Home / Flooding & Risk / Explainer

Flooding & Risk · Explainer

The 100-year flood explained: recurrence intervals and probability

A 100-year flood is not a flood that arrives once a century. It is a statement about probability in any single year. That difference is the source of most public confusion about flood risk, and it changes how the number should be used.

A river in flood spreading over its banks and surrounding ground.
Fig. 0   The 100-year flood is defined by annual exceedance probability, not by a fixed calendar interval.
Ayres River Editorial TeamReviewed by a water-resources engineer
Published 22 June 2026
9 min read · ~880 words
AI-assisted, human-reviewed. Research synthesis and a first draft were produced with AI tooling. The text and the probability figures were checked and edited by the editorial team against the cited references before publication. See cited sources for the underlying material.

After two serious floods in the same neighbourhood within a few years, a familiar question follows: weren't these supposed to be 100-year events? The answer is that they can be, and the apparent contradiction comes from how the term is defined. A 100-year flood describes a probability, not a schedule.

The phrase is shorthand for a flood magnitude with a one in 100 chance of being equalled or exceeded in any given year. That probability, one percent per year, is the figure that engineers and floodplain managers actually work with. It is increasingly written as the one-percent annual exceedance probability, or AEP, precisely because the calendar language misleads.

Recurrence interval and annual exceedance probability

The recurrence interval, also called the return period, is the long-run average time between floods of a given size or larger. A 100-year flood has a 100-year recurrence interval and a one-percent AEP. The two are reciprocals: AEP equals one divided by the return period. A 50-year flood carries a two-percent annual chance; a 500-year flood, a 0.2-percent chance.

The key property is that each year is treated as an independent draw. The one-percent chance resets every year regardless of what happened last year, the way a fair die does not remember its last roll. So a one-percent flood can occur in consecutive years, or twice in a decade, without violating the definition. It is unlikely in any single year, not forbidden across many.

1% AEP (100-yr) Q rarer →
Fig. 1   A schematic flood-frequency curve. Discharge rises as events become rarer; the 100-year flood is read off at the one-percent annual exceedance probability.

Where the number comes from

The figure is estimated by flood-frequency analysis. The annual peak discharges from a gauging record are fitted to a statistical distribution, and the discharge corresponding to the one-percent probability is read from the fitted curve. In the United States, federal practice follows Bulletin 17C, which fits a log-Pearson Type III distribution to the annual peak series.

This is where uncertainty enters. Most records are far shorter than the return periods being estimated. Reading a 100-year value off thirty or forty years of data is an extrapolation well beyond the observed range, so the estimate carries confidence bounds that are often wide. The published number is a best estimate, not a measured ceiling, and our walk-through of how channel behaviour responds to the discharges that shape it is a reminder that the river itself is also changing.

What a non-stationary climate does to the number

Flood-frequency analysis assumes the record is stationary, meaning the underlying probability distribution does not drift over time. A changing climate and a changing catchment both challenge that assumption. Where heavy-rainfall regimes are shifting or where land use has altered runoff, a flood estimated at one percent from the historical record may correspond to a larger annual chance going forward.

The practical response is to treat the published value as a living estimate, to update it as records lengthen, and to design with margin where the consequences of being wrong are severe. Many agencies now map the 0.2-percent (500-year) event alongside the one-percent line to make the tail of the distribution visible. For the catchment-management side of living with that uncertainty, see our piece on nature-based solutions for flood risk reduction.

None of this makes the 100-year flood a useless idea. It remains a clear, comparable benchmark for regulation and design. It just has to be read as what it is: a probability statement carrying real uncertainty, not a promise about the calendar.

Cited sources

References & underlying data

  1. The 100-Year Flood — It's All About Chance. U.S. Geological Survey, General Information Product 106. USGS
  2. Guidelines for Determining Flood Flow Frequency — Bulletin 17C. U.S. Geological Survey, Techniques and Methods 4-B5, 2019. USGS
  3. The 100-Year Flood and Flood Insurance. Federal Emergency Management Agency (FEMA). FEMA Flood Maps
  4. National Water Information System (NWIS), annual peak streamflow. U.S. Geological Survey. USGS NWIS