Image: Adobe Stock
PSL and the Unified Forecast System
Overview
Sophisticated computer models are vital tools for forecasters, providing weather estimates for the near and distant future that help people day-to-day and serve crucial information to decision-makers to improve community safety. These models and their underpinnings are continually evolving as research advances. PSL plays a pivotal role in improving these models through innovative research and development.
PSL works with the Unified Forecast System (UFS), a community-based framework for modeling Earth Systems, across a wide range of timescales. The lab collaborates with a community of developers that contribute to the UFS, including other NOAA contributors, to improve the forecast model and data assimilation system, as well as developing diagnostic tools and supporting reanalysis/reforecast datasets.
Select a tab above to learn more about PSL's work in specific areas related to the UFS.
What is the UFS? Learn more about the Unified Forecast System
PSL's UFS work contributes to NOAA Research's science priorities and foundational capabilities, including:
- improving approaches to modeling the planet,
- integrating emerging technologies, and
- transitioning research into operations and applications.
Learn more about NOAA Research's science priorities and foundational capabilities
Projects
PSL participates in the development of several UFS-based applications and contributes to several projects centered around the UFS.
UFS-R2O
The Unified Forecast System - Research to Operations (UFS-R2O) Project is a major ongoing undertaking to build partnerships across the research and operational environmental modeling communities with a focus on transitioning UFS-based applications into the National Weather Service (NWS) operational modeling suite. The project is based out of NOAA’s Office of Science and Technology Integration in the NWS, and is in collaboration with NOAA Research’s Weather Program Office.
PSL participates actively via development and coordination leadership of atmospheric physics, stochastic physics, data assimilation and reanalysis.
UFS-Arctic
PSL is developing a fully-coupled regional Arctic configuration of the Unified Forecast System: UFS-Arctic. In addition to research, its use includes support for safe maritime transportation, search and rescue, oil spill response, commercial fishing, coastal communities, aviation safety, emergency and ecosystem management, tourism, the energy and mineral industries, and research.
The development is part of a larger longstanding research effort within PSL on Arctic science and data and will replace current experimental sea ice forecasting capabilities.
Seasonal Forecast System
PSL is a major contributor to the development of the Seasonal Forecast System (SFS), which is a seasonal time-scale application of the UFS. We are developing model physics parameterizations building on the GFSv17 application, creating the initial condition for the retrospective forecasts needed for forecast calibration and skill assessment. The initial conditions are being developed by both a coupled reanalysis and replay. Additionally, atmospheric physics development targeting the lower resolution to reduce sea surface temperature (SST) biases within the SFS has key contributions from PSL.
Project EAGLE
NOAA EAGLE is a collaboration between National Weather Service, NOAA Research, and the Earth Prediction Innovation Center (EPIC) program to develop AI weather forecast tools and advance them into near real time practice. Our team actively contributes to the curation of multiple training datasets within NOAA [link to the reanalysis page], development of the data access tools, and the development of the community tools for training of the EAGLE models within the Anemoi framework. Our recent scientific focus is on development of the stretch-grid Nested-EAGLE model with the 6km grid refinement over CONUS and on the development of the coupled AI global model trained on the UFS-Replay dataset.
Hourly Data Assimilation
Increasing the frequency of analysis updates in the global data assimilation (DA) system is expected to improve the representation of fast-moving systems such as hurricanes, as well as provide a pathway to consolidate regional and global applications. PSL has led an effort to evaluate the impacts of moving from 6-hourly to 1-hourly assimilation windows in the global system, and found that hourly DA can improve the fit to observations, particularly in wind. PSL has also tested overlapping assimilation windows to handle real-time observational data latency and found similar improvements. Work is ongoing to investigate the feasibility of using AI to provide hourly-updated initial conditions for both physical and AI model forecasts, building upon the existing 6-hourly traditional DA system.
