Abstract
An experiment was conducted in Boulder, Colorado in the summer of 1983 to forecast probability of precipitation over a 70 000 square km area for periods up to 8 hours beyond forecast time. The forecasts were issued from an advanced workstation that included high-resolution computer displays of Doppler radar and satellite images and numerous graphics products. The forecasts were compared with probabilities given by climatology, conditional climatology, and persistence. Brier scores and skill scores were used to evaluate the forecasts. Results show that positive forecast skill was achieved for most of the forecast periods. However, the forecasters normally tended to overforecast precipitation from rare events. It is speculated that future forecast experiments will show improved skill as more feedback and experience are gained.
