地球化学矿致异常空间分析与定量评价
|
摘要
地球化学勘查作为一门重要的找矿方法在矿产资源勘查中占据重要的地位,尤其适宜于贵金属Au、Ag以及有色金属Cu、Pb、Zn等矿种的勘查。在地球化学勘查活动中,对获取的地球化学数据进行矿致异常的识别和评价及其矿质来源的辨析是十分关键的一个环节,直接关系到下一步勘查决策的制定和找矿实践的成败。通常认为地球化学异常是指某一元素(同位素)含量在采样介质中富集或贫化的一种异常现象,若元素趋于富集称为正异常,反之则称为负异常。由于成矿元素是一个逐步富集的过程,由矿床或矿化引起的异常(称为矿致异常)往往是正异常,也是评价的主要对象。就矿致异常的提取和评价而言,空间分析技术和定量评价方法的研究一直是国内外勘查地球化学家角逐的热点。
近年来,不同学科的主题数据呈几何指数急剧膨胀,互联网技术和数据库技术也得到极大发展,使得传统的数据分析成为一门独立的交叉学科,称为数据挖掘或数据科学,其基本含义为通过某一主题数据的分析和挖掘,从而获取有价值的专业知识,实现从数据工厂到数据黄金的变革。当勘查地球化学家对勘查地球化学空间数据进行空间分析和定量评价时,实际上是扮演着数据科学家或数据挖掘师的身份,即针对多来源和多尺度的高维地球化学数据进行分析,从而挖掘其中隐藏的地球化学找矿知识。
基于上述研究背景,以现代统计学、稳健统计学、成分数据分析和空间分析理论与技术为指导,充分利用GIS和R编程语言为数据分析工具,针对勘查地球化学主题数据实现两个目标,一是提出地球化学矿致异常空间分析和定量评价的新方法与新技术,二是所挖掘的找矿知识直接服务于区域和矿区两种尺度的矿产勘查决策。
根据上述研究目标,研究内容主要包括：
(a)深入认识和系统总结地球化学空间数据固有的基本特征,尤其是离群性、封闭性和自相关性等基本特征的理论分析和实证研究；
(b)设计一套地球化学空间数据检查的解决方案和总结几种常用的数据变换方法,重点讨论对数比数据变换的优越性和必要性；
(c)开展单元素和多元素的探索性空间数据分析,包括单元素分布形式的各种判定方法和多元素之间相关关系、组合规律和空间趋势变化的可视化显示；
(d)对比不同方法计算单元素异常下限的结果,实现多元素离群点自动检测和可视化显示技术,辨析Cu、W等地球化学矿致异常的矿质来源和示踪其空间过程(成矿作用);
(e)研究聚类分析、判别分析和回归分析等基于数据驱动的矿致异常评价方法,分析其应用效果并评价结果的有效性。研究基于知识驱动(“标准样本”找矿模型)的矿致异常定量评价新方法。
本论文选取九瑞铜矿和大湖塘钨矿作为研究对象,数据源以九瑞和大湖塘地区的区域地球化学数据为主(九瑞地区1081件组合样,大湖塘地区3142件组合样),还包括在大湖塘地区野外采集的6条风化土壤剖面分析数据和收集的石门寺矿区4号勘探线剖面分析数据(6516件样品),数据具有多介质(水系沉积物、土壤和岩石)、多元素(高达39种)和多尺度(区域和矿区)的特点。
在充分认识九瑞和大湖塘地区不同尺度地质矿产特征的基础上,分别对Cu、W矿致异常的空间分析与定量评价展开了系统研究。取得了如下6个方面的研究成果：
1)针对勘查地球化学数据分析元素种类多且样本量大的特点,尤其是区域地球化学数据分析的元素达39个,首次系统明确了勘查地球化学数据具有10个方面的固有基本特征,依次为：非负性、连续性、批次性、变异性(异质性)、自相关性(依赖性)、尺度不变性(自相似性)、叠加性(混合性或不均一性)、删失性(截尾性)、离群性和封闭性。在充分认识这10个基本特征的基础上,设计出5个地球化学数据检查的可选方案和总结了5种重要的数据变换方法,进一步指出了地球化学成分数据为消除其封闭性产生的伪相关,须进行对数比数据变换。
2)结合地球化学数据的基本特征,提出了一套单元素和多元素探索性空间分析的完整解决方案。
(a)关于单元素的探索性分析,提出了单元素分布形式的4种判定方法,主要为图解法、偏度和峰度概要统计参数、Kolmogorov-Smirnov和Shapiro-Wilk正态检验以及C-A分形法,并指出了各种方法的优缺点。直观的图解法包括一维散点图、直方图、密度图、经验分布函数图、QQ图、累积概率图、PP图和箱图,其中直方图、累积概率图和Tukey箱图等应用十分广泛,尤其适用于探查勘查地球化学数据的离群性、删失性和叠加性等。另外,指明了直方图中分组数目的合理划分是其制图分析的关键参数。最后,推荐联合多个图解共同探索单元素的分布形式和数据结构,这样才能全面认识数据的基本特征。
(b)除了单元素探索性分析,讨论了多元素可视化探索性分析的4种图解,分别是相关图、平行坐标图、空间趋势图和空间距离图。相关图即是两两元素之间相关系数的矩阵化表达,对比相关图中的Pearson相关系数和稳健相关系数发现,稳健相关系数一般小于Pearson相关系数,其反映的是数据的主体信息。当进一步用颜色和图案直观区分相关图中相关系数的大小后,这对于多元素之间相关系数的对比将十分便捷和高效。平行坐标图用于探查多元素之间的组合关系。设计出多元素在任意方向和区带上关联分析的空间趋势图和空间距离图,为地球化学矿致异常的空间分析和定量评价提供了一个新的思路。
3)地球化学矿致异常正确识别的前提是背景与异常的合理分割,对单元素而言,实际上是异常下限的计算,本次研究采用了基于经典统计学的均值标准差法(Mean+2SD)、基于稳健统计学的中位数绝对偏差法(Median+2MAD)和Tukey箱图法(Q3+1.5·IQR)以及基于百分数的累积频率法(98%)等4种计算方法。对比这4种方法所计算的异常下限结果发现：基于经典统计学的均值标准差法不适宜于计算成矿元素含量的异常下限；基于稳健统计学的中位数绝对偏差法计算的异常下限能够最大程度避免极端值的影响；基于百分位数的累积频率法始终能各自识别出占总样本数一定比例(如2%)的高值样本和低值样本；推荐异常下限的计算方法使用基于稳健统计学的中位数绝对偏差法和Tukey箱图法,但二者在计算之前须进行Log10变换或Logit变换。
除了单元素的异常下限计算,实现了多变量离群点的自动检测和可视化技术,基于城门山铜矿床和石门寺钨矿床的特征元素组合,在九瑞地区检测出Cu+Mo+Au+Ag+W+Sb+Zn元素组合的离群点,呈高度聚集分布和随机离散分布,其高度聚集的离群点对应已知的大型铜矿床或Cu元素成矿概率较大的地区；在大湖塘地区检测出W+Cu+F+Ag+As元素组合的离群点,也呈高度聚集分布和随机离散分布,其高度聚集的离群点对应已知的大型钨矿床或W元素成矿概率较大的地区。
4)在Cu、W矿致异常识别的基础上,按照多尺度和多介质的聚焦找矿思路,辨析了Cu、W矿致异常的矿质来源,有如下3点认识：
(a)通过辨析九瑞地区Cu异常和大湖塘地区W异常的矿质来源发现,“高、大、全”Cu异常的矿质来源于深源中酸性岩浆岩,如形成的城门山矽卡岩型-斑岩型铜矿床和武山矽卡岩型铜矿床,而“高、大、全”W异常的矿质来源则来源于壳源的酸性花岗岩,如形成的石门寺岩浆热液型钨矿床、狮尾洞岩浆热液型钨矿床和香炉山矽卡岩型钨矿床。
(b)通过区域→空间子区再到矿床邻域的多尺度分析,以元素的地球化学性质为理论依据,按照元素组合的关联性和区分性,获得了城门山铜矿、武山铜矿、曾家垅锡矿、石门寺钨矿、狮尾洞钨矿和香炉山钨矿的特征元素组合,依次为：Cu+Mo+Au+Ag+W+Sb+Zn, Cu+Au+Ag+Pb+As+Sb, Sn+F+As+Sb, W+Cu+F+Ag+As, W+Cu+F+Ag+As+Sb, W+Cu+F+Ag,可作为区域上发现类似矿床的找矿标志。
(c)从元素“量”的视角出发,以大湖塘地区6个典型土壤风化剖面为例,发现了W等成矿元素含量在垂向土壤风化剖面上呈逐渐递增的变化趋势,进一步证明了土壤B层是该区最佳的采样层位。再以石门寺矿区4号勘探线剖面为例,首先对主成矿元素W、Cu在不同含量区间和不同岩石单元中的含量展开分析,揭示了钨矿体和铜矿体有两个显著的浓集中心和二者富矿体的空间叠合度高且主要分布在热液角砾岩和含钨石英脉中；然后利用Fry图解分析了W、Cu成矿元素空间迁移轨迹的方向和倾角,其迁移轨迹均为从NE向SW方向运移,但倾角略有不同,分别为25°～30°和30°。这为W、Cu矿体在剖面上的空间分布形态提供了定量证据。
(d)选择九瑞地区的Cu+Mo+Au+Ag+W+Sb+Zn元素组合和大湖塘地区的W+Cu+F+Ag+As+Sb元素组合,分别开展主成分分析(PCA)和因子(FA)分析。结果表明：双标图和主成分(因子)得分图是示踪Cu和W等成矿作用过程的有力工具,由于成分数据受数据离群性的影响和封闭性的约束,故在数据分析前推荐原始数据进行clr变换,就结果的解释度而言,PCA比FA能够获得更高的解释度。
5)以九瑞和大湖塘地区的区域地球化学数据为例,研究了聚类分析、判别分析和回归分析等3个基于数据驱动的矿致异常定量评价方法。结果发现：聚类分析能够用于优化九瑞和大湖塘地区的指示元素组合；判别分析适宜于九瑞地区3个空间子区(Cu矿化区、Sn矿化区和深海相沉积地层区)的样本分类,且Sn矿化区的分类误判率最低；通过回归分析,分别建立了城门山、武山、曾家垅、石门寺、狮尾洞和香炉山等6个典型矿床邻域内的多元线性回归方程,经过回归系数t检验、回归方程F检验和校正可决系数(Adjusted R2)等3个指标的综合判别以及进一步的诊断分析,认为6个多元线性回归方程对于Cu、Sn和W等成矿元素含量的预测精度较高。这将为基于知识驱动的定量评价奠定坚实的基础。
6)提出了基于知识驱动(“标准样本”找矿模型)的矿致异常定量评价新方法——相似系数法,主要应用于定量判别相似矿致异常的矿化类型。阐释了地球化学相似系数的内涵,制定出相似系数法的5个实施步骤,依次是：选典型矿床→挑元素组合→制“标准样本”→算距离值和转相似系数→绘相似系数图。重点对欧氏距离公式中的数据变换、“标准样本”指示元素的含量赋值和权重系数的计算、欧氏与兰氏距离公式、相似系数符号图的分级方案、相似系数等值区图的插值参数以及相似系数的自相关模式等几个方面进行了详细的讨论与分析。经反复试验和深入比较后获知：
(a)在欧氏距离计算公式中,对数变换优于匀化变换。
(b)标准样本”中各个指示元素含量的赋值推荐采用异常平均值,并建议使用相对权重方法计算各个元素的权重系数。
(c)九瑞地区优选兰氏距离公式作为计算2个铜矿“标准样本”(城门山和武山)相似系数的最佳距离公式,而大湖塘地区优选欧氏距离公式作为计算3个钨矿“标准样本”(石门寺、狮尾洞和香炉山)相似系数的最佳距离公式。
(d)当绘制相似系数符号图或等值区图时,为了便于对比,其分级方案推荐采用5级累积频率法(50th,90th,95th,98th)。
(e)以九瑞和大湖塘地区的区域地球化学数据为例,当绘制5个“标准样本”的相似系数等值区图时,建议其最佳搜索半径为4.5km,对应于最大全域型Moran's Ⅰ指数：再由5个“标准样本”相似系数的局域型LISA聚集图得出：高高聚集模式(High-High)和高低聚集模式(High-Low)是下一步的重点勘查区。
按照证据层组合决策的思想,最终给出九瑞和大湖塘地区具体的勘查决策建议为：
在九瑞地区,下一步铜矿勘查优先部署在城门山兰氏距离相似系数≥0.298且武山兰氏距离相似系数≥0.405的样本分布区,其次部署在仅城门山兰氏距离相似系数≥0.298或武山兰氏距离相似系数>≥0.405的样本分布区；在大湖塘地区,下一步钨矿勘查优先部署在同时满足石门寺欧氏距离相似系数≥0.602、狮尾洞欧氏距离相似系数≥≥0.578、香炉山欧氏距离相似系数≥0.717等3个条件的样本分布区,其次部署在同时满足前述3个条件中任意2个的样本分布区,最后部署在只满足前述3个条件中任意1个的样本分布区。
综上所述,本文的3个创新点为：
(a)针对多元素的可视化探索性分析,设计出多元素在任意方向和任意形状区带上关联分析的空间趋势图和空间距离图,为地球化学矿致异常的空间分析和定量评价提供了一个新的思路。
(b)提出了基于知识驱动(“标准样本”找矿模型)的矿致异常定量评价新方法—相似系数法,主要应用于定量判别相似矿致异常的矿化类型。阐释了地球化学相似系数的内涵,制定出相似系数法的5个实施步骤,依次是：选典型矿床→挑元素组合→制“标准样本”→算距离值和转相似系数→绘相似系数图。
(c)提出了矿致异常多尺度和多介质源辨析的新途径,查明了九瑞地区Cu矿致异常和大湖塘地区W矿致异常的矿质来源,并揭示了Cu、W成矿元素的空间过程(成矿作用),从而为矿致异常的定量评价提供了理论基础。
Geochemical exploration as an important prospecting method has a strong position in mineral exploration, which is very suitable to explore precious metals (gold, silver) and non-ferrous metals (copper, lead, zinc and so on). In geochemical exploration activities, recognition and evaluation of geochemical anomalies caused by ore bodies and discrimination of ore sources are both critical. When a kind of elements (isotopes) is riched or depleted in sampling media, it is called geochemical anomalies that can be devided into positive anomalies and negative anomalies. Since ore-forming elements are a gradual process of enrichment, geochemical anomalies caused by ore bodies are usually positive anomalies and the main evaluation object. In the case of recognition and evaluation of geochemical anomalies caused by ore bodies, its spatial analysis and quantitative evaluation are two cutting edge problems.
In recent years, theme data from different disciplines are sharply increasing and database technology has been greatly improved, thus the traditional data analysis has become an independent interdisciplinary called data mine or data science. The purpose of data mine is to discover professional knowledges, so data gold can be refined from data factory. When exploration geochemists analyze and evaluate geochemical data, meanwhile they are also playling the role of data scientists or data mining engineers. That is, through analysis of high-dimensional geochemical data derived from two or more sources, prospecting knowledges can be obtained.
Based on above researth background, this thesis regards modern statistics, robust statistics, compositional data analysis and spatial analysis theory as guides. Both GIS software and R contributed packages are used for univariate and multivariate statistical analysis. Finally the researth will achieve two goals; one is proposing new quantitative methods, the other is prospecting knowledges served as exploration decision-making.
According to the research objectives, the researth contents mainly include as follows:
(a) Insight and summary of several inherent properties of exploration geochemical data, especially theoretical analysis and empirical study of outlier, closure, and spatial autocorrelation;
(b) A solution of geochemical data check and summary of various data transformations; discussion of the superiority of three kinds of logratio data transformation;
(c) Exploratory data analysis of single element and multielement, concerning statistical and spatial distribution of single element, and graphic visualization of multielement;
(d) Contrastive analysis of geochemical anomaly threshold calculated by four different ways, implementation of automated multielement outlier detection and visualization, analysis of a new source identification method about geochemical anomalies caused by ore bodies, both PCA and FA used to trace spatial process of copper and tungsten (metallogenesis);
(e) data-driven quantitative methods such as cluster analysis, discriminant analysis, and regression analysis, as well as a novel knowledge-driven quantitative method called similarity method.
This dissertation selects the Jiurui copper and Dahutang tungsten as two study cases. There are different data types among research area, namely1081composite samples of regional geochemical data in Jiurui ore field,3142composite samples of regional geochemical data in Dahutang ore field,6weathered soil profiles and6516chemical samples in No.4exploration section located at Shimensi ore mine. These data have distinctive multimedia (rock, soil, and stream sediment), multielement (up to39elements), and multiscale (regional and local area).
Based on fully understanding geological and metallogenic characteristics of Jiurui and Dahutang ore field, then making further efforts to analyze and evaluate Cu and W geochemical anomalies caused by ore bodies, ultimately, the eventual research findings of six aspects are generalized as below:
1) Ten salient features of exploration geochemical data have been summarized in detail for the first time, which are nonnegativity, continuity, batchwise, heterogeneity, autocorrelation, scale invariance, superposition (mixing), truncation (censored), outlier, and closure. And besides, five alternatives of geochemical data check have been proposed and five significant methods of data transformation have been summarized. Moreover, three kinds of logratio data transformations are the best solution to solve closure problem about compositional data.
2) Combined with some basic characteristics of geochemical data, a integral solution is provided about exploratory data analysis of single element and multielement respectively.
(a) For exploratory data analysis of single element, four kinds of methods are used to explore statistical distribution and spatial distribution, which are graphic method, skewness and kurtosis, Kolmogorov-Smirnov and Shapiro-Wilk normal test, and C-A fractal method. Their strengths and weaknesses are analyzed. The intuitively graphic method includes the one-dimensional scatterplot, the histogram, the density trace, plot of the empirical cumulative distribution function (CDF-plot), the quantile-quantile plot (QQ-plot), the cumulative probability plot (CP-plot), the probability-probability plot (PP-plot), and boxplots. The histogram, CP-plot, and Tukey boxplot have good performance in outlier and censored data detection, so they are very popular in geochemical domain. In addition, rational division of the block number is crucial in histogram. So, in order to comprehensively inspect data structure and behavior, it is recommended to combine multiple diagrams.
(b) Except exploratory data analysis of single element, there are four multivariate graphics can be used to reveal the interrelation of multielement, whose names are called as correlogram, parallel coordinates plot, spatial trend plot, and spatial distance plot. The function of correlogram is used to display correlation coefficient matrix. Comparison of Pearson correlation coefficient and robust correlation coefficient, it has been shown that the former is larger than the later. Futher, when colours and icons are used to denote the size of correlation coefficients, it is very efficient and convenient to compare correlation coefficients. The parallel coordinates plot is a graphical representation of all profiles of all observations in just one plot, thus it is a powerful tool for multielement analysis. The spatial trend plot is designed to study the distribution of all the measured variables along a transect line via selecting the x-(for the east-west-transect) or y-coordinate (for the north-south-transect) for the x-axis of the plot and any other variable for the y-axis and plotting these as xy-plots. Whilst the spatial distance plot is designed to study systematic changes of multielement with distance from a defined point in all directions and irregular region.That is, these two plots provide a new idea for spatial analysis and quantitative evaluation of geochemical anomalies caused by ore bodies.
3) One of the essential prerequisites of geochemical anomaly recognition is to calculate a threshold between background and anomaly. The geochemical threshold is a critical thing for single element, which can be calculated by four formulas. The first formula is MEAN±2SD, which is based on classical statistics. The second formula is MEDIAN±2MAD, which is based on robust statistics. The third formula is Q3+1.5·IQR, which is based on Tukey boxplot. The last formula is uppermost2percent of the data defined as "outliers" for further inspection, which is based on percentiles. Compared with four results, findings demonstrate that the first formula is not appropriate for ore-forming elements and yet the second and the third formula are recommended because of their robustness. When one geochemist employs the second and the third formula to calculate geochemical threshold, it is better to transform original data by log10or logit. The problem of the last formula is that even percentiles will change with size or location of the survey area.
In addition to geochemical anomaly of single element, the algorithm of automated multielement outlier detection and visualization has been realized. Taking two element associations represented Chengmenshan copper deposit (Cu+Mo+Au+Ag+W+Sb+Zn) and Shimensi tungsten deposit (W+Cu+F+Ag+As) for example, the interpretation of multielement outliers is a challenge, but maps and graphics can help to understand the patterns of multielement outliers.
4) Based on recognition of geochemical anomaly caused by ore bodies for Cu and W, the sources of Cu and W geochemical anomlies caused by ore bodies have been identificated via multiscale and multimedia geochemical data. It shows that:
(a) The ore forming material of Cu anomalies is derived from deep source related with intermediate-acid magmatic rock, and conversely the ore forming material of W anomalies is erived from shallow source related with acidic granite.
(b) Using multiscale analysis (from regional area to local area) and principle of element classification, every typical deposit has been summarized a suite of element association. The element association of Chengmenshan copper deposit is Cu+Mo+Au+Ag+W+Sb+Zn; the element association of Wushan copper deposit is Cu+Au+Ag+Pb+As+Sb; the element association of Zengjialong tin deposit is Sn+F+As+Sb; the element association of Shimensi tungsten deposit is W+Cu+F+Ag+As; the element association of Shiweidong tungsten deposit is W+Cu+F+Ag+As+Sb; the element association of Xianglushan tungsten deposit is W+Cu+F+Ag. These element associations can be regarded as prospecting criteria in regional geochemical exploration.
(c) From the point of view of element concentration, in the case of six weathered soil profiles in Dahutang ore field, the concentration of ore forming elements appears to increase slowly in vertical direction, so B-horizon soils are the best sampling horizon. Further more, taking No.4exploration section located at Shimensi ore mine for example, two main ore forming elements have been analyzed in different concentration intervals and rock types. Both copper ore body and tungsten ore body have two significant concentration centres and their bonanzas are overlapped especially in hydrothermal breccia and quartz vein bearing tungsten. Afterward Fry plot has been applied to reveal migration direction and dip angle of the trajectory of W and Cu. The results show that their migration directions are from north-east to south-west and their dip angles are25°～30°and30°respectively.
(d) Seven indicator elements (Cu+Mo+Au+Ag+W+Sb+Zn) selected in Jiurui ore field and six indicator elements (W+Cu+F+Ag+As+Sb) selected in Dahutang ore field have been carried out principal component analysis (abbreviated to PCA) and factor analysis (abbreviated to FA). It has been realized that biplots and scores map are useful to trace minerogenesis process of Cu and W. Because geochemical compositional data are affected by outliers and constrained by closure, it is better to transform original data via centered logratio tansformations (abbreviated to clr) prior to statistic analysis. As far as interpretability of statistical results, the interpretability of PCA is higher than the interpretability of FA.
5) Three data-driven methods have been studied, which is aimed at quantitative evaluation of Cu and W geochemical anomalies caused by ore bodies in Jiurui and Dahutang ore field. These three methods are cluster analysis, discriminant analysis and regression analysis. Cluster analysis can be used to optimize element association for regional geochemical data. Discriminant analysis is effective to classify observations in three subareas called copper (Cu) mineralized area, tin (Sn) mineralized area, and deep-sea facies sedimentary strata area, and the consequence is that error rate of classification in Sn mineralized area is the lowest. Six multiple linear regression equations have been established in six corresponding the nearest neighbour of Chengmenshan, Wuhan, Zengjialong, Shimensi, Shiweidong, and Xianglushan deposit. According to comprehensive analysis (including t test of regression coefficients, F test of regression equation, and adjusted R square), prediction accuracy of all six multiple linear regression equations is good. These conclusions can provide a strong foundation for next knowledge-driven analysis.
6) A novel knowledge-driven method has been created to evaluate Cu and W geochemical anomalies caused by ore bodies in two ore fields named Jiurui and Dahutang, and its name is known as similarity coefficient method and its fuction is to identify similar mineralization type. Subsequently, the essential connotation of geochemical similarity coefficient is explained and five executive steps have been made. The first step is to select some typical deposits. The second step is to sieve element association. The third step is to make "standard sample". The fourth step is to calculate distance value and then to convert into similarity coefficient. The last step is to draw map of similarity coefficient. A few key details have been profoundly discussed, such as data transformations, concentration assignment of "standard sample", calculation of weight coefficient, distance formula of Euclidean and Canberra, classification scheme in symbol map of similarity coefficient, spatial interpolation parameter in contour map of similarity coefficient and spatial autocorrelation pattern of similarity coefficient. Knowledges have been acquired by trial and error and intensive comparison as below.
(a) Logarithmic transformation is better than normalized transformation in Euclidean distance formula.
(b) The concentration of indicator element is assigned arithmetical mean of geochemical anomaly and its weight is preferred to calculate by relative weights algorithm in "standard sample".
(c) Improved Canberra distance formula is preferable to calculate similarity coefficient of two "standard samples" of copper deposit in Jiurui ore field, whilst improved Euclidean distance formula is more suitable to calculate similarity coefficient of three "standard samples" of tungsten deposit in Dahutang ore field.
(d) When drawing symbol and contour map of similarity coefficient, for the purposes of comparison, classification scheme is adopted by percentile classes. Geochemists are often more interested in the tails of the distribution of similarity coefficient, so to highlight the tails the50th,90th,95th,98thpercentiles can be used.
(e) Taking regional geochemical data for example, when drawing contour map of similarity coefficient in Jiurui and Dahutang ore fields, the optimum search radius is4.5km corresponding to maxium global Moran's I. Five LISA cluster maps shows that High-High pattern and High-Low patern are next key exploration target areas.
According to thoughts of integrated decision-making, next prospecting strategies of Jiurui and Dahutang ore fields are suggested as below.
In Jiurui ore field, if Canberra similarity coefficient of observations calculated by Chengmenshan copper "standard sample" is not less than0.298and Canberra similarity coefficient of observations calculated by Wushan copper "standard sample" is not less than0.405, these observations can be regarded as the first consideration in next copper prospecting targets. However, if one of above two conditions is not satisfied, these observations can be regarded as the second choice in next copper prospecting targets. In Dahutang ore field, if three conditions are satisfied, which are Euclidean similarity coefficient of observations calculated by Shimensi tungsten "standard sample" is not less than0.602, Euclidean similarity coefficient of observations calculated by Shiweidong tungsten "standard sample" is not less than0.578and Euclidean similarity coefficient of observations calculated by Xianglushan tungsten "standard sample" is not less than0.717, these observations can be regarded as the first consideration in next tungsten prospecting targets. If one of above three conditions is not satisfied, these observations can be regarded as the second choice in next tungsten prospecting targets. If two of above three conditions is not satisfied, these observations can be regarded as the third choice in next tungsten prospecting targets.
In conclusion, there are three inovations as follows:
(a) There are two powerful spatial plots to be used to display the spatial relationship of multielement. The spatial trend plot is designed to study the distribution of all the measured variables along a transect line via selecting the x-(for the east-west-transect) or y-coordinate (for the north-south-transect) for the x-axis of the plot and any other variable for the y-axis and plotting these as xy-plots. Whilst the spatial distance plot is designed to study systematic changes of multielement with distance from a defined point in all directions and irregular region.That is, these two plots provide a new idea for spatial analysis and quantitative evaluation of geochemical anomalies caused by ore bodies.
(b) A novel knowledge-driven method has been created to evaluate Cu and W geochemical anomalies caused by ore bodies in Jiurui and Dahutang ore field, whose name is known as similarity coefficient method and whose fuction is to identify similar mineralization type. Subsequently, the essential connotation of geochemical similarity coefficient is explained and five executive steps have been made. The first step is to select some typical deposits. The second step is to sieve element association. The third step is to make "standard sample". The fourth step is to calculate distance value and then to convert into similarity coefficient. The last step is to draw map of similarity coefficient.
(c) A fresh idea is proposed that the sources of geochemical anomlies caused by ore bodies have been identificated via multiscale and multimedia geochemical data. The sources of Cu and W geochemical anomlies caused by ore bodies have been identificated and the minerogenesis process of Cu and W is traced by PCA and FA.
近年来,不同学科的主题数据呈几何指数急剧膨胀,互联网技术和数据库技术也得到极大发展,使得传统的数据分析成为一门独立的交叉学科,称为数据挖掘或数据科学,其基本含义为通过某一主题数据的分析和挖掘,从而获取有价值的专业知识,实现从数据工厂到数据黄金的变革。当勘查地球化学家对勘查地球化学空间数据进行空间分析和定量评价时,实际上是扮演着数据科学家或数据挖掘师的身份,即针对多来源和多尺度的高维地球化学数据进行分析,从而挖掘其中隐藏的地球化学找矿知识。
基于上述研究背景,以现代统计学、稳健统计学、成分数据分析和空间分析理论与技术为指导,充分利用GIS和R编程语言为数据分析工具,针对勘查地球化学主题数据实现两个目标,一是提出地球化学矿致异常空间分析和定量评价的新方法与新技术,二是所挖掘的找矿知识直接服务于区域和矿区两种尺度的矿产勘查决策。
根据上述研究目标,研究内容主要包括：
(a)深入认识和系统总结地球化学空间数据固有的基本特征,尤其是离群性、封闭性和自相关性等基本特征的理论分析和实证研究；
(b)设计一套地球化学空间数据检查的解决方案和总结几种常用的数据变换方法,重点讨论对数比数据变换的优越性和必要性；
(c)开展单元素和多元素的探索性空间数据分析,包括单元素分布形式的各种判定方法和多元素之间相关关系、组合规律和空间趋势变化的可视化显示；
(d)对比不同方法计算单元素异常下限的结果,实现多元素离群点自动检测和可视化显示技术,辨析Cu、W等地球化学矿致异常的矿质来源和示踪其空间过程(成矿作用);
(e)研究聚类分析、判别分析和回归分析等基于数据驱动的矿致异常评价方法,分析其应用效果并评价结果的有效性。研究基于知识驱动(“标准样本”找矿模型)的矿致异常定量评价新方法。
本论文选取九瑞铜矿和大湖塘钨矿作为研究对象,数据源以九瑞和大湖塘地区的区域地球化学数据为主(九瑞地区1081件组合样,大湖塘地区3142件组合样),还包括在大湖塘地区野外采集的6条风化土壤剖面分析数据和收集的石门寺矿区4号勘探线剖面分析数据(6516件样品),数据具有多介质(水系沉积物、土壤和岩石)、多元素(高达39种)和多尺度(区域和矿区)的特点。
在充分认识九瑞和大湖塘地区不同尺度地质矿产特征的基础上,分别对Cu、W矿致异常的空间分析与定量评价展开了系统研究。取得了如下6个方面的研究成果：
1)针对勘查地球化学数据分析元素种类多且样本量大的特点,尤其是区域地球化学数据分析的元素达39个,首次系统明确了勘查地球化学数据具有10个方面的固有基本特征,依次为：非负性、连续性、批次性、变异性(异质性)、自相关性(依赖性)、尺度不变性(自相似性)、叠加性(混合性或不均一性)、删失性(截尾性)、离群性和封闭性。在充分认识这10个基本特征的基础上,设计出5个地球化学数据检查的可选方案和总结了5种重要的数据变换方法,进一步指出了地球化学成分数据为消除其封闭性产生的伪相关,须进行对数比数据变换。
2)结合地球化学数据的基本特征,提出了一套单元素和多元素探索性空间分析的完整解决方案。
(a)关于单元素的探索性分析,提出了单元素分布形式的4种判定方法,主要为图解法、偏度和峰度概要统计参数、Kolmogorov-Smirnov和Shapiro-Wilk正态检验以及C-A分形法,并指出了各种方法的优缺点。直观的图解法包括一维散点图、直方图、密度图、经验分布函数图、QQ图、累积概率图、PP图和箱图,其中直方图、累积概率图和Tukey箱图等应用十分广泛,尤其适用于探查勘查地球化学数据的离群性、删失性和叠加性等。另外,指明了直方图中分组数目的合理划分是其制图分析的关键参数。最后,推荐联合多个图解共同探索单元素的分布形式和数据结构,这样才能全面认识数据的基本特征。
(b)除了单元素探索性分析,讨论了多元素可视化探索性分析的4种图解,分别是相关图、平行坐标图、空间趋势图和空间距离图。相关图即是两两元素之间相关系数的矩阵化表达,对比相关图中的Pearson相关系数和稳健相关系数发现,稳健相关系数一般小于Pearson相关系数,其反映的是数据的主体信息。当进一步用颜色和图案直观区分相关图中相关系数的大小后,这对于多元素之间相关系数的对比将十分便捷和高效。平行坐标图用于探查多元素之间的组合关系。设计出多元素在任意方向和区带上关联分析的空间趋势图和空间距离图,为地球化学矿致异常的空间分析和定量评价提供了一个新的思路。
3)地球化学矿致异常正确识别的前提是背景与异常的合理分割,对单元素而言,实际上是异常下限的计算,本次研究采用了基于经典统计学的均值标准差法(Mean+2SD)、基于稳健统计学的中位数绝对偏差法(Median+2MAD)和Tukey箱图法(Q3+1.5·IQR)以及基于百分数的累积频率法(98%)等4种计算方法。对比这4种方法所计算的异常下限结果发现：基于经典统计学的均值标准差法不适宜于计算成矿元素含量的异常下限；基于稳健统计学的中位数绝对偏差法计算的异常下限能够最大程度避免极端值的影响；基于百分位数的累积频率法始终能各自识别出占总样本数一定比例(如2%)的高值样本和低值样本；推荐异常下限的计算方法使用基于稳健统计学的中位数绝对偏差法和Tukey箱图法,但二者在计算之前须进行Log10变换或Logit变换。
除了单元素的异常下限计算,实现了多变量离群点的自动检测和可视化技术,基于城门山铜矿床和石门寺钨矿床的特征元素组合,在九瑞地区检测出Cu+Mo+Au+Ag+W+Sb+Zn元素组合的离群点,呈高度聚集分布和随机离散分布,其高度聚集的离群点对应已知的大型铜矿床或Cu元素成矿概率较大的地区；在大湖塘地区检测出W+Cu+F+Ag+As元素组合的离群点,也呈高度聚集分布和随机离散分布,其高度聚集的离群点对应已知的大型钨矿床或W元素成矿概率较大的地区。
4)在Cu、W矿致异常识别的基础上,按照多尺度和多介质的聚焦找矿思路,辨析了Cu、W矿致异常的矿质来源,有如下3点认识：
(a)通过辨析九瑞地区Cu异常和大湖塘地区W异常的矿质来源发现,“高、大、全”Cu异常的矿质来源于深源中酸性岩浆岩,如形成的城门山矽卡岩型-斑岩型铜矿床和武山矽卡岩型铜矿床,而“高、大、全”W异常的矿质来源则来源于壳源的酸性花岗岩,如形成的石门寺岩浆热液型钨矿床、狮尾洞岩浆热液型钨矿床和香炉山矽卡岩型钨矿床。
(b)通过区域→空间子区再到矿床邻域的多尺度分析,以元素的地球化学性质为理论依据,按照元素组合的关联性和区分性,获得了城门山铜矿、武山铜矿、曾家垅锡矿、石门寺钨矿、狮尾洞钨矿和香炉山钨矿的特征元素组合,依次为：Cu+Mo+Au+Ag+W+Sb+Zn, Cu+Au+Ag+Pb+As+Sb, Sn+F+As+Sb, W+Cu+F+Ag+As, W+Cu+F+Ag+As+Sb, W+Cu+F+Ag,可作为区域上发现类似矿床的找矿标志。
(c)从元素“量”的视角出发,以大湖塘地区6个典型土壤风化剖面为例,发现了W等成矿元素含量在垂向土壤风化剖面上呈逐渐递增的变化趋势,进一步证明了土壤B层是该区最佳的采样层位。再以石门寺矿区4号勘探线剖面为例,首先对主成矿元素W、Cu在不同含量区间和不同岩石单元中的含量展开分析,揭示了钨矿体和铜矿体有两个显著的浓集中心和二者富矿体的空间叠合度高且主要分布在热液角砾岩和含钨石英脉中；然后利用Fry图解分析了W、Cu成矿元素空间迁移轨迹的方向和倾角,其迁移轨迹均为从NE向SW方向运移,但倾角略有不同,分别为25°～30°和30°。这为W、Cu矿体在剖面上的空间分布形态提供了定量证据。
(d)选择九瑞地区的Cu+Mo+Au+Ag+W+Sb+Zn元素组合和大湖塘地区的W+Cu+F+Ag+As+Sb元素组合,分别开展主成分分析(PCA)和因子(FA)分析。结果表明：双标图和主成分(因子)得分图是示踪Cu和W等成矿作用过程的有力工具,由于成分数据受数据离群性的影响和封闭性的约束,故在数据分析前推荐原始数据进行clr变换,就结果的解释度而言,PCA比FA能够获得更高的解释度。
5)以九瑞和大湖塘地区的区域地球化学数据为例,研究了聚类分析、判别分析和回归分析等3个基于数据驱动的矿致异常定量评价方法。结果发现：聚类分析能够用于优化九瑞和大湖塘地区的指示元素组合；判别分析适宜于九瑞地区3个空间子区(Cu矿化区、Sn矿化区和深海相沉积地层区)的样本分类,且Sn矿化区的分类误判率最低；通过回归分析,分别建立了城门山、武山、曾家垅、石门寺、狮尾洞和香炉山等6个典型矿床邻域内的多元线性回归方程,经过回归系数t检验、回归方程F检验和校正可决系数(Adjusted R2)等3个指标的综合判别以及进一步的诊断分析,认为6个多元线性回归方程对于Cu、Sn和W等成矿元素含量的预测精度较高。这将为基于知识驱动的定量评价奠定坚实的基础。
6)提出了基于知识驱动(“标准样本”找矿模型)的矿致异常定量评价新方法——相似系数法,主要应用于定量判别相似矿致异常的矿化类型。阐释了地球化学相似系数的内涵,制定出相似系数法的5个实施步骤,依次是：选典型矿床→挑元素组合→制“标准样本”→算距离值和转相似系数→绘相似系数图。重点对欧氏距离公式中的数据变换、“标准样本”指示元素的含量赋值和权重系数的计算、欧氏与兰氏距离公式、相似系数符号图的分级方案、相似系数等值区图的插值参数以及相似系数的自相关模式等几个方面进行了详细的讨论与分析。经反复试验和深入比较后获知：
(a)在欧氏距离计算公式中,对数变换优于匀化变换。
(b)标准样本”中各个指示元素含量的赋值推荐采用异常平均值,并建议使用相对权重方法计算各个元素的权重系数。
(c)九瑞地区优选兰氏距离公式作为计算2个铜矿“标准样本”(城门山和武山)相似系数的最佳距离公式,而大湖塘地区优选欧氏距离公式作为计算3个钨矿“标准样本”(石门寺、狮尾洞和香炉山)相似系数的最佳距离公式。
(d)当绘制相似系数符号图或等值区图时,为了便于对比,其分级方案推荐采用5级累积频率法(50th,90th,95th,98th)。
(e)以九瑞和大湖塘地区的区域地球化学数据为例,当绘制5个“标准样本”的相似系数等值区图时,建议其最佳搜索半径为4.5km,对应于最大全域型Moran's Ⅰ指数：再由5个“标准样本”相似系数的局域型LISA聚集图得出：高高聚集模式(High-High)和高低聚集模式(High-Low)是下一步的重点勘查区。
按照证据层组合决策的思想,最终给出九瑞和大湖塘地区具体的勘查决策建议为：
在九瑞地区,下一步铜矿勘查优先部署在城门山兰氏距离相似系数≥0.298且武山兰氏距离相似系数≥0.405的样本分布区,其次部署在仅城门山兰氏距离相似系数≥0.298或武山兰氏距离相似系数>≥0.405的样本分布区；在大湖塘地区,下一步钨矿勘查优先部署在同时满足石门寺欧氏距离相似系数≥0.602、狮尾洞欧氏距离相似系数≥≥0.578、香炉山欧氏距离相似系数≥0.717等3个条件的样本分布区,其次部署在同时满足前述3个条件中任意2个的样本分布区,最后部署在只满足前述3个条件中任意1个的样本分布区。
综上所述,本文的3个创新点为：
(a)针对多元素的可视化探索性分析,设计出多元素在任意方向和任意形状区带上关联分析的空间趋势图和空间距离图,为地球化学矿致异常的空间分析和定量评价提供了一个新的思路。
(b)提出了基于知识驱动(“标准样本”找矿模型)的矿致异常定量评价新方法—相似系数法,主要应用于定量判别相似矿致异常的矿化类型。阐释了地球化学相似系数的内涵,制定出相似系数法的5个实施步骤,依次是：选典型矿床→挑元素组合→制“标准样本”→算距离值和转相似系数→绘相似系数图。
(c)提出了矿致异常多尺度和多介质源辨析的新途径,查明了九瑞地区Cu矿致异常和大湖塘地区W矿致异常的矿质来源,并揭示了Cu、W成矿元素的空间过程(成矿作用),从而为矿致异常的定量评价提供了理论基础。
Geochemical exploration as an important prospecting method has a strong position in mineral exploration, which is very suitable to explore precious metals (gold, silver) and non-ferrous metals (copper, lead, zinc and so on). In geochemical exploration activities, recognition and evaluation of geochemical anomalies caused by ore bodies and discrimination of ore sources are both critical. When a kind of elements (isotopes) is riched or depleted in sampling media, it is called geochemical anomalies that can be devided into positive anomalies and negative anomalies. Since ore-forming elements are a gradual process of enrichment, geochemical anomalies caused by ore bodies are usually positive anomalies and the main evaluation object. In the case of recognition and evaluation of geochemical anomalies caused by ore bodies, its spatial analysis and quantitative evaluation are two cutting edge problems.
In recent years, theme data from different disciplines are sharply increasing and database technology has been greatly improved, thus the traditional data analysis has become an independent interdisciplinary called data mine or data science. The purpose of data mine is to discover professional knowledges, so data gold can be refined from data factory. When exploration geochemists analyze and evaluate geochemical data, meanwhile they are also playling the role of data scientists or data mining engineers. That is, through analysis of high-dimensional geochemical data derived from two or more sources, prospecting knowledges can be obtained.
Based on above researth background, this thesis regards modern statistics, robust statistics, compositional data analysis and spatial analysis theory as guides. Both GIS software and R contributed packages are used for univariate and multivariate statistical analysis. Finally the researth will achieve two goals; one is proposing new quantitative methods, the other is prospecting knowledges served as exploration decision-making.
According to the research objectives, the researth contents mainly include as follows:
(a) Insight and summary of several inherent properties of exploration geochemical data, especially theoretical analysis and empirical study of outlier, closure, and spatial autocorrelation;
(b) A solution of geochemical data check and summary of various data transformations; discussion of the superiority of three kinds of logratio data transformation;
(c) Exploratory data analysis of single element and multielement, concerning statistical and spatial distribution of single element, and graphic visualization of multielement;
(d) Contrastive analysis of geochemical anomaly threshold calculated by four different ways, implementation of automated multielement outlier detection and visualization, analysis of a new source identification method about geochemical anomalies caused by ore bodies, both PCA and FA used to trace spatial process of copper and tungsten (metallogenesis);
(e) data-driven quantitative methods such as cluster analysis, discriminant analysis, and regression analysis, as well as a novel knowledge-driven quantitative method called similarity method.
This dissertation selects the Jiurui copper and Dahutang tungsten as two study cases. There are different data types among research area, namely1081composite samples of regional geochemical data in Jiurui ore field,3142composite samples of regional geochemical data in Dahutang ore field,6weathered soil profiles and6516chemical samples in No.4exploration section located at Shimensi ore mine. These data have distinctive multimedia (rock, soil, and stream sediment), multielement (up to39elements), and multiscale (regional and local area).
Based on fully understanding geological and metallogenic characteristics of Jiurui and Dahutang ore field, then making further efforts to analyze and evaluate Cu and W geochemical anomalies caused by ore bodies, ultimately, the eventual research findings of six aspects are generalized as below:
1) Ten salient features of exploration geochemical data have been summarized in detail for the first time, which are nonnegativity, continuity, batchwise, heterogeneity, autocorrelation, scale invariance, superposition (mixing), truncation (censored), outlier, and closure. And besides, five alternatives of geochemical data check have been proposed and five significant methods of data transformation have been summarized. Moreover, three kinds of logratio data transformations are the best solution to solve closure problem about compositional data.
2) Combined with some basic characteristics of geochemical data, a integral solution is provided about exploratory data analysis of single element and multielement respectively.
(a) For exploratory data analysis of single element, four kinds of methods are used to explore statistical distribution and spatial distribution, which are graphic method, skewness and kurtosis, Kolmogorov-Smirnov and Shapiro-Wilk normal test, and C-A fractal method. Their strengths and weaknesses are analyzed. The intuitively graphic method includes the one-dimensional scatterplot, the histogram, the density trace, plot of the empirical cumulative distribution function (CDF-plot), the quantile-quantile plot (QQ-plot), the cumulative probability plot (CP-plot), the probability-probability plot (PP-plot), and boxplots. The histogram, CP-plot, and Tukey boxplot have good performance in outlier and censored data detection, so they are very popular in geochemical domain. In addition, rational division of the block number is crucial in histogram. So, in order to comprehensively inspect data structure and behavior, it is recommended to combine multiple diagrams.
(b) Except exploratory data analysis of single element, there are four multivariate graphics can be used to reveal the interrelation of multielement, whose names are called as correlogram, parallel coordinates plot, spatial trend plot, and spatial distance plot. The function of correlogram is used to display correlation coefficient matrix. Comparison of Pearson correlation coefficient and robust correlation coefficient, it has been shown that the former is larger than the later. Futher, when colours and icons are used to denote the size of correlation coefficients, it is very efficient and convenient to compare correlation coefficients. The parallel coordinates plot is a graphical representation of all profiles of all observations in just one plot, thus it is a powerful tool for multielement analysis. The spatial trend plot is designed to study the distribution of all the measured variables along a transect line via selecting the x-(for the east-west-transect) or y-coordinate (for the north-south-transect) for the x-axis of the plot and any other variable for the y-axis and plotting these as xy-plots. Whilst the spatial distance plot is designed to study systematic changes of multielement with distance from a defined point in all directions and irregular region.That is, these two plots provide a new idea for spatial analysis and quantitative evaluation of geochemical anomalies caused by ore bodies.
3) One of the essential prerequisites of geochemical anomaly recognition is to calculate a threshold between background and anomaly. The geochemical threshold is a critical thing for single element, which can be calculated by four formulas. The first formula is MEAN±2SD, which is based on classical statistics. The second formula is MEDIAN±2MAD, which is based on robust statistics. The third formula is Q3+1.5·IQR, which is based on Tukey boxplot. The last formula is uppermost2percent of the data defined as "outliers" for further inspection, which is based on percentiles. Compared with four results, findings demonstrate that the first formula is not appropriate for ore-forming elements and yet the second and the third formula are recommended because of their robustness. When one geochemist employs the second and the third formula to calculate geochemical threshold, it is better to transform original data by log10or logit. The problem of the last formula is that even percentiles will change with size or location of the survey area.
