P. R. Christensen, ^* D. L. Anderson, S. C. Chase, R. T. Clancy, R. N. Clark, B. J. Conrath, H. H. Kieffer, R. O. Kuzmin, M. C. Malin, J. C. Pearl, T. L. Roush, M. D. Smith

The Thermal Emission Spectrometer spectra of low albedo surface materials suggests that a four to one mixture of pyroxene to plagioclase, together with about a 35 percent dust component provides the best fit to the spectrum. Qualitative upper limits can be placed on the concentration of carbonates (<10 percent), olivine (<10 percent), clay minerals (<20 percent), and quartz (<5 percent) in the limited regions observed. Limb observations in the northern hemisphere reveal low-lying dust hazes and detached water-ice clouds at altitudes up to 55 kilometers. At an aerocentric longitude of 224° a major dust storm developed in the Noachis Terra region. The south polar cap retreat was similar to that observed by Viking.

P. R. Christensen and D. L. Anderson, Arizona State University, Tempe, AZ 85287, USA.
S. C. Chase, Santa Barbara Remote Sensing, Goleta, CA 93017, USA.
R. T. Clancy, University of Colorado, Boulder, CO 80309, USA.
R. N. Clark, U.S. Geological Survey, Denver, CO 80225, USA.
B. J. Conrath, Cornell University, Ithaca, NY 14850, USA.
H. H. Kieffer, U.S. Geological Survey, Flagstaff, AZ 86001, USA.
R. O. Kuzmin, Vernadsky Institute, Moscow, Russia.
M. C. Malin, Malin Space Science Systems, San Diego, CA 92121, USA.
J. C. Pearl and M. D. Smith, Goddard Space Flight Center, Greenbelt, MD 20771, USA.
T. L. Roush, Ames Research Center, Moffet Field, CA 94035, USA.
^*   To whom correspondence should be addressed. E-mail: phil.christensen{at}asu.edu

The specific objectives of the TES experiment are: (i) to determine and map the composition of surface minerals, rocks, and ices; (ii) to study the composition, particle size, and spatial and temporal distribution of atmospheric dust; (iii) to locate water-ice and CO₂ condensate clouds and determine their temperature, height, and condensate abundance; (iv) to study the growth, retreat, and total energy balance of the polar cap deposits; (v) to measure the thermophysical properties of the martian surface materials; and (vi) to characterize the thermal structure and dynamics of the atmosphere (5).

The TES instrument consists of three subsections: (i) a Michelson interferometer that collects spectra from 1700 to 200 cm¹ (~6 to 50 µm) at either 5 or 10 cm¹ spectral resolution, (ii) a bore-sighted bolometric thermal radiance spectrometer (4.5 to ~100 µm), and (iii) a solar reflectance spectrometer (0.3 to 2.7 µm). The TES has a noise equivalent spectral radiance near 1.2 × 10⁸ W¹ cm² str¹ cm¹. This corresponds to a signal-to-noise ratio (SNR) of 490 at 1000 cm¹ (10 µm) viewing a 270 K scene. Absolute radiometric accuracy was estimated from pre-launch data to be about 4 × 10⁸ W¹ cm² str¹ cm¹. However, in-flight observations indicated that a small, systematic calibration offset of ~1 × 10⁷ W¹ cm² str¹ cm¹ is present in the TES data. This error is primarily due to slight variations in the instrument background energy between observations taken of space for calibration and those viewing the planet at an angle 90° away (6).

The data presented here were collected between 14 September and 30 November 1997, corresponding to orbits 2 to 53 and covering the southern spring season from aerocentric longitude (L_s) 182° to 228°. Throughout this period Mars Global Surveyer (MGS) was in a highly elliptical orbit such that the spatial resolution and observation angular velocity varied widely (7). During each periapsis pass the TES pointing mirror is pointed as close to nadir as possible, resulting in observations with emission angles between 0° and 30°. The spatial resolution at periapsis is as low as 1 km in the cross-track direction, but is 8 to 10 km down-track due to smear produced by the spacecraft velocity. Over the period observed, the local time of the periapsis observations decreased from 18.3 H (where H refers to martian hours, such that 24 H equals one martian day) to 15.7 H, while the latitude of periapsis increased from 30°N to 39°N. As a result of these viewing conditions, the near periapsis surface temperatures typically ranged from 180 to 220 K, below the 260 to 290 K temperature for which the TES was designed to observe.

