Southeast Asian haze is a recurring episode of regional air pollution in which smoke from burning vegetation and organic soils accumulates over populated areas and crosses national borders. The best-known episodes originate largely in Sumatra and Indonesian Borneo and affect Indonesia, Malaysia, Singapore, and sometimes neighboring countries. Mainland Southeast Asia also experiences extensive seasonal smoke pollution, with distinct agricultural practices, fire seasons, and atmospheric circulation. In both settings, the haze consists principally of suspended particles and gases rather than water droplets, although humidity can greatly increase its optical effects.
The phenomenon connects land management with public health, atmospheric processes, and regional diplomacy. Severe episodes disrupt transport, schooling, outdoor work, and economic activity, while fires damage forests and release carbon stored over centuries or millennia. Indonesian peat fires are particularly important because they can burn beneath the surface for prolonged periods and produce dense smoke even when visible flames are limited.
Haze-producing fires are not adequately explained as exceptional natural disasters. Drought and seasonal winds influence their severity and geographical reach, but ignition, drainage, vegetation clearance, and the management of agricultural landscapes strongly determine where fires occur and whether they become persistent. The regional problem therefore involves both immediate smoke exposure and the longer history of environmental change that makes landscapes combustible.
Regional extent and the nature of the pollution
In maritime Southeast Asia, major smoke-source regions include eastern and southern Sumatra and the Indonesian provinces of Kalimantan on Borneo. Extensive lowlands in these areas contain peat-swamp forests, agricultural estates, smallholder farms, and previously cleared land. Smoke can affect communities close to the fires before traveling across the Strait of Malacca, the South China Sea, or other surrounding waters. Singapore, Peninsular Malaysia, and Malaysian Borneo experience different exposure patterns because their positions relative to individual fire clusters and prevailing winds differ.
Mainland Southeast Asian smoke pollution has a different seasonal geography. Burning in Myanmar, northern Thailand, Laos, Cambodia, and Vietnam is associated with crop-residue disposal, shifting cultivation, forest fires, and other land-management activities. Pollution commonly intensifies during the late dry season, often between February and April in northern parts of the mainland. Mountain valleys can trap emissions near settlements. Maritime haze, by contrast, is especially associated with the relatively dry southwest monsoon period and its extension during unusually dry years. These are related regional problems, but they should not be treated as a single plume or a uniform agricultural system.
The defining pollutants include particulate matter, carbon monoxide, volatile organic compounds, and other combustion products. Fine particles designated PM2.5 have aerodynamic diameters of micrometres or less. They can remain airborne long enough to travel hundreds or thousands of kilometres. Biomass smoke contains substantial organic material, with variable amounts of black carbon, inorganic salts, and other components. Its composition changes with fuel type, combustion temperature, and the time elapsed since emission. Smoke also mixes with urban, industrial, and transport pollution, complicating the interpretation of measurements in cities.
Haze is therefore not synonymous with either fog or all forms of poor air quality. Fog consists mainly of suspended water droplets, whereas smoke haze is an atmospheric aerosol produced by combustion and subsequent chemical processing. Nevertheless, smoke particles absorb water at high relative humidity, increasing light scattering and reducing visibility. A city may consequently experience a marked deterioration in visibility without a proportionate increase in dry particle mass. Conversely, unhealthy concentrations can occur when visibility appears less dramatic. Air-quality measurements, satellite observations, and meteorological information are needed to characterize an episode rather than relying on the appearance or smell of the air alone.
Peat soils and persistent underground fires
Peat forms where waterlogged conditions slow the decomposition of plant remains, allowing organic material to accumulate. In tropical Southeast Asia, peatlands commonly support forests rather than the open, moss-dominated vegetation associated with many northern peatlands. Some deposits are several metres thick and form raised peat domes whose central portions depend mainly on rainfall. Their natural hydrology keeps much of the organic soil saturated and relatively resistant to sustained burning.
Drainage fundamentally changes this condition. Canals constructed for timber extraction, agriculture, or plantation development lower the water table and allow air to enter previously saturated material. Microbial oxidation then consumes organic matter, releasing carbon dioxide and contributing to land subsidence. During dry periods, an increasingly thick surface layer becomes available as fuel. The moisture content, density, and structure of that layer influence whether an ignition remains a small surface fire or develops into a deeper, persistent burn.
