Rainfall Variability and Spatio Temporal Dynamics of Flood Inundation during the 2008 Kosi Flood in Bihar State, India
Suraj Kumar Singh,
The devastating flood in Kosi River initiated on 18 August 2008 resulted in heavy loss of human life and natural resources in the Bihar state of India. About 2.5 millions people become homeless and 2516 km2 of agricultural land including fallow land is damaged by the flood inundation. The spatio-temporal dynamics of flood inundation during a period of two months from 20 August to 21 October 2008 has been carried out. The temporal satellite images of RADARSAT were used to map the spatial-inundation pattern of Kosi flood whereas, IRS P6 LISS III optical satellite data of October 2004 was used to map the land use characteristics of the region in a non flood situation to evaluate the loss of agricultural land inundated by Kosi flood. Rainfall variability in the entire Kosi river catchment was assessed by comparing the monthly rainfall during monsoon season for a period of 1998-2009. The variability in rainfall pattern during the pre and post Kosi flood event was examined to develop insight into genetic aspect of flooding. The study revealed a loss of 2135 km2 of standing crops due to highest flood inundation of 3089 km2 as on 5 September, 2008. Rainfall pattern indicates much higher rainfall in the lower Kosi catchment which probably induced the conditions causing breach of embankment and resulted in the initiation of Kosi flood on 18 August 2008.
The Kosi, known as Kaushiki in Sanskrit, is one of the most ancient rivers
of India. It rises in the Himalaya, drains the foothills to the east of Kathmandu
in Nepal and plains of Bihar state in India before joining the river Ganga.
Its three tributaries within the Himalaya, the Sun-Kosi, Arun and Tamur, confluence
before entering to the plain region further downstream. The combined discharge
of the three tributaries of Kosi has traveled through a deep gorge extending
for 10 km length downstream of their confluence. The turbulent water with high
sediment flux emerges out of the canyon and form the worlds largest, 180
km long and 150 km wide alluvial fan. Due to the fanning of material, the single
river acquires a distributary channel pattern in the plain areas (Agarwal
and Bhoj, 1992).
Gole and Chitale (1966) described the Kosi system as
an inland delta built by large sediment flux which was also attributed to be
the primary factor causing westward shifting of Kosi and extensive flooding.
The recurrent flooding in the plain areas of lower Kosi catchments in Bihar
state induces condition of water logging which is resulting in development of
vast tract of productive agricultural lands into wastelands. Pandey
et al. (2010a) examined the impact of natural and anthropogenic features
on flood induced waterlogging in parts of Bihar plains and revealed that canals
and railway line induced highest waterlogging conditions. The flooding also
raises the levels of water table which result in development of salt affected
soils in this region. Low topography, low carrying capacity and avulsive behavior
of the Kosi river are attributed to frequent and prolonged flooding in this
region (Chowdary et al., 2008). To avert conditions
of disastrous flood events in this region and to reduce flood induced environmental
hazards scientific understanding of the river hydrodynamics is essential.
The dynamic nature of the Kosi River has attracted attention for over a century
and a variety of mechanisms have been suggested ranging from tectonic tilting
and nodal avulsions (Gole and Chitale, 1966; Arogyaswamy,
1971; Agarwal and Bhoj, 1992) and discharge peakedness
and autocyclic processes (Leier et al., 2005,
Bridge and Karssenberg, 2005).
