中国降雨侵蚀力的时空分布及重现期研究

降雨侵蚀力是土壤侵蚀模型USLE的一个重要因子。基于中国中东部水蚀区18个气象站1961(1971)-2000年逐分钟降水数据和全国范围内774个气象站1961-2016年逐日降水数据,采用克里金插值方法,得到全国多年平均年、多年平均24个半月、不同重现期年和次侵蚀力空间分布特征,可满足USLE模型对侵蚀力因子相关参数输入的要求。交叉验证结果表明:以上所有指标的空间插值模型精度较好,模型有效系数NSE不低于0.74,偏差百分比PBIAS低于1%,均方根误差与观测值标准差的比值RSR小于等于0.51。侵蚀力年内变化曲线具有较好的区域相似性,使用K均值聚类分析方法将中国侵蚀力年内变化特征划分为4个区域,每个区域概化出一条侵蚀力年内变化曲线。

Spatiotemporal distribution and return period of rainfall erosivity in China

Rainfall erosivity is an indicator of the potential capacity of rainfall to cause soil erosion and one of the most important factors in the soil erosion model USLE and its revised versions. Based on the precipitation data with one-minute interval collected from 18 meteorological stations in the middle eastern water erosion areas of China from 1961(1971) to 2000, relationships between daily rainfall erosivity and event rainfall erosivity for different return periods were explored. Based on the daily precipitation data from 774 meteorological stations nationwide from 1961 to 2016, average annual, half-month, annual and event erosivity for the 2-year, 5-year, 10-year, 20-year, 50-year and 100-year return periods were calculated for the stations with observations. The Kriging interpolation method was used to estimate the rainfall erosivity for sites without observations and the spatial distribution maps of rainfall erosivity obtained could meet the requirements of the USLE models for the input of erosive force related parameters. The results showed that: 1) there was a good linear relationship between event and daily rainfall erosivity for corresponding return periods, and coefficients of determination were all greater than or equal to 0.96. The coefficients for converting daily rainfall erosivity into event rainfall erosivity were 1.12, 1.15, 1.17, 1.19, 1.22, and 1.24, respectively for the 2-year, 5-year, 10-year, 20-year, 50-year and 100-year return periods, which could be useful when event precipitation data was not available. 2) The leave-one-out cross-validation results showed that spatial interpolation models for all the above indices had good precision. Spatial interpolation models for the average annual, half-month, and annual erosivity of different return periods performed better than those for the event erosivity of different return periods. For the average annual, half-month, and annual erosivity of return periods, Nash-Sutcliffe efficiency coefficient (NSE) was greater than or equal to 0.80, deviation percentage (PBIAS) was less than 1%, and the ratio of root mean square error to the observed standard deviation (RSR) was less than 0.45; for the event erosivity of different return periods, NSE was not less than 0.74, PBIAS was less than 1%, and RSR was less than and equal to 0.51. 3) The seasonal variation curve of erosivity had good regional similarity. The K-means clustering analysis method was used to cluster seasonal variation curves of rainfall erosivity into four categories. One curve of seasonal variation for each category was obtained by averaging values for all stations in the same category. Similar with the seasonal variation of precipitation caused by monsoon climate, four categories were all with predominant peak in summer and autumn. The type I curve was characteristic of "short" with the erosivity more dispersed during the year and d two peaks; the type II reached a peak in June and decreased rapidly after June; the type III had a peak in July; and the type IV was characteristic of "slim and high" with a high degree of concentration and a significant peak appearing in the late July.