During the period 11–15 July 1981, heavy rainfall occurred over the Sichuan Basin in China, resulting in severe floods that took a large toll in human life and property damage. Mesoscale analyses by Kuo, Cheng and Anthes have shown that the flood was directly related to the development of a long-lived mesoscale southwest (SW) vortex over the basin. In this paper we present the results of numerical experiments aimed at 1) testing the capability of a limited-area mesoscale model to predict the evolution of the SW vortex and the accompanying heavy precipitation, 2) examining the structure of the simulated vortex using the model data, and 3) elucidating the role of various physical processes in the evolution of the SW vortex. Principal findings are: 1) The control experiment, which utilized an 80-km grid spacing and simple physical parameterizations, was able to simulate the evolution of the mesoscale SW vortex and the accompanying heavy precipitation. The simulation captured many observed features as analyzed by Kuo, Cheng and Anthes. The SW vortex formed completely within the southwesterly monsoon current, remote from the baroclinic frontal system to the north. At its mature stage, the SW vortex possessed a column of cyclonic vorticity, extending from the surface to 250 mb with very little vertical tilt. The core of the vortex was characterized by strong vertical motion, a considerably higher θe compared with its environment. The horizontal momentum was uniformly distributed within the vortex layer. The vorticity field and the vertical motion field were in phase, which was favorable for the development and the persistence of the SW vortex. The model predicted a maximum 48-h precipitation (ending at 0000 UTC 14 July) of 213 mm over the Sichuan basin, which compared favorably with the observed precipitation. 2) Latent heat release was essential for the development of the SW vortex and the resulting precipitation. A simulation without latent heating produced a much weaker SW vortex and little vertical motion. The total 48-h precipitation was only 35 mm, an order of magnitude less than that of the control simulation. The results suggest a strong interaction between cumulus convection and the SW vortex. 3) Surface sensible and latent heat fluxes were important to the precipitation forecast. An experiment with surface energy fluxes removed predicted a maximum 48-h rainfall of 70 mm, one-third of the value of the control simulation. The Intensity of the SW vortex was weakened as a result of decreased latent heat release. 4) Further sensitivity experiments showed that the SW vortex observed in this case was a terrain-induced standing eddy. Its formation was not influenced by diabatic processes (e.g., latent heating and surface energy fluxes), although these processes were important for its development. 5) The differential frictional effect, hypothesized over the past decade to be a plausible mechanism for the formation of the SW vortex, was shown to be unimportant in this case. The surface friction acted as both a vorticity sink and an energy sink. When it was removed, the SW vortex evolved sooner with considerably stronger intensity. 6) A trajectory diagnosis and a model experiment with modified topography showed that as the southwesterly monsoon current impinged upon the mesoscale Yun-Gui Plateau, which extends from the southeastern corner of the main Tibetan Plateau, the low-level flow was blocked. The flow aloft then descended into the Sichuan Basin on the lee side of the mesoscale plateau, creating cyclonic relative vorticity over the basin by stretching of earth's background vorticity. This explains why the Sichuan Basin is a climatically favorable location for the origin of the SW vortex.