The Same Air, Five Verdicts
Give five official air-quality tables the same 24-hour average of fine particles, 58 micrograms per cubic metre, and they return five verdicts: Unhealthy in the United States, High in the UK, 良 in China, Satisfactory in India, and Very poor on the European index's daily table, which the EEA retired in July 2025. Slide the number and watch them agree on their cleanest word only from 0 to 9. Then watch an agency re-grade its own past: scored again with the US table of 2024, 14.0% of the 2023 site-days in EPA's own data file change category.
At 58: US Unhealthy, UK High, China 良, India Satisfactory, EU daily table Very poor.
Each bar is one table laid along the same 24-hour axis, its words shaded by rank within that table: the shading is this page's own, not any agency's colour. The sixth bar is the European index as it stands today. It grades a single hour, so the 24-hour slider moves no marker on it.
- All five give their cleanest word (Good, Low, 优, Good, Good) only from 0 to 9.
- From 56 to 60 the US says Unhealthy, the fourth of its six words, while China and India, which also have six, both say their second. Nowhere else on the slider are those three tables two words apart.
- Each number on a card is the PM2.5 part alone. A full report takes the worst of several pollutants, and the words are not health advice here: this page calls no level safe and no agency right.
Why the slider moves in whole numbers
The five tables do not even agree on how to round. The US rule says PM2.5—truncate to 1 decimal place
, and then Round the index to the nearest integer.
EPA's own manual rounds a half up: one of its worked examples ends 125.5 = 126
. China's standard rounds the concentration to a whole number under GB/T 8170 and then rounds every index up: 环境空气质量指数及空气质量分指数的计算结果应全部向上进位取整数，不保留小数。
(Our translation: the AQI and every sub-index are all rounded up to a whole number, keeping no decimals.) Defra says that rounding has to be done only once
and that All comparisons with the index should be undertaken with concentrations rounded to integers.
Its own illustration: For example, 50.486 μg/m³ is rounded to 50 μg/m³. Rounding this value in a first step to one decimal place gives 50.5 μg/m³ and rounding in a second step to and integer would yield 51 μg/m³.
The EU and Indian tables are printed in whole numbers with gaps between their bands, from 5 to 6 and from 30 to 31, and we found no rule for what falls in between in the documents we read.
So the slider moves in whole micrograms, where none of those rules can change a word. They still move the index numbers: the US, UK and Chinese cards compute theirs by their own agency's rule, the Indian card rounds its sub-index to the nearest whole number by this page's choice, named in the check, and the European card shows no number. One more choice is named in the check at the end: China's Table 1 starts at an index of 1, so a concentration of 0, whose index is 0, is counted here with its first level.
The agencies' own worked examples, recomputed in your browser
| Whose example | Input | They print | This page computes |
|---|---|---|---|
| EPA, Technical Assistance Document, May 2026 | PM2.5 35.9 | 102 | 101.98 → 102 |
| 40 CFR Part 58, Appendix G, equation 2 | PM10 210 | 128 | 128.22 → 128 |
| CPCB, About National Air Quality Index | PM2.5 31, 60, 45 | 51, 100, 75 | 51, 100, 74.66 → 75 |
| EEA, ETC HE Report 2024/17, Table 3.8 | daily 25, 50, 75 | 47, 93, 139 | 46.701, 92.926, 139.151 |
The tables, as this page transcribed them (PM2.5, µg/m³)
- US, from 6 May 2024: 0.0–9.0 Good (0–50); 9.1–35.4 Moderate (51–100); 35.5–55.4 Unhealthy for Sensitive Groups (101–150); 55.5–125.4 Unhealthy (151–200); 125.5–225.4 Very Unhealthy (201–300); 225.5 and above Hazardous (301 and above, 500 at 325.4).
- US, before 6 May 2024: 0.0–12.0 Good (0–50); 12.1–35.4 Moderate (51–100); 35.5–55.4 Unhealthy for Sensitive Groups (101–150); 55.5–150.4 Unhealthy (151–200); 150.5–250.4 Very Unhealthy (201–300); 250.5–350.4 Hazardous (301–400); 350.5–500.4 Hazardous (401–500).
- UK, April 2013 table: Low 0–11 (1), 12–23 (2), 24–35 (3); Moderate 36–41 (4), 42–47 (5), 48–53 (6); High 54–58 (7), 59–64 (8), 65–70 (9); Very High 71 or more (10).
