Climate Change and Water Resources in South Asia - Chapter 3 ppt

3
Are Floods Getting Worse in the Ganges,
Brahmaputra and Meghna Basins?
3.1 INTRODUCTION
The Ganges, Brahmaputra and Meghna/Barak (GBM) river systems occupy about
175 million hectares (mha) of South Asia (Fig. 3.1) and supports more than 500 million
people (Verghese and Iyer, 1993). They are unique in the world with respect to water and
sediment supplies, channel processes, and instability. While the Brahmaputra ranks fourth
among the largest rivers of the world with regard to mean annual discharge, the Ganges
ranks thirteenth (Mirza, 1997). The estimated annual sediment yield of the Brahmaputra is
1,028 tons/km
2
, the highest among the world’s largest rivers. On the other hand, the
sediment yield of the Ganges is only 502 tons/km
2
although its basin area is two times that
of the Brahmaputra (Barua, 1994). The swinging and avulsion of the courses of the Ganges
and Brahmaputra Rivers in recent history have significant influence on the morphology of
their alluvial floodplains (Rahman, 1993; Brammer, 1996). They are characterized by high
flows during the monsoon and low flows during the dry season. For example, the ratio of
monsoon flow to dry season flow of the Ganges River at Hardinge Bridge in Bangladesh is
6:1 (Mirza and Dixit, 1997). The high flows often cause floods in many parts of these vast
river basins.
Sitting at the confluence of the three major rivers, Bangladesh (area 148,000 sq. km)
is considered to be the most flood-affected country in the world followed by India. Every
year, slightly over one-fifth of its land area becomes flooded and in extreme cases, more
than two-thirds of the country is affected. In upstream India (area 3,280,000 sq. km),
floods annually inundate an area larger than half of Bangladesh.
Available information shows that in recent years, flood damage in Nepal, India and
Bangladesh is increasing. Substantial increases in flood damage in Nepal during the 1980s
were reported by the Asia Development Bank (ADB, 1991). For India, the Center for

Science and Environment (CSE, 1992) reported that the annual flood damage had
increased 40 times from the 1950s to the 1980s (Fig. 3.2). According to Mirza (1991a),
compared to the 1960s and 1970s, flood damage in Bangladesh was the greatest in the
1980s (Fig. 3.3). These increases have largely been attributed to worsening flood events
(increased river discharge and spatial extent) in the GBM basins in India and Bangladesh
M. MONIRUL QADER MIRZA
R. A. WARRICK
N. J. ERICKSEN
G. J. KENNY
Reprinted with permission from Environmental Hazards 3 (2001), pp.37-48.
Copyright © 2005 Taylor & Francis Group plc, London, UK
(CSE, 1992; BBJTO, 1989; RBA, 1980; and Ives, 1991). These claims, do not, however,
appear to be based on systematic analyses of relevant data. Therefore, this paper examines
whether floods in the GBM basins are getting worse by applying statistical tests to:
1) the peak discharge data of the three rivers recorded at various stations in India, Nepal
and Bangladesh; and 2) the flooded area data. The latter were used to determine changes
in spatial severity of flooding in India and Bangladesh. Peak discharge recording stations
and period of records are shown in Table 3.1. It is possible that the reported increases in
flood problems are due to increased human activities in flood-prone areas. But that
element of flood hazard is not the subject of this paper.
Fig. 3.1 The Ganges, Brahmaputra and Meghna basins. Location of some discharge measurement
stations have also been shown.
Fig. 3.2 Flood damage in India during 1953-1999 (Source: CWC, 1989; ADRC, 2000a).
0
10000
20000
30000
40000
50000
1950 1960 1970 1980 1990 2000

Damage (in million rupees)
Year
56 ARE FLOODS GETTING WORSE?
Copyright © 2005 Taylor & Francis Group plc, London, UK
Fig. 3.3 Flood damage in Bangladesh during 1954-1998 (Source: Mirza, 1991a; ADRC, 2000a).
3.2 HYDRO-METEOROLOGY OF THE GBM BASINS
Of the three river basins, the Ganges is the largest. Its 109.5 mha basin area is distributed
over China, Nepal, India and Bangladesh. The Ganges River rises South of the main
Himalayan divide near Gangotri at a height of 4,500 m in the Uttar Pradesh (UP), India. In
Nepal, India and Bangladesh, mean annual precipitation in the basin is 1,860 mm, 908 mm
and 1,568 mm, respectively. Mean annual runoff of the Ganges River at Farakka, India and
Hardinge Bridge, Bangladesh, is estimated to be 415 x 10
3
million cubic meters (mcm)
and 352 x 10
3
mcm, respectively (Mirza, 1997). The highest annual peak discharge
(80,230 m
3
sec
-1
) was recorded at Hardinge Bridge in 1998 (See Fig. 3.1 for the location of
some of the stations referred to in this paper.)
The Brahmaputra basin area is 58 mha. It is regarded as one of the world’s largest
braided river systems in terms of discharge, sediment transport, and channel processes
(JMBA, 1989). The river originates at an elevation of 5,150 m in a large glacier mass in the
Kailash range of the Himalayas, very close to Manassarovar Lake. Mean annual
precipitation in the basin area in India and Bangladesh is 2,500 mm and 2,400 mm,
respectively. Mean annual runoff of the Brahmaputra at Pandu, India and Bahadurabad,
Bangladesh is estimated to be 511 x 10

3
mcm and 643 x 10
3
mcm, respectively. The highest
peak discharge was 98,600 m
3
sec
-1
recorded at Bahadurabad in 1988.
The Meghna/Barak basin is the smallest of the three basins, with an area of 8 mha. The
headstream of the river in India is known as Barak and originates on the Southern slope of
the mountain range to the North of Manipur, India. In Bangladesh, the river is known as
Meghna and flows Southwest to meet the Padma (combined flow of the Ganges and
Brahmaputra Rivers) at Chandpur. Mean annual precipitation of the basin in India and
Bangladesh is 2,640 mm and 3,574 mm, respectively. Mean annual runoff of the Barak in
India is 41 x 10
3
mcm (measured at Badarpurghat) (Kothyari and Garde, 1991). At Bhairab
Bazaar in Bangladesh, the mean annual runoff is estimated to be 151 x 10
3
mcm. The
highest peak discharge at Bhairab Bazaar was 19,900 m
3
sec
-1
recorded in 1993.
M. M. Q. MIRZA ET AL. 57
Copyright © 2005 Taylor & Francis Group plc, London, UK
Table 3.1 Statistical properties of the peak discharge and flooded area data
* non-random.

River Station Period of
Record
Latitude
(deg. N)
Longitude
(deg. E)
Mean
(m
3
sec
-1
)
Coefficient of
Variation
(CV)
Lag-1
Autocorrelation
Coefficient

Hardwar 1885-1971 29.58 78.10 6,639.00 0.47 - 0.42*
The Ganges Farakka 1949-1980 25.00 87.91 56,516.00 0.17 + 0.05
Hardinge
Bridge
1934-2000 23.06 89.03 51,184.00 0.18 + 0.22
The Kosi Barahkshetra 1948-1978 - - 10,190.00 0.44 - 0.18
The Brahmaputra Pandu 1955-1974 26.20 91.50 50,524.00 0.20 + 0.20
Bahadurabad 1956-1999 25.15 89.66 67,389.00 0.18 +0.03
The Meghna Bhairab Bazaar 1964-1998 24.03 59.98 14,072.00 0.19 - 0.23
The Surma-Meghna Kanairghat 1969-1993 - - 2,224.00 0.16 + 0.35

Flooded Area (mha)
India
Bangladesh




1953-1997
1954-1999

7.28
3.03

0.47
0.68

+0.17
+0.16

58 ARE FLOODS GETTING WORSE?
Copyright © 2005 Taylor & Francis Group plc, London, UK
3.3 THE FLOOD PROBLEM
Flooding of catastrophic proportions often occurs in the GBM river basins. Extreme
precipitation in the monsoon, together with the physical settings of the river basins has
caused many severe floods in the last few decades. Causes and characteristics of floods
vary between the highlands in Nepal, the middle ground in India, and the flat deltaic terrain
in Bangladesh.
In Nepal, the flood problem is mainly restricted to the Terai region along the border it
shares with India. Rivers in the Terai region are very unpredictable and cause heavy flood
damage as a result of intensive downpours on the Southern slopes of the Siwalik Himalaya

(SAARC, 1992). The high Himalayan Mountains of Nepal are affected by Glacier Lake
Outburst Floods (GLOF). These floods do not cause much damage to human settlements
because the upper mountainous areas are sparsely populated. In contrast, floods in the
Terai occur regularly and cause considerable damage to densely populated floodplains.
In India, floods in the Ganges region are caused by the following factors either singly
or in combination: excessive precipitation, inadequate river channel capacity, obstruction
in streams, inadequate waterways at confluences, human encroachments and lack of
adequate drainage, failure of flood control embankments and deforestation (Rangachari,
1993; Chowdhury, 1989; Dhar and Nandargi, 1998). Similar factors cause floods in the
Brahmaputra region, but they are compounded by local physiographic features. The
region is interspersed with a large number of streams, flooding from which inundates
the intervening narrow valleys. The riverbeds in some cases are higher than the surround-
ing valley land. Therefore, any breach or spilling causes deep flooding in the valleys. The
Brahmaputra region in India is highly prone to earthquakes and this often causes
landslides. These seriously disturb the drainage system. The Barak region lies between the
Khasi and Jaintia Hills in the North and Mizo Hills in the South. The river often overflows
its banks inundating low areas on either side. There is a series of bowl-shaped depressions,
locally known as “Haors”, which fill with floodwater. The gradient of the river is extremely
flat and the outfall at the border with Bangladesh is congested.
In recent years, the role of deforestation in the upstream areas in causing flooding in
the downstream areas of the GBM basins has triggered interesting debates (BWDB, 1987;
Carson, 1985; Thompson and Warburton, 1985; Hamilton, 1987; Hofer, 1998; Ives and
Messerli, 1989; Messerli and Hofer, 1995; Rogers et al., 1989). BWDB (1987) indicated
that deforestation in the upstream contributed significantly to the increased rates of
sediment supply and accretion. However, the existing publications do not report any
significant recent increase in the sediment load of the larger rivers and their tributaries, or
in the magnitude of annual flooding and levels of river discharge (Ives and Messerli, 1989).
Thompson and Warburton (1985) questioned the linkage between massive floods in the
plains and land-use activities upstream in the Himalayas. However, they noted that there
was some technical uncertainty encountered when analyzing the human components of

erosion, flooding and shifting hydrological patterns. Hofer (1998) concluded that land-use
changes in the Himalayas were not responsible for the floods far downstream in India
and Bangladesh. In the aftermath of the devastating floods in Bangladesh in 1988,
Rogers et al. (1989) remarked that there were no grounds for considering deforestation in
the Himalayas as a significant cause of the flooding in the delta of the river system. Carson
(1985:36) mentioned, “…Flooding and sedimentation problems in India and Bangladesh
are a result of the geomorphic character of the rivers and man’s attempts to contain the
rivers. Deforestation likely plays a minor, if any, role in the major monsoon flood events on
the lower Ganges.” The role of deforestation in the sedimentation and flooding processes
M. M. Q. MIRZA ET AL. 59
Copyright © 2005 Taylor & Francis Group plc, London, UK
in South Asia is a highly contentious issue and it needs adequate scientific research and
attention.
Of the three countries affected by flooding in the GBM river basins, Bangladesh is the
most vulnerable because of its geographic location, high monsoon cross-border flow, and
the physiographic characteristics of its deltaic floodplains. Half of the country is under
12.5 m above the mean sea level (CBJET, 1991). Because of its flatness, floodwaters
cannot drain quickly. The three rivers together may generate as much as 200,000 m
3
sec
-1
of
peak discharge. The problem becomes more complicated when the peak flow of each of
the three rivers synchronises. In such a case, the flooded area may exceed 60% of the
country (as occurred in 1988 and 1998), about three times the normally flooded area
(Fig. 3.4). Although the river levels fall rapidly from September through November, water
levels on adjoining floodplains fall more slowly because of low gradients, congested
drainage, and substantial depression areas. The latter may stay submerged until December
to January and some throughout the whole dry season (November to May).
Floods cause considerable damage in the GBM basins and four main economic

sectors - agriculture, housing, industry and transportation infrastructure are particularly
vulnerable. Flood related damage puts considerable strain on the economies of the
countries that share the GBM basins. This is particularly true in terms of diversion of
resources for recovery activities and the loss in growth of Gross Domestic Products (GDP).
For example, during the 1998/1999 fiscal year in Bangladesh, GDP growth declined to
4.6% from 5.2% of the previous year due to the devastating floods of 1998. Industry
sector growth, however, decreased by 3.4% during the same period as a result of
flood-induced disruptions in the manufacturing subsector (ADB, 2000).
0
20
40
60
80
100
120
1954
1961
1965
1969
1973
1977
1981
1985
1989
1993
1999
Ye ar
Flooded Area (000 sq.km)
Fig. 3.4 Flooded area in Bangladesh during 1954-1999 (Source: BWDB, 2000b).
In Nepal, government statistics show an increasing trend in damage to public and

