SEARCH

SEARCH BY CITATION

Keywords:

  • spectroscopy;
  • remote sensing;
  • life in extreme environments;
  • Mars;
  • instruments and techniques

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[1] The Life in the Atacama project examined six different sites in the Atacama Desert (Chile) over 3 years in an attempt to remotely detect the presence of life with a rover. The remote science team, using only orbital and rover data sets, identified areas with a high potential for life as targets for further inspection by the rover. Orbital data in the visible/near infrared (VNIR) and in the thermal infrared (TIR) were used to examine the mineralogy, geomorphology, and chlorophyll potential of the field sites. Field instruments included two spectrometers (VNIR reflectance and TIR emission) and a neutron detector: this project represents the first time a neutron detector has been used as part of a “science-blind” rover field test. Rover-based spectroscopy was used to identify the composition of small scale features not visible in the orbital images and to improve interpretations of those data sets. The orbital and ground-based data sets produced consistent results, suggesting that much of the field sites consist of altered volcanic terrains with later deposits of sulfates, quartz, and iron oxides. At one location (Site A), the ground-based spectral data revealed considerably greater compositional diversity than was seen from the orbital view. One neutron detector transect provided insight into subsurface hydrogen concentrations, which correlated with life and surface features. The results presented here have implications for targeting strategies, especially for future Mars rover missions looking for potential habitats/paleohabitats.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[2] The Life in the Atacama (LITA) project included rover field tests designed to look for life in the arid environment of the Atacama Desert in Chile. Over three field seasons, six separate tests (referred to as Sites A–F) were conducted with a field team in Chile responsible for rover logistics and assistance in data acquisition. For the last 2 years (2004–2005, Sites B–F), a remote science team at Carnegie Mellon University (Pittsburgh, PA) planned daily rover traverses and analyzed the returned data [Cabrol et al., 2007]. These later tests were designed to simulate a Martian rover mission where the remote science team has limited knowledge of the field site: discussion of similar data sets analyzed after the field test at Site A is also included. Due to the blind nature of the study, previously published geologic analyses and ground truth information were not used in analysis of orbital or rover data, nor in rover traverse planning. Results discussed in this paper relied only on orbital and field test data sets for this reason; a limited comparison with ground truth results is discussed only for validation of the remote analysis. One of the key aspects of the field tests was the remote identification of life through rover investigation of potential habitats. The purpose of this portion of the larger study was to support the search for possible habitats by identifying the composition of the surface and subsurface material.

[3] Potential sites of interest for rover investigation were initially identified via orbital data sets of the field areas made available to remote science team approximately one week prior to each field test. These data sets included both visible/near-infrared (VNIR) multispectral and hyperspectral reflectance and thermal infrared (TIR) multispectral emissivity. Resulting processed images included geomorphologic maps, false-color images sensitive to concentrations of chlorophyll, and end member images indicating regional composition. Areas of interest within a simulated landing ellipse were identified on these images prior to the start of each field test and were considered priority targets for the traverse. Once the simulated rover landing site was identified, local targets were examined using rover instrumentation to characterize these locales, identify potential areas of further interest, and to refine orbital data analysis. The path traveled by the rover from one locale to another is termed a transect, and the total path from the simulated landing site locale to the final locale is referred to as the rover traverse. During remote operations, the location of the rover was estimated by triangulation using landmarks visible in both rover and satellite imagery [see Cabrol et al., 2007]. Because triangulation was not possible at all locales due to lack of imagery, the rover traverses presented here are derived from GPS points acquired by the field team and made available to the remote science team after field tests were complete.

[4] The structure of this paper will mimic the structure of the data analysis itself. First, we present our methods and results from orbital data sets, and then the methods and results from the rover data sets. These results are compared and discussed both in the context of their implications for the mineralogy of the field sites, how these results might be improved for future tests, and the validity of these results in terms of ground truth comparison. Finally, we discuss the implications of our results for future Mars rover missions.

[5] The Atacama Desert, located in northern Chile, is one of the driest deserts on the Earth. This arid zone lies along a central valley bounded on the west by a coastal mountain range and to the east by the pre-Andean cordillera. The regional geology consists of Mezosoic and Cenozoic sedimentary units interspersed with volcanic deposits and intrusives [Dingman, 1967; Reutter et al., 1996; Pueyo et al., 2001; Hartley and Chong, 2002; Dunai et al., 2005]. Sedimentary units include fluvial, lacustrine, alluvial fan, and evaporite deposits. In places, these evaporites take the form of large surface halite deposits [Stoertz and Ericksen, 1974] or thick sequences of saline soils consisting mostly of sulfates with localized concentrations of economic minerals such as nitrates, iodates, and copper [e.g., Bao et al., 2004; Palacios et al., 2005]. Recent microbiological studies have focused on organisms surviving in the extreme environment of the Atacama [e.g., Warren-Rhodes et al., 2006; Wierzchos et al., 2006], and have extended a possible comparison to Mars [e.g., Navarro-González et al., 2003]. The field sites in this study were chosen to represent different environments within the Atacama: for details and specific locations, see Cabrol et al. [2007] and Warren-Rhodes et al. [2007a, 2007b]. The background information in this section was not known to the remote science team until after rover operations were complete.

2. Orbital Data Sets

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[6] Orbital data in both the visible and infrared wavelength regions were used for analysis of regional geology. VNIR data were used for photogeologic observations, identification of potential concentrations of chlorophyll, rover traverse planning, and locale triangulation. TIR data were used to examine the regional composition. Three different satellite instruments were used as sources for the VNIR data sets: the Hyperion hyperspectral imager aboard NASA's Earth Observing 1, GeoEye's IKONOS satellite, and the Terra satellite's Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER).

2.1. Hyperion

[7] Hyperion is a hyperspectral imager with a spatial resolution of 30 m/pixel and 242 bands covering a wavelength range from 0.355 to 2.577 μm [Pearlman et al., 2003]. The visible wavelengths can be used to create a “true color” representation of the surface (red, green, and blue wavelength bands displayed in red, green, and blue (RGB) respectively), and the infrared wavelengths contain information about the composition. A single Hyperion strip was used for analysis of Sites A and D, although the image does not include the rover traverse at Site D.

2.2. IKONOS

[8] The IKONOS data set [Cook et al., 2001] has a spatial resolution of 1 m/pixel with four bands: blue, green, red, and near infrared (NIR) (0.445–0.516 μm, 0.506–0.595 μm, 0.632–0.698 μm and 0.757–0.853 μm respectively). The NIR, red, and green bands were utilized to make a “false color” image with the NIR, red, and green bands in RGB respectively. Concentrations (on the order of a pixel in size) of chlorophyll, which typically has a low reflectance in visible wavelengths and a high reflectance in the NIR, appear red in this color scheme [Vincent, 1997, pp. 111–112]. IKONOS data were only available for Sites B and D.

2.3. ASTER

[9] Both VNIR and TIR data sets from the ASTER instrument [Abrams, 2000] were used during field tests at Sites B–F. The VNIR data have 3 bands (Band 1, green: 0.52–0.60 μm, Band 2, red: 0.63–0.69 μm, and Band 3N, nadir-pointing NIR: 0.78–0.86 μm) and a resolution of 15 m/pixel. Images could be displayed in a fashion similar to the IKONOS false-color image (ASTER bands 3N, 2, and 1 in RGB, respectively).

[10] TIR data from the ASTER instrument were used to understand the composition of each field site. In the TIR, ASTER has 5 bands ranging from 8.1 to 11.7 μm and a spatial resolution of 90 m/pixel. Level 2 processed surface emissivity data (georeferenced and atmospherically corrected [Gillespie et al., 1998]) were analyzed using ITT Visual Information Solutions' ENVI software (versions 4.0 and 4.1) to isolate spectral end members in the data set. Processing steps included a forward minimum noise fraction (MNF) transform [Green et al., 1988] and creation of a pixel purity index [Boardman et al., 1995]. The result was a set of spectral end members taken from the original emissivity data that represented a majority of the pixels in the scene. Because of the small number of spectral bands, only a limited number of end members can reasonably be extracted from these data. It is unlikely that every minor composition will be represented, but major surface constituents should be identifiable. These spectral end members were then analyzed using linear deconvolution (via the method of Ramsey and Christensen [1998]) to determine specific mineral and/or rock end members, as these derived spectra likely represented a mixture of compositions.

[11] A spectral library of likely compositions for this field area (known to be an arid region near a volcanic range) was compiled for use in TIR deconvolutions (Table 1). Sources for this library included mineral and rock reflectance spectra from the Johns Hopkins University (JHU) library [Salisbury et al., 1991] converted to emissivity, spectra of rock samples from Ward's Natural Sciences acquired at Arizona State University (ASU), zeolite spectra collected at ASU [Ruff, 2004], and the Thermal Emission Spectrometer (TES) mineral library also acquired at ASU [Christensen et al., 2000] (currently available at http://tes.asu.edu/speclib/index.html). Spectra were resampled to ASTER wavelengths using instrument band passes included in ENVI.

Table 1. List of 108 End Members Used for TIR Linear Deconvolutions, Grouped by Source
ASU-TESaJHUbWard'scZeolitesd
Acmite LACMNH-6800 218Halloysite WAR-5102 granularAlunite (alunite.2)ObsidianCrystalline thomsonite
Actinolite HS-116.4BHalloysite WAR-5102 solidAragonite CaCO3(aragonite.1)PumiceCrystalline heulandite
Albite WAR-0235Hectorite SCHa-1 powderBarite BaSO4(barite.1)Rock AnhydriteCrystalline stilbite
Almandine BUR-120AHectorite SCHa-1 solidCerussite PbCO3(cerussite.1)Rock GypsumAnalcime tuff
Andalusite WAR-0482Hedenbergite NMNH-16168Chabazite (chabazite.1)ScoriaMordenite tuff
Andesine BUR-240Hematite BUR-2600Goethite (goethite.1) Chabazite tuff
Anhydrite ML-S9Hornblende BUR-2660Heulandite (heulan.1) Erionite tuff
Anorthite BUR-340Hypersthene NMNH-B18247Illite/smectite (illsmec.1) Ferrierite tuff
Anorthoclase WAR-0579Illite IMt-2 granularLeucite (leucite.1) Clinoptilolite tuff
Anthophyllite BUR-4760Ilmenite WAR-4119Lizardite (lizard.1) Phillipsite tuff
Antigorite NMNH-47108Jadeite WAR-9909Mordenite (morden.1)  
Augite NMNH-119197Kaolinite KGa-1b granularPyrophyllite (pyroph.1)  
Biotite BUR-840Kaolinite KGa-1b solidPyrrhotite (pyroph.1)  
Bronzite BUR-1920Kutnahorite C43Pyrite (pyrite.1)  
Bytownite WAR-1384Kyanite (tan) WAR-1002bRutile (rutile.1)  
Ca-montmorillonite STx-1 granularLabradorite BUR-3080ASanidine (sanidine.1)  
Ca-montmorillonite STx-1 solidMagnetite WAR-0384Sepiolite (sepiolite.1)  
Calcite C22Malachite C24Spessartine (spessart.2)  
Calcite ML-C10Microcline BUR-3460Vermiculite (vermicul.1)  
Chalk C6Montmorillonite SCa-3 powderVesuvianite (vesuvian.1)  
Chlorite WAR-1924Muscovite WAR-5474   
Clinochlore BUR-1340Na-montmorillonite SWy-2 granular   
Dickite WAR-5101 powderNa-montmorillonite SWy-2 solid   
Dioperationside NMNH-107497Nontronite WAR-5108 granular   
Dolomite C17Nontronite WAR-5108 solid   
Dolomite C20Oligoclase BUR-060   
Enstatite DSM-ENS01Palygorskite PF1-1 granular   
Enstatite NMNH-38833Palygorskite PF1-1 solid   
Epidote BUR-1940Perthite WAR-5802   
Fayalite WAR-RGFAY01Phlogopite HS-23.3B   
Fe-smectite SWa-1 granularQuartz BUR-4120   
Fe-smectite SWa-1 solidSaponite ASU-SAP01 granular   
Ferrohornblende HS-326.4BSerpentine BUR-1690   
Flourite BUR-2080CSiderite C50   
Forsterite BUR-3720ATalc BUR-4640C   
Glaucophane WAR-0219Tremolite var. jade WAR-0979   
Gypsum ML-S5    

[12] The deconvolution algorithm used requires the number of spectra used in the input library to equal or be less than the number of bands in the data set. This significantly limited the set of library spectra that were used in any single analysis (maximum of 5 for ASTER TIR data), so multiple deconvolutions with different sets of library spectra were performed to expand the number of end members that were tested. A blackbody end member (emissivity = 1 at all wavelengths) was used in some deconvolutions to try to account for variations in absorption band depth between the ASTER-derived spectra and the library spectra. Decreasing particle size causes absorption band depths to shallow, and can cause band centers to shift wavelengths [Moersch and Christensen, 1995; Mustard and Hays, 1997; Cooper and Mustard, 1999]. The latter effect is not accounted for by addition of a blackbody spectrum in deconvolution, but requires extremely small particle sizes (on the order of the observing wavelength or smaller).

[13] Mineral/rock end member spectra identified from deconvolutions of spectral end members were then used (with a blackbody spectrum) to linearly deconvolve the original data. Because of the limitations of linear deconvolution, no more than 4 end members could be used. The result is a series of grayscale images showing the concentration of each mineral/rock end member, plus additional bands showing the blackbody concentration needed to match library spectra and the root mean square (RMS) error between the ASTER spectrum and the deconvolution model. The blackbody and RMS values do not directly describe surface properties, but these two bands did provide information regarding the surface and are discussed in the results section. The final result for each field site is a color “end member image”, which was created by displaying the concentrations of three selected end members in RGB. Concentrations of a fourth end member were displayed as a white overlay when necessary. Variations in color within the image indicate variations in potential mineralogies at each field site. Although the end member list (see Table 1) is quite specific in terms of composition, the limited spectral resolution of the ASTER TIR bands is not sufficient to make detailed mineralogic descriptions. Instead, each end member is considered to represent a group of rocks or minerals, so a generalized term such as “clays” (referring to most phyllosilicate minerals) was used to describe those end members.

[14] The ASU-TES mineral library used here contains few sulfate end members (anhydrite and varieties of gypsum): additional sulfate spectra have been added to this library [Lane, 2007] but were not available when this analysis was completed. Although conversion to emissivity of the JHU spectral library mineral spectra, which are biconic reflectance data, is not appropriate for quantitative spectral analysis [Salisbury et al., 1991], post-operations examination of the shape of the converted spectra shows that they are qualitatively consistent with laboratory emissivity spectra of the same minerals acquired by Lane [2007]. To confirm that the inclusion of converted JHU end members did not affect results, field TIR data were re-analyzed with the JHU end members replaced with laboratory emissivity spectra of the same minerals plus additional sulfate spectra [Lane, 2007]. The results from these analyses were consistent with those presented here.

2.4. Orbital Data Interpretation

2.4.1. Site A

[15] Site A is located along the western edge of Salar Grande, a large salt deposit surrounded by mountain ranges. A subset of the Hyperion image over this field area is shown in Figure 1a, with the associated MNF image in Figure 2a. Atmospheric correction was completed using a single ground truth spectrum with a known location on the image. Color variations in the MNF image suggests a variety of compositions and were used to identify regions of interest (ROIs). The mean spectra from these ROIs taken from the original data are shown in Figure 3. The relative similarity of these spectra suggest a homogeneous composition consisting of hydrated minerals, probably clays. The only significantly different spectrum (ROI #1) has a large absorption near 0.9 μm, which is consistent with the presence of iron.

image

Figure 1. Grayscale reflectance images of all six field sites. Images for sites B–F include the simulated landing ellipse (where possible, in white), rover traverse (black), and locale numbers for locations discussed in the rover data section. North is to the top of all images. (a) Grayscale version of Hyperion true color image for Site A, including the rover traverse (in white, beginning at the northernmost point on that line) and ROIs identified using the MNF image in Figure 2a. Spectra from these ROIs are shown in Figure 3. Salar Grande is the light colored flat unit that makes up most of the scene. (b) IKONOS satellite image (NIR band) of Site B. (c) ASTER Band 3N (NIR) image of Site C. (d) IKONOS NIR image of Site D. Clouds are visible at the very western edge of the image. (e) ASTER Band 3N image of Site E. The high albedo area discussed in the text is located in the northeast corner of the image, and is crossed by the northern part of the rover traverse. (f) ASTER images (Band 3N) of Site F. The location of Site C operations is just south of the westernmost end of the traverse.

Download figure to PowerPoint

image

Figure 1. (continued)

Download figure to PowerPoint

image

Figure 2. Color images used to examine regional composition. Images for Sites B–F are ASTER TIR-derived end member images with the simulated landing ellipses (where possible, in white) and rover traverses (black) overlain. North is to the top of all images. (a) Forward MNF transform of Hyperion image of Site A. Color variations in this image were used to identify ROIs (see Figure 1a) whose mean spectra appear in Figure 3. (b) End member image for Site B. Carbonate percentage (0–23%) is shown in red, the felsic volcanic end member (38–92%) in green, and the intermediate volcanic end member (4–53%) in blue. (c) End member image for Site C. Quartz percentage (0–20%) is shown in red, talc (35–85%) in blue, and albite (15–66%) in green; areas of significant gypsum concentration (20–40%) are shown as white crosses, and are limited to the southwestern edge of the simulated landing ellipse. (d) Site D end member image. Particulate sulfate end member percentage (3–18%) is shown in red, volcanic tuff (zeolite) percentage (3–78%) shown in green, and clay end member percentage (15–95%) shown in blue. (e) Site E end member image. Particulate sulfate end member percentage (7–42%) is shown in red, quartz (3–8%) in green, and volcanic end member (tuff + scoria; 51–100%) in blue. The strong blue signature in the upper right (overlying the high albedo area in Figure 1e) correlates with a high RMS error, which suggests that the end members selected here do not match that area well. (f) Site F end member image. The particulate sulfates end member (5–57%) is displayed in red, quartz (4–13%) in green, and clays (36–94%) in blue.

Download figure to PowerPoint

image

Figure 2. (continued)

Download figure to PowerPoint

image

Figure 3. Mean reflectance spectra from numbered ROIs drawn in Figure 1a, offset for clarity (offsets noted in the figure key).

Download figure to PowerPoint

2.4.2. Site B

[16] Site B is located in a valley just east of the western coastal range and west of Salar Grande (see Figure 1b). The slopes coming off the bounding mountain ranges are dissected, and the floor of the valley appears to partly consist of a bajada. VNIR false color imagery suggested a concentration of chlorophyll on the western side of the coastal range. Unfortunately this area was too steep for the rover, so the identification of chlorophyll in the orbital imagery could not be confirmed by the rover.

[17] Patchy deposits in the false color IKONOS image, possibly representing a reddish deposit, appeared to be associated with drainages. These patches were not visible in the ASTER false color image even though the features were several ASTER pixels in size and the image wavelengths are similar. However the ASTER data (acquired 11/17/2000) was taken at a different time from the IKONOS data (acquired 9/13/2002), and these patches may represent temporal variations in the surface, possibly transient deposits of reddish iron oxide. Because the field test took place in a similar time of year as the IKONOS data were acquired (Site B field test 9/10–18/2005), these patchy deposits presented an interesting target, and rover results are discussed in a subsequent section.

[18] Compositional end members initially identified at Site B via the TIR-derived end member image (Figure 2b) included volcanic rocks of both felsic and intermediate compositions and carbonates (specifically limestone). Areas of carbonate concentration were considered targets of interest, but no rover spectral data were collected in these areas. The identified volcanic rocks are consistent with the spectral identifications made from the rover; rover results (discussed in section 3.2.2), however, are more consistent with altered volcanic rocks and intermediate volcanic rocks rather than the originally identified felsic compositions. The potential iron oxide patches identified in IKONOS VNIR imagery could not be correlated with ASTER TIR results, as the TIR data were acquired at the same time as the ASTER VNIR image where the patches were not observed and iron oxide TIR spectra convolved to ASTER band passes lack absorption features.

2.4.3. Site C

[19] The field location for Site C (see Figure 1c) is located in the interior of the Atacama, and contains clusters of hill and drainages, some of which appear to be incised. No areas of chlorophyll were identified in the Site C ASTER false color image.

[20] Mineral end members identified originally at Site C (Figure 2c) were talc, albite, quartz, and gypsum. The area traversed by the rover does not appear to be very compositionally diverse, but a confluence of various sediment sources is visible in the end member image as a mottled purple color near the southeast edge of the simulated landing ellipse.

