一种基于PCA和ANN的土壤水力性质估计方法

本文根据土壤基本性质，利用主成分分析和人工神经网络相结合的方法（PANN）构建了预测田间持水量和凋萎系数的土壤转换函数，并将其结果与传统的神经网络模型（ANN）进行了比较。结果表明，由于PANN 消除了神经网络输入层参数的相关性，降低了网络拓扑的复杂度，从而具有更好的预测能力。

A method based on principal component analysis and artificial neural network for estimating soil hydraulic properties

Prediction of the field capacity and the permanent wilting point is of importance due to actual needs of hydrological model for solving large scale soil moisture problems. The aim of this study is to develop pedotransfer functions for predicting field capacity and permanent wilting point through a new methodology based on artificial neural network using principal components as inputs. The developed model is compared with artificial neural network based on the original data. The result shows that the proposed method has a better predictive ability because it eliminates the correlation of parameters in the input layer of the neural network and reduces the complexity of the network topology.