Professional Documents
Culture Documents
MANOJ M. MUJUMDAR2
2
Introduction
Flood estimates are required for the design and economic appraisal of a variety of engineering
works, including dam, spillways, bridges and flood protection works. Flood estimates are also required for the
safe operation of flood control structures, for taking emergency measures such as maintenance of flood
levees, evacuating the people to safe localities etc. Two main approaches are available for flood estimation viz.,
deterministic approach and statistical approach. Deterministic approach assumes that input, say, the
precipitation is related to the output in a predefined manner and there is no uncertainty involved in arriving at
the output, say the discharge., whereas the statistical approach treats the inter-relationship between processes as
governed by theory of statistics. The inter-relationship between processes is established through the measures of
correlation, the processes considered may be multivariate or, univariate. For example, the rainfall-runoff process
may be considered as multivariate while, the consideration of maximum annual peak series falls under
univariate process. Flood frequency analysis deals with univariate process comprising of maximum peak flow
values. Before discussing the frequency analysis it is necessary to distinguish between the terminologies of
prediction and forecasting from the consideration of their field use.
Estimation of floods arises in hydrology from two different aspects of prognostication. The first which
is usually called forecasting, involves the estimation of the values of some hydrological variable at a prescribed
future instant or a forecast of the time when a particular value will occur. Forecasting the level of a flood
discharge or a water level in a river some days ahead of the event is an obvious example. The second aspect,
usually known as prediction is concerned not with the value of a single future event, but with the frequency of
occurrence of some prescribed critical condition. For example, how often a particular level will be exceeded.
Problems of first category arise most directly in the operation of hydrological controls in the broadest sense
including flood warning. Problems of second category are associated with the designs rather than the operation
of such controls. Design flood estimation may also be obtained through deterministic approaches say rainfallrunoff models. Therefore essentially predication deals with "how often" problem while the forecasting deals
with "when" problem.
ISSN : 0975-5462
5252
Akshay R. Thorvat et al. / International Journal of Engineering Science and Technology (IJEST)
2.
The problem of flood frequency prediction i.e. Estimation of the relationship between the magnitude of
peak flow and us corresponding return period is a central one in the field of applied hydrology. Attempts to
solve this problem are usually based on (a) the analysis of a record of peak flow data at the site in question and
(b) use of previously established relationship between the characteristics of other catchments in the region and
the parametric values of the corresponding magnitude return period relationship. The latter approach is known
as regional frequency analysis used for estimating floods at sites where there is a very short peak flow record or
no record available.
2.1.
Hydrologic data are mostly available as samples of limited sizes. Using statistical principles we extract
the needed information from the available sample data and conclude about the characteristics of the population.
Since any survey or any attempt cannot exhaust all possible events of a variable, we assume that the sample is
representative of population.
2.1.2.
We assume that that the sample of peak flows available is independent of each other and they are
assumed to be evolved from a purely random process.
2.1.3.
When a series of events arranged in time show no systematic variations in time (e.g., a seasonal
variation or an increasing or diminishing trend) so that we may say that the probability of an event in a period (t)
is independent of the location, the series is said to be homogeneous The factors which affect the homogeneity of
peak flows are the development in the catchments over time such as deforestation, urbanization, flood control
works, earthquakes etc.
2.2.
There have been significant developments and studies in the area of regional flood frequency analysis
in India as well as abroad. Estimation of regional flood frequency parameters is performed for a specific site for
two reasons:
Because of the sample variations present in the short hydrologic records, frequency estimates of rare events
based on at site frequency analysis are subjected to large error and thus unreliable. This error can be
reduced by combining data from many more sites,
There are many more sites in the same region where hydrologic data are not available but design flood
estimates are needed for the design of small structures. In such a situation regional flood frequency analysis
helps in transferring the knowledge arrived from gauged sites to ungauged sites.
2.3.
The random variability of such hydrologic variables as stream flow and precipitation has been
recognized for centuries. The general field of hydrology was one of the first areas of science and engineering to
use statistical concepts in an effort to analyze natural phenomena. The use of statistics in hydrology provides the
information about various parameters and distribution of random variables of importance to design and
operation of structures. These parameters and distributions are estimated as approximations from the available
data because they cannot be determined exactly.
Hydrologic data are mostly available as samples of limited sizes. Statistics is the main discipline
enabling the extraction of needed information from data and the derivation of conclusions about the
characteristics of hydrologic random variables Statistical estimates are numerical properties of samples. They
are necessary in statistical modeling, or for direct use in hydrology. To be effective in application of statistics in
hydrology the civil engineer or hydrologist must understand the fundamentals of statistical methods which are
employed in existing hydrologic techniques.
ISSN : 0975-5462
5253
Akshay R. Thorvat et al. / International Journal of Engineering Science and Technology (IJEST)
Normal distribution
Exponential distribution
2.4.
Flood frequency analysis is a tool being widely used for predicting the future flood at different
recurrence intervals. The reliable estimates of the magnitude and frequency of occurrence of flood are essential
to the proper design of hydraulic structure across a river as well as to identify the flood risk area. Mainly there
are two methods of estimating the floods; i.e., deterministic and statistical approach. In the deterministic
approach, the rainfall-runoff relationships established based on the physical concepts of the various hydrological
processes are used to estimated the floods. In the statistical approach, the past records of flood peaks are
subjected to the statistical analysis which provides the distribution pattern for the flood peaks. The frequency
analysis is a statistical technique by means of which it is possible to estimate the floods of various magnitudes
and their frequencies. The flood frequency analysis for a river site with a long record can be based almost
exclusively on the flood record at that site.
3.
Data Requirement
All frequency techniques are totally data dependent. An assumption must be made of a theoretical
frequency distribution suitable for the population events and the statistical parameters of the distribution must be
computed from the sample data. Two types of sample data, namely (i) annual peak flood series and (ii) partial
duration series may be used for flood frequency analysis.
Annual peak flood series is arrived at from the recorded flood peaks by picking up only one event from
each year of the record. Annual peak flood series ensures complete randomness of the data and thus assumption
of randomness is satisfied. But a disadvantage of using this series for analysis is that the second or third highest
events in a particular year may be higher than some of the year's annual peak floods and still they are totally
disregarded in the analysis. Such a disadvantage is remedied by using the partial duration series in which all the
events above a certain threshold are included in the analysis. However care should be taken not to include those
peaks which are dependent as the assumption of randomness would be violated. This can be achieved by
ensuring that consecutive flood peaks are separated by a recession of a suitable length of time. The procedures
for dealing with dependent data are still in research stage.
As a preliminary step the basic data should be screened and adjusted to remove, as far as possible, any
non-conformity that may exist. The following are the more important considerations (CWC, 1969).
Effect of man made changes in the regime of flow should be investigated and adjustment be made as
required.
For small catchments areas a distinction should be made between daily maxima and instantaneous or
momentary, flood peaks.
ISSN : 0975-5462
5254
Akshay R. Thorvat et al. / International Journal of Engineering Science and Technology (IJEST)
Changes in the stage discharge relation render stage records non homogeneous and unsuitable for
frequency analysis studies. It is therefore preferable to work with discharges and if stage frequencies
are required, refer the results to the most recent rating.
Any useful information contained in data publications and manuscripts should be made use of after
proper scrutiny.
The records used for the frequency analysis should satisfy certain assumptions in order to have
meaningful estimates;
Sample of data should be long enough to provide reliable estimates of the parameters.
4.
The Krishna basin for which sufficient annual peak flood series at number of gauging station were
available was selected as the study area. The total catchments area of the basin considered for the analysis is
90,000 sq.km and is located between longitude of 73E to 72E and latitude of 15N to 19N and it comprises
the part of Maharashtra and Karnataka states. The figure 1 shows the river system and gauging stations with all
its tributary of river Krishna. The drainage area of these gauging sites varies from 540 sq.km to 70,000 sq.km.
The main tributaries of river Krishna are river Bhima, Ghataprabha, Malaprabha and Tunga-Bhadra. The TungaBhadra basin which forms a part of the Krishna basin has not been included in the study.
5.
Methodology
The method used in the present study to carry out the regional flood frequency analysis involves the
USGS method. The USGS method for estimating the floods of given recurrence intervals for ungauged
catchments consists of following sequential steps:
Select gauged catchments within region having more or less similar hydrological characteristics to that
of the ungauged catchments.
Establish flood frequency curves for each gauging station using EV-I distribution probability paper.
Rank ratios of selected return period floods to the mean annual flood at each station, and
Compute median flood ratio for each of the selected return period of step (5), multiply by the estimated
mean annual flood of the ungauged catchments and plot them against recurrence interval on Gumble
probability paper.
The end result of these 6th steps is a flood frequency curve for an ungauged catchments.
ISSN : 0975-5462
5255
Akshay R. Thorvat et al. / International Journal of Engineering Science and Technology (IJEST)
Fig. 1.
6.
