Climate Change and Water Resources in South Asia - Chapter 5 pdf

5
The Implications of Climate Change on River
Discharge in Bangladesh
5.1 INTRODUCTION
5.1.1 WATER RESOURCES PROBLEM OF BANGLADESH
Bangladesh lies in the delta of three large rivers - the Ganges, Brahmaputra and Meghna
(GBM), which is often termed as a “land of rivers and water.” With a complex network of
230 rivers, including 57 cross boundary rivers, about 92.5% of the 175 million hectares
(mha) of combined basin area of the GBM Rivers (Fig. 5.1) is beyond the boundary of
Bangladesh and is located in China, Nepal, India and Bhutan. Therefore, Bangladesh acts
as a drainage outlet for the cross-border runoff. More than 90% of the annual runoff is
generated outside of Bangladesh. However, there is a high seasonal difference in the
availability of water. For example, for the Ganges River, the ratio of dry and monsoon
runoff is 1:6 (Fig. 5.2). This illustrates that Bangladesh has an abundance of water in the
monsoon while the country still faces surface water scarcity in the dry season. Irrigated
agriculture is highly dependent on dry season surface water availability. On average,
annually floods engulf roughly 20.5% of the area of the country, or about 3.03 mha (Mirza,
2003). In extreme cases, floods may inundate about 70% of Bangladesh, as it occurred
during the floods of 1988 and 1998 (Ahmed and Mirza, 2000). Hydrological droughts are
very common in the rivers of Bangladesh.
The magnitude of precipitation over the GBM basins is very high and more than
three-quarters occurs during the summer monsoon (June-September) (Table 5.1). The
resulting huge volume of cross-border monsoon runoff, together with the locally
generated runoff and some physical factors, either singly or in combination, causes floods
in Bangladesh. The physical factors, either singly or in combination, include snow and
glacier melt, El Niño Southern Oscillation (ENSO) induced conditions, loss of drainage
capacity due to the siltation of principal distributaries, backwater effect, unplanned
infrastructure development, deforestation and the synchronization of flood peaks of the
major rivers. Recently Mirza (2003), compared three recent extreme floods (1987, 1988
and 1998) in Bangladesh and found that the intense monsoon precipitation was the
principal cause of flooding. However, there are differences in opinions concerning the role

of deforestation in upstream areas in the flooding process in Bangladesh. Deforestation of
steep slopes in the Himalayas is assumed to lead to accelerated soil erosion and landslides
M. MONIRUL QADER MIRZA
A part of this chapter was published in the Climatic Change 57 (2003), pp.287-318 and reprinted
with permission.
Copyright © 2005 Taylor & Francis Group plc, London, UK
during monsoon precipitation. This in turn is believed to contribute to devastating floods
in Bangladesh (Khalequzzaman, 1994; Hamilton, 1987). Hofer (1998) concluded that
land-use changes in the Himalayas were not responsible for floods in India and Bangladesh.
With regard to sedimentation, the existing publications do not report any significant recent
increase in the sediment load of the major rivers and their tributaries (Ives and Messerli,
1989).
Fig. 5.1 The Ganges, Brahmaputra and Meghna basins.
Fig. 5.2 Hydrograph of the Ganges (lighter solid line) and Brahmaputra (thicker solid line) Rivers for
the typical water year 1967-1968. The values are in m
3
/sec. Data source: Bangladesh Water
Development Board (BWDB, 1995).
0
10000
20000
30000
40000
50000
60000
70000
80000
4/1/67
5/1/67
6/1/67

7/1/67
8/1/67
9/1/67
10/1/67
11/1/67
12/1/67
1/1/68
2/1/68
3/1/68
104 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
Table 5.1 Mean annual precipitation in the Ganges, Brahmaputra and Meghna basins
Basin

Country Mean Annual
Precipitation (mm)

Ganges

Nepal
India
Bangladesh

1,860
450-2,000
1,570

Brahmaputra

Tibet (China)
Bhutan
India
Bangladesh

400-500
500-5,000
2,500
2,400

Meghna/Barak


India
Bangladesh

2,640
3,575


Source: Mirza, 1997.
Bangladesh generally experiences four main types of floods: flash, riverine, rain and
storm-surge (Fig. 5.3). Eastern and Northern areas of Bangladesh adjacent to its border
with India are vulnerable to flash floods. Rivers in these regions are characterized by sharp
rises and high flow velocities resulting from exceptionally heavy rainfall occurring over the
hilly and mountainous regions in the neighboring India. Riverine floods occur when flood
water of the major rivers and their tributaries and distributaries spill. With the onset of the
monsoon in June, all of the major rivers start swelling to the brim and bring flood water
from the upstream basin areas. Rain floods are caused by intense local rainfall of long
duration in the monsoon months. Heavy pre-monsoon rainfall (April-May) causes local

runoff to accumulate in depressions. Later (June-September), local rainwater is
increasingly ponded on the land by the rising water levels in the adjoining rivers. Coastal
areas of Bangladesh, which consist of large estuaries, extensive tidal flats, and low-lying
offshore islands, are vulnerable to storm-surge floods, which occur during cyclonic storms.
Cyclonic storms usually occur during April-May and October-November.
Flood is a necessity as well as a danger in Bangladesh. For example, normal floods
help the growth of rice crops because of the fertilization produced by nitrogen supplying
blue-green algae, which grow in the ponded clear flood water (World Bank, 1989). The
extra moisture provided by large floods to higher lands also benefits rabi crops such as
vegetables, lintels, onion, mustard, etc. (Brammer, 1990). Rabi refers to a cropping season
from November-May. But, high flood levels can cause substantial damage to key
economic sectors: agriculture, infrastructure and housing. Based on the reported crop
damage due to floods, average annual loss is estimated to be 0.47 million tons (Paul and
Rasid, 1993). However, in a year of an extreme flood such as 1998, food grain loss may
exceed 3.5 million tons (Ahmed, 2001). The total monetary loss caused by the extreme
floods of 1998 and 1988 was US$ 3.4 billion and US$ 2.0 billion, respectively or 10% of
the GDP of Bangladesh in the respective years (Bhattacharya, 1998; World Bank, 1989).
For a country like Bangladesh with a transitional economy and a low per capita income
($360 in 2001) (World Bank, 2003), this amount of loss is very high. Although flood
affects people of all socio-economic status, the rural and urban poor have been the hardest
hit.
M. M. Q. MIRZA 105
Copyright © 2005 Taylor & Francis Group plc, London, UK
Fig. 5.3 Bangladesh and various flood types.
5.1.2 RATIONALE OF THE RESEARCH
Future climate change may affect water resources availability and extreme hydrological
events such as floods in Bangladesh in many ways. The IPCC (2001) indicated a likelihood
of increased intensity of extreme precipitation over the South Asian region. All climate
models simulate an enhanced hydrological cycle and increases in annual mean rainfall
over South Asia (under non-aerosol forcing). In all periods of simulation (GHG and

