A change detection approach to flood inundation mapping using multi-temporal Sentinel-1 SAR images, the Brahmaputra River, Assam (India): 2015-2020

Flood is one of the major disasters occurring worldwide, which occurs frequently and affects many lives and property. In the present study, an automatic flood inundation mapping approach based on Sentinel-1 SAR datasets was used for the assessment of monsoonal flood events of the Brahmaputra River from 2015 to 2020. Google Earth Engine was used to prepare potential flood inundation maps using a change detection technique by processing SAR images. In total, around 144 SAR images were analysed for the study period, and it was found that the flood inundation extent was more in 2015 and 2016 at about 6 lakh hectares; thereby, it decreased in 2017 and 2018 near to 3.5 lakh hectares. Again, an increase in inundation extent was observed of about 6 lakh hectares in 2019 and 2020, respectively. The results were evaluated by applying binarisation thresholding, removing permanent water bodies and shadows from SAR images to delineate the actual flooded areas. It shows an encouraging overall validation accuracy of 93.6% and 95.15% for the pre-flood events of 2019 and 2020, respectively. There was a change in the trend of inundation extent observed in 2015, and it was confirmed with the Integrated Multi-Satellite Retrievals for GPM (IMERG) precipitation dataset. The results were further analysed for damage assessment, and it was figured out that the flood event in July 2020 resulted in the highest crop area affected. The present study shows the technological advancements over the traditional approach of flood mapping and focuses on rapid flood assessment. The generated flood extent database can be used further by the hydrologist to generate the inundation probability maps based on the forecasting rainfall and hydrological model generated river discharge measurements.