Mars Surface Mineralogy From Hubble Space Telescope Imaging

Transcript

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO. F_A,PAGES 9109-9123, APRIL 25, 1997 Mars surface mineralogy from Hubble Space Telescope imaging during 1994-1995: Observations, calibration, and initial results JamesF. Bell III, • MichaelJ. Wolff,2 PhilipB. James, 3 R. ToddClancy, 2 StevenW. Lee,4 and LeonardJ. Martins Abstract. Visible to near-infraredobservations of Mars were madewith the Hubble Space Telescope(HST) during1994-1995with the goalsof monitoringseasonalvariabilityof the surface andatmosphere andmappingspecificspectralunitsto constrain theplaner'ssurfacemineralogy. Thispaperpresentsthedetailsof thecollectionandcalibrationof the data,concentrating specifically on thenear-IRdatathatwereobtainedexclusivelyfor thesurfacemineralogyaspectof our HST Mars observingprogram. We alsopresentsomeinitial resultsfrom the calibrateddata set. Our calibrationprocedures includedthe standard"pipeline"processing steps,supplemented by specialprocedures requiredfor usewith the linearrampfilterson theWide Field/PlanetaryCamera 2 instrument,andan additionalpointspreadfunctiondeconvolution procedureappliedin orderto realizethe full potentialspatialresolutionof the images(23 to 64 km/pixel betweenAugust 1994 andAugust1995). The calibrationresultsin a setof imagesprojectedontoa standardmap grid andpresented in radiancefactor(I/F) units,havingan estimated--5% photometricaccuracybasedon theperformance of HST andcomparisons with previousground-based andspacecraft Mars spectra. Initial scientificanalysesof thesedatareveal (1) distinctred/bluecolor unitswithin the classical brightregions,similar to thoseseenin Viking Orbiter imagesandpossiblyrelatedto variationsin nanophase and/orcrystallineferricmineralabundance; (2) near-IRspectralslopevariations correlatedwith albedoon a largescale(darkeris "bluer"near-IR slope)but exhibitingwider variationsamongmany of the small-scalefeaturesvisiblein the data;(3) an absorptionat 860 nm that occursin all regionsbut which is 3 to 5% strongerin many of the classicaldark regionsthan in the brightregions,possiblybecauseof a greaterabundanceof a well-crystallineferric phaselike hematiteor a very low Ca pyroxeneor opx/cpxmixture;and (4) an absorptionfrom pyroxeneat 953 nm with a banddepththatis inverselycorrelatedwith albedo(brightregions0 to 5% deep; darkregions7 to 15% deep)andwhich showsthe highestbanddepthvaluesin individualcraters, calderas,and othersmall geologicunitsthat are resolvedin the images. Introduction The surfacemineralogy of Mars providesa window into past and present geologic, geochemical, and hydrothermal processeson the planet. Surface minerals play an important role in the transport, storage, and processingof volatiles. For example, water can be stored either temporarily (adsorbed "surface" water) or semi-permanently (bound or "structural" water) on mineral surfacesand within mineral crystal lattices. Atmospheric gases such as CO 2 or SO2 can be sequestered within minerals through aqueous or other processes, thus providing a sink for ancient atmosphericconstituents[e.g., Fanale et aI., 1992]. The surface-atmospherictransport of volatiles on seasonal timescales is an important part of the Mars volatile cycle and can substantially influence the climatic and radiative behavior of the surface-atmosphere system[e.g., Jakoskyand Haherie, 1992]. Previous investigators have used a combination of groundbased and spacecraft observations to detect or infer the presenceof many different minerals on the Martian surface or in the airborne dust (recent detailed reviews can be found in the works by SoderbIom [1992], Roush et aI. [1993], and Bell [1996]). Iron-bearing minerals make up the majority of these phasesbecauseof (1) the relatively high iron abundanceof the Martian surface [TouImin et aI., 1977; Clark et aI., 1982] and (2) the highly spectroscopically active nature of iron in a variety of different oxides, oxyhydroxides, and silicates [e.g., Burns, 1993]. In addition, observational detection of iron 111111•...,1 •...,111 L.i.i.J o •...,Li.o 1 •...,1 ,h,,,., LIIL&11 that of many other materials ß '• .... L&i•o is l.J ,.,•,-h ...... •,:-•Department of Astronomy,Centerfor Radiophysics and Space becausemost of the relevant mineral absorptionfeatures occur in the visible to short-wave near-IR (-- 400 to 1200 nm) where Research,Cornell University, Ithaca, New York. 2Space ScienceInstitute,Boulder,Colorado. the solar reflected flux is highest, and recent advances in 3Department of Physicsand Astronomy,Universityof Toledo, telescopic and spacecraft instrumentationhave allowed highToledo, Ohio. quality CCD observations to be routinely obtained at these 4Laboratory for Atmosphericand SpacePhysics,Universityof wavelengths. Colorado, Boulder, Colorado. We are attempting to provide additional new information on 5LowellObservatory, Flagstaff, Arizona. the mineralogy of the Martian surface using multispectral observationswith the Hubble Space Telescope (HST). The Copyright1997 by the AmericanGeophysicalUnion HST Papernumber 96JE03990. 0148-0227/97/96JE-03990509.00 instrument suite available for these observations covered only the UV to short-wave near-IR spectral region, so the observations were optimized for the detection and 9109 9110 BELL ET AL.: MARS SURFACE MINERALOGY characterization of iron-bearing oxide, oxyhydroxide, and silicate minerals. In this paper we discussthe details of the specificobservationsthat were obtained during the 1994-1995 apparition of Mars, the data reduction and calibration efforts, and someinitial resultsof our mineralogicinvestigation. Observations and Filter FROM HUBBLE SPACE TELESCOPE informationcan alsobe extractedfrom thesedata [e.g.,Bell et al., 1995; James et al., 1996], as described in more detail below. With additionalobservingtime grantedto us in HST cycle 5, we were able to expandour Mars program(GO Program 5832) to include four more wavelengthsfrom 740 to 1042 nm. Two of these wavelengths (740 and 860 nm) were obtained Selection usingthe WFPC2 linearrampfilters (LRFs) andthe WF4 chip The data were obtained in 15 orbits of HST over 11 different Mars orbital positionsbetween August 23, 1994, and August 21, 1995, using the Wide Field/Planetary Camera 2 (WFPC2) instrument. These dates correspondedto coverage between Mars areocentriclongitudes of Ls=336ø to Ls=145ø, or late northern winter through mid northern summer. This time period allowed for excellent study of the north polar regions, as the sub-Earth latitude ranged up to 26øN; conversely, coverageof surfaceregionsbelow about 60øS was not possible for most of our observations. The opposition was aphelic (Mars near aphelion at opposition), and the largest apparent angulardiameterof Mars in our data set is 13.5 arcsec(roughly a factor of 2 worse than during a perihelic opposition). This correspondsto a best case spatial resolution of 2Dsin(F2/2) = 22.7 km/pixel, where D is the Earth-Mars distance in kilometers (104 million km at closest approach in February 1995) and F2 is the angular resolution of the WFPC2 camera (0.045 arcsec/pixelfor the Planetary Camera chip PC1). This is approximately the same spatial resolution that was obtained by the imaging spectrometerinstrument (ISM) on the Phobos 2 missionin 1989 [Bibring et al., 1990] and is slightly better than the resolution obtained during the Mariner 6 and 7 mission far-encounter phase [Leighton et al., 1969]. The observational circumstances of our data set are summarized in Table 1, and Figure 1 provides an illustration of our image coverageand resolution. WFPC2 has a variety of filters and observing modes available for HST observers[Burrows, 1995]. Our program during HST observingcycle 4 (GO program 5493) utilized the PC1 chip and five filters from 255 to 673 nm (Table 2), specifically concentrating on monitoring of atmospheric activity and surface reflectivity and color changes [James et al., 1996]. A limited amount of compositional/mineralogic (Table 2). The LRFs are a setof continuouslyvariablenarrow- bandpassfilters that allow imaging at --1.1 to 1.3% spectral resolutionat almost any wavelengthbetween370 and 980 nm [Burrows, 1995]. While it is a greatadvantageto be able to image at precisewavelengthsthat are most diagnosticof the specific spectral features being sought, there are some importanttrade-offsanddisadvantages to usingthe LRFs. For example,images at different wavelengthsare obtainedby placing the object at specificparts of each of the four WFPC2 detectorarrays;thusobtainingimagesat a specificwavelength may mean the loss of a factor of 2 in spatial resolution(for example,usingthe WF4 chip insteadof the PC1 chip resultsin a spatial scale of 0.0996 arcsec/pixel) or may result in vignetting problems if the object must be located close to the edge of a chip. Additionally, the calibrationof images obtainedwith the LRF is still being developed,and thususers are forced to develop bootstrapcalibrationtechniques(see Appendix A) for the images until the proper in-flight calibrationdata are obtained and processedby the Space TelescopeScienceInstitute (STScI). The four additionalwavelengthsobtainedin Cycle 5 providediagnostic information on Mars surfacemineralogy. Specifically, 860nm waschosen to detectandmapthespatial extentof the 6Al-->4T2(4G ) electronictransitionbandof Fe3+ (ferric iron) that is primarily characteristicof the iron oxide mineralhematite(ct-Fe203)[Shermanand Waite, 1985;Morris et al., 1985;Bell et al., 1990]. Imagingat 953 nm waschosen in orderto detectand map the spatialextentof the "l-micron" (1-gm) absorptionfeaturein pyroxenesarisingfrom Fe2+ (ferrousiron) ions primarily in M2 crystallographic sites [e.g., Burns,1970, 1993; Adams,1974]. The imagesat 740 nm and 1042 nm provide continuum measurementsfor these Fe2+andFe3+ absorption bands,andtheimagesat 1042nm Table 1. 1994-1995 HST Mars ImagesFrom GO Programs5493 and 5832 UT Datea YYMMDD Time,b Wavelengths, UT Diameter, arcsec nm SE Latitude, deg Phase, deg L•, deg Resolution,c kin/pixel HST Cycle 4 Data 940823 2311 255,336,410,502,673 5.2 5.1 34.4 335.7 59.1 940919 1526 255,336,410,502,673 5.7 11.8 36.5 349.7 54.0 941020 1154 255,336,410,502,673 6.5 17.7 38.3 5.3 46.8 941118 0546 255,336,410,502,673 7.8 21.0 38.0 19.1 39.3 950102 1007 255,336,410,502,673 11.2 21.8 28.0 39.7 27.3 950224 1711 255,336,410,502,673 13.5 17.3 10.0 63.1 22.7 950225 0117 255,336,410,502,673 13.5 17.2 10.3 63.6 22.7 950225 0918 255,336,410,502,673 13.4 17.2 10.5 63.7 22.7 950408 1933 255,336,410,502,673 9.8 18.1 32.6 81.9 31.4 950528 0157 255,336,410,502,673 6.8 23.0 41.7 104.1 45.2 950706 0336 255,336,410,502,673,740,860,953,1042 5.6 25.7 38.9 122.1 55.1 950706 1139 255,336,410,502,673,740,860,953,1042 5.5 25.8 38.9 122.2 55.2 950711 2335 255,336,410,502,673,740,860,953,1042 5.4 25.8 38.1 124.8 54.0 950802 2137 255,336,410,502,673,740,860,953,1042 5.0 25.3 34.4 135.4 61.0 950821 0937 255,336,410,502,673,740,860,953,1042 4.8 23.5 30.8 144.6 64.4 HST Cycle 5 Data aRead940823 as August23,1994. bTimegivenis theapproximate middleof the20-to 35-min.totalobserving sequence. CResolution is the maximumspatialresolutionat the sub-Earthpointfor imagesobtainedon the PC chip. BELL ET AL.' MARS SURFACE MINERALOGY 330 o. FROM HUBBLE SPACE TELESCOPE 0 o, 30 ø 9111 60 ø Mars -, reocentri c Longitude (Ls) 60 ø 90 ø 150 .ø 120 ø 1 ..................... ..:. .... •":•""v:;•' ........... '**":•::•::::':'::' ':'•$?• •:z::' .:. ......... '-::..::½•.:•-:•:....._.• &::•.: ß -2' .•::.......... :•:• • •'•½:" •'•":'•? .2: •Z..• .... . .•2•:• :::• •.•:•4• •::½ ..........%: ..... .•.•.- ,,•--- Figure 1. Schematic representation of the 1994-1995 Mars •mages obtained by HST. Shown are the seasonal date (L s) of each set of multispectral observations,the relative size of the planet when it was observed,and the times when multiple central meridianswere observed. The observationsbetweenLs=335ø and 104ø are the HST Cycle 4 five-color imaging sets,and the observationtaken from Ls=122ø to 145ø are the HST Cycle 5 nine-color imaging sets (Table 1). also provide additional characterization of the l-gm band and a way to test for the possible presence of olivine on Mars [Burns, 1970; Huguenin, 1987]. In combination with the images at 410, 502, and 673 nm that characterize the shape and curvatureof the near-UV reflectancedropoff, the additional four filters obtainedduring HST cycle 5 provide an adequateset of wavelengthsfor the investigationof iron-bearing minerals on the surface of Mars. Data Reduction Instrumental Table 2. HST WFPC2 Filters and Exposure Times Xcenter , FWHM, Filter nm nm Exp.Time1sec. Cycle 4 Cycle 5 F {t W cm-2gm-1 F255W 256 41 350.0 180.0 0.01553 F336W 333 37 6.0 3.0 0.08667 F410M 409 15 4.0 2.0 0.16822 F502N 501 3 7.0 4.0 0.18775 F673N 673 5 0.7 0.6 0.15185 FR680N b 740 10 -- 0.11 0.12919 FR868N b 860 11 -- 0.12 0.09746 F953N 955 5 -- 4.0 0.07696 F 1042M 1044 61 -- 3.0 0.07125 Filter data from Burrows [1995]. aSeeAppendixB. bLinearrampfilter. and Calibration Corrections Raw images were processed using standard HST data reductionprocedures asoutlinedby Lauer [1989] and Holtzman et al. [1995a,b] and using calibration files produced by the WFPC2 InstrumentDefinition Team (IDT). The stepsincluded correctionfor analog-to-digitalconversionerrors, subtraction of bias,superbias,and superdarkframes,correctionfor shutter shading effects, correction for pixel-to-pixel sensitivity variations (flatfielding), and correction of bad pixels and cosmic ray hits. We remove cosmic rays through a combination of automated (low- and high-pass filters) and manual processes. Individual and small groups of undefined pixels are repaired by performing an iterated fourth-order polynomial least-squaresfit to the neighboring pixels. The high signal-to-noise ratio (SNR; see below) and extended nature of our target precludethe need for correctionfor charge transfer efficiency variations [Holtzman et al., 1995a]. The instrumental SNR of our short exposure time HST visible and near-IR Mars images is quite high and is limited, 9112 BELL ET AL.: MARS SURFACE MINERALOGY ultimately, by quantization of the 12-bit analog-to-digital electronics and by scattering of bright Mars light in the telescopeand instrumentoptics. An estimateof the SNR in the raw data can be made by examining the standarddeviation of sky pixelsfar from Mars in the imagesandof scattered-light pixelsadjacentto the limb of Mars. The variationof the sky far from Mars and in all wavelengthsis ñ0.6 to ñ0.8 raw data numbers(DN; the gain was 7 e-/DN for most of our images), and typical scatteredsky variationis ñ2.8 to ñ7.5 DN. If we assumethat the scatteredlight componentalso occursover the FROM HUBBLE SPACE TELESCOPE The second aspect of the absolute photometric calibration processrequiresa correctionfor the shapeof the point spread function (PSF), which leads to "smoothing" of planetary albedo variations because of the telescope and instrument optics. For example, if uncorrected,this effect will result in dark regionson Mars being spectrally contaminatedwith light from adjacentbright regionsor polar deposits. Similar effects occurin regionsof the imagesaroundsteepintensitygradients associated with clouds or other albedo features. It is often extremely difficult to account for this PSF effect in grounddisk of Mars itself, then it is this level of variation that based telescopic or spacecraft imaging instrumentation. governsthe effectiveSNR of the final data. For example,in However, one of the "advantages" of the HST primary's the July 1995 image data the scattering-limitedSNR ranges sphericalaberrationis that the PSF is now bettercharacterized from a low of 150-240 for dark surfaceregionsat 740 nm and for HST and WFPC2 than for perhapsany other optical system 1042 nm to a high of 740-880 for bright regionsat 673 nm ever constructed. The PSF deconvolutionsof our HST images and 953 nm. were performedusing 40 iterationsof the dampedRichardsonLucy algorithm with a thresholdnoise parameterof 3 [White, 1994a,b]. Details and examplesof this procedureare presented Absolute Photometric Calibration by Wolff et aI. [ 1997]. The absolutephotometriccalibrationof WFPC2 data must take into account both the time-variable nature of the system Registration and Photometric Correction throughput,as well as the extendedand inhomogeneous nature (surfacealbedo features) of Mars. For the first component,we rely on the analysesby Holtzman et aI. [1995a,b]and Bagget et aI. [1996] of the extensive standard star observation and monitoringprogramscarried out by STScI. The photometric Perhaps the most difficult aspect of dealing with Mars multispectral data is the fact that the planet rotates significantlyduring the time it takes to obtain a typical set of images. Thus it is not possibleto simply