666 Implementation of Radar Data Assimilation Capabilities within Ensemble–Variational Hybrid GSI for the Stand-Alone Regional FV3-Based Convection-Allowing Forecasting System

Tuesday, 14 January 2020
Hall B (Boston Convention and Exhibition Center)
Chong-Chi Tong, CAPS, Norman, OK; and Y. Jung, C. Liu, and M. Xue

In 2018, the stand-alone regional (SAR) version of the National Weather Service’s new unified modeling system, the Finite-Volume Cubed-Sphere Dynamical Core (FV3), became available. The SAR-FV3 will eventually replace a suite of current operational regional modeling systems (e.g., NAM, SREF, RAP, and HRRR) and an experimental Warn-on-Forecast system (WoFS) that continuously assimilate radar data. Therefore, efforts are needed to develop advanced data assimilation (DA) capabilities including radar DA for the SAR-FV3 to accelerate this transition to the next-generation FV3-based rapidly updating convection-allowing prediction system.

In this study, radar DA capabilities developed at CAPS and implemented within the Grid-point Statistical Interpolation (GSI) analysis system for EnKF, EnVar, and hybrid EnVar are interfaced with the SAR-FV3. This system has been applied to a 30 April 2019 tornado outbreak in the Central Plains. A 40-member ensemble is initialized at 1800 UTC and radar observations including the MRMS reflectivity (Z) and WSR-88D radial wind (Vr) are assimilated for an hour from 1900 UTC to 2000 UTC at 15-minute intervals using 3DVar, EnKF, and hybrid EnVar (refer to hybEnVar hereafter), followed by 3 hour deterministic and/or ensemble forecasts. The model configurations are set as close as to those of the 3-km HRRR forecast system to ensure a smooth transition to the operational system.

Results show that the 3DVar and hybEnVar exhibit much better fit to the observed Z but the EnKF presents much smaller error growth rates during the forecasts, suggesting the analysis is more balanced. With respect to Vr, the performances are rather similar among experiments. Subjective evaluation of reflectivity forecasts of different DA experiments suggest that EnKF captures the storm evolution better than 3DVar or hybEnVar during the 3 hour forecast. The neighborhood probability (NP) of midlevel updraft helicity (UH) overlaid with the SPC tornado reports also exhibits better correspondence of the EnKF forecast. More evaluation results will be presented at the conference.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner