We studied low-albedo deposits in the floors of craters within the Amazonis Planitia region using infrared data from Mars Global Surveyor Thermal Emission Spectrometer (TES) and Mars Odyssey Thermal Emission Imaging System (THEMIS), high-resolution visible images from the Mars Orbiter Camera and THEMIS, and topographic data from the Mars Orbiter Laser Altimeter. These deposits are found within subdued to sharp-rimmed craters that impacted into thick basaltic lava flows from Olympus Mons. These lava flows flooded the region and were subsequently eroded by aeolian processes. TES spectral modeling of the low-albedo deposits reveals a mineralogy that is dominated by mafic minerals (olivine, pyroxene) with a derived bulk chemistry that ranges from ultramafic to mafic (∼40–52 wt % SiO2) in composition. These bulk compositions are comparable to some Martian lithologies but represent some of the lowest silica contents identified on Mars. The compositional range could be produced by either primary igneous processes or subsequent alteration by aeolian processes. The spatial distribution of the compositions reveals weak to absent groupings and does not distinguish between the igneous or aeolian scenarios. In some craters, possible local sources have been identified within the crater walls, floors and interiors; local sources may be present in additional craters but obscured by dust. No regional sources were identified within a reasonable distance in terms of sand transport but may also be obscured by dust. Therefore, the low-albedo deposits probably resulted from aeolian erosion of local intracrater basaltic materials that were subsequently redistributed within the floors of these craters.
 Low-albedo intracrater deposits on Mars were first identified in Mariner images in the early 1970s [e.g., Arvidson, 1974; Sagan et al., 1973] and appear in more than 25% of Martian craters ≥25 km in diameter within 50° of the equator [Christensen, 1983]. The localized deposits are both lower in albedo and higher in thermal inertia than the surrounding materials, which is commonly interpreted to mean that they are dominated by sand-sized particles [Christensen, 1983; Edgett and Christensen, 1994]. Although these deposits are found in both low- and high-albedo regions [e.g., Christensen, 1983; Schneider and Hamilton, 2006a], studying low-albedo deposits in high-albedo (dusty) regions may be important because, if they are locally derived materials, they provide insight into the composition of an area whose lithologic composition is otherwise obscured by dust cover.
Edgett and Christensen [1992, 1994] studied low-albedo intracrater deposits found in various regions of Mars. Using three-point emissivity data derived from the Viking Infrared Thermal Mapper (IRTM) [Kieffer et al., 1972], they concluded that the windblown sands on Mars may be dominated by mafic and ultramafic minerals and rock fragments [Edgett and Christensen, 1992, 1994]. These investigators observed distinct regional thermophysical and/or aeolian dune characteristics and attributed them to variability in the availability of sand and regional wind conditions. In addition, Edgett and Christensen  observed that craters located across Mars, contained thick, transverse dune fields with a thermal inertias of 270 ± 40 (thermal inertia units are J × m−2 × s−0.5 × K−1, and will be assumed for all subsequent thermal inertia values), consistent with an average grain size of medium to coarse sand (250–1000 μm). Subsequent studies of Proctor Crater using Mars Global Surveyor Thermal Emission Spectrometer (TES) data calculated an average thermal inertia value of 277 ± 17, consistent with coarse sand, and determined that the sands are basaltic in composition [Aben, 2003; Fenton et al., 2003]. Mafic-rich zones with elevated concentration of olivine were also identified within craters (>20 km in diameter) within the northern plains of Vastitas Borealis by OMEGA, which were revealed to be localized aeolian deposits in Mars Orbiter Camera (MOC) images [Bibring et al., 2005].
1.2. Low-Albedo Intracrater Deposits Within Amazonis Planitia
Schneider and Hamilton [2006a] conducted a detailed study of a low-albedo, intracrater deposit in Amazonis Planitia at 17.92°N, 190.26°E using visible, thermal and spectral data. Their study found that the mineralogy of the low-albedo deposit is dominated by a mafic mineralogy (dominantly pyroxene and olivine, with lesser amounts of oxides) and that the bulk chemistry derived from the modeled mineralogy indicates an ultramafic lithology with the lowest silica contents currently found on Mars [Schneider and Hamilton, 2006a]. That study also determined, on the basis of the distribution of the dark sediment within the low-albedo deposits as well as the lack of an obvious source for the dark materials in the surrounding region, that this deposit is locally derived [Schneider and Hamilton, 2006a].
 The Amazonis Planitia region consists of eroded ancient cratered terrain that was subsequently flooded by a thick series of (probably basaltic) lava flows from Olympus Mons. This was followed by intense aeolian activity during which time the previous materials were eroded and buried in situ under a blanket of aeolian debris [Morris and Dwornik, 1978]. In general, the Amazonis Planitia region is dust covered, meaning that the fine surface dust component completely obscures the spectral signature of underlying rocks [Ruff and Christensen, 2002]. Therefore, we study the Amazonis intracrater low-albedo deposits to determine if they are locally derived materials and, if they are, to determine more about the lithologies present in this dusty region.
 We use infrared data from TES and Thermal Emission Imaging System (THEMIS), high-resolution visible images from the Mars Orbiter Camera (MOC), and topographic data from the Mars Orbiter Laser Altimeter (MOLA) to study the low-albedo deposits within the Amazonis Planitia craters (Table 1 and Figure 1). Both TES and THEMIS are thermal infrared spectrometers, and were designed to have complementary spectral and spatial resolutions. Because of its higher spatial resolution, THEMIS is best viewed as a spectral unit mapper, whereas TES is better at mineral identification because of its higher spectral sampling [Christensen et al., 2003]. Together, TES and THEMIS constitute a powerful tool for studying the thermal signatures of the surface of Mars.
Table 1. Information for Craters Containing Low-Albedo Intracrater Deposits Within the Amazonis Planitia Examined in This Study
 The Thermal Emission Spectrometer acquired infrared data (∼1655–∼200 cm−1, or ∼6–50 μm) with a spatial resolution of ∼3 × 6 km at a spectral sampling of 5 or 10 cm−1 [Christensen et al., 1992]. TES also acquired broadband visible (0.3–2.7 μm) and thermal (5–100 μm) data providing albedo and thermophysical information of the Martian surface [Christensen et al., 2001]. Calibration of the spectrometer and bolometers, and use of quality fields assigned to the data to aid in the selection of high-quality data for analysis, are described by Christensen et al. .
 The Thermal Emission Imaging System acquires infrared images with a spatial resolution of 100 m pixel−1 in nine surface-sensing spectral bands ∼1 μm wide. These bands are centered at eight wavelengths ranging from ∼1475–796 cm−1 (∼6.8–12.6 μm). Bands 1 and 2 have identical filters centered at ∼1475 cm−1 (∼6.8 μm) to obtain a better signal-to-noise ratio in this spectral region [Christensen et al., 2004]. THEMIS also has a visible/near infrared imaging subsystem that acquires images with a spatial resolution of ∼19 m pixel−1 at up to five wavelengths (between 0.426 and 0.860 μm) [Christensen et al., 2003]. Calibration of THEMIS visible data is described by Christensen et al. .
 The Mars Orbiter Camera, onboard the Mars Global Surveyor, is a three-component system including both one narrow-angle and two wide-angle cameras [Malin et al., 1992]. The narrow-angle camera can obtain images with up to 1.4 m pixel−1 spatial resolution for localized studies; the wide-angle cameras provide low-resolution (7.5 km pixel−1) images for global studies as well as intermediate-resolution (∼250 m pixel−1) images for regional studies. In addition, different color filters within the two wide-angle cameras provide color images of the surface and atmosphere [Malin et al., 1992].
 The Mars Orbiter Laser Altimeter determined the global topography and gravity of Mars [Zuber et al., 1992; Smith et al., 2001]. This data set has been used to produce global elevation maps with up to 128 m pixel−1 spatial resolution having a vertical precision of ∼37.5 cm [Smith and Zuber, 1998]. In this study, we have used the 128 pixel degree−1 MOLA Mission Experiment Gridded Data Records (MEGDR) topography data (version b) [Smith et al., 2003].
3.1. TES Data Selection and Spectral Modeling
 We examined TES spectra, THEMIS albedo, and THEMIS thermal inertia for 23 low-albedo deposits in Amazonis Planitia (∼150°–205°E and 0°–35°N) craters initially identified by Schneider and Hamilton [2006b] (Table 1). In addition, we performed a search for additional craters containing the low-albedo deposits within the defined region; however, none were found. This study retains the crater numbering system established by Schneider and Hamilton [2006b]; however, where appropriate, the official names of craters are listed in Table 1.
 In order to obtain the best quality data available, we selected spectra that had low-albedo (<0.19), high target temperature (>255 K), low atmospheric dust and water opacity (<0.20 and <0.10, respectively), no solar panel movement, no major phase inversions, low risk of phase inversions, and no high-gain antenna motion. In addition, the data were screened for a type of noise called “ringing,” an artifact that occurs in TES data when changes in the target temperature occur rapidly between observations, such that the electronics cannot reestablish the base level of the interferometer between observations [Christensen et al., 2001]. Ringing is present in some of the Amazonis Planitia data because the low-albedo deposits have significant temperature differences (∼20 K) from the surrounding terrain. Because many deposits were observed in only a few TES pixels, we calculated an average spectrum from all acceptable spectra identified to maximize the signal-to-noise ratio of the data for each low-albedo deposit. We also utilized TES spectra from the bright (dusty) crater floor adjacent to each deposit in our analyses; for these data, we were careful to select spectra from the same detectors as used in the observation of the deposit to avoid differences in sensitivity between detectors. Of the 23 Amazonis Planitia craters examined, we found satisfactory TES spectra for eleven of the low-albedo intracrater features (Table 1). For each of the 11 intracrater deposits, we extracted between 2 and 13 TES spectra from 1 to 3 orbits for each crater.
 The average TES spectrum for each low-albedo deposit and the associated bright crater floor were modeled via the linear least squares method of Ramsey and Christensen ; atmospheric components were accounted for and removed according to the methods of Smith et al.  and Bandfield et al. [2000b]. This approach fits the measured TES apparent emissivity spectrum using a spectral library of atmospheric and surface components. This strategy is based on the finding that atmospheric and surface spectral components add linearly in emissivity space for daytime TES spectra; these data are acquired when the high temperature contrast between the Martian surface and the atmosphere allows for the linear approximation to be accurate [Bandfield et al., 2000b; Smith et al., 2000]. The algorithm outputs a modeled spectrum, a list of library spectra used in the best fit model, the fractional amount of each component included in the best fit, and a root mean square (RMS) error.
 Because we intended to compare our results to the work of Schneider and Hamilton [2006a], we started with the same subset of library spectra and made slight refinements. Our final spectral library is detailed in Table 2. The major differences between our final spectral library and the one used by Schneider and Hamilton [2006a] are (1) carbonate library spectra were excluded and (2) the olivine library spectra included a different sampling of intermediate olivine compositions (based on the work of Koeppen and Hamilton ).
