Journal of Geochemical Exploration 136 (2014) 93–101

Contents lists available at ScienceDirect

Journal of Geochemical Exploration

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

Identification of weak geochemical anomalies using robust

neighborhood statistics coupled with GIS in covered areas
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: Robust neighborhood statistics, such as median, median absolute deviation (MAD), and local relative enrichment
Received 24 June 2013 index (LREI), were used to model spatial variations of geochemical landscapes and to recognize weak
Accepted 10 October 2013 geochemical anomalies in covered terrane by means of a case study from the Chaobuleng Fe polymetallic district
Available online 19 October 2013
covered by grassland, in Inner Mongolia (China). The results of this study show that: (1) the robust neighborhood
statistics of median and MAD can well discriminate between different geochemical landscapes (with outcrop
Keywords:
Exploration geochemistry
areas characterized by high values of Fe2O3, and the covered areas linked with a low value of Fe2O3) and delineate
Weak anomalies the boundaries between the outcrop and covered areas; (2) the LREI can reduce the effects of covers such as soils
Robust neighborhood statistics and grassland on geochemical mapping and identify the weak geochemical anomalies caused by decaying and
GIS masking effects of covers; and also (3) the resulting neighborhood statistics are influenced by the size of
Covered area neighborhood.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction However, these methods do not consider the spatial variations of

geochemical patterns, which are the most important characteristics of
Attention has increasingly been focusing on mineral exploration in geochemical landscapes. Fractal/multifractal modeling of geochemical
covered areas due to excellent potential for mineral exploration, as data has become a popular approach for mapping of geochemical data
indicated by the fact that few deposits have been discovered in the and separation of different geochemical populations (e.g., Afzal et al.,
covered areas in comparison with the number of discoveries in the 2010, 2011, 2013; Agterberg, 2012; Bölviken et al., 1992; Cheng, 2007;
outcrop areas. There are three challenges that recently were proposed Cheng et al., 1994, 2000; Zuo and Cheng, 2008; Zuo et al., 2009, 2012,
by Cheng (2012a, 2012b) for mineral exploration in covered areas, 2013b).
including identification of weak anomalies, decomposition of complex The singularity, a multifractal term that is defined as the charac-
and mixing anomalies, and integration of missing or incomplete terization of anomalous behavior of singular physical processes that
information, because of difficulty to observe geological phenomena often result in anomalous amounts of energy release or material accu-
and to collect rock samples owing to covers (Zuo et al., 2013b). It has mulation within narrow spatial-temporal intervals (Cheng, 2007,
been observed that elemental concentrations in stream sediments in 2012a), has been demonstrated to be a powerful tool to identify weak
the covered areas are low due to decaying and masking effects of covers geochemical anomalies (e.g., Arias et al., 2012; Cheng, 2012a, 2012b;
(Cheng, 2012b; Zuo et al., 2013b). Therefore, identification of the weak Xiao et al., 2012; Zuo et al., 2009, 2013b). The singularity index, which
geochemical anomalies is one of the most challenging tasks for the measures the relative changes of elemental concentration within a
processing of geochemical data in covered areas (Cheng, 2012a). small neighborhood, can be regarded as a type of local neighborhood
Delineation of various geochemical anomalies plays a significant role statistic. For a raster map, the local neighborhood statistics calculated
in mineral exploration and environmental studies. Various methods, for each input cell location are values for within a specific neighborhood
consisting of frequency-based and spatial frequency-based methods, around it. The shape of neighborhood, such as annulus, circle, rectangle,
are popular in processing the geochemical data (Zuo et al., 2013a). wedge, irregular and weight kernel files around each cell can be used to
Conventional methods, such as calculating the mean and standard calculate these statistics. Then, the size and unit of distance can be set
derivation (SD), probability graphs (Sinclair, 1974), explorational data to define the shape of neighborhood. The simple statistics, including
analysis (EDA) (Carranza, 2010; Chiprés et al., 2009; Reimann, 2005a; mean, majority, maximum, median, minimum, minority, range, SD,
Reimann, 2005b; Tukey, 1977), and multivariate data analysis (Yousefi, sum, and variance, can be calculated. Meanwhile, more complex statistics
et al., 2012; Zuo, 2011a, 2011b; Zuo et al., 2009; Zuo et al., 2013a), have like the coefficient of variation (CV = mean/SD) also can be constructed
been widely applied in various fields of geochemical exploration. from these simple statistics.
The local neighborhood statistics unlike global statistics, which
⁎ Tel./fax: +86 27 67885096. consider all the values of a dataset, involve only data from within the
E-mail address: [email protected]. small neighborhood around a particular spatial location and can reveal

0375-6742/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.gexplo.2013.10.011
94 R. Zuo / Journal of Geochemical Exploration 136 (2014) 93–101

more about the relationship between the specific location and its
neighbors (Zhang et al., 2007). In addition, the local neighborhood sta-
tistics can provide less statistical information which is less biased than
that of global statistics, such as mean and SD, because geochemical data
generally satisfy non-normal distributions and contain outliers (Zhang
et al., 2007). The neighborhood statistics have been demonstrated to be
an effective way for exploring the spatial variations of geochemical
patterns and for identification of boundaries between different rock
types (Zhang et al., 2007). In this study, neighborhood statistics coupled
with GIS were used to model the spatial variation of Fe2O3, and to identify
the weak geochemical anomalies in covered areas.

2. Neighborhood statistics

2.1. Robust neighborhood statistics

Simple neighborhood statistics, like mean and SD, are often used in
different branches of applied geochemistry. For example, the local
mean of the cells in a neighborhood is used to measure the local
background of elemental concentration. The SD and CV are suitable pa-
rameters to quantify the spatial variations of geochemical variables.
Areas linked with similar geochemical background may show small
values of SD and CV; whereas, areas characterized by distinctly different
geochemical background may yield large values of these statistics.
Zhang et al. (2007) observed that the local mean influenced the value
of SD but did not influence the CV, suggesting CV as a useful index to
quantify the spatial variation of geochemical variables. However, these
statistics are sensitive to outliers; therefore, robust neighborhood
statistics, such as median and median absolute deviation (MAD), should
be applied. The MAD is a robust estimator of SD calculated using Eq. (1)
(Reimann et al., 2008; Rousseeuwa and Crouxa, 1993).

MADi ¼ 1:4826  mediani ðjMEDIANi −xi jÞ ð1Þ

The robust coefficient of variation (RCVi) is defined as

MADi
RCV i ¼ ð2Þ
MEDIANi

where xi represents the element concentration at a location of zi and

Fig. 1. Two artificial examples: an enriched geochemical pattern (A), and a depleted
MEDIANi represents the value of median with specific small neighboring
geochemical pattern (B) (after Zuo et al., 2009).
cells around zi.

2.2. Local relative enrichment index

like the singularity index, can detect the enrichment and depletion of
element concentration. The value of LREIi for the enriched pattern
The local relative enrichment index (LREIi) at location of zi can be
(Fig. 1A) increases and for the depleted pattern (Fig. 1B) it decreases
defined as
with increasing size of rectangle neighborhood from 3 × 3 to 9 × 9
xi −MEDIANi xi (Fig. 2), indicating that the value of LREIi is influenced by the
LREIðzi Þ ¼ ¼ −1: ð3Þ neighborhood size. Furthermore, by replacing the MEDIANi with MEANi
MEDIANi MEDIANi
in Eq. (3), the values of LREIi were calculated as 0.24 and −0.47 at the
The conditions of LREIi N 0 and LREIi b 0 represent local enrichment central locations of Fig. 1A and Fig. 1B with a window size of 3 × 3,
and depletion of element concentration, respectively. A large positive respectively. For the enriched and depleted maps, the values of LREIi
value of LREIi indicates a high degree of enrichment. Two artificial calculated using robust neighborhood statistics were observed as larger
geochemical patterns (Fig. 1, after Zuo et al., 2009) were constructed and smaller than the values of LREIi estimated through classic neigh-
as examples to demonstrate the usage of the LREIi compared with the borhood statistics, respectively (Fig. 2). This suggests that the LREIi
singularity index. Singularity indexes were observed as 1.42 (b 2) and estimated using the robust neighborhood statistics is much more
2.96 (N2), representing an enrichment (Fig. 1A) and a depletion pattern sensitive to detect geochemical anomalies than the value of LREIi cal-
(Fig. 1B) of element concentration, respectively (Zuo et al., 2009). culated using the classic neighborhood statistics.
Rectangular neighborhoods of 3 × 3, 5 × 5, 7 × 7, and 9 × 9 were used
to calculate the LREIi for the central location of Fig. 1A and B. For the 3. Case study and data
enriched pattern, at the central location of Fig. 1A the element
concentration value was found to be 10, and the median value with a In this study, to compare the results obtained using the robust
rectangular neighborhood of 3 × 3 gave 8, yielding a positive value of neighborhood statistical method with those obtained through the sin-
LREIi (0.25 N 0). Meanwhile, the value of LREIi for the depleted pattern gularity mapping technique, data from literature were used (Zuo et al.,
at the central location of Fig. 1B was observed as – 0.5 (b0) with a 2013b). The study area, Chaobuleng Fe polymetallic district in Inner
rectangular neighborhood of 3 × 3. These results suggest that the LREIi, Mongolia of northern China, covered by grassland, was chosen as the
 R. Zuo / Journal of Geochemical Exploration 136 (2014) 93–101 95

