About 94,500 hectares burned in July alone, against 107,465 hectares across January through June. That is the number making headlines. The more useful one is that in the three fire years the government just compared 2026 to, roughly three-quarters to more than 90 percent of the annual burned area arrived after July.
On Tuesday 18 August, Indonesia’s Minister of Forestry, Raja Juli Antoni, told Commission IV of the DPR that 202,004 hectares of forest and land had burned nationally between January and July 2026. A month earlier, on 17 July, his ministry had published the figure for January through June: 107,465.47 hectares.
Subtract one from the other and July alone accounts for roughly 94,539 hectares — about 88 percent of what the entire first half of the year produced, compressed into thirty-one days.
Before going further: that July figure is a subtraction, not a published number. The two inputs come from different releases three weeks apart, and Indonesian burned-area statistics are revised as field verification catches up with satellite detection. Treat roughly 94,500 hectares as a good estimate of July’s contribution, not a settled figure. Everything else below rests on numbers the ministry published directly.
January through June averaged about 17,911 hectares per month. July ran at roughly 5.3 times that rate. Put another way, one month out of seven produced 46.8 percent of the year’s burned area so far.
The ministry’s 17 July release gives the provincial breakdown for the first half of the year. The five largest were West Kalimantan (28,680.47 ha), Riau (15,477.95 ha), East Nusa Tenggara (10,538.30 ha), Maluku (7,091.07 ha) and South Papua (6,281.81 ha).
The same release reports that hotspot monitoring through SIPONGI, the ministry’s fire information system running on high-confidence Terra/Aqua detections, was up 166.99 percent through 17 July against the same period in 2025. It also notes that South Kalimantan declared a karhutla emergency alert status running from 6 July to 31 October, and that 180 Manggala Agni personnel were staged across three operational districts there.
Hotspots and burned area are different measurements. A hotspot is a satellite detection of a thermal anomaly; burned area is a mapped extent, verified later. A large percentage change in hotspots does not translate into the same percentage change in hectares, and the two series move on different lags. Both are in this article because the ministry publishes both, not because they are interchangeable.
Raja Juli presented the 202,004-hectare figure against the same January–July window in three previous years:
So 2026 is running about 47 percent above 2019, about 117 percent above 2023, and at roughly a third of 2015. His framing was that this shows fire control improving even as conditions worsen. In his words, as reported by Bisnis.com: “So in 2026 [through July], 202,000 hectares. Already exceeding 2023, exceeding 2019. But with the El Niño signs that were described as the same as 2015. This is still below 2015, and that is our struggle.”
He also gave full-year totals for those years: about 2.61 million hectares in 2015, 1.64 million in 2019, and about 1.16 million in 2023.
That is a reasonable case, and the downward trend in annual totals across cycles is real. But the comparison has a problem inside it, and it is not the one you would expect.
The three comparison years are being treated as a set of El Niño fire seasons. By NOAA’s own records, one of the three was not. Here is where the ocean actually stood during July–August–September, the heart of each fire season, using the Relative Oceanic Niño Index that NOAA has used for official ENSO monitoring since February 2026:
| Year | RONI, Jul–Aug–Sep ERSSTv6 | Burned area, Jan–Jul | Burned area, full year |
|---|---|---|---|
| 2015 | +1.5 | >600,000 ha | ~2.61 million ha |
| 2019 | −0.1 | 137,007 ha | ~1.64 million ha |
| 2023 | +0.8 | 92,924 ha | ~1.16 million ha |
| 2026 | JAS not yet published MJJ +1.0 | 202,004 ha | — |
RONI values from NOAA CPC’s historical episode table (ERSSTv6). Burned-area figures as presented by the Minister of Forestry, 18 August 2026. RONI table retrieved 19 August 2026; CPC notes recent values may be revised.
2019 was not an active El Niño fire season by RONI; the earlier 2018–19 episode had ended by the late dry season. By JAS 2019 — when Indonesia was burning — RONI stood at −0.1. That year still finished at about 1.64 million hectares, more than 2023 managed with El Niño conditions in place.
Which is the whole point, and it is the point worth taking away from this: burned area in Indonesia does not scale with the state of the Pacific. The worst of the three comparison years for ENSO was 2015 and it burned the most; the second worst outcome came in a year whose fire season ran without active El Niño conditions. Something other than sea surface temperature is doing a great deal of the work.
This article does not attempt to establish what drove 2019 specifically. Other climate modes and, more importantly, land use decisions were in play. What can be said from NOAA’s published table is narrower and still useful: 2019’s fire season did not run under active El Niño conditions, so a three-year trend that treats it as one is not measuring quite what it appears to measure.
Now the number that should be leading coverage of this. Take each comparison year’s January–July total as a share of its full-year total:
| Year | Through July | Full year | Share in by 31 July | Share still to come |
|---|---|---|---|---|
| 2015 | >600,000 ha | ~2.61 million ha | >~23% | <~77% |
| 2019 | 137,007 ha | ~1.64 million ha | ~8.4% | ~91.6% |
| 2023 | 92,924 ha | ~1.16 million ha | ~8.0% | ~92.0% |
Arithmetic on the totals presented to Commission IV, 18 August 2026. The 2015 and 2023 annual figures were given as approximations, and the 2015 January–July figure is reported only as more than 600,000 hectares, so its shares are bounds rather than point values.
In every one of those years, the great majority of the annual burn came after 31 July. In 2019 and 2023, more than ninety percent of it did. Indonesia’s fire season peaks in the late dry season — broadly August into October — and July is the ramp, not the event.
So 202,004 hectares through July is not this year’s fire season. It is the prologue to it.
There is a second, less comfortable implication. The through-July share ranged from about 8 percent to more than 23 percent across just three years — a spread of roughly threefold. That is not a stable ratio. It means the January–July total constrains the annual outcome very weakly. Anyone converting 202,004 hectares into a confident projection for 2026 is doing something the historical record does not support.
To be explicit about what this is not: the shares above are arithmetic on four reported totals. They are not a forecast, and they are not a model. They are included to show how wide the range of outcomes has been from a similar point in the calendar, which is an argument for waiting rather than an argument for any particular number.
El Niño does not start fires in Indonesia. Almost all of them are set deliberately, to clear land. What El Niño does is change the odds that a set fire escapes and keeps going.
The mechanism is documented in NOAA’s current diagnostics. The ENSO Diagnostic Discussion of 13 August 2026 states that convection and rainfall “were enhanced from the central to eastern Pacific and were suppressed over Indonesia.” That is the link, stated plainly by the agency: the ascending branch of the Walker circulation shifts east, and the rainfall that would normally sit over the Maritime Continent goes with it.
Three things follow. Fuel dries out, so vegetation that would not carry fire in a normal July will carry it. The burnable window lengthens at both ends, so a season that might run six weeks runs twelve. And drained peatland — which holds enormous quantities of carbon and, when wet, simply will not burn — drops below the moisture threshold where subsurface smouldering becomes possible. Peat fires are the ones that produce the regional haze, that resist suppression because the combustion is underground, and that keep burning through rainfall.
The ministry’s own framing matches this. Its 17 July release describes peatland rewetting through water-gate management to keep the water table up, cloud seeding operations that began 15 July based on BMKG cloud-potential forecasts, and an enforcement track running administrative sanctions against concession holders, civil suits for environmental restoration, and criminal prosecution of individuals and corporations. Those are all interventions on the human side of the equation, because that is the side that can be intervened on.
Which is why the comparison across cycles is genuinely informative even after the 2019 problem is accounted for. Between 2015 and 2023 the annual totals fell by more than half. Some of that is ENSO luck. Some of it is that moratoria, restoration, patrol capacity and prosecution changed the human inputs. Disentangling the two is not something a burned-area table can do.
