API (Malaysia) and PSI (Singapore, and most of Southeast Asia's other indices) are both built the same way: a 24-hour average pulled from a small number of fixed stations. On 8 September 2026, Serian, Sarawak read 302, Hazardous. Kuching, about 65km away, read 249, Very Unhealthy, the same day. The gap is not two different air masses. It's what happens when 68 stations have to speak for a country the size of Malaysia, and a rider on the road between two of them gets neither number.
Part of our Haze and air quality by city guide.
Serian's monitor sat at 302 at 10am on 8 September. That's Hazardous, Malaysia's worst band, the one that triggers talk of school closures and movement restrictions (Malay Mail, 2026).
Kuching's monitor, the same morning, read 249. Very Unhealthy, a full category below. Same haze event, same state, one grade apart, reported by the same source on the same day.
Nobody's air changed at the state border. What changed is which station a person happened to be standing near.
Key takeaways
- On 8 September 2026, Serian's API hit 302 (Hazardous) while Kuching, roughly 65km away, read 249 (Very Unhealthy) the same day (Malay Mail, 2026).
- Malaysia's Department of Environment runs 68 monitoring stations for the whole country (aqicn.org, citing DOE), and Sarawak alone covers about 124,450 km², close to the land area of England.
- In a 2025 study of 441 motorcycle-taxi drivers across six Thai provinces, personal PM2.5 exposure showed no significant correlation with official government station readings at all (Samana et al., 2025).
- Mobile monitoring in Bengaluru found black carbon readings of 10 µg/m³ on residential streets versus 56 µg/m³ on highways within the same city, a gap no single fixed station can see (Upadhya et al., 2024).
The research behind this
Peer-reviewed measurements this post is built on. Each card says what was measured, what was found, and what it can't tell you.
Bengaluru mobile monitoring
- What they measured
- About a year of mobile monitoring for black carbon, ultrafine particles and CO2 across 150 unique kilometres of Bengaluru roads, with roughly 22 repeat passes per road segment at 30 metre resolution.
- What they found
- Black carbon averaged 10 µg/m³ on residential streets, 22 µg/m³ on arterial roads and 56 µg/m³ on highways, all inside the same city on the same day.
- Can't tell you: this is one Indian city, and black carbon isn't a regulated pollutant there either, so it shows the resolution problem rather than proving the same spread exists in Sarawak.
Bangkok motorcycle-taxi exposure
- What they measured
- Personal air sampling on 441 motorcycle-taxi drivers across six central Thai provinces, January to March 2023, worn through full work shifts.
- What they found
- Mean personal PM2.5 ranged from 132.2 µg/m³ in Bangkok to 410.9 µg/m³ in Pathum Thani, with no significant correlation between personal readings and the nearest official monitoring station.
- Can't tell you: dry-season Thailand only, and the null correlation shows the station and the rider's lungs disagree without isolating exactly why.
Cincinnati neighbourhood sensor network
- What they measured
- 12 low-cost sensors placed through one neighbourhood, calibrated against the nearest EPA reference monitor, April to June 2022, with 3 residents also wearing personal monitors for a week.
- What they found
- 11 of 12 neighbourhood sites read a higher air quality index than the EPA site, with some locations logging up to 6 days above healthy thresholds that the EPA monitor never crossed in the same window.
- Can't tell you: a US neighbourhood, not Southeast Asia, and only 3 people wore personal monitors.
Hanoi motorcyclist black carbon
- What they measured
- Portable aethalometers carried simultaneously on motorcycles and inside car and bus cabins through Hanoi traffic, a pilot study comparing commute modes directly.
- What they found
- Motorcyclists inhaled a mean of 29.4 µg/m³ of black carbon, against 10.1 µg/m³ for bus passengers on the same roads at the same time.
- Can't tell you: a small pilot sample, and black carbon sits outside every major Southeast Asian air index, so this shows a blind spot, not a coverage gap specifically.
What the studies actually measured
None of the four studies above were designed to embarrass a government index. They were designed to find out what people actually breathe, and they all landed on the same structural problem from different angles.
Upadhya and colleagues drove monitoring equipment over 150km of Bengaluru's roads for a year. They weren't testing Malaysia's API.
They were asking whether one fixed station can represent a whole city. The answer was a 5.6 times difference in black carbon between a residential street and a highway a few kilometres away (Upadhya et al., 2024).
Samana and colleagues put personal samplers on 441 real motorcycle-taxi drivers in Thailand for full shifts, then checked their readings against the nearest government air quality monitor. The two numbers had no significant statistical relationship (Samana et al., 2025). The station and the rider's lungs were, statistically, strangers.
Kubis, Edwards and Ryan did the same test in a single US neighbourhood with a dense sensor network, and found the same pattern: nearly every local sensor read worse than the one official monitor a few streets away (Kubis et al., 2026).
And Quang's team in Hanoi strapped monitors onto motorcycles and inside buses, at the same time, on the same roads, and found riders breathing close to three times the black carbon that bus passengers breathed feet away (Quang et al., 2020).
Four cities, four teams, zero overlap in funding or purpose. Same finding: a fixed station a few kilometres or a few streets away can miss what's actually in the air where a person stands.
Ride with filtered air
Filters haze PM2.5 on every ride
Easi Breezi mounts on your scooter or motorbike and feeds H11 HEPA-filtered air into your helmet. Bike-powered, no batteries, no charging.
One index, two neighbouring towns, two different verdicts
Malaysia's Air Pollutant Index and Singapore's PSI are built the same way everywhere in the region: pollutant concentrations averaged over 24 hours, converted into a single number, at a limited set of fixed stations.
On 8 September, that system produced 302 for Serian and 249 for Kuching, roughly 65km apart on the same highway, the same morning (Malay Mail, 2026). Both numbers were real. Both stations worked correctly.
The gap didn't come from a broken sensor. It came from measuring two different points in a very large, unevenly smoky state.
Malaysia's Department of Environment runs the Air Pollutant Index Management System (APIMS) through 68 stations covering the entire country (aqicn.org, citing DOE). Sarawak alone is about 124,450 km², not far off the land area of England.
A handful of those 68 stations sit in Sarawak. Most of the state, including most of the road between Serian and Kuching, has no station at all.
That single point matters. 302 triggers a different government conversation than 249 does, even though both numbers describe air most doctors would call dangerous. And both numbers are still just two data points standing in for a Division-sized area between them.
If you're riding between two towns during haze season, assume the air between the two official readings is worse than the better one and could be worse than both. Averages and station gaps both bias toward understating what you'll pass through.
Why a 24-hour average hides what you breathe at 5pm
Station sparsity is only half of the gap. The other half is time.
Both API and PSI report a 24-hour average. If pollution spikes for two hours during evening rush and stays lower the rest of the day, the daily number dilutes that spike into something that looks manageable. A rider who happens to be on the road during the spike breathes the peak, not the average.
This is exactly what Samana's team found in the field: real riders, measured for real shifts, showing exposure levels the official average simply didn't predict (Samana et al., 2025).
The World Health Organization's 2021 global air quality guidelines admit a second blind spot. There's still no formal guideline value for black carbon or ultrafine particles, the exact pollutants Quang's Hanoi study measured at tailpipe height (WHO, 2021).
Put together: the number misses the peak, misses the location, and misses two pollutant classes a rider disproportionately breathes.
The haze keeps coming back faster
This isn't a one-season problem. Severe Southeast Asian haze years have landed in 2015, 2019, 2023 and now 2026, a pattern that used to run on four-year gaps and has compressed to three.
The 2015 haze cost Indonesia an estimated $16.1 billion, about 1.9% of GDP that year, according to World Bank analysis (World Bank, 2016). The 2019 event cost roughly $5.2 billion, 0.5% of GDP, a smaller fire season with a smaller bill (World Bank, via Mongabay, 2019).
2026 is already tracking closer to 2015 than 2019. By 27 July 2026, satellite monitoring had recorded more than 96,000 fire hotspots nationwide in Indonesia, ahead of the roughly 88,000 recorded by the same date in 2019 and about 58,000 in 2023 (Mongabay, 2026).
Close to 103,000 hectares burned in the first half of the year alone, the same source reports.
Malaysia's station count hasn't grown to match. If the interval between severe haze years keeps shrinking while the network stays flat, the gap between what 68 stations can see and what a rider actually breathes has no reason to close on its own.
This is a projection based on the pattern above, not a published forecast from any agency.
What this means if you're riding through haze season
None of this means the API or PSI numbers are fake. Serian's 302 was real, and it correctly told a police headquarters and a state government that conditions were dangerous. What it can't do is tell a specific rider on a specific road at a specific hour what's actually in the air around their helmet.
That gap is precisely what active filtration is built to close. An intake at the rider's actual face, filtering in real time, doesn't need to know whether the nearest station reads 249 or 302.
It responds to what's genuinely arriving at the intake. Not to a number that was already hours and kilometres away from being accurate by the time it was published.
Who can you trust for a second opinion?
