BMKG's July 30 briefing put 5,019 fire hotspots on the board as of July 29, with 54.5 percent of Indonesia's land area already in its dry season and the peak forecast for August. The government has been spending money and flying aircraft for months in advance. This is anticipatory action in the ordinary sense, resources deployed before the expected peak, which raises the question nobody wants: if it works, how would anyone know?
Three weeks ago we covered the UN moving $100 million before El Niño peaks, and the concept behind it: spend on the forecast rather than on the aftermath. That post described the idea. Indonesia is now running the experiment, with its own equipment, on its own territory, and the operational record is specific enough to check.
It is also being reported with three different percentages that answer three different questions, and with a hotspot figure whose date moved by two weeks in translation. Here is the clean version.
BMKG's own press release of July 30 gives an operational log rather than an intention. Seventeen weather modification operations, which the agency abbreviates OMC, have been run for forest and land fire mitigation during 2026 across Aceh, Riau, Jambi, South Sumatra, the Riau Islands, West Kalimantan, Central Kalimantan and South Kalimantan. Three separate reservoir-filling operations have been completed in North Sumatra, the Riau Islands and Central Sulawesi, with four more targeted at the Lake Toba catchment and the Citarum, Brantas and Mamasa river basins.
The named dates are the useful part, because they are checkable. Operations finished in Aceh on July 26 and in Palembang on July 30. Seeding was still running in Palangkaraya from July 27 to August 2, and in Pekanbaru from July 30 to August 4. Two new operations posts are being stood up, one in Pekanbaru for fires in Riau and one in Bandung for drought in West Java. Funding for the current South Sumatra and Central Kalimantan work comes from the Ministry of Forestry and the disaster agency BNPB.
On the agricultural side, the national news agency ANTARA reported on July 31, citing official statements, that the Ministry of Agriculture has distributed 100,000 water pumps, rehabilitated 900,000 hectares of irrigation networks, deployed drought-resistant seed varieties and is drawing on roughly 800,000 hectares of moisture-retaining swampland. Government-managed rice reserves were put at 5.24 million tonnes as of late July, with standing crops estimated at 10 to 11 million tonnes. We could not locate a primary ministry release carrying those four figures, so treat them as officially stated and wire-reported rather than independently documented.
This is where coverage keeps slipping, and the numbers are close enough together to make the slip easy.
| Figure | What it counts | What it answers |
|---|---|---|
| 54.5% | 509 seasonal zones already in dry season, through mid-July | Where the dry season has arrived |
| 48.84% | 369 seasonal zones whose dry season peaks in August | When the worst of it lands |
| 48.77% | 437 seasonal zones facing a longer than normal dry season | How long it lasts versus climatology |
The middle two differ by seven hundredths of a percentage point and describe entirely unrelated quantities. One is a forecast of peak timing; the other is a forecast of duration relative to normal. They also rest on different zone counts, 369 against 437, which is the giveaway that they are not the same measurement rounded differently. A further 25.41 percent of land area peaks in September, and BMKG's June release put the July peak at 12.26 percent.
For the remainder of the picture, BMKG reports 77 zones, 11.8 percent, still in their rainy season, and 113 zones, 33.7 percent, in areas classified as having a single season rather than a wet and dry pair. High rainfall is not expected to return gradually until October or November.
One more figure worth having: the longest recorded dry spell as of the briefing was 80 consecutive days without rain, at Katerban in Kaligesing, Purworejo Regency, Central Java. That is a single station, not a regional average, and it should be quoted as such.
BMKG's Indonesian-language release states plainly that as of July 29, 2026 it had detected 5,019 hotspots, dominated by West Kalimantan, South Papua and Riau. ANTARA's English wire copy rendered the same figure as "5,019 wildfire hotspots as of mid-July 2026." The count is identical; the date is two weeks apart. If you are citing the number, cite the July 29 date from the primary release.
Two more cautions belong with it. A hotspot is a satellite thermal anomaly, not a confirmed fire, and hotspot totals are sensitive to which sensor and which confidence threshold are used, which is why different Indonesian agencies routinely publish different counts for the same week. And the word is doing double duty in the current coverage: ANTARA also ran a July 31 story about Indonesia "tackling 294 drought hotspots," which refers to agricultural areas under drought stress, not thermal detections. Those two "hotspot" numbers cannot be compared.
