Assessing the Skill of Operational Atlantic Seasonal Tropical Cyclone Forecasts

Abstract

Since 1984, W. Gray of Colorado State University and a team of researchers have been issuing seasonal tropical cyclone forecasts for the North Atlantic Ocean. Prior to this, little work had been done in the area of long-term tropical cyclone forecasting because researchers saw minimal potential skill in any prediction models and no obvious benefits to be gained. However, seasonal forecasts have been attracting more attention as economic and insured losses from hurricane-related catastrophes rose sharply during recent decades. Initially, the forecasts issued by Gray consisted of output from simple statistical prediction models. Over time, the models became increasingly more complex and sophisticated, with new versions being introduced in 1992, 1993, 1994, 1996, and 1997. In addition, based on a combination of experience with the statistical models and other qualitative considerations such as examinations of analog years, the statistical forecasts were modified to create adjusted seasonal forecasts. This analysis assessed the skill demonstrated, if any, of both the statistical and adjusted forecasts over the benchmarks of climatology and persistence and examined whether the adjusted forecasts were more accurate than the statistical forecasts. The analysis indicates that, over the past 18 yr, both the statistical and adjusted forecasts demonstrated some skill over climatology and persistence. There is also evidence to suggest that the adjusted forecast was more skillful than the statistical model forecast.