ASTROBIOLOGYVolume 2, Number 3, 2002© Mary Ann Liebert, Inc.

Application of Hyperspectral Infrared Analysis ofHydrothermal Alteration on Earth and Mars

MATILDA THOMAS* and MALCOLM R. WALTER

ABSTRACT

An integrated analysis of both airborne and field short-wave infrared hyperspectral mea-surements was used in conjunction with conventional field mapping techniques to map hy-drothermal alteration in the central portion of the Mount Painter Inlier in the Flinders Ranges,South Australia. The airborne hyperspectral data show the spatial distribution of spectrallydistinct minerals occurring as primary minerals and as weathering and alteration products.Field spectral measurements, taken with a portable infrared mineral analyzer spectrometerand supported by thin-section analyses, were used to verify the mineral maps and enhancethe level of information obtainable from the airborne data. Hydrothermal alteration zoneswere identified and mapped separately from the background weathering signals. A main zoneof alteration, coinciding with the Paralana Fault zone, was recognized, and found to containkaolinite, muscovite, biotite, and K-feldspar. A small spectral variation associated with a ring-like feature around Mount Painter was tentatively determined to be halloysite and interpretedto represent a separate hydrothermal fluid and fluid source, and probably a separate system.The older parts of the alteration system are tentatively dated as Permo-Carboniferous. Theremote sensing of alteration at Mount Painter confirms that hyperspectral imaging techniquescan produce accurate mineralogical maps with significant details that can be used to identifyand map hydrothermal activity. Application of hyperspectral surveys such as that conductedat Mount Painter would be likely to provide similar detail about putative hydrothermal de-posits on Mars. Key Words: Hyperspectral infrared analysis—Hydrothermal deposits—Min-eralogical mapping—Mount Painter, Australia. Astrobiology 2, 335–351.

335

INTRODUCTION

ANCIENT HYDROTHERMAL ENVIRONMENTS areprime targets in the search for earliest life on

Earth and former life on Mars (e.g., Bock andGoode, 1996). Hyperspectral infrared analysistechniques provide a useful tool in the explora-tion and mapping of such environments on bothplanets. Mount Painter in the northern Flinders

Ranges of South Australia is one of the largestknown terrestrial hydrothermal systems and thusprovides a suitable environment in which to ex-plore the application of these techniques. Weused a dataset from an airborne hyperspectralshort-wave infrared (SWIR) spectrometer to makea mineral map of the Mount Painter Palaeozoichydrothermal system and then ground-truthedthe map using a hand-held spectrometer. We pos-

Australian Centre for Astrobiology, Department of Earth and Planetary Sciences, Macquarie University, NorthRyde, New South Wales, Australia.

*Present address: Geoscience Australia, Canberra, Australian Capital Territory, Australia.

tulate that a similar exploration strategy wouldbe highly effective on Mars, in that case usingsatellite and robotic rover-mounted instruments.

Spectroscopy is the measurement and analysisof portions of the electromagnetic spectrum toidentify spectrally distinct and physically signif-icant features of a material (Fig. 1). Spectral dataare measured using spectral sensors, whichrecord radiation (usually solar) reflected from thesurface of materials. Because many materials ab-sorb radiation at specific wavelengths, it is pos-sible to identify them by characteristic absorptionfeatures, which appear as troughs in a spectralcurve (Kruse, 1994).

Wavelength ranges most suitable for the dis-crimination of geological materials include thevisible and near-infrared (VNIR), SWIR, and themid- or thermal infrared (TIR) (Fig. 2). Spectralvariation is the result of different compositions,the degree of ordering, mixtures, and the grainsize of different rocks and minerals (Table 1)(Huntington, 1996). Owing to their multiple va-lence states, transition elements, such as Fe, Cu,Ni, Cr, Co, Mn, V, Ti, and Sc, display the most

prominent spectral features in the VNIR wave-length range (Kruse, 1994).

The SWIR wavelength region between 2,000and 2,500 nm is particularly suitable for mineralmapping. The 2,000–2,400 nm wavelength regioncan show many absorption features characteris-tic of certain hydroxyl- and carbonate-bearingminerals and mineral groups that are character-istic of hydrothermal alteration. These mineralgroups may include pyrophyllite, kaolinite, dick-ite, micas, chlorites, smectite clays, alunite,jarosite, calcite, dolomite, and ankerite (Linton,1998).

Using spectroscopy, and in particular hyper-spectral imaging technology, it is possible tomake accurate maps of surface mineralogy in-cluding boundaries, relative abundances, andmineral assemblages. Hyperspectral mappingtechniques can identify individual species of ironand clay minerals, which can provide detailed in-formation about hydrothermal mineralizationand alteration zones (Huntington, 1996). Hyper-spectral remote sensing has a distinct advantageover other remote sensing methods due to its abil-

THOMAS AND WALTER336

FIG. 1. Schematic diagram of the imaging spectrometry concept. Images are acquired simultaneously of up to sev-eral hundred narrow spectral bands, providing a complete reflectance spectrum for every pixel in the imaging spec-trometer scene (Courtesy NASA).

ity to differentiate fine spectral variations be-tween different types of rock, making it suited tomapping surface lithology (Wang et al., 1998). Hy-perspectral imaging spectrometers measure andrecord light reflected from the surface in numer-ous contiguous channels and using narrow band-widths (see Fig. 1; Kokaly et al., 1998).

Wang et al. (1998) showed that using hyper-spectral remote sensing techniques, types ofminerals and altered minerals that are closelyrelated to the mineralizing process can be de-termined. Wang et al. (1998) also noted that the“formation of altered minerals can be analysed,that hydrothermal mineralised altered zonescan be delineated, and industrial deposits canbe found directly under favourable conditions”(p. 87) and that “hyperspectral remote-sensingis more effective in identifying earth surface

lithology than any other remote sensing tech-nique” (p. 93).

