Comment on “Changes in the rates of North Atlantic major hurricane activity during the 20th century”

Abstract

Atlantic major hurricanes have a profound impact on human and natural environments. Understanding and predicting these storms’ occurrence on daily, intraseasonal, interannual, multidecadal and climatic timescales can be of tremendous benefit and should be pursued vigorously. Elsner et al. [2000] (hereafter as EJN) address multidecadal changes of Atlantic basin major hurricanes - those tropical cyclones with maximum sustained surface winds of at least 50 m/s. However, aspects of EJN analysis are questionable because its treatment of the Atlantic major hurricane database may be erroneous, because its proposed link to the North Atlantic Oscillation is unsubstantiated, and because it fails to cite or acknowledge key earlier studies. Treating the record of Atlantic major hurricanes since 1900 as trustworthy in its entirety is very questionable. The assumption that the records of major hurricane activity are complete before the advent of aircraft reconnaissance leads directly to the erroneous conclusion that major hurricanes became much more numerous starting in 1943. That year - not coincidentally - was the beginning of routine aircraft flights into hurricanes [Sumner 1944]. EJN state that “the observational bias is smallest for the strong storms considered here”. It is unclear how the authors come to this conclusion. Major hurricanes typically stay at that intensity for an average of 2-3 days and have their 50 m/s or higher winds concentrated in the mesoscale (10s of km) eyewall [Landsea 1993; Powell and Houston 1998]. Before aircraft and satellite reconnaissance, it was simply not possible to detect such small transient features offshore consistently or reliably. Holland [1981] demonstrated that even when numerous ships and buoys are in a strong tropical cyclone, the actual intensity is likely to be drastically underestimated without aircraft or satellite data. Another way to understand the problem is a comparison of the relationship between all major hurricanes to the subset that have hit the United States. (The latter should be sampled adequately back to the beginning of the 20th Century.) These two records are correlated at r=0.41 for the years 1943-99 (less than 17% of the variance) versus r=0.78 for 1900-1942 (over 61% of the variance). (These two correlation coefficients show a strong significant difference in a two-tailed test with a p-value of less than 0.03.) Since in nature the relationship should be relatively steady over time, one can conclude that before the mid-1940s, the Atlantic major hurricane records were less likely to include those that did not strike the United States. Solow and Moore [2000] come to similar conclusions as stated above by analyzing the ratio of landfalling U.S. hurricanes to all Atlantic basin hur