Youngsun Jung;
Ming Xue;
Gang Zhao;
Fanyou Kong;
Kevin Thomas;
Timothy Supinie;
Keith Brewster;
Nathan Snook
Center for Analysis and Prediction of Storms (CAPS) and School of Meteorology, University of Oklahoma (OU)
The 8th EnKF Data Assimilation Workshop
CAPS Real-time Storm-Scale EnKF Data Assimilation and Forecasts for the NOAA Hazardous Weather Testbed Spring Forecasting Experiments: Towards the Goal of Operational Ensemble-Variational Data Assimilation for Convection-Permitting Models
Talk:
2018_CAPS_Realtime_Jung.pptx
As part of the NOAA-sponsored CSTAR (Collaborative Science and Technology Applied Research) and Testbed projects, CAPS has been generating realtime Contiguous U.S. (CONUS)-domain storm-scale ensemble forecasts (SSEFs) that include unique radar data assimilation (DA) at 4 km (3-km since 2015) grid spacing every spring since 2007 as part of the NOAA Hazardous Weather Testbed (HWT) Spring Forecast Experiment (SFE). In 2017, the CAPS SSEFs consist of two ensembles. The 23-member 3DVar-based ensemble assimilates radar along with surface and upper-air observations using a 3DVar/cloud-analysis system at 0000 UTC every day. A separate 10-member ensemble is initialized from cycled EnKF analyses that are run over a 6-hour period from 1800 to 0000 UTC using a combination of the NCEP Gridpoint Statistical Interpolation (GSI)-based EnKF system and an EnKF system developed at CAPS. The EnKF DA uses 40 members with multiple PBL schemes, and the Thompson microphysics scheme with perturbed graupel density. The CAPS EnKF system assimilates radar data between 2300 and 0000 UTC at 15 minute intervals while the GSI-EnKF assimilates all operational data used by the Rapid Refresh except for satellite and aircraft data at hourly intervals.
The forecasts from the EnKF analyses are evaluated in terms of the precipitation forecasting skills and other metrics, and loosely compared to forecasts initialized from 3DVAR/cloud analysis at 0000 UTC. Evaluation results for the 2017 season show that forecasts from the ensemble mean analyses outperform individual ensemble members in general although the precipitation forecasts exhibit large sensitivity to the choice of microphysics schemes. When radar data are not assimilated, the equitable threat scores are lower than the corresponding forecasts that assimilated radar data for the first 24 hours of forecast. The power spectra of composite reflectivity suggest that the EnKF analyses and forecasts are much better balanced while the 3DVar/cloud analysis-initialized members go through rapid adjustments in the first hour to re-balance among the model variables. We believe this effort represents the first time that full-volume radar data are assimilated in realtime over a CONUS-size domain using EnKF at native grid resolution for extended periods. More detailed evaluation results will be presented at the workshop.
As part of the NOAA-sponsored CSTAR (Collaborative Science and Technology Applied Research) and Testbed projects, CAPS has been generating realtime Contiguous U.S. (CONUS)-domain storm-scale ensemble forecasts (SSEFs) that include unique radar data assimilation (DA) at 4 km (3-km since 2015) grid spacing every spring since 2007 as part of the NOAA Hazardous Weather Testbed (HWT) Spring Forecast Experiment (SFE). In 2017, the CAPS SSEFs consist of two ensembles. The 23-member 3DVar-based ensemble assimilates radar along with surface and upper-air observations using a 3DVar/cloud-analysis system at 0000 UTC every day. A separate 10-member ensemble is initialized from cycled EnKF analyses that are run over a 6-hour period from 1800 to 0000 UTC using a combination of the NCEP Gridpoint Statistical Interpolation (GSI)-based EnKF system and an EnKF system developed at CAPS. The EnKF DA uses 40 members with multiple PBL schemes, and the Thompson microphysics scheme with perturbed graupel density. The CAPS EnKF system assimilates radar data between 2300 and 0000 UTC at 15 minute intervals while the GSI-EnKF assimilates all operational data used by the Rapid Refresh except for satellite and aircraft data at hourly intervals.
The forecasts from the EnKF analyses are evaluated in terms of the precipitation forecasting skills and other metrics, and loosely compared to forecasts initialized from 3DVAR/cloud analysis at 0000 UTC. Evaluation results for the 2017 season show that forecasts from the ensemble mean analyses outperform individual ensemble members in general although the precipitation forecasts exhibit large sensitivity to the choice of microphysics schemes. When radar data are not assimilated, the equitable threat scores are lower than the corresponding forecasts that assimilated radar data for the first 24 hours of forecast. The power spectra of composite reflectivity suggest that the EnKF analyses and forecasts are much better balanced while the 3DVar/cloud analysis-initialized members go through rapid adjustments in the first hour to re-balance among the model variables. We believe this effort represents the first time that full-volume radar data are assimilated in realtime over a CONUS-size domain using EnKF at native grid resolution for extended periods. More detailed evaluation results will be presented at the workshop.
Thanks to McGill's Atmospheric and Oceanic Sciences department for hosting this web-site.