private property from floods and landslides in recent years. The estimated damage to
property increased from US$ 1.0 million in 1983 to US$ 100.0 million in 1989 (ADB,
1991).
In India, out of the 34 mha of “flood-prone” area, some 23 mha are in the GBM
basins. Fifteen Indian states and the union territory of Delhi lie in the basin. However, four
states alone account for over two-thirds of the flood-prone area: Uttar Pradesh, Bihar,
West Bengal and Assam (Rangachari, 1993). Data compiled by India’s Central Water
Commission (CWC) show that during 1953-1987, the average area affected by floods was
60 ARE FLOODS GETTING WORSE?
Copyright © 2005 Taylor & Francis Group plc, London, UK
7.66 mha (22.5% of the flood-prone area), of which 3.51 mha were cropped (CWC, 1989).
Recent data published by the Ministry of Water Resources, Government of India shows
that during 1953-1997 annual average flood affected area has declined to 7.42 mha (MWR,
2000) from that of the 1953-1987 average (Fig. 3.5). This is due to a decline in flooded area
in the period 1987-1997. Available evidence indicates increasing flood damage in recent
years in India. State governments estimated that flood damage in 1987 and 1988 was US$
1.5 billion and US$ 2.5 billion, respectively (Rangachari, 1993). In 1953 it was 524 million
rupees and remained around that level until the middle of the 1960’s when damage tracked
upward (Fig. 3.2) (CWC, 1989).
Fig. 3.5 Flooded area in India during 1953-1997 (Source: MWR, 2000).
In Bangladesh, the area prone to floods in the GBM basin is 6.14 mha. This is 42% of
the country’s geographical area. On an annual average, 20.5% of Bangladesh (3.03 mha)
becomes inundated. The loss caused by floods in Bangladesh in a normal year is about
US$ 175 million; but in extreme cases, the damage may exceed two billion dollars. The
1998 flood damage was the worst in history, totaling in the range of US$ 2 billion to
2.8 billion (ADRC, 2000a; MOFA, 1998). The industry and infrastructure sectors were
worst hit, followed by agriculture (MDMR, 1998). The flood damage in Bangladesh for
the period 1954-1998 is shown in Figure 3.3.
Flood damage estimation methods in Nepal, India and Bangladesh only take into
account the direct damage. Death, trauma, accidents, post-flood health and nutrition

problems are not considered direct damage as their monetary valuations are unaccounted
for. Almost every year, a significant number of people die due to floods. During
1953-1987, the annual average loss of human lives in India due to floods was 1,439
(CSE, 1992). In West Bengal, India 1,262 people had died and another 117 were reported
missing during the devastating monsoon floods of 2000 (UNICEF, 2000). In Nepal, during
the period of 1981-1999, a total of 5,453 people lost their lives with the highest, 2,307
people, in 1993 (ADRC, 2000a). In the catastrophic floods of 1998 in Bangladesh, the
number of reported deaths was 1,050 (ADRC, 2000a). The number of deaths caused by
floods in India, Bangladesh and Nepal is summarized in Table 3.2.
Flood damage is an indicator of flood hazard, which in turn, is a function of potential
flood events in relation to human use of flood-prone land. This includes activities aimed at
alleviating the flood problem, such as embanking river channels and elevating floor levels
of buildings. Thus, flood hazard effects (registered as property damage, social disruption,
and human injury) rise or fall with changes in the parameters of the flood event
0
25
50
75
100
125
150
175
200
1953
1957
1961
1965
1969
1973
1977

1981
1985
1989
1993
1997
Ye ar
Flooded Area (000 sq.km)
M. M. Q. MIRZA ET AL. 61
Copyright © 2005 Taylor & Francis Group plc, London, UK
(e.g., discharge and areal extent) or human use of flood-prone land (e.g., type and density),
or both.
In the wake of increasing flood damages in the GBM basins, special emphasis has
been given to flood problems in both India and Bangladesh. The Government of India
created the Rashtriya Barh Ayog (National Commission on Floods) in 1976. Devastating
floods in India in 1987 led to the setting up of two committees to look into the problem
(Rangachari, 1993; Mirza, 1991a). Bangladesh formulated a Water Master Plan in 1964
that recommended 59 flood control projects as a result of the consecutive floods of 1954
and 1955 (Mirza, 1991b). However, from the mid-seventies to the late-eighties, flood
control received little attention in Bangladesh. In response to the devastating flood of
1988, Bangladesh carried out 28 studies under the “Flood Action Plan (FAP)” during the
period 1989-1995. All of these efforts were based on the claim by government (as well as
non-government agencies) that floods in the GBM basin areas were getting worse
(CSE, 1992; BBJTO, 1989; RBA, 1980; Ives, 1991).
As floods are generally accompanied by over bank spilling, assessment of peak
discharges in the major rivers is the best way to determine whether changes in flood events
have occurred or not. In order to detect changes, Cumulative Deviation, Worsley
Likelihood Ratio, Kruskal-Wallis and Mann-Whitney U tests (Annex 3.1) were applied to
the peak discharge series of the three rivers. Similar tests were also applied to the flooded
areas in India and Bangladesh to see if there were associated changes in the spatial (areal)
extent of flooding.

Table 3.2 Number of deaths due to floods in India, Bangladesh and Nepal during the period
1953-2000
Year India Bangladesh Nepal

1953 37
1954 279 112 60
1955 865 129
1956 462
1957 352
1958 389
1959 619
1960 510
1961 1,374
1962 348 117
1963 432 30
1964 690
1965 79
1966 180 39
1967 355
1968 3,497 221 276
1969 1,408
1970 1,076 87 350
1971 994 120
1972 544 50
1973 1,349 427
1974 387 1,987
1975 686