A case study area. The study area is composed of 40.2% Quaternary

sediments (Q), 33.4% Tertiary sediments (N), 17.0% granites, and 9.4% of
local relative enrichment index

1.6 other formations (Fig. 3). More detailed information on geology and
mineral deposit can be found from Zuo et al. (2013b).
1.2 The stream sediment data of Fe2O3, at a density of one sample per
4 km2, were collected from Chinese National Geochemical Mapping
(CNGM) project including 39 major and trace geochemical elements,
0.8 to demonstrate the usage of robust neighborhood statistics. The X-ray
fluorescence technique was used to determine the value of Fe2O3 with
detection limit of 0.05%. More information about sampling, analysis
0.4
and quality of the geochemical data are available in literature (Xie
et al., 1997, 2008). The spatial distribution of Fe2O3 with resolution of
0 1 km × 1 km created using the multifractal inverse distance weighted
1 3 5 7 9 11 method (Cheng, 1999) with support of GeoDAS software (Cheng,
The size of neighborhood 2000) (Fig. 4), shows red patterns (large values) in outcrop areas hosted
Median Mean by granites and formations (mapped 26.4% of total study area), and blue
patterns (small values) in the covered areas composed of Quaternary
B sediments and Tertiary sediments (mapped of 73.6% of study area).
These patterns indicate the presence of a low concentration value of
local relative enrichment index

-0.4
Fe2O3 in the covered areas due to the decaying and masking effects of
covers (Zuo et al., 2013b).

-0.6
4. Results

4.1. Spatial variation of Fe2O3

-0.8
Robust neighborhood statistics of local MEDIANi, local MADi and local
RCViwere calculated using square windows of 3km×3km, 5km×5km,…,
-1 and 31 km × 31 km. The 3 km × 3 km and 31 km × 31 km neighborhoods
1 3 5 7 9 11 occupy 0.028% and 2.98% of the total study area, respectively. The
The size of neighborhood quantile classification was set to display color geochemical maps. The
local median patterns with sizes of neighborhoods of 5 km × 5 km,
Median Mean
7 km × 7 km, 9 km × 9 km, and 11 km × 11 km (Fig. 5) were observed to
Fig. 2. Plots of the size of neighborhood versus the local relative enrichment index for the be similar to the original map of Fe2O3 that mainly reflects two distinct
enriched pattern (A) and depleted pattern (B). geochemical landscapes: the outcrop areas characterized by a large
value of Fe2O3, and the covered areas linked to a small value of Fe2O3.

Fig. 3. Simplified geological map of Chaobuleng district, Inner Mongolia (China).

Modified from GSIIM, 2010 and Zuo et al., 2013b.
96 R. Zuo / Journal of Geochemical Exploration 136 (2014) 93–101

Fig. 4. Map showing the spatial distribution of Fe2O3.

After Zuo et al., 2013b.

