Thermal Infrared Properties of the Martian Atmosphere 1. Global Behavior at 7, 9, 11, and 20 #m

Transcription

1 VOL. 84, NO. B6 JOURNAL OF GEOPHYSICAL RESEARCH JUNE 10, 1979 Thermal Infrared Properties of the Martian Atmosphere 1. Global Behavior at 7, 9, 11, and 20 #m TERRY Z. MARTIN,: ALAN R. PETERFREUND, x'9' ELLIS D. MINER a HUGH H. KIEFFER, x'4 AND GARRY' E. HUNT Infrared observations of Mars by the Viking infrared thermal mapper (IRTM) are presented for both conditions of a relatively clear and a dust-laden atmosphere. The 7-, 9-, 11-, and 20-#m bands of the IRTM respond differently to radiation emitted through and by a dusty atmosphere, permitting characterization of the global atmospheric state, monitoring of secular changes, and derivation of optical depth information. Surface temperature behavior is found to be greatly modified by the diminution of insolation and thermal blanketing resulting from global dust storms. Brightness temperature at 7 m (T?) is employed to estimate surface temperatures in the presence of dust absorption. The difference T?-T9 is strongly indicative of airborne dust when thermal contrast exists between the surface and atmosphere. The diurnal behavior of T7-T9 reveals changes in that contrast; the sign of the differential reverses as the surface, warmer than the atmosphere in daytime, becomes cooler than the atmosphere at night. IRTM observations of local areas at varying emission angle yield optical depths indicative of global trends. Two global dust storms in 1977 produced large optical depth changes; at 9 #m the optical depth became as large as 2.0. INTRODUCTION The above-mentioned surface-sensing channels of the The effect of Martian atmosphericonditions upon mea- I RTM were selected primarily for high-accuracy determinasurements made by the Viking infrared thermal mapping tion of surface-emitted radiance over the entire expected range (IRTM) experiments has been much greater than originally of Martian temperature: approximately K. The 20- anticipated. The instruments, designed to study primarily sur- #m band has useable sensitivity down to less than 120 K, and face thermal and reflective behavior, have returned a wealth of the 7-#m band does not experience saturation below 335 K. data with profound relevance to atmospheric thermal struc- The IRTM views Mars with four co-aligned telescopes that ture, scattering properties, the nature of local dust storms, and simultaneously obtain data in six wavelength bands, for seven modification of surface behavior by atmospheric phenomena. spatial elements arranged in a chevron [Chase et al., 1978; Outside of the CO2 15-#m band, airborne dust is principally Kieffer et al., 1977]. The 7-, 9-, and 15-#m 'bands share one responsible for atmospheric effects evident in IRTM data; dust telescope and chevron, and consequently those measurements can have a substantial impact on the Martian spectrum from are spatially separate from each other. The placement of the visible wavelengths to the middle infrared. During the Viking various bands in wavelength was intended to permit, in clear primary mission (June-November 1976, Ls 85ø2147 ø) atmo- conditions, discrimination of surface emissivity differences spheric opacity was low enough that substantial study of sur- arising from the nature of the silicate vibrational spectrum. face characteristics was possible [Kieffer et al., 1977]. In 1977, Atmospheric modification of surface emission was well docuhowever, Mars approached perihelion; atmospheric dynamic mented by the infrared interferometer spectrometer (IRIS) exactivity increased, and large quantities of dust were injected periment on Mariner 9; comparison of IRTM bands with irito'the atmosphere in two global and many local storms. This IRIS spectra (Figure 1 ) shows the extento which atmospheric paper and the three which follow it in this volume (Martin and effects were expected to appear in IRTM data. Kieffer [1979]; Peterfreund and Kieffer [1979]; Hunt [1979]; The 7-#m band may be considered a 'surface' channel for referred to here as papers 2-4) treat different aspects of the most purposes, since there is comparatively little absorption in IRTM sensing of atmospheric phenomena. In this paper are that band by airborne dust or ice [Hand et al., 1972; Curran et presented results for the infrared wavelengths at which the al., 1973]. The surfac emissivity expected to be near 1.0 at 7 instrument normally detects surface emission: 7, 9, 11, and 20 #m in the absence of extensive fields of carbonate or nitrates m. Paper 2 addresses brightness temperatures in the 15- m [Hunt and Salisbury, 1974, 1975]. Although there is little accuemission band of atmospheric CO2. Paper 3 deals with racy in the 7-#m measurements for surface temperatures below IRTM measurements of local dust storms, including maps of 160 K, that band is expected to give the best estimate of storm structure. Paper 4 is an analysis of the IRTM response surface temperature when airborne dust is present. During to cloudy conditions and the usefulness of brightness temper- much of the Viking extended mission, T, (the 7-#m brightness ature measurements. temperature) is typically the highest IRTM temperature in equatorial regions during the day and the lowest at night. The other bands receive proportionally more radiance from sust Department of Earth and Space Sciences, University of California, pended dust, which is normally cooler than the surface in Los Angeles, California : Presently at the Department of Geology, Arizona State University, daytime and (in the lowest few kilometers) warmer at night. Tempe, Arizona The 9-#m band (Figure 1) covers the low-wavelengthalf of a Jet Propulsion Laboratory, Pasadena, California a broad vibrational feature typical of silicates that extends 4 Presently at U.S. Geological Survey, Flagstaff, Arizona from 8.0 to 12.0 #m in IRIS spectra. Tg is expected to be the a Laboratory for Planetary Atmospheres, Department of Physics and Astronomy, University College London, London WCIE 6BT brightness temperature most affected by airborne dust and by England. deviations from unit surfac emissivity. Often, Tg is the lowest measured temperature other than T s during midday; assum- Copyright 1979 by the American Geophysical Union. ing a typical positive lapse rate for the daytime atmosphere, it Paper number 8B / 79/008 B