62 ARE FLOODS GETTING WORSE?

Copyright © 2005 Taylor & Francis Group plc, London, UK
Table 3.2 Continued
Source: India: 1953-1987 (CWC, 1989) and 1988-1999 (ADRC, 2000a). Death toll for 2000 was
taken from (ADRC, 2000b; UNICEF, 2000). Bangladesh: ADRC, 2000a except for 1954, 1955,
1962, 1968, 1970 and 1974 (Islam, 2000). Nepal: until 1999 (ADRC, 2000a.). Figure for 2000 was
taken from ADPC, 2000.
3.4 THE DATA
Annual peak discharge data for the Ganges, Brahmaputra and Meghna Rivers were
collected from UNESCO (1976), IAHS-AISH (1984), Bangladesh Water Development
Board (BWDB) (1995, 2000a), French Engineering Consortium (FEC) (1989a), Raghunath
(1985) and Nepal Water Conservation Foundation (NWCF, 1996). Flooded area data were
collected from the Central Water Commission (CWC, 1985), Ministry of Water Resources
(MWR, 2000) and BWDB (1993, 2000b). The periods of records of the collected peak
discharge data for various rivers and stations varied, as shown in Table 3.1, but fall within
the period 1885 to 2000. For India, flood damage data were collected from the Central
Water Commission (CWC, 1989) and Asian Disaster Reduction Center (ADRC, 2000a).
Flood damage data for Bangladesh were taken from Mirza (1991a) and Asian Disaster
Reduction Center (ADRC, 2000a).
Two observations were missing in the peak discharge data series for the Ganges River.
One observation was missing for each of the Farakka (1969) and the Hardinge Bridge
(1971) sites. These were filled by determining the correlation coefficient and then applying
Year India Bangladesh Nepal

1976 1,373 103
1977 11,316 13
1978 3,396 17 130
1979 3,637
1980 1,913 655
1981 1,376 750
1982 1,573 92

1983 238 245 186
1984 1,661 1,200 200
1985 1,804 300 46
1986 1,200 150 22
1987 1,835 3,680 358
1988 2,050 2,379 27
1989 1,097 180
1990 203 231 25
1991 1,024 450 51
1992 572 15
1993 1,862 366 2,307
1994 2,845 43
1995 1,479 900 140
1996 1,506 55 768
1997 2,526 179
1998 2,131 1,050 311
1999 500 48 170
2000 2,159 100 144


M. M. Q. MIRZA ET AL. 63
Copyright © 2005 Taylor & Francis Group plc, London, UK
the method of Salinger (1980).
1
Similarly, one observation was also found missing for
each of the Pandu (1964) and the Bahadurabad (1971) sites for the Brahmaputra River.
These were also filled by the method applied for the missing data of the Ganges River. For
the Meghna River at Bhairab Bazaar, four observations (1977-1980) were missing. These
missing observations were filled by applying the precipitation-peak discharge regression
model (Mirza, 1997).

2
Errors involved with the discharge measurement, processing and
storage of data are not generally reported and documented. Standard equipment, methods
and specifications are being used in discharge and water level measurements in Nepal,
India and Bangladesh. However, in Bangladesh, due to changing bed forms, velocity
measurements from non-anchored boats and inaccurate measurement of depths for current
meter may cause ≤10% and ≤15%-20% uncertainty in discharge and water level
measurements, respectively (Sir William Halcrow and Partners, 1991; FAP 24, 1993). The
magnitudes of errors in the measurement of discharge and water levels in India and Nepal
are not known. But, due to similar characteristics of river channels, they are assumed to be
the same as those of Bangladesh.
Statistical properties of the peak discharge data are shown in Table 3.1. Annual peak
discharge of the Ganges River at Hardwar is found to be highly variable, followed by the
Kosi River at Barahkshetra. The almost equal coefficients of variation of the Brahmaputra,
Meghna and Surma-Meghna Rivers indicate that they drain the catchment areas with
similar characteristics. Lag-1 autocorrelation coefficient was determined using the
equation shown below. This coefficient is used to determine the presence of “persistence”
in the data. A negative value of r
1
is indicative of marked high frequency (i.e. short-period)
oscillations. On the other hand, positive values indicate Markov linear type persistence
(Mirza et al., 1998). The presence of this type of persistence in a peak discharge series
means that a large (or small) peak discharge for one year is more likely to be followed by
a large (or small) for the next year:
where
X
i
is annual peak discharge at year i, n is the sample size, and is mean peak
discharge.
The randomness of the series can be tested to identify presence of trend or cycle using

the one-tail 95% confidence limit of the Gaussian distribution (Mitchell et al., 1966). The
1
The missing observation for one year was calculated using the ratio of the mean peak discharge of the
two stations with a missing record to the adjacent data-possessing station multiplied by the peak
discharge of that year.
2
The regression model for estimating annual peak discharge is Q
p
= -10531 + 3.41*P1 + 5.69* P2
(R
2
= 87%). Where P1 is average precipitation in the North Assam meteorological sub-division and
P2 is the average precipitation in the South Assam meteorological sub-division and Bangladesh part
of the basin.
test value
()
r
t
1
is computed from:
64 ARE FLOODS GETTING WORSE?
Copyright © 2005 Taylor & Francis Group plc, London, UK
If the sample value of
r
1
is larger than the value of
()
r
t
1

, then the series is considered
3
The barrage was built by India across the Ganges River at Farakka (18 km from the Bangladesh
border) in order to divert 1,133 m
3
sec
-1
of water to restore the navigability of the port of Kolkata. It
was commissioned in April of 1975.
3.5 STATISTICAL ANALYSES METHODS
The Cumulative Deviation (Buishand, 1982), Worsley Likelihood Ratio (Worsley, 1979),
Kruskal-Wallis (Coshall, 1989) and Mann-Whitney U (Kite, 1988; Essenwanger, 1985)
tests (Annex 3.1) were applied to detect changes in the mean of the peak discharge and
flooded areas. Details of these tests are described below.
For

individual station and the flooded areas, the entire time series was considered for
the Cumulative Deviation and the Worsley Likelihood Ratio tests. The available data were
divided into 10-year periods for the Kruskal-Wallis test. For the Mann-Whitney U test,
data were divided into two equal segments. The test statistics of the Mann-Whitney and
Kruskal-Wallis tests were calculated based on the ranks of the observations, not their
actual values. The cumulative deviation and Worsley Likelihood Ratio tests assume that
the observations are random and normally distributed. However, the tests can be applied
with some reasonable departures from the normality. The Lag-1 autocorrelation values
were within the desired range (except the Ganges River at Hardwar). They indicated that
the observations of peak discharge at most of the stations were approximately random and
normally distributed.
For all four tests, critical values for a two-sided probability were used. The null
hypothesis was rejected only if a change was detected at a 95% confidence level. Results
of the statistical tests are presented in Table 3.3.

3.6 RESULTS
All four statistical tests indicate a downward trend in the peak discharge of the Ganges
River at Hardwar in India. Note that this is a discharge measurement station located very
far upstream. The Kruskal-Wallis test detected increases at Farakka, 18 km upstream of
the Bangladesh border. Three statistical tests (Cumulative Deviation, Worsley Likelihood
Ratio and Kruskal-Wallis) indicated increases in the Ganges River flood peak discharge at
Hardinge Bridge, in Bangladesh (Table 3.3). As the discharge of the Ganges River at
Farakka has been regulated by a barrage
3
since 1975, these changes might have not
occurred entirely as a result of natural phenomena. The Kosi River (an important tributary
of the Ganges) in Nepal showed no change in peak discharge.
Similarly, all four statistical tests demonstrated decreases in the peak discharge of the
Brahmaputra River at Pandu in India. However, there was no change indicated at
Bahadurabad further downstream. In the Surma-Meghna at Kanairghat, the statistical tests
indicated increases in the peak discharge. Downstream at Bhairab Bazaar, there was no
change in peak discharge of the Meghna River.
M. M. Q. MIRZA ET AL. 65
non-random.
Copyright © 2005 Taylor & Francis Group plc, London, UK
Table 3.3 Results of the statistical tests applied to the peak discharge data of the Ganges, Brahmaputra and Meghna Rivers and the flooded
areas in India and Bangladesh
Note: The shaded rows denote locations within the border of Bangladesh.

Peak
Discharge/Flooded
Areas

Station/Basin

Statistical Tests
Worsley
Cumulative Likelihood Kruskal-Wallis Mann-Whitney
U Deviation Ratio


a
. Peak Discharge
Hardwar Change (-) Change (-) Change (-) Change (-)
The Ganges Farakka No Change No Change Change (+) No Change
Hardinge Bridge Change (+) Change (+) Change (+) No Change
The Kosi Barahkshetra No Change No Change No Change No Change
Pandu Change (-) Change (-) Change (-) Change (-)
The Bhahmaputra Bahadurabad No Change No Change No Change No Change
The Surma-Meghna Kanairghat Change (+) Change (+) Change (+) Change (+)
The Meghna
Bhairab Bazaar No Change No Change No Change No Change
b.
Flooded Area

India
The Ganges
The Brahmaputra
The Meghna
Change (+)
No Change
No Change
No Change
No Change
Change (+)

Change (+)
No Change
No Change
Change (+)
No Change
No Change
Bangladesh
- Change (-) No Change No Change Change (-)


66 ARE FLOODS GETTING WORSE?
Copyright © 2005 Taylor & Francis Group plc, London, UK
With regard to flooded areas, three statistical tests - Cumulative Deviation,
Kruskal-Wallis and Mann-Whitney U showed detectable increases in flooded areas in the
Ganges basin. No change was detected in the Brahmaputra basin. Only the Worsley
Likelihood Ratio test identified an increase in the Meghna basin in India. In Bangladesh,
Cumulative Deviation and the Mann-Whitney U tests indicated decreases in the flooded
areas in Bangladesh.
Overall, the results do not indicate any conclusive change in the peak discharge or
flooded area time series within Bangladesh (Table 3.3). However, at the upstream stations
in India two rivers showed an increase in peak discharge. This was not registered at
downstream stations in Bangladesh (except at Hardinge Bridge where three of the four
tests showed increases). As mentioned above, however, these increases are likely to be due
to the regulation of discharge by the Farakka Barrage.
Are changes in climate responsible for increases in flood peaks and flooded areas in
the upstream areas of the Ganges, Brahmaputra and Meghna Rivers? Firm evidence of a
long-term regional trend in area-averaged precipitation is yet to be found. Mooley and
Parthasarathy (1983) examined above- and below-average annual precipitation extremes
between 1871 and 1980 for 360 precipitation stations all over India, except for the
Northern mountainous districts. They included the Gangetic Plain, Bengal and Assam in

their analysis. No statistically significant trends or oscillations were found. Their
conclusion was that annual precipitation totals were distributed randomly overtime.
During 1988-1997, India experienced nine abnormal monsoons; six were higher than
normal and three were 1% to 10% less than a normal (Dhar and Nandargi, 1998). There is
no indication of any trend towards increased monsoon precipitation in the upstream river
basins that could have exacerbated flooding in either upstream areas or in Bangladesh in
downstream.
3.7 DISCUSSION
The results should be considered with caution, since the period of analysis from the data
available at the eight river gauging stations ranges from 20 years to 87 years. Nevertheless,
claims that worsening flood problems in the GBM basins are due to increasing flood events
are not borne out by this analysis. The results do not indicate any conclusive change in the
peak discharge or flooded area overtime.
Factors other than increasing flood event characteristics (peak discharge and flooded
area) must be responsible for the increase in flood damage (Figs. 4.2 and 4.3) recorded for
countries within the GBM basins. Two main factors have been identified as possible
contributors to the record of increasing flood damage: (i) improvement in flood damage
assessment techniques; and (ii) increases in human settlement in flood-prone areas.
3.7.1 IMPROVEMENT IN FLOOD DAMAGE ASSESSMENT TECHNIQUES
Flood damage assessment techniques have undergone a series of improvements in the
last five decades, particularly in India and Bangladesh. In the past, the Ministry of Water
Resources in India carried out flood damage assessment based on: area flooded, frequency
of floods, probable depth and duration of inundation, crop area damaged and value of
damaged crops, amount of damage to houses, number of cattle heads lost and damage to
public utilities. A bottom-up reporting process was followed. For example, block
supervisors used to report to sub-divisional officers. Between 1957-1964, several
high-level committees were constituted to improve the techniques of data collection,
M. M. Q. MIRZA ET AL. 67
Copyright © 2005 Taylor & Francis Group plc, London, UK
processing and reporting (RBA, 1980). These committees emphasized the need to send

out teams at the end of every flood season to contact other agencies collecting relevant
data, visit flood affected areas to ascertain the damage from local populations, and make
their own assessment. Rashtriya Barh Ayog (RBA, 1980) recommended that complete
enumeration methods be followed by the State Governments, and made other
recommendations to reduce over- and under-estimations. It also recommended use of
remote sensing techniques for flood damage assessment whenever possible.
Bangladesh has also improved its flood monitoring and assessment procedures since
the early 1960s, when flood control was given an institutional shape. The First Water
Master Plan of 1964 (EPWAPDA, 1964) prepared the foundation for flood control
processes in Bangladesh. The National Water Plan (MPO, 1986) and the Flood Action
Plan (FPCO, 1993) have significantly improved the flood damage data collection,
processing, interpretation and reporting techniques. For example, since 1980, the
Government of Bangladesh has been using Advanced Very High Resolution Radiometer
(AVHRR) data for assessing flooded areas and damage. The Bangladesh Water
Development Board has a comprehensive network of data collection through its field
offices (Rasid and Pramanik, 1993).
3.7.2 INCREASED HUMAN SETTLEMENT IN FLOOD-PRONE AREAS
It is estimated that by 1991, about 530 million people lived in the GBM basins in Nepal (4%),
India (75%) and Bangladesh (20%). Most of them dwelt on floodplains susceptible to
annual floods. In India, the number of people affected by the average annual floods rose
from 16 million to 53 million in the 30-year period leading up to the late 1980s (CSE,
1992). In Bangladesh it has increased from 13 million to 38 million for the same period.
This translates into about 10 million additional households at risk from annual flooding.
When people living in areas reached by larger than normal floods are considered, the
population and dwellings now at risk due to population growth, escalates dramatically.
Although the exact totals are not known, the rapid increase in the number of dwellings at
risk over the last 30 years will have contributed significantly to the increase in the amount
of flood damage reported in that period.
People living in, or associated with these at threat floodplains draw their livelihood
from them. Such activities include agriculture, industries, public infrastructure, and

commercial outlets. For example, during the 1950s and 1960s, India’s cultivated area
increased from about 119 mha in 1950-1951 to 141 mha in 1970-1971. Since then, it has
remained relatively stationary, although in some areas it intensified through the “green
revolution” (CSE, 1992). In India, agriculture accounts for 48% of the average annual
flood damage. The average crop area affected annually by floods increased sharply from
less than 2 mha (30% of total) in the 1950s to over 4.5 mha (>50%) in the 1980s
(CSE, 1992).
Similarly, the agricultural area in Bangladesh has expanded in the last 30 years. The
net-cropped area increased from 8.43 mha in 1972/1973 to 8.61 mha in 1989/1990. The
French Engineering Consortium (FEC, 1989b) has estimated that agriculture and
dwellings each accounts for roughly 30% of the total damage. The remaining 40% is from
damage to infrastructure, such as roads, railways, and water works.
While it is highly likely that damage estimates have improved and accounted for a
significant portion of the escalating flood damage figures reported for countries within the
GBM basins, it is most likely that increasing settlement and associated economic activities
in flood-prone areas is the more important contributor. It is these human use elements of
68 ARE FLOODS GETTING WORSE?
Copyright © 2005 Taylor & Francis Group plc, London, UK
the flood hazard to which an explanation must be sought for rising flood damage and not
the flood events themselves.
3.8 CONCLUSIONS
The findings in this article have a number of policy level implications for the government
agencies of the countries that share the GBM basins and the donors/aid agencies involved.
First, governments should formulate and implement policies to discourage further
settlement in the flood-vulnerable areas. This requires concrete action plans in terms of
delineation of the floodplains based on risk factors, enactment of proper legislation and
allocation of suitable lands to settle additional people. Second, plans should be formulated
to protect economic and industrial centers in order to reduce flood damage. Adoption and
implementation of such plans requires political decisions as well as significant investment.
Third, flood damage mitigation programmes should provide more emphasis to local level

actions. Historically, a top-down flood management approach has been in practice in the
countries that share the GBM basins. People at the lowest level are seldom consulted in
developing any flood management plan. Therefore, in many cases, stipulated benefits are
not delivered to the vulnerable population and areas (Adnan, 1991). Fourth, flood
research should also focus on understanding the relationship between climate variations
and societal factors in the observed flood damage. Presently, flood research in the
GBM basins is more focused on structural solutions rather than exploring any possible
relationship between hydro-meteorological and societal factors. Cooperation between
upstream and downstream countries in flood management could also reduce vulnerability
to a great extent (Verghese and Iyer, 1993). Although some cooperative arrangements
exist between Bangladesh and India, other countries of the GBM basins - Nepal and Bhutan
should be brought under a comprehensive regional flood management plan. This requires
executive initiative and decision from the highest political levels. Fifth, flood damage
adjustment research should receive more focus at the government level. Although
independent researchers (Paul, 1997; Rasid and Mallik, 1995) have conducted some
studies on this topic, flood research specifically on damage adjustment at the government
level is almost absent. Direct government funded research institutions could undertake the
necessary research studies. This will require the allocation of additional financial and
human resources by governments. Sixth, donors and aid agencies should put emphasis on
projects that build capacity at local levels to reduce vulnerability and damage. In the
countries of the GBM basins, donors and aid agencies usually exert a substantial influence
in shaping socio-economic development policies of the respective governments.
Therefore, they could bring about a significant shift in the flood management policies by
focusing more on projects for capacity building at local levels, in pursuance of a bottom-up
planning process.
M. M. Q. MIRZA ET AL. 69
Copyright © 2005 Taylor & Francis Group plc, London, UK
Annex 3.1
(i) Cumulative Deviation Test (Buishand, 1982): The purpose of this test is to detect the
existence of a jump after m observations:

Given the observations
XX
n
1

, we let
The test statistic
is percentage points of the statistic Q are given in
(ii) Worsley Likelihood Ratio Test (Worsley, 1979):
computing
The test statistic is
Worsley (1979) gives critical
values for the test statistic W.
(iii) Kruskal-Wallis Test (Coshall, 1989): The test is applied for determining equality of
sub-sample means. For this test the time series data is divided into m sub-samples with
(3.1)
(3.3)
(3.2)
Buishand (1982).
lengths n
j
( j = 1, 2, , m) and R
ij
is the rank of the ith observation of the jth sub-sample
in the ordered complete sample. The test statistic is:
70 ARE FLOODS GETTING WORSE?
The basic assumption for this test is that the observations are independent and
normally distributed. The test can still be applied, however, when there are slight
departures from normality.
Copyright © 2005 Taylor & Francis Group plc, London, UK

where, R
j
is the total ranks in the jth sub-sample, i.e. R
j
= ∑ R
ij
, and N = is the total
sample size.
When ties are involved in the ranking procedure,
H
is divided by where
T = t
3
- t and t is the number of tied observations in a tied group of scores.
Under the null hypothesis, for larger n
j
, the statistic follows the Chi-Square
Distribution with (m-1) degrees of freedom.
(iv) Mann-Whitney U Test (Kite, 1988; Essenwanger, 1985): This test is suggested for
determining progressive change in the mean value with time. If two sub-samples are not
well mixed and the entire sample is ranked, the elements of one sub-sample display
relatively low rank numbers while those of the other display relatively high rank numbers.
The test statistic U, reflects the low value arising when the sub-samples are not well mixed.
Therefore, if the observed U value is less than a certain critical value U
cr
, the hypothesis
that there is no location difference between the sub-samples is rejected.
The test statistic is calculated as the smaller of U
1
and U

