Abstract
A new model for making probability forecasts of accumulated spot precipitation from weather radar data is presented. The model selects a source region upwind of the forecast spot. All pixels (horizontal size 2 × 2 km2) within the source region are considered, having the same probability of hitting the forecast-spot. A pixel hitting the forecast spot is supposed to precipitate there a short time (about 10 min.). A drawing is performed, and a frequency distribution of accumulated precipitation during the first time step of the forecast is obtained. A second drawing gives the frequency distribution of accumulated precipitation during the first to second time step, a third one during the first to third, and so on until the end of the forecast period is reached. A number of forecasts for 1-h accumulated precipitation, with lead times of 0, 1, and 2 h, have been performed and verified. The forecasts for 0-h lead time got the highest Brier skill scores, +50% to 60% relative to climatological forecasts for accumulated precipitation below 1 mm.
