Quaternary International 304 (2013) 142e155

Contents lists available at SciVerse ScienceDirect

Quaternary International
journal homepage: www.elsevier.com/locate/quaint

Assessing morphological changes of the Ganges River using satellite

images
Md Altaf Hossain, Thian Yew Gan*, Abul Basar M. Baki
Dept. of Civil & Environmental Engineering, University of Alberta, NREF 3-033, Edmonton, AB T6G 2W2, Canada

a r t i c l e i n f o a b s t r a c t

Article history: Using eight dry season satellite images of Landsat MSS (1973e1984), Landsat TM (1993e2003), and IRS LISS
Available online 27 March 2013 (2009), this study assessed morphological changes of the Ganges River within Bangladesh. As a typical al-
luvial river, morphological changes of the Ganges River had been a common phenomenon. In the 1973e2009
study period, with its morphologically dynamic nature, the Ganges River had undergone considerable
changes in its river banks and its vegetated islands. The bankline movement rates (m/y)and erosioneac-
cretion rates (ha/y) of the river were analyzed under four time periods: 1973e1984,1984e1993,1993e2003,
and 2003e2009. The results indicate that both the left and the right banks of Ganges have changed signif-
icantly due to varying erosion and accretion rates that had occurred. On a whole, the left bank was more
prone to accretion while the right bank to erosion. The magnitude of bank erosion along the Ganges River is
closely related to the erodibility of riverbank materials which vary spatially. In 1973e1984, the average river
width and the overall islands area had decreased considerably due to a net river bank accretion which
resulted in an increase of 6790 ha of land area. However, afterwards, both river width and islands area
continue to increase to the end of this study period (2009) because of a net erosion (loss of 18,830 ha of land).
Furthermore, from 1984 to 2009, the sinuosity of Ganges had generally increased and was related to
increased river width caused by a net bank erosion. Compared to other major rivers, Ganges has a very high
erosion rate probably because its river banks are primarily composed of highly erodible materials and its high
discharge rate especially during the Monsoon season.
Ó 2013 Elsevier Ltd and INQUA. All rights reserved.

1. Introduction sometimes human-induced channel changes often create various

socio-economic and environmental problems for society, often
Bangladesh is a riverine country with more than 7 percent of its leading to undue hardship to people living along such rivers and
lands occupied by river systems (Hossain et al., 2008). The bed and their islands. Therefore, a better understanding on morphological
banks of major rivers of Bangladesh which include Jamuna, Ganges, changes of alluvial rivers, particularly channel changes through
Padma, and the Lower Meghna rivers primarily compose of alluvial erosion and accretion processes, as well as techniques to detect
materials, with their channels characterized by highly dynamic such changes would be useful for effective planning and manage-
behavior (CEGIS, 2003). Floodplains of these alluvial rivers are ment of these alluvial environments.
predominantly formed with flood-related sediments while their Similar to other major rivers of Bangladesh, the Ganges River is
bank materials consist mostly of fine-grained and cohesive sedi- also dynamic within its morphologically active corridor formed by
ments (Azuma et al., 2007). In such alluvial rivers, through the river over many centuries (Halcrow et al., 1993). Because of
continuous erosioneaccretion processes, the channels frequently bank erosion and accretion, planform evolution, major lateral
change from reach to reach (Kammu et al., 2008). In addition, de- shifting and channel width adjustments generally occur rapidly in
velopments such as building of bank protection structures on river the Ganges River (CEGIS, 2003). This severe nature of riverbank
banks, artificial cutoffs, construction and operation of barrages, erosioneaccretion has brought untold sorrow to the densely
dams and land use changes also modify natural geo-morphological populated country of Bangladesh. Every year, thousands of flood-
dynamics of these river systems significantly (Surian, 1999; Fuller plain dwellers of Bangladesh become landless and homeless due to
et al., 2003; Rinaldi, 2003). The aforementioned natural and riverbank erosion. Other than these direct social and economic
impacts, the risk of frequent and rapid river bank changes also
* Corresponding author. result in a significant constraint against further development of
E-mail address: tgan@ualberta.ca (T.Y. Gan). public and private enterprises along the river banks.

1040-6182/$ e see front matter Ó 2013 Elsevier Ltd and INQUA. All rights reserved.
http://dx.doi.org/10.1016/j.quaint.2013.03.028
 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155 143