2 _, MARTIN ET AL..' MARS VOL TILES O [ anisothermality: existence of spatially varying surface temperatures within the field of view; and (3) surface variation in emissivity with wavelength. It is important to note that sources 1-3 all contribute spectral variation that changes with time of day, and that each phenomenon has a different diurnal dependence. There is also important interplay between effects. For example, a larger dust opacity will pro p& ;t pe I variation, but that dust will also lessen the streng li q[type 2 Variation, because the contrast in temperature betw a; ifferently heated areas on the groundecreases with great r.th/ rmal blanketing by the atmosphere. IRTM data are compared with brightness temperature spectra from the IRIS experiment in Figure 2. It can be seen that the relative behavior of the IRTM temperatures is controlled by airborne dust for these daytime observations in much the same way as are the IRIS spectra, which. show a greatly re. duced 9-pm absorption at rev 174 (L, 344ø), when the 1971 major dust storm had largely dissipated. lh this paper we emphasize the contrast'between the thermal appearance of Mars in relatively clear conditions and the spectral behavior arising from globally distributed airborne dust. The effects of water in ice clouds and fogs have been briefly addressed earlier [Kieffer et al., 1977]. SURFACE TEMPERATURES The urface temperature, which depends upon thermal inertia and albedo, is a valuable measurement for geologic studies. Since atmospheric phenomenare driven by surface temperature, this parameter is basic in global circulation models. A homogeneous thermal model [Kieffer' et al., 1977] has been I used extensively in previous IRTM investigations [Kieffer et FREQUENCY, CM '! al., 1976a, b, 1977; Kieffer, 1976]. We present here the longi- Fig. 1. IRTM spectral response and Mariner 9 IRIS spectra. (a) tudinally averaged diurnal behavior of surface temperature Spectrum from Mariner 9 rev 423 (L, = 39 ø), showing the flat charac- measurements for clear and dusty conditions. teristic blackbody spectrum of the surface when little atmospheric The period September 9-19, 1976 (L, 120ø-125 ø), has been opacity exists. The prominent feature at 667 cm - is the vibrational isolated for several studies of IRTM data discussed in this set fundamental band of CO:. (b) Rev 18 (L, = 301ø), local time 12.55, of papers. This part of the Viking primary mission is characshowing absorption due to dust suspended in an atmosphere cooler than the surface. (c) Rev 676 (L, = 98 ø), local time 15.39, showing terized by a relatively clear atmosphere, with Viking Lander 1 absorption from water-ice clouds over the Tharsis ridge. (d) Theoreti- visual optical depth in the range [Pollack et al., 1977]. cal absorption spectrum for water ice. (e) IRTM bands are labeled The 7-urn brightness temperature is Chosen to represent surwith the wavelength in micrometers. face temperature in this period; although very' low temperatures are encountered in southern winter and Ta0 has more follows that the fraction of in-band radiance due to the atmosphere is higher at 9/ m than for the 7-, 11-, or 20-/ m bands. The 11-/ m band covers the high-wavelength part of the accuracy, we use T because of the 'known substantial dust opacity at 20 #m (Figure l) and because T will also be required for later, more dusty, time periods. T 0 measurements silicate feature, providing with T measurementsome leverage in southern winter have been presented earlier [Kieffer et al., in determining spectrum shape in this important region. Since 1977]. Where simultaneous data in both bandpasses are availthe H.O ice absorption (Figure 1) extends down to 10/ m able, T?-To.0 never exceeds 3 K in this L, ø period, [Curran et al., 1973], T]] is expected to show some depression except near the polar cap edge (see below). when H.O ice clouds are viewed above a warm surface. The 20-/zm band is used for determination of accurate surface temperatures below 140 K, where the other bands suffer Measured temperatures have been collected and binned according to latitude and local time, with longitudinal information ignored. Thus data from regions of low and high inertia from decreasing blackbody radiance. At 125 K, digitization may be superimposed, though their temperatures the same implies an uncertainty of only 0.6 K in T.0. Thus T.0 is fre- t. ime of day are very different. The result is a gross longitudiquently used for estimation of low surface temperatures, even nally averaged picture of planetary thermal behavior. It should when dust or condensates may be present in the atmosphere. be noted that all longitudes are not necessarily viewed for a The degree to which the 20-/ m band is affected by these given latitude band, so that the longitudinal mixing is not materials is exemplified by the absorption spectra shown in Figure 1' both dust and H.O ice have considerable opacity in this band. The several IRTM brightness temperatures undergo relative complete. The T? data (Figure 3) portray the thermal behavior of Mars in southern winter; the extensive polar cap reaches to about -50 ø. Diurnal variation is very. well behaved, with temperatures indreasing rapidly after local dawn, as predicted changes arising from many phenomena, which are enumerated by the thermal mocj.,el (Figure 4), In equatorial latitudes the here for reference in the following text: (1) atmospheric ab- amplit?,de is well ø Eer 100 K. Within a single sorption and emission by gases, dust, and condensates; (2) time.!l in, the standard deviation of measured T? values is latitude-local

