Improvement and Evaluation of Cligen for Storm and Rainfall Erosivity Generation.

CLIGEN is a stochastic weather generator, which produces daily estimates of precipitation and individual storm parameter including time to peak, peak intensity, and storm duration. These parameters are required to run other models such as the Water Erosion Prediction Project (WEPP) model. CLIGEN has proven to be effective when predicting daily estimates, but not for the individual storm parameters, like the storm duration. Therefore a study was conducted to evaluate and improve CLIGEN estimates. Individual rainfall events were identified from 1-hour pluviograph records collected in 30 sites located in Central Chile. More than 415 years of data were used, and 18,012 storms were analyzed. In addition, rainfall erosivities were computed for all the storms by the prescribed method to compare the energy provided in the rainfall events. Using the measured data, a procedure was developed to improve the CLIGEN estimates by calibrating the input parameter controlling the storm duration and intensity. The model was tested before and after calibration with the rainfall data from the 30 sites, for the wet and the dry season. Based on a monthly rainfall analysis, the number of storms and total rainfall depth, which are not affected by the calibration process, were accurately estimated with CLIGEN. However, before calibration and especially in the wet season, measured and generated storm durations and maximum intensities were consistently over and underestimated in most of the sites and months. Thus, monthly and annual rainfall erosivities were underestimated with CLIGEN. After the calibration, measured and generated storm durations were closer, even though the coefficient of variation in the generated durations were between 0.02 and 0.4, compared to the 0.8 to 1.2 observed in the measured data. Nevertheless, measured and generated maximum storm intensities were almost identical for most of the sites and months, especially in the wet season. As a result of this, annual erosivity estimates showed a significant improvement in 29 of the 30 sites. Monthly erosivity estimates were also closer to the measured values in almost every site and month. Therefore, this calibration procedure proved to be an alternative to generate more reliable storm patterns. This paper explains in detail this procedure and analyzes these and other parameters related to the individual storm generation process.