Introduction

Different data integration algorithms, which have been used for mineral potential mapping in recent years, are categorized as either data driven or knowledge driven (Bonham-Carter 1994; Pan and Harris 2000; Carranza 2009, 2011; Abedi et al. 2013). Examples of data-driven techniques are artificial neural networks (Porwal et al. 2003a; Abedi and Nouruzi 2012), weights-of-evidence (Bonham-Carter et al. 1989; Asadi and Hale 2001; Ford et al. 2015), support vector machine (Zuo and Carranza 2011; Abedi et al. 2012b; Geranian et al. 2015), and random forests (Carranza and Laborte 2015a, b, c; McKay and Harris 2015). In these and other data-driven techniques, known mineral deposits in a region of interest are used as ‘training points’ to establish spatial relationships with particular geological, geochemical, and geophysical features (Carranza et al. 2008; Abedi et al. 2013). These methods are appropriate for well-explored regions and the trained model is used for extracting features that are similar to those associated with known mineral deposits in a study area. Knowledge-driven techniques are suitable for greenfields or less-explored regions (Carranza 2011). In these techniques, the prepared predictor maps are combined based on the expert opinion. The most commonly used methods of this category are index overlay (Bonham-Carter et al. 1989; Carranza et al. 1999; Yousefi and Carranza 2015) and fuzzy logic (An et al. 1991; Porwal et al. 2003b; Abedi et al. 2012c). Porwal et al. (2004) and Asadi et al. (2015) used a hybrid fuzzy-neural network data integration modeling. In this type of modeling, expert knowledge is used for weighing the spatial proxies, and the weights are fine-tuned using a neural network which is trained by known mineral occurrences.

Mineral potential mapping is a multi-criteria decision making (MCDM) activity (Abedi et al. 2012a, d, 2013). Various MCDM algorithms have been used as a tool for mineral potential mapping by Pazand et al. (2011, 2012) and Abedi et al. (2012a). They proposed ELECTRE (Abedi et al. 2012d) and PROMETHEE (Abedi et al. 2012e) algorithms as knowledge-driven techniques for mineral potential mapping and used these methods for finding the best location for extra drilling in a copper deposit. Pazand et al. (2012) proposed the TOPSIS method as a tool for preparing prospectivity maps for porphyry copper exploration. The analytic hierarchy process (AHP) is another MCDM technique, which needs a pairwise comparison matrix for determination of weights of exploration criteria, that is used in mineral potential mapping (Hosseinali and Alesheikh 2008; Pazand et al. 2011; Abedi et al. 2013). Pazand and Hezarkhani (2015) proposed a combined AHP–TOPSIS method to map porphyry Cu potential in Siahrud area, NW Iran.

In this paper, the integrated AHP–TOPSIS algorithm, initially proposed by Pazand and Hezarkhani (2015), was used to map high-potential zones in an area in central Iran with very few known mineral deposit occurrences. The proposed algorithm consists of two main steps. In the first step, the criteria weight vector is obtained from the AHP. In the second step, the TOPSIS algorithm is used for integration of different data sets. TOPSIS is based on the distance of attributes (pixels are considered as attributes) to positive (best alternative) and negative (the worst alternative) ideal solution. The ideal solutions are determined by the user or from existing data. Simplicity and no need of prior knowledge by decision maker are the most important advantages of this method in comparison to other MCDM algorithms such as ELECTRE (Abedi et al. 2012d), PROMETHEE (Abedi et al. 2012e), and the well-known fuzzy logic method (Porwal et al. 2003a, b). The only requirement in the TOPSIS method is the criteria weight vector as other knowledge-driven algorithms that are determined in the first step by AHP method.

