GIS-based prospectivity-mapping based on geochemical multivariate analysis technology: A case study of MVT Pb–Zn deposits in the Huanyuan-Fenghuang district, northwestern Hunan Province, China

Accepted Manuscript GIS-based prospectivity-mapping based on geochemical multivariate analysis technology: A case study of MVT Pb–Zn deposits in the Huanyuan-Fenghuang district, northwestern Hunan Province, China Ku Wang, Nan Li, Leon Bagas, Shengmiao Li, Xianglong Song, Yuan Cong PII: DOI: Reference:

S0169-1368(17)30299-8 https://doi.org/10.1016/j.oregeorev.2017.09.015 OREGEO 2349

To appear in:

Ore Geology Reviews

Received Date: Revised Date: Accepted Date:

9 April 2017 15 September 2017 21 September 2017

Please cite this article as: K. Wang, N. Li, L. Bagas, S. Li, X. Song, Y. Cong, GIS-based prospectivity-mapping based on geochemical multivariate analysis technology: A case study of MVT Pb–Zn deposits in the HuanyuanFenghuang district, northwestern Hunan Province, China, Ore Geology Reviews (2017), doi: https://doi.org/10.1016/ j.oregeorev.2017.09.015

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

GIS-based prospectivity-mapping based on geochemical multivariate analysis technology: A case study of MVT Pb–Zn deposits in the Huanyuan-Fenghuang district, northwestern Hunan Province, China1 Ku Wang1, Nan Li1,2,*, Leon Bagas1,2, Shengmiao Li3, Xianglong Song1, Yuan Cong1 1. MLR Laboratory of Metallogeny and Mineral Resource Assessment, Institute of Mineral Resources, Chinese Academy of Geological Sciences, Beijing, 100037, China 2. Centre for Exploration Targeting, ARC Centre of Excellence for Core to Crust Fluid Systems, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia 3. Mineral Branch of Hunan Institute of Geological Survey, Changsha, China

Abstract This paper demonstrates a partial least-squares regression (PLS) method for geochemical modelling, and then uses the models and geological favourable features to obtain mineral potential maps. The PLS is one of multivariate analysis technologies, which can identify variations in associations and correlations among geochemical elements and mineralisation. The method is here used to calculate principal components as well as to identify correlations between Pb-Zn (mineralization) and 25 stream sediment elements for constructing geochemical models in the Huayuan-Fenghuang district of northwestern Hunan Province, China. The models showing the distribution of geochemical anomaly are useful in interpreting the distribution of faults and the Cambrian Qingxudong Formation (ore-bearing formation), and to better define the architecture on mineralisation in the study area. In addition, the models and other favourable features (proxies) are easily integrated into single possibility map by Boost Weights-of-Evidence (Boost WofE) approach for targets. Keywords: Mineral potential maps; Partial least squares regression; Singularity mapping; Boost Weights-of-Evidence; Mineral systems

*

Corresponding author: Nan Li, E-mail: [email protected], [email protected] (Nan Li).

1. Introduction Geographic information system (GIS)-based prospectivity mapping is widely used to target mineral potential maps at multiple spatial scales, and various methods have been proposed and developed (Bonham-Carter et al., 1989; Singer, 1993; Bonham-Carter, 1994; Singer and Kouda, 1996; Cheng et al., 1996, 2007, 2015; Cheng and Agterberg, 1999; Wang, 1999; Zhu, 1997; Brown et al., 2000; Knox-Robinson, 2000; Porwal and Carranza, 2001; Zhao, 2002; Zhao et al., 2003; Porwal et al., 2006a, 2006b; Zhao, 2007; Singer and Menzie, 2010). With increasing requests for mineral exploration at deeper levels and the constant enhancement of GIS technology, these methodologies have been used in two- and three-dimensional datasets (c.f. Porwal, 2006; Chen et al., 2007, 2012, 2014; Wang et al., 2011; Xiao et al., 2012, 2015; Perrouty et al., 2012, 2014; Li et al., 2015, 2016, 2017). Based on many studies as above, Porwal and Carranza (2015) summarize three major tasks in model-based (GIS-based) mineral potential mapping, namely: (i) identification of a conceptual geological and metallogenic model; (ii) collection of data and construction of models as GIS-layers; and (iii) the integration of proxy predictors that present interpretations of features such as geological, geochemical, geophysical data, and remote sensing. Firstly, the conceptual geological model is selected or summarized before GIS-based mineral resources assessment. Mineral deposit models and mineral systems are two of these models. Cox and Singer (1986) describe 85 mineral deposit models at the deposit scale. It has been extended to multiple scales, and Wyborn et al. (1994) proposed the mineral system concept of Magoon and Dow (1991) to target mineral deposits in green fields. Mineral system analysis in the past two decades has been applied to describe different geological settings at multiple scales, including Archean orogenic Au deposits (e.g. Goleby et al., 2004; Blewett et al., 2010), Zn–Pb–Ag deposits (e.g. Hoatson et al., 2006), Ni–Cu–(PGE) mineralisation (e.g. Beresford et al., 2007;

Begg et al., 2009), uranium mineralisation (e.g. Skirrow, 2009), the genesis of opal deposits (e.g. Rey, 2013), and the enrichment of Fe in banded-iron formation (BIF) (e.g. Angerer et al., 2014). In addition, the constituents of continental- or global-scale mineral systems include the source rocks, generation and migration of mineralised fluids, structural architecture, and sites of deposits (Wyborn et al., 1994; Price and Stoker, 2002; Walshe et al., 2005; Cleverley, 2006; McCuaig and Hronsky, 2014). Secondly, geological conceptual studies decide what datasets to be collected and what modelling technologies to be addressed for compiling predictive proxies. Geochemical anomaly is one of critical indicators for prospecting, especially for supergene. Modelling of geochemical anomalous has been widely used in mineral prospecting and exploration targeting. The reason is that geochemical anomalies can pinpoint the location mineral deposits and associated pathfinder elements (e.g. Agterberg, 2007; Carranza, 2009; Grunsky, 2010). Many papers and books have been published on the: (i) distribution and identification of geochemical anomalies, including the use of normal or log-normal distributions with self-similarity and self-affinity; (ii) fractal distributions (Bölviken et al., 1992; Cheng et al., 1994; Zuo et al., 2009; Agterberg, 2014); (iii) concentration–area models (C-A) by Cheng et al., (1994); (iv) spectrum–area models (S-A) by Cheng (1999); (v) singularity indexes (Cheng, 2007); (vi) concentration–volume models (C-V) (Afzal et al., 2011); (vii) modified singularity mapping techniques (Zuo et al., 2015); and (viii) multivariate analysis settings that figure out association and correlations between geochemical elements and mineralization (Reimann et al., 2008; Zuo et al., 2013; Deutsch et al., 2016). An important example here is Principal component analysis (PCA), which can be used to construct components or hidden information integrating multi-dimension data (based on predictor correlation structures; Hartnett et al., 1998; Li, 2016). In contrast, the PLS regression approach is another multivariate analysis method that considers the relationship between the predictors and responses by the construction of latent

variables. Finally, according to what proxies are constructed, geologists can devise a specific combination method (Bonham-Carter et al., 1989; Cheng et al., 1996, 2007, 2015; Cheng and Agterberg, 1999; Knox-Robinson, 2000; Porwal and Carranza, 2001), for example by choosing knowledge-driven method or data-driven method. This paper demonstrates a modelling process using the PLS method and singularity mapping technique based on components of PLS to establish proxy predictors for MVT Pb-Zn potential mapping in the Huayuan-Fenghuang area of northwestern Hunan Province, China. Finally, the geochemical models are integrated by Boost WofE approach into posterior-map. For geological perspective, the interpretation of correlation from PLS has helped to improve the understanding of mineralisation on a regional scale and to help in exploration targeting. 2. Study district 2.1. MVT lead-zinc introduction Mississippi Valley Type (MVT) Pb-Zn mineralisation is an important resource of base metals around the world. Most MVT Pb–Zn deposits are hosted by carbonate-dominated sequences that are typically located at the edges of basins, at orogenic forelands, and at thrust belts inboard of clastic-dominated passive margin sequences (Leach et al., 2005). Joly et al. (2015) summarises the critical elements of sedimentary exhalative (Sedex) and MVT deposits for prospectivity mapping using a mineral system approach. The following assumptions are considered: (i) Meso- to Paleoproterozoic sedimentary and mafic volcanic rocks can act as key sources; (ii) the presence of syn-volcanic and syn-sedimentary structures, such as growth and intersecting faults, may focus fluid flow and localise sulfide mineralisation; (iii) the presence of mineral anomalies that could be indicative of hydrothermal

