The Upgrade of GRAPE_TYM in 2016 and Its Impacts on Tropical Cyclone Prediction

The model reference profile and the vortex initialization scheme in GRAPES_TYM of China National Meteorological Center are modified in 2016 to improve the ability of tropical cyclone (TC) track and intensity prediction.The reference atmosphere profile is often applied in numerical weather prediction model to guarantee model integration stability and the accuracy. The reference atmosphere profile of isothermal temperature is replaced by a profile based on the horizontal mean of the model initial condition in GRAPES_TYM. It could decrease the amplitude of the perturbation of potential temperature and pressure and increase the accuracy and stability of model integration.The TC vortex initialization is a key factor to TC track and intensity numerical prediction. The TC vortex initialization scheme in GRAPES_TYM includes two parts:Vortex relocation (the vortex in the analyzed field is moved to the location analyzed by forecasters) and intensity correction. The modification of vortex initialization scheme includes two aspects:The relocation of vortex is removed, the radius of correction of the intensity is reduced to 4° from 12° in order to weaken the influence on the outer circulation of TC, which is assumed well analyzed by global model with higher resolution that provides the initial and boundary conditions for regional model.Experiments are carried out twice a day using 2014-2016 main TCs which last more than 72 h. Results show that upgrade of reference atmosphere profile could reduce the northward bias and the mean track errors especially for the mid-turning typhoon around 140°E. The modification of the radius of the intensity correction from 12° to 4° could improve TC track prediction especially for 0-72 h. GRAPES_TYM with the upgrade of reference profile and vortex initialization scheme could reduce the mean track errors by 10%(24 h), 12%(48 h), 16%(72 h), 14(96 h), and 15%(120 h) compared with the operational system.Results from 2014-2016 are also compared with results of NCEP global forecast system (NCEP-GFS). Mean track errors of GRAPES_TYM are larger than those. The track error differences are 9.2 km(24 h), 17.2 km(48 h), 18.4 km(72 h), 41.1 km(96 h) and 60.3 km(120 h), which are generally within 50 km except for 120 h prediction. GRAPES_TYM performs better than NCEP-GFS for TCs moving westward or northwestward and landing at the coast of China. The mean TC intensity errors of GRAPES_TYM within 72 h are smaller than those of NCEP-GFS.It can be found from the above results that the regional model improvements are more important for TC track prediction compared with improvements of model initial conditions which come from the global model, especially when the resolution of global model get higher and more data are assimilated. Therefore, more efforts should be put into the regional model optimization and improvement in the future.