We have used a detailed cirrus cloud model to evaluate the physical processes responsible for the formation and persistence of subvisible cirrus near the tropical tropopause and the apparent absence of these clouds at midlatitudes. We find that two distinct formation mechanisms are viable. Energetic tropical cumulonimbus clouds transport large amounts of ice water to the upper troposphere and generate extensive cirrus outflow anvils. Ice crystals with radii larger than 10--20 μm should precipitate out of these anvils within a few hours, leaving behind an optically thin layer of small ice crystals (&tgr;vis≂0.01--0.2, depending upon the initial ice crystal size distribution). Given the long lifetimes of the clouds, wind shear is probably responsible for the observed cloud thickness ≤1 km. Ice crystals can also be generated in situ by slow, synoptic scale uplift of a humid layer. Given the very low temperatures at the tropical tropopause (≂-85 ¿C), synoptic-scale uplift can generate the moderate ice supersaturations (less than 10%) required for homogeneous freezing of sulfuric acid aerosols. In addition, simulations suggest that relatively large ice crystal number densities should be generated (more than 0.5 cm-3). The numerous crystals cannot grow larger than about 10--20 μm given the available vapor, and their low fall velocities will allow them to remain in the narrow supersaturated region for at least a day. The absorption of infrared radiation in the thin cirrus results in heating rates on the order of a few K per day. If this energy drives local parcel temperature change, the cirrus will dissipate within several hours. However, if the absorbed radiative energy drives lifting of the cloud layer, the vertical wind speed will be about 0.2 cm-s-1, and the cloud may persist for days with very little change in optical or microphysical properties. The fact that these clouds form most frequently over the tropical western Pacific is probably related (through the nucleation physics) to the very low tropopause temperatures in this region. Simulations using midlatitude tropopause temperatures near -65 ¿C suggest that at the higher temperatures, fewer ice crystals nucleate, resulting in more rapid crystal growth and cloud dissipation by precipitation. Hence, the lifetime of thin cirrus formed near the midlatitude tropopause should be limited to a few hours after the synoptic-scale system that initiated cloud formation has passed. ¿ American Geophysical Union 1996