Publication

Abstract : Densely populated urban centers like Coimbatore city, India, with overlooked seismic events, require robust seismic risk assessments that extend beyond standard models. This study develops an integrated framework to evaluate hazard, vulnerability, and site- effect amplification potential across 24 indicators, explicitly incorporating preparedness as a moderating factor in vulnerability. A Convolutional Neural Network-Extreme Gradient Boosting (CNN–XGB) hybrid model was used to map eleven physical indicators for site-effect amplification, demonstrating strong performance (precision, recall, and F-1 score of 0.82) by combining deep learning spatial recognition with ensemble predictive accuracy. The hazard map was prepared using the Deterministic Seismic Hazard Assessment (DSHA) method. An integrated vulnerability map was developed using the Analytic Network Process (ANP), to synthesize ten interdependent physical and social vulnerability indicators, including household density, building intensity, demographics, socio-economics, and emergency service accessibility. Preparedness was mapped using ward-level data from the 27-item Individual Earthquake Readiness Index (IERI) and the 20-item Community Earthquake Readiness Index (CERI), based on survey responses. Spatial data from government records, Sentinel and SRTM (30-meter resolution) were rasterized, classified by the Jenks method, and integrated into a high-resolution seismic risk map. Results reveal critical hotspots in the central and southwestern wards, where high hazard and vulnerability coincide with low-to-moderate preparedness. Findings emphasize the need for preparedness-inclusive models enabling proactive, location-specific policy interventions, such as extending workplace readiness to households, shared emergency-box systems, equitable leadership promotion, and retrofitting. This adaptable framework supports anticipatory disaster risk reduction strategies in rapidly urbanizing regions by strengthening social resilience.