A spatially distributed hydrological model, dedicated to flood simulation, is developed on the basis of physical process representation (infiltration, overland flow, channel routing). Estimation of model parameters requires data concerning topography, soil properties, vegetation and land use. Four parameters are calibrated for the entire catchment using one flood event. Model sensitivity to individual parameters is assessed using Monte-Carlo simulations. Results of this sensitivity analysis with a criterion based on the Nash efficiency coefficient and the error of peak time and runoff are used to calibrate the model. This procedure is tested on the Gardon d'Anduze catchment, located in the Mediterranean zone of southern France. A first validation is conducted using three flood events with different hydrometeorological characteristics. This sensitivity analysis along with validation tests illustrates the predictive capability of the model and points out the possible improvements on the model's structure and parameterization for flash flood forecasting, especially in ungauged basins. Concerning the model structure, results show that water transfer through the subsurface zone also contributes to the hydrograph response to an extreme event, especially during the recession period. Maps of soil saturation emphasize the impact of rainfall and soil properties variability on these dynamics. Adding a subsurface flow component in the simulation also greatly impacts the spatial distribution of soil saturation and shows the importance of the drainage network. Measures of such distributed variables would help discriminating between different possible model structures.

Flash floods are natural phenomena with environmental, social and economic impacts. To date, few numerical models are able to simulate hydrological processes at catchment scale at a reasonable time scale to describe flash events with accurate details. Considering a ~810 km² Mediterranean river coastal basin (southwestern France) as a study case, the objective of the present study was to assess the ability of the sub-daily module of the lumped Soil and Water Assessment Tool (SWAT) to simulate discharge (1) time-continuously, by testing two sub-basin delineation schemes, two catchment sizes, and two output time-steps; and (2) at flood time-scale, by comparing the performances of SWAT to the performances of the event-based fully distributed MARINE model when simulating flash flood events. We showed that there was no benefit of decreasing the size of the minimum drainage area (e.g., from ~15 km² down to ~1 km²) when delineating sub-basins in SWAT. We also showed that both the MARINE and SWAT models were equally able to reproduce peak discharge, flood timing and volume, and that they were both limited by rainfall and soil data. Hence, the SWAT model appears to be a reliable modelling tool to predict discharge over long periods of time in large flash-flood-prone basins.

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 regression 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 data sets 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. Multiple evaluation metrics based on flood‐oriented hydrological signatures also indicate that the use of an ANN leads to better performances than a multi‐linear regression in a validation context. ANN enables to learn a non‐linear descriptors‐to‐parameters mapping which provides better model controllability than a linear mapping for complex calibration cases.

To advance the discovery of scale-relevant hydrological laws while better exploiting massive multisource data, merging artificial intelligence with process-based modeling has emerged as a compelling approach, as demonstrated in recent lumped hydrological modeling studies. This research proposes a general spatially distributed hybrid modeling framework that seamlessly combines differentiable process-based models with neural networks. Specifically, we focus on hybridizing the hydrological module – built atop a differentiable kinematic wave routing over a flow direction grid – using a process-parameterization network that refines internal water fluxes, with all conceptual parameters estimated by a regionalization network trained simultaneously. We evaluate flood modeling performance and analyze the interpretability of learned conceptual parameters and corrections of internal fluxes using two high-resolution datasets (dx=1 km and dt=1 h). The first dataset involves 235 catchments in France, used for local calibration–validation and model structure comparisons between the classical Génie Rural (GR)-like model and the hybrid approach. The second dataset presents a challenging multi-catchment modeling setup in flash-flood-prone areas to demonstrate the framework's regionalization learning capabilities. The results show that the hybrid models achieve superior accuracy and robustness compared to classical approaches in both spatial and temporal validation. Analysis of the spatially distributed parameters and internal fluxes reveals the hybrid models' nuanced behavior, their adaptability to diverse hydrological responses, and their potential to uncover physical processes.

A method of multiple working hypotheses was applied to a range of catchments in the Mediterranean area to analyse different types of possible flow dynamics in soils during flash flood events. The distributed, process-oriented model, MARINE, was used to test several representations of subsurface flows, including flows at depth in fractured bedrock and flows through preferential pathways in macropores. Results showed the contrasting performances of the submitted models, revealing different hydrological behaviours among the catchment set. The benchmark study offered a characterisation of the catchments' reactivity through the description of the hydrograph formation. The quantification of the different flow processes (surface and intra-soil flows) was consistent with the scarce in situ observations, but it remains uncertain as a result of an equifinality issue. The spatial description of the simulated flows over the catchments, made available by the model, enabled the identification of counterbalancing effects between internal flow processes, including the compensation for the water transit time in the hillslopes and in the drainage network. New insights are finally proposed in the form of setting up strategic monitoring and calibration constraints.