Abstract:
Up until the beginning of this century, the National Weather Service (NWS) had no official method by which to forecast hurricanes. Through the hard research of S. J. Kimball, and later refinements done by Appleman and Jarrell, a method for producing probabilistic hurricane strike forecasts had been derived. Differences of opinion exist among forecasters and between forecasters and users, regarding the meaning of a "good forecast". There are three types of goodness applied to a weather forecast. In this paper, these three type of goodness will be discussed in detail and in respect to the results from my verification of the probabilistic hurricane strike forecasts from 1994 and 1995. With the use of the Brier Score (BS), it is possible to determine the validity for probabilistic forecasts. In this study, I utilized the decomposition of the BS, which allowed me to view exactly what aspect(s) needed improvement. This study showed that in 1994, forecasts were, on average, 4% better than climatology. In 1995, the forecasts, overall, were approximately 17% better than the climatological probability for the season. Possible reasons for the results of my study, and suggestions for improvement among probabilistic hurricane strike forecasts, are discussed.
Paper available upon request.