The application of geostatistics in erosion hazard mapping

Geostatistical interpolation or kriging of soil and vegetation variables has become an important alternative to other mapping techniques. Although a reconnaissance sampling is necessary and basic requirements of geostatistics have to be met, kriging has the advantage of giving estimates with a minimized error. In erosion hazard assessment this technique may therefore be applied instead of the conventional choropleth mapping, in order to improve the reliability of the input data for an erosion model. In a study, which was performed in the Ardeche district (Southern France) at a detailed scale (1:5000), the two methods are compared directly. It demonstrates a straightforward way to decide which method will give the best overall results at different sampling densities for erodibility controlling variables and the USLE K-factor. To this end a land unit map and several kriging maps of the investigated area (±2,5 km2) have been prepared. Quotients of kriging estimation variances and within-unit variances of the land unit map were calculated as a relative measure of error reduction. It was found that the kriging estimation variance is 66% (on average) of the within-unit variance for the USLE K-factor at a 150 meter sampling distance. This percentage can be further reduced to about 25% at a 50 meter sampling distance. It is concluded that kriging is an efficient option for mapping erodibility at this scale.