For the second time in two years, Ellicott City drowned in a  “thousand-year” rainstorm Sunday. The concept of a thousand-year rain event, while intended to place the rarity of an extreme event into perspective, is controversial.

The headline of Capital Weather Gang’s detailed explanation of Sunday’s flood, “The second 1,000-year rainstorm in two years engulfed Ellicott City,” by Jeff Halverson, drew swift criticism.

Article’s headline is a clear cut case of statistical illiteracy, complained Tee Queue, a Facebook reader.

But hydrologists, geologists and meteorologists have applied the concept for decades. That said, it is often misunderstood. “From a practical standpoint, the concept can be misleading without a proper understanding of how such a number is derived,” said Halverson, who is also a professor of meteorology at the University of Maryland Baltimore County.

A 1,000-year rain event, as its name implies, is exceptionally rare. It signifies just a 0.1 percent chance of such an event happening in any given year. “Or, a better way to think about it is that 99.9 percent of the time, such an event will never happen,” explained Shane Hubbard, a meteorological researcher at University of Wisconsin’s Space Science and Engineering Center.

But people often fail to appreciate that when scientists declare that a storm is a 1,000-year, 500-year or 100-year event, it does not mean this extreme rainfall will necessarily happen that infrequently. These return intervals just express probabilities, which lead some to underestimate the risks they signify.

“The fact is, the math behind these numbers also tell us a 500-year event at a given place has about a 10 percent chance of happening over a 50-year period,” wrote Brian Bledsoe, a professor of civil and environmental engineering at the University of Georgia in a commentary for the Capital Weather Gang.

He added: “[M]ost people are still surprised, if not astonished, to learn that the 100-year flood at a given location has more than a 1 in 4 chance of occurring within the term of a 30-year mortgage. For most of us, this 26 percent chance our home will be flooded before we have a chance to pay it off is troubling if not unacceptable.”

Destroyed parked cars are stranded in the mud May 28 in Ellicott City, Md. This comes two years after another flash flood killed two people and devastated the historic downtown. (Katherine Frey/The Washington Post)

These metrics are also limited based on the assumptions on which they’re calculated.

For one, rainfall and flood data generally go back only 100 years or so, so scientists must extrapolate available data back in time to determine what 500-year and 1,000-year events actually represent.

Furthermore, the climate is changing, and precipitation events have become more intense in recent decades, so what constitutes different return frequencies (100-year, 500-year, 1,000-year and so forth) is probably changing.

Climate change studies have found that what’s considered a 500-year rainstorm today may become much more frequent in coming decades.

An additional complication is that a 1,000-year rainstorm is not the same thing as a 1,000-year flood, and the two are sometimes used interchangeably. Whereas the rarity of a rain event can be objectively analyzed based on how much water falls from the sky, the frequency of flood events can be a moving target because of land-use decisions that make a particular location more or less vulnerable.

Although rainfall of six inches in three hours has a probability of occurrence of less than or equal to one in 1,000 in Ellicott City according to the National Weather Service, the town has flooded 15 times since 1768 indicating the flood return interval is much higher.

Halverson even wonders whether six inches of rain in three hours in Ellicott City is truly as rare as the Weather Service’s statistics suggest. “There were two of these 1,000-year events in two years,” Halverson said. “That is not distorting the truth. But does the statistic anymore hold water?”

Despite their shortcomings, the University of Wisconsin’s Hubbard, who analyzes geographic information to help decision-makers plan for floods, stands by the use of these return interval metrics. “For a community, they help put these events into perspective and understand the impact,” he said.

He added that they have “tremendous” value for flood planning and designing infrastructure to be able to withstand events up to a certain intensity. “Decision-makers have to be able to pick a number and say this is the number we need to be prepared for,” he said.