GeoAnalitikEvery figure in this study is as of 16 October 2026 and will not be refreshed.
Each source’s window and clock, the satellite gaps and their size, the checks against Indonesia’s SiPongi database, ASMC’s counts and Copernicus’ reanalysis, the back-trajectory method, what the study left out, and every credit and licence.
Six windows, one study date
No source’s data runs past the study date, but some files were downloaded or read after it; the credits give each access date. Fire days are UTC dates; DOE readings and the CAMS forecasts run on Malaysian time; rain days run on each fire point’s own clock. Only the timing test re-dates detections to Malaysian time.
| Source | Window | Clock | Note |
|---|---|---|---|
| Satellite detections, NASA FIRMS | 1 Jan – 15 Oct of each year | UTC dates | VIIRS S-NPP alone for the long series; its 2026 gaps in Table 2 |
| Air quality, DOE APIMS | 1 Sep 10:00 – 16 Oct 20:00 2026 | MYT | 45 days with readings; 4 Sep missing. A station’s daily figure is its highest hourly API |
| Modelled smoke, Copernicus CAMS forecasts | 1 Jul – 15 Oct of 2015, 2019, 2023 and 2026 | MYT | Forecasts, not readings; seasons are not ranked on them (Table 5) |
| Rain, ERA5 at the 12 fire points | 2015 to 10 Oct 2026 | Each fire point’s local day (WIB, WITA, WIT) | A reanalysis at a point, not a rain gauge |
| Back-trajectories, ERA5 winds | Arrivals 1 Sep 00:00 – 15 Oct 21:00 2026 | UTC; arrivals counted on their MYT day | Every 3 h, 120 h back |
| SiPongi database (Indonesia’s forestry ministry) | WIB days 2–9 Oct 2026 | WIB | Read 17 Oct 2026, 09:22 UTC; a snapshot, SiPongi may reviseSource: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) |
Table 1. The window and clock of each source. MYT is Malaysian time (UTC+8); WIB, WITA and WIT are Indonesia’s western, central and eastern clocks.
NOAA ended S-NPP delivery at 13:00 UTC on 2 Nov 2026, after this study’s window; the window is unaffected, but this S-NPP-only series cannot continue.
The satellite gaps hide about 20,600 more Indonesian detections (4.0%), or about 25,800 (5.0%) with the night of 12 October filled; Indonesia stays second
A detection is not a fire: it is one pixel flagged hot on one satellite pass, so only counts from the same satellite compare across years, and the long series is VIIRS S-NPP alone. In 2026 S-NPP has no detection anywhere in the study box on 46 days (10–12 Mar, 28 Apr – 2 Jun, 11–16 Jul and 3 Aug), none by day on 4 Aug and 13 Oct and none by night on 10 Jul and 12 Oct, so every 2026 S-NPP count is a lower bound.
Each gap is filled from NOAA-20 on the same half-day and island group, scaled by the two satellites’ median daily ratio (1.00, over the 232 days on which neither misses a pass). On the night of 12 Oct NOAA-20 missed Sumatra and Kalimantan as well; filled from NOAA-20’s nights either side, it adds about 5,200 more. Filling only raises 2026, so Indonesia stays second of the 15 years either way.
| Island group | Hidden by the gaps | With the night of 12 Oct filled |
|---|---|---|
| Sumatra | about 4,800 | about 5,600 |
| Kalimantan | about 7,600 | about 11,900 |
| Papua and West Papua | about 4,100 | about 4,200 |
| Rest of Indonesia | about 4,100 | about 4,100 |
| Indonesia | about 20,600 (4.0%) | about 25,800 (5.0%) |
Table 2. Indonesian VIIRS S-NPP detections the gaps hide, 1 Jan – 15 Oct 2026, by island group, rounded to the hundred. An estimate. FIRMS may yet fill May. The 5,848 near-real-time detections in the 3–6 Aug window (1.1%) would drop out if FIRMS replaced them with standard data.
SiPongi’s database matches our 2026 pull on land; its press counts stop part-way through a day
SiPongi, the Indonesian forestry ministry’s system, serves FIRMS detections, so this check shows that our pull is complete, not that FIRMS is right, and only for its high-confidence points: MODIS from confidence 82, S-NPP and NOAA-20, with no NOAA-21, on WIB days 2–9 Oct. Every one of the 9,060 points in its database for those days, read on 17 Oct 2026, 09:22 UTC, is in our pull, and 9,059 of them lie inside our land outlines. Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id)
Its counts reach the press two ways: the ministry’s evening releases, counted “up to” a stated time with no window length given, and Databoks’ morning reads of “the last 24 hours”. Cut where its data stood, the database gives both morning reads to the last detection, and the evening releases of 4 and 7 Oct at an hour before the stated one; the releases of 3 and 5 Oct fall between two of its detection times, and 8 Oct’s only with data to late morning (Table 3).
Counted on SiPongi’s classes, every FIRMS detection type, ours match the database to the detection to 21:15 WIB on 5 Oct and in Databoks’ 7 Oct window. On 3 and 4 Oct and in Databoks’ 5 Oct window ours run higher, nearly all of it NOAA-20 detections over the sea, which the database leaves out; on land both count the same NOAA-20 detections in those windows.
| Day (WIB) | Published | SiPongi database | Ours, SiPongi’s classes |
|---|---|---|---|
| 2 OctEvening release, no time stated | 1,271Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | Not compared: the release states no time | – |
| 3 OctEvening release, to 21:10 WIB | 666Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 987 986 on land; the published count falls between 18:30 (628) and 20:06 (814) WIBSource: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,203987 on land |
| 4 OctEvening release, to 21:10 WIB | 527Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 693 693 on land; the published count is the database’s at 19:08 WIBSource: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 757695 on land |
| 5 OctEvening release, to 21:15 WIB | 1,359Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,433 1,433 on land; the published count falls between 19:44 (1,196) and 19:46 (1,395) WIBSource: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,4331,433 on land |
| 5 OctDataboks morning read, data to 05:52 WIB | 860Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 860 860 on land; the published count is the database’s at 05:52 WIBSource: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 924862 on land |
| 6 OctEvening release, no time stated | 1,201Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | Not compared: the release states no time | – |
| 7 OctDataboks morning read, data to 04:27 WIB | 1,074Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,074 1,074 on land; the published count is the database’s at 04:27 WIBSource: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,0741,074 on land |
| 7 OctEvening release, to 21:00 WIB | 1,176Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,212 1,212 on land; the published count is the database’s at 17:47 WIBSource: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,2211,219 on land |
| 8 OctEvening release, to 21:10 WIB | 387Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,363 1,363 on land; the published count falls between 11:32 (360) and 11:49 (445) WIBSource: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id) | 1,3671,366 on land |
Table 3. SiPongi’s published high-confidence counts against its own database and ours, WIB days 2–9 Oct 2026, each over the 24 hours to the published figure’s stated time, or to the last detection the page held. Ours are counted on SiPongi’s classes.
Counts only, never points, maps or screenshots. The match shows our pull is complete, not that FIRMS is right.
The Star’s ASMC count sets 2026 to 13 October against all of 2015 on two satellites; on one satellite and the same window, 2026 is 0.69 of 2015
The Star’s chart of ASMC data gives Sumatra and Kalimantan a 2026 count of 38,511 “as of Oct 13”, VIIRS daytime high confidence with the satellite not stated, against 58,106 for 2015. ASMC’s own annual page gives that 2015 figure for the full calendar year, on Suomi-NPP; from 2019 its series is NOAA-20. Our S-NPP counts, one satellite throughout, give the same-window ratio.
| Ratio | 2026 | 2015 | What it compares |
|---|---|---|---|
| 0.66 | 38,511ASMC via The Star, to 13 Oct | 58,106ASMC, full year, Suomi-NPP | To date against a full year, two satellites |
| 0.69 | 338,790Ours, S-NPP, 1 Jan – 13 Oct | 490,236Ours, S-NPP, 1 Jan – 13 Oct | Same window, one satellite |
| 0.52 | 338,790Ours, S-NPP, 1 Jan – 13 Oct | 656,254Ours, S-NPP, full year | To date against a full year, one satellite |
Table 4. Sumatra and Kalimantan, 2026 to 13 Oct against 2015. 0.66 is not a same-window ratio. ASMC’s 2015 figure is full-year Suomi-NPP.
Seasons are not ranked on modelled smoke: Copernicus’ reanalysis puts 2015 and 2019 the other way round from its forecasts, for East Malaysia and for the Peninsula
CAMS, ECMWF’s operational forecast, changed model cycle, grid, aerosol species, fire emissions and PM2.5 formula between seasons, so the study checked the forecasts’ order of 2015, 2019 and 2023 against EAC4, CAMS’ reanalysis, which runs one model version throughout. The rule, fixed before any result was seen: a side keeps the forecasts’ order only if EAC4 gives the same order on excess PM2.5 and on smoke-only aerosol optical depth, and every pairwise ratio of excesses is within ±25% of the forecasts’. Both sides fail, so the study puts neither side in order. EAC4 ends in Dec 2025, so 2026 stays forecast-only either way.
| Season | Forecast PM2.5 | EAC4 PM2.5 | Forecast smoke AOD | EAC4 smoke AOD |
|---|---|---|---|---|
| East MalaysiaOrder on excess PM2.5: forecasts 2015, 2019, 2023; EAC4 2019, 2015, 2023. On smoke AOD: forecasts 2015, 2019, 2023; EAC4 2015, 2019, 2023 | ||||
| 2015 | 63 | 65 | 1.57 | 1.46 |
| 2019 | 51 | 74 | 0.87 | 1.16 |
| 2023 | 15 | 27 | 0.19 | 0.39 |
| PeninsulaOrder on excess PM2.5: forecasts 2019, 2015, 2023; EAC4 2015, 2019, 2023. On smoke AOD: forecasts 2015, 2019, 2023; EAC4 2015, 2019, 2023 | ||||
| 2015 | 55 | 87 | 1.32 | 1.31 |
| 2019 | 71 | 78 | 0.59 | 1.12 |
| 2023 | 29 | 35 | 0.18 | 0.41 |
Table 5. Peak-week station-average means, 1 Jul – 15 Oct of each season, each product on its own: PM2.5 in µg/m³, and organic-matter plus black-carbon aerosol optical depth at 550 nm (smoke AOD). Seasons in time order. The DOE monthly anchor (OpenDOSM) covers 2019 against 2023 only.
That anchor sides with EAC4 on the one pair it can test: DOE’s national September PM2.5, via OpenDOSM, was 2.65 times as high in 2019 as in 2023, against 2.42 in EAC4 and 2.25 in the forecasts. OpenDOSM’s national PM2.5 starts in 2018, so it cannot test 2015.
Back-trajectories: 1,984 of the 2,160 centre runs stayed in the box to 120 h, and 90% of the return runs came back within 2 km
The study’s own code on ERA5 hourly winds follows HYSPLIT’s published method, not its software: the two-step position update of Stein et al. (2015) and the 27-member offset-meteorology ensemble of the ARL User’s Guide, in pressure coordinates, without turbulence or convection. Arrivals every 3 h at Johor Bahru and Singapore, at 500, 1,000 and 1,500 m above ground, run 120 h back. An arrival’s label counts when 18 of the 27 members agree, and the study would name one area only if it held 70% of the agreed haze-day arrivals; none does (How it travelled).
8.5% of the trajectories leave the wind box within 120 h, all outside the fire table’s area: 4,272 of the 4,972 across the southern edge, 690 across the western and 10 across the eastern. Integrating forward from the 120 h point of each centre run still in the box returns to the receptor within 2 km in 1,788 of the 1,984, which tests the arithmetic, not the winds.
| Check | Count |
|---|---|
| Centre runs | 2,160 |
| Still in the wind box at 120 h | 1,984 |
| Back within 2 km of the receptor, integrated forward | 1,788 (90.1%) |
| Trajectories, all members, leaving the box within 120 h | 4,972 of 58,320 (8.5%) |
| Of those, leaving inside the fire table’s area | 0 |
Table 6. Back-trajectory checks, arrivals 1 Sep – 15 Oct 2026. Our own code following Stein et al. 2015 and the ARL User’s Guide; HYSPLIT software not used.
Offshore and sea-pass rules
Near-real-time rows carry no FIRMS type, so a detection in a cell where FIRMS typed a VIIRS detection as a volcano, another static source or offshore is dropped. The static-source mask drops 2.6% of the Jul–Oct 2026 near-real-time detections; rebuilt without a year’s Jul–Dec, it keeps 97.9–99.4% of that year’s. A detection more than 0.5 km off land is dropped too, and so is every untyped detection off land in a granule with 50 or more beyond 0.5 km.
