Soil erodibility is among the trickiest erosion factors to estimate. This is especially true for heterogeneous Mediterranean environments, where reliable and dense soil data are rarely available, and interpolation methods give very low accuracies. Towards estimating soil erodibility, research so far has resulted in several alternatives mainly based on empirical formulas, on physics-based equations or on inference with expertise. The aim of this work was to compare erodibility patterns derived by using the empirical United States Department of Agriculture (USDA) formula and by inference from a geological map in a Mediterranean agricultural site. The Kolymvari area, located in the western part of Crete, an area covered by olive groves and citrus orchards, was selected as the study site for this work. Comparison of the spatial patterns of soil erodibility derived from the two alternatives showed significant differences (i.e.
, a mean normalized difference value of 0.52), while a test of the “inference” alternative indicated very low accuracies (0.1475 RMS error). A comparison, however, of the spatial patterns of erosion values derived from both alternatives indicated that dissimilarities of the two soil erodibility maps faded out. Moreover, the highly risky areas provided by both alternatives were found to be identical for 88% of the whole study site.