A Collaborative Approach for Algorithm Operationalization

Effective Research to Operations (RTO) for environmental science algorithms requires access to operational algorithms, near-real-time data and operationally representative test environments. Traditional approaches to providing these capabilities come with a high entry cost that limits engagement by many environmental sciences departments. This paper will demonstrate a new approach based on formal UML modeling best practices, open standards and an adaptable algorithm development framework which provide an open, adaptable environment for the operationalization of environmental research.  This approach has a low cost of entry and yet provides the discipline needed for developing and testing of high quality research algorithms that are ready for rapid and reliable transition to operations.

The Algorithm Workbench (AWB) is a flexible algorithm development, test and production built on three key concepts: (1) a component-based algorithm model; (2) standard interface between the algorithms and the underlying processing infrastructure (the data model interface of DMI); and (3) an extensible plug-in architecture that facilitates addition of new data types, formats and other interface.  The AWB is presented here not as a point solution, but as a step towards a standards-based approach to environmental science algorithm development and sustainment.

Joint Polar Satellite System (JPSS) produces dozens of meteorological data products useful to weather and climatology scientists. The processing chain to result in these products is complex, with interdependency among the products themselves as well as with databases external to the JPSS system. Data flow management through this system presents a challenge with multiple sources and users of various data products. On the JPSS program, UML models of the algorithm production processing were developed and maintained with the MagicDraw UML modeling environment.  The JPSS Program uses the model to manage responsibilities for products, to create and manage requirements specifications, and to scope downstream impacts to changes in a particular product.

In this paper were present the use of MagicDraw with the AWB.  The component-based algorithm design inherent in AWB maintains all necessary algorithm properties and interrelationships in meta-data.  We show how this meta-data can be employed to automatically create UML descriptions of the as-built system.  This is a critical step towards automated documentation and round-trip engineering to manage responsibilities for products, to create and manage requirements specifications, and to scope downstream impacts to changes in a particular product. With the upcoming transition to an Enterprise science processing paradigm, the combined use of AWB & MagicDraw provides an ability to holistically manage the data processing flow across NOAA's missions.