What This Solves
Performs statistical flood frequency analysis using Log-Pearson Type III and Gumbel distributions to estimate flood magnitudes for various return periods from annual peak flow data.
Best Used When
- You have a record of annual peak flows at a gauged site and need design flood estimates
- You need to estimate the 10-, 25-, 50-, or 100-year flood at a stream gauge location
- You want to fit a statistical distribution to observed flood data per USGS Bulletin 17C guidelines
Do NOT Use When
- You are estimating peak flows at an ungauged site without stream gauge data — Use USGS Regional Regression Calculator
- You need a design flow for a small urban drainage area (not a gauged stream) — Use Rational Method Calculator
Key Assumptions
- Annual peak flows are independent and identically distributed random variables
- The record is stationary (no significant trends from land use change, climate change, or regulation)
- Log-Pearson Type III is the standard distribution per USGS Bulletin 17C
- Low outliers are handled using the Grubbs-Beck test or conditional probability adjustment
- The skew coefficient can be weighted with regional skew per Bulletin 17C guidelines
Input Quality Notes
Longer records produce more reliable estimates. Records under 10 years should be supplemented with regional data. Check for regulation changes, diversions, or land use shifts that may make older data non-representative.
Estimate the magnitude of the 2- through 500-year flood from a record of annual peak flows using statistical flood frequency analysis. This tool fits the Log-Pearson Type III and Gumbel distributions to your data following USGS Bulletin 17C, the standard US method for FEMA studies and hydraulic design.
Expert Analysis Tool
This calculator performs statistical flood frequency analysis per USGS Bulletin 17C. Requires historical annual peak flow data. Results should be verified by a qualified hydrologist for regulatory applications.
About Flood Frequency Analysis
Flood frequency analysis uses historical annual peak flow data to estimate the magnitude of floods for various return periods (recurrence intervals). The analysis fits a probability distribution to the observed data and extrapolates to estimate flows with lower probabilities of occurrence.
Log-Pearson Type III is the standard method recommended by the U.S. Geological Survey (USGS) and FEMA for flood frequency analysis in the United States. The Gumbel distribution is a simpler alternative often used internationally.
Distribution Comparison
| Aspect | Log-Pearson Type III | Gumbel (EV1) |
|---|---|---|
| Parameters | 3 (mean, std dev, skew) | 2 (location, scale) |
| Skewness | Variable (data-driven) | Fixed (1.14) |
| Standard | USGS, FEMA (US) | International |
| Best for | Most US watersheds | Symmetric distributions |
Data Requirements
Required:
- Minimum 10 annual peak flow values
- Annual maximum instantaneous peaks
- Consistent measurement method
- Representative of current conditions
Recommended:
- 20+ years for reliable skew estimate
- Regional skew for weighted analysis
- Outlier screening before analysis
- Verification with nearby gages
Data Source: Annual peak flow data can be obtained from the USGS National Water Information System (NWIS).
For educational purposes only. Not a substitute for professional engineering judgment.
How flood frequency analysis works
The Log-Pearson Type III (LP3) method works in log space. The annual peak flows Qi are transformed to base-10 logarithms, and the mean, standard deviation and skew of those logs are computed. The discharge for a given return period T is then:
log₁₀(QT) = ȳ + KT · sy ⟶ QT = 10(ȳ + KT · sy)
- ȳ — mean of the log-transformed peaks, ȳ = (1/n)·Σ log₁₀(Qi)
- sy — standard deviation of the logs, sy = √[ Σ(yi − ȳ)² / (n − 1) ]
- Gs — station skew of the logs, Gs = n·Σ(yi − ȳ)³ / [ (n − 1)(n − 2)·sy³ ]
- KT — the Pearson III frequency factor, read from the table below for the skew and the exceedance probability P = 1/T
- T — return period in years; exceedance probability P = 1/T
When a regional skew is supplied, the calculator computes a weighted skew Gw by blending station and regional skew in inverse proportion to their mean square errors (Bulletin 17C):
Gw = (MSEGr·Gs + MSEGs·Gr) / (MSEGs + MSEGr)
The Gumbel (Extreme Value Type I) alternative works directly on the untransformed flows. Its scale and location parameters are α = s·√6/π and μ = x̄ − 0.5772·α (0.5772 is the Euler–Mascheroni constant), and the design flow is QT = μ + α·y, where the reduced variate y = −ln(−ln(1 − 1/T)). Gumbel has a fixed skew of about 1.14.
