Air Quality Index Calculation: How the AQI Is Computed
Last month Priya, an environmental consultant working in Dubai, opened two air quality apps for the same city on the same afternoon. One showed a US AQI of 78. The other showed 42.
She checked the timestamps and the station list twice. Neither app was broken.
If you’ve ever compared a sensor dashboard against a government report, you’ve hit this wall. “The AQI” sounds like one number produced by one method. It’s not. There is no single air quality index calculation; there are several, and they disagree.
Several national and regional indices are in active use, each with its own breakpoints, averaging windows and colour bands. Two of the largest systems, the US EPA scale and China’s national index, changed their rules in 2024 and 2026 respectively.
This guide walks through the air quality index calculation from the ground up. By the end you will be able to compute an AQI from a pollutant concentration by hand, read a breakpoint table critically, work out which index a dashboard is actually using, and see why the sensor feeding the calculation matters as much as the arithmetic.
We will cover the interpolation formula, the breakpoint tables and the 2024 revision, averaging windows and NowCast, the dominant-pollutant rule, a four-way index comparison, and the data chain from a field sensor to a published number. Explore how outdoor IoT air quality monitoring stations produce those raw readings →
What the Air Quality Index Actually Is
The Air Quality Index is a dimensionless, single-number summary of air quality. It takes several measured pollutant concentrations, converts each one into a sub-index through a fixed set of breakpoints, and maps the result onto a colour-coded scale that communicates health risk at a glance.
The reference pollutants are PM2.5, PM10, ozone (O₃), carbon monoxide (CO), sulfur dioxide (SO₂) and nitrogen dioxide (NO₂). Not every country uses all six, and not every station measures all six. That gap alone is one reason two sources can disagree about the same city.
The AQI isn’t a concentration. An AQI of 100 doesn’t mean 100 µg/m³ of anything. It’s a position on a ruler, and the ruler is drawn differently in different places.
It’s also not a health diagnosis, and it isn’t directly comparable across borders without a conversion. The chemistry behind those readings is its own subject: gas sensors for outdoor air quality monitoring cover that layer. Here we stay on how the number is built.
The Air Quality Index Calculation Formula: Piecewise Linear Interpolation
Every major index uses the same core maths: piecewise linear interpolation. You locate the concentration between two breakpoints, then scale it onto the index values that correspond to those breakpoints.
The formula is:
Ip = [(IHi - ILo) / (BPHi - BPLo)] x (Cp - BPLo) + ILoEach term has a specific job:
- Cp is the concentration of the pollutant, truncated (not rounded) to the precision of the breakpoint table.
- BPLo and BPHi are the concentration breakpoints that bracket Cp.
- ILo and IHi are the index values that bracket the result.
- Ip is the resulting sub-index for that pollutant.
“Piecewise linear” means each band of the table is a straight line, but the full function is a staircase with a kink at every breakpoint. The scale is deliberately not proportional to concentration, so a doubling of a pollutant does not double the AQI.
A worked example makes this concrete. Say a station reports PM10 = 210 µg/m³. That value falls between the breakpoints 155 and 254 µg/m³, which map to index values 101 and 150. Plugging in:
Ip = [(150 - 101) / (254 - 155)] x (210 - 155) + 101
Ip = [49 / 99] x 55 + 101
Ip = 0.4949 x 55 + 101 = 128So 210 µg/m³ of PM10 gives an AQI of 128, which lands in the “Unhealthy for Sensitive Groups” band. For developers, the same logic reduces to four operations: truncate the concentration, find the bracketing band, interpolate, and round to the nearest integer. The US EPA AirNow AQI documentation is the canonical reference for the method.
AQI Breakpoints: The PM2.5 Table and the 2024 Revision
Breakpoints are not a fixed ratio of concentration to index. They come from epidemiological evidence and from each country’s ambient air quality standard, which is why they move when the science or the regulation moves.