Improvements in hyperspectral technologieshave not been limited to remote instruments. Ad-vances in technology have also led to the devel-opment of highly accurate, high-resolution fieldspectrometers such as the portable infrared min-eral analyzer (PIMA). By ground-truthing air-and space-gathered data with a high-resolutionfield instrument, such as the PIMA, the spectralresolution and the accuracy of the classificationof the remotely sensed data can be determined.Portable infrared spectrometers are particularlysuited to mapping hydrothermal alteration zones,which are typically characterized by layer sili-cates, including clay minerals. Field portablespectroscopy is a viable and economically feasi-ble tool for accurately determining hydrothermal

HYPERSPECTRAL ANALYSIS OF HYDROTHERMAL ALTERATION 337

FIG. 2. Electromagnetic spectrum. The top diagram locates various bands in a relative sense, the next line is an ex-pansion of the AB portion. The bottom line shows approximate band locations for some of the operational multi-spectral and hyperspectral systems (modified from Rinker, 1994).

TABLE 1. GEOLOGICALLY SIGNIFICANT REGIONS OF THE ELECTROMAGNETIC SPECTRUM (FROM KRUSE, 1994)

Wavelength Wavelength (nm) Associated molecularregion range Mineralogy feature

VNIR 400–1,100 Fe and Mn oxides, rare Crystal field absorptionearths charge transfer absorption

SWIR 1,100–2,5000 Hydroxyls, carbonates, Al(OH)2, Fe(OH)2,sulfates, micas, Mg(OH)2, NH4, SO4amphiboles absorption, CO3

TIR 8,000–14,000 Carbonates, silicates Si-O bond distortion

alteration in both ancient and active environ-ments (Yang et al., 2000).

GEOLOGY OF THE MOUNT PAINTER REGION

The Mount Painter Province is located in thenorthern Flinders Ranges, South Australia, at thenortheastern margin of the Neoproterozoic-Cam-brian Adelaide Rift Complex (Fig. 3). At presentthe ranges comprise an elevated, rugged terrainrising from the plains of Lake Frome in the east,

eventually merging into the plains of the GreatArtesian Basin to the north.

The Mount Babbage and Mount Painter Inliers,together with the Broken Hill and Olary Domains,make up the Curnamona Province. This region isan extensional, probably back arc, continental mar-gin magmatic area of Paleoproterozoic to Meso-proterozoic age. Neoproterozoic-Cambrian clasticsedimentary and carbonate rocks of the AdelaideRift Complex unconformably overlie the older suc-cession (Foster et al., 1994). The Mount Painter In-lier comprises crystalline basement rocks, theMount Painter complex, which includes Palaeo- to

THOMAS AND WALTER338

FIG. 3. Geology of the Mount Painter and Mount Babbage Inliers. The various suites are continuous, extendingfrom high to lower metamorphic grade areas, around folds, across major faults and zones of geophysical disconti-nuity, and through areas that have been subjected to intense retrograde deformation (Drexel and Major, 1990).

Mesoproterozoic metasediments and metavol-canics (the Radium Creek Metamorphics), andMesoproterozoic granites, pegmatites, and minoramphibolite dykes. Ordovician granites and peg-matites make up a “younger granite suite,” whichintrudes into the Proterozoic basement. The thirdcomponent is a convoluted and puzzling collectionof breccias, of which a granitic breccia is the mostcommon (Fig. 3). Many of these breccias containuranium and minor sulfide mineralization and arecemented with hematite. The breccias occur as ir-regular bodies within the basement complex, adja-cent to a zone of extensive faulting, which also con-tains Ordovician granites and pegmatites (Lambertet al., 1982). The breccias have been assigned an ageof 280 Ma using palaeomagnetic dating methods(Idnurum and Heinrich, 1993).

A fault system, the Paralana Fault zone, con-sisting of northeast-trending faults, runs alongthe eastern margin of the Mount Painter Inlier(Fig. 3). The Paralana Fault zone is generally be-lieved to be the main conduit for hydrothermalfluid dispersal. The fault zone has been intermit-tently active since the Neoproterozoic formationof the Adelaide Rift Complex; seismic studiessuggest the area is still active (Wellman andGreenhalgh, 1988; Foster et al., 1994). A modernhot spring (Paralana Hot Springs) is evidence ofcontinuing hydrothermal activity along the Par-alana Fault zone. The source of this fluid, and ofpast hydrothermal fluids, is difficult to ascertain,although it is most likely to be predominantly me-teoric in origin (Foster et al., 1994).

Apatite fission track analyses (Foster et al.,1994) confirm that significant hot spring activityoccurred along the Paralana Fault zone duringthe mid-Tertiary, ending with a stage of uplift

HYPERSPECTRAL ANALYSIS OF HYDROTHERMAL ALTERATION 339

TABLE 2. MINERALS IDENTIFIED BY PIMA IIMEASUREMENTS IN THE FIELD AREA

KaoliniteBiotitePhengitePhlogopiteHalloysiteCalciteEpidoteGypsumIllite

MuscoviteMontmorillioniteIntermediate chloriteChloriteFe chloritePalygorskiteNontroniteParagoniteJarosite

during or after the Miocene. Mid-Tertiary hy-drothermal activity caused mineralization atdepth and redistributed earlier mineralized zonesin the Mount Painter Inlier (Foster et al., 1994).However, the main time of hydrothermal activ-ity is considered to be Permo-Carboniferous,based on fitting the directions of remanent mag-netism in hydrothermal minerals to the apparentpolar wander path.

GROUND OBSERVATIONS

The area of interest for field study extended .100km2, and ground observations and sampling, us-ing a PIMA II spectrometer, was conducted over anarea of .50 km2. Some of the minerals identified inthe field using the PIMA are listed in Table 2. ThePIMA utilizes SWIR, having 601 contiguous spec-tral bands in the 1,300–2,500 nm wavelength range.The PIMA has a 2-nm sampling interval with ,10nm resolution; this is high enough to identify sub-tle absorption features that can impart importantinformation, such as the degree of crystallinity anddiscrete mineralogical–chemical variations.

