Iron mineralogy of the surface of Mars from the 1 μm band spectral properties

Authors


Abstract

[1] We study the 1 μm absorption from OMEGA/MEX spectra to map Martian iron mineralogy at a global scale. This band is covered on the left by the VNIR (visible and near infrared) OMEGA channel and on the right by the SWIR (short wavelengths infrared) one. We first perform a systematic spatial coregistration of the two channels after an improvement of the VNIR radiometric calibration. The update of the VNIR Instrumental Transfer Function (ITF) and the internal stray-light estimation is based on the spectra of the Phobos red units and of the water ice north polar cap of Mars, which have been fitted according to an iterative process. The level of the signal in the blue wavelength range, previously systematically overestimated due to a stray-light residual and the general shape of the spectrum forλ > 0.7 μm are improved . Global maps of the 1 μm signature have been derived from 9 new spectral indices. The largest values of the 1 μm band integral are found in Noachian terrains and in the dunes around the north polar cap. In the south polar region, an area centered at ∼155°W and ∼83°S is mapped as a distinctive spectral unit, dominated by pyroxene. The northern lowlands of Mars together with other dark terrains located in the northern hemisphere show very low values of some spectral indices due to the negative spectral slope in the NIR. This behavior is consistent with the presence of weathered basalts with a possible glassy or amorphous component. Among the hydrated terrains, the only ones that can be isolated by studying the 1 μm band are those located in Terra Meridiani, Aram Chaos and Capri Chasma, enriched in sulfate and hematite. On the other hand, the sulfates of the dark dunes surrounding the northern polar cap and the phyllosilicates of the bright hydrated deposits of Mawrth Vallis cannot be isolated combining the parameters used in this study. This suggests that their distinctive mineralogy does not affect the 1 μm band, remaining similar to the global Martian average shape.

1. Introduction

[2] This work is based on the recovery of the whole 1 μm absorption from OMEGA (Observatoire pour la Minéralogie, l'Eau, les Glaces et l'Activité [Bibring et al., 2004]) spectra. OMEGA is the imaging spectrometer on board the ESA mission Mars Express (MEX) in orbit around Mars since December 2003 and still operating. OMEGA is composed of three channels: the VNIR channel working in the visible-near infrared wavelengths (0.35–1.05μm), the SWIR channel operating in the 0.92–2.70 μm range and the LWIR channel covering the 2.7–5.1 μm one. They do not observe exactly the same areas at the same instant due to their slight misalignment and different point source response (physically characterized by the point spread function, PSF hereafter). Previous global investigations of the 1 μm band [Mustard et al., 2007; Mangold et al., 2007; Poulet et al., 2007, 2008; Pelkey et al., 2007] were based on the use of spectral parameters derived from each channel separately. Our aim is thus to study at a global scale the Martian iron mineralogy through the 1 μm absorption by combining the VNIR and the SWIR data sets.

[3] Several mineral species of interest for Mars have an absorption signature in the 1 μm wavelength range:

[4] 1. In near-infrared reflectance spectra, olivine spectra are identifiable from a complex, strong absorption at about 1μm that is a composite of three overlapping bands due to the presence of Fe2+ [Singer, 1981; Burns, 1993]. This absorption moves toward longer wavelengths with increasing fayalite percentage (i.e., Fe content) and grain size [Burns, 1970; Adams, 1974; King and Ridley, 1987; Mustard and Hays, 1997; Poulet et al., 2009; Clénet et al., 2011]. On Mars, olivine has been found at the Spirit and Opportunity landing sites [Christensen et al., 2004; Morris et al., 2004] and in many other locations [e.g., Hoefen et al., 2003; Poulet et al., 2007; Mustard et al., 2005; Ody et al., 2012].

[5] 2. Pyroxenes [(Ca, Fe, Mg)Si2O6] are among the most common silicate minerals in both the upper crust and the surface of Mars. In visible and near-infrared reflectance spectra, they are dominated by two main broad absorptions near 1 and 2μm [Hunt and Salisbury, 1970; Adams, 1974; Klima et al., 2007]. Pyroxenes are divided into two types on the basis of calcium abundance and crystallographic structure: low-calcium pyroxenes (LCP, orthopyroxene <5% mol% Wo, orthorombic) and high-calcium pyroxenes (HCP, clinopyroxene, >5% mol% Wo, monoclinic). The minima of both absorptions change as a function of concentration of Fe2+ and Ca2+. In general, with increasing Fe and Ca content, both the absorptions shift to longer-wavelength [Adams, 1974; Cloutis and Gaffey, 1991]. Several factors can control the intensity of the absorptions including the concentrations of the cations and the asymmetry of the crystallographic sites [Burns, 1993; Klima et al., 2007] as well as the grain size. Pyroxenes have been mapped in the dark cratered terrains of the southern hemisphere and Syrtis Major [e.g., Bandfield et al., 2000; Bibring et al., 2006; Poulet et al., 2007].

[6] 3. Iron oxides also exhibit several signatures in the 0.4–1.4 μm wavelength range. The spectra of some ferric species are characterized by an intense absorption edge extending from 0.40 to 0.75 μm due to the Fe3+ electronic transitions [Poulet et al., 2007]. Ferric nanophase iron oxides are one of the most common species distributed on Mars and cover mostly the bright regions [Bibring et al., 2006]. The strong correlation with the spatial distribution of the dust suggests that these oxides can be used as an indicator of dust contamination [Poulet et al., 2007]. Other ferric oxides, in the form of crystalline hematite are present in the region of Terra Meridiani [Christensen et al., 2000].

[7] 4. Several iron-bearing hydrated minerals have been identified in restricted sites. Both Fe-bearing sulfates and phyllosilicates are the major phases found on Mars [e.g.Gendrin et al., 2005; Poulet et al., 2005]. These minerals are identified due to a 1.9 μm hydration band together with other minor bands depending on the type of hydrated mineral. The study of the 1 μm band can better constrain their oxidation state and their iron content.

[8] A better understanding of the spatial distribution and the iron content of these different mineral phases can thus be obtained by investigating the spectral characteristics of the 1 μm band.

[9] In the first part of this paper (section 2) we describe the approach adopted to refine the VNIR radiometric calibration, and we present the method to coregister the VNIR and SWIR channels. In the second part (sections 3 and 4) we describe the spectral indices used to study the 1 μm band, and we discuss the implications of their distribution on the iron mineralogy of the surface of Mars.

2. Method

[10] In this section, after a brief explanation of the need for VNIR calibration reassessment (section 2.1), we describe the procedure followed to update it (section 2.2) and the method implemented for coregistering the VNIR and SWIR channels (section 2.3). The latter step is essential to recover the whole spectral region where the channels overlap, allowing study of the 1 μm absorption.

2.1. VNIR Calibration Issues

[11] Before integrating the OMEGA instrument on the MEX spacecraft, an extensive on ground calibration campaign was performed [Bonello et al., 2005]. At that time the VNIR instrumental transfer function (ITF) was not completely defined because the signal of the blackbody used as the calibration source caused a saturation problem for λ < 0.75 μm. Moreover, the VNIR channel operates in a pushbroom mode with the slit defining a field of view (FOV) of 8.8° and, due to internal parasitic reflections, the signal measured by VNIR on a single spatial pixel (instantaneous field of view IFOV ≈ 1.2 mrad) is contaminated by the light coming from the whole FOV. The internal stray-light intensity and behavior as a function of wavelength depends both on the Mars albedo and the instrument optical design (i.e., stray-light goes on the sensor after internal reflections either interacting with the grating or not).

[12] In the laboratory the stray-light for a source out of the field of view (external stray-light) was well-tested, although stray-light inside the instrument itself was never estimated. Thus, when the spacecraft arrived at Mars, it was necessary to recover the ITF forλ < 0.75 μm, validate it for λ > 0.75 μm and evaluate the internal stray-light contamination. For this purpose, we first recovered the ITF-over the whole VNIR spectral range by using previous Phobos and Martian spectra [Bellucci et al., 2006]. Then, keeping frozen the retrieved ITF, an estimation of the internal stray-light contamination (assumed to be composed by one component) was performed by minimizing the radiance residual at 1μm between the VNIR and SWIR channels for both dark and bright Martian regions.

[13] By matching the VNIR and SWIR channels on a global scale to study the 1 μm band, we noticed a slight (<10%) but systematic mismatch of the radiance levels between the two OMEGA channels still persisting after the spatial coregistration, detailed in section 2.3. We also noticed that the VNIR Phobos spectra exhibited the presence of a spurious absorption centered at about 0.85 μm in both red and blue units. The strength of such a band was a function of the intensity of the signal itself, and it disappeared when rationing the spectra to spectra with the same radiance level.

[14] This anomalous behavior has led us to develop a new VNIR radiometric calibration. The results shown below clearly indicate that it is more reliable than the previous one. In the new approach, the stray-light is assumed to be composed of two additive terms, and spectra of the Phobos red units and Martian northern perennial cap have been fitted in a iterative process in order to retrieve the updated ITF. In this way we have avoided the use of data acquired by other instruments. They are in fact characterized by different fields of view and can introduce, if not perfectly calibrated themselves, spurious residuals. More details on the method followed to converge to the final estimation of new ITF and the contamination due to the two stray-light components are given in the next section.

2.2. VNIR Calibration Updating: Process and Results

[15] In the previous calibration version, we assumed the VNIR internal stray-light to be proportional only to the dispersed signal of the whole FOV, containing each pixel of the slit. In the new procedure reported here, we followed a different approach, and we have assumed that the stray-light is composed of two additive components (SL = k1 × SL1 + k1 × SL2):

[16] 1. The first component (SL1), after some internal reflections, goes directly to the CCD detector without interacting with the grating. As a first approximation, it can be modeled as a constant term (in each spectral channel) proportional to the signal integrated over the entire VNIR spectral range and 8.8° field of view;

[17] 2. The second component (SL2) interacts with the grating and is then proportional (in each spectral channel) to the Martian spectrum as seen by the whole slit.

[18] The signal from Mars SMARS(c0, x0, y0) in a given VNIR spectral channel c0 and pixel (x0, y0) of the image can be retrieved by removing the total stray-light contamination from the VNIR signal SVNIR(c0, x0, y0):

display math

with:

display math
display math

SL1 and SL2 terms are calculated for each cube, while k1 and k2 coefficients are the same for all the cubes. They were estimated in an iterative process as described in the following, together with the new ITF. The flowchart that summarized each step we followed in the new approach is shown in Figure 1.

Figure 1.

Flowchart of the whole process followed to refine the VNIR calibration as described in Sec. 2.2. k1 and k2are the coefficients of the linear combination of the two stray-light components (SL1 and SL2 respectively, SLindicates their total contribution). They were recomputed for each i-th iteration together with the intermediates ITFs, before converging to the final values ITFfin, k1 fin and k2 fin. The NP Martian spectra refer to the water ice spectra of the Northern polar cap. In STEP 4, the slope of SL1 was added as free parameter.

[19] First of all (STEP 1), we made an estimation of the stray-light upper limit, assuming that the stray-light contamination is interlay due to the first component:

display math

For this purpose, we considered only the spectral range from 0.36 to 0.40 μm (sampled in the first 5 VNIR spectral channels), where the detector sensitivity as well as the Martian expected signal are sufficiently low to justify the assumption that the lowest signal measured in these wavelengths is given by the VNIR noise plus the internal stray-light contamination. In fact, between 0.36 and 0.40μm the VNIR ITF is approximately 10% of its maximum, occurring at 0.70 μm. Moreover, in this range the Mars surface has its deepest Fe absorption and reflectance from Mars is about 20% of that at 0.70 μm in the case of the bright regions. The lowest signal measured from 0.36 to 0.40 μm can be thus taken as an estimation of the total stray-light upper limit.

[20] To derive the stray-light upper limit, we have considered a large number of OMEGA cubes, representative of the whole data set, excluding observations of the night side, deep space and limb. The minimum signal found in the single cubes ranges between 3.8 × 10−4% and 9.5 × 10−4% of the total signal collected by the whole CCD, composed of 128 × 96 pixels. These values correspond to about 4.7% and 11.7% of the signal when considering single CCD pixel. The lowest value was used in equation (4) as kupin order to compute the stray-light upper limit.

[21] Once the stray-light upper limit was evaluated, a preliminary version of the new VNIR ITF (before iterating the process) was retrieved by fitting the Phobos spectra from the red units between 0.45 and 0.76μm and the Martian spectra from the water icy North Polar Cap between 0.75 and 1.0 μm (STEP 2). Phobos spectra of the red units were chosen to start the process because they are featureless. Phobos blue unit data were not used because they are characterized by a flat spectrum after 0.65 μm [Murchie and Erard, 1996]. A quadratic fit was performed in the spectral region between 0.45 and 0.76 μm, then it was extended to the whole VNIR range (see Figure 2a). The UV-blue range was excluded in the fit in order to avoid a real feature that characterizes the Phobos spectra, confirmed also by the SPICAM spectra [Bertaux et al., 2011], the origin of which is still under investigation. The NIR interval has been avoided due to the spurious band appearing there with the previous calibration procedure. For comparison, in Figure 2b we also reported the same Phobos mean spectrum as derived with the final version of the new VNIR calibration procedure. In order to improve the fit in the range where the uncertainty in the retrieved Phobos spectra is the highest, the Martian spectra between 0.75 and 1.0 μm from the water icy North Polar Cap have been also fitted (see Figure 3). The water ice spectrum is flat and featureless [Warren, 1984] in the considered range. Fine particles of the airborne dust can introduce a negative slope in the Martian spectra [e.g., Vincendon et al., 2007], but they do not introduce deep bands around 1 μm. We assumed the difference between the measured spectra and fit was due to stray-light, and we have used data during the Northern summer with a low expected dust contamination. At this stage, in the range of wavelengthsλ < 0.45 μm we used the previous ITF, retrieved as described in Bellucci et al. [2006] by the Phobos 2 measurements [Murchie and Erard, 1996].

Figure 2.

(a) The black curve (and corresponding standard deviations) is the result of the average of 164 VNIR spectra (with comparable radiance factor at 0.7 μm) computed with the previous calibration and extracted from the Phobos redder unit. The absorption at bluer wavelengths is real, while the one in the NIR range is a residual of the calibration. The red line is the quadratic fit of the spectrum as derived using the spectral range between 0.45 and 0.76 micron (dotted lines) and extended to the whole VNIR range. (b) Comparison between the mean spectrum shown in Figure 2a (thin curve) and the new mean spectrum (thick curve and corresponding standard deviations) computed with the final, updated calibration procedure.

