Predicting soil loss in central and south Italy with a single USLE-MM model

Abstract

The USLE-MM estimates event normalized plot soil loss, Ae,N, by an erosivity term given by the runoff coefficient, QR, times the single-storm erosion index, EI30, raised to an exponent b1 > 1. This modeling scheme is based on an expected power relationship, with an exponent greater than one, between event sediment concentration, Ce, and the EI30/Pe (Pe = rainfall depth) term. In this investigation, carried out at the three experimental sites of Bagnara, Masse, and Sparacia, in Italy; the soundness of the USLE-MM scheme was tested. A total of 1192 (Ae,N, QREI30) data pairs were used to parameterize the model both locally and considering all sites simultaneously. The performances of the fitted models were established by considering all erosive events and also by distinguishing between events of different severity. The b1 exponent varied widely among the three sites (1.05–1.44) but using a common exponent (1.18) for these sites was possible. The Ae,N prediction accuracy increased in the passage from the smallest erosion events (Ae,N ≤ 1 Mg ha−1, median error = 3.35) to the largest ones (Ae,N > 10 Mg ha−1, median error = 1.72). The QREI30 term was found to be usable to predict both Ae,N and the expected maximum uncertainty of this prediction. Soil erodibility was found to be mainly controlled by the largest erosion events. Development of a single USLE-MM model appears possible. Sampling other sites is advisable to develop a single USLE-MM model for a general use.