- Email: [email protected]
The application of geochemical pattern recognition to regional prospecting: A case study of the Sanandaj–Sirjan metallogenic zone, Iran
Journal of Geochemical Exploration 108 (2011) 183–195
Contents lists available at ScienceDirect
Journal of Geochemical Exploration j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j g e o ex p
The application of geochemical pattern recognition to regional prospecting: A case study of the Sanandaj–Sirjan metallogenic zone, Iran Seyed Ahmad Meshkani a,b,⁎, Behzad Mehrabi a, Abdolmajid Yaghubpur a, Younes Fadakar Alghalandis c a b c
Department of Geology, Tarbiat Moallem University, 15614, Tehran, Iran Geological Survey of Iran, P.O.Box: 13185-1494, Tehran, Iran School of Earth Sciences, University of Queensland, Brisbane, Australia
a r t i c l e
i n f o
Article history: Received 5 September 2010 Accepted 28 January 2011 Available online 18 February 2011 Keywords: Cluster analysis K-means Sanandaj–Sirjan metallogenic zone Pb–Zn deposits Spatial pattern Iran
a b s t r a c t In regional exploration programs, the distribution of elements in known mineral deposits can be used as a guide for the classification of deposits, search for new prospects and modeling ore deposit patterns. The Sanandaj–Sirjan Zone (SSZ) is a major metallogenic zone in Iran, containing lead and zinc, iron, gold, copper deposits. In the central part of the SSZ, lead and zinc mineralization is widespread and hitherto exploration has been based on geological criteria. In this study, we used clustering techniques applied to element distribution for classification lead and zinc deposits in the central part of the SSZ. The hierarchical clustering technique was used to characterize the elemental pattern. Elements associated with lead and zinc deposits were separated into four clusters, encompassing both ore elements and their host rock-forming elements. It is shown that lead and zinc deposits in the central SSZ belong to two genetic groups: a MVT type hosted by limestone and dolomites and a SEDEX type hosted by shale, volcanic rocks and sandstone. The results of elemental clustering were used for pattern recognition by the K-means method and the respective deposits were classified into four distinct categories. K-means clustering also reveals that the elemental associations and spatial distribution of the lead and zinc deposits exhibit zoning in the central part of the SSZ. The ratios of ore-forming elements (Sb, Cd, and Zn) vs. (Pb and Ag) show zoning along an E–W trend, while host rockforming elements (Mn, Ca, and Mg) vs. (Ba and Sr) show a zoning along a SE–NW trend. Large and medium deposits occur mainly in the center of the studied area, which justify further exploration around occurrences and abandoned mines in this area. The application of a pattern recognition method based on geochemical data from known mineralization in the central SSZ, and the classification derived from it, uncover elemental zoning, identify key elemental associations for further geochemical exploration and the potential to discover possible target areas for large to medium size ore deposits. This methodology can be applied in a similar way to search for new ore deposits in a wide range of known metallogenic zones. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Geochemical pattern recognition has been used in exploration geochemistry for many years. Since the 1970s, pattern recognition techniques have been applied to recognize the economic mineral resource information hidden in geological and geochemical data (Briggs, 1978; Castillo-Munoz and Howarth, 1976; Collyer and Merriam, 1973; Gustavsson and Bjorklund, 1976; Howarth, 1973) and to establish multivariate geochemical background patterns (Lindqvist et al., 1987). It has also been applied to geochemical hydrocarbon exploration (Granath, 1988), recognition of polluted sites (Hanesch et al., 2001), and to investigate the relations between regional geochemistry and large ore deposits (Xie et al., 2004).
⁎ Corresponding author at: Department of Geology, Tarbiat Moallem University, 15614, Tehran, Iran. Tel.: + 98 21 44011288. E-mail address: [email protected] (S.A. Meshkani). 0375-6742/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.gexplo.2011.01.006
Typical pattern recognition methods used in geochemical exploration consist mainly of discriminant analysis and cluster analysis (Ji et al., 2007). The principal aim of cluster analysis is to partition multivariate observations into a number of meaningful homogeneous multivariate groups. With geochemical data, cluster analysis can be used in different ways: it can be used to cluster the variables (e.g. to detect geochemical relations between the variables) and it can be used to cluster the observations (e.g. to assign samples to certain types) to come to more homogenous data subsets for further data analysis. Furthermore, there are methods that try to group the data by simultaneously clustering objects and variables (Friedman and Meulman, 2004; Ji et al., 1995, 2007; Leisch, 1999; Raftery and Dean, 2004). In this study, multielement chemical analyses of representative samples from the major ore deposits, prospects and showings, were processed and a database created. Using hierarchical cluster analysis, ore- and rock-forming elements were separated. The derived elemental associations were put through K-means clustering and
184
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
the outcomes plotted as scatter plot. Well-separated element groups possess practical geochemical significance, representing distinct types of mineralization, were plotted on a map using ArcGIS. The Sanandaj–Sirjan Zone (SSZ) is the major metallogenic zone in Iran (Fig. 1). The SSZ is a magmatic–metamorphic belt with a NW–SE trend located between the Zagros and Urmieh–Dokhtar volcanic zone of Iran (Stocklin, 1968). Based on its metallogenic characteristics, it can be divided into northern, central and southern subdivisions. The central subdivision is a well-known major Pb–Zn mining district, the Malayer–Esfahan Zone (MEZ) (Ghorbani, 2007; Momenzadeh, 1976). More than 100 ore deposits, mineralization showings and mineral occurrences have been recognized in the MEZ. They are mainly hosted by limestone, dolomite or shale and rarely by volcanic rocks or sandstone of Jurassic to Lower Cretaceous age, and exhibit stratabound, stratiform, layered, vein and lenticular shapes (Burnol, 1968; Momenzadeh, 1976). A conventional approach to classification of the ore deposits of a metallogenic province is based on genetic-age distribution (Meyer, 1981; 1988) and statistical–genetic concept (Cox and Singer, 1986). In
our research, we used a cluster analysis (CA) method for recognition of the Pb–Zn distribution in the MEZ. The hierarchical clustering technique was used to characterize elemental pattern of well-studied Pb–Zn deposit of the MEZ. In this technique all variables were clustered based on inter grouping of similarities or differences (Swan et al., 1995). Collyer and Merriam (1973) used cluster analysis of data from depleted and operating W-mines (in North America) and developed an effective guide for future exploration. Briggs (1978) applied pattern recognition techniques to uranium deposits of the Casper, Wyoming quadrangle and of the Colorado plateau for regional prospecting. Samples from 104 Pb–Zn ore deposits and occurrences within the MEZ were analyzed for 43 elements using ICP-OES techniques by Amdel Laboratories Ltd (Australia). Hierarchical clustering showed two major clusters which are related to ore-forming elements and rock-forming elements respectively. The K-means technique (McQueen, 1967) was employed for grouping ore deposits. The K-means calculation was carried out on each of the elemental groups identified by hierarchical
Fig. 1. Major structural zones of western Iran showing the studied area. Modified from Sahandi et al., 2002.
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
cluster analysis. Scatter plots show significant clustering and discrimination of mineral deposits. The significant groupings were used for developing new models and patterns. 2. Tectonic setting and regional geology Iran geology is assembled of continental fragments initially rifted from Gondwana as the Paleotethys and Neotethys oceans developed, subsequently amalgamated as both oceans closed (Alavi, 2004; Berberian and King, 1981; Sengör, 1987). The subduction of the later ocean plate involved collision of the Arabian plate with Iran and growth
185
of the Zagros Mountains. Stampfli and Borel (2002) suggested that the Iran portion of Neotethys was already closed in the late Cretaceous, following obduction along the Arabian margin. The suture lies in southern Iran between the Zagros and a complex arc (Sanandaj–Sirjan Zone and Urumieh–Dokhtar Magmatic Arc) in Central Iran (Berberian and King, 1981). At least three parallel tectonic zones are known in the Zagros Orogen. These are from the northeast to southwest (Fig. 1): 1) the Zagros Fold-Thrust Belt (ZFTB), 2) the Sanandaj–Sirjan Zone (SSZ) and, 3) the Urumieh–Dokhtar Magmatic Arc (UDMA) (Alavi, 1994; Berberian and King, 1981). The SSZ is 1500 km long, and 150 to 200 km wide, and is related to the opening of the Tethys Ocean and its
Fig. 2. Generalized geological map of studied area showing major and minor lead and zinc deposits. Modified from Haghipour and Aghanabati, 1985.
186
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