The median pattern becomes smooth and the range of median narrows were found to be positive, representing element enrichment (Zuo
when the size of neighborhood increases from 5 km × 5 km to et al., 2013b).
11 km × 11 km because the larger window size occupies a larger area Both points 1 and 4 provided smaller values of median than those of
involving more cells probably representing the regional geochemical points 2 and 3 when the size of neighborhood was smaller than
background. 17km×17km (Fig. 9A), probably because points 1 and 4 contain smaller
The MADi patterns, estimated using neighborhood sizes of concentration of Fe2O3 due to covers. When the size of window was
5 km × 5 km, 7 km × 7 km, 9 km × 9 km, and 11 km × 11 km, mainly increased, the neighborhoods around points 1 and 4 encompass many
highlight the boundaries between various geochemical landscapes more cells overlain in the outcrop areas linked to a large value of
(Fig. 6) and exhibit a strong spatial variation of Fe2O3 occurring in or Fe2O3. Additionally, the neighborhoods around points 2 and 3 encompass
near the boundaries between the outcrop areas and the covered areas. a larger area over the covered areas linked to a small concentration of
Low spatial variation of Fe2O3 was observed within the outcrop and the Fe2O3, resulting in the value of median at the location of point 4 greater
covered areas. The MADi pattern also became smoother and the range than the value of median at point 2 (Fig. 9A). The MADi and RCVi values
of MADi became narrower when the neighborhood size increased from at points 1, 2 and 4 increased when the size of window increased,
5 km × 5 km to 11 km × 11 km. Comparing the RCVi pattern (Fig. 7) with indicating an increasing spatial variation of geochemical pattern around
the MADi pattern created using the same window size of 9 km × 9 km, points 1, 2, and 4. The MADi and RCVi values at point 3 initially increased
RCVi displayed a clearer pattern than that of MADi, suggesting that RCVi and then decreased when the size of window was increased (Fig. 9B and
is a better parameter to model the spatial variations of geochemical C), reflecting a complex spatial variation of Fe2O3 around point 3.
patterns and to detect the boundaries between distinct geological The values of LREIi at points 1 and 4 were observed to be below zero,
patterns. representing depleted patterns, whereas the values of LREIi at points 2
and 3 were greater than zero, representing enriched patterns. This
4.2. Identification of weak anomalies suggests that the LREIi, like the singularity index, can well identify
weak anomalies in the covered areas. When the size of the neigh-
The spatial distribution of LREIi with the neighborhood size of borhood increased from 3 km × 3 km to 31 km × 31 km, the values of
9 km × 9 km (Fig. 8) was observed to be similar to the singularity map LREIi at points 1 and 4 decreased and the values of LREIi at points 2
produced by Zuo et al. (2013b). It was observed that the LREIi pattern and 3 increased. The value of LREIi at point 2 was found to be larger
within outcrop areas had no difference with the LREIi pattern within and smaller than that at point 3 when the window size was smaller
the covered areas, suggesting that the LREIi can remove the effects of and larger than 19 km × 19 km, respectively. These results indicate
distinct geochemical background. The LREIi is thus a useful tool to that the resulting robust neighborhood statistics are influenced by
identify weak geochemical anomalies through eliminating the effects the window size. Therefore, the values of LREIi were comparable
of covers. Four points (1–4, Zuo et al., 2013b) were selected to compare when the same neighborhood sizes were used.
the values of LREIi estimated using various window sizes. Points 1 and 2
were selected from the covered areas linked to weak geochemical 5. Discussion
backgrounds, and points 3 and 4 were selected from the outcrop areas
linked to strong geochemical backgrounds. The singularity values of Neighborhood statistics provide an important kind of spatial analytic
points 1 and 4 were observed to be negative, representing elemental tool in support of GIS. They are not only simple and easy to implement,
concentration depletion; while the singularity values of points 2 and 3 but also can well detect the spatial variations of geochemical landscapes
 R. Zuo / Journal of Geochemical Exploration 136 (2014) 93–101 97

Fig. 5. Maps showing the spatial distribution of median for Fe2O3 calculated with various sizes of neighborhoods: (A) 5 km × 5 km, and (B) 11 km × 11 km.