Figure 3.

Thin curve (0.95–1.85 μm): SWIR spectrum. Thin curve (0.35 and 1 μm): VNIR spectrum from the North polar cap (cube 902_1, Ls = 95.3) as derived from the previous calibration. Thick curve: VNIR spectrum as derived from the new calibration at the first iteration, after the fit of the Phobos spectra. Red thick curve: fit of the new Martian spectrum between 0.72 μm 0.85 μm and then extended up to 1 μm. Green thick curve: VNIR spectrum as derived from the new calibration at the end of the iterative process, with ITFfin, k1 fin and k2 fin.The VNIR-SWIR match at 1μm has been used to define the coefficients of the two stray light components.

[22] In STEP3, we introduced both stray-light components. The coefficients of the linear combinations were retrieved by minimizing the reflectance differences between the VNIR and SWIR channels at 1μm for several cubes. We used cubes that had already been spatially coregistered (see section 2.3). The new VNIR ITF was retrieved by fitting iteratively the Phobos spectra from red unit and the Martian spectra from the water icy North Polar Cap. In each iteration we also refined the coefficients of the two stray-light components. The two OMEGA channels overlap in the wavelength range between 0.92 and 1.07μm, covered by the last VNIR spectral channels (with a sampling of 7 nm) and by the first SWIR ones (sampling = 15 nm). Because the VNIR signal drops down after 0.98 μm (spectral channel 84), we compared the VNIR and SWIR reflectance at 0.97 μm, where the standard deviations of the mean signal from both OMEGA channels are comparable. The considered wavelength is measured by VNIR spectral channel 82 and by SWIR spectral channel 4. The reflectance differences were computed at this fixed wavelength after VNIR/SWIR spatial coregistration, as described in section 2.3. At this stage, we refined the new ITF also for λ < 0.45 μm taking into account the fit of the Phobos spectra retrieved between 0.45 and 0.76 μm and extended to longer wavelengths.

[23] After several iterations we found that it was hard to converge to a unique ITF in the 0.85–1.00 μm range that worked properly for both the Martian dark and bright terrains. Dark and bright regions on Mars are characterized by a different shape in the 1 μm band, as discussed further in the second part of this paper (sections 3 and 4). In general, in the dark regions the absorption is deeper, with variable concavity on both the VNIR and SWIR sides of the band. On the other hand, the bright regions can present a weaker band that can be linearly fit to first order on the VNIR side (∼0.8 < λ < 1.0 μm). For this reason, a combination of NIR ITF values and stray-light coefficients, which apparently works properly for bright regions, can produce incorrect results when tested on dark terrains, and vice versa. In order to get through this difficulty, a further step has been included (STEP4), adding the SL1 slope as a free parameter.

[24] STEPs 3 and 4 have been iterated on several OMEGA cubes acquired of dark, bright and mixed terrains, as well as with different swath widths (mainly 128, 64 and 32 pixels) and different integration times. We considered the iterative process concluded when a unique ITF for both dark and bright regions was found without introducing sudden divergences in the signal in the highest or lowest signals in the scene, rather than fixing a unique threshold for the offset value at 1 μm. This because the reflectance differences in the overlapping λ range can vary slightly from one cube to another depending on several factors (e.g., signal level, integration time, etc.). Furthermore, a good match at 1 μm in one cube on the average does not guarantee a physical-meaning shape in the single spectrum of some particular terrains, especially those rich in olivine.

[25] At the end of the process we found that the weights of the first and second stray-light components were respectively approximately of 66% and 34%, and the SL1 slope about 5%. The influence of the first component is stronger in the blue range, while the second one particularly affects the NIR side. The final ITF is shown in Figure 4a. Its uncertainty is better than 5% between 0.45 and 0.85 μm, and it increases up to 20% at the edges of the range due to the decrease of the SNR and the uncertainty in the stray-light contamination. As already mentioned, the VNIR sensitivity after spectral channel 84 (λ = 0.98 μm) becomes significantly lower than 5% of its maximum, and the corresponding data must be discarded. The ratio between the new ITF and the previous one is shown in Figure 4b. Their differences are about 2% for 0.44 < λ < 0.76 μm, and they increase up to 5% for λ < 0.44 μm and 0.76 < λ < 0.94 μm, while they diverge for λ > 1 μm. The ITFs ratio shows clearly that the presence of a band at such wavelengths in the Phobos spectra was due to a calibration residual, which influenced also the previous shape of the VNIR Martian spectra. Moreover, since for λ > 0.85 μm the ITF decreases very rapidly, small changes for the ITF in this range can produce strong effects in the appearance of the Martian spectra. For this reason it is hard to determine if the 5% slope that we introduced in stray-light component SL1 is a mathematical artifact to permit convergence using this method or if it is a real instrumental effect, due to variation as a function of the wavelength in the efficiency of the internal parasitic reflections.

Figure 4.

(a) The new VNIR ITF. (b) Ratio between the new VNIR ITF and the old one.

[26] Examples of the new VNIR spectra are shown in Figure 5. The general shape of the spectrum is clearly improved in comparison with the old VNIR calibration, in particular for the shortest wavelengths. The spectral match between the VNIR and SWIR channels also appears improved. Figure 6 compares the histograms of the I/F difference (in percentage) at 0.97 μm for the new calibration and for the previous one. In both cases the VNIR and SWIR channels have been spatially coregistered. The new histogram is now centered on the zero and Gaussian shaped, although with longer wings in comparison with a Gaussian fit. The wings correspond to pixels located at the edges of craters and valleys, or at mineralogical unit boundaries, where the differences between the PSFs of the VNIR and SWIR channels have a stronger influence.

Figure 5.

Comparison of VNIR spectra as derived from the previous calibration procedure (thin line) and the new one (thick line) on dark (Ls = 36.1°), bright (Ls = 33.4°) and polar icy regions (Ls = 97.2°).

Figure 6.

Histogram of the I/F differences at 0.97 μm between the SWIR and VNIR channels for cube 422_4 on the Nili Fossae region. Thick curve: new calibration results. Thin curve: old one. In both cases the VNIR and SWIR channels have been spatially coregistered.

[27] The new VNIR calibration procedure is included in OMEGA software version SOFT08 and will be made publicly available at the ESA web site: http://www.sciops.esa.int/index.php?project=PSA&page=mex.

2.3. VNIR–SWIR Coregistration and Mapping

[28] Due to a slight optical misalignment and different PSFs of the VNIR and SWIR channels, they do not observe exactly the same area at the same instant. As a result, the VNIR and SWIR footprints corresponding to the same pixel are not geographically coincident. Therefore, the following spatial coregistration procedure has been applied:

[29] 1. The VNIR image is shifted with respect to the SWIR image in each spatial direction X-Y (along-across track, respectively) and for the range of ±7 pixel with a step of 1 pixel;

[30] 2. For each step a new matrix containing the difference between the two superimposed images is calculated at 0.97 μm;

[31] 3. For each matrix the standard deviation is calculated. The minimum of these standard deviations corresponds to the best spatial coregistration between the two channels.