Deputy for Climatology Ardhasena Sopaheluwakan said accumulation charts show hotspot numbers from January to mid-July 2026 rising faster than in the previous El Niño years of 2015, 2019 and 2023. Read that carefully. It is a claim about the shape of a cumulative curve, not a forecast of burned area, and 2015 remains the benchmark catastrophe by outcome. We made the same distinction in July's look at the "Godzilla" label and what Indonesia's fire numbers actually show, and it has not stopped being the most misused sentence in this story.
BMKG reported its monthly Niño 3.4 index at 1.56 at the end of June, and classified that as above its strong-event threshold of 1.5. NOAA's familiar strong-El Niño classification is normally applied to a seasonal three-month magnitude rather than to a single monthly reading, so the threshold here is BMKG's own and should be attributed that way. Meanwhile NOAA CPC's weekly relative table, which this site tracks, shows Niño 3.4 at +1.5°C for the week centered July 29, with Niño 1+2 at +3.2°C, Niño 3 at +2.1°C and Niño 4 at +0.2°C.
None of those are in conflict and none is stale. The relative index subtracts the tropical mean SST anomaly, so it reads lower than the conventional index by construction, and the size of that gap is visible in the same week: for the week centered July 29 the conventional Niño 3.4 anomaly was +2.3°C against the relative +1.5°C, a difference of 0.8°C. A monthly mean and a weekly value are separately not interchangeable. Substituting one for the other, in either direction, produces a wrong sentence, and it happens constantly. Our explainer on why NOAA changed how El Niño is measured covers the mechanics.
BMKG also flags a potential positive Indian Ocean Dipole in the September to December window, citing an Indian Ocean sea surface temperature value of -0.387. The release does not specify which index that figure comes from, and published IOD values for mid-2026 vary by dataset and baseline, so we are quoting it as BMKG's stated value rather than assimilating it to any particular index. What matters for Indonesia is the co-occurrence: a positive IOD alongside El Niño tends to suppress rainfall further over the maritime continent. Our IOD explainer covers why that combination is the one to watch.
Ardhasena added the qualification that most coverage drops entirely: although El Niño continues into early 2027, its impact on Indonesia is concentrated in the period when it overlaps directly with the dry season. An event that persists is not the same as an event that keeps hurting.
Buried at the end of BMKG's sectoral guidance is a recommendation that never makes the international coverage. Alongside warnings to agriculture, water, energy, health and forestry, the agency notes that operators in fisheries and salt farming can use this dry season to their advantage, optimising salt production and taking advantage of upwelling to improve fish catches.
That is not a silver lining offered to soften bad news. Dry weather favours evaporation in salt ponds, while seasonal upwelling can increase nutrient availability and support fisheries. A country-level account of El Niño that contains only losses is not a complete account, and BMKG's own release does not pretend otherwise.
Anticipatory action has a measurement problem baked into it. If the pumps, the seeding and the irrigation work as intended, the observable outcome is a season less bad than it would otherwise have been, and a disaster that is reduced or avoided does not reveal the untreated counterfactual directly. You cannot run Indonesia's 2026 dry season twice.
This is not an argument against spending early. The economic case for anticipatory action rests on ex ante expected value, not on post hoc attribution, and the alternative is to wait for the burned area and then pay more. But it does mean that claims of success and claims of failure will both be cheap this year, and that anyone asserting either should be asked what evidence would have changed their mind.
The season will not be short of data, either. Rainfall, seeded and unseeded areas, fire detections, burned area, peat moisture, agricultural output and comparable historical seasons all get recorded, and matched comparisons, historical controls, meteorological adjustment and occasionally randomised seeding trials can estimate an effect, imperfectly. What none of it delivers is the untreated version of this particular season.
So the most useful evidence available in real time is comparative rather than causal. If the hotspot accumulation curve finishes the season above 2015 while burned area finishes below it, that combination is consistent with suppression working. It is not proof, because ignition conditions, rainfall timing and peat moisture all differ between events. It is simply the closest thing to a test this situation permits without a formal evaluation, and it is worth stating in advance rather than reconstructing afterward.
The useful thing about this case is that Indonesia published its homework in advance. Dates, locations, counts and funding sources are all on the record before the peak arrives, which means the claims can be checked rather than merely believed. That is rarer than it should be. You can follow the Pacific side as it updates on our teleconnections dashboard.