TABLE 3. PIMA II ATTRIBUTES VERSUS PROBE-1

PIMA II PROBE-1

� High spectral resolution in the SWIR � Fast and efficient wide area imaging� bands (laboratory-quality spectra) � for mineral mapping and

� environmental monitoring� Measures 601 contiguous spectral � 128 contiguous spectral bands in the� bands from 1,200 to 2,500 nm � 450–2,500 nm range� Signal-to-noise ratio of 3,500; 1 � Signal-to-noise ratio of .500; 1� 1 cm spatial sampling � 2–10 m spatial resolution� Built-in wavelength calibration target � Pre- and/or postflight wavelength and� Can run on both Palmtop and Laptop � radiometric calibration� computers � Operates in any light aircraft equipped� Can provide a spectral analysis in the � with a standard camera port� field in 1–3 min � Can be customized to user

� requirements

The PIMA data were collected to provideground-truthing for the airborne data, whichwere collected using a PROBE-1 instrument. Themain differences between the two instrumentsare listed in Table 3.

PIMA field spectral measurements were made,and hand-samples were collected of rocks that dis-played typical mineralogy for each major outcroparea. A total of 78 spectra were obtained from 60samples. Preliminary processing was carried outin the field using PIMA View software on a Palm-top computer. Background spectra were removedto obtain hull-quotient spectra that could then beused to determine the second-derivative spectra toenhance spectral features and facilitate prelimi-nary identification of minerals (Fig. 4).

Most rocks are made up of mixtures of miner-als in varying proportions. Multiple mineralogiesmay cause overlapping of absorption features,making the spectra difficult to interpret. Second-derivative spectra are crucial in the interpretationof minor and overlapping absorption features,particularly when coupled with a complex back-ground (Yang et al., 2000). The PIMA data wereprocessed using The Spectral Geologist software,which identifies spectra by comparing sampledspectra with extensive spectral libraries.

AIRBORNE OBSERVATIONS

The main aims for processing were to focus onhydrothermal alteration mineral assemblages

THOMAS AND WALTER340

FIG. 4. Examples of reflectance and hull-quotientspectra. A baseline is fitted to the spectral curve (thindashed line). The spectrum is normalized to this line toremove the effect of a spectral background, producing thehull-quotient spectrum. Note that the minor feature, a, be-comes more pronounced on the second derivative spec-trum (from Yang et al., 2000).

and to highlight end-member relationships.Mapping alteration minerals and abundancesaround Mount Painter and Mount Gee wasaimed at finding mineralization patterns thatcould be used to interpret different hydrother-mal episodes. Another aim was to try to identifyalteration assemblages that could be consideredcharacteristic of a hydrothermal system. The dataset, courtesy of Anglo-American Corporation inJohannesburg, was collected by the Australian-designed hyperspectral airborne imaging spec-trometer PROBE-1 (also called Hymap). PROBE-1 has 128 channels in the VNIR and SWIR. ThePROBE-1 data were collected in 3.2-km-wideswaths of varied lengths (average 5 20 km)flown at ,2,200 m (Fig. 5 shows the location ofthe swaths). Each swath, once corrected and cal-ibrated, has 128 reflectance bands that can thenbe used for further processing. Corrections, in-cluding mean terrain height calibrations, meanaircraft height calibrations, and atmospheric (in-cluding water vapor) corrections, were per-formed using an ATREM-based program calledHycorr developed at CSIRO North Ryde. Thispreliminary processing reduces the data to theapparent surface reflectance, the most useful forgeological mapping purposes.

Processing was performed using ENVI v3.2 soft-ware (ENvironment for Visualizing Images),which is specifically designed for image process-ing of satellite and aircraft remote-sensing hyper-spectral data. ENVI is suitable for processing hy-perspectral data because of its many uniqueinteractive analysis tools and multiple dynamicoverlay capabilities. One particularly useful fea-ture of the ENVI software is that it allows users tomake and apply their own customized analysisstrategies. Only the SWIR data from 2,000 to 2,500nm (SWIR2) were used because this range is themost suitable for studying alteration minerals; aSWIR2 subset was used for the following process-ing method applied separately to each swath:

1. Creation of an albedo mask to eliminate pos-sible complications from nongeological fea-tures

2. Normalization of the data to remove per pixelalbedo variation

3. Minimum noise fractionation transforma-tion—to de-correlate and order the spectralbands according to a decreasing signal-to-noise ratio (spectral reduction); uses onlygood pixels to drive statistics; compresses thedata

4. Hyper pixel purity index for spatial data re-duction—finds spectrally purest pixels in thedata by making repeated projections in n-di-mensional scatter plots onto a lower-dimen-sional subspace

5. n-dimensional visualizer rotates the pixelsthrough n-dimensions, producing an ani-mated data cloud by random projections ofn-dimensional space; end-members foundand identified

6. Creation of regions of interest (ROIs)7. Creation of spectral library of end-members

representing the spectrally purest pixels8. Location of known material in an unknown

background—mixture-tuned matched filter-ing

9. Redefinition of ROIs based on low infeasibil-ity and high matched filtering scores (gives

distribution and abundance maps for end-members)

10. Make ROI masks—one for each end-member11. Vectorization of ROIs using CSIRO’s in-

house custom mask creation program de-signed specifically to automate the creationof vectors from ROIs

12. Selection and arrangement of vector layers ofend-members to be shown on the final map

The mineral compositional end-member distri-butions are superimposed onto a base image (aSWIR band) to provide a reference to the groundsurface. Like the PIMA, PROBE-1 providesenough resolution to differentiate between typesof micas. Comparisons with PIMA data wereused to check these minor spectral variations tosee if they had geological significance. In Swath

HYPERSPECTRAL ANALYSIS OF HYDROTHERMAL ALTERATION 341

FIG. 5. Location of PROBE-1 data swaths, each 3.2 km wide. The survey was flown in a north–east/south–westdirection.