Transition of the Global Data Assimilation System to JEDI
PSL has been leading the development of ensemble solvers within the Joint Effort for Data assimilation Integration (JEDI), NOAA’s next-generation operational data assimilation system. These solvers have undergone extensive testing across a diverse range of the UFS applications, including the atmosphere, ocean, sea ice, land, and atmospheric composition. PSL is currently providing scientific and technical support to the NOAA Office of Modeling and Development (previously the Environmental Modeling Center) as it works to transition the full JEDI suite including the ensemble solvers for future standalone and coupled UFS operational applications from the Global Data Assimilation System (GDAS). Meanwhile, PSL continues to advance and implement innovative data assimilation methodologies within JEDI through established collaboration with partners in academia, as well as national and international research and operational institutions.
Water in the West
Under the Water in the West project, our teams have made several contributions aimed at improving medium-range precipitation prediction skill over the Mountain West within the UFS framework. In this project, PSL contributed to the development of the variable-resolution UFS system, featuring grid refinement over the western United States and northeastern Pacific (UFS-AR and AR-AFS), through the development of scale-aware physics parameterizations and workflows. In parallel, ongoing work is focused on developing novel diagnostic methods to evaluate the representation of the multiscale interaction present in atmospheric rivers (ARs), driven by diabatic processes. PSL also contributes to AI model development. We developed both a variable resolution version of the AI model (with grid refinement over CONUS) and a coupled (ocean-atmosphere-ice) AI model. From the data assimilation perspective, we contributed to the development of the in-core memory ensemble forecasting within the JEDI framework and to understanding the role of the all-sky microwave assimilation within the core of the atmospheric rivers. We are now actively investigating the opportunities to use AI to improve data assimilation of satellite observations within the core of the ARs.
Precipitation Grand Challenge
The NOAA Precipitation Prediction Grand Challenge (PPGC) aims to substantially improve prediction of high-impact precipitation across timescales through advances in coupled Earth system modeling, data assimilation, observations, and physics parameterizations. A major focus of the effort is the development of next-generation, rapidly cycling prediction systems capable of better representing the multi-scale processes that govern extreme precipitation, including convection, atmospheric rivers, land-atmosphere feedbacks, and cloud microphysical interactions.
At PSL, researchers are contributing to the PPGC through development of advanced model physics and coupled prediction capabilities within the Unified Forecast System. PSL is helping advance research toward hourly updated global data assimilation to improve initialization of rapidly evolving precipitation systems and better capture interactions across scales, from convective storms to larger-scale moisture transport and tropical variability. In addition, with support from the Joint Technology Transfer Initiative(JTTI), PSL is developing next-generation physics approaches that improve the representation of unresolved processes in both regional and global coupled prediction systems.
Data Assimilation
PSL has a long history of developing fundamental tools for ensemble data assimilation, including original development of the Ensemble Kalman Filter (EnKF) algorithms in the Gridpoint Statistical Interpolation (GSI) data assimilation system (Hamill et al., 2001, Whitaker et al., 2008, Bishop et al., 2017).
GSI is a widely used data assimilation system that underpins current operational capabilities for NOAA and NASA, including a variety of global, regional, and tropical cyclone applications. As NOAA is transitions from GSI to the new Joint Effort for Data assimilation Integration (JEDI) infrastructure, PSL is migrating the fundamental science of EnKF solvers from GSI to JEDI (Frolov et al., 2024). EnKF JEDI solvers are now undergoing testing in several weather forecast centers including NOAA, NASA, the U.S. Navy, the U.S. Airforce, National Center for Atmospheric Research (NCAR), UK MetOffice , and the Australian Bureau of Meteorology .
In more recent work, in close collaboration with the National Weather Service, PSL scientists developed a major upgrade to the land data assimilation used in NOAA's global Numerical Weather Prediction (NWP) system, GFSv17. For the GFSv17 update, PSL led the development of NOAA's first global soil moisture and soil temperature analysis (Draper et al., 2025). This is an ensemble-based strongly coupled land/atmosphere assimilation system, and represents a major advance in land data assimilation methodology within NWP (Draper et al., 2021). PSL upgraded the snow depth analysis used in GFSv17 to 2D Variational data assimilation of satellite snow cover and station snow depth, implemented within the JEDI framework (Gichamo et al. 2022; Shlyaeva et al., 2024). Testing prior to the launch of GFSv17 showed that both of these new land DA schemes significantly improve the GFSv17 forecast skill.
PSL is currently developing and testing new scientific capabilities within JEDI, including multiscale localization, correlated observational errors, coupled data assimilation, and hourly data assimilation for UFS (Slivinski et al., 2022). In addition to advancing traditional data assimilation wool within the JEDI framework, PSL is also actively investigating how AI can be used to provide more skillful initial conditions for UFS (Slivinski et al., 2025).
References and Related Publications
2025-Present
Draper, C., Whitaker, J. & Barlage, M. (2026) Exploring strongly coupled land/atmosphere data assimilation for numerical weather prediction. Quarterly Journal of the Royal Meteorological Society, 152(774), e70033. https://doi.org/10.1002/qj.70033
Slivinski, L. C., and J. S. Whitaker, 2025: Investigating the Impact of Rapid Cycling in a Global Data Assimilation System. Mon. Wea. Rev., 153, 2933–2945, https://doi.org/10.1175/MWR-D-25-0097.1.
Slivinski, L. C., Whitaker, J. S., Frolov, S., Smith, T. A., & Agarwal, N. (2025). Assimilating observed surface pressure into ML weather prediction models. Geophysical Research Letters, 52, e2024GL114396. https://doi.org/10.1029/2024GL114396
2020-2024
Frolov, S., Shlyaeva, A., Huang, W., Sluka, T., Draper, C., Huang, B., et al. (2024). Local volume solvers for Earth system data assimilation: Implementation in the framework for Joint Effort for Data Assimilation Integration. Journal of Advances in Modeling Earth Systems, 16, e2023MS003692. https://doi.org/10.1029/2023MS003692
Stanley, Z. C., Draper, C., Frolov, S., Slivinski, L. C., Huang, W., & Winterbottom, H. R. (2024). Vertical localization for strongly coupled data assimilation: Experiments in a global coupled atmosphere-ocean model. Journal of Advances in Modeling Earth Systems, 16, e2023MS003783. https://doi.org/10.1029/2023MS003783
Gichamo, T. Z., and C. S. Draper, 2022: An Optimal Interpolation–Based Snow Data Assimilation for NOAA’s Unified Forecast System (UFS). Wea. Forecasting, 37, 2209–2221, https://doi.org/10.1175/WAF-D-22-0061.1.
Slivinski, L. C., Lippi, D. E., Whitaker, J. S., Ge, G., Carley, J. R., Alexander, C. R., & Compo, G. P. (2022). Overlapping Windows in a Global Hourly Data Assimilation System. Monthly Weather Review, 150(6), 1317–1334. https://doi.org/10.1175/mwr-d-21-0214.1
Draper, C. S. (2021). Accounting for land model error in numerical weather prediction ensemble systems: toward ensemble-based coupled land/atmosphere data assimilation. Journal of Hydrometeorology, 22(8), 2089–2104. https://doi.org/10.1175/jhm-d-21-0016.1
Pre-2020
Bishop, C. H., Whitaker, J. S., & Lei, L. (2017). Gain form of the ensemble transform Kalman filter and its relevance to satellite data assimilation with model space ensemble covariance localization. Monthly Weather Review, 145(11), 4575–4592. https://doi.org/10.1175/mwr-d-17-0102.1
Frolov, S., C. H. Bishop, T. Holt, J. Cummings, and D. Kuhl, 2016: Facilitating Strongly Coupled Ocean–Atmosphere Data Assimilation with an Interface Solver. Mon. Wea. Rev., 144, 3–20, https://doi.org/10.1175/MWR-D-15-0041.1.
