A Non-Stationary Multivariate Framework for Assessing Compound Coastal Hazards at Global Scales

Coastal regions are increasingly exposed to compound hazards driven by the joint occurrence of extreme sea levels, waves, river discharge, and atmospheric forcing, with risks further amplified by long-term sea-level rise. Accurately quantifying these low-probability, high-impact events requires statistical frameworks capable of representing both multivariate dependence and non-stationary behavior across space and time. Here, we present an integrated approach for global to regional coastal hazard assessment that combines non-stationary extreme value analysis with multivariate dependency modeling. The framework builds on transformed-stationary representations of evolving marginal extremes and incorporates time-varying dependence structures to capture changes in cross-hazard relationships under shifting climate conditions. Event-based sampling strategies and statistical diagnostics are used to isolate relevant extremes and assess the significance of observed trends and uncertainties. Applied to large-scale datasets of coastal and hydrometeorological variables, the methodology reveals substantial temporal and spatial variability in compound hazard characteristics, highlighting the limitations of stationary and univariate assumptions. Ongoing developments extend the framework toward a unified multihazard modeling chain that consistently integrates oceanic, atmospheric, and terrestrial drivers. By embedding diverse physical processes within a coherent statistical structure, this work advances the representation of compound coastal extremes and provides a robust foundation for next-generation hazard assessments. The proposed approach supports the development of more realistic risk scenarios, offering critical insights for adaptation planning and resilience strategies under present and future climate conditions.