基于磁感式土壤表观电导率空间变异性的插值方法比较

针对目前黄河三角洲地区存在的土壤盐渍障碍问题，以该地区典型地块为研究对象，运用磁感大地电导率仪(EM38)，结合GIS与地统计学研究了不同样点密度下土壤表观电导率的空间变异特征，确定了最佳的空间插值模式，并采用偏差指数对各分布图层的空间相似性进行了评价。结果表明：不同样点密度下的土壤表观电导率均呈中等变异强度，并服从对数正态分布；各样点密度下的土壤表观电导率均表现为强空间相关性，其空间变异主要表现在小于10 m的田间尺度上，且对于预测精度，泛克立格>普通克立格>反距离权重>局部多项式。偏差指数法分析表明，各插值方法分布图的空间相似性均随样点密度的降低而下降；对于相同的样点密度减小比例下的空间信息保留度，泛克立格>普通克立格>反距离权重>局部多项式，采用泛克立格法可以在保证预测精度的基础上合理降低样点密度。该研究为黄河三角洲地区盐渍土地磁感式田间数据采集的合理密度确定以及优化空间插值模式选取提供了理论依据与技术参考。

Comparison of interpolation approaches based on spatial variability of apparent soil electrical conductivity with an electromagnetic induction

To solve the problem of soil salinization existing in the Yellow River Delta at present, electromagnetic induction EM38 was applied to study the spatial variability of apparent electrical conductivity under various sample densities with the GIS technology and geostatistics method, which was conducted in characteristic field of the Yellow River Delta. The optimized spatial interpolation approach was determined through systematic comparisons, and distortion index(DI) was introduced to evaluate the spatial similarity of apparent electrical conductivity distribution patterns. Results indicate that apparent electrical conductivity data under various sample densities all exhibite moderate spatial variation and follow logarithmic normal distribution, and apparent electrical conductivity under various sample densities all show strong spatial autocorrelation with the spatial variability at the field scale of less than 10 m. As to prediction accuracy, the sequence is universal kriging > ordinary kriging > inverse distance weighed > local polynomial. Distortion index analysis show that the spatial similarity of apparent electrical conductivity distribution patterns all decline with the decrease of sample density. In terms of the extent of spatial information retention under the same sample density reduction, the sequence is universal kriging > ordinary kriging > inverse distance weighed > local polynomial. It means that the sample density can be rationally reduced and prediction accuracy will be maintained with the application of universal kriging. With respect to the field data collection by using electromagnetic induction EM38, the research results can serve as a theoretical and technical reference to the determination of rational sample density and optimized spatial interpolation model in saline soils of the Yellow River Delta.