Bernd Diekkrüger, Attanda Mouinou Igué and Thomas Gaiser


Besides other information, environmental modeling requires data concerning soil properties and their spatial distribution. The quality and the applicability of soil maps not only depend on their scales, but also on the concept of the mapping procedures. To study the effects of different soil mapping approaches on the performance of the Soil and Water Assessment Tool (SWAT), a modeling study was carried out in the Zou catchment (6980 km²) which is a tributary of the Ouémé catchment (Republic of Benin, West Africa). Two different maps of the hierarchical Soil and Terrain digital database approach (SOTER; old and new) was considered in conjunction with the ORSTOM approach (soil mapping at the reconnaissance level, with one dominant soil type per mapping unit) all at the same scale. The effects of less coarser new SOTER mapping units and soil layer aggregation on the model results were evaluated. Sensitivities regarding the model input parameters with respect to the different soil databases are quantified. Based on daily water discharge measurements at the catchment gauging station, the SWAT model was calibrated (period 2001 to 2004) and validated (period 2005 to 2006) with reasonable results (coefficient of determination, model efficiency, and index of agreement, about 0.70 for weekly discharge) but with a slight underestimation of the total water yield for the coarser old SOTER database. With respect to the annual discharge, the calibration compensates for the differences in the different soil databases. The validation identified the less coarser new SOTER map as the best (with respect to modeling of weekly discharge) used in our study (showing higher model efficiency (0.59), higher coefficient of determination (0.61), and higher index of agreement (0.87)). Combined effects of the coarser old SOTER mapping units vs. aggregated soil layers have a measurable influence on lateral flow and sediment yield within the study area. High spatial variability in surface runoff and sediment yield patterns caused by the different mapping approaches were simulated although calibration of the discharge resulted in similar quality measures. Changes in the model parameters discriminated clearly the effects of soil mapping approaches from those of the mapping unit/parameter aggregation and have shown that SWAT could be calibrated successfully for discharge with the three databases, but differences between the soil mapping approaches with respect to discharge could not be balanced out.