24th Conference on Hurricanes and Tropical Meteorology

10A.1

Accuracy of U.S. Hurricane Landfall forecasts in the Atlantic basin 1976–1998

Mark D. Powell, NOAA/AOML/HRD, Miami, FL; and S. D. Aberson

U. S. landfall position and timing forecast errors were compiled for 1976-1998 from a set of nearly 600 official forecasts of landfall for the Atlantic basin. Of all forecasts issued, about 10 % are for actual landfalling storms and 2 % are for storms initially forecast to landfall in the U. S. before turning out to sea. In a given year, on average, 200 forecasts are issued for about 9 storms, with 2-3 landfalls in the U.S. and one "near miss". Landfall position errors are significantly smaller than those compiled for the 1970-1996 period from the general forecast population at similar time periods. Landfall position forecasts demonstrate skill relative to climatology and persistence at all time periods while landfall time forecasts are skillful at all but the 55-72 h period. Although annual landfall forecast errors plotted vs. year suggest an improvement trend, none of the fits were statistically meaningful when corrected for forecast difficulty. These findings are consistent with an hypothesis that it is difficult to improve upon forecasts that take advantage of proximity to the upper air observing system. With realtime wind field analyses available on a regular basis, it may be possible to determine the hurricane warning area by the size of the damaging part of the wind field plus 1.5-2.0 standard deviations of the landfall position uncertainty, and reduce warning areas as the time of predicted landfall approaches. Landfall timing errors indicate an early bias of 1-4 h within 36 h of predicted landfall. Landfall timing uncertainty is ~12 h at 24, 36, and 48 h suggesting that it may be possible to disseminate hurricane warnings about 12-24 h earlier (at 48 h rather than 24-36 h before predicted landfall) without substantial loss of lead time accuracy. Landfall position forecasts are most accurate for strong, slow storms moving perpendicular to the coast. However the most accurate forecasts of the time of landfall forecasts occur in weak, fast storms moving perpendicular to the coast. Changes to the U. S. Weather Research Program "Hurricanes at Landfall" track forecast goals are recommended in light of these results.

Session 10A, Tropical Cyclone at Landfall III (Parallel with Sessions 10B and J7)
Thursday, 25 May 2000, 1:15 PM-3:15 PM

Next paper

Browse or search entire meeting

AMS Home Page