"REMOTE SENSING AND GIS FOR MAPPING AND MONITORING THE"
REMOTE SENSING AND GIS FOR MAPPING AND MONITORING THE EFFECT OF LAND USE/COVER CHANGE ON FLOODING IN GREATER DHAKA OF BANGLADESH Ashraf M Dewan1 and Yasushi Yamaguchi2 1 Department of Earth and Environmental Sciences, Nagoya University, Japan 2 Graduate School of Environmental Studies, Nagoya University, Japan ABSTRACT The objectives of this paper are to quantify the land use/cover changes and to analyze the effect of such changes on flooding in Greater Dhaka of Bangladesh. Using topographic maps and multi-temporal remote sensing data, land use/cover change analysis has been performed. The result revealed that significant portion of low-lying areas, water bodies and agricultural lands are transformed to built-up areas, hence increasing flood risk potential. For example, built-up areas have increased to about 343.7% in 2005 compared to the 1960 built-up areas (11.1%). Flood area estimation for the three greatest floods of 1988, 1998 and 2004 were obtained from the Landsat TM, and RADARSAT SAR data. It is found that 47.1% area was flooded in 1988 while in 1998 and 2004, inundation areas were 53.2% and 43%, respectively. This information was subsequently integrated with the hydro-meteorological information and found that flooding in Greater Dhaka during monsoon is aggravating. Analysis showed that land use/cover change playing an important factor in intensifying the flood process in Greater Dhaka of Bangladesh. KEY WORDS: Land use/cover, Greater Dhaka, Floods, Water level, Landsat, IRS, Radarsat SAR 1991). It lies in the sub-tropical monsoon zone and is under the 1. Introduction humid climatic condition. The city experiences about 2000 mm Land use/cover studies using remote sensing data have been annual rainfall, of which more than 80% occurs during the received immense attention worldwide due to their importance monsoon season (June-September). Historically, the city is in global change analysis (Cihlar, 2000). Both human-induced endowed with rivers, numerous khals (ephemeral water bodies) and natural land cover changes can influence the global change and canals that drain water from the upper reaches during the because of its interaction with terrestrial ecosystem (Houghton, monsoon season. As population increased, these areas have 1994), biodiversity (Sala et al., 2000) and landscape ecology been encroached on, thus compounding flood problems. (Reid et al., 2000). In addition, it reflects the human impacts on environment at various temporal and spatial scales (Lopez et al., 3. Materials and Methods 2001). Therefore, accurate and up-to-date land use/cover The Landsat, RADARSAT and IRS-1D data were acquired information is essential for environmental planning, to for this study. Two Landsat TM data (03 February and 15 understand the impact on terrestrial ecosystem (Muttitanon and October, 1988) were acquired from Bangladesh Space Research Tripathi, 2005), and to achieve sustainable development and Remote Sensing Organization (SPARRSO). In addition, (Alphan, 2003). Furthermore, timely and reliable data on land two RADARSAT SAR data covering the 1998 (25 August) and use/cover may facilitate the formation of integrated resource 2004 (24 July) peak floods were acquired and used. One IRS- management policies. 1D LISS III (26 December) data was also collected from Centre Even though most of the developed countries are well for Environment and Geographic Information System (CEGIS) equipped and updated with detailed land use/cover information, and used for 2005 land cover classification. lack of geospatial database for land use/cover persists in Hydrological data of daily rainfall and water level (1965- developing countries and may hinder proper development 2005) of the governing rivers of the study area were collected planning. Hence, space-borne remotely sensed data may be from Bangladesh Water Development Board (BWDB). particularly useful in developing countries where recent and reliable spatial information is lacking (Dong et al., 1997). 