Numerical Weather Prediction in the Next Decade - Convective Forecasts with a Global Atmospheric Model?

Bill Skamarock
10:30 a.m. August 21 in Room 2155

Convection permitting NWP uses regional models and short forecasts periods because of computational constraints and the inherent limitations of downscaling global forecasts. Using global models that permit local refinement, existing computers are capable of producing CONUS-scale explicit convective forecasts over intermediate-range periods within operational time windows. The Model for Prediction Across Scales (MPAS) uses an unstructured spherical centroidal Voronoi mesh that allows for smooth cell-size variations between coarse and fine resolution regions on the mesh, and we been using it to produce experimental convective-scale global model forecasts. We will present MPAS forecast results from the May 2015 NOAA Hazardous Weather Testbed forecast experiment, and MPAS forecasts supporting the Plains Elevated Convection at Night (PECAN) experiment, that demonstrate these MPAS capabilities. Two issues are addressed with the MPAS configuration in these forecast experiments. First, forecast results show that the problems inherent in traditional grid-nesting approaches are not evident in the MPAS variable-mesh forecasts. Second, we are using the scale-aware version of the Grell-Freitas convective parameterization, and we find that it transitions appropriately from parameterizing essential all the unstable deep convection at hydrostatic-scale mesh spacings to allowing the convection to be explicitly simulated at nonhydrostatic mesh spacings on the variable resolution mesh. We will discuss our experience with these issues within the context of these and other forecast tests.