Authors: Yan Xue (NWS OSTI Modeling Program), Steven Simon (NWS OSTI Modeling Program & IBSS)
SFS Application Team co-Leads: Neil Barton (NWS/NCEP/Environmental Modeling Center), Phil Pegion (OAR/Physical Sciences Laboratory), Xiaqiong (Kate) Zhou (NWS/NCEP/Environmental Modeling Center)
Acknowledgements: The development of SFS has been made possible by the dedicated contributions of many SFS developers within NOAA and the wider community, with essential funding support provided by the NWS Office of Science and Technology Integration Modeling Program and the OAR Weather Program Office’s Subseasonal-to-Seasonal (S2S) Program.
The Seasonal Forecast System (SFS) Beta, version 1.0 Prototype (SFSbeta) was officially launched in Near-Real-Time (NRT) in March 2026. Jointly supported by the National Weather Service (NWS) Modeling Program and the Office of Oceanic and Atmospheric Research (OAR) Weather Program Office (WPO), this milestone, under the SFS Project, is a critical step in advancing NOAA’s seasonal forecasting mission. It marks the transition from the 15-year-old Climate Forecast System, version 2.0 (CFSv2) to a modern, fully coupled system based on the Unified Forecast System (UFS).
SFS, version 1.0 (SFSv1) is designed to support decision-makers across the entire Weather Enterprise. It aims to provide enhanced predictions up to 12 months in advance for temperature, precipitation, drought, wildfire, heat and cold waves, hurricanes, and marine and sea ice hazards.
Under the Hood: System Components
The SFS integrates four fully coupled Earth system components:
- Atmosphere: Finite-Volume Cubed-Sphere (FV3) Dynamical Core and Common Community Physics Package (CCPP)
- Ocean: Modular Ocean Model, version 6.0 (MOM6)
- Sea Ice: Los Alamos Sea Ice Model, version 6.0 (CICE6)
- Land Surface: Noah Land Surface Model (LSM) with Multi-Parameterization (Noah-MP)
The SFS atmospheric physics package is largely based on that of Global Forecast System, version 17 (GFSv17), featuring specific updates to improve computational efficiency and better simulate large-scale climate variability across subseasonal and seasonal timescales. Key enhancements include:
- Dynamics: Computational costs are reduced by replacing the non-hydrostatic dynamical core with a hydrostatic FV3 core.
- Gravity Wave Parameterization: The European Centre for Medium-Range Weather Forecasts (ECMWF) non-stationary gravity-wave-drag parameterization was implemented to improve the representation of the Quasi-Biennial Oscillation (QBO) and polar stratospheric jets.
- Planetary Boundary Layer (PBL) Scheme: The Total Turbulence Energy Eddy-Diffusivity Mass-Flux (EDMF) PBL scheme has been adapted for the system.
- Aerosol Forcings: A 5-year sliding NASA’s Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA2) aerosol climatology is adopted to represent historical trends of aerosol forcing.
- Deep Convection: The addition of an improved prognostic convective closure enhances forecast skill for the Madden-Julian Oscillation (MJO).
- Ozone prediction: Improvements to the prognostic ozone scheme better represent the ozone hole in the Southern polar stratosphere from September through November.
Operational Progress and Reforecast Skill
Starting from March 1, 2026, the SFSbeta in NRT has been employed as experimental forecast guidance to support monthly and seasonal outlooks at the Climate Prediction Center (CPC). Moving forward, this SFS prototype will be integrated into the North American Multi-Model Ensemble (NMME) to support multi-model seasonal forecast guidance.
Hindcasts initialized from March 1, April 1, and May 1 during the 1991-2020 period show that SFSbeta’s reforecast skill consistently improves upon CFSv2. These improvements are evident in U.S. surface air temperature, precipitation, and sea surface temperature (SST), including the Nino3.4 index. See figure 1 comparing the SST anomaly correlation coefficient (ACC) for CFSv2, SFS Beta 1.0, and ECMWF’s Seasonal Forecast System, version 5 (SEAS5).
Open Data and Continuous Refinement
To gather crucial stakeholder feedback, the monthly atmospheric and oceanic reforecast and NRT forecast data are publicly available on the AWS Cloud server through NOAA Open Data Dissemination (NODD). This open-data approach allows the SFS development team to gather feedback from both internal and external NOAA stakeholders, deepen their understanding of model biases, and continuously refine the system.
As the team prepares for the final SFSv1 configuration and implementation, SFSbeta NRT forecasts will undergo rigorous updates. Based on performance metrics and valuable stakeholder feedback, the model configuration, physics parameterizations, and initialization methods will continue to evolve. Additional information regarding SFSBeta NRT specifications, background model development, and new update features is available on the SFSBeta webpage.
