Author(s): Diego Armando Urrea-Mendez; Dina Vanessa Gomez-Rave; Manuel Del Jesus Penil
Linked Author(s):
Keywords: Flood frequency estimation; Hydrological–hydraulic modelling; Surrogate models; Multivariate analysis
Abstract: Flood-frequency estimation has traditionally relied on univariate approaches that assume linear relationships between rainfall, discharge, and often inundation extent, overlooking the nonlinear and multivariate nature of fluvial flooding. Continuous hydrological–hydraulic modelling provides a physically based representation of flood processes, but its use is often limited by high computational requirements and the need for long, high-quality datasets. This study introduces an integrated framework that combines continuous hydrological–hydraulic simulation with surrogate modelling and multivariate techniques to represent the joint behaviour of key hydrometeorological variables. The surrogate components approximate the outputs of continuous models with substantially lower computational demand, enabling the identification of combinations of precipitation, discharge, hydrograph volume, and spatial patterns associated with extreme events. The central hypothesis of this work is that embedding surrogate models within a multivariate structure improves the estimation of flood frequency compared with traditional approaches.
Year: 2026