Physics Test Harness

Physics Testing Harness Hierarchy
Conceptual illustration of the simple-to-complex hierarchical structure of the physics test harness. Colors denote potential responsibility for carrying out each type of testing. Physics developers are expected to be involved throughout the process and conduct less computationally expensive testing on their own machines. DTC staff can support and carry out testing up to and including low and medium resolution (LR/MR) global forecasts with data assimilation. Operational centers are expected to contribute support and computational resources if complex testing is warranted.

The GMTB physics testing harness provides a common infrastructure for testing physics parameterization developments for potential applicability for NWP. It consists of several “tiers” of modeling configurations arranged in a simple-to-complex hierarchy. Testing a physical parameterization innovation within such a structure allows for the sequential addition of model feedbacks that add layers of complexity to the interpretation of results as testing proceeds toward an operational configuration. Priority development of the physics testing harness has been focused on the single column model and low/medium resolution global reforecast tiers (using cold starts and cycling with data assimilation).

Single Column Model

  • Represents the simplest way to test a physics suite as an entity while eliminating bias from a dynamical core. An atmospheric column is “forced” by observed or modeled large-scale meteorological conditions within individual case studies.

  • The current version uses NOAA’s Interoperable Physics Driver (IPDv3) to connect to the 2017 operational GFS physics.

  • Documentation for obtaining, running, and analyzing output from the GMTB SCM is available. Data to run several test cases is included and users are encouraged to develop and run new case studies.

Global Forecast/Reforecast

  • A workflow for conducting global NWP runs that expands upon NOAA/EMC’s operational configuration has been implemented and used within the DTC and is available for interested collaborators.

  • The latest workflow for pre-processing, running the GSM-based GFS, and post-processing from NOAA/EMC has been combined with a python graphics package and verification from DTC’s MET software. It uses NOAA’s Theia HPC platform.

  • Testing within the DTC has utilized the workflow for both cold start and cycled runs with data assimilation.

Future Development

The tools within the physics test harness will undergo substantial changes to keep up with NOAA’s shift to the FV3 dycore and continued changes to physics. The effort required for introducing new physics schemes for testing will be considerably reduced as the tools transition to using a new version of the Interoperable Physics Driver and the Common Community Physics Package ecosystem.