C M Y K

article

Flood Mapping : A Case Study of C M

Y A System for Survival K Remote sensing technology has proven that it is the most effective means for mapping of affected areas quickly, accurately and repetitively covering wide areas, apart from its cost effectiveness.

V. BHANUMURTHY, G. SRINIVASA RAO, B. SIMHADRI RAO and P. MANJUSREE Water Resources Division, Water Resources & Oceanography Group National Remote Sensing Agency

mong all the natural disasters that countries in the world and is affected by represents an acutely flood-prone region faces, floods are the most different natural disasters through out the characterised by awesome hazards of floods. frequent and often devastating. Floods year. Among all the natural disasters that India Although occurrence of flood has been an A age-old phenomenon in the riverine areas of occur in India mostly during the south-west faces, floods are the most frequent and often monsoon season causing heavy loss to lives devastating. Floods cause disruption of this region, yet the extent of damage caused and property. When a flood event occurs, it telecommunications network, road and rail by the flood has increased significantly in is very important to get dynamic flood traffic and damage public utilities, industries, recent years particularly after the great Assam information timely and accurately for the houses and property resulting in huge losses of 1950 which recorded 8.7 on safety of life and property. This information in financial terms. Loss of human lives and the Richter scale. With more than 40 percent can be used in developing comprehensive cattle due to floods also cause distress. The of its land surface susceptible to flood relief effort. Remote sensing technology has loss in terms of both lives and property affects damage, the total flood-prone area in the proven that it is the most effective means for the overall economy of the country. Floods Brahmaputra valley is 3.2 Mha (Goswami, mapping of flood affected areas quickly, occur in India mostly during the south-west 2001). accurately and repetitively covering wide monsoon season starting from June to When a flood event occurs, it is very important areas, apart from its cost effectiveness. This September. During this period more than 75% to get dynamic flood information timely and paper discusses the methodology for of the country’s annual rainfall occurs, accurately for the safety of life and property. preparing and delivering the flood maps to causing the major rivers to discharge huge This information can be used in developing the user agencies in near real time. The volumes of runoff, resulting in floods. comprehensive relief effort (Corbley, 1993). various elements involved in this process such About 40 Mha or nearly 1/8th of India’s It is very difficult to collect the information as data acquisition under emergency geographical area is flood prone. The annual during floods from ground due to disruption programming, generation of flood inundation average area affected due to floods is about of communications and transport network. map and damage statistics, dissemination of 7.57 Mha and could be as high as 17.5 Mha In recent years, satellite remote sensing the information to the concerned relief in the worst year and the affected crop area technology has proven that it is the only agencies was discussed. This was is about 3.5 Mha (National Flood means for mapping of flood affected areas demonstrated through a case study of Assam Commission, 1980). The Brahmaputra and quickly, accurately and repetitively covering floods during 2002. Ganga river basins are the most flood prone wide areas, apart from its cost effectiveness. 1.0 Introduction river basins in the country. The Brahmaputra This information helps relief agencies for carrying out relief measures and restoration India is one of the most disaster prone basin, particularly its valley in Assam,

Fig. 1 Pre Flood image of the flood prone areas Fig. 2 Combined Radarsat image depicting the flood scenario

