摘要
以田间实测土壤体积含水率为标准,对以Penman-Montieth公式计算蒸散量ET为主的农田水量平衡法和以气象资料与作物叶面积指数为输入向量的BP神经网络法进行了土壤含水率预测对比研究,比较分析了这二种方法的预测精度。结果表明:农田水量平衡法比BP神经网络法的预测精度高。与实测值相比,前者的平均绝对误差、平均相对误差和均方差分别为1.04、4.84和1.245,而后者分别为1.06、5.08和6.657,二者的一致性指数CI分别为0.9505和0.9459。两种方法对土壤含水率预测值与实测值的拟合关系分别为y=0.7821x+4.7127(R2=0.8390)和y=0.7302x+6.1104(R2=0.8615),表明预测值与实测值的相关关系都达到了极显著水平,均可实现精度与实用的统一。
Based on the real tested data of soil volumetric water contents,we studied the performance for predicting soil moisture with two different methods,which include water balance method with PenmanMontieth formula(named as ET method) and BP artificial neural network method with main meteorology data and LAI as input vectors(named as BP method).The results showed that the predicted performance with ET method was higher than that of BP method.Compared with real tested values,mean absolute error,relative mean abso...
出处
《石河子大学学报（自然科学版）》
CAS
2007年第6期757-761,共5页
Journal of Shihezi University(Natural Science)
基金
新疆兵团科技计划项目(2007YD24
2006YD43和2006GJS13)
关键词
土壤墒情
预测
农田水量平衡模型
BP神经网络模型
对比
soil moisture
prediction
farmland water balance model
back-propagation neural network model
comparative study
