偏最小二乘模型在黄土坡面土壤侵蚀中的应用研究

Research on the Application of Partial Least Squares Model to Soil Erosion on Loess Slope

选取不同坡度(5°、10°、15°、20°)和不同雨强(50、60、70、80 mm/h)的16场降雨,利用偏最小二乘模型研究了坡度、雨强、降雨历时、土壤初始含水率、剪切力5个因素对黄土坡面土壤侵蚀过程的影响程度。结果表明:①变量对数化处理后,偏最小二乘模型模拟精度和预测优度均达到0.8以上,径流率和产沙率相关系数为0.844;②VIP值显示,时间X3、坡度X1是驱动降雨初期径流率Y1和产沙率Y2变化的重要因素,其中时间X3是最重要的影响因素;雨强对径流率和产沙率影响均较大,标准回归系数分别为0.408、0.341,初始含水率、剪切力的标准回归系数最大仅为0.145,表明二者对径流率和产沙率的影响都比较有限;③精度分析结果表明,模型模拟和预测精度都较高,与产沙率相比,径流率的预测精度更高,平均相对误差绝对值为18.813%,模拟预测的效果更好。

Sixteen rainfalls with different slopes(5°,10°,15°,20°)and different rain intensities(50,60,70,80 mm/h)are selected,and the slope and rain intensity are studied by using a partial least squares model.The degree of influence of rainfall duration,initial soil moisture content,and shear force on the soil erosion process on the loess slope.The results show that:①After the logarithmization of variables,the simulation accuracy and prediction goodness of the partial least squares model are above 0.8,and the correlation coefficients of runoff rate and sediment yield are 0.844;②VIP value shows that times,slope it is an important factor that drives the changes of runoff rate and sediment yield in the initial period of rainfall,of which times is the most important influencing factor,and the strong influence on runoff rate and sediment yield is greater.The standard regression coefficients are 0.408 and 0.341,respectively.The maximum standard regression coefficient of water content and shear force is only 0.145,indicating that both have a relatively limited impact on runoff rate and sediment yield;③The accuracy analysis results show that the model simulation and prediction accuracy are both high,and Compared with the sand rate,the prediction accuracy of runoff rate is higher,the absolute value of the average relative error is 18.813%,and the simulation prediction effect is better.