Presented at the Thirteenth International Conference and Workshops on Applied Geologic Remote Sensing, Vancouver, British Columbia, Canada, 1-3 March 1999.

ABSTRACT

Mineral exploration companies are being asked to embrace new airborne and satellite borne hyperspectral data, often without having fully appraised existing remote sensing data.  Landsat TM data have been around for many years and are very cheap in comparison to the cost of acquiring new data.  Aslo, airborne data exist which may also be bought cheaply "off the shelf."  Much of this archival data existed prior to recent advances in image processing and spectral analysis and, when looked at in a new light, may still have much to offer.

Landsat TM, GERIS-64 channel and Geoscan MKII AMSS over the same area in Northern Chile, were processed with the aim of identifying mineral end-members important to exploration.  Whereas all data sets had previously been processed and interpreted using RGB colour algorithms as analogue mineral or alteration indicators, this work set out to identify important end members directly from the data.  Mineral alteration assemblage maps were produced and subsequently verified in the field using detailed spectral analysis of fresh and weathered rock surfaces and soils.

The work demonstrates that archival data can focus exploration attention to the extent that meaningful alteration maps may assist in the localisation of exploration targets and first pass mapping in support of geochemical exploration.  The new generation of higher spectral resolution instruments can later be focussed at higher spatial resolution over promising exploration targets where greater mineral discrimination can further advance exploration.

1.0 INTRODUCTION

A new generation of airborne hyperspectral scanners are now operating commercially around the world.  The Short wavelength infrared Full Spectrum Imager (SFSI), and multiple versions of the Hymap instrument are operating anywhere from Mexico to Mongolia.  These instruments offer the mineral exploration user the ability to discriminate and map not only individual minerals but also variations within certain minerals.  In the near future, the ARIES project will offer the exploration industry satellite borne hyperspectral data.

Mineral exploration users are being asked to embrace these new instruments, often without having made full use of existing remote sensing data.  Landsat TM data has been around for many years and is very cheap compared to the cost of flying new hyperspectral surveys.  Furthermore, airborne data exist which can be bought cheaply "off the shelf" in a "ready processed" value added format.  Much of these data existed prior to recent advances in image processing and spectral analysis techniques and, when viewed in a new light, may still have an important role to play.

Geoscan AMSS MKII data were widely acquired in late 1980’s and early 1990’s and much of these data are now available on the open market.  Similarly there exist other airborne data such as those acquired by GER on behalf of specific clients, not all of which were processed by any other means than simple algorithms designed to enhance broad mineral groups.  The aim of this work was to study archival data sets of both GER DAIS-63 and Geoscan AMSS MKII over the same ground in Northern Chile with the aim of evaluating their amenity to the type of advanced processing which has led to the current enthusiasm for hyperspectral imagery and in particular the new generation of instruments.   If it could be shown that existing data of only moderate spectral resolution, could still provide meaningful mineral maps, then they may provide a lower cost alternative to regional exploration than the newer and more expensive hyperspectral scanners.

2.0 THE STUDY AREA

The area chosen for the study was the area immediately south of Cerro Cenizas, located in Antofagasta Province, First Region, Northern Chile, 55km south east of Sierra Gordo, a small village some 60km SW of Calama on the main Calama - Antofagasta road (figure 1).  The area had previously been explored by Empresa Minera Mantos Blancos (EMMB) and remains under claim.  The area is underlain by the Upper Triassic - Lower Jurassic Agua Dulce Formation and Cerro Negro Beds (TrJad and TrJcn respectively in figure 1) which both comprise andesitic lavas, breccias and tuffs with intercalated sandstones and rhyolites.  The both are intruded by Cretaceous granites (Kg) and Tertiary porphyry stocks (Tg).

Good quality ground mapping and mineral alteration mapping was available for the study as well as two airborne data sets.  The GERIS-63 data was flown along a single flight line on March 17th 1991.  The pixel size is approximately 8m, the swath width is slightly more than 3km, and the run length 13.5km in a south-south westerly direction (figure 1).  The Geoscan AMSS MKII data was flown just over 1 year later on May 20th 1992 with a ground resolution of 8m and a swath width of 8km.  The GERIS data are proprietary to EMMB and the Geoscan data are commercially available.

Figure 1: Location map showing the study area, GERIS 63 and Geoscan flight lines and local geology.

3.0 THE LANDSAT TM DATA

Landsat 5 TM data for path 233 row 076 were calibrated in ENVI using a dark subtraction routine to remove atmospheric backscatter effects and published pre-launch gain and offset information to convert the digital values in the file to radiance.  Data sub-sets were created according to standard 1:100,000 scale quadrangle map series and processed to produce clay-iron indices.  The Crosta technique (Loughlin, 1991) was used to generate hydroxyl (H), Red Iron Oxide (Fr) and Yellow Iron Oxide (Fy) indices which were then density sliced to identify those pixels with the maximum likelihood of comprising iron and clay rich mineralogy. The top 5% were thresholded into a region of interest.

