Rainfall input generation for the European Soil Erosion Model

Abstract

A procedure to generate rainfall input for the EUROpean Soil Erosion Model is presented. To develop such a procedure, first of all the influence of rainfall event amount, rainfall event duration, and time to peak intensity of event rainfall on soil losses, calculated with EUROSEM, has been tested for several rainfall stations. Results revealed that every tested rainfall parameter had highly significant influence on computed soil loss. Therefore, distributions for each station of the dataset and for each of these rainfall parameters were calculated. To simulate rainfall event amounts, a mixed exponential distribution was applied. After transformation of rainfall event durations, their distribution could be simulated using a normal distribution. The location of the peak intensity was estimated using a kernel estimator, because no specific distribution characteristics could be identified. According to the respective distribution functions, parameter values for each of the tested rainfall event characteristic were then generated. These values were used to select rainfall events with identical parameter values out of the rainfall station-specific dataset. Computed soil losses for events selected this way were compared with soil losses calculated with available station specific rainfall event data. Comparisons for the respective means and medians generally revealed good agreement. A comparison of 75% quartiles resulted in less good agreement, especially for test conditions with high soil losses. In general, the applied procedure was capable of simulating station-specific soil losses and of reflecting different environmental conditions for the respective stations. Therefore, it seems possible to produce site specific appropriate rainfall input for EUROSEM, only with the knowledge of distributions for the investigated basic rainfall parameters. These are normally easier to obtain than long term rainfall information with high temporal resolution which would otherwise be necessary. In order to improve the procedure and make it practically useful, it will be necessary to account for seasonal changes of distributions of basic rainfall event parameters.