3, a ring-like structure displays a muted spec-trum, similar to both kaolinite and halloysite. Theidentification of halloysite is tentative, but PIMAsamples from the same area also identified a hal-loysite variant that is spectrally different from theother kaolinites in the area. Although the spectraare extremely similar, it is interesting to note thatthe two variations have distributions that corre-late with (1) a possible maar structure centeredon Mount Painter with a halloysite variant and(2) a probable later stage of hydrothermal activ-ity associated with the Paralana Fault with theregular kaolinite. Figure 6 shows the differentabundances for the two kaolinites (green 5 pos-sible hallosite variant, yellow 5 regular kaolinite)identified in Swath 3. Analysis of minimum noisefractionation images revealed a ring-like feature,possibly the physical boundary of a maar struc-ture associated with an extensive first stage of al-teration; and a second feature showing a distinc-tive diagonal linear feature (yellow) thatcorresponds with the Paralana Fault, interpretedas a second subsequent stage of alteration in theregion. More detailed geophysical information isneeded to confirm the presence and structure ofthe proposed maar.

GROUND AND AIRBORNE DATARESULTS COMPARISON

Correlations between thin-section observationsand the PIMA results are generally excellent, withmost of the identified PIMA minerals being eas-ily located in thin section, and vice versa. Theclassification made by The Spectral Geologistsoftware was confirmed by detailed manual in-spection of the PIMA spectra. In the case of themany samples containing white micas, the PIMAnot only registered a spectrum diagnostic of mica,but was able to make a more detailed classifica-tion based on crystallographic structure. ThePIMA is able to differentiate between types of mi-cas and between pure muscovites and muscoviteswith significant amounts of large cation substi-tution (aluminium replaced by iron or magne-sium) in the crystal lattice (K. Yang, personalcommunication, 2000). In most cases, if a carefulchoice is made in selecting a representative sam-ple for measurement, the PIMA will provide use-ful mineralogical data.

While the PIMA produces laboratory-qualityspectra, PROBE-1 data do not have the same spec-

tral resolution. PROBE-1 is designed for regionalsurveys, and PIMA provides ground truth, sogreatly enhancing the accuracy of the mineralogi-cal interpretations. The PIMA II field spectrome-ter confirmed the main mineralogies identified bythe airborne survey and provided more detailedmineral information that could be extracted fromthe PROBE-1 images. The PIMA was also able toconfirm the 5-m spatial resolution of PROBE-1 byidentifying and mapping a mica-rich rock unit thathad been uncovered by the building of a road-cut-ting. A distinct mica end-member was mappedthat correlated with the 5-m-wide road (as seen onaerial images). This section of road was observedin the field to have uncovered a highly micaceousrock that occurred only in this road cutting, andonly for the 5-m width of the road.

Two or more minerals can often be inferredfrom indications in PIMA spectra. The same istrue for PROBE-1 data, but separating out indi-vidual minerals is even more difficult as severalminerals typically occur together in a landscapeand can appear as overlapping and “mixed”—this is enhanced by the lower spatial resolutionof the airborne dataset. The PIMA II analysesagreed with many of the minerals identified us-ing PROBE-1. The PIMA data also helped sup-port the sometimes difficult unmixing of spectraand mineral identification made with the PROBE-1 data. The advantage of the PROBE-1 data is thefacility of detailed regional-scale mapping. Beingable to map minor mineralogical variations, suchas the mica end-member discovered along theroad-cutting, makes this type of airborne surveywell suited to mapping surface geology and, inparticular, geological features such as alterationzones. The ability to map large areas, particularlyin a region with rugged terrain such as theFlinders Ranges, has made possible detailed min-eralogical mapping for every 5–10 m2 of theground surface. The general agreement betweendifferent data types is sufficient to confirmPROBE-1’s spatial and spectral resolution andthus the accuracy of the end-member maps.

OTHER ASPECTS OF THEHYDROTHERMAL DEPOSITS

Native sulfur on Mount Painter

An intriguing yellow outcrop was sampled atstop 511.3 (342°1509E 54°8009N) on the northern

THOMAS AND WALTER342

HYPERSPECTRAL ANALYSIS OF HYDROTHERMAL ALTERATION 343

FIG. 6. (a) Botryoidal sulfur, alternating bands of silica (clear) and sulphur (yellow). (b) Enlargement from photo-graph above showing banding and botryoidal texture.

FIG. 7. Looking southwest to Mount Painter showing sample locations. Route starts at southern flank behind andbelow 5111.1. Scale is approximate only.

side of Mount Painter (see Fig. 7 for location). Thisoutcrop, only some 4–5 m2, has not previouslybeen reported. Three samples were collected,spectrally analyzed, and thin-sectioned. Analysisof the PIMA data using The Spectral Geologistsoftware registered “null,” because the principalrock component is quartz, which has no diag-nostic spectral feature at PIMA wavelengths. Pet-rographic study of two of the samples revealedextensive recrystallization of silica (probablyoriginally chalcedonic), and laminations, vugs,cavity-fill, displacive and pervasive replacementtextures in silica, and concentrically banded andbotryoidal structures (see Fig. 8). Electron micro-probe analysis of the yellow material in the lam-inations and botryoidal microstructures identi-fied “native” sulfur, with the rest of the scannedimage being silica, and minor iron. The electronmicroprobe results show that the sulfur is not as-sociated with either the iron or the silica. The sul-fur occurs in a ,2-mm band in sample MU58919.Laminations, some of which are disjointed or bro-ken, appear to have undergone brittle, and pos-sible some minor ductile, deformation.

The presence of native sulfur suggests highlyacid oxidizing conditions, typical of the condi-tions that are found associated with solfataras,steaming grounds, and hot springs.

Origin of sulfur banding

Sulfuric acid can be produced naturally by (a)inorganic and bacterial oxidation of sulfides, (b)disproportionation reaction of sulfur dioxide inmagmatic hydrothermal systems, or (c) acidicbrines in crater lakes on top of active volcanoes(Kusakabe et al., 2000). Mechanism c has beendocumented in acid sulfate hydrothermal alter-ation systems but is more typically associatedwith a crater lake and volcano. The banded sul-fur rock samples MU58919 and MU58921 inhand-sample and in thin section show sedimen-tary characteristics such as graded bedding andsoft-sediment deformation structures akin to thebanded sulfur rocks described by Kusakabe et al.(1986, 2000).

