Flood mapping, damage assessment, and susceptibility zonation in northeastern Bangladesh in 2022 using geospatial datasets
Mizanur Rahman, Mohammad Kamruzzaman, Limon Deb, H. M. Touhidul Islam
Abstract
This study assesses flood inundation, impacts, and susceptibility zones in northeastern Bangladesh during the 2022 flood. The region is highly vulnerable to recurrent flooding Due to its geographic position and climate change impacts. The Sentinel-1 SAR data on the Google Earth Engine (GEE) platform was used to generate flooded areas using a simple change detection technique with thresholding. This analysis was further supported by incorporating cropland, population, national highway, and DEM datasets for a comprehensive damage assessment. Findings show that 55.76 % (10,993.09 km 2 ) of the area was inundated, impacting 10.69 million people and causing severe displacement and health hazards. Sylhet, Kishoreganj, and Brahmanbaria districts were the most affected, with 2.73 million impacted in Sylhet alone. Additionally, 67.87 % of agricultural land was flooded, particularly in Sunamganj, and 43.38 % of national highways (535.08 km 2 ) were damaged. A flood susceptibility zonation map identified high-susceptibility areas like central Sunamganj and parts of Kishoreganj to assist authorities in resource allocation and mitigation. The flood extent model achieved strong predictive accuracy (AUC: 0.97 % RF, 0.96 % LR, and 0.94 % DT), providing crucial insights for regional flood management and guiding communities with limited modeling capacities. • 55.76 % of the northeastern region was inundated, affecting Sunamganj, Sylhet, Kishoreganj, Netrokona, Brahmanbaria, Habiganj, and Moulvibazar. • 10.69 million people were affected, with Sylhet as the worst-hit district. • 67.87 % of croplands were impacted, with Sunamganj was the most affected. • 43.38 % of national highways were affected by the flood. • Flood-prone areas include Sunamganj (2109.46 km 2 ), Kishoreganj (1041.73 km 2 ), and Sylhet (1033.20 km 2 ). • Accuracy of flood prediction models: Random Forest (97 %), Linear Regression (96 %), and Decision Tree (94 %). • The study provides essential flood assessment insights for regions with limited modeling capabilities.