Infrared imaging spectroscopy is a relatively new technique in the study of planetary surfaces and atmospheres made possible in part by developments in infrared array technology (Gillespie et al. 1990a, b, Pinet & Chevrel 1990, Bell 1992, Bell & Crisp 1991a, b, 1992, 1993). The primary strength of this method is that the entire visible planetary disk is imaged at each wavelength allowing all areas to be measured at one time. This allows for easy spot-to-spot comparisons of variations in spectral features. The great strength of this technique does come with a price.
First, this technique uses a circular variable filter (CVF) to isolate the band pass. A CVF is an arc shaped interference filter that has its thickness varying continuously along the arc thus allowing only a certain wavelength bandpass to transmit based on where the telescope beam passes through the filter. In general CVF's have much lower spectral resolutions ( Dl/l = 1.5% in this case) as compared to grating or prism based spectrometers ( Dl/l ~ 0.1%) and so the technique is limited to relatively broad, deep features that will not disappear when convolved with the instrumental spectral response.
Second, although all spatial information can be gathered at one time, the CVF must be stepped to gather all the spectral information. The problem with this is that it is possible for the sky conditions to change over the time it takes to make a complete spectral pass. Most important is the possibility of water content in the beam changing enough to affect the spectral measurements differently at each measured wavelength. This restriction is less of a problem when sky conditions are stable and the telluric atmospheric water content is low.
The third problem, applicable to this work, is that as the data are being taken Mars is rotating. This rotation will cause some minor registration, or feature alignment, difficulty. The spectral scans for this work were completed in 20 to 30 minutes which corresponds to at most about 7.3° of rotation or 433km at the sub-Earth point. The seeing was on the order of 0.8" when Mars had a disk size of 12.3" and about 1" when Mars had a disk size of 13.6". These seeing limits correspond to 453km and 500km at the sub-Earth point respectively. The effect will be that surface albedo feature edges will be enhanced when images from near the distant ends of the spectral scan are divided such as in the band depth and ratio maps.
In the late 1950's Sinton (1959) extended the spectral coverage to the mid-infrared and discovered three absorption bands centered at 3.45, 3.58, and 3.69µm that were found to match absorptions in some terrestrial plants. His observational procedure was to take data at five wavelengths several times in a night and then the next night, shift the wavelength set slightly to fill in wavelengths not covered the previous night. Repeating this procedure over seven nights gave good coverage through the 3.3 to 3.6µm range. Calibration spectra were acquired by observing sunlight reflected from a clean aluminum wire the next morning. Sinton then took spectra of various plant life and found that lichens best matched the general shape of the Mars/Sun spectra. By using lichen spectra along with a modeled surface thermal spectrum and an adjustable constant level offset, a least-square error analysis could produce a good fit to the data. The absorptions appeared more prominent in the dark region Syrtis Major and less visible in the bright region Amazonis, implying more vegetation in the dark regions.
Moroz (1964) confirmed the absorptions in his full disk spectra. His spectra were taken with a prism spectrometer that had a l/Dl resolution of 40. His claim was that the 3.43µm feature could be explained by carbonates but he had no explanation for the other bands. Shirk et al. (1965) found that isotopic water vapor (HDO and D2O) have absorptions near these "Sinton bands" and concluded that the observed Martian features were due to these molecules in the Martian atmosphere. However, matching the Sinton bands with absorptions of HDO and D2O would require a D/H ratio of one, which is much greater than the currently measured value of (9±4)×10-4 (Owen et al. 1988). According to Rea et al. (1965), it may be possible that the absorptions are due to terrestrial HDO and D2O since the Mars and Sun spectra were taken under different humidity and airmass conditions. Their conclusion was that next day solar observations were not a good way to calibrate infrared spectra in this spectral range. Curiously, no follow up observations searching for these bands were performed to see if the features truly can be explained as telluric atmospheric absorption.