alteration; and (iv) the presence of evaporites acting as a source of ligands are important for the transportation of base and precious metals. 2.2. Regional geological setting The Huayuan-Fenghuang area contains major MVT-type mineral deposits in northwestern Hunan Province (Fig. 1; Table 1). The region is located at the southeastern margin of the Yangtze Block, and is part of a major mineralised belt that extends through the Hunan and Guizhou provinces (Yang and Lao, 2007a). The area contains thick sedimentary sequences with minimal regional metamorphism in a 10 km-thick marine sequence that includes an 1800 m-thick carbonate unit that hosts the ore deposit. The Huayuan–Zhangjiajie structural belt is important in the northwestern Xuefeng Orogen, and has been recognised as a major ore-controlling structure within the study area. The fault zone trends between 030° and 050°, with dip directions between 45° to 80° SE, is up to 200 m in width, and has been multiply deformed during the Paleozoic and Cretaceous (Li, 1991; Yang and Lao, 2007a; Tang et al., 2012a). The early history of the structure dates back to the Late Meso- to Neoproterozoic when north- to northwest subduction of an oceanic plate ceased with the development of the Xuefeng Orogen. The orogeny was accompanied by a series of NE-trending thrusts, with the youngest event characterised by the development of tight folds and associated NE- to NEE-trending faults (Liu, 1985; Yang, 1987). Tang et al. (2013) propose that long-lived basin-controlling structures and growth faults controlled the facies-type of the sequences deposited. This was succeeded by extensional tectonic in a NW-SE orientation leading to a drop in the sea level with a paleo-coastline trended NE–SW, and with the depth of water increasing gradually from the north to the south (Mei et al., 2006). The sea regression resulted in the deposition of algal-reef carbonates and associated Pb–Zn mineralisation (Yang and Lao, 2007b). The depositional

environment progressed from an open platform in the west through a slope to a basin-margin facies in the east (Fig. 1). 2.3. Mineral deposits and prospecting model The mineral deposits in the Huayuan-Fenghuang area, such as at Bamaozhai, Tudiping, Limei and Laohuchong, include structurally controlled and MVT-type deposits hosted by algal-reef carbonate interbedded with calc-arenite (Luo et al., 2009; Cheng et al., 2011). Sedimentary structures include bioclastic beds, oolitic and oncolite textures, and erosional channels containing sandstone and locally derived conglomerate (Fu, 2011; Yang and Lao, 2007b). 2.3.1. Typical deposit: Limei The Limei deposit is a typical example of MVT mineralisation in the region, and was chosen here for detailed studies. The deposit is up to 200 m thick hosted by the Early Cambrian Qingxudong Formation, which consists of thickly bedded sandstone inter-bedded with algal-reef carbonate (Chen, 2011; Figs. 2, 3). The lower part of the deposit is richer in grade and larger in size, and the grade and size decreases upward (Liu and Zheng, 2000). In addition, the relative proportion of Zn and Pb changes from north to south, with a Zn/Pb ratio of 10:1 in the north to 2:1 in the south (Chen et al., 2011). The major minerals at Limei are euhedral sphalerite ((Zn, Fe, S) with a Zn grade of 0.71 to 5.08%), galena (PbS with a Pb grade 0.5 to 2.43), pyrite (FeS2), and wulfenite (PbMoO4) (Fig. 4; Cai et al., 2014; Chen et al., 2011). The sulfides are commonly accompanied by poorly crystalline calcite, barite, dolomite, fluorspar, and gypsum gangue (Chen et al., 2011). The sphalerite is typically hosted by irregularly distributed calcite veins, the galena is commonly between 2 and 20 mm in diameter, pyrite is granular and crystalline, and wulfenite forms thin tabular crystals with a bright orange-red to yellow-orange

colour (Luo et al., 2009; Chen et al., 2011). 2.3.2. Ore genesis model The Qingxudong Formation was deposited in a rift-type setting, and was locally metasomatised by mineralised fluid migrated upward along first-order deep-seated faults and on to second-order splay faults where they were deposited in structurally and chemically prepared sites (Fig. 5; Kuang et al., 2015). The Qingxudong Formation was one of the paleo-oil reservoirs in the area. Geologists (Luo et al., 2009; Xia and Fu, 2010; Duan et al., 2014) inferred the oilfield brine was saline, heated at depth, and extracted the ore-forming materials from the Lower Cambrian Shipai Formation clastic rocks during convection. Influenced by Caledonian Movement, the balance of paleo-oil and ore-forming fluid was destroyed, which attributed to the precipitation of Pb-Zn deposit, and in this process the oilfield brine of the paleo-oil reservoir probably provided Pb-Zn deposits with ore-forming sulphur (Liu et al., 2012). The ore-forming time of Pb-Zn deposits is nearly the same as that of the accumulation and destruction of the paleo-oil reservoir during Caledonian. 2.3.3. Indicators for mineralisation The important geological controls for Pb–Zn mineralisation are structures, sedimentary units, and algal-reef carbonate (Chen et al., 2008; Yang, 2003; Yang and Lao, 2007a). The major mineralised structures are NNE- to NE-trending faults, and the mineralised carbonate is commonly located near the intersection of these faults (He and Huang, 2016; Liu, 1985; Tang et al., 2012b). The secondary mineralised structures are NE–trending anticlines. The major lithological unit is algal-reef carbonate, and the richest part of the mineralisation is in oolitic calc-arenite at the top of the unit (Zheng and Zeng, 1988; Fu, 2011; Tang et al., 2012a). Features such as the location of

convecting hot brine in the source-region for the deposits cannot be easily depicted on exploration prospectivity maps. However, approximations of such features are the presence of anomalous stream-sediment geochemistry patterns that might reflect the movement of the fluid and deposition processes of the mineralisation (Cai et al., 2014; Chen et al., 2008). Alteration is also an important indicator, which includes calcite veining, barite, pyrite, and dolomit-rich zones. These minerals are used as targets indicators of hydrothermal fluid in the study area (Duan et al., 2014; Yang and Lao, 2007). 3. Geochemical modelling 3.1. Dataset collected and compiled As part of this study 1578 stream sediment samples were collected, which covered the entire Huayuan-Fenghuang district using a grid of 2 x 2 km2 (c.f. Xie et al., 1997). The samples were assayed for major (SiO2, Al2O3, K2O, Na2O, CaO, MgO, and Fe2O3) and trace elements (Ag, As, Au, B, Ba, Be, Bi, Cd, Cr, Cu, F, Hg, Li, Mn, Mo, Nb, Ni, P, Pb, Sb, Sn, Ti, V, W, and Zn). 3.2. Methodology for geochemical modelling 3.2.1. Multivariate statistical analysis Several multivariate analysis methods are used to identify geochemical anomalies following the examples of Ergon (2003), Reimann et al. (2008) and Botre et al. (2016). Among these, Principal Component Analysis (PCA) method was applied to the data integrating multivariate geochemical anomalies related to spatial distribution of geological features and mineralisation (c.f. Carranza, 2010; Grunsky, 2010; Wang et al., 2015); The approach decreases the numbers of variables by constructing “component” variables (c.f. Hartnett et al., 1998; Li et al., in press). Mineralising processes are typically multi-stage and have many sources

so that the results always show complex paragenetic relationships in space and time (c.f. Tripathi, 1979; Singer and Kouda, 2001; Voroshilov, 2009). It was pointed out earlier that geochemical anomalies and alteration act as important predictors for exploration targeting, and significant anomalies can be recognised after establishing the background values (e.g. Aitchison, 1986; Filzmoser et al., 2009; Filzmoser et al., 2010; Reimann et al., 2012; Zuo et al., 2013; Zuo, 2014). However, most of the frequency-based methods do not consider the spatial variation of a geochemical field, consequently significant but weak geochemical anomalies cannot be easily detected (Zuo et al., 2013). The PLS method has been applied in many different communities including analysis

of remote sensing (e.g. Rapaport et al., 2015;

