Thermal infrared observations of the condensing Martian polar caps: CO ice temperatures and radiative budget

Similar documents
Lecture 3: Global Energy Cycle

MARS CLIMATE DATABASE VERSION 4.3 VALIDATION DOCUMENT - DRAFT -

Seasonal variations of the martian CO over Hellas as observed by OMEGA/Mars Express ABSTRACT

1. Weather and climate.

Dust in the Atmosphere of Mars 2017 (LPI Contrib. No. 1966)

The climates of planet Mars controlled by a chaotic obliquity

Modeling Optical Properties of Martian Dust Using Mie Theory

Pfs results at Mars. By V.Formisano and the PFS Team

9/1/14. Chapter 2: Heating Earth s Surface and Atmosphere. The Atmosphere: An Introduction to Meteorology, 12 th. Lutgens Tarbuck

Lecture 2: Global Energy Cycle

The Relative Humidity of Mars' Atmosphere

Radiation in the atmosphere

Fundamentals of Atmospheric Radiation and its Parameterization

Key issue in Mars atmosphere and climate. François Forget LMD, IPSL, Paris

Earth s Energy Budget: How Is the Temperature of Earth Controlled?

Name(s) Period Date. Earth s Energy Budget: How Is the Temperature of Earth Controlled?

Planetary Atmospheres

Planetary Atmospheres

Electromagnetic Radiation. Radiation and the Planetary Energy Balance. Electromagnetic Spectrum of the Sun

Surface Observations Including from the 2012 Mars Curiosity Rover. Martian Atmosphere

providing 100-m per pixel resolution in nine ~1.0 µm wide infrared bands centered from

Lecture 2: Global Energy Cycle

Solar Flux and Flux Density. Lecture 2: Global Energy Cycle. Solar Energy Incident On the Earth. Solar Flux Density Reaching Earth

MCD General Description

Lecture 2 Global and Zonal-mean Energy Balance

Chapter 3. Multiple Choice Questions

ATMS 321: Sci. of Climate Final Examination Study Guide Page 1 of 4

Chapter 2. Heating Earth's Surface & Atmosphere

Torben Königk Rossby Centre/ SMHI

CONSTRUCTION OF A 4D WATER ICE CLOUD DATABASE FROM MARS EXPRESS / OMEGA OBSERVATIONS DERIVATION OF THE DIURNAL MARTIAN CLOUD LIFE CYCLE

Radiation in climate models.

Meteorology Pretest on Chapter 2

The Main Point. Basic Properties of Mars. Observations. Lecture #19: Mars

Lecture 3: Atmospheric Radiative Transfer and Climate

Friday 8 September, :00-4:00 Class#05

Chapter 10 Planetary Atmospheres Earth and the Other Terrestrial Worlds. What is an atmosphere? Planetary Atmospheres

General Comments about the Atmospheres of Terrestrial Planets

Water Ice Clouds over the Martian Tropics during Northern Summer

Evolution of the Martian Climate and atmospheric escape

Outline. Planetary Atmospheres. General Comments about the Atmospheres of Terrestrial Planets. General Comments, continued

Data and formulas at the end. Exam would be Weds. May 8, 2008

MAPH & & & & & & 02 LECTURE

The Cosmic Perspective Planetary Atmospheres: Earth and the Other Terrestrial Worlds

Planetary Temperatures

Chapter 10 Planetary Atmospheres Earth and the Other Terrestrial Worlds

Radiative Equilibrium Models. Solar radiation reflected by the earth back to space. Solar radiation absorbed by the earth

Chapter 10 Planetary Atmospheres: Earth and the Other Terrestrial Worlds. What is an atmosphere? Earth s Atmosphere. Atmospheric Pressure

Chapter 10 Planetary Atmospheres: Earth and the Other Terrestrial Worlds

Meteorology Practice Test

Spectrum of Radiation. Importance of Radiation Transfer. Radiation Intensity and Wavelength. Lecture 3: Atmospheric Radiative Transfer and Climate

Sensitivity of climate forcing and response to dust optical properties in an idealized model

2. Meridional atmospheric structure; heat and water transport. Recall that the most primitive equilibrium climate model can be written

Lecture 9: Climate Sensitivity and Feedback Mechanisms

Understanding the Greenhouse Effect

The orbital forcing of climate changes on Mars

The Mars Climate Database (MCD version 5.2)

Chapter 10 Planetary Atmospheres Earth and the Other Terrestrial Worlds

Solar Insolation and Earth Radiation Budget Measurements

Energy Balance and Temperature. Ch. 3: Energy Balance. Ch. 3: Temperature. Controls of Temperature

Energy Balance and Temperature

Chapter 10 Planetary Atmospheres: Earth and the Other Terrestrial Worlds. What is an atmosphere? About 10 km thick

Warming Earth and its Atmosphere The Diurnal and Seasonal Cycles

MULTIPLE CHOICE. Choose the one alternative that best completes the statement or answers the question.

Lecture Outlines PowerPoint. Chapter 16 Earth Science 11e Tarbuck/Lutgens

Energy Systems, Structures and Processes Essential Standard: Analyze patterns of global climate change over time Learning Objective: Differentiate

Today. Events. Terrestrial Planet Atmospheres (continued) Homework DUE. Review next time? Exam next week

HEATING THE ATMOSPHERE

Simple energy balance climate models

MULTIPLE CHOICE. Choose the one alternative that best completes the statement or answers the question.

Prentice Hall EARTH SCIENCE. Tarbuck Lutgens

Solar-terrestrial coupling evidenced by periodic behavior in geomagnetic indexes and the infrared energy budget of the thermosphere

Planetary Atmospheres (Chapter 10)

Mars photometry with OMEGA observations

Simplifying the martian carbon dioxide cycle: An empirical method for predicting surface pressure

Chapter 3- Energy Balance and Temperature

THE EXOSPHERIC HEAT BUDGET

AT350 EXAM #1 September 23, 2003

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Physics Problem Solving 10: The Greenhouse Effect. Section Table and Group

Improving the representation of the martian water cycle in the Global Climate Model of the LMD*

Lecture 11: Meridonal structure of the atmosphere

ATMOS 5140 Lecture 1 Chapter 1

Investigation of the nature and stability of the Martian seasonal water cycle with a general circulation model

Atmospheric Lidar The Atmospheric Lidar (ATLID) is a high-spectral resolution lidar and will be the first of its type to be flown in space.

ATMS 321 Problem Set 1 30 March 2012 due Friday 6 April. 1. Using the radii of Earth and Sun, calculate the ratio of Sun s volume to Earth s volume.

Atmospheric Radiation

The Energy Balance Model

MODELLING THE MARTIAN WATER CYCLE

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. E9, 5098, doi: /2003je002058, 2003

The Arctic Energy Budget

Lecture 7: The Monash Simple Climate

2010 Pearson Education, Inc.

Impact of the 2002 stratospheric warming in the southern hemisphere on the tropical cirrus clouds and convective activity

Terrestrial Atmospheres

COURSE CLIMATE SCIENCE A SHORT COURSE AT THE ROYAL INSTITUTION

Insolation and Temperature variation. The Sun & Insolation. The Sun (cont.) The Sun

Ozone retrieval from SPICAM UV and near IR measurements : a first global view of ozone on Mars

CLIMATE CHANGE Albedo Forcing ALBEDO FORCING

Observation of Martian Polar CO2 clouds using the MOLA and TES Instruments: An Exploratory Study. By Kaj Williams April 2003

Hand in Question sheets with answer booklets Calculators allowed Mobile telephones or other devices not allowed

Monday 9 September, :30-11:30 Class#03

Transcription:

Published in: Journal of Geophysical ResearchVOL.11-E7 16,865 16,879, july 25, 1996. Thermal infrared observations of the condensing Martian polar caps: CO ice temperatures and radiative budget François Forget Laboratoire de Météorologie Dynamique du Centre National de la Recherche Scientifique, Ecole Normale Supérieure, Paris, France James B. Pollack NASA Ames Research Center, California Abstract. The physical processes that control the formation of the Martian seasonal polar caps are not completely understood. On the one hand, climate models reproducing the annual variations in atmospheric pressure caused by the condensation of the polar caps have shown that the amount of CO actually trapped in the polar regions in winter is lower than expected from simple energy balance considerations. On the other hand, the available spacecraft observations of the condensing polar caps are complex and puzzling. They are characterized by highly variable low-emission zones exhibiting anomalously cold brightness temperatures. To better understand these results, we have carefully reanalyzed the Viking infrared thermal mapper (IRTM) measurements obtained during the polar night in both hemispheres. First, by removing the signature of the low-emission zones in the data, we have retrieved the actual surface temperatures of the polar caps. We find that they were lower than the frost point of CO for the topography of the polar regions usually used in models, especially in the south polar region. However, our analysis reveals that the low-emission zones were more frequent and more intense in the northern hemisphere. They strongly altered the polar radiative budget which is computed and analyzed here, and thus the CO condensation rate. We conclude that the models tendency to overestimate the amount of CO ice condensing in the polar caps is explained by different causes in each hemisphere. In the north, the models did not simulate the low-emission zones and underestimated the heat advected to the polar cap region during the dust storms, especially by the upper atmosphere polar warming. In the south, they overestimated the polar cap surface temperatures and also did not simulate the low-emission zones. 1. Introduction During the polar night in each polar region of Mars, carbon dioxide, the major ingredient of the Martian atmosphere, condenses when temperatures become cold enough to form CO ice deposits on the surface. This seasonal CO cycle represents a unique and important aspect of the current climate of the planet. Its influence is not restricted to the high Martian latitudes: every year, as much as 4% of the atmosphere takes part in this seasonal CO cycle, causing large-amplitude seasonal fluctuations of the surface pressure Deceased, June 13, 1994. which were measured by the Viking Landers. Understanding the processes that control the CO cycle is thus important for understanding the past and current Martian climates. The process of formation of the polar caps seems to be simple. In theory, the amount of CO which condenses in a combined atmosphere + surface column is determined by its energy balance equation, which gives to first order during the polar night,! #" %$'&)( (1) where! * + is the total CO condensation rate, ( is the latent heat of CO (,.- /13254#6 J kg798 ), the thermal in-

2 FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS frared flux of the system toward space, : the heating rate due to heat transport by the atmosphere (advection), and! ;" the subsurface heat conduction flux (positive when the subsurface is warmer than the surface). Note that the total rate of CO condensation in the winter polar region (atmospheric plus surface) decreases monotonically as the infrared emission to space decreases. Based upon such energy balance considerations, many models have been developed to simulate the formation and the recession of the Martian polar caps. Most recent models have been able to reproduce the general forms of the cap seasonal evolution including the annual pressure variations recorded by the Viking Landers [e.g., James and North, 1982; Wood and Paige, 1992; Pollack et al., 1993; Hourdin et al., 1993, 1995]. However, these studies found that the amount of CO actually condensing in the polar cap was about 3% lower than predicted by equation 1 if the caps were assumed to be blackbody emitters. To fit the Viking pressure data, they thus used low values (e.g.,.7) for the modeled cap emissivities. Such values are likely to be artificial. For instance, Paige [1985] showed that the cap emissivities were close to unity at least during the early fall and late winter seasons. In fact, the actual physical processes that explain the observed low condensation rate are still poorly known. Our approach in this paper is to come back to the available observations of the condensing polar caps to learn more about these processes. Two complementary data sets are currently available: the Mariner 9 infrared interferometer spectrometer (IRIS) spectra and the Viking infrared thermal mapper (IRTM) brightness temperatures. Until now, these data have been difficult to analyze because of their puzzling characteristics. The IRTM instrument measured 2- < m brightness temperatures showing considerable structures with anomalously low values in the winter polar regions [Kieffer et al., 1976], far below 148 K, the temperature appropriate for condensed carbon dioxide in vapor pressure equilibrium at the expected atmospheric pressure. The location and brightness temperatures of these areas (hereafter also called low emission zones ) sometimes varied on timescales of days [Kieffer et al., 1977]. The low emission zones were also characterized by their complex spectral signature observed by Mariner 9 IRIS [Paige et al., 199]. Various hypotheses have been suggested to explain these observations [see Kieffer et al., 1977; Paige, 1985]. Recently, Forget et al. [1995] (referred to as paper 1 hereafter) used a combination of radiative transfer modeling and IRIS and IRTM data analysis to show that the low brightness temperatures were likely to be created by the radiative properties of CO snow falls and that both falling snow particles and fresh snow deposits could contribute to create the observed features (Fig. 1). In both case, the infrared emission is decreased because CO snow grains larger than 1 < m can be efficient scatterers at infrared wavelengths, whether they are airborne or on the ground. Conversely, CO ice deposits composed of nonporous solid ice having directly condensed on the ground or having undergone frost metamorphism should behave almost like blackbody emitters. In fact, it is likely that such a solid CO ice layer will be transparent in the infrared and that the ground beneath will radiate through, with a high emissivity. Such processes reducing the observed brightness temperatures are likely to have a strong impact on the radiative balance of the polar regions. Thus they should partly explain the low condensation rate inferred from the Viking Lander pressure measurements. Paige and Ingersoll [1985] previously used a subset of the IRTM data to determine annual radiation budgets and infer annual CO frost budgets for the core regions of the north and south residual caps. In the present paper, we extend this study to the entire seasonal polar caps during the fall and winter seasons and apply the results of paper 1 to process the IRTM data. After a brief description of the IRTM data set and its interpretation (section 2), we present results of detailed analyses of the IRTM data carried out to determine the actual surface temperatures of the seasonal polar caps (section 3), estimate the spatial and temporal distributions of the low emission zones (section 4), and evaluate the polar region infrared emission toward space and the various components of this emission (section 5). Indeed, the infrared emission may not only be affected by the scattering properties of the CO fresh surface snow and clouds, but also by the absorption and emission of the gaseous atmosphere in the 15- < m band or the presence of aerosol such as airborne water ice or dust. Finally, we compare our results concerning the energy balance of the polar caps with results of polar cap models (section 6). 2. Viking IRTM Data 2.1. Data Set The infrared thermal mappers (IRTM) of the two Viking orbiters observed the entire planet during almost two Martian years [Kieffer et al., 1977]. The instruments operated in five broad infrared channels, but only the 11- < m, 15- < m, and 2- < m channels are reliable to study very cold areas like the polar caps. The 7- < m and 9- < m channels are too noisy, whereas the temperature equivalent noise of = 8>8 is about 1.6 K at 15 K, 2.6 K at 14 K, and 3.4 K at 135 K. This noise is less than.5 K for = >?. The 11- < m and 2- < m channels are unaffected by gaseous absorption. Thus, they usually measure the surface brightness temperature, more or less affected by the presence of aerosol (dust, H O, or CO ice) which can absorb, reemit or scatter the ground emission. The 15- < m channel, with a spectral response centered on a strong gaseous CO absorption feature, measures atmospheric temperatures over a broad range of altitudes with maximum sensitivity near the.6-mbar pressure level. We have analyzed IRTM data measured during the fall and winter seasons and also looked at the data obtained during the spring. We included data poleward of 4@ latitude (lat) with an emission angle lower than 6@. More than four million observations of this kind are available, where

FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS 3 a b c d Tb = Tcap Tb << Tcap Tb < Tcap Tb = Tcap CO2 snow backscatt. Metamorphism Emissivity : ε = 1 ε < 1 ε < 1 Ground condensation ε = 1 Figure 1. Schematic representation of the CO snow fall scenario suggested to explain the low brightness temperatures Tb observed by Viking in the polar night [see Forget et al., 1995]. (a) Outside of the low emission zones, the cap emissivity A is close to 1. (b) During a snow fall, the airborne CO ice particles scatter the thermal radiation back to the ground and, once on the ground, decrease the surface emissivity A. (c) After the end of the snow fall, the emissivity A remains below 1 until metamorphism and ground condensation increase the CO ice grain size, and thus A. (d) Ultimately, the ice layer becomes nonporous and transparent, and the infrared radiation emitted by the ground can radiate through with A BC2. one observation consists of one brightness temperature at 11 < m (= 8>8 ), one at 2 < m (= D? ), and in one case out of seven, one at 15 < m (= 8D6 ). Using data from the first northern fall and winter observed by Viking (January 1977 to November 1977, 2FE;4;@G (!H GJI;K#4;@ ) and a combination of the first southern winter (June 1976 to January 1977 for E%,#@LG ( H GM25E#4;@ ) and of the following year southern fall (November 1977 to May 1978 for 4 @ G ( H GNE;, @ ) provides a sufficient coverage to derive a climatology of both polar caps during the polar night seasons. The field of view of each channel was defined by focal plane stops 5.2 mrad in diameter. The orbital geometry of the Viking missions was elliptical. The corresponding footprint diameter generally ranged from 7 km to 35 km in the north polar region, from 1 km to 2 km in the south polar region during the fall and from 4 to 17 km during the winter. The asymmetry of the orbital geometry could have artificially biased our results. In order to evaluate this effect, we compared the observations of the same area obtained with very different spatial resolutions. As we did not find any significant differences, we believe that our results are not significantly biased by the orbital geometry. 2.2. Interpretation of the IRTM Data The difference = 8>8 =O >? (also called spectral contrast ) is a useful quantity for characterizing the radiative properties of the atmosphere and the surface (the spectral contrast of a blackbody would be zero). In paper 1, we analyzed the different processes which can affect the brightness temperatures measured above the polar caps. At the polar cap boundaries, high values of = 8P8 = >? were usually observed in both hemispheres (e.g., Figure 3 in paper 1). They must be due to a combination of water ice clouds (known to be present especially near the polar cap edge) and nonuniform temperatures (viewing a scene whose area includes both relatively warm bare ground and cold patches of ice produces substantial spectral differences) [Christensen and Zurek, 1984]. Since it would be difficult to simulate these two combined effects, we restricted our study to the inner polar caps, where the ground is more uniformly covered by CO frost and the incident solar flux is too weak to significantly warm the possibly unfrosted areas. Inside the cap and outside the dust-storm periods, the brightness temperatures were affected in the low emission zones, characterized by a strong correlation between =9 >? and = 8P8 =9 >?, with = 8>8 =9 D? increasing linearly while =9 >? decreases. As explained in the introduction, this systematic behavior was shown to be consistent with the spectral signature of scattering precipitating CO clouds with particle radius larger than 1 < m, or that of CO snow deposits with millimeter-sized grains (paper 1). Unlike the IRIS spectra which, can only be reproduced with CO ice particles mixed with small amounts of water or dust (as expected for the northern winter cap observed by the Mariner 9 mission), the correlation between = >? and = 8P8 = D? can be simulated with both pure or contaminated CO ice. As discussed in paper 1, dust is not expected to have a strong impact on = and 8P8 = D? under regular conditions.

4 FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS However, two global dust storms occurred during the northern autumn and winter of the first Viking year. The presence of dust is detected during the peak of the second dust storm over the cap south of 8@ by a rise in =9 D? and especially = 8P8 due to sensing the surface through a partly opaque dusty atmosphere warmed by heat advection from lower latitudes [Martin et al., 1979; Jakosky and Martin, 1987]. This second storm and, to a lesser extent, the first one were accompanied by an intense and more durable warming of the upper atmosphere in high northern latitudes, often called polar warming [Martin and Kieffer, 1979; Jakosky and Martin, 1987]. = 8D6 typically increased from 13 K to 18 K at the pole and to well above 22 K near the cap edge. This phenomenon is thought to be caused by complex atmospheric dynamical processes linked with forced planetary waves [Barnes and Hollingsworth, 1987] or gravity waves [Barnes, 199]. 3. Polar Cap Surface Temperatures To date, the puzzling presence of the low emission zones and their high variability prevented a direct remote estimation of the thermodynamic temperatures of the polar caps. In most previous studies and models these temperatures were often assumed to be about 148 K or 15 K because CO frost was expected to be in solid gas equilibrium at the mean atmospheric pressure. Nevertheless, based on IRTM measurements obtained in the vicinity of the residual caps in early fall and late winter, Paige and Ingersoll [1985] suggested that the actual surface temperatures of these regions were somewhat lower, near 146 K at the north pole and near 142 K at the south pole. 3.1. Method To determine the actual polar cap surface temperatures in a given region of the polar caps, it is necessary to process the IRTM observations in order to remove the radiative signature of the low emission zones. Two independent techniques can be used. First, one can select the maximum temperature observed during the polar night outside the dust storm periods. (Temperatures obtained later during the spring are difficult to process because they are affected by the airborne dust warmed by the sun). Second, one can make use of the fact that the low emission zones can be identified by their spectral signature = 8P8 = D?RQ 4. When = 8P8 = D?, the polar cap emissivity should be close to unity, making the measured brightness temperature close to the actual surface frost temperature. More generally, the systematic linear correlation between = and 8>8 = >? observed in the low emission zones and predicted by models in paper 1 can be used to extrapolate the polar cap surface temperature from all the observed brightness temperatures. We have chosen to use this second approach to calculate the actual surface frost temperature, and then to compare the results with the maximum brightness temperatures observed locally. In practice, we sorted the IRTM data temporally into 1@ solar longitude and spatially into 5@ latitude bins and 15@ (below 7@ lat), 3@ (between 7@ and 85@ lat), and 9@ (above 85@ lat) longitude bins. In each bin, we performed a linear regression in the [= >? ; = 8P8 = >? ] space and assumed that the surface temperature was the value where = >? 4. That way, we eliminated a possible bias due = 8P8 to the high noise level on =. In each spatial bin the surface 8>8 temperatures computed with this method at various seasonal dates are generally very close to each other, with a small expected seasonal modulation due to the strong seasonal pressure variations. These temperatures are also always close to the maximum temperature observed locally. For the regions near the poles, this is illustrated by Fig. 2, which shows the evolution of the difference between the observed brightness temperatures and the retrieved surface temperatures. Fig. 3 and 4 show the polar cap ground ice temperatures at winter solstice. All the measurements are included and averaged: The surface temperatures were temporally extrapolated to each winter solstice by using the CO vapor pressure relationship and by employing the surface pressure variations simulated by a climate model fitting closely the observed Viking lander seasonal pressure changes [Hourdin et al., 1995]. On the basis of the convergence between the surface temperature estimations made at different seasonal dates, and from the correlation between the retrieved temperatures with the maximum brightness temperatures observed in a given area, we estimate that the uncertainties in surface temperature are lower than S K. 3.2. Interpretation The most striking characteristic of the actual polar cap temperatures retrieved from the IRTM data is their systematic decrease toward the pole in both hemispheres. The southern polar cap is almost 5 K colder than the northern cap. Temperatures at the pole are about 145.5 K in the north and 141.1 K in the south, corresponding to CO partial pressures of 4.2 mbar and 2.2 mbar, respectively. This is consistent with Paige and Ingersoll s [1985] estimations of the surface temperatures for the regions near the pole in the vicinity of the residual caps (their estimation was based on only two sequences of observations obtained at (!H IUT;E - K%@ in the north and (!H 25E.- @ in the south, chosen because the IRTM brightness temperatures were then nearly coincident). Assuming a scale height of 9 km, Paige and Ingersoll estimated that the pole altitudes corresponding to the surface temperatures should be +2.5S 2 km in the north and +8S 2 km in the south. In fact, interpreting the cap temperatures in term of topography is not straightforward, since the surface pressure is also controlled by the atmospheric circulation: In the winter hemisphere, the latitudinal temperature gradient and the condensation of the atmosphere in the polar caps induce a low atmospheric pressure zone at high latitude [Pollack et al., 199]. To take this effect into account, we have used

\ V FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS 5 5 ] a) North T2 - Tsurf -5-1 -15 18 V 21 W 24 X 27 Y 3 Z 33 [ 36 Ls (deg) Amount of data : >1 5 5 5 ` b) South T2 - Tsurf -5-1 -15 Z 3 Z 6 Z 9 Z 12 ^ 15 _ 18 Ls (deg) Figure 2. Difference between =9 >? and the polar cap surface temperatures computed with the method described in this paper. Included are data obtained poleward of 85@ latitude, shown as a function of season (solar longitude ( H ). For readability, data are sorted in 1 K and 5? ( H bins.

6 FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS 27 7N 8N 18 Figure 3. Surface temperatures of the northern polar cap at winter solstice, computed from IRTM data. 9 7S 8S Figure 4. Surface temperatures of the southern polar cap at winter solstice, computed from IRTM data. 18 9 27 the Martian atmospheric general circulation model (GCM) of the Laboratoire de Météorologie Dynamique described by Hourdin et al. [1993, 1995]. Fig. 5 shows the zonal averaged frost point temperatures computed by the GCM (resolution K#TabT;Ea T ) compared to the surface temperature derived from the IRTM observations. In all of the simulations, the atmospheric mass was adjusted to match the surface pressure observed at the Viking Lander 1 site at winter solstice. The most variable (and uncertain) parameter being the dust optical thickness, experiments were conducted with various mean atmospheric opacity. The polar surface temperatures were found to be not significantly sensitive to this parameter, set to.2 for the results shown in Fig. 5. Because the large-scale Martian topography is still poorly known, two rather different data sets have been used by modelers: The Mars consortium and the digital terrain map (DTM, officially called digital elevation model, or DEM). Recently, Smith and Zuber [1996] have reanalyzed occultation data from the Mariner 9 and Viking spacecraft and used the Viking Lander data to compute a low-resolution, eighthdegree and order spherical harmonic model for the topography of Mars. Results obtained with these three data sets are shown in Fig. 5. The surface temperatures retrieved from the IRTM data appear to be colder than the modeled frost point of CO. This is especially true with the consortium data set in the southern hemisphere. However, this data set probably suffers from the lack of data south of 6@ S at the time of his completion [Wu, 1978]. Similarly, the Smith and Zuber data set may suffer from poor resolution and the lack of radio occultations at very high latitude. The other three modeled curves (DTM in the south, DTM and consortium in the north) are less than 2 K warmer than the retrieved temperatures, except poleward of 8@ latitude, where the differences reach 3 K. These discrepancies may be explained by uncertainties in the topography data sets, in the retrieved surface temperatures, and/or by other physical mechanisms decreasing the apparent surface temperatures (e.g. the depletion of CO gas in the polar region or the low temperature of the emitting level inside the CO ice layer; see details below). As explained above, the relative topography of the polar regions is poorly known, especially because of the lack of radio occultation measurements at high latitude [Kliore et al., 1973; Lindal et al., 1979]. With an atmospheric scale height of 7.1 km (typical of an atmosphere at about 14 K), a 1 K difference in the frost point temperatures corresponds to altitude difference of roughly 1 km. Locally, errors larger than 3 km, corresponding to the 3 K difference, may be possible. However, explaining the surface temperature discrepancies by topography errors only would require a systematic error, increasing toward the poles in both hemispheres. One geological origin of such a shape could be the long-term winter deposition of sediment. These polar deposits have been interpreted to control the topography much more than any underlying regional features. They could be several kilometers thick [see Thomas et al., 1992] and raise progressively the global topography toward the pole. Accurate topographic

g FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS 7 South po h lar cap North po h lar cap 152 152 mperature (K) Cap surface te 15 148 146 144 142 15 148 146 144 142 DTM topography Consortium topography Smith&Zuber(1996) topography Tcap retrieved from IRTM 14 14-9c -8d -7e -6f 6 7 8 9 Latitude (deg) Latitude (deg) Figure 5. Comparison between the zonally averaged polar cap surface temperatures retrieved from the IRTM data with surface temperatures calculated using the DTM, consortium, and Smith and Zuber [1996] topography data sets. These temperatures are the results of atmospheric general circulation simulations performed at winter solstices with the Laboratoire de Meteorologie Dynamique s Martian model. The atmospheric mass is adjusted to match the surface pressure observed at the Viking Lander 1 site. measurements from the laser altimeter aboard the Mars Surveyor spacecraft should be available in 1997 and should help us interpret the cap temperatures observations.. A systematic error in our retrieved polar cap temperatures cannot be ruled out. The purpose of our retrieval technique was to eliminate the low emission zones by selecting temperatures with = 8P8 = >? (found to be close to the maximum temperature observed locally). By doing so, we did not eliminate the possibility that the surface could have a maximum emissivity slightly lower than that for both = 8>8 and =9 D? and still yield equal values. Because Planck s function is not linear, that would require a lower emissivity at 11 < m than at 2 < m. This is not expected for CO ice particles (see paper 1). However, that could be the case for a rocky surface lying below nonporous solid CO ice transparent in the infrared. For example, emissivity values of about.94 and.97 at 11 < m and 2 < m would give = 8P8 = >? =OikjFlnm 2 K. Other physical mechanisms may affect the apparent surface temperature. First, if the circulation processes do not bring in or mix fresh atmosphere quickly enough, the condensation of CO could lead to a dilution of CO gas in the atmosphere normally composed of 95.3% CO and 4.7% noncondensible gas [Kieffer et al., 1976, 1977; Hess, 1979]. The corresponding decrease of the CO partial pressure will lower the frost point temperature. Following the observations of the anomalously low brightness temperatures of the polar caps by Viking, Hess [1979] explored the meteorological consequences of this idea. He concluded that unrealistic circumstances would be required to explain the observed temperatures. In particular, he showed that the atmosphere would be statically unstable unless the very strong anomaly in the CO mixing ratio required to explain the anomalous temperatures extended to heights of tens of kilometers (in any case, this explanation would not be consistent with the spectral properties of the low emission zones). However, the possibility of a small depletion cannot be ruled out. A decrease of the CO abundance down to 8% would lower the surface temperature by 1 K. According to Hess [1979], such a dilution would be stable, assuming that the anomaly in the CO mixing ratio extended vertically up to a few hundred meters, which is realistic. The fact that the CO ice layer may be quite transparent in the infrared may also decrease the apparent surface temperature of the polar caps. The thermal infrared radiation will originate from inside the CO frost or from the ground under the ice. Since this emitting level will not be in contact with the atmosphere, and yet, radiatively cooled, it may be colder than the frost point temperature (D. Paige, private communication, 1995). Note that such a process should exhibit a spectral signature, since the emitting level depth will depend on the absorption coefficient of the ice.

8 FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS To conclude, we can say that, on the basis of the available infrared observations and topography data, it is difficult to interpret the surface temperatures retrieved from the IRTM data. Nevertheless, these temperatures are useful for modeling and data analysis because they represent the potential, maximum emitting temperature of the surface of the polar caps. Within this context, they bring to the fore a strong hemispherical asymmetry, the south polar cap being almost 5 K colder than the north polar cap. According to our analysis, this difference is primarily due to higher elevations in the south compare to the north. 4. Distribution of the Low Emission Zones During the Viking Years Fig. 6 illustrates the distributions of the anomalously low brightness temperature events during the Viking years as a function of latitude and seasonal date. The low emission zones are identified by the difference between = >? and the cap surface temperatures retrieved as described above. The low emission events seemed to occur much more frequently in the northern hemisphere compared to the southern hemisphere. The difference is especially striking between 7@ and 85@ latitude, where very few low emission zones were observed in the southern hemisphere. In the north, brightness temperatures as low as 125 K occurred around ( H =295@ and ( H =32@. In the south, =9 D? reached a low of 128 K several times during the polar night seasons. Lower brightness temperatures may have been measured by the IRTM instruments at emission angles higher than 6@, but we did not analyze these data. The low emission zones apparently almost disappeared, but not totally, in the northern hemisphere during the second global dust storm between ( H = 27@ and ( H = 29@ (note, however, that we did not account for the increased dust opacity). The dust storm was then followed by a period of minimum brightness temperatures. A detailed study of the temporal evolution of the data shows that an exceptionally high frequency of occurrence of extremely low brightness temperatures was reached around (!H = 3@, just after the peak of dust optical depth observed by the Viking Landers. This contrasts with the high temperatures observed at the same time in the upper atmosphere near the.6-mbar pressure level. In fact, this level was probably clear of dust and warmed by dynamical mechanisms, unlike the lower atmosphere, which was then dusty and radiatively cooled. As mentioned in paper 1, the apparent correlation between airborne dust and low emission zones can indeed be explained by the infrared emissivity of the dust particles which strongly increase the cooling rate of the polar night atmosphere and thus favor the condensation of the CO ice in the atmosphere rather than on the ground [Pollack et al., 199]. It also seems likely that the dust particles can serve as condensation nuclei. By their radiative properties, the resulting snow clouds and snow falls should then decrease the polar cap brightness temperatures. This link between airborne dust and CO atmospheric condensation may also explain the strong hemispheric asymmetry with regard to the low emission zones. Indeed, during the years observed by Viking, airborne dust is thought to have been more abundant in the northern hemisphere during the fall and winter seasons than in the southern hemisphere during the same seasons [see Martin and Richardson, 1993]. In both hemispheres, we have identified a few unambiguous low emission zones at latitudes as low as 55@, around the winter solstices. In the southern hemisphere, the last low emission events occurred around ( H = 15@ (but there are no data between ( H = 15@ and ( H = 165@ ). In the northern hemisphere, several cases were observed after the vernal equinox, poleward of 8@ N until (!H = 1@. A last cold brightness temperature feature occurred at ( H = 37.5@, at a relatively low latitude of about 77@ N. Almost all of these spring low emission zones were positioned very close to surface features such as the permanent polar cap edge, Chasma Boreale, or an isolated crater. Similarly, Paige [1985] analyzed a set of observations exhibiting a low emission zone near the edge of the polar night at only 66@ N ((!H = 262@ ), and also found that the position of the low brightness temperature feature corresponded exactly with that of an isolated crater. These observations suggest that some low emission zones are created by the orography. Within the context of our snow-fall theory, several physical processes could explain such a behavior. For instance, adiabatic cooling of the atmosphere in ascending motion created by orography may be able to condense carbon dioxide locally and create the observed feature. Despite the fact that the south polar region is more heavily cratered than the north polar region, no low emission zones were observed in the southern hemisphere after the end of the winter, probably because the atmosphere was then very dust laden and warmed by the sun. 5. Polar Cap Radiative Budget During the Polar Night 5.1. Technique The radiative budget of the polar caps during the polar night can be determined from the IRTM observations. In a previous study, Paige and Ingersoll [1985] used this data set to estimate the annual polar cap infrared fluxes for the core regions of the north and south permanent polar caps. However, since the purpose of their work was to establish the global heat balance of these regions, they did not analyze in detail the spatial and temporal variations of these fluxes. Because the low emission zones do not behave like blackbody emitters, computing the thermal integrated fluxes toward space of the condensing polar caps from the IRTM data is not straightforward. Outside the 15- < m CO band, the IRTM instrument only measured brightness temperatures around 11 and 2 < m. Paige and Ingersoll [1985] assumed that the brightness temperatures at other wavelengths were bracketed by = >? Spok= 8>8 = D? ). In fact, the available IRIS spectra suggest that the brightness temperature of the low

FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS 9 North 1% 5% % 19. 21. 23. 25. 27. 29. 31. 33. 35. 1. 1% 5% % 19. 21. 23. 25. 27. 29. 31. 33. 35. 1. 1% 5% % 19. 21. 23. 25. 27. 29. 31. 33. 35. 1. 1% 5% % 19. 21. 23. 25. 27. 29. 31. 33. 35. 1. 1% 5% % 19. 21. 23. 25. 27. 29. 31. 33. 35. 1. Ls (deg.) 65 < lat < 7 7 < lat < 75 75 < lat < 8 8 < lat < 85 85 < lat < 9 South 1% 5% % 1. 3. 5. 7. 9. 11. 13. 15. 17. 19. 1% 5% % 1. 3. 5. 7. 9. 11. 13. 15. 17. 19. 1% 5% % 1. 3. 5. 7. 9. 11. 13. 15. 17. 19. 1% 5% % 1. 3. 5. 7. 9. 11. 13. 15. 17. 19. 1% 5% % 1. 3. 5. 7. 9. 11. 13. 15. 17. 19. Ls (deg.) -2 K -4 K -6 K -8 K -1 K -12 K T2 - Tcap Figure 6. Distributions of the anomalously low brightness temperature measurements as a function of latitude and seasonal date. In each time and latitude bin, the percentage indicates the fractions of IRTM observations measured in a given = >? =OqsrPt bracket, with =OqsrPt the CO ice cap surface temperature computed with the method describe in this paper, taking into account the seasonal variations of the global atmospheric pressure.

w 1 FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS emission zones between 2 < m (5 cm798 ) and 4 < m (25 cm78 ) are lower than =O >? (e.g. Figure 1 in paper 1). In paper 1, we showed that the polar IRIS spectra can be simulated using radiative transfer models of precipitating CO cloud with particle radius larger than 1 < m or CO snow deposits with millimeter-sized grains, with in both cases, small amounts of water or dust mixed with the CO ice particles. As explained in section, such models should thus be able to simulate the spectra corresponding to the IRTM data obtained inside the polar caps outside the duststorm periods (or more generally, outside the period when =9 >? and = 8>8 were warmed by the related heat and dust advection). Therefore we have used these models to compute the polar cap integrated fluxes toward space from the IRTM data outside of the 15- < m CO band region [13.5-16.5 < m]. Rather than processing the IRTM observations one by one, we have binned these data in several dimensions for the following reasons: (1) by averaging = 8P8 in each bin, we tend to strongly reduce the noise level of this data; (2) the processing of millions of individual data would have been very costly in terms of computational time; (3) choosing appropriate bins allowed us to keep track of all necessary information. To reach this last goal, each bin was defined in five dimensions: seasonal date (!H (bin size 5@ ), latitude (5@ ), longitude (15@ ), = >? (2 K), and emission angle cosine (.2). By binning on = D?, we kept track of different low-emission structures within individual temporal/spatial bins. In each bin, the standard deviation of = 8>8 was typically 1 K and did not exceed 3 K. Binning on the emission angle was necessary because the retrieval of the surface snow and cloud model parameters depends on the emission angle. As mentioned above, we also discarded observations that were obtained at emission angles greater than 6@. The next step was the determination of the models parameters from = 8>8 and = D?. The cloud and surface snow models were used separately. With only two data points, we had to keep only two varying parameters for each model ; for the cloud model, these were the surface temperature and the cloud optical depth; for the surface snow model, the surface temperature and the CO snow particle size (the surface CO snow emissivity is mainly controlled by the frost grain size). With all other parameters set constant except those two, all we had to do was to run the models iteratively to find the unique pair of varying parameters providing the unique spectrum corresponding to = 8>8 and = >?. The other parameters were set to the values used to match the IRIS spectra in paper 1. The modeled cloud was thus composed of 5- < m CO ice particles mixed with 1 precipitable microns (pr- < m) of 1- < m water ice particles. The modeled surface snow was composed of CO grains mixed with.5% of dust (u -, < m) and.5 % of water ice (u 2F4 < m). With both models, we found that the computed spectra were not strongly dependent upon these parameters, except when the amount of contaminants was small. Such a situation could have occurred during the year observed by Viking, especially in the southern hemisphere. Thus we also used the tt m -2 sr -1 m) Radiance (Wa 2 1-4 1.5 1-4 1-4 5 1-5 Figure 7. pure CO 2 snow contaminated CO 2 snow contaminated CO 2 cloud pure CO 2 cloud IRIS spectrum 2µm 11µm v 2v 4v 6v 8v 1 12 Wavenumber (cm -1 ) Typical IRIS radiance spectrum (obtained at 75@ N,15@ W, (!H =331@, and emission angle 59.5@ ) compared to simulated spectra. These are not best fit spectra but were synthethized in order to match the IRIS spectrum over the 11- and 2- < m IRTM channels. models for the extreme case of pure CO surface snow and cloud. Fig. 7 shows simulated spectra compared with a typical IRIS spectrum of a low emission zone (also shown with brightness temperatures in paper 1, Figure 1). The simulated spectra were all obtained with = >? =141.9 K and = 8P8 =145.9 K. These values were deduced from the IRIS spectrum, although = 8P8 was not directly computed because the IRIS spectrum is too noisy in the 11- < m region. We = D? T K because it is the average value in chose = 8>8 the northern hemisphere for = D? =141.9 K (see Figure 6, paper 1). The contaminated snow and cloud spectra computed from these two values are close to the IRIS spectrum, giving us confidence in the accuracy of our extrapolation method, at least for conditions similar to the Mariner 9 observations. The pure CO ice models give rather different results. For given values of = D? and =, the pure 8P8 CO surface snow model tends to strongly overestimate the integrated radiance because the computed pure CO snow spectra reach a minimum in the IRTM 2- < m channel. On the other hand, the pure CO cloud model underestimates the integrated radiance. Since the actual spectral behavior of the low emission zones has not been observed when the CO ice is less contaminated than during the northern winter observed by Mariner 9, we used these pure CO ice models to estimate an upper and a lower limit on the radiative fluxes. We used a six-point Gaussian quadrature to integrate intensities at the top of the atmosphere over the upward hemisphere to obtain the net outgoing thermal flux to space (assumed to have zero downward IR flux) outside the 15- < m region. Between 13.5 and 16.5 < m, we had to estimate the thermal fluxes from the only available information: =. Based 8x6

FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS 11 on polar IRIS data analysis and results from a narrowband model of the CO 15- < m band developed by Hourdin [1992] for Mars, we have found that /#4-4 Uy;z 7O{?P?} P~U' x W m7 sr798 was a fairly good approximation of the intensity of the 13.5-16.5 < m channel at the top of the atmosphere, from which its contribution to the corresponding net infrared outgoing thermal flux is readily found. 5.2. Polar Cap Infrared Fluxes Toward Space Fig. 8 shows a comparison of longitudinally averaged thermal infrared fluxes emitted by the polar caps of each hemisphere toward space during the Viking years. The abscissa scale is (!H from @ to 18@ in the south and 18@ to 36@ in the north. For both hemispheres, the left side is the autumnal equinox, the right side the spring equinox, and the winter solstice is at the middle. The thick curves (solid or dashed) correspond to the two contaminated models (which give almost similar results). The two thin lines illustrate upper and lower limits on the fluxes computed with the pure CO ice model. At all latitudes, the emission of the southern cap was weaker than that in the north, mainly because of its lower surface temperature (Figures 3 and 4). As explained in section, this difference is due to a higher topography in the south. The southern emission was quite constant during the studied period, except at high latitudes, where it was strongly affected by the low emission events in the first part of the winter. In the north, dust storm (( H = 21@ -23@ ) and, especially, ƒ (( H = 27@ -3@ ) and the associated polar warming events corresponded to strong emission periods contrasting with colder periods (dust-storm periods were not studied equatorward of 8@ N, where an increase of =9 >? and = 8>8 was then detected). 5.3. Atmospheric and Snow Fall Radiative Forcing The method used to compute the polar cap thermal fluxes toward space allows us to analyze its components. On the one hand, the ground emission is reduced by the low emission zones thought to be created by the nonblackbody properties of the CO fresh snow and precipitating clouds. On the other hand, the CO gas 15- < m band absorption and emission reduces or increases the upward thermal fluxes, depending on the temperature profile of the atmosphere relative to the ground. We call radiative forcing the reduction or increase (in W m7 ) of the radiative fluxes toward space related to these two effects, as it is often used in the terrestrial literature [e.g. Ramanathan et al., 1989]. For instance, the total radiative forcing due to both effects is the difference between the computed thermal flux toward space and the potential cap surface emission: ˆ = qsrpt (2) with =OqsrPt the surface temperature, as shown in Figures 3 and 4, taking into account its seasonal evolution due to the seasonal pressure variations. By restricting the definition of to the CO gas 15- < m band, we * get >Š the radiative forcing due to the CO gas absorption Š rpi Œ 8DŽ 6 798D 6> ˆ š rpi : 8D 6> 8DŽ 6P œ o'= qsrpt.ž}ÿ*d %Ÿ (3) with Plank s function, Ÿ the wavelength in micrometers and the radiative flux at the top of the atmosphere8xž in6 78x the 13.5-16.5 6P < m band. As explained above, we assume that 8xŽ 6 798D 6P Uy [/#4-4 z 7 {?P?} P~ k ]. Similarly, the radiative forcing ik x due to the CO snow and clouds creating the low emission zones is the difference between the flux emitted toward space outside the 13.5 to 16.5- < m band and the integration of the Planck function at =OqsrPt over the same spectral intervals. By definition, we thus simply have: ik x PŠ r>i (4) Š Averaged * values of the radiative forcing, r>i and ik s in six latitude belts are shown in Fig. 9. They correspond to six of the longitudinally averaged curves shown in Fig. 8. In the southern hemisphere, the atmospheric CO did not affect the infrared emission. The total forcing was dominated by the effect of the low emission zones, limited to the regions near the pole. At lower latitudes this effect was much less important and the emission flux was close to = qsrpt. The situation in the northern hemisphere was much more complex because of the occurrence of two dust storms/polar warming events during this particular year. South of 7@ N, the upper atmosphere was always warmer than the ground, increasing the radiative fluxes by a constant forcing of at least several W m7. Poleward of 7@ N, between the first and the second dust storms ((!H = 23@ - 273@ ) the thermal flux was mostly affected by the low emission of the CO fresh snow and clouds. During the onset of the second dust storm starting at ( H = 273@, this effect disappeared, while the upper atmosphere was strongly warmed at all latitudes by the polar warming. The warming corresponds to a strong CO gas forcing, which reached a maximum of 5 W m7 poleward of 8@ N. At lower latitudes the forcing reached 8-1 W m7 (not shown), and was still of the order of 3-5 W m7 when a large number of very low emission zones were observed after the first phase of the storm above 7@ N. Finally, after ( H = 31@, things got back to the prestorm situation with a thermal emission reduced by the radiative properties of the CO snow and clouds above 7@ N, and increased by the emission of the upper atmosphere inversion below 7@ N. 5.4. Impact on the CO Condensation Rate Impact of the low emission zones. The impact of the nonblackbody radiative properties of the CO fresh snow and clouds on the condensation rate of CO is illustrated in Fig. 1, which shows the seasonal evolution of the radiative forcing ratio * i' s & o ik x. This

ª ª ª 12 FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS Ls (Southern Hemisphere) «3 6 9 12 15 18 Ls (Southern Hemisphere) «3 6 9 12 15 18 North North es (W/m 2 ) Energy flux 3 25 2 South 3 25 2 South 85 < latitude < 9 8 < latitude < 85 15 15 es (W/m 2 ) Energy flux 3 25 2 3 25 2 75 < latitude < 8 7 < latitude < 75 15 15 es (W/m 2 ) Energy flux 3 25 2 3 25 2 65 < latitude < 7 6 < latitude < 65 15 ± 18 21² ³ µ 24 27 3 33 36 Ls (Northern Hemisphere) 15 ± 18 21² ³ µ 24 27 3 33 36 Ls (Northern Hemisphere) Figure 8. Wavelength integrated thermal infrared fluxes of the Martian polar regions toward space, as a function of time during the fall and winter seasons averaged on different latitude belts. The two thin lines illustrate the upper and lower limits on the radiative fluxes computed with the pure CO ice model (see text).

FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS 13 xes (W m -2 ) Energy flu 5-5 North 85 < lat < 9 5-5 CO 2 snow radiative forcing CO 2 gas forcing Total forcing South 85 < lat < 9 21¹ º» ¼ ½ ¾ 24 27 3 33 36 ¼ ¼ ¼ À Á  3 6 9 12 15 18 5 5 xes (W m -2 ) Energy flu -5 North 75 < lat < 8-5 South 75 < lat < 8 21¹ º» ¼ ½ ¾ 24 27 3 33 36 ¼ ¼ ¼ À 3 6 9 12 15 Á  18 5 5 xes (W m -2 ) Energy flu -5 North 65 < lat < 7-5 South 65 < lat < 7 21¹ º» ¼ ½ ¾ 24 27 3 33 36 ¼ ¼ ¼ À Á  3 6 9 12 15 18 L s (degree) L s (degree) Figure 9. Radiative forcing of the CO fresh snow and gaseous atmosphere on the polar cap thermal emission. This quantity shows how much the potential cap surface emission was affected by the CO atmosphere in the 15- < m band and by the low emission zones thought to be created by the radiative properties of the CO fresh snow and precipitating clouds (see text).

