Estimating soil loss of given return period by USLE-M-type models

Many field investigations have clearly shown that rare and severe events control total soil erosion occurring over a long time period with up to 92% of total soil erosion over a 7-year period resulting from just three daily events. Therefore, soil conservation strategies should be developed taking into account large events rather than long-term average erosion. From an engineering point of view, establishing the soil loss of a given return period is needed. This can be obtained by the frequency analysis of soil loss measurements or by suitable soil erosion models. The USLE-M modified and USLE-M based are two empirical Universal Soil Loss Equation-Modified (USLE-M) type models which were developed using runoff and soil loss measurements collected at the Sparacia (South Italy) station. According to USLE-M type models, the return period of the estimated soil loss coincides with that of the rainfall-runoff erosivity index. In this investigation, using the maximum annual values of event soil loss measured on bare plots at Sparacia, this hypothesis was recognized to be not valid. To overcame this limit, a parameterization method which imposes the coincidence of the two return periods was applied. The models, which were parameterized by the maximum annual values of event soil loss, produced only moderate improvements in the prediction of the soil loss of given return period as compared with their previously published versions, which were parameterized by all available measurements. This result induced to hypothesize that parameterization of the USLE-MM and the USLE-MB models useful to estimate the soil loss of given return period in other sites can be also performed with all available data, not only with the maximum annual event soil losses which would require a larger number of recording years to obtain a suitable sample size. However, the developed database should span a wide soil loss measurement range. In general, at Sparacia both USLE-MM and USLE-MB were able to yield accurate predictions for return period ranging from 4 to 20 years.