Comparing AI-Based Flood Risk Estimation and Risk Perception Among the Community: A Study from a Floodplain Region of India

The methods of flood risk estimation are undergoing a paradigm shift due to the emergence of Geospatial Artificial Intelligence (Geo AI). Undoubtedly, it is capable of estimating flood hazards more precisely than previous methods, and it has the potential to serve as a significant tool for spatial planning. However, the execution of such planning is very difficult without community acceptance. Therefore, integrating AI-based flood risk estimation with community perception can be a pragmatic approach to building a disaster-resilient society. Apart from this importance, existing literature is rarely focused on AI and community perception in flood risk studies. This research develops a framework to compare flood risk estimation and community risk preconception. The framework is applied through a case study of villages of North Bihar, located at floodplain region of India. The region is affected by severe floods almost every year due to the rapid onset of rainwater from the Himalayas. The study follows three steps: Frist, it develops a flood hazard model using GIS and Remote Sensing derived geohydrological parameters, along with Google Earth Engine (GEE) based Random Forest Techniques. The cluster of building footprints within severe flood hazard is identified as high flood risk. The study is validated through past damage data and media reports. In the second step, risk perception is evaluated, with data collected through a field survey of 200 households in the region. In the third step, the AI-based estimated risk and the community’s risk perception are compared using spatial matrix analysis. The findings indicate a coexistence of agreement and disagreement between AI-based flood risk estimates and the community’s risk perception. Finally, the study suggests enhancing disaster communication to bridge the gap between estimated risk and community perception.