Assessment of vegetation damage of the Sundarbans and inundation in the southwest coastal region of Bangladesh for three recent cyclones using satellite images

Abstract

Bangladesh is highly vulnerable to tropical cyclones due to its unique geographical location, flat topography, and high density of population. Over the last few decades, climate change and sea-level rise have been increasing the destructive impact of cyclones, their rainfall, and the storm surges they cause. The recent rapid intensification of tropical cyclonesin the Bay of Bengal mostly affected offshore islands, coastal areas, and coastal forests. Remote sensing plays a vital role in damage assessment over forest areas where accessibility is limited. The availability of high-resolution (~10m) multi-spectral satellite images of Sentinel-2 provides an opportunity to assess vegetation damages as well as SAR images of Sentinel-1 offer the development of near-real-time storm surge inundation maps. By using Sentinel-1 and-2 images, this study was carried out for vegetation damage assessment of Sundarbans and storm surge inundation mapping for 7 south coastal districts for the three recent cyclones: Fani, Bulbul, and Amphan.Images before and after the occurrence of each cyclone were used for calculating the Normalized Difference Vegetation Index (NDVI) and areas under different NDVI classes. To assess the cyclonic storm surge inundation, SAR Water Index (SWI)was calculated using Sentinel-1 data, followed by a pre-processing phase, a supervised classification method to obtain the binary mask of the water bodies and calculating the extents. The validation for each of the obtained results was done by comparing the areas of classified image analysis with the reference value to obtain confusion matrices, overall accuracy, and kappa indexes.It has been found from the analysis of NDVI that about 10% area of Bangladesh’s part of the Sundarbans (approximately 500 sq. km.) was affected due to these cyclones.In contrast, SWI analysis indicates that over 3,000 sq. km. of land,which accounted for 21% of the study area,was inundated after these cyclones and destroyed homes, polders, embankments, roads, bridges, culverts, and electricity poles.Moreover, the higher efficiency and promptness of the overall accuracy percentage (<92%) and kappa index (<0.80) showed the strength of agreement is almost perfect in this study. Vegetation damage assessment can provide valuable assistance to the decision-makers of concerned authorities for mangrove forest management as well as quantifying the spatial heterogeneity of cyclone flooding, which is critical for both an immediate response and understanding of their needs.