Short-term Forecasting Through Cycling Assimilation Of China Coastal Radar Data For Typhoon Meranti At Landfall

Accurate prediction of the track, intensity, structure and precipitation of typhoon has remained a challenging problem in typhoon research area. Over the past decade, typhoon track forecasts have improved steadily, but typhoon intensity and structure forecasts have improved slowly. The lack of accurate initial conditions capturing the internal structure of typhoons has been attributed as one of the main factors. Coastal Doppler radar is the only platform that can observe the three-dimensional structure of typhoons near landfall with sufficiently high temporal and spatial resolutions This study examines the impact of cycling assimilation of radar data from eight coastal ground-based Doppler radars of China, for the analysis and prediction of Typhoon Meranti (2010) making landfall at a cloud-resolving resolution using the ARPS3DVAR and cloud analysis package.Firstly, the impact of radial velocity (Vr) and reflectivity (Z) data individually and in combination are examined. The radar data are assimilated over a6-h period before Meranti landfall through1-hour assimilation cycles, and then twelve-hour predictions are made. All experiments that assimilate radar data produce better structure, intensity, track and precipitation analysis and prediction than the one without radar data assimilation. Vr data lead to a larger improvement to the intensity and track forecast than Z data, while additional Z data further improve the precipitation forecast. Overall, assimilating both Vr and Z data from multiple radars gives the best forecasts. Three local rainfall maxima related to typhoon circulations and their interactions with the complex terrain in the southeast China coastal region are also captured.Secondly, the analysis increments produced by the3DVAR/cloud analysis system, and the dynamic and thermodynamic responses during the forecast step is studied. It is shown that the errors in Vr and Z decrease during the assimilation cycles. The most significant improvement to the model vortex occurs in the first and second cycles, when the background error is larger. In later cycles, the corrections contain mostly subvortex structures. Although the ARPS3DVAR system does not directly update pressure and temperature when analyzing radar data, the pressure and temperature field can responds to the analyzed winds through model adjustments during the assimilation cycles, which leads to the reduction in MSLP and increase in warm core. As a result, a vortex with the balance between the dynamic and thermodynamic fields is established in the final analysis, which is close to the observation.Finally, Sensitivity experiments suggest that the assimilation frequency also affects the analysis and forecast. The experiments with lower assimilation cycles predicted a somewhat weaker typhoon with larger track errors later on than the experiment with hourly cycles, but they still performed much better than the experiment without radar data. Assimilating MSLP from the best track in addition to radar resulted in a vortex whose MSLP is much closer to observed in the analysis, but the benefit was mostly lost within the first hour of free forecast, mainly because of the lack of balance between the pressure and temperature fields analyzed by the3DVAR. Additional assimilation of the conventional observation data helps improve the large-scale environmental conditions in the initial field. Assimilating data from a single Doppler radar with good coverage of the typhoon inner core region is also quite effective, except that it takes one more cycle to establish circulation analyses of a similar quality as the multiple radar case; although, the forecasts using multiple radars are still the best. Increasing frequency of moisture adjustment in the cloud analysis can improve the intensity forecasts, but degrade the track forecast and overpredict precipitation.