During each periapsis pass the TES observes the atmosphere across the limb of the planet at several northern latitudes. These observations provide 8 km vertical resolution for a spacecraft altitude of 125 km. The TES pointing mirror is used to provide observations from the surface to a tangent point >75 km above the surface.

Away from periapsis MGS is in a 360°/100 min roll with the nadir axis pointed away from Mars. During the portion of each roll in which the TES scan plane intersects the planet, the TES pointing mirror is stepped across the planet disk, producing a raster image of Mars (Fig. 1). These observations provide views of the south polar cap at a distance of ~16,000 km, corresponding to a surface resolution at 0° emission angle of ~135 km. During the period reported, apoapsis decreased from 53,900 to 40,800 km, corresponding to an improvement in surface resolution from 460 to 350 km, the sub-spacecraft latitude varied from 30°S to 40.6°S, and the local time at the sub-spacecraft point varied from 6.4 to 3.6 H, with surface temperatures typically <200 K. 

The composition of the surface minerals, dust, water-ice clouds, and atmospheric gases can be determined from the TES spectra (Fig. 2). The CO₂ band centered at 667 cm¹ was used to determine the atmospheric temperature-pressure profiles. Water vapor vibration-rotation bands at wavenumbers >1350 cm¹ and from 200 to 300 cm¹ were also identified. The TES spectra represent combinations of these components in varying amounts. The most prominent nongaseous spectral feature observed by the Mariner 9 IRIS was a broad absorption band between 200 and 1300 cm¹ produced by atmospheric dust with a maximum absorption strength near 1100 cm¹ (2). This feature was also observed by the TES (Fig. 2). Its depth increases with increasing emission angle, indicating that the feature is due to an atmospheric aerosol. Water-ice has a broad absorption below 1000 cm¹, with a maximum absorption near 800 cm¹, and a narrower feature with a maximum absorption near 230 cm¹. The TES spectrum (Fig. 2) is similar to IRIS spectra of water-ice and matches synthetic water-ice spectra (8).

Four nearly consecutive spectra from Syrtis Major show a spatial pattern that indicates that the TES observed a surface deposit ~15 km in size with different spectral characteristics from the surrounding material. Previous spectral observations of Mars provided evidence for regional (50 to 300 km) variations in surface composition, in particular Phobos ISM visible/near-IR (9) and Viking IRTM data (6). However, the TES data provides evidence for compositional variability at ~10 km spatial scales.

Seven periapsis tracks have crossed over large regions composed of relatively low albedo surface materials, including three passes over Syrtis Major (Fig. 3), one over a low albedo region in Hebes Chasma (Fig. 4), and two passes over Sinus Meridiani. Spectra from these regions are similar, indicating that the surfaces are composed of similar minerals.

A spectrum of Hebes Chasma is similar to laboratory spectra (10, 11) of augite, a Ca-rich clinopyroxene, in the 800 to 1350 cm¹ region. The identification of pyroxene is consistent with previous visible/near-IR Earth-based (12-15) and spacecraft (9, 16) observations. The depths of the absorptions in the 400 to 550 cm¹ region do not match the laboratory spectra, although the subtle feature near 475 cm¹ is similar to the features seen in the augite spectrum (Fig. 3). These differences in the spectral contrast between the 800 to 1350 cm¹ region and the 400 to 550 cm¹ region, as compared to laboratory data (Fig. 3), are not understood at this time. The surface reflection of downwelling atmospheric radiation, which is peaked at between 300 cm¹ and 450 cm¹ for atmospheric materials at temperatures between 160 and 240 K, may provide a partial explanation.