In recent decades, various studies had been conducted on the sensitive trans-boundary issue between the two countries
morphology of major rivers using remotely sensed images, e.g., (Fig. 1). Because the banklines of the river channel changes
Kammu et al. (2008) for the Mekong River; Chu et al. (2006) for the frequently along the Indo-Bangladesh border, fixing the admin-
Yellow River; Wang and Lu (2010) for the Lower Yangtze River; istrative boundary between the two countries has been contro-
Sarker and Thorne (2006) and Sarker et al. (2003) for major river versial. In particular, three locations (upstream of Godagari,
systems of Bangladesh; and Thorne et al. (1993), ISPAN (1993), and opposite side of Godagari downstream and opposite side of
Baki and Gan (2012) for the BrahmaputraeJamuna River. A number Rougtha downstream) along the border between India and
of studies had been conducted on the Ganges River, such as the geo- Bangladesh (Fig. 1) have experienced a growing bank erosione
morphological characteristics of the Ganges River within accretion problem over the last several decades. For a long time
Bangladesh by Hossain (1989), the hydro-morphological behavior these three locations have become the most unstable reaches of
of Ganges by Halcrow et al. (1993) under the framework of Ban- the river, highly vulnerable to erosion. According to Sarker
gladesh’s Flood Action Plan (FAP), the morphological aspects of (2004), the average annual maximum erosion rate of these
Ganges by ISPAN (1995) using satellite images, and the dimensions three locations was 315 m/y between 1980 and 1999.
of Ganges and its morphological behavior around the off-take of the Before entering into Bangladesh, the river splits into two
Gorai River by Delft Hydraulics and DHI (1996a, 1996b). branches at Farakka, called Hooghly-Bhagirathi and Ganges and
Many recent studies conducted on the Ganges River are based they both empty into the Bay of Bengal but through different routes
on satellite images, such as the river and char (island) dynamics (Fig. 1). The main left arm enters Bangladesh about 18 km below
(EGIS, 2000), and morphological evolution (CEGIS, 2003) of Ganges. Farakka and flows toward the south-eastern direction before
Studies regarding human impact on rivers are such as the meeting with Brahmaputra (Jamuna) River at Aricha (Fig. 1), while
morphology of the Ganges-Gorai system under human intervention the right arm continues flowing south to West Bengal, India as
by Sarker (2004), and morphology of Ganges under the influence of Bhagirathi-Hooghly. Soon after it enters into Bangladesh, the main
natural and man-made hard points by Hossain (2006) based on branch of Ganges forms the common boundary between India and
satellite images of 1973e2003, borehole, and stage discharge data. Bangladesh for about 110 km. The total length of the river along the
The objective of this study is to detect and assess morphological centre line for this study is about 220 km (depending on the sin-
changes of the Ganges River from 1973 to 2009 by analyzing multi- uosity) and the valley length is about 190 km starting from the
temporal satellite images using GIS and remote sensing techniques. Indo-Bangladesh border up to the confluence with Brahmaputra
This study focused on quantifying morphological changes in-terms (Jamuna) River at Aricha. The flow of the Ganges is highly seasonal
of bankline shifting, erosion/accretion rates, river width, sinuosity, with most of the peak flow occurring during the monsoon (Junee
and islands area over the periods of more than 30 years. From September), e.g., the flow ratio between the dry (JanuaryeMay) and
quantifying the spatial extent and rates of bankline changes and the monsoon seasons at the Hardinge Bridge in Bangladesh is about
islands area via remote sensing, the purpose is to better understand 1:6 (Mirza and Dixit, 1997). Fig. 2a shows the time series of the
the erosion and accretion processes of Ganges River from the mean annual discharge (1973e2009) and Fig. 2b the average
perspective of channel width, sinuosity, and islands, and the effects monthly water levels and discharge (1973e2009) of the Ganges
of human intervention such as the implementation of hard points River at the Hardinge Bridge. The 3-year moving average line shows
on the River. Results of this study will contribute to our under- that since early 1973 the discharge of the Ganges River had been
standing and our ability to predict the morphological behavior of decreasing until about 1982 when it had been rising until 1987,
the Ganges River, which will help to mitigate potential human then declined till 1994, and again increased until 2003 (Fig. 2a).
suffering and adverse socio-economic impacts expected from After 2003, the mean annual discharge declined sharply by about
bankline and island changes of the Ganges River. 50% in 6 years. As expected, the mean annual water level varied in a
manner similar to that of the mean annual discharge. Between 1973
2. Ganges River and 2009, the maximum and minimum discharge rates recorded
for the Ganges was 78,000 m3/s and 800 m3/s, respectively. The
The Ganges River, one of the large anabranching rivers in the highest and lowest water levels recorded between 1973 and 2009
world (Kleinhans, 2010), drains a basin area of more than were 15.19 m þ PWD (Public Works Department of Bangladesh)
1.1 million km2 which covers over four countries: China, Nepal, and 4.22 m þ PWD, respectively (annual average ¼ 8.65 m þ PWD).
India and Bangladesh (Mirza, 2004). In terms of basin area, India The average water level during monsoon and dry seasons was
has the largest share of 79% while Bangladesh and China each have 11.35 m þ PWD and 6.39 m þ PWD, respectively, which means a
a share of about 4% and the remaining 13% is in Nepal (Mirza, 2004). range of 4.96 m in water level between these two seasons. There-
This river originates in the Gangotri glacier at an elevation of about fore, the planform of the monsoon season is quite different from
7010 m on the southern slope of Himalayas and traverses south and that of the dry season. The average water level slope of the Ganges
south-eastward in India for about 2000 km before it reaches the River within Bangladesh is about 5 cm/km (EGIS, 2000).
Indo-Bangladesh border (Adhikari et al., 2000). The hydrologic The bed material of the river consists mainly of fine sand with a
cycle and water resources of the Ganges basin are mainly governed median diameter (D50) of about 0.14 mm, but along the down-
by the southwest monsoon which usually begins in early June (mid- stream direction it generally decreases in size at about 0.03 mm/
July) in the southern (northern) part of the Ganges basin. The 100 km (Delft Hydraulics and DHI, 1996a). By suspended sediment
climate of the basin varies from semi-arid in the west to humid in transport as the dominating mode of bed material transport, the
the east, with a mean annual precipitation ranging from 450 mm to river carries about 300e500 million tons of sediment every year, of
2000 mm (basin average 1070 mm), of which about 84% of pre- which about one third is fine sand and the remaining two-third is
cipitation occurs during the monsoon season e June to September silt and clay (Delft Hydraulics and DHI, 1996b). Bore hole data
(Mirza, 2004). analysis along the Ganges River revealed that clay layers of various
color with a thickness of about 5e20 m were found in different
2.1. Ganges River within Bangladesh locations in the left bank (Hossain, 2006). Along the right bank in
Bangladesh, the presence of any substantial layers of clay is much
As the Ganges River forms a significant length of the border less common than along the left bank (Hossain, 2006). Borehole
between Bangladesh and India, the river itself has been a data for upper reaches of the Ganges are not available. However,
144 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155

Fig. 1. Location of the Ganges study area.

(a) 16000
according to Umitsu (1993), the sediment in this area was probably
13 deposited within the last 5000 years through delta formation and
eastward shifting of the Ganges.
Water level (m PWD)

12000
Discharge (m3/s)

11 Within Bangladesh, as a meandering river characterized by

sweeping meanders, the flow of Ganges is divided within its
8000 meander belt in which the planform varies temporarily and
9
spatially (CEGIS, 2003). However, sweeping meanders and forma-
tion of a braided belt are limited within the active corridor of this
4000 7
Mean annual discharge river (EGIS, 2000), where the boundaries consist mostly of much
Mean annual water level
Moving average older, weathered and cohesive materials (Sarker, 2004). Within the
0 5 active corridor, frequently re-worked sediments are loose and un-
consolidated and so are highly susceptible to erosion. Under these
19

19

19

19

19

19

19

19

19

20

20

20

circumstances, and together with a strong seasonal fluctuation of

07
72

75

78

81

84

87

90

93

96

01

04

Year
(b) river discharges, the Ganges River is subjected to frequent and
40000 16.0 significant erosion and accretion to be analyzed below.
Discharge
35000 14.0
Water Level
3. Data and analysis methods
Water Level (m+PWD)

30000 12.0
Discharge (m3/s)

25000 10.0
3.1. Satellite and water data
20000 8.0

15000 6.0 Multi-temporal Landsat and one IRS-LISS satellite images ac-
10000 4.0 quired by optical sensors during the dry season (JanuaryeMarch) of
5000 2.0 Bangladesh (see Table 1) from 1973 to 2009 were supplied by CEGIS
(Center for Environmental and Geographic Information Services) of
0 0.0
Dhaka, Bangladesh. Only satellite images acquired during the dry
Ja

Ju
Fe

Ap

Ju

Au

Se

season were selected because during dry season, vegetation cover

ar

ct

ov

ec
ay
n

l
b

Date and other ground conditions, particularly the water level, are
Fig. 2. (a) Variation of mean annual discharge and water level (1973e2009) and (b)
relatively consistent from year to year which is essential for
average monthly water levels and discharge (1973e2009) of the Ganges River at the assessing the inter-year change of erosion and accretion of Ganges
Hardinge Bridge. River. In addition, during dry season the chances of getting a
 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155 145

relatively cloud free atmosphere will be higher and the planform For each image transformation, as a measure of the accuracy of the
generally shows the boundary and pattern of channels within the geo-referencing procedure, the root mean square error (RMSE) sta-
braid belt clearly. Seven scenes of Landsat and one scene of IRS-LISS tistics was calculated. The RMSE of the image transformation was not
(1973, 1980, 1984, 1993, 1997, 1999, 2003 and 2009) covering the permitted to exceed the size of a pixel, which for Landsat-MSS data is
Indo-Bangladesh border to Aricha (confluence point of Ganges and 80 m, for Landsat-TM is 30 m, and for IRS LISS III is 24 m. All satellite
Brahmaputra/Jamuna River) were selected for this study. To relate images were projected onto the Bangladesh Transverse Mercator
erosion and accretion processes to streamflow data, water level and (BTM) projection system, whose specifications are: (1)
discharge rates at the Hardinge Bridge point during 1973e2009 Ellipsoid ¼ Everest 1830, (2) Projection ¼ Transverse Mercator, (3)
were also used in this study. Central meridian ¼ 90 E, (4) False easting ¼ 500,000 m, and (5) False
northing ¼ 2,000,000 m (ISPAN, 1992).

Table 1 3.3. Satellite data analysis

Satellite images used for this study.

Satellite data Acquisition date Band number The primary task of this study was to delineate banklines and
Landsat-MSS (80 m  80 m) February 21, 1973 1, 2, 3 boundaries of vegetated islands from all geo-referenced images. A
February 21, 1980 bankline, defined as the feature that separates the outer margin of a
February 25, 1984 river channel from the floodplain, was delineated manually ac-
Landsat-TM (30 m  30 m) March 11, 1993 2, 3, 4 cording to the criteria outlined in EGIS (1997) and Hasan et al.
February 18, 1997
(1997). All selected satellite images were carefully analyzed for
January 19, 1999
January 26, 2003 banklines and boundaries using the ArcViewÒ GIS software at a
IRS LISS III (24 m  24 m) January 15, 2009 scale of 1:50,000.
To quantify changes in river bank locations that have occurred
between any two images using ArcView, a total of 381 cross-sections
3.2. Satellite images at 0.5 km intervals along the 190 km long study reach of the Ganges
were drawn and coordinates of all intersection points between the
All satellite images were processed and analyzed by CEGIS of banklines and cross-section lines determined (Fig. 3). Usually, the
Dhaka, Bangladesh using image analysis and GIS software such as sections used to measure the banklines shifting are set to be at 90 to
ERDAS and ArcView. These images were geo-referenced on the basis the axis of the channel (EGIS,1997; Baki and Gan, 2012). However, for
of the 1997 Landsat Image mosaic of Bangladesh, which itself was a channel undergoing meandering and erosion, the channel tends to
geo-referenced against “SPOT photo maps”, produced in 1989e1990 change its flow direction frequently year after year, and so it will be
from multi-spectral SPOT images, printed at a scale of 1:50,000. In difficult to use sections at 90 to the axis of the channel to predict the
geo-referencing, a set of ground control points (GCPs) was collected banklines shifting. In other words, for the Ganges River subjected to
from the geo-referenced 1997 image mosaic. For each GCP, the map frequent and significant erosion/accretion, if we were to set the
coordinates were obtained from the image mosaic of 1997 and cross-sections at 90 to the axis of the channel, the locations of the
entered into a data file together with the file coordinates of the same cross-sections would shift from year to year. To avoid such a prob-
GCP identified in the digital satellite image. GCPs were taken from lem, we set the river cross-sections at 90 to the valley direction. By
recognizable, permanent features such as edges of large bodies and so doing, based on changes to the cross-sections detected from the
road intersections, and are well dispersed for accurate rectification. satellite images, we would be able to consistently track temporal

N
Bankline 1st year

Sections across river

at 0.5 km intervals

Accreted area

Eroded area

Bankline 2nd

Intersection points

Channel 1st year

Fig. 3. A schematic diagram showing how the riverbank movement from 1973 to 2009 was computed. Sections are taken at 0.5 km intervals along the 190 km valley length of the
Ganges River.
146 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155

changes to left and the right banklines. A similar method was used by those that “migrate upon their floodplains via organized, pre-
Kotoky et al. (2005) to estimate the erosion and deposition rates for dictable arrangements of net erosion and deposition associated
the upper reach of Brahmaputra River within India. For the with a single set of bars on alternate sides of the channel, main-
meandering problem of the Mekong River, Kammu et al. (2008) taining statistically similar channel dimensions” (Eaton et al.,
estimated the banklines shifting over the years with reference to 2010). They have implications with respect to the stability of the
selected permanent points on both banks. inter-channel islands (chars) and with respect to the potential for
From the co-ordinates of two intersection points on a single rapid shifts due to avulsion. The results on the morphological
cross-section line, linear change of the position between two changes of the Ganges River, an anabranching system (according
banklines was calculated. In this analysis, a negative value was to the criteria given in Latrubesse (2008) and Eaton et al. (2010)),
interpreted as erosion, while a positive value as accretion and a are discussed below.
very low or zero value as stable areas. Changes to river banklines
were performed for four time periods, 1973e1984, 1984e1993, 4.1. Erosion and accretion of the Ganges River (1973e2009)
1993e2003, and 2003e2009. The total eroded area between any
two time periods were calculated by overlaying banklines of two 4.1.1. Bankline shifting
respective years. Bands 1, 2, and 3 of Landsat-MSS and bands 2, 3, The bankline shifting rates due to erosion and accretion for the
and 4 of both Landsat-TM and LISS III images that give a colour Ganges River for four time intervals between 1973 and 2009 are
composite of visible-infrared bands useful for land/water delinea- shown in Table 2, and rates of bankline shifting with respect to
tion were used in this study. The river width was calculated from distance from the Indo-Bangladesh border for the left and right
the distance between two co-ordinates of intersection points on banks are plotted in Fig. 4a and b, respectively. These figures show a
both banks along a single cross section line. The sinuosity was considerable movement of both banklines resulted from accretion
calculated from a ratio of path length to direct length. Islands (“riverward”) as well as erosion (“landward”) between 1973 and
within the river system that are distinguishable from the satellite 2009. From Table 2 and Fig. 4a and b, it seems that overall the left
images were also identified in the study. Changes to river bank was more prone to accretion than to erosion, while the
morphological attributes such as river width, sinuosity, and islands opposite happened to the right bank. Between 1973 and 2009,
area were quantified for images of 1973, 1980, 1984, 1993, 1997, bankline shifting rates could vary from less than a m per year to
1999, 2003, and 2009. several hundred meters per year, as is evident in high standard
Given the resolution of images of Landsat-MSS, Landsat-TM and deviations (values in parenthesis) shown in Table 2. A previous
IRS LISS III, our assessments of changes of erosion and accretion work on the Jamuna River has established that the presence of
along the river banks are approximate and subjected to an order of sedimentary features downstream of a bend was indicative of a
accuracy comparable to the image resolutions. In some locations high rate of bank erosion (EGIS, 2002). For 1973e2009, the average
along the Ganges River, the bank erosion and accretion are not left bank erosion and accretion rates were 54 m/y and 105 m/y
recognizable in the Landsat-MSS image of 80 m  80 m resolution. respectively, while the corresponding rates for the right bank were
Although Landsat-TM images (30 m resolution) are typically more 78 m/y and 71 m/y, respectively.
useful than Landsat-MSS images in mapping homogeneous, near-
urban land covers, at times they could be less useful in heteroge- Table 2
neous urban areas (Haack et al., 1987). Errors associated with the Rates of bankline shifting due to erosion (“landward movement”) and accretion
estimation of bankline changes based on remotely sensed data will (“riverward movement”) at different time intervals of 1973e2009.
partly depend on the spatial resolution of satellite images, such that Year Left bankline shifting Right bankline shifting
coarser resolution images tend to result in more errors in esti-
Erosion Accretion Erosion Accretion
mating the bankline migration of any river system, and vice versa. (m/y) (m/y) (m/y) (m/y)
Other possible sources of error include visual delineation of bank-
Avg. Max. Avg. Max. Avg. Max. Avg. Max.
lines, differences in water levels between images and other sys-
tematic or random errors. However, the overall resultant errors 1973e1984 63 (86)a 332 169 (236) 760 92 (84) 395 163 (152) 459
1984e1993 61 (91) 353 62 (142) 676 122 (132) 569 83 (109) 464
should be relatively small compared to the overall, sizeable changes 1993e2003 58 (120) 604 141 (253) 848 63 (86) 476 24 (26) 91
in the river banks of Ganges River. 2003e2009 32 (55) 310 46 (65) 287 36 (68) 447 13 (21) 158