The proposed algorithm is demonstrated here for mapping favorable areas of porphyry Cu–Au mineralizations in the northern part of the 1:100,000 scale Salafchegan geological map sheet, which contains very few known porphyry Cu–Au occurrences. The Salafchegan area is a sparsely vegetated and mountainous region, located in the central part of Urumieh–Dokhtar volcanic arc (UDMA), the main porphyry copper belt of Iran (Fig. 1). Dalli, with a total inferred resource of 20 million tons at 0.5 % Cu and 0.65 g/t Au, is the largest porphyry Cu–Au deposit in the central UDMA. Spectral remote-sensing and catchment basin analysis of stream sediment geochemical data were employed, respectively, to map hydrothermal alterations and geochemical signatures that could be associated with this type of mineralization. Magnetic data were interpreted to map intrusive rocks and structures associated with porphyry Cu–Au mineralization. Then, an integrated AHP–TOPSIS model was used as an innovative integration modeling approach to integrate geological, spectral, magnetic, and geochemical datasets. This integration model takes advantages of the MCDM, and the results were used to construct a predictive model to locate high-potential areas of porphyry Cu–Au mineralizations for district-scale exploration in the study area.

Figure 1
figure 1

Generalized geologic map of Iran (modified after Aghanabati (2004) and location of study area

Full size image

Background Theory of Proposed AHP–TOPSIS Algorithm

In the AHP–TOPSIS data integration approach, the weight vector of criteria for input exploration data is calculated primarily using the AHP method. Then, if an acceptable consistency ratio is obtained, the alternatives are ranked using the TOPSIS method. The background theory of AHP and TOPSIS methods is explained in the following paragraphs.

AHP

The AHP was proposed by Saaty (1977, 1980) to model subjective decision-making processes based on multiple attributes in a hierarchical system (Tzeng and Huang 2011). In the AHP method, a decision problem is considered as a hierarchical structure and, according to their properties, it is decomposed into elements and levels corresponding to the common characteristic of the elements. The first level is the final goal of the problem, and the intermediate levels correspond to criteria and sub-criteria, while the lowest level contains the decision alternatives.

The elements of each level are compared pairwise to a specific element in the immediate upper level. Table 1 shows the pairwise comparison scale of real numbers from 1 to 9 used in the AHP that was first developed by Saaty (1977). It allows converting the qualitative judgments into numerical values with intangible attributes.

Table 1 Ratio scales in AHP method (from Saaty 1977)
Full size table

The pairwise comparison matrix for “n” criteria {c 1, c 2, c 3, …, c n } in the second level, with respect to the final goal, is formed as

$$A = \left[ {\begin{array}{cccc} {{a_{11}}}&{{a_{12}}}& \cdots &{{a_{1n}}}\\ {{a_{21}}}&{{a_{22}}}& \cdots &{{a_{2n}}}\\ \cdots & \cdots & \cdots & \cdots \\ {{a_{n1}}}&{{a_{n2}}}& \cdots &{{a_{nn}}} \end{array}} \right],$$

where a ij represents the pairwise comparison rating between the elements i and j of a level with respect to the upper level. The entries a ij are governed by the following rules:

$${a_{ij}}{\rm{ > }}\;0;\quad {a_{ij}}{\rm{ = }}\;1/{a_{ji}};{\mkern 1mu} \quad {a_{ii}} = 1\;\forall \;i.$$

There are some simple methods for calculating the weight vector from the pairwise comparison matrix A, and the eigenvalue approach proposed by Saaty (1977, 1980) is one of them. Based on the eigenvalue approach (Saaty 1980, 2000), the weight vector can be estimated by finding the principal eigenvector w of the matrix A, thus

$$Aw = {\lambda _{{\rm{max}}}}w.$$

When the vector w is normalized, it becomes the weight vector of elements of one level with respect to the upper level. \({\lambda _{{\rm{max}}}}\) is the largest eigenvalue of the matrix A. In cases where the pairwise comparison matrix satisfies transitivity for all pairwise comparisons, it is said to be consistent and it verifies the following relation:

$${a_{ij}} = {a_{ik}} \times {a_{kj}};\quad \forall i,\;j,\;k.$$

Saaty (1980) has shown that to maintain reasonable consistency when deriving weight vector from paired comparisons, the number of factors being considered must be less than or equal to nine. AHP allows inconsistency, but provides a measure of the inconsistency in each set of judgments. The consistency of the judgmental matrix can be determined by a measure called the consistency ratio (CR), defined as

$${\rm{CR}} = \frac{{{\rm{CI}}}}{{{\rm{RI}}}},$$

where CI is called the consistency index and RI is the random index. Furthermore, Saaty (1980, 2000) provided average consistencies (RI values) of randomly generated matrices (Table 2). The CI for a matrix of order n is defined as

$${\rm{CR}} = \frac{{{\lambda _{{\rm{max}}}} - n}}{{n - 1}}.$$
Table 2 Values of RI for different matrix sizes (Tzeng and Huang 2011)
Full size table

In general, a consistency ratio of 0.1 or less is considered acceptable. If the value is higher, the judgments may not be reliable and should be elicited again.

