Comparison of long-term field-measured and RUSLE-based modelled soil loss in Switzerland

Long-term field measurements to asses model-based soil erosion predictions by water are rare. We have compared field measurements based on erosion assessment surveys from a 10-year monitoring process with spatial-explicit model predictions with the Revised Universal Soil Loss Equation (RUSLE). Robust input data were available for both the mapped and the modelled parameters for 203 arable fields covering an area of 258 ha in the Swiss Midlands. The 1639 mapped erosion forms were digitized and converted to raster format with a 2 m resolution. A digital terrain model using 2 m resolution and a multiple flow direction algorithm for the calculation of the topographic factors and the support practice factor was available for modelling with the RUSLE. The other input data for the RUSLE were determined for each field. The comparison of mapped and modelled soil loss values revealed a substantially higher estimation of soil loss values from modelling by a factor of 8, with a mean mapped soil loss of 0.77 t/ha/yr vs. modelled soil loss of 6.20 t/ha/yr. However, high mapped soil losses of >4 t/ha/yr were reproduced quite reliably by the model, while the model predicted drastically higher erosion values for mapped losses of <4 t/ha/yr. Our study shows the value of long-term field data based on erosion assessment surveys for model evaluation. RUSLE-type model results should be compared with erosion assessment surveys at the field to landscape scale in order to improve the calibration of the model. Further factors related to land management like headlands, traffic lanes and potato furrows need to be included before they may be used for policy advice.