Development of a tropical cyclone tracker and applications to tropical cyclones occurred in 2008 in North Western Pacific

To investigate whether tropical cyclones (TCs) are simulated correctly and verify forecast skill for TCs in numerical weather prediction models, it is necessary to derive trajectories of TCs from model output data and compare them with TC best track information. Because TC tracks are not model output variables, they can be diagnosed from basic variables as pressure, wind, and temperature of a model. A method to decide the TC center is to find out the center manually by looking at the weather charts. However, this manual method is not efficient and sometimes impossible to pinpoint the TC center. Furthermore, determining TC center manually from ensemble prediction results becomes increasingly inefficient. Thus an objective analysis of TC center and trajectory becomes an indispensable work. In this study, a TC tracker that has been developed (hereafter YSU TC tracker) for various model output format including netCDF used in WRF, binary data used in MM5, and GRIB used in GFS and ECMWF YOTC, with various projection including Lambert conformal and cylindrical equidistant (lat/lon) grid is applied for the model output to investigate the performance of the tracker. In detecting step, YSU TC tracker used mean sea level pressure, vorticity and geopotential height at 700 and 850hPa. In tracking step, a search area of next forecast time is determined with specified radius around TC center position of the next forecast time, guessed by using thresholds of cyclone movement such as direction, range of angle, and search radius based on TC speed between two consecutive analyses. The YSU TC tracker developed above was applied on the TCs occurred in 2008 in North Western Pacific. As a forecast data, 0.5° by 0.5° resolution NCEP GFS and ECMWF YOTC data on cylindrical equidistant grid were used. To investigate the performance of the YSU TC tracker, the YSU TC tracker was applied to the GFS data and the results are compared with those from the NCEP tracker. Average track positional errors from the best track data were similar for both trackers, implying that the YSU TC tracker performs as well as the NCEP tracker. To investigate the average forecast skill of the NCEP GFS and ECMWF YOTC data, the YSU TC tracker was applied on both data. The average track positional errors for both data were similar at the initial time, but the average track positional error of the NCEP GFS becomes larger than that of ECMWF YOTC data as the forecast time increases. More comprehensive results will be presented in the meeting.