Here is an interesting article about Moore, Oklahoma, which has been hit by four incredibly powerful tornadoes in sixteen years, which is statistically all but impossible. Beyond the sheer spectacle of it, and the fact that I’ve spent some time in central Oklahoma, the statistical side of it is interesting to me, as I sometimes find myself asked whether some system should be designed to withstand a storm that happens 10 times a year on average, 4 times a year on average, once every 10 years on average, once every 25 years on average, etc. We don’t have a million years of data to base these things on, and even if we did the climate seems to be changing, and even if it were not there is ultimately a judgment call involved about how much risk is too much given our finite resources we have to divide up among so many things.

…tornadoes are pretty rare. One thousand a year, scattered across the continent, does not produce many data points at the scale of an individual city. Most days, there aren’t tornadoes anywhere. That problem is exacerbated by the third issue: Scientists really only have about 50 years of really good tornado documentation. Essentially, Brooks told me, scientists can’t tell us whether what’s happened in Moore is abnormal because they don’t know what a “normal” amount of violent tornadoes is. With all of that, Brooks said, there’s not a good way to clearly tell the difference between patterns and pareidolia. After all, the human brain is primed to find significance in the random. In the creaky corners of our neural pathways, a jumble of rocks can become an old man, a coat hanger can become a drunk octopus, a bunch of craters on the moon give us a friendly smile. It’s so easy for a few random events to make one small town look like a tornado magnet. It would be harder not to see it.