numerical models, starting with the thermal model designed by Leighton and Murray (1966) who originally pre-Although CO 2 snowfall has never been directly observed on Mars, it has been suggested that such precipitation may explain dicted the existence of the martian CO 2 cycle. Recent modthe puzzling infrared measurements obtained by Mariner 9 and eling efforts include more sophisticated energy balance Viking during the polar night in each hemisphere. The radiative models (e.g., James and North 1979, Wood and Paige 1992, effect of the snow would strongly alter the radiative balance Pollack et al. 1993a) and climate simulation with general of the condensing polar caps and thus the CO 2 cycle and the circulation models or GCMs (Pollack et al. 1981(Pollack et al. , 1990, global climate. We have simulated this phenomenon with a Hourdin et al. 1993Hourdin et al. , 1995)). Unlike the energy balance modgeneral circulation model (GCM). For that purpose, a new els which usually assume that the CO 2 directly condenses parameterization of CO 2 condensation in the atmosphere and on the ground, GCMs can estimate the amount of CO 2 on the ground has been developed, paying particular attention condensing in the atmosphere. For instance, Pollack et al.
to mass and energy conservation and allowing for the possible sublimation of sedimenting CO 2 ice particles. Atmospheric con -(1990) showed that a fraction of the total CO 2 condensation densation may result from radiative cooling on the one hand could take place in the atmosphere, especially when the (especially when the atmosphere is dust laden) and from adiaatmosphere is dust laden because of the increased atmobatic cooling in upward motions on the other hand. This latter spheric emissivity. However, in every GCM study so far, process can be very efficient locally. On this basis, we have the condensed CO 2 has been assumed to precipitate instanmodeled the effect of the CO 2 snowfall on the infrared emission taneously to the surface without changing the properties by decreasing the local emissivities when atmospheric condenof the atmosphere and the cap. sation was predicted by the model. This parameterization is The available observations of the condensing polar caps based on physical considerations (radiative transfer through suggest that reality may be more complex. The infrared the CO 2 ice particles, snow metamorphism on the ground). thermal mapper (IRTM) instrument aboard Viking mea-Without tuning the model parameters, we have been able to sured 20-Ȑm brightness temperatures showing consideraccurately reproduce the general behavior of the features observed by Viking in the thermal infrared. These modeling re-able structures with anomalously low values in the winter sults support the CO 2 snowfall scenario suggested from the polar regions (Kieffer et al. 1976), far below 148 K, the observations. Overall, this new parameterization, used in comtemperature appropriate for condensed carbon dioxide in bination with the digital terrain model topography and with vapor pressure equilibrium at the expected atmospheric allowance for a varying atmospheric dust content, allows the pressure. The location and brightness temperatures of GCM to simulate the CO 2 condensation-sublimation cycle realthese areas (hereafter also called ''low emission zones'') istically. In particular, the seasonal variations of the surface sometimes varied on time scales of days (Kieffer et al. pressure recorded by the Viking Landers can now be reproduced 1977). The low emission zones were also characterized by without artificially decreasing the condensation rate as was a complex spectral signature as observed by Mariner 9 done in previous studies.