Multivariate design events for compound flooding analysis in estuaries

Understanding Compound Flood (CF) hazard in estuaries requires moving beyond univariate approaches toward multivariate frameworks that capture the joint behavior of multiple drivers. Although the relevance of such approaches is increasingly recognized, most existing methods remain limited to bivariate analyses. Extending to higher dimensions poses conceptual and computational challenges, particularly in estimating Joint Return Periods (JRP) and defining representative design events. This limitation is especially relevant in estuarine systems, where the hazard may result from the combined action of interacting drivers — including precipitation, river discharge, storm surge, and waves — that rarely occur in isolation. In this context, restricting the analysis to two variables may overlook relevant dependencies, reinforcing the need for models that account for higher-order interactions. This study examines the role of multivariate dependence structures within a six-dimensional case-study, comparing different copula families to evaluate their suitability for CF hazard analysis. Focusing on the Santoña estuary, we assess how model choice influences the estimation of joint events and the selection of representative conditions for design. Among the models explored, vine constructions incorporating extreme-value copulas led to more coherent joint estimates, offering improved stability across dependence scenarios. Rather than seeking a universally optimal model, the analysis illustrates how the choice of dependence structure can influence the representation of joint extremes. The proposed framework supports physically interpretable and statistically consistent multivariate design events for compound hazard analysis in coastal settings. • R-vine copulas capture dependence in compound flooding best • Multivariate return periods capture uncertainty with a set of equally probable events • Mixture distributions are essential to capture marginal distributions