, it is essential to find the best fit method of 3.5. Derivation of IDF equation
frequency analysis. The three different most widely used
GOF tests, viz., Kolmogorov-Smirnov (K-S), Anderson- The IDF formulae depict a relation among maximum
Darling (A-D) and Chi-Squared (χ2) tests, are imple- rainfall intensity as a dependent variable, with other rel-
mented to select the most appropriate model. The said evant parameters being rainfall duration and frequency
tests compute test statistics, which are used to check the as independent variables. There are many frequently
suitability of a given distribution to fit the data. The test used functions relating these variables mentioned in the
statistics are described in Table 1. The EasyFit software literature of hydrology applications (Chow et al. 1988).
employs the method of moments for the Gumbel and LPIII There are some necessary steps for forming the equation
distributions, and the maximum likelihood method for for estimating the intensity of rainfall for a specific return
the LN distribution for the estimation of the parameters period and given rainfall duration: these rely mainly on the
(Sharma et al. 2016a). The estimated parameters for differ- results attained from the IDF curves. The general form of
ent parametric distributions were fitted to 24-hour series at empirical equation can be written as shown in equation (8):
all the stations. The test statistics were computed and the
probability distributions were ranked based on the lowest CT rm
I= (8)
values of the test statistics. T de

3.4. Construction of IDF curves where I is average rainfall intensity for any duration Td, Tr
is the return period, and C, e and m are constants (Elsebaie
The Intensity-Duration-Frequency (IDF) curves were 2012).
worked out for each station. As shorter rainfall durations

Table 1. Description of goodness-of-fit tests (Sharma et al. 2016b)

GOF test Test statistic Interpretation
Kolmogorov-Smirnov D = max ǀ P(Xm) – F(Xm)ǀ The test decides if a sample comes from a hypoth-
(K-S) test where P(Xm) is the cumulative probability for each of the ordered esized continuous distribution. The hypothesis
observations Xm using Weibull’s formula, and F(Xm) is the theoretical regarding the distributional form is rejected at the
cumulative probability for each of the ordered observations Xm using chosen significance level (α) if the test statistic,
the assumed distribution. D, is greater than the critical value.
Anderson-Darling A2 = –n – S The hypothesis regarding the distributional form
S = ∑ n [ln F(Yk) + ln{1 – F(Yn + 1 – k)}]
2k – 1
(A-D) test is rejected at the chosen significance level (α) if
where Y1, Y2,..., Yn = data series, F = cumulative distribution function the test statistic, A2, is greater than the critical value.
(CDF), and n = size of the sample.
k
Chi-Squared The hypothesis that the data are from a popula-
χ2 = ∑ (Oi – Ei)2
(χ2)test i=1
Ei tion with the specified distribution is rejected,
where Oi = observed frequency for class i, and Ei = expected fre- if χ2 › χ2(α, k-c), where χ2(α, k-c) is the critical test statistic
quency for class i. value with k-c degrees of freedom and a significant
level (α).
 Generation of Rainfall Intensity-Duration-Frequency curves for the Barak River Basin 5

3.6. Thiessen Polygon method 4. Results and discussion

The Thiessen Polygon method estimates average rain- The annual maximum series was extracted from daily
fall aspects over an area. The recorded rainfall at each rain rainfall and disaggregated to smaller durations using the
gauge station is given a weightage according to the posi- IMD reduction formula for all the 23 stations. The AMS
tion of the station with respect to the boundary of the area. was subjected to preliminary analysis for consistency
The average intensity of rainfall over an area is expressed using the Double Mass Curve method. Figure 2 shows
in equation (9): a comparison between observed and expected 24-hour
rainfall based on the Gumbel, LPIII and LN methods for
I1 A1 + I2 A2 + ... + In An station 236925. The performances of the models were
I= (9)
A1 + A2 + ... + An decided against a model performance index, which gave
very high model efficiency, a coefficient of determina-
where I1, I2,…, In are the intensities of recorded rainfall at tion (R2) that is above 90%. The goodness-of-fit of the
the respective stations 1, 2,.., n and A1, A2,…, An are the distribution was worked out using EasyFit software 5.6.
areas of the Thiessen Polygons, respectively. The Thiessen It shows the model parameters and the goodness-of-fit of
Polygon method is used to produce an average IDF curve the Gumbel, LPIII and LN distributions. The results of
for the whole study area (Sarkar et al. 2010). goodness-of-fit tests for 24-hourseries using the K-S, A-D
and χ2 tests for the stations are shown in Table 2. From the
3.7. Isopluvial map results, LPIII can be considered more reliable for the study
area. After completion of the above steps, the IDF curves
The rainfall intensity values at different sites inside were generated for each station at time steps 30-minute,
the Barak River Basin were interpolated using the Geo- 1-hour, 2-hour, 3-hour, 6-hour, 12-hour and 24-hour and
statistical Analyst in GIS. To show spatial variability of an average IDF curve for the whole area was deduced
rainfall intensity, Isopluvial maps were generated using using the Thiessen Polygon method in ArcGIS. Figure
the Inverse Distance Weighted (IDW) method (Sarkar et 3 shows the Thiessen Polygon map of the basin. Thiessen
al. 2010). To interpolate a value for any unmeasured site, Polygon area weightages were assigned to each station
IDW executes the assumption that each measured point and the quantiles at each station were used to generate an
has a local impact that decreases with distance. average IDF curve for the basin. Figure 4 shows the IDF

Table 2. Goodness-of-fit results for 24-hour series

Kolmogorov-Smirnov Anderson-Darling Chi-Squared
Station ID Distribution
Statistic Rank Statistic Rank Statistic Rank
Gumbel 0,13801 3 0,65068 3 2,5056 3
248928 Log normal 0,13005 2 0,58594 2 1,3188 1
Log Pearson type III 0,12183 1 0,52785 1 1,3437 2
Gumbel 0,15941 3 0,87762 3 7,9743 3
248931 Log normal 0,15519 2 0,77755 2 7,7261 2
Log Pearson type III 0,13173 1 0,57116 1 2,346 1
Gumbel 0,10941 3 0,53773 3 3,6693 3
245928 Log normal 0,09271 1 0,40278 2 1,6901 2
Log Pearson type III 0,09327 2 0,34376 1 0,72095 1
Gumbel 0,08388 3 0,41902 3 1,3578 1
245931 Log normal 0,07392 2 0,29631 2 1,571 2
Log Pearson type III 0,06961 1 0,26244 1 2,1869 3
Gumbel 0,09764 3 0,49631 3 0,5119 1
242925 Log normal 0,07467 2 0,12534 1 0,69228 2
Log Pearson type III 0,5342 1 0,12534 1 0,69228 2
Gumbel 0,13861 3 0,53783 3 4,702 3
242928 Log normal 0,13222 2 0,46006 2 2,5999 2
Log Pearson type III 0,11038 1 0,41661 1 0,03525 1
6 V. Basumatary, B. Sundar Sil

