It's also a communication problem, because labels like "100-year/1000-year event" are easily misunderstood.
* they're derived from an estimated probability of the event (independently) happening each year. It doesn't mean that it won't happen for n years. The probability is the same every year.
* the probabilities are estimates, trying to predict extreme outliers. Usually from less than 100s of years of data, using sparse records that may have never recorded a single outlier.
* years = 1/annual_probability ends up giving large time spans for small probabilities. It means that uncertainty between 0.00001% and 0.00002% looks "off by 500 years".
https://practical.engineering/blog/2025/9/16/an-engineers-pe...
I find a useful exercise is to have a cheat sheet of historic flood heights in some area, tell someone the first record high, ask them how high they would make the levee and how long they think it would last. Peoples' sense for extremal events is bad.
That's a great exercise. Where I live a lot of people died because in the past we were not able to make that guess correctly. A lot was learned, at great expense.