1000 year storm

Emphases Mine

If you add that language in, then yes, you’d be wrong. It’s a 1% chance of occurring in any given year. That means 1% this year, 1% next year, 1% the following, etc etc etc. Indeed, there’s nothing stopping this event in South Carolina from happening again in a week or two, the atmosphere has no memory.

My point is that calling an event with a 1% chance of occuring per annum as a “[once in] 100 year” event is nonsensical. If an even has a 1% chance of occuring per annum, ~there is a 63% chance it will occur at least once in a 100 year period.

Stranger

And it may occur twice or more. The point is, it will occur on average once per 100 years. And what’s so bad about the “63% probability”? If you’d prefer 100% you’ll need to make that time period VERY long. :smiley:

The really interesting case is where there’s inadequate historical evidence to guess the probability of a rare event. So far there’s been zero response to my post about Hurst and Mandelbrot’s power-law. Was my post wrong? Do I really need to explain why the central limit theorem just doesn’t model these sorts of events?

Ah … I see what you’re saying … yes, in any given 100 year period, the chances of a 100-year event in 63%. That can be calculated from the 1% per year number. It still boils down to someone looking at a hundred years of river gauge data and pointing to the highest and calling that the 100-year flood, and feel lucky to have a hundred years data.

Weather data tends to have very large standard deviations, so if it’s a “1,000-year flood”, that’s plus or minus 500 years or so.

Well, sort of. If the odds of occurance are 1% per annum, then yes, the likelyhood of occurrance in a 100 year period is 63%. But, perhaps incorrectly on my part, I would interpret a “100 year event” as an event that would not occur more than once in any 100 year period to some statistical degree of probability (e.g. a single tailed three sigma value). Saying that an event with a 1% per annum probability of occurance is a “100 year event”, as if the per annum probability is addative, is really nonsensical.

Stranger

Yes indeed, that’s the reason I wanted to get the actual definition posted early in this thread: