Prototype real-time mesoscale prediction during the 1991-92 winter season and statistical verification of model data

Abstract

With the advent of faster, more powerful computers has come an opportunity to perform real-time mesoscale numerical weather prediction. The Regional Atmospheric Modeling System (RAMS) developed at CSU has been modified to become a prototype real-time forecast model. Surprisingly, a substantially complex model configuration can attain real­ time forecasts on CSU workstations, as well as a CRAY supercomputer, although more sacrifices must be made on local machines in order to maintain the real-time restrictions. Real-time forecasting began in November 1991 with two main objectives at the time. The first was to predict orographically-forced precipitation in the Colorado region. In this effort the RAMS model was run throughout the winter season of 1991-92 and continues again for the winter season of 1992-93. Through investigation of a case study day (9 March 1992) in which a major winter storm produced blizzard conditions along the Colorado Front Range, the true potential of RAMS real-time forecasting is demonstrated. The second goal was to provide real-time forecasts of cirrus-level clouds to investigators involved in the FIRE II field program in Kansas from 13 Nov. to 6 Dec. 1991. Again a case study day is analyzed in order to assess the possibility that RAMS could improve local mesoscale forecast capabilities. Lastly, a comprehensive statistical analysis of all model data is discussed. The analysis uses multivariate randomized block permutation methods (MRBP) to illustrate model forecast skill for a variety of categories. These categories are comprised of case study model verification and entire season verification for forecasts of lengths 12, 24, 36, 48 hours as well as a brief comparison with established numerical models.