United States Flood Peak Dataset

175,008

Observations

2,236

Series

2,236

Subjects

1806–2025

Period covered

What this looks like

One real series from the dataset, and real rows from the head of its published extract. Neither is a mock-up: the rows below are the same rows, unchanged, that are in the file you would download.

Annual maximum river flow at CUMBERLAND RIVER AT NASHVILLE, TN

The longest single series this dataset holds — 167 instantaneous values, one of 2,236. The dataset is not this chart.

Drawn to show the shape and resolution of what the dataset contains, not to make a point about CUMBERLAND RIVER AT NASHVILLE, TN. Gaps are breaks, never interpolated.

View the data behind this chart

Underlying data
Period	CUMBERLAND RIVER AT NASHVILLE, TN (m³/s)
1808	0
1826	0
1827	0
1830	0
1832	0
1833	0
1834	0
1835	0
1838	0
1839	0
1840	0
1841	0
1842	0
1843	0
1844	0
1845	0
1846	0
1847	0
1849	0
1850	0
1851	0
1852	0
1853	0
1854	0
1855	0
1856	0
1857	0
1859	0
1861	0
1863	0
1865	0
1867	0
1868	0
1869	0
1870	0
1871	0
1872	0
1873	0
1874	0
1875	0
1876	0
1877	0
1878	0
1879	0
1880	0
1881	0
1882	0
1883	0
1884	0
1885	0
1886	0
1887	0
1888	0
1889	0
1890	0
1891	0
1892	0
1893	3,341
1894	3,455
1895	2,401
1896	3,568
1897	4,191
1898	3,143
1899	3,313
1900	1,849
1901	3,228
1902	3,879
1903	3,285
1904	3,002
1905	2,277
1906	2,622
1907	3,171
1908	2,050
1909	3,285
1910	2,693
1911	3,115
1912	4,078
1913	4,332
1914	2,860
1915	3,058
1916	3,596
1917	3,993
1918	4,502
1919	3,879
1920	3,794
1921	2,447
1922	3,625
1923	3,455
1924	3,511
1925	2,449
1926	2,529
1927	5,748
1928	3,398
1929	4,531
1930	2,243
1931	2,078
1932	4,219
1933	3,738
1934	3,568
1935	3,370
1936	3,540
1937	5,267
1938	2,427
1939	4,332
1940	2,778
1941	2,141
1942	2,180
1943	3,879
1944	3,058
1945	3,511
1946	4,276
1947	2,860
1948	4,389
1949	2,591
1950	4,219
1951	2,628
1953	2,019
1954	2,597
1955	0
1956	0
1957	0
1958	0
1959	0
1975	4,304
1986	0
1988	0
1989	0
1990	0
1991	0
1993	1,977
1994	3,540
1995	2,679
1997	3,794
1998	2,718
1999	2,860
2000	2,302
2001	2,526
2002	3,115
2003	4,078
2004	3,766
2006	2,144
2007	1,838
2008	2,483
2009	2,945
2010	5,324
2011	2,053
2012	1,974
2013	2,481
2014	1,798
2015	2,158
2017	2,036
2018	2,693
2019	3,568
2020	2,917
2021	3,568
2022	2,724
2024	3,200
2025	2,820

Sample rows

12 of 483,211 rows Taken from the start of observations-river-flow-usgs-nwis.csv. These are sample rows, not the dataset.

Sample rows from the United States Flood Peak Dataset extract
`metric_id`	`subject_id`	`subject_kind`	`value`	`censored`	`censor_operator`	`censor_threshold`	`censor_limit`	`censor_reported`	`unit`	`period_start`	`period_end`	`period_kind`	`basis`	`confidence`	`source_id`	`source_ref`	`qualifier`	`superseded`	`retrieved`
catchment-area	gauge-usgs-01010500	geography	6980.017957	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01010500, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01011000	geography	3180.505399	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01011000, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01013500	geography	2253.289656	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01013500, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01014000	geography	14700.772514	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01014000, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01018500	geography	1061.895125	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01018500, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01019000	geography	580.157337	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01019000, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01020000	geography	3405.834365	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01020000, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01021500	geography	1186.214555	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01021500, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01023000	geography	383.31824	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01023000, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01029500	geography	2812.727088	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01029500, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01031500	geography	766.636481	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01031500, contributing drainage area	contributing area	—	2026-08-31
catchment-area	gauge-usgs-01033000	geography	841.746136	—	—	—	—	—	km²	2026-08-31	—	instant	reported	high	usgs-nwis	USGS site 01033000, contributing drainage area	contributing area	—	2026-08-31

Download these rows as CSV — the same twelve, with the column header, for testing a loader before committing to the full file.

Methodology

Collected from the USGS peak streamflow service, which publishes one instantaneous maximum per water year for each station together with the codes that qualify it. Those codes decide what may be held. A peak the publisher marks as itself a maximum daily average is not an instantaneous peak and is not held under this measure, because a daily average is systematically smaller than the peak of the same day and a series mixing the two understates every flood it covers. A peak marked as bounded rather than measured — the discharge known only to be above or below a figure, because backwater or an out-of-range stage made it unmeasurable — is held as a censored observation carrying that bound rather than as a number nobody measured. Peaks the publisher marks as estimates are held, because an estimated peak is its best account of a real flood, and the qualification travels with the series. Where a water year lists more than one peak, the larger is that year’s maximum. A peak dated to an unknown day is not held, because placing it on the first of the month would be inventing the date of a flood. Cubic feet per second are converted to cubic metres per second and the publisher unit is recorded alongside.

