A scalable data driven geospatial framework for climate risk assessment

Abstract

Abstract Traditional flood risk management approaches often rely on historical data, limiting their ability to account for the increasing severity and frequency of climate-induced hazards. This study presents a scalable, data-driven framework that integrates geospatial analysis and machine learning to dynamically assess climate risks. The framework enables decision-makers to identify vulnerabilities, quantify flood risk under evolving climate scenarios, and develop informed adaptation strategies. Using bias-corrected CMIP5 climate projections as use case, the framework is demonstrated through a case study in Texas, where community flood risk prediction is done under multiple emission scenarios. Results indicate that under RCP 8.5, community vulnerability is projected to increase by 14%, leading to an estimated 28% rise in economic damages ($1.8B per decade by 2050) and heightened socio-economic disruptions, including displacement and infrastructure failures. By identifying the most influential climatological factors that impact community resilience, our approach stresses the urgent need for global action to mitigate extreme climate scenarios. It shows the scalability and flexibility of the framework, emphasizing its potential as decision-support tool, and a step towards a digital twin system for climate risk assessment and adaptation planning.