3 ß _ 2832 MARTIN ET AL.: MARS VOLATILES I I ' I I I I - _MARINER Ls REV REV "_ t s.= I i i I I i i I i i i i i i i ' 13, WAVE NUMBER, CM-1 Fig. 2. IRTM relative temperature behavior compared to IRIS spectra for two states of atmospheric dust abundance. IRTM dat are normalized to T7, which is offset vertically by I5 K (the ordinate t;c spacing) for each data set. Data are labeled by the L, of observation; local time is 12 H, and the latitudes included are -25 ø to -30 ø. ; rather high at,midday, typically 7 K. That number expresses the planet in a few days. Viking lander visual optical depths the effect of albedo and inertia variations for the latitude band exceeded 2.0 at + 22 ø [Pollack et al., 1979]. We anticipate a in question. large reduction in peak surface temperatures at this time be- For comparison with the data presented above, we now cause of the transfer of insolation to the atmosphere itself. show.the behavior of T? during the week s a, fter the inception of 61so, nighttim e surface temperatures should be higher because the second and more intense of the two global dust storms of of tlie increased infrared opacity due to dust. Thus the diurnal This time period (June 20-29, L, ø) follows variation should decrease, and the temperature contrast besouthern summer solstice and perihelion. The storm, starting tween higfi and low thermal inertia areas will decrease. in southern atitudes on about June 5, grew to. cover much of We employ T? again to estimate surface behavior because of L s ø 280, _,o-,,o,,o 0/It :::" :5 qo ' I I I I I I I I I I I ' HOUR Fig. 3. Dependence on latitude and local time of 7-t m brightness temperatures L8 = 120ø-125 ø.

4 . MARTIN ET AL.: MARS VOLATILES TMODEL L s ø +10 o , o 2 ' HOUR 'Pig. 4. Sa ½ as Figure 3, for computed model surface tcmpcratur½. the low dust opacity at that Wavelengthi'H owev, er; it is obvious which depict surface temperatt]res for clear.and duat-la. den (Figure l)that the dust cloud contributes to the measured atmospheric conditions encoumer d by Mariner 9. The time radiance in the 7-#m band. For a 7-#m optical depth of 0.2, for bf maximum is the same'(13 H) for the 'clear' peri0ds,' and example (see discussion below), it is calculated that the dust the' maximum'of 275 K in {he Mariner 9 data. is just 5 K below cloud has an emissivity of 0.05 and a transmissivity 0.95 (paper the maximum in Figure 3. The data are Widely separated'in L 4). Thus for a noontime T7 measurement of 267 K and an (-340' versus-1200); the latitudc of maximum is s.ubsolar in assumed cloud temperature of 248 K (&hich is the 9-#m tem- both cases, 91though 'the IRTM data sho TM little!atitudinhl perature), we have underestimated the surface temperature by 'dependence from +5ø to +35 ø during daytime.,that iõ, as -3 K. At night, T7 will overestimate the surface temperm (, shown in the. model temperatures (Figure 4), the r sult of with an error probably not exceeding 1 K. prolonged warming in the nor th during the preceding spring. The 7-#m temperatures (F gure 5) for the veryduõty condi - The dusty-period T,.data from Mariner 9 and the IRTM tions sho TM the expected behavi6r: midday temperatures are (Figure 5) cover similar' L rang ; and show similar nighttimi about K below the model prediction for that season minimum and daytime maximum temperatures (-210 and 265 (Figure 6), and predawn temperatures are a6out K higfi, K, respectively; in southern la!;i tudes). However, the extreme for equatorial latitudes. The diurnal ampli'i ud e has decreased southern extension 0f the maximum in thee IRTM data is not from 100 K to about 55 K, and the standard deviation n,ear present in the other data set;. probably that is the rcsul[0f the noon is down to 3 K in equatorial latitudes.. Mariner 9 dat a. encompassing a less dust-laden period (the There is a shift in peak tempei'ature for latitudes -20 ø t$ storm had. certainly diss{pat6d substantially by rev 85, the end -40 ø of about I H (1 H = Yo. Martian day), to 14 H. of the period covered). maximum temperature regim extends much farther south The change in surface temperature behavior effected by a than in the model. Both of these effects are, qn accord with dust storm is.impo'r. tant in several respects. The driving force the 15-#m band atmospheric temperatures measured at the for lower atmosphere radiativ6-convective activity is weaksame time (paper 2). To the extent that the warm atmosphele ened. Boundary layer gradients are diminished, and static radiates downward, heating the surface, and considering the airborne dust contribution to T? radiance, we expect a phase lag in T data, and an extension of diurnal maxima southward. The model, it should be noted, is inaccprate near the polar cap edge, and gives rise to a false latitudinal gradient in stability increases. Deposition of solar energy in the dust-laden atmospherenhances the coupling Of surface-atmosphere temperatures still further, so that the conditions for violent raising of dusthat must earlier hav existed are quelled, allowing dust storm dissipation to proceed. The thermal contrast is also the region south of-70 ø. decreased across boundaries of dissimilar thermal inertia and The T? data of Figures 3 and 7 bear comparison to the IRIS 7.7-#m brightness temperature data of Hand et al. [1972], albedo; winds stimulated by differential heating of air above these terrains must decrease in magnitude, further sapping the

5 2834 MARTIN ET AL.: MARS VOLATILES. poiential for continued raising of dust. The atmosphere be- the surface is not sensed. Thus T7 - T can provide valuable comes a thermal blanket, moderating surface thermal behav- information on lapse rates within dense local dust storms ior. (paper 3) or fresh global storms. At night the surface cooling ::- SPECTRAL CHARACTERISTICS relative to the atmosphere may cause the sign of T - Tg to reverse. In situations when T - T is small, which may be clear or isothermal dusty conditions, T, is of great value in The efulness of the IRTM bands has been mentioned distinguishing the true behavior of the atmosphere, since T, above. We now discuss the value of differential temperature measurements: the differences T - Tj, where i and j denote will be much lower in the clear case. The difference T - T is of potential value also for the wavelength passbands. Since the IRTM instrument is com- discrimination of areas with nonblack surface emission due to prised of four co-aligned