2
and
(3.4)
(3.5)
(3.6)
(3.7)
M. M. Q. MIRZA ET AL. 71
where n
1
ϭ size of the first sub-sample, n
2
ϭ size of second sub-sample, R
1
= sum of ranks
attributed to members of the first sub-sample in the ranks total sub-sample.
When both sub-samples are larger than 20, U is approximately normally distributed
with mean n
1
n
2
/2 and standard deviation
The test statistic
Z
is calculated as:
Copyright © 2005 Taylor & Francis Group plc, London, UK
The Mann-Whitney test is based on a continuous distribution and if tied observations
are encountered, a correction must be made. If t is the number of observations tied for a
where
is the sum of T’s over all groups of tied observations.
(3.8)

given rank and
T
= 1/12 (t
3
- t) then the Z statistic is
72 ARE FLOODS GETTING WORSE?
Copyright © 2005 Taylor & Francis Group plc, London, UK
REFERENCES
Adnan, S.: Floods, People and Environment: Institutional Aspects of Flood Protection Programmes
in Bangladesh, Research and Advisory Services, Dhaka, Bangladesh, 1991.
Asian Development Bank (ADB): Disaster Mitigation in Asia and the Pacific, ADB, Manila, 1991.
Asian Development Bank (ADB): Annual Report 1999, ADB, Manila, 2000.
Asian Disaster Preparedness Center (ADPC): Bulletin on Nepal Floods, ADPC, Bangkok, 2000.
Asian Disaster Reduction Center (ADRC): Disaster Database ( 2000a.
Asian Disaster Reduction Center (ADRC): Latest Disaster Information ( />2000b.
Bangladesh-Bhutan Joint Team of Officials (BBJTO): Report on Flood Control and Flood
Mitigation. Government of Bangladesh and His Majesty’s Government of Bhutan, Dhaka and
Thimpu, 1989.
Bangladesh Water Development Board (BWDB): Flood in Bangladesh: Investigation, Review and
Recommendation for Flood Control, BWDB, Dhaka, 1987.
Bangladesh Water Development Board (BWDB): Annual Flood Report 1992, BWDB, Dhaka, 1993.
Bangladesh Water Development Board (BWDB): Discharge Data for Various Rivers in Bangladesh,
BWDB, Dhaka, 1995.
Bangladesh Water Development Board (BWDB): Discharge Data for Various Rivers in Bangladesh,
BWDB, Dhaka, 2000a.
Bangladesh Water Development Board (BWDB): Flooded Area Data, BWDB, Dhaka, 2000b.
Barua, D. K.: On the Environmental Controls of Bangladesh River Systems. Asia Pacific Journal on
Environment and Development 1(1) (1994), pp.81-98.
Brammer, H.: The Geography of the Soils of Bangladesh, University Press Ltd., Dhaka, 1996.
Buishand, T. A.: Some Methods for Testing the Homogeneity of Rainfall Records. Journal of

Hydrology 58 (1982), pp.11-27.
Carson, B.: Erosion and Sedimentation Processes in the Nepalese Himalaya. International Center for
Integrated Mountain Development (ICIMOD), Occasional Paper No.1, ICIMOD, Kathmandu,
1985, p.36.
Central Water Commission (CWC): State-Wise Data on Flood Affected Areas and Damages 1953 to
1984, CWC, New Delhi, 1985.
Central Water Commission (CWC): Water and Related Statistics, CWC, New Delhi, 1989.
Center for Science and Environment (CSE): Floods, Floodplains and Environmental Myths, CSE,
New Delhi, 1992.
China -Bangladesh Joint Experts Team (CBJET): Study Report on Flood Control and River Training
Project on the Brahmaputra River in Bangladesh, Vol. 1 and Vol. 2. Government of Bangladesh
and Government of China, Dhaka and Beijing, 1991.
Chowdhury, A. M.: Flood 1988. In: Flood in Bangladesh (M. Ahmad Edited), Community
Development Library, Dhaka, 1989, pp.235-240.
Coshall, J.: The Application of Non-Parametric Statistical Tests in Geography, University of East
Anglia, Norwich, Norfolk, U.K., 1989.
Dhar, O. N. and Nandargi, S.: Floods in Indian Rivers and Their Meteorological Aspects. In: Kale,
V. S. (Editor), Flood Studies in India, Geological Society of India, Bangalore, 1998, pp.1-25.
East Pakistan Water and Power Development Authority (EPWAPDA): Master Plan. International
Engineering Company Inc., San Francisco, 1964.
Essenwanger, O. M.: World Survey of Climatology. General Climatology 1B, Elsevier, Amsterdam,
1985.
Flood Action Plan (FAP) 24: Hydrological Study: Phase I. Flood Plan Coordination Organization
(FPCO), Dhaka, 1993.
Flood Plan Coordination Organization (FPCO): Disaster Preparedness Program. FPCO, Dhaka, 1993.
French Engineering Consortium (FEC): Pre-Feasibility Study for Flood Control in Bangladesh,
Annexes, FEC, Paris, 1989a.
French Engineering Consortium (FEC): Pre-Feasibility Study for Flood Control in Bangladesh,
Vol. 5, Financial and Economic Aspect, FEC, Paris, 1989b.
M. M. Q. M

IRZA ET AL. 73
Copyright © 2005 Taylor & Francis Group plc, London, UK
Hamilton, L.: What are the Impacts of Himalayan Deforestation on the Ganga-Brahmaputra
Lowlands and Delta? Assumptions and Facts. Mountain Research and Development 7(3) (1987),
pp.256-263.
Hofer, T.: Do Land-Use Changes in the Himalayas Affect Downstream Flooding? Traditional
Understanding and New Evidences. In: Kale, V. S. (Editor), Flood Studies in India, Geological
Society of India, Bangalore, 1998, pp.119-141.
IAHS-AISH: World Catalogue of Maximum Observed Floods. IASH-AISH Publ. No. 143, UNESCO,
Paris, 1984.
Islam, N.: Flood ’98 and the Future of Urban Settlements in Bangladesh. In: Ahmad, Q. K., Chowdhury,
A. K. A., Imam, S. H., Sarker, M. (Eds.), Perspectives on Flood 1998, University Press Ltd.,
Dhaka, 2000, pp.52-65.
Ives, J. D. and Messerli, B.: The Himalayan Dilemma: Reconciling Development and Conservation,
Routledge, New York, 1989.
Ives, J. D.: Floods in Bangladesh: Who is to Blame? New Scientist, April 13
th
, 1991, pp.30-37.
Jamuna Multipurpose Bridge Authority (JMBA): Jamuna Bridge Project, Phase II Study: Feasibility
Report, Vol. II, JMBA, Dhaka, 1989.
Kite, G. W.: Frequency and Risk Analysis in Hydrology. Water Resources Publications, Fort Collins,
USA, 1988.
Kothyari, U. C. and Garde, R. J.: Annual Runoff Estimation for Catchments in India. Journal of
Water Resources Planning and Management 117(1) (1991), pp.1-10.
Master Plan Organization (MPO): National Water Plan. Ministry of Irrigation, Water Development
and Flood Control, Dhaka, 1986.
Messerli, B. and Hofer, T.: Assessing the Impact of Anthropogenic Land-Use Changes in the
Himalayas. In: Chapman, G. P., Thompson, M. (Eds.), Water and the Quest for Sustainable
Development in the Ganga Valley, Global Development and the Environment Series, Mansell,
London, 1995, pp.64-89.

Ministry of Disaster Management and Relief (MDMR), Bangladesh: Estimated Sector-Wise
Damages Due to Flood 1998, MDMR, Dhaka, 1998.
Ministry of Foreign Affairs (MOFA), Bangladesh: Damages Caused by Flood 1998
( 1998.
Ministry of Water Resources (MWR): Flood Affected Areas ( />flood_affected.htm), 2000.
Mirza, M. M. Q.: Environmental Impacts of Floods, Droughts and Deforestation in the Himalayan
Region. In: Veziroglu, T. N. (Editor) Energy and Environmental Progress-I., Vol. 6. Nova
Science Publishers, New York, 1991a, pp.551-570.
Mirza, M. M. Q.: Flood Action Plan of Bangladesh - The Embankment Issue. Water Nepal 2(2/3)
(1991b), pp.25-28.
Mirza, M. M. Q.: Modeling the Effects of Climate Change on Flooding in Bangladesh, Unpublished
D.Phil Thesis. International Global Change Institute (IGCI), University of Waikato, Hamilton,
New Zealand, 1997.
Mirza, M. M. Q. and Dixit, A.: Climate Change and Water Resources in the GBM Basins. Water
Nepal 5(1) (1997), pp.71-100.
Mirza, M. M. Q., Warrick, R. A., Ericksen, N. J. and Kenny, G. J.: Trends and Persistence in
Precipitation in the Ganges, Brahmaputra and Meghna basins in South Asia. Hydrological
Sciences Journal 43(6) (1998), pp.845-858.
Mitchell Jr., J. M., Dzeedzeevskii, B., Flohn, H., Hofmeyr, W. L., Lamb, H. H., Rao, K. N. and
Wallen, C. C.: Climatic Change. Technical Note No. 79, WMO, Geneva, 1966.
Mooley, D. A. and Parthasarathy, B.: Droughts and Floods Over India in Summer Monsoon Seasons
1871-1980. In: A. Street-Parrot, M. Beran, R. Ratchiffe (Eds.), Variations in the Global Water
Budget. D. Reidel, Dordrecht, the Netherlands, 1983, pp.239-252.
Nepal Water Conservation Foundation (NWCF): Peak Discharge Data for the Kosi River at
Barahkshetra, NWCF, Kathmandu, 1996.
Paul, B. K.: Flood Research in Bangladesh in Retrospect and Prospect: A Review. Geoforum 28(2)
(1997), pp.121-131.
Raghunath, H. M.: Hydrology. Wiley Eastern, New Delhi, 1985.
74 ARE FLOODS GETTING WORSE?
Copyright © 2005 Taylor & Francis Group plc, London, UK

Rahman, M. A.: The Tale of a Delta, the Rivers, the Donor Dictature and FAP: The Congenital
Profligacy. In: Proceedings of Open Discussion on Flood Action Plan, 12 November. Institution
of Engineers, Dhaka, Bangladesh, 1993.
Rangachari, R.: Flood Management. In: B.G. Verghese, R.R. Iyer (eds.), Harnessing Eastern
Himalayan Rivers: Regional Cooperation in South Asia, Konark Publishers, New Delhi, 1993,
pp.86-98.
Rasid, H. and Mallik, A.: Flood Adaptations in Bangladesh - Is the Compartmentalization Scheme
Compatible with Indigenous Adjustments of Rice Cropping to Flood Regimes. Applied
Geography 15(1) (1995), pp.3-17.
Rasid, H. and Pramanik, A. H.: Areal Extent of the 1988 Flood in Bangladesh: How Much Did the
Satellite Imagery Show? Natural Hazards 8 (1993), pp.189-200.
Rashtriya Barh Ayog (RBA): Report, Vol. 1. Ministry of Energy and Irrigation, New Delhi, 1980.
Rogers, P., Lydon, P. and Seckler, P.: Eastern Waters Study: Strategies to Manage Flood and Drought
in the Ganges-Brahmaputra Basin, ISPAN, Virgina, USA, 1989.
Salinger, J. M.: New Zealand Climate: The Instrumental Method, Unpublished Ph.D Thesis, Victoria
University of Wellington, New Zealand, 1980.
Sir William Halcrow and Partners: River Training Studies of the Brahmaputra River. First Interim
Report, Annex I, Part 4: Analysis of Discharge Measurements, Sir William Halcrow and
Partners Ltd., Dhaka, 1991.
South Asian Association for Regional Cooperation (SAARC): Regional Study on the Causes and
Consequences of Natural Disasters and the Protection and Preservation of the Environment,
SAARC Secretariat, Kathmandu, 1992.
Thompson, M. and Warburton, M.: Uncertainty on a Himalayan Scale. Mountain Research and
Development 5(2) (1985), pp.115-135.
United Nations Educational Scientific and Cultural Organization (UNESCO): World Catalogue of
Very Large Floods, UNESCO, Paris, 1976.
United Nations International Children’s Emergency Fund (UNICEF): West Bengal Flood Emergency
Situation Report, 24 Oct 2000, UNICEF, India, 2000.
Verghese, B. G. and Iyer, R. R.: Harnessing Eastern Himalayan Rivers: Regional Cooperation in
South Asia, Konark Publishers, New Delhi, 1993.

Worsley, K. J.: On the Likelihood Ratio Test for a Shift in Location of Normal Populations. Journal
of American Statistical Association 74 (1979), pp.365-367.
M. M. Q. MIRZA ET AL. 75
Copyright © 2005 Taylor & Francis Group plc, London, UK