Other minerals, in addition to pyroxene, are required to account for all of the features observed in the TES spectra. In particular, additional materials with absorptions between 1150 and 1250 cm¹ and between 975 and 1050 cm¹ are required. Linear deconvolution (17) of the TES spectra gives a best-fit mixture of 45% augite, 7% bronzite (an Mg-rich orthopyroxene), 12% labradorite (a plagioclase feldspar), and 36% atmospheric dust component (Fig. 5). These abundances are interpreted to be the areal fraction of each component. Again, the absolute depth of the spectra do not match in the 400 to 550 cm¹ region. These results suggest that a significant amount of feldspar is present on the martian surface. The spectrum from Hebes Chasma has been used to estimate the upper limit abundances of carbonate (calcite, <10%), olivine (forsterite, <10%), clay minerals (<20%), and quartz (<5%).

A four to one mixture of pyroxene to plagioclase provides the best fit for the surface materials in Hebes Chasma. Both clino- and orthopyroxenes appear to be present, with clinopyroxene making up 85% of the total pyroxene. Dust is an important component, either on the surface or in the atmosphere. If the dust is all in the atmosphere, then the surface is composed of 80% pyroxene and 20% plagioclase. The presence of pyroxene and plagioclase is consistent with the bulk composition of most of the martian meteorites (10, 16, 18).

The low albedo materials are relatively free of weathering products, including hematite and clay minerals. These observations suggest that the low albedo materials on Mars are relatively pristine samples of igneous rocks. The estimated low abundances of calcite (<10%) in the limited regions observed indicates that carbonates are not ubiquitous on the martian surface and, if present, occur in specific locations that either favored their initial deposition or their subsequent preservation.

A map of the relative spatial variation of the pyroxene-rich component was made using a spectral shape characterization. This method determines a curvature parameter in the emissivity () spectrum from 900 to 1100 cm¹ that is related to the depth of the pyroxene band using: where the subscripts refer to the wavenumber (Fig. 3). This parameter increases as pyroxene abundance increases (Fig. 6). The spatial variability in the spectral properties shows a correlation with surface morphology and albedo. In particular, the pyroxene abundance increases from bright surfaces in Isidis into Syrtis Major and decreases toward the Hellas Basin. Localized patches of low albedo material have higher abundances of the pyroxene and plagioclase mixture, indicating either a concentration of these minerals by reworking, or an absence of obscuring dust on the surface.

Atmospheric temperature profiles were obtained from 10 cm¹ resolution measurements within the 15-µm CO₂ absorption band using a constrained linear inversion algorithm (19). The necessary atmospheric slant-path CO₂ transmittances were prepared using absorption coefficient distribution functions derived from line-by-line monochromatic calculations. The required surface pressures are derived from elevations relative to the nominal 6.1 mbar surface taken from the USGS Digital Topographic Map, with an adjustment for the seasonal CO₂ sublimation cycle (20). For nadir-viewing spectra, profiles were obtained from the surface up to about the 0.1 mbar level with a vertical resolution of about 0.75 pressure scale height. The formal 1 random error due to instrument noise propagation in individual spectra is 1 K at all levels. By combining several thousand individual retrievals from a portion of a single orbit, meridional cross sections were constructed (Fig. 7A). The observed thermal structure agrees qualitatively with Mars general circulation modeling results for similar seasonal conditions (21).

