What is the Unified Forecast System (UFS)?
The Unified Forecast System (UFS) is a community-based, coupled, comprehensive Earth modeling system. It is designed to provide numerical guidance for applications in the forecast suite of the National Centers for Environmental Prediction (NCEP). The UFS numerical applications span local to global domains and predictive time scales from hours to greater than one year. The UFS is also intended to be an exploratory model that can include research elements that are not part of production suite applications.
When is UFS going to be ready? Is anyone running UFS now?
Some of the UFS applications are already being used in operations and research, and others are still in development.
The UFS Medium-Range Weather Application 1.0 was released in March 2020 to the community. It includes an atmospheric model that is an updated version of the one used in the operational Global Forecast System (GFS). When we refer to GFS we mean the end-to-end global operational system that includes the model, data assimilation, post processing, products and verifications.
The atmospheric component of the UFS is running experimentally at NWS/NCEP at 13 km horizontal resolution globally out to 15 days. Development to include other components (ocean, atmospheric composition…) is underway. Progress and plans for UFS can be found in the Strategic Implementation Plan, linked here.
Most of the UFS coupled applications are already in the testing phase. Note that most of the component models targeted for the coupled UFS model (atmosphere, ocean, ice etc.) are community models. Many of these models have already become the focal stand-alone models in operations, and as community models are used in many places other than NCEP operations.
How can I get the UFS code?
Information on accessing the first release of the UFS, the Medium-Range Weather Application 1.0, is available here.
In the future, additional UFS applications will be accessible to the community.
What are the component models in the UFS?
In a comprehensive modeling system such as UFS, different component models representing atmosphere, ocean, ice, and other physical processes are connected or “coupled” together. Coupling means that the inputs one component model needs are provided by one or more of the others.
The component models currently used in the UFS are the Global Forecast System 15 (GFSv15) atmosphere, the Modular Ocean Model 6 (MOM6), the WAVEWATCH III wave model, the Los Alamos sea ice model 5 (CICE5), the Noah and Noah-MP land models, the Goddard Chemistry Aerosol Radiation and Transport (GOCART) aerosol model, the Ionosphere-Plasmasphere Electrodynamics (IPE) model, and the Advanced Circulation (ADCIRC) model for storm surge, tides, and coastal circulation.
Can there be different implementations of particular model components in the UFS; for example, multiple ocean models – or is there a single officially supported implementation of each component?
It will likely be difficult to have a single implementation of every component. However, as UFS evolves the system is being simplified by reducing the number of implementations of each component to as few as possible, given operational requirements.
This topic is discussed in the white paper for the Unified Modeling Task Force of the NOAA Research Council.
Will the UFS approach provide better forecasts?
There are several aspects of a community approach to operational modeling that are likely to result in a more rapid improvement of operational modeling, and hence in better forecasts. First, having both operational support groups, and additional research groups work on the same model will increase the talent pool and resources. This is very likely to accelerate improvement. Second, by doing research with operational models, using operational configurations, input, product generation, validation and verification, the resulting improvements can much more rapidly be adopted for operations. In this case, operational centers do not have to recode approaches, and can directly accept some of the external testing provided.
As an example of potential acceleration of transition to operations look at NCEP’s experiences with the HWRF hurricane model. The initial implementation of the model took close to a decade, partially because the entire model infrastructure needed to be operationalized. With more recent operations based projects to improve hurricane intensity forecasting, it typically take an annual cycle to get proven research into operations. With this approach, we follow the operational lead of the UK Met Office who pioneered unified modeling across forecast scales.
Are there any results that show improvements for UFS now?
Yes, an evaluation of the atmospheric component of the UFS was performed as part of its upcoming transition to operations. One of the most important metrics of weather forecasting, the 500 hpa anomaly correlation, showed significant improvement relative to the current GFS baseline. Relative to the same baseline, important improvements to hurricane track and intensity were also realized.
A summary of improvements and remaining challenges can be found in the NCEP Office of the Director Decisional Brief found at this link.
How can I access UFS data and forecasts?
Experimental forecasts and near real-time verification from GFSv15 are found here, under the Data and Real-Time Graphics headings.
Real time forecast graphics can also be found at several sites external to NOAA, including
What is FV3?
NOAA GFDL’s Finite Volume Cubed Sphere (FV3) is a dynamical core, the part of the atmospheric model that describes the equations of fluid motion. FV3 was selected from a number of different dynamical cores through a competitive process.
FV3 was designed for scientific integrity and excellent computational performance. It is also well established at all of the spatial and temporal scales that the UFS will operate at. With regard to scientific integrity, the dynamical core is able to represent fronts and behavior of atmospheric constituents in a way that maintains physical consistency. Therefore, it ultimately provides a better basis for the incorporation of atmospheric physics. Importantly, the dynamical core has a novel treatment of the vertical coordinate, which has benefits for both the treatment of physics and computational performance.
The FV3 dynamical core implementation includes incorporating FV3 into the UFS software infrastructure and developing advanced physics and data assimilation techniques to accompany it that match or exceed the skill of its current operational counterpart. In addition, NWS is working with federal partners, universities, and the community to make FV3 a fully accessible community model.
How did the selection process for the dynamical core work?
The FV3 core was chosen in a two phase, objective evaluation process that included prescribed tests, agreed upon in advance by all of the participants. The second phase of the testing was after a winnowing of several dynamical cores. In the second phase, the FV3 was chosen based on scientific and computational performance in a set of standardized tests. The evidence of the scientific and computational testing both supported the selection of FV3.
The full test reports are at this link.