The UFS effort will simplify and improve the operational Production Suite run at NOAA’s National Centers for Environmental Prediction. The scope of the Production Suite demands that UFS numerical applications span local to global domains and predictive time scales from sub-hourly analyses to seasonal predictions. Thus, the UFS is necessarily based on coupled models, with a unified coupling infrastructure based on the Earth System Modeling Framework (ESMF) and a unified Data Assimilation (DA) System based on the Joint Effort for Data Assimilation Integration (JEDI) infrastructure. The UFS is ensemble-based, with robust procedures for gathering model biases and other performance information through a standardized reforecast and reanalysis process. Crucial to the concept of the UFS is that applications are essentially specific configurations of a common code base, not separate codes built from the ground up. The figure above shows the set of applications that form the UFS.
Science applications and activities are described in the UFS Strategic Implementation Plan (SIP), a document prepared by contributors from a wide range of partner organizations. This page is drawn from the front matter of the SIP document.
The basis for UFS development is the Finite Volume 3 (FV3) atmospheric dynamical core, a System Architecture, and companion Infrastructure that are required for building applications out of the individual models (atmosphere model, ocean model, wave model, etc.).
FV3-Based Forecast Systems
First to be delivered among the set of UFS applications is the Global Forecast System 15 (GFSv15) based on the FV3 dynamical core. A FV3-based global ensemble prediction system for sub-seasonal timescales will be implemented in 2020, with consideration given to a high-resolution version and a better treatment of uncertainty in the ensemble system. The FV3-based seasonal forecast system will provide model guidance out to 9 months. Global prediction will extend into the deep atmosphere with the development of Deep Atmospheric Dynamics (DAD) for an FV3-based Whole Atmosphere Model (WAM) coupled to an Ionosphere, Plasmasphere, and Electrodynamics Model (IPE).
Atmospheric Physics and Composition
The development of advanced physics for the anticipated 2021 version of GFS will include an assessment of several physics packages appropriate not just for global weather models, but for convection-allowing models and seasonal prediction. The advanced physics and all subsequent UFS physics packages will be expressed in the framework of a Common Community Physics Package, which is a collaborative framework for developing physical parameterizations.
Physics development will be complemented by the development of a general unified atmospheric composition modeling system that will better account for both trace gas effects of radiation and aerosol effects on radiation and clouds. It will also improve the handling of satellite observations by properly accounting for aerosol and trace gas effects during data assimilation. The development of a generic atmospheric composition component and its accompanying DA system will focus on integration into the UFS architecture so that two-way interactive coupling with atmospheric physics and consistent coupling with dynamics can be supported. A unified emission system with the capability of providing model-ready, global anthropogenic and natural source emissions inputs for aerosol and gas phase atmospheric composition across scales will also be developed.
Model development for the UFS is accompanied by a revitalization of global DA techniques. This starts with enabling the DA system to effectively use observational data. Tasks include developing new data types, extending existing data types, and creating new techniques that can take advantage of data that’s available but not fully used. Hybrid (ensemble + variational) DA in its various forms is the current state-of-the-science for environmental prediction and is expected to remain so over the next several years. This implies that the DA framework will be required to support several current technologies for a variety of applications. The UFS DA effort will utilize the JEDI framework to build out a project for global numerical weather prediction inter-comparison between hybrid 4DEnVar (current technology) and Hybrid 4DVar (with adjoint) and do so within a rapidly updating (about every hour) global analysis system for atmospheric applications.
The fully coupled Earth system models in the UFS will be accompanied by a coupled DA capability that can initialize the models to improve predictability from weather to subseasonal to seasonal timescales. This includes a comprehensive land DA system for a comprehensive land surface model, to be coupled with other components in the UFS.
Regional and Hurricane Modeling
In addition to global modeling, the UFS will provide a capability for regional modeling to provide high-resolution numerical guidance. A stand-alone regional model forms the technical core of this activity, from which a regional modeling application suite is being developed. These applications include an FV3-based replacement for the NAM/RAP and RAP/HRRR along with their associated ensemble prediction systems and data assimilation capabilities. A new Hurricane Analysis and Forecasting System (HAFS) provides an operational analysis and forecast out to seven days, with reliable and skillful guidance on Tropical Cyclones (TC) track and intensity. Key to this new prediction system is the development of grid nests that move with individual storms within the global model, and a coupling capability for these nests. UFS will integrate with a comprehensive NOAA Water Initiative through collaborative development of UFS and the National Water Model. The goal is to provide people and governments better access to the water information they need for their unique circumstances. Planning is also underway for integrating the UFS with regional quasi near-real-time ecological forecasts, such as for Harmful Algae Blooms (HABs) and hypoxia.
Verification and Validation
A key element of the UFS vision is the development of a unified suite of metrics that cover all prediction scales. These will be based on the community Model Evaluation Tools (METplus) developed at NCAR. The metrics form the core of an evidence-based evaluation of all components. Evidence-based evaluation is needed to ensure that new modeling systems are better than those being replaced. Plans for Testing and Evaluation have evolved substantially through community workshops such as the one hosted in summer 2018 by the Developmental Test Center.
Envisioning an Integrated Community System
Taken together, this portfolio of projects represents a path toward building out the UFS into a true community-based modeling system for numerical Earth system prediction.