标题:
区域性水土空间变异的人工神经指示克立格技术研究Research on the Zonal Water-Soil Resource Spatial Variation Using Artificial Neural
Indicator Kriging Technology
作者:
刘全明, 陈永强
关键字:
区域性, 水土空间变异, 人工神经指示克立格Zonal, Water-Soil Resource Spatial Variation, Artificial Neural Indicator Kriging
期刊名称:
《Journal of Water Resources Research》, Vol.3 No.2, 2014-04-09
摘要:
本文基于非参数地质统计学的指示Kriging法思想，借助人工神经网络高度非线性逼近的技术优点，以河套灌区内一典型实验区的土壤水盐空间变异性为案例，进行了非参数统计与人工智能技术的创新性融合研究。经与普通克立格、BP神经网络、指示克立格与析取克立格估计值对比分析发现：人工神经指示克立格法具有指示克立格的优点，即对数据无统计假定，不涉及特异值识别与处理，又可以高度逼近线性与非线性函数，很好地解决非线性估值问题，可用于土壤水盐监测及相关工作中。
Based on the spatial variation of soil water-salt at one representative experimental section in Hetao irrigation zone, directed by Indicator Kriging of non-parameter Geostatistic theory, using the high non-linear approximating ability of ANN, this paper innovated fusion of the non-parameter statistic and artificial intelligence technologies. Compared with the Ordinary Kriging, BP, Indicator Kriging, and Distinctive Kriging estimating values, it is founded that Artificial Neural Indicator Kriging has the characteristics of indicator kriging: 1) no need to make statistical assumptions for raw data; 2) not involving identification and specificity values treatment; 3) a high degree of approximation of linear and nonlinear functions; which can solve nonlinear problems effectively, so it can be used to monitor soil water and salt and the related work.