The concept of a Nation unified data assimilation system differs from a framework such as NEMS and ESMF, which provide tools to build, interface, and articulate software components. These tools reside at an abstract level and can therefore be considered as a set of building blocks, which are used in turn to design an endtoend system. JEDI can leverage on ESMF capabilities and provide unique features for data assimilation such as unified observation forward operators, background error covariance modeling, etc. A common repository may be required and it would hold centralized contributions from multiple origins. Furthermore, the unified system will organize various components together; ensuring they are standardized and avoiding duplicated work. It is important to note that a unified system does not mean a single configuration to be imposed. Although the choice of architecture and level of modularity will restrict the field of all possible options for data assimilation, large flexibility is envisaged to allow fundamentally different configurations to be actualized within the unified system.
The core aspect will be to modularize, standardize and mutualize data assimilation code across the partners of the Joint Center for Satellite Data Assimilation (JCSDA). These partners are NASA/GMAO, NOAA/NWS, NOAA/NESDIS, NOAA/OAR, Naval Research Laboratory, Oceanographers of the Navy, and US Air Force. The expected benefit is the increased ability to directly share and reuse software developed or tested by the community of partners. This will result in reduced overlap in development work, facilitate transition from research to operational implementation, which is expected to optimize and accelerate the assimilation of observations and in particular satellite data. Spearheading such an ambitious collaborative project is consistent with the mission of the JCSDA, uniquely situated at the crossroad of key organizations.
By enlarging the circle of partners to the academic community and the scientific research community in general, the transition rate from research to operations (R2O) is expected to increase considerably. This requires developing a flexible environment to quickly implement, evaluate and compare scientific algorithms. The “Blueprints for NextGeneration Data Assimilation Systems” workshop convened jointly by the NCAR DA Program and the JCSDA in March 2016 was a first step in collectively addressing this issue.
Gradually elevating existing software to modern coding standards and increasing its modularity will simplify the work toward code performance, maintenance and scalability on multiple and complex machine architectures. Concurrently scientists will be able to focus their attention on more compartmented regions of the data assimilation code, which will result in increased scientific productivity.
The drive toward Earth System Modeling is associated to either weakly or strongly coupled systems. Several programs such as the Next Generation Global Prediction System (NGGPS) and Earth System Prediction Capability (ESPC) are investigating and developing coupled DA, which on top of major scientific uncertainties represents a significant technical challenge. JEDI is expected to work as an enabler for scientific breakthroughs such as coupled DA and DA across spatiotemporal scales.