非参数投影寻踪回归在坡面土壤侵蚀产沙量预测中的应用

Application of Projection Pursuit Regression Model in Predicting Soil Erosion on Slope Land

根据土壤侵蚀机理,采用投影寻踪回归分析的方法,对坡面侵蚀产沙过程进行了预测。以四川某水土保持实验站观测数据为实例,以最大30 min降水量、最大1 h降水量、径流量等因素作为坡面产沙量预测的投影指标,采用可变阶的正交Hermite多项式拟合岭函数,并用单纯形算法优化投影指标函数,进而构建基于投影寻踪回归分析的坡面产沙量预测模型,通过MATLAB语言编制程序进行了仿真试验。结果表明,该模型具有较高的拟合度和良好的预测精度,为坡面产沙模拟预测模型的建立提供参考。

According to the soil erosion mechanism,the process of sediment yield on the slope was forecasted using the projection pursuit regression analysis.With the observed data of some soil and water conservation experimental station in Sichuan Province as an example,the maximum 30 min rainfall,maximum 1 h rainfall and runoff amount were selected as the projection indicators,the variable-order orthogonal Hermite polynomial was used to fit ridge function and the simplex algorithm was used to optimize the projection index function.Then the projection pursuit regression model was established to predict the erosion and sediment yield on slope land,and the simulation testing was carried out through the MATLAB language.The result of the sediment yield predicted by projection pursuit regression model was compared with the survey data,indicating that the model had higher prediction and accuracy and provided a reference for the prediction model of sediment yield on the slope land.