In a forthcoming paper, Powell and Aberson (2001, hereafter PA2001) argue that while Atlantic Basin forecasts may have improved overall, there has not been an improvement in NHC forecasts of tropical cyclone landfalls. Although the NHC does not issue landfall forecasts, PA2001 inferred forecast landfall locations and timings by interpolation from official NHC forecast positions. They then calculated errors in forecast landfall location and timing. They report that the resulting trend lines showed no, or no statistically significant improvements over the period. The implication of PA2001 is that official NHC forecasts for storms threatening the United States are not improving.
In this study, we examine the time tendency of track forecast errors for tropical storms and hurricanes threatening the United States over the period 1970-2000. In order to more directly compare with the results of McAdie and Lawrence, and because there is no unique way to infer a landfall location and time from an NHC track forecast, we verify these forecasts in the traditional way (e.g., the forecast error is the great circle distance separating the 24 h forecast position from the 24 h verifying position). To identify forecasts for "threatening" tropical cyclones, we use the subset of forecasts issued when tropical storm or hurricane watches or warnings were in effect for the mainland United States. Compared to the relatively small number of actual U.S. landfalls, this set of forecasts has the advantage of providing a larger, statistically more stable sample of cases from which to determine long-term trends.
Preliminary results show that the long-term improvement trends for forecasts issued when U. S. watches or warnings are in effect are comparable to those of the Atlantic Basin as a whole.
Supplementary URL: