A Robust Calibration and Evaluation Framework for Dynamic Catchment Characteristics in Hydrological Modeling

Abstract

Hydrological models often face challenges in accurately simulating hydrological processes within dynamic catchments due to simplifications of model structure. In a dynamic catchment where hydrological processes exhibit significant intra-annual or inter-annual variability, accurately capturing dynamic behaviours across different flow regimes is still challenging for models. To address these challenges, this study investigates calibration issues in dynamic catchments with a focus on two key aspects: the influence of objective function design on flow-phase-specific performance, and the limitations of sub-period calibration with dynamic parameters. Seven calibration experiments were designed to explore issues related to time-invariant parameters, objective function configurations, parameter correlations, dimensionality in global optimization, and abrupt parameter shifts. The experiments were conducted using the MOPEX dataset, which includes 219 basins across the United States, and were evaluated based on performance metrics, as well as state variables and fluxes. Among all calibration schemes, sub-period calibration with dynamic parameters exhibited the most reliable performance. Static parameter approaches often averaged catchment responses and poorly represented extreme flows, whereas enabling temporal variability to only a subset of parameters yielded limited improvement. In contrast, multi-parameter dynamic schemes significantly improved NSE and LNSE values and enhanced parameter transferability across flow phases, where the high-dimensional calibration strategy balanced dynamic adaptability with physical consistency, while the parallel calibration maintained accuracy through gradual parameter transitions despite higher variability in some catchments. This study demonstrates that sub-period calibration with dynamic catchment characteristics outperforms traditional static parameters by effectively capturing flow-regime variability and sustaining robust performance under changing catchment conditions, offering a generalizable solution for simulating hydrological processes in dynamic catchments.