论文部分内容阅读
针对新疆渭干河库车河三角洲绿洲土壤盐分动态监测中存在的方法问题,首先用灰色关联度模型分析影响形成土壤盐渍化的各因子,并确定其与土壤盐分之间的关联度,然后将人工智能计算技术引入土壤盐分的预测中,经过多次调整网络结构和参数,建立了预测表层土壤盐分的BP神经网络模型和RBF神经网络模型。结果表明:以潜在蒸散量、地下水埋深、地下水矿化度、土壤电导率、总溶解固体、pH值、坡度和土地利用类型8个因素为输入因子,以土壤含盐量为输出因子的BP网络模型和RBF网络模型可有效模拟土壤盐分与其影响因子之间的内在复杂关系,并且有较高的精度。BP网络模型预测误差略低于RBF神经网络。本研究可为分析和预测土壤盐渍化动态规律提供一种有效可行的新途径,是对传统土壤盐分动态研究的补充。
Aiming at the existing problems of soil salinities monitoring in the oasis of the Kuqa River Delta in Weigan River, Xinjiang, firstly, the factors affecting the formation of soil salinization were analyzed with the gray relational degree model and the correlation between them and soil salinity was determined. Then, Artificial intelligence technology is introduced into the prediction of soil salinity. After many times adjustment of network structure and parameters, a BP neural network model and a RBF neural network model for predicting the surface soil salinity are established. The results showed that the factors of potential evapotranspiration, groundwater depth, groundwater salinity, soil electrical conductivity, total dissolved solids, pH, slope and land use types were the input factors and the salinity of soil was the output factor The network model and the RBF network model can effectively simulate the inherent complex relationship between soil salinity and its influencing factors and have higher accuracy. BP network model prediction error is slightly lower than the RBF neural network. This study can provide a new and effective method for analyzing and predicting the dynamic of soil salinization, which is a supplement to the traditional research on soil salinity dynamics.