Later spectral work in the 2-4µm range by Sinton (1967) includes the discovery that Mars has an overall, broad, deep absorption centered around 3.1µm. He ascribed this absorption to hydrated silicate minerals.
The Mariner 6 and 7 spacecraft which flew by Mars on 31 July and 5 August 1969 respectively, each carried an infrared spectrometer which scanned from 1.88-14.3µm. Using the segments from 1.88-3.68µm and 2.99-6.00µm, Herr and Pimentel (1969) found absorptions at 3.03 and 3.31µm between 61-80S that can be associated with n1 band of ammonia and the n3 band of methane, respectively. These identifications were made using a laboratory spectrum of a mixture containing about 0.2m-atm of each of the gasses pressure-broadened by 1atm of air. However, for Mars they concluded that the absorptions can be better explained as being due to solid CO2, as the line assignments are closer and since the amounts of ammonia and methane in the Martian atmosphere are negligible (Kieffer et al. 1992), thus implying that the southern polar cap is composed at least partially of dry ice. This prompted more research into the spectral characteristics of CO2 and H2O frosts of various crystal sizes, under various environments as well as various mixtures of the two. Study of these frost reflectances began with work by Kieffer (1970a, b) and measuring frost reflectances and optical properties is still an ongoing project (e.g., Fink & Sill 1982, Warren 1984 & 1986, Calvin 1990, Roush et al. 1990, Hansen 1996). Some of these results will be used in this study as comparison spectra and to calculate model spectra.
In 1973 Houck et al. investigated the hydrated mineral absorption feature and concluded that the abundance of water, to within a factor of 3, was only 1% by weight and that the spectra were consistant with most common hydrated mineral spectra from the previous literature (Hunt et al. 1950, Miller and Wilkins 1952, Hovis and Callahan 1966) with a mean particle size from 10 to 300µm. For their study Houck et al. used integrated disk spectroscopic data from an airborne telescope. The Mars spectra were normalized to solar spectra taken at the same time to create reflectivity plots. The feature was then compared to rock and mineral absorption features and the shape was only matched by bound water. The wavelength of the minimum of the feature they found, 2.85±0.1µm, is a standard assignment of the O-H stretching vibrational fundamental.
Pimentel et al. (1974) attempted to differentiate between bound water features of hydrated minerals and water in the form of ice. Using the Mariner 6 and 7 infrared spectra they looked at two different intensity ratios. The first ratio, , where I is the measured intensity at the particular wavelength, they determined was a good indicator of both bound water and ice. It is the ratio of the O-H band minimum to an IR continuum so when this ratio is less than one there a water detection. This ratio is less than one for much of the planet as seen in previously mentioned infrared spectral work. The second ratio, , is a good indicator of the actual state of the water. Based upon their laboratory spectral results, that the ratio is always less than one for a hydrated mineral and greater than one for water ice. In one experiment, 3.2mm of solid CO2 was deposited on the hydrated mineral in an attempt to mask this feature. This serves to increase the reflectivity at 3.1µm even more than at 2.9µm. However, as little as 0.14µm of solid H2O deposited on that mixture is enough to reverse the R2 value to greater than one. Their work then compiled values of these ratios from several scans of the planet and the value of R2 exceeded unity near the polar cap edge. The work in this dissertation will look briefly at these ratios in newer infrared spectra of water ices and Martian images.
Studies of the Martian atmosphere, detailed and referenced below show an interesting situation. During polar winters the temperatures drop low enough that a significant fraction of the atmosphere condenses to form the seasonal polar caps. This causes a large drop in the local atmospheric pressure which then drives the Martian meteorology.