Fernández-Espinosa, 2016; Jaqueline et al., 2016), geochemical (e.g. Li et al., 2015; Botre et al., 2016), and mineral exploration (e.g. Makvandi et al., 2016). In general terms, the method uses latent variables to model relationships between blocks of independently observed variables. This assumes the data observed are generated by a process driven by a small number of latent variables, which are not directly observed or measured (Chen, 2012). Theoretically, the PLS method is based on regression models, principal components analysis, and canonical correlation analysis (Zhang et al., 2009). It is usually used to handle uncorrelated linear transformations termed latent components for variables that have high covariance with dependent variables (Chen, 2012). Latent components are useful in sorting variables into multi-independent or multi-dependent sets eliminating noise in the data (Farifteh et al., 2007). PLS is therefore a valuable tool for modelling geochemical anomalies related to mineralisation, and is used to calculate correlations between two “sets” of variables in X and Y space based on “component” variables. Values are integrated into “component” variables using a linear algorithm based on two weight-vectors, which are termed u and v. The method builds successive and orthogonal components for each set to ensure that the covariance between the pair of components is maximal, as expressed

in the equation:

max uh

= 2

vh

u′h X′h−1Yh−1 v h

2

=1

(1)

where X 0 = X and Y0 = Y (whose columns have been standardised), and where X h −1 and Yh −1 are the deflated matrices obtained by subtracting from variables X and Y . 3.2.2. The principles of singularity mapping The Singularity mapping technique is based on local singularity exponents, and is used to compile geochemical maps at different scales quantifying the spatial features of “concentration” and “depletion” (e.g. Cheng, 2007; Zhao, 2012; Zuo and Wang, 2016). The total amount of metals in a study, A, can be denoted as
(), whereas the concentration of an element in the same area can be denoted as (). There is a power-law relationships between mass
() or concentration (), according to the Singularity theory, and A can be expressed by the equations: 

<
() >= 

(2)



< () >=  

(3)

where c is the fractal density constant, and the symbol < > means “expectation”, implying that the power-law relationship is statistically significant (Cheng, 2007; Wang et al., 2012; Wang et al., 2013, 2015; Zhao et al., 2013). The exponent “α” preserves the shape of the function, and when α is a single value it indicates a mono-fractal elemental distribution, whilst a multiple value indicates the presence of a multi-fractal elemental distributions. There are three different conditions based on the values of α, where: (1) α = 2 implying that the average element concentration is stable; (2) 0 < α < 2 implying there is enrichment in elements due to geological processes, which is spatially consistent with known mineral deposits in area A; and (3) α > 2 or a < 0 is

indicative of the depletion of elements in area A. 3.3. Modelling in the study area 3.3.1. Pre-processing Pre-processing should be completed before the PLS method can be used for compiling Singularity maps. The GeoSAS software, which is utilises state-of-the-art GIS and RS technologies to solve spatial problems, is used with the Singularity mapping technique to define the location of significant geochemical anomalies (Zuo et al., 2015). Given that all of the geochemical datasets contain dependent variables and compositional data that are parts of a whole containing relative information, there are possible closure problem that should be rectified before the data is processed (Aitchison, 1986; Pawlowsky-Glahn et al., 2011; Buccianti et al., 2006; Reimann et al., 2012). Three log-ratio transformations are used to handle compositional data opening, which are called additive log-ratio (ALR) transformation, centred log-ratio (CLR) transformation, and isometric log-ratio (ILR) transformation (Aitchison, 1986; Egozcue et al., 2003). These transformations have both merits and disadvantages under different conditions. The ALR transformation is subjective and depends on the selection of the denominator (Aitchison, 1986). Although CLR can avoid this problem, the transformation results in matrix singularity problems, because the resulting data are collinear and the sum of values of CLR variables is zero (c.f. Reimann et al., 2012; Zuo et al., 2013). The ILR transformation can give a correct representation of compositional data in Euclidean space and is thus used in this contribution for data pre-processing. 3.3.2. Geochemical processing All 25 trace elements analysed from the study area were seen as

independent (observed) variables presented as ‘X’. The Pb and Zn components of mineralisation are seen as dependent variables presented as ‘Y’. Thirteen components have been extracted based on the results of cross-validation, after which the predicted error sum of squares (PRESS) was calculated for each component. The “hill-climbing” approach of Russell and Russell and Norvig (2003) was used to select favourable components and search for the next component using the TANAGRA 1.4.49 software. Five principal components (Components 1 to 5) were used containing 97.621% of the cumulative contribution to the dependent variable ‘Y’ (Table 2). Based on the analysis of two columns of data called “Contribution rate to X” and “Contribution rate to Y” (Table 2), the Component 1 contains more robust information than the others in explaining the distribution of the dependent variables Pb and Zn. Integrated element associations can also be extracted from Component 1 (Fig. 6), where all of 25 geochemical elements have positive values (P1), and the elements Pb, Zn, Cd, and As are the most favourable for mineralisation (represented on the Y-axis). Furthermore, P1 is analysed by interpolation and then establishing a score map that indicates the spatial distribution of the first component (Fig. 7). The Singularity index (α) is calculated based on square areas of different sizes on a kilometre scale, and Fig. 10 shows Singularity index maps for P1. When these maps are compared against the geological map, it becomes apparent that known mineralisation is located along major fault zones (Figs. 1 and 8). This means that areas with 0 < α < 2 values, such as around Limei, are highly prospective for mineralisation where regional-scale faults intersect. These sites potentially provided pathways for mineralised hydrothermal fluid. 3.3.3. Identification of geochemical anomalies The Singularity index (α) was used to extract anomalous concentrations of

Pb, Zn, Cd, As, Mn, Ti, Bi, Li, W, Be, Cr, and Cu included in the first component (P1) of the PLS method. The Inverse distance weighting (IDW) method is used to assess geochemical data at a 2 km spatial resolution and in the generation of geochemical distribution raster maps. The Window-based method in GeoSAS is used to assess the singularity indexes (α) for mineralised elements (Fig. 9). 4. Prospectivity mapping 4.1 Multiple spatial datasets Multiple spatial datasets were collected containing geological and geochemical data at a regional scale (Table 3). The datasets include tectonic structures, mineralised formations, known deposits, and stream sediment geochemistry (introduced in Section 3; Fig. 10). 4.2. Geological modelling The model of stream sediment dataset discussed in Section 3 was used to highlight the location of geochemical anomalies. In this section, we assess the anomalies using the information entropy method and fault density. 4.2.1. Entropy process Mineralised strata are important indicators used to focus on mineral deposit (c.f. Section 2.3.3). Such strata can act as proxy predictors for mineralisation and are thus modelled for mineralisation. The entropy process is an effective tool for studying tectonic structures controlling orebodies and geological processes (Chi, 1985). For example, Chi and Zhao (2000) used the combined entropy anomalies of geological units to target mineralisation in the Yunnan Province of China. In this paper, mineralised strata are important indicators used to focus

onto mineral deposit (see Section 2.3.3). Such strata can act as important proxy predictors for mineralisation and are thus modelled. The method produces grid-unit-based entropy contours generated by initially obtaining the total area for all individual sedimentary beds in each grid-unit, before calculating the area sum for each grid unit. The ratio of the area and the stratum from one grid-unit (xi (i = 1, 2, 3, …, n)) are calculated using the following equation by Chi and Zhao (2000): n

∑ x ln x i

E jk = −

i

i =1

ln n

(4)

where n is the number of geological elements in one unit, j is the row number, and k is the column number of the unit. The size of the grid-unit chosen is based on the distribution of geological strata, making sure each grid-unit contains at least four strata. The presence of higher entropy in a grid-unit relates to an increased geological complexity. The entropy of geological formations are calculated in this contribution using a 2 X 2 km² grid (Fig. 11). 4.2.2. Fault density Faults are also indicators for mineralisation (refer to Section 2.3). The fault density is defined as the total length of faults in a grid cell, which is defined as follows (Zhao et al., 2011; Elish and Mohammed, 2015):

di =

1 pi ∑l Ai j =1 ij

(5)

where pi is the total number of the faults in the grid cell (i), l is the total ij length of the jth fault in the grid cell(i), A(i) is the area of the grid cell(i). The study area is divided into 1512 units by each cell with 2 X 2 km². The main fault density is then mapped using MRAS software with favourable rocks shown in Figs. 12 and 13.

4.3. Combination of multiple spatial datasets 4.3.1. Boost WofE The Boost WofE is a kind of weights of evidence model that was developed to calculate conditional weights like ordinary weights, but on the basis of a progressively weighted training sample (Cheng, 2012, 2015). The method provides a solution for the requirement of ordinary weights of evidence model on conditional independency among evidential layers (Cheng, 2015). Boost WofE can be used to integrate quantitative exploration criteria in sequence with updates, within dependent weights for the first exploration criterion and the conditional weights for the subsequent evidence. 4.3.2. Targeting Mineral potential maps in the Huayuan-Fenghuang region were assessed using Boost WofE. Resource estimation is completed by the improved volume method by Li et al. (in press) using 3D models for the Limei deposit and Qingxudong Formation at a district scale. An empty mineral potential map is constructed by dividing the study area into 1512 units using a 2 x 2 km² a cell, and proxies are chosen from Table 4. In the case of the Limei deposit, favourable sedimentary units are not independent of favourable strata layers, and the latter layer was removed. The remaining seventeen proxies were integrated into posterior probabilities using the Boost WofE model (Table 5). The inflection points on the cumulative graph shown in Fig. 14 indicates that the posterior probability in each cell can be classified into three intervals (Fig. 15), with targets at: (1) on a fault near Tuanjie Town; (2) around Qianchang; (3) north of Shuiyin; (4) near Buchou; and (5) near the Wangpo Mountain.

5. Discussion The geochemical models presented in Section 3 are here geologically interpreted, and compared using the PLS and PCA methods. 5.1. Mineral system approach in the Huayuan-Fenghuang district 5.1.1. Source It is envisaged that mineralised meteoric fluid containing Pb-Zn migrated along deep structures to the bottom of basins in the Huayuan-Fenghuang area. The highest concentrations of Pb and Zn are found in Neoproterozoic to Early Cambrian formations (Yang and Lao, 2007b; Fu, 2011), but not in economic concentrations. 5.1.2. Pathways Major pathways carrying mineralised fluids in the study area are likely to be deep-seated faults and splays off them. Deep-seated faults can also act as sites of deposition, especially at their inflection points. Given that the Huayuan–Zhangjiajie Fault hosts mineralisation in the study area, ore-forming fluids would have migrated along faults, driven by tectonic activity and elevated temperatures during Paleozoic to Cretaceous orogenic events. 5.1.3. Traps The development of MVT Pb–Zn mineralisation is obviously dependent on the present mineralised horizons, and lithological variations along strike (c.f. Cai et al., 2014). The mineralisation in the study area is located in Early Cambrian algal-reef carbonate beds of the Qingxudong Formation. The characteristic sedimentary structures in the mineralised part of the formation are characteristic of a slope facies that is rich in organic material and is also likely to be a source for the sulfur in the mineral deposits (Cai et al., 2014).

5.1.4. Geological perspective based on geochemical modelling As discussed above, pathways and traps are two important indicators for targeting MVT Pb–Zn deposits using prospectivity maps at a regional scale (Fig. 17). The results of geochemical modelling show that the spatially varied distribution of these two indicators are obvious and have been included in the prospectivity maps. The multivariate analysis of the geochemical anomalies shown in Fig. 8 indicates that Pb, Zn, Cd, and as correlate positively with the location of economic concentrations of Pb–Zn. High positive anomalies are spatially related to the algal-reef carbonate in the Qingxudong Formation, which obviously acted as structural or geochemical traps. Almost all known Pb–Zn deposits in the study area are located in areas with high positive anomalies (Fig. 8). In contrast, the Neoproterozoic rocks, Ordovician Bitiao Formation, and Cretaceous rocks have negative anomalies, which implies that these rocks might be the source for the mineralisation. The Pb–Zn deposits are associated with the fault density in the study area, showing that mineralisation is best developed where faults are best developed. Furthermore, intersecting faults are also invariable located near geochemical anomalies, confirming their role as pathways for mineralised fluids. This also confirms that the conceptual geological models presented in Section 2 are valid. 5.2. Comparison between the Partial least-squares regression (PLS) and Principal component analysis (PCA) methods In this section, we compared the PLS and PCA methods at two points that are modelling and combination. Firstly, the first six components in the geochemical dataset were calculated and selected by the PCA method and have Eigen-values greater than the other components (Fig. 18). The first component calculated using the

PCA method is similar to the results using the PLS method (Figs. 7 and 19). All elements have positive values in the first component, but correlations among elements differ from the results of the PLS method. The mineral association determined using the PCA method is Cr – Ni – Cu – Be – Ti – V – As – Bi – Nb – W – P – B, but the method cannot explain the relationships between these elements and mineralisation (Pb-Zn) elements. As mentioned in Section 3.3.3, the mineral association using the PLS method is Pb – Zn – Cd – As – Mn – Ti – Bi – Li – W – Be – Cr – Cu, which does recognise the mineralisation elements. Consequently, the observed mineral association is properly ordered using the PLS method, but the PCA method does not achieve this level of accuracy. Furthermore, the PLS method not only calculates the element associations in order, it explains the relationships between mineralisation and the association. Secondly, twelve elements are used in this contribution as predictive layers (Tables 4 and 5) from PLS method. The advantage is more detail related to the known mineralization are integrated into the final prospectivity map. 6. Conclusion and future work This paper demonstrates geochemical modelling based on the PLS and singularity

method,

and

target

MVT

Pb–Zn

prospectivity

in

the

Huayuan-Fenghuang district, China. Firstly, the mineral system approach and genetic models are used to delineate what spatial datasets should be collected for MVT Pb–Zn deposits at the regional scale. The indicators acting as vectors pointing to deposits include the structural architecture, favourable lithological units (i.e. algal-reef carbonate that act as chemical traps), geochemical anomaly data (including stream-sediment analyses), geological maps at scale of 1:200,000. Secondly, the mineral association of Cr – Ni – Cu – Be – Ti – V – As – Bi – Nb – W – P – B was constructed by the PLS and singularity methods. Thirdly, the geological anomalies were modelled using information entropy and fault density, and

finally, Boost WofE was used to integrate the proxy predictors and assess MVT lead-zinc prospectivity with three levels that depicted the rank of the potential for MVT Pb–Zn mineralisation in the study area. The mineral potential maps indicate the following areas are targets for MVT Pb-Zn exploration: (1) faulting in the Tuanjie area; (2) the Qianchang area between Longtan and Maoer; (3) north of Shuiyin; (4) around Buchou; and (5) around the Wangpo Mountain between Huanghe to Linfeng. A final part of this study is the analyses and summaries of the results as prospectivity maps from geological perspective, and a comparison between the PLS and PCA multivariate analysis methods. The study has been found that the partial least-squares (PLS) method not only extracts the principal components from trace elements, but also distinguishes how importance is for each element relating to mineralisation. In the future, the PLS method will be tested in different study areas for other deposit types to verify its effectiveness for mineral prospecting. Acknowledgements This study is financially supported by the China National Mineral Resources Assessment Initiative (Project No. 1212011120140), the National Natural Science Foundation of China (Project No. 41672330; NSFC, http://www.nsfc.gov.cn/), and the National Key Technology Research and Development Program of the Ministry of Science and Technology of China (Project No. 2006BAB01A01). We acknowledge the use of MRAS, GeoSAS, and TANAGRA for the construction of geological and geochemical models presented in this contribution. We also appreciate the valuable comments of anonymous reviewers and editors. References Afzal, P., Alghalandis, Y.F., Khakzad, A., Moarefvand, P., Omran, N.R., 2011. Delineation of mineralization zones in porphyry Cu deposits by fractal concentration-volume

modeling. Journal of Geochemical Exploration. 108(3), 220-232. Agterberg, F.P., 2007. Mixtures of multiplicative cascade models in geochemistry. Nonlinear ProcGeoph. 14, 201-209. Agterberg, F.P., 2014. Geomathematics: Theoretical Foundations, Applications and Future Developments. Springer International Publishing, Cham (ZG), pp. 413-466. Aitchison, J., 1986. The Statistical Analysis of Compositional Data. Chapman & Hall Ltd., London, 416 pp. Angerer, T., Duuring, P., Hagemann, S.G., Thorne, W., McCuaig, T.C., 2014. A mineral system approach to Archean and Paleoproterozoic banded iron formation hosted iron ore of Western Australia. Geol. Soc. Lond. Spec. Publ. 393, 81-115. Begg, G., Hronsky, J.M.A., O'Reilly, S., Griffin, W., Hayward, N., 2009. Plumes, cratons, and nickel sulphide deposits, Proceedings of the 10th Biennial SGA Meeting of the Society for Geology Applied to Mineral Deposits, Townsville, Australia, pp. 378-380. Beresford, S.W., Fiorentini, M., Hronsky, J.M.A., Rosengren, N., Bertuch, D.I., Saleem, A., Cas, R.A.F., Stone, W.E., Duuring, P., Bekker, A., Barley, M., 2007. Komatiite-hosted Ni–Cu–(PGE) deposits: understanding deposit and camp footprints, Proceedings of Geoconferences (WA) Inc. Kalgoorlie ‘07 Conference. Geoscience Australia Record 2007, Kalgoorlie, Western Australia, 159-163 pp. Blewett, R.S., Henson, P.A., Roy, I.G., Champion, D.C., Cassidy, K.F., 2010. Scale-integrated architecture of a world-class gold mineral system: The Archaean eastern Yilgarn Craton, Western Australia. Precambrian Research. 183(2), 230-250. Bölviken, B., Stokke, P.R., Feder, J., Jössang, T., 1992. The fractal nature of geochemical landscapes. Journal of Geochemical Exploration. 43(2), 91-109. Bonham-Carter, G.F., Agterberg, F.P., 1989. Weights of evidence modelling: a new approach to mapping mineral potential. Geological Survey of Canada. 89, 171-183. Bonham-Carter, G. F., 1994. Geographic Information systems for Geoscientist: Modeling with GIS. Elsevier, Oxford, U.K., 398 pp. Botre, C., Mansouri, M., Nounou, M., Nounou, H., Karim, M.N., 2016. Kernel PLS-based GLRT method for fault detection of chemical processes. Journal of Loss Prevention in the Process Industries. 43, 212-224.

Brown, W.M., Gedeon, T.D., Groves, D.I., Barnes, R.G., 2000. Artificial neural networks: a new method for mineral prospectivity mapping. Australian Journal of Earth Science. 47(4), 757-770. Buccianti, A., Mateu-Figueras, G., Pawlowsky-Glahn, V., 2006. Compositional data analysis in the geosciences: from theory to practice. Geological Society Special Publications. 22(1), 139-141. Cai, Y., Yang, H., Duan, R., Lu, S., Zhang, L., Liu, C., Qiu, X., 2014. Fluid Inclusions and S, Pb, C Isotope Geochemistry of Pb-Zn Deposits Hosted by Lower Cambrian in Western Hunan-Eastern Guizhou Area. GEOSCIENCE. 28(1), 29-41. Carranza, E.J.M., 2009. Geochemical Anomaly and Mineral Prospectivity Mapping in GIS. Elsevier, Amsterdam, 368 pp. Carranza, E.J.M., 2010. Improved Wildcat Modelling of Mineral Prospectivity: Wildcat Modelling of Mineral Prospectivity. Resource Geology. 60(2), 129-149. Chen, J., Lv, P., Wu, W., Zhao, J., Hu, Q., 2007. A 3D method for predicting blind orebodies, based on a 3D visualization model and its application. Earth Science Frontiers. 14(5), 54-62. Chen, J., Wang, C., Shang, B., Shi, R., 2012. Three-Dimensional Metallogenic Prediction in Yongmei Region Based on Digital Ore Deposit Model. Scientific and Technological Management of Land and Resources. 29(6), 14-20. Chen, J., Yu, P., Shi, R., Yu, M., Zhang, S., 2014. Research on three-dimensional quantitative prediction and evaluation methods of regional concealed ore bodies. Earth Science Frontiers. 21(05), 211-220. Chen, S., 2012. The Comparison of Classification Model with Partial Least Square Based Dimension Reduction. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 173-180. Chen, M., Hu, M., Bao, Z., Bao, J., 2011. Discussion on the Geological Features and Genesis of the Limei Pb-Zn ore Concentration Belt in North-western Hunan Province. Geology and Exploration. 47(2), 251-260. Chen, M., Sun, J., Fu, Y., Bao, Z., Bao, J., 2008. Geology and Geochemistry of Lead-Zinc Ore Zone and Prospecting Potential in Longshan County, Western Hunan. Resources Environment & Engineering. 22(02), 151-158.

Cheng, Q., Agterberg, F.P., Bonham-Carter, G.F., 1996. A spatial analysis method for geochemical anomaly separation. Journal of Geochemical Exploration. 56(3), 183-195. Cheng, Q., Agterberg, F.P., 1999. Fuzzy Weights of Evidence Method and Its Application in Mineral Potential Mapping. Natural Resources Research. 8, 27-35. Cheng, Q., 1999. Spatial and scaling modeling for geochemical anomaly separation. Journal of Geochemical Exploration. 65(3), 175-194. Cheng, Q., 2007. Mapping singularities with stream sediment geochemical data for prediction of undiscovered mineral deposits in Gejiu, Yunnan Province, China. Ore Geology Reviews. 32(1-2), 314-324. Cheng, Q., 2012. Application of a newly developed boost weights of evidence model (Boost WofE) for mineral resources quantitative assessments. Jilin Univ. Earth Sci. 42(6), 1976-1985. Cheng, Q., 2015. BoostWofE: A new sequential weights of evidence model reducing the effect of conditional dependency. Math. Geosci. 47, 591-621. Cheng, Q., Bonham-Carter, G., Wang, W., Zhang, S., Li, W., Xia, Q., 2011. A spatially weighted principal component analysis for multi-element geochemical data for mapping locations of felsic intrusions in the Gejiu mineral district of Yunnan, China. Computers & Geosciences. 37(5), 662-669. Cheng, Q., Agterberg, F.P., Ballantyne, S.B., 1994. The separation of geochemical anomalies from background by fractal methods. Journal of Geochemical Exploration. 51(2), 109-130. Chi, S., 1985. The application of entropy function in Geological exploration. Earth Science. 10(02), 131-139. Chi, S., Zhao, P., 2000. Application of combined-entropy anomaly of geological formations to delineation of preferable ore-finding area. Geoscience. 14(04), 423-428. Cleverley, J.S., 2006. Using the chemistry of apatite to track fluids in Fe-oxide Cu-Au systems. Geochimica et Cosmochimica Acta. 70(18), A105. Cox, D.P., Singer, D.A., 1986. Mineral Deposit Models. U.S. Geological Survey Bulletin. United States Government Printing Office, Washington, 379 pp.

Deutsch, J.L., Palmer, K., Deutsch, C.V., Szymanski, J., Etsell, T.H., 2016. Spatial Modeling of Geometallurgical Properties: Techniques and a Case Study. Natural Resources Research. 25(2), 161-181. Duan, Q., Cao, L., Zeng, J., Zhou, Y., Tang, Z., Li, K., 2014. Rb-Sr Dating of Sphalerites from Shizishan Pb-Zn Deposit in Huayuan Ore Concentration Area, Western Hunan, and Its Geological Significance. Earth Science Journal of China University of Geosciences. 39(08), 977-986. Egozcue, J.J., Pawlowsky-Glahn, V., Mateu-Figueras, G., Barceló-Vidal, C., 2003. Isometric logratio transformations for compositional data analysis. Mathematical Geology. 35(3), 279-300. Elish, M.O., Mohammed, M.A., 2015. Quantitative analysis of fault density in design patterns: An empirical study. Information and Software Technology. 66, 58-72. Ergon, R., 2003. Constrained numerical optimization of PCR/PLSR predictors. Chemometrics And Intelligent Laboratory Systems. 65(2), 293-303. Farifteh, J., Van der Meer, F., Atzberger, C., Carranza, E.J.M., 2007. Quantitative analysis of salt-affected soil reflectance spectra: A comparison of two adaptive methods (PLSR and ANN). Remote Sensing of Environment. 110, 59-78. Fernández-Espinosa, A.J., 2016. Combining PLS regression with portable NIR spectroscopy to on-line monitor quality parameters in intact olives for determining optimal harvesting time. Talanta. 148, 216-228. Filzmoser, P., Hron, K., Reimann, C., 2009. Univariate statistical analysis of environmental (compositional) data: Problems and possibilities. Science of the Total Environment. 407(23), 6100-6108. Filzmoser, P., Hron, K., Reimann, C., 2010. The bivariate statistical analysis of environmental (compositional) data. Science of the Total Environment. 408(2010), 4230-4238. Fu, S., 2011. Discussion on formation rules of high-grade Pb-Zn ore in Western Hunan. Nonferrous Metals (Mining Section). 63(6), 27-35. Goleby, B.R., Blewett, R.S., Korsch, R.J., Champion, D.C., Cassidy, K.F., Jones, L.E.A.,

