In 2019 JCSDA was awarded a Disaster Relief Appropriations Supplemental (DRAS) grant from NOAA to accelerate JEDI development and integration, with an emphasis on connecting that development to operational use and systems. The main goals were to accelerate towards use in operational NWS production suites, improving R2O and O2R, and advancing tools for use of observations, all of which were accomplished. With the DRAS funding JCSDA was able to add an incredible amount of workflow and infrastructure support, develop and combine the components of JEDI SkyLab, improve background error representation, optimize software, develop marine data assimilation and improve the Community Radiative Transfer Model (CRTM).
The original goal of DRAS was to “accelerate the development of software from projects led by the Joint Center for Satellite Data Assimilation (JCSDA), including the Joint Effort for Data
assimilation Integration (JEDI)...[to] improve Data Assimilation (DA) and model performance intended to result in more accurate and longer-range forecasting”. JCSDA accomplished this with flying colors, surpassing every milestone from background error covariance modeling to a JEDI-based ocean sea-ice DA prototype.
DRAS enabled an enormous amount of progress to be made in JEDI. The biggest accomplishment was the development and maturation of SkyLab, the comprehensive JEDI testbed. SkyLab allows new ideas, systems, and scenarios to be tested. It has immense flexibility for data assimilation science testing, and can also run operational-sized experiments, adding even more functionality. With DRAS, JCSDA was able to develop the generic workflow used by SkyLab and all of the key components that transfer from SkyLab to many other JEDI uses.
One such component is the System-Agnostic Background Error Representation (SABER). SABER is a repository of background error covariance “blocks” that can be used by any model interfaced with JEDI. These blocks can be mixed and matched for different applications, depending on which is the best fit for the experiment or model. Some of the blocks already developed include spectral covariance/localization, GSI-like covariance, explicit diffusion covariance/localization, and the Background Error on an Unstructured Mesh Package (BUMP). BUMP includes NICAS, a grid-point smoother for correlation or localization, and uses a standard-deviation operator, a statistical vertical balance operator, and smooth estimates of the derivatives to do variable changes from streamfunction and velocity potential to horizontal wind components. The SABER blocks are put together in a chain that can be mixed and matched depending on the needs of the experiment. Hybrid ensemble-climatological covariances can be done in SABER as well, with no constraint on the number of hybrid components. Nearly all SABER blocks are model-agnostic and configurable.
Several other key SkyLab components were developed through DRAS. The Interface for Observation Data Access (IODA) and the Experiments and Workflow Orchestration Kit (EWOK) together ingest, convert, and store observations and backgrounds, and can be used in several fully cycling DA and Ensemble Data Assimilation (EDA) systems. They are highly configurable and portable; they can be used with any platform, observations, model, or algorithm. The Research Repository for Data and Diagnostics (R2D2) is a software system that performs data management, registration, and configuration. R2D2 stores observations and SkyLab experiment results for a rolling total of over 10 million data files across more than 30 data stores and 7 hubs, which are accessible from 6 HPC hosts. Altogether, JEDI SkyLab is an invaluable tool for research and development.
For the last five years DRAS has provided some of the core funding for JCSDA’s Sea Ice, Ocean, Coupled Assimilation (SOCA) team, allowing them to develop a number of useful and impressive capabilities. With this funding SOCA scratch-built a system now transitioning to operations which has been proven, through extensive testing by NOAA, to perform significantly better than the current operational system (figure 1). The overarching goal for this system is to achieve operational-quality ocean and sea ice data assimilation (DA) for GFSv17/GEFSv13, slated for code freeze by the end of 2024. The SOCA prototype uses marine hybrid 3DEnVar, which outperforms current operational analysis (figure 1).
SOCA’s work on the operational ocean and sea ice DA requirement for DRAS has benefited all of JEDI; for example, the explicit diffusion operator developed for SOCA’s prototype is now being used as an alternative to BUMP_NICAS by the JEDI Land DA and Atmospheric
Composition groups when short correlation lengths are needed. The ocean linear variable changes and balance operators from SOCA are also being moved upstream for use in other projects such as ROMS (Regional Ocean Modeling System) JEDI. Advances made in this prototype will also benefit the hurricane DA system used by NOAA, which has just been updated to run the MOM6 ocean model which SOCA is built around; the next version will most likely use SOCA for ocean initialization.
Another JCSDA team that was able to achieve significant progress with the DRAS funding is the Community Radiative Transfer Model (CRTM). The biggest accomplishment is the improved aerosol interface. Starting with one set of default aerosol optical assumptions at the beginning of the project, the CRTM team developed and implemented additional aerosol look-up tables, updated the aerosol optical data structure, improved data transparency with NetCDF support, and introduced a flexible user interface within JEDI. CRTM/JEDI now supports six aerosol look-up tables (both released and internal) that are utilized by five models, providing scientifically consistent aerosol radiative transfer simulations across the major data assimilation platforms (see figure 2).
Along with the aerosol tables, CRTM also developed methods to interface improved ultraviolet (UV) bi-directional reflection distribution function models via the Community Surface Emmisivity Model (CSEM), compute Geostationary Operational Environmental Satellite Advanced Baseline Imager (GOES ABI) visible reflectances and demonstrate that capability in SkyLab. CRTM was also able to adapt to interface with the Geostationary Interferometric Infrared Sounder (GIIRS), the first example of a Chinese instrument in CRTM and UFO.
Finally, under the DRAS grant the Observations team achieved three major accomplishments. Firstly, through adding visible reflectance assimilation and a ground-based radar operator, secondly by reaching operational-level volumes of observation data in Skylab, and thirdly through the implementation of the all-sky error model that has been demonstrated at NOAA.
Visible reflectance assimilation is a cutting-edge area for JCSDA, as it is among only a few teams globally attempting visible reflectance data assimilation. Geostationary satellites and various polar orbiters have a long history of using visible spectrum sensors, which allows us to leverage a wealth of both current and historical data.
Future work will move forward with using visible reflectance to aid in forecasting dust clouds, smoke, storms and stratus clouds. The ground-based radar operator was developed in partnership with University of Oklahoma’s Center for Analysis and Prediction of Storms (CAPS). This operator incorporates the MRMS radar reflectivity product, which uses combined data from many NEXRAD radars rather than single-site radars. This data will increase accuracy for high-resolution forecasts of mesoscale convective systems, and will aid in short-term extreme weather prediction by centers like NOAA’s National Severe Storms Laboratory (NSSL).
Starting with the release of SkyLab version 6.0 JCSDA SkyLab incorporated operational levels of data, with the stability of the system proving to be robust in a variety of applications. The NOAA microwave all-sky approach was integrated and tested, the correct hydrometeor treatment is now retained in the model interface, and the GMI and AMSR/2 sensors were integrated. Other Skylab work enabled by the DRAS grant includes progress towards performing observing system experiments showing the impact of an observation set, which have traditionally been very computationally demanding and expensive. Performing these experiments in Skylab offers excellent possibilities for completing impact assessments.
The DRAS grant has empowered numerous JCSDA teams to achieve significant advancements that benefit all our partners and, in some cases, push the boundaries of data assimilation techniques. The contributions made possible by this grant are invaluable.