UNIFIED FORECAST SYSTEM : Convective-Allowing Forecast and Data Assimilation System

UNIFIED FORECAST SYSTEM : Convective-Allowing Forecast and Data Assimilation System


1. Preamble

The UFS Convective-Allowing Model (CAM) Working Group’s charge is to create, using the FV3 dynamical core, a convective-allowing ensemble-based data assimilation and forecast system for the 0-3 day range to replace This development is a collaborative effort between NWS (NCEP/EMC), OAR (NSSL, GSD, GFDL), NCAR, the DTC, and academia. Experience gained from the development of earlier operational and experimental CAM systems such as the GSD’s HRRR/HRRRE, the NAM nests from EMC, the NSSL Experimental Warn-on-Forecast System for ensembles (NEWS-e), the NCAR experimental CAM ensemble, and GFDL’s FV3-based CAM efforts, will guide this project as it evolves. The anticipated operational system, the Rapid Refresh Forecast System (RRFS), will be a single-core (FV3) CAM ensemble-based data assimilation and forecast system. The RRFS is planned for operational implementation in FY2023.


2. Background and Development Timeline

As part of its commitment to the implementation of a unified forecast system in NCEP’s Production Suite, NOAA is planning to replace NCEP’s myriad mesoscale and convective-allowing systems with new guidance systems based on the FV3 dynamic core. As of July 2018, the NCEP meso/CAM scale modeling suite currently consists of the following components:

  1. North American Mesoscale (NAM) system: Runs the Non-hydrostatic Multiscale Model on B-Grid (NMMB). The NAM consists of a North American 12-km parent domain run to 3.5 days and 4 non-moving nests run to 2.5 days at 3-km resolution over the CONUS, Alaska, Puerto Rico, and Hawaii. An additional 1.5 km nest is run over a pre-selected part of the CONUS or Alaska to 1.5 days for fire weather support operations. The NAM features an 6-h data assimilation cycle with hourly analysis updates for the 12 km parent domain and the 3 km CONUS/AK nests.
  2. High-Resolution Window (HiResW):  Consists of ~ 3-km runs of the NMMB model and two configurations of the ARW model over the CONUS, Alaska, Hawaii, Puerto Rico, and Guam (single ARW run only).
  3. High-Resolution Ensemble Forecast (HREF) system: Current/time-lagged HiResW and NAM CONUS nests are used to generate ensemble products. Current version is an 8-member ensemble for CONUS with HiresW and NAM CONUS nest members; HREF for Alaska, Hawaii, and Puerto Rico are a 6-member ensemble with HiResW members only.
  4. Short-range Ensemble Forecast (SREF) system: Runs at 16 km over North America and currently consists of 26 members (13 NMMB, 13 ARW) with physics/initial condition diversity. The replacement of the SREF and deterministic NAM systems will be based on whether they can be replaced by improved forecast guidance from the FV3-GFS and FV3-GEFS.
  5. Rapid Refresh (RAP) and High-Resolution Rapid Refresh (HRRR): The RAP and HRRR are run hourly out to 21 hr and 18 hr, respectively. RAP is run at 13-km resolution over North America (identical to the NAM parent domain), while HRRR is run at 3 km over CONUS. As of July 2018, the 00/06/12/18z HRRR cycles were extended to 36-hr, and the 03/09/15/21z RAP cycles were extended to 39-hr. A HRRR-Alaska system was also added, running every 3 h.

NAM development was frozen after the March 2017 upgrade, and RAP/HRRR development with the WRF-ARW model will cease after the RAPv5/HRRRv4 upgrade in 2020. However, operational execution of these modeling and associated DA systems will continue until comparable FV3-based systems are able to give similar performance. The transition of these deterministic modeling systems to FV3-based configurations will be prioritized as follows; these milestones are also summarized in the chart below:


FY2019-FV2020 : HRRR v4 implementation and SAR-FV3 development

  • Development of the FV3-based stand-alone regional model (SAR-FV3) to bring its capabilities and performance up toward the current CAM systems
  • Finalize and implement RAPv5/HRRRv4
  • Add the extended HRRR forecasts to HREF, possibly replace NMMB members in HREF with SAR-FV3 members
  • Preliminary ensemble DA testing with SAR-FV3
  • Begin evaluation in NOAA testbeds


FY2020-FY2021 :SAR-FV3 development/testing for Meso/CAM

  • Finalize porting of HRRR physics into CCPP for use in SAR-FV3
  • Finalize FV3-based RAP
  • Continue evaluation of FV3 SAR against existing systems
  • Continue development of DA capabilities with SAR-FV3
  • Continue evaluation in NOAA testbeds


FY2021-FY2023 : Evolution to a FV3-based single core Rapidly Refreshed Forecast System (RRFS) : FV3 CAM ensemble with DA

  • Ensemble analysis and forecast system development with SAR-FV3 and JEDI
  • Development of stochastic physics for single core
  • Continue evaluation in NOAA testbeds
  • Pursuant to favorable evaluation, implement v1.0 RRFS [2023]



3. Ongoing Efforts


EMC : Developed a limited area regional version of the FV3 and adapted the NCEP Unified Post-Processor (UPP) and Grid-scale Interpolation (GSI) analysis for the FV3SAR. Currently running two real-time forecast experiments:


  •        A control run of a 60-h forecast over the CONUS of a 3 km limited area standalone FV3 with initial and boundary conditions from the operational FV3GFS
  •       A parallel run of the 3 km FV3 for dynamics/physics tuning tests


EMC is also developing a hourly ensemble data assimilation capability with the limited area FV3.


GSD : Developing RAP-sized FV3 domain and modifying pre-processing to use RAP instead of GFS / Fv3GFS input for lateral boundary conditions. Working with EMC on developing better grid-spacing uniformity for the continental FV3-RAP domain


EMC/GSD/NSSL/Developmental Test Bed Center (DTC) : Develop commonly-shared FV3SAR workflow


4. Links to Relevant Documents/Web Sites