The AQI is a definition: China vs US breakpoints

Thirty-five micrograms of PM2.5 per cubic metre sits exactly on China’s line between an “excellent” day and a “good” one: an AQI of 50. The US index scores the same air 99, at the top of “moderate”. Both come from a table, and both are arithmetically correct.

An index is a definition.

Two tables, one concentration

China’s index comes from HJ 633—2026, the United States’ from the breakpoints EPA republished in May 2026. Both turn a 24-hour PM2.5 concentration into an index by interpolating in straight lines between fixed anchors; the anchors differ.

China’s anchors for fine particles are index 50 at 35 µg/m³, 100 at 60, 150 at 115, 200 at 150, 300 at 250, 400 at 350 and 500 at 500; its PM10 anchors are 50, 120, 250, 350, 420, 500 and 600 (HJ 633—2026, approved 2026-02-14, in force 2026-03-01, Table 3 on printed page 3, read 2026-10-12). The US anchors are 50 at 9.0 µg/m³, 100 at 35.4, 150 at 55.4, 200 at 125.4, 300 at 225.4 and 500 at 325.4 (AQI technical assistance document, May 2026, Table 6, read 2026-10-12).

Index 50 sits at the transitional grade-I daily limit for PM2.5, 35 µg/m³, and index 100 at the transitional grade-II daily limit, which the 2026 revision moved from 75 to 60 µg/m³ (GB 3095—2026, Table 1 on printed page 3; MEE Q&A, 2026-02-24). EPA moved its good/moderate line to 9.0 in 2024, when the annual standard moved to 9.0 µg/m³, and left the next line at 35.4 because the 24-hour standard stayed at 35 (EPA fact sheet, February 2024, effective 2024-05-06).

The US truncates a concentration to one decimal place and rounds the index to the nearest integer; China rounds up, keeps no decimals, on a scale of 1 to 500 (technical assistance document, May 2026, section IV; HJ 633—2026, 4.2.6).

What changed in China on 1 March 2026

On 2026-02-14 the environment ministry and the market regulator issued GB 3095—2026, the ambient air quality standard, in one announcement and HJ 633—2026 with HJ 663—2026, the companion technical rules, in a second. Both take effect on 2026-03-01, and GB 3095—2012 with HJ 633—2012 were abolished that day (公告 2026年第13号, 2026-02-14; 公告〔2026〕14号, 2026-02-14).

The limits arrive in two stages: from 2026-03-01 to 2030-12-31 the transition values apply, for PM2.5 30 µg/m³ annual and 60 daily at grade II and 15 and 35 at grade I; from 2031-01-01 the full limits apply nationwide, 25 annual and 50 daily at grade II and 10 and 25 at grade I (GB 3095—2026, Table 1; MEE Q&A, 2026-02-24). The ministry calls these comparable to the WHO 2021 second interim target, 25 µg/m³ annual and 50 for 24 hours (WHO global air quality guidelines, 2021-09-22, Table 0.1).

Because the index is bound to the limits, tightening moved numbers. The PM2.5 value of 100 now sits at 60 µg/m³, down from 75, and the PM10 anchor at index 100 moved from 150 to 120 (HJ 633—2012, Table 1, abolished 2026-03-01). A reading of 65 µg/m³ scored 88 and read “good” under the old rule; it scores 105 and reads “light pollution” now, on unchanged air. The revision also adds a sensitive-group list by pollutant, removes the definition and determination method of the non-attainment pollutant, and adds data-completeness rules (HJ 633—2026, 前言, 5.3).

Which number, from which station, how often

China. HJ 633—2026 allows the index to be calculated per monitoring station, per county-level unit or per city. The daily report covers a calendar day and must be public by 12:00 the next day. The real-time report covers one hour, may be published after each hour with at most one hour of lag, and is built from the current one-hour averages of all six pollutants — including PM10 and PM2.5, whose hourly values go into the daily-average breakpoints of the table (HJ 633—2026, 4.2.1, 4.2.3, 4.2.4, 5.2). The 2012 edition carried the rolling 24-hour averages of both particle sizes among its real-time columns (HJ 633—2012, 5.1.3), and the 2026 revision changed the real-time particle sub-index calculation. Nationally, the monthly reports cover 339 cities at prefecture level and above (MEE monthly report, July 2025 data, published 2025-08-25), and the 2025 national PM2.5 average was 28 µg/m³, with 89.3 per cent of days rated good or excellent (Xinhua via gov.cn, 2026-01-16).

United States. Metropolitan statistical areas above 350,000 people must report an AQI daily, seven days a week on days with data. The required report is a full 24-hour value, in effect yesterday’s, from the highest monitor in the reporting area rather than an average. Hourly reporting is recommended, not required; AirNow’s NowCast averages the previous 12 hours for particles, with a weight factor capped between 0.5 and 1, valid on at least two of the last three hours, last changed for PM2.5 on 2013-08-01 (technical assistance document, May 2026, sections II, IV). EPA’s FAQ notes that monitors within one area can land in different NowCast categories.

