H111-0010
Use of Polymorphic Model Realizations to Allow Dynamic Reconfiguration of Continental Scale Modeling

Friday, 11 December 2020
Poster
Robert Bartel1, Donald W Johnson2, Nels J J Frazier1, Fred Ogden3, Christopher Tubbs4, Shengting Cui1, David Mattern5 and David L Blodgett6, (1)NOAA Affiliate, ERT Inc, Office of Water Prediction, National Water Center, Tuscaloosa, AL, United States, (2)NOAA Affiliate, Tech Global, Office of Water Prediction, National Water Center, Analysis and Prediction Division, Tuscaloosa, AL, United States, (3)NOAA/NWS Office of Water Prediction, National Water Center, Tuscaloosa, AL, United States, (4)NOAA Affiliate, ERT Inc, Office of Water Prediction, National Water Center, Laurel, United States, (5)NOAA Affiliate, Lynker Technologies, Office of Water Prediction, Tuscaloosa, AL, United States, (6)USGS, Baltimore, MD, United States
Abstract:
The NextGen National Water Model framework approaches the problem of continental scale modeling using polymorphic model realizations. The framework defines a standard interface for model realization classes, where the interface prescribes that individual realization objects will each represent physical processes for a particular geographic area. This characteristic of the interface allows the continental scale to be broken down into more distinct subproblems. However, the interface does not dictate the specifics for how physical processes are represented. This allows multiple realization classes to be implemented, which are easily substitutable for one another, but where each internally represents physical processes in a distinct manner. Using this design simplifies the process for adding new modeling behavior, as only an understanding of the model realization interface is needed, as opposed to a more thorough understanding of the complete internals of the framework. The design also makes it possible to create realizations that conform to external APIs, such as the Community Surface Dynamics Modeling System Basic Model Interface. Having realization implementations that comply with such standards provides an easy means for plugging in existing or new compatible models. The flexibility provided by the design leads to opportunities for both improved computational efficiency and more accurate results, by facilitating use of either a more finely tuned or more general and efficient model realization for a particular subproblem as appropriate.