3.1 Satellite data pre-processing Studies demonstrated that the transformation of land All the images used in this study were geometrically use/cover either by human activity or natural origin can corrected using a Landsat TM image of 1997 as a reference, and profoundly impact the hydrological cycle by accelerating were rectified to Bangladesh Transverse Mercator (BTM) volume and rate of surface runoff (Weng, 2001; Shi et al., system using thirty-five uniformly distributed ground control 2007), mounting flood risk (Nirupama and Simnovic, 2007), points. The root mean square error (RMSE) varied from 0.25- degradation of water quality (Xian et al., 2007), and causing 0.3 pixels for the optical data, and was slightly higher (0.45- erosion (Weng, 2001). Considering these facts, the objectives of 0.48 pixel) for the radar data. Finally, a first-order polynomial this study are to quantify the land use/cover changes and to fit was applied and all the data were resampled to 30 meter pixel analyze the effect of land use/cover change on flooding in using the nearest neighbour method for an integrated analysis. A Greater Dhaka of Bangladesh using remote sensing and GIS GAMA-MAP filter (5x5 window size) was used to subdue techniques. speckle from the RADARSAT data, a random granular effect in the radar data. 2. Environmental setting of the Study Area The study area is the Greater Dhaka City, Bangladesh. 3.2 Reference data Latitude and longitude of the lower left and upper right corners A number of reference data sets were constructed in this of the study area are 23°68/ N 90°33/ E and 23°90/ N, 90°50/ E, study. Due to the retrospective nature of the study, we had to respectively. Topographically, the area is a flat land and is rely on a variety of methods to develop reference data for located mainly on an alluvial terrace, popularly known as the training and accuracy assessment of land use/cover maps. These Modhupur terrace of the Pleistocene period. The surface include high resolution satellite data (e.g. SPOT Pan of 1989; elevation of the area ranges between 1 and 14 m (FAP 8A, IKONOS Pan of 2003), topo-sheets of different years, existing sampling method. Then using the geographical locations of land use maps and field data. features available on land use maps, high resolution images and SOB topo-sheets, map validation has been performed. To assess 3.3 Land use/Cover Classification the accuracy of 2005 map, ground truth data obtained from the The 1960 land cover map has been developed from the field were employed. 110 reference data were used to validate topographic maps complied from the aerial photographs taken the derive 2005 land cover map. in 1958. Two topo-sheets (e.g.79 I 5/6) were scanned and displayed on computer screen. Using ArcGIS (v 9.1), the land 3.5 Classification of Flood-time images cover map was digitized, edited and levelled. Besides, a large A completely cloud-free Landsat TM image of 1988 was scale map (1:20000) depicting the study area and its used to estimate inundation in the 1988 flood. Additionally, two surrounding was employed to identify various land cover types RADARSAT SAR (ScanSAR narrow) data were used to in the GIS environment. The land use map of 1962 by Khan and analyze the 1998 and 2004 floods. Two specific dates for the Islam (1964) was of value to develop the 1960 land cover data. RADARSAT were acquired as they represented the peak All the satellite data were thoroughly studied using flooding in the study area (Dewan et al., 2005; The Daily Star, spectral and spatial profile to ascertain the digital number (DN) 2004). The Landsat TM band 4 (NIR) was classified by using a of different land cover categories prior to classification. A threshold technique. Empirical threshold values were modified version of the Anderson Scheme (Anderson et al., determined from image spectral property or DN values. Pixel 1976) was adopted to study the land cover types from remotely value >43 for non-water and ≤43 for water were used to identify sensed data at various scales and resolutions. Six separable land the land and water boundaries in the TM band 4. For the use/cover types have been identified, that is, water bodies, RADARSAT data, a rule-based approach (Dewan et al., 2005) wetland/lowland, built-up, cultivated land, vegetation, and bare was used to estimate flooding in the 1998 and 2004 floods. soil/landfill sites. After the extraction of flooded and non-flooded areas from the Training samples were selected through reference data individual flood-image, each classified flood-time image was and ancillary information. Sixty-seventy training sites varying then superimposed on a dry seasonal classified data to estimate in size from 286 to 8914 pixels were used to locate training the net inundated areas in Greater Dhaka of Bangladesh. It is pixels on the images. Except for the bare soil/landfill sites, necessary to note here that the dry seasonal water bodies were training samples for each class were 5-12 subclasses. The discarded for flood area estimation. Hence, only the flooded training samples were then evaluated by using class histogram areas were attained. Finally, the inundation area percentage (%) plots. Training samples were refined, renamed, merged, and was estimated using the following equation: deleted after evaluation of class histogram and statistical a parameters for normality. A supervised classification by the Inundation area percentage (%) = × 100 ………... (1) maximum likelihood algorithm was subsequently applied to a+b obtain land cover from each data. where, a = inundated area, b = non-water area Misclassification was observed in the classified land cover categories obtained from the supervised classification. For 4. Results and Discussion example, certain urban settlements were misclassified as bare soil/landfill sites due to their similar spectral characteristics. 4.1 Land use/cover change analysis Likewise, misclassification also observed between It is found that land use/cover in the study area has been wetland/lowlands and cultivated land categories and between changed significantly since 1960 (Fig. 1). Analysis revealed that water and lowland/wetland categories. It may be noted that, the built-up area has increased to about 343.7% compared to the initially wetland category was identified as a separate class but 1960 built-up areas (11.1%). The bare soil/landfill category also eventually it has been merged with the lowland as it was not increased extensively. In contrast, other land use/cover classes possible to separate from lowland category as both of them including water bodies, cultivated land, low-lying lands, and consist similar spectral properties. Post-classification vegetation have been reduced greatly and converted to built-up refinement, therefore, was used to improve the accuracy. To areas. Table 1 shows a summary of the major land use/cover surmount the difficulty of misclassification, a number of conversions that have been taken place from 1960 to 2005. In strategies were considered. For example, thematic information the 1970 and 1980s, major development occurred on the (e.g. water bodies, cultivated land, bare soil/landfill site) was cultivated lands and by filling up of water bodies. However, first extracted from the V-S-W index (Yamagata et al., 1997). A recent development is being taken place in the low-lying areas. rule-based technique using ancillary/thematic information For example, relative conversion of the lowlands/wetlands (DEM, municipal maps, bare soil/landfill sites and water during the period of 1988-2005 revealed that nearly 20% has bodies) and GIS tools such as AOI (area of interest) by visual been transformed to the built-up category. It should have interpretation were then employed to correct previously incredible consequences on the hydrological environment of misclassified land cover categories. It is necessary to mention Greater Dhaka. Derived land use/cover types from 1960 to 2005 that ground truth information was also of importance in the are shown in Fig. 2. refinement process. Application of those techniques Accuracy of the land cover maps revealed that the TM substantially improved the classification accuracy. Finally, a 3 resulted in the overall accuracy of 86.4% while the 2005 by 3 majority filter was applied to classified land use/cover classification resulted 88.2% of overall accuracy. Table 1 images to reduce the salt-and-pepper effect before the images shows the total accuracy of the land cover maps and their were further analyzed for land use/cover changes (Lillesand and corresponding kappa coefficient. The result showed that all the Kiefer, 1999). data met the minimum USGS total accuracy requirements (Anderson et al., 1976). 3.4 Map Validation Classified land cover maps from satellite data were further 4.2 Flood estimation from satellite images used for validation using ground truth data obtained from Flood estimation from satellite data revealed that in 1988, various sources. For instance, for the 1988 land cover map, a 47.1% areas was flooded. It is similar to the result obtained by total of 125 pixels was generated using stratified random Sado and Islam (1997). Using a MOS-MESSR of 17 October Table 1 Area and Percentage of Land use/Cover types from 1960 to 2005 Land use/Cover types 1960 1988 2005 Area (ha) % Area (ha) % Area (ha) % Water bodies 2952.5 7.1 2101.5 5.1 2107.4 5.1 Lowlands/wetland 13549.0 32.6 12715.6 30.6 7163.5 17.2 Cultivated land 14351.9 34.5 9024.9 21.7 6049.9 14.6 Vegetation 5605.4 13.5 5793.8 13.9 3932.6 9.5 Built-up area 4631.8 11.1 10858.9 26.1 20551.0 49.4 Bare soil/landfill sites 473.6 1.1 1069.4 2.6 1759.7 4.2 Total 41564.0 100.0 41564.0 100.0 41564.0 100.0 Overall accuracy (%) - - 86.4 88.2 Kappa Coefficient - - 0.84 0.86 1988, they reported that at least 41.1% areas were flooded in the “pre-urban flood” and the three events of 1988, 1998, and 2004 1988 flood. This study and the study by Sado and Islam (1997) floods are termed as “post-urban flood”. The interpretation of show nearly 5% discrepancy of flooded area estimation. This pre-and post-urban flood hydrographs showed that the could be due to the fact of cloud presence in the MOS-MESSR hydrographs were at their natural state in 1970, 1974, and 1980 data. In 1998, the flooded area was 53.2% of the total area when impervious surfaces were comparatively lower. As more which resembles to the result by Dewan et al. (2005). On the areas become impervious and more flood control works have other hand, in 2004, 43% of the study area was flooded in the been put in place, the hydrographs had been artificial thus, peak flood date of 24 July. As no more quantitative information peaking up earlier and remaining on the top of the danger level available on the area flooded in 2004, comparison was not (DL) for more days than the pre-urban times (Fig. 4). Rainfall possible. The flood maps of three years (e.g. 1988, 1998 and can be another cause of having worst flood in recent times. The 2004) are shown in Fig. 3. Estimation of built-up areas under comparative analysis of those pre- and post urban rainfall flood during those three biggest flood events revealed that confirmed that rainfall has not been changed significantly and in flooding in the urban places has been increasing at an alarming July pre-urban rainfall was much higher than that of post-urban. rate due to putting more properties in the previously floodplain However, a slight increase of rainfall in June and September in and rural lands. In 1988, 1484.1 hectare of built-up areas was post-urban situation has been observed (Fig. 4). flooded whereas in other two events this figure shot up to 2991 and 4503.2 hectares of land. 4.4 Integration of remote sensing and hydrological analysis Remote sensing analysis of land use/cover change demonstrated that there has been major increase of built-up 25000.0 areas over the past four decades in Greater Dhaka. It is clear from Table 1 that urban sprawl is being taken place in the study 20000.0 1960 1988 2005 area, producing more impervious surfaces, hence accelerating the rate and volume of surface runoff. The most important 15000.0 feature is that the flood peaks were going up very quickly and Area (ha) remained in the above of danger level for more days than the past events (e.g. pre-urban situation). On the other hand, rainfall 10000.0 records demonstrated very subtle changes in the post-urban condition. As impervious surfaces increased significantly, water 5000.0 levels in the surrounding rivers were augmented to be above of the normal flood level for longer time. So, it can be said that 0.0 increase of impervious surfaces as a result of land use/cover changes clearly enhanced the river flows. 5. Conclusions Land use/cover change analysis was performed using Land use/cover types topographic maps and multi-temporal remotely sensed data. A maximum likelihood of supervised classification was employed Fig 1 Trends of land use/cover change in Greater Dhaka. to quantify land use/cover changes in Greater Dhaka of Bangladesh. Water bodies, low-lying lands and agricultural 4.3 Analysis of flood hydrographs and rainfall lands in Greater Dhaka have been transformed at a substantial Flooding in the study area is the product of water levels of rate, and being converted to built-up lands. Thus, impervious the surrounding rivers and rainfall in the monsoon season surfaces are increasing and mounting flood risk potential. This (Dewan et al., 2005). Only studying three independent flood information was subsequently integrated with hydro- events by remote sensing and flood hydrographs can not meteorological data to analyze the impact of land use/cover substantiate the fact of land use/cover change consequences on changes on flood. The result demonstrated that flooding in flooding. Therefore, in this study we incorporated another three Greater Dhaka is aggravating and is attributed to the rigorous greatest flood events of 1970, 1974, and 1980 to establish the changes of land use/covers. hypothesis. The hydrographs of those past events termed as Figure 2 Multi-temporal Land use/cover maps. Figure 3 Inundation maps of Greater Dhaka, Bangladesh. References Yamagata, Y., Sugita, S. and Yasuoka, Y., 1997. Development of Vegetation-Soil-Water Index algorithms and Alphan, H., 2003. Land use change and urbanization in Adana, applications. Journal of Remote Sensing Society of Japan, Turkey. Land Degradation and Development, 14, pp. 575- 17(1), pp. 54-64. (in Japanese with English abstract) 586. Anderson, R., Hardy, E.E., Roach, J.T. and