A Comparison of Four Operational Global Numerical Weather Prediction Models in Predicting the Genesis of Typhoon Megi (2010)

Operational forecasts of tropical cyclone (TC) genesis are currently based on global model forecasts. Although the track forecasts are the first priority, the genesis forecasts are also important because in some cases genesis occurs not too far away from the coast and the time between the genesis and the landfall is not too long. Regardless of the importance, large differences exist among forecasts of TC geneses by global models. This study attempts to understand why some global numerical weather prediction (NWP) models can predict TC genesis well while others cannot. The archived forecasts for the genesis of Typhoon Megi (2010) from four operational global models (ECMWF, GFS, JMA and UKMO) are compared to elucidate the causes of the success or failure of a forecast.

Overall, the ECMWF shows the best skill among the four global models and predicted the genesis with the longest leading time. The failed forecasts tend to show lower relative humidity in the mid-troposphere, weaker vorticity or stronger vertical wind shear than successful forecasts. CAPE in the failed forecasts tends to be higher than the successful forecasts, indicating that the convective scheme is not active enough to release CAPE in the failed forecasts. The area of overlapping region of the Okubo-Weiss-Zeta (OWZ) parameter at 850hPa and 500hPa exists before the genesis in the successful forecasts. In addition, the Genesis Potential Index (GPI) modified by using OWZ parameter instead of absolute vorticity is shown to be a better indicator of developing and non-developing systems than the original GPI. The sensitivity experiments using the modified GPI are conducted to evaluate the relative importance of each parameter to the successful/failed forecasts.