[32] A new OMEGA cube is built on the basis of the coregistration, losing the data at the edges of the cube where the two channels do not overlap. The shift in the X direction is usually 3–5 pixels, and in the Y direction is about 4–6 pixels, depending on the spacecraft velocity. This method does not take into account the geometric distortion due to the different VNIR and SWIR optical aberrations. The global maps are generated with a resolution of 0.14°/pixel corresponding to ∼8.5 km/pixel at the equator.

[33] After the spatial coregistration, a residual difference in the I/F at 1 μm between the two channels can still remain due to the difference in the PSF's. To avoid this problem, each VNIR spectrum is rescaled to the value of the I/F at ∼1 μm of the corresponding SWIR spectrum. The exact wavelength at which the two OMEGA channels are matched, inside the overlapping range between 0.92 and 0.98 μm, is selected for each spectrum on the basis of the spectral channels with the lowest standard deviation with respect to the smoothed spectra. The difference |ΔR| between the VNIR and SWIR reflectance at ∼1 μm is usually 1–2% and only about 1% of data show a |ΔR| > 10%. The spectra with the largest mismatch between the two channels are located along the walls of the craters and valleys where the influence of PSF differences is expected to be larger. In this study the data with ΔR > 10% are discarded.

3. Spectral Parameters

[34] In general, the VNIR spectrum of Mars is dominated by an intense and featureless ferric absorption edge from UV (∼0.40 μm) up to a marked reflectance peak at 0.68–0.85 μm with decreasing reflectivity toward longer wavelengths. On a global scale, after the removal of CO2 atmospheric absorptions, the OMEGA spectrum in the SWIR is generally characterized by two important absorptions. The first one is at about 1 μm and the second one between 2.00 and 2.30 μm, depending on the mineralogy. The 1 μm absorption is covered by both OMEGA channels, the left wing by the VNIR and the right one by the SWIR. Therefore, the reflectance maxima defining the continuum of the 1 μm band at shorter and longer wavelengths falls in the VNIR and SWIR channels respectively.

[35] Each spectrum of the surface is the result of areal or intimate mixing of different minerals, and the resulting spectral reflectance properties are a complex combination of the spectra of each mineral end-member. Several studies have shown that it is possible to explore Martian mineralogical diversity utilizing specific spectral parameters [Bell et al., 2000; Murchie et al., 2000; Christensen et al., 2000; Mustard et al., 2007; Mangold et al., 2007; Poulet et al., 2007, 2008; Pelkey et al., 2007].

[36] In order to extract compositional information from the spectra, we have introduced some spectral indices and studied the correlations between them all over the OMEGA data set. They are the band integral (integrated band depth) around 1 μm (BI1000), the height of the 1 μm left/right shoulders (SHVNIR1000, SHSWIR1000), the band width at 1 μm (BW1000), the relative height of the peak at 0.685 μm (RHP685) and a spectral parameter that takes into account the relative spectral ratio between 1.743 and 1.543 μm with respect to the ratio between 1.300 and 1.213 μm (OLVINDEX). As complementary parameters, we also discuss the reflectance peaks in both VNIR and SWIR ranges, the ratio of their magnitudes (COATINDEX) and their corresponding wavelength positions. Herein we refer to the reflectance peak in the visible channel and in the near infrared channel as VNIRpeak and SWIRpeak, respectively, and in term of wavelength position as λVNIRpeak and λSWIRpeak. We indicate reflectance with the symbol R. The exact formulation of the spectral indices is given in Table 1. In particular, BI1000 is computed as the area enclosed for λVNIRpeak < λ < λSWIRpeak, i.e., between the spectrum and the line that connects VNIRpeak to SWIRpeak.

Table 1. Summary of OMEGA Spectral Indexes Used in This Studya
NameSpectral IndexFormulaDetection ThresholdRationale
  • a

    NPD, northern polar dunes.

  • b

    Sulfate of Aram Chaos and Capri Chasma, hematite of Terra Meridiani.

λVNIRλ position of the peak in the VNIR-<0.77weathered basalt/glass
>0.88nanophase ferric oxide
λSWIRλ position of the peak in the SWIR->1.48HCP, olivine, hematiteb
RHP685relative height of peak at 0.685 μmR685/(0.5*R625 + 0.5*R782)<1.033nanophase ferric oxide
>1.038mafic and weathered basalt/glass
BW1000band width at 1 μmλSWIRpeak -λVNIRpeak0.70–0.88pyroxene (favors HCP), hematite
>0.88olivine
BI1000band integral at 1 μm math formula<0.015nanophase ferric oxide
>0.030mafic terrains, NPD, sulfateb
SHVNIR1000shoulder height at 1 μm in the VNIRRVNIRpeak/R937<1.01nanophase ferric oxide
>1.09pyroxene (favors HCP), olivine, NPD
SHSWIR1000shoulder height at 1 μm in the SWIRSWIRpeak/R1124>1.09pyroxene (favors HCP), olivine, sulfate,b hematiteb
COATINDEXcoating indexSWIRpeak/VNIRpeak1.08–1.16nanophase ferric oxide
R1743/R1543>1.16sulfate,b hematiteb
OLVINDEXolivine indexR1300/R1213>0.99olivine

[37] The main difference between the listed parameters (excluding OLVINDEX and RHP685) with respect to those usually reported in literature is that the spectral channels used for their computation are not at fixed wavelengths but they are calculated for each OMEGA spectrum. In fact, λVNIRpeak, λSWIRpeak, and consequently the positions of VNIRpeak and SWIRpeak, can vary from one pixel to another according to mineralogy.

[38] Global maps of the parameters summarized in Table 1 are shown in Figure 710, 12, 13, and 1517. They are built by making a mosaic of single OMEGA orbits. The observations have different solar longitude, local time and spatial resolution. We have restricted our analysis to the orbits with widths of 64 and 128 pixels since they have the wider spatial coverage. Spectra with H2O and CO2 ice absorptions have not been taken into consideration because they could cause confusion in the data interpretation. In order to select/discard these spectra, we have used the method of Carrozzo et al. [2009] for H2O and Poulet et al. [2007] for CO2 ices. CO2 atmospheric contribution is removed by following the method in Langevin et al. [2005]. The maps are built by selecting the orbits with a distance from the spacecraft to the planet of <5000 km. Moreover, since we are analyzing both VNIR and SWIR data sets, to reduce the processing time we have used the initial part of the OMEGA data set up to orbit #2500 corresponding to the beginning of year 2006. The data affected by the global dust storm, occurring 1 year later, are thus excluded.

Figure 7.

λVNIRpeak. over MOC albedo map. The maps are divided in three parts: North Polar Region (+65° < Lat < +90°) South Polar Region (−90° < Lat < −65°) and central region (−65° < Lat < +65°). For each map the scale is indicated. The color bar is not linear in order to underline the different spectral units.

Figure 8.

Same as Figure 7 except for parameter λSWIRpeak.

Figure 9.

Same as Figure 7 except for parameter COATINDEX.

Figure 10.

Same as Figure 7 except for parameter BI1000.

[39] As far as the poles are concerned, the north polar maps (Lat > 65°N) are derived by using the data with solar longitude 103° < Ls < 110°, corresponding to the early northern summer, while the south polar maps (Lat < 65°S) are derived by using those with 277° < Ls < 280° (early southern summer). The Ls range has been chosen in order to have a good coverage in a short range of time during the retreat of polar caps.