Whitaker, J. S., T. M. Hamill, X. Wei, Y. Song, and Z. Toth, 2008: Ensemble Data Assimilation with the NCEP Global Forecast System. Mon. Wea. Rev., 136, 463–482, https://doi.org/10.1175/2007MWR2018.1.
Hamill, T. M., Whitaker, J. S., & Snyder, C. (2001). Distance-dependent filtering of background error covariance estimates in an ensemble Kalman filter. Monthly Weather Review, 129(11), 2776–2790. https://doi.org/10.1175/1520-0493(2001)129<2776:ddfobe>2.0.co;2
Artifical Intelligence for Weather Forecasting
PSL was an early adopter of artificial intelligence (AI) technologies for weather forecasting within NOAA (Frolov et al. 2025). Our work has focused on contributing to the NOAA-wide Project EAGLE. We actively contribute to the curation of multiple training datasets within NOAA, development of data access tools, and the development of the community tools for training of the EAGLE models within the Anemoi framework. Our recent scientific focus is on development of the stretch-grid Nested-EAGLE model with the 6km grid refinement over the contiguous United States (CONUS) and on the development of the coupled AI global model trained on the UFS-Replay dataset.
Nested and coupled models
PSL developed several AI forecasts models that are designed to uniquely benefit from the NOAA initial conditions, such as the GFS, HRRR, and the coupled GFVSv17. This resulted in two models: Nested-EAGLE and Ocean-EAGLE.
Nested-EAGLE
The Nested-EAGLE combines GFS (sub-sampled at 25km resolution) and HRRR (sub-sampled at 6km resolution) initial conditions and appears to excel at representing 2-meter temperatures and 10-meter winds over CONUS.
The companion coupled atmosphere-ocean-ice-land model is in development using the UFS-Replay dataset and should be consistent with the initial conditions from the upcoming GFSv17 coupled model.
Ocean-EAGLE
With an aim of developing coupled emulators for seasonal-to-seasonal applications, we have developed a global ocean-only emulator (Ocean-EAGLE) for medium-range predictions, trained on NOAA’s UFS-Replay ocean dataset.
Using a 24 hour time step, single initial condition, and without any autoregressive training, we have produced a robust emulator that provides skillful forecasts for 10–15 day lead times (Agarwal et al, 2026). We have further demonstrated the use of Mahalanobis distance as loss (called M-Loss) that improves the forecast skill compared to the Mean Squared Error loss by explicitly accounting for the correlations between the target variables. This loss acts as a statistical-dynamical regularizer for the slow, covarying dynamics of the global oceans, offering a better background forecast for downstream tasks such as coupled data assimilation.
This work is extended further by coupling independent ocean-only and atmosphere-only emulators (both trained on UFS-Replay) by exchanging the states of the coupled variables during the inference. This development uses European Centre for Medium-Range Weather Forecasts’s Anemoi ecosystem for model development and NOAA’s EAGLE tools for postprocessing and analysis. Using a suite of baselines to isolate the impact of coupling, we have shown that the two-way coupling significantly impacts the forecast of key surface ocean quantities such as sea surface temperature (SST) and sea surface height (SSH) on 10-15 days lead times. The atmospheric variables however show a relatively lower impact on standard error metrics at these time scales, but this is under active investigation using longer lead times and process-level diagnostics. The framework is also set for further advancement using a more optimized set of coupled variables, increased spatial resolution, and refined ocean-only and atmosphere-only components using iterative fine tuning.
