This season has no fixed end date, and the date question has its own page. Over the next five years the average haze year should get somewhat cleaner, and the worst year should not.
Finished peat restoration would have cut the PM2.5 from the 2015 fires by 24% (Kiely et al., 2021). It does nothing about the drought that lights the fires.
At 8 am today the monitor in west Singapore read a 24-hour PM2.5 of 178 µg/m³, which the national index scores as 228, "Very Unhealthy" (NEA, 18 October 2026). The WHO limit for a 24-hour average is 15 (WHO, 2021).
Indonesia has spent years trying to stop this smoke at the source. So will the haze get better? The studies give a two-part answer, and the second part is the one riders need to hear.
Key takeaways
- Completed peat restoration would have cut 2015 burned area by 6% but PM2.5 emissions by 24%, and avoided about 12,000 premature deaths (Kiely et al., 2021).
- In Kalimantan, even if every canal block worked, rewetting would avoid up to 6.4% of burned area (Castro et al., 2026).
- Applied to today's west Singapore reading, a 24% cut takes 178 µg/m³ to about 135, from 11.9 times the WHO limit to 9.0 times (this post's synthesis from NEA data, Kiely et al., 2021 and the WHO limit, formula shown below).
- Extreme positive Indian Ocean Dipole events, a drought driver, are projected to rise from one per 17.3 years to one per 6.3 (Cai et al., 2014).
The research behind this
Six peer-reviewed studies carry the argument. Each card says what was measured, what was found, and what it cannot tell you.
Kalimantan canal-block counterfactual
- What they measured
- A statistical counterfactual of canal-block peat rewetting on fire occurrence in West and Central Kalimantan, 2017 to 2022.
- What they found
- If every canal block worked, rewetting could avoid up to 6.4% of burned area. Overlapping rewetted areas worked best.
- What it can't tell you
- Two provinces only, and the outcome is fire area, not PM2.5. Many blocks did little.
Cost of the 2015 fires, with restoration finished
- What they measured
- Fire emissions and land-cover data used to cost the 2015 fires, then re-run as if the 2.49 million hectare restoration target were already complete.
- What they found
- Area burned down 6%, CO2 down 18%, PM2.5 emissions down 24%, about 12,000 premature deaths avoided.
- What it can't tell you
- It assumes restoration is finished and effective, in one year only.
Fire sensitivity to drought, 2015
- What they measured
- NASA satellite fire, aerosol and carbon monoxide records, airport visibility back to 1991, and dry-season rainfall.
- What they found
- 2015 was the most severe season in the NASA record, and fire sensitivity turns nonlinear once dry spells fall below 4 mm of rain a day.
- What it can't tell you
- One extreme year dominates, and it predates most restoration work.
Extreme Indian Ocean Dipole events
- What they measured
- An ensemble of climate models run under a high-emissions scenario (RCP8.5).
- What they found
- Extreme positive events become about 3 times as frequent: one per 17.3 years in the 20th century, one per 6.3 in the 21st.
- What it can't tell you
- RCP8.5 is a high-emissions case, and a drought driver is not a fire.
Strong El Niño under warming
- What they measured
- 33 El Niño events from 1901 to 2017, grouped by where they began, plus climate-model projections.
- What they found
- Events have started further west with more extreme cases since the 1970s, and models show strong events rising if the Pacific temperature gradient strengthens.
- What it can't tell you
- The authors call model disagreement on that gradient a major roadblock.
Haze days and cardiac arrest
- What they measured
- 8,589 out-of-hospital cardiac arrests in Singapore, 2010 to 2015, matched against daily PSI.
- What they found
- "Unhealthy" PSI days carried a risk ratio of 1.37 for cardiac arrest. Each extra 30 PSI added 5.8% to 8.1%.
- What it can't tell you
- Ambient PSI is not a rider's dose, and an association is not a mechanism.
What the studies actually measured
The papers answer two different questions. One asks how much smoke people can still remove. The other asks how often the weather will make smoke likely.
Kiely's team took the 2015 fires and re-ran them with the government's 2.49 million hectare restoration target already finished. That one change cut area burned by 6%, CO2 by 18% and PM2.5 emissions by 24% (Kiely et al., 2021). The paper reports the gap between 6% and 24% but does not name the cause, so this post does not either.
Castro's 2026 team looked at what canal blocks really did in West and Central Kalimantan between 2017 and 2022. The best case, with every block working, was up to 6.4% of burned area avoided (Castro et al., 2026). Blocks that overlapped worked best, and scattered single blocks did little.
Field's team studied the weather side. In 2015 the fires were the worst in the NASA satellite record, and the carbon monoxide plume lasted 2 months. They found fire sensitivity jumps once dry spells run below 4 mm of rain a day (Field et al., 2016).