[40] Spectral properties are influenced by a set of residual atmospheric/instrumental effects or factors such as the presence of dust aerosols, thin water ice clouds, and surface coating materials. Also, due to dynamics of the surface environment, intimate or areal mixing of surface materials can change from orbit to orbit. For these reasons, it is not possible to use the parameters as a measure of the abundance of the minerals even if it is possible to use them as qualitative indicators of their presence [Pelkey et al., 2007]. In the same way, the spectral parameter thresholds discussed in this work are not compared in an absolute sense because they are sensitive both to the coregistration and the parameter calculation procedures. Thresholds have been evaluated through a visual check of the spectra at several locations showing similar values of the indices used in this study. Some parameters are not sufficient alone to identify a specific mineral. For this reason, it is also necessary to put them by means the use of scatterplots to get additional clues on the mineralogy.

4. Results and Discussion

4.1. The λVNIRpeak and λSWIRpeak Parameters

[41] The λVNIRpeak parameter varies between 0.68 and 0.85 μm, with the lowest values of λVNIRpeak observed in the dark terrains of the ancient Noachian crust and in the northern hemisphere lowlands, as well as in the dark dunes around the North Polar Cap (Figure 7). As the surface albedo increases, λVNIRpeak shifts toward higher wavelengths. The highest values are found in the high albedo regions characterized by the presence of nanophase oxides [Bibring et al., 2006; Poulet et al., 2007].

[42] As far as the λSWIRpeak map is concerned (Figure 8), the atmospheric aerosols (both dust and ice particles) seem to have a stronger influence in this spectral range in comparison to the visible counterpart. In fact, the spectral negative slope introduced by the dust [e.g., Vincendon et al., 2007] can mask the 1 μm band and shift the λSWIRpeak to shorter wavelengths. As a consequence, spectra acquired at different Ls, but covering same area show different λSWIRpeak values, introducing further confusion in the λSWIRpeak map. This parameter varies between ∼1.24 and ∼1.85 μm. The lowest values are located in dark areas of the northern hemisphere, for example in Acidalia Planitia. Values of λSWIRpeak > 1.48 μm map olivine, hematite, HCP and iron-rich LCP with the highest values coinciding with the occurrence of olivine. Usually, LCP with a lower content of iron peaks at lower wavelengths, similar to the ones of the nanophase oxides of the bright terrains.

4.2. COATINDEX

[43] The SWIRpeak/VNIRpeak (herein called COATINDEX) correlates linearly with the strength of the ferric absorption edge at 0.55 μm [Poulet et al., 2007]. Because the magnitude of this spectral feature measures either the degree of oxidation and the amount of dust contamination on the surface [Bell et al., 2000; Morris et al., 2000; Farrand et al., 2006], COATINDEX also can be considered as a parameter representative of both these effects, with the exception of the spectra of Terra Meridiani, Aram Chaos and Capri Chasma that exhibit intermediate values of the ferric absorption at 0.55 of Poulet et al. [2007] together with the largest COATINDEX values (Figure 9). In Terra Meridiani and Aram Chaos, TES instrument revealed the presence of hematite [Christensen et al., 2000]. The Opportunity Rover landed in the first location and showed that hematite was in the form of spherules of the size of 0.6–6 mm [Herkenhoff et al., 2004], hence extreme COATINDEX values can be indicative of a bigger grain size.

[44] Dark regions are characterized by SWIRpeak lower than VNIRpeak (COATINDEX < 1). The lowest values are found within the northern lowlands, whose spectral behavior is discussed with more detail in section 4.7. Values lower than 1 are also found within southern highlands Noachian terrains and Syrtis Major, where mostly pyroxenes have been mapped by means of other parameters. Interestingly, library spectra of pyroxenes are characterized by SWIRpeak/VNIRpeak > 1. An explanation of the fact that SWIRpeak/VNIRpeakshows values <1 when measured on Martian dark terrains could be due to the wavelength-dependent transparency of a thin ferric oxide coating. In fact the dust is redistributed on global scale [Yen et al., 2005] with the result that most of the terrains have a dust coating. Furthermore, a ferric rind could be formed with time because of the reactions with external agents [Squyres et al., 2006; Skok et al., 2010]. In ferric minerals the radiation penetration increases with wavelength [Singer, 1980]. Therefore, the ferric minerals are strongly absorbing in the ultraviolet-visible spectral region but very transparent in the near-infrared one. As a result, in the ultraviolet-visible spectral region, when considering an oxidized thin layer over an unoxidized material (for example a rind or coating of ferric-bearing material over a mafic rock, i.e., an oxide over a pyroxene), the spectrum is dominated by the bright ferric oxide. On the other hand, in the near-infrared range, the spectrum is controlled by the light returned from the underlying unoxidized substrate material [Singer and Roush, 1983; Bell et al., 2000]. The resulting effect on the spectrum is a decrease of reflectance in the near-infrared when the substrate material is darker than the coating. The thicker the rind/coating, the more its properties will influence the spectrum of the underlying rock [Morris et al., 2000]. Also the presence of the dust in the atmosphere can produce similar effects [Vincendon et al., 2007].

4.3. BI1000

[45] The global map of the BI1000 parameter is shown in Figure 10. Between 65°S and 65°N, the mafic terrains exhibit the highest values. This behavior is principally due to the shape of the left shoulder of the band, which decreases in magnitude in the terrains enriched in ferric phases. However, BI1000 is not diagnostic of a given mafic composition, because the increase of the band integral can also result from large grain size [King and Ridley, 1987] so that a spectrum with low mafic mineral content and large grain size can be similar to a spectrum with high mafic mineral content and small grain size. For this reason, BI1000 is not able to distinguish between the fayalite/forsterite olivine end-members. This parameter alone is not able to separate olivine from pyroxene. For this purpose, it is necessary to compare BI1000 with another spectral index in order to distinguish these mineral classes (seesection 4.9).

[46] In the north polar region the largest values of BI1000 are associated with the dark regions of the circumpolar dune field as well as with the dark area between 0 and 90°W. In the south polar region, an area centered at ∼155°W and ∼83°S shows a mean value of BI1000 of about 0.035. It is ∼60% higher than the average value of other South Polar regions. This area, near the perennial South Polar Cap and never mapped before as a distinct spectral unit, is found inside one of the flattest regions of Mars as shown in Figure 11 (left). In order to discriminate the mineralogy of this specific area, we selected the following regions of interest (ROIs): the red one, taken inside the smother terrains and showing large BI1000 values of about 0.035, the green one, taken inside one crater at lower latitude (LON = 274°W LAT = 68°S) with similar BI1000 values, and finally the blue one, chosen outside the spectral units mapped on the flat terrains with BI1000 values of about 0.020, typical of the other south polar regions. The mean spectra derived from these ROIs are shown in the right panel of Figure 11a. We used the blue spectrum as denominator in the ratios with the green and red spectra. Results are shown in Figure 11b, where the ratio spectra are compared with a laboratory spectrum of an iron-rich LCP (Fs90% - Wo10%), which match both the minimum of the band at 1μm and the maximum at about 1.5 μm. Differences between the OMEGA ratio spectra appear in the visible range and in the right shoulder of the 1 μm absorption.