- UK, January 2012 table: Low 0–11 (1), 12–23 (2), 24–34 (3); Moderate 35–41 (4), 42–46 (5), 47–52 (6); High 53–58 (7), 59–64 (8), 65–69 (9); Very High 70 or more (10).
- China, HJ 633-2026: IAQI 0, 50, 100, 150, 200, 300, 400, 500 at 0, 35, 60, 115, 150, 250, 350, 500; 优 (一级) 1–50, 良 (二级) 51–100, 轻度污染 (三级) 101–150, 中度污染 (四级) 151–200, 重度污染 (五级) 201–300, 严重污染 (六级) above 300.
- China, HJ 633-2012: IAQI 0, 50, 100, 150, 200, 300, 400, 500 at 0, 35, 75, 115, 150, 250, 350, 500; the same six categories, the first written 0–50.
- India, CPCB: 0–30 Good (0–50); 31–60 Satisfactory (51–100); 61–90 Moderately polluted (101–200); 91–120 Poor (201–300); 121–250 Very Poor (301–400); 250+ Severe (401–500).
- EU, daily table until 4 July 2025: 0–10 Good; 11–20 Fair; 21–25 Moderate; 26–50 Poor; 51–75 Very poor; 76–800 Extremely poor.
- EU, hourly table since 4 July 2025: 0–5 Good; 6–15 Fair; 16–50 Moderate; 51–90 Poor; 91–140 Very poor; >140 Extremely poor.
Two edges needed a reading. India prints Very Poor as 121–250 and Severe as 250+, so 250 sits in both; its sub-index there is 400, the top of Very Poor, and this page puts it in Very Poor. Severe has no upper breakpoint, so above 250 the Indian card shows the word and no number. Chinese category names are the standard's own; the glosses beside them are ours, because HJ 633-2026 gives no English names.
The fifth table now counts hours
Until 4 July 2025 the European Environment Agency graded PM2.5 on a day, as the other four tables on this page still do: its index used the 24-hour running means for the past 24 hours for PM10 and PM2.5
. That day it announced: Hourly data is now used for all key pollutants to estimate the index, including particulate matter pollutants (PM2.5 and PM10), which were presented until now as 24-hour averages.
A single hour and a day's average are different measurements of the same air, so the current European table has its own ruler and its own slider here, and never a card on the 24-hour axis.
One hour at 40: Moderate, word 3 of 6 on the European index since 4 July 2025.
The EEA did not simply relabel its old bands. Its technical report set Good and Fair at the World Health Organization's 2021 guideline values for PM2.5, 5 for a year and 15 for a day, and used them for single hours as they stand: the long-term guideline level marks the change between “good” and “fair”, and the short-term guideline level marks the change between “fair” and “moderate”
. For the three bands above, it kept its old daily edges, 25, 50 and 75, and converted them to hours with a straight line fitted to European stations from 2018 to 2022, between a day's mean and that day's highest hour: 1.849 * dailyMean + 0.476
. That gives 47, 93 and 139. The published table rounds 93 and 139 to the nearest ten, 90 and 140, and moves 47 to 50 to coincide with an information threshold and an alert threshold in the revised EU Ambient Air Quality Directive.
Top: a day's mean. Bottom: an hour. Lines show where the EEA's own function sends its old daily edges (25, 50, 75 to 47, 93, 139) and where the published hourly edges sit (50, 90, 140): open circles mark where the function lands, filled squares the published edges. Good and Fair (5, 15) drop straight down, unconverted, on dashed blue lines. Put through the same function they would have become 9.7 and 28.2: that is this page's arithmetic, not the EEA's.
A 24-hour mean is not one thing either
Even the shared axis hides choices. For real-time reporting EPA uses its NowCast: The PM NowCast is an average of the previous 12 hours.
When the air is changing, the latest hours weigh more. Defra grades the daily mean concentration for historical data, latest 24 hour running mean for the current day
, and needs 18 of a day's hourly means to make a daily one; CPCB needs 16. Since 1 March 2026 China's real-time report grades a single hour against its daily column: for a 1-hour PM2.5 value, its formula says 取表3中日平均的高位值
(our translation: take the upper breakpoint from the daily-mean column of Table 3), one of the changes the new standard lists: 调整了实时报中颗粒物的空气质量分指数计算方法
(our translation: adjusted how the real-time report calculates the sub-index for particles). Apart from the European hourly slider above, this page grades PM2.5 only as a 24-hour mean: a day's mean on the US, UK, Chinese and Indian tables, and on the European daily table the running 24-hour mean the EEA graded every hour until 4 July 2025.