overlay images at different wavelengthsto create ratios or band depth maps; the calibration for the WFPC2 discrete filter observations was images must first be registeredusing map projection software. derivedusingthe SYNPHOT referencefiles providedby STScI We performedthe transformationfrom image (x,y) coordinates (July 1995 update[Bagget et aI., 1996]). These files allow a to projectedlatitudeand longitudeusingautomatedsoftwarewe calibration to be determined that converts corrected DN to flux developedin the IDL programminglanguage. First, the central units [e.g., W cm-2 gm-1) for each filter in our observing (sub-Earth) pixel is automatically found by iteratively program. More details on the derivationof the photometric searchingfor the limb of the planet and fitting an elliptical calibration for the discrete filter observations can be found in the work by Wolff et aI. [1997]. The photometriccalibration of the LRF has not yet been fully determinedby the STScI calibrationprogram.Thus we deviseda bootstrapcalibrationtechniquefor theseimagesthat relies on calibrated ground-based and spacecraft spectra of curveuntil a X2 fit parameteris minimized. Next, the date and time from the image headers are automatically used to determine planetary ephemeris information necessary for the map projection (i.e., sub-Earth and subsolar latitude and longitude, north polar angle, distance). Finally, the images can be projected to one of 14 types of cylindrical, equal-area, Mars and the information from STScI that is available on the conic, or polar azimuthal projections. Our software uses fully LRF system throughput. This LRF photometric calibration ellipsoidal map projection formulae [Snyder, 1985; schemeis outlined in Appendix A. Bugayevskiy and Snyder, 1995], as HST images are sharp The flux values for both the discrete filter images and the enoughto detect the small but nonzero flattening of Mars, and LRF imageswerethenconvertedto radiancefactoror I/F using spherical projection formulae would result in detectable the methods of Roush et al. [1992] and Bell et aI. [1994] (I is mapping errors. the actual irradiancereceived from Mars within each HST pixel Our map projection software also outputs images of the and •F is the theoretical irradiance received within each HST incidence and emission angles for each pixel, thus allowing pixel from a perfectlydiffusingLambertiansurfaceilluminated by the Sun and viewed at normal geometryat the heliocentric photometric corrections to be applied in order to properly interpret absolutereflectancelevels in all regions within about distance of Mars [Hapke, 1981]). Details are presented in 60 ø of the sub-Earth and subsolar points. For the Appendix B. photometrically corrected images discussed here, we used a The absolutephotometricerrors in this calibration process simple Minnaert correction with a constant "typical" k are conservatively estimated to be approximately 2 to 5% for parameter of 0.7 [e.g., Harris, 1961; de Grenier and Pinet, the discretefilter observations[Holtzman et al., 1995b; Wolff 1995]. et aI., 1997], but are likely 5 to 10% for the LRF observations The end result of the data reductionand calibrationprocess because of the additional assumptions and uncertainties is a setof co-registered and map-projected imagecubes(spatial discussedin Appendix A. We have not applied throughput x spatial x spectral), calibrated to absolute I/F units, that can correctionsfor UV contamination effects, thus worsening the photometricaccuracy for the F255W filter data by possibly be analyzed using a variety of spectroscopicand imaging spectroscopyanalysis tools. more than 5%. However, this filter is not critical for mineralogicstudies,and flux values at 255 nm and 336 nm will vary by 5-10% or more anyway, dependingon the dust and cloud opacity and the Martian airmass [e.g., Clancy et aI., 1996a, Wolff et al., 1997]. The magnitude of the UV contamination effect is less than 1% for all of the other filters that we used [Holtzman et aI., 1995b]. Despite these uncertainties, at this current level of accuracy, these data represent some of the best calibrated Mars observations ever obtained from Earth. Resultsand Interpretations Spectra Some representativespectrafrom our July 1995 nine-color observations are shown in Figure 2. Figure 3 presents a comparisonbetween the HST spectra and previous calibrated Mars reflectancespectra obtained by McCord and WestphaI BELL ET AL.' MARS SURFACE MINERALOGY FROM HST Bright Region Spectra 0.70.•'" •'" •"' •'" HUBBLE 9113 HST Dark Region Spectra 0.70 •"' SPACE TELESCOPE I'"l'"l'"l'"l'" (A) 0.60 0.60 6 9 5 0.50 0.50 7 5 9 0 4O 0.40 6 5 7 ,.5 4 . 0.50 O3O • 6 5 0.20 0.20 4 5 0.10 o.1o 2 1 0.00 400 600 ,., 0.00 1,,,I,,,•,,,•,,,•,,, 200 1 800 1000 1200 Wavelength(nm) 200 400 600 800 1000 1200 Wavelength(nm) Figure 2. Representative radiancefactor(I/F) spectraextractedfrom the observations on July 6, 1995, at 1139UT. (a) Spectrafrombrightregions:(1) Moab, (2) Xanthe,(3) Chryse,(4) Tempe,(5) NorthPolarCap, (6) Ares/TiuValles (NASA Mars Pathfinderlandingsite). (b) Spectrafrom dark regions:(1) NorthernAcidalia, (2) SouthernAcidalia, (3) Sinus Meridiani, (4) Oxia Palus, (5) Margaritifer Sinus, (6) Mare Australe, (7) westernsideof northpolarsandsea,(8) easternsideof northpolarsandsea,(9) SinusSabaeus.Each spectrum is from a 3x3 pixel box, and the errorbar shownrepresents the varianceof the spectrawithin that box. Each spectrumis offsetby 0.05 unitsfrom the one below. [1971] and Mustard and Bell [1994]. The HST spectraare generallyconsistentwith previousmeasurements.There is a systematic increasein the reflectivityof the HST spectraat the shortestwavelengthsas comparedto the 1969 and 1988-1989 data. This is likely a manifestationof the increasedcloudiness of Mars during the aphelic apparition of 1994-1995 [e.g., Martin et al., 1995; Clancy et al., 1996b; Jameset al., 1996] relative to the earlier observations that were obtained closer to perihelion. The effect of cloudsis to increasethe reflectivity preferentially at the shortestwavelengthsbecause of the increasedRayleigh scatteringefficiency and also becausethe surfaceitself is extremelydark in the blue and near-UV. The HST nine-color spectra display interesting and diagnosticcharacteristicsthat are consistentwith previous spectroscopic investigations [e.g., McCord and Westphal, 1971; McCord et al., 1977a; Singer et al., 1979; Bell et al., 1990]: (1) The slope of the near-UV absorption edge that gives Mars its distinctive ruddy color varies with reflectivity such that bright regions are typically "redder" than dark regions; (2) There is a reflectivity maximum near 750 nm with a position that is not a function of absolutereflectivity; (3) There is a broad absorption band in the short-wave near-IR between the reflectivity maximum near 750 nm and the longwavelength extent of our data at 1042 nm. The near-IR spectral slope between 750 nm and 1042 nm is "red" (reflectivity increasing at longer wavelength) for bright regionsbut is neutralto "blue"for dark regions; (4) There is a weak absorption/inflection at 673 nm superimposed on the 9114 BELL ET AL.' MARS SURFACE MINERALOGY FROM HUBBLE SPACETELESCOPE Typical Dark Regions TypicalBrightRegions .4 .4 McCordand Westphal (1971): Spot 69-1 McCordand Westphal (1971): Spot 69-6 Mustardand Bell (1994): Spot 88-41 Mustard and Bell (1994): Spot 88-22 HST July 1995: ChrysePlanitia HST July 1995: MargaritiferS•nus LL c'..,. -o cO 0 2 ß O (A) _ (B) _ 0 200 400 600 800 1000 1200 Wavelength (nm) 2OO 400 600 800 1000 1200 Wavelength (nm) Figure 3. Comparisonof spectraextractedfrom our HST data set and previousground-based and spacecraft calibratedspectraof Mars. The McCord and Westphal[1971] data are geometricalbedoat 5ø phaseangle,the Mustard and Bell [1994] compositespectraare in reflectance,and the HST data are in I/F. near-UV absorptionedge in the HST spectraof bright regions that is usually absent in the spectra from dark regions (however, see spectrum4 (Oxia Palus) in Figure 2b); (5) The reflectivity of the residual north polar cap is considerably higher than that of the surface at wavelengthsshorterthan 673 nm and is comparable to or slightly higher than that of the surfaceat 673 nm and longer. Despite the usual whitish/bluish appearancein typical images of Mars, the cap is in fact quite red. Similar overall spectral character is also observed in the other