Figure 11.

(a) Here we show the mean spectrum taken in the flattest area (in red) of the South Polar Region compared to the mean spectrum of other terrains. The red ROI falls in the terrains where some spectral indices like BI1000, RHP685 and SHVNIR1000 have values higher than in the other South Polar areas (see Figure 10, 15, 16). The magenta curve is the contour that delimits the area with BI1000 ≥ 0.03. The ROIs are superimposed on the topography map on the left based on data from the Mars Orbiter Laser Altimeter (MOLA) (http://geopubs.wr.usgs.gov/). (b) Rationed spectra from Figure 11b. The absorptions near 1 μm and in the range 2.1–2.2 μm are consistent with iron-rich pyroxene taken from Relab spectral library (file ID: DL-CMP-088), reported in the bottom part.

[47] The region seems to correspond to the unit Apb, namely bedded plains material, interpreted to be aeolian deposits of young age [Scott and Carr, 1978]. Because LCP pyroxene is preferentially found in old terrains [Mustard et al., 2005] with a decrease in abundance with younger units [Poulet et al., 2009], it is unusual to find this mineralogy in terrain of Amazonian age. Another possible explanation for the nature of the red spectrum in Figure 11a is that it is a mixture of LCP and HCP that could also explain the position and the shape of the pyroxene bands at 1.0 and 2.2 μm.

4.4. BW1000

[48] This parameter has been defined to study olivine and, more generally, the distribution of mafic terrains distribution (Figure 12). Since the λSWIRpeak moves to longer wavelengths as the olivine concentration increases [Singer, 1981; Cloutis and Gaffey, 1991], BW1000 is much higher for the olivines than for any other minerals of interest for Mars. λSWIRpeak position has a significant effect on BW1000 with respect to the λVNIRpeak since the latter varies in a shorter range of wavelengths. As already mentioned, olivines show the highest values for the λSWIRpeak, and thus they are also characterized by the largest BW1000 values. In principle, since a positive correlation exists between iron content and BW1000 [King and Ridley, 1987; Mustard et al., 2007], this spectral index could be used to determine the composition of olivine, discriminating between forsterite (Fo90) and fayalite (Fo10). But the same effect can be given by different grain size in shaping the olivine 1 μm band. The presence of other minerals in the mixture, for instance pyroxene, can shift the λSWIRpeaktoward wavelengths uniquely diagnostic of the forsteritic composition, not allowing the use of BW1000 to discriminate between the types of olivine. Nevertheless, BW1000 can be used to distinguish pyroxenes from the olivines, the former exhibiting lower values compared to the latter. Because BW1000 value increases with increasing the calcium content, we can also use this spectral index to discriminate HCP and iron-rich LCP with respect to LCP. In fact, in HCP and iron-rich LCP,λSWIRpeak occurs at longer wavelengths than LCP with a lower iron content [Cloutis and Gaffey, 1991; Mustard et al., 2007, Klima et al., 2007]. Thus, BW1000 increases with the Ca and Fe content of the pyroxenes. This parameter can map the presence of mafic composition for BW1000 > 0.70, occurring in the dark Noachian terrains, as well as inside some craters of the south polar region. Also the hematite of Terra Meridiani, having the highest λSWIRpeak together with the olivine of Nili Fossae, shows similar BW1000 values. To summarize, the highest values of BW1000 parameter map olivine, the intermediate ones pyroxene and the lowest ones the terrains without mafics.

Figure 12.

Same as Figure 7 except for parameter BW1000.

4.5. OLVINDEX

[49] On the Earth olivine sometimes occurs with pyroxene and the resulting spectrum tends to preserve the pyroxene band [Adams, 1974]. On Mars we observe the same spectral behavior: in the rocks where both olivine and pyroxene are present, the iron absorption of the pyroxene dominates the spectrum. In an olivine/pyroxene mixture, the resulting OMEGA spectrum shows the presence of a slope between 1.743 and 1.543 μm (typical of the olivine) and between 1.300 and 1.213 μm (typical of the pyroxene) that can be used to map them. The ratio (R1743/R1543)/(R1300/R1213) (OLVINDEX) shows a strong correlation with olivine occurrence mostly in agreement with previous works [i.e., Poulet et al., 2007], excluding some areas not covered by the data set considered in our study, for example around both Hellas and Argyre Planitia. The global olivine distribution (OLVINDEX > 0.99) is shown in Figure 13. Because the areas where the olivine is found are very small (a few pixels) we show them with a circle bigger than the real size.

Figure 13.

OLVINDEX over MOC albedo map. Green color circles indicate the terrains where the value of the parameter is >0.99. Spectral indices maps over MOC albedo map. The maps are divided in three parts: North Polar Region (+65° < Lat < +90°) South Polar Region (−90° < Lat < −65°) and central region (−65° < Lat < +65°).

4.6. RHP685

[50] In Figure 14 the OMEGA mean spectra of olivine (green, Nili Fossae), pyroxene (blue, Syrtis Major) and nanophase oxide (red, Arabia) are shown. In the VNIR range, the olivine spectrum is characterized by a sharp reflectance peak at 0.685 μm. As the composition moves to pyroxene type, the peak shifts to longer wavelengths with a smoother shape. It disappears in the nanophase oxide spectra typical of the bright regions. The relative height of the peak with respect to the continuum, taken as the average between two fixed wavelengths (0.625 and 0.782 μm), is named RHP685 index, and its map is shown in Figure 15. The higher values (>1.04) correspond to Noachian dark terrains, as expected. Comparable values are found also on the northern dark terrains, such as Acidalia Planitia and in the circumpolar regions.

Figure 14.

In green: mean spectrum of an area taken from the orbit 2239_4 (Nili Fossae, olivine); in blue: mean spectrum of an area taken from the orbit 2459_4 (Syrtis Major, pyroxene); in red: mean spectrum of an area taken from the orbit 2375_5 (Arabia Terra, nanophase ferric oxide). The dashed lines indicate the wavelengths used in order to calculate RHP685. In the green and blue spectra an offset it has apply for clarity.

Figure 15.

Same as Figure 7 except for parameter RHP685.

[51] Although this parameter has been optimized to map olivine on the basis of the spectral shape in the VIS range, even choosing a threshold >1.048 (value derived from the terrains with OLVINDEX > 0.99) the ambiguity between olivine and pyroxene (preferentially HCP) still persists.

[52] In general, since RHP685 shows a linear correlation with respect to BI1000 and SHVNIR1000, this parameter provides a means to map the mafic composition using only the VNIR channel.

4.7. SHVNIR1000 and SHSWIR1000

[53] The maps of these parameters are reported in Figures 16 and 17. SHVNIR1000 shows a correlation with the albedo. The lowest values characterize the high albedo regions, while the highest ones map the dark areas of ancient terrains and the sand dunes around the North Polar Cap. The area centered at ∼155°W and ∼83°S, close to the South Polar Cap and already discussed in §4.3 also shows larger values than the surrounding regions.

Figure 16.

Same as Figure 7 except for parameter SHVNIR1000.

Figure 17.

Same as Figure 7 except for parameter SHSWIR1000.

