U.S. Air Quality
EPA monitor data through 2025 · Updated August 13, 2026
American air is far cleaner than it was in 2000. Across the 35 metropolitan areas EPA has tracked longest, unhealthy-air days fell about 80% between 2000 and 2025. But the official record is harder to read than it should be: the scoring rules changed twice without being marked on any chart, most areas missed their legal deadlines for cleaner air, and in the East the long-reliable link between cutting emissions and lowering pollution has broken down. These three pieces work through what the raw monitor data actually shows.
Three investigations
Is air quality getting better or worse?
EPA changed what an Air Quality Index of 100 means, in 2013 and again in 2015. The same 2010 air from the same monitors scores 632 unhealthy days under the old rules and about 1160 under today's. Both scorings, every year, 2000–2025, including the detail that keeps it honest: one of those changes fixed a seven-year inconsistency in EPA's own reporting.
Did the 2015 ozone standard work?
EPA tightened the ozone standard to 70 parts per billion and the Clean Air Act attached legal deadlines. 83% of the first cohort missed 2021 and 96% of the second missed 2024. Against that, most of the modelled health benefit was delivered. Both are true, and the reason turns on the difference between emissions and the air people breathe.
Why has ozone stopped improving?
Eastern nitrogen oxide emissions fell 44% after 2014, about the same rate as the decade before, with measured power-plant emissions down 60%. Ambient ozone barely moved. The obvious explanation, that the cuts ran out, is false, and the data rules it out.
One distinction runs through all three. Emissions are the tons of pollution released, which is what regulation controls. Ambient levels are the concentration in the air people breathe, which is what health standards are written in. Ozone is not emitted at all: it forms in the atmosphere from other pollutants reacting in sunlight, so the two are linked indirectly and not proportionally. Conflating them is the most common error in reporting on air quality, and it is what makes the recent record look contradictory when it is not.
The emissions behind it
Everything above measures ambient levels: what monitors detect in the air. This measures emissions: what comes out of stacks and tailpipes. Burning fuel releases the pollutants regulated for air quality, so emissions fall when combustion gets cleaner or when less fuel is burned. They are the input; ambient levels are the result, and the two have come apart.
| Pollutant | 2002 | 2024 | Change |
|---|---|---|---|
| SO₂ | 12,742 | 897 | -93% |
| NOₓ | 7,493 | 2,263 | -70% |
| PM10 | 1,425 | 900 | -37% |
| PM2.5 | 1,103 | 838 | -24% |
| CO | 4,604 | 4,730 | +3% |
| VOC | 539 | 630 | +17% |
Thousand tons a year, fuel-combustion source categories only. Ozone is not on this chart because nothing emits it: it forms in the atmosphere from nitrogen oxides and volatile organic compounds, both of which are emitted. That is precisely why falling emissions have stopped producing falling ozone. Fuller treatment on U.S. energy facts.
The longer record, 1970 to 2024
EPA's record starts in 1970 for five of the six, and in 1990 for fine particulates, which were not measured before then. Indexed to 1990 = 100 so the shape of each decline is visible: on an absolute scale sulfur dioxide, at more than twenty times the smallest series in 1970, flattens the rest into the axis. Values above 100 are years before the base.
| Pollutant | Record starts | First year | 1990 | 2002 | 2024 | Since first year | Since 1990 |
|---|---|---|---|---|---|---|---|
| SO₂ | 1970 | 23,456 | 20,290 | 12,742 | 897 | -96% | -96% |
| NOₓ | 1970 | 10,061 | 10,894 | 7,493 | 2,263 | -78% | -79% |
| VOC | 1970 | 721 | 1,005 | 539 | 630 | -13% | -37% |
| PM10 | 1970 | 2,871 | 1,196 | 1,425 | 900 | -69% | -25% |
| CO | 1970 | 4,632 | 5,511 | 4,604 | 4,730 | +2% | -14% |
| PM2.5 | 1990 | 909 | 909 | 1,103 | 838 | -8% | -8% |
Four of these lines contain breaks in how EPA estimates them, not changes in the air. Volatile organic compounds fall 55% in 2002, fine particulates rise 103% and coarse particulates 75% in 1999, carbon monoxide rises 63% in 1980 and falls 35% in 1990. Those steps are accounting revisions and should not be read as emissions moving. Sulfur dioxide and nitrogen oxides carry no such break, which is why the indexed chart above starts in 2002 and why these two are the ones EPA measures at the stack rather than re-estimating. Sulfur dioxide is the largest single air-quality gain in the record: scrubbers, the acid-rain trading program, low-sulfur fuel rules and the shift away from coal-fired power.
How to read the VOC line
Volatile organic compounds end the chart higher than they started, which invites the conclusion that this pollution is getting worse. Three things in the data say otherwise, and they are worth separating because VOCs are one of the two ingredients ozone forms from.
Fuel combustion is a small part of the VOC problem. It was 4.9% of EPA's all-source national VOC total in 2024, against 3.1% in 2002. The national total fell 26% over those years, from 17,333 to 12,783 thousand tons, as highway vehicles fell 81% and off-highway engines fell 65%. Nothing here says U.S. VOC emissions rose. They fell by roughly a quarter.
Inside fuel combustion, the sources disagree. Power plants fell 44% and industry fell 22%. The entire increase, and more, sits in one category: residential and commercial buildings, up 43%. That single category is what pulls the combustion total up 17%.
And that category is an estimate, not a measurement. EPA ties the movement in it to residential wood burning, which no monitor counts: it is modelled from fuel use and emission factors, both of which EPA has revised. Residential wood energy use fell 9% between 2002 and 2024 in EIA's fuel data, so the fuel burned did not move anything like the emissions charged to it. Read the rising line as an inventory revision more than a change in the air. Carbon monoxide, the other line that rises, splits the same way: residential and commercial buildings up 32% while power plants and industry fell.
Sources: EPA national tier-1 emissions inventory, source-category detail, and EIA Monthly Energy Review series WDRCBUS for residential wood. The ambient side of the VOC question, what monitors actually record, is in why ozone stopped improving.
Look up a place
The same data, one page per geography: days above AQI 100 every year since 2000 under both rulebooks, what drove the bad days, and how much of the year was actually monitored.
- All 52 states and territories, ranked by unhealthy days. Each state page also lists its metro areas.
- 456 metropolitan areas, ranked. Metro areas are the geography EPA actually reports the index for.
- Download the full dataset as CSV, free to reuse with attribution.
How this was built
Every figure is recomputed from EPA's raw monitor records rather than copied from a summary table, because the central question, what the same air looks like under two rulebooks, cannot be answered from published summaries. The calculation is checked against EPA's own published numbers before use: 94–98% exact on annual metro counts, 716 of 716 counties on ozone design values. Methods, limits and every correction we have made.