subsequent convergence during the Cretaceous and Tertiary due to continental collision between the Afro-Arabian and the Eurasian plates (Agard et al., 2005; Ghasemi and Talbot, 2006; Mohajjel et al., 2003). The predominantly Mesozoic and Tertiary marine and continental sedimentary sequences, major unconformities and structural framework of SSZ are comparable to those of the Central Iranian block in the east. The post-Permian rocks of the SSZ are considered as having a Eurasian affinity by most authors (Agard et al., 2005; Berberian and King, 1981; Mohajjel and Fergusson, 2000; Stocklin, 1968), except Alavi (1994, 2004) who favors an Arabian affinity. Late Jurassic to Tertiary Alpine compressive tectonic events have imparted the characteristic structural trend to this zone. It resulted in northwest-oriented parallel belts of sedimentary and metamorphic rocks; northwest-trending folds partly overturned to the south and southwest; southwest-verging thrusts, and northwest-trending high-angle reverse faults, which resulted in thickening of the crust and Tertiary uplift of the SSZ (Berberian and Berberian, 1981; Ghasemi and Talbot, 2006; Mohajjel et al., 2003; Stocklin, 1968). 3. Geological setting of the studied area Precambrian rocks composed of metasediments and metavolcanics, dolomite, sandstone and shale (major outcrops in Aligodarz– Golpaygan area) form the basement of the studied area (Thiele et al., 1968). The basic igneous rocks are evidence of Permian rifting in the SSZ and are followed by conglomerate, recrystallized chert-bearing limestone and dolomite. The Permian sequence of the SSZ is comparable with that of Central Iran (Berberian and King, 1981). An orogenic event in the Middle Triassic time caused widespread metamorphism and magmatism in the SSZ with production of acidic tuff and intermediate to acidic lavas in association with carbonates and shale (Alrik and Virlogeux, 1977). The Jurassic sequence comprises volcanic rocks at the base followed by carbonates, shale and grey sandstone formations, all affected by low grade metamorphism in the late Jurassic (Berberian, 1973). Clastic red beds of Cretaceous age overlie Triassic and Jurassic sequence with clear disconformity, followed by dolomite, orbitolina-bearing limestone, glauconitic sandstone and shale. Tertiary sequences are mainly flyshtype sediments of the Eocene to Oligocene age with some outcrops of amphibole-pyroxene diorites, andesite and volcaniclastics. The Tertiary sequences in the SSZ are limited to local exposures (Mohajjel, 1997). Quaternary deposits are mainly composed of alluvium and terrace sediments. The magmatism in the SSZ is related to rifting and later subduction and occurred from Precambrian to Paleocene (Berberian and Berberian, 1981). The Triassic intrusive bodies are exposed mainly in southern SSZ while Upper Cretaceous–Paleocene bodies are in the northern part. At the northeastern border of SSZ, scarce granitoid bodies of Eocene (Moritz et al., 2006) and Miocene (Stockli et al., 2004) age are exposed. The Precambrian–Paleozoic magmatic rocks are mainly basalt and diorite, and have experienced low grade metamorphism. The Triassic magmatic rocks are andesite and diorite while the Cretaceous rocks are dominantly andesitic. 4. Mineral deposits (Fig. 2) The SSZ is the major metallogenic zone with Pb–Zn, Au, Cu, Fe, W–Sn mineralization. Non-metallic deposits such as feldspar, talc, barite, silica, fluorspar, garnet and dimension stones are also located in the SSZ (Ghorbani, 2007; 2008). The Pb–Zn mineralizations in the SSZ are mainly MVT (e.g. Irankuh, Anjireh–Tiran, and Emarat) and SEDEX (e.g. Gol-e-Zard and Hossein Abad) types In the central SSZ, the Malayer–Esfahan zone contains major carbonate-hosted Pb–Zn deposits (Fig. 2). The host rocks are mainly Early Cretaceous limestones, though there are some deposits hosted by Triassic, Jurassic and even Eocene carbonates. The Pb–Zn mineralization is
limited to three stratigraphic units (Momenzadeh, 1976): I) Jurassic strata below the Cretaceous–Jurassic angular unconformity e.g. the Gol-e-Zard deposit (Farhadi Nejad et al., 1999), II) Lower Cretaceous orbitolina-bearing limestone, which hosts the major deposits such as Ahangaran (Zamanian, 1993), Lakan (Momenzadeh, 1976) and Ravanj (Modaberi and Rastad, 1998), and III) Middle Cretaceous limestone and dolomites which host major deposits such as Emarat (Ehya et al., 2010), Irankuh (Ghazban et al., 1994; Rastad et al., 1980) and Anjireh–Tiran (Taherian, 1993). The characteristics of the major Pb–Zn deposits in the central SSZ are presented in Table 1. 5. Sampling and chemical analysis 5.1. Sampling Sampling of the active and abandoned mines and prospects were carried out by taking representative samples from previously recognized ore zones. The number of unweathered samples collected from each deposit was linked to the size of mineral deposits. 200 ore samples, with 10% duplicated from 15 deposits, were collected from 104 mines and occurrences as chip and channel samples weighing from 2 to 5 kg. Samples were crushed, homogenized, divided and pulverized down to less than 74 μm, and the rest were kept as an archive. 5.2. Chemical analysis Chemical analysis was performed by Amdel Laboratories Ltd. (Australia) using ICP-OES. In this method, 500 mg of sample in replicate were weighed and digested in 6 ml of nitric acid and 1 ml of hydrochloric acid, and measured for 43 elements. The analyzed elements include 9 major elements: Al, Ca, Fe, K, Mg, Mn, Na, P, S and 34 trace elements: Ag, As, Au, B, Ba, Be, Bi, Cd, Ce, Co, Cr, Cs, Cu, Hg, La, Li, Mo, Nb, Ni, Pb, Rb, Sb, Sc, Sn, Sr, Te, Th, Tl, U, V, W, Y, Zn, and Zr. Additionally, high concentrated elements i.e. Pb and Zn, were analyzed by AAS, taking a 200 mg subsample which was digested in HF and aqua regia in the Geological Survey of Iran (GSI) Lab. 6. Data processing Univariate and multivariate analysis techniques were applied to the matrix of concentration values. This enabled to indentify interrelations and grouping of the datasets and to facilitate their subsequent interpretation (Burgos et al., 2006; Meloun et al., 2005; Reimann and Garrett, 2005). The chemical analysis data were stored in a database and all censored data were set to a value of 3/4 and 4/3 of the detection limits, respectively for lower and higher values (Helsel, 2004; Sanford et al., 1993). For Au and Sb, less than 10% of the data were higher than the upper detection limit and for Ba and Sr a similar proportion were less than lower detection limit. There were some elements (Be, Cr, Li, Sn, Tl and W) with over 10% of the data below the lower detection limit, and these have been omitted from cluster analysis. The following descriptive statistics of centering, dispersion and shape were used (Table 2): mean, median, standard deviation, kurtosis, skewness, minimum, maximum, detection limit (DL) and percentage of censored data (% b DL). Most of the statistical methods require a normal distribution of the input data to yield valid statistical results. We started with using logarithmic technique (Reimann and Filzmoser, 2000; Swan et al., 1995), for normalization of our datasets. Since this procedure did not result in good normalization of our data, the Cox & Box transformation (Box and Cox, 1964) has been used to transform the data to a normal distribution form. Here, for each variable the optimal parameter for the Cox & Box transformation has to be determined which is a time consuming procedure with large datasets, and so, MATLAB 2007
Table 1 Summary of major characteristics of Pb–Zn deposits of central SSZ. UTM
Host/country rocks
Age of host
Metal associations
Type
Mining
Reserves and grade
References
Iran Kuh
559326 E 3598381 N 522175 E 3620275 N 511580 E 3623240 N 502822 E 3581782 N 458632 E 3697258 N 429216 E 3709905 N 396546 E 3714236 N 385121 E 3701603 N 381842 E 3728979 N 387701 E 3737137 N 370774 E 3746612 N 354010 E 3770012 N 315506 E 3783633 N 472273 E 3740133 N
Dolostone , limestone , calcareous shales
Lower Cretaceous
(Zn, Pb), Ag, Cd, Ba, Mn
MVT
Active
10 Mt @ %11 Zn, %2.5 Pb
Sandy limestone , dolomitic limestone
Lower Cretaceous
(Pb, Zn), Ag, Sb
MVT
Non_active
b 1 Mt @ %5 Pb, %1 Zn
Ghazban et al.(1994) and Rastad et al.(1980) Momenzadeh (1976)
Dolostone , limestone , calcareous shales
Lower Cretaceous
(Zn, Pb), Ag, Cu, Sb, Cd
MVT
Active
1.5 Mt @ %4 Zn, %1 Pb
Taherian (1993)
Schist , limestone , sandstone
Lower Cretaceous
(Zn, Pb, Cu), Cd
SEDEX
Under exploration
Not available
–
Mafic lava , tuff , limestone
Cretaceous
(Pb, Zn, Cu), Ag, Cd, Sb
SEDEX
Non_active
Not available
Momenzadeh (1976)
Dolomitized limestone , pyroclastic volcanic
Cretaceous
(Pb, Zn, Ag), Cu
MVT
Non_active
Not available
Ghorbani (2007)
Phylite–slate–quartzite
Jurassic
(Pb, Zn), Fe, Au
SEDEX
Under exploration
Not available
Momenzadeh (1976)
Sandy dolomitic, shaly slate, sandstone
Jurassic
(Pb, Zn, Cu), Ag, Au
SEDEX
Under exploration
Not available
Farhadi Nejad et al. (1999)
Sandy dolomitic, limestone
Cretaceous
(Pb, Zn), Ag, Cd, Ba
MVT
Active
b 1 Mt @ 4% pb, 1.5% Zn
Ghorbani (2007)
Sandstone, shale
Lower Cretaceous
(Zn, Pb), Ag, Cd, Ba, Sb
MVT
Active
b 1 Mt @ %5 Zn, %2.5 Pb
–
Shaly and silica limestone, dolostone
Lower Cretaceous
(Pb, Zn), Ag, Cd, Cu
MVT
Active
1.5 Mt @ %5 Pb, %2 Zn
Ehya et al.(2010)
Limestone, shale
Triassic–Jurassic
(Pb, Zn), Ba
MVT
Active
3 Mt @ %3 Zn, %2.5 Pb
Momenzadeh (1976)
Sandy dolomite , quartzite , Limestone
Lower Cretaceous
(Pb, Zn, Ag), Ba, Fe, Cu
MVT
Active
b 1 Mt @ %3 Pb, %1 Zn, 200 g/t Ag
Zamanian (1993)
Dolostone , limestone
Lower Cretaceous
(Pb, Zn, Ag), Ba
MVT
Active
2 Mt @ %4 Pb,%1 Zn,50 g/t Ag
Modaberi and Rastad (1998)
Khaneh Surmeh Anjeereh Tiran Gardaneh Rokh Saleh Paygambar Dareh Noqreh Baba Qoleh Gol-e-Zard Lakan Robat Emarat Vishan Ahangaran Ravanj
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
NAME
187
188
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
Table 2 Calculated statistical parameters of 104 ore samples from selected Pb–Zn deposits in the study area (by Statistica code). Elements
Mean
Median
Minimum
Maximum
Std.Dev
Skewness
Kurtosis
DL
% b DL
Al Ca Fe K Mg Mn Na P Ti Ag As Au Ba Be Cd Ce Co Cr Cu La Li Mo Nb Ni Pb Rb Sb Sc Sn Sr Th Tl U V W Y Zn
1.05 8.81 10.27 0.32 1.11 0.35 0.08 0.03 0.04 43 504 0.1 4318 0.5 124 13 14 118 4801 6 16 3 1 36 38,303 16 906 2.6 4.9 466 13 2.6 10 25 9 7 33,602
0.73 5.86 4.8 0.23 0.45 0.15 0.03 0.02 0.02 15 52 0.008 266 0.4 23. 6 7 18 816 2.7 8.5 2 0.2 7 21,000 12 58 1.9 1.6 186 5.4 0.6 2.8 16 0.3 4 5637
0.048 0.05 0.27 0.018 0.011 b0.0005 0.003 b0.01 b0.01 0.14 1 0.0008 10 b0.1 b0.1 b5 b1 b1 b1 b5 b0.1 b0.5 0.1 b1 45 0.25 b0.1 0.2 b0.5 b1 b0.2 b0.1 b0.2 1 b0.5 0.5 45
5.39 30.5 58.5 1.93 8.57 8.7 1.36 0.27 0.38 532 198,50 2 350,000 4 1093 153 186 7030 67,125 68 305 23 25 1389 198,650 85 43,985 20 160 9013.3 66 82 123 148 860 68 329,950
0.99 8.57 14.5 0.31 1.55 1.04 0.2 0.03 0.06 80 2229 0.3 34,261 0.5 214 22 24 690 10,311 9 32 3 3 142 45,818 13 4613 3 15 1029 17 8 20 26 84 9 62,537
2.1 0.9 2.3 2.3 2.2 4.8 3.6 6 3.7 4.1 7.5 5.6 10.2 4 2.5 4.2 4.9 9.9 3.8 4.1 7.4 3.6 5.9 8.7 1.8 2.1 8.4 3.2 9.3 6.1 1.5 8.9 3.6 2.6 10.2 3.6 2.7
5.3 −0.1 4.9 6.9 5.5 25.1 15.4 41.5 15.1 20.1 60.4 32.2 103.6 23.9 6.5 20.8 29.2 100 17.5 20 63.8 15.4 41.9 81.1 3.3 6.7 76.4 13.3 90.8 46.8 1.2 86.6 14.3 7 104 19.1 7.4
0.01 0.01 0.01 0.01 0.01 0.0005 0.001 0.01 0.01 0.1 0.5 0.0001 1 0.1 0.1 5 1 1 1 5 0.5 0.5 0.1 1 1 0.1 0.1 0.1 0.5 1 0.2 0.1 0.2 1 0.5 0.5 1
0 0 0 0 0 3.2 0 3.2 2.1 0 0 0 0 15 2.1 1 2.1 12.5 1 1 10.5 8.3 0 2.1 0 0 0.6 0 18 0.1 3.2 33 4.2 0 50 0 0
Major elements in (%) and trace elements in (ppm). In addition the detection limit (DL) and the percentage of samples below detection limit (% b DL-censored data) are given.
