Rating the winter weather forecast

Authors


Abstract

For the second year in a row, the National Weather Service (NWS) winter forecast fell disappointingly short of the mark because of an unforseen atmospheric event that shaped an entire season of weather. This winter it was a mass of bitter cold arctic air that caused December's record-breaking low temperatures and established the character of January and February weather patterns, according to Donald L. Gilman, chief of the NWS longrange predictions branch. As a result, says Gilman, the weather prediction issued in November (Eos, December 20, 1983, p. 985) was “not a good forecast.” Once the model makers started seeing the cold temperatures in December, however, they were able to update their figures for a new 3-month forecast extending through March that turned out to be quite accurate.

NWS issues a 3-month forecast each month. The most publicized forecast, according to Gilman, is the November one that covers the entire winter season. At the end of each prediction period, the forecast is given a numerical accuracy score. A rating of zero would mean that the forecast was no better than what might be predicted from general climatological data. Above zero, though, means that the forecast is “skilled,” and a score of 100 would mean 100% accuracy. (Scores generally range between +10 and −15.) Using this method of figuring probability scores, the November forecast earned a low score of −5, ranking it 13th among the 18 predictions issued by NWS since these forecasts began a couple of years ago. Last winter's forecast, sabotaged by a surprisingly strong El Niño, ranked 14th.