Cycle 5 nine-color HST image sequencesobtained in August 1995. The five-color image sequences obtained betweenAugust 1994 and May 1995 (Cycle 4) are not able to detect spectralvariationslongward of 673 nm, but the spectra from 255 to 673 nm exhibit color variations consistent with the Cycle 5 data. Global Color Variations Rather than individually examine tens of thousands of spectra, a more fruitful way to explore spectral variations is through the use of color ratio images and similar imageoriented analysis techniques(see Appendix C). James et al. [1996] provided an initial analysis of red to blue (673 nm to 410 nm) color variations and spectral units from the HST Cycle 4 images obtainedin February 1995. Analysis of twodimensional (2-D) histogram scatter plots of 673 nm versus 410 nm I/F valuesby Jameset al. [1996] revealeda numberof distinctcolor units and showedthat the ubiquitouscloud cover observed near aphelion can substantially influence the interpretation of multispectral images obtained in blue and near-UV wavelengths. Specifically, clouds produce elevated reflectance values shortward of 502 nm and thus frustrate efforts to derive the true surface color ratio values. We presentadditionalvisible-wavelengthcolor ratio data in Figure 4 and Plate 1. Figures 4a and 4b show the reflectances at 673 nm and 410 nm for the Syrtis Major-centered hemisphereas imaged in February 1995, and Figure 4c shows the ratio of these two images. Plate l a presents a 2-D histogramplot of the 673/410 nm color ratio (ordinate) versus the reflectanceat 673 nm (abscissa). The ratio image and histogram both reveal a number of distinct color units, delineatedin Plates l a and lb. The 2-D histogramshows a large and diffuse cluster correspondingto the dark, moderate 673/410 nm ratio regions Syrtis Major, Hesperia Planum, and VastitasBorealis (blue in Plates la and lb); a compactcluster corresponding to the bright, very high 673/410 nm ratio region encompassingparts of Elysium and Utopia Planitia (magenta);and anotherrather diffuse clustercorrespondingto the bright, high 673/410 nm ratio regions Arabia and Isidis BELL ET AL.' MARS SURFACE MINERALOGY FROM HUBBLE SPACE TELESCOPE 9115 Planitia (yellow). Plate l a also shows there to be a substantial partsof theplanetnot imagedin Cycle4 (Figure5a). The fine amountof spectralmixing between thesecolor units, and Plate detailsare slightlydifferentbetweenthe February1995 Cycle ø) andJuly1995Cycle5 data(Ls=122 ø) because of the lb demonstratesthat the mixing occurs primarily along the 4 (Ls=63 boundaries between the individual units. Other outlier color units in the 2-D histogram correspond primarily to the polar cap and other bright condensateregions (green), and regions along the limb and at high emission angles, where wavelength-dependentlimb darkening accountsfor most of the color variation (red and cyan). These same general color ratio units can also be identified in our Cycle 5 HST images in the Syrtis region and in other different distribution of clouds at this later seasonal date. In general,our Cycle 5 HST imageshavethe ability to extend previous ground-based andspacecraft colorunitresultsintothe short-wave near-IR, where additional mineralogic information can be obtained from the broad region of 750-1050 nm absorption thatresultsfrom a combination of ferricandferrous minerals.Figures5b, 5c, and5d showexamplesof three-color ratio and 2-D histogrampairsfrom our near-IR data. The ratio between 740 nm and 1042 nm (Figure 5b) is a measure of the overall near-IR spectral slope. The near-IR slope is sensitive to mineralogy (especially Fe2+-bearing minerals with a strong 1-pm absorption feature), the opacity and composition of Mars atmospheric aerosols, and the presence of particle coatings or rinds [e.g., Fischer and Pieters, 1993; Erard et al., 1994]. Figure 5b shows that the 740/1042 nm color units occur in two primary clusters (linked by a well-defined mixing trend) and that the low-albedo regions have approximately 20% higher 740/1042 nm ratio values than the high-albedo regions. This result can also be seen in the individual spectraof Figure 2: darker regions have flat or negative near-IR spectral slopes, while brighter regions have positive near-IR spectral slopes. The ratio between 860 nm and 953 nm (Figure 5c) is a measure of the relative strengths of the 860-nm ferric absorption band and the 953-nm ferrous absorption band. Figure 5c (and Figure 2) reveals that the low-albedo regions generally exhibit flat spectra between 860 and 953 nm, althoughthere is a roughly +5% 860/953 nm color ratio value variation among dark regions. The brightest regions exhibit "red" 860/953 nm spectral slopes (860/953 ratio values < 1.0), and also show +5% color ratio variations. There is a clear mixing trend between these endmembercolor ratio units. The ratio between953 nm and 1042 nm (Figure 5d) provides a way to characterize the shape of the 1-pm pyroxene absorptionfeature, which is centered near 920 to 950 nm for low-Ca orthopyroxenes and near 950-1000 nm for high-Ca clinopyroxenes[e.g., Adams, 1974; Pinet and Chevrel, 1990; Mustard et al., 1993]. The global variation in the 953/1042 nm ratio is small (mean+1o = 0.92+0.03), and there is only a O* weak correlation between albedo and 953/1042 nm ratio. However, Figure 5d demonstratesa stunningexampleof one of the greatestassetsof our data set: spatial resolution. A number of small surface regions (from 60 to 100 km in size) exhibit significantly lower 953/1042 nm ratio values than the rest of the planet, meaning that these regions have increased953-nm +60' +30':' c Figure 4. Red/blue color ratio resultsfrom the February 1995 HST observations near opposition. All images are shown in a Molleweide projection of the Martian eastern hemispherenorth of-60 ø, with grid lines at every 30ø of longitude and15ø of latitude. (a) Map of calibratedI?F at 673 nm, showing prominent Syrtis Major dark albedo feature (center) as well as the bright regionsIsidis and Arabia, the retreatingnorthpolar cap, and the Hellas Basin(-45ø, 300ø), (b) Map of calibratedI?F at 410 nm, showingthe loss of surfacealbedofeature contrastin the blue, exceptfor the north polar cap, cloudsforming along the morningand evening limbs and in the southpolar region, and a discretecloud over the volcano Elysium (+20ø, 220ø), (c) Ratio of 673 nm (Figure 4a) to 410 nm (Figure 4b). The image has been enhancedso that black corresponds to a ratio value of 0.5 and white corresponds to a ratio value of 6.2. 9116 BELL ET AL.' MARS SURFACE MINERALOGY FROM HUBBLE SPACE TELESCOPE I ! I I I :::"l'" '1' I ! -- m -- -- m -- --- -- m -- ,-- m -- C) -- m --- -- --. .. --- m --.. ,-,-- .. .-.. 0.00 0.10 0.20 0.30 0.40 67,%nm I/F Values ,i ß ß ß,,......,:'....,::..:,:+60 o. ß....70 ß ß ß :0 -•,• ß ....... ß: ............... ß ß e" ß ......:.½ .......: ß ß ß ß :;.......•...... +3;0o...• ß ß ß ß ß ß ß ß ß ß ß ß ß ß ß ß........ ß; ......... ß• ........ •ß........ •ß ........ -ß ......... ß• .... ß ß . : * : ........ ß................... ß ß ß=........ ß. ß ...... ß =......... . ß ; ........ % ßß ß '.. : .**......... =......... • ........ .:....... ßß ß ß : % : - ' ..... 30' .... -:......... ß ß ß : 0 ß.... eß ..... ß o ßß ' .......... e'lle : :ß.................. ß el ß ß ß . ß........... ,...............•.......•.......•..... ,........ '"'"'% .....:......:'.....•'-60' ...... 0ø : .... '" 300' .•.'..•:: :'.... •; "'" 240 ø 180' Plate 1. Red/bluecolorratioresultsfrom the February1995HST observations nearopposition.All images are shownin a Molleweideprojectionof the Martianeasternhemisphere northof-60 ø, with grid linesat every 30ø of longitude and 15ø of latitude. (a) Two-dimensional (2-D) histogram of the673/410nm ratio(Figure4c) versusradiancefactorat 673 nm (Figure4a). This andsubsequent 2-D histograms are shownso thata higher frequencyof occurrenceis seen as a darker and densercluster. Here, black correspondsto five or more occurrences of eachdata value in eachparticular(x,y) bin, and white corresponds to 0 occurrences. Different clustersof datahavebeengroupedby color,(b) Spectralunit mapderivedfrom the histogramunitsin Plate l a. This unit map providesa representative exampleof how spectralvariationsseenin ratio imagescan be correlatedwith specificsurfaceregionsor geologic/albedounits. See text for details. absorption.Theseregionsare associated with specificimpact craters in the Acidalia hemisphere,with the Nili Patera and Meroe Pateracalderaeof Syrtis Major, and with the small dark teardrop-shapedfeature to the northeast of Syrtis Major (referredto as Nubis Lacusand/orAlcyoniusNodusin classical albedomaps;it doesnot appearto have any uniquegeologic characteristic). Absorption Band Depth Variations While color ratios provide diagnostic information on spectral slope variations, we chose our Cycle 5 imaging wavelengthsto allow for the mappingof the actual absorption features associatedwith iron-bearing minerals on the Martian surface. Specifically, we can use the images at 740 nm and BELL ET AL.' 2t.0• 150' 90* 30•W MARS SURFACE MINERALOGY •0'* 270' 210' FROM HUBBLE SPACE TELESCOPE 21.0 • t50' 90 •*W 5'•0' 27'0• 9117 ZlO" 1.10 1,05 .,.., E .