functions were used with some short coding for normalization of our datasets. Normalized data were plotted as box plots. Figs. 3 and 4 present the box plot of data before and after normalization respectively, showing means, upper and lower limits, quartiles, outliers and extremes in the dataset. The statistical analyses were carried out using multivariate statistics, hierarchical cluster and K-means techniques (Templ et al., 2008). K-means is known as a very fast clustering technique in terms of its computing requirements (Hartigan, 1975; McQueen, 1967). The objective of K-means is to partition all observations into n groups by minimizing a predefined criterion. Starting from n initial guesses for the cluster centers, each object is transferred to another group until an “error measure” (e.g., the sum over the squared distances from each observation points to its cluster center) cannot be further reduced. As it would be computationally impractical to calculate the overall minimum of the objective function, a procedure for finding a local minimum may be used (Anderberg, 1973; Hartigan and Wong, 1979). In general, the K-means method is not sensitive to the choice of the initial cluster centers and the simplest approach is to take the first n observations of a data matrix. However, it is important to consider an initial number of clusters. A well-known method used to determine the number of clusters is the Hartigan (1975) criterion which is based on iteration analysis until finding the Hartigan value (HV) smaller than 10. The formula is: HV =
SSk −1 ðn−k−1Þ SSk −1
ð1Þ
where HV is Hartigan value, SSk is the sum of squares, k is the number of clusters and n is the number of data points. As shown in Fig. 5, four is the best choice of cluster number in our database for K-means analysis. 7. Discussion Hierarchical clustering algorithms were used for reconstruction of elemental patterns of Pb–Zn mineralization in the central SSZ. A hierarchical algorithm yields a dendrogram representing the nested grouping of patterns and similarity levels in which groupings change. The Phonon line draws based on equal similarity factor for selecting major groups with the most similarity in clustering. The cluster relations were created by a complete linkage method and similarity levels (Swan et al., 1995) established by Pearson correlation coefficient. The results are presented in a dendrogram (Fig. 6) showing the cogenetic clusters. The best separation line is around 1.2 and is selected based on paragenetic similarities of deposits. Fig. 6 depicts four master clusters, which are related to each other as pairs. The first cluster (H1) includes Ag, Pb, Sb, Cd and Zn all of which are ore-forming elements associated with sediment hosted Pb–Zn deposits (Leach et al., 2005). These can also be used as pathfinders in stream sediment and lithogeochemical exploration for blind deposits. The Pb–Sb–Ag and Zn–Cd limbs are significantly correlated, as these groups of elements appear in lead and zinc rich minerals respectively. These elements are as major or minor elements of sedimentary-hosted Pb–Zn deposits (Leach et al., 2001). The presence of Sb is unusual for this type of deposit, though it has been detected in some of the Pb–Zn deposits of the SSZ. The second cluster (H2) includes Ba, Sr, Ca, Mg and Mn, and is considered as a rock-forming element group associated with
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
189
Fig. 3. Raw data box plot of lead and zinc deposits from central part of SSZ.
limestone and dolomite host rocks, as well as with barite and calcite gangue minerals. Mn is present as a minor element. The first and second cluster ties reflect the genetic models for carbonate-hosted Pb–Zn deposits (Cox and Singer, 1986) depicting elements associated with MVT and SEDEX mineralization. The third cluster (H3) includes Al, K, Rb, Sc, Ti, V, Na, Nb, Ce, La and Y, elements associated with shale and volcanic rocks which host some of the Pb–Zn deposits in the central SSZ (Momenzadeh, 1976). The REE are mainly associated with monazite-bearing Jurassic shales (Alipour and Meshkani, 2005). The fourth cluster (H4), including As, Au, Mo, Co, Ni, Cu, P, Fe, Th and U, is mainly associated with MVT, SEDEX and massive sulphide deposits (Berger, 2000; Sangster and Leach, 1995). The metamorphosed shale and carbonate host rocks in the central SSZ contain Au,
Cu, Co and As mineralization (Ghorbani, 2007). The third and fourth clusters depict Pb–Zn deposits hosted by shale, sandstone and volcanic rocks, therefore they can guide exploration for these type of deposits in the district. For example, the Gol-e-Zard deposit is hosted by shale and sandstone and contains Cu, Au and Ag. Having identified meaningful clusters in the data using hierarchical cluster analysis, these data were used for further refining and discrimination of Pb–Zn deposits by the K-means technique. The clusters of ore- and rock-forming elements were transferred to the Statistica program and K-means calculations were carried out based on K = 4. The results are presented as a series of scatter plot in Fig. 7. Variation in this plot is related to differences in elements and their correlations. As an example, if we compare a Cu and a Pb–Zn deposit, both with dolomitic host rocks, they will fit into one group
Fig. 4. Normalized data box plot of lead and zinc deposits from central part of SSZ.
190
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
Fig. 5. Diagram illustrating number of clusters based on Hartigan (1975) formula.
based on rock-forming elements, but will be separated based on the ore-forming elements into two different groups. The Kc1 scatter plot is based on all elements and there is no meaningful separation, as depicted in Fig. 7a. The first, second and third clusters of the deposits show negative slopes in the scatter plot (Fig. 7a), while the fourth has a positive slope. Considering the database and dendrogram it was predicted that members of the fourth cluster have affinities to MVT and SEDEX deposits and are mainly located in the northwest of the SSZ. On the other hand, deposits in the first, second and third clusters show only MVT affinities and are mainly present in the central to southern part of the SSZ. The ore elements related to MVT deposits (Kc2), shown in Fig. 7b, depicting a meaningful division for four clusters. All depicted clusters follow the same trend except the fourth cluster (Fig. 7b). Variations of lead-related elements are almost insignificant in all clusters, while zinc-related elements show significant variations. Thus, there are deposits with similar Pb + Sb + Ag values but variable amounts of Zn + Cd in each cluster. These variations are probably due to geological and geochemical conditions for deposition, tectonic activity or emplacement of deposits in the central SSZ. Most of deposits grouped in the second and third clusters are large to medium in size while those in the fourth cluster are small deposits based on volume
or tonnage. It can be inferred that those deposits with low concentrations of both elemental cluster are small deposits. In the Kc3 diagram (Fig. 7c) there is a reasonable discrimination of deposits in the first and second clusters from those of the third and fourth clusters. It seems that in the host rock in the deposits of the first cluster we are dealing decrease in carbonate content, while those of the third and fourth clusters are hosted by carbonate, mainly dolomitic, host rocks. It should be noted that most of the deposits in the central SSZ are hosted by carbonate rocks and are well separated from shale, sandstone and volcanic hosted deposits. The Kc4 diagram (Fig. 7d) provides no meaningful separation between deposits, in spite of similar positive correlations, although it confirms that the REE are related to volcanic and shale hosted deposits. Group Kc5 (Fig. 7e) demonstrates a meaningful discrimination of deposits based on the negative and positive correlations of the first and second clusters with the third and fourth clusters respectively. The elemental concentrations of the third and fourth clusters are high, and the corresponding deposits are mainly located in northern part of the studied area. Two groups of deposits from the central SSZ, that were welldiscriminated by K-means clustering analysis, were selected for investigation of their spatial pattern of distribution in ArcGIS (Figs. 8 and 9). Ore elements, (Pb, Zn, Cd, Ag, and Sb) show significant differences between element associations (Fig. 8). In previous studies, these deposits were considered as one class of mineralization (Momenzadeh, 1976; Rasa and Kazemi Mehrnia, 2005), but the K-means clustering make it possible to separate them into meaningful groups. From the spatial distribution and mineral deposit zoning, it can be inferred that deposits of the second and fourth clusters are more frequent in the center of studied area, illustrated by Kc2 diagram results (Fig. 7b). Deposits of the first and third clusters are more frequent in the eastern segment of the studied area. Together these data show a distinctive mineral zonation in the central SSZ. The highest Pb+Sb+Ag concentration in mineral deposits is observed in the central segment of the study area and shows a decrease from the center to both east and west. The third cluster deposits (K3) are restricted to the center of the studied area, and have high concentrations of Pb + Sb + Ag, probably reflecting the neighboring volcanic rocks and plutonic intrusions (Fig. 8). Large and medium deposits occur mainly in this region and
Fig. 6. Dendrogram of hierarchical cluster analysis for major, minor and trace elements of lead and zinc deposits in central SSZ, separating elements into four distinct groups (H1 to H4).
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
191
Fig. 7. Scatter plot of each group of elements using K-means clustering to indicate the separation of mineral deposits in Kc1 (a), Kc2 (b), Kc3 (c), Kc4 (d) and Kc5 (e).
less commonly in the first and second clusters. These results suggest further exploration of occurrences and abandoned mines in the center of the studied area would be appropriate, as the potentials for medium to large Pb–Zn deposits is greatest here. Host rock elements (Mg, Ca, Sr, Ba and Mn) also show regional zonation (Fig. 9). A NW–SE trend controls the frequency of deposits of
the first to fourth cluster. This is supported by geological evidence as in the northern part of the studied area metamorphic rocks (recrystalized limestone, metasandstone and metavolcanics) are exposed and show an increase in metamorphic grade from SE toward NW (Ghorbani, 2002). Deposits of the fourth cluster are mainly hosted by dolomite and limestone which correlates well with the geology of the studied area.
192
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
Fig. 8. Symbol map showing lead and zinc deposits group based on K-means clustering (Kc2) after hierarchical clustering of (H1) in the central SSZ and the respective scatter plot.
8. Conclusion The results of our investigation shows that it is possible to discriminate and classify ore deposits at the regional scale by using clustering techniques and elemental associations, and so provides a practical exploration tool. It is also possible to investigate the spatial patterns of mineral deposits and use group of elements associations for more precise discrimination of ore deposit types. The Pb–Zn mineral deposits of the central SSZ were classified using multivariate cluster analysis. Using hierarchical clustering, elements
were grouped into four clusters, and ore elements and rock-forming elements were effectively separated. As, Au, Co, Ni, Cu, Fe, Cd, Sb, Ag, Pb, Zn and Mo clustered as an elemental association linked with Pb–Zn mineralization while Ca, Mg, Mn, Ba and Sr grouped as carbonate hosts rocks. The ore elements can be used as exploration tools in geochemical prospecting and exploration in the studied area to identify deposit types, not just anomalies. Associations of elements indicate the presence of MVT and SEDEX type mineralization in the central SSZ. Elemental clustering was also used for mineral deposit discrimination
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
193
Fig. 9. Symbol map showing lead and zinc deposits group based on K-means clustering (Kc3) after hierarchical clustering of (H2) in the central SSZ and the respective scatter plot.
by the K-means methods of Hartigan (1975). Based on Sb + Ag + Pb vs. Zn + Cd ratios, deposits separated into four clusters, with large to medium deposits less abundant in the first and second clusters. Therefore, the obtained clustered element association can be used as a geochemical exploration tool for large to medium deposits, identifying which abandoned mines, recognized occurrences and prospects are worthy of further detailed exploration. Based on Ca + Mg + Mn and Ba + Sr elemental ratios, likewise deposits clustered into four groups, indicating the role of carbonate host rocks in mineralization. The symbol map (Figs. 8 and 9) shows a well-defined spatial zoning of
mineral deposits occurrences in the studied area with E–W trends for ore elements and NW–SE trends for rock-forming elements. The study has demonstrated that multivariate techniques and sequential clustering are very effective in separating large and small deposits and in discriminating different type of neighboring deposits, which in previous classifications were considered as one type. As well, enhancing the classification of mineralization, the pattern recognition methods were able to uncover elemental zoning, identify key elemental associations for further geochemical exploration and indicate exploration target areas for large to medium size ore deposits.
