Indonesia's haze season follows the dry season, which usually runs from July to October. In 2026 BMKG expects fire risk into early October, with late, short rains. In Palembang the smoke is worst before sunrise: on 17 September the city's monitor read 681.6 µg/m³ at 4 am and 79.4 at 3 pm (BMKG via Antara, 2026).
Part of our Haze and air quality by city guide.
Ride out of your gang in Palembang before dawn this month and the streetlights wear halos of smoke. The ojol shift has already started, and the air has not cleared.
At that hour on 17 September, BMKG's Musi 2 station in Palembang read PM2.5 at 681.6 µg/m³. By 3 pm it read 79.4 (BMKG via Antara, 2026). Same city, same day, same smoke. The only difference was the clock.
Key takeaways
- On 17 September 2026, Palembang's Musi 2 monitor read PM2.5 at 681.6 µg/m³ at 4 am and 79.4 µg/m³ at 3 pm, so the dawn air was 8.6 times as dense as the afternoon air (BMKG via Antara, 2026).
- The 4 am reading was 45 times the WHO 24-hour PM2.5 guideline of 15 µg/m³ (WHO, 2021).
- Retail masks tested on real faces leaked 3 to 68% of outside particles inward, and only one of four averaged under 10% (Cherrie et al., 2018).
- Indonesia counted 174,694 acute respiratory infection cases in seven haze-hit provinces from August to 21 September 2026 (Ministry of Health via Malay Mail, 2026).
The research behind this
Peer-reviewed measurements this post is built on. Each card says what was measured, what was found, and what it cannot tell you.
Haze and child respiratory health in Palembang
- What they measured
- Ten years (2011 to 2020) of monthly BMKG air data in Palembang, set against child respiratory infection and pneumonia cases.
- What they found
- 366,632 child respiratory infection cases and 27,574 child pneumonia cases, with rises in PM2.5 predicting more pneumonia.
- What it can't tell you
- City-level monthly data, so no personal exposure and no figures for riders.
What Sumatran peat smoke is made of
- What they measured
- PM2.5 mass, particle counts and chemistry at three Sumatran sites during a peat fire episode: next to the fires, semirural, and a downwind city.
- What they found
- PM2.5 reached 1,600 µg/m³ near the fires, mostly organic carbon particles.
- What it can't tell you
- One episode, and the peak values are near the source, not in the city.
Why dry years burn so hard
- What they measured
- Satellite fire and smoke records, long-term airport visibility and dry-season rainfall for Sumatra and Kalimantan.
- What they found
- Fire jumps sharply once rainfall drops below about 4 mm a day. 2015 ranked just behind 1997, alongside 1991 and 1994, among the worst episodes on record.
- What it can't tell you
- Regional scale with no health data, and the record ends at 2015.
How retail masks perform on real faces
- What they measured
- Nine retail PM2.5 masks bought in Beijing. Four were worn by volunteers in a diesel-exhaust chamber, sitting and moving, with sensors inside and outside the mask.
- What they found
- Inward leakage ran from 3 to 68% while sitting. Only one mask averaged under 10%.
- What it can't tell you
- Diesel exhaust in a lab, not peat smoke, and nobody wore a helmet.
How much a mask cuts smoke exposure
- What they measured
- A model combining mask filter efficiency, fit, leak rate and particle size, tested on a US wildfire season.
- What they found
- A surgical mask cut exposure by a factor of 1.7. An N95 with a 5% leak cut it by more than 14.
- What it can't tell you
- Modelled, not measured on riders, and it assumes the mask is worn well all the time.
What the studies actually measured
Start with the only peer-reviewed study built on Palembang itself. Haryanto and colleagues matched ten years of BMKG air data against child health records. They counted 366,632 child respiratory infection cases and 27,574 child pneumonia cases from 2011 to 2020 (Haryanto et al., 2026).
They describe Palembang as a city hit by peatland fire smoke every year. In their models, rises in PM2.5 predicted more child pneumonia.
What is in that smoke? A field team sampled Sumatran peat smoke at three distances from the fires. Next to the fires PM2.5 reached 1,600 µg/m³, with 170,000 particles in every cubic centimetre, and most of the mass was organic carbon (See et al., 2007). The smoke still degraded air in the downwind city.
Why some years are so much worse comes from rainfall. Field and colleagues found fire in Sumatra and Kalimantan climbs steeply once dry-season rain falls below about 4 mm a day (Field et al., 2016). They expect these episodes to return in future droughts, usually El Niño years.
The cost at scale is large. A model of the 2015 fires found 69 million people persistently exposed to unhealthy air (Crippa et al., 2016). A later study put the average toll of peat-fire smoke at about 33,100 adult premature deaths a year across Sumatra and Kalimantan (Hein et al., 2022).
Two more studies cover the mask that riders are handed. Cherrie's team found retail masks leaked 3 to 68% inward on volunteers (Cherrie et al., 2018). Kodros's model gives a surgical mask a 1.7 times cut in exposure (Kodros et al., 2021).
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.
When haze season starts and ends, and why 2026 is running long
Indonesia's dry season typically runs from July to October, and that is when the peat burns (Malay Mail, 2026). BMKG first recorded hazardous PM2.5 in parts of Palembang on 24 August this year (ECHO via ReliefWeb, 2026).
The end date is the part riders keep asking about. BMKG now expects fire risk to continue into early October, with more than 60% of Indonesia's land starting its rainy season later than average (The Star, 2026).
The rains themselves will be late and short. BMKG puts rainy season onset in November for 24.66% of the country and December for 17.31%, driven by a strong El Niño and a positive Indian Ocean Dipole (Tempo, 2026).
The 2026 fires have burned 895,657 hectares since January, 34% more than 2019 (The Guardian, 2026). Health ministry counts across seven haze provinces reached 174,694 respiratory infection cases by 21 September, over 40,000 of them toddlers, up from more than 113,000 on 9 September (Malay Mail, 2026). That is roughly 5,000 new cases a day over those 12 days.
In Palembang itself, the mayor reported 9,313 respiratory infection cases in the first 20 days of August, about 466 a day (detikSumbagsel, 2026). The city moved schools from preschool to junior high online from 9 September, with no end date set (Antara Sumsel, 2026).
If you want the longer trend behind these seasons, our post on whether haze is getting worse every year walks through it.
Palembang's dawn air is 8.6 times as dense as its afternoon air
The Musi 2 station uses beta attenuation, the standard reference method for PM2.5. On Thursday 17 September it read 681.6 µg/m³ at 04:00 WIB, in the black "hazardous" band. At 15:00 WIB the same station read 79.4 µg/m³ (BMKG via Antara, 2026).
Divide one by the other and the dawn air carried 8.6 times as much fine particle mass (BMKG via Antara, 2026). Against the WHO 24-hour guideline of 15 µg/m³, the 4 am reading was 45 times the limit and the 3 pm reading was 5.3 times (WHO, 2021).
| Time | PM2.5 (µg/m³) |
|---|---|
| 04:00 WIB | 681.6 |
| 15:00 WIB | 79.4 |
| WHO 24-hour guideline | 15 |
Here is what that swing means per ride. At the same breathing rate, a 45-minute ride at 4 am took in the same PM2.5 as about 6 hours 26 minutes of riding at 3 pm.
Moving one hour of riding from dawn to mid-afternoon cut that hour's concentration by about 88%. Both figures are our arithmetic on the two BMKG readings (BMKG via Antara, 2026).
Two readings on one day are not an hourly average. The numbers above describe 17 September, not every day. But the direction matches what BMKG and the city warn about: the smoke sits lowest and thickest overnight.
Night air traps the smoke at rider height
The mechanism is ordinary weather physics. After dark the ground loses heat fast, and the air touching it becomes cooler than the air above. That warm layer on top acts like a lid, called a temperature inversion.
Smouldering peat keeps smoking all night, and under the lid the smoke has nowhere to go. It pools in the lowest layer of air, which is where you ride.
After sunrise the sun heats the ground, the lid breaks, and the smoke mixes up through a much deeper layer of air. The same smoke, spread through more air, reads as a lower number by mid-afternoon.
Why the same smoke reads higher at dawn
If you can shift a trip, shift it out of the pre-dawn hours. On 17 September the 3 pm air held about one eighth of the 4 am smoke (BMKG via Antara, 2026).
Work out your own dawn ride
Pick a time, a mask and your ride length. The tool uses only the two Musi 2 readings and the two mask factors cited above.
Your ride on 17 September, Palembang
Based on BMKG's Musi 2 readings for that day. Illustrative only.