telescopes, each covering different compositional factors. However, the Martian atmosphere is spectral' regions, simultaneous measurements of a single rarely clear enough that the airborne dust contribution can be areare p0ss!ble in several bands. When datare binned by ignored. Unless strong spatial correlations between geologic latitude and,local hour, as, for example, in Figures 3-6, the features and T - T can be established, suggesting control of abundane of data permits forming differencesuch as T7 - which are not possible on an instantaneous basis for a given field of view because the fields of view in these bands are not coincident [Chase et al., 1978]. Since the 9-#m band temperature is much more strongly affected by ai'rborne dust than TT, the difference T - Tg is of obvious interest. This differencexpresses, to first order, the depth of the dust-induced absorption (or emission) band, in Kelvin units. Given the presence of thermal contrast between the surface and the lower atmosphere, T - T is indicative of that difference at the surface, it is likely that T7 - T (or in fact other differentials) will be interpreted in terms of atmospheric effects. When minimal quantities of dust are present in the atmosphere, water ice clouds may form. The difference Tg - Txx then becomes useful, since the 1 l-#m band encounters a strong water ice absorption feature (Figure l) and T is unaffected to first order (paper 4). This difference was employed to study clouds near the volcano Ascraeus Mons in the Viking primary mission [Kieffer et al., 1977]. It was originally anticipated that the presence of dust in the air. Such dust will affect the thermal T and Tx could be used together to note changes in surface contrast, so the relation between T - T and dust content is not Completely straightforward (see paper 4). T will always refer to lower altitude than T, even when dust is so thick that thermal emission associated with the silicate band; the shape of that feature is known to correlate with mineralogic properties [Hunt and Salisbury, 1974, 1975]. The aforementioned I I I I I _ T? L s ø 190-2O SUN I I HOUR Fig. 5. Same as Figure 3, for 7-t m temperature at L8 = 283o-289 ø.

6 MARTIN ET AL.' MARS VOLATILES TMODEL ø lo O 27O 28O SUN -2O -30-7O HOUR Fig. 6. Same as Figure 5, for computed model surface temperature. atmospheric effects of dust and ice have thus far precluded use below, and paper 4). With increasing emission angle, the of T9 - T for geologic studies. relative contribution of the atmosphere will increase in each One more important brightness temperature difference band. The differing opacities at the several wavelengths and should be mentioned: T7 - T.0. The large wavelength range the angular variation provide many variables of use in deriving between these two bands emphasizes the effect of ani- vertical temperature profiles. Moreover, the infrared optical sothermality, the variation of temperature within a field of depth of the atmosphere can be gaged from examination of view. Because radiance is a stronger function of temperature I RTM temperatures. One particular approach to the'deriva- 7 #m than at 20 #m, the effective brightness temperature of a tion of optical depths is employed in the discussion below. A mixed-temperature scene will be higher at 7 #m than at 20 #m, more thorough treatment of a larger body of data is planned; leading to a positive difference T7 - T.0. The physical source of the global and seasonal behavior of optical depths would be the temperature variation can be thermal inertia, albedo, shad- valuable in storm generation and global circulation modeling. owing, or combinations of these. Conceivably, internal heat We now present differential temperature data for two very source such as cooling lava lakes could be responsible. The different Martian seasons and examine the effectiveness of this response of the IRTM to small 'hot spots' has airearly been representation. The format is identical to that used for surface discussed (Kieffer et ai. [1977]; see also discussion of surface temperatures above; data are averaged over longitude and block model). Given adequate diurnal coverage of a particular expressed versus local time and latitude. area, the difference T7 - T.0 could be employed to infer the First, for the case of a'relatively clear atmosphere, we examnature of fine-scale topography below the instrumental spatial ine the difference T, - T9 for part of the primary mission resolution. T.0, of course, suffers from airborne dust effects discussed earlier (Figure 7). The positive T? - T, values occur- (Figure l). The largest T, - T.0 to date attributable to ani- ring south of-40 ø may be partly due to.anisothermality sothermality is 16 K, encountered at the south polar cap edge effects at the polar cap edge, but in view of the low tempernear noon near L8 228 ø. atures there, T, and the differential measurement both have In a dusty atmosphere, the various IRTM bands all sample high uncertainty. At -30 ø there is virtually no diurnal Change different levels. The atmospheric band at 15 #m(paper 2) will in T, - T,. Further north, a very slight diurn change probe somewhat higher levels when high-altitude dust is pres- with maximum amplitude about 6 K. At night, T9 exc ds T?; ent, since the dust opacity will block radiance from the lowest the pattern reverses near midday, implying that thermal emislevels to a small extent. The Other bands will generally sample sion from airborne dust affects T, and that the atmosphere is the surface to varying degrees (see relative opacities given warmer than the surface at night. Because in-band radiance is occurs,

