JEDI Skylab - Observation Assessment and Evaluation

The Joint Center for Satellite Data Assimilation (JCSDA) is a partnership between agencies (NOAA, NASA, U.S. Air Force and Navy, and the UK Met Office). A primary project within the JCSDA is the Joint effort for data assimilation integration (JEDI), which strives to create a fully abstracted data assimilation system. This abstracts the numerical environmental prediction models from the observation forward operators, for instance, and is developed in collaboration with the partners in an agile framework. A second key project at the JCSDA is the Community Radiative Transfer Model (CRTM), which for the work shown here is the forward operator used for satellite radiances, reflectances, and reflectivities. At the JCSDA the JEDI Skylab system has been created to demonstrate these capabilities to validate these for partners and their subsequent research-to-operation (R2O) efforts. An additional goal of the JCSDA and JEDI Skylab is to use the demonstrator to explore new observations that may be helpful in the future.

In these demonstrations, the JCSDA extensively uses the MPAS (NCAR) and FV-3 (NOAA) models, with initializations from the NOAA GFS and GEFS. With these NWP models, the CRTM, and other forward operators in the UFO, we have developed and demonstrated various observation filters, variational bias correction and observation error models for wide-ranging observations covering those currently used by the partner agencies. While we continue to assist and collaborate with these partners to assume all the capabilities and requirements of their operational implementation we have leveraged these same filters and options for new observation types demonstrating the reusability and genericity of the JEDI code. We will show some examples of extending capabilities in JEDI to examine new observations such as Windborne balloonsonde, GNSS-R ocean surface wind (OSW), and space-based radar (reflectivities). With the Skylab demonstrator, we offer an end-to-end capability to produce forecasts and metrics, evaluating their fit to observations and various types of analysis, including self-analysis, control analysis, and external analysis. We will present our results from work done using these tools within a JEDI Skylab environment, verified against the ERA-5 re-analysis and to other observation types. This JEDI demonstration is highly flexible through YAML configuration files, creating opportunities for new users to engage more readily with the JEDI system.