and identify weak geochemical anomalies. These techniques can be also disadvantage of neighborhood statistics is that the obtained results are
applied in many other fields. functions of the size of neighborhood used. The window size is a
Neighborhood statistics involving data within small neighborhoods significant factor in neighborhood statistics. On the one hand, the
have the advantages of revealing more detailed local structures and neighborhood should contain enough data to make the local statistics
spatial variations of geochemical patterns, and of reducing the effects valid. On the other hand, a large window size should be avoided because
of various geochemical backgrounds. Therefore, they are useful to the local neighborhood statistics could become global statistics and the
identify the weak geochemical anomalies in the covered areas. The local structures and spatial variations of geochemical variables could not
98 R. Zuo / Journal of Geochemical Exploration 136 (2014) 93–101

Fig. 6. Maps showing the spatial distribution of the median absolute deviation for Fe2O3 calculated with various sizes of neighborhoods: (A) 5 km × 5 km, and (B) 11 km × 11 km.

be detected well when a relatively large window is used (Zhang et al., average value of elements concentration, c is a constant value, a is the
2007). Meanwhile, when the size of neighborhood increases, the singularity index. In this equation ε is a normalized distance measure,
smoothing effect of local neighborhood statistics becomes stronger. such as block cell edge, and E is the Euclidian dimension (Cheng,
Comparing neighborhood statistics with the singularity index, it was 2007). The singularity index a can be estimated from the slope of the
found that the singularity index at a specific location zi is independent of straight line fitted to paired data of X and ε in a log–log graph. It
the element concentration value at the location of zi and the size of measures the gradient of relative change within small neighborhoods.
window used. The singularity index can be estimated within small For a specific location zi, the value of the singularity index does not
neighborhoods using the equation X = c εa − E, where X represents the depend on the element concentration at that location, but relies on
 R. Zuo / Journal of Geochemical Exploration 136 (2014) 93–101 99

Fig. 7. Map showing the spatial distribution of the robust coefficient of variation for Fe2O3 calculated using a window size of 9 km × 9 km.

the changes around zi. When this change is large, the absolute value of Geochemical data usually include outliers. Conventional statistics,
the fitted slope is large; and vice versa. From the multifractal theory such as mean and SD, are sensitive to outliers. Therefore, robust statistics
viewpoint, the singularity index is also uncorrelated to the size of like the median, MAD and RCV, should be applied. However, few case
neighborhood, meaning that it is independent of scale. Therefore, it is studies on application of robust neighborhood statistics to process geo-
a powerful tool to identify geochemical anomalies regardless of geo- chemical data analysis in support of GIS have been reported probably
chemical background and can detect deep information because of its because commercial GIS software, like ArcGIS that includes only median
independence of scale. Although the resulting neighborhood statistics function, do not provide other robust statistical functions. Therefore,
are influenced by the elemental concentration and size of window, they robust statistical scripts should be developed in the commercial GIS
are easily implemented and understood. software.

Fig. 8. Map showing the spatial distribution of the local relative enrichment index for Fe2O3 calculated using a window size of 9 km × 9 km.
100 R. Zuo / Journal of Geochemical Exploration 136 (2014) 93–101

A No.1 No.2 No.3 No.4

B No.1 No.2 No.3 No.4

Median absolute deviation

6 2
5 1.6
4
Median

1.2
3
0.8
2
1 0.4

0 0
1 5 9 13 17 21 25 29 33 1 5 9 13 17 21 25 29 33
The size of neighborhood (km) The size of neighborhood (km)

C D

Local relative enrichment index

No.1 No.2 No.3 No.4 No.1 No.2 No.3 No.4
robust coefficient of variation

0.6 1.2

0.8
0.4
0.4

0
0.2
-0.4

0 -0.8
1 5 9 13 17 21 25 29 33 1 5 9 13 17 21 25 29 33
The size of neighborhood (km) The size of neighborhood (km)

Fig. 9. Plots of the size of neighborhood versus the median (A), median absolute deviation (B), robust coefficient of variation(C) and local relative enrichment index (D) for various
locations.

