论文部分内容阅读
在美国土壤水分物理性质数据库(UNSODA 2.0)的基础上,考虑土壤质地不分类和分类2种情况,分别构建了基于支持向量回归机(SVR)的土壤传递函数模型,比较了在土壤质地不分类和分类情况下预测土壤水力学参数(水分特征曲线和饱和导水率)的效果,并与建立在相同数据库上的基于神经网络的Rosetta模型的预测效果进行了比较。结果表明:土壤质地不分类的情况下,输入参数越多,基于SVR模型的预测效果越好;土壤质地分类情况下,基于SVM分类建模的预测结果普遍好于不分类情况。无论土壤质地是否分类,样本和输入参数相同的条件下,基于SVR的模型预测的效果都优于Rosetta模型。
Based on the United States Soil Moisture Physical Properties Database (UNSODA 2.0), the soil transfer function model based on Support Vector Regression (SVR) was constructed considering both the unclassified and classified soil texture. The soil texture was classified And the prediction of soil hydraulic parameters (water characteristic curves and saturated hydraulic conductivity) under the classification conditions were compared with the prediction results of the Rosetta model based on neural networks established on the same database. The results show that the prediction results based on SVR model are better when the soil texture is not classified, and the prediction results based on SVM classification modeling are generally better than the unclassified soil texture classification. The results of SVR-based model prediction are better than that of Rosetta model, regardless of soil texture classification, sample and input parameters.