Peat commonly burns through smouldering combustion, a slow, flameless process that differs from the rapid flaming of dry leaves and branches. Heat moves through porous organic material while oxygen reaches the reaction zone through pores, cracks, and exposed surfaces. A fire can spread below an apparently extinguished surface, reappear elsewhere, and continue for weeks or months. Because combustion is incomplete, peat fires emit large quantities of carbon monoxide and organic particles relative to the amount of material consumed. A comparatively inconspicuous ground fire can therefore be a major source of regional smoke.
This behavior makes suppression difficult. Water applied to the surface may not reach deeper combustion zones, and water sources may be scarce during the same drought that sustains the fire. Access is often obstructed by canals, fallen trees, unstable ground, and remote terrain. Substantial rainfall can eventually raise moisture levels over a wide area, but isolated showers need not extinguish deep burning. Fire crews must also distinguish between visible flames in vegetation and the less visible heat retained in the soil.
Repeated burning changes the landscape further. Fire can remove the upper peat layer, kill trees, and leave increasingly open vegetation that dries readily. Subsequent fires may occur before a forest canopy has recovered. Drainage and combustion thus reinforce one another: drainage increases flammability, fire damages vegetation and soil, and damaged landscapes can become more susceptible to future burning. Peat haze is consequently a hydrological and land-management problem as well as a firefighting problem.
Land conversion, agricultural burning, and responsibility
Fire is an inexpensive means of removing vegetation, disposing of residues, and preparing land for cultivation. In parts of Southeast Asia it has long been used within agricultural systems adapted to particular local conditions. The regional haze problem, however, reflects more than the existence of traditional burning. Commercial land conversion, logging, roads, canals, settlement, and changes in land tenure have altered the scale of ignition and the ability of landscapes to contain fire. A small burn in a drained peatland can have consequences far beyond the intended plot.
Expansion of oil-palm cultivation and pulpwood plantations has been closely associated with landscape transformation in Sumatra and Kalimantan. Timber extraction may precede conversion, leaving logging debris and increasing access to previously remote areas. Subsequent drainage and clearance can create dry fuels across large areas. These relationships do not mean that every fire within an agricultural region was deliberately set by a plantation company. Estates, independent growers, contractors, settlers, and neighboring land users may occupy overlapping or disputed spaces, and fire can spread across their boundaries.
Establishing responsibility requires separating several questions: where a fire occurred, who controlled or managed that land, who ignited it, and whether management practices allowed it to spread. A satellite hotspot inside a concession boundary is evidence of thermal activity, not by itself proof of the identity or intention of the person responsible. Conversely, uncertainty about the ignition source does not remove questions about drainage, fuel management, supervision, or compliance with environmental obligations. Reliable concession maps and documented land rights are therefore important to both scientific interpretation and legal investigation.
Economic incentives also differ among land users. Mechanical clearance and residue management can require machinery, labor, and capital that are unevenly available. Larger enterprises may be able to finance alternatives but operate through complex contracting and purchasing arrangements. Smallholders may cultivate areas outside formal licensing systems or sell crops into supply chains that obscure their origins. Policies directed only at individual ignitions can miss the market and tenure conditions that make repeated burning attractive.
The effects extend beyond the direct conversion of forest. Drainage infrastructure can alter water levels on neighboring land, while roads and canals facilitate additional clearance. Deforestation and forest degradation therefore interact with fire rather than functioning solely as separate pressures. Effective analysis distinguishes carefully managed agricultural fire on suitable land from uncontrolled burning in vulnerable landscapes, while also recognizing that exceptionally dry conditions can make normally contained practices more hazardous.
Drought, monsoons, and the movement of smoke
Regional climate establishes the seasonal opportunity for fire, but year-to-year ocean–atmosphere variability can greatly extend it. During many El Niño events, changes in tropical Pacific temperatures and the Walker circulation reduce rainfall over parts of Indonesia. The resulting moisture deficit lowers water tables, dries vegetation, and lengthens the period during which fires can spread. Strong El Niño conditions accompanied the major 1997–1998 and 2015 fire crises, although the intensity and spatial pattern of drought differed between them.