[54] The same correlation with the albedo is not observed for the SHSWIR1000 where low reflectance terrains as Acidalia Planitia have the lowest SHSWIR1000 values on Mars, while other low albedo terrains, such as those of the Noachian crust, are characterized by high-intermediate SHSWIR1000. The behavior in the right shoulder at 1μm together with the differences in other spectral parameters (see BW1000 and BI1000 maps) suggests a peculiar nature of the terrains as shown by the first time by Bandfield et al. [2000]. The composition of the northern dark plains is still under discussion: first interpreted as andesite by means of TES spectra (6.3–50.0 μm range [Bandfield et al., 2000]), they were afterward interpreted as weathered basalt [Wyatt and McSween, 2002] or basaltic material covered by dust or glassy coating to explain the blue slope observed in the OMEGA spectra [Mustard et al., 2005; Poulet et al., 2007; Salvatore et al., 2010]. Recently, Horgan and Bell [2012]have shown that OMEGA spectra of Siton Undae (located at higher latitudes than Acidalia Planitia) are consistent with features arising both from high abundances of iron-bearing glass and silica-enriched leached rinds on glass.

[55] In Figure 18 a spectrum from Acidalia Planitia is shown together with similar spectra taken from dark regions that have the same behavior for the SHVNIR1000 and SHSWIR1000 parameters. These spectra are the ones from Utopia Planitia and Kasei Valles (KV1), as well as from dark deposits in the floor of two low latitude craters in Oxia Palus and Lunae Planum, previously ascribed to dark impact glasses having a typical blue slope with no mafic absorption bands [Schultz and Mustard, 2004]. These spectra are also compared with: (1) an OMEGA spectrum dominated by pyroxene mineralogy taken in another deposit of Kasei Valles (KV2), close in location to spectrum KV1; (2) an OMEGA spectrum taken over the northern dark sand dunes that also exhibits hydration features; and, finally, (3) the library spectrum of an oxidized basalt.

Figure 18.

Comparison of OMEGA spectra from different types of dark terrains. The negative slope in the IR range matches with the oxidized basalt library spectrum showed in the bottom part.

[56] Excluding the pyroxene-like spectrum from KV2, the slopes of the other OMEGA spectra are comparable with the spectrum of the oxidized basalt in near-infrared range. In particular, the reflectance decrease at these wavelengths could be due to the presence of an oxidized coating/rind covering the basalt (probably pyroxene) as discussed insection 4.2 or from a glass component as suggested by Poulet et al. [2007] and Horgan and Bell [2012]. This seems to confirm, together with the recent results from CRISM [Salvatore et al., 2010], that the mineralogy of these peculiar areas on Mars is likely linked to a weathered basaltic component, rather than to a more andesitic one [Bandfield et al., 2000]. In fact, i) OMEGA spectra in Figure 18stand out for their negative slope in the near-IR (excluding KV2); ii) they are characterized by the visible spectral domain typical of the mafic terrains on Mars, with theλVNIRpeak at wavelengths <0.75 μm. Importantly, the same negative slope in the IR range is also present in spectra taken from the circumpolar dark sand dunes, which are characterized by hydration bands [Poulet et al., 2008] and every Martian year are subjected to the deposition and sublimation of ice. This spectral similarity in the NIR range puts some constrains on the weathering processes that the northern lowlands underwent. Distinct from the southern and older mafic terrains, the northern lowlands may be linked to alteration processes possibly including seasonal ices, as also suggested by the morphologies of those areas [Zealey, 2009; Horgan and Bell, 2012].

4.8. BI1000 Versus SHVNIR1000

[57] In Figure 19 we show the linear correlation between the band integral (BI1000) and the left shoulder height at 1 μm (SHVNIR1000). This important relation allows, as a first approximation, to use SHVNIR1000 as a measure of BI1000 in case the absorption at 1 μm is incomplete due to the VNIR and SWIR misregistration. In the same figure the relation with the Martian terrain albedo is shown. The general trend is indicated by the arrows, bright terrains having low BI1000 and low SHVNIR1000, while the dark ones exhibit high values.

Figure 19.

BI1000 as a function of SHVNIR1000. A linear dependence between the two spectral indices exists. The arrow show the albedo behavior.

4.9. BI1000 Versus SHSWIR1000

[58] In Figure 20 we show the distribution of BI1000 as a function of the right shoulder height at 1 μm (SHSWIR1000). Unlike the previous case (BI1000 versus SHVNIR1000), the linear dependence between the band integral and the SWIR shoulder at 1 μm is less obvious, especially for low values of BI1000 (typically <0.04).

Figure 20.

Correlation between BI1000 and SHSWIR1000. The various minerals present on the Martian surface cover a specific portion of the scatterplot. Mafic minerals cover the right part of the figure (cyan and blue curves), while oxides are on the left (yellow curve). It is possible also to select three classes of hydrated minerals. They are the hematite of Terra Meridiani (red curve), the sulfates of Aram Chaos and Capri Chasma (orange curve) and the gypsum of the Olympia Planitia around the North Polar Cap (dark orange). The terrains of the northern lowlands and those of southern high-latitude have the lowest SHSWIR1000 (bottom part inside the magenta curve). The curves that defines the area of a specific mineral is approximate and are drawn using spectral indices of other authors (details in the test). They are drawn using not only the 1μm features.

[59] Nanophase oxides in the bright regions occupy the left upper part of the scatterplot. They fall inside the yellow cluster selected by considering the points with 1.08 < COATINDEX < 1.16 that can be used to map them (see also Figure 9). Bright regions are thus characterized by 1.01 < SHSWIR1000 < 1.08 and BI1000 < 0.04. This distribution matches the one from the Fe3+ parameter of Poulet et al. [2007] with the threshold >0.0025, coinciding with the points inside the yellow cluster.

[60] The low albedo terrains that show a mafic composition cover the right upper part of the scatterplot, where the two indices have a linear trend. Moreover, HCP mapped through the corresponding Pelkey et al. [2007] parameter (threshold >0.0025) tends to dominate in the upper right side. Spectra with HCP fall inside the cyan cluster. This distribution matches the one we obtain with 1.5 < λSWIRpeak < 1.6, by excluding the hematite and olivine rich terrains (see also section 4.1). LCP mapped through the corresponding Pelkey et al. [2007] parameter (threshold >0.0015) is located to the bottom left of the HCP, and data points fall inside the blue cluster, although with a large superimposition with the cyan cluster. The parameters used to select the two clusters are defined by the use of wavelengths around the 2 μm band in addition to the 1 μm band and are optimized to map the two different pyroxene classes. In contrast, the parameters used in the plot shown in Figure 20 (as well as the λSWIRpeak) use the 1 μm band alone. The fact that the cyan and the blue clusters overlap over a wide area is indeed surprising considering that LCP have a much lower spectral diversity than HCP [Adams, 1974] and could be indicative of the fact that LCP occur mainly as a mixture. This, together with the fact that grain size plays a role in shaping the 1 μm band, does not permit use of parameters to discriminate the possible presence of intermediate Ca/Fe pyroxene. Correct interpretation can be done only by means of mineral deconvolution [e.g., Poulet et al., 2009], which is not the purpose of the present work.