GHG + aerosol forcing), summer precipitation shows an increase. The magnitude of
increase in summer precipitation with GHG + aerosol forcing is smaller than that seen in
the GHG forcing. The difference in change with aerosol forcing is due to its dampening
106 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
effect on Indian summer monsoon precipitation (Lal et al., 2001; Cubasch et al., 1996;
Roeckner et al., 1999).
Annual runoff may increase as a result of increased precipitation. However,
uncertainty remains in dry season availability of river flow as it is related to a number of
factors. They include: amount of monsoon precipitation and ground water recharge, amount
of snowfall, temperature gradient, snowmelt, evaporation, upstream water demand, etc.
More frequent extreme precipitation could increase the possibility of flash floods.
Increased precipitation in the GBM basins may increase the magnitude, depth and spatial
extent of riverine and rain floods. Based on a series of theoretical and model-based
studies, including the use of a high resolution hurricane prediction model, it is likely that
peak wind intensities will increase by 5% to 10% and the mean and peak precipitation
intensities by 20% to 30%, in some regions (IPCC, 2001). Therefore, stronger
storm-surges can aggravate coastal flooding. Of all of these flood types, the riverine floods
are the most pervasive and have long-term impacts on land-use, the economy and most
development strategies for Bangladesh. Thus, it is with the changes in riverine flooding
that the effects of climate change may be most strongly felt.
In the past, a number of studies on climate change and its possible implications on
Bangladesh have been undertaken (Ahmad and Warrick, 1996; ADB, 1994; and Resource
Analysis, 1993). The consensus was that over the past 100 years, the broad region
encompassing Bangladesh had warmed by 0.5
o
C (Ahmad and Warrick, 1996). However,
overall increases in precipitation were not found (Mirza et al., 1998). These studies also
indicated that with increases in precipitation in Bangladesh and surrounding areas due to
climate change, flooding in Bangladesh might worsen. However, no specific research has

assessed changes in flooding in terms of magnitude, depth and spatial extent in Bangladesh
taking into account possible changes in precipitation in the cross-border basin areas of the
GBM Rivers.
5.2 OBJECTIVES
As indicated above, the annual runoff in the GBM basins may be changed due to possible
changes in future climate and it may also exacerbate the flood problem in Bangladesh. Most
experiments using GCMs show increases in monsoon precipitation as a consequence of
enhanced greenhouse effect. However, it is not known exactly what the magnitude of climate
change will be in the future or how it will affect precipitation, and thereby flooding in Bangladesh.
Therefore, a study was carried out under the BDCLIM (Bangladesh Climate) project to
examine possible changes in flooding in Bangladesh under climate change. The BDCLIM is a
large integrated model system developed for assessing the effects of future climate change
scenarios on Bangladesh (Warrick et al., 1996).
Taking into account the range of uncertainty in the climate scenarios, the overall goals
of this research include: 1) determining the sensitivity of mean annual and mean peak
discharge at the boundary of Bangladesh to future climate change and 2) estimating the
consequent changes in depth and spatial extent of flooding in Bangladesh.
5.3 METHODOLOGY
In order to meet the first objective, four major steps were followed. First, an empirical
relationship between precipitation and discharge was determined. Second, climate
change scenarios were constructed for the three river basins using the results of CSIRO9
(McGregor et al., 1993), UKTR (Murphy and Mitchell, 1995), GFDL (Whetherland and
M. M. Q. MIRZA 107
Copyright © 2005 Taylor & Francis Group plc, London, UK
Manabe, 1986), and LLNL (Whener and Convey, 1995) GCMs in the SCENGEN
software of the Climatic Research Unit (CRU), University of East Anglia, U.K. (CRU,
1995). Third, the climate change scenarios were applied to empirical models in order to
determine the magnitude of changes in discharge at the boundaries of Bangladesh. Fourth,
the MIKE 11-GIS hydrodynamic model was forced with current and future peak discharges
to simulate river flood stages and depth and spatial extent of flooding within Bangladesh.