Gumbel 0,16298 2 1,5884 3 10,573 3

239925 Log normal 0,17289 3 1,5554 2 6,5491 2
Log Pearson type III 0,09612 1 0,40763 1 2,2212 1
Gumbel 0,12932 3 0,64651 3 0,94613 2
239928 Log normal 0,10524 2 0,63567 2 2,9543 3
Log Pearson type III 0,08962 1 0,27501 1 0,31525 1
Gumbel 0,08462 1 0,38324 2 0,4914 1
236925 Log normal 0,09097 2 0,4251 3 0,78046 2
Log Pearson type III 0,09369 3 0,33147 1 2,3965 3
Gumbel 0,10068 3 0,33659 3 2,3461 3
236928 Log normal 0,09098 2 0,22878 2 1,2423 2
Log Pearson type III 0,08768 1 0,21161 1 1,2122 1
Gumbel 0,11299 3 0,33578 3 0,37423 1
242931 Log normal 0,09237 2 0,2565 2 4,1936 3
Log Pearson type III 0,09084 1 0,25227 1 0,46289 2
Gumbel 0,12029 3 0,45016 3 1,5712 3
242934 Log normal 0,10499 1 0,37494 1 0,18537 1
Log Pearson type III 0,11099 2 0,40717 2 0,49741 2
Gumbel 0,12177 3 0,62073 3 1,6715 3
245934 Log normal 0,09974 2 0,34381 2 0,25393 2
Log Pearson type III 0,70606 1 0,21391 1 0,12574 1
Gumbel 0,09372 3 0,34973 3 1,9823 3
248934 Log normal 0,08133 1 0,30776 2 0,78512 2
Log Pearson type III 0,08523 2 0,25791 1 0,2821 1
Gumbel 0,09752 2 0,4269 3 3,6883 3
248938 Log normal 0,09615 1 0,33164 1 1,9197 1
Log Pearson type III 0,10945 3 0,33164 1 1,9197 1
Gumbel 0,10406 1 0,63067 2 2,6008 2
239931 Log normal 0,11698 2 0,55521 1 2,5886 1
Log Pearson type III 0,12174 3 4,2461 3 N/A*
Gumbel 0,0787 2 0,24682 2 1,6027 2
239934 Log normal 0,09874 3 0,34245 3 1,2849 1
Log Pearson type III 0,07278 1 0,2434 1 2,5893 3
Gumbel 0,08013 2 0,19673 3 2,0192 2
251928 Log normal 0,07632 1 0,17554 2 1,353 1
Log Pearson type III 0,0813 2 0,1718 1 2,8569 3
Gumbel 0,07447 2 0,21814 2 0,6259 1
251931 Log normal 0,08375 3 0,24479 3 1,1995 3
Log Pearson type III 0,07111 1 0,21691 1 0,84156 2
Gumbel 0,09743 3 0,31712 3 1,5269 2
251934 Log normal 0,0792 2 0,19523 2 2,5315 3
Log Pearson type III 0,07176 1 0,13867 1 0,50659 1
Gumbel 0,09235 2 0,24434 2 4,5212 3
251938 Log normal 0,09653 3 0,24497 3 4,3074 2
Log Pearson type III 0,08289 1 0,20154 1 2,5582 1
Gumbel 0,05709 3 0,09683 2 1,3036 3
254938 Log normal 0,0556 2 0,10079 3 0,26242 2
Log Pearson type III 0,04818 1 0,0863 1 0,17781 1
Gumbel 0,06839 1 025347 2 0,94356 2
254941 Log normal 0,07274 2 0,26136 3 0,99382 3
Log Pearson type III 0,07422 3 0,20767 1 0,35569 1
*Each bin in the Chi-Squared test must contain at least 5 data points, so certain adjacent bins sometimes need to be joined together.
This can result in a small number of bins that would, in turn, be not enough for the test to be applied.
 Generation of Rainfall Intensity-Duration-Frequency curves for the Barak River Basin 7

Fig. 2. Comparison between observed and simulated

24-hour rainfall depth based on:
(a) Gumbel, (b) Log Pearson Type III and
(c) Lognormal distributions for station 236925

Table 3. Rainfall Intensities [mm per hour] for different durations Table 4. Rainfall intensity parameters for
and return periods by the LPIII method for the entire basin the Barak Basin area

Duration Return period [year] Parameters LPIII Gumbel LN

[hour] 2 5 10 25 50 100 C 39,02 40,87 40,24
0,5 67,89 90,19 105,70 126,27 142,37 158,82 m 0,21 0,21 0,19
1 42,73 56,80 66,63 79,72 90,00 100,55 e 0,67 0,67 0,66
2 26,92 35,78 41,97 50,22 57,59 63,34
3 20,54 27,31 32,03 38,33 43,88 48,34
study, in most cases the LPIII distribution fitted best and
6 12,94 17,20 20,18 24,14 27,99 30,45
was employed to develop IDF curves for each station
12 8,16 10,85 12,73 15,22 17,19 19,20
and an average IDF curve was developed by the Thies-
24 5,14 6,73 7,90 9,46 10,68 11,94
sen Polygon method for the entire region. An empirical
rainfall intensity formula for the Barak River Basin was
curves on a logarithmic scale for the Gumbel, LPIII and developed – this plays a key role in the popular Rational
LN methods. Finally, using the data from Table 3, the aver- Formula for computing runoff. This formula is recom-
age IDF equation can be derived. Table 4 shows the results mended for designing and planning hydraulic structures
of the empirical parameters from the Gumbel, LPIII and in the Barak River Basin area. Isopluvial maps for the
LN methods. The average empirical IDF equation for the spatial distribution of rainfall intensities were drawn for
LPIII distribution can be written as in equation (10): different durations and return periods. It can be deduced
that if rainfall intensity values are needed as information
I = 39,02T r0,21 T d–0,67 (10) for the entire basin, then the empirical relation can be
used. If on site rainfall intensity values are desired, then
The spatial distribution of the LPIII model of the IDF Isopluvial maps can be used corresponding to the location
curves for 30-minute rainfall intensities for return periods of the site. It is observed that rainfall intensity increases
of 2-year, 5-year, 10-year, 25-year, 50-year and 100-year with the increase in return period. This explains why large
in the Barak River Basin is highlighted in Fig. 5. structures like bridges and dams are planned for higher
return periods while small structures such as irrigation
5. Conclusions tanks and culverts are planned for low return periods.
Also, intensities of rainfall decrease with the increase
Information deducted from IDF curves is useful for in duration of rainfall for a specified return period.
engineers and hydrologists engaged in the planning and This signifies that intensive rainfall of short durations can
designing of water resources related projects. For this have calamitous consequences for the environment.
8 V. Basumatary, B. Sundar Sil