1973e2009 54 (88) 105 (174) 78 (93) 71 (77)

4. Discussions of results
a
Values in the parenthesis indicate standard deviation computed from 381 rea-
ches of 0.5 km intervals shown in Fig. 3.
According to Kleinhans (2010), Ganges is a large, anabranching
river system, which is either a single-channel system with recently
abandoned flood-conveying branches or more commonly, multi- Fig. 4a shows that between 1973 and 1984, the left bank was
channel systems in which the inter-avulsion period is larger relatively stable except at four locations where banklines shifted
than the avulsion duration. Other than the lower Mississippi, significantly in both directions. On the other hand, for this period,
Latrubesse (2008) said that all mega-rivers, typically with a the entire right bank was unstable with its bankline shifting
discharge larger than 17,000 m3/s, a high river width/depth ratio haphazardly at different magnitudes (Fig. 4b). The highest accretion
frequently exceeding 200, low slope <0.00015, and high bed load, rate (riverward movement) of the left bankline was about 8.40 km/
have anabranching channel patterns. They do not satisfy the usual y at Pakka Narayanpur. On the right bank, the river was eroded
discriminators between braiding and meandering (Kleinhans, (landward movement) more than 4.30 km/y at certain locations,
2010), possibly because there exists a stable state of anabranch- which might be related to the immergence of a big island just
ing with large stable vegetated islands that do not seasonally opposite of this bend.
adjust with annual flow variability (Latrubesse, 2008). Eaton et al. Between 1984 and 1993, significant bankline shifting occurred
(2010) proposed theoretical thresholds based on power laws at two locations in the left bank. As shown in Fig. 4a, at Pakka
which relate critical slope to discharge and bank strength to Narayanpur the left bankline encroached the floodplain by about
discriminate multiple-thread channels that are stable anabranch- 3.18 km (353 m/y) while near Aricha the bankline advanced river-
ing channels from unstable, braided channels. The former are ward for over 6 km (676 m/y). The reason for the major erosion at
 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155 147

Fig. 4. Changes to erosion and accretion rates along the (a) left bank and (b) right bank of the Ganges between 1973 and 2009.

Pakka Narayanpur was likely due to unconsolidated bank material bank had two locations where accretion tended to dominate over
deposited earlier (prior to 1984) by accretion that occurred at Ari- erosion.
cha. At the same time period, the shifting of the right bankline The relationships between bankline shifting due to erosion and
mainly happened at the river reach running from 35 km to 110 km accretion and the annual discharge are presented in Fig. 5. The cor-
(Fig. 4b). Within this river reach, the shape of the bankline curve relation between left and right bank erosion rates and annual dis-
fluctuates more or less as a sinusoidal function (Fig. 4b), which charges were 0.60 and 0.61, respectively. In contrast, the correlation
means that between 1984 and 1993, erosion and accretion occurred between left and right bank accretion rates and annual discharges
more or less alternatively along the right bank. were only 0.01 and 0.07, respectively. As expected, the correlation
Between 1993 and 2003, the left bank of the Ganges River shifted between both banks’ erosion rates and the annual discharge is
substantially due to erosion that occurred along the river reach higher than that with accretion because erosion is primarily a
starting from the Indo-Bangladesh border up to about 20 km function of the annual discharge rate while accretion is primarily a
downstream and accretion thereafter for another 15 km at God- function of the deposition of suspended sediment load. For the large
agari. During this period, serious erosion resulted from rapid braided Jamuna River, Baki and Gan (2012) found that the correlation
streamflow washing away loosely packed material at Pakka Nar- between bank erosion/accretion rates and annual discharges were
ayanpur and depositing massive sediment just downstream of the 0.26e0.54 for short-term and 0.01e0.44 for log-term changes.
eroded bank. As a result, in Pakka Narayanpur, the Ganges River Apparently on a long-term basis, this relationship can vary widely
shifted more than 6 km into the floodplain at one location while the mainly because these river systems in Bangladesh are subjected to a
downstream location extended by about 8.5 km into the river fairly wide range of interannual climate deviations (FAO, 2007).
channel (Fig. 4a). Along the right bank, the bankline shifting was
more erratic with varying degrees of erosion and accretion, and the 4.1.2. Area of erosion and accretion
maximum “intrusion” into the floodplain was over 4.75 km (Fig. 4b). Summary statistics and graphical representation of riverbank
In the last study period (2003e2009), other than river reaches changes of the Ganges due to erosion and accretion for 1973 to
from 10 km to 35 km on the left bank and between 65 km and 2009 are presented in Table 3 and Fig. 6, respectively. Fig. 6
90 km on the right bank exhibited significant bankline shifting, generally shows a high spatial variability of erosion and accre-
both river banks were relatively stable compared to the previous tion on both sides of the riverbank. From Table 3, it can be seen
three study periods. Apparently for the whole study period, two that erosion rate (ha/y) gradually increased from 1973 to 2003
locations for the left bank, the river reach starting from the Indo- and then decreased during 2003e2009 on the left bank which
Bangladesh border up to Godagari and that at the end of the had mostly occurred on the side of Bangladesh. However, erosion
study reach near Aricha were very unstable. On the other hand, the rate on the right bank decreased both in Bangladesh and in the
right bank exhibited fairly erratic behavior along almost the entire Indian side during the whole study period. In terms of accretion
study reach. Overall, the right bank of the Ganges River had un- rate, the left bank had not exhibited any consistent trend
dergone more erratic shifting towards the floodplain while the left throughout the study period while for the right bank, accretion
148 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155

Fig. 5. The regression relationships between annual discharge and left/right banks erosion/accretion rates with a 90% confidence limit.

occurred very rapidly in 1973e1993 but it slowed down in 2003e Ganges River. Similar to river bank erosion rates, the average
2009, in both the Bangladesh and the Indian side. Fig. 7 shows eroded area showed a relatively high correlation (R ¼ 0.76) with
the regressional relationship between the average eroded/ the mean annual discharge of the Ganges than that of the average
accreted area (ha/y) and the mean annual discharge of the accreted area (R ¼ 0.09).

Table 3
Erosioneaccretion along the Ganges River for four study periods.

Duration Location Erosion Accretion

Total (ha) Rate (ha/y) Total (ha) Rate (ha/y)

1973e1984 (11 years) Left bank Bangladesh 5110 465 18,040 1640
India ea e e e
Right bank Bangladesh 7894 718 5689 517
India 6840 622 2906 264
Total reach 19,845 1804 26,634 2421
1984e1993 (9 years) Left bank Bangladesh 5777 642 3971 361
India e e e e
Right bank Bangladesh 6223 691 2983 331
India 5485 609 3808 423
Total reach 17,485 1943 10,762 1196
1993e2003 (10 years) Left bank Bangladesh 7603 760 7241 724
India 353 35
Right bank Bangladesh 4866 487 482 48
India 5549 555 157 16
Total reach 18,018 1802 8233 823
2003e2009 (6 years) Left bank Bangladesh 2013 336 1710 285
India 50 8 501 84
Right bank Bangladesh 2259 377 349 58
India 708 118 161 27
Total reach 5030 838 2710 454
a
The left bank of Ganges River was totally within the Bangladesh territory in 1973e1993 period and that’s why there was no erosion or accretion happened to the left bank
of Ganges in India.
 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155 149

Fig. 6. River bank changes along the Ganges River for 1973e1984, 1984e1993, 1993e2003, and 2003e2009, respectively.

From 1973 to 1984, the entire reach of Ganges (both left and been 1943 ha/y and 1196 ha/y, respectively, which are similar to
right banks) eroded 19,845 ha but accreted 26,634 ha, and so there 2240 ha/y and 1010 ha/y estimated by EGIS (2000). Therefore,
was a net gain of 6790 ha of new land, and all of which occurred on differences between results of this study and that of EGIS (2000)
the Bangladesh side, mainly at Pakka Narayanpur (Fig. 6). Between are marginal even though different analysis methods had been
1984 and 1993, erosion was more dominant than accretion on used. From 1993 to 2003, erosion dominated over accretion
both banks of the river. The total amount of land eroded during especially for the right bank on the Indian side, where most of the
this period was about 17,485 ha compared to an accretion of about net floodplain loss of about 9785 ha had occurred. However,
10,762 ha and so there was a net loss of 6723 ha of land. Among all in contrast to the first three study periods, in 2003e2009 the
the land eroded, more than 75% occurred in Bangladesh while the Ganges River was found to be relatively stable with a relatively
remaining 25% occurred in India. For 1984e1993, the average bank small amount of erosion and accretion occurring on both banks.
erosion and accretion for the entire study reach estimated had This was possibly because the mean annual discharge of Ganges
150 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155

Fig. 7. Regression relationships between the reach average eroded/accreted areas versus annual average discharge of the Ganges River with a 90% confidence limit.

River had declined sharply after 2003 by about 50% in 6 years forth in a sinusoidal manner (CEGIS, 2003). Previous studies
(Fig. 2a). (Halcrow et al., 1993; CEGIS, 2003; Hossain, 2006) on the Ganges
In summary, for the entire study reach, the erosion rate River have identified a brown and gray clay layer with a thickness of
increased by about 8% in the second study period (1984e1993) as 15e30 m overlying brown silt and fine sand along the boundary of
compared to the first (1973e1984), but from the second to the third the active corridor that are resistant to erosion. The presence of
study period (1993e2003) it decreased by more than 7%, and then cohesive material, however, is not continuous along both margins
it decreased by more than 53% between the third and the final of the active corridor. On the left bank, cohesive materials which
study periods. In contrast, between 1973e1984 and 1984e1993, the include consolidated and weathered Pleistocene aged sediments
accretion rate decreased by about 51% and it further decreased by are very common, but they are sparse along the right bank (Sarker,
more than 31% from 1984e1993 to 1993e2003, and from which to 2004). In this study, those places with such sediments are referred
2003e2009, accretion decreased by about 45%. Therefore, in to as natural hard points because they are relatively resistant to
contrast to the past, under a drier climate, the Ganges River has erosion, as shown in red in Fig. 8. In addition to natural hard points,
become relatively stable in recent years. several man-made protection structures have been constructed as
an effort to minimize erosion along the Ganges River over the last
4.2. Impact of natural hard points on bank erosion 100 years for both the Bangladesh and the Indian sides (Hossain,
2006). Unfortunately, the amount of protection provided by some
Over 1973e2009, the Ganges River has created a morphologi- of these structures was below expectation and some structures
cally active corridor within which the river meanders back and even failed within a few years. Due to insufficient data related to

Fig. 8. Locations of natural hard points (cohesive bank material) and manmade hard points (protection structures) along the Ganges River in 2003.
 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155 151

manmade structures, we cannot quantitatively estimate the presented in Table 4 indicate that during the study period, the local
amount of protection manmade structures had provided against width of Ganges River varies from a minimum of 1.0 km to a
erosion. Therefore, we only examined the effectiveness of natural maximum of 13.86 km. Over these 36 years, the river has changed
hard points against erosion, which can be estimated when the river its width significantly primarily because of erosion and accretion in
flow directly against the structures. both banks. For example, from 1973 to 1980, the average river
In this study, we only consider erosion rates for those years width had decreased slightly from about 4.94 km to 4.65 km due to
when the river flowed directly against the natural hard points. The major accretion at the Pakka Narayanpur area and downstream of
spatial and temporal distributions of bank erosion are closely Hardinge Bridge along the left bank (Fig. 10). However, after that
related to the spatial variations of sediment materials resistant to the river began to widen with varying annual rates until 2003 (see
erosion, or the presence of natural hard points shown in Fig. 9. From Table 4) because during this period (1980e2003) erosion was more
the results it seems that the erosion rate at natural hard points dominant than accretion within the study reach (Table 2). The
varies from about 0 to 21, 0 to 20, 0 to 21, and 0 to 19 m/y on the left average rate of widening of the river in 1980e2003 was about
bank and about 0 to 22, 0 to 21, 0 to 20, and 0 to 20 m/y on the right 31 m/y, and by that the average width of the river had increased
bank during 1973e1984, 1984e1993, 1993e2003, and 2003e2009, from about 4.65 km to 5.27 km (Table 4). Correspondingly, EGIS
respectively (Fig. 7). For these four study periods, the average (2000) reported the average width of the Ganges within
values were about 5 m/y for all four study periods on the left and 7, Bangladesh as 4.37 km in 1984 and 4.69 km in 1993, or an average
5, 7, and 4 m/y on the right bank, respectively. In contrast, the widening rate of 36 m/y from 1984 to 1993. Although the widening
average erosion rates for the Ganges River as a whole were about of the Ganges River had not been as high as other rivers of
63, 61, 58, and 32 m/y on the left bank and about 92, 122, 63, and Bangladesh (EGIS, 2000), the impact of this widening of the Ganges
36 m/y on the right bank during the same four study periods, River was still significant because an increase in the river width are
respectively (Table 2). Therefore, erosion rates along the natural associated with a loss of floodplain and a lateral migration of the
hard points are substantially lower than average erosion rates of meandering bend of the river, which had been of concern for
the Ganges River in all the study periods. The presence of cohesive Bangladesh with a large population living on limited land areas. In
sediment (natural hard points) at the boundary of the corridor addition, changes in the land use and problems due to deforesta-
significantly reduces the maximum bank erosion rate from a few tion in the upper catchment of the Ganges River would result in
hundred meters to several meters per year. Therefore, natural hard more sediment load transported to lower reaches of this river,
points have significant impact on the planform development and leading to a more dynamic river morphologically.
morphological changes of the Ganges River.
Table 4
Changes in the Ganges River channel width for 1973e2009.
4.3. Changes in river channel for 1973e2009
Year Maximum Minimum Average
width (km) width (km) width (km)
Changes of the river channel of Ganges were estimated in terms
of changes of width, sinuosity and island area of the study reach 1973 10.82 1.86 4.94
1980 10.03 1.73 4.65
estimated from eight satellite images dated between 1973 and
1984 10.13 1.94 4.66
2009. Discussions of results are herein given below. 1993 11.74 1.00 4.84
1997 12.07 1.49 4.87
4.3.1. River width 1999 12.64 1.55 4.98
2003 13.35 1.61 5.27
Eight dry season satellite images from 1973 to 2009 were used
2009 13.86 1.05 5.27
to study changes to the width of Ganges River. The results

(a) (b)
1973-1984 1973-1984
20 20

10 10

0 0
1984-1993 1984-1993
20 20

10 10
Erosion (m/y)
Erosion (m/y)

0 0
1993-2003 1993-2003
20 20

10 10

0 0
2003-2009 2003-2009
20 20

10 10

0 0
0 50 100 150 200 0 50 100 150 200
Distance (km) Distance (km)

Fig. 9. The spatial and temporal variations of erosion rates at natural hard points (cohesive bank material) along the study reach of the Ganges River for four study periods at the (a)
Left bank, and the (b) Right bank.
152 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155

10000 5.4
5.3
Net erosion /accretion (ha)

5.2
Accretion

Avg. River width (km)

5000 5.1
5.0
4.9
0
4.8
1973-1980 1980-1984 1984-1993 1993-1997 1997-1999 1999-2003 2003-2009
4.7
-5000 4.6
Erosion 4.5
4.4
-10000 4.3
Year

Net eosion/accretion Avg. width

Fig. 10. Changes to the average river width of Ganges under a net accretion between 1973 and 1984, and a net erosion between 1984 and 2009.

Fig. 8 shows how the net erosion/accretion process had that the sinuosity between the Farakka Barrage and Hardinge
temporally changed the average river width of the Ganges over Bridge for the past 200 years (1780e1973) was around 1.20. How-
1973e2009. Between 1973 and 1984, the figure shows that a net ever, the lower reaches between the Hardinge Bridge and Aricha of
accretion had decreased the average width of the river, but after the Ganges River were more sinuous with a sinuosity of about 1.46
that the average width of the Ganges continued to increase up to during the 18th century. Since the beginning of the 20th century,
2009 because net erosion had dominated the morphology of the this reach has become nearly straight and its sinuosity remains
Ganges after 1984. To maintain the total sediment discharge of any close to 1.10 (Sarker, 2004).
anabranching river system, any reduction in the total channel Using the eight satellite images acquired, we computed the
width result from the formation of islands could be more than sinuosity of the Ganges River (Table 5 and Fig. 11). Our results show
compensated by an increase in the transport rate (Nanson and that the sinuosity between Farakka Barrage and Hardinge Bridge
Knighton, 1996). has gradually increased from 1973 to 2003. This increase in sinu-
osity is possibly due to the temporary blocking of sediment by the
4.3.2. Sinuosity Farakka Barrage (Sarker et al., 2003). The sinuosity of this reach had
Other than bankline shifting and changes to island areas, sinu- been increasing slowly partly because of a reduction of the river
osity is another important morphological parameter which de- bed slope achieved through bed aggradation at the downstream of
scribes the dynamic nature of the river system. Previous studies the Farakka barrage which became operational in 1974. This phe-
showed that the sinuosity of the Ganges River had changed nomenon has also been observed in some other rivers of India
considerably over time and location, e.g., changes to the River’s (Galay, 1983). However, in general the increase in sinuosity of a
sinuosity differ between that observed at Farakka Barrage and river is related to an increase in the bank erosion because it causes a
Hardinge Bridge and that between Hardinge Bridge and Aricha. decrease in streamflow velocity, as was the case for the Ganges
Sarker (2004) analyzed historical maps of 1780 and 1860 and found River between 1980 and 2003 (Fig. 12). The meandering pattern in

1.45
1.40
1.35
1.30
Sinuosity

1.25
1.20
1.15
1.10
1.05
1.00
1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009
Year
Farakka to hardinge Br. Hardinge Br. to Aricha Overall

Fig. 11. Changes in the sinuosity of different reaches of the Ganges River for 1973e2009.
 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155 153

4.3.3. Islands area

Anabranching in any river system occurs where a river is divided
by islands (Schumm, 1985) or a river that consists of multiple
channels characterized by vegetated semi-permanent alluvial
islands excised from existing floodplain (Nanson and Knighton,
1996). In the Ganges River, under the influence of active erosion
and accretion, small sandbars tend to grow to relatively large
islands within the river reach, or as land attached to river banks.
When these lands are vegetated, locally they are known as “chars”
which often create opportunities for people of highly populated
country like Bangladesh to establish settlements and pursue agri-
cultural activities. ISPAN (1995) categorized all chars in major rivers
of Bangladesh into two groups: island chars and attached chars
(Fig. 13). Island chars are lands that, even in dry season, can only be
reached from the mainland by crossing a main channel, while
attached chars are accessible from the mainland without crossing a
main channel during the dry season, but they will be inundated or
Fig. 12. A simple linear regression between the average river width and sinuosity of surrounded by water during a typical flood. However, for simplicity
the study reach for 1980e2003. sake, both island chars and attached chars are herein classified as
islands. According to Nanson and Knighton (1996), “islands may be
excised by channel avulsion from extant floodplain, developed from
an anabranching river system could adjust to decreased bed- within-channel deposition or formed by prograding distributary
sediment/water discharge ratios by increasing their sinuosity channels within splays or deltas”. Changes to island areas derived
(Nanson and Knighton, 1996). On the other hand, as mentioned from satellite images are presented in Table 6 and Fig. 14. The re-
above, for the river reach between Hardinge Bridge and Aricha, the sults show that on a whole islands area of the Ganges had
sinuosity had maintained at about 1.10, implying that the amount of decreased significantly from 1973 to 1984. Further, during this
land erosion and accretion in this reach has not changed much in period, the river had also narrowed down due to a net accretion of
this 1973e2009 study period. This particular reach of Ganges had the river banks (Table 2). As a result, increased river flow had lead
maintained its modest sinuosity partly because of guide bunds of to considerable erosion around boundaries of islands.
the Hardinge Bridge, bank protection structures and the presence of
old cohesive bank materials at several locations downstream of the
bridge. However, between 1973 and 2003, the overall sinuosity of Table 6
the Ganges study reach had consistently increased, which resulted Changes to islands from 1973 to 2009.
in an increase in the overall length of this 190 km study reach.
Year Total water Islands (Char) Islands area Average
surface (ha) area (ha) over total width (km)
water surface
Table 5
Changes in the sinuosity of the Ganges River from 1973 to 2009. 1973 108,584 26,455 24 4.94
1980 102,674 25,880 25 4.66
River reach Sinuosity 1984 101,793 20,767 20 4.66
1993 108,516 27,389 25 4.84
1973 1980 1984 1993 1997 1999 2003 2009
1997 108,614 30,341 28 4.87
F. B. to H. Br. 1.17 1.23 1.27 1.33 1.33 1.37 1.39 1.33 1999 112,110 31,371 28 4.98
H. Br. to Aricha 1.08 1.12 1.10 1.08 1.10 1.12 1.12 1.08 2003 118,302 46,067 39 5.27
Overall 1.16 1.21 1.24 1.25 1.25 1.28 1.30 1.21 2009 120,648 42,453 35 5.27
Note: F.B. ¼ Farakka Barrage, H. Br. ¼ Hardinge Bridge.

Fig. 13. Chars (islands) classification (adapted from ISPAN, 1995).

154 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155

5.4 sediment load of about 520 million tonnes/year (UNEP, 1995). Ac-
y = 3E-05x + 4.1275 cording to Latrubesse (2008), large anabranching river systems are
Avg. width of river (km)

R2 = 0.89 unique in that it is generally difficult to discriminate their channel

5.2 pattern based on the analysis of standard variables such as slope,
grain size, bankfull discharge, width/depth ratio, grain size, stream
power, etc., that are applied to small or medium size river systems.
5
5. Summary and conclusions

4.8 Over the years, under the combined impact of erosion, accretion,
and human interventions, the large anabranching Ganges River
within Bangladesh has experienced significant hydro-
4.6 morphological changes. However, to accurately quantify morpho-
15000 25000 35000 45000 logical changes of an active alluvial river such as the Ganges can be
challenging. A practical approach to detect the long-term,
Area of vegetated Islands (ha)
morphological changes of the Ganges River is by analyzing multi-
Fig. 14. A linear regression fit between island area and the average width of the ple satellite images. In this study, multiple Landsat MSS, TM and IRS
Ganges River. LISS images of 1973e2009 acquired over the Ganges River located
within Bangladesh were analyzed. During this study period, the
Ganges River had shifted its bankline both by eroding the flood-
However, after 1984, islands had increased consistently both in plain (landward) and by accretion (riverward) almost every year to
terms of area and in percentage. Our results are consistent with the a significant amount. In particular, the river reach from the Indo-
study of EGIS (2000) that also reported an increase of vegetated Bangladesh border to the upstream of the Hardinge Bridge point
islands area in the Ganges from 1984 to 1993. This increase in experienced more bankline shifting compared to the river reach
islands area during 1984e2003 were partly due to increased from downstream of Harding Bridge to Aricha. In 1973e2009, the
sediment supply from river bank erosion and soil erosion from maximum bank erosion and accretion rates that had occurred were
agricultural land upstream of the study area. In an alluvial river, if 604 m/y and 848 m/y, respectively. The erosion rate (ha/y)
the discharge and slope do not decrease significantly over time increased from 1973e1984 to 1984e1993, but after that it
when a river width increases, the rates of erosion should generally continued to decrease from 1993 to 2009. In contrast, the accretion
decrease because the velocity of river flow should decrease and vice rate had been decreasing since 1973 to 2009.
versa. In contrast, as a river becomes wider, the rate of accretion to The Ganges River underwent two fairly distinct patterns of
islands should increase, if other factors remain unchanged. Because morphological changes which happened in 1973e2003 and in
the net effect of anabranching is to enlarge the total channel width 2003e2009. On a whole, the Ganges River was morphologically
which decreases the average depth and velocity, hence decreasing more active from 1973 to 2003 than from 2003 to 2009. The erosion
the specific stream power and the bed shear stress without any rate varied widely along the River. Within the active corridor of the
change in the channel slope. This phenomenon was also observed Ganges, if there are natural hard points where cohesive sediment
in the Ganges, that as the river slowly widened, its islands area also dominated, the bank erosion rate decreased from several hundred
increased slowly (Fig. 14). meters to several meters per year. Therefore, natural hard points
played a significant role on the planform and the morphological
4.4. Comparison of data with other major rivers in the world changes of the Ganges. From 1973 to 2003, the river between Far-
akka Barrage and Aricha had become longer by about 20 km
Ganges is one of the largest rivers in the world in terms of because its sinuosity had increased from 1.16 to 1.30. This channel
catchment area, length and width. The total catchment area of this length increase caused a net loss of floodplain of about 9717 ha
river is about 1.095  106 km2. Compared to the erosion and ac- along its course, both in India and in Bangladesh. On the other
cretion of 221 river systems of varying sizes across the world, Van hand, between 1973 and 1984, both the average river width and the
de Wiel (2003, p. 45) developed an empirical relationship (best islands area had decreased significantly. Then from 1984 to 2003,
regression fit) between bank erosion rate (E) and drainage area (A), the river width and island areas began to increase and after that,
which is E ¼ 0.053A0.44 (ibid). According to this equation, the there had been minimal change. Apparently for Ganges, its river
average erosion rate of the Ganges River should only be 24 m/y, or width and overall islands area are strongly correlated to each other
about three times smaller than the observed average erosion rate of (R2 ¼ 0.89). Also, the correlation between the banks’ erosion rates
66 m/y for the study period (Table 2). In other words, the average and the reach average eroded areas with the mean annual dis-
annual bank erosion rate of the Ganges is much higher than that of charges of the Ganges were stronger than that with the banks’
a river with a similar catchment area but of a global average erosion accretion rates and the reach average accreted areas.
rate. Similarly, the average erosion rate for the Brahmaputra basin Lastly, the Ganges River has a significantly higher erosion rate
(573,500 km2) was only 18.0 m/y which was substantially smaller than other major rivers in the world, probably because its river
than the observed average bank erosion rates of Jamuna River (Baki banks are primarily composed of highly erodible materials and it
and Gan, 2012). The reason for its erosion rate being substantially has a high discharge rate especially during the Monsoon season.
higher than the global average erosion rate is probably because its Furthermore, the net erosion of Ganges has been closely related to
river banks are primarily composed of highly erodible materials the increase of its river width, sinuosity and the amount of vege-
and the discharge rate of the Ganges is high especially during the tated islands area within the river channel.
Monsoon season. Compared to 23 major rivers of Asia with a mean
(standard deviation) runoff depth of 380 (290) mm/year and a Acknowledgements
mean (standard deviation) suspended sediment load of about 150
(220) million tons/year, the Ganges River has a runoff depth of For this study, satellite images, hydro-meteorological and sedi-
619 mm/year, and it produces the third highest annual suspended ment data were provided by CEGIS (Center for Environmental and
 Md.A. Hossain et al. / Quaternary International 304 (2013) 142e155 155

Geographic Information Services), Dhaka, Bangladesh. The authors Hossain, M.M., 1989. Geomorphic Characteristics of the Padma Upto Brahma-
putra Confluence. Final Report. Institute of Flood Control and Drainage
acknowledge the facilities provided by CEGIS. The study is partly
Research, Bangladesh University of Engineering and Technology, Dhaka,
funded by the Natural Science and Engineering Research Council of Bangladesh.
Canada, and the third author is also funded by a scholarship from Hasan, A., Haque, I., Huq, P.A.K., Martin, T.C., Nishat, A., Thorne, C.R., 1997. Defining
the University of Alberta, Canada. the banklines of the Brahmaputra-Jamuna River using satellite imagery. In:
Watts, J. (Ed.), Proceeding of the 3rd Inc. on Flood Hydraulics at Stellenbosch,
South Africa. H.R.Wallingford, Oxford, UK, pp. 349e358.
References ISPAN, 1992. Technical Notes Series 1. ISPAN/USAID, Flood Plan Coordination Or-
ganization, Dhaka, Bangladesh.
Adhikari, K.D., Ahmed, Q.K., Malla, S.K., Pradhan, B.B., Rahman, K., Rangachari, R., ISPAN, 1993. The Dynamic Physical and Human Environment of the Riverine
Rasheed, K.B.S., Verghese, B.G. (Eds.), 2000. Cooperation on Eastern Himalayan Charlands: Brahmaputra-Jamuna, Geographic Information System. Prepared for
Rivers: Opportunities and Challenges. Konark Publishers Pvt Ltd, Delhi, India, the Flood Plan Coordination Organization (FPCO), Dhaka, Bangladesh.
pp. 1e23. ISPAN, 1995. The Dynamic Physical and Human Environment of the Riverine
Azuma, R., Sekiguchi, H., Ono, T., 2007. In: Studies of High-resolution Morphody- Chars and Charland, Socioeconomic RRA Reports for the Brahmaputra-
namics with Special Reference to River Bank Erosion vol. 50(C). Annuals of Jamuna, Meghna, Ganges and Padma Rivers, and the Charland Flood Proof-
Disaster Prevention Research Institute, Kyoto University, Japan, pp. 199e209. ing Study. Prepared for Flood Plan Coordination Organization (FPCO), Dhaka,
Baki, A.B.M., Gan, T.Y., 2012. Riverbank migration and island dynamics of the Bangladesh.
braided Jamuna River of the Ganges Brahmaputra basin using multi-temporal Kammu, M., Lu, X.X., Rasphone, A., Sarkkula, J., Koponen, J., 2008. Riverbank
Landsat images. Quaternary International 263, 146e161. changes along the Mekong River: remote sensing detection in the Vientiane-
CEGIS (Center for Environmental and Geographic Information Services), 2003. Nong Khai area. Quaternary International 186, 100e112.
Ganges River: Morphological Evolution and Prediction. Prepared for Water Kleinhans, M.G., 2010. Sorting out river channel patterns. Progress in Physical Ge-
Resources Planning Organization (WARPO). Ministry of Water Resources, Gov- ography 34 (3), 287e326.
ernment of the People’s Republic of Bangladesh, Dhaka, p. 70. Kotoky, P., Bezbaruah, D., Baruah, J., Sarma, J.N., 2005. Nature of the bank erosion
Chu, Z.X., Sun, X.G., Zhai, S.K., Xu, K.H., 2006. Changing pattern of accretion/erosion along the Brahmaputra river channel, Assam, India. Current Science 88 (4),
of the modern Yellow River (Huanghe) subaerial delta, China: based on remote 634e640.
sensing images. Marine Geology 227, 13e30. Latrubesse, E.M., 2008. Patterns of anabranching channels: the ultimate end-
Delft Hydraulics and DHI (FAP 24), 1996a. Bed Material Sampling in Ganges, Padma, member adjustment of mega rivers. Geomorphology 101, 130e145.
Old Brahmaputra and Jamuna. Special Report No. 8, River Survey Project (FAP Mirza, M.M.Q., 2004. The Ganges water diversion: environmental effects and
24), Prepared for Flood Plan Coordination Organization (FPCO). Ministry of implications-an introduction. In: Mirza, M.M.Q. (Ed.), The Ganges Water
Water Resources, Government of the People’s Republic of Bangladesh, Dhaka. Diversion: Environmental Effects and Implications. Kluwer Academic Pub-
Delft Hydraulics and DHI (FAP 24), 1996b. Sediment Rating Curves and Balances. lishers, Dordrecht, The Netherlands, pp. 1e12.
Special Report No. 18, River Survey Project (FAP 24), Prepared for Flood Plan Mirza, M.M.Q., Dixit, A., 1997. Climate change and water management in the GBM
Coordination Organization (FPCO). Ministry of Water Resources, Government of basins. Water Nepal 5, 71e100.
the People’s Republic of Bangladesh, Dhaka. Nanson, G.C., Knighton, A.D., 1996. Anabranching rivers: their cause, character and
Eaton, B.C., Millar, R.G., Davidson, S., 2010. Channel patterns: braided, anabranching, classification. Earth Surface Processes and Landforms 21, 217e239.
and single-thread. Geomorphology 120, 353e364. Rinaldi, M., 2003. Recent channel adjustments in alluvial rivers of Tuscany, central
EGIS (Environmental and GIS Support Project for Water Sector Planning), 1997. Italy. Earth Surface Processes and Landforms 28, 587e608.
Morphological Dynamics of the Brahmaputra-Jamuna River. Water Resources Sarker, M.H., 2004. Impact of upstream human interventions on the morphology of
Planning Organization, Ministry of Water Resources, Government of People’s the Ganges-Gorai system. In: Mirza, M.M.Q. (Ed.), The Ganges Water Diversion:
Republic of Bangladesh, Dhaka, p. 76. Environmental Effects and Implications. Kluwer Academic Publishers, Dor-
EGIS (Environmental and GIS Support Project for Water Sector Planning), 2000. drecht, The Netherlands, pp. 49e80.
Riverine Chars in Bangladesh: Environmental Dynamics and Management Is- Sarker, M.H., Thorne, C.R., 2006. Morphological response of the Brahmaputra-
sues. University Press Limited (UPL), Dhaka, Bangladesh, p. 88. Padma-Lower Meghna river system to the Assam earthquake of 1950. In:
EGIS, 2002. Developing and Updating Empirical Methods for Predicting Morpho- Sambrook Smith, G.H., Best, J., Bristow, C.S., Petts, G.E. (Eds.), Braided
logical Changes of the Jamuna River. In: Technical Note Series 29. Environment Rivers: Process, Deposits, Ecology and Management. International Associ-
and Geographic Information System (EGIS), Dhaka, Bangladesh. ation of Sedimentologists Special Publication 36. Blackwell, London,
FAO, 2007. Climate Variability and Change: Adaptation to Drought in Bangladesh, A pp. 289e310.
Resource Book and Training Guide. Asian Disaster Preparedness Center, Food Sarker, M.H., Huque, I., Alam, M., 2003. Rivers, chars and char dwellers of
and Agriculture Organization of the United Nations, Rome, p. 66. Bangladesh. International Journal of River Basin Management 1 (1), 61e80.
Fuller, I.C., Large, A.R.G., Milan, D.J., 2003. Quantifying channel development and Schumm, S.A., 1985. Patterns of alluvial rivers. Annual Review of Earth and Plane-
sediment transfer following chute-off in a wandering gravel-bed river. Geo- tary Sciences 13, 5e27.
morphology 54, 307e323. Surian, N., 1999. Channel changes due to river regulation: the case of the Piave
Galay, V.J., 1983. Causes of River bed degradation. Water Resources Research 19 (15), River, Italy. Earth Surface Processes and Landforms 24, 1135e1151.
1057e1090. Thorne, C.R., Russell, A.P.G., Alam, M.K., 1993. Planform Pattern and Channel Evo-
Haack, B., Brayant, N., Adams, S., 1987. An assessment of Landsat MSS and TM data lution of the Brahmaputra River, Bangladesh. In: Geological Society, London,
for urban and near-urban land-cover digital classification. Remote Sensing of Special Publications, vol. 75, pp. 257e276.
Environment 21, 201e213. Umitsu, M., 1993. Late Quaternary sedimentary environments and land forms in the
Halcrow and Partners Ltd., DHI, EPC Ltd., Sthapati Sangshad Ltd. (FAP 4), 1993. Ganges Delta. Sedimentary Geology 83, 177e186.
Morphological Studies: Southwest Area Water Resources Management Project. UNEP, 1995. Water Quality of World River Basins. UNEP Environment Library No. 14.
Report Produced for Flood Plan Coordination Organization (FPCO), Dhaka, UNEP, GEMS Collaborating Centre for Freshwater Quality Monitoring and
Bangladesh. Assessment, National Water Research Institute, Environment Canada, Burling-
Hossain, M.M., Baki, A.B.M., Ahmad, S.S., 2008. The Flood Problem and Mitigation ton. Ontario, Canada.
Methods in Bangladesh. 4th International Symposium on Flood Defence: Van de Wiel, M.J., 2003. Numerical Modelling of Channel Adjustment in Alluvial
Managing Flood Risk, Reliability and Vulnerability. Toronto, Ontario, Canada, Meandering Rivers with Riparian Vegetation. Ph.D. thesis. Dept. of Geography,
May 6e8, 2008. Univ. of Southampton.
Hossain, M.A., 2006. A Study on the Influence of Hard Points on the Morphology of Wang, J.J., Lu, X.X., 2010. Estimation of suspended sediment concentrations using
the Ganges River. M. Sc. thesis. Department of Water Resources Engineering, Terra MODIS: an example from the Lower Yangtze River, China. Science of the
Bangladesh University of Engineering and Technology, Dhaka, p. 67. Total Environment 408, 1131e1138.