Trivariate flood frequency analysis using copula functions: assessing non-stationary hydrological risks under climate change

Abstract

This study presents a trivariate flood frequency analysis (FFA) framework that integrates copula theory with both parametric and non-parametric marginal distributions to model the joint behavior of peak discharge, flood volume, and duration. The approach captures nonlinear interdependencies among flood characteristics without assuming specific marginal forms, enhancing realism in flood modeling. The model was applied to long-term hydrological records from the Zayandeh-Roud Basin in central Iran, a semi-arid watershed facing intensifying climatic and anthropogenic pressures. Performance evaluation based on root mean square error (RMSE), Akaike information criterion (AIC), and Bayesian information criterion (BIC) confirmed that the proposed trivariate model outperforms traditional univariate and bivariate approaches in predictive accuracy and parsimony. The framework also supports real-world decision-making by improving the reliability of flood risk estimation, infrastructure planning, and reservoir operations in non-stationary conditions. Future scenarios generated using the LARS-WG weather generator and MRI-ESM2-0 model under CMIP6 SSP2-4.5 and SSP5-8.5 pathways indicate increased compound flood risks in the coming decades. Moreover, the study highlights the importance of accounting for structural uncertainty in climate projections to ensure more robust and adaptive long-term flood risk management strategies under changing environmental conditions.