Indian Journal of Geo Marine Sciences Vol. 48 (03), March 2019, pp. 363-368 Thresholding technique for flood extent mapping using dual polarization ENVISAT ASAR data P. Gopi 1 , Kiran Yarrakula 1* , & Y.R.S. Rao 2 1 Centre for Disaster Mitigation and Management, Vellore Institute of Technology, Vellore, India, 2 National Institute of Hydrology, Kakinada * [E-mail: kiranyadavphysik@gmail.com] Received 02 August 2017; revised 08 January 2018 The research paper explores an optimum water extraction technique using synthetic aperture radar (SAR) data to give the flood inundation map for irregular flood scenario, in which the planned algorithm depends on thresholding technique using MRS amplitude SAR data. For classification, supervised maximum likelihood method is used, and gamma filtering technique was used for speckle removal. Since water layer in SAR data has low amplitude value, thresholding technique for amplitude SAR data could efficiently take out water layer. The mean backscatter basics of water layers in SAR data is not to be compared with inundated regions caused by sharp release and is not considered as a terrain value. The speckle filtering method was used for the power generated noise subtraction using a window size 5x5. The threshold values were calculated from backscatter values and their average was taken as threshold, which gives clear difference of backscatter values between the land and the water in the river and the flood plain areas. The maximum likelihood classifier provides 99% overall accuracy. The Landsat 5 TM data was used in this research for delineating water feature and flood accurately. The results obtained from SAR data were validated with Landsat TM data. Finally, the flood inundation map using optical data and SAR data was generated. [Keywords: Landsat TM, Synthetic aperture radar (SAR), Thresholding technique, Flood extent] Introduction Flood is the form of local inundation due to heavy rain or due to accumulation of runoff/discharge after heavy rain in the catchments due to breach of embankment. In India, during 1953 to 2004, about 397 Mha area has been affected due to floods as reported by CWC (2005). During these years, about 185 Mha crop land has been affected, damaging 6.4 crore houses, affecting lives of 1.7 billion people and killing 49 lakh cattles. Thus, the understanding of rainfall pattern, effect of climate change, river flow dynamics and mapping of flood plain during monsoon season is the key to minimize the risk of flood around a river basin. The present study concentrates on the analysis of discharge pattern of the Krishna River and estimates the flood inundated areas in and around the river basin 1 . Synthetic Aperture Radar(SAR) data proved invaluable for the spatial characterization of floods 2 due to its all-weather/ all-day imaging capability and the low backscatter recorded for water due to specular reflection 3 . Two kinds of information can be extracted from SAR image: First one is the flood inundation area extent mapping using normal MLC (maximum likelihood classifier) method 4 and second one is the flood inundation area by thresholding method. The SAR derived flood extent can be effectively used as a calibration target to constrain the classification and minimize its prediction uncertainty 5 . Study area description The study area selected was severely affected by floods in 2006 and 2009. At Krishna barrage, Vijayawada, AP, floods occurred 13 times from 1903 to 2009. The study area of lower Krishna lies between latitude 15°35'20.18"N to 16°49'44.92"N and Longitude 79°58'30.75"E to 81°14'38.07"E. Data used Two sets of SAR data at pre-and post-flood stages have been planned based on the flood occurrence. The first date coincided with the pre-flood stage (03 rd July 2006) and the second date was acquired post flood stage (26 th August 2006). The two data sets provide normal water features and flood inundated areas accurately. For this study, ENVISAT ASAR images were acquired from the European Space Agency. ENVISAT ASAR imageries during 2006 floods in