How Big Is a 100-Year Flood? New USGS Numbers for Georgia and the Carolinas

Engineers sizing a bridge or a culvert, and planners drawing a floodplain, need one number above all: how big a flood a stream is likely to produce, and how often. For rural streams in Georgia, South Carolina and North Carolina, the U.S. Geological Survey has updated that number using peak flows measured at 965 streamgages through 2017 — and has good news about how stable it is. Of 331 long-term gages, 276 (83%) show no statistically significant trend in their annual peak flows.

The study was done with the transportation departments of all three states and the North Carolina Department of Crime Control and Public Safety. This page summarizes the USGS fact sheet; the full results and data are in a companion report and data release (Feaster and others, 2023; Kolb and others, 2023).

What a "100-year flood" means

USGS expresses flood size as an annual exceedance probability (AEP) — the chance that a flood at least that large happens in any given year.

So a "100-year flood" is a flood with a 1% chance of being equaled or exceeded in any year — not one that arrives once a century.

Are floods getting bigger?

Not on these rural streams, on the evidence so far. Among the 331 gages that were still operating in 2017 and had 30 or more years of record, the Kendall's tau test found:

That matters because a real long-term trend would undermine the whole method, which estimates future flood odds from past records. The results show no substantial long-term change in peak-flow patterns for rural streams in the three states.

How the estimates were made

Estimates at the 965 gages with at least 10 years of record follow Bulletin 17C (2019), which replaced Bulletin 17B, the federal standard since 1982. Both fit a log-Pearson Type III distribution to annual peak flows. Bulletin 17C adds three improvements:

1. the expected moments algorithm, which can use interval estimates, censored values and multiple thresholds of observation;
2. the multiple Grubbs–Beck test, which flags several potentially influential "low outliers" rather than one;
3. confidence intervals that account for historical and paleoflood information and regional skew.

The distribution depends on three properties of a gage's record: its mean, standard deviation and skew. Skew is uncertain for short records, so it is weighted with a regional skew — and the new regional value is much more reliable than before:

Better methods and longer records at USGS gages explain much of the improvement.

Figure 1. The five hydrologic regions and the rural streamgages considered for the regional analysis. USGS.

Streams with no gage

Most streams have no gage, so USGS built regression equations to estimate flood flows anywhere from a basin's drainage area and how much of it lies in each of five hydrologic regions, confirmed from earlier studies as still appropriate:

• HR1 — Piedmont and Ridge and Valley
• HR2 — Blue Ridge
• HR3 — Sand Hills
• HR4 — Coastal Plain (the base region)
• HR5 — Lower Tifton Upland

Of the 965 gages, 801 fed the regression — 670 in the three states and 131 from neighboring Alabama, Florida, Tennessee and Virginia. The other 164 were left out as redundant, where one gaged basin nests inside another of similar size. Working across states widens the range of conditions and keeps estimates consistent for projects whose basins cross a state line. The Blue Ridge curve has a different slope from the others, so the equations include a cross-product term for it.

The equations explain 94.1% (2-year flood) to 89.5% (500-year flood) of the variation (pseudo R²). Their average standard error of prediction is 35.8 to 44.4%: there is about a 68% chance that the true value at an ungaged site lies within that margin of the estimate.

For a basin lying entirely within one region, the equations reduce to Q = a × DA^b, with Q in cubic feet per second and DA, the drainage area, in square miles:

For basins spread across regions, the general form uses the percentage of the basin in regions 1, 2, 3 and 5 (PCT1, PCT2, PCT3, PCT5), with the Coastal Plain as the base:

These apply to unregulated rural streams.

Figure 2. The 100-year (1% AEP) flood against drainage area, by hydrologic region. USGS.

What comes next: the cities

Flood statistics depend on the length and conditions of the record, so they are usually refreshed about every 10 years. By the end of water year 2023 (September 30, 2023), 12 years will have passed since the urban statistics for the three states were updated. A next study could combine the 801 rural gages with urban records through at least 2023, test variables such as impervious area or percent development, and perhaps produce one set of equations for rural and urban basins alike.

Sources

Based on "Magnitude and frequency of floods for rural streams in Georgia, South Carolina, and North Carolina, 2017—Summary," U.S. Geological Survey Fact Sheet 2023–3011, summarizing Feaster and others (2023), Scientific Investigations Report 2023–5006, and Kolb and others (2023), data release; a work of the United States government in the public domain.