Effect of LULC data resolution on hydrological and erosion modeling using SWAT model

The effect of input data resolution on hydrological model performance was not considered in many studies despite its crucial importance, especially for over-parameterized hydrological models. Therefore, in the present study, three land cover sources with different resolutions (Sentinel-2, Landsat 8, and CCI-LC, with 10 m, 30 m and 300 m, respectively) were used as input data for the Soil & Water Assessment Tool (SWAT) model, to assess their impact on hydrological and erosion modeling. The SWAT model was performed over 19 years (2000-2018) in El Grou watershed (3504 km(2)) under a multi-site calibration approach (two stations). Statistical analysis showed good and satisfactory results for model calibration and validation. The values obtained are between 0.65 and 0.68 for R-2 and NSE (Nash-Sutcliffe coefficient), and 0.57-0.63 for RSR (ratio of the root mean square error). Comparing the three models' performance (before and after calibration) indicates a significant similarity whatever the used LULC. Also, the three LULC-based simulations show a predominance of evapotranspiration with 68% and a significant sediment loss of around 1 million tons. This highlights a significant deterioration of the soils and vegetation cover, which requires a sustainable intervention to limit the effects of this phenomenon and protect soils and dams against siltation. Finally, the findings of this study highlight that low-resolution LULC data are often suitable to obtain good results while reducing processing and calibration efforts, which saves time and resources; However, the choice remains open, depending on the study purpose and the expected results.