In addition to geochemical anomaly of single element, the algorithm of automated multielement outlier detection and visualization has been realized. Taking two element associations represented Chengmenshan copper deposit (Cu+Mo+Au+Ag+W+Sb+Zn) and Shimensi tungsten deposit (W+Cu+F+Ag+As) for example, the interpretation of multielement outliers is a challenge, but maps and graphics can help to understand the patterns of multielement outliers.
4) Based on recognition of geochemical anomaly caused by ore bodies for Cu and W, the sources of Cu and W geochemical anomlies caused by ore bodies have been identificated via multiscale and multimedia geochemical data. It shows that:
(a) The ore forming material of Cu anomalies is derived from deep source related with intermediate-acid magmatic rock, and conversely the ore forming material of W anomalies is erived from shallow source related with acidic granite.
(b) Using multiscale analysis (from regional area to local area) and principle of element classification, every typical deposit has been summarized a suite of element association. The element association of Chengmenshan copper deposit is Cu+Mo+Au+Ag+W+Sb+Zn; the element association of Wushan copper deposit is Cu+Au+Ag+Pb+As+Sb; the element association of Zengjialong tin deposit is Sn+F+As+Sb; the element association of Shimensi tungsten deposit is W+Cu+F+Ag+As; the element association of Shiweidong tungsten deposit is W+Cu+F+Ag+As+Sb; the element association of Xianglushan tungsten deposit is W+Cu+F+Ag. These element associations can be regarded as prospecting criteria in regional geochemical exploration.
(c) From the point of view of element concentration, in the case of six weathered soil profiles in Dahutang ore field, the concentration of ore forming elements appears to increase slowly in vertical direction, so B-horizon soils are the best sampling horizon. Further more, taking No.4exploration section located at Shimensi ore mine for example, two main ore forming elements have been analyzed in different concentration intervals and rock types. Both copper ore body and tungsten ore body have two significant concentration centres and their bonanzas are overlapped especially in hydrothermal breccia and quartz vein bearing tungsten. Afterward Fry plot has been applied to reveal migration direction and dip angle of the trajectory of W and Cu. The results show that their migration directions are from north-east to south-west and their dip angles are25°～30°and30°respectively.
(d) Seven indicator elements (Cu+Mo+Au+Ag+W+Sb+Zn) selected in Jiurui ore field and six indicator elements (W+Cu+F+Ag+As+Sb) selected in Dahutang ore field have been carried out principal component analysis (abbreviated to PCA) and factor analysis (abbreviated to FA). It has been realized that biplots and scores map are useful to trace minerogenesis process of Cu and W. Because geochemical compositional data are affected by outliers and constrained by closure, it is better to transform original data via centered logratio tansformations (abbreviated to clr) prior to statistic analysis. As far as interpretability of statistical results, the interpretability of PCA is higher than the interpretability of FA.
5) Three data-driven methods have been studied, which is aimed at quantitative evaluation of Cu and W geochemical anomalies caused by ore bodies in Jiurui and Dahutang ore field. These three methods are cluster analysis, discriminant analysis and regression analysis. Cluster analysis can be used to optimize element association for regional geochemical data. Discriminant analysis is effective to classify observations in three subareas called copper (Cu) mineralized area, tin (Sn) mineralized area, and deep-sea facies sedimentary strata area, and the consequence is that error rate of classification in Sn mineralized area is the lowest. Six multiple linear regression equations have been established in six corresponding the nearest neighbour of Chengmenshan, Wuhan, Zengjialong, Shimensi, Shiweidong, and Xianglushan deposit. According to comprehensive analysis (including t test of regression coefficients, F test of regression equation, and adjusted R square), prediction accuracy of all six multiple linear regression equations is good. These conclusions can provide a strong foundation for next knowledge-driven analysis.
6) A novel knowledge-driven method has been created to evaluate Cu and W geochemical anomalies caused by ore bodies in two ore fields named Jiurui and Dahutang, and its name is known as similarity coefficient method and its fuction is to identify similar mineralization type. Subsequently, the essential connotation of geochemical similarity coefficient is explained and five executive steps have been made. The first step is to select some typical deposits. The second step is to sieve element association. The third step is to make "standard sample". The fourth step is to calculate distance value and then to convert into similarity coefficient. The last step is to draw map of similarity coefficient. A few key details have been profoundly discussed, such as data transformations, concentration assignment of "standard sample", calculation of weight coefficient, distance formula of Euclidean and Canberra, classification scheme in symbol map of similarity coefficient, spatial interpolation parameter in contour map of similarity coefficient and spatial autocorrelation pattern of similarity coefficient. Knowledges have been acquired by trial and error and intensive comparison as below.
(a) Logarithmic transformation is better than normalized transformation in Euclidean distance formula.
(b) The concentration of indicator element is assigned arithmetical mean of geochemical anomaly and its weight is preferred to calculate by relative weights algorithm in "standard sample".
(c) Improved Canberra distance formula is preferable to calculate similarity coefficient of two "standard samples" of copper deposit in Jiurui ore field, whilst improved Euclidean distance formula is more suitable to calculate similarity coefficient of three "standard samples" of tungsten deposit in Dahutang ore field.
(d) When drawing symbol and contour map of similarity coefficient, for the purposes of comparison, classification scheme is adopted by percentile classes. Geochemists are often more interested in the tails of the distribution of similarity coefficient, so to highlight the tails the50th,90th,95th,98thpercentiles can be used.
(e) Taking regional geochemical data for example, when drawing contour map of similarity coefficient in Jiurui and Dahutang ore fields, the optimum search radius is4.5km corresponding to maxium global Moran's I. Five LISA cluster maps shows that High-High pattern and High-Low patern are next key exploration target areas.
According to thoughts of integrated decision-making, next prospecting strategies of Jiurui and Dahutang ore fields are suggested as below.
In Jiurui ore field, if Canberra similarity coefficient of observations calculated by Chengmenshan copper "standard sample" is not less than0.298and Canberra similarity coefficient of observations calculated by Wushan copper "standard sample" is not less than0.405, these observations can be regarded as the first consideration in next copper prospecting targets. However, if one of above two conditions is not satisfied, these observations can be regarded as the second choice in next copper prospecting targets. In Dahutang ore field, if three conditions are satisfied, which are Euclidean similarity coefficient of observations calculated by Shimensi tungsten "standard sample" is not less than0.602, Euclidean similarity coefficient of observations calculated by Shiweidong tungsten "standard sample" is not less than0.578and Euclidean similarity coefficient of observations calculated by Xianglushan tungsten "standard sample" is not less than0.717, these observations can be regarded as the first consideration in next tungsten prospecting targets. If one of above three conditions is not satisfied, these observations can be regarded as the second choice in next tungsten prospecting targets. If two of above three conditions is not satisfied, these observations can be regarded as the third choice in next tungsten prospecting targets.
In conclusion, there are three inovations as follows:
(a) There are two powerful spatial plots to be used to display the spatial relationship of multielement. The spatial trend plot is designed to study the distribution of all the measured variables along a transect line via selecting the x-(for the east-west-transect) or y-coordinate (for the north-south-transect) for the x-axis of the plot and any other variable for the y-axis and plotting these as xy-plots. Whilst the spatial distance plot is designed to study systematic changes of multielement with distance from a defined point in all directions and irregular region.That is, these two plots provide a new idea for spatial analysis and quantitative evaluation of geochemical anomalies caused by ore bodies.
(b) A novel knowledge-driven method has been created to evaluate Cu and W geochemical anomalies caused by ore bodies in Jiurui and Dahutang ore field, whose name is known as similarity coefficient method and whose fuction is to identify similar mineralization type. Subsequently, the essential connotation of geochemical similarity coefficient is explained and five executive steps have been made. The first step is to select some typical deposits. The second step is to sieve element association. The third step is to make "standard sample". The fourth step is to calculate distance value and then to convert into similarity coefficient. The last step is to draw map of similarity coefficient.
(c) A fresh idea is proposed that the sources of geochemical anomlies caused by ore bodies have been identificated via multiscale and multimedia geochemical data. The sources of Cu and W geochemical anomlies caused by ore bodies have been identificated and the minerogenesis process of Cu and W is traced by PCA and FA.
引文
[1]龚鹏.矿产资源地球化学模型建立与定量预测方法技术研究[D].中国地质大学(武汉),
[2]王学求.勘查地球化学近十年进展[J].矿物岩石地球化学通报.2013,32(2)：190-197.
[3]奚小环,李敏.现代地质工作重要发展领域：“十一五”期间勘查地球化学评述[J].地学前缘.2013,20(3)：161-169.
[4]Cohen D R, Kelley D L, Anand R, et al. Major advances in exploration geochemistry,1998-2007[J]. Geochemistry:Exploration, Environment, Analysis.2010,10(1):3-16.
[5]Carranza E J M. Geochemical anomaly and mineral prospectivity mapping in GIS[M]. Elsevier,
[6]Howarth R J. Statistics and data analysis in geochemical prospecting[M]. Elsevier scientific publishing company,1983.
[7]Fletcher W K. Analytical methods in geochemical prospecting[M]. Elsevier Scientific Publishing Company,1981.
[8]Kauranne L K, Salimnen R, Eriksson K. Regolith exploration geochemistry in arctic and temperate terrains.[M]. Elsevier Science Publishers,1992.
[9]Butt C, Zeegers H. Regolith exploration geochemistry in tropical and subtropical terrains[M]. Elsevier Amsterdam and London,1992.
[10]Hale M. Geochemical remote sensing of the sub-surface[M]. Access Online via Elsevier,2000.
[11]Hein J R. Life cycle of the Phosphoria Formation:from deposition to the post-mining environment[M]. Gulf Professional Publishing,2003.
[12]Govett G J S. Rock geochemistry in mineral exploration[M]. New York.:Elsevier Scientific Pub. Co. Amsterdam,1983:461.
[13]Hale M, Plant J A. Drainage geochemistry. Handbook of exploration geochemistry[M]. Elsevier, Amsterdam,1994.
[14]Mukhopadhyay R, Ghosh A K, Iyer S D. The Indian Ocean nodule field:Geology and resource potential[M]. Access Online via Elsevier,2007.
[15]Dunn C E. Biogeochemistry in Mineral Exploration, Handbook of exploration and Environmental Geochemistry, V.9[M]. Elsevier,2007.
[16]Xie X, Mu X, Ren T. Geochemical mapping in China[J]. Journal of Geochemical Exploration. 1997,60(1):99-113.
[17]奚小环,李敏.中国区域化探若干基本问题研究：1999—2009[J].中国地质.2012,39(2)：267-282.
[18]黄芳.水土污染空间分析及源辨析[D].杭州：浙江大学,2010.
[19]Krige D G. A Statistical Approach to Some Mine Valuation and Allied Problems on the Witwatersrand[D]. University of the Witwatersrand,1951.
[20]Krige D G. A statistical analysis of some of the borehole values in the Orange Free State goldfield[J]. Journal of the Chemical and Metallurgical Society of South Africa.1952,53:47-64.
[21]Matheron G. Principles of geostatistics[J]. Economic geology.1963,58(8):1246-1266.
[22]肖斌,侯景儒.归来庄金矿床综合化探金异常的地质统计学研究[J].北京大学学报：自然科学版.2000,36(4)：531-538.
[23]黄静宁,赵鹏大.东天山地区Cu-Au异常信息提取与评价[J].地球科学：中国地质大学学报.2011,36(2)：317-326.
[24]叶水盛,刘光胜.克立格方法在区域化探数据处理中的应用[J].长春科技大学学报.2000,30(2)：198-200.
[25]陈建国,王仁铎,陈永清.利用分形统计学提取化探数据中的隐蔽信息并圈定地球化学异常[J].地球科学：中国地质大学学报.1998,23(2)：175-178.
[26]胡光道,王仁铎,李正琪.对数泛克立金法在处理化探数据中的应用[J].地球科学.1985, 10(2)：141-150.
[27]Panahi A, Cheng Q, Bonham-Carter G F. Modelling lake sediment geochemical distribution using principal component, indicator kriging and multifractal power-spectrum analysis:a case study from Gowganda, Ontario[J]. Geochemistry:Exploration, Environment, Analysis.2004, 4(1):59-70.
[28]Moran P A. Notes on continuous stochastic phenomena[J]. Biometrika.1950,37(1/2):17-23.
[29]Cliff A D, Ord J K. Spatial processes:models & applications[M]. Pion London,1981.
[30]Cressie N A. Statistics for Spatial Data[M]. Wiley Series in Probability and Mathematical Statistics:Applied Probability and Statistics. John Wiley & Sons,1991.
[31]Cressie N A. Statistics for Spatial Data, revised edition[M]. Wiley, New York,1993.
[32]Anselin L. Local indicators of spatial association-LISA[J]. Geographical analysis.1995,27(2):
[33]Fischer M M, Wang J. Spatial data analysis:models, methods and techniques[M]. Springer,
[34]Haining R P. Spatial data analysis:theory and practice[M]. Cambridge University Press,2003.
[35]Haining R P, Kerry R, Oliver M A. Geography, spatial data analysis, and geostatistics:an overview[J]. Geographical Analysis.2010,42(1):7-31.
[36]张朝生,陶澍,袁贵平,等.天津市平原土壤微量元素含量的空间自相关研究[J].土壤学报.1995,32(1)：50-57.
[37]Zhang C, Selinus O. Spatial analyses for copper, lead and zinc contents in sediments of the Yangtze River basin[J]. Science of the total environment.1997,204(3):251-262.
[38]张朝生,章申,何建邦.长江水系沉积物重金属含量空间分布特征研究——空间自相关与分形方法[J].地理学报.1998,53(1)：87-96.
[39]夏学齐,陈骏,廖启林,等.南京地区表土镉汞铅含量的空间统计分析[J].地球化学.2006,35(1)：95-102.
[40]梁二,王小彬,蔡典雄,等.河南省士壤有机碳分布空间自相关分析[J].应用生态学报.2007,18(6)：1305-1310.
[41]刘永生,李瑞敏.不同方法下土壤重金属元素空间最大变异距离研究——以浙江上虞为例[J].岩矿测试.2007,26(4)：331-335.
[42]Zhang C, Luo L, Xu W, et al. Use of local Moran's I and GIS to identify pollution hotspots of Pb in urban soils of Galway, Ireland[J]. Science of the Total Environment.2008,398(1): 212-221.
[43]霍霄妮,李红,孙丹峰,等.北京耕作土壤重金属含量的空间自相关分析[J].环境科学学报.2009,29(6)：1339-1344.
[44]李勇,周永章,张澄博,等.基于局部Moran's I和GIS的珠江三角洲肝癌高发区蔬菜土壤中Ni、Cr的空间热点分析[J].环境科学.2010,31(6)：1617-1623.
[45]杨奇勇,杨劲松,余世鹏,等.不同尺度下耕地土壤Cr含量的空间自相关性分析[J].应用与环境生物学报.2011,17(3)：393-397.
[46]刘庆,夏江宝,谢文军.半方差函数与Moran's I在土壤微量元素空间分布研究中的应用—以寿光市为例[J].武汉大学学报(信息科学版).2011,36(9)：1129-1133.
[47]刘永生.华北平原土壤重金属元素空间自相关研究[D].北京：中国地质大学(北京),2012.
[48]Barreto-Neto A A, Barros Da Silva A. Methodological criteria for auditing geochemical data sets aimed at selecting anomalous lead-zinc-silver areas using geographical information systems[J]. Journal of Geochemical Exploration.2004,84(2):93-101.
[49]Carranza E J M. Mapping of anomalies in continuous and discrete fields of stream sediment geochemical landscapes[J]. Geochemistry:Exploration, Environment, Analysis.2010,10(2): 171-187.
[50]Rose A W, Dahlberg E C, Keith M L. A multiple regression technique for adjusting background values in stream sediment geochemistry[J]. Economic Geology.1970,65(2):156-165.
