Landsea, Trenberth, Connolley... the list of famous scientists drawn irresistibly to draw trend lines through hurricane numbers grows. Anyway, this follows some recent stuf on sci.env. I found Atlantic Storms
sorted by Year (1851-2002), extracted the raw data, and plotted it. I have concerns though: it seems to me that there might be biases in missing early storms. And indeed if you look at the East Pacific page, it seems there are early biases that they warn you of. But they don't warn about the Atlantic numbers. So lets trust them, at least for these porpoises.
So we get the
[A word about stat sig: as two comments pointed out (thanks) I was being naive about the sig: you need to take autocorrelation of the data into account. I know this full well, but I was in a hurry. To do this, you can (a) do it properly (deflate the d.o.f.) or (b) try averaging the data for a longer period. Thats what the middle pic is: I've averaged it for 10 year periods, so the autocorr mostly disappears. And the trend stays sig now.]
But... if you look closely, it looks like there is a jump in numbers in 1920-1930 ish, perhaps? So, I ran it through my jump-detection software (Reference: Lund and Reeves, J Climate, p2547, 2002) and get... this pic. Top graph is the "is-there-a-jump" function, which intersects the line of 95% probability, so yes it thinks there is. Bottom graph is the softwares best guess at the jump point, and trend lines before and after assuming a discontinuity there. The software thinks the jump is between 1930 and 1931. The trend, from 1931 on, is not significant any more [and nor is the pre-1930 trend. Note that this *can* be trusted, even with autocorr: taking into account the corr would only make it even less sig]. Its interesting that the software picks up the jump that my eye (at least) does. And I *didn't* run the software first and then decide on a jump: I looked at it by eye, thought "hmm I wonder what the software sez" and lo!, we agree.
What, if anything, does this mean? Probably, not very much. You'd need to be far more careful about homogenising the data, and checking back to the sources, if you were to take this seriously. You might want to see what happened around 1930 to observing practices, ship movements, etc, around there. This is just for fun.