Witmer, R.E., 1976. A land use and land cover classification system for use with remote sensor data, USGS Professional Paper 964, Sioux Falls, SD, USA. Chilar, J., 2000. Land cover mapping of large areas from satellites: status and research priorities. International Journal of Remote Sensing, 21 (6&7), pp. 1093-1114. Dewan, A.M., Kankam-Yeboah, K. and Nishigaki, M., 2005. Flood mapping and damage evaluation in Greater Dhaka, Bangladesh with Remote Sensing. Asian Profile: an international journal, Vol. 33 (5), pp. 495-512. Dong, Y., Forster, B. and Ticehurst, C., 1997. Radar backscatter analysis for urban environments. International Journal of Remote Sensing, 18(6), 1351-1364. Flood Action Plan (FAP) 8A, 1991. Master Plan Study for Greater Dhaka Protection Project. Japan International Cooperation Agency, Dhaka. pp. A1-A7. Houghton R.A., 1994. A worldwide extent of land-use change. Bio-Science, 44, pp. 305–313. Khan, F.K. and Islam, N., 1964. High class residential areas in Dhaka city, 1608-1962, Oriental Geographer, January, 1- 40. Lillesand, T.M. and Kiefer, R.W., 1999. Remote Sensing and Image Interpretation, (4th edition), John Wiley and Sons, New York, pp. 724. Lopez, E., Bocco, G., Mendoza, M. and Duhau, E., 2001. Predicting land cover and land use change in the urban fringe a case in Morelia City, Mexico. Landscape and Urban Planning, 55, pp. 271-285. Muttitanon, W. and Tripathi, N.K., 2005. Land use/cover changes in the coastal zone of Bay Don Bay, Thailand using Landsat 5 TM data. International Journal of Remote Sensing, 26(11), pp. 2311-2323. Nirupama, N. and Simonovic, S. P., 2007. Increase of flood risk due to urbanization: a Canadian example. Natural Hazards, 40, pp. 25-41. Reid, R.S., Kruska, R.L., Muthui, N., Taye, A., Wotton, S. and Wilson, C.J., 2000. Land-use and land-cover dynamics in response to changes in climatic, biological and socio- political forces: the case of southwestern Ethiopia. Landscape Ecology, 15, pp. 339–355. Sado, K. and Islam, M.M., 1997. Satellite remote sensing data analysis for flooded area and weather study-case study of Dhaka City, Bangladesh. Annual Journal of Hydraulic Engineering, 41, pp. 945-950. Sala O.E., Chapin F.S., Armesto J.J., Berlow E., Bloomfield J. and Dirzo R., 2000. Biodiversity: Global biodiversity scenarios for the year 2100. Science, 287, pp. 1770–1774. Shi, P. J., Yuan, Y., Zheng J., Wang, Jing-Ai, Ge, Y. and Qiu G.Y., 2007. The effect of land use/cover change on surface runoff in Shenzhen region, China. Catena, 69, pp. 31-35. The Daily Star, 2004. http://www.thedailystar.net/2004/07/26/d4072601011.htm (last accessed May 20, 2007). Weng, Q., 2001. Modeling urban growth effects on surface runoff with the integration of remote sensing and GIS. Environmental Management, 28(6), pp. 737-748. Xian, G., Crane M. and Su, J., 2007. An analysis of urban development and its environmental impact on the Tampa Bay Watershed. Journal of Environmental Management, (in Press). 7 8 7 Danger level 7 6 6 6 Danger level Danger level Water level (pwd/m) 5 5 Water Level (m/pwd) Water Level (m/pwd) 5 4 4 4 3 3 3 2 2 2 1988 1998 2004 1970 1974 1980 1970 1974 1980 1 1 1 0 0 0 15 22 29 36 43 50 57 64 71 78 85 92 99 106 113 120 127 134 141 148 1 8 1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106 113 120 127 134 141 148 1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106 113 120 127 134 141 148 Days Days Days (a) River Buriganga: Pre-urban flood (d) River Balu: Pre-urban flood (a) River Buriganga: Post-urban flood 8 7 6 8 9 Danger level ater level (m/pwd) 5 7 8 6 Danger level 7 4 Water Level (m/pwd) Danger level Water Level (m/pwd) 6 3 5 W 5 2 4 1988 1998 2004 4 3 1 3 2 0 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145 153 1970 1974 1980 2 1 1 1988 1998 2004 Days 0 0 (d) River Balu: Post-urban flood 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145 153 1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103 109 115 121 Days Days 1800 (b) River Tongi: Pre-urban flood (b) River Tongi: Post-urban flood 1600 Pre urban (Total) Post urban (Total) 1400 8 9 7 1200 8 Danger level 6 7 Danger level 1000 Water level (m/pwd) Water level (m/pwd) 5 6 5 800 4 4 3 600 3 2 1970 1974 1980 2 1988 1998 2004 400 1 1 200 0 0 1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106 113 120 127 134 141 148 1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103 109 115 121 Days Days 0 June July August September (c) River Turag: Pre-urban flood (c) River Turag: Post-urban flood Mont hs (e) Pre and Post urban total monsoonal rainfall Figure 4 Pre- and Post-urban flood hydrographs and rainfall in Greater Dhaka of Bangladesh