• 0.95 o }... Ec0.90 ............. :*'- • ".,.';':.::':i-i ,'"'-:,-:•.:..,..,...,....;,: -:.... :,::.;,::,:½:;•?%.:.,..;::d' .. :: "- 0.85 :% .:.::..: 0,80 O. '0 0.20 0.30 0_40 0 50 Figure5. Initialcolorratioresults fromtheJuly1995HSTobservations. All of theratioimages presented herearestretched to encompass thesamerangeasthey axisof theircorresponding histogram, andareshown ona Molleweide mapprojection centered on30øWlongitude, 0ølatitude.Pixelshaving anincidence angleor emission anglegreaterthan60ø wereremoved fromboththemapsandthehistograms. An estimate of the 1-• x axisandy axiserrorbaris shownfor each2-D histogram (seeAppendixC). (a) The 673/410nm colorratio imageand 2-D histogramof 673/410 nm ratio versus673-nmradiancefactor, (b) The 740/1042 nm color ratioandhistogram of ratioversus 1042-nm radiance factor,(c) The860/953nmcolorratioandhistogram of ratioversus953-nmradiance factor, (d) The 953/1042nm colorratioandhistogram of ratioversus953-nm radiance factor. 1042 nm as continuumpointsand spatiallymap the strength of the 860-nm (ferric) and 953-nm (ferrous)absorptionbands usingthe banddepthmappingtechniquesdefinedby Clark and Roush [1984] and Bell and Crisp [1993]. Details of the method and the techniqueused to estimateerrors in the band depthmapsare providedin AppendixC. Figure 6a displays a map of the 953-nm pyroxene observations absorption feature relative to a linear continuum defined absorptionin Mars dark regions[e.g., Bell, 1992; Murchie et between740 nm and 1042 nm. This map revealsthe powerof HST as a mineralogic mapping tool. It is apparentthat the 953-nm band is much strongerin the dark regionsthan in the brightregions,consistentwith the spectraof Figure 2 and the colorratio dataof Figure5. However,the strengthof the 953- al., 1993; Mer•nyi et al., 1996]. strengthof the 860-nm band seen in our HST Mars spectra exhibitsa weak inversecorrelationwith albedo:dark regions generallyhave a stronger860-nm featurethan bright regions by an averageof 3 to 4%. This relationshipis difficult to derivefrom examiningjust a few spectra(as in Figure2). This result is consistentwith previousground-basedand Phobos2 that showed evidence for increased "ferric" Anotherinterestingresultis shownin Figure 6c, which is a map of the depth of the 673-nm band seenin the spectraof Figure 2 relative to a linear continuumdefined between 502 nm and 740 nm. The map and the histogramreveal the curvatureof nm bandvarieswithinthe darkregions:the bandis strongest the Martian spectrumbetween 502 and 740 nm, which may be in SyrtisMajor andothernearbyequatorialdark regions,but it relatedto the degreeof crystallinity of surfaceiron mineralogy is weaker in the northern dark regions Acidalia and [e.g., Guinnesset al., 1987; Morris and Lauer, 1990; Morris et Utopia/Borealis. al., this issue]. While there is no systematictrend above the Figure 6b shows a map of the 860-nm ferric oxide errors in 673-nm band depth versus albedo, the 673-nm band absorption feature relative to the same linear continuum varies suchthat the spectraof most surfaceregionsare convex between740 nm and 1042 nm. This map reveals that the (negativeband depth in Figure 6c). Some bright regionsand a 9118 BELL ET AL.' MARSSURFACEMINERALOGYFROMHUBBLESPACETELESCOPE 70.'"' 1 .'t0 1,0:5 0,95 '• 1,00 .,,• 0,90 ,. '% 0,95 ::::'*:L:; .... E 0,9-0 o 85 0.85 0.10 8. i5 0.20 0.25 0 30 0,35 0.40 0.10 O, •.5 0.20 0.25 0.,30 0,35 0.40 953 953 nm I/F volue.• Figure (continued) similar to the ratio map of Pinet and Chevrel [1990]: bright regions are spectrally flat in the near-IR, while dark regions are "blue." However, Figures 5d and 6a reveal individual craters and caldera with much stronger 953-nm pyroxene absorption features than typical dark regions. Comparison Discussion with the 953/1042 nm ratio image (Figure 5d) indicates that The resultspresentedabove provide new informationon the these regions of increased 953-nm absorptionare likely not spectroscopic variability of the Martian surface. For example, causedby the presenceof olivine or very high-Ca pyroxene color ratio images and 2-D histogram analysisreveals that the becausethere is no associatedstrong increasein the 953/1042 classical bright regions can be subdivided into at least two nm ratio value for these areas. The lack of atmospheric distinct red/blue color units that occur in spatially distinct interference/turbulence and the exceptional ability to model regionsin Isidis and Utopia Planitia (Plate la and Figure 5a). and correct the HST PSF [see Wolff et al., 1997] allow this This result is consistentwith the color ratio and 2-D histogram level of spatialdetail to be obtainedin the near-IR for the first analysesof Viking Orbiter imaging by Soderblom et al. time. The implication is that there are small regions of the [1978] and McCord et al. [1982]. While the Viking data have Martian surface that exhibit a very strong pyroxene higher spatial resolution and thus can reveal finer details absorptionband either because (1) there is simply a higher associatedwith specific craters or other features,it was abundanceof pyroxene in the rocks and soils of these regions; restrictedin wavelength to only three bandsbetween 450 nm (2) the pyroxenein theseregionsis "fresher"or less alteredto and 590 nm. The HST images can thus extend the color ratio ferric phasesthan in other regions, and/or (3) the particle size results into the near-IR and provide additional diagnostic of the pyroxene-bearing surface minerals is larger, thus informationkeyed to specific mineralogicvariations. leadingto increased953-nm absorption. Our HST near-IR color ratio results are consistent with The ubiquity of the 860-nm absorption feature and its with albedo came as somewhat of a previous ground-based near-IR imaging observations by inverse correlation McCord et al. [1977b] and Pinet and Chevrel [1990], and they surprise, especially considering the well-known correlation extend these results by providing higher spatial resolution between red/blue color ratio and albedo. However, the measurements that are free from the typical terrestrial apparent increase in ferric mineral content in some dark atmospheric contamination problems encountered between regions has been noted by previous ground-basedobservers, 750 and 1050 nm (e.g., 02, H20). For example, the overall and these HST data provide confirmation. How could dark appearanceof the HST near-IR slope image (Figure 5b) is regions have a stronger 860-nm absorption band yet a smaller number of isolated intermediate and dark regions exhibit concave spectra(band depth > 0) between 502 and 740 nm (cf. Figure 2), however. BELL ET AL.' MARS SURFACEMINERALOGY FROM HUBBLE SPACETELESCOPE . 9119 . ::::::::::::::::::::::::: :::::S• : x:.3• :•.' .............. ,'•:• .... %.. 