Sie sind auf Seite 1von 8

Hydrological Sciences-Bulletin-des

Sciences Hydrologiques, 23, 3, 9/1978

A simple version of Gumbel's method for flood estimation


G . A L - M A S H I D A N I , P A N D E , B . B . L A L and M . F A T T A H M U J D A Department of Civil Engineering University of Mosul, Mosul, Iraq Received 6 February 1978, revised 28 June 1978

Abstract. Gumbel's method has been simplified in such a manner that one can obtain the magnitude of a given return period flood without recourse to looking at a table and working out the value of the coefficient of variation of the given data. The results obtained by the simplified version are compared with those obtained from using both the original approach and those from Powell's modification of Gumbel's method.

Version simplifie de la mthode de Gumbel pour l'estimation des crues

Rsum. On a simplifi la mthode de Gumbel de sorte qu'on puisse obtenir la grandeur d'une crue de priode de retour donne sans avoir recours un tableau et sans calculer la valeur du coefficient de variation des donnes disponibles. On a compar les rsultants obtenus en utilisant la version simplifie ceux obtenus par la mthode originelle et ceux de la modification de Powell de la mthode de Gumbel.

INTRODUCTION

The different methods of flood estimation have been summarized most adequately by Wolf (1966). The applicability of frequency methods to the study of floods has been widely recognized by numerous researchers in the field. Yet many others tend to criticize them on the following grounds: (a) insufficient records on which to base any extrapolation; (b) ignorance of the statistical laws on which extrapolations may be based; (c) inhomogeneity of records (e.g. recorded flows may be the result of various incoherent climatological mechanisms); (d) climatological changes over long periods. Statisticians, however, agree that floods of small frequency are random variables and they argue that even the highest design floods are strictly random variables and should be treated as elements of statistics of extremes. Besides, to a practical engineer the interpolation and extrapolation of flood frequencies provide an easy answer on which to base their designs. For the purpose of flood estimation the Pearson type III, the Gumbel extreme value distribution, and lognormal distribution seem to have found a wider applicability than many other distributions. Kaczmarek (1957) shows that the Gumbel type distribution is more applicable for the Vistula River in Poland than the other types. The writers have similarly found that this distribution is the most suitable for the River Tigris at the Mosul Gauging Station (Mujda, 1978).
03O3-6936/78/0900-O373$O2.O0 1978 Blackwell Scientific Publications 373

374

G. Al-Mashidani, Pande, B. B. Lai and M. Fattah Mujda

Details of Gumbel's method normally described in the literature can be written as follows: QT = Q(l+KCv) (1) where QT = the probable discharge with a return period of T years Cv = coefficient of variation = a/O Q = mean flood K = frequency factor = (yT - yn)/an an = standard deviation of data yT= ~Mn (T/T- 1) yn, an = expected mean and standard deviations of reduced extremes to be found from Gumbel's table. Therefore, to use the method, one has to refer to a table of values to determine y n and an and one had to calculate the value of Cv for the given data. The use of Gumbel's method is rather time consuming. The design offices in most of the developing countries are still not equipped with computers and because of this there is a need to simplify the Gumbel technique.

M O D I F I C A T I O N OF GUMBEL'S METHOD

Possibly motivated by the idea of determining the value of K without referring to a table, Powell defined K as follows: K= (6)'A (0.5772 + lnln T/T - 1)
(2)

What is different from Gumbel is that the value of K now does not depend upon the number of years of record, but the difficulty of calculating the value of Cv remains as before. Possibly the value of K is but a very weak function of the number of years of record and a strong function of the return period. A graph of QT/Q versus Cv for Iraqi rivers show that QT/Q for all rivers is only a function of Cv (Fig. 1). This means

F I G . 1. QT/Q versus Cv for Iraqi rivers.

Version of Gumbel's method for flood estimation

375

that K is a constant, and does not depend upon the number of years of record. Therefore Powell's method is applicable for Iraqi rivers. The value of K for the Powell method (Nash & Shaw, 1966) is given in Table 1.
TABLE 1. Value of in Powell's method T 1 2 3 4 5 6 7 8 9 K _
-0.16 0.25 0.52 0.72 0.88 1.01 1.12 1.21

T 10 15 20 25 30 40 50 60 70

K
1.30 1.64 1.86 2.04 2.20 2.40 2.61 2.73 2.88

T 80 90 100 200 400 500 600 800


1000

K
2.94 3.07 3.14 3.68 4.08 4.30 4.52 4.76 4.94

Chow introduced another modification of Gumbel's method. According to him: T = fl + / J log 1 0 log 1 0 (7'/7'~l) (3) where a and b are parameters estimated by the method of moments from the observed data. In practice it means that the values of a and b may be determined by the least squares technique from the given records by taking the value of 0T for a corresponding value of T. The value of T is defined as T = (N + \)jm. Where N is number of years of record and m is the rank of a flood arranged in descending order of magnitude. Here, again one need not consult any table but the labour involved in determining the value of and b from the given data by the method of least squares is simply enormous.