The MIKE 11 is a professional engineering software tool that simulates flows, water
quality and sediment transport in river basins, estuaries, irrigation systems, channels and
other water bodies. The Danish Hydraulic Institute (DHI) developed the software.
The GIS interface was developed and applied during the Flood Action Plan (FAP) Study
(1990-1995) in Bangladesh. The model has been calibrated and validated in a Bangladesh
context by the Surface Water Modeling Center (SWMC), Dhaka and is currently being
used for water resource development, planning and management.
5.3.1 DEVELOPMENT OF EMPIRICAL DISCHARGE MODELS
As a first step for determining the sensitivity of mean peak discharge at the boundary of
Bangladesh, different approaches of modeling were envisaged. The empirical modeling
approach was compared to the water-balance, lumped-parameter and physically-based
distributed models and found to be preferable on the basis of the constraints imposed by
the large areal extent of the river basins and the lack of available data and resources.
Sensitivity analyses for three selected stations in the Ganges, Brahmaputra and Meghna
River basins was carried out using the model R = P - E. Here R = runoff, P = precipitation
and E = actual evapo-transpiration, which was calculated using the relationship
E =
))(1(
2
PE
P
P
+
(Pike, 1964), where PE = potential evapo-transpiration. The analysis showed that runoff
was far more sensitive to precipitation than to temperature (Mirza, 1997; Mirza and Dixit,
1997) (Fig. 5.4). Therefore, temperature was excluded as an explanatory variable for empirical
model building but it may be considered as an explanatory variable as part of a future
research undertaking.
The results of the sensitivity analysis also shows that, in percentage terms, runoff is
more sensitive to precipitation and temperature changes in relatively dry stations than wet

stations. As an example, in the case of the New Delhi station (a drier station in the Ganges
basin) no change in temperature and a 4% increase in precipitation changes runoff by
+11%, while for the Gauhati and Syhet (the wetter stations in the Brahmaputra and Meghna
basins, respectively), the changes in runoff are +6% and +8%, respectively. In the extreme
case, a 5
o
C increase in temperature and a 20% increase in precipitation could increase
runoff by 29% at the New Delhi station, whereas for Gauhati and Syhet stations the
expected changes are 22% and 21%, respectively.
Accordingly, time-series data for precipitation were collected from various primary
and recognized secondary sources for the three river basins. Sources of precipitation data
were: 1) Carbon Dioxide Information Analysis Center (CDIAC)/Oak Ridge National
Laboratory (ORNL), Tennessee, USA; 2) Climatic Research Unit (CRU), University of
East Anglia, U.K.; 3) Nepal Water Conservation Foundation (NWCF), Kathmandu;
4) The Bangladesh Water Development Board (BWDB), Dhaka; 5) United Nations; and
108 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
(c)
Fig. 5.4 Sensitivity of runoff to temperature and precipitation changes in the: (a) Ganges basin
(New Delhi), (b) Brahmaputra basin (Gauhati) and (c) Meghna basin (Syhet).
(a)
(b)
M. M. Q. M
IRZA 109
Copyright © 2005 Taylor & Francis Group plc, London, UK
6) Center for Ocean-Land-Atmosphere Studies (COLA), Maryland, USA. Discharge data
was received from the Bangladesh Water Development Board. Details of these datasets
are given in Mirza, 1997. Selection of the dataset for the development of empirical models
was made with regard to length of record, spatial coverage and missing observations. The
selected datasets were the COLA dataset and the NWCF dataset for the Ganges basin; the

COLA dataset and selected four stations from the UN dataset within Bangladesh for the
Brahmaputra basin; and the COLA dataset and the UN and BWDB datasets for the Meghna
basin. Missing observations were between 1%-12% for the NWCF, UN and BWDB datasets.
These observations were filled in by applying the method stated by Salinger, 1980. After
filling in the missing observations, the means and standard deviations were computed for
the complete time series and compared with those of the incomplete time series. The
difference in the means and standard deviations were found to be statistically insignificant
at a 5% level of significance.
The precipitation and discharge data were examined with respect to their adequacy of
empirical modeling. Statistical tests show that the precipitation observations in all
meteorological sub-divisions are normally distributed. Over the periods of record, one
meteorological sub-division (The East Madhaya Pradesh) (V10 in Fig. 5.5a) in the Ganges
basin shows a statistically significant decreasing trend. In the Brahmaputra basin, a
decreasing trend is found only in the precipitation time series of South Assam (V2 in
Fig. 5.5b). However, the basin-wide average precipitation series does not show any
discernible trend. On the other hand, each of these two sub-divisions covers a small area
over the respective river basin. Therefore, they would not have a major effect on the
predictive capability of the empirical models. Precipitation observations of all
meteorological sub-divisions are found to be random, with a few exceptions. Analysis
shows the presence of Markov linear type “persistence” only in the observations of the
North Assam and South Assam meteorological sub-divisions in the Brahmaputra basin
(Mirza, 1997).
Annual mean and peak discharge series have been found to be normally distributed for
the GBM Rivers. Statistical tests indicate that the difference in mean annual discharge
of the Ganges River at Hardinge Bridge for the pre- and post-Farakka period is not
statistically significant. Therefore, on an annual basis, the regulation effect of the Farakka
Barrage (Fig. 5.1) can be overlooked (Mirza, 1997). The barrage was constructed at Farakka
(18 km from the border of Bangladesh) and commissioned in April of 1975 to divert
1,134 m
3

/sec water to make the Hooghly-Bhagirathi River channel (on which the port of
Kolkata is situated) navigable (Mirza, 2002).
A sequence of empirical models that describe the relationship between precipitation
and annual mean and peak discharge was developed. One of the advantages of such a
relationship is, for example, that in absence of precipitation data, peak discharge can be
estimated from known values of annual discharge. Initially, in order to examine the
independence of the explanatory variables, annual mean discharges of the Ganges River at
Hardinge Bridge and Brahmaputra River at Bahadurabad in Bangladesh (Fig. 5.1) were
regressed on the meteorological sub-division wide annual precipitation data. Initial
examination indicated the presence of multi-collinearity in the precipitation data. This is
the condition where at least one explanatory variable is highly correlated with another
explanatory variable or with some combination of other explanatory variables (Maidment,
1993). Multi-collinearity may cause a number of consequences. (1) In extreme cases, the
least square point estimates can be far from the true values of the regression parameters,
and some estimates may even have the incorrect sign; (2) increases in standard error of
regression coefficient estimators occur as the correlations among the independent
110 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
Fig. 5.5a Independent and contiguous precipitation regions of the Ganges basin.
variables increase; (3) serious rounding errors in the calculation of the least square point
estimates are produced; and (4) significance tests and confidence intervals for regression
coefficients, due to increases in the standard errors of coefficient estimates, are affected.
The principal components analyses (Dunteman, 1989; Manly, 1986) of the precipitation
data were carried out to minimize the problems of collinearity and to generate relatively
independent, contiguous precipitation regions (Table 5.2 and Fig. 5.5c). Selection of
components and a procedure for regionalization are discussed in Cattel, 1966; Kaiser,
1960; Morgan, 1971; Ogallo, 1989 and Regemortel, 1995.
Multiple regression models were then developed for estimating mean annual discharge
for the Ganges and Brahmaputra Rivers. For the Meghna River, a multiple regression
model was developed between annual precipitation and the peak discharge. This was due

to the absence of adequate annual discharge data. In order to determine mean annual peak
discharge in relation to mean annual discharge, regression models between annual mean
and peak discharges were developed for the Ganges and Brahmaputra Rivers. Standard
procedures (Berry and Feldman, 1985; Bowerman and O’Connell, 1990; Cook and Wesberg,
M. M. Q. MIRZA 111
Copyright © 2005 Taylor & Francis Group plc, London, UK
Fig. 5.5b Independent and contiguous precipitation regions of the Brahmaputra basin.
1982) were followed to examine the model parameters. The precipitation annual mean
discharge regression models for the Ganges, Brahmaputra and Meghna basins are given in
Table 5.3.
Table 5.2 New variables (regions) derived by the principal components analysis

River Basin

Variables
(Sub-Divisions)
New Variables

Region 1 Region 2 Region 3

Ganges V1-V11* V3, V5-V11 V1 and V4 V2
Brahmaputra V1-V4** V1, V2 and V4 V3 -
Meghna V1-V3*** V2 and V3 V1 -

* V1 - Sub-Himalayan West Bengal; V2 - Gangetic West Bengal; V3 - Bihar Plateau; V4 - Bihar
Plain; V5 - East Up; V6 West Up; V7 - Haryana; V8 - East Rajasthan; V9 - West Madhaya
Pradesh; V10 - East Madhaya Pradesh; and V11- Nepal
** V1 - North Assam; V2 - South Assam; V3 - Sub-Himalayan West Bengal; and V4 - Teesta
Basin in Bangladesh
*** V1 - North Assam; V2 - South Assam; and V3 - Meghna basin (Bangladesh part)

112 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
5.3.2 CONSTRUCTION OF CLIMATE CHANGE SCENARIOS
For the first objective outlined above, the second step was to construct climate change
scenarios for the three river basins. Seven alternatives for scenario generation suggested
by Carter et al., 1994 and WMO, 1987 were reviewed. These alternatives include direct
use of GCM runoff changes, GCM-generated regional temperature and precipitation
changes, addition of GCM-predicted changes to baseline conditions, scaling of the
standardized patterns of change from GCMs, temporal analogues, spatial analogues and
hypothetical scenarios. For this research, the empirical models were developed based on
the spatial distribution of precipitation in the three river basins. Therefore, preference was
given to the method of scenario construction, which predicts spatial changes in
precipitation. For this purpose, the results of the GCMs are useful in that they indicate
possible spatial changes in climate.
For the scenario construction, a method of scaling “standardized” patterns of
precipitation derived from GCMs was adopted. Hulme (1994) recommended
standardizing GCM results for climate change scenario construction in order to overcome
the problem of variation of equilibrium global mean temperature and overcome the
problem of variation of equilibrium global mean temperature and precipitation changes
from GCM to GCM. This arises mainly because of the way the GCMs treat clouds and
oceans. Moreover, some of the atmospheric GCMs (For example, LLNL and MPILSG -
Fig. 5.5c Independent and contiguous precipitation regions of the Meghna basin.
M. M. Q. M
IRZA 113
Copyright © 2005 Taylor & Francis Group plc, London, UK
Table 5.3 Precipitation-annual mean discharge and annual mean-peak discharge regression models

Model Ganges Brahmaputra Meghna

Annual

Precipitation-
Mean Discharge

Q
a
= -6856 +8.02* P1 + 3.55 * P2 + 5.41*
P3 (R
2
=81.5%) (1)

where Q
a
is the estimated annual mean
discharge and P1, P2 and P3 are the area
weighted annual precipitation for the
Region 1, Region 2 and Region 3,
respectively.

The model excludes the part of the basin in
China.



Q
a
= 7201 + 5.23 * P
(R
2
= 62.33%) (2)

where Q
a
is the

estimated annual mean
discharge at Bahadurabad and P is the
area weighted annual precipitation in the
basin.

The model excludes the part of the basin
in Bhutan and China.

Multiple regression of Region 1 and
Region 2 produced negative parameter
(not significant) for the latter, which was
unrealistic from physical point of view.
This might have caused by error inherent
in the data. Therefore two regions were
treated as one homogeneous region.


Q
p
= -10531 + 3.43 *P1 + 5.69 *
P2 (R
2
=87.1%) (3)

where Q
p

is annual peak
discharge and P1 is the area
weighted annual precipitation in
Region 1 (North Assam) and P2
is the average of the
precipitation in Region 2 (South
Assam and the Bangladesh part
of the basin upstream of Bhairab
Bazar).

Annual Mean-
Peak Discharge
Q
p
= 14,844 + 3.26 Q
a
(R
2
=49.3%) (4)

where Q
p
is annual peak discharge

Q
p
= 6,816 + 3.00 Q
a
(R
2

=51.5%) (5)

-

114 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
MPI large scale geotropic ocean) were coupled with Ocean General Circulation Models
(OGCMs), while others include prescribed ocean heat. This produces substantial
inter-model differences in their simulation of current climate and response to a doubling of
CO
2
.
The results of 11 GCMs were compared, and four were selected - CSIRO9
(3.2 x 5.6
o
L9), UKTR (2.5 x 3.75
o
L19), GFDL (2.25 x 3.75
o
L14) and LLNL
(4 x 5
o
L15). This maximizes the range of predicted changes in precipitation amounts and
spatial variability within the GBM basins window. The other selection criterion was
goodness-of-fit of a GCM with respect to regional bias (control-observed). The CSIRO9,
UKTR and GFDL models showed a slight negative bias for summer precipitation but
showed a close fit, compared to the other GCMs. Note that the selected four GCM
experiments were based on only GHG forcing. These spatial patterns of precipitation change
were then “standardized” to account for the different climate sensitivity values of the GCMs.
This gave a pattern of change per degree of global warming. The standardized patterns

were then scaled for global mean temperature changes of 2
o
C, 4
o
C and 6
o
C giving a total of
12 scenarios (4 GCMs x 3 DTs) for each river basin and the Bangladesh window.
5.3.3 APPLICATION OF THE CLIMATE CHANGE SCENARIOS TO THE
EMPIRICAL MODELS
The third step was to apply the constructed climate change scenarios to the empirical
models to determine the magnitude of changes in discharges at the boundary of Bangladesh.
This was carried out in two stages: (1) changes in the mean annual discharge were
estimated by applying the scenarios of changes in the mean precipitation generated from
the results of four GCMs. These GCMs represent the range of uncertainty in climate model
projections of future climate change; and (2) the calculated mean annual discharge was
used to estimate changes in the mean annual peak discharge.
5.3.4 ESTIMATION OF CHANGES IN DEPTH AND EXTENT OF FLOODING IN
BANGLADESH
The fourth step was to force the MIKE 11-GIS model with current and future peak
discharges to simulate depth and spatial extent of flooding in Bangladesh. This was carried
out in three stages: (1) current and future mean precipitation was scaled for the Bangladesh
window as the MIKE 11-GIS model needed input of local precipitation for the simulation
purpose (Table 5.4); (2) current and future mean peak discharges were scaled to 1991
discharges for the Ganges, Brahmaputra and Meghna Rivers (Table 5.5). For scaling
purposes, the year ‘1991’ was selected because the monsoon of that year represented a
temporal distribution which was considered fairly ‘typical’ with regard to the usual
peaking time of the three rivers; and (3) the MIKE 11-GIS model was forced with current
and future peak discharges to simulate present and future depth and extent of flooding.
5.4 ESTIMATION OF CHANGES IN ANNUAL DISCHARGE

The precipitation change scenarios for the Ganges and Brahmaputra basins are applied to
the empirical models in order to assess the possible changes in the mean annual discharge
for 2
o
C, 4
o
C and 6
o
C increases in global temperature. For the Ganges basin, the empirical
model (Table 5.6) was developed between annual precipitation in the basin area in India
and Nepal and annual mean discharge at Hardinge Bridge in Bangladesh. This station is
M. M. Q. MIRZA 115
Copyright © 2005 Taylor & Francis Group plc, London, UK
located very close to the border of Bangladesh with India. Therefore, the measured
discharge takes account of the total cross-border inflow. Annual mean discharge is
predicted from the area weighted annual precipitation from three regions comprising the
total basin area (excluding parts of the basin area in China and Bangladesh).
116 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
An empirical model was developed between annual precipitation in the Brahmaputra
basin area (in India and Bangladesh) and the annual mean discharge of the Brahmaputra at
Bahadurabad, Bangladesh. The discharge measurement station (Bahadurabad) takes into
account the discharge generated in the Bangladesh part of the basin. The discharge is
predicted from the area weighted annual precipitation in the four meteorological
sub-divisions of the Brahmaputra basin.
The overall results with regard to changes in the mean annual discharge for the Ganges
and Brahmaputra Rivers are presented in Table 5.6. It is evident from Table 5.6 that the
mean discharge of the Brahmaputra River is less sensitive than the Ganges River to the
changes in precipitation. The results of the empirical model support the contention that
runoff or discharge of a wetter basin will be less sensitive to climate changes than a

relatively drier basin. Details regarding the changes in mean annual discharges for the
Ganges and Brahmaputra basins for the four selected GCMs are discussed below.
The Ganges Basin
Three precipitation change scenarios for global temperature increases of 2
o
C, 4
o
C and 6
o
C
were considered for determining changes in the mean annual discharge. Changes in the
magnitude of mean discharges from present to the future conditions for the Ganges basin
for the four GCMs (CSIRO9, UKTR, GFDL and LLNL) are shown graphically in
Figure 5.6. For all four models, the figure shows an increase in discharge as global
temperature and thereby basin-wide precipitation, increases. The figure also shows a marked
M. M. Q. MIRZA 117
Copyright © 2005 Taylor & Francis Group plc, London, UK
increase in discharge (for each of the scenarios) across the four models, with LLNL
showing least change and UKTR showing most change. Three of the models show
significant increases in mean discharge. The application of scenarios from the UKTR model
shows that a 39% change in precipitation (at a 6
o
C rise in global mean temperature)
produces a 63.4% change in the mean discharge. In absolute terms, this implies an increase
in mean discharge from the current 11,606 m
3
/sec to 18,970 m
3
/sec.
In contrast, the LLNL model shows the least changes for all temperature scenarios.

