Wednesday, 10 January 2018: 2:15 PM
615 AB (Hilton) (Austin, Texas)
US NWS operational convection allowing forecast systems NAM CONUS nest and HRRR both use the GSI-based data assimilate (DA) system in a hybrid mode through the passive use of the readily available GFS ensemble members to provide a single analysis to initialize the single deterministic forecasts. Critical convective scale observations such as radar reflectivity are assimilated indirectly using a separate approach called cloud analysis /diabatic digital filter. The underlying assumptions adopted in these approaches were not able to produce appropriate storm scale details in the analysis and subsequent forecast. The University of Oklahoma (OU) team in collaboration with NCEP/EMC and ESRL/GSD further developed the GSI based EnKF /hybrid DA system to assimilate multi-scale observations including direct assimilation of radar reflectivity for both models and use of the models’ own EnKF to provide background error covariances. Extensive retrospective experiments for the full CONUS domain have found that the newly extended GSI hybrid DA system with the advanced radar DA method outperforms the system using the cloud analysis/DFI.
With the development of the GSI-based EnKF/hybrid DA system, both HRRR and NAM CONUS nest systems now are able to automatically initialize convection allowing ensemble forecasts. The newly developed convective scale GSI based hybrid DA system and its associated ensemble were run for the first time in real time during the 2017 HWT Spring Forecast experiment for the NMMB model. It was found that the convection allowing ensemble forecasts initialized from the newly developed system are skillful measured by traditional and object based metrics. It is also found that this single model, single physics ensemble is under-dispersive. To facilitate the design of the NWS operational convection allowing ensemble, the impact of an ensemble of dual dynamic cores relative to an ensemble using a single core and multi-physics configuration is evaluated with retrospective cases. These results, our efforts of transitioning the newly extended hybrid DA system to operations (R2O) and plans for FV3 model will be discussed in the conference.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner