An evaluation of flood inundation mapping from MODIS and ALOS satellites for Pakistan

Abstract

The paper presents a moderate resolution imaging spectroradiometer (MODIS) time-series imagery-based algorithm for detection and mapping of seasonal and annual changes in flood extent, and tests this using the flooding of the Indus River Basin in 2010 – one of the greatest recent disasters that affected more than 25 million people in Pakistan. The algorithm was applied to produce inundation maps for 10 annual flood seasons over the period from 2000 to 2011. The MODIS flood products were validated in comparison with advanced land observing system (ALOS) sensors, which have both advanced visible and near infrared radiometer and phased array type L-band synthetic images using the flood fraction comparison method. A simple threshold method is created to cluster the data to identify the flood pixels in the imagery. Calculations are then made to estimate a flood area for each resolution. A statistical study is performed to analyze false positive and false negative rates using the ALOS sensors as ‘ground truth’. Comparison of two flood products at a grid size of 10 km resulted in the coefficient of determination range of 0.72–0.97. This research points to a relevant spatial resolution that could be effectively used to obtain accurate mapped products of the extent of the inundated area. The approach can be used to quantify the damage caused by floods.