神经网络模型在预测土壤pH值中的应用研究

Soil pH Forecast Based on Artificial Neural Network

该文通过对西藏察雅县 10 5层土样资料 (1988年 )建立CaCO3-pH神经网络模型 ,与刘世全等所做的回归模型在拟合精度和预测性能方面作了比较 ,结果显示 ,BP网络在拟合性能方面不亚于回归方法 ,在预测性能上要优于回归方法。该文对将神经网络引入土壤环境系统的研究中作了有意义的尝试 ;所建立CaCO3-pH间的关系模型 ,是研究污染物在土壤中的降解和转化的重要基础 ,对评价周边环境因素对土壤的综合作用也有重要意义。本文的结论说明 ,神经网络对于研究土壤系统的目标因子和相应的影响因子间的关系方面 ,是较为适用的数学手段。

A neural network model forecasting CaCO3-pH relation in soil is established by using 105 soil samples in Chaya county in Tibet autonomy. The model is compared with regression model established by Liu Shi-quan on fitting accuracy and forecasting quality. The results show that the fitting accuracy is similar for BP network and regression method, and the forecasting quality of BP model is better than regression model. The application of neural network in studying the soil environment is a significant try. The BP model on soil CaCO3-pH established in the study may form an important base for the study of the pollutants' degradation and conversion in soil. It is also one of the key factors in the evaluation of functions of environment on soil. This study suggests that neural network can be an efficient mathematic method in the study of relationships among the factors that affect the soil environment.