Search

Estimating spatially distributed hydrological parameters in ungauged catchments poses a challenging regionalization problem and requires imposing spatial constraints given the sparsity of discharge data. A possible approach is to search for a transfer function that quantitatively relates physical descriptors to conceptual model parameters. This paper introduces a Hybrid Data Assimilation and Parameter Regionalization (HDA-PR) approach incorporating learnable regionalization mappings, based on either multi-linear regressions or artificial neural networks (ANNs), into a differentiable hydrological model. This approach demonstrates how two differentiable codes can be linked and their gradients chained, enabling the exploitation of heterogeneous datasets across extensive spatio-temporal computational domains within a high-dimensional regionalization context, using accurate adjoint-based gradients. The inverse problem is tackled with a multi-gauge calibration cost function accounting for information from multiple observation sites. HDA-PR was tested on high-resolution, hourly and kilometric regional modeling of 126 flash-flood-prone catchments in the French Mediterranean region. The results highlight a strong regionalization performance of HDA-PR especially in the most challenging upstream-to-downstream extrapolation scenario with ANN, achieving median Nash-Sutcliffe efficiency (NSE) scores from 0.6 to 0.71 for spatial, temporal, spatio-temporal validations, and improving NSE by up to 30% on average compared to the baseline model calibrated with lumped parameters. ANN enables to learn a non-linear descriptors-to-parameters mapping which provides better model controllability than a linear mapping for complex calibration cases.

Classical calibration methods in hydrology typically rely on a single cost function computed on long-term streamflow series. Even when hydrological models achieve acceptable scores in NSE and KGE, imbalances can still arise between overall model performance and its ability to simulate flood events, particularly flash floods. In this study, the potential of using multi-scale signatures is explored to enhance multi-criteria calibration methods for spatially distributed flood modeling, which remains considerable challenges. We present a novel signatures and sensitivity-based calibration approach implemented into a variational data assimilation algorithm capable to deal with high dimensional spatially distributed hydrological optimization problems. It is tested on 141 flash flood prone catchments mostly located in the French Mediterranean region. Our approach involves computing several signatures, including flood event signatures, using an automated flood segmentation algorithm. We select suitable signatures for constraining the model based on their global sensitivity with the input parameters through global signature-based sensitivity analysis (GSSA). We then perform two multi-criteria calibration strategies using the selected signatures, including a single-objective optimization approach, which transforms the multi-criteria problem into a single-objective function, and a multi-objective optimization approach, which uses a simple additive weighting method to select an optimal solution from the Pareto set. Our results show significant improvements in both calibration and temporal validation metrics, especially for flood signatures, demonstrating the robustness and delicacy of our signatures-based calibration framework for enhancing flash flood forecasting systems.