Mesoscale models have become a valuable tool in the study of a wide variety of weather phenomena including extratropical cyclones (e.g., Kuo and Reed 1988; Kuo et al. 1992; Reed et al. 1993; Steenburgh and Mass 1996), mesoscale convective systems (e.g., Anthes et al. 1982; Zhang and Fritsch 1986, 1988; Tripoli and Cotton 1989), and thermally-driven circulations such as slope/valley flows (e.g., Bossert and Cotton 1994a,b). Increasingly, real-time versions of these models, as well as high resolution versions of operational models such as the 29-km and 10-km Eta, are being used for mesoscale prediction. Although it has been speculated that fixed surface forcing would enhance mesoscale predictability in regions of complex terrain (Paegle et al. 1990), considerable debate remains concerning the value added by increased resolution in such regions. Real-time simulations run over the Pacific Northwest have shown a noticeable improvement in skill scores as resolution was increased from 36 to 12 km (Colle et al. 1999), but additional resolution has shown less of an impact to date (C. Mass, personal communication). A comparison of model forecasts from the 48-km, 29-km, and 10-km versions of the Eta model over the western United States revealed that higher-resolution forecasts showed improved skill only for heavy-precipitation events (McDonald 1998). However, such improvements in traditional skill scores, such as root-mean-squared error and the equitable threat score, were accompanied by a high false-alarm rate that would greatly reduce the operational utility of such simulations.

Although initial condition uncertainty is a major source of error in real-time simulations, studies suggest that more accurate numerical representation of dynamical and microphysical processes in complex terrain will yield forecast improvements. For example, Gaudet and Cotton (1998) showed significant improvement in precipitation skill with the use of a bulk microphysics parameterization instead of a simple "dump-bucket" scheme in a real-time version of RAMS used over Colorado. This study also showed, however, that the bulk microphysical parameterization produces excessive precipitation at low-elevation locations and too little precipitation at higher elevation locations. Systematic bias errors have also been found in real-time simulations by the MM5, which have produced too much precipitation on the windward slopes of the Cascades and too little to the lee (Colle et al. 1999). Excessive windward precipitation has also been found in the 10 km Eta (Colle et al. 1999). These biases may be related to errors in the specifications of ice crystal fall speed, cloud droplet spectra, and other cloud characteristics. For example, simulations of precipitation to the lee of the Cascade Mountains have shown considerable sensitivity to the specification of ice particle fall speed (Colle and Mass 1999).

Datasets derived from IPEX will be used to validate and improve the simulation of dynamical and microphysical precipitation processes over the narrow, steeply sloped Wasatch Mountains. The large concentration of surface and upper-level observations will allow for validation of mesoscale model (MM5) wind and temperature fields, including the flow field near the mountains and the structure of gravity waves excited by the barrier. Bulk microphysics parameterizations will also be examined, including the accuracy of model-predicted cloud liquid water concentrations and specification of cloud droplet spectra and ice particle size distribution. We propose that some of the errors in model QPFs are due to an inadequate treatment of these cloud characteristics, as well as the treatment of aggregation and specification of ice-particle fall speeds. IPEX datasets and a snow growth model (Mitchell 1988; Mitchell et al. 1996) will be used to obtain better estimates of these characteristics and improve their representation in mesoscale models.

Improvements in short-range (< 12 h) mesoscale prediction over complex terrain will also require significant advances in data assimilation. Current operational or quasi-operational data assimilation systems such as the Rapid Update Cycle (40 km resolution) and Local Analysis and Prediction System (10 km resolution) struggle with issues related to the data assimilation over the western United States and there have been few attempts to perform data assimilation on scales that are adequate to resolve atmospheric processes in complex terrain. Under the direction of Prof. John Horel, the Advanced Regional Prediction System (ARPS) Data Assimilation System (ADAS) is being used to generate high-resolution analyses over Utah. Data is assimilated hourly at a 1-km resolution over a 220 x 220 km region that encompasses the IPEX target areas. Data from IPEX will make it possible to assess the strengths and weaknesses of this and other analysis systems. ADAS analyses with and without the supplemental observations will help assess the sensitivity of the analyses to data voids and help design optimal strategies for observing systems in complex terrain. Short-range forecasts utilizing such data-assimilation systems will also be tested.