What this dataset cannot tell you

Read before relying on these figures

Rating curves are extrapolated at extreme discharges, so the largest value in a peak series carries the largest measurement uncertainty. That is inherent to the measurement and applies to every published annual maximum series, not to this one particularly.
A peak belongs to a gauge on a reach, never to a river as a whole. Two gauges on the same river can differ by an order of magnitude.
This is not a flood frequency analysis and Hydrionis publishes no return periods. A return period is a statistical claim requiring a stated distribution, a stated fitting method and a sample long enough to support both, and presenting one without those would put a confident number on the least certain part of the record.
Some peaks are affected by regulation, diversion or dam failure, and some occurred before the station was operated systematically. The publisher’s codes record this; a peak series is a record of what happened at a gauge, not of what a natural catchment would have produced.
Coverage is the United States. The absence of a gauge elsewhere is an absence of collected peak data rather than an absence of floods.

Coverage

Subjects	United States Geological Survey streamgauges whose daily discharge record opens before 1930 and spans at least fifty years — 2,236 held
Period	1806–2025
Units	m³/s
Update cadence	Annually
Identifier	`us-flood-peaks` v1.0.0

Fields

Data dictionary →

Every metric this dataset carries, with the unit it is stored in and the geography level it is reported at. Full definitions are in the data dictionary.

Identifier	Field	Unit	Reported at
`river-flow-annual-max`	Annual maximum river flow The highest instantaneous discharge recorded at a gauging station in a water year.	m³/s	river-gauge

Rights and access

Licence position

Open. Derivative use true; redistribution true; commercial reuse true.

Applies to material authored by US federal agencies. Third-party material carried within a federal product may be separately protected.

Permitted uses

Public display
Derived analysis
API delivery
Bulk download
Commercial use

Available to: Public, Developer, Research, Professional, Enterprise. These follow from the rights of the contributing publishers, not from a Hydrionis pricing decision — no access tier can widen what a publisher withheld.

Contributing sources

United States Geological Survey — National Water Information System (US Government work — public domain)

Attribution and citation

Cite and reuse →

Two different obligations, deliberately kept apart. One is a condition of somebody else's licence; the other is a request Hydrionis has no standing to enforce.

Suggested citation for Hydrionis

Hydrionis (2026), United States Flood Peak Dataset, version 1.0.0, accessed 2026-09-02, hydrionis.com/data/datasets/us-flood-peaks/

Cite Hydrionis where you rely on the compilation, the normalisation, the eligibility judgements or a chart. Where you use an unmodified published value, the publisher’s attribution is the one that is required; crediting Hydrionis as well is welcome and is not something Hydrionis can require of you.

Recommended, not required. Hydrionis does not claim rights over unmodified public data it relayed, and doing arithmetic on somebody else’s measurements does not make them Hydrionis’s. Cite the version and the access date: United States Flood Peak Dataset changes as publishers release and as methods are revised, and a citation without both cannot be checked later. Full guidance is on cite and reuse.

None of the contributing licences requires attribution. Crediting the publishers is still good practice and they are listed above.

Changelog

Version 1.0.0 — 2026-08-31

First release: the instantaneous annual maximum series.

Additions

175,008 instantaneous annual peaks at 2,236 stations, each dated to the day it occurred.
Peaks the publisher bounds rather than measures are held as censored observations carrying the bound.

Use this data

Download river-flow observations from USGS (483,211 rows) (CSV)
Machine-readable JSON /api/v1/datasets/us-flood-peaks.json
How to cite this, and who you must credit
How these figures were produced
Request this at a different scale, format or geography

Every bulk extract ships with a companion metadata file carrying this dataset’s version, its sources, their licences, the attribution you are required to give and the suggested citation — so the terms travel with the file rather than staying on this page.

About this data

The fields a professional user needs to judge whether this is fit for a particular purpose. Deliberately not a score.

About this data

Everything below is a property of the data rather than a rating of it. There is deliberately no single confidence score: how much a gap matters depends on what you are about to do with the series, and only you know that.

Source	United States Geological Survey — National Water Information System · US Government work — public domain
How the values came to exist	Mixed across the series — the basis is recorded on every individual value.
Period held	1806–2025
Known gaps	Rating curves are extrapolated at extreme discharges, so the largest value in a peak series carries the largest measurement uncertainty. That is inherent to the measurement and applies to every published annual maximum series, not to this one particularly. A peak belongs to a gauge on a reach, never to a river as a whole. Two gauges on the same river can differ by an order of magnitude. This is not a flood frequency analysis and Hydrionis publishes no return periods. A return period is a statistical claim requiring a stated distribution, a stated fitting method and a sample long enough to support both, and presenting one without those would put a confident number on the least certain part of the record.
Continuity	Version 1.0.0, released 2026-08-31. 1 recorded version; breaking changes and corrections are listed in the changelog above.
Last read from the publisher	2026-08-31 · publisher updates annually
Redistribution	Permitted by the upstream licences
Commercial reuse	Permitted by the upstream licences. No Hydrionis access tier can widen this.
Methodology	How this was produced