AI-accelerated data assimilation algorithms
In addition to developing AI forecast models, our team is actively engaged in the development of the AI-accelerated data assimilation algorithms. Our early attempt to combine AI forecast models with traditional data assimilation methods exposed limitations of the current AI models (Slivinski et.al 2025). As a result, we are now investigating end-to-end AI modeling systems that combine AI data assimilation with AI forecasting. An early prototype of such an end-to-end system can be illustrated by the HealDA system that was developed in collaboration with the NVIDIA and MITRE (Gupta et.al 2026).
References
Agarwal, N., Smith, T.A., Frolov, S. and Slivinski, L.C., 2026. Skillful Global Ocean Emulation and the Role of Correlation-Aware Loss. arXiv preprint arXiv:2604.18727.
Frolov, S., Garrett, K., Jankov, I., Kleist, D., Stewart, J. Q., & Ten Hoeve, J. (2025). Integration of Emerging Data-Driven Models into the NOAA Research-to-Operations Pipeline for Numerical Weather Prediction. Bull. Amer. Meteor. Soc., 106, E430–E437. https://doi.org/10.1175/bams-d-24-0062.1
Gupta, A., Subramaniam, A., Pritchard, M. S., Kashinath, K., Frolov, S., Lieberman, K., Miller, C., Silverman, N., & Brenowitz, N. D. (2026). HealDA: Highlighting the importance of initial errors in end-to-end AI weather forecasts. arXiv. https://doi.org/10.48550/arXiv.2601.17636
Slivinski, L. C., Whitaker, J. S., Frolov, S., Smith, T. A., & Agarwal, N. (2025). Assimilating observed surface pressure into ML weather prediction models. Geophysical Research Letters, 52, e2024GL114396. https://doi.org/10.1029/2024GL114396
Physics
PSL contributes to physics development across multiple applications within the Unified Forecast System (UFS), spanning a wide range of spatial and temporal scales, from high-resolution storm-scale prediction to global coupled subseasonal and seasonal forecasting systems.
A major focus in PSL is the development of moist physics for the atmospheric component of the UFS. Our research emphasizes multiscale interactions between atmospheric physics and dynamics, including:
- Arctic cloud and boundary layer processes,
- tropical convection and wave interactions associated with the Madden-Julian Oscillation (MJO) and other equatorial modes, and
- precipitation processes linked to atmospheric rivers and complex terrain.
Leveraging observations
PSL’s physics development is closely connected to observations and process-level understanding. We leverage expertise from PSL-led field campaigns and observing programs, including the Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC) and the Atlantic Tradewind Ocean-Atmosphere Mesoscale Interaction Campaign (ATOMIC), to develop forcing datasets for the Community Common Physics Package (CCPP) Single Column Model (SCM) and to evaluate physical parameterizations. We also make use of observational networks across the western United States and other regions to assess model performance across scales and environments.
New and improved parameterizations
A central aspect of our work is the development of scale-aware physics parameterizations that can operate consistently across model resolutions and applications. In particular, we contribute to the development of scale-aware cumulus convection schemes designed to represent convective processes across the “gray zone” between parameterized and explicitly resolved convection.
Convective cloud systems play a fundamental role in regulating tropical variability and predictability, yet many of the key cloud-environment interactions occur on scales that remain unresolved in operational weather and climate models. PSL is developing new convective parameterizations for the UFS that improve the representation of convection–moisture coupling, entrainment, and detrainment processes. This includes research on moisture-sensitive closure formulations and innovative plume frameworks that account for the internal structure of convective updrafts and their surrounding environment. These efforts are designed to improve simulations of the Madden-Julian Oscillation (MJO), tropical waves, and other important sources of subseasonal-to-seasonal predictability.
Tools and testing frameworks
To support this work, PSL develops and applies a hierarchy of research tools and testing frameworks. These include a wave-harness capability coupled to the SCM to better understand interactions between atmospheric waves and physical parameterizations, contributions to the development of a UFS aquaplanet configuration in collaboration with EPIC (Earth Prediction Innovation Center), and telescoping nested modeling frameworks that bridge scales from numerical weather prediction (NWP) down to large-eddy simulation (LES). Together, these tools allow us to isolate physical processes, investigate multiscale physics-dynamics interactions, and accelerate model development and testing across the broader UFS community.
