TECHNICAL PAPERS

SATELLITE, AIRBORNE AND GROUND SPECTRAL DATA APPLIED TO MINERAL EXPLORATION IN PERU

Presented at the Twelfth International Conference and Workshops on Applied Geologic Remote Sensing, Denver, Colorado, 17-19 November 1997.

ABSTRACT

The Peruvian Andes present significant logistical difficulties to mineral exploration. Whereas Landsat TM data has helped focus exploration interest, zones of clay-iron spectral anomalism may still cover large areas and range over 3-km in vertical relief. Furthermore Landsat clay-iron signatures may relate to supergene rather than hypogene alteration.

A GER 63-channel hyperpectral survey flown over an area containing many mining and exploration projects aimed to further exploration by mapping alteration minerals, discriminating hydrothermal alteration from supergene and generate new targets. Ground spectra were acquired over well mapped alteration zones in advance to assist calibration and verification of the airborne data.

Difference algorithms on raw data highlighting argillic, propylitic and ferruginous alteration signatures were successfully used to identify target areas, for which the data were separated out and calibrated. Calibrated airborne spectra were verified against ground data.

Linear spectral unmixing readily separated quartz-sericite, quartz-alunite, and kaolinitic hydrothermal alteration assemblages from montmorillonite or smectite clays in supergene zones. Iron oxide minerals along with jarosite were also mapped. Mineral claims were lodged on better projects as defined by geology, structure and hydrothermal alteration.

The exercise successfully focussed mineral exploration onto discrete targets which met specific geological, structural and mineralogical criteria. Landsat TM clay-iron spectral anomalies were further mapped into alteration mineral assemblages subsequently confirmed by detailed field mapping.

1.0 INTRODUCTION

The Peruvian Andes present enormous logistical difficulties to mineral exploration. Generally however, they are relatively free of vegetation and lend themselves well to geological mapping and mineral exploration by remote sensing techniques. Landsat Thematic Mapper (TM) data has been widely used to generate exploration targets using the wavelengths characterised by clay and iron absorption. Whereas Landsat TM data used in this way undoubtedly focusses exploration interest on a regional scale, zones of clay-iron spectral anomalism may still cover large areas, range over 3-km in vertical relief and still present major access problems for follow-up. Furthermore, some strong clay-iron signatures may relate to supergene rather than hypogene alteration.

Thus, additional criteria are required to target alteration systems, rank them, separate the hydrothermal from the supergene, map important alteration assemblages and localise the most prospective zones. Such information may be gathered from existing geological data or through ongoing fieldwork. However, due to the nature of the terrain, both may prove unreliable and expensive. Alternatively, remote sensing data of higher spatial and spectral resolution capable of discriminating specific alteration mineralogy and small, discrete alteration zones could be applied.

Cia. De Minas Buenaventura S.A. own and operate a number of mines in the Greater Huancavelica area of south central Peru. The region is high, rugged and difficult to access, comprising two geologically distinct domains. The first of these is situated in the north east of the area where Devonian-Permian basement rocks occur in the cores of domes flanked by the Triassic-Jurassic Pucará Group, with Cretaceous-Tertiary sedimentary formations occurring in synclinal structures. In the south west, Tertiary volcanic and volcano-sedimentary sequences dominate and extend unconformably over the Mesozoic and Palaeozoic rocks. They are intruded by pene-contemporaneous andesites and dacites. Most of the known mineralisation in the area is hosted by the Tertiary volcanic rocks and associated with extensive hydrothermal alteration although significant lead-zinc-silver mineralisation is also hosted by the Pucará Group.

The Geophysical and Environmental Research Corporation (GER) were contracted by Cia. De Minas Buenaventura to carry out a survey using their 63-channel Digital Airborne Imaging Spectrometer (DAIS). The DAIS data was acquired in October 1996 over a 5,000 sq-km area centred just south of Huancavelica (figure 1) and comprises 25 bands in the visible and near infra-red (VNIR), 32 in the short wave infra-red (SWIR) and 6 in the thermal infra-red (TIR) portions of the electromagnetic spectrum (Table 1). Because of the mountainous terrane, the ground resolution varies from approximately 10m over higher ground to 14m over deeper valleys. Fourteen flight lines were recorded, each orientated 148^oN and approximately 70-km long and 6-km wide.