To avert the frequent flooding events in the plain areas, Kosi river flow was
controlled through embankments constructed in the year 1963 to protect about
2800 km2 of land in North Bihar and Nepal (Sinha
et al., 2008). The 18 August 2008 avulsion of the Kosi River recorded
an Eastward jump of about 120 km which is an order of magnitude higher than
any single avulsive shift recorded in historical times. The avulsion was triggered
by a breach in the eastern afflux bund of the Kosi at Kusaha, 12 km upstream
of the Kosi barrage. The Sanjai, Sursara and Bhenga canal streams, on which
the Kosi flowed earlier, were fed afresh, leading to the spread of water in
areas located along the banks of these streams. The river then abandoned the
prescribed western channel and occupy a straight course near the center of its
alluvial fan. The water in this course spread out widely and inundated towns,
villages and cultivated fields in the densely populated region of Bihar State
in India. The five worst affected districts in Bihar are Supaul, Madhepura,
Araria, Saharsa and Purnea. The rampaging Kosi has left at least 3.02 million
people marooned in these five districts of Bihar (Unicef
India, 2008). The new course has a much lower carrying capacity, the water
flows like a sheet, 15-20 km wide and 150 km long as a result of which a very
large area remains inundated/waterlogged for more than four months after the
breach (Sinha, 2009). Pandey et
al. (2010b) have described that areas with high waterlogging risk in
Northern Bihar plains also corresponds to high flood hazards and vulnerability
due to poor socio-economic conditions in these areas.
Since, the launch of the Earth observation satellites, remote sensing data
have been used frequently for analysing and mapping of floods (Rango
and Salomonson, 1974; Phillipson and Hafker, 1981).
Although optical remote sensor data can be used effectively for flood mapping,
they have major drawback due to the necessity of obtaining imagery in time and
extensive cloud cover during flooding obscures the inundated area and makes
it difficult to acquire good quality optical imagery. The availability of microwave
data from ERS/Radarsat can penetrate clouds and is capable of detecting and
delineating flooded areas (Kussul et al., 2008).
Multi-date radar data together with optical remote sensing data provide real
time solution in mapping, monitoring and management of floods and related hazards
during flood transgression and regression phase (Pandey,
2008). Satellite based precipitation measurement obtained through TRMM is
being widely used for monitoring of floods in regions that regularly experiences
extreme precipitation and flood events (Asante et al.,
The IRS P6 LISS III optical satellite image of non flooded period of 22 October-2004 was used to map the landuse/landcover characteristics in this region to evaluate the loss of agricultural land inundated by Kosi flood. The Tropical Rainfall Measuring Mission (TRMM) data was analyzed for the period of 1998-2009 to examine the spatio-temporal variability of rainfall in the Kosi catchment during monsoon period as well as to evaluate change in rainfall pattern during pre-post flood event.
MATERIALS AND METHODS
Study area: The area under investigation comprises the lower Kosi catchment a part of which was affected by 2008 Kosi flood. It is located between 86° 20' 00" to 87 °30 00" E longitude and 25 °20' 00" to 26° 45' 00" N latitude, covering a total area of 10536 km2 mainly in the state of Bihar (Fig. 1). The total catchment area of Kosi river is 54419 km2 majority is located in Nepal.
Data used and methodology: The availability of satellite data products
provided by various organizations world-wide which can be accessed online has
increased the understating about happening of major natural disasters in any
part of the world. Such information can be utilized collectively to obtain genetic
aspects of such disasters in near real time. In the present study we have utilized
the RADARSAT data derived images showing flood affected regions due to Kosi
flood in the year 2008 which was available in the NRSC image gallery (www.nrsc.gov.in).
These images of different dates from 20 August to 21 October 2008 have been
used in the analysis. The data was downloaded from NRSC image gallery and georeferenced
with reference to state and district boundary maps of Bihar. Survey of India
(SOI) political map on 1:4 million was used for delineation of boundary of Bihar
state whereas district boundary maps were obtained from the census hand book
of Bihar, 2001. IRS P6 LISS III satellite data of 22 October, 2004 was procured
from NRSC, GoI to map the land use-land cover in the area during the end of
Kharif season (June-October) in a non flood period. Satellite data was processed
using ERDAS IMAGINE software whereas the spatial data was created under ArcGIS
platform for spatial analysis. To assess the spatio-temporal variability of
rainfall, TRMM Multi-satellite Precipitation Analysis (TMPA) product was downloaded.
