ABSTRACT
The purpose of this paper is to assess the value added, as it relates to the accuracy of a particular long range weather forecast. The forecast under consideration is Accuweather’s 45 day forecast. This forecast predicts both the high and low temperature for every day, in a given location (Philadelphia, PA), for up to 45 days in the future. In order to determine whether there is value to such a detailed, long range forecast, the differences between the actual high and low temperatures and the 44 day lag time predicted high and low temperatures were calculated, over a period of one year. Additionally, the differences between the actual high and low temperatures and the historical mean high and low temperatures were calculated for the same period of time. The accuracies of the 44 day forecast highs and lows were compared to the accuracy of simply using the historical mean temperatures as the forecast highs and lows. It was found that the historical mean high and low were statistically significant better predictors (p-values < 0.05) of the daily high and low temperatures than were Accuweather’s 44 day lag time predictions.
Keywords
long range weather forecasts, value added, forecast accuracy.