Research Highlights
Related Publications
2026
Solomon, A, J-W Bao, L. Bengtsson, J. Han (2026): Simulating Cloud-driven Turbulence in Stratocumulus Cloud Systems over Sea Ice in Weather-scale Models. Submitted to Journal of Advances in Earth Modeling.
Bengtsson, L., I-K. Hu, S. Tulich, J. Han, M. Zhang, E. Grell, W. Li, I. McCoy, S. Baidar, E. Thompson, A. Brewer (2026): A Prognostic Updraft Velocity Formulation Accounting for Perturbation Pressure Forcing in a Convective Parameterization. Submitted to AMS Monthly Weather Review.
Bengtsson, L. (2026): Moisture-Flux-Sensitive Convection Strengthens MJO Preconditioning. Geophysical Research Letters. https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2026GL122033
Groot, E., H. Christensen, X. Sun, K. Newman, W. Lfarh, R. Roehrig, L. Bengtsson, & J. Simonson (2026): How different are deterministic physics suites when coupled to fixed model dynamics and why? Submitted to Journal of European Meteorological Society.
Maithel, V., B. Wolding, S. N. Tulich, M. Gehne, J. Dias, X.-W. Quan, & L. Bengtsson (2026): Plume Model Assessment of Tropical Convection in the NOAA Unified Forecast System. Wea. Forecasting, 41, 489-503, https://doi.org/10.1175/WAF-D-25-0152.1
2020-2025
Hu, I-K., Chen, X., Bengtsson, L., Thompson, E. J., Dias, J., & Tulich, S. N. (2026). Utilizing ATOMIC observations for assessing marine shallow cumuli in single column models. Journal of Advances in Modeling Earth Systems, 18, e2024MS004814, https://doi.org/10.1029/2024MS004814
Bengtsson, L., Tulich, S. N., Dias, J., Wolding, B., Hall, K. J. C., Gehne, M., et al. (2025): The crucial role of the initial state in MJO prediction. Geophysical Research Letters, 52, e2025GL115833, https://doi.org/10.1029/2025GL115833
Bengtsson, L., and J. Han (2024): Updates to NOAA’s Unified Forecast System’s cumulus convection parameterization scheme between GFSv16 and GFSv17. Wea. Forecasting, 39, 1559–1570, https://doi.org/10.1175/WAF-D-23-0232.1
Bernardet, L., L. Bengtsson, A. Reinecke, F. Yang, M. Zhang, K. Hall, et al. (2024): Common Community Physics Package: Fostering Collaborative Development in Physical Parameterizations and Suites. Bull. Amer. Meteor. Soc., 105, E1490–E1505, https://doi.org/10.1175/BAMS-D-23-0227.1
Stefanova, L.; Meixner, J.; Wang, J.; Ray, S.; Mehra, A.; Barlage, M.; Bengtsson, L.; Bhattacharjee, P. S.; Bleck, R.; Chawla, A.; Green, B. W.; Han, J.; Li, W.; Li, X.; Montuoro, R.; Moorthi, S.; Stan, C.; Sun, S.; Worthen, D.; Yang, F.; Zheng, W. (2022): Description and Results from UFS Coupled Prototypes for Future Global, Ensemble and Seasonal Forecasts at NCEP. Office Note (National Centers for Environmental Prediction (U.S.)); 510. https://doi.org/10.25923/knxm-kz26
Bengtsson, L., Gerard, L., Han, J., Gehne, M., Li, W., & Dias, J. (2022): A Prognostic-Stochastic and Scale-Adaptive Cumulus Convection Closure for Improved Tropical Variability and Convective Gray-Zone Representation in NOAA’s Unified Forecast System (UFS). Monthly Weather Review, 150(12), 3211–3227.