The near-real-time files hold whole day passes over the sea that no other satellite saw: S-NPP’s on 14 Oct (658, 657 of them off Papua) and 12 Oct (158, 155 off Maluku), and NOAA-20’s on 3 Oct (452) and 9 Oct (277, 216 off West Papua). In the typed years only 3 to 17 in every 100,000 type-0 detections lie more than 0.5 km out, so the rule drops almost nothing a type would keep. The two rules take 961 and 141 detections off Indonesia’s 2026 S-NPP count, 0.2% in all, and 47 off Malaysia’s.
| Satellite | Indonesia, 0.5 km rule | Indonesia, sea-pass rule | Malaysia, 0.5 km rule |
|---|---|---|---|
| VIIRS S-NPP | 961 | 141 | 47 |
| VIIRS NOAA-20 | 790 | 56 | 157 |
| VIIRS NOAA-21 | 254 | 9 | 6 |
| MODIS | 12 | 0 | 0 |
Table 7. 2026 detections the offshore rules removed, by satellite.
MODIS typing
FIRMS’ older MODIS typing filed many fire-season detections as static land sources (type 2): in 2015, 9,921 of its 11,846 type-2 detections (84%) had an S-NPP type-0 detection within 1 km the same day, and in 2014, 4,441 of 5,883 (75%); every other year since 2012 stays under 20%. Counted with one rule for every year, 2026’s 118,644 is still third, behind 2015 (135,984) and 2006 (124,709).
Not used: NADMA’s air-quality relay, IQAir and WAQI, and the Indian Ocean Dipole
| Source | Why not |
|---|---|
| NADMA MyDIMS air-quality relay | On the 35,551 station-hours it shares with DOE, only 5.45% of its values match DOE’s exactly (5.0–6.3% paired with the hour before or after; mean gap 12–14 API points), and in 2025 it never exceeds 115. |
| IQAir and WAQI | US EPA AQI, on a different scale and averaging window; never mixed with DOE’s API. |
| NOAA PSL Dipole Mode Index (the Indian Ocean Dipole) | NOAA PSL computes it from the Met Office’s HadISST1.1, whose terms page grants the Open Government Licence v2 while its dataset metadata names the Non-Commercial Government Licence v2. It fed no figure, so it was dropped rather than used under a licence in doubt. |
Table 8. Sources left out, and why.
Limits
- Every figure is as of the study date and is not refreshed, though FIRMS may fill its gaps, CPC revises recent seasons and SiPongi may revise its database.
- Seasons are not ranked on modelled smoke; detection counts compare only on the same satellite.
- CAMS is a model: use it for timing and comparison, never as a reading.
- Back-trajectories show paths, not concentrations.
- The study names no company or concession.
Credits and licences
- Fire detections. “We acknowledge the use of data and/or imagery from NASA's Fire Information for Resource Management System (FIRMS) (https://earthdata.nasa.gov/firms), part of NASA's Land, Atmosphere Near real-time Capability for Earth observations (LANCE) (https://earthdata.nasa.gov/lance) and NASA's Earth Science Data and Information System (ESDIS).” Products: MODIS C6.1 NRT, doi:10.5067/FIRMS/MODIS/MCD14DL.NRT.0061; VIIRS S-NPP NRT, doi:10.5067/FIRMS/VIIRS/VNP14IMGT_NRT.002; VIIRS NOAA-20 NRT, doi:10.5067/FIRMS/VIIRS/VJ114IMGT_NRT.002. Also used: MODIS standard processing (MCD14ML, Collection 6 to 2022 and 6.1 from 2023), VIIRS NOAA-21 NRT (VJ214IMGTDL_NRT) and the VIIRS standard-processing files (no DOI found). Image captions: “NASA FIRMS”.