No single source replaces a government station, but several free tools let you cross-check the official number against something closer to street level:
- IQAir's AirVisual map aggregates government and independent stations worldwide, so you can compare a nearby town's reading against your own. Caveat: still station-based, still subject to the same siting gaps.
- Sensor.Community is a volunteer low-cost sensor network with far denser coverage in some cities than official networks. Caveat: individual sensors can drift without calibration checks.
- PurpleAir's public map shows real-time readings from thousands of consumer sensors, the same type of network Kubis and colleagues validated against EPA data in Cincinnati. Caveat: raw PurpleAir readings typically run high and need a correction factor to compare directly with official AQI.
- Nafas focuses on Indonesian cities with a denser local sensor network than government stations alone provide. Caveat: coverage outside its core cities is thin.
- Easi Breezi's live haze page at easibreezi.com/pages/haze pulls a modelled, more localized estimate rather than the nearest single station. Caveat: it's a model estimate, not an official PSI or API reading, and says so on the page.
- Check two nearby stations, not one. If you're travelling between towns during haze season, look up the reading at both ends of your route, and at the halfway point if one exists.
- Treat any 24-hour average as a floor, not a ceiling. Assume the worst hour of your ride is higher than the daily number suggests, especially around rush hour.
- Cross-check with an independent map like IQAir or Sensor.Community before a long ride through a haze-affected state.
- Filter at the source, not the forecast. A system that reacts to your actual intake air doesn't depend on how recent or how local the nearest official reading is. Easi Breezi's clip-on helmet air purifier filters what's actually at your face in real time.
Frequently asked questions
What is the difference between API and PSI?
API (Malaysia) and PSI (used in Singapore and historically Malaysia) are both composite air quality indices built from multiple pollutants averaged over 24 hours. They use different breakpoint scales and slightly different pollutant weightings, but both report a single number per station, and both share the same core limitation: sparse stations and long averaging windows.
Is API or PSI a better measure of air quality?
Neither is inherently better. Both are 24-hour, station-based averages, so both share the same structural blind spots: they miss short pollution spikes and can't represent conditions far from the nearest monitor. The more useful comparison is checking multiple nearby stations and an independent source, not picking one index over the other.
Why did Serian and Kuching show such different API readings on the same day?
Serian and Kuching each have their own DOE monitoring station, roughly 65km apart. Haze concentration varies with local wind, terrain and fire proximity.
On 8 September 2026, that variation was large enough to put one town in the Hazardous band and the other in Very Unhealthy, a full category apart, despite sharing the same regional haze event.
Does a higher API number always mean worse air where I am?
Not necessarily where you specifically are; it means worse air was measured at that station. If you're between two stations with different readings, especially in a large, sparsely monitored state like Sarawak, your actual exposure could sit anywhere between the two numbers, or occasionally outside that range.
Sources
- Malay Mail (2026). "Serian air quality 'hazardous' again just a day after emergency ends." malaymail.com
- aqicn.org / Malaysia Department of Environment. "APIMS network, 68 monitoring stations." aqicn.org/network/my.apims
- Upadhya, A. et al. (2024). "Long-term mobile monitoring reveals within-city air pollution variability." Science of the Total Environment. doi.org/10.1016/j.scitotenv.2024.169987
- Samana, K., Ito, K., Suthienkul, O., Ketsakorn, A. (2025). "Analysis of PM-bound polycyclic aromatic hydrocarbons exposure among motorcycle taxi drivers in six central provinces in Thailand in winter." PLOS ONE 20(12): e0336587. doi.org/10.1371/journal.pone.0336587
- Kubis, Edwards, Ryan (2026). "Neighbourhood-scale low-cost sensor network versus regulatory monitoring." Science of the Total Environment. doi.org/10.1016/j.scitotenv.2026.181837
- Quang, T. et al. (2020). "Black carbon exposure of motorcyclists versus car and bus passengers." Atmospheric Environment. doi.org/10.1016/j.atmosenv.2020.118029
- World Health Organization (2021). "WHO Global Air Quality Guidelines." who.int
- World Bank (2016). "The Cost of Fire: An Economic Analysis of Indonesia's 2015 Fire Crisis." worldbank.org
- Mongabay (2019). "Indonesia fires cost nation $5 billion this year: World Bank." news.mongabay.com
- Mongabay (2026). "Indonesia shows elevated fire hotspot activity in 2026 as super El Nino threat looms." news.mongabay.com
Written by Ash, mechanical engineer and founder of Easi Breezi, building an active HEPA filtration system for motorcycle helmets (Patent Pending). Based between Hong Kong and Bali, riding daily in the traffic this blog writes about.