The enhanced Model Evaluation Tools (METplus) development team at the DTC is pleased to announce the release of the multi-component verification capability called METplus. It includes METplus version 3.1, MET version 9.1 and METviewer version 3.1. METplus contains a suite of Python wrappers and ancillary scripts to enhance the user's ability to quickly set-up and run MET. METviewer is a database and display system for aggregating and plotting MET output. Users can now check everything out and build through the METplus website.
METPLUS 3.1 and components may be obtained via the METplus Downloads Page.
- Major Enhancements: Enable METplus to only process certain months of a year for seasonal verification, added curl possibility to build_components build MET script, added many more examples of using Python Embedding (e.g. for Feature Relative evaluation and use of StatAnalysis to evaluation point forecasts), GenVxMask, and computing Surrogate Severe and Practically Perfect Prognosis fields.
- Current supported use cases and examples (new are bolded):
- Simple examples of running met_tool_wrappers: ASCII2NC, CyclonePlotter, EnsembleStat, GenVxMask, GridDiag, GridStat, MODE, MTD, PB2NC, PCPCombine, Point2Grid, PointStat, RegridDataPlane, SeriesAnalysis, StatAnalysis, TCGen, TCMPRPlotter, TCRMW, TCPairs, TCStat
- More complex examples relevant to model_applications (new are bolded): Climate, Convection Allowing Models, Ensemble, Marine and Cryosphere, Medium Range, Precipitation, Space Weather, Subseasonal to Seasonal, Tropical Cyclone and Extra-Tropical Cyclone
- Full details can be found in the release notes.
- METplus 3.0 Online Tutorial covers much of the functionality of this release, including examples of how to run the components individually.
MET 9.1 is also available via the METplus Downloads Page and as a Docker Container via DockerHub.
- Major Enhancements: Documentation: Transitioned to the Web. Statistics: complement version of the Ranked Probability Score (RPS_COMP), the uncentered Anomaly Correlation (ANOM_CORR_UNCNTR). Improved: Grid-Diag, Point2Grid, and TC-Gen tools, Fortify: Compliance with zero high risk findings.
- Documentation: may be found at https://ncar.github.io/MET/
- Full details can be found in the release notes.
- For MET 9.1, please refer to the METplus 3.0 Online Tutorial covers much of the functionality of this release, including examples of how to run MET.
METVIEWER 3.1 is also available via the METplus Downloads Page and as a Docker Container via GitHub.
- Major Enhancements: Support for MET 9.1 output and an optional alpha-release of Python support for aggregation, event equalization, pairwise differencing, and computing confidence intervals. Two plot templates, Performance and ROC diagrams, are also available in Python. The use of Python can be toggled through the “Common” tab in the lower right portion of the user interface.
- Current Plot Templates: Time-Series, Bar, Box, ROC, Reliability, Spread-Skill, Performance and Taylor Diagrams, Rank, PIT, and Relative Position Diagrams, Economic Cost-Loss Diagram, Contour Diagram.
- Full details can be found in the release notes.
- For METviewer 3.1, please refer to the METplus 3.0 Online Tutorial covers much of the functionality of this release, including examples of how to run METviewer.
SUPPORT FOR US-BASED HPCS
Over the next week, METplus 3.1 and MET 9.1 will be installed on several US-based high performance computing platforms (e.g. Cheyenne, Jet, Hera, WCOSS, and Stampede) to make it more accessible to both the research and operations community. Please see the list of existing builds. METviewer is intended to be installed on your local server.
