Development of a Downscaling Scheme for a Coarse Scale Soil Water Estimation Method
Many river basins worldwide, especially in semi-arid regions, are adversely impacted by poor hydrological infrastructure or are poorly characterized due to limited or no hydrologic data. This condition challenges water-management authorities, who benefit from reliable prediction of the hydrological dynamics that can be made by means of hydrological models. Because of the lack of sufficient or reliable data, often such models are difficult to calibrate and to validate. This study addresses this data limitation by formulating and testing an independent validation tool for hydrological models that can be applied to downscale macro-scale soil water data derived from a remotely sensed scatterometer dataset. This proposed method uses the concept of hydrological response units (HRU) to analyze the spatial variability within one scatterometer footprint. The HRUs are treated as model entities in the process oriented hydrological model J2000 that was applied to the Great Letaba River catchment (ca. 4.700 km²) in South Africa. The soil water time series results were then compared to the remotely sensed data set and the downscaling scheme derived. First, the analysis conducted on footprint scale highlights the similarities in predicting the soil water generation over the long term and in seasonal terms. It also exhibits that the absolute values of both time series can not be used for further investigation, due to differences in the observed soil water volume. Second, the resulted simulated soil water time series were used to establish the downscaling method. Here, the study provides promising results that allow the downscaling of the coarse scale soil water calculated dataset, based upon the landscape related parameters of land cover, soil group and precipitation. The study findings indicate that, by linking the two concepts, hydrological modeling and remote sensing, water management authorities should be able to reduce certain prediction uncertainties of the applied models.