focus on forecasting1

Abstract

A forecast experiment was conducted from 1 June to 12 August 1983 at the Program for Regional Observing and Forecasting Services (PROFS) in Boulder, Colorado. The exercise included probabilistic forecasts of severe-weather convective outlooks prepared with the use of existing technology and severe-storm warnings made with the use of advanced technology to display a wide variety of data sets on the PROFS workstation. The convective outlooks were evaluated using signal-detection-theory analyses and Brier scores. Considerable skill using signal-detection theory was found for both severe-thunderstorm and tornado outlooks, treated as separate outlooks. On the other hand, Brier skill scores relative to sample climatology show small skill for the thunderstorm outlooks and no skill for the tornado outlooks. The signal-detection analyses provide valuable baseline values for future evaluation studies.Skill scores for the warnings issued at PROFS are compared with a similar set of skill scores for Warnings from the Denver WSFO (Weather Service Forecast Office). Considerably higher skill is found in PROFS severe-thunderstorm-warning scores. This result is attributed to advanced technology and the controlled environment in the PROFS forecast office. For tornado warnings, the probability of detection was slightly higher and the false-alarm rate was much lower at the Denver WSFO than at PROFS. The high false-alarm rate at PROFS is ascribed to the lack of familiarity with the new Doppler-radar data, the overreaction to those data, and the incompleteness of the radar data set. An analysis of area-and time-weighted skill scores is also presented to illustrate the concept that frequently issued time-and site-specific severe-storm warnings have a marked potential for improved service to the public.forecasts of severe weather (MacDonald, 1984). The summer forecast exercise included the following objectives: 1) to test a new version of the forecasting workstation, upgraded since the 1982 system reported by Reynolds (1983); 2) to obtain expert opinion on the scientific and technological obstacles to further progress in convective forecasting; 3) to develop procedures for evaluating progress in nowcasting and very-short-range forecasting of convection as these terms are defined by Zipser (1985). Three types of forecasts were prepared daily: 1) a convective-weather outlook for probabilities of tornadoes and severe thunderstorms; 2) warnings for severe thunderstorms, tornadoes, and flash floods as necessary; and 3) scheduled forecasts of significant weather probabilities for precipitation, wind, and visibility. Williams and McCoy (1984) have presented a preliminary report on evaluation of the probability forecasts. Schlatter et al. (1985) discuss topics related to the first two objectives of the 1983 exercise listed above.This paper presents an evaluation of the daily severeweather convective outlook forecasts and the severe-weather warnings (nowcasts to short-term forecasts) prepared during the experiment.