On a rapidly rotating planet, away from the equator, the large-scale zonal (east-west) atmospheric motion can be described by a so-called "gradient-wind" in which the north-south pressure gradients are balanced by Coriolis and centrifugal forces. Assuming hydrostatic balance in the vertical direction, the zonal wind field can be calculated from the north-south temperature gradients (Fig. 7B). We assumed that the wind speed is zero at the planetary surface and, because the gradient-wind approximation is invalid near the equator, no wind speeds are shown between ±12° latitude. The strong winds centered near 55°N (Fig. 7B) represent an intense circumpolar vortex associated with the large temperature gradients at high northern latitudes in the fall hemisphere. At altitudes above the 1 mbar pressure level, temperatures increase from the equator to 45°N. This is not expected for an atmosphere in radiative equilibrium and indicates the response of the temperature field to atmospheric motions. The temperature maximum may be a result of adiabatic compression and consequent heating from the downward branch of a meridional circulation (Hadley cell) with accompanying upwelling at low southern latitudes. This heating, combined with radiative cooling of the polar regions, may maintain the temperature gradient poleward of about 50°N, which is associated with the high-speed vortex. Since orbit insertion (L_s = 180°), winds in the vortex have tripled, and the vortex moved northward ~10°.

The strength of the spectral signatures of dust are correlated with emission angle, indicating that a large proportion of the dust is suspended in the atmosphere. These signatures were used to infer column-integrated opacity as a function of location and time. Using the derived temperature profiles and surface temperature estimates obtained from the nearly transparent spectral region near 1300 cm¹, optical depths of the dust were determined for several points in the spectrum. We assumed that the dust was uniformly mixed vertically in the atmosphere and that the surface emissivity is unity. The dust opacities were determined most accurately when there was a significant difference between the surface temperature and the mean atmospheric temperature in the lowest scale height of the atmosphere. Figure 8 shows a map of the average column-integrated dust opacity observed between 5 October and 21 October 1997. Most of the horizontal structure is caused by topographical features. The more dusty (dark) areas are topographic lows such as Hellas basin at 45°S latitude, 290°W longitude, and the less dusty (light) areas are topographic highs, such as the Tharsis region centered at the equator and 120°W longitude. These column-integrated opacities are consistent with the dust opacities at visible wavelengths derived from the Pathfinder camera (22) when the appropriate conversion from IR to visible opacity (23-25) is taken into account.

1

If the dust is uniformly distributed in the atmosphere, the effect of topography can be removed from the dust opacity data by scaling to an equivalent 6.1 mbar surface. Examples of scaled column-integrated dust opacity maps from three selected orbits (Fig. 9) illustrate the variations in dust loading that have occurred. The dust opacity retrieved on 21 October 1997 was nearly spatially uniform, with an optical depth of 0.20 to 0.25 (Fig. 9A); transient dust activity (optical depths to 0.5) then began appearing occasionally near the receding cap edge. Beginning on orbit 49 (L_s = 224°; 25 November), the TES began observing an increase in the temperature of the lower atmosphere and an increase in atmospheric dust opacity. This activity developed over the next few days to a storm of regional extent. On 27 November, shortly after the dust storm had started in the Noachis Terra region, scaled optical depths of order unity covered an area that extended about 50° in longitude, centered near the zero meridian, and from 20°S to 60°S latitude (Fig. 9B). The dust opacity far from the storm was also elevated (typically 0.35) compared to that observed on 21 October. By 30 November, when the storm had expanded and intensified further, the core area had extended to the southeast, and a secondary area of dustiness was present north of the Hellas basin at about 20°S, 290°W (Fig. 9C). Thereafter, expansion to the southeast continued, resulting in another region of persistent activity centered near 60°S, 240°W. Other new, less active sites near 60°S, 90°W and 65°S, 160°W waxed and waned through the first 10 days of December. Subsequently, all activity equatorward of 65°S gradually ceased. By the third week of December, the background dustiness in the southern hemisphere, which had reached opacities of about 0.5 to 0.6, was declining. By 10 January 1998 (orbit 88, L_s = 235°), overall levels were at their prestorm values, with only scattered transient activity at high southerly latitudes.

