Abstract
Late evening weather forecasts by telecasters at major network television stations in seven United States cities, and corresponding forecasts from the National Weather Service, were monitored by meteorology students for 6 months in 1985–86. These forecasts were of temperature and precipitation for the three periods of “tonight,” “tomorrow,” and “tomorrow night.” The accuracy of temperature forecasts was evaluated with three indexes: mean absolute error, root mean square error, and percentage of errors over 10°F. For precipitation, the indexes were Brier score (accuracy) and reliability.
The accuracy of temperature forecasts was not greatly different for the telecasters and the NWS. Three of 20 pairings show a statistically significant difference according to the sign test; this is not much more than would be expected by chance. For precipitation similar results were obtained: only 1 of 20 Brier score pairings is statistically significant. The NWS has higher reliability scores, although no test exists for determining the statistical significance of this difference.