Location Map of Study Area Showing CWC Drainage Basins and River Network
Preliminary Statistics
Huvinhedgi
Galgali
Kurundwad
Arjunwad
Karad
Krishna
Krishna
Krishna
Krishna
Krishna
55150
22560
15190
12660
5462
5955.613
6134.493
4257.826
3563.538
2555.545
2228.376
2556.943
1342.008
1677.299
1529.461
Coefficient
of
Variation
0.374
0.417
0.315
0.421
0.599
Yadgir
Takli
Narsingpur
Dhond
Boriomerga
Wadakbal
Kokangaon
Shirdhon
Sarati
Jewangi
Cholachguda
Bagalkot
Gokak Falls
Daddi
Gotur
Bastewade
Terwad
Warunji
Koyna Nagar
Bhima
Bhima
Bhima
Bhima
Borinala
Sina
Bornala
Doddahalla
Nira
Kagna
Malaprabha
Ghataprabha
Ghataprabha
Ghataprabha
Hiranyakeshi
Vedganga
Panchganga
Koyna
Koyna
69863
33916
22856
11660
2640
12092
1640
630
7200
1920
9373
8610
2776
1150
1100
640
2425
1690
920
3681.1
2930. 618
2635.612
3960.32
370.809
724.55
111.025
39.857
972.5
422.94
1081.9
1097.62
1350.569
1082.15
746.213
604.159
1564.22
1547.382
989.28
2935.770
2565.100
2168.210
3505.076
492.591
748.690
100.420
40.860
724.480
357.960
569.890
897.790
1648.960
460.970
396.200
169.940
442.440
856.230
940.340
0.798
0.875
0.823
0.880
1.328
1.033
0.905
1.150
0.745
0.846
0.527
0.817
1.220
0.426
0.531
0.281
0.283
0.553
0.950
Station name
ISSN : 0975-5462
Stream
Catchment
Area km2
Mean
m3/sec
Standard
Deviation
Coefficient
of
Skewness
0.143
1.633
0.342
0.522
0.833
1.310
0.709
0.857
1.330
1.441
1.733
0.617
1.744
0.192
0.903
1.005
1.680
2.630
1.010
1.216
0.009
0.413
0.860
0.910
5256
Akshay R. Thorvat et al. / International Journal of Engineering Science and Technology (IJEST)
Table 2.
Station Name
Huvinhedgi
Galgali
Kurundwad
Arjunwad
Karad
Yadgir
Takli
Narsingpur
Dhond
Boriomerga
Wadakbal
Kokangaon
Shirdhon
Sarati
Jewangi
Cholachguda
Bagalkot
Gokak Falls
Daddi
Gotur
Bastewade
Terwad
Warunji
Koyna Nagar
Stream
Regression Equations
Krishna
Krishna
Krishna
Krishna
Krishna
Bhima
Bhima
Bhima
Bhima
Borinala
Sina
Bornala
Doddahalla
Nira
Kagna
Malaprabha
Ghataprabha
Ghataprabha
Ghataprabha
Hiranyakeshi
Vedganga
Panchganga
Koyna
Koyna
QT = 5023.71+1675.76 YT
QT = 5042.48+ 1980.34 YT
QT =3668.5+ 1059.75 YT
QT =2834.2 + 1311.51 YT
QT =1873.12 + 1226.31 YT
QT =2399.47+ 2304.57 YT
QT =1812.07 + 2011.38 YT
QT =1670.86 + 1734.83 YT
QT =2470.79+ 2678.5 YT
QT =172.91 + 355.86 YT
QT =415.45 + 555.83 YT
QT =66.6 + 79.3 YT
QT =21.25 + 33.45 YT
QT =665.14 + 552.72 YT
QT =264.2 + 285.45 YT
QT =826.12 + 459.94 YT
QT =1093.21 + 994.24 YT
QT =737.36 + 1102.68 YT
QT =882.85 + 358.39 YT
QT =573.04 + 311.40 YT
QT =537.58 + 119.72 YT
QT =1384.96 + 322.368 YT
QT =1172.27+ 674.53 YT
QT =598.33+ 703.0 YT
Table 3.
Regional FF Curve
Takali
Sr No.
Return
Period
1
2
3
4
5
6
7
ISSN : 0975-5462
5
10
20
50
100
200
500
Estimated
Actual
Parameter
4829.019
6338.414
7786.261
9660.351
11064.718
12463.961
14309.995
605.442
420.681
996.926
763.262
1372.447
1091.873
1858.521
1517.228
2222.764
1835.971
2585.679
2153.551
3064.475
2572.537
Sum
Average
Absolute Error
Site-1
Site-2
0.125
0.157
0.176
0.192
0.201
0.207
0.214
1.2738
0.1819
0.087
0.120
0.140
0.157
0.166
0.173
0.180.
1.0233
0.1460
5257
Akshay R. Thorvat et al. / International Journal of Engineering Science and Technology (IJEST)
Table 4.
Regional FF Curve
Gotur
Sr No.
Return
Period
1
2
3
4
5
6
7
5
10
20
50
100
200
500
Estimated
Actual
Parameter
1040.121
1273.804
1497.959
1788.104
2005.526
2222.156
2507.957
Table 5.
