How the 2015 Southeast Asian fire complex began: a study says many of them, not one

On 17 July 2026, researchers published findings in npj Natural Hazards showing that the catastrophic 2015 fire season across equatorial Southeast Asia was not, for the most part, a single-source catastrophe. Tracing the earliest detectable hotspots in satellite records, the team found that a large majority of the most damaging fires had multiple independent origin points rather than a single spark that then spread. The work reframes a disaster long treated in shorthand as a haze-and-deforestation story driven by one or two bad actors, and it sharpens the policy question from who lit the match to who lit the matches.

The 2015 event released an estimated 11.3 million tonnes of carbon according to prior published estimates cited in the new paper, blanketed Singapore, Malaysia and Thailand in hazardous particulate matter, and cost Indonesia an estimated $16 billion in damages. The new analysis does not dispute those totals. It argues instead that the shape of the disaster was different from what most prevention strategies assume: not a handful of large, attributable ignitions but a distributed wave of small starts, many of which coalesced into the conflagrations that closed airports and pushed air quality past hazardous thresholds in six countries.

What the satellite record actually shows

The study used daily active-fire detections from NASA's MODIS and VIIRS instruments to identify ignition locations across Sumatra, Kalimantan (Indonesian Borneo) and peninsular Malaysia during the southern dry season of 2015. Each detected hotspot was mapped against land-cover type, peat-soil distribution and concession boundaries (oil-palm, timber, smallholder agriculture). The authors then reconstructed probable fire-perimeter growth using a perimeter-expansion model calibrated to wind and fuel conditions for the period.

The headline finding: of the fires that ultimately burned more than 1,000 hectares, the majority showed two or more spatially and temporally distinct ignition points within the first 72 hours of detection. In roughly a third of those large events, ignition points were separated by more than five kilometres and several days, ruling out simple perimeter spread from a single start. A smaller subset, fewer than one in five of the large fires, did appear to originate from a single ignition that then ran under continuous fuel loads, including drained peat domes where smouldering combustion persisted for weeks.

In plain terms: the dominant pattern was not one big fire that got out of hand. It was many fires, started in different places at different times, that grew into a coordinated regional smoke crisis because weather and fuel conspired to merge their smoke footprints.

The single-ignition story, and where it breaks

The conventional account of equatorial Southeast Asian fires has leaned heavily on a small set of high-profile cases: the 1997–98 fires blamed largely on plantation concessionaires during an El Niño drought; the recurring peat fires in former Mega Rice Project concessions in Central Kalimantan; and the recurring burning of drained peatlands that Indonesia's own president called a "national disaster" in 2015. That account is not wrong on the facts. It is wrong, the new study argues, on the generalisability of those facts to the broader fire complex.

If most large fires have multiple origins, then prevention strategies built around catching a single culprit for each disaster will systematically miss the structure of the problem. The policy levers that work against a single-source ignition (targeted enforcement against a known concession, a satellite alert routed to one company) are blunt instruments against a wave of small, distributed starts. The study does not exonerate any actor; it argues the actor-mix is broader and more fragmented than the prevailing narrative implies, with smallholder agricultural burning and drainage-canal ignition both contributing meaningfully to the multi-origin fires in the sample.

This complicates, rather than dismantles, the standard line. Indonesian and regional authorities have spent the past decade building attribution capacity: the Sipongi peat monitoring dashboard, the moratorium on new peat concessions, and the haze-free ASEAN agreement signed in 2014 (before this fire season) all rest on the assumption that fires are knowable to a single responsible party. The new data suggests the institutional design may need to be matched to a problem that is smaller per-fire but more fires at once.

Structural frame: a fire problem that scales with land use, not with weather alone

The deeper finding sits inside the structural pattern of equatorial Southeast Asian land use. Peat domes drained for oil-palm and acacia plantations remain the slow-burning backbone of the worst haze episodes; that is established. What the origin-point analysis adds is the observation that the rate at which fires are lit on those drained landscapes is itself a function of how fragmented the landscape is. Highly concessioned, monoculture blocks tend to produce single-source large fires because the fuel is continuous. Mosaic landscapes of smallholder plots, concession edges, and remnant forest produce many small starts that aggregate.

Both patterns sit inside the same regional economy: Indonesian and Malaysian palm-oil expansion, pulpwood plantations supplying paper mills in both countries, and the smallholder agriculture that fills the matrix between concession boundaries. The fire problem is, in other words, a problem of how a commodity-driven landscape is broken up on the ground. Continuous fuel produces one big disaster; fragmented fuel produces many small ones whose smoke merges into the same regional crisis. The carbon totals and the public-health costs end up comparable. The policy implications diverge sharply.

What changes for prevention, and what doesn't

For prevention agencies in Jakarta, Kuala Lumpur and Singapore, the practical shift is from catching the largest fires to interrupting the rate of new ignitions during dry spells. The study points to early-warning systems tied to fire-danger indices rather than to single-perimeter monitoring, and to drainage-canal water management that keeps peat water tables high enough that ignition does not translate into smouldering. Neither lever is new; both have been underfunded relative to the fire-suppression capacity that has grown since 2015.

For the palm-oil and pulp-and-paper sectors, the data complicates the certification logic that has dominated the past decade. Roundtable on Sustainable Palm Oil (RSPO) certification and similar schemes are designed around auditable single-actor supply chains. A fire complex in which most of the burning originates on concession edges and smallholder land is partially outside that audit perimeter. The credible industry response will need to extend into landscape-level monitoring rather than parcel-level certification alone, which the larger Indonesian and Malaysian producers have begun to do but unevenly.

For Singapore and the regional downwind states, the diplomatic question is unchanged: how to ensure Jakarta's sovereignty over land-use decisions inside its borders coexists with cross-border accountability for transboundary haze. The new data does not settle that question. It does sharpen it: the fire complex that closed Singaporean schools in 2015 was, in the new framing, less a single failure of Indonesian enforcement and more a distributed property of a regional land economy. The diplomatic arithmetic is harder when the responsible parties are many.

What remains uncertain

The paper's reconstruction depends on MODIS and VIIRS active-fire detection, which under-counts smouldering peat fires relative to flaming front fires because peat combustion produces lower thermal signatures. The multiple-origin finding is therefore most robust for the flaming-front fraction of the complex, and least robust for the long-tail peat emissions that drive much of the public-health damage. The authors flag this honestly. Independent corroboration, ideally from Sentinel-1 SAR-based peat-disturbance mapping and from on-the-ground ignition reports by Indonesian and Malaysian fire services, will be needed before the multi-origin conclusion is treated as definitive for the full fire complex.

The 2015 season was also an El Niño year, which compressed the dry period and intensified fuel conditions. Whether the multiple-origin pattern holds in non-El Niño fire seasons, or whether it is itself a property of extreme drought conditions, is an open question the paper raises but does not resolve. The answer will determine whether the prevention implications travel beyond the worst-case years.

Desk note: Monexus framed this as a structural finding about how a disaster is built up from many small starts, not as a blame-attribution story. The Western wire line tends to compress complex fire seasons into single-culprit narratives; the new data argues that compression is the wrong frame for prevention policy.