194
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
Acknowledgements The authors wish to thank Professor Bruce Yardley, School of Earth and Environment, Leeds University, for extensive review, constructive comments and significant improvement of the manuscript. We would also like to thank Professor Simon Pirc, Professor Ladoslav A. Palinkas and the other reviewers of the Journal of Geochemical Exploration whose constructive comments significantly improved the earlier version of manuscript. We also like to thank the Geological Survey of Iran (GSI) for the support during current project. Constructive comments and support of Mr. B. Borna, GSI, is highly appreciated. We are also thankful to Mr. A. Mousavi for drawing Fig. 1 and Mr. P. Afzal for helpful discussions. References Agard, P., Omrani, J., Jolivet, L., Mouthereau, F., 2005. Convergence history across Zagros (Iran): constraints from collisional and earlier deformation. Int. J. Earth Sci. 94, 401–419. Alavi, M., 1994. Tectonics of the Zagros orogenic belt of Iran: new data and interpretations. Tectonophysics 229, 211–238. Alavi, M., 2004. Regional stratigraphy of the Zagros fold–thrust belt of Iran and its proforeland evolution. Am. J. Sci. 304, 1–20. Alipour, A., Meshkani, S.A., 2005. Identification of monazite mineral in Jurassic shale of Iran. 22nd Symposium on Geosciences, Geological Survey of Iran, pp. 197–203. Alrik, G., Virlogeux, D., 1977. Petrographie et geochimie des rockes metamorphique et magmatiquse de la region de Deh Bid-Bavanat, chaine de Sanandadj–Sirjan. These 3 eme cycle,universite Scientifique et Medicale de Grenoble, France. 239 p. Anderberg, M.R., 1973. Cluster Analysis for Applications. Academic Press. Inc., New York. 203 pp. Berberian, M., 1973. Two important deformational and metamorphic phases in the belt northeast of the Zagros thrust line (Iran); a brief structural review of the Sanandaj– Sirjan belt. Geological Survey of Iran. internal report No. 27, 35 p. Berberian, F., Berberian, M., 1981. Tectono-plutonic episodes in Iran. In: Gupta, H.K., Delany, F.M. (Eds.), Zagros–Hindu Kush–Himalaya Geodynamic Evolution. Geodynamics Series, V. 3. American Geophysical Union, Washington, D.C, pp. 5–32. Berberian, M., King, G.C.P., 1981. Towards a paleogeography and tectonic evolution of Iran. Can. J. Earth Sci. 18 (2), 210–265. Berger, B.R., 2000. Classification of mineral deposits. In: Seal II, R.R., Foley, N.K. (Eds.), Geoenvironmental Models of Mineral Deposits: U.S Geological Survey Open-File Report 02-195. Box, G.E.P., Cox, D.R., 1964. An analysis of transformations. J. R. Stat. Soc. B 26, 211–252. Briggs, P.L., 1978. Pattern recognition applied to Uranium prospecting. Submitted to the Department of Earth and Planetary Sciences. 233pp. Burgos, P., Madejón, E., Pérez de Mora, A., Cabrera, F., 2006. Spatial variability of the chemical characteristics of a trace-element-contaminated soil before and after remediation. Geoderma 130, 157–175. Burnol, L., 1968. Contribution al'etude des gisments de plomb et zinc de L'Iran. Essais de classification paragenetique. Geological Survey of Iran Report No. 11. 113 p. Castillo-Munoz, R., Howarth, R.J., 1976. Application of the empirical discriminant function to regional geochemical data from the United Kingdom. Geol. Soc. Am. Bull. 87, 1567–1581. Collyer, P.L., Merriam, D.F., 1973. An application of cluster analysis in mineral exploration. Math. Geol. 4, 213–223. Cox, P., Singer, D.A., 1986. Mineral Deposit Model: U.S. Geological Survey Bulletin 1693. 379 p. Ehya, F., Lotfi, M., Rasa, I., 2010. Emarat carbonate-hosted Zn–Pb deposit, Markazi Province, Iran: a geological, mineralogical and isotopic (S, Pb) study. J. Asian Earth Sci. 37 (2), 186–194. Farhadi Nejad, T., Rastad, E., Hamedi, M.A., 1999. Mineralogy, geochemistry and genesis of Gol-e-Zard Zn–Pb deposit. 2nd Symposium of Geological Society of Iran, Mashad, pp. 328–329 (In Persian). Friedman, J.F., Meulman, J.J., 2004. Clustering objects on subsets of variables (with discussion). J. R. Stat. Soc. B 4, 815–849. Ghasemi, A., Talbot, C.J., 2006. A new tectonic scenario for the Sanandaj–Sirjan Zone (Iran). J. Asian Earth Sci. 26, 683–693. Ghazban, F., McNutt, R.H., Schwarcz, H.P., 1994. Genesis of sediment hosted Zn–Pb–Ba deposits in the Irankuh district, Esfahan area, west-central Iran. Econ. Geol. 89, 1262–1278. Ghorbani, M., 2002. Introduction to the economic geology of Iran. National Geosciences Database of Iran (NGDIR). 854 p. (In Persian). Ghorbani, M., 2007. Economic geology of mineral and natural resources of Iran, Vol. 1. Arian Zamin. 492 p. (In Persian). Ghorbani, M., 2008. Economic Geology of Mineral and Natural Resources of Iran, Vol. 2. Arian Zamin. 639 p. (In Persian). Granath, G., 1988. Pattern recognition in geochemical hydrocarbon exploration: a fuzzy approach. Math. Geol. 20, 673–691. Gustavsson, N., Bjorklund, A., 1976. Lithological classification of tills by discriminant analysis. J. Geochem. Explor. 29, 89–103. Haghipour, A., Aghanabati, A., 1985. Geological Map of Iran. 1:2,500,000. Geological Survey of Iran.
Hanesch, M., Scholger, R., Dekkers, M.J., 2001. The application of fuzzy c-means cluster analysis and non-linear mapping to a soil. Phys. Chem. Earth A. 26 (11–12), 885–891. Hartigan, J.A., 1975. Clustering Algorithms. John Wiley and Sons, New York. 108 pp. Hartigan, J.A., Wong, M.A., 1979. A K-means clustering algorithm, Journal of the Royal Statistical Society. C Appl. Statistics. 28, 100–108. Helsel, D.R., 2004. Non-detects and Data Analysis: Statistics for Censored Environmental Data. John Wiley and Sons, New York. Howarth, R.J., 1973. The pattern recognition problem in applied geochemistry. In: Jones, M.J. (Ed.), Geochemical Exploration 1972. Institution of Mining and Metallurgy, London, pp. 259–273. Ji, H., Zhu, Y., Wu, X., 1995. Correspondence cluster analysis and its application in exploration geochemistry. J. Geochem. Explor. 55, 137–144. Ji, H., Zeng, D., Shi, Y., Wu, Y., Wu, X., 2007. Semi-hierarchical correspondence cluster analysis and regional geochemical pattern recognition. J. Geochem. Explor. 93, 109–119. Leach, D.L., Bradley, D., Lewchuck, M.T., Symons, D.T.A., de Marsily, G., Brannon, J.C., 2001. Mississippi Valley-type lead–zinc deposits through geological time: implications from recent age-dating research. Miner. Deposita 36, 711–740. Leach, D.L., Sangster, D.F., Kelley, K.D., Large, R.R., Garven, G., Allen, C.R., Gutzmer, J., Walters, S., 2005. Sediment-hosted lead–zinc deposits: a global perspective. Economic Geology 100th Anniversary Volume. Society of Economic Geologists, Inc, pp. 561–607. Leisch, F., 1999. Bagged Clustering. Working Paper # 51, SFB “Adaptive Information Systems and Modeling in Economics and Management Science”. http://www. wuwien.ac.at/am. Lindqvist, L., Lundholm, I., Nisca, D., Esbensen, K., Wold, S., 1987. Multivariate geochemical modelling and integration with petrophysical data. J. Geochem. Explor. 29, 279–294. McQueen, J., 1967. Some methods for classification and analysis of multivariate observations. Fifth Berkeley Symp. Math. Stat. Probab. V1, 281–297. Meloun, M., Sánka, M., Nèmec, P., Krítková, S., Kupka, K., 2005. The analysis of soil cores polluted with certain metals using the Box–Cox transformation. Environ. Pollut. 137, 273–280. Meyer, C., 1981. Ore-forming processes in geologic history. Economic Geology. 75th Anniversary Volume, pp. 6–41. Meyer, C., 1988. Ore deposits as guides to the geologic history of the earth. Annu. Rev. Earth Planet. Sci. 16, 47–71. Modaberi, S., Rastad, E., 1998. Facies analysis as an approach for depositional environment of Ravanj Pb–Ag stratabound deposit, Delijan. 2nd Symposium of Geological Society of Iran, Mashad, pp. 443–444 (In Persian). Mohajjel, M., 1997. Structure and tectonic evolution of Palaeozoic–Mesozoic rocks, Sanandaj–Sirjan Zone, western Iran. PhD Thesis, University of Wollongong, Australia. Mohajjel, M., Fergusson, C.L., 2000. Dextral transpression in the late Cretaceous continental collision, Sanandaj–Sirjan Zone, Western Iran. J. Struct. Geol. 22, 1125–1139. Mohajjel, M., Fergusson, C.L., Sahandi, M.R., 2003. Cretaceous–Tertiary convergence and continental collision, Sanandaj–Sirjan Zone, western Iran. J. Asian Earth Sci. 21, 397–412. Momenzadeh, M., 1976. Stratabound lead zinc ores in the lower Cretaceous and Jurassic sediments in the Malayer–Esfahan District (West Central Iran): Litology, metal content, zonation and Genesis, PhD Thesis, University of Heidelberg, 300 pp. Moritz, R., Ghazban, F., Singer, B., 2006. Eocene gold ore formation at Muteh, Sanandaj– Sirjan tectonic zone, western Iran: a result of late-stage extension and exhumation of metamorphic basement rocks within the Zagros orogen. Econ. Geol. 101, 1497–1524. Raftery, A.E., Dean, N., 2004. Variable selection for model-based clustering. Technical Report 452, Department of Statistics, University of Washington. Rasa, I., Kazemi Mehrnia, A.K., 2005. Carbonate-hosted Base Metal Deposits. Roozbahan Pub.co, Tehran. 262 pp. Rastad, E., Fontbote, L., Amstutz, G.C., 1980. Relation between tidal flat facies and diagenetic ore fabrics in the strataband Pb–Zn–(Ba–Cu) deposits of Irankuh, Esfahan, West Central-Iran. Revista del Institu to de Investigationes Geological Diputation provincial Universsidad de Barcelona 34, 311–323. Reimann, C., Filzmoser, P., 2000. Normal and lognormal data distribution in Geochemistry: death of a myth, consequences for the statistical treatment of geochemical and environmental data. Environ. Geol. 39, 1001–1014. Reimann, C., Garrett, R., 2005. Geochemical background — concept and reality. Sci. Total Environ. 350, 12–27. Sahandi, M.R., Soheily, M., Sadeghi, M., Delavar, S.T., Jafari-Rad, A., 2002. 1:1,000,000 Geological Map of Iran. Geological Survey of Iran, Tehran. Sanford, R.F., Pierson, C.T., Crovelli, R.A., 1993. An objective replacement method for censored geochemical data. Math. Geol. 25, 59–80. Sangster, D., Leach, D.L., 1995. Evidence for a genetic link between SEDEX and MVT deposits. In: Leach, D.L., Goldhaber, M.B. (Eds.), Extended Abstracts, International Field Conference on Carbonate-hosted Lead–Zinc Deposits, St. Louis Missouri, June 1–4, pp. 260–263. Sengör, A.M.C., 1987. Tectonics of the Tethysides: orogenic collage development in a collisional setting. Annu. Rev. Earth Planet. Sci. 15, 213–244. Stampfli, G.M., Borel, G.D., 2002. A plate tectonic model for the Paleozoic and Mesozoic constrained by dynamic plate boundaries and restored synthetic oceanic isochrons. Earth Planet. Sci. Lett. 196, 17–33. Stockli, D.F., Hassanzadeh, J., Stockli, L.D., Axen, G., Walker, J.D., Dewane, T.J., 2004. Structural and geochronological evidene for oligo-miocene intra-arc low-angle detachment faulting in Takab–Zanjan area, NW Iran. Geological Society of America, Abstract with program V36, 319.
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195 Stocklin, J., 1968. Structural history and tectonics of Iran: a review: American Association of Petroleum. Geol. Bull. 52, 1229–1258. Swan, A.R.H., Sndilands, M., McCabe, P., 1995. Introduction to Geological Data Analysis. Blackwiell Science. 446 pp. Taherian, A., 1993. Mineralogy, geochemistry and genesis of Anjireh–Tiran Pb–Zn deposit, W. Najaf Abad. Unpublished MSc. Thesis, Tehran Tarbiat Moallem University 287 p. (In Persian). Templ, M., Filzmoser, P., Reimann, C., 2008. Cluster analysis applied to regional geochemical data: problems and possibilities. Appl. Geochem. 23, 2198–2213.
195
Thiele, O., Alavi-Naini, M., Assefi, R., Houshmandzadeh, A., Seyed-Emami, K., Zahedi, M., 1968. Explanatory text of the Golpaygan Quadrangle E7. 1:250,000. Geological Survey of Iran, Tehran, Iran. 24 p. Xie, X., Liu, D., Xiang, Y., Yan, G., Lian, C., 2004. Geochemical blocks for predicting large ore deposits — concept and methodology. J. Geochem. Explor. 84, 77–91. Zamanian, H., 1993. Mineralogy, paragenesis and genesis of Ahangaran Ag–Pb deposit, Malayer. Unpublished MSc. Thesis, Tehran Tarbiat Moallem University 280 p. (In Persian).