[51]Hawkes H E. The downstream dilution of stream sediment anomalies[J]. Journal of Geochemical Exploration.1976,6(1):345-358.
[52]Ihaka R, Gentleman R. R:A language for data analysis and graphics[J]. Journal of computational and graphical statistics.1996,5(3):299-314.
[53]Hengl T. A practical guide to geostatistical mapping, University of Amsterdam [M]. Amsterdam, 2009.
[54]Grunsky E C. A program for computing RQ-mode principal components analysis for S-PLUS and R[J]. Computers & Geosciences.2001,27(2):229-235.
[55]Grunsky E C. R:a data analysis and statistical programming environment-an emerging tool for the geosciences[J]. Computers & Geosciences.2002,28(10):1219-1222.
[56]Reimann C, Filzmoser P, Garrett R, et al. Statistical data analysis explained:applied environmental statistics with R[M]. Wiley, Chichester, UK,2008:1-362.
[57]Grunsky E C. The interpretation of geochemical survey data[J]. Geochemistry:Exploration, Environment, Analysis.2010,10(1):27-74.
[58]Sarkar D. Lattice:multivariate data visualization with R[M]. Springer,2008.
[59]Garrett R G. The'rgr'package for the R Open Source statistical computing and graphics environment-a tool to support geochemical data interpretation[J]. Geochemistry:Exploration, Environment, Analysis.2013:2011-2106.
[60]郭科.复杂地质地貌区多尺度地球化学异常识别的非线性研究[D].成都：成都理工大学,2005.
[61]陈春香,胡光道,付水兴.基于遗传算法的石英滩金矿区域地球化学研究[J].地质与勘探.2006,42(6)：77-80.
[62]陈春香,姜兰芳.基于遗传算法的岩石地球化学元素关系评价[J].桂林工学院学报.2007,27(1)：36-39.
[63]陈聆.地球化学矿致异常非线性分析方法研究[D].成都：成都理工大学,2011.
[64]王明志,董方灵,梁群峰,等.模糊决策树技术在化探异常提取中的应用[J].中国地质.2011,38(3)：762-769.
[65]刘洪滔,陈三明,朱家群.地球化学信息挖掘在云南易门铜矿区找矿预测中的应用[J].桂林理工大学学报.2011,31(3)：328-338.
[66]韩天成,张振飞,吕新彪,等.东天山中段区域化探异常评价方法研究[J].地质与勘探.2011,47(5)：885-893.
[67]李晓晖,袁峰,贾蔡,等.基于多维分形模型与指示克里格方法的地球化学异常识别研究[J].地理与地理信息科学.2011,27(6)：23-27.
[68]刘江涛,成秋明,王建国.结构方程模型在地球化学数据分析中的应用[J].地球科学—中国地质大学学报.2012,37(6)：1191-1198.
[69]张冬梅,金辉,刘伟.基于Kriging的多重GEP演化建模趋势分析圈定区域化探异常研究[J].应用基础与工程科学学报.2012,20(3)：526-538.
[70]胡云中,任天祥,马振东,等.中国地球化学场及其与成矿关系[M].北京：地质出版社,2006：212.
[71]刘勇,周贤旭.基于MapGIS综合找矿信息香炉山钨矿田钨矿资源预测[J].东华理工大学学报(自然科学版).2010,33(3)：262-269.
[72]黄修保,裴荣富,梅燕雄,等.江西九瑞坳下—丁家山地质特征及找矿前景分析[J].地质与勘探.2011,47(4)：531-542.
[73]赵胜利,秦润君,曹其琛,等.九瑞区大城门—仙姑台斑岩型铜矿成矿地质条件[J].云南地质.2012,31(4)：518-522.
[74]王会敏,李永明,罗春林,等.江西九瑞地区铜多金属矿整装勘查区成矿规律与找矿靶区优选[J].资源调查与环境.2012,33(4)：245-253.
[75]项新葵,陈茂松,詹国年,等.赣北石门寺矿区钨多金属矿床成矿地质条件[J].地质找矿论丛.2012,27(2)：143-155.
[76]黄兰椿,蒋少涌.江西大湖塘钨矿床似斑状白云母花岗岩锆石U-Pb年代学、地球化学及成因研究[J].岩石学报.2012,28(12)：3887-3900.
[77]丰成友,张德全,项新葵,等.赣西北大湖塘钨矿床辉钼矿Re-Os同位素定年及其意义[J]. 岩石学报.2012,28(12)：3858-3868.
[78]左全狮.江西九岭山西段大湖塘—李扬斗成矿区成矿地质条件分析及进一步找矿前景评价[J].资源环境与工程.2006,20(4)：348-353.
[79]林黎,占岗乐,喻晓平.江西大湖塘钨(锡)矿田地质特征及远景分析[J].资源调查与环境.2006,27(1)：25-32.
[80]林黎,余忠珍,罗小洪,等.江西大湖塘钨矿田成矿预测[J].东华理工学院学报.2006(S1)：139-142.
[81]项新葵,汪石林,詹国年,等.石门寺“一区三型”钨铜钼矿床地质特征[Z].中国浙江杭州：浙江省国土资源厅、浙江省地质学会,2011.
[82]项新葵,刘显沐,詹国年.江西省大湖塘石门寺矿区超大型钨矿的发现及找矿意义[J].资源调查与环境.2012,33(3)：141-151.
[83]项新葵,王朋,孙德明,等.赣北石门寺钨多金属矿床同位素地球化学研究[J].地球学报.2013,34(3)：263-271.
[84]成秋明.空间模式的广义自相似性分析与矿产资源评价[J].地球科学——中国地质大学学报.2004,29(6)：733-743.
[85]李德仁.空间数据挖掘的理论与应用[M].北京市：科学出版社,2005：569.
[86]Pearson K. On a Form of Spurious Correlation Which May Arise When Indices Are Used in the Measurement of Organs[C]. London:Proceedings of the Royal Society of London,1897.
[87]Miesch A T. The constant sum problem in geochemistry[M]. Computer applications in the earth sciences, Springer,1969,161-176.
[88]Chayes F. On correlation between variables of constant sum[J]. Journal of Geophysical Research.1960,65(12):4185-4193.
[89]Pawlowsky-Glahn V, Egozcue J J. Compositional data and their analysis:an introduction[J]. Geological Society, London, Special Publications.2006,264(1):1-10.
[90]Krumbein W C, Watson G S. Effect of trends on correlation in open and closed three-component systems[J]. Journal of the International Association for Mathematical Geology. 1972,4(4):317-330.
[91]Butler J C. Occurrence of negative open variances in ternary systems[J]. Journal of the International Association for Mathematical Geology.1975,7(1):31-45.
[92]Butler J C. Principal components analysis using the hypothetical closed array[J]. Journal of the International Association for Mathematical Geology.1976,8(1):25-36.
[93]Skala W. A mathematical model to investigative distortions of correlation coefficients in closed arrays[J]. Journal of the International Association for Mathematical Geology.1977,9(5): 519-528.
[94]Skala W. Some effects of the constant-sum problem in geochemistry[J]. Chemical Geology. 1979,27(1):1-9.
[95]Aitchison J. The statistical analysis of compositional data[M]. Chapman & Hall, Ltd.,1986.
[96]Egozcue J J, Pawlowsky-Glahn V, Mateu-Figueras G, et al. Isometric logratio transformations for compositional data analysis[J]. Mathematical Geology.2003,35(3):279-300.
[97]Filzmoser P, Hron K, Reimann C. The bivariate statistical analysis of environmental (compositional) data[J]. Science of the Total Environment.2010,408(19):4230-4238.
[98]Filzmoser P, Hron K, Reimann C. Univariate statistical analysis of environmental (compositional) data:Problems and possibilities[J]. Science of the Total Environment.2009, 407(23):6100-6108.
[99]Templ M, Filzmoser P, Reimann C. Cluster analysis applied to regional geochemical data: Problems and possibilities[J]. Applied Geochemistry.2008,23(8):2198-2213.
[100]Buccianti A, Mateu-Figueras G, Glahn V P. Compositional data analysis in the geosciences: from theory to practice[M]. Geological Society,2006.
[101]Howarth R J, Garrett R G. Statistical analysis and data display at the Geochemical Prospecting Research Centre and Applied Geochemistry Research Group, Imperial College, London[J]. Geochemistry:Exploration, Environment, Analysis.2010,10(3):289-315.
[102]Carranza E J M. Analysis and mapping of geochemical anomalies using logratio-transformed stream sediment data with censored values[J]. Journal of Geochemical Exploration.2011,110(2): 167-185.
[103]Zuo R, Xia Q, Wang H. Compositional data analysis in the study of integrated geochemical anomalies associated with mineralization[J]. Applied Geochemistry.2012,28:202-211.
[104]Van Den Boogaart K G, Tolosana-Delgado R. Analyzing Compositional Data with R[M]. Springer,2013.
[105]周蒂.地质成分数据统计分析——困难和探索[J].地球科学.1998,23(2)：147-152.
[106]Ahrens L H. A fundamental law of geochemistry[J]. Nature.1953,172:1148.
[107]Ahrens L H. The lognormal distribution of the elements (2)[J]. Geochimica et cosmochimica acta.1954,6(2):121-131.
[108]Ahrens L H. The lognormal distribution of the elements (a fundamental law of geochemistry and its subsidiary)[J]. Geochimica et Cosmochimica Acta.1954,5(2):49-73.
[109]Miller R L, Goldberg E D. The normal distribution in geochemistry[J]. Geochimica et Cosmochimica Acta.1955,8(1):53-62.
[110]Aubrey K V. Frequency-distributions of elements in igneous rocks[J]. Geochimica et Cosmochimica Acta.1956,9(1):83-89.
[111]Ahrens L H. Lognormal-type distributions-Ⅲ[J]. Geochimica et Cosmochimica Acta.1957, 11(4):205-212.
[112]Jizba Z V. Frequency distribution of elements in rocks[J]. Geochimica et Cosmochimica Acta. 1959,16(1):79-82.
[113]Vistelius A B. The skew frequency distributions and the fundamental law of the geochemical processes[J]. The Journal of Geology.1960,68(1):1-22.
[114]Shaw D M. Element distribution laws in geochemistry[J]. Geochimica et Cosmochimica Acta. 1961,23(1):116-134.
[115]Ahrens L H. Element distributions in igneous rocks-Ⅵ. Negative skewness of SiO2 and K[J]. Geochimica et Cosmochimica Acta.1963,27(9):929-938.
[116]Ahrens L H. Lognormal-type distributions in igneous rocks-Ⅴ[J]. Geochimica et Cosmochimica Acta.1963,27(8):877-890.
[117]Ahrens L H. Lognormal-type distributions in igneous rocks-Ⅳ[J]. Geochimica et Cosmochimica Acta.1963,27(4):333-343.
[118]Ahrens L H. Element distributions in igneous rocks-Ⅶ:A reconnaissance survey of the distribution of SiO2 in granitic and basaltic rocks[J]. Geochimica et Cosmochimica Acta.1964, 28(3):271-290.
[119]Ahrens L H. Element distributions in specific igneous rocks-Ⅷ[J]. Geochimica et Cosmochimica Acta.1966,30(1):109-122.
[120]成秋明.多维分形理论和地球化学元素分布规律[J].地球科学——中国地质大学学报.2000,25(3)：311-318.
[121]Cheng Q, Agterberg F P, Ballantyne S B. The separation of geochemical anomalies from background by fractal methods[J]. Journal of Geochemical Exploration.1994,51(2):109-130.
[122]Allegre C J, Lewin E. Scaling laws and geochemical distributions[J]. Earth and Planetary Science Letters.1995,132(1):1-13.
[123]Friedman J H, Tukey J W. A projection pursuit algorithm for exploratory data analysis[J]. Computers, IEEE Transactions on.1974,100(9):881-890.
[124]Tukey J W. Exploratory data analysis[M]. Philippines:Addison-Wesley Publishing Company,Inc.,1977:1-688.
[125]Haining R, Wise S. Exploratory Spatial Data Analysis, NCGIA Core Curriculum in GIScience[Z].1997:2013.
[126]Haining R, Wise S, Ma J. Exploratory spatial data analysis in a geographic information system environment[J]. Journal of the Royal Statistical Society:Series D (The Statistician).1998,47(3): 457-469.
[127]Kiirzl H. Exploratory data analysis:recent advances for the interpretation of geochemical data[J]. Journal of Geochemical Exploration.1988,30(1):309-322.
[128]Howarth R J. Statistical applications in geochemical prospecting:a survey of recent developments[J]. Journal of Geochemical Exploration.1984,21(1):41-61.
[129]Venables W N, Ripley B D. Modern Applied Statistics with S[M]. Springer,2002:1-495.
[130]Tennant C B, White M L. Study of the distribution of some geochemical data[J]. Economic Geology.1959,54(7):1281-1290.
[131]Sinclair A J. Selection of threshold values in geochemical data using probability graphs[J]. Journal of Geochemical Exploration.1974,3(2):129-149.
[132]Sinclair A J. A fundamental approach to threshold estimation in exploration geochemistry: probability plots revisited[J]. Journal of Geochemical Exploration.1991,41(1):1-22.
[133]李随民,姚书振,韩玉丑Surfer软件中利用趋势面方法圈定化探异常[J].地质与勘探.2007,43(2)：72-75.
[134]陈希清,杨晓君,陈富文,等.应用MAPGIS数字高程模型提取区域地球化学异常信息的方法探讨[J].地球学报.2009,30(1)：119-125.
[135]石文杰,魏俊浩,张德才,等.基于数字高程模型因子分析的地球化学异常提取[J].物探与化探.2012,36(1)：103-108.
[136]Matschullat J, Ottenstein R, Reimann C. Geochemical background-can we calculate it?[J]. Environmental geology.2000,39(9):990-1000.
[137]Reimann C, Garrett R G. Geochemical background-concept and reality[J]. Science of the total environment.2005,350(1):12-27.
[138]Reimann C, Filzmoser P, Garrett R G. Background and threshold:critical comparison of methods of determination[J]. Science of the Total Environment.2005,346(1):1-16.
[139]Mrvic V, Kostic-Kravljanac L, cakmak D, et al. Pedogeochemical mapping and background limit of trace elements in soils of Branicevo Province (Serbia)[J]. Journal of Geochemical Exploration.2011,109(1):18-25.
[140]Mrvic V V, Kostic-Kravljanac L M, Zdravkovic M S, et al. Background limit of Zn and Hg in soils of Eastern Serbia[J]. Journal of Agricultural Sciences, Belgrade.2010,55(2):157-163.
[141]Devesa-Rey R, Diaz-Fierros F, Barral M T. Normalization strategies for river bed sediments:A graphical approach[J]. Microchemical Journal.2009,91(2):253-265.
[142]Sanei H, Goodarzi F, Hilts S. Site-specific natural background concentrations of metals in topsoil from the Trail region, British Columbia, Canada[J]. Geochemistry:Exploration, environment, analysis.2007,7(1):41-47.
[143]Diez M, Simon M, Dorronsoro C, et al. Background arsenic concentrations in Southeastern Spanish soils[J]. Science of the total environment.2007,378(1):5-12.
[144]Escalante A, Dipple G M, Barker S, et al. Defining trace-element alteration halos to skarn deposits hosted in heterogeneous carbonate rocks:Case study from the Cu-Zn Antamina skarn deposit, Peru[J]. Journal of Geochemical Exploration.2010,105(3):117-136.
[145]Mandelbrot B. How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension[J]. Science.1967,156(3775):636-638.
[146]Mandelbrot B B. The fractal geometry of nature[M]. Macmillan,1983.
[147]Bolviken B, Stokke P R, Feder J, et al. The fractal nature of geochemical landscapes[J]. Journal of Geochemical Exploration.1992,43(2):91-109.