[21] The gypsum end member was present only in localized areas in the TIR data and the rover traverse did not pass through any of these regions. Areas of high quartz concentration were identified as targets of interest by the remote science team, and one such area was targeted and reached during the course of the traverse. The talc end member likely represents a variety of clay minerals that mimic the talc spectrum at ASTER spectral resolution. An albite spectrum, when deconvolved to ASTER band passes, has a relatively shallow absorption in Band 11 (8.475–8.825 μm), which is similar to the strong absorption seen in sulfates near 8.5 μm at this spectral resolution (see Figure 4). Based on rover examination of the field site, it appears that the “albite” end member more likely represents a particulate sulfate: the 8.5 μm absorption present in this end member is the result of a shallowing of the strong sulfate absorption by volume scattering effects. This particular end member appears to be concentrated along channels and drainages, many of which have light albedos in visible imagery, which would be more consistent with the particulate sulfate interpretation.

image

Figure 4. Spectra of albite, gypsum, and anhydrite from the ASU-TES end member library deconvolved to ASTER band passes. The similarity of these spectra may have led to confusion between albite and sulfate as an end member at Site C, with the albite spectrum more closely matching the natural surface spectrum due to shallowing of the sulfate absorption feature due to small particle size.

Download figure to PowerPoint

2.4.4. Site D

[22] The field area for Site D is located in a valley near a break in the western coastal range (see Figure 1d); the rover traverse begins in a smaller side valley and extends into a larger valley that contains basin infill. Several alluvial fans appear to overlie a high albedo unit which is exposed in the southern portion of the valley. No areas of chlorophyll were identified in either IKONOS or ASTER images of the field site. Although the site is close enough to Site B that portions of both sites appear in a single ASTER image, the chlorophyll signatures observed near the coast at Site B do not extend to the western edge of Site D.

[23] The TIR end member image (Figure 2d) is dominated by a clay end member, with minor concentrations of particulate sulfate (as suggested by an albite end member, see previous section about Site C) and zeolites (interpreted as a volcanic tuff end member). The majority of the rover traverse plotted in the figure overlies predicted clay-rich terrains, but a few pixels that contained a significant sulfate concentration were targeted by the traverse.

[24] An interesting side product of the TIR linear deconvolution is the amount of the blackbody spectrum modeled for each pixel. Because the blackbody spectrum is added to account for particle size differences between the laboratory and natural spectra, this can be considered as a proxy for particle size. In this case, however, the largest amount of blackbody was used over Salar Grande (see Figure 5): this particular feature is a large evaporite deposit (mostly halite, with associated sulfates) covered by a patchy coating of soil (probably clay-rich). In that particular location, the shallowing of the natural spectrum is probably not entirely due to particle size, but to the lack of absorption features in the halite spectrum in the TIR. This particular result is interesting in comparison with features discussed below at Site E.

image

Figure 5. Blackbody percentages used in linear deconvolution of Site D end member map (Figure 2d). The large bright area in north center of the image is Salar Grande (near Site A), which suggests that the high blackbody percentage in that area is indicative of halite. The rover traverse is overlain as a white line. North is to the top of the image.

Download figure to PowerPoint

2.4.5. Site E

[25] Site E is located in an area of mountainous terrain (see Figure 1e) with an area of dissected terrain just to the south and what appeared to be a flat plain to the east. The rover traverse begins near the top of an alluvial fan, then proceeds eastward onto the plain and north to an interesting high albedo feature.

[26] False color ASTER images of the Site E test area revealed no chlorophyll, but contained interesting albedo features (see Figure 1e). First, a large, topographically low, very flat, and high albedo feature (unit pvb) near the northern part of the image was identified as a potential area of water ponding, and was initially interpreted as a salar (or salt flat) by the remote science team. Second, another relatively bright unit (pmb) was identified just outside the southern margin of the simulated landing ellipse as another potential area of moisture influence. Finally, a set of high albedo streaks trending east-west across the flat plain were identified near the southern portion of this image. These light streaks widen to the east, and appear to indicate an area of wind transport, which would have the potential to carry airborne moisture (fog) into the field area from the west. Although the distance to the coast (assumed to be located west of the field area) was not known during operations, clouds were observed in the western edge of the ASTER image outside the simulated landing ellipse.

[27] End member mineralogy maps at Site E (Figure 2e) suggested a mixture of volcanic compositions (scoria and a zeolite tuff), plus quartz and particulate sulfates (again, based on alkalic feldspar end members interpreted here as sulfates). The rover traverse begins in a fairly mixed terrain and ends in a bright blue patch. This patch correlates with the unit pvb observed in the VNIR image. The bright blue result in the end member image also correlates with a high RMS error in the linear deconvolution: in this case, the deconvolution algorithm had rejected all but one potential library spectrum, which resulted in the assignment of 100% to the single positive result. This is an indication that the spectral library used in the deconvolution does not match the surface spectrum well: in the case of pvb, the absorption is quite shallow and is not matched well by any spectral library end members, even including the shallowing effects of a blackbody (Figure 6). Another similar bright blue patch is associated with unit pmb. The mean spectrum of unit pmb has a different shape than the one acquired from pvb, but also has a shallow band depth and does not match library end members well. Both units were identified as targets of interest not only due to the potential water contribution, but also because of the ambiguity of the composition results.

image

Figure 6. Example spectra from ASTER TIR data over Site E. The solid spectrum corresponds with the high albedo, high RMS area discussed in the text (unit pvb), which has a shallower absorption than the two nearby plains spectra (dotted and dashed spectra). The dash-dot spectrum corresponds to another albedo unit defined as an area of interest based on poor linear deconvolution fits to orbital data (unit pmb).

Download figure to PowerPoint

2.4.6. Site F

[28] Site F is located northeast of Site C in an area of clustered hills, topographic basins, and drainages (see Figure 1f). False color ASTER images indicated a small area of chlorophyll concentration, which was targeted and reached during operations. Various high albedo regions were also identified as targets of interest.

[29] The same ASTER data sets used for Site C cover Site F, but the TIR data were completely reprocessed using insights from previous field sites. An updated end member image covering the Site F traverse appears in Figure 2f; Site C is located just south of the westernmost point of the Site F traverse. In this image, the particulate sulfate end member is again visible throughout the image, although apparent concentrations appear along horizontal lines in the image due to noise. The other end members (clays and quartz) suggest the presence of altered igneous rocks.

3. Ground-Based Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[30] The rover used in the 2004 and 2005 field tests (sites B–F), named Zoë, has onboard a VNIR spectrometer, and is complemented by a human-operated thermal emission spectrometer [see also Cabrol et al., 2007]. At the final test site (F), the field equipment also included a human-operated neutron detector. The rover used during the 2003 field season, Hyperion, did not have an on-board spectrometer, but was followed by a human-operated VNIR spectrometer similar to the rover-mounted model on Zoë.

[31] Targeting of the rover-mounted spectrometer was accomplished by specifying an azimuth angle (in the rover reference frame) and an elevation for pointing of the pan/tilt head (0° defined as horizontal and −90° as straight down). This pointing was mimicked by the field team when using the human-operated spectrometers, with the instrument fixed at the point where the rover had stopped. Two types of strategies were used to specify spectrometer targets: first, a “blind” strategy where the target (either a general locale or a specific target such as a plowed surface) had not been imaged previously and was targeted based solely on azimuth/elevation, and a second “directed observation” strategy where the azimuth/elevation coordinates of specific targets of interest were determined from previously acquired rover images. The neutron detector observes an area underneath the instrument and is not targeted: instead, the field team reproduced a requested rover transect with the neutron instrument mounted on a wagon.

3.1. Instrumentation

3.1.1. Visible/Near-Infrared Spectrometer

[32] An Analytical Spectral Devices FieldSpec FR spectrometer mounted on the rover receives data from a fiber optic mounted on the pan/tilt head [Cabrol et al., 2007]. The optic has a field of view of a few centimeters at a distance equal to the height of the rover mast (1.8 m). This spectrometer collects light in 2151 wavelength bands with a range of 0.35–2.5 μm. Each group of spectral acquisitions includes a dark current spectrum and the spectrum of a white reference standard. The acquired data can be converted to reflectance by the simple formula Reflectance = (Target Spectrum – Dark Current)/(White Spectrum – Dark Current). General pointing of the spectrometer optics was performed by moving the pan/tilt head. Specific targeting was difficult as the parallax error between the pointing of the spectrometer and the pointing of the pan/tilt head was not well-characterized and it was logistically impractical to correct this issue during operations. Multiple spectral targets could be requested in a “pan” mode, where the pan/tilt head moved in a series of steps in pan and tilt to create a “raster” image. Ideally, this would be used to identify areas where the composition of a target of interest is different from its surroundings. Specific compositions were identified by comparison of the location and shape of absorption bands present in the field spectrum with those of laboratory samples available in VNIR spectral libraries [Grove et al., 1992; Clark et al., 1993]. Absolute depths of absorption bands were not considered diagnostic, as both hydration state and particle size can affect the depth.

3.1.2. Thermal Infrared Emission Spectrometer

[33] The TIR spectrometer (Design and Prototypes, Ltd Model 101) used at Sites B–F has an effective wavelength range of 7.5–13.0 μm (1330–770 cm−1). It was impractical to mount this instrument on the rover due to the need for multiple calibration targets and to cool the detector with liquid nitrogen. Instead, the instrument was carried and operated by the field team: the spectrometer was mounted on a tripod at the approximate location of the rover with a foreoptic at the same height as the rover's pan/tilt head. Blind targets for the TIR spectrometer were requested in a similar manner to those for the VNIR spectrometer via azimuth and elevation angles forwarded to the spectrometer operators. In addition, because the instrument was human-operated, specific targets on interesting rocks and/or surfaces identified in visible images could be reliably acquired. Because the spectrometer was fixed at the approximate height of the rover mast, the field of view of the instrument varied as it would if the spectrometer were mounted on the rover. Most observed rock targets did not fill the entire field of view (minimum diameter ∼15cm). Other limitations to the TIR targeting were the time required to acquire calibration spectra at any given locale: spectra of a “cool” blackbody, a warm blackbody, and a diffusive reflecting metal foil target at a known temperature are required for temperature/emissivity separation and atmospheric removal. The instrument also took a significant amount of time to set up, align with the rover direction, and break down again. This limited the number of TIR spectra that could be acquired in any one day.

[34] Reduction of the TIR data to surface emissivity consisted of converting the spectrometer measurements to calibrated radiance through use of the two blackbody spectra, then removal of the atmospheric component through the use of the foil standard spectrum and polynomial fitting to remove random noise, and finally separation of the temperature and emissivity components [Horton et al., 1998]. As random noise was present in all of the TIR spectra collected, they were subjected to a smoothing filter to remove small scale noise before analysis; extremely noisy TIR spectra were excluded.

[35] Resulting emissivity spectra were examined first by eye to identify distinctive absorption features and were then analyzed using linear deconvolution [Ramsey and Christensen, 1998]. The end members used for deconvolution of field TIR spectra were identical to those used to analyze ASTER data (see Table 1), resampled from laboratory wavelengths to those of the field spectrometer.

3.1.3. Neutron Detector

[36] The prototype neutron detector developed for this study had undergone several preliminary control tests and models were generated to interpret counts returned from the field [Moersch and Drake, 2003; Hardgrove et al., 2006; Cabrol et al., 2007]. The test at the final LITA site (F) provided an opportunity to learn how neutron data taken in the field can be used in a mission-like scenario to provide a more complete understanding of the planet's surface and shallow subsurface. This marks the first use of a neutron detector as part of a “science-blind” field trial with a suite of other instruments.

[37] Neutron detector data collection was requested by the remote science team via the same interface as all other instruments: a starting and ending locale on a transect were identified in the plan sent to the rover in the field. Actual collection was accomplished with both the isotopic fast neutron source and the neutron detector mounted on a wagon which was pulled along the requested rover transect by the field team, stopping at pre-defined intervals to allow the collectors to integrate and acquire data. The neutron instrument had two detector tubes: a cadmium jacketed tube counted only incident neutrons with high enough energies to penetrate the jacket, and a bare tube that counted any incident neutron. These neutrons are sourced from an area of approximately a meter or less, depending on the composition of and amount of hydrogen in the near-subsurface. The ratio of the counts from each tube provides a measure of the lowest energy neutrons, which have been slowed by collisions with hydrogen.

[38] The relationship between subsurface hydrogen content and neutron detector counts is modeled using the Monte Carlo Neutral Particle eXtended code (MCNPX), which was used to simulate the neutron detector's response to increasingly hydrated environments [Pelowitz, 2005]. The simulated count ratios from the model are used to form a look-up table that converts the observed detector count ratios to a percentage of hydrogen (by weight) in the near-subsurface.

3.1.4. Additional Rover Instrumentation

[39] Additional rover instruments included a stereo panoramic imager (SPI), navigational cameras (navcam), and fluorescence imager (FI). The SPI consisted of three cameras mounted on the rover pan/tilt head: the foreoptic for the VNIR spectrometer is mounted near the rightmost of the SPI cameras. SPI images (both color and grayscale) were used to examine the region around the rover via 360° panoramic image sets. Navcam, a set of stereo cameras used for hazard detection by the rover, was used to take periodic grayscale images along a transect to record areas where the rover did not pause to acquire data.

[40] The FI used a flash lamp to briefly illuminate a 10 cm square area underneath the rover and a high sensitivity camera to image that area. Filter wheels were used to limit the incident and measured light to specific band passes, which were chosen specifically to correspond to wavelengths where biologic materials are known to fluoresce. The FI was capable of detecting chlorophyll, proteins, lipids, and DNA. In addition, the FI camera was also used to acquire RGB images of the surface material. Additional information about these instruments and further discussion of rover capabilities can be found in other publications discussing this study [e.g., Cabrol et al., 2007] (A. Hock et al., Life in the Atacama: A scoring system for mapping habitability and the robotic exploration for life, submitted to Journal of Geophysical Research, 2007; S. Weinstein et al., Application of pulsed-excitation fluorescence imager for daylight detection of sparse life in tests in the Atacama Desert, submitted to Journal of Geophysical Research, 2007).

3.2. Ground Data Interpretation

[41] Locale-specific results based solely on rover spectroscopy for Sites A–F are given in Table 2 and example spectra from each site are shown in Figure 7. All spectra are single field acquisitions. For each locale, the number of spectral targets acquired is listed, as well as the general results for the mineralogy. When specific targets have interesting results (e.g., comparisons of plowed versus unplowed surfaces), they are discussed separately. Rover locales where spectral data were not collected are omitted from the tables.

image

Figure 7. Example spectra from each field site: all spectra are single field acquisitions. VNIR results are plotted in the left graph: gaps occur where atmospheric contributions have been removed. Smoothed TIR spectra are shown in the right column. Most spectra are offset (as noted in y-axis label): offset values are shown in the figure key. Locales for sites B–F are labeled on the corresponding image in Figure 1. (a) Example VNIR spectra acquired during rover operations at Site A, offset for clarity. The Sol 1 spectrum is dominated by calcite with minor sulfate. The Sol 2 and Sol 4 spectra represent an anhydrite; the Sol 2 spectrum shows evidence of iron oxide. The Sol 3 spectrum appears to be desert varnish. The Sol 5 spectrum is relatively bland, but has some shallow absorptions that are similar to those in iron-bearing clay minerals. (b) Example spectra from rover operations at Site B listed by locale and azimuth/elevation angles (in degrees) of target. All VNIR four spectra are consistent with a phyllosilicate mineralogy, specifically palygorskite for Locales 3 and 4, and hectorite for Locale 11. Derived compositions for the TIR spectra based on linear deconvolution were: Locale 3 (0, −20) - hematite, clay, feldspar; Locale 3 (180, −20) - anhydrite, clay, feldspar; Locale 4 - clay, minor goethite; and Locale 11 - zeolite and clay. (c) Example spectra from rover operations at Site C. Three VNIR spectra are shown: the Locale 25 spectrum is broadly consistent with clay, Locale 34 with kaolinite clay (doublet near 1.4 μm), and Locale 26 with a desert varnish. Linear deconvolution of TIR spectra suggests: Locale 40: Dark rock - clay, obsidian; Locale 40, light soil: sulfate; Locale 36 - clays and zeolites with minor sulfates; and Locale 34 - quartz, sanidine (feldspar), mica (phlogopite). The Locale 34 TIR spectrum shows evidence of the quartz doublet located near 9 μm. (d) Example spectra from rover operations at Site D. The VNIR spectra all have relatively low contrast: the results from Locales 010, 050, and 210 are most consistent with clay (possibly illite). The spectrum from Locale 060 contains absorption features consistent with gypsum/anhydrite; the spectrum from Locale 050 may include a minor contribution from sulfate as well. TIR spectra (right plot) from Locales 110 and 200 are consistent with clays/zeolites. A series of spectra acquired at Locale 050 are mostly consistent with alteration materials (clays and zeolites), with sulfates present only in the unplowed soil. (e) Example spectra from rover operations at Site E. All four VNIR spectra shown here are consistent with clay minerals of varying compositions and hydration states. TIR spectra are evidence for sulfates, via an absorption feature near 8.5 μm in the soil and subsurface (Locale 330, light soil and hole). Spectra of rock targets are more consistent with alteration materials (Locale 330 - dark gravel, Locales 360 and 440). (f) Example spectra from rover operations at Site F. VNIR targets include a plant (Locale 880) with the expected chlorophyll reflectivity increase near 0.7 μm, and a spectrum from a group of white rocks (Locale 701) that indicates calcite and chert/chalcedony. The remaining spectra (Locales 880, 810, and 730) show shallow absorption features consistent with iron oxides and clays. The TIR spectra contain 2 rock targets (Locales 730 and 880) with an odd feature near 10.76 μm that is attributed to uranium ions in iron oxides. Linear deconvolution results include varying amounts of clays, obsidian, feldspars, and iron oxide.

Download figure to PowerPoint

image

Figure 7. (continued)

Download figure to PowerPoint

Table 2. Rover Spectroscopy Results for All LITA Field Sitesa
 LocaleNumber of VNIRNumber of TIRClaysZeolitesSulfatesFe-OxideSilicaOther
  • a

    Data are grouped by site and locale (listed by sol for Site A). The number of spectra from each spectrometer are noted. For each column, ‘All’ indicates the composition was present in all spectra, ‘Y’ indicates some spectra suggest that composition, and ‘N’ indicates it was not present. Dashes in the silica column indicate that no TIR spectra were collected, as quartz does not have diagnostic absorption bands in the VNIR unless it is hydrated. The' Other' column contains additional compositions identified at that locale. Question marks indicate ambiguous results.

Site A
 Sol 130NNV. minor anhydriteY 
 Sol 120NNAnhydriteY 
 Sol 230NNAnhydriteN 
 Sol 220NNAnhydriteY 
 Sol 310NNAnhydriteY 
 Sol 310Fe-bearingNNY? 
 Sol 320NNNNDesert varnish
 Sol 420NNAnhydriteN 
 Sol 420NNNNDesert varnish
 Sol 420NNAnhydriteY 
 Sol 420NNNNDesert varnish
 Sol 520Dehydrated illite/smectiteNNN 
 Sol 510NNNFine-grainedDesert varnish?
 Sol 510Fe-bearing?NNN 
Site B
 144AllYMinorYNFeldspar, pyroxene
 244AllYMinorMinorN 
 344AllNYYNFeldspar, pyroxene
 441AllNNMinorN 
 541AllNNYN 
 601AllYNYNFeldspar, pyroxene
 740AllNNN 
 844AllYYYNFeldspar, pyroxene
 1001AllNNNNFeldspar, pyroxene
 1144AllYNNNFeldspar, pyroxene, pyrite
 1444AllYNYNFeldspar, pyroxene
 1941AllNNYN 
 2044AllYNMinorNFeldspar, pyroxene
Site C
 2548AllYYMinorObsidian 
 2641AllNNYMinor obsidianDesert varnish
 2903All (minor)AllNNN 
 3087AllYin soilMinorObsidian 
 34810AllYNYQuartzFeldspar (sanidine)
 3603AllYMinorMinorN 
 3802AllNNNN 
 4022AllNYNObsidian 
 4101AllNNNObsidian 
Site D
 00021010AllNNNN 
 010360AllNNN 
 0401401AllNNNN 
 0501407AllY<20%, none post plow∼10%N 
 060360AllNNNN 
 09009Y?Y?Y?N?N? 
 1102020All (illite)NNN 
 12112418Y?N?Y?Y?N?Feldspar?, pyroxene?
 14009YNMinorNN 
 190360Y?N?N?N? 
 200369YNMinor?NN 
 210360All (illite)NNN 
 2401240AllNNN 
Site E
 33005YY<20%GoethiteObsidianFeldspar
 3601212Y (illite)NY, white rock onlyNNDesert varnish
 44066AllNNN 
 52030DehydratedNNN 
 53030DehydratedNNN 
Site F
 70140NNNCoating?Hydrated, Amorphous?Feldspar, calcite
 73033NNNYAmorphousDesert varnish
 8102424YYMinorCoating?Hydrated, AmorphousFeldspar, pyroxene, desert varnish
 88055YNNCoating?NChlorophyll
3.2.1. Site A

[42] At this field site, all VNIR spectra were acquired by the field team using an artificial light source with a spectrometer identical to the rover instrument used at later sites. A TIR spectrometer was not used. The orbital data (discussed earlier in section 2.4.1) suggest only a couple of different surface compositions (iron oxide and clay), but the field-based spectra show a more diverse surface (see Table 2 and Figure 7a). A majority of the returned spectra have absorption features that are consistent in position with those in gypsum or anhydrite. The absorption band depths are intermediate between the two end members, suggesting a partially hydrated state. A small fraction of the returned spectra are consistent with calcite, although the observed absorption bands are shallow and may indicate a small particle size. Another small fraction are consistent with iron-bearing clay minerals as predicted by the orbital data. Several of the returned spectra contain a broad absorption near 0.9 μm that is present in iron oxides. The overall picture from field-based spectra is an area dominated by evaporitic minerals and iron oxides, with a small component of clay. This result is quite different from the orbital view.

3.2.2. Site B

[43] For the field test at Site B (see Table 2 and Figure 7b), the time required to collect a single VNIR spectrum was approximately a minute due to hardware issues. For this reason, it was impractical to request a large number of spectra at any given locale, so VNIR spectral acquisitions were typically limited to a single “compass pan”: this is a series of four spectra taken at the cardinal points of the rover reference frame (0°, ±90°, 180°) with an elevation of −15 to −25°. These relatively shallow elevation values were necessary to avoid targeting the rover body and were modified when steeper initial values caused the solar panels at azimuth = 180° to be in the field of view.

[44] An additional targeting strategy involved a single spectrum taken directly in front of the rover (azimuth = 0°, elevation = −90°). Attempts were made at one locale to target rocks that appeared in the visual panorama, but due the aforementioned parallax issue it was impossible to determine if the rock targets had been observed. There is no indication based on absorption band depth or location that the rover spectrometer acquired any spectra of a target that was predominantly rock rather than soil.

[45] The VNIR spectra acquired at this site are relatively flat, with shallow absorption features indicative of bound water and a few identifications of iron-bearing minerals. A majority of these spectra are interpreted to represent a clay mineralogy. The overall blandness of the results, both in terms of spectral and spatial variation, may well be the result of the targeting strategy used rather than a homogeneous surface mineralogy. The VNIR spectra in general also have an atmospheric component that was not removed by calibration which increases in intensity as the atmospheric humidity appears to increase (coincident with an increase of clouds and fog visible in rover images). The atmospheric component was used as a proxy for humidity when weather data were not available.

[46] TIR results can be summarized as a surface that is dominated by alteration materials, mostly zeolites and clays, that are probably weathered from a primary volcanic source. Small amounts of pyroxene and feldspar are found in several locales, which likely represent the parent rock. Sulfates are present at a few locales, but not in large quantities. Quartz, identified as a potential habitat for hypolithic organisms [Warren-Rhodes et al., 2007c], was not found at Site B.

[47] The potential iron oxide patches identified as targets of interest on the IKONOS orbital image were visited early in the rover traverse. An example spectrum from this locale (3) is shown in Figure 7b, and the location labeled on Figure 1b. Both the VNIR and TIR spectra are most consistent with a clay mineralogy, but minor amounts of iron oxides are present in the TIR spectra. Although this is not a direct confirmation of the orbital data results, it was noted that the deposits appeared to be transient, which is consistent with the spectral signatures observed here.

3.2.3. Site C

[48] For the field test at Site C, the VNIR hardware problems had been resolved, and the time per spectrum reduced to a number of seconds. This allowed the science team more flexibility in targeting options, although the previous targeting strategies were retained. Attempts were made to create spectral “raster images”, but these were unsuccessful at this site. The remaining spectra were targeted in the manner described previously: either in a compass pan, or in a series of spectra taken at the foot of the rover (see results in Table 2 and Figure 7c).

[49] VNIR results are similar to those at Site B, indicating a surface that is dominated by alteration, mostly to clays. At Locale 34, the clay appears to be kaolinite based on the presence of a doublet in the spectrum near 1.4 μm. Desert varnish (library spectrum composition Na4Mn14O27•9H2O + αFe2O3 [Clark et al., 1993]) was identified in a spectrum from Locale 26.

[50] TIR results also suggest alteration materials, mostly zeolites and clay. Morphology of local rock targets more conclusively identifies a potential volcanic origin. Vesicular rocks were visible at the simulated landing site, and a volcanic tuff was identified at Locale 36 based on visual morphology in FI images. Linear deconvolutions of spectra from this locale are dominated by silicate alteration products with small amounts of sulfate (see Figure 7c). The presence of sulfate in several of the TIR spectra from this site can be noted not only with linear deconvolution, but by the presence of an absorption at 8.5 μm that is characteristic of these minerals. In a few cases, the absorption is weak and is not well-modeled by linear deconvolution results: this is likely due to differences in particle size causing the absorption to become shallower (and therefore less consistent with laboratory library spectra) possibly combined with low concentrations in the instrument field of view. Sulfate concentrations appear to be confined to the soils in many places: this is consistent with the identification of albite as an end member in orbital data.

[51] Silica was identified in TIR spectra at various locales in Site C, consistent with the identification of quartz as an end member in orbital data. As with sulfates, quartz has diagnostic absorption features, including a large absorption near 9 μm that commonly appears as a doublet in laboratory spectra. A similar feature is noted in some of the Site C field spectra, particularly those from Locale 34 (see Figure 7c) which was targeted based on increased quartz abundance in the end member image (Figure 2c). Silica is present in the form of obsidian at other locales.

3.2.4. Site D

[52] A majority of the VNIR spectra collected at Site D were compromised either by issues with target planning, calibration, changing sun angle, or weather. The spectra that could be analyzed are relatively low contrast. Nearly every VNIR spectra collected of a natural surface is consistent with a clay mineralogy, most likely illite: only a few spectra consistent with sulfates were observed (see Table 2 and Figure 7d).

[53] TIR spectra collected at Site D are also dominated by clays with some evidence of sulfates. In a few cases, the presence of an absorption feature at 8.5 μm indicates that these minerals are present, but are not modeled by linear deconvolution due to the shallow depth of the feature. Of the locales where sulfates were detected by linear deconvolution, locale 050 is interesting, as the sulfates appear in the spectrum of the surface, but were not present in detectable levels in the plowed surface where the top few millimeters of soil and rock have been removed. This was contrary to the expected relationship (soil overlying sulfate-rich soil), and suggests that this portion of the field area (a small valley leading into a larger valley where a majority of the rover traverse was located) may contain only a surficial coating of sulfates (possibly wind-blown).

[54] Reconstruction of the rover traverse based on GPS points collected by the field team suggests that the rover did not reach the concentration of sulfate identified in orbital data (near Locales 200 and 210 in Figures 1d and 2d). Examination of images taken from the FI suggests that area traversed by the rover near those yellow pixels is underlain by a saline crust that, in places, has heaved upward and is exposed at the surface [Warren-Rhodes et al., 2007b]. It is possible that the saline crust observed in these images is not sufficiently exposed at the surface in other areas in the valley and would not be detected by remote methods.

3.2.5. Site E

[55] A series of mechanical problems caused during transport of the rover from Site D to Site E rendered the VNIR spectrometer non-operable, but VNIR spectral acquisition at a few locales was accomplished via a human-mounted instrument using the same targets as the TIR spectrometer. As only a few targets were acquired, the results are presented simultaneously: VNIR results were collected at 4 locales, although TIR results were collected at only 3 locales due to additional hardware problems with that instrument (results in Table 2 and Figure 7e).

[56] Linear deconvolution of TIR results at the simulated landing site (Locale 330) suggested mostly zeolites (possibly volcanic tuff, as observed in the orbital end member image of this area), with minor amounts of feldspars and obsidian. A subsurface layer of white rock, exposed as a semi-resistant layer in a small pit, is likely sulfate based on spectral results: because the surrounding soil filled a majority of the field of view, only 20% sulfate was modeled by deconvolution.

[57] Results from a nearby locale (Locale 360) are slightly different: VNIR spectra are consistent with illite clay, which was also modeled in TIR deconvolutions. Desert varnish is suggested by the VNIR spectra as well (similar to the earlier identification at Site C). In this area, sulfates appear to be concentrated in the soils and not in the rocks. Typically, linear deconvolution results at earlier sites did not suggest a large concentration of sulfates (<25% in most cases), and VNIR spectra are more consistent with clay than with sulfate. If the sulfates are concentrated in soils rather than in rocks, then the majority of targeted spectra (which were usually aimed at rocks that filled most of the field of view) would likely not contain a great deal of sulfate. In blind targeted spectra, the spectral signatures of larger rocks and pebbles at the surface may be dominating the results over soils, which have smaller particle sizes and will be partially covered and shadowed by the rocks.

[58] The high albedo/high RMS “volcanic tuff” area discussed earlier (labeled as pvb on Figure 1e) was observed by the VNIR spectrometer along the southern margin of the unit (Locales 520 and 530). These spectra are consistent with an illite clay spectra collected in a small gully at an earlier locale (compare Locales 440 and 530 in Figure 7e), but with shallower absorption bands. Because these bands are associated with vibrations in water or hydroxyl molecules, this may represent dehydration of the clays. Camera images from this unit suggested that this misnamed salar was really a playa, consisting of dried mud that formed polygonal cracks. The mud should have a fine particle size, which would result in an extreme shallowing of characteristic absorption features in the spectrum [Moersch and Christensen, 1995; Mustard and Hays, 1997; Cooper and Mustard, 1999]; this is consistent with the results obtained from the orbital end member images, and another possible explanation for the shallow absorption features observed in the VNIR field data. Unit pmb, also identified as a target of interest due to ambiguous orbital composition results, was not reached by the rover traverse.

3.2.6. Site F: Spectroscopy

[59] Mechanical problems from Site E persisted to Site F, and the VNIR spectrometer mounted on the rover had been removed entirely and was not used during operations. This data set was again collected by the field team using a human-operated instrument, which limited the amount of spectra that could be acquired. In addition, local weather conditions (high winds) caused havoc with field equipment, which also limited the results. In total, spectral data were acquired at four locales (see Table 2 and Figure 7f).

[60] The majority of rock spectra collected at the four locales were consistent with volcanic rocks (suggested by the presence of feldspars) with alteration coatings: these alteration rinds were composed of clays, iron oxides, and/or desert varnish. In addition to these altered volcanic rocks, hydrated amorphous quartz (chert or chalcedony) was identified at all four locales, and a group of white rocks at the simulated landing site (Locale 701) were spectrally consistent with calcite. In two locales, TIR spectra suggesting iron oxides contained an odd absorption feature centered at 10.76 μm (929 cm−1). This feature did not appear in any available library spectra, but is consistent with the position of an absorption that occurs in transmission spectra of hematite when uranium ions substitute for iron ions in the mineral structure [Duff et al., 2002]. This would not be a surprising result in this particular area: a nearby mine was clearly visible in the orbital images which was confirmed as a working copper mine during ground truth. The remaining spectrum at Site F was a deliberate target on a plant observed in rover images, which exhibited the expected chlorophyll signature.

[61] A large concentration of chlorophyll observed in the ASTER false color image was identified as a target of interest, and was eventually reached by the rover during its traverse. This area turned out to be a large grouping of plants surrounding an old tailings pond. The large concentration of plants at this locale, as well as earlier detections of plant life that were not directly visible in orbital images confirms that the method used to identify potential chlorophyll concentrations does require a significant concentration of surface vegetation.

3.2.7. Site F: Neutron Detector

[62] Site F was the only location where the neutron detector was deployed during rover operations; due to equipment issues in the field, data collection was limited to a single transect (between locales 880–890). Neutron detector integrations took place at 200 m intervals, resulting in seventeen measurement locations along a total transect of approximately 3,200 m. The transect follows the rover's path around obstacles, which may not be along the straight line between the two locales. The percent hydrogen abundance for each point along the transect was derived from measured count rates and the MCNPX model assuming a homogeneous soil component. Figure 8 shows the percent hydrogen for the transect on a sub-frame of the orbital ASTER image for the region. This image shows that higher count ratios (>5.5%) were observed in locations where drainage channels are present, specifically at N03, N05, N06 and N09.

image

Figure 8. ASTER greyscale image (Band 3N) with associated plot of hydrogen abundance values versus distance (m) along the neutron transect. North is to the left side of the ASTER image, which is rotated to match the plot. Locations of neutron integrations are noted on the image, and correspond to data points on the plot.

Download figure to PowerPoint

[63] Several apparently anomalous values are present, specifically at N02 (400 m), N08 (1,400 m) and N10 (1,800 m). Count ratios acquired at N02 indicate a lower hydrogen abundance for this location, although the ASTER image shows the data were acquired in the proximity of a drainage channel. At N08, count ratios revealed a high hydrogen abundance, although the ASTER image shows the data were acquired relatively distant from any apparent drainage channel. Other than navcam imagery taken along the transect, there is no additional data available to aid in characterizing the count ratios observed at these locations. Navcam images near N02 and N08 show low and high quantities of vegetation, respectively, which suggest that subsurface hydrogen abundance may not be well correlated with drainage channels visible in the ASTER image. N10 is very near the edge of a large drainage channel seen in the ASTER image, but the count ratio is quite low. Navcam images show increased levels of vegetation at N10, although a local decrease in clay minerals could also explain the observed decrease in count ratio.

[64] Several good correlations between ASTER, navcam and the neutron detector were seen along the transect. Vegetation was not observed in the navcam images taken near N04 (600 m) and N07 (1,200 m). These locations are not within drainage channels in the ASTER image and correlated well with the decreases in count ratio seen at N04 and N07. N09 shows increased hydrogen abundance, close proximity to a drainage channel in the ASTER image, as well as increased vegetation in the navcam images. TIR spectra at the starting locale (Locale 880, N01) suggest a surface mineralogy consisting of amorphous or fine-grained quartz, feldspars, and clays. VNIR soil spectra show evidence for clay minerals in the form of bound water absorption bands. Evidence for both water and clay minerals suggests bound-hydrogen is relatively abundant within the near-surface soil. Data from the FI indicates biologic activity is present at this location as well. As observed with the neutron detector, N01 corresponds to a hydrogen abundance of ∼4% by weight. Other than the exception at N10, hydrogen abundances greater than 4% are associated with an increased presence of vegetation in navcam images until approximately 2,600 m into the transect. At this point, vegetation levels returned to levels similar to those seen at N01. The FI measurement made at the ending locale (N17) returned no positive biologic signals. Although no vegetation was seen in navcam images, hydrogen abundance derived from count ratios reached a constant level of approximately 1% higher than those observed at the starting locale, which suggests an increase in clay mineralogy or an increase in the abundance of subsurface water.

4. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[65] The orbital spectral results provided a good background for interpretation of the local geology of the rover traverse paths. Although the exact compositions derived prior to the field tests did not prove to be completely accurate, the derived spectral end members were broadly consistent with compositions determined from rover spectroscopy, and in turn could be improved by re-interpretation based on results from rover data sets. This is evident in the ASTER data processed for Sites B and D, and Sites C and F: in both cases, results from the 2004 field season allowed for improved analysis of the same data sets for use during the 2005 field season. In most cases, the diversity of surface composition suggested by orbital analysis is less than observed on the ground, as might be expected. This is partially due to the limits of the data sets used: the coarse spectral resolution of the ASTER TIR data precludes identification of large numbers of spectral end members. This is most evident in these results, however, at Site A, where the hyperspectral Hyperion data suggested the compositional diversity might be limited to different species of clays and possibly iron oxide (see Figure 3). The spectra acquired as part of the rover test, however, identified three major surface components (anhydrite, calcite, and iron oxide) plus two more minor ones (clay, probably iron-bearing; and desert varnish). Regardless of the ground truth accuracy of the results, the orbital data sets were extremely useful for identifying areas of interest, both in terms of derived mineralogy (TIR) and geomorphology (VNIR).

[66] An interesting digression in the orbital data analysis is the comparison of the ASTER TIR linear deconvolutions from Sites D and E over high albedo features suspected to be formed by water (Salar Grande at Site D, and both the observed playa pvb and the unvisited unit pmb at Site E; see Figures 5 and 2e). In both cases, the spectral end members used in deconvolution did not match the surface spectra very well: in the Site D data, the spectra acquired over Salar Grande are best modeled by including a large percentage of a blackbody spectrum, although the spectra of both pvb and pmb at Site E are not well matched by any end member, even with a blackbody. The difference in the deconvolution results is likely related to differences in the composition. At Salar Grande, a large deposit of halite is covered by a thin, patchy layer of soil, so it would be reasonable to expect that a spectrum taken of this unit would resemble a clay spectrum plus a halite spectrum: in the TIR, halite has no absorption features, and acts like a blackbody. Thus the deconvolution results in an acceptable match between the model (clay + blackbody) and the surface spectrum. The playa at Site E (unit pvb), however, is likely composed of materials with a very fine particle size. When the particle size of a substance gets very small, the effects that cause the absorption features to shallow and shift behaves non-linearly, and cannot be simply modeled as the addition of a blackbody spectrum [Moersch and Christensen, 1995; Mustard and Hays, 1997; Cooper and Mustard, 1999]. In this case, although it is possible that the chosen end members are not sufficient to match this spectrum, it is likely that the very small particle size of the playa has introduced non-linear particle size effects into the spectrum that cannot be modeled by linear deconvolution. The result is a poor model fit. Although no rover results were acquired for unit pmb at Site E, this area was visited during ground truth expeditions and is discussed below.

[67] Rover spectral results (Figure 7) are broadly consistent throughout each field site between both instruments. Weather conditions affected both instruments, as increases in atmospheric contributions and random noise were observed with increasing humidity at both sites. Spectra taken without a specific target (i.e., those taken in “compass pans”) tended to be spectrally bland compared to results from spectra taken of specific targets. This is likely a result of small particle sizes affecting the depth of absorption features in soil samples. This observation underlines the importance of acquiring spectra by directly targeting large rocks, or in a “raster image” method, which would allow for identification of spectra that are different from the background soil. Another possible explanation for the general spectral blandness of the VNIR results is dehydration of the surface (reducing the depth of OH bands) and a lack of Fe2+ in minerals exposed at the surface. Although this would explain the lack of strong spectral features, there are minerals exposed at the surface that should have distinctive VNIR spectra (hematite and sulfates, for example), which suggests that particle size and targeting strategy may well have a significant effect. This is also illustrated by the spectra gathered at Site A (Figure 7a), as the ASD spectrometer used for that field test was human-operated and could accurately target rock samples that filled the field of view. These spectra were perhaps the most mineralogically diverse data set of all the field sites.

[68] The TIR results were more useful for deriving the mineralogy of the surfaces covered by the rover. In general, deconvolution results are consistent across multiple spectra at a given site, and at nearby locales. The results are broadly consistent with the orbital data sets, especially considering the differences in spatial/spectral resolutions. The zeolites and clay minerals identified are also consistent with the bound water absorption features observed in a majority of the VNIR spectra collected by the rover: a possible explanation for the non-identification of sulfates in the VNIR results is also found in the low percentages of sulfates used in linear deconvolution models of TIR results (<20%), the particulate nature of these sulfate deposits as observed in rover images, and the likelihood that these sulfates are quite dehydrated, thus reducing the depth of the bound water bands typically observed in sulfate spectra in the VNIR wavelength region.

[69] The overall picture of the mineralogy of the field sites derived both from the rover and orbital data sets is of a terrain that is dominated by a surface layer of clay alteration. This alteration layer appears to be mostly in the form of soil or dust, although it is likely that many of the observed rock targets could also have an alteration rind. The presence of a thin alteration layer is also suggested by a lack of correlation between the variations in mineralogy shown in end member images and geomorphologic maps derived from orbital images (e.g., discussion of Site D by Warren-Rhodes et al. [2007b]). The pre-altered source of many of the observed rocks appears to be volcanic, as evidenced by the observation of vesicular rocks and volcanic tuff in some locales (i.e., Site C), and the occurrence of primary volcanic minerals such as feldspar and pyroxene in linear deconvolution results from TIR spectra.

[70] Sulfates appear in small concentrations throughout the field sites, with the exception of Site A, and are sometimes present only in the soil at the surface (i.e., Site D, where sulfates appear in the pre-plowed surface at Locale 050 but not the spectrum taken after the rover plow had removed the surface soil) and sometimes below the surface soil (i.e., Site E, where a layer of sulfates appeared exposed in a pit at the simulated landing site). The large number of anhydrite targets observed at Site A is not surprising considering its location next to a large evaporite deposit, however.

[71] Another significant observation is the effect of targeting on the results. Simple blind targeting of the spectrometers based on azimuth/elevation locations tended to result in relatively uniform, low contrast spectra of ubiquitous soils rather than rock targets. Although linear deconvolution of TIR spectra is capable of separating different end members from a mixed spectrum, there is no spatial information in the result: the minerals identified by deconvolution cannot be specifically identified as rock or soil components unless deliberate targets on soils are acquired as well as targeted spectra on rockier surfaces. In the VNIR, this issue is observed in one extreme when spectra are targeted at elevation angles close to horizontal: when grazing the surface and covering a large field of view, the spectral results are often dominated by low contrast background materials. In the other extreme, the field of view of the VNIR spectrometer when pointing straight down in front of the rover is on the order of a few cm in size. In this case, it was not possible to accurately identify which particular portion of a camera image was observed by the spectrometer, and it was therefore impossible to correlate compositions identified in the VNIR spectra with FI biology results. The correlation of habitat and identified life would be stronger if such composition data were available.