Bengtsson, L., Dias, J., Tulich, S. N., Gehne, M., & Bao, J.-W. (2021): A stochastic parameterization of organized tropical convection using cellular automata for global forecasts in NOAA's Unified Forecast System. Journal of Advances in Modeling Earth Systems, 13, e2020MS002260. https://doi.org/10.1029/2020MS002260
Pre-2020
Bengtsson, L., J. Bao, P. Pegion, C. Penland, S. Michelson, and J. Whitaker (2019): A Model Framework for Stochastic Representation of Uncertainties Associated with Physical Processes in NOAA’s Next Generation Global Prediction System (NGGPS). Mon. Wea. Rev., 147, 893–911. https://doi.org/10.1175/MWR-D-18-0238.1
Bengtsson, L., J. Dias, M. Gehne, P. Bechtold, J. Whitaker, J. Bao, L. Magnusson, S. Michelson, P. Pegion, S. N. Tulich, and G. N. Kiladis (2019): Convectively Coupled Equatorial Wave Simulations Using the ECMWF IFS and the NOAA GFS Cumulus Convection Schemes in the NOAA GFS Model. Mon. Wea. Rev., 147, 4005–4025. https://doi.org/10.1175/MWR-D-19-0195.1
Reanalysis
PSL develops or helps develop several reanalysis products that use and improve the UFS.
NOAA NASA Joint Archive (NNJA) of Observations for Earth System Reanalysis
PSL, in collaboration with the National Weather Service, NASA, and artificial intelligence startup partner Brightband, develops and maintains the NOAA-NASA Joint Archive (NNJA).
NNJA is an experimental effort to provide a unified, curated dataset of Earth system observations from 1979 to the present, hosted on AWS S3. Designed to support Earth System reanalysis research, the archive enables consistent comparison across reanalysis efforts by minimizing variability in input observations. The dataset includes ocean, land, ice, and atmospheric data in formats such as BUFR, IODA, and NetCDF, along with quality control tools and metadata like observation errors and black/white lists. Future developments aim to expand data coverage to near real-time, enhance usability through APIs, and provide robust diagnostics, offering broad value to the Earth system science community.
UFS Replay
PSL has developed a new “replay” dataset to support improvements to the Global Ensemble Forecast System (GEFSv13) and Global Forecast System (GFSv17). It is a modern reanalysis-like dataset for coupled atmosphere, ocean, ice, land, and waves.
By nudging a coupled model toward ERA5 and ORAS5 reanalysis data, this dataset enhances the accuracy of retrospective forecasts and supports machine learning applications. It provides initial conditions for the UFS model and AI training data from 1979 to 2025. The effort aims to improve medium-range global weather predictions and reduce model biases through innovative data assimilation techniques.
Visit the UFS Replay project page
20th Century Reanalysis
The 20th Century Reanalysis is a flagship atmospheric reanalysis that uses advanced data assimilation and surface pressure observations to create a comprehensive 4D weather dataset spanning 1806 to 2015. This allows researchers to analyze modern atmospheric patterns within a deep historical context. This was a collaborative project between NOAA, the Department of Energy, and the Cooperative Institute for Research in Environmental Sciences (CIRES) at the University of Colorado Boulder.
Visit the 20th Century Reanalysis project page
Seasonal Forecast System Reanalysis
PSL is using the Global Forecast System (GFSv17) to generate historic Earth system reconstructions that span from 1979 to the present. These reconstructions provide a consistent, gap-filled estimate of past atmospheric, oceanic, ice, and land conditions, simulating the atmosphere as accurately as possible using modern data assimilation techniques and today’s best forecast models. While the initial customer for this reanalysis is the initialization of the Seasonal Forecast System, we are actively exploring how this effort could be expanded to provide high-quality, high-resolution model simulations for training of the AI models for Project EAGLE.
