Improved Prediction Of Landfalling Typhoon In China Based On Assimilation Of Radar Radial Winds

With the continuous development of numerical models and data assimilation algorism,the tropical cyclone(TC,also called typhoon in the Northwest Pacific Ocean)track forecast error decrease apparently over the last decades,but its intensity,especially the fine wind and rainfall prediction still needs progress.Since there is no operational aircraft mission in the Northwest Pacific Ocean currently,ground-based radar is the only way that performs high-temporal-spatial-resolution observations of the fine structure of the landfalling TC.Effective use of radar observations is of great importance to improve the numerical prediction of TCs.Existing radar assimilation algorism have shown excellent improving capabilities in certain TC case study,but the improving sensitive observations are unclear.Besides,there is weakness in the thinning method of radar observation that used commonly in previous researches.This work mainly focuses on improving landfalling TC's intensity,wind,and rainfall forecasts with radar radial wind data assimilation.The sensitivity of assimilating inner-core and outer-core radar observation to the improvement of the TC analysis and forecasts are analyzed first.For the weakness of the current thinning method,a new evenly-spaced thinning method is developed in this work.Finally,the above technique is applied in a real-time TC nowcasting system,which is of great significance for improving the ability of landfalling TC.The specific content and conclusions are as follows:With the radius of the 34 kt wind radium,the TC Moranti(2010)is divided into two regions: the inner core area and the outer core area.The radar radial wind observations located in the inner and outer areas are assimilated individually,and an experiment assimilating all observations is also conducted for comparision.It is found that both inner and outer observation have a positive impact on TC analysis and prediction.However,the inner core observation has a more obvious improvement on the initial position and intensity of the TC.Inner-core observation can significantly strengthen the TC vortex and correct the TC position by fewer assimilation cycles.While outer core observations are more effective for outer rainband correction.Furthermore,although the inner-core observation accounts for about 20% of the total,its improvement on the position and intensity analysis of the TC is larger than that of assimilating the entire data,so does TC forecasting improvement.For the forecasts of precipitation in TC inner-core,the TS of assimilating inner core observations is almost the same as assimilating all data.For the forecasts of precipitation near TC outer rainband,assimilating inner-core or outer observations have similar TS,lower than experiment that assimilating all data.The above experiments show that the inner core observation is the key area for TC analysis and forecasting in the improved model.Towards radar observations,a new data thinning method called evenly-spaced thinning method(ESTM)is developed,in which the radar observations are projected into a horizontal grid that is equivalent to the model resolution.Compared to traditional Radial Spatial Thinning Method(RSTM)that commonly used in many papers,ESTM enhance the data usage rate of the TC inner core and avoid the additional error created.The SOs created by SETM are almost evenly distributed in horizontal grids of model background,resulting in more observations located in TC inner-core region are involved in SO.Compare to RSTM,more cyclonic wind innovation,larger analysis increment of height and horizontal wind in lower level can be seen in ESTM assimilation experiment of TC Mujigae(2015).The analysis and forecasts of Mujigae's location and intensity are improved by ESTM.The rainfall correlation coefficient of no DA,RSTM-En KF,and SETM-En KF is 0.59,0.71,and0.77,respectively,showing proper SO processing method can improve rainfall forecast.In order to further validate the performance of radial wind assimilation with SETM,all landfalling TCs that made landfall in the Chinese mainland in 2017 are examined.It is shown that TC landing position error and intensity error is reduced by33% and 25% compared with the experiments without radar data assimilation.The skill score analysis of rainfall,which is verified by space correlation coefficients,ETS,shows that the scores of extreme rainfalls(larger than 80mm/3h)could be twice compared to no data assimilated.Based on the radar data assimilation sensitive analysis and the radar data thinning method research,this work further developed a Typhoon rapid Refresh Analysis and Nowcasting System(TRANSv1.0).The WRF numerical model and the ensemble Kalman filter assimilation scheme are integrated in TRANSv1.0.The TRANSv1.0assimilates Chinese ground-based Doppler radar radial velocity observations.The system hourly update when the TC enters the radar observation network,and hourly perform 12 h deterministic forecast.By examining the error of the real-time forecast of the 6 landing TCs in 2020,it is found that the mean track error is 42.8km,and the average intensity error is 4.4m/s(4.5h Pa).The 12-hour precipitation forecast scores for light rain,moderate rain,heavy rain,and extremely heavy rain are 0.66,0.50,0.23,and 0.17,respectively.Compared with the current official model,the TRANSv1.0model has advantages in extremely heavy rainstorm prediction.It fills up the official model's weakness on forecast heavy rains and extreme rains.It has a good capacity for predicting TC's rapid intensification and the winds before and after the TC makes landfall in the north.After the evaluation by the CMA,TRANSv1.0 is now operational running in the Chinese Academy of Metrological Sciences.The products are providing to CMA real-timely for guidance in official TC forecasts.