Journal of Geochemical Exploration 139 (2014) 170–176

Contents lists available at ScienceDirect

Journal of Geochemical Exploration

journal homepage: www.elsevier.com/locate/jgeoexp

Identification of geochemical anomalies associated with mineralization

in the Fanshan district, Fujian, China
Renguang Zuo ⁎
State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences, Wuhan 430074, China

a r t i c l e i n f o a b s t r a c t

Article history: The Fanshan district is a prospective area to explore for epithermal-type Cu–Au mineralization that is similar to
Received 11 March 2013 the Zijinshan Cu–Au deposit in the neighboring region which is the largest Cu–Au deposit in southern China. In
Accepted 21 August 2013 this study, the robust principal component analysis (RPCA) method and spectrum-area (S-A) fractal model were
Available online 28 August 2013
applied to multi-element data from 2042 stream sediment samples collected in this region. The isometric logratio
(ilr) transformation was used to open the geochemical data for reducing the effects of data closure problem. The
Keywords:
Geochemical exploration
resulting PC1 obtained using RPCA displays two different compositional assemblages: (I) Pb, Zn, Sn, W, Mo, Bi, Hg,
Compositional data and Ag, perhaps representing hydrothermal-type Zn–Pb polymetallic mineralization, and (II) As, Au, Cu, Sb and
Data closure problem Mn, likely representing epithermal-type Cu–Au mineralization. The decomposed anomaly map from the PC1
Logratio transformation obtained from the S-A model illustrates that highly anomalous areas occurring around the intrusions are
Robust principal component analysis correlated with hydrothermal-type Zn–Pb polymetallic mineralization, and areas with lower anomalies such as
Spectrum-area fractal model Dafanshan and Xiaofanshan are perhaps associated with Cu–Au polymetallic mineralization, and should be
further studied to explore for epithermal-type Cu–Au mineralization.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction
Extensive high quality, multi-element, and multi-scale geochemical
data sets have been collected since the Chinese geochemical mapping
program was launched in 1978 (Xie et al., 2008). It plays a significant
role in mineral resource exploration in China as a number of mineral
deposits have been discovered (Xi and Li, 2012). The processing of
geochemical data for detecting multivariate geochemical patterns or
signals associated with mineralization in support of mineral resource
exploration is challenging.
In this regard, two important issues should be addressed. The first
one is the data closure problem for geochemical data analysis. Geo-
chemical data are typically compositional data with the components
measured as proportions or percentages of some whole data and only
provide relative information because the information is the proportion
of the element in the complete sample (Aitchison, 1982, 1986;
Aitchison et al., 2000; Buccianti and Pawlowsky-Glahn, 2005). The
data closure problem may result in spurious correlations or associations
between geochemical variables and complicate the interpretation of
different correlations between the same variables among different
subcompositions. For instance, a scatterplot of SiO2 versus Al2O3 for
Kola Project C-horizon soils indicates that SiO2 has a negative
⁎ Tel./fax: +86 27 67885096.
E-mail address: zrguang@cug.edu.cn.
correlation (r = −0.65) with Al2O3 (Fig. 1a) (Reimann et al., 2008).
When the closed data is opened using the additive logratio transforma-
tion (alr) proposed by Aitchison (1982) (TiO2 as denominator) to elim-
inate the spurious relationships between geochemical variables, the
relationship between SiO2 and Al2O3 becomes positive (r = 0.91)
(Fig. 1b). The same phenomenon also occurs in stream sediment data
collected from Fujian Province, China (Zuo, 2013). The Harker diagram
for the raw data also shows a negative relationship (r = −0.83)
between SiO2 and Al2O3 (Fig. 2a) because the major component SiO2
increases and all other components decrease due to the fixed sum
(100%) (Reimann et al., 2008). The Harker diagram for the opened
data using alr (CaO as denominator) demonstrates that SiO2 has a
positive correlation (r =0.81) with Al2O3 (Fig. 2b). The positive rela-
tionship can be interpreted due to the presence of both Al and Si in
most common rock-forming minerals and their weathering products
(Reimann et al., 2008). The differences between Fig. 1a and b, and
between Fig. 2a and b display a serious problem when analyzing
inter-element relationships in geochemical data.
Palarea-Albaladejoa and Martín-Fernández (2013) showed the
biplots for the Hongite dataset from Aitchison (1986). The biplot is a
powerful tool to display the relationships between variables via loadings
and observations via the scores (Gabriel, 1971). It can simultaneously
visualize scores (data points or sample identifiers) and loadings
(vectors) of the principal components (PC). The correlations between
variables are measured by the angle between any two vectors. The
dataset consists of 25 mineral samples of the composition x = [albite,

0375-6742/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.gexplo.2013.08.013
 R. Zuo / Journal of Geochemical Exploration 139 (2014) 170–176 171

Fig. 1. Scatterplot of SiO2 versus Al2O3 for Kola Project C-horizon soils: (a) the raw data, and (b) after alr transformed data.
After Reimann et al., 2008).

Fig. 2. Scatterplot of SiO2 versus Al2O3 for Fujian stream sediment data: (a) the raw data, and (b) after alr transformed data.

Fig. 3. Simplified geological map.

Modified from Geological Survey Institute of Fujian, China, 2011.
172 R. Zuo / Journal of Geochemical Exploration 139 (2014) 170–176

Fig. 4. Maps showing stream sediment sampling locations. Data collected from a mineral resource survey in Fanshan district by the Geological Survey Institute of Fujian, 2011.

Fig. 5. Maps showing the spatial distribution of Cu (a), Pb (b), Zn (c) and Au (d).
Data collected from a mineral resource survey in Fanshan district by the Geological Survey Institute of Fujian, 2011.
 R. Zuo / Journal of Geochemical Exploration 139 (2014) 170–176 173

blandite, cornite, daubite, endite] in wt %. When considering the full can be classified into three alteration zones: (1) quartz–pyrite–alunite,
components, the Pearson's correlation coefficients between cornite and (2) silica and (3) sericite–pyrophylite. The main alteration is the
endite, and between albite and blandite are 0.42 and 0.51, respectively. secondary alunite–quartzlithification that also contains sulfidation.
However, if respectively considering the subcompositions s1 = [cornite, The datasets used in this study were a digital 1:50,000 geological
daubite, endite] and s2 = [albite, blandite, endite], the Pearson's correla- map, and a regional stream sediment geochemical dataset. The geo-
tion coefficients between cornite and endite, and between albite and chemical dataset is based on analyses of 2042 stream sediment samples
blandite become −0.74 and −0.94, respectively (Palarea-Albaladejoa collected at a density of 4.37 samples per 1 km2 (Fig. 4) during a miner-
and Martín-Fernández, 2013). It is difficult to interpret the different cor- al resource survey in Fanshan district by the Geological Survey Institute
relations between the same components when working with different of Fujian (GSIF). The following 13 elements were analyzed (numbers in
subcompositions from a geological perspective. Therefore, geochemical parentheses denote detection limits in μg/g): Ag (0.03), As (1), Au
data should be opened prior to analysis, such as the principal component (0.0003), Bi (0.1), Cu (1.5), Hg (0.005), Mn (30), Mo (0.5), Pb (5), Sb
analysis and multivariate statistical analysis. Even when there is only one (0.2), Sn (1), W (0.5), and Zn (15). The sample preparation, analyses
component, the effects of the data closure problem also should be and quality control methods follow Chinese Geochemical Survey
considered because the values of a selected element represent its propor- Specifications of DZ/T 0011-91 and ZD0130.6-94. The concentrations
tions in the complete sample (Filzmoser et al., 2009). of Ag and Sn were determined using atomic emission spectrometry;