Unbound sulfur is stable under relatively highredox potentials, low temperatures, and high to-tal sulfur concentrations, while hydrogen sulfideis stable only under low redox potentials, hightemperatures, and low total sulfur concentrations(Kusakabe et al., 2000). High-temperature salinefluids from magma interact with overlying crys-

talline rocks at different levels and times duringthe evolution of a system, and can lead to suit-able conditions for high sulfidation ore deposits(Kusakabe et al., 2000). Strongly acidic fluids pro-duced by SO2 disproportionation or inorganic ox-idation of sulfides produce acid alteration ofcountry rocks. Crater-lake bedded sulfur sedi-ments are the closest match to the bedded andbotryoidal sulfur rocks at Mount Painter (Kusak-abe et al., 1986, 2000), although banded sulfur de-posits are common in advanced argillic alterationzones and other hot spring precipitates and sur-face encrustations. Other comparable materialsare the “tiny spheres of sulfur” (Francis et al.,1980) and “elemental yellow sulfur coatings” onsand grains from high-temperature hydrother-mal vents off the island of Milos, Greece (Dandoet al., 1998).

Most of the silica was probably deposited in acryptocrystalline form, or as a gel, as in modernhydrothermal systems (Cady and Farmer, 1996).The jasperoidal unit at Mount Gee could havebeen deposited as a gel (e.g., Eugster and Jones,1967), and many textures in the Mount Painterand Mount Gee quartzes are similar to the botry-oidal silica surfaces observed at the Sleeper de-posit in Nevada (Saunders, 1994), or the recrys-tallized silica gels from Yellowstone NationalPark (Fournier et al., 1991).

Carbon and oxygen isotope analyses

Results for carbon and oxygen isotope analy-ses are shown in Fig. 9. The chart includes 23analyses presented in a paper by Lambert et al.(1982) and a sample from Sunshine Pound ana-lyzed for this project. The analyses are from cal-cite in various hydrothermally altered rock typesaround the central area of the Mount Painter re-gion. The d13C ranges from 222.3‰ to 23.3‰,and d18O from 24‰ to 123.1‰. There is no ap-parent systematic change of isotopic compositionwith location or drill hole depth, and the widespread of values is probably caused by mixing oftwo different waters containing CO2

1 from dif-ferent sources. The results are consistent with aninverse d13C/d18O relationship, confirming amixed origin of the fluids (Lambert et al., 1982).

Most calcite samples from the Mount Painterbreccias have d13C less than 210‰, which sug-gests CO2 generated from organic carbon (Lam-bert et al., 1982). The higher d18O values in themore strongly 13C depleted samples correspond

THOMAS AND WALTER344

with values expected for precipitation from flu-ids derived from argillaceous (carbonaceous)metasediments, as a result of melting or high-grade metamorphic processes. This type of fluid,with d18O around 112‰, could produce hy-drothermal carbonate with a d18O of 123‰ at atemperature of around 170oC (Lambert et al.,1982).

It is likely that the CO2 was derived by solu-tion of magmatic carbonate minerals and from at-mospheric gases in groundwater (Lambert et al.,1982). The sample analyzed in this study, fromthe Sunshine Pound calcite, has a d13C value of23.3‰ and d18O value of 21.3‰. Although thed13C value is the highest, the Sunshine Pound val-ues plot well into the category of probable evo-lution from meteoric waters. Fluctuations in tem-perature of precipitation of the minor carbonatequantities in the Mount Painter samples can ac-count for the scatter of compositions around themixing line.

DISCUSSION

The spectral maps of clays, carbonates, and mi-cas reveal information not previously recorded.These maps show alteration zones around MountPainter and the fault, which can be considered in-dicative of different mineralization styles and,thus, different fluid chemistries. The abundancesof two kaolinite end-members shows that al-though they overlap, they also occur separatelyand delineate different features. The features areinterpreted as being due to different hydrother-mal fluids rather than weathering because theyinclude typical hydrothermal mineral associa-tions. There must also be weathering productspresent, but more work will be required to dif-ferentiate them.

Spectral changes in clay minerals, muscovites,biotites, carbonates, and chlorites were used tomap alteration minerals and to identify differentmineralogical zones that could relate to differenthydrothermal episodes and/or systems.

Mount Painter/Mount Gee can be classified asan epithermal hydrothermal system on the basisof characteristic features: typical epithermal min-erals kaolin, pyrophyllite, and zeolites; chalcedonic(although now recrystallized) silica; vuggy silica;and lattice textures produced by silica pseudo-morphs of bladed calcite. Vuggy silica, extensivein the Mount Gee hydrothermal quartz unit, can

be a good indicator of high-sulfidation deposits(Heald et al., 1987; White and Hedenquist, 1990),but vein and other textural information, com-bined with the scarcity of sulfides, suggests thatthe Mount Painter system is a low-sulfidationsystem (Table 4). Chalcedony indicates local sil-ica saturation and suggests that boiling may haveoccurred. A temperature of between 190°C and100°C and a depth of ,100 m below the watertable is required for its deposition. Chalcedonyalso indicates, owing to its cryptocrystalline na-ture, rapid cooling (Clark and Govett, 1990; Whiteand Hedenquist, 1990). The former presence ofbladed calcite is consistent with this interpreta-tion.

Changes of mineralization characteristics withinthe study area are interpreted to be the result oftwo different hydrothermal fluids from differentsources.

Figure 10 shows a conduit structure that per-mits systems with both separate and shared fluidconduits. It would be expected that any naturalfluid conduits would be utilized by both systems.The repeated and overprinted alteration in thearea makes it difficult to separate hydrothermalsystems or fluid types. However, petrographicdifferences are enough to assume significantlydifferent fluids and hydrothermal conduit sys-tems.

MARTIAN ANALOGUE

The majority of exploration outside Earth hasbeen, and will at least for the next few decades,continue to be via remote means. Mars is a focusfor planetary research, as well as a prime targetin the search for extraterrestrial life.

