We use the Advanced Regional Prediction System (ARPS), a mesoscale numerical weather prediction system developed by the Center for Analysis and Prediction of Storms (CAPS) at the University of Oklahoma, in a variety of modes to provide forecasts to 48 hours to customers in a broad range of industries, including broadcast meteorology, energy utilities, commercial and general aviation, and emergency management. We perform nationwide ARPS runs at 32 km four times each day. Using the ARPS Data Analysis System (ADAS), a variety of data are used to initialize the forecasts: METAR, mesonet, satellite, rawinsonde, wind profiler, and aviation reports (MDCRS). Custom runs on finer grids (e.g., 6 km) provide individual customers with highly localized, storm-scale forecasts. NEXRAD radar data are also objectively analyzed on these grids to provide appropriate initial conditions.

Results are distributed to customers in digital or graphical form, the latter via a web interface. Aviation variables such as turbulence and icing potential are produced. Model research and development, performed in collaboration with CAPS, is ongoing. Results from model verification studies will be discussed.

We discuss the benefits and challenges of computing on a distributed-memory, massively-parallel computing system. A workstation version, designed initially for broadcast meteorologists, provides customers the ability to run the model on their own computing system. Fields from the nationwide ARPS run and raw input data, including radar data, are sent on demand to the workstation, which then performs the objective analysis and runs the model. All aspects of the forecast can be configured using a web interface.