Not an official forecast. This is an experimental, independent model. For health advice, follow NEA.
Not yet validated. These yes/no odds have not been tested against past seasons with the current model, so treat them as rough.
Each % is the share of the 100,000 simulated futures where it happens. "Above 300" and "above 200" are checked at every simulated time step (every 6 hours to day 6, every 12 hours after), using the 24-hour average PSI. Beyond day 6 the ranges are wider. In the replays of 2015, 2019 and 2023, "above 300 within 6 days" was too high: on all 5 occasions it gave 30% or more, PSI did not go above 300. When "above 200" was given 60% or more, it happened about 4 times in 10.
Issued Sat 17 Oct, 2:04pm SGT. Independent, experimental model; not an official NEA forecast.
| Day | Best guess PSI | Likely range | Above 100 | Above 200 | Above 300 |
|---|---|---|---|---|---|
| Sun 18 Oct | 219 | 155–258 | 94% | 58% | 8% |
| Mon 19 Oct | 252 | 147–303 | 85% | 64% | 26% |
| Tue 20 Oct | 293 | 128–354 | 79% | 67% | 47% |
| Wed 21 Oct | 259 | 119–342 | 82% | 60% | 38% |
| Thu 22 Oct | 213 | 118–299 | 82% | 52% | 25% |
| Fri 23 Oct | 188 | 103–284 | 76% | 48% | 21% |
| Sat 24 Oct | 189 | 113–283 | 82% | 46% | 20% |
| Sun 25 Oct | 177 | 111–278 | 78% | 44% | 19% |
| Mon 26 Oct | 165 | 108–258 | 78% | 36% | 15% |
| Tue 27 Oct | 149 | 101–215 | 75% | 29% | 9% |
| Wed 28 Oct | 132 | 101–179 | 77% | 18% | 4% |
| Thu 29 Oct | 122 | 104–169 | 78% | 19% | 4% |
| Fri 30 Oct | 123 | 105–180 | 76% | 21% | 5% |
| Sat 31 Oct | 126 | 94–180 | 68% | 18% | 4% |
| Sun 1 Nov | 114 | 91–174 | 65% | 15% | 2% |
The model gives a 38% chance the 24-hour PSI is back to 100 or below by Fri 23 Oct. The haze season most likely ends around 11 Nov, when steady rain sets in over the fire areas (experimental estimate).
Gets worse: 79% chance. Peaks 265–420 around Tue 20 Oct. Clears early: 18% chance. PSI 100 or below by Mon 19 Oct. Clears later in the week: 2% chance. PSI 100 or below by Thu 22 Oct.
No. This is an independent, experimental model. For official readings and health advice, follow NEA at haze.gov.sg.
It traces the air arriving in Singapore back through up to 56 weather forecasts (NOAA GEFS and ECMWF's AI model), adds smoke from fires seen by NASA satellites, and simulates 100,000 possible next 15 days, weighted to match NEA's latest readings. The odds rerun every hour; the full model runs every 6 hours.
In replays of the 2014, 2015, 2019 and 2023 haze seasons, about 85% of NEA readings fell inside the forecast's 90% range. For 1 to 6 days ahead the best guess was about as good as assuming PSI stays as it is today; for 7 to 15 days ahead it was clearly better. Every live forecast is scored daily on the Track record page.