Trends in Forecasting the Tracks of Atlantic Tropical Cyclones

Abstract

Errors in official 24 h forecasts of tropical cyclone motion over the Atlantic for the period 1954–80 are examined for the purpose of isolating any long-term trends in the data. It is shown that the magnitude of a forecast error has been primarily a function of forecast difficulty, i.e., how well storm motion adheres to climatology and persistence. Another contributing factor (in the negative partial correlation sense) is shown to be the storm's initial longitude—a measure of the adequacy of initial analyses. After adjusting 24 h forecast error for these two factors, it is shown that errors have declined gradually over the 27 years from near 124 n mi in 1954 to near 107 n mi in 1980—a 13.7% reduction. Most of the decline since the mid-1960s is attributed to better specification of initial storm motion through satellite imagery. Although the decline of forecast errors is encouraging, a disturbing aspect is that the rate of decline appears to have slowed in recent years. This leveling-off is attributed to a loss in the ability to assess environmental steering through mid-level analysis deficiencies that have been compounding since 1965 and, more recently, to a plateauing in the ability to obtain still better storm initial-motion vectors. To assure a continued monotonic decline in 24 h forecast errors, mid-level initial analysis (500 mb) over the essentially data-void tropical cyclone basins must be improved.