The obvious reply, and each agency against its own past
The obvious reply is that every country has its own scale and its own words, so of course they differ. But the scales are not fixed inside one agency either, and when a table changes, the record can follow the table, not the air: EPA's does, and in 2013 Defra said historic index values should be recalculated. Move this slider and read each agency against its own previous table.
At 70: the UK and China words differ between tables; the US word does not.
- United States. The final rule of 6 March 2024 took effect on 6 May 2024 and lowered the top of Good from 12.0 to 9.0, and the tops of Unhealthy and Very Unhealthy from 150.4 and 250.4 to 125.4 and 225.4. On the slider's 301 whole numbers the word changes at 53: 10 to 12, 126 to 150, and 226 to 250.
- United Kingdom. In April 2013 Defra moved the start of the Moderate, High and Very High bands up by one microgram (its report:
The minimum concentration for the moderate band has increased by one µg/m3 to ensure that the low/moderate division is compatible with the Limit Values for PM10, SO2 and NO2
), so a few readings got gentler words. The index changes at 35, 47, 53 and 70, the band at 35, 53 and 70. The April 2013 report then said:The revised DAQI is shown in Table 1 and replaces that launched in January 2012. Historic index values should be recalculated to the point when the DAQI was first implemented.
We did not find out whether the archive was recalculated; that is the instruction, not an outcome. - China. HJ 633-2012 put an IAQI of 100 at 75; HJ 633-2026, in force from 1 March 2026, puts it at 60, so 61 to 75 moved from 良 to 轻度污染. The ministry repealed the old standard the same day:
以上标准自2026年3月1日起实施。自标准实施之日起，《环境空气质量指数（AQI）技术规定（试行）》（HJ 633—2012）和《环境空气质量评价技术规范（试行）》（HJ 663—2013）废止。
Of the air-quality standard it revised alongside, GB 3095-2026, an official of its Department of Atmospheric Environment said, answering reporters' questions:《标准》修订会导致达标城市比例等评价结果有所下降，但是这种下降并不意味着空气质量变差
(our translation: the revision of the Standard will cause assessment results, such as the share of cities meeting it, to fall somewhat, but the fall does not mean that air quality has got worse). - European Union. The EEA changed what it grades, from a day to an hour, so its old daily words and its new hourly ones cannot be set side by side value for value at all.
The record follows the table: EPA's own 2023 file, graded twice
EPA says of its AirData service: Yes, every AQI summary from AirData is based on the current AQI breakpoints, regardless of year.
So the 2023 PM2.5 figures it serves today carry index values from the table that took effect in May 2024. This page's counts come from one of EPA's two daily files of PM2.5 mass for 2023, daily_88101_2023.zip, which EPA lists with 847,162 rows as of 25 June 2026. EPA's heading for it is PM2.5 FRM/FEM Mass; the other file, daily_88502_2023.zip, holds PM2.5 from monitors of other kinds and is not used here. The check at the end of this page, run with --deep, reads the file and first confirms that it computes EPA's arithmetic: on every one of the 452,388 rows that carry an AQI and are not marked as excluded events, its index equals EPA's.
Taking each monitoring site's highest value each day gives 306,215 site-days. Graded with the table in force before 6 May 2024 and again with the one after, 42,901 of them (14.0%) get a different word: 42,795 move from Good to Moderate, 98 from Unhealthy to Very Unhealthy, and 8 from Very Unhealthy to Hazardous. Good site-days fall from 256,184 to 213,389. The air of 2023 did not change.
| Word | Before 6 May 2024 | From 6 May 2024 |
|---|---|---|
| Good | 256,184 | 213,389 |
| Moderate | 47,441 | 90,236 |
| Unhealthy for Sensitive Groups | 1,528 | 1,528 |
| Unhealthy | 958 | 860 |
| Very Unhealthy | 103 | 193 |
| Hazardous | 1 | 9 |
42,901 of 306,215 site-days (14.0%) change word.
The counts rest on choices, each named in the check below. They are monitor-site days, not days in the United States, and a different but reasonable choice gives a different count: keeping the original T640 rows gives 46,268 of 306,422 (15.1%).