[61] In green we have colored the points with OLVINDEX > 0.99 (see section 4.5). They are located in a wide range crossing the pyroxene portion of the scatterplot. This also may indicate that most of the olivine is present in association with pyroxene. The fraction of olivine located in Nili Fossae has the highest values of BI1000 and SHSWIR1000, showing the lowest contribution from other minerals.

[62] The lowest values of the scatterplot (magenta curve) delineate the low albedo regions of the northern plains including Acidalia Planitia together with the impact crater ejecta and the southern high-latitude terrains.

[63] The red and orange curves include points that cluster separately from the main trend of the scatterplot, mapping three separated regions on Mars: Terra Meridiani (red curve), Aram Chaos and Capri Chasma (orange curve). In Terra Meridiani, the largest deposit on Mars of crystalline ferric oxides occurs. Hematite best explains the spectral features observed in this geographic area [Christensen et al., 2001, 2004; Squyres et al., 2004; Poulet et al., 2008]. The spectra belonging to this region depart from the main distribution since they are characterized by intermediate values of BI1000 (between 0.02 and 0.05) coupled to higher values for SHSWIR1000 (>1.1). In the terrains of Capri Chasma and Aram Chaos previous studies have revealed deposits of mono and poly hydrated sulfates together with hematite [Glotch and Christensen, 2005; Hutchison et al., 2005; Bibring et al., 2007; Massè et al., 2008; Noe Dobrea et al., 2008; Roach et al. 2009; Flahaut et al., 2010; Lichtenberg et al., 2010]. Their data cluster separately from the general trend having 0.06 < BI1000 < 0.08 and SHSWIR1000 > 1.18. Only the Nili Fossae spectra dominated by olivine have SHSWIR1000 values larger than this threshold. Although the largest sulfate deposit occurs in the dunes near the North Polar Cap [Langevin et al., 2005], the scatterplot is not able to isolate these terrains. Their location in Figure 20 is pointed out by the brown curve, drawn by means of the occurrence of the 1.9 μm band: some of them show BI1000 values as high as the points in the orange curve, but with SHSWIR1000 < 1.06, that characterizes both LCP and magenta points as well. A possible interpretation can be that the gypsum bearing terrains contain both Acidalia-type and pyroxene-bearing sands that determine the 1μm band shape, also consistent with previous work [Horgan et al., 2011].

[64] In the same way, the phyllosilicates, another important class of hydrated minerals identified on Mars [Poulet et al., 2005] do not group separately in the scatterplot. In fact, although the phyllosilicates spectra can have an absorption in 0.9–1.0 micron range, its shape and depth has values similar to those of other minerals present in the main trend of the scatterplot.

[65] As far as the correlation of SHVNIR1000 versus SHSWIR1000 is concerned, we observe a clustering similar to the one shown in Figure 20. This is due to the linear behavior between BI1000 and SHVNIR1000 discussed in the previous section.

5. Summary

[66] By using both VNIR and SWIR OMEGA channels, the distribution of the 1 μm absorption on Mars (including the terrains surrounding the permanent caps of both hemispheres) was studied. For this purpose, the VNIR calibration procedure was updated. We have created 9 new maps using new spectral parameters. The integral of the whole band depth has been correlated with other spectral parameters such as its band width, the left/right shoulder heights, the reflectance peak in both VNIR and SWIR ranges, together with their ratio and the peak wavelengths, the height of reflectance peak at 0.685 μm and a spectral parameter that takes into account the relative spectral ratio between 1.743 and 1.543 μm with respect to the ratio between 1.300 and 1.213 μm. Global mineralogy mapped through the 1 μm absorption is in agreement with previous works [e.g., Poulet et al., 2007; Pelkey et al., 2007]. In some cases, the use of one spectral index coupled to another one allowed identification of the most important classes of minerals that cluster in specific areas of the scatterplot. The combinations between the parameters are listed in Table 2. The highest values of the 1 μm band integral are found in olivine rich terrains, because this mineral possesses the largest 1 μm band width. In principle, the band width could be used to determine the composition of olivine, discriminating forsterite (Fo90) from fayalite (Fo10), because a positive correlation exists between iron content and the band width. However, the presence of other minerals, e.g., pyroxenes, can shift the reflectance peak in the SWIR channel toward wavelengths diagnostic of the forsterite, preventing the use of this parameter for discriminating between the two types of olivine.

Table 2. Summary of Parameter Combinations That Can Be Used to Identify a Class of Minerala
CouplingRationale
  • a

    NPD, northern polar dunes.

  • b

    Sulfate of Aram Chaos and Capri Chasma, hematite of Terra Meridiani.

BI1000 vs SHSWIR1000pyroxene, nanophase ferric oxide, sulfate,b hematite,b olivine, NPD, weathered basalt/glass
BI1000 vs SHVNIR1000bright and low terrains
SHVNIR1000 vs SHSWIR1000pyroxene, nanophase ferric oxide, sulfate,b hematite,b olivine, NPD, weathered basalt/glass
COATINDEX vs SHVNIR1000pyroxene, nanophase ferric oxide, sulfate,b hematite,b olivine, NPD, weathered basalt/glass
COATINDEX vs SHSWIR1000pyroxene, nanophase ferric oxide, sulfate,b hematite,b olivine, NPD, weathered basalt/glass
COATINDEX vs BI1000pyroxene, nanophase ferric oxide, sulfate,b hematite,b olivine, NPD, weathered basalt/glass

[67] A specific area in the south polar region (centered at ∼155°W and ∼83°S) appears in the BI1000 map. It is characterized by values 60% higher than the average value of other south polar regions and correlates with one of the flattest terrains on Mars. Spectral ratio analysis shows a higher content of pyroxenes in this area. In particular, the presence of a peak at 1.6 μm and bands centered at around 1.05 and 2.3 μm indicates an enrichment of HCP or iron-rich LCP.

[68] The northern lowlands differ from southern terrains in the NIR negative slope but exhibit VNIR spectra similar to pyroxene-rich areas. These observations combined with both recent detection of mafic minerals at higher spatial resolution by CRISM [Salvatore et al., 2010] and recent results of Horgan and Bell [2012] supports that their mineralogy is linked to weathered basalt with a possible glassy component. Moreover, the spectral similarity of Acidalia with the northern circumpolar sand dunes, apart from the hydration features, suggests that the weathering processes that took place there could be related to past glacial activity, as also suggested by the superficial morphology showing glacial structures.

[69] Most of the hydrated minerals present on Mars do not exhibit significant 1 μm absorptions. Among the hydrated terrains, the only ones that can be isolated by coupling the 1 μm band integral and the 1 μm right shoulder height are those located in Terra Meridiani, Aram Chaos and Capri Chasma. They are enriched in sulfate and hematite and have a higher right wing of the band in comparison with mafic dark terrains with the same 1 μm band integral. Moreover, Aram Chaos and Capri Chasma spectra, which have 1 μm absorption larger than those of Terra Meridiani terrains, group together.

Acknowledgments

[70] We would like to thank Harold Clénet, an anonymous reviewer, and the JGR Associate Editor Bethany Ehlmann for their constructive comments and suggestions. We are also grateful to the whole OMEGA team and ESA staff for scientific, operational and technical supports. F.G.C., F.A., and G.B. are thankful to ASI for the financial support through contract N. I/060/08/0 and to Cristian Carli for providing Klima library spectra.

Ancillary