7 2836 MARTIN ET AL.: MARS VOLATILES not a linear function of temperature, the atmospheric effect on T7 - To is more pronounced at night than during the day, when surface radiance dominates in both channels. Apparently, dust affects IRTM temperatures even in relatively clear conditions such as the Viking primary mission (Ls near this latitude, so that there may also be a higher dust abundance there. Latitudes distant from the subsolar point (-24 ø ) should suffer additional reduction in surface temperature variation, and thus in T,. However, as we have seen, the diurnal maximum in T, at -60 ø is nearly as great as at -15 ø (Figure 5); similar behavior obtains for Tg (Figure 9). This 85ø-147 ø). However, the effect is small in IRTM passbands other than 9/ m, and the conclusions reached [Kieffer et al., 1977] by ignoring atmospheric effects are still valid. Atmospheric effects must be modeled in order to interpret much of the IRTM data obtained during the Viking extended mission (Ls 165ø-360ø). We now examine the most dusty time period of the Viking mission to date: the days following the second global storm of implies that there is less dust present in the southernmost latitudes; the surface could warm more readily, as revealed by T,, and the Tg would refer to a lower level in the atmosphere. The evidence given by T 5 (paper 2) is interesting in this regard, since it, too shows a very southerly peak temperature at 16 H. The delayed peak in T at -60 ø suggests that the 9-/ m channel is sampling regions between those seen at 7 and June 1977 ( L, 286ø). Here, airborne dust certainly affects 15/ m. The behavior at 0.5 mbar revealed by Tx5 is probably brightness temperatures profoundly. The difference T7 - T, (Figure 8) is chosen again because of the great disparity in dust opacity at the two wavelengths. We estimate the vertical optical depth at 9 um to reach a maximum of 2.0 (see below and paper 4); most of the magnitude of the differential is due to strong dust absorption at 9 um. The diurnal variation seen in equatorial latitudes for the earlier data has now become a governed by tidal phenomena, which do not substantially influence surface temperatures but may be responsible for shifts in heating characteristics throughout the atmosphere. If we take T7 to represent surface temperature and Tx to be the temperature at 25 km, it is possible to estimate very roughly the level sampled by T. This process will be inaccurate to the extent that we ignore boundary layer discontin- global characteristic, and greatly magnified. Again, T, - T, is uities, the error in T, as a surface temperature indicator, and negative at night; the difference is as much as -12 K at -70 ø the nonlinearity of the temperature profile. Considering the latitude. The midday differential is a minimum at the south- time of peak surface temperature at latitude -35 ø, 14 H, ernmost latitudes and increases consistently with latitude to we have for example T, = 265 K, T = 247 K, and Tx = 225 about 22 K at -15 ø. The maximum diurnal change occurs K. With a linear lapse rate assumed, the value of T implies a near -15 ø also, implying that the magnitude of the difference level of 11 km. The same process applied to data at 14 H, depends primarily on insolation. The heating available at the -62 ø, yields a level near 7 km for Tg. These numbers imply surface is a strong function of latitude in dusty conditions that T is a potentially valuable indicator of conditions in the [Leoin et al., ]977]. However, the storm probably originated lower atmosphere. The total set of IRTM temperatures, con- 0] I I I I I I 0 I.! x?-b /0 I I I I HOUR Fig. 7. Dependence on latitude and local time of the difference T7 - Tg for Ls = 120ø-125 ø.

8 MARTIN ET AL..' MARS VOLATILES O 1' 7 - T 9 L s ø 2O SUN O -6O -7O HOUR 24 sidering time and spatial variation, can strongly constrain models of thermal structure. We next examine the seasonal and diurnal variation of T7 - Te for certain latitude bands, to define the way in which this valuable parameter indicates the change in degree of global atmospheric dustiness. The data are averaged over longitude as before (Figure 10). The earliest period, Ls 122 ø, correspohds to the large-scale presentation made in Figures 3 and 7 of primary mission data. We see that T? - Te has a slight diurnal variation for the more northerly latitude of less than 3 K. No data for T are available in the -55 to -60 ø band because of the low temperatures associated with the CO2 frost cover there. At Ls 189ø,.diurnal amplitude is increased at both latitudes, implying a general increase in atmospheric dust abundance. Local storms around the receding south polar cap are thought to be the source of this slow rise in dustiness during southern winter (paper 3). The global nature of the increase is ascertained by the fact that both Viking landers recorded a rise in visual optical depths over the same period [Pollack et al., 1977]. At L 228 ø, the first global dust storm (L 207 ø) has begun, and the resultant raised dust has been distributed globally. The difference T, - T, has grown in diurnal amplitude at both latitudes to more than 20 K. Both positive and negative differ- Fig. 8. Same as Figure 7, for difference T7 - Te at Ls = 283ø-289 ø. ences occur, indicating that the atmosphere is warmer than the surface at night. By L, 259 ø, the dust has dissipated to a large extent in the interim period between the major storms, and T? - T reflects the corresponding change in thermal contrast between dust cloud and surface. The second global dust storm (L, 275 ø) injected even more dust into the atmosphere than the first, as evidenced by the greater infrared optical depths (see below). T? - T values for Ls 288 ø are the largest encountered, both in absolute value and in diurnal amplitude. There is an interesting tendency for T? to approach T at night in the -25 to -30 ø latitude band. Possibly, strong atmospheric backwarming of the surface could keep it from cooling effectively at night. Alternatively, the 7- and 9-#m bands may sample layers bracketing a high-level temperature inversion. At the more southerly latitude, T? - T goes to -10 K at night. The possibility of lower dust abundance in the south, mentioned earlier, may explain the ability of the surface layers to cool more completely. Following the second storm, the differential between T? and T decreasesystematically, shown by the data for L 334 ø and 352 ø. By October 1977 (L, 352 ø) the atmosphere has apparently cleared substantially, although at -25 ø to-30 ø latitude, it is still not as clear as it was during the L, 122 ø period.

