--> Application --> Geology

Alteration Extraction Using Remote Sensing Data for Mineral Exploration

Maryam Dehghani
Remote Sensing Specialist
Remote Sensing Group, Geological Survey of Iran (GSI), Meraj St., Azadi Sq., Tehran, Iran
Tel: +98 21 6459217, Fax: +98 21 877 9476

Claude Durocher
31 Prennan Avenue, Etobicoke, Ontario, M9B 4B7
Tel: (416) 234-9053, Fax: (416) 233-8284

John Gingerich
5623 Goldenbrook Drive, Missisagua, Ontario L5M 3W2, Canada
Tel: 905-821-2872 Cell: 416-450-2784

For more than one decade, remote sensing technology has been an important tool for mineral exploration in Iran. Satellite images can be widely used in different geological applications as they provide information referred to the ground surface as well as in different parts of electromagnetic spectrum. The main goal of this paper is to demonstrate the superiority of the remote sensing technology in mineral exploration in Khoy-Oshnavieh area located in the north-west of Iran. Two primary components, VMS and Au, were considered to be explored in the region. Several kinds of remote sensing data were used to prepare different information layers which may play important roles in mineral exploration. The most important layers are lineaments and structures in the study area as well as alteration. Although there are some limitations related to the quality of the remote sensing data, they seemed to be the only reliable data in this project. This paper mainly focuses on techniques used to process different remote sensing data to extract different types of alteration, clay and iron-oxide, using ETM+ and ASTER. The methods that were used in this project consisted of Crosta, selective principal component analysis and band ratio for alteration extraction. The alteration information and other information layers such as geology maps, geophysics and geochemistry data were then integrated to produce the primary exploration model in GIS. Two different models including index overlay and fuzzy logic were tested to find the promised areas as targets to be checked. The quality of the results was evaluated after checking the field. Remote sensing alteration mapping proved to be the most reliable source of information.

1. Introduction
Mineral exploration is one of the goals of the Geomatics Group in Geological Survey of Iran. The study area to be studied is Khoy-Oshnavieh that is one of 20 regions of Iran identified by the Geologic Survey of Iran (GSI) during their strategic exploration review as having significant mineral potential. The Khoy-Oshnavieh study area is located in the Northwestern corner of Iran along the Iraq-Turkey-Armenia border. Given the large area and limited data for the region, it was decided to undertake a detailed Geomatics study of the area to assist exploration target generation. Remote sensing and GIS data modeling technology have the potential to accelerate discovery through lower cost and rapid identification of the most prospective geology.

The mineral potency of the study area was evaluated according to two primary components, VMS and Au. The primary objective of this project was to use the GSI database containing different sorts of information such as remote sensing, geology, geochemistry and geophysics, to develop Geomatics targets in Khoy-Oshnavieh area. One of the challenges is to develop the approaches in the use of Geomatics tools to get more robust results in exploration targeting.

Firstly, the essential information layers from different available data sources were prepared. These layers were selected according to the interested components, VMS and Au, and the effective elements for those kinds of mineralization. The second step devoted to integrating the gathered information layers in a GIS environment. In developing the data models, a geologic guideline prepared by GSI experts, was used to construct the respective target models [1]. Using this guideline, the GSI Geomatics group developed the critical factors associated with the models and the recognition of the criteria from a data analysis perspective used to form the structure and weightings of the model. Using basic model concepts and examining each dataset independently, we developed a simple deposit model that requires 4 keys elements (Host geology, Metal engine – source, structure and ore forming process – alteration/mineralization). Taking into account that there were only five basic datasets (RS, Geology, Magnetics, Geochemistry and mineral occurrences) we have generated a model with over twenty input layers. Figure 1 illustrates the developed model.

Fig. 1 Data Model Generated by the GSI

As said before the needed information layers were extracted using five basic datasets. Unfortunately, most of our available datasets did not have high quality. The most reliable of all was remote sensing data. As it can be seen from the model shown in figure 1, there are several information layers which can be prepared using remote sensing images such as structures, alterations, host geology and sources. Structures, host geology and sources were extracted using visual interpretation comparing to other datasets such as geological maps and geophysics data.

According to the geological guideline prepared by GSI, two sorts of alteration may be decisive to targeting. These kinds of alteration, iron-oxide and clay, can be easily extracted using the remote sensing data with the high degree of reliability.

Alteration extraction used to be done using visual interpretation by making different color compositions such as 5, 3 and 1 as red, green and blue. In the visual interpretation, the high experience of image interpreter was needed to recognise the altered areas in the image. Moreover, the alteration areas which may be critical to exploration are mostly small in size – a few pixels. These small areas were omitted through the visual interpretation, since the interpreter was only able to detect the massive alteration areas in the image. Furthermore, the visual interpretation was done in an adhoc manner and was a subjective method. Therefore, it was found that to accentuate only the visual interpretation resulted in missing the important exploration targets. Hence, it was tried to start an automatic approach to minimize missing the altered areas.

