Adaptation of METplus Wrappers at GSD for Ensemble Verification

Verification continues to be essential to the development of Global Systems Division (GSD) models, particularly the experimental High Resolution Rapid Refresh Ensemble (HRRRE). Model developers require flexible verification in order to analyze aspects of ensemble performance beyond typical deterministic verification techniques applied to the ensemble mean and/or individual members. To that end, the GSD Assimilation Development Branch (ADB) developed an ensemble verification system that incorporates aspects of the Developmental Testbed Center’s (DTC) METplus package. Originally using in-house developed scripts and later transitioning to the officially supported METplus Python wrappers, this system runs in real-time and generates performance metrics based on MET EnsembleStat and GridStat output for experimental and operational ensemble systems. These metrics are ingested into a MySQL database, which was originally designed as the backend for DTC’s METviewer web application for statistical visualization. However, the METviewer GUI itself has a steep learning curve, due to the wide scope of its functionality, and can impede developers who wish to create performance plots efficiently. As a result, ADB is developing an Ensemble web application that combines the web application framework of the ADB Model Assessment Tool Suite (MATS) and the METviewer MySQL database with the high-level goal of reproducing the common plot types utilized in ensemble verification (ROC, reliability, etc.)in a more manageable user interface for model developers. In the past year, this work has been incorporated into the Unified Forecast System’s (UFS) METexpress project, a collaborative effort between GSD and DTC’s METplus team.

In this presentation, we will highlight the development of the ADB ensemble verification system, including requirements, methodology, metrics, and infrastructure. Additionally, we will discuss the ongoing development of the METexpress Ensemble application.