30 July-August 2003 Geospatial Today article

Figure 3. Flood inundation map of assam state Figure 4. Submerged crop area in marigaon district of normalcy. National Remote Sensing submerged barring the natural as well as is 240,000 Sq.Km. and in is Agency (NRSA) has been carrying out Near artificial highlands within it by this wave of 47,000 Sq.Km. The Brahmaputra basin Real Time Flood Mapping project for more flood. In overall, this wave of flood has extends over an area of 580,000 Sq.Km. up than a decade by mapping flood affected affected a population of more than 5.3 million to its confluence with in Bangladesh. The total areas in the country in near real time. Under people of 19 districts in the state. length of river from its origin to outfall is 2,897 this project, information on floods has to be in Assam was gauged at Km and its course within India is 918 Km provided to relief agencies in the form of flood six different locations namely Dibrugarh, (Report on Brahmaputra floods, 1996). maps and flood damage statistics at the Neamatighat, Tezpur, Guwahati, Goalpara and The maximum discharge at Pandu near earliest possible time, so that they can be stations which is maintained by Central Guwahati was recorded 72,794 Cumec on effectively utilized for relief management. Water Commission. The water level in 23.8.1962 and a minimum discharge was This paper discusses the methodology for Brahmaputra river reached the peak at recorded 1,757 Cumec on 22.2.1968. The preparing the flood inundation maps using Dibrugarh on 22-July-2002. The average dry season discharge is 4,420 Cumec the satellite data and delivering the maps and Brahmaputra reached the flood peak at (Report on Brahmaputra floods, 1996). flood damage statistics to the user agencies Neamatighat and Tezpur on 23-July-2002, It is a braided and unstable river in its entire in near real time. The various elements Guwahati and Goalpara on 26-July-2002 and length in Assam. The instability of the river is involved in this process such as data Dhubri on 27-July-2002 correspondingly mainly attributed to high sediment charge, acquisition under emergency programming, (Central Water Commission, 2002). steep slopes and traverse gradient. During its generation of flood inundation map and 3.0 Study Area and Flood Problem course in Assam valley it is joined by major damage statistics, dissemination of the The Assam state is comprised of tributaries Subansiri, Jiabharali, Puthimari, information to the concerned relief agencies Brahmaputra and Barak valley and covered Pagladiya, Beki, Manas, Aie, Sankosh, etc., was discussed. This was demonstrated by latitude of 240 to 280 N and longitude of from Northern Himalayan region and through a case study of Assam, where a major 890 45’ to 960 E. The geographical area of Buridihing, Desang, Dikhow, Dhansiri, Kopili, flood wave occurred during the month of July the state is 78,523 Sq.Km. The Brahmaputra etc., from southern Hill ranges. These 2002. valley extending from Sadia in the east and tributaries contribute sufficient discharges to 2.0 Flood Scenario in 2002 Dhubri in the west covers a geographical area the Brahmaputra river. During the year 2002 Assam state was of 57,116 Sq.Km. and Barak valley including 3.2 Rainfall affected by a major flood wave in the month southern hill districts covers a geographical Normally, North-eastern states in the country of July. This wave of flood devastated the area of about 21,407 Sq.Km (Goswami, which covers an area of the heaviest rainfall entire Assam state paralysing the transport 2001). The Brahmaputra river has its source in the world, experiences rainfall due to south- network and affecting millions of people. The at Manas Sarovar in Himalayan region and west monsoon and out of total annual rainfall flood situation in the state worsened with has an extensive catchment area both in the 85% occurs during the monsoon months of several districts reeling under the onslaught Himalayan region and in state of Assam. As June to September. Average annual rainfall of the rising Brahmaputra and its tributaries the river travels down from China as Sungpo in the Brahmaputra valley ranges from 1750 due to heavy downpour in the upper by name, enters India through the Siang mm in the to about 6400 mm catchments of Brahmaputra river. The worst- district of to the plains of in the north-east hilly region. Besides, the hit district was cut off from the rest Assam, it joins with two other mighty rivers, valley gets a good amount of rainfall in of the State as the National Highway (NH) 52 Dehing and Lohit near Sadiya and from there months of April and May due to was breached at Kekuri and Samarajan by it assumes the name Brahmaputra. activities. flood water of Jiadhal river. Around 13 districts 3.1 Brahmaputra River

3.3 Flood Problem C of the state were severely affected by floods. The Brahmaputra river is one of the largest The entire 430 Sq.Km. area of the Kaziranga The flood problems of Brahmaputra valley as M river in the world. Its catchment area in Tibet

National Park famous for rhinoceros, was well as Barak valley in Assam, are mainly as Y is 293,000 Sq.Km. that in India and Bhutan

