Flood susceptibility mapping using machine learning boosting algorithms and Geospatial techniques – A Case Study of Subarnarekha River Basin

Floods are a major natural disaster, particularly in the Subarnarekha River basin in eastern India, where severe monsoon season flooding poses significant risks to communities, agriculture, and infrastructure. Preventing floods is challenging, but technological advancements like machine learning in geospatial analysis offer promising methods for identifying and managing flood-prone areas. This study employs machine learning boosting algorithms and 15 conditioning factors, such as elevation, rainfall, and drainage density, to assess flood susceptibility in the Subarnarekha River basin. Using 25 years of historical flood data (1998–2022) for training and validation, the models are evaluated using metrics like precision, recall, F1 score, and area under the curve (AUC), with AUC values ranging from 0.91 to 0.95. Adaboost proves to be the most effective model with a 95% AUC, followed by XGboost (93%), Gradient Boosting (92%), Catboost (92%), and Stochastic Gradient Boosting (91%). The analysis reveals varying flood hazard conditions, with low hazards in the upper reaches and high susceptibility in coastal areas due to heavy rainfall and runoff. This study highlights the value of machine learning techniques in improving flood risk assessment and management strategies. By leveraging these advanced methods, authorities can develop more effective flood mitigation plans and enhance early warning systems. This integration of technology provides a proactive approach to disaster management, potentially saving lives and reducing economic losses in flood-prone regions.