Coda: the agencies' own arithmetic
Recomputing the agencies' worked examples turns up two printed figures that do not follow from the documents' own rules. Both are about ozone, not PM2.5, and neither changes a word. EPA's Technical Assistance Document (May 2026) works a 1-hour ozone value of 0.162 ppm and concludes: In this case, the index is 148 (the maximum of 148 and 126) and the main pollutant is O3.
Its own equation gives 49 ÷ 0.039 × 0.037 + 101 = 147.487, which its rule rounds to 147. The September 2018 edition prints the same 148. Both 147 and 148 are Unhealthy for Sensitive Groups.
The worked example written into the regulation itself, 40 CFR Part 58 Appendix G, reads an 8-hour ozone value of 0.130 ppm as iii. Find the breakpoints for 8-hour O3 at 0.130 ppm as 0.116 ppm and 0.374 ppm, corresponding to index values 201 and 300;
and ends vii. Find the maximum, 206. This is the AQI.
The same appendix's Table 2 puts 0.130 in its 0.106 to 0.200 row, which gives 226.277, rounded to 226. Both 206 and 226 are Very Unhealthy.
What these words are not
Each agency says where its edges come from, and they are different things. The US rule says: In most cases, the index value of 100 is associated with the numerical level of the short-term standard (i.e., averaging time of 24-hours or less) for each pollutant.
China's 100 at 60 matches the transitional 24-hour limit of 60 that GB 3095-2026 applies from 1 March 2026 to 31 December 2030. CPCB set India's first two edges with Indian background levels in mind: the first two categories, Good and Satisfactory, the breakpoints are kept as 30 and 60 µg/m3
. The UK index, Defra says, was recommended by the Committee on the Medical Effects of Air Pollutants (COMEAP). The EEA took its first two edges from WHO guideline values. None of this makes one table right, and this page offers no health advice and calls no level safe.
The US table is the one in force as of 1 November 2026. The 2024 rule that carries it was challenged in court; on 26 June 2026 the US Court of Appeals for the D.C. Circuit denied the petitions for review and the EPA's own motion to vacate the rule. Two petitions asking the Supreme Court to review that decision, Nos. 26-277 and 26-298, were pending on 1 November 2026, with responses due by 17 November 2026.
The check
Every verdict, index, count and percentage on this page comes from engine.mjs and the tables typed into it, and the 2023 counts from airdata-2023.json, a histogram this page's authors made from EPA's file. Your browser recomputes the instruments, the worked examples and the record as the page loads. One program checks all of it:
curl -sO https://artwaste.land/checks/verify-the-same-air-five-verdicts.mjs node verify-the-same-air-five-verdicts.mjs
It types the nine tables again from the sources, grades all 301 whole numbers with a second, plainer implementation that shares no code with the engine, and finds every figure on this page in its own sentence. It checks the quotations against quotes.json, which holds each passage with its source. With --deep it downloads EPA's zip, checks its sha256, and redoes the 2023 recount from the raw rows. With --mutate it plants wrong figures in a copy and must fail.
The free choices, each of which moves a result:
- Whole numbers only, from 0 to 300, so that no unstated rounding rule matters. China's index 0 is counted with its first level.
- India's 250 is put in Very Poor (its sub-index there is 400), and India's index number is rounded to the nearest whole number, because the CPCB documents we read state no rule; its own example, 45 to 75, fits rounding to the nearest and rounding up alike.
- The European card uses the daily table the EEA used until 4 July 2025. The current hourly table is shown only on its own hourly ruler.
- Chinese category names are glossed in English by this page.
- The 2023 recount: a site-day is one monitoring site on one date, valued at the highest daily mean among its rows that carry an AQI. Rows marked as excluded events are dropped, because each also appears as included. A negative daily mean, as on 1,376 of the rows that carry an AQI, is read as 0.0, the bottom of the table. The original T640 and T640X rows are dropped, as EPA's advisory recommends.
- The zip this page used has sha256
4b40502e8147419643a1120c666c4aa952633094f21f65475844172bcc78833band was downloaded on 1 November 2026. EPA refreshes these files twice a year, so a later copy may count differently.
What it does not check: that the documents say what this page's copies of them say (follow the links below), that HJ 633-2026's image-only pages were read correctly, whether the UK archive was ever recalculated, or anything about health.