Cai's models, Wang's historical record and Thirumalai's palaeoclimate work all ask how often that dry weather returns. They agree on the direction, which is more of it. They disagree on how much (Cai et al., 2014; Wang et al., 2019; Thirumalai et al., 2024).
Ride with filtered air
Filters haze PM2.5 on every ride
Easi Breezi clips onto any helmet and feeds you H11 HEPA-filtered air. Bike-powered, no batteries, no charging.
Protected peat is not the same as wet peat
Indonesia's restoration effort is large. The Environment Ministry counts about 4.15 million hectares of concession peat with roughly 35,000 canal blocks, and says more than 17,000 further blocks are needed outside concessions (Mongabay, July 2026).
The awkward number sits next to it. Pantau Gambut counted more than 28,000 hotspots inside peat hydrological units in the first half of 2026. Nearly two-thirds fell on peat designated as protected (Mongabay, July 2026).
Protected peat means the restoration worked, so fires there prove it failed.
"Protected" is a legal label for a zone. It says nothing about whether canals were blocked or the water table came back. Neither reading is supported by that figure alone.
Restoration trims the smoke per fire, not the fire's cause
Restoration works on one input: how much peat is dry enough to burn. Rewetting a block of drained peat lowers the fuel available when a fire starts. It cannot make it rain.
Two levers that decide how much smoke reaches the city
Conceptual, not measured data
Illustrative only. Drought sets how many fires start, and dry peat sets how much smoke each one makes.
The most smoke completed restoration would have removed from a 2015-class year, from only 6% less burned area.
Source: Kiely et al., 2021Page and Hooijer add a point that keeps restoration worth doing even where the smoke gain is modest. Drained peat gives off carbon all year at a scale similar to the fires themselves (Page and Hooijer, 2016). Rewetting is a climate win on top of a smoke win.
A quarter off today's haze is still nine times the WHO limit
Here is the arithmetic behind the headline. It is a synthesis, so the formula is printed.
Take west Singapore this morning: PM2.5 of 178 µg/m³ (NEA, 2026). Assume ambient smoke scales in proportion to source emissions, which is a fair first approximation for primary smoke and a rough one for weather-driven transport.
Apply Kiely's 24% cut (Kiely et al., 2021): 178 × (1 − 0.24) = 135 µg/m³. On the NEA breakpoints that scores about 184, one band down from "Very Unhealthy" to "Unhealthy".
The 178 figure is from NEA, 2026 and 15 is from the WHO, 2021. The 135 figure is this post's synthesis from Kiely's 24%.
The same sum for the other four Singapore regions reads like this. Every row stays well above the 15 µg/m³ limit.
| Region | Today PM2.5 | After 24% cut | Approx. PSI, today to after |
|---|---|---|---|
| West | 178 | 135 | 228 to 184 |
| Central | 137 | 104 | 186 to 152 |
| East | 130 | 99 | 179 to 146 |
| South | 125 | 95 | 174 to 142 |
| North | 82 | 62 | 129 to 108 |
µg/m³, NEA, 2026 readings of 18 October 2026, 08:00. "After" is a synthesis. The PSI figures use NEA's published PM2.5 breakpoints.
Text version of the chart: WHO limit 15, after a 24 percent cut 135, today measured 178, all in micrograms per cubic metre of PM2.5 for west Singapore.Treat 24% as a ceiling. Restoration will not be finished by 2031, so the real cut will be smaller than the one used here.
Why the drought side is not improving
The fire count depends on dry weather, and the climate papers point one way. Under greenhouse warming, extreme El Niño genesis is projected to become more frequent (Thirumalai et al., 2024). The size of the rise is disputed, and Wang's team calls model disagreement a major roadblock (Wang et al., 2019).
The Indian Ocean Dipole is the second drought driver, and it has a clean number. Cai's models put extreme positive events at one per 17.3 years in the 20th century and one per 6.3 years in the 21st (Cai et al., 2014).
Turning that into odds for 2027 to 2031 uses a standard formula: the chance of at least one event in 5 years is 1 − (1 − 1/T)^5. With T = 17.3 that is 25.7%. With T = 6.3 it is 57.9%.
Synthesis from Cai et al. (2014) event rates under RCP8.5, a high-emissions scenario. It is a probability of the drought driver, not of a haze crisis.
Text version of the chart: the chance of at least one extreme positive Indian Ocean Dipole event in five years is 25.7 percent at the 20th-century rate and 57.9 percent at the 21st-century rate.The 2026 season shows the driver at work. NOAA's 8 October discussion gives an 83% chance that this El Niño reaches historic strength, with Niño-3.4 at +2.1 °C (NOAA CPC, 8 October 2026).