Downstream Applications and Research Systems
PSL contributes to a wide variety of downstream applications within the Unified Forecast System ecosystem, ranging from idealized research tools to fully coupled operational forecast systems. These applications enable hierarchical testing and development of model physics, Earth system coupling, data assimilation, and scale-aware prediction capabilities across weather and climate timescales.
Our activities span global and regional prediction systems, Arctic and atmospheric river applications, seasonal prediction, reanalysis, and idealized frameworks designed to better understand multiscale atmospheric processes and physics-dynamics interactions.
Single-Column Model and Aquaplanet hierarchical testing tools
PSL uses a hierarchy of modeling tools to accelerate the development and evaluation of physical parameterizations in NOAA’s Unified Forecast System (UFS). At the foundation is the Common Community Physics Package (CCPP) Single Column Model (SCM), a lightweight framework that runs the same physics as the full atmospheric model in a single atmospheric column. By combining the SCM with observations from field campaigns, researchers can create controlled experiments that reproduce real-world cloud, precipitation, and boundary-layer evolution. This approach provides an efficient environment for diagnosing model biases, testing new parameterizations, and improving the representation of atmospheric processes before they are implemented in full forecast systems.
To bridge the gap between highly idealized tests and full global simulations, PSL is also developing intermediate-complexity modeling frameworks. These include an enhanced SCM capable of representing interactions between convection and large-scale atmospheric waves, as well as an idealized “aquaplanet” version of the UFS developed in collaboration with NOAA’s Earth Prediction Innovation Center (EPIC).
By removing the complexities of land, topography, and other real-world influences, these frameworks provide powerful tools for understanding how physical parameterizations shape tropical variability, cloud organization, and large-scale atmospheric circulation. Together, these hierarchical testing capabilities help ensure that advances in UFS physics are grounded in both observations and process-level understanding.
Global UFS applications: SFS, GFS and GEFS
PSL contributes to the development and advancement of the global UFS applications, including the Seasonal Forecast System (SFS), the Global Forecast System (GFS), and the Global Ensemble Forecast System (GEFS). Our contributions span atmospheric physics development, coupled Earth system modeling, data assimilation, and reanalysis generation in support of both research and operational forecasting.
A major focus within PSL is the development of advanced and scale-aware physical parameterizations for the atmosphere, including cumulus convection, cloud microphysics, planetary boundary layer processes, and cloud-radiation interactions. PSL also contributes to initialization, uncertainty representation and coupled data assimilation activities (see the Data Assimilation tab), as well as the generation and evaluation of reanalysis products used for model development, diagnostics, and predictability studies.
Through these efforts, PSL helps improve prediction skill across a wide range of timescales, from medium-range weather forecasting to subseasonal and seasonal prediction. This work is carried out in close partnership with other OAR laboratories, the National Weather Service, and the broader UFS community.
Regional UFS applications: UFS-Arctic, HAFS, AR-AFS, RRFS and the SRW application
PSL contributes to several regional UFS applications, including UFS-Arctic, the Hurricane Analysis and Forecast System (HAFS), the Atmospheric River Atmospheric Forecast System (AR-AFS), the Rapid Refresh Forecast System (RRFS), and the Short-Range Weather (SRW) Application.
A central component of PSL’s regional modeling activities is the development of advanced physical parameterizations within the Common Community Physics Package (CCPP), a shared physics infrastructure used across multiple UFS applications. By developing unified and scale-aware physics capabilities within the CCPP framework, PSL helps support consistent model improvements across convection-permitting, regional, and global prediction systems.
PSL also contributes to the development of coupled regional prediction systems tailored to specific environments and hazards. In particular, UFS-Arctic focuses on high-resolution coupled atmosphere-sea ice prediction and Arctic boundary layer processes, supporting improved forecasts of sea ice, clouds, and surface exchanges in polar regions.
The work is conducted in partnership with other NOAA Research laboratories, the National Weather Service, and the broader UFS community.