This paper focuses on the introduction of methods used to extract alterations. In the next section, the proposed algorithm in order to highlight the alteration areas is presented and section 3 is devoted to the results as concluding remarks achieved through checking field.

2. The proposed algorithm for alteration extraction
In the visual interpretation of the image, the interpreter extracts the interested information based on the knowledge as well as experience. However, in a digital approach some samples of interested information are given to a digital system. The information extraction is automatically done based on those samples [2]. By integrating both the visual and digital methods to interpret the image, the accuracy of the results will increase.

In this project, two digital methods including band ratio and principal component analysis were used to extract two types of alteration. The available remote sensing data within the datasets are ETM+ with the Geotif format as well as ASTER Level 1B images. The ETM+ image of the LANDSAT 7 satellite contains 7 bands in the visible and infrared regions of the electromagnetic spectrum with the spatial resolution of 30 (m). Another multispectral data in the remote sensing database was ASTER image. It has 9 bands totally in the visible and infrared parts of the electromagnetic spectrum with the spatial resolution of 15 (m) and 30 (m), respectively.

In the preprocessing step, the radiometric corrections regarding atmosphere and sensor failures were done on the images. Geometrically, the images were corrected using the topographical maps with the scale of 1:50000. The precision of the geometric correction was 1 to 2 pixels for ETM+ and sub-pixel for ASTER images. In this point, the preprocessed images are ready to be processed.

2.1. Iron-oxide and Clay alteration extraction
According to the iron-oxide spectral signature, shown in figure 2, a remarkable reflection of electromagnetic wave is occurred in the third band of ETM+ with the wavelength of 700 (nm) as well as a considerable absorption in band 1 with the wavelength of 400 (nm).

Fig. 2 The spectral signature of hematite (iron-oxide) and alunite (clay) illustrated in white and red color, respectively. ETM+ and ASTER spectral bands are shown below the spectral signatures.

The maximum and minimum reflections of the iron-oxide characterize it among the other objects in the image. This characteristic made us use these ultimate points to develop an index. The most appropriate index used to extract the iron-oxide alteration areas is the ratio of band 3 to band 1. This index (band3/band1) can easily highlight the iron-oxide in the ETM+ image. According to the spectral signature of the iron-oxide, the second component of the principal component analysis of bands 3 and 1 may include the most amount of difference between these two bands. This component leads to find the iron-oxide. This method is called selective principal component. Another approach used for iron-oxide extraction is feature oriented principal component method. That is using the forth principal component of principal component analysis of bands 1, 3, 4 and 5. This component can highlight the iron-oxide in the image as well. These three indexes were used together to extract the iron-oxide alteration areas. Since the ASTER data does not contain the blue part of visible spectrum, it is rarely used for iron-oxide extraction.

After preparing these three indexes, r(3/1), PC2(1,3) and PC4(1,3,4,5), a thresholding method was used. As its name suggests, this algorithm uses different threshold values to find the most probable area of iron-oxide in the image. The threshold values were determined according to a couple of known iron-oxide areas in the image as training data. The best values for threshold were selected using the indexes’ histograms and the iron-oxide samples were highlighted while using those thresholds. This process was done by trial-and-error.

By applying a simple Boolean logic using the threshold values, the layers with the most probability of iron-oxide alteration called sub-layers were created. Eq. (1) presents Boolean logic function.

where f ( i, j ) is a pixel in each layer of r(3/1), PC2(1,3) or PC4(1,3,4,5) , T is the threshold value and g ( i, j ) is the corresponding pixel in the sub-layer. Three extracted alteration layers in binary format were then added together in order to form the final alteration layer. Since, there were some objects in the image owning spectral overlaps with the iron-oxide, final alteration layer contained noise. These objects were clouds, snowy area, sandstones and vegetations. A mask included all these objects applied on the final alteration layers to remove the noise. The combination of different methods was used to create the mask such as classification and vegetations, clouds and snow indexes. Then the visual interpretations using the geological maps incorporated to eliminate the remained noise to form the final iron-oxide alteration layer.

The other alteration layer to be prepared was clay alteration. According to the clay spectral signature, shown in figure 2 the clay alteration was extracted in a similar way to iron-oxide. Due to the maximum reflection and absorption points of the clay spectral signature, the tools used for clay alteration extraction are the ratio of bands 5 and 7, r(5/7), the second principal component of the principal component analysis of bands 5 and 7, PC2(5/7), and the forth principal component of the principal component analysis of bands 1, 4, 5 and 7, PC4(1,4,5,7).