NASA’s Thermal Emission Spectrometer in-strument that was flown on the Mars GlobalSurveyor mission detected an extensive depositof crystalline hematite (a-Fe2O3). The depositcovers an area of 350 3 350–750 km, has sharplydefined boundaries, and is centered near SinusMeridiani at 2°S latitude between 0° and 5°Wlongitude (Cristensen et al., 2000). Positive iden-tification of crystalline hematite requires thepresence of fundamental absorption featurescentered near 300, 450, and .525 cm21 and theabsence of silicate fundamentals in the 1,000cm21 region (Cristensen et al., 2000). A radiativetransfer model was used to remove atmosphericinterference from CO2, dust, and water ice. To

HYPERSPECTRAL ANALYSIS OF HYDROTHERMAL ALTERATION 345

THOMAS AND WALTER346

FIG. 8. Mount Painter spectra and endmember maps. (a) Identified end-member spectra, including variants of theone mineral. (b) Black and white background image is a shortwave infrared band included to provide reference tothe ground surface. The endmember maps were created using ENVI 3.2 software.

a

b

THOMAS AND WALTER348

FIG. 9. Carbon and oxygen isotope analyses. Breccia and calcite-quartz vein analyses are from Lambert et al. (1982).Sunshine Pound calcite analyses are from this study.

TABLE 4. DISTINCTION BETWEEN HIGH- AND LOW-SULFIDATION SYSTEMS [MODIFIED

FROM WHITE AND HEDENQUIST (1990) AND CORBETT AND LEACH (1994)]

Low sulfidation High sulfidation

Fluid Low salinity Mostly low, some high salinityH2S-dominant SO2-dominantReduced Oxidized

Alteration Generally neutral alteration adjacent Characteristic zoned pervasive acidto structures dominated by alteration from: residual (vuggy)sericitic/illitic clays ï peripheral quartz ï alunite ï kaolinpropylitic veins dominated by minerals ï illite minerals ïquartz 6 carbonate1 propylitic1–4

Intensive white mica in regions with Deep deposits show intensehigh water ; rock ratios1 pyrophilite-white mica alteration

Near surface deposits may havepervasive clay alteration1–3

Associated minerals Low % pyrite High % pyriteGalena, sphalerite. chalcopyrite3 Engarite–luzonite

Metals: economic accessory Au 6 Ag Au 6 CuPb, Zn, Cu AsAs, Te, Hg, Sb at high levels Te at high levels

Character of mineralization Veins common with crystalline Hosted in clasts or matrix inphases at depth, banded at shallow competent wall rock alteration1–4

levels3Characteristic textures Crustification,2,3 banding,3 Vuggy silica (fine-grained quartz)2,3

colloform banding (S),2 Massive silica (fine-grainedchalcedony,2,3 vugs,2,3 vein quartz)1–4breccia,1,3,4 silica pseudomorphsafter bladed calcite1,3,4

Structure Preexisting fractures at depth Dilational structure and permeablesubsidiary dilational structures at lithological control3higher levels Diatreme breccias common3,4

Magmatic, diatreme, and eruptionbreccias2–4

1Present elsewhere in field area.2Present at Mount Painter.3Present at Mount Gee.4Present at Radium Creek.

enhance the unique spectral features at SinusMeridiani the average neighboring spectrumwas removed. The depth and shape of the fun-damental bands indicate that the hematite iscrystalline and relatively coarse-grained (.5–10mm) and thus separate from the fine-grained(diameter ,5–10 mm), red, crystalline hematitethat is considered to be a minor spectral com-ponent in Martian bright regions like Olympus-Amazonis (Morris et al., 1997; Cristensen et al.,2000).

The large hematite deposit at Sinus Meridianicould have been deposited from hydrothermalfluids. The deposit may represent a deposit formedby an ancient thermal plume, which periodicallywelled up at the surface. The phenomenon re-sponsible for the vast size of Martian volcanoes,like Olympus Mons, is the apparent lack of platetectonics. This same phenomenon could also beused to explain the very large size of the proposedhematite deposit at Sinus Meridiani. Other can-didates for hydrothermal deposits have also beenrecognized (Cady and Farmer, 1996).

CONCLUSIONS

Integrated analysis of both air- and ground-col-lected SWIR hyperspectral measurements can beapplied to interpreting hydrothermal alterationin arid regions of Australia. This study suggeststhat it would also be possible to apply similartechniques to searching for hydrothermal alter-ation products on the surface of Mars. Currenttechnology is already at a level where airborne orspaceborne hyperspectral measurements of Marscould be collected and would produce similarlevels of detail about surface mineralogy as in thisstudy.

Mapping ancient hydrothermal systems withhyperspectral imaging techniques is effective andcan provide detailed information. The ability tomap minor spectral features in alteration assem-blages can identify structures that would be difficult to locate using other techniques. The hy-perspectral maps produced show distinctive ar-eas of mineralization commonly associated withthe Paralana Fault zone. The Paralana Fault zone

HYPERSPECTRAL ANALYSIS OF HYDROTHERMAL ALTERATION 349

FIG. 10. Low Sulfidation Gold Classification and Fluid Flow Model. From Corbett and Leach (1994).

is still likely to be the most prolific and active hy-drothermal fluid conduit, but mineralogical andhyperspectral evidence suggests it is not the onlyhydrothermal system in the area. The combinedPROBE-1 and PIMA II hyperspectral data suggestthere is some evidence for two different hy-drothermal fluids, responsible for the interpretedmineral variations.

