For those of a statistical bent, it is instructive to grab a list of hurricane frequencies over the last, say, sixty years, bin them into a histogram and overlay a Poisson distribution with the same mean as the frequency data (appropriately vertically scaled, of course). The fit is really good.
There were 371 named hurricanes in the 60 years 1948-2007. For a Poisson distribution with mean lambda events in a given time interval, the PDF of getting k events in the interval is (lambda^k exp(-lambda)/k! (this is the continuous version). Here, we have lambda=371/60 = 6.18. Scaling by 60 for the number of years’ worth of data we have and superimposing the plot over the histogrammed data shows that hurricane frequencies are as near as dammit random. A linear fit to the time series data gives a VERY weak upward trend which would be nearly eliminated by a few years of low activity.
Statistics as weak as these being used to bolster any kind of alarmism have the scent of voodoo about them.