[148]Cheng Q. The perimeter-area fractal model and its application to geology[J]. Mathematical Geology.1995,27(1):69-82.
[149]Cheng Q. Selection of Multifractal Scaling Breaks and Separation of Geochemical and Geophysical Anomaly[J]. Journal of China University of Geosciences.2001,12(1):54-59.
[150]Xu Y, Cheng Q. A fractal filtering technique for processing regional geochemical maps for mineral exploration[J]. Geochemistry:Exploration, environment, analysis.2001,1(2):147-156.
[151]Goncalves M A, Mateus A, Oliveira V. Geochemical anomaly separation by multifractal modelling[J]. Journal of Geochemical Exploration.2001,72(2):91-114.
[152]Zheru Z, Huahai M, Cheng Q. Fractal geometry of element distribution on mineral surfaces[J]. Mathematical geology.2001,33(2):217-228.
[153]Cheng Q, Russell H, Sharpe D, et al. GIS-based statistical and fractal/multifractal analysis of surface stream patterns in the Oak Ridges Moraine[J]. Computers & geosciences.2001,27(5): 513-526.
[154]Panahi A, Cheng Q. Multifractality as a measure of spatial distribution of geochemical patterns[J]. Mathematical geology.2004,36(7):827-846.
[155]Ge Y, Cheng Q, Zhang S. Reduction of edge effects in spatial information extraction from regional geochemical data:a case study based on multifractal filtering technique[J]. Computers & geosciences.2005,31(5):545-554.
[156]Ali K, Cheng Q, Chen Z. Multifractal power spectrum and singularity analysis for modelling stream sediment geochemical distribution patterns to identify anomalies related to gold mineralization in Yunnan Province, South China[J]. Geochemistry:Exploration, Environment, Analysis.2007,7(4):293-301.
[157]Xie S, Cheng Q, Ke X, et al. Identification of geochemical anomaly by multifractal analysis[J]. Journal of China University of Geosciences.2008,19(4):334-342.
[158]Xie S, Cheng Q, Xing X, et al. Geochemical multifractal distribution patterns in sediments from ordered streams[J]. Geoderma.2010,160(1):36-46.
[159]Zuo R. Decomposing of mixed pattern of arsenic using fractal model in Gangdese belt, Tibet, China[J]. Applied Geochemistry.2011,26:S271-S273.
[160]Zuo R. Identifying geochemical anomalies associated with Cu and Pb-Zn skarn mineralization using principal component analysis and spectrum-area fractal modeling in the Gangdese Belt, Tibet (China)[J]. Journal of Geochemical Exploration.2011,111(1):13-22.
[161]谢淑云,鲍征宇.地球化学场的连续多重分形模式[J].地球化学.2002,31(2)：191-200.
[162]谢淑云.地球化学场的分形与多重分形特征[D].武汉：中国地质大学(武汉),2003.
[163]Li C, Ma T, Shi J. Application of a fractal method relating concentrations and distances for separation of geochemical anomalies from background[J]. Journal of Geochemical Exploration. 2003,77(2):167-175.
[164]Agterberg F P, Cheng Q. Introduction to special issue on "Fractals and Multi fractals" [J]. Computers & Geosciences.1999,25(9):947-948.
[165]Zuo R, Carranza E J M, Cheng Q. Fractal/multifractal modelling of geochemical exploration data[J]. Journal of Geochemical Exploration.2012,122:1-3.
[166]陈建国,王仁铎,陈永清.利用分形统计学提取化探数据中的隐蔽信息并圈定地球化学异常[J].地球科学.1998,23(2)：175-178.
[167]成秋明.多重分形与地质统计学方法用于勘查地球化学异常空间结构和奇异性分析[J].地球科学——中国地质大学学报.2001,26(2)：161-166.
[168]李庆谋,成秋明.分形奇异(特征)值分解方法与地球物理和地球化学异常重建[J].地球科学——中国地质大学学报.2004,29(1)：109-118.
[169]李堃,胡光道,刘才泽.个旧地区地球化学元素的多重分形特征及找矿预测[J].矿产与地质.2006,20(4-5)：498-502.
[170]孙忠军.矿产勘查中化探异常下限的多重分形计算方法[J].物探化探计算技术.2007,29(1)：54.
[171]成秋明,张生元,左仁广,等.多重分形滤波方法和地球化学信息提取技术研究与进展[J].地学前缘.2009,16(2)：185-198.
[172]徐明钻,朱立新,马生明,等.多重分形模型在区域地球化学异常分析中的应用探讨[J].地球学报.2010,31(4)：611-618.
[173]娄德波,肖克炎,左仁广,等.分形滤波技术在新疆黄山-镜儿泉镍铜成矿带中的应用[J].地球学报.2012,33(1)：83-90.
[174]迟清华,鄢明才.应用地球化学元素丰度数据手册[M].北京：地质出版社,2007：148.
[175]Garrett R G. The chi-square plot:A tool for multivariate outlier recognition[J]. Journal of Geochemical Exploration.1989,32(1):319-341.
[176]Filzmoser P, Garrett R G, Reimann C. Multivariate outlier detection in exploration geochemistry[J]. Computers & Geosciences.2005,31(5):579-587.
[177]Filzmoser P, Hron K, Reimann C. Interpretation of multivariate outliers for compositional data[J]. Computers & Geosciences.2012,39:77-85.
[178]Kruskal J B. Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis[J]. Psychometrika.1964,29(1):1-27.
[179]Fry N. Random point distributions and strain measurement in rocks[J]. Tectonophysics.1979, 60(1):89-105.
[180]Vearncombe J, Vearncombe S. The spatial distribution of mineralization:applications of Fry analysis[J]. Economic Geology.1999,94(4):475-486.
[181]Cleveland W S. Visualizing data[M]. Hobart Press,1993.
[182]Reimann C, Filzmoser P, Garrett R G. Factor analysis applied to regional geochemical data: problems and possibilities[J]. Applied Geochemistry.2002,17(3):185-206.
[183]Johnson R A, Wichern D W. Applied multivariate statistical analysis[M]. Prentice hall Upper Saddle River, NJ,2002.
[184]de Caritat P, Grunsky E C. Defining element associations and inferring geological processes from total element concentrations in Australian catchment outlet sediments:Multivariate analysis of continental-scale geochemical data[J]. Applied Geochemistry.2013,33:104-126.
[185]孟宪伟,窦明晓,余先川.地球化学场分解的理论与方法[J].地球科学进展.1994,9(6)：59-64.
[186]时艳香,纪宏金,陆继龙,等.水系沉积物地球化学分区的因子分析方法与应用[J].地质与勘探.2004,40(5)：73-76.
[187]王永华,龚鹏,龚敏,等.成矿带1：20万水系沉积物地球化学分区的方法及地质意义：以西藏冈底斯铜多金属成矿带为例[J].现代地质.2010,24(4)：801-806.
[188]van Helvoort P, Filzmoser P, van Gaans P F. Sequential factor analysis as a new approach to multivariate analysis of heterogeneous geochemical datasets:an application to a bulk chemical characterization of fluvial deposits (Rhine-Meuse delta. The Netherlands)[J], Applied geochemistry.2005,20(12):2233-2251.
[189]Jackson J E. A user's guide to principal components[M]. Wiley, com,2005.
[190]Kaiser H F. The varimax criterion for analytic rotation in factor analysis[J]. Psychometrika. 1958,23(3):187-200.
[191]Bartlett M S. Tests of significance in factor analysis[J]. British Journal of Statistical Psychology. 1950,3(2):77-85.
[192]Efron B. Bootstrap methods:another look at the jackknife[J]. The annals of Statistics.1979:
[193]Eastment H T, Krzanowski W J. Cross-validatory choice of the number of components from a principal component analysis[J]. Technometrics.1982,24(1):73-77.
[194]Basilevsky A T. Statistical factor analysis and related methods:theory and applications[M]. John Wiley & Sons,2009.
[195]Gabriel K R. The biplot graphic display of matrices with application to principal component analysis[J]. Biometrika.1971,58(3):453-467.
[196]Cook D, Swayne D F. Interactive and dynamic graphics for data analysis:with R and GGobi[M]. Springer,2007.
[197]Grunsky E C, Smee B W. The differentiation of soil types and mineralization from multi-element geochemistry using multivariate methods and digital topography[J]. Journal of Geochemical Exploration.1999,67(1):287-299.
[198]Cheng Q, Bonham-Carter G F, Hall G, et al. Statistical study of trace elements in the soluble organic and amorphous Fe-Mn phases of surficial sediments, Sudbury Basin.1. Multivariate and spatial analysis[J]. Journal of Geochemical Exploration.1997,59(1):27-46.
[199]余先川,王世称,王桂安.稳健协同克立格因子分析及其在化探中的应用[J].地球科学——中国地质大学学报.1998,23(2)：171-174.
[200]孟宪伟.地球化学变量的空间自相关与互相关及空间因子分析[J].物探化探计算技术.1993,15(3)：192-203.
[201]Gordon A D. Classification[M]. Chapman&Hall, CRC,1999:256.
[202]Hubert L, Arabie P. Comparing partitions[J]. Journal of classification.1985,2(1):193-218.
[203]Halkidi M, Batistakis Y, Vazirgiannis M. Cluster validity methods:part I[J]. ACM Sigmod Record.2002,31(2):40-45.
[204]Milligan G W, Cooper M C. An examination of procedures for determining the number of clusters in a data set[J]. Psychometrika.1985,50(2):159-179.
[205]Calinski T, Harabasz J. A dendrite method for cluster analysis[J]. Communications in Statistics-theory and Methods.1974,3(1):1-27.
[206]Hartigan J A. Clustering algorithms[M]. John Wiley & Sons, Inc.,1975.
[207]Kaufman L, Rousseeuw P J. Finding groups in data:an introduction to cluster analysis[M]. John Wiley & Sons,2009.
[208]Chork C Y, Rousseeuw P J. Integrating a high-breakdown option into discriminant analysis in exploration geochemistry[J]. Journal of Geochemical Exploration.1992,43(3):191-203.
[209]Gnanadesikan R. Methods for statistical data analysis of multivariate observations[M]. John Wiley & Sons,2011.
[210]Howarth R J. Statistics and data analysis in geochemical prospecting[M]. Elsevier scientific publishing company,1983.
[211]Rousseeuw P J, Van Driessen K. Computing LTS regression for large data sets[J]. Data Mining and Knowledge Discovery.2006,12(1):29-45.
[212]Rousseeuw P J. Least median of squares regression[J|. Journal of the American statistical association.1984,79(388):871-880.
[213]Hoerl A E, Kennard R W. Ridge regression:Biased estimation for nonorthogonal problems[J]. Technometrics.1970,12(1):55-67.
[214]Hastie T, Tibshirani R, Friedman J J H. The elements of statistical learning[M]. Springer New York,2001.
[215]Jolliffe I. Principal component analysis[M]. Wiley Online Library,2002.
[216]王斌会.多元统计分析及R语言建模第2版[M].广州市：暨南大学出版社,2011：321.
[217]龚鹏,胡小梅,李娟,等.建立地质-地球化学找矿模型——以西藏甲玛铜多金属矿床为例[J].地质通报.2013,32(10)：1601-1612.
[218]龚鹏,马振东.矿产预测中区域化探异常的识别和评价[J].地球科学(中国地质大学学报).2013,38(S1)：113-125.
[219]Garrett R G, Grunsky E C. Weighted sums-knowledge based empirical indices for use in exploration geochemistry[J]. Geochemistry:Exploration, environment, analysis.2001,1(2): 135-141.
[220]Johnson J W. A heuristic method for estimating the relative weight of predictor variables in multiple regression[J]. Multivariate Behavioral Research.2000,35(1):1-19.
[221]Lance G N, Williams W T. Computer programs for hierarchical polythetic classification ("similarity analyses")[J]. The Computer Journal.1966,9(1):60-64.
[222]Lance G N, Williams W T. Mixed-Data Classificatory Programs I-Agglomerative Systems[J]. Australian Computer Journal.1967,1(1):15-20.
[223]向运川,任天祥,牟绪赞等.化探资料应用技术要求[M].北京市：地质出版社,2010：82.
[2]王学求.勘查地球化学近十年进展[J].矿物岩石地球化学通报.2013,32(2)：190-197.
[3]奚小环,李敏.现代地质工作重要发展领域：“十一五”期间勘查地球化学评述[J].地学前缘.2013,20(3)：161-169.
[4]Cohen D R, Kelley D L, Anand R, et al. Major advances in exploration geochemistry,1998-2007[J]. Geochemistry:Exploration, Environment, Analysis.2010,10(1):3-16.
[5]Carranza E J M. Geochemical anomaly and mineral prospectivity mapping in GIS[M]. Elsevier,
[6]Howarth R J. Statistics and data analysis in geochemical prospecting[M]. Elsevier scientific publishing company,1983.
[7]Fletcher W K. Analytical methods in geochemical prospecting[M]. Elsevier Scientific Publishing Company,1981.
[8]Kauranne L K, Salimnen R, Eriksson K. Regolith exploration geochemistry in arctic and temperate terrains.[M]. Elsevier Science Publishers,1992.
[9]Butt C, Zeegers H. Regolith exploration geochemistry in tropical and subtropical terrains[M]. Elsevier Amsterdam and London,1992.
[10]Hale M. Geochemical remote sensing of the sub-surface[M]. Access Online via Elsevier,2000.
[11]Hein J R. Life cycle of the Phosphoria Formation:from deposition to the post-mining environment[M]. Gulf Professional Publishing,2003.
[12]Govett G J S. Rock geochemistry in mineral exploration[M]. New York.:Elsevier Scientific Pub. Co. Amsterdam,1983:461.
[13]Hale M, Plant J A. Drainage geochemistry. Handbook of exploration geochemistry[M]. Elsevier, Amsterdam,1994.
[14]Mukhopadhyay R, Ghosh A K, Iyer S D. The Indian Ocean nodule field:Geology and resource potential[M]. Access Online via Elsevier,2007.
[15]Dunn C E. Biogeochemistry in Mineral Exploration, Handbook of exploration and Environmental Geochemistry, V.9[M]. Elsevier,2007.
[16]Xie X, Mu X, Ren T. Geochemical mapping in China[J]. Journal of Geochemical Exploration. 1997,60(1):99-113.
[17]奚小环,李敏.中国区域化探若干基本问题研究：1999—2009[J].中国地质.2012,39(2)：267-282.
[18]黄芳.水土污染空间分析及源辨析[D].杭州：浙江大学,2010.
[19]Krige D G. A Statistical Approach to Some Mine Valuation and Allied Problems on the Witwatersrand[D]. University of the Witwatersrand,1951.
[20]Krige D G. A statistical analysis of some of the borehole values in the Orange Free State goldfield[J]. Journal of the Chemical and Metallurgical Society of South Africa.1952,53:47-64.
[21]Matheron G. Principles of geostatistics[J]. Economic geology.1963,58(8):1246-1266.
[22]肖斌,侯景儒.归来庄金矿床综合化探金异常的地质统计学研究[J].北京大学学报：自然科学版.2000,36(4)：531-538.
[23]黄静宁,赵鹏大.东天山地区Cu-Au异常信息提取与评价[J].地球科学：中国地质大学学报.2011,36(2)：317-326.
[24]叶水盛,刘光胜.克立格方法在区域化探数据处理中的应用[J].长春科技大学学报.2000,30(2)：198-200.
[25]陈建国,王仁铎,陈永清.利用分形统计学提取化探数据中的隐蔽信息并圈定地球化学异常[J].地球科学：中国地质大学学报.1998,23(2)：175-178.
[26]胡光道,王仁铎,李正琪.对数泛克立金法在处理化探数据中的应用[J].地球科学.1985, 10(2)：141-150.
[27]Panahi A, Cheng Q, Bonham-Carter G F. Modelling lake sediment geochemical distribution using principal component, indicator kriging and multifractal power-spectrum analysis:a case study from Gowganda, Ontario[J]. Geochemistry:Exploration, Environment, Analysis.2004, 4(1):59-70.
