中文摘要：

模拟降水和试验践踏通过各种不同强度的组合方式改变土壤抗蚀性,从而起到对草地土壤侵蚀的增减作用。在同一模拟降水量条件下,随着践踏强度的逐渐加大,土壤可蚀性K值依次增大,表明践踏增大了放牧侵蚀的风险;但K值的增幅显然与模拟降水量相关,践踏强度由轻度递增到重度,K值的增幅在干旱、自然降水、平水、丰水条件下依次为干旱〉自然降水〉平水和丰水,表明模拟降水和践踏对K值的影响存在交互效应,模拟降水具有减缓K值随践踏强度增大的趋势。从简单相关关系来看,K值与践踏强度呈极显著正相关（相关系数0.741）,与降水呈负相关（相关系数-0.378）,但K值并不是可以由践踏强度和模拟降水量二元线性回归可以解说地。与传统回归模型相比较,BP网络模型能更好地刻画土壤可蚀性K值的复杂非线性特性,具有自学习、自组织、自适应和容错性等一系列优点,因而,以牧草生长期单位面积累计践踏量和模拟降水量为自变量的土壤可蚀性K值的ANN（artificialneuralnetworks）关系模型具有较好的拟合结果和预测能力,说明直接从输入到草地生态系统的外侵蚀营力着手,跨越系统内土壤可蚀性变化的内在的复杂的隐含过程建立的输出端——土壤可蚀性K值与土壤侵蚀外营力的ANN关系模型,是准确确定土壤可蚀性K值的一次成功尝试。

英文摘要：

Soil erodibility differed with various combinations of experimental trampling and simulated rainfall.Under the same simulated rainfall condition,the soil erodibility factor,k,increased as the trampling intensity increased,suggesting that trampling enhanced grazing-induced erosion risk.However,the amplitude of the k value was apparently related to the levels of simulated rainfall：the k value changed in the sequence aridnaturalaveragehigh.There was an interaction between the effects of trampling and simulated rainfall on the k value.Simulated rainfall reduced the trend of k value increase while an increase in trampling intensity did the opposite.A positive correlation was found between the k value and trampling intensity （Coefficient=0.741）,and a negative correlation between k value and rainfall level （Coefficient=-0.378）.Compared with traditional regression models,the artificial neutral network （ANN） model shows many advantages,such as self-studying,self-organizing,self-adapting and fault tolerance.Therefore k was calculated using the ANN model,based on the independent variables of accumulated trampling effects per unit area and simulated rainfall.The ANN model allows the complicated soil erodibility mechanisms to be skipped and to directly establish the relationship between soil erodibility and factors influencing it.e.g.trampling and rainfall.This study is an innovative attempt to evaluate soil erodibility more precisely.