An Evaluation of the MM5, RAMS, and Meso-Eta Models at Subkilometer Resolution Using VTMX Field Campaign Data in the Salt Lake Valley

This study presents what is, to the authors' knowledge, the first intercomparison and evaluation of three state-of-the-art mesoscale numerical models, the fifth-generation Pennsylvania State University–NCAR Mesoscale Model (MMS), the Regional Atmospheric Modeling System (RAMS), and the NCEP Meso-Eta, at horizontal resolution finer than 1 km. Simulations were carried out for both weak and strong synoptic forcing cases during the Vertical Transport and Mixing (VTMX) field campaign conducted in the Salt Lake valley in October of 2000. Both upper-air and surface observations at high spatial and temporal resolution were used to evaluate the simulations with a focus on boundary layer structures and thermally driven circulations that developed in the valley. Despite differences in the coordinate systems, numerical algorithms, and physical parameterizations used by the three models, the types of forecast errors were surprisingly similar. The common errors in predicted valley temperature structure include a cold bias extending from the surface to the top of the valley atmosphere, lower than observed mixed-layer depths when the observed mixed layers were relatively high, and much weaker nocturnal inversion strengths over the valley floor. Relatively large wind forecast errors existed at times in the midvalley atmosphere even in the case of strong synoptic winds. The development of valley, slope, and canyon flows and their diurnal reversals under weak synoptic forcing were captured better by RAMS and MM5 than by Meso-Eta. Meso-Eta consistently underpredicted the strengths of these terrain-induced circulations and the associated convergence and divergence over the valley floor. As operational mesoscale modeling moves toward subkilometer resolution in the near future, more detailed forecasts of the circulation patterns and boundary layer structure can be produced for local-scale applications. However, this study shows that relatively large forecast errors can still exist even with sufficiently fine spatial resolution, indicating that the future for accurate local forecasting still lies in improved model parameterization of longwave radiation and turbulent mixing.