What the mask labels are defined against

GB 2626—2019 is China’s mandatory standard for non-powered particle respirators, issued 2019-12-31, still current, with 2020-07-01 recorded as its implementation date (national standards platform, read 2026-10-12). It splits filters by aerosol: KN filters for non-oily particles, tested with sodium chloride, and KP filters for oily and non-oily particles, tested with an oil aerosol. The levels are KN90 at 90.0 per cent or better, KN95 at 95.0 per cent and KN100 at 99.97 per cent, the same numbers for KP; the national standards institute says the 2019 edition left classification, marking and filtration levels unchanged from 2006 (CNIS, 2020-02-12).

NIOSH certification under 42 CFR 84 states its conditions: twenty filters per model; N-series filters challenged with sodium chloride aerosol at 30 ± 10 per cent relative humidity, after 25 ± 1 hours of preconditioning at 38 ± 2.5 °C and 85 ± 5 per cent humidity; a single filter at 85 ± 4 litres per minute, the challenge running until at least 200 ± 5 mg of aerosol has reached it. Efficiency floors: 95, 99 and 99.97 per cent (42 CFR 84.174, read 2026-10-12).

The European standard for filtering half masks tests penetration at 95 litres per minute against sodium chloride and paraffin oil, with maximum penetration of 20 per cent for FFP1, 6 per cent for FFP2 and 1 per cent for FFP3 (EN 149:2001, Table 1).

Medical masks are a separate family: GB 19083—2010 sets three filtration grades against non-oily particles — 95, 99 and 99.97 per cent — with synthetic-blood penetration and fit items, and YY 0469—2011 requires bacterial filtration efficiency of at least 95 per cent and non-oily particle filtration of at least 30 per cent (CNIS, 2020-02-12; a provincial drug regulator, 2021-08-20). GB/T 32610—2016, the daily protective standard, is keyed to the index: protection levels A to D, each tied to an AQI category, with a protection effect of at least 90, 85, 75 and 65 per cent (CNIS, 2020-02-12).

The Chinese and US protocols use the same challenge aerosol and flow — 0.075 µm sodium chloride at 85 litres per minute — but differ in conditioning and loading. Three planar masks put through all three protocols came out in different classes: two met FFP1 under EN 149, one reached KN90 under GB 2626—2019, none reached N95 under NIOSH (Wang et al., 2022).

Where these numbers come from

Primary documents, read on 2026-10-12.

• China’s index and limits: HJ 633—2026 and GB 3095—2026, in force from 2026-03-01, plus the two announcements and the Q&A page for the phase dates. Both PDFs are page images without a text layer, so the tables were read from rendered pages (printed page 3 in each; HJ 633—2026’s Appendix A is on page 6) and the same cells read twice, full page and crop; both readings agreed on every PM cell reported here.
• The comparison column: my arithmetic. Neither column is copied. I interpolated linearly between the two anchors surrounding each concentration and applied each side’s rounding rule — up to an integer for China, nearest integer for the US after truncating to one decimal place. At 25 µg/m³: China 25 ÷ 35 × 50 = 35.7, up to 36; the US 51 + (100 − 51) ÷ (35.4 − 9.1) × (25 − 9.1) = 80.6, to 81. The 65 µg/m³ case: 87.5 → 88 under the 2012 table, 104.5 → 105 under the 2026 one.
• Mask standards: 42 CFR 84.174 and the EN 149 preview read in full; the Chinese standard levels from the standards platform and the national standards institute’s survey, cross-read with a provincial drug regulator.
• Scope: none of this is guidance on whether to wear anything; the question is what each standard specifies and how it is measured.

What I could not check

• The GB 2626—2019 and GB/T 32610—2016 texts themselves. The standards platform’s viewer would not load; their levels come from the national standards institute’s survey, and the 85 litres per minute and 0.075 µm details from a peer-reviewed comparison of the three protocols.
• Whether the implementation date was postponed. A 2020 notice extending the transition to 2021-07-01 is reported by testing bodies; the notice itself was not reachable, and the standards platform still lists 2020-07-01. 2021 is unverified.
• The EN 149 amendment. I read EN 149:2001; the +A1:2009 edition usually cited was not reachable.
• My reading of the scanned HJ 633—2026 and GB 3095—2026 tables. Two readings agreed on every cell, but the figures were transcribed by eye, not extracted as text.
• What happens to the AQI table when the 2031 limits take effect. The published anchors match the transitional limits (35 and 60 µg/m³ at index 50 and 100); the 2031 grade-II daily limit is 50, and nothing I found says whether the index table changes then.
• EPA’s own breakpoint page, which returned an error on three attempts today; the US anchors come from the May 2026 technical assistance document.
• Live values and platform behaviour. I read the labels of China’s public real-time page, not a snapshot, so no hourly figure can sit beside the daily one for the same hour; how provinces choose among the three calculation levels was not tested.
• Anything about where I live. Deliberately absent.