Groenewald, P.B., Henson, P., 2004. Deep seismic reflection profiling in the Archaean northeastern Yilgarn Craton, Western Australia: implications for crustal architecture and mineral potential. Tectonophysics. 388(1-4), 119-133. Grunsky, E.C., 2010. The interpretation of geochemical survey data. Geochemistry: Exploration, Environment, Analysis. 10(1), 27-74. Hartnett, M.K., Lightbody, G., Irwin, G.W., 1998. Dynamic inferential estimation using principal components regression (PCR). Chemometrics and Intelligent Laboratory Systems. 40(2), 215-224. He, F., Huang, J., 2016. Geological characteristics and genesis of strata bound Pb-Zn deposits in the Hunan-Guizhou-Guangxi Triangular Area. Mineral resources and geology. 30(02), 208-210+244. Hoatson, D.M., Jaireth, S., Jaques, A.L., 2006. Nickel sulfide deposits in Australia: Characteristics, resources, and potential. Ore Geology Reviews. 29(3-4), 177-241. Jaqueline M.D.P., José E.M.B., Paulo, J.S.B., 2016. A PLS regression model using flame spectroscopy emission for determination of octane numbers in gasoline. Fuel. 176, 216-221. Joly, A., Porwal, A., McCuaig, T.C., Chudasama, B., Dentith, M.C., Aitken, A.R.A., 2015. Mineral systems approach applied to GIS-based 2D-prospectivity modelling of geological regions: Insights from Western Australia. Ore Geology Reviews. 71, 673-702. Knox-Robinson, C.M., 2000. Vectorial fuzzy logic: a novel technique for enhanced mineral prospectivity mapping, with reference to the orogenic gold mineralisation potential of the Kalgoorlie Terrane, Western Australia. Australian Journal of Earth Science. 47(5), 929-941. Kuang, W., Xiang, S., Xiao, W., Chen, W., Yang, S., Yu, P., Chen, N., 2015. Metallogenic geological characteristics and genesis of lead-zinc deposits in northwestern Hunan. Mineral Deposits. 34(05), 1072-1082. Leach, D.L., Sangster, D.F., Kelley, K.D., Large, R.R., Garven, G., Allen, C.R., Gutzmer, J., Walters, S., 2005. Sedement-hosted lead-zinc deposits: A global perspective, Economic Geology 100th Anniversary Volume, pp. 561-607.

Li, F., Liu, J., Cao, L., 2015. A comparative QSAR study on the estrogenic activities of persistent organic pollutants by PLS and SVM. Emerging Contaminants. 1(1), 8-13. Li, H., 2016. Accurate and efficient classification based on common principal components analysis for multivariate time series. Neurocomputing. 171, 744-753. Li, N., Bagas, L., Li, X., Xiao, K., Li, Y., Ying, L., Song, X., 2016. An improved buffer analysis technique for model-based 3D mineral potential mapping and its application. Ore Geology Reviews. 76, 94-107. Li, N., Bagas, L., Lindsay, M., Wedge, D., Bai, L., Song, X., 2017. An irregular triangle mesh buffer analysis method for boundary representation geological object in three-dimension. Earth Science Informatics. 10(2), 149-167. Li, R., Wang, G., Carranza, E.J.M., 2016. GeoCube: A 3D mineral resources quantitative prediction and assessment system. Computers & Geosciences. 89, 161-173. Li, Z., 1991. A Preliminary Discussion on the Origin of Pb-Zn Ore deposits in Western Hunan and Eastern Guizhou. Geology of Guizhou. 8(4), 363-371. Liu, J., Zou, X., Tang, C., Cui, S., Xia, J., Gan, J., Zhao, W., Jin, S., 2012. Preliminary Discussion on Relationship between Pb-Zn Deposits and Paleo-oil Reservoirs in Western Hunan Province and Eastern Guizhou Province. Geology and Mineral Resources of South China. 28(3), 220-225. Liu, W., 1985. Evolution of Hunan-Guizhou Fault Zone and the Features of Mineralization. Geological Review. 81(3), 224-231. Liu, W., Zheng, R., 2000. Characteristics and movement of ore-forming fluids in the Huayuan lead-zinc deposit. Mineral Deposits. 19(02), 173-181. Luo, W., Yin, Z., Kong, L., Dai, T., 2009. Discussion on the Geological Features and Genesis of the Limei Pb-Zn Ore Concentration Belt in North-western Hunan Province. Geological survey and research. 33(3), 194-202. Magoon, L.B., Dow, W.G., 1991. The petroleum system from source to trap. The American Association of Petroleum Geologists, Tulsa, U.S.A., 655pp. Makvandi, S., Ghasemzadeh-Barvarz, M., Beaudoin, G., Grunsky, E.C., McClenaghan, M.B., Duchesne, C., Boutroy, E., 2016. Partial least squares-discriminant analysis of trace element compositions of magnetite from various VMS deposit subtypes:

Application to mineral exploration. Ore Geology Reviews. 78, 388-408. McCuaig, T.C., Hronsky, J.M.A., 2014. The mineral system concept: the key to exploration targeting. Society of Economic Geologists, Inc.Special Publication, pp. 153-175. Mei, M., Zhang, C., Zhang, H., Meng, X., Chen, Y., 2006. Sequence-Stratigraphic Frameworks and Their Forming Backgrounds of Paleogeography for the Lower Cambrian of the Upper-Yangtze Region. Geoscience. 20(2), 195-208. Pawlowsky-Glahn, V., Buccianti, A., Pawlowsky-Glahn, V., 2011. Compositional Data Analysis : Theory and Applications. John Wiley & Sons, Incorporated, Hoboken, 402 pp. Perrouty, S., Aillères, L., Jessell, M.W., Baratoux, L., Bourassa, Y., Crawford, B., 2012. Revised Eburnean geodynamic evolution of the gold-rich southern Ashanti Belt, Ghana, with new field and geophysical evidence of pre-Tarkwaian deformations. Precambrian Research. 204-205, 12-39. Perrouty, S., Lindsay, M.D., Jessell, M.W., Aillères, L., Martin, R., Bourassa, Y., 2014. 3D modeling of the Ashanti Belt, southwest Ghana: Evidence for a litho-stratigraphic control on gold occurrences within the Birimian Sefwi Group. Ore Geology Reviews. 63, 252–264. Porwal, A.K., 2006. Mineral potential mapping with mathematical geological models, International Institute for Geo-information Science and Earth Observation, Enschede, The Netherlands, 279 pp. Porwal, A.K., Carranza, E.J.M., 2001. Extended Weights of Evidence Modelling for Predictive Mapping of Base Metal Deposit Potential in Aravalli Province, Western India. Exploration Mining Geology. 10, 273-287. Porwal, A.K., Carranza, E.J.M., Hale, M., 2006a. A Hybrid Fuzzy Weights-of-Evidence Model for Mineral Potential Mapping. Natural Resources Research. 15, 1-14. Porwal, A.K., Carranza, E.J.M., Hale, M., 2006b. Bayesian network classifiers for mineral potential mapping. Computers & Geosciences. 32(1), 1-16. Porwal, A.K., Carranza, E.J.M., 2015. Introduction to the Special Issue: GIS-based mineral potential modelling and geological data analyses for mineral exploration, Ore Geology Reviews. 71, 477-483. Price, G.P., Stoker, P., 2002. Australian Geodynamics Cooperative Research Centre's