One caution about the ocean figures presented alongside the fire data, because readers will encounter them.
Bisnis.com reports BMKG’s Niño 3.4 index at +2.45 for the third dasarian of July and +2.19 as the monthly value. In the hearing the minister said the current index was “already 2.9,” that July–September 2015 was “only plus 1.5,” that November 2015 reached +2.57, and that “we hope we will reach 2.47 like 2015.” Those four numbers do not reconcile with each other, and none of them reconciles cleanly with the +2.19 reported in the same article. The discrepancies may reflect transcription, translation, or the mixing of different Niño indices and averaging periods. Whatever the cause, they are a reason not to build a 2026-versus-2015 severity judgment on them.
The deeper issue is that “Niño 3.4” is not one number. For July 2026, three agencies published three values for the same region in the same month:
These values are not necessarily contradictory. Differences in source dataset, climatological baseline, temporal averaging, and relative-versus-conventional treatment can each account for part of the spread — which is the subject of yesterday’s article, and the reason NOAA now monitors ENSO on RONI rather than on the conventional index. For cross-year comparisons of event severity, the published seasonal tables are the tool. NOAA’s current RONI table (ERSSTv6) gives 2015 as +1.5 in JAS, +1.8 in ASO, +2.0 in SON and +2.3 in NDJ. Those are the numbers a 2026-versus-2015 comparison has to be made against. RONI for 2026 currently runs through May–June–July at +1.0; the JAS value covering this fire season has not been published yet.
What NOAA has said about 2026 is that El Niño is strengthening, with a greater than 90 percent chance of a very strong event during Northern Hemisphere fall and winter, and a 69 percent chance during October–December of a historic event exceeding every El Niño back to 1950 on a three-month RONI basis.
Four things, in rough order of how much they will tell you.
The August and September burned-area releases. August through December carried roughly three-quarters to more than 90 percent of the annual load in the comparison years; the monthly releases are the only way to see where inside that window 2026 lands. If August comes in near July’s ~94,500 hectares, 2026 is tracking an ordinary bad year. If it comes in at two or three times that, the 2015 comparison starts to matter.
Whether peat provinces take over the totals. The January–June leader board was headed by West Kalimantan and Riau, both peat-heavy, but also included East Nusa Tenggara, Maluku and South Papua, which are largely savanna and dryland burns. Savanna fires produce hectares. Peat fires produce haze, health emergencies and diplomatic incidents. The composition matters more than the total.
Transboundary haze. The signal to watch is not smoke over Sumatra and Kalimantan, which is routine at this point in the year, but sustained air quality degradation and visibility reductions in Singapore and peninsular Malaysia. That has not become a significant regional issue in 2026 so far.
NOAA’s next updates. The ENSO Diagnostic Discussion is due 10 September, which is when the strength probabilities are revised. The RONI table updates separately, normally near the start of each month. The value to wait for is JAS 2026 — the first chance to place this fire season’s ocean state next to 2015’s +1.6 on the same measurement.
The July number is real and it is steep. But the honest summary of where Indonesia stands is that the fire season being discussed has not happened yet, and the record of the last three cycles says the part that has happened tells you remarkably little about the part that hasn’t.
Separately, and not an error at the time, NOAA has since migrated RONI from ERSSTv5 to ERSSTv6 and revised the full 1950–present series. The 2015 sequence originally published here (+1.6 JAS, +1.9 ASO, +2.2 SON, +2.4 NDJ) were the correct ERSSTv5 values when this piece ran; the article now carries the current ERSSTv6 figures (+1.5, +1.8, +2.0, +2.3). The ERSSTv5 series remains published by CPC at /roni/v5/.
Neither change alters this article’s argument. JAS 2019 remains −0.1 on both series, so 2019’s fire season still did not run under active El Niño conditions, and the ordering of the three comparison years is unchanged.