A Study On The Forecasting Performance And Ensemble Perturbation Of Tropical Cyclone Track Using GRAPES_TCM

In this study,the verification of the tropical cyclone track forecasting performance of GRAPES_TCM is first shown . Thereafter, some primary research on ensemble forecasting on tropical cyclone track are performed , including investigating the influence of the initial position uncertainty of tropical cyclone (TC) on TC track prediction in the broclinic model ,as well as design and evaluation of a vortex perturbation method .GT is used to post-forecast tracks of TCs occurred over the Northwestern Pacific and the South China Sea during 2005. A series of detailed performance verifications for the numerical tropical cyclone tracks prediction is presented. The results show that GT has a good performance on TC track prediction. The total mean forecast errors (MFE) at 24h, 48h and 72h are 135.8km, 230.7km and 336.0km respectively. The result also shows that the MFE at 12h is somewhat large (about 100km). The Significant improvement of GT's track prediction for recurved cases can be obtained through the modification of the forecast tracks with systematic bias. The model is found to have very good forecast skill for the cases that occurred in the west of 130E, especially for those close to the southeast of china. In total average, the forecast track of GT is faster (slower) before (after) 48 hours than the best track, and the speed bias at 12 hour is larger during the forecast period (approximately 1 m/s).To investigate the influence of the initial position uncertainty of tropical cyclone (TC) on TC track prediction, Two series of ensemble experiments (IPUE1 and IPUE2) are performed with 15 cases, in which the initial positions of TCs are perturbed with the perturbation magnitude being 1 (upper bound) and 0.5 (average value) respectively. The mean spreads in both experiments maintain their magnitudes in the first half of the 72 hours prediction and increase slowly in the second. They show no significant changes in the whole prediction, which is consistent with the results in barotropic model. In both experiments ,the proportion of the spread to the ensemble mean error is higher than 80% at 12 hour, and thereafter decreases dramatically. As far as IPUE2 is concerned, its smallest spread in 72 hour is 27% of the ensemble mean error. Therefore, it is necessary to perturb the initial positions of TCs when performing the ensemble prediction of tropical cyclone track in the baroclinic model. Further analysis shows that , displacement of the initial vortex will regullarly move the track in a certain distance before 36 hour or so, thereafter it cause a irregular modification , although the regular one is largely retained. In addition, herein the maintenance of the position bias by the forecast track, the bias (relative to best track) in early stage(say,6h) can be used to revise the forecast positions in later stage. Some results show that the revise is useful.The vortex perturbation method proposed here is replacing the initial vortex with the ones picked from historical analysis data and numerical simulation results to form an ensemble of initial fields . In comparison with the methods often used before (such as BGM, LAF, perturbing the synthetic vortex) ,this one can almost totally get rid of the destructed initial vortex, so that the availability of the ensemble prediction is ensured. This method can eliminate the certain error of the initial intensity, and add proper perturbations to the intensity. And it perturbs both the dynamic and thermodynamic variables with possibly proper vortex structure. Some results show that this method can obtain a good correlation between spread and forecast skill.