- Boundaries. geoBoundaries: Runfola, D. et al. (2020) geoBoundaries: A global database of political administrative boundaries. PLoS ONE 15(4): e0231866, https://doi.org/10.1371/journal.pone.0231866 (www.geoboundaries.org), CC BY 4.0. Indonesia's provinces, Malaysia's states and the outlines of Timor-Leste, Papua New Guinea and Thailand contain information from OpenStreetMap (© OpenStreetMap contributors, https://www.openstreetmap.org/copyright), via Wambacher's OSM Boundaries (wambachers-osm.website/boundaries) and geoBoundaries, available under the Open Database License (ODbL). Philippines outline: “Philippines - Subnational Administrative Boundaries” (https://data.humdata.org/dataset/cod-ab-phl), National Mapping and Resource Information Authority (NAMRIA) and Philippines Statistics Authority (PSA), contributed by OCHA, via the Humanitarian Data Exchange, CC BY 3.0 IGO. Australia outline: © Commonwealth of Australia (Australian Bureau of Statistics), ASGS Edition 3 digital boundaries, CC BY 4.0. Brunei outline: Wikimedia Commons (public domain). Singapore: “Contains information from Master Plan 2014 Subzone Boundary (No Sea) (Urban Redevelopment Authority) accessed on 16 Oct 2026 from data.gov.sg which is made available under the terms of the Singapore Open Data Licence version 1.0 https://data.gov.sg/open-data-licence” (taken through geoBoundaries' SGP ADM0, built 12 Dec 2023, which records no access date; the date is ours). Singapore is drawn as a point only. Changes: geoBoundaries simplified the Philippines, Papua New Guinea, Thailand and Australia outlines; we repaired and simplified all of them (tolerance 0.0005°) and used them to assign fire detections to provinces, states and countries. The ODbL files are free online at geoBoundaries' commit-pinned URLs:
- github.com/wmgeolab/geoBoundaries/raw/9469f09/releaseData/gbOpen/IDN/ADM1/geoBoundaries-IDN-ADM1.geojson
- github.com/wmgeolab/geoBoundaries/raw/9469f09/releaseData/gbOpen/MYS/ADM1/geoBoundaries-MYS-ADM1.geojson
- github.com/wmgeolab/geoBoundaries/raw/9469f09/releaseData/gbOpen/TLS/ADM0/geoBoundaries-TLS-ADM0.geojson
- github.com/wmgeolab/geoBoundaries/raw/9469f09/releaseData/gbOpen/PNG/ADM0/geoBoundaries-PNG-ADM0_simplified.geojson
- github.com/wmgeolab/geoBoundaries/raw/9469f09/releaseData/gbOpen/THA/ADM0/geoBoundaries-THA-ADM0_simplified.geojson
- Map and chart notice. “Map data from OpenStreetMap (https://www.openstreetmap.org/copyright), via geoBoundaries, available under the Open Database License (ODbL)”, carried under every map in this study and beside every chart or table whose province or state labels come from the OpenStreetMap-based layers.
- Peat. “Global peatland extent.” Accessed on 16 Oct 2026 from Global Nature Watch (CC BY 4.0; Miettinen et al. 2016; Gumbricht et al. 2017). Sampled at fire-detection points by the study.
- Residents. Bondarenko M., Kerr D., Sorichetta A., and Tatem, A.J. 2020. Census/projection-disaggregated gridded population datasets for 189 countries in 2020 using Built-Settlement Growth Model (BSGM) outputs. WorldPop, University of Southampton, UK. doi:10.5258/SOTON/WP00684 (CC BY 4.0). Summed to 1 km by the study. WorldPop's data are “not intended, and should not be used, for purposes that discriminate against, exploit, surveil, or otherwise harm individuals or vulnerable populations”.
- Copernicus. “Contains modified Copernicus Atmosphere Monitoring Service information 2026. Contains modified Copernicus Climate Change Service information 2026. Neither the European Commission nor ECMWF is responsible for any use that may be made of the Copernicus information or data it contains.”
- Copernicus Atmosphere Monitoring Service (2021): CAMS global atmospheric composition forecasts. Copernicus Atmosphere Monitoring Service (CAMS) Atmosphere Data Store, DOI: 10.24381/04a0b097 (Accessed on 16-Oct-2026 and 18-Oct-2026).
- Copernicus Climate Change Service (2025): ERA5 hourly time-series data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). DOI: 10.24381/1cf1ad76 (Accessed on 16-Oct-2026).
- Copernicus Climate Change Service (2026): ERA5 time-series data on pressure levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). DOI: 10.24381/af48f136 (Accessed on 16-Oct-2026).
- Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., Thépaut, J-N. (2023): ERA5 hourly data on pressure levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). DOI: 10.24381/cds.bd0915c6 (Accessed on 16-Oct-2026 and 18-Oct-2026).
- Copernicus Climate Change Service, Climate Data Store, (2023): ERA5 hourly data on pressure levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). DOI: 10.24381/cds.bd0915c6 (Accessed on 18-Oct-2026).
- Copernicus Climate Change Service, Climate Data Store, (2023): ERA5 hourly data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). DOI: 10.24381/cds.adbb2d47 (Accessed on 18-Oct-2026).
- Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., Thépaut, J-N. (2023): ERA5 hourly data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS), DOI: 10.24381/cds.adbb2d47 (Accessed on 18-Oct-2026).
- Copernicus Atmosphere Monitoring Service (2020): CAMS global reanalysis (EAC4). Copernicus Atmosphere Monitoring Service (CAMS) Atmosphere Data Store, DOI: 10.24381/d58bbf47 (Accessed on 18-Oct-2026).
- Inness, A., Ades, M., Agustí-Panareda, A., Barré, J., Benedictow, A., Blechschmidt, A.-M., Dominguez, J. J., Engelen, R., Eskes, H., Flemming, J., Huijnen, V., Jones, L., Kipling, Z., Massart, S., Parrington, M., Peuch, V.-H., Razinger, M., Remy, S., Schulz, M., and Suttie, M.: The CAMS reanalysis of atmospheric composition, Atmos. Chem. Phys., 19, 3515–3556, https://doi.org/10.5194/acp-19-3515-2019, 2019.
- All Copernicus data used here (CAMS forecasts, EAC4, ERA5) are licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Changes: ERA5 winds, geopotential, boundary-layer height and surface pressure were subset to the study area and interpolated by the study to compute its own back-trajectories (the study's calculation, not a Copernicus or ECMWF product); CAMS forecast and EAC4 PM2.5, PM10 and aerosol optical depth were sampled by the study at the DOE station and fire points; the ERA5 point series were taken as the CDS serves them (nearest grid point to 8 smoke-path and 12 fire points); both were summed or averaged to daily, weekly and monthly values by the study.
- Back-trajectory method. Follows HYSPLIT's published scheme: two-step advection as in Stein, A. F., Draxler, R. R., Rolph, G. D., Stunder, B. J. B., Cohen, M. D., and Ngan, F. (2015), NOAA's HYSPLIT atmospheric transport and dispersion modeling system, Bull. Amer. Meteor. Soc., 96, 2059–2077, https://doi.org/10.1175/BAMS-D-14-00110.1, and the 27-member offset-meteorology ensemble described in the NOAA Air Resources Laboratory HYSPLIT User's Guide (https://www.ready.noaa.gov/hysplitusersguide/S253.htm); HYSPLIT software not used.
- DOE monthly means. “Monthly Air Pollution” (air_pollution). Data source(s): Department of Environment, Department of Statistics Malaysia. Via OpenDOSM (https://open.dosm.gov.my/data-catalogue/air_pollution), data as of 31 Dec 2024, last updated 30 Sep 2026. Accessed on 18 Oct 2026. Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). The study used the national monthly PM2.5 only as a ratio check (Sep and Jul–Oct, 2019 against 2023): it averaged months and took ratios and changed no values.
- El Niño. ONI and RONI: NOAA Climate Prediction Center (public domain).
- Air quality. Department of Environment Malaysia (Jabatan Alam Sekitar), Air Pollutant Index Management System (APIMS), https://eqms.doe.gov.my/APIMS/main. “Data dan maklumat tertakluk di bawah Terma Penggunaan Data Terbuka Kerajaan Malaysia 1.0” (https://www.doe.gov.my/wp-content/uploads/2021/07/TermaPenggunaan.pdf).
- Singapore PSI. “Contains information from Pollutant Standards Index (PSI) (National Environment Agency) accessed on 17 Oct 2026 from data.gov.sg which is made available under the terms of the Singapore Open Data Licence version 1.0 https://data.gov.sg/open-data-licence”.
- SiPongi. Source: SIPONGI KEMENHUT (sipongi.gakkum.kehutanan.go.id), beside every SiPongi-derived figure, including the release counts and the burned area quoted through Databoks or Kompas. Counts only: SiPongi's raw points, maps and screenshots never go on a page or into a download (its terms reserve them); only the study's own counts do.
- Licences. CC BY 4.0: https://creativecommons.org/licenses/by/4.0/; CC BY 3.0 IGO: https://creativecommons.org/licenses/by/3.0/igo/; ODbL 1.0: https://opendatacommons.org/licenses/odbl/1-0/.
- No endorsement. No credited source or agency endorses this study, and no figure here is official; the study carries no DOE, ministry or agency logo. These credits go with every copy of every page. Files offered for download carry names only (province, state, country), never boundary geometry. Every cited fact was checked against its source URL on 19 Oct 2026.