In Brazil’s Amazonas state, satellite hotspots have more than doubled 2025’s weak count, but remain below the 1998 to 2025 average and far below 2024. The October surge is real; “record fire season” is not.
In Brazil’s Amazonas state, INPE counted 7,865 satellite hotspots from 1 January to 13 October, against 3,639 in the same period of 2025. That is 116 percent more, but it is 12 percent below INPE’s 1998 to 2025 mean for the same dates and about one third of the same-date 2024 count. Meanwhile Censipam’s Painel do Fogo returned 1,679 fire events in Amazonas for the selected 1 to 14 October window, up 253 percent on a year earlier, but its latest event date was 13 October, so that is not a fully observed 14-day comparison for 2026. The October surge is real. A record fire season is not what the data show.
Our 17 September piece was about water, and it did not forecast fires. This one is about what is happening on the land, which satellite system is saying it, and what El Niño can and cannot be credited with.
The central numbers here come from the two systems themselves, read on 14 October. For INPE we read Programa Queimadas’ Situação Atual and Estatísticas pages for Amazonas, which count hotspots from the reference satellite, with data to 13 October. For Censipam we read the Painel do Fogo FAQ and the indicators panel’s data feed for Amazonas. For the El Niño context we read the INMET and INPE spring outlook and the September Painel El Niño bulletin, and the abstracts of two peer-reviewed papers. News reporting supplies only what the minister and researchers said and what the response looks like.
The reporting checks out against the sources. The 7,815 and 3,611 that Amazonas’s audit prosecutors printed on 13 October are INPE’s counts a day earlier: 7,865 less the 50 hotspots of 13 October, and 3,639 less the 28 of 13 October 2025. Reuters’s 1,679 events and 253 percent are exactly what the Painel do Fogo returns for 1 to 14 October, 1,679 against 476. We did not use Agência Brasil’s 1,380 for the first ten days of October because the panel returns 1,203 for 1 to 10 October.
They are not the same measurement, and the difference is the story. INPE counts hotspots. The Painel do Fogo, by its own FAQ, counts fire events rather than hotspots, and it was built to help dispatch firefighters, not to keep a long series.
| INPE Programa Queimadas | Censipam Painel do Fogo | |
|---|---|---|
| What is counted | Hotspots: a satellite detection of heat at the moment of an overpass | Fire events: a polygon formed when at least 3 hotspot buffers of 400 m overlap over at least 1 km² in rural areas |
| Satellites | One reference satellite (the MPC bulletin names AQUA_M-T) | VIIRS on S-NPP, NOAA-20 and NOAA-21, and MODIS on Aqua and Terra |
| Built for | A consistent series since 1998. INPE advises comparing monthly totals, not daily values | Dispatching firefighting teams. The panel does not display past events; its indicators panel uses the stored history |
| Amazonas, latest read | 1 Jan to 13 Oct: 7,865 hotspots (3,639 in 2025) | 1 to 14 Oct: 1,679 events (476 in 2025) |
Put on matching windows, the two systems agree on direction and disagree on size. In September INPE’s hotspots rose 296 percent and Censipam’s events 114 percent. In October so far INPE is up 350 percent and Censipam about 263 percent. Reuters’s 253 percent compares 13 days of 2026 events with 14 days of 2025, because the panel’s latest event date was 13 October when we read it, so a like-for-like figure is a little higher. Neither number can be converted into the other, and neither can be divided by the year-to-date 116 percent, which covers 285 days against these two weeks.
| Window | INPE hotspots, 2026 vs 2025 | Censipam events, 2026 vs 2025 |
|---|---|---|
| 1 to 30 September | 3,726 vs 942 (+296%) | 2,388 vs 1,114 (+114%) |
| 1 to 13 October | 1,608 vs 357 (+350%) | 1,679 vs 462 (+263%) |
| 1 to 14 October, as published | Not published for this window | 1,679 vs 476 (+253%) |
A hotspot says a satellite saw heat when it passed. It says nothing about area burned, emissions or why it was burning, and INPE’s own table notes that comparing daily values is not recommended, only monthly totals. The daily series shows why. The 540 hotspots in two days that Agência Brasil reported as more than half of October 2025’s total are INPE’s 494 on 10 October and 46 on 11 October. The 494 alone was 57 percent of all of October 2025’s 874, and 7 October had zero.
INPE’s table lays out the baseline. 2025 was a very quiet year: 3,639 by 13 October, against 23,497 on the same date in 2024. The 1998 to 2025 mean for the same dates is 8,922, and 14 of the 28 earlier years were higher than 2026’s 7,865. So 2026 sits near the middle of the record, not at the top.
| INPE Amazonas, reference satellite | 2025 | 2026 | 1998 to 2025 mean |
|---|---|---|---|
| January to August | 2,340 | 2,531 (+8%) | about 4,970 (2026 is 51%) |
| September | 942 | 3,726 (+296%) | 3,211 (2026 is 116%) |
| 1 to 13 October | 357 | 1,608 (+350%) | 1,433 for the whole month (2026 is already 112%) |
| 1 January to 13 October | 3,639 | 7,865 (+116%) | 8,922 (2026 is 88%) |
The shape of the year is the finding. From January to August 2026 Amazonas ran 8 percent above 2025 and about half of the long-term mean, with July at 14 percent of its mean and August at 59 percent. Then September came in 16 percent above its mean, and the first 13 days of October already exceed the whole-month October mean by 12 percent. Between 1 September and 13 October the count rose 4,035 above 2025, which is 95 percent of the entire year-to-date surplus of 4,226.
Our reading. The honest sentence is that Amazonas fires have surged since September, from a dry-season start that was well below normal, and that the year to date still sits below its long-term mean. Both halves belong together. A reader who sees only “doubled” will picture 2024, and 2026 is nowhere near it on INPE’s count.
The “below average” half has a shelf life. The mean for 1 January to 31 October is about 9,600, which is 1,752 above where the count stood on 13 October. October’s first 13 days ran at 124 hotspots a day, which would close that gap by month end if the pace held. INPE’s daily values are volatile, so that is arithmetic to check, not a forecast.
El Niño raises the risk of the conditions that favour fire and smoke, and it does not ignite anything. CPC’s 8 October discussion calls a strong to very strong El Niño likely through January to March 2027, and the weekly relative Niño 3.4 index was +2.4°C for the week centred 7 October. What CPC establishes is that the event is strengthening and that a stronger El Niño raises the odds of its characteristic impacts. It does not say this Amazonas surge was made hotter, drier or less windy by El Niño, and neither does anything else we read.
What Brazil’s agencies say. The joint INMET and INPE spring outlook, with the regional forecast credited to CPTEC, INMET and FUNCEME, favours below-average rainfall across nearly all of the North for October to December 2026, with deficits of up to 200 mm in Roraima, north-northeast Amazonas and parts of Pará, and temperatures up to 2°C above average. The September Painel El Niño bulletin, issued jointly by INMET, INPE, ANA, Cemaden, SGB, Sedec and Censipam, reports June to August rainfall below the historical average in most of the North and Northeast, and a rapid worsening of drought in Roraima, Acre and Amazonas in August. It calls that rainfall pattern consistent with other El Niño events, but it does not attribute the North’s dryness to El Niño outright, and the outlook note does not either. The bulletin also reports that fire events in July and August were about 44 and 37 percent below their 2020 to 2025 averages before rising, which matches the shape INPE’s hotspot record shows.
What the literature says. Peer-reviewed work supports a tilt in the odds, not a rule. Jiménez-Muñoz and colleagues found that during the 2015 to 2016 El Niño the typical El Niño drought appeared only in eastern Amazonia while the west turned unusually wet, a dipole they link to where Pacific warming peaked. Marengo and Espinoza’s review finds that some Amazon rainfall extremes were linked to El Niño and La Niña, but not only to them, with moisture transport from the tropical Atlantic also shifting the water cycle. So the effect on a given part of the basin depends on the event, and an El Niño being strong does not by itself say how dry Amazonas will be.
Folhapress ties the region’s drought and heat to El Niño and says weaker winds help smoke accumulate over Manaus. That is the newspaper’s attribution, and we read no wind measurement. The MPC bulletin has the Rio Negro at Manaus at 17.62 m on 13 October, still falling about 8 cm a day, and an INMET low-humidity alert with relative humidity as low as 20 to 30 percent in the east of the state, but the bulletin does not tie each figure to a named gauge or station.
Our reading. The forecast and the observed regional rainfall are consistent with what El Niño tends to bring to northern Brazil, and that is why this site says El Niño raises the risk. We read no Amazonas-specific rainfall anomaly, soil-moisture or fuel-moisture measurement and no attribution study of this surge, so this site does not say El Niño dried the fuel, made the season hotter, drier or less windy, or caused any share of the hotspots. INPE’s own monthly record adds a reason for caution: July and August were well below average and the surge began in September, so the hotspot count alone does not read as a season steadily driven by dryness. Our September piece also carried World Weather Attribution’s finding that for the 2023 Amazon drought El Niño and climate change contributed comparably to the rainfall deficit, while warming was the dominant reason the drought reached an exceptional intensity.
The statement on the record is from Environment and Climate Change Minister João Paulo Capobianco. After a flyover on 11 October he said he believes the fires were set deliberately, possibly to degrade the biome and clear land, and that Amazon fires do not start on their own. Agência Brasil reports he saw fires in hard-to-reach forest, and Reuters says his ministry has promised an investigation. Ane Alencar of IPAM told Folhapress that much of the burning follows the BR-319 highway and reflects land occupation, that fire is used in production during dry periods, and that forests do not burn without people using fire. Reuters adds that Amazon fires are rarely natural and are usually set to clear land for farming or ranching.
The limit is that a minister’s belief and an investigation promised are not findings about any specific fire. A hotspot or a fire event looks the same whether it is arson, clearing or an agricultural burn that escaped.
Manaus’s Civil Defense issued a health-risk alert on 9 October, and Agência Brasil reports provisional emergency declarations in 36 Amazonas municipalities. Reuters reports that outdoor events were cancelled in Manaus and public schools closed on Wednesday. Readings from a university network of low-cost sensors, quoted in the MPC bulletin, ran about 39 to 63 µg/m³ in Manaus on the morning of 13 October after being above 130 the day before, with an instantaneous reading of about 324 µg/m³ in Autazes. These are not regulatory stations and not 24-hour averages. A joint task force with 290 firefighters is working the fires.
Our reading. The weather and the people do different jobs. Hot, dry conditions raise the risk that fire spreads, light winds let smoke linger over a city, and ignition and land use set where fire starts. Satellite systems register the result in different units, and none of them can apportion it. That is the case for saying El Niño raises the risk and people ignite the fires, and for assigning no share.
Read INPE’s Amazonas monthly table for October once the month closes, and the Painel do Fogo indicators panel for the same dates. Look for an Amazonas rainfall anomaly or a soil-moisture or fuel-moisture measure from Cemaden or INMET before saying anything about dry fuel, and for the next monthly Painel El Niño bulletin, and for the investigation the Ministry of Environment has promised.