1

Limb spectra were used to obtain temperature profiles to higher altitudes (~70 km) and with better vertical resolution (~0.5 pressure scale height) than can be obtained from nadir observations. Discrete water-ice clouds were detected and quantitatively analyzed using limb data (Fig. 10). Measurements in the CO₂ band were used to retrieve a temperature profile (Fig. 10A). Water ice displays a number of spectral features with one of the strongest centered near 230 cm¹. The radiances within this absorption band at 243 cm¹ show distinct vertical structure (Fig. 10B); these were used along with the temperature profile to retrieve the ice number density profile (Fig. 10C). Discrete clouds centered at 45 and 53 km altitude are evident. Detached ice clouds are frequently observed in limb spectra acquired in the northern hemisphere during the fall season considered here.

The low temperatures of the martian winter poles represent a challenging target for thermal spectroscopy. At 145 K the SNR is <60 at all wavenumbers and is <1 beyond 1200 cm¹. Therefore, corrections for small radiance errors are important (26). To mitigate the low SNR for single spectral points, TES spectra were used to synthesize a seven-band thermal emission mapper, with the bands tailored to fit this application; these include a dust band and adjacent continuum bands, "red" and "blue" surface bands where the atmosphere is relatively clear, and a split atmospheric temperature band (27). In addition, some of the IRTM bands (28) were synthesized; the corresponding brightness temperatures are labeled T₂₀, T₁₁ and so on, where the subscript indicates the effective wavelength in micrometers.

Within the CO₂ cap (T₂₀ < 155 K and T₇ < 205 K) and near the cap edge, where the atmosphere is warmer than the surface, particulates in the atmosphere were seen in emission (Fig. 11). The small increase in brightness temperature from 300 to 250 cm¹ is attributed to water-ice clouds (8). An average "cloud" spectrum, based on T₁₁- T₂₀ > 7 K, shows strong dust and water-ice features. In contrast, an average spectrum from the warm frost-free region surrounding the cap (180 K < T₂₀ < 210 K) is relatively featureless; opacities derived under such conditions have large uncertainty.

1

1

1

1

1

From L_s = 185° to 223°, the seasonal south polar CO₂ cap retreated continually, and the atmosphere above the cap gradually warmed. There were indications of local, temporary incursions of moderately dusty ( ~ 0.25) air over the cap, with atmospheric dust generally being greatest near the cap edge. The outer few hundred kilometers of the cap do not appear as a blackbody at the expected CO₂ condensation temperature, but, in the spectral regions with relatively low atmospheric absorption, have brightness temperatures increasing to higher wavenumber. This behavior is indicative of a mix of frost and warmer frost-free ground, consistent with the mottled visual appearance reported for the springtime cap (29, 30).

Throughout much of this period, the upper atmospheric temperature (T₁₅, equivalent to the 0.6 mbar pressure level) had a consistent thermal gradient of about 0.2 K per degree latitude away from the pole. The minimum temperature, at the pole, rose gradually from 148 K at L_s = 185° to 180 K at L_s = 225°, then rapidly to 200 K by L_s = 225°. By L_s = 222° the dust opacity had reached 0.3 over nearly the entire cap, and atmospheric temperatures showed a strong diurnal dependence. At this time, the T₁₅ minimum was located at 75°S and 6 H. The surface brightness temperature contrast was correlated with time of day and extended over the entire cap, indicating partial frost coverage.

Early telescopic observations (30, 31) had indicated that the late spring recession is not symmetric about the pole, and Mariner 9 (32) and Viking (29) observations clearly showed that the retreating edge is irregular. TES observations for pairs of orbits were combined, and the edge of the cap defined on the basis of brightness temperatures observed in the daytime between 10 and 16.5 H. Near the cap edge, the TES field of view contains a range of temperatures, and the brightness temperature increases to higher wavenumber. The cap edge was defined to correspond to the location where the temperature began increasing rapidly in latitude away from the pole (Fig. 12). This cap edge temperature was 165 K in T₂₀ and 187 K in T₇. The range of cap edge latitude over all longitudes for any date in this period is significant, averaging about 6°. Even at the earliest season observed (L_s = 185°), the cap was discernibly asymmetric, and became more asymmetric through the observation period (Fig. 13). The cap radius we determined is smaller than reported from reflectance observations (30) and this difference may be the result of subpixel mixing with nonlinear weighting of the different surface components (frost and bare ground) in the IR, versus linear weighting of these components in the visible.