210ø 1_0 90• 30•W 3%½ 270ø 210ø 0.20 0.16 0,10 0.I0 0.05 0,00 0,05 O. lO 0.20 0.30 0.40 9'53 nm I/F vo:u•: 0.10 0.15 0.20 0.25 030 0.35 0.40 860 nm I/F volues Figure 6. Initial banddepthmappingresultsfrom the July 1995 HST observations.All of the banddepth imagespresentedhere are stretchedto encompassthe same range as the y axis of their corresponding histogramand are shownon a Molleweidemap projectioncenteredon 30øW longitude,0ø latitude. Pixels having an incidenceangle or emissionangle greaterthan 60ø were removedfrom both the maps and the histograms.An estimateof the 1-o x axisandy axiserrorbar is shownfor each2-D histogram.SeeAppendix C for a discussion of the banddepthmappingtechniqueand error analysismethod. (a) The 953-nm banddepth defined relative to a linear continuum between 740 nm and 1042 nm. There is a clear trend above the error indicatingthat the low albedo regionsgenerally exhibit a 5-10% deeper 953-nm absorptionthan bright regions, (b) The 860-nmbanddepthdefinedrelativeto a linearcontinuumbetween740 nm and 1042 nm. There is a weak correlation above the 1-o level between 860-nm band depth and albedo such that low-albedo regionsexhibita =3-5% deeper860-nmbandthan high-albedoregions, (c) The 673-nm banddepthdefined relativeto a linear continuumbetween502 nm and 740 nm. There is no systematiccorrelationbetween673nm banddepthand albedo,althoughthe largestvariationis seenamongintermediate-and low-albedoregions. shallowervisible spectral slope? One possible solution is that the visible spectral slope is dominated by poorly crystallineferric-richmaterialsimilarto the nanophase ferric oxidepigmentsstudiedby Morris andLauer[1990], Morris et al. [1993, 1997], and others. This pigmentingmaterial does not have an 860-nm absorptionfeature but it does have a strongred spectralslope. The 860-nm bandcouldthen arise from a well-crystalline ferric oxide phase like hematite occurringpreferentiallyin the lower albedo regions. The additionalabsorption at 860 nm in the dark regionsmay result from moreof this ferric phaseoccurringin theseregions. This possibilityis supported by the observation of greater673 nm band depth variability in low albedo regions (Figure 6c) becausemany ferric oxides also exhibit an absorptionin the 600 to 700-nm region as well as the strongerfeaturecentered near 850 to 900 nm. Alternately, the additional 860-nm absorptionin the dark regions may be causedby a band centeredlongwardof 860 nm that has a broad wing extending down through 860 nm. In this case, the origin of the additional860-nm absorptioncould be ascribedto a very low Ca pyroxene,an orthopyroxene-clinopyroxene mixture [e.g., Mustard and Sunshine, 1995] or an iron oxyhydroxide phase like goethite. Given the coarsespectralsamplingprovidedin our HST data set and the broad overlap in absorptionbands possiblefor variousferric and ferrousphases[e.g., Morris et al., 1995], it is not possibleto uniquely determinethe origin of this additional 860-nm absorption. The optimal way to providethis discriminationwould be with a combinationof high spectral resolution and high spatial resolution observations. 9120 BELL ET AL.' MARS SURFACE MINERALOGY FROM HUBBLE SPACE TELESCOPE from 5% to 15% deep. The spatialdistributionof this band is not a strongfunction of albedo, unlike the red/blue spectral slope. These HST data confirm previous ground-based observations, indicating that many of the classical dark regions(includingAcidaliaand Syrtis)exhibit--3 to 5% deeper 860-nm absorptionfeature than the classicalbright regions. The origin of this additional860-nm absorptionin the dark regionsmay be related to a greater abundanceof a wellcrystalline ferric phase or a very low Ca pyroxene or orthopyroxene/clinopyroxenemixture. 4. Mappingof the depthof the 953-nm ferrousabsorption band showslarge variationsin the depth of the 1-pm feature. Bright regionstypically have either no band or only a weak absorption, while dark regionsexhibit a 953-nm bandranging from --7% to 15% deep. The increasesin 953-nm banddepth within individual craters, calderas, and other small geologic units in the dark regionsare interpretedas indicatingeither an increasein the abundance, "immaturity," and/or particle size of pyroxenein theseareas. ..c •0.00 Appendix A' Photometric Calibration of LRF Images The scheme that we developed to determine the LRF photometric calibration proceeded as follows. We used the STScI/WFPC2 Exposure Time Calculator (ETC) software for extended objects to estimate the expected Mars DN value for each LRF wavelength (at the appropriate gain and exposure time settings). The current version of the ETC uses preflight calibration data in order to provide an estimate of the photometricperformance of the LRFs and thus does not take into accountthe likely differences between preflight and inflight performance and calibration. To estimate the photometric calibration of the LRF images, we input the E .-0.05 -0.1 o -0,15 o. 15 0•10 0 20 o. 25 0.50 67,,,5nm I/F values Figure 6. appropriateMars surfacebrightness(mag/arcsec 2) and use a (continued) spectraltype G5 stellar spectrumas the sourceto approximate reflected sunlight. An estimated photometric scale factor (in W cm-2gm-• DN-•) is thenobtainedto convertDN to flux by Conclusions This paper presents the details of the collection and calibrationof imagesof Mars obtainedby HST during 19941995, as well as some initial analyses. The images were obtainedbetween 255 nm and 1042 nm as part of a long-term HST investigation of seasonalphenomenaon Mars. The primarygoalof obtainingthe near-IRimagesdiscussed hereis the study of Martian surface mineralogy. The calibration exerciseperformedon the data resultedin a spectacular set of imagesthatexhibit--5% photometricaccuracyfrom 410 nm to 1042 nm. Our initial scientific examination of these calibrated data shows the images to compare favorably with previous ground-based and spacecraft imaging and spectroscopic observations.Some of the most salientresults found to date include the following. 1. Discrimination of at least two distinct red/blue color units within the classicalbright regions, with one unit having a 20-25% higher 673/410 nm ratio value than the other, possibly becauseof an increase in nanophaseferric iron abundance. 2. Mapping of near-IR spectral variations on a scale unprecedented in previousEarth-basedtelescopicobservations and coveringmany regionsnot yet imagedby spacecraftin the near-IR. The maps reveal general correlation between 1042/740 nm spectralratio and albedo, and the increased spatial resolution allows the near-IR color properties of individual cratersand small geologicunits to be investigated. 3. Mapping of the depth of the 860-nm ferric absorption band reveals that all surface regions exhibit a band ranging dividing the input Mars surface brightnessby the DN value estimatedby the ETC. In comparison with calibrated ground-based spectra of Mars, the ETC-based LRF calibration schemeyields 740 and 860-nm flux values that are systematicallyapproximately20% too high. This is likely the result of compoundingerrorsfrom the various system throughput estimates made by STScl in developing the ETC in the absence of a completely characterizedLRF in-flight calibration and from the derivation of the solar flux convolved through the LRF bandpasses(the transmission functions of the LRF filters were only approximated, and the in-flight system efficiency of the WFPC2-LRF combination has not yet been determined; see Appendix B). Given the various uncertaintiesinvolved, a 20% absoluteerror is not unexpectedlylarge. In order to correct for this systematicoffset as best we can, we usedthe ground-based/ISMcompositereflectancespectraof Mustardand Bell [1994] to help determinea correctionfactor for the 740 and 860-nm LRF data. This was a two-step process: first, we used the LRF photometriccalibration values derivedaboveand the solarflux valuesdeterminedin Appendix B to calibratethe data to I/F, and then we extracteda typical bright region spectrumfrom the Isidis region in the July 6, 1995, image cube from 0336 UT (Table 1). We then examined the Mars bright region composite spectrum of Spot 41 (Olympus-Amazonis) from Mustard and Bell [1994] and determinedthe ratios of the compositespectrum'sreflectances to the HST I/F valuesat 740 and 860 nm. The averageof these ratios was found to be 0.80. Thus the LRF photometric calibrationvalues derived above are multiplied by 0.8 to yield BELL ET AL.: MARS SURFACE MINERALOGY FROM HUBBLE SPACE TELESCOPE 9121 Error propagation using standard derivative-based error a more accurate estimate of the actual I/F values of the HST data. Scalingbothof the LRF imagesin this way by the same formulae [e.g., Bevington, 1969] yields the following amountalso preservesthe value of the ratio betweenthese equationsfor calculatingerrorsin ratios (I R) and band depth wavelengths.We notethat it is not criticalthat the exactsame maps(IsD): regionof the planetis usedfor this scalingtechnique:what reallymattersis thattwo spectraarechosenthatarebothfrom IR = Is / I L "typical" bright regions (thus likely having similar mineralogy) andthatbothhavesimilarreflectance levels. The -- IL! Olympus-Amazonis spectrum of MustardandBell [1994]and øR _ •/Ors 2+IRCrL 22 (Cla) (Clb) the HST Isidis