integrated research program delivers a new minerals exploration strategy for industry. Aust. J. Earth Sci. 49(4), 595-600. Rapaport, T., Hochberg, U., Shoshany, M., Karnieli, A., Rachmilevitch, S., 2015. Combining leaf physiology, hyperspectral imaging and partial least squares-regression (PLS-R) for grapevine water status assessment. ISPRS Journal of Photogrammetry and Remote Sensing. 109, 88-97. Reimann, C., Filzmoser, P., Garrett, R., Dutter, R., 2008. Statistical Data Analysis Explained. Applied Environmental Statistics with R. Wiley, Chichester, UK, 362 pp. Reimann, C., Filzmoser, P., Fabian, K., Hron, K., Birke, M., Demetriades, A., Dinelli, E., Ladenberger, A., 2012. The concept of compositional data analysis in practice-Total major element concentrations in agricultural and grazing land soils of Europe. Science of the Total Environment. 426, 196-210. Rey, P.F., 2013. Opalisation of the Great Artesian Basin (central Australia): an Australianstory with a Martian twist. Aust. J. Earth Sci. 60, 291-314. Russell, S.J., Norvig P., 2003. Artificial intelligence: a modern approach. Upper Saddle River, N.J: Prentice Hall, 1152 pp. Singer, D.A., 1993. Basic concepts in three-part quantitative assessments of undiscovered mineral resources. Non-renewable Resources. 2, 69-81. Singer, D.A., Kouda, R., 1996. Application of Feedforward Neural Network in the Search for Kuroko Deposits in the Hokuroku District, Japan. Mathematical Geology. 28(8), 1017-1023. Singer, D.A., Kouda, R., 2001. Some Simple Guides to Finding Useful Information in Exploration Geochemical Data. Natural Resources Research. 10(2), 137-147. Singer, D.A., Menzie, W.D., 2010. Quantitative mineral resource assessments-An integrated approach: Oxford University Press, New York, 219 pp. Skirrow, R.G.E., 2009. Uranium ore-forming systems of the Lake Frome region, South Australia: Regional spatial controls and exploration criteria. Geoscience Australia Record. 40, 148. Tang, Z., Deng, F., Li, K., Wang, G., Duan, Q., Zou, X., Zhao, W., Jin, S., 2012a. Sedimentary Sequence of the Lower Cambrian Qingxudong Formation in the Western

Hunan and Eastern Guizhou Provinces and its Constrains on Lead-Zinc Mineralization. Geotectonica et Metallogenia. 36(1), 111-117. Tang, Z., Deng, F., Li, K., Duan, Q., Zou, X., Dai, P., 2012b. Stratigraphic characteristics of the Cambrian Qingxudong Formation in relation to lead-zinc mineralization in western Hunan-eastern Guizhou area. Geology in China. 39(04), 1034-1041. Tang, Z., Deng, F., Li, K., Zhao, W., Jin, S., 2013. Lithofacies palaeogeography of the Qingxudong epoch, Duyun stage, Cambrian Period and its relationship to lead-zinc deposits in western Hunan and eastern Guizhou Provinces. Geology and Exploration. 49(1), 19-27. Tripathi, V.S., 1979. Factor analysis in geochemical exploration. Journal of Geochemical Exploration. 11, 263-275. Walshe, J.L., Cooke, D.R., Neumayr, P., 2005. Five questions for fun and profit: a mineral system perspective on metallogenic epochs, provinces and magmatic hydrothermal Cu and Au depositsMineral Deposit Research: Meeting the Global Challenge. Springer, Berlin, Heidelberg, pp. 477-480. Wang, G., Zhang, S., Yan, C., Song, Y., Sun, Y., Li, D., Xu, F., 2011. Mineral potential targeting and resource assessment based on 3D geological modeling in Luanchuan region, China. Computers & Geosciences. 37(12), 1976-1988. Wang, H.W., 1999. Partial least squares regression and its application. National Defence Industry Press, Beijing, China, 300 pp. Wang, W., Zhao, J., Cheng, Q., 2011. Analysis and integration of geo-information to identify granitic intrusions as exploration targets in southeastern Yunnan District, China. Computers & Geosciences. 37(12), 1946-1957. Wang, W., Zhao, J., Cheng, Q., 2013. Fault trace-oriented singularity mapping technique to characterize anisotropic geochemical signatures in Gejiu mineral district, China. Journal of Geochemical Exploration. 134, 27-37. Wang, W., Zhao, J., Cheng, Q., Carranza, E.J.M., 2015. GIS-based mineral potential modeling by advanced spatial analytical methods in the southeastern Yunnan mineral district, China. Ore Geology Reviews. 71, 735-748. Wang, W., Zhao, J., Cheng, Q., Liu, J., 2012. Tectonic–geochemical exploration modeling

for characterizing geo-anomalies in southeastern Yunnan district, China. Journal of Geochemical Exploration. 122, 71-80. Wyborn, L.A.I., Heinrich, C.A., Jauqes, A.L., 1994. Australian Proterozoic mineral systems: essential ingredients and mappable criteria, Australian Mining Looks North-the Challenges and Choices. Australian Institute of Mining and Metallurgy Publication, pp. 109-115. Xie, X., Mu, X., Ren, T., 1997. Geochemical mapping in China. Journal of Geochemical Exploration. 60(1), 99-113. Xia, X., Fu, S., 2010. Mineralization Pattern of North Western Hunan Lead-zinc Mine. Nonferrous Metals (Mining Section). 62(2), 35-38. Xiao, K., Li, N., Sun, L., Zou, W., Li, Y., 2012. Large scale 3D mineral prediciton methods and channels based on 3D information technology. Journal of Geology. 36(3), 229-236. Xiao, K.X., Li, N., Porwal, A., Holden E.J., Leon, B., Lu, Y., 2015. GIS-based 3D prospectivity mapping: A case study of Jiama copper-polymetallic deposit in Tibet, China. Ore Geology Reviews. 71, 611-632. Yang, S., 2003. Analysis on the prospecting foreground for Pb-Zn deposits in West Hunan. Hunan Geology. 22(02), 107-111. Yang, S., Lao, K., 2007a. A tentative discussion on genesis of lead-zinc deposits in northwest Hunan. Mineral Deposits. 26(3), 330-340. Yang, S., Lao, K., 2007b. Geological characteristics and ore indicators of lead-zinc deposits in northwestern Hunan, China. Geological Bulletin of China. 26(7), 899-908. Yang, Z., 1987. A Paleo-fault Zone Traversing Southeastern China. Scientia geologica sinica. 22(3), 221-230. Zhang, Z., Yu, J., Mao, J., Rui, Z., 2009. Advances in the study of Mississippi Valley-type deposits. Mineral Deposits. 28(02), 195-210. Zhao, J., Wang, W., Cheng, Q., 2013. Investigation of spatially non-stationary influences of tectono-magmatic processes on Fe mineralization in eastern Tianshan, China with geographically weighted regression. Journal of Geochemical Exploration. 134, 38-50. Zhao, J., Wang, W., Dong, L., Yang, W., Cheng, Q., 2012. Application of geochemical

anomaly identification methods in mapping of intermediate and felsic igneous rocks in eastern Tianshan, China. Journal of Geochemical Exploration. 122(2012), 81-89. Zhao, J., Chen, S., Zuo, R., Carranza, E.J.M., 2011. Mapping complexity of spatial distribution of faults using fractal and multifractal models: vectoring towards exploration targets. Computers & Geosciences. 37(12), 1958-1966. Zhao,

P.D.,

2002.

"Three

step"

quantitative

forecasting

and

assessment

of

resource-theory of digital exploration and discussion on practice. Earth Science. 27, 482-489. Zhao P.D., Chen, J.P., Zhang, S.T., 2003. The new development of "three components" quantitative mineral prediction. Earth Science Frontiers. 2, 455-463. Zhao, P., 2007. Quantitative mineral prediction and deep mineral exploration. Earth Sci. Front. 14, 1-10. Zheng, R., Zeng, Y., 1988. Sedimentary characteristics of early Cambrian Yutang organic reefs in western Hunan. ACTA Sedimentologica Sinica. 6(2), 61-67,134. Zhu, Y.S., 1997. Prospectivity Mapping Methods. Geological Publish House, Beijing, China, 361 pp. Zuo, R., 2014. Identification of geochemical anomalies associated with mineralization in the Fanshan district, Fujian, China. Journal of Geochemical Exploration. 139, 170-176. Zuo, R., Cheng, Q. and Xia, Q., 2009. Application of fractal models to characterization of vertical distribution of geochemical element concentration. Journal of Geochemical Exploration. 102(1), 37-43. Zuo, R., Wang, J., 2016. Fractal/multifractal modeling of geochemical data: A review. Journal of Geochemical Exploration. 164, 33-41. Zuo, R., Wang, J., Chen, G., Yang, M., 2015. Identification of weak anomalies: A multifractal perspective. Journal of Geochemical Exploration. 148, 12-24. Zuo, R., Xia, Q., Wang, H., 2013. Compositional data analysis in the study of integrated geochemical anomalies associated with mineralization. Applied Geochemistry. 28, 202-211.

Figures Fig.1. Regional geological map. Fig.2. Cross-section of the Limei deposit. Fig.3. Measured geological profile at Limei deposit. Fig.4. Samples. A. galena; B. sphalerite. Fig.5. Ore genesis model of Huayuan MVT Pb-Zn deposits Fig.6. Associations of trace elements from the first component (P1) by PLS. Fig.7. Spatial distribution map from the first component (P1) by PLS. Fig.8. Spatial distribution map from P1 by singularity index. Fig.9. Fig.9. Spatial distribution maps of singularity index of Pb, Zn, Cd, As, Mn, Ti, Bi, Li, W, Be, Cr, and Cu. Fig.10. Ore-bearing formation (Qingxudong formation) in the study area. Fig.11. Combined-entropy anomaly of geological formation in the study area. Fig.12. Fault density map in the study area. A. faults; B. fault-density. Fig.13. Litho-facies map in the study area. Fig.14. Cumulative probability. Fig.15. Mineral potential map in the study area. Fig.16. Average grade of Pb and Zn in different units (Modified byYang and Lao, 2007b). Fig.17. Prospectivity mapping based on mineral systems (modified by McCuaig et al., 2014). Fig.18. Associations of trace elements by the first component (P1) by PCA. Fig.19 Spatial distribution map based on the first component (P1) by PCA.

Tables Table1. Features of Pb-Zn deposits in Huayuan-Fenghuang district, China (modified by Zhong et al., 2007). Table 2. Principal component information by PLS. Table 3. Geological survey dataset. Table.4. Indicators of prospecting model

Table.5. Boost weights of evidences.

Table1.Comparative table between typical MVT Pb-Zn and Pb-Zn deposits in NW Hunan Province, China (modified by Zhong et al., 2007). Features

Typical MVT Pb-Zn ore deposit

Pb-Zn ore deposits at north-west Hunan province, China

Regional Tectonic

In the carbonate platform along

Enriched in Paleozoic large

Setting

the basin margin, and partially in

sedimentary basin along the

the rift

south east margin of Yangze Platform

Host stratigraphy

Mainly in the Paleozoic, Triassic

Mainly hosted in Cambrian and

Lithology

thick sedimentary rock formation

Ordovician carbonate with

(e.g. dolomite and biogenic reef

thickness of ~5000 m, i.e. widely

carbonate), others in the

developed dolomite and other

Proterozoic

shallow water carbonate facies together with biogenic reef facies

Relation between

Ore forming process in close

n the platform cover folds and

tectonic activities

relation with orogenic activities

closely related to Yanshanian

and ore deposit

or tectonic collision with weak

tectonic activities with smooth

rock deformation and

strata formation

metamorphism for the ore deposits

Deposit Distribution

Widespread mineralization

Range of mineralization belts

forming large ore belts

about tens of thousands square kilometres

Ore body

Layer bound characteristic of the

Main ore body controlled by

morphology

mineral ore controlled by

stratum and stratoid morphology

stratum, stratoid and lenticular

in consistent with wall rock

morphology

occurrence both dominated by the interlayer fracture

Host structures

Various open gaps

Interlayer fracture, structural

includingstructural fissures,

fissures, and faults

unconformity surface and so on

Mineral

Simple mineral assemblage of

Simple composition such as

assemblage

sphalerite and galena, and minor

sphalerite and galena with other

amounts of disseminated pyrite

minor minerals

and chalcopyrite

Wall rock alteration

Ore grade

Ore structure

Carbonate-, dolomite-, silica-,

Mainly controlled by carbonate-,

Fe- and fluorite-alteration

dolomite- and silica-alteration

Pb+Zn: ~5-15% Zn as the main

Pb+Zn: ~3-10% Zn as the main

element accompanied by Ag and

element accompanied by Ag、Cd

Cu

and Cu.

Fine- to coarse-grained,

Various grain and metasomatic

massive, brecciated,

structure with metasomatic, patch

disseminated and veins with

and disseminated texture

metasomatic and dissolution textures

Ore deposit scale

Mainly small- and medium-size

Appraised 1 large deposit and 5

deposits distribute in groups with

small and medium size deposits

some large ore deposits

with promising prospectivity

δ34S: ~+10 – +25‰; enriched

δ34S: ~+11 – +31‰ (mostly

heavy S is derived from marine

~20‰). Enriched heavy S is

evaporates with no distinct

derived from marine sulfate. Crust

characteristic disparities

with mantle mixed Pb in orogenic

amongst the ore deposits

belts

Ore forming

Homogenization temperatures

Homogenization temperatures

temperature and

measured in mineral inclusions

measured in mineral inclusions

depth

(such as hornblende) are ~

(such as hornblende) are

80–220° with the ore forming

~99–190° corresponding to an ore

depth being over hundreds and

forming depth of between 0.9 and

thousands metres

1.38 km

Fluid

Underground thermal brine with

Thermal brine with high salinity

characteristics

salinity between ~10 and 30%

over 5% and alkane organic

Pb, Sisotope

matter

Relation between

None

None

ore deposits and intrusion

Table 2. Principal component information extracted using PLS. Input variables (X)

Target Variables (Y)

Number of

Contribution

Cumulative

Contribution

Cumulative

components

rate to X (%)

contribution rate

rate to Y (%)

contribution rate

to X (%)

to Y (%)

1

23.219

23.219

61.331

61.331

2

14.604

37.823

26.205

87.536

3

3.131

40.954

8.472

96.008

4

10.544

51.497

0.890

96.898

5

5.613

57.110

0.723

97.621

Table 3. Geological survey data in the study area. Data set type

Scale

Description of the data

Source

Details of original multi-type data in studyarea Map of mineral

1:200,000

Pb-Zn deposits

deposits Regional

Hunan Institute 1:200,000

geological map

1 map covered the entire study

of Geological

area including stratigraphy

Survey

boundary, stratigraphic column, and structures. Geochemical

1:200,000

Span of sample points is 2km;

Chinese

stream sediment

The elements include Ag, As, Au,

National

data

B, Ba, Be, Bi, Cd, Cr, Cu, F, Hg,

Geochemical

Li, Mn, Mo, Nb, Ni, P, Pb, Sb, Sn, Ti, V, W , Zn,Zr, Al2O3, CaO,

Mapping Project

Fe2O3, K2O, MgO , Na2O, SiO2.

Table.4. Indicators for the prospectivity model. indicators

Predictors

Parameters

Strata

Favorable layers

Qingxudong Formation

Combined-entropy anomaly

of geological formations

Combined-entropy>50

Favorable lithofacies

open platform facies

lithofacies

and slope facies Structures

Faults buffer

Buffer of 1km

Density of the faults

Density>0.04

Geochemical

Pb anomaly

index(α)<2

models

Zn anomaly

index(α)<2

Cd anomaly

index(α)<2

As anomaly

index(α)<2

Mn anomaly

index(α)<2

Ti anomaly

index(α)<2

Bi anomaly

index(α)<2

Li anomaly

index(α)<2

W anomaly

index(α)<2

Be anomaly

index(α)<2

Cr anomaly

index(α)<2

Cu anomaly

index(α)<2

Table.5. Results of Boost weights of evidences. W+

W-

C

Favorable layers

1.33

-1.73

3.07

Combined-entropy anomaly of

0.34

-0.48

0.82

Faults buffer

0.49

-0.45

0.94

Density of the faults

0.25

-0.28

0.52

Pb anomaly

0.99

-1.14

2.13

Zn anomaly

0.83

-1.52

2.34

Cd anomaly

0.82

-1.51

2.33

As anomaly

0.18

-0.26

0.44

Mn anomaly

0.45

-0.80

1.25

Ti anomaly

0.23

-0.43

0.66

Bi anomaly

0.45

-1.20

1.65

Li anomaly

0.16

-0.23

0.39

W anomaly

0.30

-0.63

0.92

Be anomaly

0.37

-1.08

1.44

Cr anomaly

0.38

-0.73

1.10

geological formations

Cu anomaly

0.44

-0.95

1.39

Highlights
Geochemical modelling using partial least-squares regression (PLS) method;
Proxy predictors were established by PLS, Singularity mapping, Entropy process, and Fault density;
The geochemical models presented in Section 3 are here geologically interpreted, and compared using the PLS and PCA methods.