6. Conclusions References

Afzal, P., Khakzad, A., Moarefvand, P., Rashidnejad Omran, N., Esfandiari, B., Fadakar
In this paper, robust neighborhood statistics such as MEDIANi, MADi, Alghalandis, Y., 2010. Geochemical anomaly separation by multifractal modeling in
and LREIi were introduced to explore the spatial variation of Fe2O3, to Kahang (Gor Gor) porphyry system, Central Iran. J. Geochem. Explor. 104, 34–46.
identify weak anomalies in covered areas, and for comparison with Afzal, P., Fadakar Alghalandis, Y., Khakzad, A., Moarefvand, P., Rashidnejad Omran, N.,
2011. Delineation of mineralization zones in porphyry Cu deposits by fractal
the singularity index. The following conclusions were obtained: concentration-volume modeling. J. Geochem. Explor. 18, 220–232.
Afzal, P., Dadashzadeh Ahari, H., Rashidnejad Omran, N., Aliyari, F., 2013. Delineation of
(1) robust neighborhood statistics can be used to explore more gold mineralized zones using concentration-volume fractal model in Qolqoleh gold
deposit, NW Iran. Ore Geol. Rev. 55, 125–133.
detailed local structures and spatial variations of geochemical Agterberg, F.P., 2012. Multifractals and geostatistics. J. Geochem. Explor. 122, 113–123.
landscapes; Arias, M., Gumiel, P., Martín-Izard, C., 2012. Multifractal analysis of geochemical anomalies: a
(2) the LREIi can reduce the decaying and masking effects of covers tool for assessing prospectivity at the SE border of the Ossa Morena Zone, Variscan
Massif (Spain). J. Geochem. Explor. 122, 101–112.
and identify weak anomalies in covered areas; and
Bölviken, B., Stokke, P.R., Feder, J., Jossang, T., 1992. The fractal nature of geochemical
(3) the resulting neighborhood statistics depend on the window landscapes. J. Geochem. Explor. 43, 91–109.
size. Carranza, E.J.M., 2010. Mapping of anomalies in continuous and discrete fields of stream
sediment geochemical landscapes. Geochem. Explor. Environ. Anal. 10, 171–187.
Cheng, Q., 1999. Multifractal interpolation. In: Lippard, S.J., et al. (Ed.), Proc. Annu. Conf.
Int. Assoc. for Mathematical Geol., Trondheim, Norway. 6–11 Aug. 1999. , vol. 1.
Acknowledgments Norw. Univ. of Sci. and Technol, Trondheim, pp. 245–250.
Cheng, Q., 2000. GeoData Analysis System (GeoDAS) for mineral exploration: unpublished
user's guide and exercise manual. Material for the Training Workshop on GeoDAS
The author thanks Frits Agterberg (associate editor for Journal of Held at York University. 204.
Geochemical Exploration) for improving this work, Peyman Afzal, and Cheng, Q., 2007. Mapping singularities with stream sediment geochemical data for
prediction of undiscovered mineral deposits in Gejiu, Yunnan Province, China. Ore
an anonymous reviewer for constructive comments and suggestions, Geol. Rev. 32, 314–324.
Dr. Daojun Zhang and Mr. Liheng Chang from the China University of Cheng, Q., 2012a. Singularity theory and methods for mapping geochemical anomalies
Geosciences (Wuhan) for processing a part of the data. This study is a caused by buried sources and for predicting undiscovered mineral deposits in
covered areas. J. Geochem. Explor. 122, 55–70.
part of the program for the integrated prediction of mineral deposits Cheng, Q., 2012b. Ideas and methods for mineral resources integrated prediction in
in covered areas. This research benefited from the joint financial covered areas. Earth Sci.-J. China Univ. Geosci. 37, 1109–1125 (in Chinese with
support from a research project on “Quantitative models for prediction English abstract).
Cheng, Q., Agterberg, F.P., Ballantyne, S.B., 1994. The separation of geochemical anomalies
of strategic mineral resources in China” (201211022) by the China
from background by fractal methods. J. Geochem. Explor. 51, 109–130.
Geological Survey, the Fundamental Research Funds for the Central Cheng, Q., Xu, Y., Grunsky, E., 2000. Integrated spatial and spectrum method for geochemical
Universities, China University of Geosciences(Wuhan) (Nos. CUG120501 anomaly separation. Nat. Resour. Res. 9, 43–51.
and CUG120116), the National Natural Science Foundation of China Chiprés, J.A., Castro-Larragoitia, J., Monroy, M.G., 2009. Exploratory and spatial data
analysis (EDA-SDA) for determining regional background levels and anomalies of
(No. 41372007), and the Program for New Century Excellent Talents in potentially toxic elements in soils from Catorce-Matehuala, Mexico. Appl. Geochem.
University (NCET-13-1016). 24, 1579–1589.
 R. Zuo / Journal of Geochemical Exploration 136 (2014) 93–101 101