Like the method used for iron-oxide, the threshold values were determined with respect to several known clay alteration areas in the image as training data. A threshold function was then applied using the threshold values in order to form the alteration sub-layers. Finally, applying the essential masks in noise removal process, the final clay alteration layer was obtained.

Figure 3 depicts the clay and iron-oxide alteration layers of Khoy-Oshnavieh area. In this figure the alteration extracted can be compared to the color composition of 5, 3 and 1 of ETM+ image.

(a) Clay alteration(b) iron-oxide alteration

Fig. 3 The clay and iron-oxide alteration layers overlaid on the ETM+ image of the Khoy-Oshnavieh area. The red, yellow and brown dots depict the alteration

Another remote sensing data used to extract clay alteration was ASTER image. The most important part of electromagnetic spectrum achieved by ASTER sensor is SWIR (Short Wave Infrared), from 1.60 (mm ) to 2.430 (mm ). The spectral signatures of different kinds of clay alteration are distinctly characterized in this part of spectrum. Since the band width of ASTER image is narrow, it is possible to separate different types of clay alteration.

According to the spectral signature of different sorts of clay alteration such as Kaolinite, Calcite and Sericite, several indexes were applied on ASTER image. These indexes are as follows:
  • Ratio (4/6): All types of clay alteration
  • Ratio (7/6): Pyrophyllite, Alunite
  • Ratio (8/6): Alunite
  • Ratio (9/8): Calcite, Tale, Chlorite, epidote
The extracted alteration layers as crucial layers for mineral exploration modeling were then entered the GIS in addition to other layers. The first step in GIS modeling is to organize appropriate information layers according to the model shown in figure 1 and then build a model with simple mathematics concepts such as Boolean logic. Once the model is built it should represent the collective knowledge as per our understanding of the target potential for the area.

Actually, in developing the data target model for the respective deposit opportunities, simple index overlay methods were used to better understand the relationship of the geologic model understanding and the composite signature of these targets within multiple datasets. when the geologic team is more comfortable with the simple geologic model, more complex data tools can be used (weights of evidence, fuzzy logic) but these must also be used carefully in that these methods assumes significance to the data that may also not be relevant. The results of modeling using simple and complicated models for VMS and Au are shown in figure 4.

Fig. 4 (a) The Au targets using the index overlay method. Proposed targets shown in blue. (b) The Au Fuzzy logic targets. (c) The VMS targets identified using the index overlay method. (d) The VMS targets identified using the Fuzzy logic method.

In total 37 Au targets and 16 VMS targets were selected for field follow-up. The optimization of any data model (data or knowledge driven) requires a methodology to assess the reasons (data sources, incorrect logic/tools) for false positives and optimization protocol to minimize these effects. For this reason, the second phase of the program was designed to implement a three week to optimize targeting methodologies by the assessment of targets attributes.

After checking the targets, a total of 11 targets out of 55 were recommended for more study. The low success rate was related to low quality of different data sources in dataset. Testing the defined targets, alteration mapping proved to be the most reliable information layer among the others. Although the remote sensing data especially ETM+ images did not have high quality, containing atmospheric errors, the results showed that they were the most precise data source in GSI database. Using ASTER images lead to extract different clay alteration types which were complementary to ETM+ data. Moreover, the proposed algorithm for alteration extraction performed better than visual interpretation due to the results obtained by checking the targets. It minimized the missing alteration areas in targeting process. Since the proposed approach is automatic, it does not need so much knowledge to interpret the image while its efficiency is increased. The superiority of the proposed method has made the remote sensing group use it in further mineral exploration projects.

4. Concluding remarks
A regional scale GIS contains an enormous amount of data from a wide variety of sources including geologic, geochemical, geophysical, and other exploration data sets. The fundamental goal of such a system is to identify areas of potential mineralization through spatial analysis and other modeling functions. A variety of models were tested to get the best potential targets. The original Geomatics modeling program identified 55 exploration targets (high favorability areas) with respect to VMS and Au potential. Of these targets, after field examination a total of 11 targets have been recommended for further exploration. In addition, a 12th target not identified by the analysis was revealed during the field work. The relatively low success rate was not unexpected given the sparse and/or incomplete data (showings, mines, no Au assays, etc.) and the limited regions of prospective geology.

Remote sensing alteration mapping proved to be the most reliable source of information though it also was limited due to the quality of the image (data quality, water effects, clouds/mist, vegetation, snow). Since the band width of ASTER image is narrower compared to ETM+ data, it's possible to separate different types of clay alteration. Using ASTER image, the amount of noise produced in the extracted alteration layer was also declined. Furthermore, the superiority of the proposed algorithm for alteration extraction was proved after testing the targets. Its efficiency would be substantially increased using high-quality remote sensing data. Further improvement to alteration extraction process may be achieved by integrating the automatic and visual approaches.

