Hurricane Maria

As you already know, Hurricane Maria is a Category 5 storm menacing the Leewards, and heading, likely, for Puerto Rico.

Please avoid making the mistakes that were made in talking about Irma. There will probably be no Category 5 storm hitting Puerto Rico. The storm will probably be a Category 4 before it hits. So, reporters will sloppily declare that "a category 5 storm is heading for Puerto Rico" then later Rush Limberger will say "Look there was never no such storm, see?" and so on.

But, a Category 4 storm is still nothing to sneeze into, and Puerto Rico and the other island in this storm's path are in big trouble.

As we wait for that to develop further, let's talk a bit about predicting hurricane seasons. A lot of people are arguing about whether or not global warming means more, or bigger, or whatever, Atlantic hurricanes. (Short answer: there are probably already more hurricanes in the Atlantic, and bigger ones, but they are hard to count because they are in fact rare events, and science says that there will likely be more in the future). One of the dumber counter arguments to science suggesting that there may be bigger storms, or more of them, is this: You can't even predict the weather for next weekend, so what the heck, right?

The counter argument to that is this: Ok, fine, we don't know very accurately what the weather is going to be like next weekend, but what would you say if I told you that we can do a pretty good job of telling, before the hurricane season starts, how many named storms there will be? Huh? Wouldn't that be amazing?

Turns out we can. And the fact that we can suggests that we should be trusting the models, generally, and therefore, expecting more and bigger hurricanes.

I looked at the predictions made in several recent years by several groups that do this prediction, and found out that the total average wrongness averaging across all of them is down near one hurricane, with the range of wrongness being between -8 and 4, but with most of the predictions being within just a few one way or another.

First, a quick look at the number of named tropical storms in the Atlantic per year:

People will tell you there is no trend here, but as you can see, about 44% of the variation seen in the number of storms over time is accounted for by year, so there is a good argument that there is an increasing trend. One might argue that back in the 70s we missed some Hurricanes. That, I do not buy, but if you need to believe that, you can see there is still a trend from 1980 on. So there is an increase.

But I digress. Here's the point I wanted to make with this graph. The number of hurricanes in a given year varies quite a bit, from 4 to 28 over this time period (and less over the most recent years). So, a method of prediction that gets within two or three in either direction is pretty good.

The number of named storms (many, usually most, of which will be hurricanes) that will happen in a give season in the Atlantic is predicted with reasonable accuracy by several groups. Here's a chart showing several different prediction groups compared to reality.

The light blue line is the actual number, and you can see that except for 2011 and 2012, the number of storms predicted by various groups, and the number that occur, are very similar. Let's assume 2011 and 2012 are strange years and arbitrarily ignore them (I know, this would normally be cheating but we'll come back to that in a minute).

Looking only at those years, one prediction undershot by 4, one prediction undershot by 3, and 7 overshot by 3 or 4. The other 20 predictions were off by no more than two storms.

So, why is it OK to fudge the data like that? Well, it isn't really, but the last two years of predictions have been off by one or fewer storms on average, and I'm assuming the predictions are getting better and better. In other words, if I were to lay odds on predicting three years in a row a few years in the future, I'd bet that the average difference between all the predictions an the actual observations would be less than one named storm, and I'd win that bet. For this reason I don't care so much about older data.

Notice that I'm only using predictions made prior to the start of the season, not later updates which some groups do provide.

For this year, we've had 13 named storms so far, and all the various groups predicted 14. There is plenty of time to have a couple more storms, so likely, this year will be a bit more active than expected, but just by a couple of storms.

Back to Maria for a moment, you may be wondering if this storm will hit the coast along the lower 48. It is possible, it is too early to tell, but history and the models that exist so far both suggest that it probably will not, but stay tuned.

More like this