D E T A I L S OF PROPOSED S I M P L I F I C A T I O N

As stated earlier the value of QT in Gumbel's method is:


QT = Q

, yr - yn l +c

(4)

Supposing we have a discharge Qm having a rank m, then utilization of Equation 4 leads to Qm = Q 1 + ym~yn
o C (5)

Rearranging Equations 4 and 5 and dividing the former by the latter we obtain

QT- Q Qm-Q

_yr-yn ym-yn

(6)

Values of yn, presented in Table 2, as given by Gumbel, show variations from 0.4952 for a record of 10 years, to 0.5745 for 1000 years. It can therefore be treated as constant and may be arbitrarily taken as 0.55, which is equivalent to a value of yn for a record of 50 years.

376

G. Al-Mashidani, Pande, B. B. Lai and M. Fattah Mujda TABLE 2. Value ofyn and an in Gumbel's method
N
(number of years) yn 0.4952 0.5128 0.5236 0.5309 0.5362 0.5403 0.5436 0.5463 0.5465 0.5504 0.5521

n
0.9497 1.0206 1.0620 1.0915 1.1124 1.1285 1.1413 1.1518 1.1607 1.1681 1.1747

N
65 70 75 80 85 90 95 100 200 500
1000

yn 0.5536 0.5548 0.5559 0.5569 0.5578 0.5589 0.5593 0.5600 0.5672 0.5724 0.5745

on 1.1803 1.1854 1.1898 1.1938 1.1973 1.2007 1.2038 1.2065 1.2359 1.2588 1.2685

10 15 20 25 30 35 40 45 50 55 60

Also Gumbel's method defines the return period TmdyT


T
N + l

as follows:

1 1

T=

and yT = - In In (7) '

m \T 1 Therefore, the above definitions yield N + l ( i 7 m = - l in1l n ( I \N+ \-ml Hence substitution of the above in Equation 6 results in QT-Q Qm~Q "I"* (^-0.55

(8) , , / N+l -lnlnl- ) 0.55 \./V+ 1 m) Equation 8 can be utilized for predicting discharges for higher return periods. Which value of Qm should be used in the foregoing equation however, should be given some consideration. If one chooses Qm as Qmax (with m = 1) the prediction by Equation 8 might be highly biased, since there is an uncertainty associated with the selection of max as a parameter. Statistically speaking the highest discharge in a record could possibly be an outlier, or it might even be in error, as such the dependence on this single extreme value should be avoided. It is therefore proposed to use the average of the three highest discharge values as a parameter. Thus Equation 8 can be rewritten as follows: lnlnlQT-Q

1 X

1 J,

j-0.55 \T-l' / N+ 1 ,

(9)

It can be shown (Schulz, 1973) that

^lnln(^JLJ)=lnT~(~
\T-l/ \2T

+ --L-+")
24T2 8TS '

(io)
K

'

Version ofGumbeVs method for flood estimation


O - ^ O O l r - l O ^ ' - H ' - f ,-*
<N

i-*

t-t

rsi

OJ

<N

<

<N

^-,

C4

t-H

i-H

<N

^
l O

r^ _
O ^ t ^

,_
O H H O

fs

,_< ( ^

-H

r~1

<M

( N I / " ) \ C O O O O ' ~ H

oo<Nr-oor--aNoooo

r o ^ H T j - t s CN t^- r~-

^ O C O ' O T - l t - H ^ } - t O C ~ -

S)

a
comicn(Nr-coo co^oocrj-c~-mmt^

\ r o n ^ f n ^ O H

O)

sz
^ ^

a
si
J
C3

II! s I I 5 e
5

i c IS S . 13 3

3 5

I l l ^<
<n ^f

>-

a!
*-<

S>

S .5; ;

378

G. Al-Mashidani, Pande, B. B. Lai and M. Fattah Mujda

and by neglecting the quantities in brackets on the right-hand side: In In Similarly In In T = lnT (11) (12)

'N+ 1 N +1 = ln KN+ \-m, m

Hence Equation 9 can be further simplified as QT-Q 1 3 _ 1 3 In T- 0.55 (N+\ m (13) 0.55

T 2 m- T I In

Discharges for various return periods estimated by Equations 1, 2 and 9 have been tabulated in Table 3; the data were obtained from the Ministry of Irrigation, Republic of Iraq (1976) and Varshney (1977). The relevant figures for the discharge predictions are given in Table 4.

TABLE 4. Salient data for the rivers under analysis


max (cumec) Q (cumec) 3774 6725 2919 1762 3603 766 1374 5550 a (cumec) 1660 3125 1782 820 1377 767 817 3840 N no. of years 45 43 41 27 44 31 45 37

Name of river

yn 0.5464 0.5448 0.5442 0.5330 0.5457 0.5370 0.5464 0.5418

n 1.1554 1.1455 1.1434 1.0998 1.149 1.115 1.155 1.133

7740 River Tigris at Mosul River Tigris at Fatha 16380 Greater Zab at Eski-Kelek 9710 Lesser Zab at Altun Kupri 3440 7460 Euphrates at Hit 3520 River Adhaim at the Narrows River Diyala at the discharge station 3340 River Yamuna at Tajewala (India) 15970

DIMENSIONLESS P L O T OF THE E X I S T I N G RECORD

Equation 9 can be re-written in the following form: Q-Q 1


J

J I 2
m=1

N +1 -lnlnl0.55 N+ 1 -m 1 J2, I N+\ l n l n (N+ _ _ 1 -m ,^0.55

(14)

The data for five Iraqi rivers have been plotted according to Equation 14 and this is shown in Fig. 2. All the river data should have fallen on a single line inclined at 45 to the axes. The data show some scatter, however, but the band in which all the data are plotted can be regarded as reasonable.

Version ofGumbel's method for flood estimation

379

-0-6

-0-4

-0-2.A

-0-6 1 -

FIG. 2. Data for some Iraqi rivers plotted according to Equation 14. X = right hand side of Equation 14.

DISCUSSION OF RESULTS

A study of Table 3 indicates that for five out of the seven Iraqi rivers, the prediction by the present method compares favourably with those of Gumbel and Powell. For two rivers, however, the Greater Zab at Eski-Kelek and the Adhaim at the Narrows, the method gives values approximately 10 per cent higher than those of Gumbel. In both these rivers a high flood has already occurred. For the Greater Zab a flood of 9710 m3/s has occurred in a record of 41 years, whereas Qioo predicted by GumbePs method is only 9240 m3/s. For the River Adhaim at the Narrows a maximum flood of 3520 m3/s has occurred in a record of 31 years whereas (2ioo by Gumbel has a value of 3562 m3/s, which is practically the same as the recorded flood. Powell's method gives a value of Q100 which is lower than 3520 m3/s. The limitations of Gumbel's and Powell's methods are that the predicted values of Ql00 can be lower than the value of
3

Gmax that has already occurred in the limited period of record. By treating 3

2
m =1

Qm as a parameter for the prediction discharge for higher return floods, the predicted values are not likely to be lower than Qmax. Hence, predictions by the present method can be considered to be on the safe side.

380

G. Al-Mashidani, Pande, B. B. Lai and M. Fattah Mujda

CONCLUSIONS 1. Powell's method consistently gives lower values than Gumbel's method. 2. Gumbel's method can be simplified as in Equations 9 and 13, and using these equations QT can be predicted without determining the value of Cv and looking up Gumbel's table. 3. The discharge data can be plotted according to Equation 14 along a 45 line which shows that the suggested simplification is reasonable. 4. The present method takes into account Q and
3
3

m = l

Um

as the two parameters for discharge prediction. The former takes all the recorded data into account and the latter gives a special weight to a few values of successively higher discharges available in the records, as such extrapolation can be better relied upon. 5. The simplified approach compares favourably with those of Gumbel and Powell.

ACKNOWLEDGEMENT The authors are grateful to the Associate Editors of the Bulletin who reviewed the original manuscript for their valuable suggestions.

REFERENCES Kaczmarek, Z. (1957) Efficiency of the estimation of floods with a given return period. In General Assembly of Toronto, vol. Ill-Surface Waters, Prevision, Evaporation, pp. 144-159. IAHS Publ. no. 45. Ministry of Irrigation, Republic of Iraq. Discharges for selected gaging stations in Iraq (1976) (a) 1959-1975: D.G. of Irrigation, Baghdad, Iraq, Aug. 1976. (b) 1931-1958: M/S Harza Engineering Company and Binnie Deacon & Gourley. Mujda, M.F. (1978) Flood Frequency Analysis of Tigris River at Mosul. (M.Sc. thesis in preparation), Department of Civil Engineering, Mosul University, Iraq. Nash, J.E. & Shaw, B.L. (1966) Flood Frequency as a Function of Catchment Characteristics. Paper 6, Session C, Proceedings of the Symposium organized by the Institution of Civil Engineers on River Flood Hydrology, London. Schulz, E.F. (1973) Problems in Applied Hydrology, Section 9. Water Resources Publications, Fort Collins, Colorado, U.S.A. Varshney, R.S. (1977) Engineering Hydrology. Nem Chand and Brothers, India. Wolf, P.O. (1966) Comparison of Methods of Flood Estimation. Paper 1, Session A, Proceedings of the Symposium organized by the Institution of Civil Engineers on River Flood Hydrology, London, pp. 1-23.

Das könnte Ihnen auch gefallen