Bivariate analysis of rainfall characteristics for identifying soil loss return periods at the event scale

In this work, a method was developed to associate a return period with event soil loss at the plot scale (SLe) based on thejoint return period RP of event rainfall height H and maximum 30-minute intensity I30. The analysis used rainfall and SLedata from bare-soil plots of varying lengths with a 16% slope at the SERLAB site (central Italy). H and I30 were well described by the Generalized Pareto Distribution, and their dependency structure was modeled with a non-parametric copula. Marked seasonal differences emerged in the joint probabilistic behavior of H and I30. A regression analysis between the normalized soil loss SLN,(relative to the mean value μSLefor each plot length) and RP identified linear, season-specific relationships. The slope coefficients (1.16·10−4, 2.05·10−3, 8.33·10−3, and 8.73·10−4for winter, spring, summer, and autumn, respectively) clearly indicate that a given SLN,corresponds to different RPs depending on the season. The confidence intervals of the coefficients, besides reflecting experimental uncertainty, indicate system vulnerability (e.g., due to the temporal variability of soil erodibility). Finally, the study outlines a practical procedure for applying these relationships in soil conservation planning.