【摘要】 收集太湖典型农田2010年10—12月和2011年3—6月2个时间段的逐日气象资料和土壤水分资料,运用BP(back propagation)神经网络和缺省因子分析法确定影响该地区土壤水分动态的主要气象因子(降水量、蒸发量、平均气温和平均地表温度以及平均风速),以这些主要影响因子作为输入变量建立该地区土壤水分动态模拟的BP神经网络模型。利用100组实测样本对神经网络进行训练,用剩余的64组实测样本进行检验。结果表明:0～14 cm和14～33 cm土壤含水量模拟的平均相对误差(MARE)最大为0.062 9,均方根误差(RMSE)最大为1.764,不同土壤层次的训练样本和检验样本的精度(PA)都在0.87以上。因此,BP神经网络用于太湖典型农田的土壤水分动态模拟是可行的。更多还原

【Abstract】 The meteorological data and soil moisture data of two time intervals(October,2010 to December,2010,and March,2011 to June,2011)in the typical farmland of Tai Lake region were chosen in this study.By using BP(back propagation)neural network and default factor method,the main meteorological factors affecting soil moisture dynamic were confirmed,which were precipitation,evaporation,average temperature,average surface temperature and the average wind speed.A BP neural network model for simulating the soil moisture dynamic in study area was established.100 groups of measured sample were used to train the neural network,while the remaining 64 groups were used for model test.The evaluation standard of model simulation showed a better simulated results(the maximal MARE and RMSE of soil moisture simulation in 0-14 cm layer and 14-33 cm layer were 0.062 9 and 1.764 respectively,while both precisions were more than 0.87),which indicated that BP neural network could be used for the simulation of soil moisture dynamic in typical farmland of Tai Lake region was feasible.更多还原