9 2838 MARTIN ET AL.: MARS VOLATILES The seasonal trend of T7 - Tg is remarkably consistent with the behavior of lander optical depths [Pollack et al., 1977, 1979] and with infrared opacities derived at various locations (see below). We have presented in Figure 10 only a very small fraction of the data of this kind that are available to portray global dustiness as a function of season. ATMOSPHERIC INFRARED OPACITY At frequent intervals throughout the Viking mission, special observational sequences denoted as emission phase functions have been used to measure the thermal radiation from localized areas of the planet over a wide range of emission angles. These sequences were originally intended to (1) define the surface photometric function and (2) determine the variation of surface thermal emission with emission angle. Prior to Mariner 9 it had been tacitly assumed that, except for major dust storms and large-scale cloudiness, the atmosphere was essentially transparent over most of the visual and infrared spectrum. Mariner 9 IRIS data gave the first indications that this assumption was incorrect. Thermal infrared data from Viking have further shown that even for relatively dust-free conditions, atmospheric effects are noticeable in all wavebands measured by the IRTM [Kieffer, et al., 1977]. As a result, the emission phase function sequences have proven to be a valuable source of information for a third type of study, namely, the determination of atmospheric infrared opacities. The ITRM 15- m channel is centered on a major CO: absorption band, and is indicative of atmospheric brightness temperatures near the 0.6-mbar pressure level, corresponding to an altitude of about 25 km. Under the assumption of pure absorption (no scattering), the thermal radiance at other wavelengths may be represented by Rx(Tx) = Rx(Ts)e -rxm + Rx(TaXI - e -rxm) (1) where Rx(Tx) is the radiance at wavelength 2, corresponding to the observed brightness temperature at that wavelength, Rx(Ts) is the radiance corresponding to the surface temperature, Rx(T.) is the radiance corresponding to the 'mean' atmospheric temperature at the level where the thermal opacity is present, rx is the optical depth through one air mass at wavelength 2,, and M is the air mass. Solving for rx, one obtains rx = I In IRx(T,)-Rx(T.)I Rx(Tx)- Rx(T ) (2) I I I I 'L' / L s 282t ø O SUN -2O O I 22ø1. I HOUR Fig. 9. Same as Figure 8, for T,

10 MARTIN ET AL.: MARS VOLATILES LAT = -30 TO -2.5 L S _ ß o 189 -! o v estimate. Under clear conditions, an average temperature lapse rate of about q-2.5 K/km applies, implying that nearsurface atmospheric temperature should be K warmer than Txs. However, when dust in the atmosphere reaches a visual optical depth of unity, the atmospheric lapse rate is greatly reduced. For simplicity, we will here equate atmospheric temperature to Txs, recognizing that for relatively clear conditions, Tx will be an underestimate of the relevant atmospheric temperature. For such clear conditions the brightness temperatures at 7, 9, 11, and 20 #m are nearly identical anyway, leading to low values of the infrared opacity essentially independent of the atmospheric temperature. With the above assumptions, (2) becomes F. Rx_(Ts) - Rx(T ) ] rx = cos e In LRx(Tx) - Rx(T ) (4) where is the emission angle and M has been approximated by 1/cos. A total of 39 emission phase functions occurring L T between 9 and 17 H have been analyzed to date. For each of these, a best-fit cosine power law was determined for the thermal radiances at each wavelength. Only data at emission angles less than 60 ø (M < 2) were used. From these fits, the vertical-viewing values of T, T?, Tg, Tn, T:o, and T have been determined. These determinations are generally accurate to within a few tenths of a Kelvin or better. Infrared opacities at 0 7, 9, 11, and 20 #m are then calculated. The results are given in Table I. No account has been taken of the fact that the data are from widely varying locations on the planet. In spite of this, the data reveal a relatively smooth progression of opacities as a function of time. Obviously, if any of the data had been taken in an area where a local dust storm was occurring (paper 3), deviations from this longitude-independent behav LOCAL HO0 ior would be apparent. A possible conclusion, and one that we postulate as being representative of general Martian behavior ig. 10. Diurnal beh vio b ds nd v fious se so s. during the Viking mission, is that the dust opacity increases or decreases relatively uniformly over the disk of Mars, especially during the global dust storms. If Ts >> Ta, a circumstance which occurs between local ObServations of the summit of Olympus Mons near the time times of about 9 H and about 16 H, small errors in T and/or of peak opacity from the first dust storm (L 226 ø) yield Ta have little effect on the calculated opacities. As mentioned near-zero opacities, indicating that the dust was primarily earlier, the brightness temperature at 7 #m, T7, is a fair ap- confined to lower levels in the atmosphere (the summit of proximation for actual surface temperatures, except in the Olympus M ons is near 27-km elevation). Similarly, observacase of very dusty conditions. Since