The LLNL model shows that for a 6
o
C rise in global mean temperature, there is a 1.5%
increase in precipitation, which gives an increase in mean discharge of 2.4% (from
11,606 m
3
/sec to 11,888 m
3
/sec). As seen from Figure 5.6, the other models (CSIRO9 and
GFDL) fall almost linearly between these two extremes.
1
The Brahmaputra Basin
The changes in mean annual discharge in the Brahmaputra River due to changes in the
mean annual precipitation predicted by the four GCMs are displayed graphically in
Figure 5.7. Two general patterns are evident from the figure. First, a very marked difference
in the mean discharge predicted by the UKTR and GFDL models as compared with the
CSIRO9 and LLNL models. The UKTR and GFDL models show large, step-like increases in
discharge as temperature increases (2
o
C, 4
o
C and 6
o
C). Second, the LLNL model shows a
very small increase in discharge across the temperature changes, while the CSIRO9 model
shows a decrease. The largest increase in the basin-wide precipitation are predicted by the
UKTR model. With a 30.6% precipitation change for a 6
o
C global warming, the mean
discharge may increase from 19,350 m

3
/sec, to 23,069 m
3
/sec. The smallest increase in the
0
2
4
6
Fig. 5.6 Predicted mean annual discharge for the Ganges River at Hardinge Bridge for the four
selected GCMs under global mean temperature increases of 2
o
C, 4
o
C and 6
o
C.
1
The CSIRO9 model scenarios indicate somewhat lower changes in the mean discharge for a 6
o
C
warming, from the current 11,606 m
3
/sec to 16,310 m
3
/sec. Predicted changes for mean annual
discharge for the GFDL is somewhat slightly smaller than for the CSIRO9 model. The mean annual
discharge may increase from 11,606 m
3
/sec to 14,151 m
3

/sec at 6
o
C rise in global mean
temperature.
118 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
basin-wide precipitation are predicted by the LLNL model. With a 4.2% precipitation change
for a 6
o
C global mean temperature change, the mean annual discharge may increase from
19,350 m
3
/sec to 19,863 m
3
/sec (Fig. 5.7).
0
2
4
6
Interestingly, the CSIRO9 model indicates decreases in the mean discharge as
temperature increases (2
o
C, 4
o
C and 6
o
C). This is because the CSIRO9 model predicts a
slight decrease in precipitation in the Brahmaputra basin for increases in global mean
temperature. This reduces the mean discharge of the Brahmaputra by a negligible quantity.
For example, with a 1.5% decrease in the mean precipitation (at a 6

o
C rise in global mean
temperature), the predicted mean discharge (19,335 m
3
/sec) is almost identical to the
current mean (19,350 m
3
/sec).
5.5 EFFECTS ON MEAN PEAK DISCHARGE
Using equation 4 and 5 in Table 5.3, mean peak discharges for the Ganges and Brahmaputra
Rivers were estimated from the mean annual discharges under various climate change
scenarios for the four GCMs (presented in the previous sub-section). For the Meghna,
changes in peak discharges were calculated using equation 3 (Table 5.3) with regard to
changes in the mean precipitation. Then, the relative changes (in percent) were calculated
with respect to the current mean discharges for the Ganges, Brahmaputra and Meghna
Rivers (tabulated in Table 5.7).
5.5.1 THE GANGES RIVER
For the Ganges River, three (CSIRO9, UKTR and GFDL) of the four GCMs indicate
substantial changes in the mean peak discharge under 2
o
C, 4
o
C and 6
o
C changes in global
mean temperature (Fig. 5.8). Among the GCMs, the UKTR model shows the highest possible
change in peak discharge for all temperature scenarios. For a 6
o
C rise in global mean
temperature, the increase in peak discharge is 45.6%. In terms of absolute magnitude,

Fig. 5.7 Predicted mean annual discharge for the Brahmaputra River at Bahadurabad for the four
selected GCMs under global mean temperature increases of 2
o
C, 4
o
C and 6
o
C.
M. M. Q. M
IRZA 119
Copyright © 2005 Taylor & Francis Group plc, London, UK
120 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
this change may increase the mean peak discharge to 76,686 m
3
/sec from the current mean
of 52,680 m
3
/sec. The model predicted value is slightly above the highest recorded peak
discharge (76,000 m
3
/sec) of the Ganges River that occurred in 1987. The LLNL model
shows the lowest possible change in peak discharge across all temperature scenarios
(2
o
C, 4
o
C and 6
o
C). Even at 6

o
C, the change is only 1.5%, giving a peak discharge of
53,628 m
3
/sec compared with the current 52,680 m
3
/sec. The other two models (CSIRO9 and
GFDL) fall between the UKTR and LLNL extremes.
The CSIRO9 model gives the second highest changes. With the same magnitude
in global mean temperature increase (6
o
C), the mean peak discharge increases to
68,017 m
3
/sec (29% increase) from the current mean of 52,680 m
3
/sec. The third largest
increases were predicted in the GFDL model. At a 6
o
C temperature increase, the mean
peak discharge is expected to increase to 60,898 m
3
/sec from the current mean.
5.5.2 THE BRAHMAPUTRA RIVER
Using equation 5 (Table 5.3), mean peak discharges were estimated from the mean annual
discharges under various climate change scenarios for the four selected GCMs. For the
mean annual discharge pattern in Figure 5.9, the mean peak discharge pattern in Table 5.7
and Figure 5.9 shows an enormous difference in outcomes across the four models. Again,
the UKTR and GFDL models indicate large changes in the mean peak discharge while the
CSIRO9 and LLNL models predict negligible changes.

Among the GCMs, the UKTR model shows the highest possible change (17%) in the
mean peak discharge for a 6
o
C global mean temperature rise. With this change, peak
discharge of the Brahmaputra may increase by 13% to 76,022 m
3
/sec from the current
64,866 m
3
/sec (Fig. 5.5). This would be equivalent to the mean peak discharge of the
Ganges River at Hardinge Bridge at a 6
o
C global mean temperature rise for the same GCM
(Fig. 5.9).
The GFDL model gives the next highest changes. With a 2
o
C-6
o
C temperature rise,
the predicted change in the mean peak discharge could be between 4%-12% (in terms of
absolute magnitude, between 67,487 m
3
/sec and 72,728 m
3
/sec) (Fig. 5.9).
0
2
4
6
Fig. 5.8 Predicted mean peak discharge for the Ganges River at Hardinge Bridge for the four

selected GCMs under 2
o
C, 4
o
C and 6
o
C temperature rise.
M. M. Q. M
IRZA 121
Copyright © 2005 Taylor & Francis Group plc, London, UK
The CSIRO9 model shows the smallest changes (negative) in peak discharge across
the temperature scenarios (2
o
C, 4
o
C and 6
o
C). It was mentioned previously that among the
four GCMs, only the CSIRO9 model predicts a decrease in precipitation in the Brahmaputra
basin. Perhaps it was due to inadequate representation of the mountainous topography of
the Brahmaputra basin. As a result, mean discharge and mean peak discharge may
decrease (Table 5.7 and Fig. 5.9). However, even for a 6
o
C rise in global temperature, the
reduction is expected to be slight, only 0.06%. In terms of absolute magnitude, the change
is only 47 m
3
/sec. Such a change would be expected to have negligible or no effect on the
mean flooded area and depth in Bangladesh.
5.5.3 THE MEGHNA BASIN

Most of the GCMs indicate high precipitation changes in the Meghna basin under the
climate change scenarios (Table 5.4). Using these scenarios, changes in the mean peak
discharges were estimated using equation 3 (Table 5.3). The overall results show large
increases in the mean peak discharges for the UKTR and GFDL and small increases for the
CSIRO9 and LLNL models.
The UKTR and GFDL models imply almost equal increases in the mean peak
discharge (19.9% increase) at 2
o
C temperature change (Table 5.7). This may lead to an
increase in peak discharge from the current mean of 14,060 m
3
/sec to 16,861 m
3
/sec
and 16,843 m
3
/sec, respectively for the UKTR and GFDL models. For a 6
o
C global
temperature increase, the peak discharge may increase to 22,470 m
3
/sec and 22,400 m
3
/sec
for the UKTR and GFDL models, respectively (Fig. 5.10).
The CSIRO9 and LLNL models show much smaller increases in the peak discharge
than the other two models. With a 2
o
C temperature rise, the estimated mean peak
discharge for these models is expected to increase to 15,171 m

3
/sec and 15,958 m
3
/sec,
respectively, from the current mean of 14,060 m
3
/sec. At the highest 6
o
C temperature
increase, the peak discharge for the CSIRO9 and LLNL models, show increases to
17,382 m
3
/sec and 19,744 m
3
/sec, respectively.
0
2
4
6
Fig. 5.9 Predicted mean peak discharge for the Brahmaputra River at Bahadurabad for the four
selected GCMs under global mean temperature increases of 2
o
C, 4
o
C and 6
o
C.
122 I
MPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK

5.6 EFFECTS ON DEPTH AND SPATIAL EXTENT OF FLOODING
For mean peak discharges, the Surface Water Modeling Center (SWMC) in Dhaka,
Bangladesh, using the MIKE 11-GIS model, carried out 13 simulations (one for the
“control” run and 12 runs for the “climate change scenarios”). The model covered most of
the Ganges, Brahmaputra and Meghna River basins in Bangladesh, approximately
9.11 million ha. In the three river basins, flood vulnerable area was estimated at
6.72 million ha (BWDB, 1987). This is about 80% of the total area vulnerable to flooding
in Bangladesh.
5.6.1 CHANGES IN MEAN FLOODED AREA
The MIKE 11-GIS model results show that the current mean flooded area is
3.77 million ha (Fig. 5.11) based on the mean discharge of 52,680 m
3
/sec, 64,866 m
3
/sec and
14,060 m
3
/sec for the Ganges, Brahmaputra and Meghna Rivers, respectively, together
with local rainfall in the river basins. The mean flooded area produced by the MIKE 11-GIS
model seems to be very reasonable in relation to observational records (see Section 5.1.1).
With regard to the mean flooded area, the model results indicate three main
outcomes:
• the largest change in flooded area occurs between 0
o
C and 2
o
C;
• there is a clear difference in flooded area outcomes from the UKTR and GFDL
models when compared with the CSIRO9 and LLNL models; and
• the Brahmaputra and Meghna Rivers will play a major role in future flooding.

Surprisingly, the model results indicate that most changes in the mean flooded areas
occur between 0
o
C and 2
o
C in relation to the increases in the peak discharges of the Ganges,
Brahmaputra and Meghna Rivers (Table 5.8 and Fig. 5.12) rather than at higher
temperature increases. In the range of 0
o
C-2
o
C, 2
o
C-4
o
C and 4
o
C-6
o
C increases in
0
2
4
6
Fig. 5.10 Predicted mean peak discharge for the Meghna River at Bhairab Bazaar for the four
selected GCMs under global mean temperature increases of 2
o
C, 4
o
C and 6

o
C.
M. M. Q. M
IRZA 123
Copyright © 2005 Taylor & Francis Group plc, London, UK
temperature, increases in flooded area for per degree warming is
)(
CT
A
o
∆
∆
0.44 mha to
0.55 mha, 0.015 mha to 0.09 mha and 0.015 mha to 0.075 mha, respectively. In general,
increases in peak discharge between 0
o
C-2
o
C will engulf most of the flood vulnerable
areas. Therefore, at higher temperature increases, proportionate increases in discharge will
not be able to increase the spatial extent of flooding as it will possibly be limited by the
elevation of the lands.
The second point to be made from the analyses of the flooded area is that there is a
clear distinction to be seen in the outputs from the UKTR and GFDL models when
compared to the CSIRO9 and LLNL models. The former two models show greater
discharge, and thereby higher flooded area, than the latter (Table 5.8).
Fig. 5.11 Spatial pattern of flood extent and depth for current mean peak discharge.
Results of the inter-model comparison show that, although there is little difference in
results between the UKTR and GFDL models, the UKTR model gives the largest
increases in the mean peak discharge for 2

o
C, 4
o
C and 6
o
C temperature changes.
Consequently, the MIKE 11-GIS model yields the highest changes in the mean flooded
area for the UKTR model. For a 2
o
C temperature increase, the expected change in the
mean flooded area is +29%. This is perhaps caused by higher increases in the peak
discharge of the Ganges River. This helps increase the flooded area in the Brahmaputra
basin by slowing down its drainage at Baruria Transit. The change is expected to be +39%
for a 6
o
C temperature rise (Fig. 5.13). For the GFDL model, the changes are 28% and 37% in
the flooded area, respectively.
124 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
Table 5.8 Area (in mha) inundated under 2
o
C, 4
o
C and 6
o
C temperature increases for the four GCMs
Fig. 5.12 Changes in the combined mean discharges of the Ganges, Brahmaputra and Meghna
Rivers (under control and climate change scenarios) and the mean flooded areas. Values within boxes
indicate changes for a 2
o