follows: K

July-August 2003 Geospatial Today 31

C M Y K C M Y K

article

i) Inundation of large areas due to over 5.0 Methodology Flood Inundated Area (Hectare) Receded Area District flowing banks of the rivers during the 5.1 Satellite Data (Hectare) C As on 31-July-02 As on 26&28-July-02 flood. Procurement and Preprocess M Dibrugarh 2,609* NC - ii) Drainage congestion at the confluence In order to carry out the Near Y point of tributaries during the high stage Real time flood mapping project Dhemaji 10,462* NC - K of rivers. effectively, the data should be Sibsagar 5,216 NC - iii) Extensive silt load in the rivers due to soil available for analysis at the 9,871 NC - , and large scale earliest, as soon as the satellite Lakhimpur 7,950 NC - land slides in hilly region resulting passes over the flood affected 6,704 10,732* 4,028 instability of the rivers and erosion of area. This was achieved under Nowgong 18,768 28,253 9,485 banks. emergency programming, as Marigaon 47,408 50,120 2,712 Darrang 6,783 13,048 6,265 4.0 Satellite Data Used Radarsat International has placed compressed SAR data on their Sonitpur 7,564 31,311 23,747 Presently there is a constellation of four Indian file transfer protocol (ftp) site as Cachar 22,741 27,304 4,563 Remote Sensing (IRS) satellites viz. IRS-1C, soon as the satellite passes over 16,188 17,854 1,666 IRS-1D, IRS-P3 and IRS-P4 for monitoring the the area of study. The Radarsat Hailakandi 4,229 5,897 1,668 earth. The WiFS sensor with moderate spatial SAR data was compressed using NC 23,095 - resolution of 188 m on IRS-1C, 1D & P3 has Multi resolution Seamless Image NC 6,424 - high repetivity and a ground swath of 810 Km Database (MrSID) compression Kamrup NC 39,238 - covering large areas which is very useful for technique, a lossy compression NC 4,045 - flood inundation mapping and monitoring technique, in 1:5, 1:10, 1:20 and Golpara NC 20,637 - (Rao, 2000). The OCM sensor on IRS-P4 has 1:100 compression ratios. The Dhubri NC 30,552 - a spatial resolution of 360 m and wider swath data with 1:10 compression ratio of 1420 Km with alternate day repetivity can was found to be optimal for flood also be used in case of major floods. As Total 1,66,493 3,08,510 54,134 studies and the same persistent cloud cover is a common problem compressed SAR data was Note : * Districts partly covered by the satellite data; NC Not covered by the during monsoon season in Tropical regions, downloaded through Internet at satellite data. microwave satellite data serves as an NRSA and then the data was Table-2 District-wise Comparative Flood Inundated Area Statistics for alternative in flood disaster monitoring, as it Assam State. uncompressed. The raw SAR can penetrate cloud cover. Synthetic Aperture data contains speckle i.e. noise in the image Radar (SAR) data has been applied to A suitable threshold is selected and the flood due to inherent properties of SAR sensor. The inundation layer was extracted. The flood speckle in the image was suppressed using Date Satellite Sensor Path-Row / inundation layer was edited further to remove median filter. Beam mode any isolated and stray pixels. 17-03-02 IRS-1D WiFS 111-53 The pre-flood satellite data was rectified From the three Radarsat data sets, the geometrically with respect to a pre-defined 26-07-02 Radarsat Scan SAR Narrow inundation layer was extracted using the projection system. The data was classified to 28-07-02 Radarsat Scan SAR Wide above mentioned procedure. As each of the extract the pre flood river course and bank 31-07-02 Radarsat Scan SAR Wide data set covers only a part of Assam, a final line. Figure 1 shows the pre-flood image of flood inundation layer was prepared by Table-1 Satellite Data used for the study. the flood prone areas in Assam state. The combining all the three inundation layers. mapping of flooded areas of the Amazon filtered Radarsat SAR data was registered with 5.3 Output Generation & Dissemination rainforest (Hess et al. 1995, Melack and Wang the pre-flood data for common coordinate 1998, Miranda et al. 1998), monsoon flood system. Figure 2 shows the combined The derived flood inundation layer was damage in India (Rao et al. 2002), and river Radarsat image depicting the flood scenario in Assam state. Crop Area flood waves in the Great Upper Mississippi Number of Villages District Submerged valley flood 1993 (Brakenridge et al. 1998, 5.2 Extraction of Flood Information Marooned (Hectare) Imhoff et al. 1987) and in China floods in 1998 The rectified Radarsat SAR data was classified + (Liu et al., 2002). using the threshold technique. The threshold Darrang 257 7,873 # NRSA has kept a constant watch on the flood technique is useful in extracting the flood Dhemaji 59 3,103 Jorhat# 154 7,514 scenario in Assam state. Satellite coverage inundation extent from SAR data. Based on Lakhimpur# 108 9,232 charts over Assam state was prepared for all the observation of Digital Number (DN) values Sibsagar# 155 3,173 available IRS Satellites. All the IRS satellite at various cross-sections of flood inundation, Marigaon+ 526 29,354 passes were checked for cloud cover over the threshold value is fixed and then Nalbari+ 223 2,974 affected areas and most of the data were thresholding was done using the following Dhubri* 657 18,977 found to be cloudy. Anticipating the cloud procedure to extract flood inundation extent Goalpara* 309 14,305 cover problem, the following Radarsat satellite from SAR data. data sets (Table 1) were procured in order to If DN <= T then 1 (flood inundation) Total 2448 96,505 study the impact of the flood wave during the Else 0 Note : + Based on the analysis of 26-July-02 Radarsat SAR month of July 2002 in Assam state. IRS WiFS data; * Based on the analysis of 28-July-02 Radarsat SAR data th data of 17 March 2002 was chosen as pre where, DN represents gray value in SAR data, # Based on the analysis of 31-July-02 Radarsat SAR data flood data set. T is threshold value. Depending upon the beam mode, T varies from image to image. Table-3 District-wise Flood Damage Statistics for part of Assam State.