751.043
949.832
1140.515
1387.334
1572.291
1756.572
1999.697
Absolute Error
289.078
75.260
323.972
136.800
357.444
195.830
400.769
272.239
433.236
329.497
465.584
386.497
508.260
461.811
Sum
Average
Site-1
Site-2
0.278
0.254
0.239
0.224
0.216
0.210
0.203
1.623
0.2312
0.072
0.107
0.131
0.152
0.164
0.174
0.184
0.985
0.1410
Station Name
Galgali
Arjunwad
Karad
Yadgir
Narsingpur
Dhond
Boriomerga
Wadakbal
Kokangaon
Shirdhon
Sarati
Jewangi
Cholachguda
Bagalkot
Gokak Falls
Daddi
Warunji
Koyna Nagar
Q2.33
Catchment
Area km2
22560
12660
5462
69863
22856
11660
2640
12092
1640
630
7200
1920
9373
8610
2776
1150
1690
920
ln Q2.33
ln A
6191.077
3594.876
2584.380
3736.121
2677.061
4024.320
379.309
737.831
112.594
40.651
985.718
429.761
1092.885
1669.869
1376.914
1090.716
1563.497
1006.070
Sum
Mean
Standard Deviation
Constants in Regression Equation
ISSN : 0975-5462
8.731
10.024
8.187
9.446
7.857
8.606
8.226
11.154
7.892
10.037
8.300
9.364
5.938
7.879
6.604
9.400
4.724
7.402
3.705
6.446
6.893
8.882
6.063
7.560
6.997
9.146
7.421
9.061
7.228
7.929
6.995
7.048
7.355
7.432
6.914
6.824
126.029
153.639
7.002
8.536
1.277
1.286
0.5037
2.7020
ln Q2.33 = 2.7020 + 0.5037 ln A
Q2.33 = 14.9095 A 0.5037
5258
Akshay R. Thorvat et al. / International Journal of Engineering Science and Technology (IJEST)
7.
Conclusion
Four sites namely Huvinhedgi, Kurundwad, Bastewade and Terwad are not in homogeneity area.
The Regional Flood Frequency curve for upper Krishna Basin is:
QT/Q2.33= 0.6970+ 0.5220 YT
The relationship between Q2.33 and the area for upper Krishna basin is:
Q2.33= 14.9095.A0.5037
8.
The absolute errors in flood prediction using atsite parameter & atsite mean are 14.10 % and 14.60 %
for Gotur and Takali respectively.
The absolute errors in flood prediction using regional parameter and regional mean are 23.12 % and
18.19 % for Gotur and Takali respectively.
Further research is necessary to improve regional mean values. The regional mean value largely
depends on catchment area. However Q2.33 dependent on other catchment characteristics such as
drainage density, average elevation of catchment, average slope of stream and catchment, average
annual rainfall, land use and land cover etc. should be tried for further improvements.
References
[1]
B. N. S. Chalam, M. Krishnaveni, M. Karmegam (1996), Correlation Analysis of Runoff with Geomorphic Parameters, Journal of
Applied Hydrology, Vol. IX, No.3, 4 1996, 24-31.
[2] Burn, D.H. (1990), Evaluation of regional flood frequency analysis with a region of influence approach, Water Resour. Res.,
26(10), 2257 2265.
[3] Central Water and Power Commission (1969), Estimation of Design Flood-Recommended Procedure.
[4] Cunnane, C. (1988), Methods and merits of regional flood frequency analysis. J. Hydrol., 100, 269 290.
[5] Dalrymple T. (1960), Flood frequency methods, U. S. Geol. Surv. Water supply pap, 1543A, U.S. Govt. Printing office,
Washington, D.C., 11 51.
[6] G. Bhaskaran, R. Jayakumar, J. Moses Edwin, K. Kumaraswamy, Identification of Influential Geomorphic Parameters in
Hydrological Modelling through Numerical Analysis, Journal of Applied Hydrology, Vol. XV, No. 1, 1-8.
[7] Hosking, J.R.M, Wallis, J.R. (1990), Regional Flood Frequency Analysis, IBM Research Report.
[8] Hosking, J.R.M, Wallis, J.R. (1993), Some statistics useful in regional frequency analysis Water Resour. Res., 29(2), 271 -281
[9] Lim, Y.H., Lye, L.M. (2003), Regional flood estimation for ungauged basins in Sarwak, Malaysia, Hydrological Sci. J., 48 (1), 7993
[10] National Institute of Hydrology, Roorkee, Flood Frequency Analysis Lecture notes.
[11] Seth, S.M., Kumar, R., Singh, R.D. (1995), Development of regional flood formula for Mahanadi sub zone, 3(d) NIH Technical
Report, TR (BR )- 134.
[12] Seth, S. M., Mishra, G.C., Venkatesh, B., Singh, R.D. (1997), Development of regional flood formula for Krishna basin, NIH
Report, CS (AR)-39.
ISSN : 0975-5462
5259