Sensitivity of the SWAT model to soil parametrization with different land use input maps: a case study of the Thyle catchment in Belgium

Romanowicz, Agnieszka [UCL] Vanclooster, Marnik [UCL] Rounsevell, Mark [UCL] La Jeunesse, I.

The sensitivity of the surface hydrological component of the spatially distributed hydrological model SWAT to different input parametrisation schemes was tested for a small agricultural catchment of the Thyle in the central part of Belgium. For this analysis 2 different soil parametrisation schemes, 2 different soil maps, and 2 land use maps were used as input to the SWAT model. For generating the soil parametrisation schemes, use was made of the generalised soil association map at the scale of 1:500.000, the detailed IRSIA soil map at the scale of 1:25.000 and the soil profile analytical data base AARDEWERK. Uncalibrated simulations were performed for two different periods (1977-1979 and 1999) for which a land use map was available and for which detailed hydrological data are available. The results are suggesting that the SWAT model is very sensitive to the soil parametrisation scheme and to the input data resolution. For both simulated time periods only one soil parametrisation scheme performed appropriately. However the use of a high resolution soil map will not always result in the best modelling performance. The use of detailed soil map in combination with detailed land use map performs well. However, the results obtained from the use of the detailed soil map in combination with a generic land use map are surprisingly poor. This can be explained by the equifinality problem of different parameters sets in distributed hydrological models. Therefore it is advice to use a similar resolution for the input maps and input data sets while applying the SWAT model. Acknowledgements This research was carried out within the framework of the EU Mulino project (contract no. EVK1-2000-22089). The authors would also like to acknowledge the developers of the AvSwat interface.