On the evening of 28 July 1997, more than 25 cm (10 in) of precipitation fell over southwest Fort Collins producing severe flooding and five fatalities. The Regional Atmospheric Modeling System (RAMS) Ver. 3b is used to simulate this event. The goal of this research is to identify the mesoscale preconditioning processes for simulated extreme precipitation in addition to the sensitivity of these mechanisms to variations in initial conditions. The simulations utilize four telescopically nested grids allowing for resolution of synoptic-, meso-, and convective scale motions in the respective domains. The initial atmospheric fields are supplied from the Rapid Update Cycle (RUC) operational forecast model analysis, sounding data and surface observations corresponding to 12Z 28 July 1997. Two simulations are performed, differing only in the method of soil moisture initialization. Simulation A is initialized with the soil moisture fields taken from the operational 40-km ETA forecast model analysis, while Simulation B utilizes the Antecedent Precipitation Index (API) method of soil moisture estimation.

The synoptic scale forecasts show that Simulation B produces a deeper and more intense topographic circulation converging on the Continental Divide. An accompanying increase and eastward shift in vertical mass transport relative to the Continental Divide, as well as increased Grid-2 precipitation volume are produced in Simulation B. Despite weaker synoptic-scale low-level convergence at the Continental Divide, Simulation A generates a 26-cm (10.2 in.) accumulated precipitation maximum 23 km to the southeast of Fort Collins in reasonable agreement with observations. Simulation B produces maxima of only 8-9 cm. The differences in accumulated precipitation occur despite comparable precipitation rates between simulations. Storm motion is identified as the primary contributing factor to the differences in accumulated precipitation and the influence of storm-induced cold-pools on storm propagation is investigated. While the cold-pool temperature profiles are similar between simulations, a deeper and warmer boundary layer environment in Simulation B results in a significantly deeper cold-pool with a larger negative perturbation temperature. This results in an increased cold-pool and storm propagation speed, with an attendant decrease in point precipitation magnitudes. During the quasi-stationary phase of the flood-producing storm in Simulation A, rainwater is lofted above the freezing level within the updraft, followed by freezing and accretional growth. Subsequent descent with complete melting occurs within the primary downdraft. This is consistent with radar-inferred microphysical properties of the Fort Collins storm noted by other researchers. No hail accumulation occurs at the surface, in agreement with observations.