32 July-August 2003 Geospatial Today article

indicates the severity of the flood disaster and its effect. NRSA was able to provide the flood information in a shortest span of 5 hours from the time of satellite data procurement. The procedure of information dissemination was successfully carried out for Assam state during flood disaster References Brackenridge, G.R., Tracy, B.T., and Knox, J.C., 1998. Orbital SAR remote sensing of a river flood wave. International Journal of Remote Sensing, 19, 1439-1445. Central Water Commission, Flood Bulletins of Middle Brahmaputra Division, 2002. Govt.of India. Goswami, D.C., 2001. Flood forecasting in the Brahmaputra River, India: A Case Study. (http:// www.southasianfloods.org/document/ffb) Hess, L.L., Melack, J.M., Filoso, S., and Wang, Y., 1995, Real time mapping of inundation on the Amazon Figure 5. Submerged villages in marigaon district, Assam. floodplain with the SIR-C/X-SAR synthetic aperture radar. IEEE Transactions on Geoscience and Remote integrated with district boundaries to extract villages marooned in Marigaon district, Assam the inundated area statistics in each district. State. Sensing, 33, 896-904. The flood inundation layer was superimposed 6.0 Results & Discussion Imhoff, M.L., Vermillon, C., Story, M.H., Choudhury, over a base map along with other A.M., and Gafoor, A., 1987, Monsoon flood boundary The study of Radarsat SAR data sets of 26, administrative boundaries, pre-flood river delineation and damage assessment using space 28 & 31-July-2002 shows the severity and layer etc. and a flood map was composed to borne imaging radar and Landsat data. impact of flood wave during July 2002 in the required scale. The flood map shows flood Photogrammetric Engineering and Remote Sensing, Assam. From the Table 2 it can be observed inundation extent spatially as this information 4, 405-413. can be used for planning relief operations in that Marigaon, Kamrup, Sonitpur, Dhubri, Liu, Z., Huang, F., Li, L., and Wan, E., 2002. Dynamic a better manner. The flood inundation map Nowgong, Goalghat, Darrang, Barpeta and along with the district-wise statistics has to Goalpara districts in Brahmaputra valley and monitoring and damage evaluation of flood in north- be furnished to the user departments Cachar and Karimganj districts in Barak valley west Jilin with remote sensing. International Journal electronically to avoid delay in dissemination. are the worst affected as on 26&28-July- of Remote Sensing, 23 (18), 3669-3679. By adopting this procedure of information 2002. On studying the 31-July-2002 satellite Melack, J.M., and Wang Y., 1998, Delineation of data, it can be observed that Dhemaji and dissemination, a turn around time of 5 hours flooded area and flooded vegetation in Balbina Jorhat districts in Brahmaputra valley were was achieved i.e. time from satellite data Reservoir (Amazonas, Brazil) with SAR. Verh Internat also affected and there is overall recedence procurement to transmission of flood Verein Limnaol, 26, 2374-2377. information. Figure 3 shows the flood in flood inundation when compared with Miranda, F.P., Fonseca, L.E.N., and Carr, J.R., 1998, inundation map of Assam state prepared 26&28-July-2002. In , Semivariogram textural classification of JERS-1 (Fuyo- based on the analysis of 26th, 28th and 31st maximum of flood inundation has receded July, 2002 Radarsat images and Table 2 as on 31-July-2002 by almost 75% when 1) SAR data obtained over a flooded area of the presents the flood inundation area statistics compared with 26&28-July-2002. The study Amazon rainforest. International Journal of Remote computed for the flood affected districts and shows that out of 23 districts, 19 districts in Sensing, 19, 549-556 the receded areas in selected districts. Assam were affected by this wave of flood. Rao, D.P. 2000, Disaster Management: Proceedings On combining the all the three data sets, the of Map India. (http://www.gisdevelopment.net/ 5.4 Detailed Flood Damage Assessment maximum flood inundated area was application/natural_hazards/overview/nho0004.html) using GIS estimated as 3,44,00 ha due to this flood, For Assam state detailed digital database was which accounts to about 4.4% of total Report on National Flood Commission, 1980 (http:/ developed for nine districts. The digital data- geographical area of Assam. /wrmin.nic.in/development/flood_affected.html) base consists of layers of land use, village From the Table 3 it can be observed that 2,448 Report on the flood situation of Brahmaputra and boundaries, district boundaries, roadways, number of villages were marooned and Barak valley for the year 1996, Hydrology division, railways etc. The flood inundation layer de- 96,505 ha of crop area submerged during the Flood Control Department, Government of Assam, rived from satellite data was integrated with flood wave in the nine districts of Assam. India. the database layers in GIS environment for Among the nine districts, maximum number Simhadri Rao B., Bhanumurthy V., assessing the flood damages in terms of crop of villages was marooned in Srinivasa Rao G. and Manjusree, P., December, area affected, number of villages marooned and maximum crop area was submerged in 2002. Near real time flood mapping for Orissa State and roadways submerged etc, in each dis- Marigaon district. trict. Table 3 shows the flood damage statis- using space technology. ISPRS TC-VII Symposium, C tics for the nine districts of Assam state. Fig- The Near Real time flood mapping study for Hyderabad, India. M Assam state using remote sensing technology ures 4&5 show the submerged crop area and [email protected] Y K

July-August 2003 Geospatial Today 33

C M Y K