Contents lists available at ScienceDirect
Journal of Geochemical Exploration j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j g e o ex p
The application of geochemical pattern recognition to regional prospecting: A case study of the Sanandaj–Sirjan metallogenic zone, Iran Seyed Ahmad Meshkani a,b,⁎, Behzad Mehrabi a, Abdolmajid Yaghubpur a, Younes Fadakar Alghalandis c a b c
Department of Geology, Tarbiat Moallem University, 15614, Tehran, Iran Geological Survey of Iran, P.O.Box: 13185-1494, Tehran, Iran School of Earth Sciences, University of Queensland, Brisbane, Australia
a r t i c l e
i n f o
Article history: Received 5 September 2010 Accepted 28 January 2011 Available online 18 February 2011 Keywords: Cluster analysis K-means Sanandaj–Sirjan metallogenic zone Pb–Zn deposits Spatial pattern Iran
a b s t r a c t In regional exploration programs, the distribution of elements in known mineral deposits can be used as a guide for the classification of deposits, search for new prospects and modeling ore deposit patterns. The Sanandaj–Sirjan Zone (SSZ) is a major metallogenic zone in Iran, containing lead and zinc, iron, gold, copper deposits. In the central part of the SSZ, lead and zinc mineralization is widespread and hitherto exploration has been based on geological criteria. In this study, we used clustering techniques applied to element distribution for classification lead and zinc deposits in the central part of the SSZ. The hierarchical clustering technique was used to characterize the elemental pattern. Elements associated with lead and zinc deposits were separated into four clusters, encompassing both ore elements and their host rock-forming elements. It is shown that lead and zinc deposits in the central SSZ belong to two genetic groups: a MVT type hosted by limestone and dolomites and a SEDEX type hosted by shale, volcanic rocks and sandstone. The results of elemental clustering were used for pattern recognition by the K-means method and the respective deposits were classified into four distinct categories. K-means clustering also reveals that the elemental associations and spatial distribution of the lead and zinc deposits exhibit zoning in the central part of the SSZ. The ratios of ore-forming elements (Sb, Cd, and Zn) vs. (Pb and Ag) show zoning along an E–W trend, while host rockforming elements (Mn, Ca, and Mg) vs. (Ba and Sr) show a zoning along a SE–NW trend. Large and medium deposits occur mainly in the center of the studied area, which justify further exploration around occurrences and abandoned mines in this area. The application of a pattern recognition method based on geochemical data from known mineralization in the central SSZ, and the classification derived from it, uncover elemental zoning, identify key elemental associations for further geochemical exploration and the potential to discover possible target areas for large to medium size ore deposits. This methodology can be applied in a similar way to search for new ore deposits in a wide range of known metallogenic zones. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Geochemical pattern recognition has been used in exploration geochemistry for many years. Since the 1970s, pattern recognition techniques have been applied to recognize the economic mineral resource information hidden in geological and geochemical data (Briggs, 1978; Castillo-Munoz and Howarth, 1976; Collyer and Merriam, 1973; Gustavsson and Bjorklund, 1976; Howarth, 1973) and to establish multivariate geochemical background patterns (Lindqvist et al., 1987). It has also been applied to geochemical hydrocarbon exploration (Granath, 1988), recognition of polluted sites (Hanesch et al., 2001), and to investigate the relations between regional geochemistry and large ore deposits (Xie et al., 2004).
⁎ Corresponding author at: Department of Geology, Tarbiat Moallem University, 15614, Tehran, Iran. Tel.: + 98 21 44011288. E-mail address: [email protected] (S.A. Meshkani). 0375-6742/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.gexplo.2011.01.006
Typical pattern recognition methods used in geochemical exploration consist mainly of discriminant analysis and cluster analysis (Ji et al., 2007). The principal aim of cluster analysis is to partition multivariate observations into a number of meaningful homogeneous multivariate groups. With geochemical data, cluster analysis can be used in different ways: it can be used to cluster the variables (e.g. to detect geochemical relations between the variables) and it can be used to cluster the observations (e.g. to assign samples to certain types) to come to more homogenous data subsets for further data analysis. Furthermore, there are methods that try to group the data by simultaneously clustering objects and variables (Friedman and Meulman, 2004; Ji et al., 1995, 2007; Leisch, 1999; Raftery and Dean, 2004). In this study, multielement chemical analyses of representative samples from the major ore deposits, prospects and showings, were processed and a database created. Using hierarchical cluster analysis, ore- and rock-forming elements were separated. The derived elemental associations were put through K-means clustering and
184
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
the outcomes plotted as scatter plot. Well-separated element groups possess practical geochemical significance, representing distinct types of mineralization, were plotted on a map using ArcGIS. The Sanandaj–Sirjan Zone (SSZ) is the major metallogenic zone in Iran (Fig. 1). The SSZ is a magmatic–metamorphic belt with a NW–SE trend located between the Zagros and Urmieh–Dokhtar volcanic zone of Iran (Stocklin, 1968). Based on its metallogenic characteristics, it can be divided into northern, central and southern subdivisions. The central subdivision is a well-known major Pb–Zn mining district, the Malayer–Esfahan Zone (MEZ) (Ghorbani, 2007; Momenzadeh, 1976). More than 100 ore deposits, mineralization showings and mineral occurrences have been recognized in the MEZ. They are mainly hosted by limestone, dolomite or shale and rarely by volcanic rocks or sandstone of Jurassic to Lower Cretaceous age, and exhibit stratabound, stratiform, layered, vein and lenticular shapes (Burnol, 1968; Momenzadeh, 1976). A conventional approach to classification of the ore deposits of a metallogenic province is based on genetic-age distribution (Meyer, 1981; 1988) and statistical–genetic concept (Cox and Singer, 1986). In
our research, we used a cluster analysis (CA) method for recognition of the Pb–Zn distribution in the MEZ. The hierarchical clustering technique was used to characterize elemental pattern of well-studied Pb–Zn deposit of the MEZ. In this technique all variables were clustered based on inter grouping of similarities or differences (Swan et al., 1995). Collyer and Merriam (1973) used cluster analysis of data from depleted and operating W-mines (in North America) and developed an effective guide for future exploration. Briggs (1978) applied pattern recognition techniques to uranium deposits of the Casper, Wyoming quadrangle and of the Colorado plateau for regional prospecting. Samples from 104 Pb–Zn ore deposits and occurrences within the MEZ were analyzed for 43 elements using ICP-OES techniques by Amdel Laboratories Ltd (Australia). Hierarchical clustering showed two major clusters which are related to ore-forming elements and rock-forming elements respectively. The K-means technique (McQueen, 1967) was employed for grouping ore deposits. The K-means calculation was carried out on each of the elemental groups identified by hierarchical
Fig. 1. Major structural zones of western Iran showing the studied area. Modified from Sahandi et al., 2002.
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
cluster analysis. Scatter plots show significant clustering and discrimination of mineral deposits. The significant groupings were used for developing new models and patterns. 2. Tectonic setting and regional geology Iran geology is assembled of continental fragments initially rifted from Gondwana as the Paleotethys and Neotethys oceans developed, subsequently amalgamated as both oceans closed (Alavi, 2004; Berberian and King, 1981; Sengör, 1987). The subduction of the later ocean plate involved collision of the Arabian plate with Iran and growth
185
of the Zagros Mountains. Stampfli and Borel (2002) suggested that the Iran portion of Neotethys was already closed in the late Cretaceous, following obduction along the Arabian margin. The suture lies in southern Iran between the Zagros and a complex arc (Sanandaj–Sirjan Zone and Urumieh–Dokhtar Magmatic Arc) in Central Iran (Berberian and King, 1981). At least three parallel tectonic zones are known in the Zagros Orogen. These are from the northeast to southwest (Fig. 1): 1) the Zagros Fold-Thrust Belt (ZFTB), 2) the Sanandaj–Sirjan Zone (SSZ) and, 3) the Urumieh–Dokhtar Magmatic Arc (UDMA) (Alavi, 1994; Berberian and King, 1981). The SSZ is 1500 km long, and 150 to 200 km wide, and is related to the opening of the Tethys Ocean and its
Fig. 2. Generalized geological map of studied area showing major and minor lead and zinc deposits. Modified from Haghipour and Aghanabati, 1985.
186
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
subsequent convergence during the Cretaceous and Tertiary due to continental collision between the Afro-Arabian and the Eurasian plates (Agard et al., 2005; Ghasemi and Talbot, 2006; Mohajjel et al., 2003). The predominantly Mesozoic and Tertiary marine and continental sedimentary sequences, major unconformities and structural framework of SSZ are comparable to those of the Central Iranian block in the east. The post-Permian rocks of the SSZ are considered as having a Eurasian affinity by most authors (Agard et al., 2005; Berberian and King, 1981; Mohajjel and Fergusson, 2000; Stocklin, 1968), except Alavi (1994, 2004) who favors an Arabian affinity. Late Jurassic to Tertiary Alpine compressive tectonic events have imparted the characteristic structural trend to this zone. It resulted in northwest-oriented parallel belts of sedimentary and metamorphic rocks; northwest-trending folds partly overturned to the south and southwest; southwest-verging thrusts, and northwest-trending high-angle reverse faults, which resulted in thickening of the crust and Tertiary uplift of the SSZ (Berberian and Berberian, 1981; Ghasemi and Talbot, 2006; Mohajjel et al., 2003; Stocklin, 1968). 3. Geological setting of the studied area Precambrian rocks composed of metasediments and metavolcanics, dolomite, sandstone and shale (major outcrops in Aligodarz– Golpaygan area) form the basement of the studied area (Thiele et al., 1968). The basic igneous rocks are evidence of Permian rifting in the SSZ and are followed by conglomerate, recrystallized chert-bearing limestone and dolomite. The Permian sequence of the SSZ is comparable with that of Central Iran (Berberian and King, 1981). An orogenic event in the Middle Triassic time caused widespread metamorphism and magmatism in the SSZ with production of acidic tuff and intermediate to acidic lavas in association with carbonates and shale (Alrik and Virlogeux, 1977). The Jurassic sequence comprises volcanic rocks at the base followed by carbonates, shale and grey sandstone formations, all affected by low grade metamorphism in the late Jurassic (Berberian, 1973). Clastic red beds of Cretaceous age overlie Triassic and Jurassic sequence with clear disconformity, followed by dolomite, orbitolina-bearing limestone, glauconitic sandstone and shale. Tertiary sequences are mainly flyshtype sediments of the Eocene to Oligocene age with some outcrops of amphibole-pyroxene diorites, andesite and volcaniclastics. The Tertiary sequences in the SSZ are limited to local exposures (Mohajjel, 1997). Quaternary deposits are mainly composed of alluvium and terrace sediments. The magmatism in the SSZ is related to rifting and later subduction and occurred from Precambrian to Paleocene (Berberian and Berberian, 1981). The Triassic intrusive bodies are exposed mainly in southern SSZ while Upper Cretaceous–Paleocene bodies are in the northern part. At the northeastern border of SSZ, scarce granitoid bodies of Eocene (Moritz et al., 2006) and Miocene (Stockli et al., 2004) age are exposed. The Precambrian–Paleozoic magmatic rocks are mainly basalt and diorite, and have experienced low grade metamorphism. The Triassic magmatic rocks are andesite and diorite while the Cretaceous rocks are dominantly andesitic. 4. Mineral deposits (Fig. 2) The SSZ is the major metallogenic zone with Pb–Zn, Au, Cu, Fe, W–Sn mineralization. Non-metallic deposits such as feldspar, talc, barite, silica, fluorspar, garnet and dimension stones are also located in the SSZ (Ghorbani, 2007; 2008). The Pb–Zn mineralizations in the SSZ are mainly MVT (e.g. Irankuh, Anjireh–Tiran, and Emarat) and SEDEX (e.g. Gol-e-Zard and Hossein Abad) types In the central SSZ, the Malayer–Esfahan zone contains major carbonate-hosted Pb–Zn deposits (Fig. 2). The host rocks are mainly Early Cretaceous limestones, though there are some deposits hosted by Triassic, Jurassic and even Eocene carbonates. The Pb–Zn mineralization is
limited to three stratigraphic units (Momenzadeh, 1976): I) Jurassic strata below the Cretaceous–Jurassic angular unconformity e.g. the Gol-e-Zard deposit (Farhadi Nejad et al., 1999), II) Lower Cretaceous orbitolina-bearing limestone, which hosts the major deposits such as Ahangaran (Zamanian, 1993), Lakan (Momenzadeh, 1976) and Ravanj (Modaberi and Rastad, 1998), and III) Middle Cretaceous limestone and dolomites which host major deposits such as Emarat (Ehya et al., 2010), Irankuh (Ghazban et al., 1994; Rastad et al., 1980) and Anjireh–Tiran (Taherian, 1993). The characteristics of the major Pb–Zn deposits in the central SSZ are presented in Table 1. 5. Sampling and chemical analysis 5.1. Sampling Sampling of the active and abandoned mines and prospects were carried out by taking representative samples from previously recognized ore zones. The number of unweathered samples collected from each deposit was linked to the size of mineral deposits. 200 ore samples, with 10% duplicated from 15 deposits, were collected from 104 mines and occurrences as chip and channel samples weighing from 2 to 5 kg. Samples were crushed, homogenized, divided and pulverized down to less than 74 μm, and the rest were kept as an archive. 5.2. Chemical analysis Chemical analysis was performed by Amdel Laboratories Ltd. (Australia) using ICP-OES. In this method, 500 mg of sample in replicate were weighed and digested in 6 ml of nitric acid and 1 ml of hydrochloric acid, and measured for 43 elements. The analyzed elements include 9 major elements: Al, Ca, Fe, K, Mg, Mn, Na, P, S and 34 trace elements: Ag, As, Au, B, Ba, Be, Bi, Cd, Ce, Co, Cr, Cs, Cu, Hg, La, Li, Mo, Nb, Ni, Pb, Rb, Sb, Sc, Sn, Sr, Te, Th, Tl, U, V, W, Y, Zn, and Zr. Additionally, high concentrated elements i.e. Pb and Zn, were analyzed by AAS, taking a 200 mg subsample which was digested in HF and aqua regia in the Geological Survey of Iran (GSI) Lab. 6. Data processing Univariate and multivariate analysis techniques were applied to the matrix of concentration values. This enabled to indentify interrelations and grouping of the datasets and to facilitate their subsequent interpretation (Burgos et al., 2006; Meloun et al., 2005; Reimann and Garrett, 2005). The chemical analysis data were stored in a database and all censored data were set to a value of 3/4 and 4/3 of the detection limits, respectively for lower and higher values (Helsel, 2004; Sanford et al., 1993). For Au and Sb, less than 10% of the data were higher than the upper detection limit and for Ba and Sr a similar proportion were less than lower detection limit. There were some elements (Be, Cr, Li, Sn, Tl and W) with over 10% of the data below the lower detection limit, and these have been omitted from cluster analysis. The following descriptive statistics of centering, dispersion and shape were used (Table 2): mean, median, standard deviation, kurtosis, skewness, minimum, maximum, detection limit (DL) and percentage of censored data (% b DL). Most of the statistical methods require a normal distribution of the input data to yield valid statistical results. We started with using logarithmic technique (Reimann and Filzmoser, 2000; Swan et al., 1995), for normalization of our datasets. Since this procedure did not result in good normalization of our data, the Cox & Box transformation (Box and Cox, 1964) has been used to transform the data to a normal distribution form. Here, for each variable the optimal parameter for the Cox & Box transformation has to be determined which is a time consuming procedure with large datasets, and so, MATLAB 2007
Table 1 Summary of major characteristics of Pb–Zn deposits of central SSZ. UTM
Host/country rocks
Age of host
Metal associations
Type
Mining
Reserves and grade
References
Iran Kuh
559326 E 3598381 N 522175 E 3620275 N 511580 E 3623240 N 502822 E 3581782 N 458632 E 3697258 N 429216 E 3709905 N 396546 E 3714236 N 385121 E 3701603 N 381842 E 3728979 N 387701 E 3737137 N 370774 E 3746612 N 354010 E 3770012 N 315506 E 3783633 N 472273 E 3740133 N
Dolostone , limestone , calcareous shales
Lower Cretaceous
(Zn, Pb), Ag, Cd, Ba, Mn
MVT
Active
10 Mt @ %11 Zn, %2.5 Pb
Sandy limestone , dolomitic limestone
Lower Cretaceous
(Pb, Zn), Ag, Sb
MVT
Non_active
b 1 Mt @ %5 Pb, %1 Zn
Ghazban et al.(1994) and Rastad et al.(1980) Momenzadeh (1976)
Dolostone , limestone , calcareous shales
Lower Cretaceous
(Zn, Pb), Ag, Cu, Sb, Cd
MVT
Active
1.5 Mt @ %4 Zn, %1 Pb
Taherian (1993)
Schist , limestone , sandstone
Lower Cretaceous
(Zn, Pb, Cu), Cd
SEDEX
Under exploration
Not available
–
Mafic lava , tuff , limestone
Cretaceous
(Pb, Zn, Cu), Ag, Cd, Sb
SEDEX
Non_active
Not available
Momenzadeh (1976)
Dolomitized limestone , pyroclastic volcanic
Cretaceous
(Pb, Zn, Ag), Cu
MVT
Non_active
Not available
Ghorbani (2007)
Phylite–slate–quartzite
Jurassic
(Pb, Zn), Fe, Au
SEDEX
Under exploration
Not available
Momenzadeh (1976)
Sandy dolomitic, shaly slate, sandstone
Jurassic
(Pb, Zn, Cu), Ag, Au
SEDEX
Under exploration
Not available
Farhadi Nejad et al. (1999)
Sandy dolomitic, limestone
Cretaceous
(Pb, Zn), Ag, Cd, Ba
MVT
Active
b 1 Mt @ 4% pb, 1.5% Zn
Ghorbani (2007)
Sandstone, shale
Lower Cretaceous
(Zn, Pb), Ag, Cd, Ba, Sb
MVT
Active
b 1 Mt @ %5 Zn, %2.5 Pb
–
Shaly and silica limestone, dolostone
Lower Cretaceous
(Pb, Zn), Ag, Cd, Cu
MVT
Active
1.5 Mt @ %5 Pb, %2 Zn
Ehya et al.(2010)
Limestone, shale
Triassic–Jurassic
(Pb, Zn), Ba
MVT
Active
3 Mt @ %3 Zn, %2.5 Pb
Momenzadeh (1976)
Sandy dolomite , quartzite , Limestone
Lower Cretaceous
(Pb, Zn, Ag), Ba, Fe, Cu
MVT
Active
b 1 Mt @ %3 Pb, %1 Zn, 200 g/t Ag
Zamanian (1993)
Dolostone , limestone
Lower Cretaceous
(Pb, Zn, Ag), Ba
MVT
Active
2 Mt @ %4 Pb,%1 Zn,50 g/t Ag
Modaberi and Rastad (1998)
Khaneh Surmeh Anjeereh Tiran Gardaneh Rokh Saleh Paygambar Dareh Noqreh Baba Qoleh Gol-e-Zard Lakan Robat Emarat Vishan Ahangaran Ravanj
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
NAME
187
188
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
Table 2 Calculated statistical parameters of 104 ore samples from selected Pb–Zn deposits in the study area (by Statistica code). Elements
Mean
Median
Minimum
Maximum
Std.Dev
Skewness
Kurtosis
DL
% b DL
Al Ca Fe K Mg Mn Na P Ti Ag As Au Ba Be Cd Ce Co Cr Cu La Li Mo Nb Ni Pb Rb Sb Sc Sn Sr Th Tl U V W Y Zn
1.05 8.81 10.27 0.32 1.11 0.35 0.08 0.03 0.04 43 504 0.1 4318 0.5 124 13 14 118 4801 6 16 3 1 36 38,303 16 906 2.6 4.9 466 13 2.6 10 25 9 7 33,602
0.73 5.86 4.8 0.23 0.45 0.15 0.03 0.02 0.02 15 52 0.008 266 0.4 23. 6 7 18 816 2.7 8.5 2 0.2 7 21,000 12 58 1.9 1.6 186 5.4 0.6 2.8 16 0.3 4 5637
0.048 0.05 0.27 0.018 0.011 b0.0005 0.003 b0.01 b0.01 0.14 1 0.0008 10 b0.1 b0.1 b5 b1 b1 b1 b5 b0.1 b0.5 0.1 b1 45 0.25 b0.1 0.2 b0.5 b1 b0.2 b0.1 b0.2 1 b0.5 0.5 45
5.39 30.5 58.5 1.93 8.57 8.7 1.36 0.27 0.38 532 198,50 2 350,000 4 1093 153 186 7030 67,125 68 305 23 25 1389 198,650 85 43,985 20 160 9013.3 66 82 123 148 860 68 329,950
0.99 8.57 14.5 0.31 1.55 1.04 0.2 0.03 0.06 80 2229 0.3 34,261 0.5 214 22 24 690 10,311 9 32 3 3 142 45,818 13 4613 3 15 1029 17 8 20 26 84 9 62,537
2.1 0.9 2.3 2.3 2.2 4.8 3.6 6 3.7 4.1 7.5 5.6 10.2 4 2.5 4.2 4.9 9.9 3.8 4.1 7.4 3.6 5.9 8.7 1.8 2.1 8.4 3.2 9.3 6.1 1.5 8.9 3.6 2.6 10.2 3.6 2.7
5.3 −0.1 4.9 6.9 5.5 25.1 15.4 41.5 15.1 20.1 60.4 32.2 103.6 23.9 6.5 20.8 29.2 100 17.5 20 63.8 15.4 41.9 81.1 3.3 6.7 76.4 13.3 90.8 46.8 1.2 86.6 14.3 7 104 19.1 7.4
0.01 0.01 0.01 0.01 0.01 0.0005 0.001 0.01 0.01 0.1 0.5 0.0001 1 0.1 0.1 5 1 1 1 5 0.5 0.5 0.1 1 1 0.1 0.1 0.1 0.5 1 0.2 0.1 0.2 1 0.5 0.5 1
0 0 0 0 0 3.2 0 3.2 2.1 0 0 0 0 15 2.1 1 2.1 12.5 1 1 10.5 8.3 0 2.1 0 0 0.6 0 18 0.1 3.2 33 4.2 0 50 0 0
Major elements in (%) and trace elements in (ppm). In addition the detection limit (DL) and the percentage of samples below detection limit (% b DL-censored data) are given.
functions were used with some short coding for normalization of our datasets. Normalized data were plotted as box plots. Figs. 3 and 4 present the box plot of data before and after normalization respectively, showing means, upper and lower limits, quartiles, outliers and extremes in the dataset. The statistical analyses were carried out using multivariate statistics, hierarchical cluster and K-means techniques (Templ et al., 2008). K-means is known as a very fast clustering technique in terms of its computing requirements (Hartigan, 1975; McQueen, 1967). The objective of K-means is to partition all observations into n groups by minimizing a predefined criterion. Starting from n initial guesses for the cluster centers, each object is transferred to another group until an “error measure” (e.g., the sum over the squared distances from each observation points to its cluster center) cannot be further reduced. As it would be computationally impractical to calculate the overall minimum of the objective function, a procedure for finding a local minimum may be used (Anderberg, 1973; Hartigan and Wong, 1979). In general, the K-means method is not sensitive to the choice of the initial cluster centers and the simplest approach is to take the first n observations of a data matrix. However, it is important to consider an initial number of clusters. A well-known method used to determine the number of clusters is the Hartigan (1975) criterion which is based on iteration analysis until finding the Hartigan value (HV) smaller than 10. The formula is: HV =
SSk −1 ðn−k−1Þ SSk −1
ð1Þ
where HV is Hartigan value, SSk is the sum of squares, k is the number of clusters and n is the number of data points. As shown in Fig. 5, four is the best choice of cluster number in our database for K-means analysis. 7. Discussion Hierarchical clustering algorithms were used for reconstruction of elemental patterns of Pb–Zn mineralization in the central SSZ. A hierarchical algorithm yields a dendrogram representing the nested grouping of patterns and similarity levels in which groupings change. The Phonon line draws based on equal similarity factor for selecting major groups with the most similarity in clustering. The cluster relations were created by a complete linkage method and similarity levels (Swan et al., 1995) established by Pearson correlation coefficient. The results are presented in a dendrogram (Fig. 6) showing the cogenetic clusters. The best separation line is around 1.2 and is selected based on paragenetic similarities of deposits. Fig. 6 depicts four master clusters, which are related to each other as pairs. The first cluster (H1) includes Ag, Pb, Sb, Cd and Zn all of which are ore-forming elements associated with sediment hosted Pb–Zn deposits (Leach et al., 2005). These can also be used as pathfinders in stream sediment and lithogeochemical exploration for blind deposits. The Pb–Sb–Ag and Zn–Cd limbs are significantly correlated, as these groups of elements appear in lead and zinc rich minerals respectively. These elements are as major or minor elements of sedimentary-hosted Pb–Zn deposits (Leach et al., 2001). The presence of Sb is unusual for this type of deposit, though it has been detected in some of the Pb–Zn deposits of the SSZ. The second cluster (H2) includes Ba, Sr, Ca, Mg and Mn, and is considered as a rock-forming element group associated with
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
189
Fig. 3. Raw data box plot of lead and zinc deposits from central part of SSZ.
limestone and dolomite host rocks, as well as with barite and calcite gangue minerals. Mn is present as a minor element. The first and second cluster ties reflect the genetic models for carbonate-hosted Pb–Zn deposits (Cox and Singer, 1986) depicting elements associated with MVT and SEDEX mineralization. The third cluster (H3) includes Al, K, Rb, Sc, Ti, V, Na, Nb, Ce, La and Y, elements associated with shale and volcanic rocks which host some of the Pb–Zn deposits in the central SSZ (Momenzadeh, 1976). The REE are mainly associated with monazite-bearing Jurassic shales (Alipour and Meshkani, 2005). The fourth cluster (H4), including As, Au, Mo, Co, Ni, Cu, P, Fe, Th and U, is mainly associated with MVT, SEDEX and massive sulphide deposits (Berger, 2000; Sangster and Leach, 1995). The metamorphosed shale and carbonate host rocks in the central SSZ contain Au,
Cu, Co and As mineralization (Ghorbani, 2007). The third and fourth clusters depict Pb–Zn deposits hosted by shale, sandstone and volcanic rocks, therefore they can guide exploration for these type of deposits in the district. For example, the Gol-e-Zard deposit is hosted by shale and sandstone and contains Cu, Au and Ag. Having identified meaningful clusters in the data using hierarchical cluster analysis, these data were used for further refining and discrimination of Pb–Zn deposits by the K-means technique. The clusters of ore- and rock-forming elements were transferred to the Statistica program and K-means calculations were carried out based on K = 4. The results are presented as a series of scatter plot in Fig. 7. Variation in this plot is related to differences in elements and their correlations. As an example, if we compare a Cu and a Pb–Zn deposit, both with dolomitic host rocks, they will fit into one group
Fig. 4. Normalized data box plot of lead and zinc deposits from central part of SSZ.
190
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
Fig. 5. Diagram illustrating number of clusters based on Hartigan (1975) formula.
based on rock-forming elements, but will be separated based on the ore-forming elements into two different groups. The Kc1 scatter plot is based on all elements and there is no meaningful separation, as depicted in Fig. 7a. The first, second and third clusters of the deposits show negative slopes in the scatter plot (Fig. 7a), while the fourth has a positive slope. Considering the database and dendrogram it was predicted that members of the fourth cluster have affinities to MVT and SEDEX deposits and are mainly located in the northwest of the SSZ. On the other hand, deposits in the first, second and third clusters show only MVT affinities and are mainly present in the central to southern part of the SSZ. The ore elements related to MVT deposits (Kc2), shown in Fig. 7b, depicting a meaningful division for four clusters. All depicted clusters follow the same trend except the fourth cluster (Fig. 7b). Variations of lead-related elements are almost insignificant in all clusters, while zinc-related elements show significant variations. Thus, there are deposits with similar Pb + Sb + Ag values but variable amounts of Zn + Cd in each cluster. These variations are probably due to geological and geochemical conditions for deposition, tectonic activity or emplacement of deposits in the central SSZ. Most of deposits grouped in the second and third clusters are large to medium in size while those in the fourth cluster are small deposits based on volume
or tonnage. It can be inferred that those deposits with low concentrations of both elemental cluster are small deposits. In the Kc3 diagram (Fig. 7c) there is a reasonable discrimination of deposits in the first and second clusters from those of the third and fourth clusters. It seems that in the host rock in the deposits of the first cluster we are dealing decrease in carbonate content, while those of the third and fourth clusters are hosted by carbonate, mainly dolomitic, host rocks. It should be noted that most of the deposits in the central SSZ are hosted by carbonate rocks and are well separated from shale, sandstone and volcanic hosted deposits. The Kc4 diagram (Fig. 7d) provides no meaningful separation between deposits, in spite of similar positive correlations, although it confirms that the REE are related to volcanic and shale hosted deposits. Group Kc5 (Fig. 7e) demonstrates a meaningful discrimination of deposits based on the negative and positive correlations of the first and second clusters with the third and fourth clusters respectively. The elemental concentrations of the third and fourth clusters are high, and the corresponding deposits are mainly located in northern part of the studied area. Two groups of deposits from the central SSZ, that were welldiscriminated by K-means clustering analysis, were selected for investigation of their spatial pattern of distribution in ArcGIS (Figs. 8 and 9). Ore elements, (Pb, Zn, Cd, Ag, and Sb) show significant differences between element associations (Fig. 8). In previous studies, these deposits were considered as one class of mineralization (Momenzadeh, 1976; Rasa and Kazemi Mehrnia, 2005), but the K-means clustering make it possible to separate them into meaningful groups. From the spatial distribution and mineral deposit zoning, it can be inferred that deposits of the second and fourth clusters are more frequent in the center of studied area, illustrated by Kc2 diagram results (Fig. 7b). Deposits of the first and third clusters are more frequent in the eastern segment of the studied area. Together these data show a distinctive mineral zonation in the central SSZ. The highest Pb+Sb+Ag concentration in mineral deposits is observed in the central segment of the study area and shows a decrease from the center to both east and west. The third cluster deposits (K3) are restricted to the center of the studied area, and have high concentrations of Pb + Sb + Ag, probably reflecting the neighboring volcanic rocks and plutonic intrusions (Fig. 8). Large and medium deposits occur mainly in this region and
Fig. 6. Dendrogram of hierarchical cluster analysis for major, minor and trace elements of lead and zinc deposits in central SSZ, separating elements into four distinct groups (H1 to H4).
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
191
Fig. 7. Scatter plot of each group of elements using K-means clustering to indicate the separation of mineral deposits in Kc1 (a), Kc2 (b), Kc3 (c), Kc4 (d) and Kc5 (e).
less commonly in the first and second clusters. These results suggest further exploration of occurrences and abandoned mines in the center of the studied area would be appropriate, as the potentials for medium to large Pb–Zn deposits is greatest here. Host rock elements (Mg, Ca, Sr, Ba and Mn) also show regional zonation (Fig. 9). A NW–SE trend controls the frequency of deposits of
the first to fourth cluster. This is supported by geological evidence as in the northern part of the studied area metamorphic rocks (recrystalized limestone, metasandstone and metavolcanics) are exposed and show an increase in metamorphic grade from SE toward NW (Ghorbani, 2002). Deposits of the fourth cluster are mainly hosted by dolomite and limestone which correlates well with the geology of the studied area.
192
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
Fig. 8. Symbol map showing lead and zinc deposits group based on K-means clustering (Kc2) after hierarchical clustering of (H1) in the central SSZ and the respective scatter plot.
8. Conclusion The results of our investigation shows that it is possible to discriminate and classify ore deposits at the regional scale by using clustering techniques and elemental associations, and so provides a practical exploration tool. It is also possible to investigate the spatial patterns of mineral deposits and use group of elements associations for more precise discrimination of ore deposit types. The Pb–Zn mineral deposits of the central SSZ were classified using multivariate cluster analysis. Using hierarchical clustering, elements
were grouped into four clusters, and ore elements and rock-forming elements were effectively separated. As, Au, Co, Ni, Cu, Fe, Cd, Sb, Ag, Pb, Zn and Mo clustered as an elemental association linked with Pb–Zn mineralization while Ca, Mg, Mn, Ba and Sr grouped as carbonate hosts rocks. The ore elements can be used as exploration tools in geochemical prospecting and exploration in the studied area to identify deposit types, not just anomalies. Associations of elements indicate the presence of MVT and SEDEX type mineralization in the central SSZ. Elemental clustering was also used for mineral deposit discrimination
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
193
Fig. 9. Symbol map showing lead and zinc deposits group based on K-means clustering (Kc3) after hierarchical clustering of (H2) in the central SSZ and the respective scatter plot.
by the K-means methods of Hartigan (1975). Based on Sb + Ag + Pb vs. Zn + Cd ratios, deposits separated into four clusters, with large to medium deposits less abundant in the first and second clusters. Therefore, the obtained clustered element association can be used as a geochemical exploration tool for large to medium deposits, identifying which abandoned mines, recognized occurrences and prospects are worthy of further detailed exploration. Based on Ca + Mg + Mn and Ba + Sr elemental ratios, likewise deposits clustered into four groups, indicating the role of carbonate host rocks in mineralization. The symbol map (Figs. 8 and 9) shows a well-defined spatial zoning of
mineral deposits occurrences in the studied area with E–W trends for ore elements and NW–SE trends for rock-forming elements. The study has demonstrated that multivariate techniques and sequential clustering are very effective in separating large and small deposits and in discriminating different type of neighboring deposits, which in previous classifications were considered as one type. As well, enhancing the classification of mineralization, the pattern recognition methods were able to uncover elemental zoning, identify key elemental associations for further geochemical exploration and indicate exploration target areas for large to medium size ore deposits.