That outlook stops at January to March 2027. It says nothing about the 2027 dry season, so this post does not forecast one. For the El Niño odds in detail, see the heat and haze essay. For how the gaps between crisis years are shrinking, see is haze getting worse every year.
What a bad year costs a rider's heart
Ho's Singapore registry study matched 8,589 cardiac arrests outside hospital against daily PSI. "Unhealthy" days carried a risk ratio of 1.37, and each extra 30 PSI added 5.8% to 8.1% (Ho et al., 2018).
The gap between 228 and 184 is 44 PSI. On Ho's slope, 44 / 30 × 5.8% to 8.1% is about 8.5% to 12% in cardiac-arrest risk. That is a rough extrapolation from a Singapore registry, and it describes ambient air, not a rider's dose.
The honest reading is that a finished restoration programme would lower the risk of a bad day, not remove it. The tail year still sits far above the limit.
Frequently asked questions
Will the haze get better?
Better in the average year, no better in the worst. Finished restoration would cut a 2015-class year's PM2.5 emissions by about 24%, but drought odds are rising (Kiely et al., 2021). For this season's timing, see the dedicated page.
Is the haze getting better or worse each year?
Neither steadily. Crisis years follow drought years, and the drought driver is becoming more likely. The year-to-year view is covered in our trend post.
Will restoring peatland stop the haze?
No. In the best modelled case it removes about a quarter of the smoke, and up to 6.4% of burned area across the whole region studied (Castro et al., 2026). It also cannot change the weather.
Will the haze get worse?
The drivers lean that way. Extreme positive Indian Ocean Dipole events are projected to go from one per 17.3 years to one per 6.3 (Cai et al., 2014). The size of the El Niño rise is still debated.
What to do this week
- Check the 24-hour PM2.5 as well as the index. The live haze page shows current readings for riders.
- Assume a bad year is coming. A 24% cleaner haze is still about nine times the WHO limit.
- Shift or shorten rides on the worst days. Ho's data links Unhealthy days to higher cardiac risk.
- Filter at the face. The Easi Breezi helmet air purifier uses H11 HEPA media that stops 95% of PM2.5, whatever the policy or the weather does.
Sources
- Castro, M., Ritonga, R., Masuda, Y., et al. (2026). Effective restoration can avoid peatland fires: large scale counterfactual assessment in Kalimantan, Indonesia. iScience, 29, 116041. doi.org/10.1016/j.isci.2026.116041
- Kiely, L., Spracklen, D.V., Arnold, S.R., et al. (2021). Assessing costs of Indonesian fires and the benefits of restoring peatland. Nature Communications, 12, 7044. doi.org/10.1038/s41467-021-27353-x
- Field, R.D., van der Werf, G.R., Fanin, T., et al. (2016). Indonesian fire activity and smoke pollution in 2015 show persistent nonlinear sensitivity to El Niño-induced drought. PNAS, 113(33), 9204 to 9209. doi.org/10.1073/pnas.1524888113
- Cai, W., Santoso, A., Wang, G., et al. (2014). Increased frequency of extreme Indian Ocean Dipole events due to greenhouse warming. Nature, 510, 254 to 258. doi.org/10.1038/nature13327
- Wang, B., Luo, X., Yang, Y.M., et al. (2019). Historical change of El Niño properties sheds light on future changes of extreme El Niño. PNAS, 116(45), 22512 to 22517. doi.org/10.1073/pnas.1911130116
- Thirumalai, K., DiNezio, P.N., Partin, J.W., et al. (2024). Future increase in extreme El Niño supported by past glacial changes. Nature, 634, 374 to 380. doi.org/10.1038/s41586-024-07984-y
- Page, S.E., Hooijer, A. (2016). In the line of fire: the peatlands of Southeast Asia. Philosophical Transactions of the Royal Society B, 371, 20150176. doi.org/10.1098/rstb.2015.0176
- Ho, A.F.W., Wah, W., Earnest, A., et al. (2018). Health impacts of the Southeast Asian haze problem: a time-stratified case-crossover study of ambient air pollution and sudden cardiac deaths in Singapore. International Journal of Cardiology, 271, 352 to 358. doi.org/10.1016/j.ijcard.2018.04.070
- Singapore National Environment Agency (2026). Real-time PSI API, reading of 18 October 2026, 08:00. api-open.data.gov.sg
- NOAA Climate Prediction Center (2026). ENSO diagnostic discussion, 8 October 2026. cpc.ncep.noaa.gov
- Mongabay (2026). Indonesia shows elevated fire hotspot activity in 2026 as super El Niño threat looms, July 2026. news.mongabay.com
- World Health Organization (2021). WHO global air quality guidelines. who.int
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.