The second problem is how to decompose mixed and complex Cu, Pb, Mn and Zn by inductively coupled plasma optical emission spec-
patterns into geochemical components to reflect the effects of various trometry (ICPOES); As, Sb, Bi and Hg by atomic fluorescence spectrom-
geological processes (Cheng, 2012). Geochemical landscapes are prod- etry (AFS), and W and Mo by inductively coupled plasma-mass
ucts of the superimposition of lithology, landforms, weathering, and spectrometry (ICPMS). Fig. 5 shows the spatial distribution of Cu, Pb,
other processes, and thus consist of mixed and complex patterns. Differ- Zn and Au with a resolution of 100 m created by the multifractal inverse
ent subareas for a particular study area are underlain by different rock distance weighting (MIDW) interpolation method (Cheng, 2000a) with
types and different geological processes are typically characterized by support of GeoDAS GIS software (Cheng, 2000b). The high concentra-
different geochemical backgrounds. The identification of geochemical tion values of Cu and Au mainly occur in the Nanyuan Group where
anomalies from the mixed pattern, linked with different geochemical
backgrounds in a complex study area, is a challenge for the classic
statistics based on the frequency of geochemical variable. Recent studies
have demonstrated that fractal and multifractal models which involve
both the frequency and spatial variances of geochemical data are power-
ful tools to detect multivariate geochemical patterns or signals associated
with mineralization within a complex geological background (e.g., Zuo,
2011a, 2011b; Zuo et al., 2009). For instance, the Gangdese belt, in
Tibet (China), a world-class Cu polymetallic belt, can be classified
into three mineralized subzones in terms of their geotectonic back-
ground. Each subzone is underlain by different geological processes
and thus has a distinct geochemical background. In such a complex
area, the classic statistical method (such as mean + 2 standard devia-
tion) has limitations in identifying the weak anomalies hidden in
the low geochemical background. Zuo et al. (2009) and Zuo (2011b) ap-
plied the multifractal singularity technique (Cheng, 2007) and the
spectrum-area (S-A) fractal model (Cheng et al., 2000) to recognize geo-
chemical anomalies associated with Cu polymetallic mineralization in
this belt.
The Fanshan district has a high potential for epithermal-type Cu–Au
mineralization similar to the large epithermal Zijinshan Cu–Au
polymetallic mineralization developed in the neighboring region (Lin
and Zheng, 2013). In this study, the robust principal component analysis
(RPCA) coupled with the isometric logratio (ilr) transformation and S-A
model were used to identify geochemical anomalies associated with
mineralization. The RPCA coupled with the isometric logratio (ilr) trans-
formation was used to open the closed data and to detect multivariate
geochemical patterns associated Cu–Au mineralization. The selected
PC related with mineralization was further decomposed using the S-A
model to explore for Cu–Au mineralization.

2. Study area and data

The study area, Fanshan district (24°10′–24°20′N, 117°15′–117°30′

E), is situated in the southern Fujian Province, southeast China. It is un-
derlain by the Jurassic Nanyuan Group and Lishan Group, Upper Jurassic
intrusions and volcanic rocks, Cretaceous Huangken Group and Zhaixia
Group (Lin and Zheng, 2013). The Nanyuan Group and Huangken Group
are mainly composed of dacite and rhyolite, respectively, and host Cu,
Au and other deposits in this region (Lin and Zheng, 2013). The main
faults in the study area have NW and NE orientations and control the
distribution of intrusions and volcanic vents (Fig. 3) (Lin and Zheng,
2013). Secondary quartzlithification was developed in this district and Fig. 6. Biplots of the PC1 and PC2 obtained by the classical PCA (a) and RPCA (b) methods.
174 R. Zuo / Journal of Geochemical Exploration 139 (2014) 170–176
Dafanshan and Xiaofanshan are located (Fig. 5a and d). The spatial dis-
tribution of Pb is similar to Zn, and their high values mostly develop
around the intrusions (Fig. 5b and c).
3. Results and discussion
3.1. Data closure problem and selection of PC associated with mineralization
The classic PCA method is based on classic variances or covariance
matrices which are sensitive to extreme values, and thus are strongly bi-
ased by outliers. The RPCA method is designed to reduce the influences
of such values based on the minimum covariance determinant
estimator instead of the classic covariance matrix and thus can reduce
the effects of outliers. Moreover, geochemical data are typically compo-
sitional data and should be opened prior to analysis (Filzmoser et al.,
2009; Zuo et al., 2013a). The RPCA method was used to combine multi-
ple geochemical variables using the R package robCompositions (Templ
et al., 2012). Here, the RPCA method firstly opens the raw data using the
isometric logratio (ilr) transformation, developed by Egozcue et al.
(2003) to reduce the effects of the data closure problem, and then
combines multiple geochemical variables (Zuo et al., 2013a).
Fig. 7. Map showing the spatial distribution of the first principal component of the logratio transformed data (a); log–log plot of spectrum density (S) versus the number of cells with
greater than or equal to S (log base e) (b).
The PC1 and PC2 obtained through the classic PCA account for 21.7%
and 14.3% of the total variances, respectively. The relationship between
PC1 and PC2 obtained through the classic PCA (Fig. 6a) illustrates that
PC1 is composed of negative loadings from all the elements and PC2 is
composed of positive loadings from W, Mo, Bi, Sn, Mn, Pb, and Hg, and
negative loadings from As, Au, Ag, Sb, Cu, and Zn. The composition of
PC1 obtained via the classic PCA method shows that all the geochemical
elements have positive contributions and all the sample points are
located around zero, which is a consequence of the effects of data
closure.
PC1 and PC2 obtained through RPCA account for 16.4% and 13.4% of
the total variance. The composition of PC1 obtained through RPCA
method shows two different assemblage compositions (Fig. 6b). The
first assemblage composed of positive loadings from Pb, Zn, Sn, W,
Mo, Bi, Hg, and Ag, perhaps represents hydrothermal-type Zn–Pb
polymetallic mineralization that is associated with intrusions and
NE and NW trending faults. The second assemblage composed of
negative loadings from As, Au, Cu, Sb, and Mn, possibly represents
epithermal-type Cu–Au mineralization, which is similar to the min-
eralization in the Zijinshan Cu–Au deposit developed in the neigh-
boring region.
 R. Zuo / Journal of Geochemical Exploration 139 (2014) 170–176 175

3.2. Identification geochemical patterns associated with mineralization
Fig. 7a shows the spatial distribution map of PC1 obtained through
the RPCA method. Low PC1 scores linked with the assemblages of Cu,
Au, As, Sb, and Mn, is mainly observed in the Nanyuan Group. Large
scores of PC1, related to the assemblage of Pb, Zn, W, Mo, Sn, Bi, Hg,
and Ag is mostly developed within the intrusions (Fig. 7a). The map of
PC1 should be further studied because it is a product of multiple geolog-
ical processes such as lithology, landforms, and mineralization. In this
study, the S-A fractal model (Cheng et al., 2000), a powerful tool to
decompose the mixed pattern in the field of applied geochemistry
(e.g., Cheng et al., 2010; Zuo, 2011a, 2011b; Zuo and Xia, 2009; Zuo
et al., 2012, 2013a, 2013b), was used to further study PC1. The PC1 pat-
tern (Fig. 7a) was firstly converted into the frequency domain using the
fast Fourier transformation. A pair of data consisting of the power spec-
trum density and phase can be obtained. The spectral density (S) and
the number of cells with values greater than or equal to S were plotted
on a log–log graph. Three straight lines with two cutoff values, obtained
from the least squares method (Fig. 7b), can be fitted using these pairs
of data components. The left-hand line gives y = −1.34x + 15.40
(Ln S ≤ 7.56, or S ≤ 1913.4) and represents the noise component
(Cheng et al., 2010). The middle line represents the anomaly and gives
y = −1.80x + 18.74 (7.56 b Ln S b 8.08, or 1913.4 b S b 3233.84).
The right-hand line is for background and gives y = −2.13x + 21.43
(Ln S ≥ 8.08, or S ≥ 3233.84). These three frequency components
were converted back into the spatial domain using the inverse Fourier
transformation and three spatial patterns: noise, anomaly (Fig. 8a),
and background maps (Fig. 8b).
The anomaly map (Fig. 8a) shows that most of the highly anomalous
areas, where the majority of known Zn–Pb occurrences are located,
occur in the vicinity of intrusions, indicating that the anomalies are
linked with hydrothermal-type Zn–Pb polymetallic mineralization.

