A B S T R A C TIn this article, we describe the design and the validation of the Mescan precipitation analysis system developed for climatological purposes under the EURO4M project. The system is based on an optimal interpolation algorithm using the 24-h aggregated gauge measurements from the surface network. The background fields are the total accumulated precipitation forecasts at different resolutions from the ALADIN or HIRLAM mesoscale models, downscaled to 5.5 km grid spacing, chosen to match the time period of the climatological gauge reports. The validation of the Mescan system is carried out over the French territory employing various metrics and by providing forcing to a hydrological model to produce river discharges. The investigations have shown that the precipitation analyses have almost the same quality as the well-validated SAFRAN analysis system. In addition, the analysis of the precipitation variance spectra computed on the same horizontal domain has indicated that at short wavelengths the downscaled fields have significantly lower variability than a field produced by time integrating a forecast model. The Mescan precipitation analysis system has successfully been used to produce 24-h total accumulated precipitation re-analyses on a 5.5 km grid over Europe for the period 2007Á2010.

Abstract. Low-flow simulation and forecasting remains a difficult issue for hydrological modellers, and intercomparisons can be extremely instructive for assessing existing lowflow prediction models and for developing more efficient operational tools. This research presents the results of a collaborative experiment conducted to compare low-flow simulation and forecasting models on 21 unregulated catchments in France. Five hydrological models (four lumped storagetype models -Gardenia, GR6J, Mordor and Presages -and one distributed physically oriented model -SIM) were applied within a common evaluation framework and assessed using a common set of criteria. Two simple benchmarks describing the average streamflow variability were used to set minimum levels of acceptability for model performance in simulation and forecasting modes. Results showed that, in simulation as well as in forecasting modes, all hydrological models performed almost systematically better than the benchmarks. Although no single model outperformed all the others for all catchments and criteria, a few models appeared to be more satisfactory than the others on average. In simulation mode, all attempts to relate model efficiency to catchment or streamflow characteristics remained inconclusive. In forecasting mode, we defined maximum useful forecasting lead times beyond which the model does not bring useful information compared to the benchmark. This maximum useful lead time logically varies between catchments, but also depends on the model used. Simple multi-model approaches that combine the outputs of the five hydrological models were tested to improve simulation and forecasting efficiency. We found that the multi-model approach was more robust and could provide better performance than individual models on average.

Abstract. This paper describes the impact of the various changes made to the Safran–Isba–Modcou (SIM) hydrometeorological system and demonstrates that the new version of the model performs better than the previous one by making comparisons with observations of daily river flows and snow depths. SIM was developed and put into operational service at Météo-France in the early 2000s. The SIM application is dedicated to the monitoring of water resources and can therefore help in drought monitoring or flood risk forecasting on French territory. This complex system combines three models: SAFRAN, which analyses meteorological variables close to the surface, the ISBA land surface model, which aims to calculate surface fluxes at the interface with the atmosphere and ground variables, and finally MODCOU, a hydrogeological model which calculates river flows and changes in groundwater levels. The SIM model has been improved first by reducing the infrared radiation bias of SAFRAN and then by using the more advanced ISBA multi-layer surface diffusion scheme to have a more physical representation of surface and ground processes. In addition, more accurate and recent databases of vegetation, soil texture, and orography were used. Finally, in mountainous areas, a sub-grid orography representation using elevation bands was adopted, as was the possibility of adding a reservoir to represent the effect of aquifers in mountainous areas. The numerical simulations carried out with the SIM model covered the period from 1958 to 2018, thereby providing an extensive historical analysis of the water resources over France.

Abstract. The new AquiFR hydrometeorological modelling platform was developed to provide short-to-long-term forecasts for groundwater resource management in France. This study aims to describe and assess this new tool over a long period of 60 years. This platform gathers in a single numerical tool several hydrogeological models covering much of the French metropolitan area. A total of 11 aquifer systems are simulated through spatially distributed models using either the MARTHE (Modélisation d'Aquifères avec un maillage Rectangulaire, Transport et HydrodynamiquE; Modelling Aquifers with Rectangular cells, Transport and Hydrodynamics) groundwater modelling software programme or the EauDyssée hydrogeological platform. A total of 23 karstic systems are simulated by a lumped reservoir approach using the EROS (Ensemble de Rivières Organisés en Sous-bassins; set of rivers organized in sub-basins) software programme. AquiFR computes the groundwater level, the groundwater–surface-water exchanges and the river flows. A simulation covering a 60-year period from 1958 to 2018 is achieved in order to evaluate the performance of this platform. The 8 km resolution SAFRAN (Système d'Analyse Fournissant des Renseignements Adaptés à la Nivologie) meteorological analysis provides the atmospheric variables needed by the SURFEX (SURFace EXternalisée) land surface model in order to compute surface runoff and groundwater recharge used by the hydrogeological models. The assessment is based on more than 600 piezometers and more than 300 gauging stations corresponding to simulated rivers and outlets of karstic systems. For the simulated piezometric heads, 42 % and 60 % of the absolute biases are lower than 2 and 4 m respectively. The standardized piezometric level index (SPLI) was computed to assess the ability of AquiFR to identify extreme events such as groundwater floods or droughts in the long-term simulation over a set of piezometers used for groundwater resource management. A total of 56 % of the Nash–Sutcliffe efficiency (NSE; Ef) coefficient calculations between the observed and simulated SPLI time series are greater than 0.5. The quality of the results makes it possible to consider using the platform for real-time monitoring and seasonal forecasts of groundwater resources as well as for climate change impact assessments.

Abstract. Low-flow simulation and forecasting remains a difficult issue for hydrological modellers, and intercomparisons are needed to assess existing low-flow prediction models and to develop more efficient operational tools. This study presents the results of a collaborative experiment conducted to compare low-flow simulation and forecasting models on 21 unregulated catchments in France. Five hydrological models with different characteristics and conceptualizations were applied following a common evaluation framework and assessed using a common set of criteria. Two simple benchmarks were used to set minimum levels of acceptability for model performance in simulation and forecasting modes. Results showed that, in simulation as well as in forecasting modes, all hydrological models performed almost systematically better than the benchmarks. Although no single model outperformed all the others in all circumstances, a few models appeared more satisfactory than the others on average. In simulation mode, all attempts to relate model efficiency to catchment characteristics remained inconclusive. In forecasting mode, we defined maximum useful forecasting lead times beyond which the model does not contribute useful information compared to the benchmark. This maximum useful lead time logically varies between catchments, but also depends on the model used. Preliminary attempts to implement simple multi-model approaches showed that additional efficiency gains can be expected from such approaches.