[72] In the neutron detector data set, correlations were observed between derived hydrogen abundances and areas along the transect where more water would be expected. The transect crossed at least two large drainages (observed as areas of high albedo in the ASTER image shown in Figure 8); in both cases, the apparent hydrogen abundance increases near the drainages. There is also an increased amount of vegetation in the northern part of the traverse, although the vegetation and the hydrogen abundance does not correlate as well. This may well be due to factors other than subsurface hydrogen that are not observed here, however.

5. Ground Truth Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[73] The final workshop of the LITA project included day trips to each of the sites discussed here. In most locations the rover locales were still marked and the rover traverse could be followed (at some of the later sites the rover wheel tracks were still visible). This provided an opportunity to examine the field areas in the context of conclusions made from rover data sets; for further detail of ground truth, see Thomas et al. [2007].

[74] The general overview in the previous section is supported by these brief expeditions, as well as by geologic maps of the field sites [Carta Geologica de Chile, 1978, 1981; Escobar et al., 1980]: although subsurface layers of sulfate are quite apparent to anyone walking over the surface (causing them to sink as the sometimes friable saline layer collapses), the surface soils do not commonly appear to contain a great deal of sulfate and the surface rocks often appear to have igneous origins with later alteration. A number of igneous intrusives (mostly intermediate in composition) were observed in a drainage at Site B near the northwesternmost locale. The volcanic tuff identified by the field team in the later locales of Site C appears to extend throughout the area and is associated with hydrothermal quartz deposits not identified during operations. Intrusive units appear to outcrop in the mountains around Site D, and evaporite deposits in the valley are more extensive than was interpreted based on orbital data. Large chunks of welded ignimbrite make up much of the alluvial fan traversed at Site E. This ignimbrite appears to be responsible for the VNIR spectral shape identified as clay at this site, and is likely to be chemically altered even on a relatively fresh surface. The ubiquity of the clay minerals observed in nearly every collected spectrum is observed in the field as a pervasive coating of fine silt.

[75] There were units observed on the ground that were not identified during operations, however. Although the rover traverse did not reach outcrops observed in the field, the intrusive units at Sites B and D should have contributed to float observed along the rover traverse, but were not specifically identified during rover operations. The extent of the volcanic tuff at Site C was not realized during operations and the hydrothermal quartz was not identified either due to lack of a hydrated amorphous quartz in the TIR spectral library or masking by an alteration layer. The saline crust heaves that likely were responsible for indications of sulfate in the Site D end member image appeared to be more extensive than observed from orbit, and were not conclusively identified by field-based spectroscopy: it is possible that these sulfates are very dehydrated, or that the area observed by the spectrometers was simply too large. In either case, the absorption bands diagnostic of sulfate may have been overwhelmed by the surface soil spectrum. The ignimbrite at Site E appears to overlie a limestone, although neither orbital nor rover data sets suggested the presence of any carbonate. Many missed identifications are likely due to lack of data: the limestones at Site E did not outcrop at any of the locales, and even though outcrops were observed in drainages seen in navcam images, it would have been very difficult to return to those specific points for data collection. In addition, laboratory VNIR spectra of that limestone unit have very shallow absorption features, which may have been obscured by noise in the rover data sets.

[76] Although not part of the rover traverse, ground truth investigations at Site E included examination of the unit identified as pmb on the orbital images (see Figure 1e). This unit was not modeled well by analysis of orbital TIR data, and had been identified as a target of interest partially for this reason (also see discussion by Thomas et al. [2007]). The reason for this discrepancy was immediately clear in the field: this unit is a salt deposit with a thin covering of fine sediment (Salar de Navidad). Laboratory VNIR spectra of samples of this unit have relatively weak absorption bands consistent with gypsum: although halite is clearly a major component of the sample, this mineral has no absorption features in either the VNIR or TIR and would not be directly identifiable in these spectra. Although a relatively large percentage of a blackbody end member was modeled in linear deconvolutions over Salar Grande (as compared to the rest of the image, see Figure 5), the same is not true for Salar de Navidad. In that case, high RMS errors suggest that the algorithm found no combination of end members that fit the data: the spectrum of this unit (Figure 6) does not match well with any of the library end members. The difficulties in determining a particular composition indicated by the large values of a blackbody spectrum modeled for the units at Site E and RMS error (pmb, which turned out to be Salar de Navidad, and pvb, which was identified as a playa) turned out to be useful as identifiers of locations of interest for rover exploration.

6. Implications for Mars

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[77] The intent of the Life in the Atacama project was to use the results of science-blind rover field tests to better understand the limitations of a rover-driven study for application to future Mars missions. Many of the data sets used in the study are similar to those available for Mars: for example, the ASTER data discussed here are similar in spatial and spectral resolution to both the visible and infrared capabilities of the Thermal Emission Mapping Spectrometer (THEMIS) aboard Mars Odyssey [Christensen et al., 2003]. Many of the results discussed throughout this special issue should have direct implications to the interpretation of Martian data sets.

[78] Of particular interest is the observation that spectral diversity observed from orbit does not necessarily accurately reflect the diversity of compositions at the surface. This is perhaps no surprise considering the relatively large spatial resolution of the ASTER TIR data set (90 m/pixel) in comparison to the field of view of the rover spectrometers (no more than a few meters), but it is often assumed that the linear addition of the spectra of surface components will allow for the identification of these components from orbit. The orbital end member images discussed here are limited to 4 potential end member compositions when a blackbody is included to account for particle size effects: with spectrally coarse data sets, the parameters of the data set limit the breadth of the possible results. In addition, the limits of the linear deconvolution algorithm, which is reliable only for concentrations above 10% [Feely and Christensen, 1999], reduce the number of “minor” components that can be modeled.

[79] There was one field site for which hyperspectral data were available. At Site A, the orbital data set used had over 100 wavelength bands with usable data (excluding those affected by atmospheric water). The predicted compositional diversity over this site was limited to two minerals, although the rover data set included five different compositions with less than 30 total spectra collected. The evaporite minerals identified in the field were absent from the orbital data analyses. The spectral diversity observed on the ground was not apparent from orbit. This is a key point for identification of future landing sites on Mars: an area that appears spectrally bland from orbit may well prove much more diverse once it can be observed at a smaller scale. This has been demonstrated on Mars by the Opportunity rover in Meridiani Planum, which appeared relatively homogeneous from orbital Thermal Emission Spectrometer data (3 × 5 km per pixel hyperspectral TIR data) but much more diverse from the rover at the surface [i.e., Squyres et al., 2006]. This study demonstrates that this conclusion extends to higher resolution data sets (ASTER, 90 m/pixel TIR and Hyperion, 30 m/pixel VNIR).

[80] The spectral targeting strategy employed during LITA operations also affected the results. As discussed earlier, “blind” targeting of unknown locales had a tendency to produce spectrally bland results. This was most apparent at Site B, where both the VNIR and TIR spectrometers were targeted with only azimuth/elevation coordinates at most locales. The predominance of clay minerals in these results suggests that the soil component dominated the spectral results. Without specific associated images that could be analyzed for specific rock and soil percentages, it is very difficult to determine which compositions most likely belong to which surface component. The most spectrally diverse results (Site A) were obtained by direct observations of specific rock targets: patches of soil can easily be targeted this way as well. The lack of spatial information in single spectrum observations suggests that these should be reserved for specific observations on identified targets, rather than employed in a “blind” manner. These results might also have implications for autonomous spectral fitting algorithms [e.g., Gilmore et al., 2007] designed to identify and examine targets of interest in spectral data without scientist intervention.

[81] This conclusion does not apply, however, for multiple spectra observations with known spatial relationships. Although the acquisition of spectral “raster” images was not successful during these field tests, this method has been shown to be quite successful on Mars (e.g., results from the Spirit rover discussed by Christensen et al. [2004]). When the spatial relationship of the spectrometer and the imagers is well known, the data sets can be correlated and the results more robust. This is particularly key in the search for small scale habitats: in the case of this study, the results from the field spectrometers could not be accurately correlated with images from the FI, which makes compositional analysis of these habitats more difficult.

7. Summary

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[82] The surface and near subsurface composition of the areas of the Atacama Desert explored during rover operations for the Life in the Atacama project were examined using VNIR and TIR spectroscopy, and the subsurface hydrogen concentration over one traverse was probed with a neutron detector. Analysis of TIR orbital data from the ASTER instrument was used prior to operations to examine the regional composition and possible particle sizes of the surface, and the results from these analyses are broadly consistent with targeted results from the rover spectrometers. These analyses show an area that is dominated by alteration minerals such as clays and zeolites, with minor concentrations of sulfates, quartz, and iron oxide. The primary unaltered rock type appears to be volcanic in origin, based on observed rock morphology, composition of alteration minerals, and identification of primary volcanic minerals such as feldspar and pyroxene, as well as obsidian, in TIR spectroscopic results.

[83] Both the orbital and rover data analyses are consistent with observations from ground truth expeditions. Some compositional units that were missed, however. These ‘misses’ were likely the result of lack of data: specific outcrops or locations were simply not observed in enough detail. This has a number of causes, including the targeting strategy employed by the science team, non-functioning instruments, and impassable terrain. The rover data sets represent specific local samples, from which a general view of the region can be made.

[84] This project included the first use of of a rover-based neutron detector in a blind science field trial. Correlations were seen between hydrogen abundances derived from the neutron detector and areas along the traverse where more water would be expected, including drainage channels and areas of vegetation. Future work will determine the moisture content of ground truth soil samples taken at each location of the traverse.