TOPSIS

TOPSIS is a well-known ranking method for MCDM that was proposed by Hwang and Yoon (1981), and then developed by Lai et al. (1994) and Yoon and Hwang (1995). The logic of the TOPSIS approach is to define ideal and anti-ideal solutions (Liang 1999), which are based on the concept of relative closeness in compliance with that the shorter (or longer) the distance of an alternative to ideal (or anti-ideal), the higher the priority ranking (Zeleny 1982).

On the basis of many criteria to rank, various situations are evaluated and compared for the MCDM problems. Therefore, a set of alternatives A = {a, b, c,…} are evaluated by n criteria C = {c 1, c 2, c 3,…} and the aim is to find the best alternative in the set of A based on the comparison of criterion vector C.

TOPSIS has a simple algorithm with attractive property in that limited subjective input is needed from decision makers (Jahanshahloo et al. 2006; Tzeng and Huang 2011; Opricovic and Tzeng 2004). The only subjective input needed is weights (Olson 2004), which, in this paper, are determined by AHP method. The procedure of TOPSIS can be summarized in the following simple steps (Hwang and Yoon 1981; Jahanshahloo et al. 2006; Tzeng and Huang 2011):

  1. (1)

    Establish a decision matrix for the ranking. The structure of the matrix can be expressed as follows:

    $$ \begin{aligned} \begin{array}{*{20}c} {c_{1} } & {} & {c_{2} } &{}& {}{}& & {c_{j} } & {} & {} &{} & {c_{n} } & {} \\ \end{array} \hfill \\ D = \begin{array}{*{20}c} {A1} \\ {A2} \\ . \\ . \\ {Ai} \\ . \\ {Am} \\ \end{array} \left[ {\begin{array}{*{20}c} {g_{{11}} } & {g_{{12}} } & \cdots & {g_{{1j}} } & \cdots & {g_{{1n}} } \\ {g_{{21}} } & {g_{{22}} } & \cdots & {g_{{2j}} } & \cdots & {g_{{2n}} } \\ . & . & \cdots & . & \cdots & . \\ . & . & \cdots & . & \cdots & . \\ {g_{{i1}} } & {g_{{i2}} } & \cdots & {g_{{ij}} } & \cdots & {g_{{in}} } \\ . & . & \cdots & . & \cdots & . \\ {g_{{m1}} } & {g_{{m2}} } & \cdots & {g_{{mj}} } & \cdots & {g_{{mn}} } \\ \end{array} } \right] \hfill \\ \end{aligned} $$

    where Ai denotes the alternatives i, i = 1, 2, 3, …., m; \({c_j}\) represents jth attribute or criterion, j = 1, 2, …, n, related to ith alternative; and \({g_{ij}}\) is a crisp value indicating the performance rating of each alternative \({A_i}\) with respect to each criterion \({c_j}.\)

  2. (2)

    Calculate the normalized decision matrix. The normalized value r ij is calculated as

    $${r_{ij}} = {g_{ij}}/\sqrt {\sum\limits_{i = 1}^m {g_{ij}^2} } ;\quad i = 1,2, \ldots,m;\quad j = 1,2, \ldots ,n.$$
  3. (3)

    Calculate the weighted normalized decision matrix. The weighted normalized value v ij is calculated as

    $${v_{ij}} = {w_j}{r_{ij}};\quad i = 1,2, \ldots ,m;\quad j = 1,2, \ldots ,n,$$

    where \({w_j}\) is the weight of the ith attribute or criterion and \(\sum\limits_{j = 1}^n {{w_j}} = 1.\)

  4. (4)

    Determine the positive ideal and negative ideal solutions.

    $$\begin{array}{l} {A^*} = \{ v_1^*, \ldots ,v_n^*\} = \{ (\mathop {{\rm{max}}}\limits_i {v_{ij}}\left| {j \in {J^\prime }),} \right.(\mathop {{\rm{min}}}\limits_i {v_{ij}}\left| {j \in {J^\prime }^\prime )\} } \right.\\ {A^ - } = \{ v_1^ - , \ldots ,v_n^ - \} = \{ (\mathop {{\rm{min}}}\limits_i {v_{ij}}\left| {j \in {J^\prime }),} \right.(\mathop {{\rm{max}}}\limits_i {v_{ij}}\left| {j \in {J^\prime }^\prime )\} } \right., \end{array}$$

    where \({J^\prime }\) is the set of benefit criteria and \({J^{\prime \prime }}\) is the set of cost criteria.

  5. (5)

    Calculate the separation measures using the n-dimensional Euclidean distance. The separation of each alternative from the ideal solution and negative ideal solution are given as

    $$D_i^{*} = \sqrt {\sum\limits_{j = 1}^n {{{({v_{ij}} - v_j^{*})}^2}} } ;\quad i = 1,2, \ldots ,m.$$
    $$D_i^{ -} = \sqrt {\sum\limits_{j = 1}^n {{{({v_{ij}} - v_j^ {-} )}^2}} } ;\quad i = 1,2, \ldots,m.$$
  6. (6)

    Calculate the relative closeness to the ideal solution. The relative closeness of the alternative A i with respect to A* is defined as

    $${\rm{RC}} = \frac{{D_i^{ - }}}{{(D_i^{ - } + D_i^{*})}},\quad i = 1,2, \ldots ,m.$$
  7. (7)

    Ranking the alternative based on relative closeness.

    The best alternative is the one that has the higher RC. By sorting the calculated relative closeness, the best alternatives are ranked.

Mineral Potential Targeting Elements

Genetic and spatial information from the known Dalli porphyry Cu–Au deposit of the study area as well as information from typical porphyry Cu–Au genetic modeling was used to identify the targeting elements (exploration criteria) for district-scale exploration in the Salafchegan area (Table 3).

Table 3 Characteristics of Dalli Cu–Au porphyry deposit compared with typical Cu–Au porphyry deposits
Full size table

Two sets of characteristics were defined from the generic Cu–Au porphyry deposit models and the Dalli deposit and conjunctively used to recognize the key targeting elements for district-scale exploration of Cu–Au porphyry mineralization in the Salafchegan area. These characteristics are as follows: (1) host lithologies of Miocene diorite, quartz-diorite to granodiorite intrusions and their contact with Eocene andesite porphyry; (2) silicic, potassic, and phyllic hydrothermal alterations surrounded by argillic and propylitic alterations; (3) ferric iron oxides of mostly hematite and goethite; (4) strong copper and gold geochemical anomalies associated with arsenic, iron, and potassium; (5) strong magnetic signatures associated with the host intrusions and mineralized potassic alterations; and (6) NE–SW trending structures.

Then, the geological map (1:100,000 scale), and Enhanced Thematic Mapper (ETM+) and Advanced Spaceborne Thermal Infrared (ASTER) satellite imagery data, as well as geochemical, geophysical, and field data were utilized to search for the above-mentioned porphyry Cu–Au targeting elements.

Lithology

A wide range of igneous rocks are host to porphyry Cu (Au, Mo) mineralization. Singer et al. (2005, 2008) compiled the host rocks of 407 porphyry copper deposits and indicated that intermediate intrusive rocks are the most important and frequent hosts to porphyry Cu (Au, Mo) mineralization (Fig. 2).

Figure 2
figure 2

Histogram of major host rocks of porphyry Cu (Au, Mo) deposits (Singer et al. 2005, 2008)

Full size image

Mineralization at Dalli is hosted by altered Miocene intermediate intrusive rocks, such as quartz-diorite, that intruded Eocene andesite porphyry. High-grade mineralization occurs at the contact of the quartz-diorite intrusion and andesite porphyry. A northeast-trending structural corridor provided the favorable site for porphyry intrusion and pathway for mineralizing hydrothermal solutions. The alteration system covers an area of 20 km2 and mostly comprises mineralized potassic and phyllic alterations, surrounded by extensive barren argillic (kaolinite) and propylitic (chlorite) alterations (Ayati et al. 2013).

Based on the above-mentioned information, the 1:100,000 scale geological map of the study area was reclassified into eight lithological classes. Some of these classes are favorable hosts to porphyry Cu–Au mineralization. Figure 3 shows the reclassified geological map of the area overlain by the known porphyry occurrences.

Figure 3
figure 3

Reclassified geological map of the study area

Full size image

Structures

Structures provide pathways for mineralizing fluids and porous zones for trapping metals. Faults, extracted from the 1:100,000 scale geological map of the study area, were used to prepare a lineament density raster map with a grid-cell size of 100 m (Fig. 4). The known porphyry Cu–Au deposits/prospects of the study area (e.g., Dalli and Zavarian) occur in zones with moderate lineament density (Fig. 4).

Figure 4
figure 4

Lineament density map of the Salafchegan area

Full size image

Hydrothermal Alteration

Hydrothermal alterations and iron oxides are important distal or direct features in porphyry Cu–Au exploration. In recent years, ASTER and ETM+ satellite imagery data have been extensively used for mapping alterations (Abrams et al. 1983; Amos and Greenbaum 1989; Sabins 1999; Galvao et al. 2005; Rowan et al. 2005; Gabr et al. 2010; Pour et al. 2011, Honarmand et al. 2012; Pour and Hashim 2012; Feizi and Mansuri 2013; Shahriari et al. 2013; Pournamdari et al. 2014; Shahriari et al. 2014). The main hydrothermal alterations associated with Dalli and other typical porphyry Cu–Au deposits are potassic, silica, hydroxyl-bearing clay, phyllic, and propylitic alterations. Iron oxides, closely associated with these alterations, are also important features of interest in porphyry Cu–Au exploration.

Spectral angle mapper (SAM) (Kruse et al. 1993) and Least Square Fit (Ls-Fit) methods were applied to ASTER and ETM+ satellite imagery data to map hydrothermal alterations and iron oxides. Hydroxyl clay minerals such as kaolinite (argillic alteration) and iron oxides were mapped from Landsat ETM+ data using the LS-Fit method, while chlorite and silicification were mapped from ASTER data using the SAM method. The hydrothermal alterations including silicification, OH-bearing clays, and chlorite were combined with mapped iron oxides to be used as an important input map in the data integration modeling (Fig. 5).

Figure 5
figure 5

Hydrothermal alterations and iron oxides mapped from Landsat ETM+ and ASTER data

Full size image

Geochemistry

A sample catchment basin analysis (CBA) was conducted using stream sediment geochemical data in the study area to generate geochemical anomalies associated with copper mineralization. Two hundred fifty stream sediment samples were collected and analyzed for 12 elements by the Geological Survey of Iran using ICP method at Zar Azma and Kan Pajuh laboratories in Iran. The samples were analyzed for Cu among those 12 elements, but not for gold or other important elements associated with Cu–Au mineralization of the study area. Therefore, we only used the Cu analytical results for mapping stream sediment geochemical anomalies for data integration modeling. To obtain a raster map of Cu anomaly, catchment basins were prepared using the locations of stream sediment samples and digital elevation model of the area. Then, the Cu concentration of each sample is assigned to its related catchment basin. Dalli and Zavarian and their surrounding areas show strong catchment basin Cu anomalies (Fig. 6).

Figure 6
figure 6

Sample catchment basin geochemical Cu map

Full size image

Magnetic Features

Magnetic data have been widely used to map porphyry intrusive rocks, structures, and hydrothermal alterations in porphyry Cu–Au exploration (Thoman et al. 2000, Holden et al. 2011; Abedi et al. 2013). According to Clark (2014), induced magnetic anomalies, caused by Au-rich porphyry Cu systems, have specific semi-predictable characteristics. A magnetic intrusion and the associated magnetite-bearing potassic alteration zone, typically create a circular positive magnetic anomaly. This sub-circular magnetic high is surrounded by an annular negative magnetic anomaly caused by propylitic and phyllic alteration zones. This negative magnetic anomaly is caused by destruction of magnetite in the volcanic country rock. The distinctive circular magnetic peak at its center contrasts sharply with the surrounding zones (Holden et al. 2011). Therefore, high-amplitude magnetic anomalies are expected to occur in porphyry intrusive rocks and potassic alteration in the porphyry Cu–Au deposits.

