API vs PSI vs ISPU: the same haze gets a different number in Malaysia, Singapore and Indonesia

API vs PSI vs ISPU: in typical haze, Malaysia's Air Pollutant Index (API) reads highest of the three for the same air. On 4 October 2026, Singapore's fine-particle (PM2.5) level averaged 51.2 µg/m³; scored on each country's official table, that is API 108 (Unhealthy) in Malaysia, PSI 96 (Moderate) in Singapore and ISPU 95 (Sedang, Moderate) in Indonesia, and 140 on the US AQI that many apps show.

All three Southeast Asian indices start "Unhealthy" at 101, so an API of 150 in Johor and a PSI of 150 in Singapore look alike. They are not the same air. Below about 150 µg/m³ of PM2.5, Malaysia's table climbs much faster. Using real October 2026 readings from Good Air's monitor archive, this guide shows what each index says for the same air.

The four indices at a glance

Each index turns pollutant concentrations into a number from 0 to 500. Each pollutant gets a sub-index, and the highest becomes the reading. In a haze that is almost always PM2.5, so this article compares the PM2.5 tables.

Malaysia's API, called IPU (Indeks Pencemar Udara) in Malay, is calculated from sulphur dioxide, PM10, PM2.5, ozone, nitrogen dioxide and carbon monoxide, with PM2.5 on a 24-hour average since 2017 (DOE, API calculation, opened 24 October 2026). Its PM2.5 table runs 0 to 50 for 0 to 12.0 µg/m³, 51 to 100 for 12.1 to 50.4, then squeezes 101 to 150 into 50.5 to 55.4 and spreads 151 to 200 over 55.5 to 150.4. Above 500 the DOE adds an Emergency level, when the public should follow National Security Council orders. On 9 October 2026 the DOE's deputy director-general, Azuri Azizah Saedon, said the API uses six pollutants including PM2.5, which often dominates in haze (New Straits Times, 9 October).

Singapore's PSI uses the same six pollutants. For PM2.5 the 24-hour breakpoints are 12, 55, 150, 250 and 350 µg/m³ for PSI 50, 100, 200, 300 and 400 (NEA, Computation of the PSI.pdf), last updated March 2014). NEA also publishes a separate 1-hour PM2.5 reading in four bands: Normal up to 55 µg/m³, Elevated above 55 up to 150, High 151 to 250 and Very High from 251 (NEA air quality FAQs, updated 22 October 2026).

Indonesia's ISPU is set by Ministerial Regulation No. 14 of 2020. Its PM2.5 breakpoints on a 24-hour average are 15.5, 55.4, 150.4 and 250.4 µg/m³ for ISPU 50, 100, 200 and 300, as reproduced with the regulation's table by the Singkawang city government's ISPU service (ISPU Singkawang, report of 29 July 2026). Our What is ISPU? explainer covers the categories and what to do at each.

The US AQI was revised on 6 May 2024. Good now ends at 9.0 µg/m³, Unhealthy for Sensitive Groups starts at 35.5, Unhealthy at 55.5, Very Unhealthy at 125.5 and Hazardous at 225.5 (US EPA fact sheet, February 2024).

Same PM2.5, four readings: the October 2026 table

We took real 24-hour PM2.5 averages from Good Air's archive of government monitors in October 2026 and ran each through the four official tables. These are not the numbers each agency published that day. They show what each index says for the same air.

\* The DOE publishes no concentrations, only the API. The Kuala Lumpur values are the PM2.5 implied by Batu Muda's published daily API of 155 and 175, through the DOE's table.

Three things stand out:

• Between 50 and 150 µg/m³, the API reads up to 50 points higher. At 55.5 µg/m³ the API is 151 while the PSI is 101 and the ISPU 100. On 22 October, Batu Muda's API of 155 corresponded to a PSI of 109.
• The PSI and ISPU almost agree once haze is thick. Above 55.5 µg/m³ they stay within one point. In cleaner air the ISPU reads lower, because its Good band runs to 15.5 µg/m³ rather than 12.
• From 150 µg/m³, all three Southeast Asian indices match. Pekanbaru on 14 October scores 218 or 219 on every one. The US AQI keeps reading higher, because its Very Unhealthy band starts at 125.5 µg/m³ since 2024.

Why a Malaysian 150 is not a Singapore 150

The jump sits in the DOE's own table. Its 101 to 150 band spans just 50.5 to 55.4 µg/m³, so a 24-hour average that passes 55.5 goes straight to 151. You can see it in the hourly readings. From 16 to 24 October 2026 (UTC), the 68 DOE stations Good Air tracks published 13,120 hourly readings: 3,746 were between 151 and 200, but only 637 between 101 and 150.

So when the Education Ministry tells Malaysian schools to stop outdoor activities above API 100 (our Malaysia haze report has the details), that line sits at about 50 µg/m³ of PM2.5. Singapore's PSI reaches 100 only at 55 µg/m³. If you live in Johor and work in Singapore, a "150" on each side of the causeway is not the same air: the Malaysian one is about 55 µg/m³ over 24 hours, the Singapore one about 103.

One caveat. An April 2020 version of the same DOE document, still online, puts the start of Unhealthy at 75.5 µg/m³ (DOE, API calculation, April 2020). The current version on the APIMS portal, and the gap in the hourly readings above, both fit the 50.5 µg/m³ table. As of 24 October 2026 we had not found a DOE notice dating the change.

Averaging time: what "now" means on each index

All four indices are defined on 24-hour PM2.5 averages, which move slowly. In a fast-moving plume the air can be far worse than the index for hours. Fatimah Ahamad of the Sunway Centre for Planetary Health has said this can mask a developing haze episode in Malaysia (SAYS, 28 August 2026).

• Singapore also publishes an official 1-hour PM2.5 reading with its own bands. Use it for the next few hours, but never put a 1-hour figure into the PSI table.
• Malaysia publishes only the API, every hour, on a 24-hour running average.
• The US AQI in apps is usually the NowCast, a weighted average of recent hours, so it reacts faster than the official indices.

That is why the table above compares daily means. A Singapore 1-hour reading and a Malaysian 24-hour API run on different clocks.

Colours and names do not line up either

Colours add to the confusion. Malaysia's DOE colours Good blue, Moderate green, Unhealthy yellow and Very Unhealthy orange (DOE IPU guide). Indonesia uses green for Baik and blue for Sedang, the second level (ISPU Singkawang). The US AQI runs green, yellow, orange, red, purple and maroon (EPA technical document, May 2026). So blue means the best air in Malaysia but the second level in Indonesia.

The names differ too. All three Southeast Asian indices have five levels: Good, Moderate, Unhealthy, Very Unhealthy and Hazardous (in Indonesian Baik, Sedang, Tidak Sehat, Sangat Tidak Sehat and Berbahaya). Only the US AQI adds a separate Unhealthy for Sensitive Groups level, from 35.5 µg/m³.

Which number should you trust?

For official decisions such as school closures, use your own country's index. Prof Mohd Talib Latif of Universiti Kebangsaan Malaysia said the DOE's API should be the primary reference, and that different formulas give readings that cannot be directly compared (New Straits Times, 9 October). Across borders, compare PM2.5 in µg/m³ or one index such as the US AQI, never an API with a PSI. For the regional picture, see Southeast Asia haze 2026 and the haze live tracker. For India's scale, read US AQI vs India's NAQI.

How we measured

• Data. Daily means of hourly PM2.5 from Good Air's archive of government monitors: NEA's five Singapore regions (1-hour PM2.5), BMKG and KLH monitors in Pekanbaru and Palembang, and the DOE's hourly API for Batu Muda in Kuala Lumpur. Days run midnight to midnight UTC, which is 8 am to 8 am in Singapore and Malaysia and 7 am to 7 am in Pekanbaru and Palembang. The window is 1 to 23 October 2026.
• Monitors. Singapore is the average of its five regions. Pekanbaru on 14 October is one BMKG monitor; on 6 October it is a BMKG and a KLH monitor. Palembang is the mean of its KLH and BMKG monitors. None of these days showed a frozen reading.
• Kuala Lumpur. The DOE publishes no concentrations. We converted Batu Muda's published daily API back to PM2.5 with the DOE's table. On both days every hourly API stayed inside the 151 to 200 band, so the conversion is linear and accurate to about 1 µg/m³.
• Indices. Each concentration was run through the DOE, NEA, KLH and EPA breakpoint tables linked above, rounded to whole numbers. Agencies' own published values for those days differ, mainly because they use running 24-hour windows and individual stations.