The Viking mission, launched in 1975, ushered in a new era of Mars observations which included seasonal and spatial variations of water vapor abundance for more than a full Martian year. Farmer et al. (1977) presented some preliminary results from the Mars Atmospheric Water Detector (MAWD). The instrument was a five channel detector behind a grating spectrometer set to record the reflectance of Mars in three water bands and two continua in the 1.380-1.385µm range. The photometry was then compared to previously calculated curves of growth for the three absorption features and their calculated water abundances are presented in units of precipitable microns; that is, a measurement of 20pr.µm would be an amount of vapor that, if condensed onto the ground, would create a global liquid layer 20µm thick. These results spanned the time between the summer solstice and the autumnal equinox. With an assumed surface atmospheric pressure of 6mbar and a surface temperature of 200K they calculated a maximum water vapor content of 90pr.µm over the dark material near the residual cap edge around the solstice. The vapor content fell rapidly to only 10pr.µm near the equinox.
Jakosky & Farmer (1982) presented the MAWD data set in its entirety. This set covers almost 1.5 Martian years, starting at Ls ~ 80° and ending at Ls ~ 245°, where Ls = 0° is the northern spring equinox, Ls = 90° is the northern summer solstice, Ls = 180° is the northern autumnal equinox and Ls = 270° is the northern winter solstice. Each of the northern and southern hemispheres has a maximum water vapor content during its local summer (~100pr.µm in the north-polar regions at Ls ~ 120° and ~15pr.µm in the south at Ls ~ 270°). The total global water vapor abundance varies from an equivalent of 1 to 2km3 of ice with the maximum occurring during northern summer and the minimum during northern winter.
One of the major results from Jakosky and Farmer's (1982) work came from tracking the local maxima through the seasons. A rise in the abundance occurs from Ls=0° to 40° for all latitudes above 20°N with the peak between 20°-50°N. The receding cap edge at this time was at 70°N so this peak in the water vapor should be due to a non-polar ice source. Their conclusion was that the regolith must be that source. Similarly, they saw a decrease in the low northern latitude abundances before the formation of the seasonal ice cap, adding support to their regolith reservoir theory. The water could be either a ground frost or water adsorbed in the soil. Unfortunately their adsorption models could not account for all the water necessary for the observed vapor abundances measured.
Another possible explanation for the anomalous high vapor measurements in the low latitudes could be due to transport of clouds to these regions from the subliming polar cap. Due to the wavelength band choices in the MAWD, it is insensitive to water ice clouds so such transport would not be detectable. In the spring, water could sublime from the polar cap but form ice clouds instead of becoming vapor. If these clouds are transported equatorward the ice could become vapor in the warmer air of the lower latitudes, thus giving a higher vapor measurement than in the polar regions. Similarly, in the autumn the air temperature could drop causing the water vapor to condense into ice clouds thus creating the low vapor measurements.
One of the major results from the Viking Infrared Thermal Mapper (IRTM) was that the temperatures measured at the southern winter polar cap reached values well below the CO2 condensation temperature of about 150K, and in some areas were as low as 125K (Kieffer, et al. 1977). The implication of this situation is that the major atmospheric gas, CO2, is condensing at the winter pole. The condensation substantially reduces the local atmospheric pressures, lowering the frost point. This great drop in local pressure drives much of the Martian meteorology. Another explanation is that the low polar atmospheric temperature measurements indicate that the atmospheric condensation should be forming CO2 clouds and Hunt (1980) showed that a t=0.5 CO2 cloud of 10µm particles over a ground with an effective temperature of 150K would produce a brightness temperature at 20µm of 120-130K as seen in the IRTM data.
In their study of polar processes in a Martian general circulation model, Pollack et al. (1990) calculated atmospheric condensation rates for several different seasons as a function of latitude, altitude, and dust opacity. They find that airborne dust favors the condensation of CO2 in the atmosphere as clouds rather than condensation directly on the surface. In the north polar winter (Ls = 279°) with a dust opacity, tr = 1, the atmospheric and surface condensation rates at about 80° north latitude are equivalent and are about 30×10-6kg/m2/sec or about 2.6kg/m2/day. If tr increases to 2.5 the surface condensation rate drops to about 20×10-6kg/m2/sec and the atmospheric condensation rate increases to about 45×10-6kg/m2/sec. With tr = 0.3 the atmospheric condensation rate drops to about 15×10-6kg/m2/sec and the surface condensation rate increases to about 35×10-6kg/m2/sec. Thus dustier conditions will produce more CO2 clouds during the formation of the north polar hood.