Time you ride
What you wear
Air you breathe: about 681.6 µg/m³, 45 times the WHO 24-hour guideline. This ride carries the same particle dose as about 386 minutes of unmasked riding at 3 pm.
| Input | Value | Source |
|---|---|---|
| PM2.5 at 04:00, 17 Sep 2026 | 681.6 µg/m³ | BMKG via Antara |
| PM2.5 at 15:00, 17 Sep 2026 | 79.4 µg/m³ | BMKG via Antara |
| Surgical mask exposure cut | 1.7 times | Kodros et al., 2021 |
| N95 at 5% leak exposure cut | more than 14 times | Kodros et al., 2021 |
| WHO 24-hour guideline | 15 µg/m³ | WHO, 2021 |
The mask at the red light cannot close an 8.6 times gap
Palembang's official protection for riders is a free mask. Traffic police handed out 2,000 masks on Jalan POM IX aimed at motorcyclists and ojol drivers passing through (Palembang Independen, 2026). A forestry association added 8,000 more at red lights (Detik Sumsel, 2026). Neither report says what type.
That matters because fit decides most of the result. In Cherrie's tests, the mask material itself let through 0.26 to 29% of particles, but total leakage on faces reached 68% (Cherrie et al., 2018). Air takes the easy path around the edges.
What is left at 4 am behind a surgical mask: about 400 µg/m³ from a 681.6 reading, using a modelled 1.7 times cut.
Source: BMKG via Antara, 2026; Kodros et al., GeoHealth, 2021; WHO, 2021Run the dawn number through Kodros's surgical mask factor and about 400 µg/m³ is still left behind the mask (Kodros et al., 2021). That is 27 times the WHO 24-hour guideline, and far worse than an unmasked rider at 3 pm.
A helmet makes the seal harder. The chin bar and the mask compete for the same space, and a mask that shifts every time you turn your head leaks more. That point is engineering reasoning, not a measured result, but it pushes in the same direction as the lab data.
| Option | Cut in exposure | Air left at 4 am (17 Sep) | Main weakness on a bike |
|---|---|---|---|
| No mask, open vents | None | 681.6 µg/m³ | Vents move unfiltered air, and none at a stop |
| Surgical mask | 1.7 times (modelled) | About 400 µg/m³ | Leaks at the edges, shifts under a chin bar |
| Sealed N95 | More than 14 times at 5% leak (modelled) | Under about 49 µg/m³ | Needs a tight seal the whole ride |
| Ride at 3 pm instead, no mask | 8.6 times vs dawn | 79.4 µg/m³ | Only helps if your schedule can move |
Table figures: readings from BMKG via Antara (2026), mask factors from Kodros et al. (2021), with the air left behind each option as our arithmetic on the two.
Helmet vents do not help here. Ram-air vents need forward speed, and they feed you whatever air is outside. At a red light the flow stops, and you sit inside the same low smoke layer the night trapped.
Peat smoke is mostly fine organic carbon (See et al., 2007). Vent mesh is built to stop bugs and grit, not particles that small. For more on mask types, see our guide to pollution masks for motorcycle riders.
The next haze season is already on the calendar
Sumatra's big haze years in the record we checked are 2015, 2019 and 2026 (Field et al., 2016; The Guardian, 2026). That is three in 12 years, roughly one every four.
This is our own base-rate estimate from those three years, not a forecast. If that pace holds, a Palembang rider should expect at least one more hazardous haze season before 2031. Field's team links these seasons to droughts, usually El Niño years, and expects them to return without land-use reform.
The same smoke will land on older lungs. Hein's team found the health burden from peat-fire smoke rises as Indonesia's population ages (Hein et al., 2022).
Children carry the first signal. In the 2015 fires, insurance records across 203 districts showed more primary care visits for breathing problems among urban children under five (Sambodo et al., 2024). In a motorbike city, many of those children ride as passengers. Our post on child air pollution exposure on a motorcycle covers the pillion seat.
Palembang is not alone. We looked at the same smoke from Pontianak last week, and you can check today's readings across the region on our live haze map for riders.
Frequently asked questions
When is haze season in Indonesia?
It follows the dry season, which typically runs from July to October (Malay Mail, 2026). In 2026 Palembang's first hazardous readings came on 24 August (ECHO via ReliefWeb, 2026).
How long will the haze last in 2026?
BMKG expects fire risk into early October, and many areas will not see rain until November or December (The Star, 2026; Tempo, 2026). Expect smoky mornings until the rains settle in.
What time of day is haze worst in Palembang?
Before sunrise, based on the readings we have. On 17 September the Musi 2 station read 681.6 µg/m³ at 4 am and 79.4 at 3 pm (BMKG via Antara, 2026).
Is it haze season right now?
If you are reading this in September or early October 2026, yes. Check a live reading before you ride, since the level changes hour to hour.
Does a surgical mask protect me on a motorbike?
Partly. A model puts the cut at 1.7 times, which at 4 am still leaves about 400 µg/m³ (Kodros et al., 2021). Fit matters most, and a helmet makes a good fit harder.
What to do this week
- Check the hour as well as the day. Look up a live PM2.5 reading before a pre-dawn ride, since the dawn number can be many times the afternoon one.
- Move the trips you can. Errands and school runs shifted to mid-afternoon cut the smoke in each breath.
- Close vents at stops. At a red light your vents only pull in the smoke layer at road level.
- Upgrade the free mask if you keep one. A sealed N95 does far more than a loose mask, if it stays sealed under the helmet.
- Filter the air inside the helmet. The Easi Breezi fan pushes air through an H11 HEPA filter that stops 95% of PM2.5, at a red light as well as at speed. It runs off the bike, with no battery to charge before a 4 am start.
The season will end when the rain comes. Until then, the clock and the air in front of your face are the two things you control.
Sources
- BMKG Stasiun Klimatologi Sumsel, via Antara Sumsel (2026). Kualitas udara di Palembang masuk kategori berbahaya akibat asap. sumsel.antaranews.com
- World Health Organization (2021). WHO global air quality guidelines. who.int
- Cherrie JW, Apsley A, Cowie H, et al. (2018). Effectiveness of face masks used to protect Beijing residents against particulate air pollution. Occupational and Environmental Medicine 75. doi:10.1136/oemed-2017-104765
- Indonesian Ministry of Health, via AFP and Malay Mail (2026). Respiratory problems climb past 170,000 cases in haze-hit Indonesia. malaymail.com
- Haryanto B, Kurniasari F, Trihandini I, et al. (2026). Climate change adaptation to smoke haze for improved child health in Southeast Asia: analysis situation in Palembang, Indonesia. Annals of Global Health 92. doi:10.5334/aogh.5211
- See SW, Balasubramanian R, Rianawati E, Karthikeyan S, Streets DG (2007). Characterization and source apportionment of particulate matter ≤2.5 µm in Sumatra, Indonesia, during a recent peat fire episode. Environmental Science & Technology 41. doi:10.1021/es061943k
- Field RD, van der Werf GR, Fanin T, et al. (2016). Indonesian fire activity and smoke pollution in 2015 show persistent nonlinear sensitivity to El Niño-induced drought. PNAS 113. doi:10.1073/pnas.1524888113
- Crippa P, Castruccio S, Archer-Nicholls S, et al. (2016). Population exposure to hazardous air quality due to the 2015 fires in Equatorial Asia. Scientific Reports 6. doi:10.1038/srep37074
- Hein L, Spadaro JV, Ostro B, et al. (2022). The health impacts of Indonesian peatland fires. Environmental Health 21. doi:10.1186/s12940-022-00872-w
- Kodros JK, O'Dell K, Samet JM, et al. (2021). Quantifying the health benefits of face masks and respirators to mitigate exposure to severe air pollution. GeoHealth 5. doi:10.1029/2021gh000482
- ECHO Daily Flash, via ReliefWeb (2026). Indonesia forest fires and haze, Sumatra and Kalimantan, 24 August 2026. reliefweb.int
- The Star (2026). BMKG forecast of forest fire risk into early October, 23 September 2026. thestar.com.my
- Tempo English (2026). BMKG 2026/27 rainy season outlook, 23 September 2026. en.tempo.co
- The Guardian (2026). Reporting on the 2026 Borneo and Sumatra fires, 21 September 2026. theguardian.com
- detikSumbagsel (2026). Kasus ISPA di Palembang tembus 9.313 dalam 20 hari. detik.com
- Pemkot Palembang, via Antara Sumsel (2026). Pemkot Palembang terapkan belajar daring dampak kabut asap. sumsel.antaranews.com
- Ditlantas Polda Sumsel, via Palembang Independen (2026). Ditlantas Polda Sumsel bagi-bagi masker ke pengendara. palembangindependen.com
- Detik Sumsel (2026). Kabut asap selimuti Palembang, 8000 masker dibagikan ke pengendara di lampu merah. detiksumsel.com
- Sambodo NP, Pradhan M, Sparrow R, van Doorslaer E (2024). When the smoke gets in your lungs: short-term effects of Indonesia's 2015 forest fires on health care use. Environmental Health 23. doi:10.1186/s12940-024-01079-x
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.