Recently, Holden et al. (2011) proposed an automatic circular feature extraction to enhance the magnetic anomalies related to porphyry copper deposits. On the other hand, derivative-based methods have been used for many years in magnetic data interpretation. In this paper, we used the tilt angle method as a tool for enhancing the magnetic anomalies, associated with porphyry intrusive rocks (Miller and Singh 1994; Verduzco et al. 2004). The tilt angle, T, is a useful filter for enhancing subtle anomalies in potential field data and was first proposed by Miller and Singh (1994) as

$$T = {\tan ^{ - 1}}\left[ {\frac{{\partial M}}{{\partial z}}/\sqrt {{{\left( {\frac{{\partial M}}{{\partial x}}} \right)}^2} + {{\left( {\frac{{\partial M}}{{\partial y}}} \right)}^2}} } \right],$$

where \(\frac{{\partial M}}{{\partial x}},\) \(\frac{{\partial M}}{{\partial y}},\) and \(\frac{{\partial M}}{{\partial z}}\) are the first-order derivatives of the magnetic field M in the x, y, and z directions, respectively. The tilt angle has many interesting properties, for example, due to the nature of the arctan trigonometric function, all tilt amplitudes are restricted to values between −90° and +90° regardless of the amplitude of the vertical or the absolute value of the total horizontal gradient (Salem et al. 2007). The positive values are located on the magnetic body and zero value along its boundary. This method can also be applied to the reduced-to-the-pole (RTP) magnetic map. In the RTP map, the inclination effect of magnetic field is corrected and high values are located on the magnetic body, while in the total magnetic intensity map magnetic sources show a dipolar anomaly and maximum values have a distance with the center of the mass.

In this study, the aeromagnetic data were extracted from the nation-wide airborne magnetic data of Iran (flight-line spacing of 7.5 km). Figure 7 shows the RTP and tilt angle maps of the study area. The tilt angle map highlighted positive magnetic anomalies that can be related to porphyry intrusive rocks associated with porphyry Cu–Au mineralization (Fig. 7).

Figure 7
figure 7

Reduced-to-the-pole (RTP) magnetic map (left) and tilt angle of RTP magnetic data (right)

Full size image

Data Integration Modeling

An integrated AHP–TOPSIS method was used to combine the input exploration data and create a final porphyry Cu–Au prospectivity map of the Salafchegan area. First, a hierarchy structure for porphyry Cu–Au exploration in the study area was designed using the input exploration maps (Fig. 8). Then, a pairwise comparison matrix was created and five input criteria were compared to indicate porphyry Cu–Au potential of the area. On the basis of the calculated weights, the relative importance of each of the criteria with respect to the porphyry Cu–Au potential is shown in Table 4. The criteria weight vector was calculated from the pairwise comparison matrix and the AHP method. The consistency ratio for this pairwise comparison is CR = 0.017 which is less than 0.1 and thus this value is acceptable (Tzeng and Huang 2011).

Figure 8
figure 8

Hierarchical structure for mapping the porphyry Cu–Au potential of the study area. Geology classes are described in Table 4. Each pixel has attributes A1 to An

Full size image
Table 4 Pairwise comparison matrix of the main criteria used for mapping porphyry Cu–Au potential
Full size table

The geological and alteration criteria have some sub-criteria that their relative weights must be determined. The geological criterion consists of seven sub-criteria (lithological units). The comparison matrix for the lithological units with respect to the geological criterion was prepared using the targeting elements generated from the Dalli Cu–Au porphyry deposit and generic porphyry Cu–Au deposit models (Table 3). Then, the weights for the seven lithological classes were calculated using the AHP method (Table 5).