Fig. 3. Thiessen Polygon map of the basin

Fig. 4. Average IDF curves for the basin using (a) Gumbel, (b) Log Pearson Type III and (c) Lognormal distributions
Generation of Rainfall Intensity-Duration-Frequency curves for the Barak River Basin 9

Fig. 5. Isopluvial maps of 30-minute rainfall intensities for different return periods
10 V. Basumatary, B. Sundar Sil

Acknowledgment Ewea H.A., Elfeki A.M., Al-Amri N.S., 2016, Development of

We would like to show our gratitude to Prof. U. Kumar, Intensity-Duration-Frequency curves for the Kingdom of
HOD, Civil Engineering Department, NIT Silchar for Saudi Arabia, Geomatics, Natural Hazards and Risk, 1-15,
sharing his pearls of wisdom with us during the course of DOI: 10.1080/19475705.2016.1250113
this research. Also, we would like to thank to other faculty Fathy I., Negm A.M., El-Fiky M., Nassar M., Al-Sayed, E., 2014,
members, staffs and reviewers for their so-called insights Intensity Duration Frequency curves for Sinai Peninsula,
and constant support. Egypt, International Journal of Research in Engineering and
Technology, 2 (6), 105-112
Bibliography Gebreslassie B., 2014, Spatial and temporal rainfall Intensity-
Duration-Frequency parameters variability for Awash River
Al-anazi K.K., El-Sebaie I.H., 2013, Development of Intensity- Basin, Ethiopia, Thesis for Master of Science, Haramaya
Duration-Frequency relationships for Abha city in Saudi University, Ethiopia
Arabia, International Journal of Computational Engineering Hosking J.R.M., Wallis J.R., 1993, Some statistics useful in
Research, 3 (10), 58-65 regional frequency analysis, Water Resources Research,
Alemaw B.F., Chaoka R.T., 2016, Regionalization of Rainfall 29 (2), 271-281, DOI: 10.1029/92WR01980
Intensity-Duration-Frequency (IDF) curves in Botswana. Isikwue M.O., Onoja S.B., Laudan K.J., 2012, Establishment of an
Journal of Water Resource and Protection, 8 (12), 1128-1144, empirical model that correlates Rainfall-Intensity-Duration-
DOI: 10.4236/jwarp.2016.812088 Frequency for Makurdi area, Nigeria, International Journal
Antigha R.E., Ogarekpe N.M., 2013, Development of Intensity of Advances in Engineering and Technology, 5 (1), 40-46
Duration Frequency curves for Calabar metropolis, South- Izinyon O.C., Ajumuka H.N., 2013, Probability distribution
South, Nigeria, International Journal of Engineering and models for flood prediction in Upper Benue River Basin –
Science, 2 (3), 39-42 Part II, Civil and Environmental Research, 3 (2), 62-74
Bhatt J.P., Gandhi H.M, Gohil K.B., 2014, Generation of Inten- Kumar R., Bhardwaj A., 2015, Probability analysis of return
sity Duration Frequency curve using Daily Rainfall Data for period of daily maximum rainfall in annual data set of Ludhi-
different return period, Journal of International Academic ana, Punjab, Indian Journal of Agricultural Research, 49 (2),
Research for Multidisciplinary, 2(2), 717-722. 160-164, DOI: 10.5958/0976-058X.2015.00023.2
Bora K., Choudhury P., 2015, Estimation of annual maximum Mirhosseini G., Srivastava P., Stefanova L., 2013, The impact
daily rainfall of Silchar, Assam, International Conference of climate change on rainfall Intensity-Duration-Frequency
on Engineering Trends and Science and Humanities, Trichy, (IDF) curves in Alabama, Regional Environmental Change,
Tamil Nadu, Imayan College of Engineering, 12-16 13 (1), 25-33, DOI: 10.1007/s10113-012-0375-5
Carlier E., Khattabi J.E., 2016, Impact of global warming on Logah F.Y., Kankam-Yeboah K., Bekoe E.O., 2013, Developing
Intensity-Duration-Frequency (IDF) relationship of precipi- Short Duration Rainfall Intensity Frequency curves for Accra
tation: a case study of Toronto, Canada, Open Journal of Mod- in Ghana, International Journal of Research in Engineering
ern Hydrology, 6 (1), 1-7, DOI: 10.4236/ojmh.2016.61001 and Computing, 1 (1), 67-73.
Chow V.T., Maidmen D.R., Mays L.W., 1988, Applied hydrol- Palaka R., Prajwala G., Navyasri K.V.S.N., Anish I.S., 2016,
ogy, McGraw-Hill, New York, 572 pp. Development of Intensity Duration Frequency curves for
Chowdhury R., Alam J.B., Das P., Alam M.A., 2007, Short dura- Narsapur Mandal, Telangana State, India, International
tion rainfall estimation of Slyhet: IMD and USWB method, Journal of Research in Engineering and Technology,
Journal of Indian Water Works Association, 9, 285-292 5 (6), 109-113
Dar A.Q., Maqbool H., 2016, Rainfall Intensity-Duration-Fre- Rasel M., Islam M., 2015, Generation of Rainfall-Duration-
quency relationship for different regions of Kashmir Valley Frequency relationship for North-Western region in Bangla-
(J&K) India, International Journal of Advance Research in desh, IOSR Journal of Environmental Science, Toxicology
Science and Engineering, 5 (1), 113-125 and Food Technology, 9 (9), 41-47, DOI: 10.9790/2402-
Das R., Goswami D., Sharma B., 2016, Generation of Intensity 09914147
Duration Frequency curve using short duration rainfall data for Rashid M.M., Faruque S.B., Alam J.B., 2012, Modelling of Short
different return period for Guwahati city, International Journal Duration Rainfall Intensity Duration Frequency (SDR-IDF)
of Scientific and Engineering Research, 7 (7), 908-911 equation for Slyhet city in Bangladesh, ARPN Journal of Sci-
Elsebaie I.H., 2012, Developing rainfall Intensity-Duration- ence and Technology, 2 (2), 92-95
Frequency relationship for two regions in Saudi Arabia, Sarkar S., Goel N.K., Mathur B.S., 2010, Development of iso-
Journal of King Saud University – Engineering Sciences, pluvial map using L-moment approach for Tehri-Garhwal
24 (2), 131-140, DOI: 10.1016/j.jksues.2011.06.001 Himalaya, Stochastic Environmental Research and Risk
 Generation of Rainfall Intensity-Duration-Frequency curves for the Barak River Basin 11