The monthly TRMM 3B43 data for the period of 1998-2009 covering the month of
June to September was downloaded in ASCII format. The data was registered to
real world coordinates in GIS to obtain variability of rainfall intensity with
reference to geographic location (Fig. 2).
map of the study area
chart of the methodology for spatio-temporal dynamics of flood inundation
The TRMM rainfall intensity values recorded at various points were interpolated
in GIS environment using IDW method of spatial interpolation to examine spatial
variability of rainfall over the entire Kosi catchment. The IDW method is selected
in the present study as it is one of the standard spatial interpolation procedures
in geographic information science (Burrough and McDonnell,
Flood dynamics: The geo-referenced RADARSAT data derived images were
segmented primarily into water and land units based on their tonal contrast.
Flood inundated areas appear black in RADARSAT images due to forward scattering
of incoming microwave radiation whereas over the land units more backscatter
cause bright signature on the satellite images (Jensen, 2005).
This tonal contrast was used to segment the flood inundated area from non flooded
areas. The RADARSAT satellite images (Fig. 3a-j)
and their corresponding classified output (Fig. 3k-t)
helped in understanding the flood dynamics during the period of observations.
The classified output in raster format was converted to vector format for spatial
analysis of flood dynamics and loss of agricultural land due to flooding in
Arc GIS platform.
It is evident that on 2 day of flooding (i.e., on 20 August, 2008) flood water
traveled a distance of only 55 km whereas, on 22 August, it reached a distance
of 93 km from breach point and finally merges with river Ganga on 27 August.
The width of flood water also changes from 12 km in the upper reaches to 45
km in the lower parts of Kosi fan areas. The flood water follows a straight
course till 22 August, 2008 after which it spreads more, further engulfing large
area down slope. It is important to note that a flood stream course originated
from the main flood course and flowing in SE direction entered into the adjacent
watershed of Mahanada river flowing east of Kosi fan area causing flooding outside
image a-j representing sequence of flood inundation (transgression and
regression) from 20 August to 21 October 2008 and the corresponding classified
output k-t derived from satellite images showing land (white) and water
(black) covered regions in the flood effected parts in the lower Kosi
The area statistics of flood dynamics obtained through classified outputs clearly
revealed that the area under flood inundation was maximum on 5 September 2008
(3089 km2) thereafter decreasing to (950 km2) on 21 October,
2008 (Fig. 4, Table 1).
use/land cover map of the study area
image based statistics of area under flood inundation and non flooded
land for selected periods from 20 August to 21 October 2008
Agricultural area loss due to flood inundation: Digital satellite image
based land use/land cover (LU/LC) mapping was performed using unsupervised classification
technique to compute the area statistics of major land use/land cover types
in the flood affected Kosi fan area. The classified map clearly depict that
majority of area is under cropland (72.18%) and associated fallow lands (17.41%).
based rainfall variability maps representing average rainfall from June
to September during 1998-2009, a-d and e-h representing monthly rainfall
from June to September in 2008. Difference image i-l represent rainfall
deficit/surplus conditions in 2008 (positive values indicate increase
in rainfall whereas negative values indicate decrease in rainfall)
The wasteland comprises (5.93%) of total area mainly consists of seasonal and
permanent waterlogged areas, whereas water bodies (4.48%) mainly comprising
the rivers of Kosi and Ganga (Fig. 5, Table
Further, the land use/land cover map was overlaid with flood inundation area
during peak flood period on 5 September 2008 to ascertain the loss to agricultural
lands due to flooding. It was estimated that an area of 2516 km2
of agricultural land (cropland: 2135 km2 and fallow land: 481 km2)
comprising the 81.45% of total flood inundation area was destroyed due to flooding.
Area statistics of various landuse/landcover categories
in the Kosi fan area
It signifies the impact of agricultural loss due to the destruction of existing
crops and unsuitability of fallow lands which could be cultivated in the Rabi
season (October-April) due to prevailing water logging condition during the
post flood periods.