spectrumsatisfythesecriteria. AppendixB: Determinationof I/F Values The flux values for both the discrete filter images and the IBD: 1- (IB / Ic) (C2a) _••foB 2+(IB/Ic)[Os(1-f) 2 2+o•f 22] OBD-Ic (C2b) LRF imageswereconverted to radiancefactoror I/F usingthe methodsof Roushet al. [1992] and Bell et al. [1994] modified Knowledgeof the errorson eachof the images(Os, os, OL) for the squarepixelsof HST (as opposed to circularapertures). should come from a rigorous formal propagationof errors This modification results in the expression for F M, the along the data reduction pipeline. However, this is not theoretical irradiance received within each HST pixel from a possiblefor our HST databecausethe errorson manyof the perfectlydiffusingLambertiansurfaceilluminatedby the Sun standardSTScI calibration productsand reductionalgorithms and viewedat normalgeometryat the heliocentricdistanceof usedin the pipelineare unknownor indeterminate. Instead,for Mars [Roushet al., 1992, equation 8], becoming our analysis(Figures5 and 6) we rely on a more empirical approachby adoptingthe error bars in the representative FM ---- 4Fs sin 2(l-l/2) •D 2 (B1) wherenFs is the solarirradianceat 1 AU, gl is the angularsize of an HST pixel in arcsec,andD is the heliocentricdistanceof Mars in AU. Values of F$ at each of our HST wavelengths (Table 2) were obtainedby using the World Meteorological Organization(WMO) solar flux spectrumof Wehrli [1985, brightand dark regionspectraof Figure3 as the "typical" errors for our HST data. These are not true instrumental errors, per se, but are measuresof the variationof homogeneous surfaceunits over a 3x3 or 5x5 pixel surface region. These errorsareusedas the [os, os, oœ]valuesin equationsClb and C2bandoRandOBDare calculated separately for bright(Figure 3a) anddark (Figure3b) regionsfor eachratio imageor band depthmap. The largerof the calculatederrorsfor brightand darkregionsis shownasthe verticalerrorbar in Figures5 and 1986]. For the discreteHST filters, the WMO solar spectral 6; the horizontal error bar on the I/F valuescomesdirectly data were convolved with the system efficiency function for from the data in Figure 2. eachWFPC2 filter as describedby Wolflet al. [1997]. For the LRFs, the systemefficiencyof the filters in flight has not yet beenprovidedby STScI or theWFPC2 IDT, sowe estimatedFs by simply convolvinga gaussianfilter transmissionprofile havinga centerat eachof the LRF wavelengths anda full width at half maximum appropriatefor each wavelength [Burrows, 1995] with the full-resolution WMO solar flux spectrum. There is obviously much uncertaintyin this determinationof F$ for theLRFs(seeAppendixA), butthe valuescanbe refined once the resultsof the ongoing STScI LRF calibrationprogram are completed. Final calibratedI/F values were derived by dividing the Mars flux values determined using the photometric calibrations Acknowledgments. We are extremely grateful to WFPC2 IDT members David Crisp and Karl Stapelfeldt for their assistancewith determiningthe best possibleflatfields for the LRF images. We thank Andy Switala and Tom Daley for crucial help in developingthe automated map projection and data analysis software, and Paul Helfenstein for assistance with the voodoo art of photometric calibration. We thank B. Ray Hawke and a mystery reviewer for providinga careful review of the initial manuscript,and JohnMustard for reviewingan on-line versionof the revisedpaper. Fundingfor this researchwas providedby grantsfrom the NASA PlanetaryGeologyand GeophysicsProgram(NAGW-5062) and the SpaceTelescopeScience Institute. This research was based on observations with the NASA/ESA Hubble Space Telescope obtained at the Space Telescope Science Institute,which is operatedby Associationof Universitiesfor Research definedby Wolffet al. [1997] andin AppendixA above(IM) by the valuesof FM determinedusingequation(B 1). in Astronomy under NASA contractNAS5-26555. Appendix C: Color Ratios, Band Depth Maps, References and Error Analysis Color ratios are formed by simple coregistration and divisionof two images.Band depth maps are calculatedusing Adams, J.B., Visible and near-infrareddiffuse reflectancespectraof pyroxenesas appliedto remotesensingof solid objectsin the solar system,J. Geophys.Res., 79, 4829-4836, 1974. threecoregistered imagesand a techniquebasedon Bell and Bagget, S., W. Sparks, C. Ritchie, and J. MacKenty, Contamination correction in SYNPHOT for WFPC-2 and WF/PC-1, WFPC2 Crisp [1993]: images on the short-wavelengthside (I s, wavelength•,s) and long-wavelength side (IL, wavelength•,L) of an absorptionband at wavelength•,B are usedto constructa continuumimage(I½) that represents the value of eachpixel at •,s alonga line definedby the pixel valuesat •,s and•,L. The lnstrum.Sci. Rep. 96-02, SpaceTelescopeSci. Inst., Baltimore, Md., 1996. Bell, J.F., III, Charge-coupleddevice imaging spectroscopy of Mars, 2, Resultsand implicationsfor Martian ferric mineralogy,Icarus, 100, 575-597, 1992. fractionaldistance,f, betweenthe absorptionband wavelength andthe short-wavelength continuumpoint is (•,s- •'S)/(•'L-•'S), Bell, J.F., III, Iron, sulfate,carbonate,and hydratedmineralson Mars, in Mineral Spectroscopy:A Tribute to Roger G. Burns,Geochem.Soc. andthusthe continuumimage I½ is simplyequalto (1 -f)Is + fIœ. Thebanddepthis thendefinedas 1 - (I B/ Is), whereIB is the imageat •,s. This definitionof band depthallows for an intuitive display of results: areas of an image having more absorptionappearbrighter. Spec. Pub. 5, edited by M.D. Dyar, C. McCammon, and M.W. Schaefer, 359-380, 1996. Bell, J.F., III, and D. Crisp, Ground-basedimagingspectroscopy of Mars in the near-infrared:Preliminary results, Icarus, 104, 2-19, 1993. 9122 BELL ET AL.: MARS SURFACE MINERALOGY Bell, J.F., III, T.B. McCord, and P.D. Owensby, Observationalevidence of crystalline iron oxides on Mars, J. Geophys. Res., 95, 14,447-14,461, 1990. Bell, J.F., III, J.B. Pollack, T.R. Geballe, D.P. Cruikshank, and R. Freedman, Spectroscopyof Mars from 2.04 to 2.44 [tm during the 1993 opposition: Absolute calibration and atmospheric vs. mineralogic origin of narrow absorption features, Icarus, 111, 106-123, 1994. SPACE TELESCOPE The performanceand calibration of WFPC2 on the Hubble Space Telescope,Publ. Astron. Soc. Pac., 107, 156-178, 1995a. Holtzman, J.A., C.J. Burrows, S. Casertano,J.J. Hester, J.T. Trauger, A.M. Watson, and G. Worthey, The photometricperformanceand calibration of WFPC2, Pub. Astron. Soc. Pac., 107, 1065-1093, 1995b. Huguenin, R.L., The silicate componentof martian dust, Icarus, 70, 162-168, 1987. Bell, J.F., III, P.B. James,L.J. Martin, R.T. Clancy, S.W. Lee, and D. Crisp, Mars surface mineralogy from Hubble Space Telescope multispectralimaging: 1994 pre-oppositiondata (abstract),Lunar Planet. Sci. Conf., XXVI, 95-96, 1995. Bevington,P.R., Data Reduction and Error Analysisfor the Physical Sciences,336 pp., McGraw-Hill, New York, 1969. Bibring,J.-P., et al., ISM observationsof Mars and Phobos:First results, Proc. Lunar Planet. Sci. Conf., 20th, 461-471, 1990. Bugayevskiy, L.M., and J.P. Snyder, Map Projections: A Reference Manual, Taylor and Francis,Bristol, Pa., 1995. Bums, R.G., MineralogicalApplicationso.fCrystalField Theory, 224 pp., CambridgeUniv. Press,New York, 1970. Bums, R.G., Origin of electronic spectraof minerals in the visible-near infrared region, in Remote GeochemicalAnalysis: Elemental and Mineralogical Composition, edited by C. Pietersand P. Englert, pp. 3-29, CambridgeUniv. Press,New York, 1993. Burrows, C.J. (Ed.), Wide Field and Planetary Camera 2 Instrument Handbook, version 3.0, Space Telescope Sci. Inst. Pub., Baltimore Md., June 1995. Clancy, R. T., M. J. Wolff, P. B. James,E. Smith, Y. N. Billawala, S. W. Lee, and M. Callan, Mars ozone measurements near the 1995 aphelion:Hubble SpaceTelescopeultraviolet spectroscopywith the Faint Object Spectrograph,J. Geophys.Res., 101, 12,777-12,784, 1996a. Jakosky,B.M., and R.M. Haberle, The seasonalbehavior of water on Mars, in Mars, edited by H.H. Kieffer, B.M. Jakosky, and M.S. Matthews,pp. 969-1016, Univ. of Ariz. Press,Tucson, 1992. James,P.B., J.F. Bell III, R.T. Clancy, S.W. Lee, L.J. Martin, and M.J. Wolff, Global imaging of Mars by Hubble Space Telescopeduring the 1995 opposition,J. Geophys.Res., 101, 18,883-18,890, 1996. Lauer, T.R., The reduction of Wide Field/Planetary Camera images, Publ. Astron. Soc. Pac., 101, 445-469, 