Atmospheric General Circulation Models (AGCMs) are capable of creating tropical storms whose statistical distributions in space and time are similar to observed ones, although the intensity of the modeled storms is much weaker and the spatial scale much larger than in observations, due to low model resolution. In this study, Western North Pacific tropical storms are detected and tracked in a low resolution AGCM (ECHAM4.5). Numerical simulations with this model using observed sea surface temperature are analysed for 17 years (1979-1995) using 13 ensemble members.

Time-dependent composites of the tropical storms are created in the Western North Pacific during the typhoon season (June to October - JJASO) using all the storms that occur in the simulations during that period in all years and all ensemble members. The composites are created by averaging simulated fields from different storms in a storm-centered coordinate. In time, all the storms are aligned on the first day they pass the detection criteria (here called day 0). The composites extend backwards and forwards for 15 days relative to day 0, including information from all storms which the tracking algorithm can track that far.

There are clear signals of the storms in all the variables that were composited, and in most of them there is a sharp transition in the mean value of the variable near day 0. Vertical wind shear and atmospheric humidity seem to be the variables with greatest influence on the tropical cyclogenesis process. The wind shear varies significantly over the storm trajectory primarily because the mean storm trajectory moves from a region of relatively high climatological shear to one of relatively low climatological shear.