Atmospheric particulate features were pervasive in spectra of the south polar region and cloud spectral features are strong near the edge of the retreating cap (Fig. 11). In place of the full temperature/pressure inversion used over the warmer parts of the planet, a broadband model was used. Where the surface temperatures were nearly uniform within a TES field of view,  was estimated using an isothermal atmosphere model, where the dust was assumed to be at the atmospheric temperature near 3 mbar pressure, where R_d is the observed radiance summed over the dust band and B_(d,T) is the Planck function at T summed over the dust band. The surface T_c is based on the average radiance within the two continuum regions on either side of the dust band and the atmosphere temperature T_a is based on two regions in the wings of the CO₂ band. This equation is inverted to derive , and the vertical opacity derived using an air-mass-factor based upon the emission angle. We assumed the emissivity of all surfaces was unity.

Near the edge of the receding polar cap, a single TES field of view commonly contains materials with different temperatures and opacities derived by assuming a gray-body surface emission are misleading. We assumed the surface could be represented by a two-component system; some coverage of frozen CO₂ and a fraction of material at a higher temperature. The temperature and fraction of the other material were determined using "red" and "blue" wavelength regions (26) thought to be little affected by dust and where the surface emissivity is high. This two-component surface is then used to compute the upwelling radiance at the surface in the dust band.

Early in the observation period, there was little contrast between the atmospheric temperature and the surface frost, and derived opacities had a large uncertainty. At L_s = 191.5°  was about 0.15 over the cap, opacities are largest near the cap edge and over the adjacent frost-free terrain. The highest opacities are partly the result of assuming the dust is at the 3-mbar level; the associated spectra show that the highest atmospheric temperatures are closer to the surface, and, hence, the dust is probably also near the surface. By L_s = 204° atmosphere/ground thermal contrast was about 15 K; localized dust clouds of opacity about 0.2 had made incursions over the cap to within 10° of the pole. At L_s = 223.5° the atmosphere was about 30 K warmer than the cap and the average opacity varied from 0.1 off the cap to 0.3 over the cap. Observations with ~45 km resolution revealed dust clouds of about 200 km extent over the cap. These observations suggest that small dust storms started near the edge, and possibly within, the seasonal polar cap.

