The severity of floods tends to get put in terms of years: a 100-year flood, a 500-year flood, a 1,000-year flood. But this isn't an assessment of "the worst flood in" that time — places like Houston don't actually have detailed weather records going back to 1017 AD, after all.
The lack of hundreds of years' worth of flood data is actually the reason we have the term "100-year" flood to begin with. When the government decided to map flood-prone areas to improve the National Flood Insurance Program in the early 1970s, the maps couldn't just use the worst flood ever recorded in a given area to judge what a "bad flood" would look like — because some areas had more records than others, and besides, just because a bad flood hadn't happened yet didn't mean it couldn't.
Instead, the standard set for mapping flood-prone areas was a compromise between the existing Army Corps of Engineers standards for dams and levees, and the (much more modest) standards that most communities had set for flood prevention. The areas deemed at risk of a bad flood were the areas that had about a 1 percent chance of flooding in any given year: in other words, the areas that would flood approximately one year out of every 100.
A 500-year flood is based on the same principle: Experts estimate that in any given year, there's a 1-in-500 (0.2 percent) chance a flood this bad will strike a particular area. In theory, that means that over 500 years, that will happen once: so there will be one flood that bad over a 500-year period.
Of course, different areas flood at different frequencies. So how bad a flood has to get to qualify as a 1-in-500-year flood is going to vary depending on where you're judging it — and a flood will probably qualify as a "500-year" occurrence in some locations but not others. This table, for example, shows the flooding in various locations in the Houston area in April 2016. You can see that the flood reached 500-year levels in certain places, but in others it was the sort of event meteorologists would predict to happen every 10 years.
When it comes to planning for future floods, you have to get a little more abstract. So FEMA maps out 100- and 500-year "floodplains" — the places that would get flooded by the kind of rainfall that has a 1 percent chance (or 0.2 percent chance) of falling on an area in any given year.
But probability never works out perfectly in practice (as you know if you've ever flipped a coin twice and gotten heads or tails both times). And it's especially hard to get the probability of an event perfect when the circumstances keep changing — as they do when we're talking about weather events, which become more or less common depending on the underlying climate. And current climate change trends could easily increase the chance of bad flooding — there's more
now (ready to condense into storm clouds and precipitation), for example, than there was 70 years ago.
As a result, FEMA has to keep updating its assessments of the floodplains — i.e., which locations should think of flooding as a 1 percent possibility, and which should think of it as something with a 0.2 percent chance of happening in any given year. The FEMA maps for Harris County had just been updated in late 2016. However, the city of Houston itself is working off a Hazard Mitigation Plan it developed in 2012 — based on where FEMA was then saying the city's 100-year and 500-year floodplains were located.
In the meantime, the reality of Houston's flooding has already shown the old models to be out of date. An area of West Houston called Memorial City, for example, was outside Houston's 500-year floodplain but flooded three times in the past decade: in 2009, 2015, and 2016.