Estimating monthly (R)USLE climate input in a Mediterranean region using limited data

Summary: This work presents three empirical models (MMFI, Morais Modification of Fournier Index; GJRM, Grimm–Jones–Rusco–Montanarella; REMDB, Diodato–Bellocchi Rainfall–Erosivity Model) where monthly-based climate data are used to estimate long-term (R)USLE (Universal Soil Loss Equation and its Revisions) rainfall erosivity factor (R m, MJmmh−1 ha−1 month−1). The objective was to evaluate two known models (MMFI and GJRM), and compare the results with the novel model REMDB meant for complex terrains. MMFI and GJRM are both based on the precipitation amount, whilst REMDB takes site latitude, elevation and precipitation seasonality also into account. The test area was the Italian region, where 30 stations (altitudes from about sea level up to 1270m, over the latitudinal range 36–46°North) with sufficient data to calculate R m according to USLE were available. The three models were evaluated against USLE rainfall erosivity over a validation data set of 14 stations, using a range of performance statistics. The REMDB estimates generally compared well with the USLE estimates according to different statistics. For REMDB, the relative root mean square error was, in average, 48.58% against 71.49% for MMFI and 66.55% for GJRM. The average modelling efficiency of REMDB was 0.51 against −0.02 (MMFI) and 0.13 (GJRM). REMDB was also superior in preventing biased errors in time, as quantified by the average pattern index versus months: 17.65MJmmh−1 ha−1 month−1, against 58.54MJmmh−1 ha−1 month−1 (MMFI) and 57.76MJmm h−1 ha−1 month−1 (GJRM). Of the two simplified models, the MMFI was the worst performer while the GJRM model performed similarly to the REMDB at two mid-altitude sites of Central Italy. [Copyright &y& Elsevier]