We propose a model with two separate (but ad-jacent) hydrothermal systems, each with manystages. The first is an extensive adularia-sericitetype system that produced hydrothermal fluid al-teration mainly along the Paralana Fault zone. Itis possible that this system had a fluid of mixedorigin as indicated by the C- and O-isotope analy-ses. The second, a localized system, was respon-sible for the Mount Painter and Mount Gee hydraulic-fracture breccias, and hydrothermalquartz deposits. This second system displayscharacteristics more typical of an acid sulfate sys-tem and likely to be related to a maar structurecentered on Mount Painter. This system, duringa later stage of cyclic temperature changes, couldhave included a solfatara, or other hot spring precipitates, at the ancient surface level. These na-tive sulfur-bearing rocks on Mount Painter rep-resent the ground surface at the time of deposi-tion, agreeing with earlier work (e.g., Coats andBlisset, 1971) based on the proposed (but uncon-firmed) sinter deposits.

The PIMA’s ability to detect mineral changeson spot surface measurements was also testedand shown to contribute accurate data to thelarge-scale mineralogical maps. The combinedstudy of PROBE-1 and PIMA II infrared spectrawas able to demonstrate that these instruments,capable of making detailed mineralogical mapson Earth, could also be applied to finding andmapping hydrothermal alteration and mineral as-semblages on Mars.

ACKNOWLEDGMENTS

We are very grateful to the Anglo-AmericanCorporation of Johannesburg for the provision ofHyMap data, and to CSIRO in Sydney for lend-ing us a PIMA spectrometer. This project wouldnot have been possible without the advice andcomputing support of CSIRO’s Mineral MappingTechnologies Group in Sydney, especially Jona-thon Huntington, Melissa Quigley, Peter Mason,and Kai Yang. Margaret and Douglas Sprigg of

Arkaroola permitted ready access to all areas oftheir property, despite the inconvenience that itsometimes caused, and Doug flew us over ourfield area to give us the benefit of an aerial re-connaissance. Ian Plimer kindly provided a sam-ple from a solfatara deposit on the island of Mi-los. This work was supported by grants from theAustralian Research Council and Macquarie Uni-versity.

ABBREVIATIONS

PIMA, portable infrared mineral analyzer; ROI,region of interest; SWIR, short-wave infrared;TIR, thermal infrared; VNIR, visible and near-in-frared.

REFERENCES

Bock, G.R. and Goode, J.A., eds. (1996) Evolution of Hy-drothermal Ecosystems on Earth (and Mars?), Ciba Founda-tion Symposium 202, John Wiley & Sons, Chichester, UK.

Cady, S.L. and Farmer, J.D. (1996) Fossilization processesin siliceous thermal springs: trends in preservationalong thermal gradients. In: Evolution of HydrothermalEcosystems on Earth (and Mars?), Ciba Foundation Sym-posium 202, edited by G.R. Bock and J.A. Goode, JohnWiley & Sons, Chichester, UK, pp. 150–173.

Clark, D.S. and Govett, G.J.S. (1990) Southwest Pacific ep-ithermal gold: a rock-geochemistry perspective. J.Geochem. Explor. 35, 225–240.

Coats, R.P. and Blissett, A.H. (1971) Regional and eco-nomic geology of the Mount Painter province. S. Aust.Geol. Survey Bull. 43, 1–426.

Corbett, G.J. and Leach, T.M. (1994) Southwest Pacific RimGold-Copper Systems: Structure, Alteration, and Mineral-ization, Special Publication No. 6, Society of EconomicGeologists, Littleton, CO.

Cristensen, P.R., Bandfield, J.L., Clark, R.N., Edgett, K.S.,Hamilton, V.E., Hoefen, T., Kieffer, H.H., Kuzmin, R.O.,Lane, M.D., Malin, M.C., Morris, R.V., Pearl, J.C., Pear-son, R., Roush, T.L., Ruff, S.W., and Smith, M.D. (2000)Detection of crystalline hematite mineralisation onMars by the Thermal Emission Spectrometer: evidencefor near-surface water. J. Geophys. Res. 105, 9623–9642.

Dando, P.R., Thomm, M., Arab, H., Brehemer, M., Hooper,L.E., Jochimsen, B., Schlesner, H., Stor, R., Miquel, J.C.,and Fowler, S.W. (1998) Microbiology of shallow hy-drothermal sites off Palaeochori Bay, Milos (HellenicVolcanic Arc). Cah. Biol. Mar. 39, 369–372.

Drexel, J.F. and Major, R.B. (1990) Mount Painter ura-nium-rare earth deposits. In: Geology of the mineral de-posits of Australia and Papua New Guinea, Volume 2,Monograph Series-Australasian Institute of Mining andMetallurgy. 14, 993–998.

THOMAS AND WALTER350

Eugster, H.P. and Jones, B.F. (1967) Gels composed ofsodium–aluminum silicate, Lake Magadi, Kenya. Sci-ence 161, 160–163.

Foster, D.A., Murphy, J.M., and Gleadow, J.W. (1994)Middle Tertiary hydrothermal activity and uplift of thenorthern Flinders Ranges, South Australia: insightsfrom apatite fission-track thermochronology. Aust. J.Earth Sci. 41, 11–17.

Fournier, R.O., Thompson, J.M., Cunningham, C.G., andHutchinson, R.A. (1991) Conditions leading to recentsmall hydrothermal explosion at Yellowstone NationalPark. Geol. Soc. Am. Bull. 103, 1114–1120.

Francis, P.W., Thorpe, R.S., Brown, G.C., and Glasscock,J. (1980) Pyroclastic sulfur eruption at Poas Volcano,Costa Rica. Nature 283, 754–756.

Heald, P., Foley, N.K., and Hayba, D.O. (1987) Compar-ative anatomy of volcanic-hosted epithermal deposits:acid-sulfate and adularia-sericite types. Econ. Geol. 82,1–26.

Huntington, J.F. (1996) The role of remote sensing in find-ing hydrothermal mineral deposits on Earth. In: Evolu-tion of Hydrothermal Ecosystems on Earth (and Mars?),Ciba Foundation Symposium 202, edited by G.R. Bock andJ.A. Goode, John Wiley & Sons, Chichester, UK, pp.214–235.

Idnurum, M. and Heinrich, C.A. (1993) A palaeomagneticstudy of hydrothermal activity and uranium minerali-sation at Mt Painter, South Australia. Aust. J. Earth Sci.40, 87–101.

