【摘要】 为筛选和构建适合苏北沿海滩涂围垦农田耕层土壤饱和水力传导率间接估算的土壤转换函数,在典型地块实测土壤饱和导水率和相关土壤基本性质的基础上,分析了11种根据基本土壤性质预测饱和导水率的转换函数方法的适用性,同时探讨了基于人工神经网络方法的土壤转换函数的预测效果。结果表明:滩涂围垦农田耕层土壤平均饱和导水率为10.04 cm/d,属低透水强度;在现有的土壤饱和导水率转换函数中,Vereecken函数是最适合滩涂围垦农区土壤、具有最佳预测精度的转换函数,其预测均方根误差为8.154,其次是Li、Campbell和Rawls函数;以砂粒、粘粒、容重和有机质作为输入因子,基于人工神经网络的土壤转换函数较Vereecken函数其预测均方根误差降至7.920,在输入因子中增加土壤盐分指标可进一步提高饱和导水率的预测精度,其预测均方根误差降为7.634。本文的研究结果显示利用人工神经网络方法建立的转换函数可有效提高滩涂盐渍农田土壤饱和导水率的预测精度。更多还原

【Abstract】 In order to select and establish the pedo-transfer function( PTF) which was suitable for the estimation of topsoil saturated hydraulic conductivity in the coastal reclaimed farmland of north Jiangsu province,soil saturated hydraulic conductivity and some soil basic properties were measured in the characteristic field in this region. The applicability of 11 widely used pedo-transfer functions( PTFs),which used soil basic properties to predict saturated hydraulic conductivity,was examined.The prediction performance of a novel PTF based upon artificial neural network( ANN) was also tested. Results indicated that the average topsoil saturated hydraulic conductivity across the study field was 10. 04 cm/d and categorized as low permeable strength. Vereecken PTF was selected as the most suitable one among the 11 widely used PTFs,followed by Li,Campbell and Rawls PTFs. Vereecken PTF had the highest prediction accuracy with the lowest prediction root mean-square-error( RMSE)of 8. 154. Using sand,clay,bulk density and soil organic matter as input variables,the novel PTF based upon ANN had better prediction performance than Vereecken PTF,and the prediction RMSE of the novel PTF decreased to 7. 920. The prediction accuracy was further improved when soil salinity was added to the input variables and the prediction RMSE decreased to7. 634. It was concluded that the novel PTF developed from artificial neural network could improve the prediction performance of saturated hydraulic conductivity in the coastal salt-affected farming area.更多还原