The AquiFR hydrogeological modeling platform. Evaluation of the 1958-2017 reanalysis for the main regional multilayer aquifers in France. J.-P. Vergnes, N. Amraoui, F. Habets, N. Roux, D. Thiéry, P. Viennot CMWR 2018 - Saint Malo 5 juin 2018

Climate impact studies Introduction Gathering hydrogeological models developed over France to provide a single numeric tool for operational and research purposes >= 2050 1958 Present 10 days 3-6 months Long-term simulation Short-term forecast Seasonal Forecast Climate impact studies

Implemented softwares SAFRAN atmospheric reanalyses Rainfall, Temperature, Wind, PET… SURFEX Land surface model Surface runoff, Groundwater recharge Spatially distributed hydrogeological model EROS lumped model for karsts MARTHE EauDyssée Piezometric heads Karst discharges

Map of the models MARTHE distributed models EauDyssée distributed models EROS lumped models Somme 1 layer Nord-Pas de Calais 10 layers Alluvions d’Alsace 3 layers Basse-Normandie 10 layers Seine 6 layers Karsts 23 basins Poitou-Charente 8 layers Loire 3 layers

Example of a spatially distributed model Jurassic-Poitou Charentes model Regular grid with 1-km resolution with 8 layers Coupling between surface water and groundwater Pumping for agriculture, drinking water or industry 1850 groundwater pumpings 100 surface water pumpings Initially calibrated on the 2000-2007 period 3d view of the model layer Area of the model 19 200 km2 Location of surface water and groundwater pumping boreholes

Evaluation of the 1958-2017 long-term simulation Why? Models calibrated independently on shorter period and with different recharge Ex : 2000-2007 for the Poitou-Charentes model Initial states for hindcasts Validate the approach on a long-term simulation for future long-term climate study Characteristics Daily simulation from 1st August 1958 to 31th July 2017 Initial state : 1st August 2006 Outputs : Monthly piezometric head maps (unconfined aquifers) 573 daily simulated piezometers levels and 23 karst discharges Groundwater-surface water exchanges

RMSE of piezometric heads RMSE (m) 58 % < 5 m 24 % < 2 m 74 % < 8 m

The Standardized Piezometric Level Index (SPLI) Based on the Standardized Precipitation Index for meteorological drought Computed from a time series of piezometric head Used in France in the Monthly Hydrological Summary for a set of chosen piezometers OBS SIM SPLI daily evolution HUMID Tableau avec les niveaux Exemple avec 1 piézomètre observé puis simulé DRY

Evaluation of the Standardized Piezometric Level Index (SPLI) Nash Coefficient 45 % > 0.5 83 % > 0 16 % < 0

Example of Monthly Hydrological Summary – August 2016 Observation Simulation

Example of karst system: La Fontaine de Vaucluse Lumped model Basin Area : 1317 km2 Calibrated over 24 years (1990-1991 –2013-2014) Gauging station : river Sorgues Daily Nash = 0.88 Monthly river discharges

Groundwater-surface water interactions Somme Model Q > 0 : river to aquifer Q < 0 : aquifer to river

Groundwater-surface water interactions Alsace Model 53 % of the recharge to the aquifer provided by the river

Conclusions Perspectives Evaluation of the AquiFR platform: Piezometric heads well simulated on the whole period of simulations Karst models are well reproduced Able to produce indices used in water resource management or climate study: Standardized Piezometric Level Index Perspectives Add new models Analyze river discharges outputs Seasonal forecast up to 6 months Make products available to users for water resource management

Les sorties d’AquiFR par modèle Fichiers de sortie pour MARTHE et ODIC : Chroniques journalières de hauteurs de nappe par piézomètre Cartes journalières des échanges nappes/surfaces Sur les rivières pour Marthe Sur les rivières ET les couches affleurantes (débordements) pour EauDyssée Cartes mensuelles de la piézométrie des couches affleurantes Cartes journalières des debits en rivière Fichiers de sortie pour EROS : Débits journaliers des 23 karsts d’Eros Charges (ne contient que des valeurs nulls)

Remarques Pour le tracé des couches affleurantes : Couches shapefiles générées dans AquiFR Nécessite le SIG GRASS et python 2.7 Pistes d’ameliorations Documentation à apporter Clarification dans la génération des fichiers utiles à l’exploitation des résultats (shapefiles, fichier de coordonnées…)