Table 2. Atmospheric, TES and Mineral End-Members Included in Spectral Modeling
Mineral Spectral Name
atmospheric dust (low CO2)
atmospheric dust (high CO2)
water ice cloud (high latitude)
water ice cloud (low latitude)
pure CO2 gas
pure H2O vapor
TES surface end-member
average high-albedo surface
maskelynite (labradorite) ASU-7951
albite WAR-0235 174
labradorite BUR-3080A 176
bytownite WAR-1384 177
anorthite BUR-340 178
anorthite WAR-5759 221
labradorite WAR-RGAND01 222
pigeonite without particle size effects
augite NMNH-9780 157
enstatite (average of HS-9.4B and NMNH-34669)
hypersthene DSM-FER01 152
bronzite NMNH-93527 168
Forsterite (Fo91) BUR-3720A
Fayalite (Fo1) WAR-RGFAY01 167
kaolinite KGa-1b solid 186
nontronite WAR-5108 solid 204
Fe-smectite SWa-1 solid 207
illite IMt-2 granular 211
 Since the original study of Schneider and Hamilton [2006a], Koeppen and Hamilton  have shown that carbonates are rarely used in models of mafic materials if olivine spectra representing a range of solid solution compositions are included in the spectral library. Although carbonates have been identified as a minor (<5%) component of the surface dust [Bandfield et al., 2003], the misidentification of carbonates by spectral modeling has been noted elsewhere on Mars [e.g., Koeppen and Hamilton, 2008] and great care must be taken when interpreting models that have included carbonates. The presence of carbonates in amounts ≥10% can be confirmed by visual inspection of surface spectra, which typically reveal the relatively narrow carbonate absorption feature at ∼890 cm−1 [Stockstill et al., 2005]. Although the models of Schneider and Hamilton [2006a, 2006b] included 10–25% carbonates, a carbonate absorption feature is not obvious in any of those surface spectra; therefore, we opted to omit carbonate spectra from our final spectral library.
 The study of Koeppen and Hamilton  also demonstrated that some compositions (e.g., Fo68 and Fo60) have very similar band shapes and minima at the 10 cm −1 spectral sampling of TES and can be accurately modeled using a linear combination of other olivine spectra. Therefore, we selected olivine spectra so that each spectrum has a unique spectral shape not easily modeled by combinations of other olivine spectra. Specifically, the olivine spectra used by Schneider and Hamilton [2006a] for intermediate olivine compositions included Fo68, Fo60, Fo52.5, Fo35; this study included library spectra for intermediate olivine compositions of Fo68, Fo52.5, Fo39 and Fo18 (Table 2). Both studies included library spectra for the two solid solution end-members, forsterite and fayalite (Table 2).
3.2. THEMIS and MOC Visible Images and MOLA Topography
 Finally, high-spatial resolution visible images allowed us to examine the morphology of the deposits and to identify potential intracrater sources for the low-albedo materials. THEMIS visible images (∼19 m pixel−1) and MOC narrow angle images (up to 1.4 m pixel−1) were used to examine the deposits and crater morphology. In addition, we used 128 pixel degree−1 MOLA topography data to examine the topographic relief of crater walls, floors and deposits.
4.1. TES Linear Least Squares Spectral Models
 As with the findings of Schneider and Hamilton [2006a, 2006b], our spectral modeling reflects lithologies that are rich in mafic minerals (olivines, pyroxenes and oxides) and plagioclase (Table 3), with the pyroxene component being dominated by clinopyroxene. Although the model for Amazonis 17 of Schneider and Hamilton [2006a] included 10% orthopyroxene, none of our models included any orthopyroxene (Table 3). In general, we see an increase in olivine abundance, decreases in pyroxene, and a similar abundance of the feldspar component included in the models relative to Schneider and Hamilton [2006a, 2006b]. Most TES surface emissivity spectra from this study strongly resemble the surface spectrum for crater 17 (Figure 2). To variable degrees, the spectra display an emissivity minimum centered at ∼910 cm−1 and a peak centered at ∼435 cm−1 that both match well to olivine laboratory spectra (Figure 2); the variability in the strength of these features may reflect differences in the olivine abundance within the deposits. Indeed, three spectra with the deepest olivine minima (i.e., 1, 10, and 12; Figure 2) contain among the highest olivine abundances among the Amazonis deposits (Table 3).
Table 3. TES Data Information and Mineral Group Model Abundancesa,b Derived by Linear Deconvolution of TES Data for Each Low-Albedo Intracrater Deposit
Model abundances have been normalized to exclude atmosphere, blackbody and surface dust and then rounded to the nearest 5%.
Oxide-removed abundances shown in parentheses.
In this table, mafic minerals include clinopyroxene, orthopyroxenes and olivine.
