During the development phase of the functional prototype Maintenance Decision Support System (MDSS), an ensemble of mesoscale models has been utilized, in two distinctly different configurations, to provide meteorological state variable and precipitation forecast quantities into the MDSS Road Weather Forecast System (RWFS). In the original ensemble configuration, pairs of forecasts from recent versions of the MM5, RAMS and WRF models were utilized on yy-km resolution grids covering the state of Iowa. The most recent version of the ensemble, utilizing the same grid, utilized lagged forecasts (over a 3 hour period) from MM5 and WRF plus forecasts from NCEP’s Eta and GFS model and output from the Eta and GFS Model Output Statistics (MOS) packages.

Although in theory utilization of an ensemble should be superior to a deterministic forecast at the same resolution, there is still considerable debate in the numerical weather prediction community as to the value of an ensemble versus a deterministic run on a higher resolution grid, particularly when computational cost/benefit analyses are taken into account. Further, given the limited geographic area used for MDSS validation, more work is needed to evaluate the efficacy of an ensemble versus a higher resolution deterministic run. Also the question of whether a physics-based ensemble is of value in this context has not yet been addressed.

Thus, in this paper we will present early results of an ongoing effort at UND related to development of an appropriate ensemble model system for use as input to a road weather forecast system. Tests of multiple configurations of forecast ensembles are performed ranging from 4 to 16 members, all configured on 10 km resolution grids, against 3 km deterministic forecasts, on up to three focus areas: 1) the Iowa functional prototype MDSS demonstration region; 2) the Pooled Fund Study (PFS) 2004-05 test area centered on Dickinson, ND; and 3) the PFS 2004-05 test area covering the Colorado Front Range from metro Denver to the Palmer Divide. In the presentation we will focus our attention on forecasts of precipitation (type and accumulated amounts), visibility (derived from an algorithm based on Bocchieri (1980)), and near-surface air temperature, as these quantities are all critical elements within a road weather forecast system. At the conference we plan to illustrate our findings with selected case studies from the three regions as well as aggregate verification statistics.