Other than the calibration of the data, the processing to this stage involved nothing new.  However, with calibrated data, it is possible to compare the anomalous pixels with spectra of real minerals and determine the style of mineral alteration identified.  Pixels from the thresholded hydroxyl region of interest were displayed in the ENVI multi-dimensional visualiser and distinct end members identified which were then analysed using the ENVI spectral analyst.  The regional distribution of these end members were mapped using a matched filter routine and anomalous clay zones characterised according to associated end member assemblages (figure 2).

Figure 2: a) Greyscale image showing Matched Filter result for phyllic alteration end member assemblage comprising kaolinite/smectite, muscovite and hematite and b) mineral spectra convolved to Landsat TM for the pixel indicated, the phyllic alteration end member and a kaolinite/smectite mix.

Spectral feature fitting and mixture tuned matched filter routines were also applied to a sub-set of the data covering specifically the area of interest covered by the two airborne data sets.  These routines used spectra of key hydrothermal alteration minerals taken from a spectral library.  The pixels with the best fit for the respective minerals were then colour coded and alteration assemblages determined (figure 3 ).   Individual alteration systems could then be characterised according to their likely mineralogy.

Figure 3: Detail of the northern alteration zone showing colour coded assemblages as mapped using the TM data (yellow - kaolinite dominated, cyan - muscovite dominated).

4.0 THE AIRBORNE DATA

4.1 GERIS-63

The GERIS-63 data comprises 63 channel data with the band positions shown in table 1. Bands 1 to 31 inclusive comprise the Visible/Near Infrared bands (VNIR) and 32 to 63 the Short Wave Infrared (SWIR).  Each single scan line comprises 544 pixels of which 512 would normally contain image data and the remainder the dark current information.  However, due to the camera aperture on the aircraft being too small for the total field of view of the instrument, in these data, only pixels 65 to 457 record actual image data.

Table 1.  GER DAIS-63 band positions. 

All bands in the dark current pixels showed a considerable instrument component in the recorded signal which was removed by dark subtraction before the residual noise levels in the data were examined.  For the VNIR, typical signal:noise levels were 20:1 and for the SWIR, 10:1.  However, noise levels were particularly high in bands 24, 25, 26, 30 and 47 which were subsequently eliminated from all spectral analyses.   Minor variations in the signal gain or offset of the instrument along the track and significant cross-track illumination variation was noted and corrected. Mis-registration of bands blurred the outlines of terrane features and effectively reduced the working resolution.

The GERIS data were calibrated using the Internal Average Relative Reflectance (IARR) Calibration routine in ENVI and a spectral feature fitting routine applied to search for specific minerals of interest.  Mineral indices were density sliced (figure 4), the top 5% thresholded and mineral assemblage maps generated.

Figure 4: Spectral fit result on GERIS 63 data for kaolinite (a) and illite (c) with best fitting pixels and highest concentrations in red, and their respective image spectra compared with USGS standard reference spectra (b) and (d) respectively.

4.2 GEOSCAN AMSS MKII DATA

The Geoscan AMSS MkII data comprise 24 channels at the band positions shown in table 2. Bands 1 to 10 inclusive comprise the VNIR bands, 11 to 18 the SWIR and 19 to 24 the Thermal Infra Red (TIR).  Each scan line comprises 752 pixels in the raw data, resampled to 1024 in the tan q correction procedure.  After correction for atmospheric backscatter, the data were calibrated using the Internal Average Relative Reflectance (IARR) Calibration routine.   Examination of the Geoscan data for noise and other artifacts identified no problems with along-track variation, instrument noise, or bad bands.

The Geoscan data were calibrated using the Internal Average Relative Reflectance (IARR) Calibration routine in ENVI and then a spectral feature fitting routine applied to search for specific minerals of interest.  Again individual mineral indices were then density sliced, the top 5% thresholded and mineral assemblage maps created.  Using the TIR bands in the Geoscan data, an index for silica was developed and applied over the area of interest.  This index was also density sliced and thresholded and in the same way as before.

Table 2. Geoscan MK2 band positions and band widths.

4.3 CONSOLIDATED INTERPRETATION OF DATA

The interpretive results of the three data sets was found to be generally consistent although varying in detail.  The broad alteration assemblages interpreted from the Landsat TM data were confirmed by the airborne data sets which were also able to break the assemblages down to specific minerals.  Where the TM data appeared to show a phyllic alteration system comprising mainly kaolinite/smectite with muscovite and hematite, (figures 2 and 3), the GER and Geoscan data confirmed kaolinite and/or dickite and the presence of extensive sericitic alteration within and around the argillic minerals (figure 5).  The airborne data also indicated a small zone of alunite and pyrophyllite to the northern end of the argillic zone (figure 5) and the Geoscan data also mapped a number of strong silicic zones away from the main area of interest.  Both instruments mapped dickite and kaolinite almost identically but interestingly, when the best fitting spectra were reviewed, in each case the fit appeared to be better for dickite.

