典型矿区非正态分布土壤元素数据的正态变换方法对比研究  

Comparison of Normalization Methods for Non-Normal Distributed Soil Elements Data in Typical Mining Area

作　　者：李晓晖[1] 袁峰[1] 白晓宇[1] 张明明[1] 贾蔡[1] 周涛发[1] 

机构地区：[1]合肥工业大学资源与环境工程学院,安徽合肥230009

出　　处：《地理与地理信息科学》2010年第6期102-105,共4页Geography and Geo-Information Science

基　　金：新世纪优秀人才支持计划项目(NCET-10-0324);安徽省科技攻关计划项目(08010302200);安徽省公益性地质(科技)工作项目(2009-13);安徽省优秀青年科技基金项目(08040106907;04045063)

摘　　要：对于铜陵矿区这类变异性强、偏度大、不符合正态分布的土壤元素样品数据,直接进行空间变异及插值分析会产生较大误差,需首先选择合适的正态变换方法进行稳健处理。该文以铜陵矿区表层土壤中的Pb元素数据为例,采用不同的正态变换方法进行数据正态变换,并通过变异函数对其变换效果进行分析。对比研究显示:对数变换无法使铜陵矿区土壤元素数据服从正态分布;Box-Cox变换虽使数据的正偏度有所减少,但仍难以通过K-S正态检验;而Johnson变换结果最优,其通过了K-S检验,对于异常数据的处理效果更优。Johnson变换与Box-Cox变换均可使变异函数表现出更好的形态,其中Johnson变换具有更强的正态变换能力,对于研究区土壤元素的变异函数稳定形态具有一定优势,是数据变异性强的成矿区地统计学数据正态变换的理想工具。For the strongly variable,large skewed and non-normal distributed soil sample data,such as those in Tongling mining area,spatial variability analysis and interpolation directly will lead to considerable errors, so it needs to select an appropriate normal transformation method to perform the robust processing firstly. In this paper,Pb element data of surface soil in Tongling mining area was taken for example,the raw data were transformed by different normal transformation methods and the effects was analyzed by variogram,expecting to provide a mensurable basis for applying geostatistics more reasonable and effective in the strong spatial variability metallogenic region. Comparison study showed that the logarithmic transformation could not make the soil element data in Tongling mining area obeying normal distribution,although Box-Cox transformation could decline the skewness of the data, it still didn＇t pass the K-S test. Johnson transformation is an optimal method and the results passed the K- S test successfully, especially for non-normal distributed data. Both Johnson transformation and Box-Cox transformation could make the variogram shape better,and Johnson transformation had stronger normalization capacity and advantage for stabilizing the shape of variogram. Johnson transformation is an ideal geostatistics normalization tool for the strong spatial variability metallogenic region.

关 键 词：地统计学 正态变换 土壤 元素 铜陵 

分 类 号：P628.1[天文地球—地质矿产勘探]