The quality of Numerical Weather Predictions (NWP) critically depends on the choice of initial conditions. In the US alone, a number of major Data Assimilation (DA) systems have been developed to provide estimates of the state of the atmosphere to initialize NWP forecasts. While these schemes contain state-of-the-art algorithms (i.e., the scientific aspect of DA), their computer representation (i.e., software engineering aspect of DA) is based on less advanced, procedure-oriented software design. Each DA system resides in separate software repositories with no chance for software-level interactions between the various systems. Arguably, this poses a major impediment to collaborative research and the efficient transition of research into operations. To remedy this situation, we suggest the community adopt the use of an Object-Oriented Design (OOD) approach to DA software development. Based on an abstract-level analysis of the common elements and processes (i.e., “objects”) in current and expected DA schemes and with 2 the introduction of OOD design elements, the various DA schemes can be reformulated to become part of a common, Community Data Assimilation Repository (CDAR). Different strategies (e.g., variational vs. ensemble filter) and techniques (e.g., observational data en- and decoding, quality control) become interchangeable across the currently disjoint DA systems, greatly expanding the technology available for both the research and user communities. Operational centers can subselect objects from CDAR to assemble and optimize the performance of each of their applications, respectively, accelerating the research to operations transition and overall NWP improvements.
Authors who have authored or contributed to this publication.