A probabilistic approach for predicting rainfall soil erosion losses in semiarid areas

The implementation of soil and water conservation structures in semiarid areas, usually poses a difficult design problem. This is, in large part, due to the high variability of rainfall and the huge potential impact of extreme hydrologic events on structures and on the landscape in general. Magnitudes of runoff and soil loss or sedimentation rates in those environments are better not assessed by conventional modelling techniques, which tend to average out event magnitude and recurrence variability in time and space. A probability-based approach is proposed here to analyse and predict rainfall erosion losses. The maximum annual storm and its associated erosivity is used as a core element in the assessment of annual interrill and rill erosion rates. Frequency and cumulative soil loss distributions are obtained by combining verified annual and maximum daily rainfall frequency distributions with a proposed erosion algorithm. This stochastic representation of erosion permits to evaluate soil losses for the maximum annual storm, as well as annual erosion rates as a function of recurrence interval. The proposed method was verified with a short series of measured soil loss data in Cape Verde. The physical basis underlying the prediction algorithm and method in general, could be sustained by experimental data and field survey evidence. The method seems applicable to arid and semiarid ecosystems with a high seasonal concentration of precipitation and with rainfall limited to only a few major storm events. (C) 2000 Elsevier Science B.V. All rights reserved.