Reimann, C., 2005a. Geochemical mapping: technique or art? Geochem. Explor. Environ. Zhang, C., Jordan, G., Higgins, A., 2007. Using neighbourhood statistics and GIS to quantify
Anal. 5, 359–370. and visualize spatial variation in geochemical variables: an example using Ni
Reimann, C., 2005b. Sub-continental-scale geochemical mapping: sampling, quality concentrations in the topsoils of Northern Ireland. Geoderma 137, 466–476.
control and data analysis issues. Geochem. Explor. Environ. Anal. 5, 311–323. Zuo, R., 2011a. Decomposing of mixed pattern of arsenic using fractal model in Gangdese
Reimann, C., Filzmoser, P., Garrett, R.G., Dutter, R., 2008. Statistical data analysis explained. Belt, Tibet, China. Appl. Geochem. 26, S271–S273.
Applied Environmental Statistics with R.Wiley, Chichester, UK 343. Zuo, R., 2011b. Identifying geochemical anomalies associated with Cu and Pb–Zn skarn
Rousseeuwa, P.J., Crouxa, C., 1993. Alternatives to the Median Absolute Deviation. J. Am. mineralization using principal component analysis and spectrum–area fractal
Stat. Assoc. 88, 1273–1283. modeling in the Gangdese Belt, Tibet (China). J. Geochem. Explor. 111, 13–22.
Sinclair, A.J., 1974. Selection of threshold values in geochemical data Using probability Zuo, R., Cheng, Q., 2008. Mapping singularities — a technique to identify potential Cu
graphs. J. Geochem. Explor. 3, 129–149. mineral deposits using sediment geochemical data, an example for Tibet, west
Tukey, J.W., 1977. Exploratory Data Analysis. Addison-Wesley, Reading, USA. China. Mineral. Mag. 72, 531–534.
Xiao, F., Chen, J., Zhang, Z., Wang, C., Wu, G., Agterberg, F.P., 2012. Singularity mapping Zuo, R., Cheng, Q., Agterberg, F.P., Xia, Q., 2009. Application of singularity mapping
and spatially weighted principal component analysis to identify geochemical technique to identify local anomalies using stream sediment geochemical data, a
anomalies associated with Ag and Pb–Zn polymetallic mineralization in Northwest case study from Gangdese, Tibet, Western China. J. Geochem. Explor. 101, 225–235.
Zhejiang, China. J. Geochem. Explor. 122, 101–112. Zuo, R., Carranza, E.J.M., Cheng, Q., 2012. Fractal/multifractal modelling of geochemical
Xie, X., Mu, X., Ren, T., 1997. Geochemical Mapping in China. J. Geochem. Explor. 60, 99–113. data. J. Geochem. Explor. 122, 1–3.
Xie, X., Wang, X., Zhang, Q., Zhou, G., Cheng, H., Liu, D., Cheng, Z., Xu, S., 2008. Multi-scale Zuo, R., Xia, Q., Wang, H., 2013a. Compositional data analysis in the study of
geochemical mapping in China. Geochem. Explor. Environ. Anal. 8, 333–341. integrated geochemical anomalies associated with mineralization. Appl.
Yousefi, M., Kamkar-Rouhani, A., Carranza, E.J.M., 2012. Geochemical mineralization Geochem. 28, 202–211.
probability index (GMPI): a new approach to generate enhanced stream sediment Zuo, R., Xia, Q., Zhang, D., 2013b. A comparison study of the C–A and S–A models with
geochemical evidential map for increasing probability of success in mineral potential singularity analysis to identify geochemical anomalies in covered areas. Appl. Geochem.
mapping. J. Geochem. Explor. 115, 24–35. 33, 165–172.