Weather Research and Forecasting Core Tests

The current Weather and Research Forecasting (WRF) Software Framework (WSF) supports two dynamic solvers: the Advanced Research WRF (ARW) developed by the Mesoscale and Microscale Meteorology (MMM) Division of NCAR, and the Nonhydrostatic Mesoscale Model (NMM) developed by the National Centers for Environmental Prediction (NCEP). WRF also offers a variety of physics packages. Parallel runs of these two dynamic solvers are available from a variety of experiments (e.g., WRF Test Plan, Spring Program, DWFE, etc…), but the differences between these parallel runs can not be solely attributed to differences between the two dynamic cores because the dynamic solvers were not configured to use the same physical parameterizations and/or initial conditions. The WRF Developmental Testbed Center (DTC) and the Global Systems Division (GSD) of NOAA's Earth System Research Laboratory (ESRL) addressed the need for a controlled comparison of these two dynamic cores through intensive retrospective testing (i.e., parallel runs of the two dynamic solvers using initial and lateral boundary conditions based on the same input data, as well as the same suite of physics parameterizations). The goal of the WRF Core Test was to determine the impact of the dynamic solvers on the forecast. With this intent, parallel runs of WRF using the two dynamic solvers were configured such that differences between the forecasts were confined to those related to the dynamic solvers, at least within the limitations of the current end-to-end system. This testing focused on four 30-day retrospective time periods corresponding to the four seasons. Twice daily (00 and 12 UTC cycles) 24-h forecasts for these retrospective periods were generated on a 13-km grid for each dynamic core using the same initial and lateral boundary conditions and the same number of vertical levels. Each dynamic core was run with the same physics package configuration. Two physics package configurations were considered during this testing. This presentation will describe the design of the experiment and discuss the differences between the dynamic cores based on standard verification measures. This discussion will also include an assessment of the statistical significance of these differences.