Table 5 Comparison matrix of the weights for the seven lithological classes estimated using the AHP
Full size table

The alteration criterion has three sub-criteria including the iron oxides, OH-bearing clay, and propylitic alterations. The relative weights of these alteration sub-criteria were also calculated using the AHP method (Table 6).

Table 6 Comparison matrix of weights for the alteration sub-criteria estimated using AHP
Full size table

For applying the proposed AHP–TOPSIS algorithm, cell values of raster maps associated with the five criteria and lithological and alteration sub-criteria were extracted to make a decision matrix. Using the developed MATLAB codes for the proposed AHP–TOPSIS method, the relative closeness to the ideal solution for each alternative (cell pixels) was calculated.

A final porphyry Cu–Au prospectivity map was created and reclassified into three classes. The thresholds for classification of favorability map were determined using the favorability—area fractal modeling in a similar way to concentration—area fractal modeling mentioned in literatures (Agterberg 1995; Cheng 1999; Carranza 2009). Figure 9 shows the log–log plot of fractal result, which indicates three straight lines with breaks in C1 and C2 that are considered as thresholds for classification. Figure 10 shows the porphyry Cu–Au prospectivity map, which accurately depicts the prominent porphyry Cu–Au mineralized zones (Dalli and Zavarian) to be located in the high favorability class. In addition, an area with high potential (similar to Dalli and Zavarian porphyry Cu–Au mineralized areas) was determined in the northwest of the area. This area needs follow-up exploration.

Figure 9
figure 9

Log–log plot of favorability area

Full size image
Figure 10
figure 10

Favorability map obtained from the combination of various geoscience datasets through the integrated AHP–TOPSIS method

Full size image

Discussion and Conclusion

In this paper, an innovative data integration approach based on an MCDM method (i.e., the integrated AHP–TOPSIS) is proposed for mineral potential mapping. This method is based on the two most popular MCDM algorithms known as AHP and TOPSIS. In this algorithm, the criteria weight vector is obtained based on the pairwise comparison matrix and AHP algorithm and the alternatives are ranked by TOPSIS. The pairwise comparison matrix is prepared by expertise in porphyry Cu–Au potential mapping. After that, the alternatives (the raster map pixels) are ranked by the TOPSIS algorithm. This strategy has some similarity to both data-driven and knowledge-driven algorithms. It needs a criteria weight vector as a knowledge-driven algorithm, but it does not employ an inference system as in knowledge-driven methods (e.g., fuzzy logic). The ranking strategy is only based on the distance of alternatives to positive and negative ideals, similar to the fuzzy c-means clustering (for two clusters) whereby the similarity of each alternative is measured by comparison to cluster prototypes (the prototypes of clusters are positive and negative ideal solutions). Therefore, the proposed method is not a pure knowledge- or data-driven method, but is a hybrid method. The proposed algorithm only needs the comparison matrix as primary information for obtaining the criteria weight vector, whereas other knowledge-driven methods, such as fuzzy logic, need more primary information of the membership function for fuzzification of input data in addition to criteria weight vector.

The AHP–TOPSIS algorithm was used for integrating datasets from a greenfield area with very few known deposits in the central part of Urumieh–Dokhtar magmatic arc, the main volcanic arc and porphyry copper belt of Iran. Five exploration layers (hydrothermal alterations, structural features, catchment basin geochemical anomaly map of copper, total magnetic intensity map of airborne data, and 1:1000000 scale lithological map) of the study area were prepared and integrated to create a final favorability map. On the basis of the fractal method, the final favorability map was classified as three classes of high favorable, favorable, and low favorable zones. The high favorable classes cover about 0.05 percent of the study area and contain the three known prominent deposits, namely Dalli south and Dalli north porphyry Cu–Au deposits and the Zavarian 1 porphyry Cu–Au prospect. The favorable classes cover about 1.5 percent of the study area and Zavarian 2 porphyry Cu–Au prospect locates in these classes. Kahak, which is an iron-skarn deposit, lies in the low favorable class. Several unknown high favorable and favorable areas are now recognized northeast of the Dalli deposits and east of the Zavarian deposits. Follow-up exploration of these high-potential areas is recommended.