Is Hurricane Frequency increasing with time?

The issue of the increase in frequency of hurricanes is one of the leading indicators with respect to a possible connection between global climate change and hurricane activity. To be sure, warmer sea surface temperatures can lead to the naive expectation that hurricanes might be come stronger (though other forces involving shear may be more important in determining storm evolution) but this by itself does not mean that the frequency of Atlantic basin hurricanes should increase as the bulk of them are generated from the easterly wave coming off the African continent. More specifically, it has been recognized since at least the 1930s (Dunn 1940) that lower tropospheric (from the ocean surface to about 5 km [3 mi] with a maximum at 3 km [2 mi]) westward traveling disturbances often serve as the "seedling" circulations for a large proportion of tropical cyclones over the North Atlantic Ocean. Riehl (1945) helped to substantiate that these disturbances, now known as African easterly waves, had their origins over North Africa. While a variety of mechanisms for the origins of these waves were proposed in the next few decades, it was Burpee (1972) who documented that the waves were being generated by an instability of the African easterly jet. (This instability - known as baroclinic-barotropic instability - is where the value of the potential vorticity begins to decrease toward the north.) The jet arises as a result of the reversed lower-tropospheric temperature gradient over western and central North Africa due to extremely warm temperatures over the Saharan Desert in contrast with substantially cooler temperatures along the Gulf of Guinea coast.

The waves move generally toward the west in the lower tropospheric tradewind flow across the Atlantic Ocean. They are first seen usually in April or May and continue until October or November. The waves have a period of about 3 or 4 days and a wavelength of 2000 to 2500 km [1200 to 1500 mi], typically (Burpee 1974). One should keep in mind that the "waves" can be more correctly thought of as the convectively active troughs along an extended wave train. On average, about 60 waves are generated over North Africa each year, but it appears that the number that is formed has no relationship to how much tropical cyclone activity there is over the Atlantic each year.

While only about 60% of the Atlantic tropical cyclones originate from easterly waves, nearly 85% of the intense (or major) hurricanes have their origins as easterly waves (Landsea 1993). It is suggested, though, that nearly all of the tropical cyclones that occur in the Eastern Pacific Ocean can also be traced back to Africa(Avila and Pasch 1995).

It is currently completely unknown how easterly waves change from year to year in both intensity and location and how these might relate to the activity in the Atlantic or the overall issue of global rises of sea surface temperature.

The table below summarizes hurricane activity, in terms of number of detected hurricanes per decade.

1900 / 40
1910 / 37
1920 / 38
1930 / 47
1940 / 40
1950 / 69
1960 / 61
1970 / 49
1980 / 52
1990 / 64
2000 / 63

There are several ways to analyze these data (described below) but the most important issue is detection completness as a function of decade. Prior to 1910, Atlantic basin hurricanes can to be visually detected and logged in the ship captains log. If some of those detected storms never make landfall, then the only record of their existence is the ship's log. Hence, one might expect some level of incompleteness in detected storms. By 1910, however, radio had been developed and actively deployed on most large ships. In this case, the captain would radio to the mainland (or the Coast Guard) that a storm was sighted in or near the shipping lanes and in turn the CG would relay that message to other ships. Via this new method of detection and recording, only hurricanes that occurred very far from the shipping lanes and who don't hit land would go unnoticed. This is not likely to be a large fraction as most Hurricanes have tracks which match up with well with the major Atlantic ocean shipping lanes. In particular, any cargo from Europe destined for ports in Virginia and south will be in hurricane territory. Using sulfur emissions a proxy for ship density (see below) one can see that there is high activity where hurricanes tend to travel.

Today hurricanes are 100% detected by satellites but an equally effect means is by radar, which begin deployment shortly after WW II ended. By 1950, with the advent of radar, the detection efficiency for hurricanes approached 100% and has remained so since then. Hence the hurricane counts from 1950 on are likely to be 100% complete but incompleteness at some level likely plagues the counts prior to 1950. However, we statistically correct for some level of incompleteness.

If we divide the data up into pre-1950 and post-1950 we find the following:

1900-1950 = 40.4 +/- 3.9 storms per decade

1950-present = 59.6 +/- 7.6

On the surface, this would indicate a 50% increase in hurricane frequency in the second half of the 20th century compared to the first half. Is this real? Probably not for two reasons:

1) incompleteness in the pre 1950 data biases the counts to lower values

2) From the graph of the data one can clearly see intradecadel variability in hurricane activity. This is a well known feature whose origin is not yet understood but is likely to be related to the known Atlantic Mean Oscillation which has a 20 year or so periodicity. That waveform is shown below:

In fact, the average storm count in 1950-1970 is 65 storms per decade but in 1970-1990 that declined to 50.5 and in both decade ranges there is no completeness problem.

While corrections for incomplete data are always difficult to make, we can use the data in the table to do so in a relatively straightforward manner by making two reasonable assumptions.

a) The counts in the period 1940-1950 are biased low because of the lack of radar detection methods

b) 1950 marked the beginning of the intradecadel maximum in hurricane activity which simply produced more storms (and this activity declines by 1970).

So we assume that 1/2 of the count different between the 1940's and 1950's are due to detection problems and the other 1/2 is do to real increase in hurricane activity. Thus 15 of the 30 more storms detected in the 1950's are due to real activity and the other 15 are due to radar detection and those 15 (out of 40) would be not detected in the 1940's. This leads to a decadal incompleteness correction of 15/40 = 37.5% which we can now use as the multiplier for the decade counts prior to 1950. Using that correction for incompleteness yields:

1900-1950 = 55.5 +/- 5.4 storms per decade

1950-present = 59.6 +/- 7.6

which at best represents a 10% increase in hurricane activity. Given this, as well as the established intradecadel atlantic ocean activity, it is hard to statistically establish with any degree of statistical integrity that the frequency of atlantic basin hurricanes is systematically increasing with time as a result of global climate change. However, we will later show that the percentage of severe hurricanes is likely increasing with time.