A framework of abrupt changes and trends detection for rainfall erosivity

Rainfall erosivity (R factor), in the Universal Soil Loss Equation (USLE) , a climate index, is used worldwide to assess and predict the potential of rainfall to cause erosion. The temporal variation in rainfall erosivity, informs of abrupt change and trend, are critical for soil loss prediction. To find a simple and effective method for accurate detection of abrupt change and trend has implication for soil and water conservation planning. In this paper, a four-step framework is proposed to detect abrupt change and trend in rainfall erosivity time series, i.e., evaluate the significance of variation in rainfall erosivity time series at three levels: no, weak and strong, abrupt change and trend detection for rainfall erosivity,  estimation of correlation coefficient between the variation component and rainfall erosivity series, remove the variation component with the largest correlation coefficient from the rainfall erosivity series, repeat the above steps for the new series until variance coefficient was insignificance. The first step is based on an index of Hurst coefficient. The trend detection is implemented using both Spearman rank and Kendall rank correlation test. For abrupt change ,three kinds of methods (Mann-Kendall, Moving T and Bayesian test) are employed.  This framework is applied to the annual rainfall erosivity series of the Three Gorges Reservoir , China. There was a large uncertainty in detecting variability with a single test method. Application of the proposed framework can reduce uncertainty  associated with soil erosion assessment and achieve more accurate regional soil and water management.