Characterizing Complex Hydrological Extremes: A Framework Integrating Nonstationarity Assessment and Copula-Based Analysis

Robust flood frequency analysis (FFA) to assess the risk of extreme floods is achievable if multiple flood characteristics, such as peak, volume, and duration, are considered. This study presents a novel framework that integrates nonstationarity (NS) detection with univariate and bivariate assessments of flood extremes using copula functions under both disjunction and conjunction scenarios. Unlike most existing studies that adopted either univariate block maxima (BM) or peaks-over-threshold (POT) approaches in isolation, this work provides a comprehensive comparison by employing both approaches within a copula-based bivariate framework across multiple catchments, addressing a key methodological gap in FFA research. The framework was applied to 31 physioclimatically diverse catchments across four major river basins in west-central India. Results reveal spatially heterogeneous trends and frequent occurrences of NS, even in the absence of statistically significant trends. The POT approach proved more effective in accounting for multiple extreme events within a season or year, yielding more-conservative and -reliable risk estimates with comparatively narrower uncertainty bounds. Dependence analysis using multiple association measures showed the strongest correlation for peak–volume (P-V) and the weakest for peak–duration (P-D) pairs. Bivariate FFA indicated that Clayton and Gumbel copulas provided the best fit across catchments and flood characteristic pairs. Pronounced differences between univariate and bivariate return levels highlight the importance of POT-based multivariate analysis for characterizing complex hydrological dynamics. Based on the results, the catchments in the Narmada basin, India, exhibited the highest flood risk in both univariate and bivariate assessments, whereas those in the Tapi basin, India, had the highest joint risk for peak and volume floods. The proposed approach provides a realistic and regionally adaptable framework for flood quantile and risk estimation. By introducing a generic methodology for comprehensive FFA, this study facilitates the development of improved early warning systems and the design of climate-resilient water infrastructure across diverse hydrological regions.