Recent Research and Development of Ensemble-Variational (EnVar) Hybrid Data Assimilation for Global, Hurricane and Convective-Scale Severe Weather Prediction

Xuguang Wang
University of Oklahoma
  2 Oct, Noon, in 2155


The 3D and 4D ensemble-variational (EnVar) hybrid data assimilation (DA) system was operationally implemented at NCEP in 2012 and 2016 respectively to improve US NWS global numerical weather prediction (NWP). This seminar will discuss our most recent research and development of the EnVar hybrid DA system targeted on improving operational numerical prediction of a wide range of scales, including global NWP, convection-allowing hurricane prediction and CONUS convective-scale severe weather prediction. Such efforts are in close collaboration with NOAA centers.

In the first part, we discuss two new developments of the global 4DEnVar hybrid DA system including the valid time shifting method to increase ensemble size and the multi-resolution ensemble 4DEnVar approach. Experiments over an extended period suggest both methods can provide a cost-effective way to further improve the global 4DEnVar analysis and the subsequent global forecasts. Detailed diagnostics will be shared. Research and development have also been made to further develop the hybrid EnVar DA system for convection-allowing regional modeling systems. In the second part of the seminar, the hybrid DA system is extended with the operational convection allowing Hurricane Weather Research and Forecast (HWRF) modeling system to improve high-resolution tropical cyclone prediction. Experiments and diagnostics have demonstrated the self-consistent HWRF hybrid DA system can improve the hurricane intensity forecasts. In addition, recent research using the HWRF hybrid DA system to identify and diagnose model errors, and to best assimilate inner core data will be discussed.

In the third part of the seminar, the hybrid DA system is also extended for convective scale severe weather (e.g. tornadic supercell, MCS, etc.) prediction over the CONUS. In particular, issues associated with the direct assimilation of radar reflectivity observations in EnVar are revealed. A method that allows direct assimilation of radar reflectivity in EnVar is proposed and implemented. Experiments for a variety of convective scale weather phenomena including tornadic supercell and Mesoscale Convective Systems (MCS) in HRRR, NAMRR, WoF (Warn on Forecast) like contexts are conducted. Experiments revealed that the new direct radar DA method for EnVar improved the tornadic supercell prediction, and also improved precipitation forecasts compared to the current operational method of assimilating the reflectivity, “the cloud analysis”. If time permits, recent efforts to assimilate GOES-R radiances for convective scale weather prediction, to optimally design CAM (convection allowing model) ensemble, and to design new object based verification for CAM will also be briefly discussed.