Combined sewer overflows · 2005–2026
The sewage stopped going into one river. It kept going into another.
Every overflow DC Water has reported to the EPA since 2005 — 4,480 records from 83 quarterly filings. Volumes are down 57% overall. Almost all of that is the Anacostia.
What a combined sewer overflow is
About a third of Washington — the older core — has a single pipe network carrying both sewage and street runoff. In dry weather all of it goes to the Blue Plains treatment plant. When it rains hard, the combined flow exceeds what the pipes can carry, and relief structures discharge the excess into the nearest river rather than letting it back up into streets and basements.
There are 47 such outfalls on the Anacostia, the Potomac and Rock Creek. DC Water measures or models the volume from each one and reports it to the EPA every quarter under its discharge permit. This page is built from every one of those reports since 2005.
Twenty-one years of overflow volume
Each bar is one year, in millions of gallons, split by receiving river. Hover or focus any year for detail.
- Anacostia
- Potomac
- Rock Creek
2018 and 2019 are each missing one quarter — those two reports are scans DC Water published with unreadable text, and no clean copy exists. 2026 covers Q1 only.
How often, not just how much
Volume and frequency tell different stories. Anacostia overflow events fell from 417 a year to three. Potomac events have barely moved.
A single large storm can produce dozens of separate outfall events. The consent decree's design target for the Potomac is four events a year.
It is not just that it rained less
Overflows happen when rain arrives faster than the system can carry it away. That means storms cause overflows, not rainfall in general — a quarter-inch a day for a week is handled comfortably; the same water in one afternoon spills.
Testing that against the record: across 52 quarters before the tunnels opened, here is how much of the variation in overflow volume each rainfall measure explains.
Share of variation explained (R²), 2005–2017, n = 52 quarters.
Storm rain explains 61%. Total rainfall explains 40%. The number of rainy days explains 1% — drizzle is effectively invisible to a combined sewer.
Adjusted for storms, the conclusions hold. Measured as gallons spilled per inch of storm rain, the Anacostia is down 90% and the Potomac is up 5%. Neither result is a weather artefact. 2025 was a quiet storm year — 20% below recent average — so the Potomac's best-looking year is partly flattered by that.
Measured against what the decree allows
The consent decree sets a numeric limit for each river: the overflow volume permitted in an average year once the system is finished. Against 2025's reported figures:
| River | Allowed (MG/yr) | 2025 actual | Reduction achieved | Required | Status |
|---|
Targets from Table ES-3 of the Long Term Control Plan, attached to the 2015 amendment to the consent decree. They describe an average rainfall year at full implementation, so a single year over or under is not by itself a violation.
How much to trust these numbers
Everything here comes from DC Water's own reports. Two checks were run on the extraction before anything was published.
Internal consistency. Each report prints subtotals per river. Every parsed report reconciles against its own subtotals, so a dropped or double-counted outfall shows up as a failure rather than passing silently.
An independent second source. DC Water republishes the same tables in a separate annual report series, typeset differently. Parsed independently: 19 of 19 overlapping quarters agree, 17 of them exactly. The two small differences are rounding in the source documents.
Most of these numbers are estimates, not measurements. DC Water meters only a handful of outfalls. For the rest it runs a computer model of what it believes the pipes did during a storm. In 2025 that meant 5.4 million gallons measured and 248 million modelled — so effectively every Potomac figure on this page is DC Water's model of DC Water's own performance, and nobody independently checks it.
Two further limits: two quarters (2018 Q4, 2019 Q1) are missing because DC Water published them as unreadable scans, and rainfall is measured at a single gauge — Reagan National — for all three watersheds.
Sources: DC Water quarterly Combined Sewer Overflow operations reports, 2005–2026, filed with the US EPA under NPDES permit DC0021199; Long Term Control Plan targets from the 2015 First Amendment to the consent decree; rainfall from NOAA GHCN-Daily station USW00013743. All source documents and datasets →