Number 34, June 2002

Reply of Powell and Aberson to Franklin

We appreciate the interest that Franklin has shown in our work and appreciate his efforts to add to the studies of tropical cyclone track forecast accuracy near the United States. Powell and Aberson concluded that forecasts of the time and location of tropical storm and hurricane landfalls in the mainland United States, Puerto Rico, and the U.S. Virgin Islands have not improved statistically (except the 24 h time of landfall forecasts) between 1976 and 2000. This result does not reflect on the accuracy of forecast tracks in general and should not be taken to mean that such forecasts for storms near or threatening the coastline have not improved during the same period of time.

Franklin attempts to quantify the accuracy and improvement trends of these forecasts. Though we do not necessarily disagree with his conclusions, we believe that there are a number of concerns with his methods and that his results do not conflict with those of the Powell and Aberson study.

Franklin looked at the subset of forecasts issued when hurricane or tropical storm watches or warnings were in effect for the mainland United States. However, some of these storms did not make landfall, and many of the forecasts verified either well inland or far from land areas. For example, for Hurricane Gabrielle (2001), the forecasts issued 24 h before landfall near Venice, FL, and earlier, are not included because warnings were not issued 21 h before landfall; however, 72 h forecasts that verified near the Grand Banks off Newfoundland are included.

Franklin specifically mentions Hurricane Michelle (2001) as being of "keen interest" because it was forecast to come close enough to Florida without making landfall to prompt the issuance of hurricane warnings. In this case, the 48 h forecast of the closest approach to Florida is not included in the Franklin subset, though the quick transition to an extratropical cyclone prevented the inclusion of forecasts verifying in the central Atlantic Ocean from his sample. We believe that the forecasts issued 24 to 48 before landfall should be included in, and that the forecasts that verify far away from the coastline should be excluded from, such a study. Further, Franklin excludes all forecasts of storms near Puerto Rico and the Virgin Islands that were included in the Powell and Aberson study.

Even so, Franklin shows that the forecasts in his chosen subset improved at an annual rate of 0.8%, 1.7%, and 1.9% at 24, 48, and 72 h respectively. We have calculated the improvement rate in the Powell and Aberson sample. Landfall location forecasts have improved at an annual rate of 0.0%, 1.6 %, and 1.6%, and the timing forecasts have improved at an annual rate of 2.2%, 0.3%, and -0.7%, at the same lead times. Since both the timing and location errors contribute to the total error, the improvement rates found in both studies are comparable, at least at 24 and 48 h; the very small 72 h sample in the Powell and Aberson study may be the cause of the degradation shown at 72 h. Franklin found that improvements in the 24 and 48 h forecasts in his sample were statistically significant at the 95% level, and the 72 h forecasts at the 90% level. Aberson and Powell showed that the 24 h forecasts of the landfall time had significantly improved at the 99% level, and the 48 h forecasts of landfall location had significantly improved at the 90% level. The 72 h forecasts did not show significance, probably due to the very small sample.

Finally, Franklin states that the public should "focus attention away from the precise forecast track of the center," though his study quantifies the errors of that exact metric. We agree with this statement, but we do not believe that this precludes evaluation of errors, either of the landfall time and location or of forecast tracks threatening land, after the event, and informing the hurricane preparedness community and the public of how well tropical cyclone impacts are forecast. As long as forecasts of the precise track of the center are issued by the National Hurricane Center, the precise track will continue to be a focus of the hurricane preparedness community and the public. Public education to the fact that dangerous conditions associated with tropical cyclones cover a large area away from the storm center must continue. This education will remain a difficult task as long as public is presented with such forecasts.

Mark Powell and Sim Aberson
Hurricane Research Division
Atlantic Oceanographic and
Meteorological Laboratory
Mark.Powell@noaa.gov
Sim.Aberson@noaa.gov