C rise in temperature.
It can be seen that the CSIRO9 model results indicate the lowest possible changes in
the flooded area, although these changes are not greatly different from the LLNL model.
For a 2
o
C increase in temperature, the mean flooded area increases by 23% under the
CSIRO9 model. For the higher temperature increases, the changes are negligible. For
example, for a 6
o
C temperature increase, the mean flooded area increases by 25%, which
in absolute terms, is only 0.06 mha more than the area expected to be inundated under a
2
o
C temperature rise. For the LLNL model, with a 2
o
C and 6
o
C temperature rise, the
flooded area may change by 24% and 27%, respectively. In absolute terms, the difference
between flooded areas under these two scenarios is only 0.1 mha (Table 5.8). For the LLNL
model, under the three global mean temperature scenarios, changes in the flooded area are
largely produced by the peak discharges of the Meghna River. Changes in peak discharges
for the Ganges and Brahmaputra are negligible.

Model
Mean Flooded Area
0
o
C 2
o

C 4
o
C 6
o
C

CSIRO9

3.77

4.65

4.68

4.71
UKTR 3.77 4.87 5.08 5.24
GFDL 3.77 4.84 5.02 5.17
LLNL 3.77 4.68 4.73 4.78

M. M. Q. MIRZA 125
Copyright © 2005 Taylor & Francis Group plc, London, UK
The third point to emerge from the analysis of flooded areas is that the Brahmaputra
and Meghna peak discharges play a major role in flooding. This can be seen from an
inspection of Figures 5.12 and 5.13, which compare flooded areas from mean peak values
under current temperature with that of the 6
o
C scenarios for the UKTR model. The peak
discharge of the Ganges slows down the drainage of the Brahmaputra River through the
Baruria Transit. This helps to increase the spatial extent, depth and duration of flood in the
Brahmaputra basin, and thus the Brahmaputra water cannot be drained out quickly.

Further downstream in Chandpur, the combined flow of the Ganges and Brahmaputra
obstructs drainage from the Meghna basin. This phenomenon creates problems in the
Meghna basin similar to those of the Brahmaputra. A significant backup of water through
the Meghna basin lasts until gradients are established which allow the drainage of flood
water from the Meghna basin.
Note that under the 2
o
C, 4
o
C and 6
o
C temperature rise scenarios, both the CSIRO9
and LLNL models imply slight decreases in the Brahmaputra peak discharge (Table 5.7).
Although, for the peak discharges of the Ganges, the CSIRO9 model implies changes
many times greater than those for the LLNL model, slight increases in the flooded area for
the LLNL model were perhaps caused by the increases in the peak discharge of the Meghna
River.
5.6.2 CHANGES IN THE INUNDATION CATEGORIES
Under the climate change scenarios, four selected GCMs indicate substantial changes in
the land inundation categories F
0
, F
1
, F
2
and F
3
(Tables 5.9 and 5.10 and Fig. 5.14).
Fig. 5.13 Spatial pattern of flood extent and depth for the UKTR model (6
o

C rise in global mean
temperature).
126 IMPLICATIONS ON RIVER DISCHARGE IN BANGLADESH
Copyright © 2005 Taylor & Francis Group plc, London, UK
In response to increased peak discharge, mean flooded area may increase significantly.
However, the rate and net changes in the mean flooded area is expected to be higher than
that of the flooded area of higher return periods. The results of simulations for the mean
and 20-year flooded area are discussed below.
The analysis of inundation categories for the 12 model simulations indicate that:
• drastic changes in most of the inundation categories may occur between 0
o
C and
2
o
C global mean temperature rise;
• rates of change are expected to be smaller with higher temperature increases;
• at 2
o
C and 4
o
C temperature changes, the UKTR and GFDL models show similar
changes in the medium flood category;
• at 2
o
C, 4
o
C and 6
o
C temperature changes, the CSIRO9 and LLNL models show a
similar pattern of change for all flood categories;

• at 2
o
C, 4
o
C and 6
o
C temperature rises, the UKTR model shows the highest changes
in the deep flood category;
• under a 6
o
C temperature rise, most of the mean flooded areas may be deeply
flooded in Bangladesh;
• land area under prolonged inundation (<9 months) may increase;
• changes in the inundation categories may result in reduced cropping intensity in
Bangladesh; and
• as a result of changes in the inundation categories, the agricultural sector of
Bangladesh may suffer substantial loss of land productivity.
Table 5.9 Various land classes in Bangladesh

Land Type of
Inundation Class
Range of Inundation
Depth
Crop Suitability


Highland (F
0
)
Less than 30 cm

(Flood Free)
Land suited to HYV T.
aman
in wet season, wheat
and HYV
boro
in
rabi
season

Medium Highland (F
1
)
30 cm to 90 cm
(Shallow Flooded)
Land suited to local
varieties of
aus
and T.
aman

in wet season;

wheat and

HYV
boro
in
rabi
season

Medium Lowland (F
2
)
90 cm to 180 cm
(Moderately Flooded)
Land suited to B.
aman
in
wet season and wheat and
HYV
boro
in
rabi
season

Lowland (F
3
) greater than 180 cm
(Deeply Flooded)
Land suited to B.
aman
in
wet season and HYV
boro

in
rab
i

season

Source: MPO, 1987.
M. M. Q. M
IRZA 127
Copyright © 2005 Taylor & Francis Group plc, London, UK