194
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195
Acknowledgements The authors wish to thank Professor Bruce Yardley, School of Earth and Environment, Leeds University, for extensive review, constructive comments and significant improvement of the manuscript. We would also like to thank Professor Simon Pirc, Professor Ladoslav A. Palinkas and the other reviewers of the Journal of Geochemical Exploration whose constructive comments significantly improved the earlier version of manuscript. We also like to thank the Geological Survey of Iran (GSI) for the support during current project. Constructive comments and support of Mr. B. Borna, GSI, is highly appreciated. We are also thankful to Mr. A. Mousavi for drawing Fig. 1 and Mr. P. Afzal for helpful discussions. References Agard, P., Omrani, J., Jolivet, L., Mouthereau, F., 2005. Convergence history across Zagros (Iran): constraints from collisional and earlier deformation. Int. J. Earth Sci. 94, 401–419. Alavi, M., 1994. Tectonics of the Zagros orogenic belt of Iran: new data and interpretations. Tectonophysics 229, 211–238. Alavi, M., 2004. Regional stratigraphy of the Zagros fold–thrust belt of Iran and its proforeland evolution. Am. J. Sci. 304, 1–20. Alipour, A., Meshkani, S.A., 2005. Identification of monazite mineral in Jurassic shale of Iran. 22nd Symposium on Geosciences, Geological Survey of Iran, pp. 197–203. Alrik, G., Virlogeux, D., 1977. Petrographie et geochimie des rockes metamorphique et magmatiquse de la region de Deh Bid-Bavanat, chaine de Sanandadj–Sirjan. These 3 eme cycle,universite Scientifique et Medicale de Grenoble, France. 239 p. Anderberg, M.R., 1973. Cluster Analysis for Applications. Academic Press. Inc., New York. 203 pp. Berberian, M., 1973. Two important deformational and metamorphic phases in the belt northeast of the Zagros thrust line (Iran); a brief structural review of the Sanandaj– Sirjan belt. Geological Survey of Iran. internal report No. 27, 35 p. Berberian, F., Berberian, M., 1981. Tectono-plutonic episodes in Iran. In: Gupta, H.K., Delany, F.M. (Eds.), Zagros–Hindu Kush–Himalaya Geodynamic Evolution. Geodynamics Series, V. 3. American Geophysical Union, Washington, D.C, pp. 5–32. Berberian, M., King, G.C.P., 1981. Towards a paleogeography and tectonic evolution of Iran. Can. J. Earth Sci. 18 (2), 210–265. Berger, B.R., 2000. Classification of mineral deposits. In: Seal II, R.R., Foley, N.K. (Eds.), Geoenvironmental Models of Mineral Deposits: U.S Geological Survey Open-File Report 02-195. Box, G.E.P., Cox, D.R., 1964. An analysis of transformations. J. R. Stat. Soc. B 26, 211–252. Briggs, P.L., 1978. Pattern recognition applied to Uranium prospecting. Submitted to the Department of Earth and Planetary Sciences. 233pp. Burgos, P., Madejón, E., Pérez de Mora, A., Cabrera, F., 2006. Spatial variability of the chemical characteristics of a trace-element-contaminated soil before and after remediation. Geoderma 130, 157–175. Burnol, L., 1968. Contribution al'etude des gisments de plomb et zinc de L'Iran. Essais de classification paragenetique. Geological Survey of Iran Report No. 11. 113 p. Castillo-Munoz, R., Howarth, R.J., 1976. Application of the empirical discriminant function to regional geochemical data from the United Kingdom. Geol. Soc. Am. Bull. 87, 1567–1581. Collyer, P.L., Merriam, D.F., 1973. An application of cluster analysis in mineral exploration. Math. Geol. 4, 213–223. Cox, P., Singer, D.A., 1986. Mineral Deposit Model: U.S. Geological Survey Bulletin 1693. 379 p. Ehya, F., Lotfi, M., Rasa, I., 2010. Emarat carbonate-hosted Zn–Pb deposit, Markazi Province, Iran: a geological, mineralogical and isotopic (S, Pb) study. J. Asian Earth Sci. 37 (2), 186–194. Farhadi Nejad, T., Rastad, E., Hamedi, M.A., 1999. Mineralogy, geochemistry and genesis of Gol-e-Zard Zn–Pb deposit. 2nd Symposium of Geological Society of Iran, Mashad, pp. 328–329 (In Persian). Friedman, J.F., Meulman, J.J., 2004. Clustering objects on subsets of variables (with discussion). J. R. Stat. Soc. B 4, 815–849. Ghasemi, A., Talbot, C.J., 2006. A new tectonic scenario for the Sanandaj–Sirjan Zone (Iran). J. Asian Earth Sci. 26, 683–693. Ghazban, F., McNutt, R.H., Schwarcz, H.P., 1994. Genesis of sediment hosted Zn–Pb–Ba deposits in the Irankuh district, Esfahan area, west-central Iran. Econ. Geol. 89, 1262–1278. Ghorbani, M., 2002. Introduction to the economic geology of Iran. National Geosciences Database of Iran (NGDIR). 854 p. (In Persian). Ghorbani, M., 2007. Economic geology of mineral and natural resources of Iran, Vol. 1. Arian Zamin. 492 p. (In Persian). Ghorbani, M., 2008. Economic Geology of Mineral and Natural Resources of Iran, Vol. 2. Arian Zamin. 639 p. (In Persian). Granath, G., 1988. Pattern recognition in geochemical hydrocarbon exploration: a fuzzy approach. Math. Geol. 20, 673–691. Gustavsson, N., Bjorklund, A., 1976. Lithological classification of tills by discriminant analysis. J. Geochem. Explor. 29, 89–103. Haghipour, A., Aghanabati, A., 1985. Geological Map of Iran. 1:2,500,000. Geological Survey of Iran.
Hanesch, M., Scholger, R., Dekkers, M.J., 2001. The application of fuzzy c-means cluster analysis and non-linear mapping to a soil. Phys. Chem. Earth A. 26 (11–12), 885–891. Hartigan, J.A., 1975. Clustering Algorithms. John Wiley and Sons, New York. 108 pp. Hartigan, J.A., Wong, M.A., 1979. A K-means clustering algorithm, Journal of the Royal Statistical Society. C Appl. Statistics. 28, 100–108. Helsel, D.R., 2004. Non-detects and Data Analysis: Statistics for Censored Environmental Data. John Wiley and Sons, New York. Howarth, R.J., 1973. The pattern recognition problem in applied geochemistry. In: Jones, M.J. (Ed.), Geochemical Exploration 1972. Institution of Mining and Metallurgy, London, pp. 259–273. Ji, H., Zhu, Y., Wu, X., 1995. Correspondence cluster analysis and its application in exploration geochemistry. J. Geochem. Explor. 55, 137–144. Ji, H., Zeng, D., Shi, Y., Wu, Y., Wu, X., 2007. Semi-hierarchical correspondence cluster analysis and regional geochemical pattern recognition. J. Geochem. Explor. 93, 109–119. Leach, D.L., Bradley, D., Lewchuck, M.T., Symons, D.T.A., de Marsily, G., Brannon, J.C., 2001. Mississippi Valley-type lead–zinc deposits through geological time: implications from recent age-dating research. Miner. Deposita 36, 711–740. Leach, D.L., Sangster, D.F., Kelley, K.D., Large, R.R., Garven, G., Allen, C.R., Gutzmer, J., Walters, S., 2005. Sediment-hosted lead–zinc deposits: a global perspective. Economic Geology 100th Anniversary Volume. Society of Economic Geologists, Inc, pp. 561–607. Leisch, F., 1999. Bagged Clustering. Working Paper # 51, SFB “Adaptive Information Systems and Modeling in Economics and Management Science”. http://www. wuwien.ac.at/am. Lindqvist, L., Lundholm, I., Nisca, D., Esbensen, K., Wold, S., 1987. Multivariate geochemical modelling and integration with petrophysical data. J. Geochem. Explor. 29, 279–294. McQueen, J., 1967. Some methods for classification and analysis of multivariate observations. Fifth Berkeley Symp. Math. Stat. Probab. V1, 281–297. Meloun, M., Sánka, M., Nèmec, P., Krítková, S., Kupka, K., 2005. The analysis of soil cores polluted with certain metals using the Box–Cox transformation. Environ. Pollut. 137, 273–280. Meyer, C., 1981. Ore-forming processes in geologic history. Economic Geology. 75th Anniversary Volume, pp. 6–41. Meyer, C., 1988. Ore deposits as guides to the geologic history of the earth. Annu. Rev. Earth Planet. Sci. 16, 47–71. Modaberi, S., Rastad, E., 1998. Facies analysis as an approach for depositional environment of Ravanj Pb–Ag stratabound deposit, Delijan. 2nd Symposium of Geological Society of Iran, Mashad, pp. 443–444 (In Persian). Mohajjel, M., 1997. Structure and tectonic evolution of Palaeozoic–Mesozoic rocks, Sanandaj–Sirjan Zone, western Iran. PhD Thesis, University of Wollongong, Australia. Mohajjel, M., Fergusson, C.L., 2000. Dextral transpression in the late Cretaceous continental collision, Sanandaj–Sirjan Zone, Western Iran. J. Struct. Geol. 22, 1125–1139. Mohajjel, M., Fergusson, C.L., Sahandi, M.R., 2003. Cretaceous–Tertiary convergence and continental collision, Sanandaj–Sirjan Zone, western Iran. J. Asian Earth Sci. 21, 397–412. Momenzadeh, M., 1976. Stratabound lead zinc ores in the lower Cretaceous and Jurassic sediments in the Malayer–Esfahan District (West Central Iran): Litology, metal content, zonation and Genesis, PhD Thesis, University of Heidelberg, 300 pp. Moritz, R., Ghazban, F., Singer, B., 2006. Eocene gold ore formation at Muteh, Sanandaj– Sirjan tectonic zone, western Iran: a result of late-stage extension and exhumation of metamorphic basement rocks within the Zagros orogen. Econ. Geol. 101, 1497–1524. Raftery, A.E., Dean, N., 2004. Variable selection for model-based clustering. Technical Report 452, Department of Statistics, University of Washington. Rasa, I., Kazemi Mehrnia, A.K., 2005. Carbonate-hosted Base Metal Deposits. Roozbahan Pub.co, Tehran. 262 pp. Rastad, E., Fontbote, L., Amstutz, G.C., 1980. Relation between tidal flat facies and diagenetic ore fabrics in the strataband Pb–Zn–(Ba–Cu) deposits of Irankuh, Esfahan, West Central-Iran. Revista del Institu to de Investigationes Geological Diputation provincial Universsidad de Barcelona 34, 311–323. Reimann, C., Filzmoser, P., 2000. Normal and lognormal data distribution in Geochemistry: death of a myth, consequences for the statistical treatment of geochemical and environmental data. Environ. Geol. 39, 1001–1014. Reimann, C., Garrett, R., 2005. Geochemical background — concept and reality. Sci. Total Environ. 350, 12–27. Sahandi, M.R., Soheily, M., Sadeghi, M., Delavar, S.T., Jafari-Rad, A., 2002. 1:1,000,000 Geological Map of Iran. Geological Survey of Iran, Tehran. Sanford, R.F., Pierson, C.T., Crovelli, R.A., 1993. An objective replacement method for censored geochemical data. Math. Geol. 25, 59–80. Sangster, D., Leach, D.L., 1995. Evidence for a genetic link between SEDEX and MVT deposits. In: Leach, D.L., Goldhaber, M.B. (Eds.), Extended Abstracts, International Field Conference on Carbonate-hosted Lead–Zinc Deposits, St. Louis Missouri, June 1–4, pp. 260–263. Sengör, A.M.C., 1987. Tectonics of the Tethysides: orogenic collage development in a collisional setting. Annu. Rev. Earth Planet. Sci. 15, 213–244. Stampfli, G.M., Borel, G.D., 2002. A plate tectonic model for the Paleozoic and Mesozoic constrained by dynamic plate boundaries and restored synthetic oceanic isochrons. Earth Planet. Sci. Lett. 196, 17–33. Stockli, D.F., Hassanzadeh, J., Stockli, L.D., Axen, G., Walker, J.D., Dewane, T.J., 2004. Structural and geochronological evidene for oligo-miocene intra-arc low-angle detachment faulting in Takab–Zanjan area, NW Iran. Geological Society of America, Abstract with program V36, 319.
S.A. Meshkani et al. / Journal of Geochemical Exploration 108 (2011) 183–195 Stocklin, J., 1968. Structural history and tectonics of Iran: a review: American Association of Petroleum. Geol. Bull. 52, 1229–1258. Swan, A.R.H., Sndilands, M., McCabe, P., 1995. Introduction to Geological Data Analysis. Blackwiell Science. 446 pp. Taherian, A., 1993. Mineralogy, geochemistry and genesis of Anjireh–Tiran Pb–Zn deposit, W. Najaf Abad. Unpublished MSc. Thesis, Tehran Tarbiat Moallem University 287 p. (In Persian). Templ, M., Filzmoser, P., Reimann, C., 2008. Cluster analysis applied to regional geochemical data: problems and possibilities. Appl. Geochem. 23, 2198–2213.
195
Thiele, O., Alavi-Naini, M., Assefi, R., Houshmandzadeh, A., Seyed-Emami, K., Zahedi, M., 1968. Explanatory text of the Golpaygan Quadrangle E7. 1:250,000. Geological Survey of Iran, Tehran, Iran. 24 p. Xie, X., Liu, D., Xiang, Y., Yan, G., Lian, C., 2004. Geochemical blocks for predicting large ore deposits — concept and methodology. J. Geochem. Explor. 84, 77–91. Zamanian, H., 1993. Mineralogy, paragenesis and genesis of Ahangaran Ag–Pb deposit, Malayer. Unpublished MSc. Thesis, Tehran Tarbiat Moallem University 280 p. (In Persian).