Assessment of spatial variability of rainfall: In the present study
for the assessment of spatio-temporal variability of rainfall in the entire
catchment (upper and lower) of Kosi river, satellite based rainfall data using
TRMM Multi-satellite Precipitation Analysis (TMPA) product was used. The TMPA
provides a calibration-based sequential scheme for combining rainfall estimates
from various satellites, at fine scales (0.25°x0.25° and 3-hourly) (Huffman
et al., 2007). The monthly TRMM 3B43 accumulated rainfall (0.25°x0.25°)
product which is the combination of TRMM Precipitation Radar (PR) and TRMM Microwave
Imager (TMI) were acquired from TRMM Online Visualization and Analysis System
tovas/). Rainfall intensity spatial maps of individual months of June, July,
August and September were prepared for the period of 1998-2009. The raster data
pertaining to monthly rainfall was further categorized into 7 classes representing
rainfall variation from less than 80 mm to more than 640 mm. In this way we
obtained maps representing average monthly variability from June to September
from the year 1998-2009. The monthly data was analyzed to obtain spatial rainfall
variation across the monsoon months to asses the rainfall induced flood vulnerability
at various stages during the monsoon season.
The spatial analysis of rainfall pattern clearly revealed that the onset of monsoon in June clearly exhibit low rainfall over the entire region among the four monsoon periods under observation. The July and August comprises wetter condition whereas in the month of September the rainfall again decreases. The monthly assessment of average rainfall condition and its comparison with rainfall event during 2008 is given below to emphasize on rainfall variability in inducing flood conditions in the area.
Comparing average rainfall of June with monthly rainfall of June 2008 revealed
that in the month of June 2008 the upper catchment received much higher rainfall
(100-480 mm) over many areas where average rainfall varies from less than 100-180
mm. Similarly the average rainfall in the month of June in the lower catchment
also exhibited substantial increase from average value of 100-180 mm to 320-400
mm in June 2008, especially in the central parts (Fig. 5a,
e). In the month of July 2008 also we observed significantly
higher rainfall especially in the lower catchment which received 480-560 mm
and at place up to 640 mm against the average rainfall of 400-480 mm (Fig.
5b, f). In the month of August 2008 the upper catchment
also received higher amount of rainfall (240-560 mm) than the average rainfall.
The lower catchment also received substantially high rainfall of 240-560 mm
as compared to 100-240 mm of average rainfall (Fig. 5c, g).
During September although their was a decrease in rainfall but still it was
much higher than the average rainfall especially in the lower catchment which
received 180-320 mm of rainfall in September 2008 as compared to 100-180 mm
during average rainfall (Fig. 5d, h).
A rainfall difference image was prepared by subtracting average rainfall value
(1998-2009) from monthly rainfall value (2008) recorded at all the observation
points in the entire catchment. The output image clearly exhibits the areas
where rainfall deficit/surplus condition prevailed in the year 2008 (Fig.
5i-l). The positive values in the map indicate increase
in rainfall whereas, negative values indicate decrease in rainfall. The upper
and lower catchment areas received comparatively very high rainfall in all the
three months i.e., June, July and August during the year 2008 than the average
rainfall. This clearly implies that the high runoff generated by these rainfall
events especially in the lower parts of the catchments resulted in high soil
saturation which probably induced the factors which lead to the breaching of
the embankment at places and resulted in a disastrous flood in Bihar.
The flood dynamics obtained through RADARSAT data derived images exhibit that
the area under flood inundation was maximum on 5 September 2008, 3089 km2
thereafter reduced to 950 km2 on 21 October, 2008. The slope in the
lower catchment areas is very gentle causing flood water to spread over large
areas. Satellite based rainfall data from TRMM clearly exhibit much higher rainfall
during the monsoon months of 2008 as compared to average rainfall which probably
induced conditions for breaching of the embankment at places. In terms of agricultural
area loss an area of 2516 km2 of agricultural land was destroyed
due to flooding. The main causes of floods in the Gangetic plains are inadequate
capacity within river banks to contain high flows, river bank erosion and silting
of river beds (Kumar et al., 2005). The excess
water with high sediment flux released during monsoon periods therefore causes
extensive floods in Gangetic plains of Bihar region. The changing climatic conditions
with more intense rainfall in future (Singh and Sontakke,
2002) may trigger more sediment flux in the upper catchment. This may get
accentuate if the snow and glacier covered regions in the upper parts of the
catchment shrink with increased temperature conditions (Bajracharya
et al., 2006). Under these conditions there would be an increase
in the cases of Glacier Lakes Outburst Floods (GLOF) (Bajracharya
and Kumar, 2005) and natural damming of the river in the upper catchment,
outburst of which may leads to flash floods in the lower catchment areas and
plain regions. The high seismic activity in the region (Sukhija
et al., 2002) also indicates possibilities of future damage of embankment
which may result in similar or more devastating floods in the future.
It is evident that high discharge moving down slope would follow the straight
course as observed in the Kosi 2008 flood event. The straight course cause hydraulic
gradient of the flood water to be nearly double of that which existed in the
earlier sinuous course (Reddy et al., 2008) and
therefore have more sediment carrying capacity. Therefore, to avert the future
disaster of this sort it is proposed to develop the straight course of the Kosi
river and adopted valley deepening activity at regular intervals. To regulate
the hydrological regime towards sustainable agricultural setup, the old river
course, as well minor distributary channels may be supplied with regular discharge
of water. Any planning for regulating the flood water discharge should be done
after proper evaluation of micro relief as well as sub surface hydrological
regimes in the Kosi fan area. Remote sensing technique is very useful as it
provide accurate, time and cost-effective method for spatio-temporal flood dynamics
assessment and evaluation of loss due to flooding. The TRMM based rainfall study
provided a continuous mapping of rainfall events throughout such a large catchment
of Kosi where ground based meteorological monitoring is not possible at large
number of places. The study provided significance of satellite based monitoring
towards analyzing flood events.
Authors are thankful to National Remote Sensing Centre, Department of Space, Government of India for providing IRS P6 LISS III satellite images and access to download the RADARSAT data derived images from NRSC image gallery and TRMM Online Visualization and Analysis System (TOVAS) for providing TRMM gridded rainfall monthly products from 1998-2009 period.
1: Agarwal, R.P. and R. Bhoj, 1992. Evolution of Kosi river fan, India: Structural implication and geomorphic significance. Int. J. Remote Sens., 13: 1891-1901.
Direct Link |
2: Arogyaswamy, R.N.P., 1971. Some geological factors influencing the behaviour of the Kosi. Rec. Geol. Surv. India, 96: 42-52.
3: Asante, K.O., R.D. Macuacua, G.A. Artan, R.W. Lietzow and J.P. Verdin, 2007. Developing a flood monitoring system from remotely sensed data for the Limpopo Basin. IEEE Trans. Geosci. Remote Sens., 45: 1709-1714.
Direct Link |
4: Bridge, J.S. and D. Karssenberg, 2005. Simulation of flow and sedimentary processes, including bifurcation and avulsion in alluvial fans. Proceedings of the 8th International Conference on Fluvial Sedimentology, Aug. 7-12, Delft, pp: 70-70
5: Burrough, P.A. and R.A. McDonnell, 1998. Principles of Geographical Information Systems. Oxford University Press, New York, Pages: 333
6: Chowdary, V.M., R.V. Chandran, N. Neeti, R.V. Bothale and Y.K. Srivastava et al., 2008. Assessment of surface and sub-surface waterlogged areas in irrigation command areas of bihar state using remote sensing and GIS. Agric. Water Manage., 95: 754-766.
7: Gole, C.V. and S.V. Chitale, 1966. Inland delta building activity of Kosi River. J. Hydraulic Divi. Proc. Am. Soc. Civil Eng., 2: 111-126.
8: Huffman, G.J., R.F. Adler, D.T. Bolvin, G. Gu and E.J. Nelkin et al., 2007. The TRMM multisatellite precipitation analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeorol., 8: 38-55.
Direct Link |
9: Jensen, J.R., 2005. Remote Sensing of the Environment: An Earth Resource Perspective. Pearson Education Inc., USA
10: Kussul, N., A. Shelestov and S. Skakun, 2008. Grid system for flood extent extraction from satellite images. Earth Sci. Inform., 1: 105-117.
Direct Link |
11: Leier, A.L., P.G. Decelles and J.D. Pelletier, 2005. Mountains, monsoons and megafans. Geology, 33: 289-292.
Direct Link |
12: Pandey, A.C., 2008. Geoinformatics for Natural Hazard Mapping with Focus on Seismotectonics: A Case Study from Haryana State, India. In: Geoinformatics in Resource Planning, Nathawat, M.S. and A.C. Pandey (Eds.). Rawat Publishing, Jaipur, pp: 124-137
13: Pandey, A.C., S.K. Singh and M.S. Nathawat, 2010. Analysing the impact of anthropogenic activities on waterlogging dynamics in Indo-Gangetic plains, Northern Bihar, India. Int. J. Remote Sens.,
14: Pandey, A.C., S.K. Singh and M.S. Nathawat, 2010. Waterlogging and flood hazards vulnerability and risk assessment in Indo Gangetic plain. Natural Hazards,
15: Phillipson, W.R. and W.R. Hafker, 1981. Manual versus digital Landsat analysis for delineating river ﬂooding. Photogram. Eng. Remote Sens., 47: 1351-1356.
16: Kumar, R., R.D. Singh and K.D. Sharma, 2005. Water resources of India. Curr. Sci., 89: 794-811.
Direct Link |
17: Rango, A. and V.V. Salomonson, 1974. Regional ﬂood mapping from space. Water Resour. Res., 10: 473-484.
Direct Link |
18: Bajracharya, S.R. and M.P. Kumar, 2005. Growth of hazardous glacial lakes in Nepal. Proceedings of the JICA Regional Seminar on Natural Disaster Mitigation and Issues on Technology Transfer in South and Southeast Asia, Sept. 30-Oct. 13, Kathmandu, Nepal, pp: 131-148
19: Bajracharya, S.R., P.K. Mool and B.R. Shrestha, 2006. The impact of global warming on the glaciers of the Himalaya. Proceedings of the International Symposium on Geo-Disasters, Infrastructure Management and Protection of World Heritage Sites, Nov. 25-26, Nepal Engineering college, Nepal, pp: 231-242
Direct Link |
20: Reddy, D.V., D. Kumar, D. Saha and M.K. Mandal, 2008. The 18 August 2008 Kosi river breach: An evaluation. Curr. Sci., 95: 1668-1669.
Direct Link |
21: Singh, N. and N.A. Sontakke, 2002. On climatic fluctuations and environmental change of the indo-gangetic plains, India. Climatic Change, 52: 287-313.
Direct Link |
22: Sinha, R., G.V. Bapalu, L.K. Singh and B. Rath, 2008. Flood risk analysis in the Kosi river basin, north Bihar using multiparametric approach of analytical hierarchy process. J. Ind. Soc. Remote Sens., 36: 335-349.
Direct Link |
23: Sinha, R., 2009. The great avulsion of kosi on 18 August 2008. Curr. Sci., 97: 429-433.
Direct Link |
24: Sukhija, B.S., M.N. Rao, D.V. Reddy, P. Nagabhushanam, D. Kumar, B.V. Lakshmi and P. Sharma, 2002. Palaeoliquefaction evidence of prehistoric large/great earthquakes in north bihar, India. Curr. Sci., 83: 1019-1025.
Direct Link |
25: Unicef India, 2008. Reuniting survivors of bihar floods. http://www.unicef.org/india/resources_4651.htm.