4.2. Lithologic Classification and Bulk Composition
 Because previous workers have determined that these deposits are dominated by igneous materials [Edgett and Christensen, 1992, 1994; Aben, 2003; Fenton et al., 2003; Schneider and Hamilton, 2006a], we selected a spectral library that was tailored for mafic igneous rocks. Therefore, we also assign a “rock” type on the basis of igneous classification schemes; because spectroscopy cannot provide information about the petrologic texture of these deposits, we consider the results using both plutonic and volcanic classification schemes. Using these igneous classification schemes allows for direct comparison of our derived compositions to other Martian lithologies as well as terrestrial rocks. First, the proportion of plagioclase, pyroxene and olivine of each deposit are plotted on the International Union of Geological Sciences classification diagram for gabbroic rocks [Streckeisen, 1974] to assign a plutonic rock type to each deposit. Equivalent bulk rock chemistries are derived by combining the known electron microprobe-measured compositions (wt % oxides) of the minerals in proportion to the abundances derived from the linear least squares model, providing a bulk composition in terms of oxide abundances [Hamilton and Christensen, 2000; Hamilton et al., 2001; Wyatt et al., 2001]. Each deposit was assigned to a lithologic class using the total-alkali silica (TAS) diagram [Streckeisen, 1978]. In general, derivation of bulk rock chemistry from linear least squares models and subsequent classification have been demonstrated to be accurate, especially when supplemented by comparisons to library rock spectra [Hamilton and Christensen, 2000; Dunn et al., 2007]. The mineral abundances for each individual spectrum were converted to major oxide abundances, as described above, and the error propagated in quadrature, to calculate the standard deviations for the mean oxide abundances for a given crater.
 In terms of plutonic lithologies, the Amazonis low-albedo intracrater deposits are mafic in composition, ranging from olivine gabbronorite to troctolite to plag-bearing ultramafic (Figure 3a and Table 3). Amazonis 1 plotted within the ultramafic field, which requires further classification using an Ol-Opx-Cpx ternary (not shown since it is the only deposit that requires further classification); this deposit is a peridotite (specifically, wehrlite). The spatial distribution of these rock types is displayed in Figure 3b. When we converted the derived mineralogies to bulk chemical compositions, the low-albedo deposits are classified as ultramafic to mafic lithologies, ranging from 29–43% SiO2 (Figure 3c). Many deposits have lower derived SiO2 abundances than the previously studied crater 17 (Figure 3c) and plot at even lower silica contents than the foidite field (craters 1, 9, 10, 12, and 15). Four craters (3, 4, 5 and 17) plot within the foidite field. Two craters plot within the picrobasalt (7) and tephrite basanite (18) fields. The spatial distribution of these rock types is displayed on the Amazonis map in Figure 3d.
4.3. THEMIS and MOC Visible Images
 In THEMIS and MOC visible images, we see a variety of morphologies for the low-albedo deposits, including areas with barchan, transverse and linear dune forms as well as areas appearing to lack dune forms (possibly due the spatial resolution of available data). In crater 3 (Figure 4a), THEMIS visible data reveal barchan dune forms in the southernmost portion of the dune field oriented to winds from the SSE and transverse dune forms in the northwest portion of the dune field with two slip faces oriented to winds from the SE and SW (Figure 4b). In crater 9 (Figure 4c), a THEMIS visible image (Figure 4d) reveals the presence of barchan dune forms aligned with winds from the NE. In crater 12 (Figure 4e), a MOC narrow-angle image (Figure 4f) reveals barchan dunes at the southern edge of the dune field oriented to winds from the ESE, whereas barchan dunes at the northern edge of the dune field are oriented to winds from the NNW. These combine into linear dunes oriented in the resultant effective wind direction, which is from the NE (Figure 4f). In crater 18 (Figure 4g), a THEMIS visible image (Figure 4h) reveals transverse dune forms formed by winds from the NW or SE There is not an apparent relationship between dune morphology and the bulk composition of the deposit.
 We also used visible images to search crater walls and floors for possible source regions and we identified potential sources of low-albedo material in four craters (1, 6 11 and 13). In crater 1, THEMIS images displayed high-thermal inertia materials in the crater floor (Figure 5a), which also appear as low-albedo materials in THEMIS visible images (Figures 5b and 5c). Specifically, THEMIS visible images show some evidence for low-albedo materials within small craters within the eastern crater floor (Figure 5b). These may be point sources for the low-albedo materials and appear similar to the point source identified in crater 17 by Schneider and Hamilton [2006a]. Additionally, it is possible that (1) the craters are recent impacts that are dark because they have exposed dark sand underlying the current floor of this crater (i.e., dark sand is widespread across the crater floor but perhaps buried by dust), or (2) the small craters are older features on the floor of this crater that have trapped dark sand as it saltated by (i.e., the source of the low-albedo materials is upwind of these features).Some low-albedo materials are observable within the central peak pit (Figure 5c). Winds from the ENE may have blown materials derived from the eastern part of the crater and in through the gap in the eastern part of the peak pit rim, trapping some low-albedo materials within the central peak. (Note: Although these materials may have been derived from local sources within the central peak or eastern crater floor, these materials may have come from some other direction or outside the crater as well; unfortunately, no sand transport pathway leading to any obvious upwind source so the source of these materials cannot be distinguished.)
 Within crater 6 (Figure 5d), the low-albedo (high-thermal inertia) material is distributed south-southwest of the central mound. A MOC narrow-angle image (Figure 5e) shows that the mound has a lower albedo than the northern crater floor and that there is darker-toned material at the mound's base. In fact, adjacent to the base of the mound may also display layers (or lineated deposits) with even darker-toned material (Figure 5e). A MOLA-derived topographic profile (Figure 5f) demonstrates that the elevation of the crater floor varies north and south of the mound, with the crater floor south of the mound appearing to have higher elevation. This may be consistent with the low-albedo material being derived from the mound, transported by northeasterly regional winds and deposited in the crater floor southwest of the mound. In addition, the northern portion of the crater may have experienced more deflation than the southern portion, and perhaps a lag deposit from such a process could be a source of low-albedo materials that have been deposited in the southern crater floor. Furthermore, the low-albedo materials may have been derived from both the central mound as well as the northern crater floor. However, THEMIS decorrelation stretch (DCS) images (Figure 5g) show strong spectral signatures of the low-albedo material on the mound and south of the mound but not on the northern crater floor (although they may be obscured by dust in this area).