Figure 1: Landsat TM image map 7, 4, 1 RGB showing the survey area in white and principal locations.

Table 1. GER DAIS-63 band positions.

2.0 PROCESSING & ANALYSIS OF THE DATA

2.1 PREPROCESSING AND GEO-REFERENCING

In its raw form, the data contain erroneous electronic information and noise. The recorded digital value (DV) can be described mathematically as follows:-

DV = R + I + E + A

Although the raw data can be treated using band difference or ratio algorithms to identify general areas of interest, before the data can be applied to detailed spectral analysis and identification of specific minerals, it must be reduced to reflectance. "I" remains constant throughout a single file, imparting a basic minimal DV for each band in each pixel which was removed by "dark subtraction". Variable noise (E) was analysed using a minimum noise fraction transform (MNF) and reduced by the inverse MNF procedure. Atmospheric backscatter was removed and the data calibrated using the Internal Average Relative Reflectance (IARR).

2.2 DATA QUALITY ASSESSMENT

For the assessment of data quality, ground spectra were collected in advance of the survey over well mapped hydrothermally altered ground at Julcani, Recuperada, Millupata and Huamanrazo (figure 1), using a GER MkV spectro-radiometer. At each site, spectra were collected from both fresh and weathered rock surfaces so as to confirm spectrally the mapped alteration mineralogy and provide a sample spectrum of surfaces that would be recorded in the airborne data. Analysis of the ground spectra prior to the survey generally confirmed the alteration mineralogy mapped by Buenaventura geologists and the spectra of the weathered surfaces were found to be representative of the fresh minerals beneath although the signal strength of the spectra was subdued.

The airborne data were assessed for quality by comparing airborne with ground spectra over easily recognisable, mineralogically homogenous areas. Direct comparison of ground and airborne spectra showed an excellent correlation (figure 2), although the average airborne response appears noisy. The advanced argillic airborne alteration curve fits very closely that of the weathered rock surface as recorded by the GER MkV (figure 2a and c). Similarly, the propylitic ground spectrum is closely followed by the noisier airborne curve (figure 2b and d).

Figure 2. a) Ground Spectra Collected From Huamanrazo Advanced Argillic Alteration Compared To b) Airborne Spectra From The Same Location.

On the basis that the airborne spectra closely approximated previously collected ground data, a preliminary mineral mapping exercise was carried out with the aim of testing the quality of the data as applied to mineral mapping and pixel un-mixing. A "spectral feature fit" routine (Clarke et al., 1987, 1991) was carried out for a suite of minerals typically found in hydrothermal alteration systems. A density slice was applied to highlight those pixels with the closest match between the airborne and reference spectra which were then compared (figure 3). In all cases, a very close match was achieved for the mineral spectra, all of which showed a distribution consistent with the alteration zones as mapped by Buenaventura geologists (figure 4).

Figure 3. A Comparison GER DAIS-63 Airborne Spectra with:- a) to d), Selected Reference Spectra For The Huamanrazo Prospect and, e) to h), PIMA Ground Spectra For All Prospects.
Figure 4: a)Plot of >62.5% spectral fit for alunite (red), propylitic minerals (sea green), phyllic minerals (aquamarine), kaolinite (yellow), jarosite (magenta), opal (blue) and pyrophyllite (coral) compared to b) mineral alteration mapping showing advanced argillic (red), argillic (yellow) and propylitic (green).

2.3 REGIONAL EXPLORATION

To reduce the volume of data, specific algorithms designed to discriminate key alteration styles, lithologies and structures were created (table 2) and mosaic images produced from which alteration zones were identified and set within a geological framework.

Table 2. Standard alteration algorithms used to produce mosaic images for the survey area.

