Tropical cyclone genesis in a global numerical weather prediction model

The physical processes responsible for tropical cyclone genesis over the western North Pacific are investigated using the operational analyses of the U.K. Meteorological Office global model for the years 1992-93. The analyses are divided into two groups depending on whether a particular analysis led to the prediction of a real genesis in the atmosphere. Composites of the winds at 850 and 200 hPa from the analyses that fall into each of the two groups (the successful and failed predictions) are then made and compared. It is found that about 3 days prior to genesis, the low-level trades north of the pregenesis vortex begin to strengthen. One day later, an upper-level westerly trough and/or the Tropical Upper-Tropospheric Trough (TUTT) starts to encroach toward the pregenesis cluster. The low-level trades near the cluster continue to intensify and a surge of southwesterly winds occurs to the south of the cluster. On the day before genesis, the southwesterlies become the dominant low-level feature. At the upper levels, the TUTT has moved to the northeast of the pregenesis vortex at this time. Most of these features are found to occur in a high percentage of the individual cases. Comparisons with the analyses that led to failed predictions also reveal these features to be statistically significant. Calculations of the angular momentum (AM) fluxes suggest that at the low levels, the enhancement of the trades and the surge in the southwesterlies contribute to the spinup of the vortex primarily through two processes: the symmetric import of planetary AM and the asymmetric import of relative AM, with the former being more dominant. As the genesis time approaches, the contribution to the latter process by the trades becomes smaller relative to that by the southwesterlies. At the upper levels, the role of the TUTT or westerly trough is to reduce the export of planetary AM by the symmetric outflow. However, the contribution by the upper-level flow toward the genesis process appears to be minimal until the day before genesis. Comparisons with the analyses in the failed prediction category using individual cases again suggest that differences between the results in these two categories are statistically significant.