 In crater 11 (Figure 5h), a MOC narrow-angle image reveals layering in the crater walls and within interior crater units (Figure 5i) that may supply the low-albedo dune materials within this crater. We also see a possible layer of low-albedo material in the walls of crater 13 (Figure 5j), revealed by a MOC narrow-angle image (Figure 5k). We also see a possible layer of low-albedo material in the walls of crater 13 (Figure 5l), revealed by a MOC narrow-angle image (Figure 5m). However, when a profile across this layer (Figure 5n) was derived from MOLA data, the topographic profile (Figure 5o) shows a slightly steeper downward slope at the location of the dark layer. Therefore, because of the orientation of the sun, the proposed layer may actually be a shadow. Although the rocky units in the crater walls are likely resistant to aeolian erosion, abrasion over long periods could contribute some low-albedo materials to the crater interiors; indeed, this particular crater lacks a sharp rimmed morphology suggesting some modification of this crater has occurred [Scott and Allingham, 1976].
5.1. Spectral Model Results
 The spectral models typically included 15–25% oxide minerals, with the oxide component dominated by magnetite. In general, magnetite is only present in igneous rocks as a minor mineral (typical abundance <5%); in ultramafic rocks (i.e., foidites, picritic basalts), magnetite is present in even lesser amounts [Raymond, 1995]. Furthermore, Fe-Ti oxides commonly only represent up to ∼3% of the Martian meteorites [McSween, 1995]. The large modeled abundances of magnetite have two possible explanations: (1) magnetite is being erroneously included in the models, or (2) these deposits have been strongly enriched in magnetite relative to an ultramafic/mafic parent rock.
Schneider and Hamilton [2006a] also noted the large amounts of oxide minerals included in their model of the Amazonis crater 17 low-albedo deposit and noted that magnetite has a negative slope with only one small feature in the TES wavelength region (Figure 6). Their study suggested that magnetite might be erroneously modeled if there were slight errors in the temperature determination used to convert TES data to emissivity [Schneider and Hamilton, 2006a]. In areas where cold materials (e.g., dusty floor) are adjacent to hotter materials (e.g., low-albedo deposit), mixing of disparate surface temperatures may occur within a TES pixel. Thus, the TES spectrum may represent a broader range of temperatures, producing a minor slope in the data [Schneider and Hamilton, 2006a]. In this case, the negative-sloping magnetite spectrum may actually improve the model fit, although magnetite is not really present.
 To test this potentially erroneous inclusion of magnetite, we substituted a slope spectrum (Figure 6) into the spectral library in place of the magnetite spectrum. The magnitude of the slope was derived assuming a 1 K error in the temperature used in the emissivity conversion process. In all cases, the spectral RMS did not show significant change between the magnetite model and the slope model. In all but one case, the amount of the slope spectrum included ranged from 10–60%. Note that in one case (Amazonis 18), the model did not include any slope spectrum; for this model, excluding magnetite provided a very small (but still significant) improvement in fit, changing the RMS error from 0.248 for the magnetite-included model to 0.234 for the magnetite-excluded model.
 In summary, in magnetite models, magnetite is included in the models in amounts exceeding what we would expect in an igneous rock and the Martian meteorites. In addition, when slope is provided in place of magnetite, the slope spectrum is commonly used in relatively high abundances to improve the fit of the models without significant changes in the abundances of other minerals. In general, neither the magnetite model nor the slope model provides significant improvements in the fit (i.e., RMS error) of the model and, therefore, no model can be chosen as a best fit. These facts are consistent with magnetite being included in the models to contribute a slope to the spectra in order model temperature variation within a pixel, as proposed by Schneider and Hamilton [2006a].
 The low SiO2 contents of the Amazonis deposits arise from the high-oxide mineral (i.e., magnetite) abundances, which may be high because of temperature effects. When oxide minerals are included, some of the derived SiO2 content is typical of foidites. A foidite is defined as an igneous rock in which feldspathoids constitute 60–100% of the light-colored components and is sometimes restricted to rocks in which feldspathoids represent 90–100% of the light-colored constituents and/or volcanic rocks [Bates and Jackson, 1987; Raymond, 1995]. Thus, it is the inclusion of feldspathoids in place of feldspars, not the increased abundance of oxide minerals, which lowers the SiO2 content of foidites. Although feldspathoids minerals were not included in the spectral library, the models display satisfactory fits and therefore do not call for the addition of feldspathoids to the spectral library. Therefore, the most conservative interpretation is that the compositions derived from modeled mineralogies should exclude the oxide contribution (Table 4 and Figures 7a and 7b). For the remainder of the discussion, we will use model abundances and derived compositions that normalize out the oxide mineral component.
Table 4. Derived Oxide Abundancesa for Low-Albedo Deposits Found Within the Amazonis Craters (Weight Percent Oxide) After Subtracting the Oxide Mineral Contribution
Oxide abundances rounded to the nearest 1%; standard deviation of the mean in parentheses rounded to the nearest 0.1%.