The table below shows the current US EPA categories with the PM2.5 24-hour breakpoints used to produce them. Version: US EPA AQI, PM2.5 breakpoints as revised in 2024.
| AQI | Category | PM2.5 24-h (µg/m³) | Colour |
|---|---|---|---|
| 0–50 | Good | 0.0–9.0 | Green |
| 51–100 | Moderate | 9.1–35.4 | Yellow |
| 101–150 | Unhealthy for Sensitive Groups | 35.5–55.4 | Orange |
| 151–200 | Unhealthy | 55.5–125.4 | Red |
| 201–300 | Very Unhealthy | 125.5–225.4 | Purple |
| 301–500 | Hazardous | 225.5–325.4 | Maroon |
The 2024 change that moved the goalposts
In 2024 the EPA revised the PM2.5 breakpoints, and the top of the “Good” band moved from 12.0 down to 9.0 µg/m³. The air did not change. The ruler did.
Consider a 24-hour PM2.5 of 25 µg/m³. Under the old table it read 78 (Moderate). Under the 2024 table the same concentration reads 81, still Moderate but climbing faster.
Concentrations that used to clear “Good” now sit inside “Moderate”, which is exactly the point of a stricter standard: it flags more days for more people. The current tables live in the EPA AQS AQI breakpoints code table.
One more detail matters as much as the number: every pollutant has its own averaging window. PM uses 24 hours; O₃ uses 8-hour and 1-hour forms; CO uses 8 hours; SO₂ uses 1 hour and 24 hours; NO₂ uses 1 hour. The AQI depends on the window as much as on the reading.
Averaging Windows and NowCast: Why “Current AQI” Differs from the Daily AQI
The official AQI uses the regulated averaging period, which is 24 hours for particulates. It’s simply not valid to compute the daily AQI from a single hourly reading, even though many dashboards do exactly that.
How the NowCast AQI Reports Near-Real-Time Conditions
NowCast is the EPA method for reporting fast-changing conditions on the same 0–500 scale. It uses the past 12 hours of measurements and applies a weight factor of 1 - (max - min) / max, clamped between 0.5 and 1.0. A factor of 0.5 behaves like a 3-hour average; a factor of 1.0 behaves like a 12-hour average. At least two of the last three hours must contain valid data for NowCast to publish.
The weight rises when the air is stable and falls when it’s changing quickly, which is why NowCast spikes far ahead of the daily number during a smoke event. Marco, who runs a community sensor network for a school district in southern Oregon, watched this play out last fire season. His NowCast jumped to Hazardous at 4 pm on a Thursday while the official 24-hour AQI still read Moderate.
The district called off outdoor practice on the NowCast signal alone. By Friday morning the daily average had caught up and confirmed the call.
What this means in practice
A momentary green NowCast doesn’t clear a red daily AQI, and the reverse is also true. Which one matters depends on your decision. If you’re planning a run or a shift, you want NowCast. If you’re reporting daily exposure for a compliance file, you want the 24-hour figure.
Air quality monitoring data platforms and APIs typically compute and display both, and the label matters more than the number. Wildfire smoke is the textbook NowCast use case.
The Overall AQI and the Dominant Pollutant
The overall AQI is the maximum of the per-pollutant sub-indices. It’s never the average. A station measuring all six reference pollutants produces six sub-indices, and the highest one becomes the AQI you see.
The pollutant that produced that maximum is called the dominant or responsible pollutant, and it’s worth naming whenever you publish a number. Suppose a station returns sub-indices of 88 for ozone, 60 for PM2.5, 45 for PM10, 30 for NO₂, 20 for SO₂ and 15 for CO. The overall AQI is 88, and ozone is the dominant pollutant. Reporting “AQI 88” without that label hides the actual health concern.
Partial-data rules vary by country, which creates another source of disagreement. India’s CPCB computes an overall AQI only when at least three pollutants are available, one of which must be PM2.5 or PM10, and each sub-index needs a minimum of 16 hours of data. Some systems average the two worst sub-indices instead of taking the maximum.
None of them average all of them. Measuring more pollutants gives a more representative AQI, which is why sensor selection drives the result.
One Number, Many Rulers: US AQI, CAQI, China HJ 633 and WHO
This is where most explainers stop, and it’s exactly where “the AQI” falls apart. The table below compares five systems that are all called “air quality” reporting. Version: current as of September 2026.
| System | Scale | Bands | Pollutants | Key averaging | Health advice? |
|---|---|---|---|---|---|
| US EPA AQI | 0–500 | 6 | 6 | 24-h PM, 8-h O₃ | Yes, per band |
| Europe CAQI | 0–100 (can exceed) | 5 | NO₂ + PM10 mandatory roadside | 1-h and annual forms | No |
| China HJ 633 | 0–500 | 6 | 6 | 24-h and 1-h | Yes, per band |
| India CPCB | 0–500 | 6 | 8 (adds NH₃, Pb) | 24-h, min 16 h | Yes, per band |
| WHO 2021 AQG | Not an index | n/a | PM, O₃, NO₂, SO₂, CO | Guideline concentrations | Guideline, not index |
The European Common Air Quality Index is the sharpest contrast. It runs 0–100 and is designed to sit close to WHO expectations, but it deliberately provides no health advice, and its roadside form is built on just two mandatory pollutants. A green CAQI and a green US AQI don’t mean the same thing.
The same air, four different numbers
Take a 24-hour PM2.5 of 25 µg/m³ and push it through four systems. Under the current tables it comes out at roughly 81 (US AQI), 42 (European), 36 (China) and 42 (India). Same air, four numbers, and that spread only grows as each jurisdiction revises its table.
China’s 2026 revision, explained for English readers
China released the revised ambient air quality standard GB 3095-2026 and the AQI technical regulation HJ 633-2026 in February 2026, and both took effect on 1 March 2026, replacing the 2012 standards. The revision tightens the PM2.5 Grade II limit to an annual 25 µg/m³ and a daily 50 µg/m³, with a transitional phase through 2030 at an annual 30 µg/m³ and daily 60 µg/m³ before full enforcement.
HJ 633-2026 also updates the IAQI concentration thresholds for PM10 and PM2.5, adjusts how daily and real-time AQI are reported, and modifies the real-time particulate sub-index calculation. Very little English-language content covers this yet, which means anyone quoting a Chinese AQI from a pre-2026 table may be using an outdated ruler. Elena, a sustainability consultant preparing an ESG report for a client with operations in Shanghai, learned this the hard way: her dashboard was still applying the 2012 thresholds, and her cross-border comparison understated local PM2.5 risk. Verify the standard number and its effective date against the China Ministry of Ecology and Environment before you publish any cross-border comparison.
For international reporting, quote the concentration and the guideline, not just “the AQI”. The WHO 2021 Global Air Quality Guidelines set PM2.5 at an annual 5 µg/m³ and a 24-hour 15 µg/m³, and they travel with fewer assumptions than any index.
Ready to see how an outdoor monitoring station produces these numbers? Review the environmental monitoring station range →
From Sensor Reading to Published AQI: The Data Chain
The formula is public and simple. The input decides whether the published AQI means anything. Here is the full chain, and it has six stages.
- Measure. A station measures the pollutant concentration. Low-cost optical PM sensors and gas sensors have known biases; the physics belongs with PM2.5 sensor accuracy for outdoor IoT.
- Quality control. Drop failed sensors and outliers, and apply correction factors to low-cost optical readings. The correction protocol itself is a separate discipline.
- Average or NowCast. Aggregate to the regulated window, or compute NowCast for real-time display. Never mix the two in one number.
- Interpolate. Push the concentration through the breakpoint table to get a sub-index.
- Take the maximum. The highest sub-index becomes the overall AQI, and its pollutant is the dominant one.
- Publish. Push the AQI, category and colour to the platform or dashboard.
Why the same PM2.5 signal produces two different AQIs
Two platforms can read the same raw signal and publish different AQIs. One posts the uncorrected output of the optical sensor; the other applies an EPA-style correction first. The second is defensible, and the first is almost always biased high. This is the single most common reason a community sensor network disagrees with the official number.
Where the hardware sits
The chain maps cleanly onto hardware. A field sensor measures; signal conditioning cleans the analogue output; an IoT data terminal or edge computer runs local collection and processing; a cloud platform stores and displays. On-node processing matters because it keeps quality control and computation alive when connectivity drops, which is common at remote stations.
