#C01 ACME Priority Metrics: A-PRIME
Poster Title | ACME Priority Metrics: A-PRIME |
|---|---|
Authors | @Kate Evans (Unlicensed) @Milena Veneziani @Salil Mahajan @Xylar Asay-Davis @Marcia Branstetter (Unlicensed) @Phillip Wolfram (Unlicensed) @Mark Petersen @Susannah Burrows @Charlie Zender @Luke Van Roekel (Unlicensed) (not in a particular order right now) |
Group | Coupled |
Experiment | Watercycle |
Poster Category | Early Result |
Submission Type | Either |
Poster Link |
Abstract
Developers of the ACME model must quickly determine if large-scale, coupled configurations of ACME are providing good results across global and local areas of interest. This allows developers to identify issues and attributes before major computer and personnel time is spent. The model’s complexity requires that many eyes of developers covering a wide range of scientific expertise be able to view and assess model quality quickly. The python-based ACME Priority Metrics (A-PRIME) software has been developed to address these requirements. It is designed for mostly autonomous execution, targeting the leadership computing centers and Livermore National Lab’s ACME1 clusters. Although A-PRIME provides only a wide view of model performance, it has been designed as a platform to hook up more substantial software that performs more robust analysis of specific model features. An example of this modular expansion is the ability to include analysis of atmospheric dynamics and extremes events, which is a scientific topic of interest by the software developers.