Advanced Filtering Techniques for Enhanced Flood Mapping Using Multispectral Imaging

The study investigates the utilisation of multispectral image processing techniques for flood mapping in the Goalpara District of Assam, India. The region experiences recurrent flooding during the monsoon season, impacting the Brahmaputra River and its surroundings. Dual-polarized Sentinel-1 Synthetic Aperture Radar (SAR) data, encompassing Vertical-Horizontal (VH) and Vertical-Vertical (VV) polarisation methods, were employed to assess flood conditions before, during, and after events occurring between April and June 2022. Various parameters, including the computation of the Normalized Difference Water Index (NDWI) and Modified Normalized Difference Water Index (MNDWI), were calculated using the Otsu Algorithm. Filtering methods such as Refined Lee, Focal Median, and Gamma Map were applied to enhance flood mapping accuracy. The study underscores the efficiency of multispectral image processing, particularly the Refined Lee filter, in accurately delineating flood extent compared to other filters. Flood mapping was facilitated through thresholding techniques applied to SAR data. Furthermore, the study identifies severe bank erosion in the Brahmaputra channel, signifying significant implications for river morphology and sediment transport dynamics. Leveraging remote sensing techniques, this research provides essential insights for proactive flood management, facilitating informed decision-making and mitigation strategies in the region. The study achieved an overall accuracy of 90% and a Kappa coefficient of 0.43, indicating high-quality flood map and ground truth data.