Does Flash Flood Model Performance Increase with Complexity? Signature and Sensitivity-Based Comparison of Conceptual and Process-Oriented Models on French Mediterranean Cases

Abstract

We compare three hydrological models of different complexities, GR4H (lumped, continuous), SMASH (distributed, continuous), and MARINE (distributed, event-based), for Mediterranean flash flood modeling. The objective was to understand how differently they simulate the catchment’s behavior, in terms of outlet discharge and internal dynamics, and how these can help to improve the relevance of the models. The methodology involved global sensitivity analysis, calibration/validation, and signature comparison at the event scale with good performances. For all models, we found transfer parameters to be sensitive in the case of Gardon and production parameters in the case of Ardeche. The non-conservative flow component of GR4H was found to be sensitive and could benefit the distributed models. At the event scale, the process-based MARINE model at finer resolution outperformed the two continuous hourly models at flood peak and its timing. SMASH, followed by GR4H, performed better in the volume of water exported. Using the operational surface model SIM2 to benchmark the soil moisture simulated by the three models, MARINE (initialized with SIM1) emerged as the most accurate. GR4H followed closely, while SMASH was the least accurate. Flexible modeling and regionalization should be developed based on multi-source signatures and worldwide physiographic databases.

Building Tools for further Investigating Acid Mining Production: Intercomparison of Four Hydrological Model Versions through a Scoring Technique on the Niger River Basin, in west Africa

“…Three years earlier, [37] performed a comparison of hydrological simulations using synthetic rainfall data. [38] includes an approach called "event scale", which has inspired [39] in constructing a comparison method of multiple models through criteria that involved a Monte Carlo simulation.…”