[28]Moran P A. Notes on continuous stochastic phenomena[J]. Biometrika.1950,37(1/2):17-23.
[29]Cliff A D, Ord J K. Spatial processes:models & applications[M]. Pion London,1981.
[30]Cressie N A. Statistics for Spatial Data[M]. Wiley Series in Probability and Mathematical Statistics:Applied Probability and Statistics. John Wiley & Sons,1991.
[31]Cressie N A. Statistics for Spatial Data, revised edition[M]. Wiley, New York,1993.
[32]Anselin L. Local indicators of spatial association-LISA[J]. Geographical analysis.1995,27(2):
[33]Fischer M M, Wang J. Spatial data analysis:models, methods and techniques[M]. Springer,
[34]Haining R P. Spatial data analysis:theory and practice[M]. Cambridge University Press,2003.
[35]Haining R P, Kerry R, Oliver M A. Geography, spatial data analysis, and geostatistics:an overview[J]. Geographical Analysis.2010,42(1):7-31.
[36]张朝生,陶澍,袁贵平,等.天津市平原土壤微量元素含量的空间自相关研究[J].土壤学报.1995,32(1)：50-57.
[37]Zhang C, Selinus O. Spatial analyses for copper, lead and zinc contents in sediments of the Yangtze River basin[J]. Science of the total environment.1997,204(3):251-262.
[38]张朝生,章申,何建邦.长江水系沉积物重金属含量空间分布特征研究——空间自相关与分形方法[J].地理学报.1998,53(1)：87-96.
[39]夏学齐,陈骏,廖启林,等.南京地区表土镉汞铅含量的空间统计分析[J].地球化学.2006,35(1)：95-102.
[40]梁二,王小彬,蔡典雄,等.河南省士壤有机碳分布空间自相关分析[J].应用生态学报.2007,18(6)：1305-1310.
[41]刘永生,李瑞敏.不同方法下土壤重金属元素空间最大变异距离研究——以浙江上虞为例[J].岩矿测试.2007,26(4)：331-335.
[42]Zhang C, Luo L, Xu W, et al. Use of local Moran's I and GIS to identify pollution hotspots of Pb in urban soils of Galway, Ireland[J]. Science of the Total Environment.2008,398(1): 212-221.
[43]霍霄妮,李红,孙丹峰,等.北京耕作土壤重金属含量的空间自相关分析[J].环境科学学报.2009,29(6)：1339-1344.
[44]李勇,周永章,张澄博,等.基于局部Moran's I和GIS的珠江三角洲肝癌高发区蔬菜土壤中Ni、Cr的空间热点分析[J].环境科学.2010,31(6)：1617-1623.
[45]杨奇勇,杨劲松,余世鹏,等.不同尺度下耕地土壤Cr含量的空间自相关性分析[J].应用与环境生物学报.2011,17(3)：393-397.
[46]刘庆,夏江宝,谢文军.半方差函数与Moran's I在土壤微量元素空间分布研究中的应用—以寿光市为例[J].武汉大学学报(信息科学版).2011,36(9)：1129-1133.
[47]刘永生.华北平原土壤重金属元素空间自相关研究[D].北京：中国地质大学(北京),2012.
[48]Barreto-Neto A A, Barros Da Silva A. Methodological criteria for auditing geochemical data sets aimed at selecting anomalous lead-zinc-silver areas using geographical information systems[J]. Journal of Geochemical Exploration.2004,84(2):93-101.
[49]Carranza E J M. Mapping of anomalies in continuous and discrete fields of stream sediment geochemical landscapes[J]. Geochemistry:Exploration, Environment, Analysis.2010,10(2): 171-187.
[50]Rose A W, Dahlberg E C, Keith M L. A multiple regression technique for adjusting background values in stream sediment geochemistry[J]. Economic Geology.1970,65(2):156-165.
[51]Hawkes H E. The downstream dilution of stream sediment anomalies[J]. Journal of Geochemical Exploration.1976,6(1):345-358.
[52]Ihaka R, Gentleman R. R:A language for data analysis and graphics[J]. Journal of computational and graphical statistics.1996,5(3):299-314.
[53]Hengl T. A practical guide to geostatistical mapping, University of Amsterdam [M]. Amsterdam, 2009.
[54]Grunsky E C. A program for computing RQ-mode principal components analysis for S-PLUS and R[J]. Computers & Geosciences.2001,27(2):229-235.
[55]Grunsky E C. R:a data analysis and statistical programming environment-an emerging tool for the geosciences[J]. Computers & Geosciences.2002,28(10):1219-1222.
[56]Reimann C, Filzmoser P, Garrett R, et al. Statistical data analysis explained:applied environmental statistics with R[M]. Wiley, Chichester, UK,2008:1-362.
[57]Grunsky E C. The interpretation of geochemical survey data[J]. Geochemistry:Exploration, Environment, Analysis.2010,10(1):27-74.
[58]Sarkar D. Lattice:multivariate data visualization with R[M]. Springer,2008.
[59]Garrett R G. The'rgr'package for the R Open Source statistical computing and graphics environment-a tool to support geochemical data interpretation[J]. Geochemistry:Exploration, Environment, Analysis.2013:2011-2106.
[60]郭科.复杂地质地貌区多尺度地球化学异常识别的非线性研究[D].成都：成都理工大学,2005.
[61]陈春香,胡光道,付水兴.基于遗传算法的石英滩金矿区域地球化学研究[J].地质与勘探.2006,42(6)：77-80.
[62]陈春香,姜兰芳.基于遗传算法的岩石地球化学元素关系评价[J].桂林工学院学报.2007,27(1)：36-39.
[63]陈聆.地球化学矿致异常非线性分析方法研究[D].成都：成都理工大学,2011.
[64]王明志,董方灵,梁群峰,等.模糊决策树技术在化探异常提取中的应用[J].中国地质.2011,38(3)：762-769.
[65]刘洪滔,陈三明,朱家群.地球化学信息挖掘在云南易门铜矿区找矿预测中的应用[J].桂林理工大学学报.2011,31(3)：328-338.
[66]韩天成,张振飞,吕新彪,等.东天山中段区域化探异常评价方法研究[J].地质与勘探.2011,47(5)：885-893.
[67]李晓晖,袁峰,贾蔡,等.基于多维分形模型与指示克里格方法的地球化学异常识别研究[J].地理与地理信息科学.2011,27(6)：23-27.
[68]刘江涛,成秋明,王建国.结构方程模型在地球化学数据分析中的应用[J].地球科学—中国地质大学学报.2012,37(6)：1191-1198.
[69]张冬梅,金辉,刘伟.基于Kriging的多重GEP演化建模趋势分析圈定区域化探异常研究[J].应用基础与工程科学学报.2012,20(3)：526-538.
[70]胡云中,任天祥,马振东,等.中国地球化学场及其与成矿关系[M].北京：地质出版社,2006：212.
[71]刘勇,周贤旭.基于MapGIS综合找矿信息香炉山钨矿田钨矿资源预测[J].东华理工大学学报(自然科学版).2010,33(3)：262-269.
[72]黄修保,裴荣富,梅燕雄,等.江西九瑞坳下—丁家山地质特征及找矿前景分析[J].地质与勘探.2011,47(4)：531-542.
[73]赵胜利,秦润君,曹其琛,等.九瑞区大城门—仙姑台斑岩型铜矿成矿地质条件[J].云南地质.2012,31(4)：518-522.
[74]王会敏,李永明,罗春林,等.江西九瑞地区铜多金属矿整装勘查区成矿规律与找矿靶区优选[J].资源调查与环境.2012,33(4)：245-253.
[75]项新葵,陈茂松,詹国年,等.赣北石门寺矿区钨多金属矿床成矿地质条件[J].地质找矿论丛.2012,27(2)：143-155.
[76]黄兰椿,蒋少涌.江西大湖塘钨矿床似斑状白云母花岗岩锆石U-Pb年代学、地球化学及成因研究[J].岩石学报.2012,28(12)：3887-3900.
[77]丰成友,张德全,项新葵,等.赣西北大湖塘钨矿床辉钼矿Re-Os同位素定年及其意义[J]. 岩石学报.2012,28(12)：3858-3868.
[78]左全狮.江西九岭山西段大湖塘—李扬斗成矿区成矿地质条件分析及进一步找矿前景评价[J].资源环境与工程.2006,20(4)：348-353.
[79]林黎,占岗乐,喻晓平.江西大湖塘钨(锡)矿田地质特征及远景分析[J].资源调查与环境.2006,27(1)：25-32.
[80]林黎,余忠珍,罗小洪,等.江西大湖塘钨矿田成矿预测[J].东华理工学院学报.2006(S1)：139-142.
[81]项新葵,汪石林,詹国年,等.石门寺“一区三型”钨铜钼矿床地质特征[Z].中国浙江杭州：浙江省国土资源厅、浙江省地质学会,2011.
[82]项新葵,刘显沐,詹国年.江西省大湖塘石门寺矿区超大型钨矿的发现及找矿意义[J].资源调查与环境.2012,33(3)：141-151.
[83]项新葵,王朋,孙德明,等.赣北石门寺钨多金属矿床同位素地球化学研究[J].地球学报.2013,34(3)：263-271.
[84]成秋明.空间模式的广义自相似性分析与矿产资源评价[J].地球科学——中国地质大学学报.2004,29(6)：733-743.
[85]李德仁.空间数据挖掘的理论与应用[M].北京市：科学出版社,2005：569.
[86]Pearson K. On a Form of Spurious Correlation Which May Arise When Indices Are Used in the Measurement of Organs[C]. London:Proceedings of the Royal Society of London,1897.
[87]Miesch A T. The constant sum problem in geochemistry[M]. Computer applications in the earth sciences, Springer,1969,161-176.
[88]Chayes F. On correlation between variables of constant sum[J]. Journal of Geophysical Research.1960,65(12):4185-4193.
[89]Pawlowsky-Glahn V, Egozcue J J. Compositional data and their analysis:an introduction[J]. Geological Society, London, Special Publications.2006,264(1):1-10.
[90]Krumbein W C, Watson G S. Effect of trends on correlation in open and closed three-component systems[J]. Journal of the International Association for Mathematical Geology. 1972,4(4):317-330.
[91]Butler J C. Occurrence of negative open variances in ternary systems[J]. Journal of the International Association for Mathematical Geology.1975,7(1):31-45.
[92]Butler J C. Principal components analysis using the hypothetical closed array[J]. Journal of the International Association for Mathematical Geology.1976,8(1):25-36.
[93]Skala W. A mathematical model to investigative distortions of correlation coefficients in closed arrays[J]. Journal of the International Association for Mathematical Geology.1977,9(5): 519-528.
[94]Skala W. Some effects of the constant-sum problem in geochemistry[J]. Chemical Geology. 1979,27(1):1-9.
[95]Aitchison J. The statistical analysis of compositional data[M]. Chapman & Hall, Ltd.,1986.
[96]Egozcue J J, Pawlowsky-Glahn V, Mateu-Figueras G, et al. Isometric logratio transformations for compositional data analysis[J]. Mathematical Geology.2003,35(3):279-300.
[97]Filzmoser P, Hron K, Reimann C. The bivariate statistical analysis of environmental (compositional) data[J]. Science of the Total Environment.2010,408(19):4230-4238.
[98]Filzmoser P, Hron K, Reimann C. Univariate statistical analysis of environmental (compositional) data:Problems and possibilities[J]. Science of the Total Environment.2009, 407(23):6100-6108.
[99]Templ M, Filzmoser P, Reimann C. Cluster analysis applied to regional geochemical data: Problems and possibilities[J]. Applied Geochemistry.2008,23(8):2198-2213.
[100]Buccianti A, Mateu-Figueras G, Glahn V P. Compositional data analysis in the geosciences: from theory to practice[M]. Geological Society,2006.
[101]Howarth R J, Garrett R G. Statistical analysis and data display at the Geochemical Prospecting Research Centre and Applied Geochemistry Research Group, Imperial College, London[J]. Geochemistry:Exploration, Environment, Analysis.2010,10(3):289-315.
[102]Carranza E J M. Analysis and mapping of geochemical anomalies using logratio-transformed stream sediment data with censored values[J]. Journal of Geochemical Exploration.2011,110(2): 167-185.
[103]Zuo R, Xia Q, Wang H. Compositional data analysis in the study of integrated geochemical anomalies associated with mineralization[J]. Applied Geochemistry.2012,28:202-211.
[104]Van Den Boogaart K G, Tolosana-Delgado R. Analyzing Compositional Data with R[M]. Springer,2013.
[105]周蒂.地质成分数据统计分析——困难和探索[J].地球科学.1998,23(2)：147-152.
[106]Ahrens L H. A fundamental law of geochemistry[J]. Nature.1953,172:1148.
[107]Ahrens L H. The lognormal distribution of the elements (2)[J]. Geochimica et cosmochimica acta.1954,6(2):121-131.
[108]Ahrens L H. The lognormal distribution of the elements (a fundamental law of geochemistry and its subsidiary)[J]. Geochimica et Cosmochimica Acta.1954,5(2):49-73.
[109]Miller R L, Goldberg E D. The normal distribution in geochemistry[J]. Geochimica et Cosmochimica Acta.1955,8(1):53-62.
[110]Aubrey K V. Frequency-distributions of elements in igneous rocks[J]. Geochimica et Cosmochimica Acta.1956,9(1):83-89.
[111]Ahrens L H. Lognormal-type distributions-Ⅲ[J]. Geochimica et Cosmochimica Acta.1957, 11(4):205-212.
[112]Jizba Z V. Frequency distribution of elements in rocks[J]. Geochimica et Cosmochimica Acta. 1959,16(1):79-82.
[113]Vistelius A B. The skew frequency distributions and the fundamental law of the geochemical processes[J]. The Journal of Geology.1960,68(1):1-22.
[114]Shaw D M. Element distribution laws in geochemistry[J]. Geochimica et Cosmochimica Acta. 1961,23(1):116-134.
[115]Ahrens L H. Element distributions in igneous rocks-Ⅵ. Negative skewness of SiO2 and K[J]. Geochimica et Cosmochimica Acta.1963,27(9):929-938.
[116]Ahrens L H. Lognormal-type distributions in igneous rocks-Ⅴ[J]. Geochimica et Cosmochimica Acta.1963,27(8):877-890.
[117]Ahrens L H. Lognormal-type distributions in igneous rocks-Ⅳ[J]. Geochimica et Cosmochimica Acta.1963,27(4):333-343.
[118]Ahrens L H. Element distributions in igneous rocks-Ⅶ:A reconnaissance survey of the distribution of SiO2 in granitic and basaltic rocks[J]. Geochimica et Cosmochimica Acta.1964, 28(3):271-290.
[119]Ahrens L H. Element distributions in specific igneous rocks-Ⅷ[J]. Geochimica et Cosmochimica Acta.1966,30(1):109-122.
[120]成秋明.多维分形理论和地球化学元素分布规律[J].地球科学——中国地质大学学报.2000,25(3)：311-318.
[121]Cheng Q, Agterberg F P, Ballantyne S B. The separation of geochemical anomalies from background by fractal methods[J]. Journal of Geochemical Exploration.1994,51(2):109-130.
[122]Allegre C J, Lewin E. Scaling laws and geochemical distributions[J]. Earth and Planetary Science Letters.1995,132(1):1-13.
[123]Friedman J H, Tukey J W. A projection pursuit algorithm for exploratory data analysis[J]. Computers, IEEE Transactions on.1974,100(9):881-890.
[124]Tukey J W. Exploratory data analysis[M]. Philippines:Addison-Wesley Publishing Company,Inc.,1977:1-688.
[125]Haining R, Wise S. Exploratory Spatial Data Analysis, NCGIA Core Curriculum in GIScience[Z].1997:2013.
