The Role of Explainable AI in Flood Modelling and Risk Assessment in the Era of Climate Change: A Systematic Review

Abstract

Flood risk is intensifying under climate change, making accurate and transparent prediction tools a global priority. While artificial intelligence (AI), machine learning (ML) and deep learning (DL) have shown strong performance in flood forecasting and susceptibility mapping. Hence, their black-box nature limits stakeholder trust and operational use. This review systematically examines the emerging role of explainable AI (XAI) in flood analysis under changing climate conditions. A structured search was conducted across Scopus, Web of Science, IEEE, Springer, and ScienceDirect, followed by PRISMA screening. Sixty-four eligible studies were analysed and grouped into four thematic domains: (i) flood forecasting, (ii) hazard and susceptibility mapping, (iii) integration with climate projections, and (iv) decision support systems. Across these domains, SHAP, LIME and attention mechanisms emerged as the most widely applied XAI techniques, improving model interpretability and stakeholder engagement. However, some key gaps persist, which includes weak operational adoption, limited integration with climate change scenarios, data scarcity in developing regions and a lack of user-oriented visualisation tools. This review concludes that XAI provides a promising pathway toward more trustworthy and actionable flood risk assessment but calls for standardised frameworks, stronger linkage with climate projections, and policy-friendly outputs to bridge research and practice.