A mesoscale model intercomparison: A case of explosive development of a tropical cyclone (COMPARE III)

The performance of current mesoscale numerical models is evaluated in a cas e of model intercomparison project (COMPARE III). Explosive development of Typhoon Flo (9019) occurred in the case in September 1990 during the cooper ative three field experiments, ESCAP/WMO-led SPECTRUM, US-led TCM-90, and f ormer USSR-led TYPHOON-90 in the western North Pacific. Sensitivity to init ial fields as well as impact of enhanced horizontal resolution are examined in the model intercomparison. Both track and intensity predictions are very sensitive to the choice of in itial fields prepared with different data assimilation systems and the use of a particular synthetic tropical cyclone vortex. Horizontal resolution en hanced from 50 km through 20 km down to a 10 km grid has a large impact on intensity prediction. This is presumably due to a better presentation of in ner structure with higher resolution. There is little impact on track predi ction in this target period when the typhoon was in its before-recurvature stage. While most models show large biases in underestimating central press ure deepening, some of the participating models with a particular initial f ield succeed in reproducing qualitatively the time evolution of central pre ssure, including slow deepening in the first half and rapid deepening in th e second half of the simulation period of 72 hours. However, differences le ading to different intensity predictions among models have yet to be identi fied. Intercomparison of the simulation results shows that wind field has a close relationship with precipitation distribution. This suggests that bet ter prediction of precipitation distribution is crucial for better predicti on of wind field, and vice versa. Through the COMPARE III experiments, it has become clear that precise simul ation of tropical cyclone structure, especially in the inner-core region, i s very important for accurate intensity prediction. Consideration, therefor e, should be given to this point, when improvements in resolution, initiali zation, and physics of numerical models for tropical cyclone intensity pred iction are reviewed.