absorption undoubtedly tions on the flank of Ascraeus Mons at L 279 ø (elevation occurs at 7 #m in extremely dusty conditions, midday T?'s ca. 9 km) show opacities 30-60% lower than might be premay be slightly depressed from actual surface temperatures. dicted on the basis of observations which surround these in Because emission phase function observations are'relatively time. These observations and others provide some information short in duration, one can use such measurements during on the scale height of mixing of the dust. If one assumes that dusty periods to determine from (2) the best fit asymptotic the distribution of dust with altitude is the same for the observalues for T and T by assuming rx remains constant during vations above Olympus Mons on March 22, 1977 (L 226ø), the observation and by observing the variation of Tx with M. and above Melas Chasma on March 23, 1977 (elevation differ- We estimate on the basis of T? variation with emission angle ence 27-3 = 24 km), one arrives at a scale height of 10 km during dusty conditions that for local times between 9 and 16 H. for the dust. This is essentially identical with the atmospheric scale height, indicating that the dust is well mixed. The self- T.= + o.3(t, - (3) consistency of the infrared opacities from differing locations is much improved if one applies an adjustment for surfac eleva- This approximation would not hold for the case where water- tion using the elevations from the 1: 25M USGS Mars Topoice fog opacity is present, but only for dust absorption. Since graphic Atlas (1976). This adjustment has been made as folground fogs are known to exist during morning hours, espe- lows for 9-#m opacities: cially early in the Viking mission [Pollack et al., 1977], data from local times earlier than about l0 H should be used rd = r e In+: /'ø (5) with caution. The maximum T? - T difference approaches 23 K (Figure 10) so that the Ts - T? difference can be as large as 7 where h is the surface elevation in kilometers and the 2 is K. included to facilitate easier comparison with VL-1 observa- Atmospheric temperatures are somewhat more difficult to tions (VL-1 elevation = -2 km). The results are given in the

11 2840 MARTIN ET AL.: MARs VOLATILES TABLE 1. Emission Phase Fuhctions: Infrared Opacities Day LT; Rev. No. Ls H Lat. Long. r9 r r2o h, km T 9' VO-I, Ill (I) (I) } i VO-I, / VO-2, (1) (1) (2) q (3) (4) (5) (6) (1) VL-I site, (2) Olympus Mons, (3) Syrtis Major, (4) Ascraeus Mons flank, (5) Hellas, and (6) VL-2 site. last column of Table I and plotted versus time in Figure 11. The effects of the two global dust storms are clearly illus- trated b 'the figure. The rapid rise in opacity near the beginning of the second storm ( Ls 275 ø) was particularly noticeable. The remaining scatter in the data is possibly due to a combination of errors in the elevations and spatial inhomogeneity of the air-borne dust quantities. lffpart cular, an observation of Hellas on July 24, 1977 (Ls 303 ø), shown by an open circle in Figure 11, yields an adjusted opacity much lower than that expected. Either the opacity in Hellas is less by a factor of 2 than at other locations on the planet at the same time, or the elevation is too low by almost 7 km, or a combination of the two effects. cities; 11- and 20-#m opacities are about equal and about 60% as large as the opacity at 9 #m. At the start of the global dust storms the 9-#m opacity is about Corresponding visual opacities from lander data were near 1.0 at the beginning of the global dust storms [Pollack et al., 1979]. If the same ratios apply at larger optical depths, the visual optical depths should have reached maxima of about 3 and 8 during the first and second global dust storms, respectively. These opacities are surprisingly close to the values estimated by Pollack et al., by backward extrapolation of observed optical depth decay rates at VL-I. Several cautionary statements are appropriate, however, Lander observations of opacity are made at low sun elevation angles; IRTM opacity measurements are generally One interesting feature of the data concerns the relative best near midday. Diurnal opacity variation like that deduced opacities at different wavelengths. The 9-#m opacities during by Thorpe [1977] from Viking orbital imagery would negate dusty periods are l;arger than corresponding opacities at 7, 1 l, any direct correlation between lander visual opacity measureand 20 #m. The 7-#m opacities are about 20% of 9-#m opa- ments and the infrared opacities given here. Such diurnal

12 _ MARTIN ET AL.: MARS VOLATILES I UTC DAY OF YEAR i I I I I I I I 35O o _ ß I ST GLOBAL STORM SfARTS 21 ) GLOBAL STORM STARTS ß I I J I I I I, AREOCENTRIC LONGITUDE OF SUN, L s Fig. 11. The 9-#m opacity derived from IRTM temperatures and adjusted to VL-1 surfac elevation by assuming a dust scale height of 10 kin. Open circle represents a measurement in the Hellas basin and is discussed in the text. The solid curve is the authors' estimate of a best fit to the data. opacity variations would most likely be due to water-ice fogs or clouds in the lower few kilometers of the atmosphere. When the amounts of atmospheric dust become sufficient (rvis > 1), the formation of water-ice clouds is inhibited, and diurnal variations should greatly diminish or disappear altogether. For comparable visual opacities, the 9-#m opacity of water-ice haze is considerably lower than for atmospheric dust. Opacities measured by VL-1 during 'clear' periods may well be due at least in part to water ice, even for the afternoon times of observation. The relative values of infrared opacities and their qualitative behavior as a function of Ls are well established by the data presented here. These relative opacities at 7, 9, 11, and 20 #m provide constraints on the composition of the suspendedust particles. If one compares the empirically determined relative infrared opacities from this paper to those calculated from theoretical considerations by Hunt [1979] (paper 4), it seems obvious that montmorillonite 219b alone does not provide a good match with the IRTM opacity data. This conclusion is in concert with that reached by Toon et al. [ 1977] from Mariner 9 IRIS data. The assumption of no atmospheric scattering (pure absorption only) may perhaps be the most seriou systematic error in the absolute values of the opacities presented here. For the dust model used in paper 4, namely, montmorillonite 219b with a size distribution given by N(r) = a? exp [-4(r/0.4) /2], where r is in micrometers, Hunt calculates an asymmetry factor of 0.6 at 9 #m. This would mean that the infrared opacities given here for 9 #m are small by a factor of about (1-0.6) - = 2.5, with corresponding errors at other wavelengths. We have not included a correction for multiple scattering in Table 1 and Figure 11 because of the complexity of the calculations and because of the dependence on the specific model parameters chosen to representhe scatterers. Acknowledgments. We wish to express our gratitude to members of the Viking Volatiles and Meteorology Working Group for valuable discussions. The programing effort of R. Mehlman and D. Finnerty made possible timely analysis of data presented here. This paper presents the results of one phase of research carried out in part at the Jet Propulsion Laboratory, California Institute of Technology, under contract NAS (JPL ), sponsored by the National Aeronautics and Space Administration. G. E. Hunt is a Viking guest investigator, supported by the U.K. Science Research Council. REFERENCES Chase, S.C., Jr., J. L. Engel, H. W. Eyerly, H. H. Kieffer, F. D. Palluconi, and D. Schofield, Viking infrared thermal mapper, Appl. Opt., 17, , Curran, R. J., B. J. Conrath, R. A. Hanel, V. G. Kunde, and J. C. Pearl, Mars: Mariner 9 spectroscopic evidence for H:O ice clouds, Science, 182, , Hanel, R., et ai., Investigation of the Martian environment by infrared spectroscopy on Mariner 9, Icarus, 17, , Hunt, G. E., Thermal infrared properties of the Martian atmosphere, 4, On predictions of the structure of the Martian atmosphere in the presence of dust storms and ice clouds from the Viking IRTM spectral measurements, J. Geophys. Res., 84, this issue, Hunt, G. R., and J. W. Salisbury, Mid-infrared spectral behavior of igneous rocks, AFCRL-TR , Air Force Cambridge Res. Lab., Bedford, Mass., Hunt, G. R., and J. W. Salisbury, Mid-infrared spectral behavior of sedimentary rocks, A FCRL-TR , Air Force Cambridge Res. Lab., Bedford, Mass., Kieffer, H. H., Soil and surface temperatures at the Viking landing sites, Science, 194, , Kieffer, H. H., S.C. Chase, Jr., E. D. Miner, F. D. Palluconi, G. Mi nch, G. Neugebauer, and T. Z. Martin, Infrared thermal mapping of the Martian surface and atmosphere: First results, Science, 193, , 1976a. Kieffer, H. H., P. R. Christensen, T. Z. Martin, E. D. Miner, and F. D. Pailuconi, Temperatures of the Martian surface and atmosphere: Viking observation of diurnal and geometric variations, Science, 194, , 1976b. Kieffer, H. H., T. Z. Martin, A. R. Peterfreund, B. M. Jakosky, E. D. Miner, and F. D. Palluconi, Thermal and albedo mapping of Mars during the Viking primary mission, J. Geophys. Res., 82, , Levine, J. S., D. R. Kraemer, and W. R. K uhn, Solar radiation incident on Mars and the outer planets: Latitudinal, seasonal, and atmospheric effects, Icarus, 31, , 1977.

13 2842 MARTIN ET AL.: MARS VOLATILES Martin, T. Z., and H. H. Kieffer, Thermal infrared properties of the and R. Kahn, Properties and effects of dust particles suspended in Martian atmosphere, 2, The 15-/ m band measurements, J. Geophys. the Martian atmosphere, J. Geophys. Res., 84, this issue Res., 84, this issue, Thorpe, T. E., Viking orbiter observations of atmospheric opacity Peterfreund, A. R., and H. H. Kieffer, Thermal infrared properties of during July-November 1976, J. Geophys. Res., 82, , the Martian atmosphere, 3, Local dust clouds, J. Geophys. Res., 84, Toon, O. B., J. B. Pollack, and C. Sagan, Physical properties of the this issue, particles composing the Martian dust storm of , Icarus, Pollack, J. B., D. Colburn, R. Kahn, J. Hunter, W. Van Camp, C. E. 30, , Carlston, and M. R. Wolf, Properties of aerosols in the Martian atmosphere as inferred from Viking lander imaging data, J. (Received May 4, 1978; Geophys. Res., 82, , revised August 3, 1978; Pollack, J. B., D. S. Colburn, F. M. Flasar, C. E. Carlston, D. Pidek, accepted November 15, 1978.)