Assessment, 24 (3), 411-423, DOI: 10.1007/s00477-009- Research, 22 (5), 785-793, DOI: 10.1029/WR022i005p00785
0330-2 Subyani A.M., Al-Amri N.S., 2015, IDF and daily rainfall gen-
Sharma R., Choudhury N., Alam R., Seleyi V., Sangtam Y., eration for Al-Madinah city, western Saudi Arabia, Arabian
2016a, Development of Intensity-Duration-Frequency Journal of Geosciences, 8 (12), 11107-11119, DOI: 10.1007/
curves for precipitation in western watershed of Guwahati s12517-015-1999-9
(Assam), International Journal of Latest Trends in Engineer- Wagesho N., Claire M., 2016, Analysis of rainfall Intensity-
ing and Technology, 6 (4), 575-579 Duration-Frequency relationship for Rwanda, Journal
Sharma P.J., Patel P.L., Jothiprakash V., 2016b, At-site flood fre- of Water Resource and Protection, 8 (7), 706-723, DOI:
st
quency analysis for Upper Tapi Basin, India, [in:] 21 Inter- 10.4236/jwarp.2016.87058
national Conference on Hydraulics, Water Resources and Zope P.E., Eldho T.I., Jothiprakash V., 2016, Development of
Coastal Engineering (HYDRO 2016), Pune, CWPRS, 1-10 rainfall Intensity Duration Frequency curves for Mumbai
Stedinger J.R., Cohn T.A., 1986, Flood frequency analysis with city, India, Journal of Water Resource and Protection, 8, 756-
historical and Paleoflood information, Water Resources 765, DOI: 10.4236/jwarp.2016.87061