More precise results are obtained using hyper-spectral images applying the spectral methods such as SAM (Spectral Angle Mapper). These methods try to find interested objects through the image based on their spectral signatures directly. The spectral analysis methods are used for detail exploration in small areas since they are computationally expensive. The Geological Survey of Iran is planning to invest in providing hyper-spectral images of a specified area in the east of Iran. The regional exploration targeting has been already done in south of khorasan zone. During the study, some targets with the high potential of mineralization have been recommended for further exploration by hyper-spectral data.

In the preparation of this paper without the Geomatics members’ support the results would not be possible. We express our deep appreciation to remote sensing, geophysics, Geochemistry group for their excellent and extensive accomplishment in preparing different information layers. Special thanks to GIS group for their good job in GIS modeling. To all these and many others who helped us, we express our sincere appreciation.

  • J. Gingerich, 2004, "The first phase of the GSI Khoy-Oshnavieh report: Data modeling".
  • J. A. Richards and X. Jia, 1999, "Remote sensing digital image analysis", Springer-Verlag Berlin Heidelberg.
  • M. Abrams and S. Hook, "ASTER user handbook", Jet Propulsion Laboratory/ California Institute of Technology.
  • M. Brown,1 T. D. Gedeon,2 D. I. Groves1 and R. G. Barnes Australian Journal of Earth Sciences Volume 47 Issue 4 Page 757 - August 2000
  • Agterberg, F.P. and Bonham-Carter, G.F., 1990, "Deriving weights of evidence from geoscience contour maps for the prediction of discrete events", Proceedings 22nd APCOM Symposium, Berlin, Germany, v.2, p. 381-395.
  • Agterberg, F.P., Bonham-Carter, G.F, and Wright, D.F., 1990, "Statistical pattern integration for mineral exploration: In Computer applications in Resource Estimation: Predictions and Assessment for Metals and Petroleum", Eds. G. Gaal and D.F. Merriam, Pergamon, Oxford, p. 1-21.
  • Bonham-Carter, G.F., 1994, "Geographic Information Systems for Geoscientists: Modeling with GIS: Pergamon", Oxford, 398 p.
  • Bonham-Carter, G.F., Agterberg, F.P. and Wright, D.F., 1988, "Integration of geological datasets for gold exploration in Nova Scotia", Photogrammetric Engineering and Remote Sensing, v. 54(11), p. 1585-1592.
  • Bonham-Carter, G.F., Agterberg, F.P. and Wright, D.F., 1989, "Weights of evidence modeling: a new approach to mapping mineral potential", In Statistical Applications in the Earth Sciences, Eds. F.P. Agterberg and G.F. Bonham-Carter, Geological Survey of Canada, Paper 89-9, p. 171-183.
  • Elliot, J.E., Wallace, C.A., Lee, G.K., Antweiler, J.C., Lidke, D.J., Rowan, L.C., Hanna, W.F., Trautwein, C.M., Dwyer, J.L. and Moll, S.H., 1992, " Mineral resource assessment map for vein and replacement deposits of gold, silver, copper, lead, zinc, manganese and tungsten in the Butte 1 degree x 2 degree quadrangle, Montana", U.S.Geological Survey, Miscellaneous Investigations Series, Map I-2050-D, 31p.
  • McCammon, R.B., 1989, "Prospector II - the redesign of Prospector", AI Systems in Government, March 27-31, 1989, Washington, D.C., p. 88-92.
  • Wright, D.F., 1996,"Evaluating volcanic hosted massive sulphide favourability using GIS-based integration models, Snow Lake area, Manitoba", Ph.D. Dissertation, University of Ottawa, 344 p.
  • Wright, D.F. and Bonham-Carter, G.F., 1996, "VHMS favorability mapping with GIS-based integration models, Chisel Lake-Anderson Lake area", In EXTECH I, A Multidisciplinary Approach to Massive Sulphide Research in the Rusty Lake-Snow Lake Greenstone Belts, Manitoba, Eds. G.F. Bonham-Carter, A.G. Galley and G.E.M. Hall, Geological Survey of Canada, Bulletin 426, p. 339-376, 387-401.
Maryam Dehghani got her M.Sc. degree in remote sensing from K. N. Toosi University of Technology. She got her B.Sc. degree in Geodesy and Geomatics Engineering from Science and Technology University. The areas of her interests are: Digital Image Processing, Remote Sensing application in geology, and radar images processing.

© All rights reserved.