While considerable studies have proved that the track and intensity forecasts of tropical cyclone (TC) relied heavily on output from numerical weather prediction (NWP) model, few researches investigated how well NWP models forecast TC genesis in the western North Pacific (WNP) basin. In order to understand the characteristics of TC genesis forecast in WNP basin by NWP models, this study derives a set of criteria to identify the formation of TC using historical data and verifies it based on ECMWF model data between 2013 and 2015. The results show that the percentile values adopted as the criteria thresholds have a significant impact on the performance of algorithm based on the criteria. A reasonable adjustment of threshold in a specific interval can effectively improve the TC genesis prediction. For example, in the WNP basin the forecast results are most sensitive to small changes in the relative vorticity on the 850 hPa level. The results of forecast test of the optimal threshold combination scheme indicate that the turning point of performance lies between 24 and 48 hours with regard to the hit rate in the 12-72 hours prior to the formation of TC. For lead time less than 24 hours, the hit rate was basically maintained at a high level above 0.7 with a small decrease. After that, the performance drops sharply before stabilizing beyond 48 hours. In addition, the performance of the TC genesis prediction in ECMWF model varies significantly from year to year and also in different WNP regions. It performs better to the east of the Philippines than over the South China Sea (SCS). On the other hand, high false alarm (FA) rates are found in the central parts of the SCS up to the waters around the Philippines and the central part of the WNP. The significant discrepancy in ECMWF’s performance can also be observed between different basins. Within the 24 hours before the genesis of a TC, the forecasts for the WNP basin verify better than those for the Atlantic basin.
Read full abstract