NUMERICAL SIMULATIONS INITIALIZED WITH RADAR-DERIVED WINDS .2. FORECASTS OF 3 GUST-FRONT CASES

Numerical simulations of three gust-front cases that occurred in north eastern Colorado during the summers of 1991 and 1992 are presented. Th e simulations are initialized with radar-derived winds and, for the tw o cases in 1992, measurements from a surface mesonet. Thermodynamic re trieval is used to calculate the buoyancy in the boundary layer. The s ensitivity of the retrieved buoyancy to the various constraints of rea l data was examined in Part I of this study. In the first case, a larg e-scale gust front moved southward over the Denver region at a speed o f 8-9 m s-1. The retrieved buoyancy field for this case exhibits a bro ad baroclinic zone, with a width of approximately 20 km centered about the radar-detected fine line. This baroclinic zone collapses to a wid th of about 5 km as the numerical model is integrated forward. The sim ulated gust front propagates at 7 m s-1, which is slightly less than t he observed speed. For the second and third cases, data from a 50-stat ion surface mesonet were also available. In the second case, two gust fronts converged in the region of Mile High Radar but failed to genera te sipificant convection. In the third case, three gust fronts converg ed and generated strong convection. Numerical simulations for both of these cases using surface and radar-derived winds are presented. A ver ification analysis is performed on the forecasts of the two cases from 1992. Both surface observations and analyzed convergence fields are u sed to verify the forecasts. For the two cases, the numerical forecast of surface winds at 60 min improved over persistence by an average of 30%. The forecast surface convergence and temperature fields improved over persistence by an average of 25% and 28%, respectively.