Fig. 8. The decomposed maps for anomaly map (a) and background component (b).
176 R. Zuo / Journal of Geochemical Exploration 139 (2014) 170–176
However, the lower anomaly areas, such as Dafanshan and Xiaofanshan,
occur mainly within Nanyuan Group where developed volcanic vents,
are associated with Cu–Au polymetallic mineralization. The background
map (Fig. 8b) shows that Nanyuan Group is characterized by low Zn and
Pb, but high Cu and Au contents, whereas the intrusions in the study
area have high concentrations of Zn and Pb, and can provide the energy
and metal source contributions for the formation of hydrothermal-type
Zn–Pb deposits.
4. Conclusions
The results of this study led to the following conclusions:
(a) Due to the effects of compositional data closure, geochemical
data should initially be opened before being analyzed; and
(b) The geochemical characteristics of Fanshan district show two
different geochemical assemblage compositions. The first one
possibly represents hydrothermal-type Zn–Pb polymetallic min-
eralization, characterized by large values of Pb, Zn, Sn, W, Mo, Bi,
Hg, and Ag, which are associated with intrusions and faults. The
second may represent epithermal-type Cu–Au mineralization,
linked to As, Au, Cu, Sb and Mn, and should be further investigat-
ed to explore for Zijinshan-type mineralization in the Nanyuan
Group where Au and Cu are enriched.
Acknowledgments
The author thanks Prof. Changjiang Li (guest editor) and two anon-
ymous reviewers for constructive comments and suggestions, and
Boxing Wu from the Geological Survey Institute of Fujian (China) for
providing geological and geochemical data and helping in the field
work. This study is a part of the program for integrated prediction of
mineral deposits in covered areas. This research benefited from the
joint financial support from a research project on “Quantitative models
for prediction of strategic mineral resources in China” (201211022) by
the China Geological Survey, the Fundamental Research Funds for the
Central Universities, China University of Geosciences (Wuhan)
(CUG120501, CUG120116), the National Natural Science Foundation
of China (Nos. 41002118, 41372007), and a Program for New Century
Excellent Talents in University.
References
Aitchison, J., 1982. The statistical analysis of compositional data (with discussion). J. R.
Stat. Soc. B 44, 139–177.
Aitchison, J., 1986. The Statistical Analysis of Compositional Data. Chapman & Hall,
London, UK (416 pp.).
Aitchison, J., Barceló-Vidal, C., Martín-Fernández, J., Pawlowsky-Glahn, V., 2000. Logratio
analysis and compositional distance. Math. Geol. 32, 271–275.
Buccianti, A., Pawlowsky-Glahn, V., 2005. New perspectives on water chemistry and com-
positional data analysis. Math. Geol. 37, 703–729.
Cheng, Q., 2000a. Interpolation by means of multiftractal, kriging and moving average
techniques. GeoCanada 2000, Proc. GAC/MAC Meeting, Calgary, AB, Canada
[CDROM]. 29 May–2 June 2000. Geol. Assoc. Can., St. John's, NF, Canada.
Cheng, Q., 2000b. GeoData Analysis System (GeoDAS) for mineral exploration:
unpublished User's guide and exercise manual. Material for the Training Workshop
on GeoDAS held at York University. (204 pp.).
Cheng, Q., 2007. Mapping singularities with stream sediment geochemical data for
prediction of undiscovered mineral deposits in Gejiu, Yunnan Province, China. Ore
Geol. Rev. 32, 314–324.
Cheng, Q., 2012. Singularity theory and methods for mapping geochemical anomalies
caused by buried sources and for predicting undiscovered mineral deposits in
covered areas. J. Geochem. Explor. 122, 55–70.
