A Bayesian Non-Stationary Framework for Bivariate Compound Flood Risk Assessment Along the Indian Coast

Compound flooding arises from the interaction of multiple flood drivers, particularly hydrologic and oceanographic drivers such as extreme precipitation (P) and elevated sea level or storm surge (SL/SS). These interconnected drivers often share common climate influences, and when they occur simultaneously or in close succession, the resulting joint probability of flooding can be substantially higher than under the assumption of independence. In this study, a bivariate non-stationary flood frequency analysis framework is developed to assess compound flooding risks along the Indian coast. The dependencies between precipitation and sea level/storm surge are modeled using copula-based approaches, while Bayesian inference is employed for parameter estimation of both marginal distributions and copula functions under non-stationarity. This enables robust uncertainty quantification while incorporating the influence of changing climate and ocean conditions. Joint return periods are evaluated for different compound flood scenarios (e.g., AND and OR cases), enabling a more realistic characterization of coastal hazard likelihood. Additionally, failure probabilities (FPs) are estimated to reflect the chance that at least one driver exceeds its critical threshold within standard infrastructure design lifetimes. The results highlight the importance of accounting for temporal changes and interdependence between precipitation and sea level/storm surge in coastal flood risk assessment. The findings provide actionable guidance for selecting resilient design criteria and support informed decision-making for coastal flood protection and long-term risk management.