5.2. Compositional Nature of the Low-Albedo Materials
 Our derived mineralogy for crater 17 can be compared directly with the derived mineralogy of Schneider and Hamilton [2006a]. Although their modeling included carbonates, we can simply subtract the modeled carbonate amount and renormalize the remaining mineral amounts to allow for direct comparison. We also have spectral and geological reasons to exclude the oxide minerals, and have removed these as well (Table 5).
Table 5. Mineral Group Model Abundancesa and the Derived Bulk Compositionb for Crater 17 From the Study of Schneider and Hamilton [2006b] (S&H) and This Study (SA&H), Minus Carbonatesc and Oxidesd
Model abundances have been normalized to exclude atmosphere, blackbody and surface dust and then rounded to the nearest 5%.
Oxide abundances rounded to the nearest 1%; standard deviation of the mean in parentheses rounded to the nearest 0.1%.
 Comparing both derived mineralogies reveals that our model included similar amounts of plagioclase, but the modeled abundances of olivine and pyroxenes differ. Specifically, as olivine increases from 20% in the model of Schneider and Hamilton [2006a] to 40% in our model, pyroxene decreases from 45% [Schneider and Hamilton, 2006a] to 30% in our model. However, the plutonic lithologic class of both models is olivine gabbronorite (Table 5). In terms of the derived bulk composition, our models vary slightly in silica content, increasing from ∼41% (±1.5%) in Schneider and Hamilton's [2006a] model to ∼45% (±1.1%) in this study (Table 5). This small variation in composition is enough to change the volcanic rock classification from picrobasalt (Schneider and Hamilton's [2006a] model) to basalt (this study, Figure 8). We note that the error bars do allow some variation in this classification; Schneider and Hamilton's [2006a] composition ranges from foidite to picrobasalt, whereas the composition from this study straddles the picrobasalt-basalt border.
 We can use the bulk composition of the low-albedo deposits to compare them to bulk chemistry data of other Martian lithologies [Lodders, 1998; Hamilton et al., 2001; McSween et al., 2003, 2004, 2006; McCollom and Hynek, 2005]. On the TAS diagram shown in Figure 8, the deposits span the compositional range from foidite to basaltic andesite (∼40–52 wt % SiO2) and match well to the range of silica contents represented by rocks measured by the Mars Exploration Rovers (MER) and the Martian meteorites and (∼37–52 wt % SiO2). However, when alkalis are taken into account, the Amazonis deposit compositional field generally plots at higher alkali contents than the compositional fields of Nakhlites and Shergottites. Despite their generally higher alkali contents, the low-albedo deposits still plot dominantly within the subalkaline fields; if their compositions are representative of their parent lithology/lithologies, this indicates that their parent magma(s) probably originated under similar mantle conditions, but the magma source(s) may have had slightly higher alkali contents. The low-albedo deposits plot at lower silica contents than derived for Surface Type II and the Mars Pathfinder S-free rock [Wänke et al., 2001; Foley et al., 2003; McSween et al., 2003].
 The low-albedo deposits represent some of the most mafic compositions on Mars. In fact, chassigny (an olivine-chromite cumulate) is the only bulk Martian composition with lower SiO2 content (∼37% SiO2). This introduces two questions: (1) does the compositional variation of the low-albedo deposits represent compositional variation of the rocks from which they were derived, or (2) does the compositional variation of the low-albedo deposits represent variations in alteration of a single lithology?
 To address these points, we have plotted the mafic oxide (FeO + MnO + MgO) content versus SiO2 of our low-albedo deposits (Figure 9a) along with compositions of phases that were included in the models [Christensen et al., 2000] and mineral compositions measured within Martian meteorites [McSween and Treiman, 1998, and references therein; Sautter et al., 2002]. This plot demonstrates how representative the library mineral compositions are of the measured mineral compositions for Martian samples. The low-albedo deposits define a fairly narrow trend of decreasing mafic oxide content with increasing silica. All of the low-albedo deposits plot within the field of possible rock compositions composed of the major rock-forming minerals (olivine, pyroxene and plagioclase) of the Martian meteorites [McSween and Treiman, 1998], suggesting that the deposits could be derived from lithologies similar to the Martian meteorites.
 Alternatively, the observed trend in Figure 9a may also represent removal of lower density minerals (plagioclase, clays, glasses) through aeolian processes leading to the concentration of the higher-density minerals (olivine, pyroxene, oxide minerals). Evidence for fractionation of heavy minerals (specifically Fe-Ti oxides) during aeolian transport was noted for the Viking and Pathfinder soils [McLennan, 2000; McSween and Keil, 2000]. In this scenario, the “starting” lithology might be similar to most silica-rich deposit (crater 18), and the removal process progressively changes the composition toward one dominated by higher-density minerals. On Figure 9a, this would result in a decrease in SiO2 and an increase in mafic oxide content, as seen in the observed trend. If this type of removal has occurred, it should be apparent on the TAS diagram (Figure 8) as well. Indeed, the compositional field of low-albedo deposits appears to follow a trend between olivine and plagioclase (Figure 8), possibly indicating that lower density lithics have been preferentially removed, leading to a concentration of higher-density olivine and pyroxene lithics in the floors of these craters. With the current data set, we are unable to distinguish the cause of the observed trend of the data. Indeed, it is also possible that the observed variation results from a combination of both primary igneous processes and subsequent alteration.
 A third possible cause of the compositional variation is aqueous alteration processes. However, olivine is commonly present in fairly large abundances, while clay minerals are commonly present in relatively low abundances (Table 3). The presence of olivine, dominated by the forsterite-rich compositions, in the low-albedo deposits suggests limited contact with aqueous solutions [Stopar et al., 2006]. In addition, the low abundance of clay minerals dominated by kaolinite and illite, which in aqueous environments form from alteration of aluminum-rich minerals (e.g., feldspars) present in felsic rocks [Deer et al., 1999], further supports limited aqueous interaction for these deposits. Furthermore, there does not appear to be a decrease in olivine with an increase in clays (Table 3, R2 = 0.2664) as would be expected if the compositional variation resulted from aqueous alteration. Therefore, it does not seem likely that the observed compositional variation resulted from aqueous alteration of the low-albedo deposits.
 If the compositional variations show some sort of pattern in how they are distributed, it might allow us to discern more about the formation and subsequent modification of these low-albedo materials. The spatial distributions of the lithologic types are displayed in Figures 3b and 7b. Figure 9b displays the spatial distribution of the data trend from most mafic (reds) to least mafic (purples). In terms of plutonic lithologies, these deposits do not display any systematic grouping, although the most olivine-rich deposits (craters 1, 9, 10 and 12) are located in the southwestern portion of the map. However, because of the lack of acceptable TES data for many craters, it is difficult to ascertain if there is a real physical separation of these lithologies. Furthermore, there do not appear to be strong groupings of the volcanic lithologies (Figure 7b) nor does there appear to be a spatial trend in the distribution in craters with increasingly greater abundances of mafic components (Figure 9b).
5.3. Origin of Low-Albedo Deposits
 Within the intracrater low-albedo deposits, we see variable dune morphologies from one crater to the next as well as within individual craters (Figures 4a, 4b, 4c, 4d, 4e, 4f, 4g and 4h). Both transverse and barchan dunes form in environments with low variation in the wind direction [Wasson and Hyde, 1983]; in the Amazonis Planitia craters, there appears to be a net sand flux toward the southwest that may result from a combination of at least two different wind directions (e.g., crater 12). In general, barchan dune forms occur where there is a lower volume of sand, whereas transverse dunes occur where the sand supply is abundant; however, barchans may also merge to form transverse dunes because of a decrease in transport of a relatively small volume of sand [McKee, 1979; Wasson and Hyde, 1983]. The presence of barchan dunes in some craters (e.g., craters 3, 9 and 12; Figures 4b, 4d and 4f) reveals an insufficient volume of sand to produce transverse dunes or increased transport of that sand (or both) within individual craters. Transverse dune forms indicate that in some craters (e.g., 18, Figure 4h) the sand supply is either currently sufficient to sustain the presence of these dunes or was sufficient at the time of dune formation, at least in the portions that have been imaged. If the dune forms are no longer being transported, the current sand supply may be insufficient to sustain the dune forms but would not result in degradation of the dune form if certain factors (e.g., wind energy, dune induration) allow the dune form to remain intact. Finally, the existence of linear dune forms (e.g., crater 12, Figure 4f) suggests that the movement of sand supply around the crater floor is affected by winds from two directions. Clearly, the intracrater wind regime within individual craters is somewhat complex, allowing materials to be transported in multiple directions within the craters. Nevertheless, the net sand flux direction for the dunes is fairly consistent with the direction of the prevailing winds in this region from the northeast [Scott and Allingham, 1976] as indicated by wind streaks in the surrounding plains.
 Previous studies of intracrater, low-albedo deposits have argued for various sources for these materials. Some studies have suggested that these deposits are derived through redistribution of mobile regional materials [Christensen, 1982, 1983; Edgett and Christensen, 1992, 1994; Fenton et al., 2003]. Possible regional sources for these deposits include Olympus Mons (∼2,700 km east and currently upwind), Apollinaris Patera (∼1,500 km south) and Elysium Mons (∼2,000 km west). However, these potential sources are all quite distant. Studies of sand transport distances on Mars have estimated the rate of particle degradation and the path lengths as a function of the particle diameter [Edgett and Christensen, 1991; Greeley and Kraft, 2001; Rogers and Christensen, 2003]. The most recent study by Rogers and Christensen  estimated that fine sand can only travel up to ∼1,000–1,800 km on Mars before it is degraded to a size that prevents saltation (<∼115 μm). No obvious sources for the low-albedo materials in the surrounding region have been identified within ∼1,800 km.
 The geologic history of the Amazonis Planitia includes erosion of ancient cratered terrain and subsequent flooding of this terrain by a thick series of (probably basaltic) lava flows from Olympus Mons. This period was followed by intense aeolian activity during which time the previous materials were eroded and buried in situ under a blanket of aeolian debris [Morris and Dwornik, 1978]. In addition, the craters from this study range from subdued to sharp-rimmed morphologies. Since they are not heavily degraded, these craters are among some of the freshest craters in this area and may postdate the lava flows [Morris and Dwornik, 1978]. Thus, the ultramafic to mafic nature of these deposits is consistent with aeolian erosion of the basaltic materials, possibly derived from the crater walls and subsequent removal of lower density phases by aeolian activity.
