A Flood Detection and Mapping Algorithm Using MODIS Data: Assessment of Extreme Flooding Events in Eastern Ganga Plains (2000-2015)

Abstract:

Flood, a hydrological extreme, is a dominant and frequent phenomena over the eastern Ganga Plains comprising of alluvial plains of Bihar and adjoining Nepal Himalaya. Flood affects major parts of Bihar where Gandak and Koshi are the major tributaries of Ganga River causing inundation during the monsoon season. Due to heavy rainfall in the Eastern Himalaya and adjoining regions, the river discharge increases several folds causing severe flood in plains. Moderate Resolution Imaging Spectroradiometer (MODIS) derived data at 250 m resolution (year 2000-2015) have been used to identify flood water and calculate daily water fraction (water cover) using model adopted from previous studies. During the monsoon season, cloud cover in daily images is found to be extremely high leading to lot of gaps in the form of missing data. To account for missing grid cell values, an adaptive polynomial filter (Savitzky-Golay) have been used to fit the time series of daily data for each grid cell. The missing values in daily images have been filled with calculated values to create daily time series of flood water. Landsat data at 30 m grid resolution have been used to verify flood water detection algorithm used in this study. Time series analysis of satellite derived data reveal a strong spatial and temporal variation in the extent, duration and frequency (inter-annual and intra-annual) of flooding event over the study region. Statistical analysis of IDF (intensity, duration, and frequency) and trend have been carried out to identify regions which show greater flood risk. Reoccurrence interval and length of flooding event in the study region is found to be high compared to other river basins in the western India. Based on the historical occurrence of flood, the study area have been classified into different flood hazard zones where flood mitigation and management need to be prioritized. MODIS based flood monitoring and mapping model used in this study can be used for monitoring and assessment of flooding events in near-real time over the study region.