Figure 5: Mineral alteration index image showing argillic minerals (red), phyllic (green) and propylitic (blue).  Also shown are the field sample sites listed in table 4, colour coded according to mineralogy determined by PIMA spectral analysis, red = alunite, pink = pyrophyllite, blue = dickite/kaolinite, emerald green = muscovite/illite and sea green = prochlorite or albite.

5.0 FIELD EVALUATION

The area of study was visited in the field and spectra collected of both fresh rock, weathered rock surfaces and soils using a PIMA spectro-radiometer.  Work focussed specifically upon the different zones of alteration as identified by the various data sets and compared the interpreted mineralogy with that mapped previously by EMMB geologists and the mineralogy determined by analysis of the ground spectra.  Interestingly, the original field map for the alteration system studied, mapped extensive sericitisation but no dickite, kaolinite, pyrophyllite or alunite.  However, the mineralogy determined by use of the PIMA not only confirmed the alteration as interpreted from the airborne imagery, but also confirmed the presence of dickite, pyrophyllite and alunite not seen in the exploration field mapping of the project (figure 5, table 3).  Analysis of spectra using the Spectral Geologist software identified no change in composition between the muscovite and illite, both of which were iron - magnesium rich and phengitic.  Soils were dominated by illite with some kaolin explaining the dominantly kao-smectitic signature in the TM data.

Table 3.

6.0 DISCUSSION & CONCLUSIONS

Essentially, the field spectral analysis identified the same alteration mineralogy in rocks and soils as was indicated by both the satellite and airborne data.  Where the TM spectra indicated a combination of kaolinite, smectite and muscovite, the airborne instruments identified dickite, pyrophyllite and alunite in addition to the phyllic assemblage indicated in the satellite data.

By today’s standards, the remote sensing data used is considered to be coarse in terms of spectral resolution.  Only the GERIS data might be considered to be hyperspectral but it has significant noise and mis-registration problems.   Nevertheless, by applying advanced spectral mapping techniques to these data it was possible to interpret much more than has hitherto been considered possible.  Even the Landsat TM data allowed some generalisations to be made about the alteration system being studied which were confirmed not just by the airborne data but in subsequent fieldwork.   The Geoscan and GERIS data sets were able to be more definite in the mineral assemblages mapped and these also were confirmed in the field.

These data, used appropriately therefore, can help identify, characterise and map alteration mineral assemblages and provide significant information useful in focussing exploration programmes.  The Landsat TM and Geoscan data are available "off the shelf" at prices considerable lower than the cost of acquiring new, airborne hyperspectral data and offer extensive coverage.  Although the GERIS data is not so available, there are a number of data sets held by early users of the technology that may not have been processed in the manner described here and may thus represent a forgotten asset for the company concerned.

Where Landsat TM data over a regional exploration area can recognise and determine the possible nature of any clay enrichment or alteration,   the GERIS and Geoscan data can rapidly confirm alteration styles and focus attention on prospective zones.  For exploration follow-up on targets so generated, the GERIS or Geoscan data can be further analysed and detailed mineral maps produced for the target.  At this stage, acquisition of new hyperspectral data using an instrument of much higher spectral resolution, capable of mapping sub-species of minerals such as high and low temperature species of alunite, kaolinite, dickite, quartz etc. can be cost effectively applied over discrete target zones of known potential.

7.0 ACKNOWLEDGEMENTS

The authors gratefully acknowledge Empresa Minera Mantos Blancos S.A. for permission to publish this work.

8.0 REFERENCES

W.P. Loughlin, "Principal Component Analysis for Alteration Mapping," Proceedings 8th Thematic Conference on Geologic Remote Sensing, Denver, Co., U.S.A., V1, pp 293 - 306, April 29th - May 2nd 1991

9.0 FIGURE CAPTIONS

Figure 1: Location map showing the study area, GERIS 63 and Geoscan flight lines and local geology.

Figure 2: a) Greyscale image showing Matched Filter result for phyllic alteration end member assemblage comprising kaolinite/smectite, muscovite and hematite and b) mineral spectra convolved to Landsat TM for the pixel indicated, the phyllic alteration end member and a kaolinite/smectite mix.

Figure 3: Detail of the northern alteration zone showing colour coded assemblages as mapped using the TM data (yellow - kaolinite dominated, cyan - muscovite dominated).

Figure 4 (left): Spectral fit result on GERIS 63 data for kaolinite (a) and illite (c) with best fitting pixels and highest concentrations in red, and their respective image spectra compared with USGS standard reference spectra (b) and (d) respectively.

Figure 5: Mineral alteration index image showing argillic minerals (red), phyllic (green) and propylitic (blue).  Also shown are the field sample sites listed in table 4, colour coded according to mineralogy determined by PIMA spectral analysis, red = alunite, pink = pyrophyllite, blue = dickite/kaolinite, emerald green = muscovite/illite and sea green = prochlorite or albite.