[85] The results discussed here have implications for future surface missions on Mars and development of autonomous spectral analysis algorithms for these missions. The observation that orbital data sets may appear spectrally bland whereas the surface is actually much more compositionally diverse is extended to higher resolution data sets, and should be considered when selecting landing sites. Once observing on the ground, it is important to understand the relationship between collected spectral data and visible images of the surface. There is no spatial information inherent in the spectral data, and these data can be difficult to interpret without visual information about the surface they represent.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information

[86] We would like to thank Wendy Calvin, Phil Christensen, Jeff Johnson, Jack Mustard, and Hugh Kieffer for use of their co-owned D&P spectrometer in the 2003 and 2004 seasons; Paul Lucey and Keith Horton of U.H. for letting us use their D&P in the 2005 season; Win Wadsworth of D&P for real-time trouble-shooting of the instrument via e-mail while we were in the field; Pedro Ramirez for help with hardware trouble-shooting in the field; Carl Pilcher and Liz Williams of NASA HQ for help in coordinating permissions for the neutron experiment; Sean Brandenburg of Transgroup Worldwide Logistics for exceptional service in helping us get the Californium source shipped to Chile; Julio Bahuer Pizarro Maluenda of the Laboratorio de Investigación y Ensayes de Materiales Universidad Católica del Norte for assistance with the source after it arrived in Chile; the staff of the U.S. Embassy in Santiago for help in facilitating the neutron experiment; and Steve Ruff for the use of his zeolite spectral library. Additional thanks to the Zoë field crew and the Eventscope team for invaluable aid in the collection and interpretation of rover data. Thoughtful comments from Jeff Byrnes and an anonymous reviewer were appreciated.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information
  • Abrams, M. (2000), The Advanced Spaceborne Thermal Emission and Reflectance Radiometer (ASTER): Data products for the high spatial resolution imager on NASA's Terra platform, Int. J. Remote Sens., 21, 847859.
  • Bao, H., K. A. Jenkins, M. Khachaturyan, and G. Chong-Díaz (2004), Different sulfate sources and their post-depositional migration in Atacama soils, Earth Planet. Sci. Lett., 224, 577587.
  • Boardman, J. W., F. A. Kruse, and R. O. Green (1995), Mapping target signatures via partial unmixing of AVIRIS data, in Proceedings of 5th JPL Airborne Earth Science Workshop, JPL Publ., 95-1, 1, 2326.
  • Cabrol, N. A., et al. (2007), Life in the Atacama: Searching for life with rovers (science overview), J. Geophys. Res., 112, G04S02, doi:10.1029/2006JG000298.
  • Carta Geologica de Chile (1978), escala 1:250,000, no. 30, Hoja Antofagasta, Inst. de Invest. Geol., Chile.
  • Carta Geologica de Chile (1981), escala 1:250,000, no. 51, Hoja Quillagua, Inst. de Invest. Geol., Chile.
  • Christensen, P. R., J. L. Bandfield, V. E. Hamilton, D. A. Howard, M. D. Lane, J. L. Piatek, S. W. Ruff, and W. L. Stefanov (2000), A thermal emission spectral library of rock-forming minerals, J. Geophys. Res., 105, 97359739.
  • Christensen, P. R., et al. (2003), The Thermal Emission Imaging System (THEMIS) for the Mars 2001 Odyssey mission, Space Sci. Rev., 110, 85130.
  • Christensen, P. R., et al. (2004), Initial results from the Mini-TES experiment in Gusev Crater from the Spirit rover, Science, 305, 837842.
  • Clark, R. N., G. A. Swayze, A. J. Gallagher, T. V. V. King, and W. M. Calvin (1993), The U.S. Geological Survey, Digital Spectral Library: version 1: 0.2 to 3.0 microns, U.S. Geol. Surv. Open File Rep. 93-592, 1340 pp.
  • Cook, M., B. Peterson, G. Dial, F. Gerlach, K. Hutchins, R. Kudola, and H. Bowen (2001), IKONOS Technical Performance Assessment, Proc. SPIE, 4381(10), 1620.
  • Cooper, C. D., and J. F. Mustard (1999), Effects of very fine particle size on reflectance spectra of smectite and palagonitic soil, Icarus, 142, 557570.
  • Dingman, R. J. (1967), Geology and ground-water resources of the northern part of the Salar de Atacama Antofagasta Province, Chile, U.S. Geol. Surv. Bull. 1219, 56 pp.
  • Duff, M. C., J. U. Coughlin, and D. B. Hunter (2002), Uranium co-precipitation with iron oxide minerals, Geochim. Cosmochim. Acta, 66, 35333547.
  • Dunai, T. J., G. A. González López, and J. Juze-Larré (2005), Oligocene-Miocene age of aridity in the Atacama Desert revealed by exposure dating of erosion-sensitive landforms, Geology, 33, 321324.
  • Escobar, T. F., A. G. Puig, and G. J. Muzzio (1980), Mapa geológico de Chile, escala 1:1,000,000, Dep. de Geol. General, Santigo, Chile.
  • Feely, K. C., and P. R. Christensen (1999), Quantitative compositional analysis using thermal emission spectroscopy: Application to igneous and metamorphic rocks, J. Geophys. Res., 104, 24,19524,210.
  • Gillespie, A., S. Rokugawa, T. Matsunaga, J. S. Cothern, S. Hook, and A. B. Kahle (1998), A temperature and emissivity separation algorithm for Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images, IEEE Trans. Geosci. Remote Sens., 36, 11131126.
  • Gilmore, M. S., R. Castaño, B. Bornstein, and J. Greenwood (2007), Autonomous mineral detectors for visible/near-infrared spectrometers at Mars, paper presented at 7th International Conference on Mars.
  • Green, A. A., M. Berman, P. Switzer, and M. D. Craig (1988), A transformation for ordering multispectral data in terms of image quality with implications for noise removal, IEEE Trans. Geosci. Remote Sens., 26, 6574.
  • Grove, C. I., S. J. Hook, and E. D. Paylor (1992), Laboratory reflectance spectra for 160 minerals 0.4–2.5 micrometers, JPL Publ., 92-2.
  • Hardgrove, C., J. Moersch, D. Drake, J. Piatek, D. Wettergreen, and N. A. Cabrol (2006), Field tests and ground truthing of a surface-based neutron detector in the Atacama Desert, Chile, Lunar Planet. Sci. Conf., XXXVII, Abstract 2320.
  • Hartley, A. J., and G. Chong (2002), Late Pliocene age for the Atacama Desert: Implications for the desertification of western South America, Geology, 30, 4346.
  • Horton, K. A., J. R. Johnson, and P. G. Lucey (1998), Infrared measurements of pristine and disturbed soils. 2. Environmental effects and field data reduction, Remote Sens. Environ., 64, 4752.
  • Lane, M. D. (2007), Midinfrared emission spectroscopy of sulfate and sulfate-bearing minerals, Am. Mineral., 92, 118.
  • Moersch, J. E., and P. R. Christensen (1995), Thermal emission from particulate surfaces: A comparison of scattering models with measured spectra, J. Geophys. Res., 100, 74657477.
  • Moersch, J. E., and D. M. Drake (2003), Neutron detector for Mars rover missions, paper presented at 3rd International Conference in Mars Polar Science and Exploration.
  • Mustard, J. F., and J. E. Hays (1997), Effects of hyperfine particles on reflectance spectra from 0.3 to 25 μm, Icarus, 125, 145163.
  • Navarro-González, R., et al. (2003), Mars-like soils in the Atacama Desert, Chile, and the dry limit of microbial life, Science, 302, 10181021.
  • Palacios, C., N. Guerra, B. Townley, A. Lahsen, and M. Parada (2005), Copper geochemistry in salt from evaporite soils, Coastal Range of the Atacama Desert, northern Chile: an exploration tool for blind Cu deposits, Geochem. Explor. Environ. A, 5, 371378.
  • Pearlman, J. S., P. S. Barry, C. C. Segal, J. Shepanski, D. Beiso, and S. T. Carman (2003), Hyperion, a space-based imaging spectrometer, IEEE Trans. Geosci. Remote Sens., 41, 11601173.
  • Pelowitz, D. B. (Ed.) (2005), MCNPX user's manual version 2.5.0, Rep. LA-CP-05-0369, Los Alamos Natl. Lab., Los Alamos, N. M.,
  • Pueyo, J. J., G. Chong, and A. Jensen (2001), Neogene evaporites in desert volcanic environments: Atacama Desert, northern Chile, Sedimentology, 48, 14111431.
  • Ramsey, M. S., and P. R. Christensen (1998), Mineral abundance determination: Quantitative deconvolution of thermal emission spectra, J. Geophys. Res., 103, 577596.
  • Reutter, K.-J., E. Scheuber, and G. Chong (1996), The precordilleran fault system of Chuquicamata, northern Chile: Evidence for reversals along arc-parallel strike-slip faults, Tectonophysics, 259, 213228.
  • Ruff, S. W. (2004), Spectral evidence for zeolites in the dust on Mars, Icarus, 131, 131143.
  • Salisbury, J. W., L. S. Walter, N. Vergo, and D. M. D'Aria (1991), Infrared (2.1–25 Micrometers) Spectra of Minerals, 294 pp., Johns Hopkins Univ. Press, Baltimore, Md.,
  • Squyres, S. W., et al. (2006), Two years at Meridiani Planum: Results from the Opportunity rover, Science, 313, 14031407.
  • Stoertz, G. E., and G. E. Ericksen (1974), Geology of salars in northern Chile, U.S. Geol. Surv. Prof. Pap. 811, 75 pp.
  • Thomas, G. W., et al. (2007), Comparing different methods for assessing ground truth of rover data analysis for the 2005 season of the Life in the Atacama Project, J. Geophys. Res. 112, G04S09, doi:10.1029/2006JG000318.
  • Vincent, R. K. (1997), Fundamentals of Geological and Environmental Remote Sensing, 370 pp., Prentice-Hall, Upper Saddle River, N. J.,
  • Warren-Rhodes, K. A., K. Rhodes, S. Pointing, S. Ewing, D. Lacap, B. Gómez-Silva, R. Amundson, E. I. Friedmann, and C. P. McKay (2006), Hypolithic cyanobacteria, dry limit of photosynthesis and microbial ecology in the hyperarid Atacama Desert, Microbial Ecol. doi:10.1007/s00248-006-9055-7.
  • Warren-Rhodes, K., et al. (2007a), Searching for microbial life remotely: Satellite-to-rover habitat mapping in the Atacama Desert, Chile, J. Geophys. Res., 112, G04S05, doi:10.1029/2006JG000283.
  • Warren-Rhodes, K.et al. (2007b), Robotic ecological mapping: Habitats and the search for life in the Atacama Desert, J. Geophys. Res. 112, G04S06, doi:10.1029/2006JG000301.
  • Warren-Rhodes, K. A., J. L. Dungan, J. Piatek, K. Stubbs, B. Gomez-Silva, and C. McKay (2007c), Ecology and spatial pattern of cyanobacterial community island patches in the Atacama Desert, Chile, J. Geophys. Res. 112, G04S15, doi:10.1029/2006JG000305.
  • Wierzchos, J., C. Ascaso, and C. P. McKay (2006), Endolithic cyanobacteria in halite rocks from the hyperarid core of the Atacama Desert, Astrobiology, 6, 18.

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Orbital Data Sets
  5. 3. Ground-Based Results
  6. 4. Discussion
  7. 5. Ground Truth Results
  8. 6. Implications for Mars
  9. 7. Summary
  10. Acknowledgments
  11. References
  12. Supporting Information
FilenameFormatSizeDescription
jgrg179-sup-0001-t01.txtplain text document3KTab-delimited Table 1.
jgrg179-sup-0002-t02.txtplain text document3KTab-delimited Table 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.