Evaluer la performance du modèle SWAT en testant différents jeux de données climatiques à diverses résolutions spatiales et temporelles : application au bassin versant de la Charente (10.000 km²)

Autor: Odile Phelpin_Leccia, Lise Andro
Přispěvatelé: Environnement, territoires et infrastructures (UR ETBX), Institut national de recherche en sciences et technologies pour l'environnement et l'agriculture (IRSTEA), Irstea Publications, Migration
Jazyk: angličtina
Rok vydání: 2019
Předmět:
Zdroj: 2019 Vienna SWAT Conference
2019 Vienna SWAT Conference, Jul 2019, Vienna, Australia. pp.7
HAL
Popis: International audience; Forcing variables relating to the hydrologic response of rainfall runoff modeling are known to play a major role in accuracy of predictions made. In this study, three hydro-climatic daily time step datasets were tested: conventional weather stations, the 8km gridded SAFRAN, and the 0.5° CFSR included in the 10,550km² Charente river basin, South-Western France. Estimation of discharge and uncertainty analysis for stream flow estimation was carried out using the Soil and Water Assessment Tool (SWAT). Multi-site calibration and parameter uncertainty analysis were performed with the Sequential Uncertainty domain parameter FItting algorithm (SUFI-2). The model was calibrated and validated by means of observed discharge data for the periods 1999 - 2008 and 2009 - 2018, respectively. The performance of the SWAT model regarding stream flows is evaluated based on five criteria: determination coefficient R², NashSutcliffe efficiency (NSE), per cent bias (PBIAS), p-factor, and r-factor (calculated on a daily basis). During calibration, the values of R² and N SE were found to be 0.73 and 0.71. When validated, they were found to be 0.54 and 0.52. These findings, which are currently undergoing final analysis, show that calibration and validation performance were at their best with conventional weather station datasets for the gauged sub-basins, and that observed and simulated daily stream flows were not much different at 95PPU (95% level of prediction uncertainty).
Databáze: OpenAIRE
Externí odkaz: