EPIC Projects

Data Assimilation

One of the major sources of error in weather and climate forecasts is uncertainty related to the initial conditions that are used to generate future predictions. Even the most precise instruments have a small range of unavoidable measurement error, which means that tiny measurement errors related to atmospheric conditions and instrument location, for example, can compound over time. These small differences result in very similar forecasts in the short term (i.e. minutes, hours), but they cause widely divergent forecasts in the long term. Data assimilation systems seek to mitigate this problem by combining the most timely observational data with other sources of data, such as historical data, to provide an analysis of possible atmospheric states and the probabilities of each. Errors in weather and climate forecasts can also arise because models are imperfect representations of reality. Data assimilation systems can use techniques including stochastic physics, which applies randomized perturbations to the physical tendencies or the physical parameters of a model, to compensate for model uncertainty.

NOAA’s operational data assimilation system for the Global Forecast System (GFS) uses the Finite-Volume Cubed-Sphere (FV3) dynamical core. The FV3 currently utilizes a hybrid 4D Ensemble-Variational (4D EnVar) solver designed to represent the analysis and background error covariances and uses stochastic physics. EPIC is involved in a number of projects that seek to improve data assimilation (DA) capabilities:

  • Unified Forecast System Research to Operations (UFS-R2O) and Joint Center for Satellite Data Assimilation (JCSDA) Projects (2019-2021):
    • UFS-R2O Medium Range and Subseasonal-to-Seasonal Applications (MER/S2S) Coupled Data Assimilation/Reanalysis & Reforecasts (FY 2020-2021)
    • Enhancements to CRTM/CSEM Microwave Ocean Emissivity and BRDF Models (FY 2021)
    • Improved CRTM for UV and Passive Microwave Hydrometeor Impacted Radiances (FY 2020)
  • Joint Effort for Data Assimilation Integration (JEDI) and Community Radiative Transfer Model (CRTM): Algorithms, Observations, Applications, and Infrastructure Projects in Collaboration with the JCSDA (FY 2019-2021)
  • Improving Coupled Ensemble Prediction and Data Assimilation for the UFS (FY 2019)
  • Improving Snow Forecasts Through Advances in Land Modeling (FY 2019)

Infrastructure Coupling

In the past, Earth scientists developed separate models to predict changes to atmosphere, ocean, ice, and land. Each model contained a set of assumptions; for example, if a particular variable was not included in the model, it was held constant. These independent systems fail to account adequately for the exchange of mass, energy, and momentum between the components, which led to significant model inaccuracies, especially at longer time scales. EPIC supports the development and improvement of a fully-coupled Earth system model framework for weather and climate predictions, which joins together multiple models for more accurate predictions. As part of this process, code infrastructure must be built to “couple,” or join, these independently-developed Earth systems model components.

EPIC is currently prioritizing improvements to the National Unified Operational Predicting Capability (NUOPC) Layer of the Earth System Modeling Framework (ESMF). The NUOPC Layer is the key to interoperability in NOAA’s operational forecasting systems. It provides generic components that can be customized and defines the rules for interactions and data sharing between components. The NUOPC layer also provides the required structure for the UFS to connect various model components into a fully coupled Earth system model. EPIC is currently supporting the following infrastructure coupling projects:

  • Earth System Modeling Framework (ESMF) Infrastructure, Software Updates, and User Support for UFS Applications (FY 2019-2021)
    • Coupling Infrastructure Integration and User Support for UFS (FY 2019)
    • UFS R2O Project: Modeling Infrastructure (FY 2020-2021)

Scientific Innovation

As science evolves, the physics and mathematics underlying certain weather and climate models may be replaced by more accurate representations of Earth systems and improved modeling and measurement capabilities. Continuous improvement of the Nation’s forecasting ability requires that these scientific innovations be translated into the existing models or that new models be developed to account for these improvements in our ability to represent a model’s underlying physical processes. 

NOAA has a variety of operational weather forecasting models that provide forecasters and other decision-makers with predictions of environmental phenomena.  An example of such a model is the Global Forecast System (GFS) version 16, which added major updates in early 2021. The GFS v. 16 Finite-Volume Cubed-Sphere (FV3) dynamical core now has double the number of vertical layers. A number of model physics innovations were also included in the upgrade, among them, improvements to the atmospheric physics that will enhance snow and precipitation forecasting capabilities. 

To support the flow of scientific updates to NOAA’s operational forecasting models, a variety of EPIC projects support innovation in Earth systems modeling: 

  • Enhancements to CRTM/CSEM Microwave Ocean Emissivity and BRDF models (FY 2021)
  • Improved CRTM for UV and Passive Microwave Hydrometeor Impacted Radiances (FY 2020)
  • Convection-Allowing Model Ensembles Optimal Configuration for Short and Longer Time Scales and Multigrid Background Error Covariance Model (FY 2019)
  • Development of Process-Level Parameterizations of Model Uncertainty in the GFS/GEFS (FY 2019)
  • Improving Tropical Boundary Layer Structure and Cloud Systems at All Scales (FY 2019)
  • Improving the Microphysics Parameterization in High-Resolution for the FV3/GFS Nested Modeling System for Tropical Cyclone Predictions (FY 2019)
  • Physics and Improved Vertical Resolution for Improving Hurricane Prediction and Tropical Convection in GFS (FY 2019)
  • Medium-range to Subseasonal to Seasonal (S2S) Prediction at Convective Scales (FY 2019)

Software Improvements & Support

NOAA and the National Weather Service (NWS) are using a community modeling infrastructure framework—the Unified Forecast System (UFS)—to drive their modeling systems. The UFS bundles several component systems and the applications that connect them into a computational model with an end-to-end workflow. 

To ensure the continued success and growth of the UFS, EPIC will ensure that the entire modeling suite follows the community development paradigm. To this end, EPIC is streamlining its development environment, code management, and user support systems. It is also funding a variety of projects to optimize the code behind UFS weather models and to ensure the interoperability of different model components. These projects include:

  • UFS Community Modeling Support Project (FY 2019-2021)
    • UFS Community Modeling Infrastructure, Code Management (FY 2019-2021)
    • Contributions to the CICE Consortium (FY 2020-2021)
  • Enhancing the Development of Land Models as a Full Component of the Coupled System (FY 2019-2021)
  • Developmental Testbed Center Training and Support for EPIC Contract (FY 2021)
  • Expanding Numerical Algorithms and Community Capabilities in FV3 (FY 2021)
  • Accelerating the UFS Evaluation Capability (FY 2019)