Kokaly, R.F., Clark, R.N., and Livo, K.E. (1998) Mappingthe biology and mineralogy of Yellowstone NationalPark using imaging spectroscopy. In: Summaries of the7th JPL Airborne Earth Science Workshop, JPL publication97–21, Jet Propulsion Laboratory, Pasadena, CA, pp.245–250.

Kruse, F.A. (1994) Imaging spectrometer data analysis—a tutorial. Proc. Int. Symp. Spectral Sensing Res. 1, 44–50.

Kusakabe, M., Hayashi, N., and Kobayashi, T. (1986) Ge-netic environments of the banded sulfur sediments atthe Takeyama Volcano, Japan. J. Geophys. Res. 91,12159–12166.

Kusakabe, M., Komoda, Y., Takano, B., and Abiko, T.(2000) Sulfur isotopic effects in the disproportionationreaction of sulfur dioxide in hydrothermal fluids: im-plications for the d34S variations of dissolved bisulfate

and elemental sulfur from active crater lakes. J. Vol-canol. Geotherm. Res. 97, 287–307.

Lambert, I.B., Drexel, J.F., Donnelly, T.H., and Knutson,J. (1982) Origin of breccias in the Mount Painter area,South Australia. J. Geol. Soc. S. Aust. 29, 115–125.

Linton, P. (1998) The application of airborne hyperspec-tral imagery to geological exploration [abstract]. In: Ab-stracts of the 14th Australian Geological Convention, Geo-logical Society of Australia, Sydney, p. 49.

Morris, R.V., Golden, D.C., and Bell, J.F., III (1997) Low-temperature reflectivity spectra of red hematite and thecolor of Mars. J. Geophys. Res. 102, 9125–9133.

Rinker, J.N. (1994) ISSSR tutorial introduction to spectralremote sensing. Proc. Int. Symp. Spectral Sensing Res. 1,5–19.

Saunders, J.A. (1994) Silica and gold textures in Bonanzaores of the sleeper deposit, Humbolt County, Nevada:evidence for colloids and implications for epithermalore-forming processes. Econ. Geol. 89, 628–638.

Wang, Q., Guo, X., and Wang, R. (1998) Lithological dis-crimination using hyperspectral remote sensing. Proc.SPIE 3502, 87–93.

Wellman, P. and Greenhalgh, S.A. (1988) Flinders/MtLofty Ranges, South Australia: their uplift, erosion, andrelationship to crustal structure. Trans. R. Soc. S. Aust.112, 11–19.

White, N.C. and Hedenquist, J.W. (1990) Epithermal en-vironments and styles of mineralisation: variations andtheir causes, and guidelines for exploration. J. Geochem.Explor. 36, 445–474.

Yang, K., Huntington, J.F., Browne, P.R.L., and Ma, C.(2000) An infrared spectral reflectance study of hy-drothermal alteration minerals from the Te Mihi sectorof the Wairakei geothermal system, New Zealand. Ge-othermics 29, 377–392.

Address reprint requests to:Prof. Malcolm R. Walter

Australian Centre for AstrobiologyDepartment of Earth and Planetary Sciences

Macquarie UniversityNorth Ryde, NSW 2109, Australia

E-mail: malcolm.walter@mq.edu.au

HYPERSPECTRAL ANALYSIS OF HYDROTHERMAL ALTERATION 351

This article has been cited by:

1. Juncal A. Cruz, Nuria Sánchez-Pastor, Alexander M. Gigler, Lurdes Fernández-Díaz. 2011. Vaterite Stabilityin the Presence of Chromate. Spectroscopy Letters 44:7-8, 495-499. [CrossRef]

2. Joël Brugger, Pierre-Alain Wülser, John Foden. 2011. Genesis and Preservation of a Uranium-Rich PaleozoicEpithermal System with a Surface Expression (Northern Flinders Ranges, South Australia): RadiogenicHeat Driving Regional Hydrothermal Circulation over Geological Timescales. Astrobiology 11:6, 499-508.[Abstract] [Full Text HTML] [Full Text PDF] [Full Text PDF with Links]

3. Jonathan D. A. Clarke, Carol R. Stoker. 2011. Concretions in exhumed and inverted channels near HanksvilleUtah: implications for Mars. International Journal of Astrobiology 10:03, 161-175. [CrossRef]

4. Dorothy Z. Oehler, François Robert, Malcolm R. Walter, Kenichiro Sugitani, Anders Meibom, SmailMostefaoui, Everett K. Gibson. 2010. Diversity in the Archean Biosphere: New Insights from NanoSIMS.Astrobiology 10:4, 413-424. [Abstract] [Full Text HTML] [Full Text PDF] [Full Text PDF with Links]

5. Michael D. West, Jonathan D.A. Clarke, Jennifer H. Laing, David Willson, James M.A. Waldie, Guy M.Murphy, Matilda Thomas, Graham A. Mann. 2010. Testing technologies and strategies for exploration inAustralian Mars analogues: A review. Planetary and Space Science 58:4, 658-670. [CrossRef]

6. Michael D. West, Jonathan D.A. Clarke, Matilda Thomas, Colin F. Pain, Malcolm R. Walter. 2010. Thegeology of Australian Mars analogue sites. Planetary and Space Science 58:4, 447-458. [CrossRef]

7. Brown AJ, Walter MR, Cudahy TJ. 2005. Hyperspectral imaging spectroscopy of a Mars analogueenvironment at the North Pole Dome, Pilbara Craton, Western Australia. Australian Journal of Earth Sciences52:3, 353-364. [CrossRef]

8. Adrian Brown, Malcolm Walter, Thomas Cudahy. 2004. Short-Wave Infrared Reflectance Investigation ofSites of Paleobiological Interest: Applications for Mars Exploration. Astrobiology 4:3, 359-376. [Abstract][Full Text PDF] [Full Text PDF with Links]

9. 2003. Astrobiology Literature Watch. Astrobiology 3:2, 429-435. [Citation] [Full Text PDF] [Full Text PDFwith Links]