The semi-arid climate and mountainous forested topography of the American West combine to make this one of the most vulnerable and difficult areas subject to forest fires in North America. Frequent summertime drought conditions and the remoteness of many areas often act to make the fire danger extreme and exceptionally difficult to combat. Add to this the uncertainty associated with rapidly changing weather conditions, and situations are created for fire fighters in which almost any blaze can behave unpredictably and suddenly become life threatening. The recent tragic deaths of four fire fighters in the northern Cascade Mts. of Washington on 11 July 2001 provides a sobering example. In such dangerous conditions, the availability of accurate up-to-the minute computer-generated "nowcasts" of local meteorology (wind, temperature, humidity, thunderstorms, etc.) in rugged mountainous areas can be of tremendous value. Unfortunately, most operational numerical weather prediction systems are run only twice per day and have grid resolutions that fail to represent accurately much of the complex topography of the western states. High-resolution research models have been developed in recent years, but generally these are restricted to limited-area fixed domains and require fairly extensive computer resources.

To address the need for timely high-resolution meteorological guidance, a versatile nowcast numerical modeling system has been developed at the Pennsylvania State University. The nowcast system is designed around an optimized full-physics version of the Penn State/NCAR mesoscale model, MM5. Using triply nested grids of 36-, 12- and 4-km (the latter domain covers 500 X 500 km), the model runs in real time on an inexpensive dual-processor PC computer and produces new meteorological nowcasts of current conditions every 30 minutes. The model is globally relocatable in less than 5 minutes and can be operated by a user with little or no meteorological or numerical training. Plugged into a real-time data stream, such as Unidata, the modeling system performs four-dimensional data assimilation (FDDA) to reduce numerical errors by blending observations into the solutions as the model runs. The result is a continuous stream of highly detailed nowcasts over complex terrain that provide timely guidance about meteorological conditions over remote areas as they develop. Further work is expected to allow assimilation of satellite-derived winds in real time, post-processing to reduce remaining errors in the model-generated products, and coupling to a plume-dispersion model for tracking smoke plumes and other airborne materials.