[126]Haining R, Wise S, Ma J. Exploratory spatial data analysis in a geographic information system environment[J]. Journal of the Royal Statistical Society:Series D (The Statistician).1998,47(3): 457-469.
[127]Kiirzl H. Exploratory data analysis:recent advances for the interpretation of geochemical data[J]. Journal of Geochemical Exploration.1988,30(1):309-322.
[128]Howarth R J. Statistical applications in geochemical prospecting:a survey of recent developments[J]. Journal of Geochemical Exploration.1984,21(1):41-61.
[129]Venables W N, Ripley B D. Modern Applied Statistics with S[M]. Springer,2002:1-495.
[130]Tennant C B, White M L. Study of the distribution of some geochemical data[J]. Economic Geology.1959,54(7):1281-1290.
[131]Sinclair A J. Selection of threshold values in geochemical data using probability graphs[J]. Journal of Geochemical Exploration.1974,3(2):129-149.
[132]Sinclair A J. A fundamental approach to threshold estimation in exploration geochemistry: probability plots revisited[J]. Journal of Geochemical Exploration.1991,41(1):1-22.
[133]李随民,姚书振,韩玉丑Surfer软件中利用趋势面方法圈定化探异常[J].地质与勘探.2007,43(2)：72-75.
[134]陈希清,杨晓君,陈富文,等.应用MAPGIS数字高程模型提取区域地球化学异常信息的方法探讨[J].地球学报.2009,30(1)：119-125.
[135]石文杰,魏俊浩,张德才,等.基于数字高程模型因子分析的地球化学异常提取[J].物探与化探.2012,36(1)：103-108.
[136]Matschullat J, Ottenstein R, Reimann C. Geochemical background-can we calculate it?[J]. Environmental geology.2000,39(9):990-1000.
[137]Reimann C, Garrett R G. Geochemical background-concept and reality[J]. Science of the total environment.2005,350(1):12-27.
[138]Reimann C, Filzmoser P, Garrett R G. Background and threshold:critical comparison of methods of determination[J]. Science of the Total Environment.2005,346(1):1-16.
[139]Mrvic V, Kostic-Kravljanac L, cakmak D, et al. Pedogeochemical mapping and background limit of trace elements in soils of Branicevo Province (Serbia)[J]. Journal of Geochemical Exploration.2011,109(1):18-25.
[140]Mrvic V V, Kostic-Kravljanac L M, Zdravkovic M S, et al. Background limit of Zn and Hg in soils of Eastern Serbia[J]. Journal of Agricultural Sciences, Belgrade.2010,55(2):157-163.
[141]Devesa-Rey R, Diaz-Fierros F, Barral M T. Normalization strategies for river bed sediments:A graphical approach[J]. Microchemical Journal.2009,91(2):253-265.
[142]Sanei H, Goodarzi F, Hilts S. Site-specific natural background concentrations of metals in topsoil from the Trail region, British Columbia, Canada[J]. Geochemistry:Exploration, environment, analysis.2007,7(1):41-47.
[143]Diez M, Simon M, Dorronsoro C, et al. Background arsenic concentrations in Southeastern Spanish soils[J]. Science of the total environment.2007,378(1):5-12.
[144]Escalante A, Dipple G M, Barker S, et al. Defining trace-element alteration halos to skarn deposits hosted in heterogeneous carbonate rocks:Case study from the Cu-Zn Antamina skarn deposit, Peru[J]. Journal of Geochemical Exploration.2010,105(3):117-136.
[145]Mandelbrot B. How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension[J]. Science.1967,156(3775):636-638.
[146]Mandelbrot B B. The fractal geometry of nature[M]. Macmillan,1983.
[147]Bolviken B, Stokke P R, Feder J, et al. The fractal nature of geochemical landscapes[J]. Journal of Geochemical Exploration.1992,43(2):91-109.
[148]Cheng Q. The perimeter-area fractal model and its application to geology[J]. Mathematical Geology.1995,27(1):69-82.
[149]Cheng Q. Selection of Multifractal Scaling Breaks and Separation of Geochemical and Geophysical Anomaly[J]. Journal of China University of Geosciences.2001,12(1):54-59.
[150]Xu Y, Cheng Q. A fractal filtering technique for processing regional geochemical maps for mineral exploration[J]. Geochemistry:Exploration, environment, analysis.2001,1(2):147-156.
[151]Goncalves M A, Mateus A, Oliveira V. Geochemical anomaly separation by multifractal modelling[J]. Journal of Geochemical Exploration.2001,72(2):91-114.
[152]Zheru Z, Huahai M, Cheng Q. Fractal geometry of element distribution on mineral surfaces[J]. Mathematical geology.2001,33(2):217-228.
[153]Cheng Q, Russell H, Sharpe D, et al. GIS-based statistical and fractal/multifractal analysis of surface stream patterns in the Oak Ridges Moraine[J]. Computers & geosciences.2001,27(5): 513-526.
[154]Panahi A, Cheng Q. Multifractality as a measure of spatial distribution of geochemical patterns[J]. Mathematical geology.2004,36(7):827-846.
[155]Ge Y, Cheng Q, Zhang S. Reduction of edge effects in spatial information extraction from regional geochemical data:a case study based on multifractal filtering technique[J]. Computers & geosciences.2005,31(5):545-554.
[156]Ali K, Cheng Q, Chen Z. Multifractal power spectrum and singularity analysis for modelling stream sediment geochemical distribution patterns to identify anomalies related to gold mineralization in Yunnan Province, South China[J]. Geochemistry:Exploration, Environment, Analysis.2007,7(4):293-301.
[157]Xie S, Cheng Q, Ke X, et al. Identification of geochemical anomaly by multifractal analysis[J]. Journal of China University of Geosciences.2008,19(4):334-342.
[158]Xie S, Cheng Q, Xing X, et al. Geochemical multifractal distribution patterns in sediments from ordered streams[J]. Geoderma.2010,160(1):36-46.
[159]Zuo R. Decomposing of mixed pattern of arsenic using fractal model in Gangdese belt, Tibet, China[J]. Applied Geochemistry.2011,26:S271-S273.
[160]Zuo R. Identifying geochemical anomalies associated with Cu and Pb-Zn skarn mineralization using principal component analysis and spectrum-area fractal modeling in the Gangdese Belt, Tibet (China)[J]. Journal of Geochemical Exploration.2011,111(1):13-22.
[161]谢淑云,鲍征宇.地球化学场的连续多重分形模式[J].地球化学.2002,31(2)：191-200.
[162]谢淑云.地球化学场的分形与多重分形特征[D].武汉：中国地质大学(武汉),2003.
[163]Li C, Ma T, Shi J. Application of a fractal method relating concentrations and distances for separation of geochemical anomalies from background[J]. Journal of Geochemical Exploration. 2003,77(2):167-175.
[164]Agterberg F P, Cheng Q. Introduction to special issue on "Fractals and Multi fractals" [J]. Computers & Geosciences.1999,25(9):947-948.
[165]Zuo R, Carranza E J M, Cheng Q. Fractal/multifractal modelling of geochemical exploration data[J]. Journal of Geochemical Exploration.2012,122:1-3.
[166]陈建国,王仁铎,陈永清.利用分形统计学提取化探数据中的隐蔽信息并圈定地球化学异常[J].地球科学.1998,23(2)：175-178.
[167]成秋明.多重分形与地质统计学方法用于勘查地球化学异常空间结构和奇异性分析[J].地球科学——中国地质大学学报.2001,26(2)：161-166.
[168]李庆谋,成秋明.分形奇异(特征)值分解方法与地球物理和地球化学异常重建[J].地球科学——中国地质大学学报.2004,29(1)：109-118.
[169]李堃,胡光道,刘才泽.个旧地区地球化学元素的多重分形特征及找矿预测[J].矿产与地质.2006,20(4-5)：498-502.
[170]孙忠军.矿产勘查中化探异常下限的多重分形计算方法[J].物探化探计算技术.2007,29(1)：54.
[171]成秋明,张生元,左仁广,等.多重分形滤波方法和地球化学信息提取技术研究与进展[J].地学前缘.2009,16(2)：185-198.
[172]徐明钻,朱立新,马生明,等.多重分形模型在区域地球化学异常分析中的应用探讨[J].地球学报.2010,31(4)：611-618.
[173]娄德波,肖克炎,左仁广,等.分形滤波技术在新疆黄山-镜儿泉镍铜成矿带中的应用[J].地球学报.2012,33(1)：83-90.
[174]迟清华,鄢明才.应用地球化学元素丰度数据手册[M].北京：地质出版社,2007：148.
[175]Garrett R G. The chi-square plot:A tool for multivariate outlier recognition[J]. Journal of Geochemical Exploration.1989,32(1):319-341.
[176]Filzmoser P, Garrett R G, Reimann C. Multivariate outlier detection in exploration geochemistry[J]. Computers & Geosciences.2005,31(5):579-587.
[177]Filzmoser P, Hron K, Reimann C. Interpretation of multivariate outliers for compositional data[J]. Computers & Geosciences.2012,39:77-85.
[178]Kruskal J B. Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis[J]. Psychometrika.1964,29(1):1-27.
[179]Fry N. Random point distributions and strain measurement in rocks[J]. Tectonophysics.1979, 60(1):89-105.
[180]Vearncombe J, Vearncombe S. The spatial distribution of mineralization:applications of Fry analysis[J]. Economic Geology.1999,94(4):475-486.
[181]Cleveland W S. Visualizing data[M]. Hobart Press,1993.
[182]Reimann C, Filzmoser P, Garrett R G. Factor analysis applied to regional geochemical data: problems and possibilities[J]. Applied Geochemistry.2002,17(3):185-206.
[183]Johnson R A, Wichern D W. Applied multivariate statistical analysis[M]. Prentice hall Upper Saddle River, NJ,2002.
[184]de Caritat P, Grunsky E C. Defining element associations and inferring geological processes from total element concentrations in Australian catchment outlet sediments:Multivariate analysis of continental-scale geochemical data[J]. Applied Geochemistry.2013,33:104-126.
[185]孟宪伟,窦明晓,余先川.地球化学场分解的理论与方法[J].地球科学进展.1994,9(6)：59-64.
[186]时艳香,纪宏金,陆继龙,等.水系沉积物地球化学分区的因子分析方法与应用[J].地质与勘探.2004,40(5)：73-76.
[187]王永华,龚鹏,龚敏,等.成矿带1：20万水系沉积物地球化学分区的方法及地质意义：以西藏冈底斯铜多金属成矿带为例[J].现代地质.2010,24(4)：801-806.
[188]van Helvoort P, Filzmoser P, van Gaans P F. Sequential factor analysis as a new approach to multivariate analysis of heterogeneous geochemical datasets:an application to a bulk chemical characterization of fluvial deposits (Rhine-Meuse delta. The Netherlands)[J], Applied geochemistry.2005,20(12):2233-2251.
[189]Jackson J E. A user's guide to principal components[M]. Wiley, com,2005.
[190]Kaiser H F. The varimax criterion for analytic rotation in factor analysis[J]. Psychometrika. 1958,23(3):187-200.
[191]Bartlett M S. Tests of significance in factor analysis[J]. British Journal of Statistical Psychology. 1950,3(2):77-85.
[192]Efron B. Bootstrap methods:another look at the jackknife[J]. The annals of Statistics.1979:
[193]Eastment H T, Krzanowski W J. Cross-validatory choice of the number of components from a principal component analysis[J]. Technometrics.1982,24(1):73-77.
[194]Basilevsky A T. Statistical factor analysis and related methods:theory and applications[M]. John Wiley & Sons,2009.
[195]Gabriel K R. The biplot graphic display of matrices with application to principal component analysis[J]. Biometrika.1971,58(3):453-467.
[196]Cook D, Swayne D F. Interactive and dynamic graphics for data analysis:with R and GGobi[M]. Springer,2007.
[197]Grunsky E C, Smee B W. The differentiation of soil types and mineralization from multi-element geochemistry using multivariate methods and digital topography[J]. Journal of Geochemical Exploration.1999,67(1):287-299.
[198]Cheng Q, Bonham-Carter G F, Hall G, et al. Statistical study of trace elements in the soluble organic and amorphous Fe-Mn phases of surficial sediments, Sudbury Basin.1. Multivariate and spatial analysis[J]. Journal of Geochemical Exploration.1997,59(1):27-46.
[199]余先川,王世称,王桂安.稳健协同克立格因子分析及其在化探中的应用[J].地球科学——中国地质大学学报.1998,23(2)：171-174.
[200]孟宪伟.地球化学变量的空间自相关与互相关及空间因子分析[J].物探化探计算技术.1993,15(3)：192-203.
[201]Gordon A D. Classification[M]. Chapman&Hall, CRC,1999:256.
[202]Hubert L, Arabie P. Comparing partitions[J]. Journal of classification.1985,2(1):193-218.
[203]Halkidi M, Batistakis Y, Vazirgiannis M. Cluster validity methods:part I[J]. ACM Sigmod Record.2002,31(2):40-45.
[204]Milligan G W, Cooper M C. An examination of procedures for determining the number of clusters in a data set[J]. Psychometrika.1985,50(2):159-179.
[205]Calinski T, Harabasz J. A dendrite method for cluster analysis[J]. Communications in Statistics-theory and Methods.1974,3(1):1-27.
[206]Hartigan J A. Clustering algorithms[M]. John Wiley & Sons, Inc.,1975.
[207]Kaufman L, Rousseeuw P J. Finding groups in data:an introduction to cluster analysis[M]. John Wiley & Sons,2009.
[208]Chork C Y, Rousseeuw P J. Integrating a high-breakdown option into discriminant analysis in exploration geochemistry[J]. Journal of Geochemical Exploration.1992,43(3):191-203.
[209]Gnanadesikan R. Methods for statistical data analysis of multivariate observations[M]. John Wiley & Sons,2011.
[210]Howarth R J. Statistics and data analysis in geochemical prospecting[M]. Elsevier scientific publishing company,1983.
[211]Rousseeuw P J, Van Driessen K. Computing LTS regression for large data sets[J]. Data Mining and Knowledge Discovery.2006,12(1):29-45.
[212]Rousseeuw P J. Least median of squares regression[J|. Journal of the American statistical association.1984,79(388):871-880.
[213]Hoerl A E, Kennard R W. Ridge regression:Biased estimation for nonorthogonal problems[J]. Technometrics.1970,12(1):55-67.
[214]Hastie T, Tibshirani R, Friedman J J H. The elements of statistical learning[M]. Springer New York,2001.
[215]Jolliffe I. Principal component analysis[M]. Wiley Online Library,2002.
[216]王斌会.多元统计分析及R语言建模第2版[M].广州市：暨南大学出版社,2011：321.
[217]龚鹏,胡小梅,李娟,等.建立地质-地球化学找矿模型——以西藏甲玛铜多金属矿床为例[J].地质通报.2013,32(10)：1601-1612.
[218]龚鹏,马振东.矿产预测中区域化探异常的识别和评价[J].地球科学(中国地质大学学报).2013,38(S1)：113-125.
[219]Garrett R G, Grunsky E C. Weighted sums-knowledge based empirical indices for use in exploration geochemistry[J]. Geochemistry:Exploration, environment, analysis.2001,1(2): 135-141.
[220]Johnson J W. A heuristic method for estimating the relative weight of predictor variables in multiple regression[J]. Multivariate Behavioral Research.2000,35(1):1-19.
[221]Lance G N, Williams W T. Computer programs for hierarchical polythetic classification ("similarity analyses")[J]. The Computer Journal.1966,9(1):60-64.
[222]Lance G N, Williams W T. Mixed-Data Classificatory Programs I-Agglomerative Systems[J]. Australian Computer Journal.1967,1(1):15-20.
[223]向运川,任天祥,牟绪赞等.化探资料应用技术要求[M].北京市：地质出版社,2010：82.