Bandfield et al. [2000a] identified a global dichotomy in composition using TES spectra, with surface type 1 (ST1) dominating in the southern cratered highlands and TES surface type 2 (ST2) in the northern plains. However, localized deposits (<400 km2) of TES basalt (ST1) have been identified in the northern lowlands [Rogers and Christensen, 2003], including the low-albedo deposit within Pettit Crater (crater 18 in this study, Table 1). That study interpreted the distribution of the two surface types to represent a stratigraphic relationship, with younger andesitic materials overlying the older basaltic materials in the equatorial and northern plains regions [Rogers and Christensen, 2003]. Although the low-albedo deposits from this study are consistent with low-silica (ST1 and more mafic) materials being exhumed by the impact process (perhaps from beneath a younger, more felsic unit) and exposed within the walls and floors of the crater, we have found no evidence for materials with a higher silica content similar to ST2 (Figure 8). If higher silica materials (i.e., ST2) exist within this region and the deposits are locally derived materials, the low-albedo deposits should represent some admixture of all local materials. However, it is important to point out that the low-albedo deposits probably represent the fraction of local materials that comminuted readily into sand-sized grains and remained resistant to any subsequent mechanical and chemical weathering. The ST2 spectrum has been also interpreted as a partly weathered basalt [Wyatt and McSween, 2002; Kraft et al., 2003; Ruff, 2004] that contains phyllosilicates; these phyllosilicates may not resist the mechanical weathering process as well as coexisting minerals. In that scenario, an admixture of basalt and partly weathered basalt might not display a compositional difference within the low-albedo deposits. Regardless, the composition of the low-albedo materials appears to represent only the lower silica materials and is not consistent with an admixture of lower and higher silica materials.
 In studying intracrater low-albedo deposits within 23 Amazonis Planitia craters, we have observed the following results:
 1. TES spectral modeling demonstrates that the mineralogy of the low-albedo deposits are dominated by mafic minerals with a derived bulk chemistry that ranges from ultramafic to mafic but includes more oxide minerals (specifically, magnetite) than expected. Careful inspection of the data revealed that magnetite may be erroneously included in the model to contribute a slope to the spectra in order model temperature variation within a pixel, as suggested by Schneider and Hamilton [2006a]. Therefore, it is most appropriate to exclude magnetite when calculating bulk composition and comparing these derived compositions to other Martian rocks.
 2. In the magnetite-removed models, the mineralogy of the low-albedo deposits are dominated by mafic minerals (olivine, pyroxene) with a derived bulk chemistry that ranges from ultramafic to mafic (∼40–52 wt % SiO2), with some of the lowest bulk silica content currently recognized on Mars. The bulk compositions are comparable to some Martian lithologies (e.g., Martian meteorites, MER rocks, Surface Type I) but are more mafic than the other bulk compositions (e.g., Surface Types II and Mars Pathfinder S-free rock). In addition, the low-albedo deposits appear to have higher alkali contents than other Martian lithologies at the same silica contents within the subalkaline fields. If the deposits are compositionally representative of their parent lithology/lithologies, this suggests that their parent magma(s) formed under similar mantle conditions but may have been derived from a source with slightly higher alkali contents.
 3. The observed compositional variation of the low-albedo deposits may represent the original compositional variation of the lithologies from which they were derived. However, the compositional variation may also represent one original (basaltic) composition that has been altered by aeolian removal of lower density phases (e.g., plagioclase, clays, glasses) and subsequent enrichment of higher-density (mafic) phases. Finally, the compositional variation may represent a combination of both processes.
 4. For the 11 craters studied using TES data, the spatial distribution of the compositions shows weak to absent grouping of like compositions/lithologies.
 5. The presence of dune forms in the intracrater low-albedo deposit reveals fairly unidirectional (southwesterly) regional winds transport the intracrater low-albedo deposits across the floor of the craters, consistent with the regional wind direction. The existence of both transverse and barchan dune forms indicate that the movement of sand supply around the crater floor is continuous and the spatial distribution of the low-albedo deposit may be changing over time. There does not appear to be a correlation between the dune morphology and the composition of the deposit.
 6. We found no evidence for a source of low-albedo materials outside the field of craters, within a reasonable distance from the field, although dust cover may preclude identification of a regional source. However, we have identified local sources for these materials within several craters, which may be supplied by erosion and deflation of the crater floor as well as erosion of crater walls and interior structures. Therefore, we believe that the source of the low-albedo materials in all craters is most likely local, but the sources in most craters are obscured by dust cover.
 7. The geologic history of the Amazonis Planitia includes flooding by a thick series of (probably basaltic) lava flows from Olympus Mons and subsequent intense aeolian erosion of these materials [Morris and Dwornik, 1978]. The craters from this study range from subdued to sharp-rimmed morphologies, are among some of the freshest craters in this area and may postdate the lava flows [Morris and Dwornik, 1978]. Therefore, the compositional range displayed by these deposits may represent primary igneous processes to produce a range of compositions or alteration of a single lithology through removal of lower density phases by aeolian activity or a combination of both.
 8. A global dichotomy in composition identified in TES data shows that the southern highlands are dominated by ST1 and the northern plains by ST2 [Bandfield et al., 2000a]. However, localized deposits of ST1 are found within the northern plains and were interpreted to represent a stratigraphic relationship, with younger andesitic materials overlying the older basaltic materials in the equatorial and northern plains regions [Rogers and Christensen, 2003]. If the low-albedo deposits are locally derived, they appear to represent only lower silica materials.
 We thank Deanne Rogers and Lori Fenton for their helpful and insightful comments during review for publication. This material is based upon work partially supported by the National Aeronautics and Space Administration through the NASA Astrobiology Institute under Cooperative Agreement NNA04CC08A issued through the Office of Space Science and the Mars Data Analysis Program grants NNX06AE15G, NAG5-13455, and NAG5-12506 to F.S.A.