The regional image mosaics were analysed for indications of hydrothermal alteration and exploration targets were generated and ranked on the basis of presence or absence of argillic, propylitic, ferruginous and siliceous alteration signatures, suitable geology and structural settings. Targets generated included all of the known mining camps and prospects within the area plus new targets both small and discrete as well extensive complexes comprising a number of alteration features. Those considered most prospective for epithermal gold mineralisation were sub-sampled from the data and analysed further through advanced spectral processing techniques described below. Some targets, Arcopunco for example (figure 5), were situated over unclaimed ground and title was secured before the advanced processing stage. Others currently under claim were placed on hold. A number of targets comprising only small or narrow linear vein-like Fe-oxide, propylitic and silica alteration signatures within Pucará limestone may be prospective for Mississippi Valley Zinc type mineralisation and will be followed up in the future.

Figure 5. Arcopunco Prospect Alteration a) Argillic, b) Propylitic, c) Fe-oxides, and d) Silica.

2.4 ADVANCED PROCESSING AND MINERAL MAPPING

Top priority exploration targets such as Arcopunco and pre-existing operational and exploration project areas were further processed to provide mineral distribution maps as an aid to ongoing exploration. The approach used here was to select a number of important minerals within the constraints of the epithermal gold exploration model and compare the apparent "fit" of the individual pixel spectra in the data with reference spectra of the selected minerals (Clarke et al., 1987, 1991). A scale for the absorption feature depth, which acts as a measure of abundance for each material, and a root mean square (rms) error value are calculated. The greater the error, the higher the rms value and hence the "fit" for each mineral is calculated as being the ratio of scale and rms values. "Spectral fit" images were generated for all selected reference minerals and displayed as grey scale images with a density slice applied to highlight pixels with the best fit. Image spectra for the better fitting pixels were then visually compared with reference spectra and the quality of the fit and validity of the mineral as a true "end member" determined. In this way an "end member" suite for a specific target or project data set was created and a linear spectral un-mix (Boardman, 1989, Boardman & Kruse, 1994) carried out for those "end member" minerals. This process assumes the reflectance at each pixel for each wavelength is a linear combination of the reflectance of each end member present in direct proportion to its abundance within that pixel and provides a measure of the relative abundances of end-member materials shown in the imagery.

The linear un-mixing routine also produces scale (abundance), and rms error values which can be displayed alone (figure 6b) or as a ratio (figure 6c). If the linear un-mix is reliable, the scale un-mix image (figure 6b), should reflect that of the scale/rms error ratio image (figure 6c), and, because the best fitting image spectra will be only the relatively pure pixels, both should show a similar but wider distribution than the spectral fit image (figure 6a). The potential for error is much greater in the un-mix process and both the linear un-mix and un-mix/rms error ratio results are relatively noisy (figure 6b & c). Re-introducing the spectral fit parameter at this stage by displaying the product of the spectral fit and un-mix/rms error ratio increases the accuracy and effectively cleans the distribution image (figure 6d). This process was applied for all "end members" and distribution maps produced both for each individually and collectively as one.

Figure 6: Density slices shown in red, yellow, green and cyan in descending order for:-     a) the spectral fit for alunite at Huamanrazo, image spectra for alunite used in figure 6a were located in the red pixels from this density slice.     b) linear un-mix scale result for alunite at Huamanrazo     c) linear un-mix scale/rms error ratio for alunite at Huamanrazo     d) the product of spectral fit and linear un-mix scale/rms error ratio for alunite at Huamanrazo.     in each case, the distribution of the top three zones compares very well with advanced argillic and argillic alteration as mapped in figure 7b.

2.5 FIELD VERIFICATION

On completion of the advanced processing and mineral mapping, a PIMA field spectro-radiometer was used to acquire ground spectra of fresh and weathered rock surfaces and soils covering areas in which specific minerals were indicated from the airborne data. These field spectra were then compared directly with those of the DAIS for the immediate vicinity of the sample site (figure 3). For all sites, the field spectra confirmed the mineralogy indicated by the mineral mapping routine, justified the processing methodology and verified the accuracy of the resultant mineral distribution maps.

3.0 DISCUSSION AND CONCLUSIONS

The work set out to further Buenaventura’s exploration effort in the Huancavelica area by building upon previous regional and prospect scale work and by identifying new areas of interest. Previous work based upon Landsat TM imagery, regional and prospect mapping and geochemistry was limited by the difficulty of access in such mountainous terrain. Furthermore, although Landsat TM imagery was useful in identifying significant clay-iron spectral anomalies, there was no distinction between supergene, hypogene or other distinct mineral alteration assemblages and targeted zones remained extensive. The GER survey aimed to provide a high quality regional spectral database at spatial and spectral resolutions which would allow these distinctions to be made.

The regional algorithm process for rapid first pass assessment of the data and identification of prospective ground was very successful and is a useful first stage whenever large data volumes are involved. The algorithms selected worked very well and discriminated different alteration styles at all of the existing prospects. Not only were alteration styles characteristic of epithermal styles of mineralisation recognised but some signatures were detected which may relate to Mississippi Valley Zinc style mineralisation.

The advanced spectral processing replicated the mapped distribution of important hydrothermal alteration minerals. Both field mapping and post - survey field spectral analysis confirmed the reliability of both data and processing techniques. The approach used proved successful in that the regional work outlined new exploration targets and generated new projects. Furthermore, detailed spectral processing on existing prospects and operational areas not only confirmed existing mapping but also provided additional exploration leads and established a high level of confidence in the technology as a tool for prospect mapping and exploration.

The programme achieved its aims by providing a significant improvement in exploration targeting and mineral mapping over what had earlier been achieved with Landsat TM data. It has significantly added to the number of exploration targets in the Huancavelica area each of which can now be evaluated and ranked through remote geological and alteration mapping, thereby focussing exploration on the better targets and minimising the cost of follow-up.

4.0 REFERENCES

BOARDMAN, J.W. 1989, "Inversion of imaging spectrometry data using singular value decomposition": In Proceedings, IGARSS’89, Twelfth Thematic Canadian Symposium on Remote Sensing, v4, p. 2069-2062

BOARDMAN, J.W. & KRUSE, F.A. 1994, "Automated spectral analysis; a geological example using AVIRIS data, north Grapevine Mountains, Nevada": In Proceedings, ERIM Tenth Thematic Conference on Geologic Remote Sensing, Environmental Research Institute of Michigan, Ann Arbor, p. I-407 - I-418.

CLARK, R.N., KING, T.V., & GORELICK, N., 1987, "Automatic continuum analysis of reflectance spectra": In Proceedings, 3^rd Airborne Imaging Spectrometer (AIS) workshop, JPL Publication 87-30, p. 138-142.

CLARK, R.N., SWAYZE, G.E., GALLAGHER, A., GORELICK, N., & KRUSE, F.A., 1991, "Mapping with imaging spectrometer data using the complete band shape least squares algorithm simultaneously fit to multiple spectral features from multiple materials": In Proceedings, 3^rd Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) workshop, JPL Publication 91-28, p. 2-3.

5.0 ILLUSTRATIONS

Figure 1: Landsat TM image map 7, 4, 1 RGB showing the survey area in white and principal locations.

Figure 2. a) Ground Spectra Collected From Huamanrazo Advanced Argillic Alteration Compared To b) Airborne Spectra From The Same Location.

Figure 3. A Comparison GER DAIS-63 Airborne Spectra with:- a) to d), Selected Reference Spectra For The Huamanrazo Prospect and, e) to h), PIMA Ground Spectra For All Prospects.

Figure 4: a)Plot of >62.5% spectral fit for alunite (red), propylitic minerals (sea green), phyllic minerals (aquamarine), kaolinite (yellow), jarosite (magenta), opal (blue) and pyrophyllite (coral) compared to b) mineral alteration mapping showing advanced argillic (red), argillic (yellow) and propylitic (green).

Figure 5. Arcopunco Prospect Alteration a) Argillic, b) Propylitic, c) Fe-oxides, And d) Silica.

Figure 6: Density slices shown in red, yellow, green and cyan in descending order for:-
    a) the spectral fit for alunite at Huamanrazo, image spectra for alunite used in figure 6a were located in the red pixels from this density slice.
    b) linear un-mix scale result for alunite at Huamanrazo
    c) linear un-mix scale/rms error ratio for alunite at Huamanrazo
    d) the product of spectral fit and linear un-mix scale/rms error ratio for alunite at Huamanrazo.
    in each case, the distribution of the top three zones compares very well with advanced argillic and argillic alteration as mapped in figure 7b.