REFERENCES AND NOTES

1. R. A. Hanel, et al., Science 175, 305 (1972) [Abstract/Free Full Text] .
2. B. Conrath, et al., J. Geophys. Res. 78, 4267 (1973) .
3. H. H. Kieffer, et al., Science 193, 780 (1976) [Abstract/Free Full Text] .
4. H. H. Kieffer, et al., J. Geophys. Res. 82, 4249 (1977) .
5. P. R. Christensen, et al., ibid. 97, 7719 (1992).
6. P. R. Christensen, ibid. 103, 1733 (1998).
7. A. Albee, F. D. Palluconi, R. E. Arvidson, Science 279, 1671 (1998) [Abstract/Free Full Text] .
8. R. J. Curran, B. J. Conrath, R. A. Hanel, V. G. Kunde, J. C. Pearl, ibid. 182, 381 (1973) [Abstract/Free Full Text].
9. J. F. Mustard et al., J. Geophys. Res. 98, 3387 (1993).
10. V. E. Hamilton, P. R. Christensen, H. Y. J. McSween, ibid. 102, 25593 (1997).
11. S. W. Ruff, P. R. Christensen, P. W. Barbera, D. L. Anderson, ibid., p. 14899.
12. T. B. McCord and J. B. Adams, ibid. 74, 4851 (1969).
13. R. B. Singer, Lunar Planet. Sci. XI, 1045 (1980) .
14. ___ and T. L. Roush, Bull. Am. Astron. Soc. 17, 737 (1985) .
15. R. B. Singer and H. Y. J. McSween, in Resources of Near-Earth Space, J. Lewis and M. Matthews, Eds. (Univ. of Arizona Press, Tucson, AZ, 1990).
16. J. F. Mustard and J. M. Sunshine, Science 267, 1623 (1995) [Abstract/Free Full Text] .
17. M. S. Ramsey and P. R. Christensen, J. Geophys. Res. 103, 577 (1998) [CrossRef].
18. H. Y. McSween Jr., Meteoritics 29, 757 (1994) [ISI].
19. R. A. Hanel, B. J. Conrath, D. E. Jennings, R. E. Samuelson, Exploration of the Solar System by Infrared Remote Sensing (Cambridge Univ. Press, New York, 1992).
20. J. E. Tillman, N. C. Johnson, P. Guttorp, D. B. Percival, J. Geophys. Res. 98, 10963 (1993) .
21. J. B. Pollack, R. M. Haberle, J. Schaeffer, H. Lee, ibid. 95, 1447 (1990).
22. P. H. Smith, et al., Science 278, 1758 (1997) [Abstract/Free Full Text] .
23. R. W. Zurek, Icarus 50, 288 (1982) .
24. T. Z. Martin, ibid. 66, 2 (1986).
25. R. T. Clancy and S. W. Lee, ibid. 93, 135 (1991).
26. The space background correction was computed for each observation sequence, but is still uncertain by about 5 x 10⁸ W¹ cm² str¹ cm¹. Reflected sunlight is comparable to the instrument noise level. For the south polar region, spectral radiance corresponding to that expected for solar reflectance by a Lambertian surface at the pole with albedo of 0.5 was subtracted from the TES observation before converting to brightness temperatures.
27. Bands were formed by summing the observed radiance over a set of TES spectral points. The dust band used 30 points over 889 to 1196 cm¹, the pair of dust continuum bands consisted of eight points at 772 to 846 cm¹ and eight points at 1238 to 1312 cm¹. The surface bands were "red"; eight points at 360 to 434 cm¹ and blue, eight points at 1300 to 1376 cm¹. The atmospheric temperature was estimated by summing two bands on each side of the 15 µm CO₂ band corresponding to a weighting function peaked near 3 mbars.
28. S. C. Chase Jr. et al., Appl. Opt. 17, 1243 (1978).
29. P. B. James, G. Briggs, J. Barnes, A. Spruck, J. Geophys. Res. 84, 2889 (1979) [ISI].
30. P. B. James, H. H. Kieffer, D. A. Paige, in Mars, H. H. Kieffer, B. M. Jakosky, C. W. Snyder, M. S. Matthews, Eds. (Univ. of Arizona Press, Tucson, AZ, 1994).
31. G. E. Fischbacher, L. J. Martin, W. A. Baum, Contract 951547 Final Report, Part A, Jet Propulsion Laboratory (1969).
32. P. B. James and K. Lumme, Icarus 50, 368 (1982) [CrossRef] [ISI].
33. We wish to thank the entire TES engineering team at the Hughes Santa Barbara Research Center, led by S. Silverman and M. Greenfield, and at Arizona State University (ASU), led by G. Mehall. We thank N. Gorelick, K. Bender, S. Anwar, M. Weiss-Malik, J. Bandfield, V. Hamilton, M. Lane, S. Ruff, and K. Qazi at ASU. We thank B. Allen II, L. Fenton, A. Gordon, L. Mazzuca, W. McMillan, S. Mason Jr., K. Walker, and W. Maguire. We thank the Goddard Software Development Team, headed by S. Dason and E. Greene, including J. Guerber, K. Horrocks, M. Kaelberer, C. Martin, R. Thompson, E. Winter. We thank T. Titus and K. Mullins at the U.S. Geological Survey in Flagstaff. Finally, we thank the entire spacecraft and mission operations teams at Jet Propulsion Laboratory and Lockheed Martin.