Black Ant’s Edge Data Acquisition Computer performs local collection, filtering and real-time processing at the device level, with end-to-end latency of 10–50 ms, and its IoT Data Collection Terminal handles multi-protocol input over RS485, M-Bus and 4–20 mA. Both pair naturally with edge computing for real-time environmental monitoring.
How to Calculate and Report AQI in an Outdoor IoT Station
If you are standing up an air quality monitoring station, this six-step checklist keeps the number defensible.
- Decide which index you are reporting, US, European, Chinese or local, and pin the version and the averaging windows. Write them down.
- Confirm which pollutants the station actually measures. An AQI built on PM2.5 alone is a PM2.5 sub-index, not a full AQI.
- Validate the raw concentrations before any maths. Quality control, correction and humidity handling all happen here.
- Aggregate to the correct window, or compute NowCast for real-time display. Do not blend the two.
- Interpolate through the breakpoint table, take the maximum, and label the dominant pollutant.
- Publish the number with its index name, averaging window and version. An unlabelled AQI is a future dispute.
A capable outdoor station needs to measure air quality alongside temperature, humidity, wind speed, noise and particulate matter, then uplink reliably. Black Ant’s environmental monitoring station carries that sensor set in an IP65 outdoor enclosure and uplinks over NB-IoT or 4G, so the readings reach your platform without a wired connection. Download the station specification or talk to an application engineer →
Frequently Asked Questions
How is the AQI calculated?
By converting each pollutant concentration into a sub-index using piecewise linear interpolation against a breakpoint table, then taking the highest sub-index as the overall AQI.
What is the AQI formula?
Ip = [(IHi – ILo) / (BPHi – BPLo)] x (Cp – BPLo) + ILo, where Cp is the truncated concentration and BP and I are the bracketing breakpoints and index values.
Why do different apps show a different AQI for the same city?
Three usual causes: they use different national indices, different averaging windows (NowCast versus 24-hour), or one is computing from uncorrected low-cost sensor data.
What is the dominant pollutant in AQI?
The pollutant whose sub-index equals the overall AQI. It is the one driving the number and the one worth naming when you report.
What is the difference between US AQI and CAQI?
The US AQI runs 0–500 across six bands and gives per-band health advice. The European CAQI runs 0–100 across five bands, requires only NO₂ and PM10 roadside, and gives no health advice.
What is the difference between the current AQI and the 24-hour AQI?
The 24-hour AQI uses the regulated averaging window. The current AQI often uses NowCast to track fast changes, which is why it moves first during a smoke event.
Did the AQI breakpoints change in 2024?
Yes. The EPA revised the PM2.5 breakpoints, moving the top of the “Good” band from 12.0 down to 9.0 µg/m³.
Which pollutants are used to calculate AQI?
PM2.5, PM10, ozone, carbon monoxide, sulfur dioxide and nitrogen dioxide are the six reference pollutants. India’s CPCB adds ammonia and lead.
Can low-cost sensors be used to calculate AQI?
Yes, but only with quality control and correction. Uncorrected optical readings are typically biased high and will overstate the band.
How do the WHO air quality guidelines relate to the AQI?
The WHO guidelines are target concentrations, not an index. Many national breakpoints are set with reference to them, but the AQI scale itself is defined by each country’s regulator.
The Formula Is Public. The Input Is Where Accuracy Is Won.
Four things are worth carrying away. The air quality index calculation itself is simple, public arithmetic that anyone can reproduce. The breakpoints are both epidemiological and political, and they move, as the 2024 EPA and 2026 China revisions show.
The averaging window changes the answer as much as the concentration does, so NowCast and the daily AQI answer different questions. And an AQI number without its index name, version and averaging window is an argument waiting to happen.
The one variable you control is the input. A textbook-correct calculation running on an uncorrected sensor is confidently wrong, and no rounding rule will save it. That’s why the hardware layer deserves as much attention as the maths layer, from the sensor module through the data terminal to the platform that publishes the number.
If you are specifying an outdoor station for a city, a campus or a community network, our team can help you match the sensor set, the uplink and the reporting requirements to the index you actually need to publish. Request the specification sheet and talk to an application engineer →