Cheng, Q., Xu, Y., Grunsky, E., 2000. Integrated spatial and spectrum method for geochem-
ical anomaly separation. Nat. Resour. Res. 9, 43–51.
Cheng, Q., Xia, Q., Li, W., Zhang, S., Chen, Z., Zuo, R., Wang, W., 2010. Density/area power-
law models for separating multi-scale anomalies of ore and toxic elements in stream
sediments in Gejiu mineral district, Yunnan Province, China. Biogeosciences 7,
3019–3025.
Egozcue, J.J., Pawlowsky-Glahn, V., Mateu-Figueras, G., BarceloVidal, C., 2003. Isometric
logratio transformations for compositional data analysis. Math. Geol. 35, 279–300.
Filzmoser, P., Hron, K., Reimann, C., 2009. Univariate statistical analysis of environmental
(compositional) data: problems and possibilities. Sci. Total. Environ. 407, 6100–6108.
Gabriel, K.R., 1971. The biplot graphic display of matrices with application to principal
component analysis. Biometrika 58, 453–467.
Lin, D., Zheng, Z., 2013. The country rock alteration characteristics and ore finding poten-
tial analysis in the Fanshan of Pinghe County, Fujian Province. Fujian Geol. 1–5
(In Chinese with English abstract).
Palarea-Albaladejoa, J., Martín-Fernández, J.A., 2013. Values below detection limit in
compositional chemical data. Anal. Chim. Acta 764, 32–43.
Reimann, C., Filzmoser, P., Garrett, R.G., Dutter, R., 2008. Statistical Data Analysis
Explained: Applied Environmental Statistics with R. Wiley, Chichester, UK
(362 pp.).
Templ, M., Hron, K., Filzmoser, P., 2012. robCompositions: robust estimation for
compositional data, R package version 1.2 http://cran.r-project.org/web/packages/
robCompositions/index.html (Accessed March 8, 2013).
Xi, X., Li, M., 2012. Regional geochemical exploration in China: from 1999 to 2009. Geol.
China 39, 267–282 (In Chinese with English abstract).
Xie, X., Wang, X., Zhang, Q., Zhou, G., Cheng, H., Liu, D., Cheng, Z., Xu, S., 2008. Multiscale
geochemical mapping in China. Geochem. Explor. Environ. Anal. 8, 333–341.
Zuo, R., 2011a. Decomposing of mixed pattern of arsenic using fractal model in Gangdese
belt, Tibet, China. Appl. Geochem. 26, S271–S273.
Zuo, R., 2011b. Identifying geochemical anomalies associated with Cu and Pb–Zn skarn
mineralization using principal component analysis and spectrum-area fractal model-
ing in the Gangdese Belt, Tibet (China). J. Geochem. Explor. 111, 13–22.
Zuo, R., 2013. Compositional analysis in the study of mineralization based on stream
sediment data. In: Pardo-Igúzquiza, E., et al. (Ed.), Mathematics of Planet Earth,
Earth Lecture Notes in Earth System Sciences, 22, pp. 91–94. http://dx.doi.org/10.1007/
978-3-642-32408-6.
Zuo, R., Xia, Q., 2009. Application fractal and multifractal methods to mapping
prospectivity for metamorphosed sedimentary iron deposits using stream sediment
geochemical data in eastern Hebei province, China. Geochim. Cosmochim. Acta 73,
A1540.
Zuo, R., Cheng, Q., Agterberg, F.P., Xia, Q., 2009. Application of singularity mapping tech-
nique to identification local anomalies using stream sediment geochemical data, a
case study from Gangdese, Tibet, Western China. J. Geochem. Explor. 101, 225–235.
Zuo, R., Carranza, E.J.M., Cheng, Q., 2012. Fractal/multifractal modelling of geochemical
exploration data. J. Geochem. Explor. 122, 1–3.
Zuo, R., Xia, Q., Wang, H., 2013a. Compositional data analysis in the study of integrated
geochemical anomalies associated with mineralization. Appl. Geochem. 28, 202–211.
Zuo, R., Xia, Q., Zhang, D., 2013b. A comparison study of the C-A and S-A models with
singularity analysis to identify geochemical anomalies in covered areas. Appl.
Geochem. 33, 165–172.