Ever since the observations of Percival Lowell, the annual cycle of Martian water has been a fascinating topic in planetary exploration. Observations by the Viking Orbiter, supplemented by Earth-based microwave and infrared observations, have given us a reasonable picture of this cycle. We are now also able to model the cycle using our Mars Climate Model, a simplified atmospheric general circulation model designed specifically for this purpose. We find that a thin adsorbing layer of the Martian regolith plays a fundamental role in the water cycle, limiting the lower atmospheric relative humidity and preventing the formation of widespread ice deposits at low latitudes. We are thus able to estimate a large-scale average value of the specific soil surface area of this regolith. Water which evaporates from the permanent north polar ice cap during summer is returned by a process of repeated evaporation and precipitation on the retreating seasonal cap the following spring, so that the global inventory of water outside the polar caps ranges within narrow limits. (There is a small net annual deposition of water ice at the south polar cap which is always at dry ice temperatures.) If ice on the residual south polar cap is exposed during the summer, it rapidly sublimes, generating vapor amounts similar to those observed in northern summer. Recovery to normal dry conditions in the southern atmosphere occurs very rapidly in the next year. Such an event could explain the otherwise anomalous Earth-based pre-Viking observations of a wet southern summer. If southern ice deposits at lower latitudes are exposed, the vapor can be transferred irreversibly through the strong Hadley cell to the north polar cap. We therefore speculate that the asymmetry of Mars' current orbit is responsible for the asymmetry of the present water distribution (with extensive permanent water ice deposits located only in the colder, aphelion summer, northern hemisphere).