Ê Ê Ï Ï 14 FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS %) Φ ( Ls (Southern Hemisphere) Ë 3 6 9 12Ì 15Í 18Î 3 2 1 85 < lat < 9 North South Ls (Southern Hemisphere) Ë 3 6 9 12Ì 15Í 18Î 3 8 < lat < 85 2 1 18Ã 21Ä 24Å 27Æ 3Ç 33È 36É Ls (Northern Hemisphere) Ls (Southern Hemisphere) Ë 3 6 9 12Ì 15Í 18Î 3 75 < lat < 8 18Ð 21Ñ 24Ò 27Ó 3Ô 33Õ 36Ö Ls (Northern Hemisphere) Ls (Southern Hemisphere) Ë 3 6 9 12Ì 15Í 18Î 3 7 < lat < 75 %) Φ ( 2 1 2 1 18Ð 21Ñ 24Ò 27Ó 3Ô 33Õ 36Ö Ls (Northern Hemisphere) 18Ð 21Ñ 24Ò 27Ó 3Ô 33Õ 36Ö Ls (Northern Hemisphere) Figure 1. Radiative forcing ratio Ø as a function of time, averaged on different latitude belts, with Ø Ù (- COÚ snow forcing)/(total flux - COÚ snow forcing). Ø shows how much less COÚ condensed in the polar region due to the radiative properties of the COÚ fresh snow and clouds.

FORGET AND POLLACK: CONDENSATION OF THE MARTIAN POLAR CAPS 15 ratio shows how much less energy was radiated to space by the polar caps because of the low emission zones. Thus it also shows how much less COÚ condensed in the polar region, as explained in the introduction with equation (1). At the poles, the impact of the low emission zones on the radiative budget was comparable in the north and in the south. However, between 7Û and 85Û lat, the impact was much stronger in the northern hemisphere. The area affected in the north was thus much larger than that in the south, since area varies proportionally to the cosine of latitude. Impact of the upper atmosphere warming. The increase of infrared emission toward space due to the emission of the northern upper atmosphere in the 15-Ü m band did not increase the amount of condensed COÚ, as could be deduced from equation (1). On the contrary, it implies that some energy was advected by the upper atmosphere from the nonpolar regions (in equation (1), the increase of ÝÞ ßáà is correlated with a larger increase of âãä ). Since the warming was well above the polar caps, it affected the condensation of COÚ by its radiative heating. Within the latitudes and periods studied in this paper (i.e., when no increase of åoú>æ and å+çpç was detected), it is reasonable to think that the corresponding downward radiative fluxes were thus of the order of the COÚ gas forcing shown in Fig. 9. That means a maximum decrease of the COÚ condensation rate up to 2% near the pole and certainly much larger nearer the cap edge. Direct impact of the dust storms. Last but not least, the condensation of COÚ during the northern winter observed by Viking was directly affected by the second dust storm during its peak, a period that we have not studied in the present paper. The rise in å Ú>æ and å+çpç south of 8Û lat and particularly near the cap edge between è!é = 273Û and è é = 3Û (see Figure 5 in paper 1 and Jakosky and Martin [1987, Figure 6]) and the very large increase of å+çxê during the same period suggest that a large amount of heat was then advected to the polar regions by the atmosphere. This amount of heat is not easy to estimate. Assuming that the dust mixing ratio was vertically uniform and that the temperature linearly increased with altitude from surface temperature to å çdê at 25 km, Martin and Kieffer [1979] computed the downward radiance over the north polar cap. They estimated that net condensation was essentially halted north of 55Û during the peak of the dust storm. Theoretical studies conducted with atmospheric general circulation models (GCMs) [Pollack et al., 199 ; Hourdin et al., 1995] confirmed that enhanced dust loading can increase the advection of heat by 3 W më Ú near the cap edge, but suggest that the presence of dust does not increase the atmospheric heating north of 7Û N. However, these GCMs were not able to simulate the polar warming events. 6. Comparison With Results of Models Simulating the Viking Lander Pressure Curves 6.1. First Viking Year As explained in the introduction, several models have been used to simulate the formation and the recession of the Martian polar caps through energy balance considerations. These models showed that the seasonal evolution of the atmospheric mass is sensitive to the modeled ice emissivity ìpí îsï and the modeled visible albedo ðñí îsï. In order to constrain these parameters, most recent studies have tried to reproduce the large seasonal fluctuations of the surface pressure measured by the Viking Landers. Indeed, to first order, these fluctuations are caused by the condensation and sublimation of the atmospheric COÚ into the polar caps and thus provide reliable quantitative information on the global energy balance of the polar regions. Most studies concluded that it was necessary to use low values for the modeled cap emissivities. For instance, with an emissivity ì í îsï set constant in time and space, best results were obtained with ì í îsï Ùóò ôöõ# by Pollack et al. [1993] and ì í îsï Ùøò.ô ù; by Hourdin et al. [1995]. Wood and Paige [1992] showed that ìpí îsï was quite sensitive to the poorly known seasonal thermal inertia of the polar regions, but also found better results with emissivities below 1. Moreover, these three teams showed that closer fits were obtained with a lower emissivity in the north than in the south (e.g. ì>í îsï (north)=.65 and ìpí îsï (south)=.75 for Pollack et al. [1993]; ìpí îsï (north)=.5 and ìpí îsï (south)=.7 for Hourdin et al. [1995]). Such values of the modeled cap emissivity should not be interpreted in terms of real COÚ ice properties. These values only mean that the amount of COÚ actually condensing in the polar caps is about 3% lower than what would be predicted by these models with ìpí îsï =1, and not that the actual COÚ ice emissivity is low. The present study suggests that the COÚ condensation rate is in fact decreased by other processes not parameterized or simulated by the models. These processes are the upper north polar atmosphere warming, the COÚ snow falls creating the low emission zones, and the low surface temperatures of the actual caps compared to those in the models. These last two processes have effects somewhat similar to the low constant emissivities in the models (both tend to decrease the upward infrared flux), although the low emission zones are highly variable in space and time. In addition to these processes, it is also likely that the direct advection of heat during the dust storms is underestimated in the models (see previous section). The impact of the overestimation of the polar cap temperatures by the models can be estimated from Fig. 5. Pollack et al. [1993] and Hourdin et al. [1995] both used the Mars consortium topography data set. The corresponding overestimation of the surface irradiance úoåñû îsü>ý, is less than +1 W më Ú over the northern polar cap. In the south, however, it is about +2 W më Ú equatorward of 6Û S, increasing regularly poleward to reach about +5 W më Ú in the south pole area.