Log-Pearson III frequency factors (KT)
Selected KT values from the Pearson Type III table (USGS Bulletin 17C, Appendix 3) for positive-skew watersheds. The full table the calculator interpolates runs from skew −3.0 to +3.0 in steps of 0.1. Negative skews reduce the upper-tail factors, raise the lower tail, and a skew of exactly 0 reproduces the standard-normal Z values.
| Skew (G) | 2-yr (P=0.50) | 10-yr (P=0.10) | 25-yr (P=0.04) | 50-yr (P=0.02) | 100-yr (P=0.01) | 500-yr (P=0.002) |
|---|---|---|---|---|---|---|
| 0.0 | 0.000 | 1.282 | 1.751 | 2.054 | 2.326 | 3.079 |
| 0.2 | −0.033 | 1.301 | 1.818 | 2.159 | 2.472 | 3.321 |
| 0.4 | −0.066 | 1.317 | 1.880 | 2.261 | 2.615 | 3.562 |
| 0.6 | −0.099 | 1.328 | 1.939 | 2.359 | 2.755 | 3.800 |
| 0.8 | −0.132 | 1.336 | 1.993 | 2.453 | 2.891 | 4.034 |
| 1.0 | −0.164 | 1.340 | 2.043 | 2.542 | 3.022 | 4.262 |
KT is dimensionless. For example, with a log mean ȳ = 3.05, log standard deviation sy = 0.20 and skew G = 0.4, the 100-year factor is 2.615, so log₁₀(Q₁₀₀) = 3.05 + 2.615 × 0.20 = 3.573 and Q₁₀₀ ≈ 3,740 cfs.
Return periods and exceedance probability
Return period vs annual chance
A T-year flood has an annual exceedance probability of 1/T. The 100-year flood is the flow with a 1% chance of being equalled or exceeded in any single year — not an event that happens only once per century. Over a 30-year period the chance of seeing at least one 100-year flood is about 26%.
Default return periods
This calculator reports the 2-, 5-, 10-, 25-, 50-, 100-, 200- and 500-year floods. The 100-year flow drives FEMA floodplain mapping; the 500-year flow is used for critical facilities and dam-safety checks. Confidence intervals widen sharply for the rarer events.
Key assumptions and limitations
Assumptions (Bulletin 17C)
- Annual peaks are independent and identically distributed
- The record is stationary and representative of future conditions
- The fitted distribution represents the population of annual peaks
- No significant flow regulation or watershed change during the record
Limitations
- Extrapolation beyond about 2× the record length is highly uncertain
- Records under 20 years give unreliable skew estimates
- Does not account for non-stationarity from climate or land-use change
- High and low outliers can strongly distort the skew and the tails
Annual peak-flow records for US streams are published by the USGS National Water Information System (NWIS). Results for regulatory use should be reviewed by a qualified hydrologist or licensed engineer.
Frequently asked questions
What is flood frequency analysis?
Flood frequency analysis is a statistical method that uses a record of annual peak flows (the single largest instantaneous discharge each year) to estimate the magnitude of floods for a range of return periods, such as the 2-, 10-, 50-, 100- and 500-year flood. A probability distribution is fitted to the observed peaks and then used to extrapolate to the rarer, larger events used in design.
How many years of data do I need?
USGS Bulletin 17C and this calculator require a minimum of 10 annual peaks, but at least 20 years is recommended because the station skew coefficient is unreliable for short records. As a rule of thumb, avoid extrapolating much beyond about twice the length of the record: estimating a 100-year flood from only 15 years of data carries large uncertainty.
Why is the Log-Pearson Type III distribution used in the US?
Log-Pearson Type III (LP3) is the standard distribution recommended by USGS Bulletin 17C and used by FEMA for flood insurance studies and most federal flood work in the United States. It is a three-parameter distribution (mean, standard deviation and skew of the log-transformed flows), so the data-driven skew lets it fit the asymmetric, heavy-tailed behaviour typical of annual flood peaks better than a fixed-shape two-parameter distribution.
What is the difference between Log-Pearson III and Gumbel?
LP3 fits three parameters (including a variable skew computed from your data) to the base-10 logarithms of the flows, while Gumbel (Extreme Value Type I) is a two-parameter distribution with a fixed skew of about 1.14. LP3 is the US regulatory standard and adapts to the shape of each record; Gumbel is simpler, common internationally, and works best when the log-flows are close to symmetric. This tool can report both side by side for comparison.
What is regional (weighted) skew and when should I use it?
The station skew from a short record is statistically noisy. Bulletin 17C improves it by blending the station skew with a published regional skew, weighting each by the inverse of its mean square error (MSE). Use the optional regional-skew inputs when you have a value from the USGS regional skew map or a state study; the default regional MSE is 0.302. If you leave it blank, only the station skew is used.
Standards & related tools
Was this calculator helpful?
Last verified: February 2026