1989. Leighton, R.B., N.H. Horowitz, B.C. Murray, R.P. Sharp, A.G. Herriman, A.T. Young, B.A. Smith, M.E. Davies, and C.B. Leovy, Television observations from Mariners 6 and 7, in Mariner-Mars 1969: A Preliminary Report,NASA Spec.Pub. 225, 37-82, 1969. Martin, L.J., J.F. Bell III, P.B. James, S.W. Lee, and D. Thompson, "Mars Watch" observations from the Lowell Observatory, and observed cloudiness in 1994-95: Possible implications for dust activity, Mars TelescopicObservationsWorkshop,LPI Tech. Rep., 95-04, 19-20, 1995. McCord, T.B., and J.A. Westphal, Mars: Narrowband photometry, from 0.3 to 2.5 microns, of surface regions during the 1969 apparition,Astrophys.J., 168, 141-153, 1971. McCord, T.B., R.L. Huguenin, D. Mink, and C. Pieters, Spectral reflectance of Martian areas during the 1973 opposition: Photoelectricfilter photometry, 0.33-1.10 [tm, Icarus, 31, 25-39, 1977a. Clancy,R. T., A. W. Grossman,M. J. Wolff, P. B. James,D. J. Rudy, Y. N. Billawala, B. J. Sandor, S. W. Lee, and D. O. Muhleman, Water vapor saturationat low altitudes around aphelion: A key to Mars climate?, Icarus, 122, 36-62, 1996b. Clark, B.C., A.K. Baird, R.J. Weldon, D.M. Tsusaki, L. Schnabel, and M.P. Candelaria, Chemical composition of Martian fines, J. Geophys.Res., 87, 10,059-10,067, 1982. Clark, R.N., and T.L. Roush, Reflectance spectroscopy:Quantitative analysis techniquesfor remote sensing applications, J. Geophys. Res., 89, 6329-6340, 1984. de Grenier,M., and P.C. Pinet, Near-oppositionMartian limb darkening: Quantification and implication for visible-near-infrared bidirectional reflectance studies, Icarus, 115, 354-368, 1995. Erard, S., J. Mustard, S. Murchie, J.-P. Bibring, P. Cerroni, and A. Coradini, Martian aerosols: Near-infrared spectral properties and effects on the observation of the surface, Icarus, 111, 317-334, 1994. Fanale, F.P., S.E. Postawko, J.B. Pollack, M.H. Carr, and R.O. Pepin, Mars: Epochal climate changeand volatile history, in Mars, edited by H.H. Kieffer, B.M. Jakosky,and M.S. Matthews,pp. 1135-1179, Univ. of Ariz. Press, Tucson, 1992. Fischer, E.M., and C.M. Pieters, The continuum slope of Mars: Bidirectionalreflectanceinvestigationsand applicationsto Olympus Mons, Icarus, 102, 185-202, 1993. Guinness,E.A., R.E. Arvidson, M.A. Dale-Bannister,R.B. Singer, and E.A. Bruckenthal, On the spectral reflectance properties of materialsexposedat the Viking landing sites,J. Geophys.Res., 92, E575-E587, FROM HUBBLE 1987. Hapke, B., Bidirectional reflectance spectroscopy, 1, Theory, J. Geophys.Res.,86, 3039-3054, 1981. Harris, D.L., Photometryand colorimetryof planetsand satellites,in The Solar SystemIII: Planetsand Satellites,pp. 272-342, editedby G.P. Kuiper and B. Middlehurst,Univ. of ChicagoPress,Chicago,Ill., 1961. Holtzman, J. A., J. J. Hester, S. Casertano,J. T. Trauger, A.M. Watson, G. E. Ballester, C. J. Burrows, J. T. Clarke, D. Crisp, and E. Evans, McCord, T.B., R.L. Huguenin,and G.L. Johnson,Photometricimagingof Mars duringthe 1973 opposition,Icarus,31, 293-314, 1977b. McCord, T.B., R.B. Singer, B.R. Hawke, J.B. Adams, D.L. Evans, J.W. Head, P.J. Mouginis-Mark, C.M. Pieters, R.L. Huguenin, and S.H. Zisk, Mars: Definition and characterizationof global surfaceunits with emphasison composition,J. Geophys.Res., 87, 10,129-10,148, 1982. Mer6nyi, E., R.B. Singer,and J.S. Miller, Mapping of spectralvariations on the surface of Mars from high spectral resolution telescopic images,Icarus, 124, 280-295, 1996. Morris, R.V., and H.V. Lauer Jr., Matrix effects for reflectivity spectra of dispersed nanophase (superparamagnetic) hematite with application to Martian spectral data, J. Geophys. Res., 95, 5101-5109, 1990. Morris, R.V., H.V. Lauer Jr., C.A. Lawson, E.K. Gibson Jr., G.A. Nace, and C. Stewart, Spectral and other physicochemicalpropertiesof submicronpowdersof hematite(c•-FeaO3),maghemite(¾-Fe203), magnetite (Fe304), goethite (c•-FeOOH), and lepidocrocite (¾-FeOOH),J. Geophys.Res., 90, 3126-3144, 1985. Morris, R.V., D.C. Golden, J.F. Bell III, H.V. Lauer Jr., and J.B. Adams, Pigmenting agents in Martian soils: Inferences from spectral, M6ssbauer,and magnetic propertiesof nanophaseand other iron oxides in Hawaiian palagonitic soil PN-9, Geochim. Cosmochim. Acta, 57, 4597-4609, 1993. Morris, R.V., D.C. Golden, J.F. Bell III, and H.V. Lauer Jr., Hematite, pyroxene,and phyllosilicateson Mars: Implications from oxidized impact melt rocks from ManicouaganCrater, Quebec, Canada, J. Geophys.Res., 100, 5319-5329, 1995. Morris, R.V., D.C. Golden, and J.F. Bell III, Low-temperature reflectivity spectra of red hematite and the color of Mars, J. Geophys.Res.,this issue. Murchie, S., J. Mustard, J. Bishop, J. Head, C. Pieters, and S. Erard, Spatial variations in the spectralpropertiesof bright regions on Mars, Icarus, 105, 454-468, 1993. Mustard,J.F., and J.F. Bell III, New compositereflectancespectraof Mars from 0.4 to 3.14 gm, Geophys.Res. Lett., 21, 353-356, 1994. BELL ET AL.: MARS SURFACE MINERALOGY Mustard, J.F., and J.M. Sunshine, Seeing through the dust: Martian crustalheterogeneityand links to the SNC meteorites,Science,267, 1623-1626, 1995. Mustard, J.F., S. Erard, J.-P. Bibring, J.W. Head, S. Hurtrez, Y. Langevin,C.M. Pieters,and C.J. Sotin, The surfaceof SyrtisMajor: Compositionof the volcanicsubstrateand mixing with altereddust and soil,J. Geophys.Res.,98, 3387-3400, 1993. Pinet, P., and S. Chevrel, Spectralidentificationof geologicalunitson the surface of Mars related to the presence of silicates from Earth-based near-infrared telescopic charge-coupled device imaging,J. Geophys.Res., 95, 14,435-14,446, 1990. Roush, T.L., E.A. Roush, R.B. Singer, and P.G. Lucey, Estimates of absoluteflux and radiancefactor of localizedregionson Mars in the 2-4[tm wavelengthregion, Icarus, 99, 42-50, 1992. Roush,T.L., D.L. Blaney, and R.B. Singer, The surfacecompositionof Mars as inferred from spectroscopic observations, in Remote GeochemicalAnalysis: Elemental and Mineralogical Composition, editedby C. Pietersand P. Englert, pp. 367-393, CambridgeUniv. FROM HUBBLE SPACE TELESCOPE 9123 Toulmin, P., III, A.K. Baird, B.C. Clark, K. Keil, H.J. Rose Jr., R.P. Christian, P.H. Evans, and W.C. Kelliher, Geochemical and mineralogicalinterpretation of the Viking inorganic chemical results,J. Geophys.Res.,82, 4625-4634, 1977. Wehrli,C. (Ed.), WorldRadiationCenterPub.615, July 1985. Wehrli, C. (Ed.), World Climate Research Programme Publication Series, No. 7, WMO ITD-No. 149, 119-126, Oct. 1986. White, R. L., Image restorationusing the Damped Richardson-Lucy method,in AstronomicalDam AnalysisSo. ftware and SystemsIII, editedby D. R. Crabtree,R. J. Hanisch,and J. Barnes, ASP Conf Ser., 61, 292-295, 1994a. White, R. L., Image restoration using the damped Richardson-Lucy method,in The Restorationo.fHST Imagesand SpectraII, editedby R. J. Hanischand R. L. White, pp. 104-110, SpaceTelescopeSci. Inst., Baltimore, Md., 1994b. Wolff, M.J., S.W. Lee, R.T. Clancy, L.J. Martin, P.B. James,and J.F. Bell III, 1995 observationsof Martian dust stormsusingthe Hubble SpaceTelescope,J. Geophys.Res., 102, 1679-1692, 1997. Press, New York, 1993. Sherman,D.M., and T.D. Waite, Electronicspectraof Fe3+oxidesand Am. Mineral., 70, J.F. Bell III, Departmentof Astronomy,Centerfor Radiophysics and Space Research,424 Space SciencesBuilding, Cornell University, Singer,R.B., T.B. McCord, R.N. Clark, J.B. Adams, and R.L. Huguenin, Mars surface composition from reflectance spectroscopy: A summary,J. Geophys.Res., 84, 8415-8426, 1979. Snyder,J.P.,Map ProjectionsUsedby the USGS,U.S. Geol. Surv.Bull., Ithaca, NY 14853-6801. (email: [email protected]) R.T. Clancy and M.J. Wolff, Space Science Institute, 1234 oxide hydroxides in the near-IR to near-UV, 1262-1269, 1985. 1532, 1985. Soderblom, L.A., The compositionand mineralogy of the martian surfacefrom spectroscopic observations: 0.3-50 [tm,in Mars, edited by H.H. Kieffer, B.M. Jakosky,and M.S. Matthews,pp. 557-593, Univ. of Ariz. Press, Tucson, 1992. Soderblom, L.A., K. Edwards, E.M. Eliason, E.M. Sanchez, and M.P. Charette, Global color variations on the martian surface, Icarus, 34, 446-464, 1978. Innovation Drive, Suite 294, Boulder, CO 80303-7814. P.B. James,Departmentof Physicsand Astronomy,University of Toledo, Toledo, OH 43606. S.W. Lee, LASP, University of Colorado,Boulder, CO 80309. L.J. Martin, Lowell Observatory,Flagstaff,AZ 86001. (Received June 7, 1996; revised December accepted December 19, 1996.) 12, 1996;