The DTC is performing a controlled comparison of forecasts generated with the Advanced Research WRF (ARW) and Non-hydrostatic Mesoscale Model (NMM) dynamic cores of the Weather Research and Forecasting (WRF) model.
The motivation for this comparison is to determine whether the medium range skill of the two cores is similar. If no significant difference is found between the forecasts from the two dynamic cores, research results obtained with one core can be directly transferred to the other.
Additionally, this test quantifies the differences in time-average forecast verification statistics from identical end-to-end forecast systems run on two different computational platforms: IBM and LINUX. Determining whether the test results are independent of platform is an important step toward qualifying the DTC’s system of distributed, multi-platform, computer resources.
Methodology
The ARW and NMM were configured to run with 13-km horizontal grid spacing over the continental United States. Both cores used 58 vertical levels, with model top at 50 hPa, and employed identical physical parameterizations. Thirty cycles (15 initialized at 00 UTC and 15 at 12 UTC) in each of four retrospective seasons were initialized from the Eta model through the WRF Preprocessing System and run out to 60 hours.
Forecasts were run through the WRF Postprocessor to remap the forecasts to a common grid. Verification statistics were computed with the National Centers for Environmental Prediction Verification System, which compares the forecast against precipitation estimates, and surface and upper air observations.
Results
The forecast images are available. The forecast verification statistics are currently being analyzed by the DTC. |