Kahn (1990) made similar observations to study the detached hazes - high altitude clouds of volatiles or dust which he presumed to be composed of water ice. Using Viking limb images he was able to measure the haze base altitude, thickness and optical depth. Using these measurements along with water vapor measurements (Jakosky and Farmer 1982) he calculated the condensation level temperature, altitude and pressure, as well as the mean ice particle radius. The haze optical depths were on the order of 0.015 to 0.04 and particle sizes ranged from 0.15 to 6µm in radius, which gave ice concentrations of 0.0021 to 0.034pr.µm. He concluded that these hazes could be a mechanism for scavenging water vapor in the atmosphere and as particle sizes grew, transporting them down to the ground where they could be more easily adsorbed into the regolith than the water vapor itself. This mechanism makes a non-polar source/sink of water vapor more plausible.
Discrete cloud layers were also detected using the infrared spectrometer on the Soviet Phobos2 spacecraft between 8 February and 23 March 1989 (Chassefière et al. 1992). The measurements were made in the equatorial region (0°-20° N latitude) near the northern spring equinox (Ls=0°-20°). The clouds detected were all above 45km altitude with a vertical extent of 3-6km. The optical depth was less than 0.03 and the particles were inferred to be composed of H2O ice as the temperature at these altitudes was above the CO2 frost point. Assuming spherical particle shapes, the effective radius is 0.15-0.85µm with a number density of ~1cm-3. As a comparison, laser measurements of terrestrial cirrus clouds (Carswell 1981) show them to have thicknesses of 1-2.5km, particle sizes of ~50µm and number densities of 0.2-1cm-3.
Current ground based studies of Martian clouds concentrate on the feasibility of detection. Johnson and Atreya (1996) looked at using far-infrared (FIR) spectral measurements to detect and discriminate between H2O and CO2 ground ices and frosts as well as clouds and hazes. They concluded that large H2O particles do not show enough spectral variation to be detectible in the FIR and that due to cloud-surface temperature contrast sensitivity, clouds could only be detected if ice amounts were close to the theoretical limit of ~2pr.µm, corresponding to a water ice polar hood of tvisible ~ 1.0. CO2 ice measurements would be more easily detectible near the 90.9 and 151.5µm lattice bands. However, they conclude that in order to achieve the kind of spatial resolutions needed to see the polar regions of Mars in the FIR, telescopes with primary mirrors on the order of 14m would be necessary.
Work by Bell et al. (1996a) concentrated on using near-infrared (NIR) measurements (1.5-4.1µm) to look at the various volatile absorption bands in this range for a diagnostic set of wavelengths useful to discriminate H2O and CO2 ices. Those results will be used heavily in this dissertation with only slight modifications based on the actual wavelengths used for the 1994-95 observations and thus will be discussed in more detail later.
The Hubble Space Telescope can make out bright features, assumed to be clouds and
ground frosts, but it is limited to wavelengths less than about 1µm. In this wavelength range it is
not possible to determine the frost or cloud composition. The infrared spectral points of the data
set used in this dissertation were chosen to lie within various diagnostic absorption bands for both
CO2 and H2O ices. Three techniques, extracted regional spectra, band-depth mapping, and
principal components analysis/linear mixture modeling, will be used to analyze the data to
determine how best to make the compositional discriminations. These three techniques, and the
strengths and weaknesses of each will be the subject of Chapter 3. Chapters 4, 5, and 6 will
present the results from the extracted regional spectra, the band depth mapping, and the principal
components analysis techniques, respectively. Finally, chapter 7 will summarize the results of this
study and give some direction for future work.