The Indian Ocean Dipole provides another important influence. Its positive phase is associated with relatively cool waters near Sumatra and Java and warmer conditions in the western equatorial Indian Ocean. This pattern can suppress rainfall over western Indonesia. A strong positive event contributed to the dry conditions of 2019. El Niño and the Indian Ocean Dipole may reinforce one another, but neither is a necessary condition for serious haze: damaging episodes can arise during shorter dry spells where ignition and fuel conditions are unfavorable.
Once smoke is emitted, winds determine its path. The monsoon changes the prevailing circulation through the year, while daily sea breezes, local storms, and terrain produce smaller-scale variations. During the southwest monsoon, air movements can carry smoke from Sumatra toward Peninsular Malaysia and Singapore, although the actual trajectory depends on the location of fires and the circulation on particular days. Smoke from Kalimantan can affect both Indonesian and Malaysian parts of Borneo and move farther across surrounding seas.
Vertical mixing matters as much as horizontal transport. A shallow, stable atmospheric boundary layer confines pollution near the ground; deeper daytime mixing can dilute surface concentrations while distributing smoke through a greater volume. Rain removes particles through wet deposition and can interrupt fire activity, but precipitation is uneven in both space and time. A downwind city may receive cleaner air after a wind shift even while fires continue, or experience renewed pollution when the circulation changes again. Local weather improvement therefore does not necessarily indicate that the regional source has disappeared.
Climate change adds a longer-term influence through warming, evaporative demand, and possible changes in drought characteristics. Its contribution to any particular haze episode must be distinguished from immediate weather variability and land-use conditions. Fire activity is not a simple thermometer of global warming: similarly dry years can produce different emissions depending on water management, ignition patterns, and the extent of previously degraded peatland. Physical climate and human land management jointly determine the severity of the resulting smoke pollution.
Development of the modern regional crisis
Large fires occurred in Southeast Asia before the late twentieth century, but expanding land conversion and the increasing availability of satellite observations made the regional character of smoke pollution progressively clearer. Severe fires affected East Kalimantan during the 1982–1983 El Niño drought. Logging and drought interacted to increase damage, demonstrating that tropical forests described as humid could become highly vulnerable after disturbance. The event helped establish the importance of considering forest condition rather than rainfall alone when assessing fire risk.
The 1997–1998 crisis became a defining episode. Prolonged drought coincided with extensive burning in Indonesia, including large areas of peat. Smoke affected several countries and disrupted aviation, shipping, and daily life. The fires also drew attention to the exceptionally large carbon stocks beneath tropical peat-swamp forests. Estimates of burned area, emissions, and economic losses varied because records were incomplete and methods differed, but the event firmly established peat combustion as an issue of both regional air pollution and global carbon accounting.
One prominent land-development example was the Central Kalimantan Mega Rice Project, initiated in the mid-1990s to develop roughly one million hectares for rice production. Extensive canal construction drained peat landscapes that were poorly suited to the intended agricultural transformation. The project was subsequently abandoned, but its canals and altered hydrology persisted. Fires in this region illustrated how the environmental consequences of an unsuccessful development scheme could continue long after its original economic objective had been relinquished.
Severe episodes did not occur only in the most prominent El Niño years. In June 2013, smoke from fires in Sumatra produced extreme pollution in Singapore and parts of Malaysia. The concentrated timing and transport of emissions showed that a relatively localized set of fires could cause a major downwind emergency. In 2015, a strong El Niño coincided with much more extensive burning. Indonesian authorities estimated that approximately 2.6 million hectares burned that year. A World Bank assessment placed Indonesia’s associated damage and losses at about US$16.1 billion, equivalent to approximately 1.9 percent of the country’s 2015 gross domestic product.
Widespread burning returned in 2019 under unusually dry conditions, including a strong positive Indian Ocean Dipole. These episodes differ in ignition patterns, rainfall, geographical extent, and response measures. Treating them as identical repetitions obscures the reasons why some years produce short but intense urban exposure while others generate months of widespread pollution. The historical record instead shows a recurrent interaction between persistent landscape vulnerability and variable atmospheric conditions.
Human exposure, health, and economic disruption
The principal health concern during haze episodes is exposure to fine particles, although the smoke mixture contains several harmful pollutants. PM2.5 can penetrate deep into the lungs, and exposure is associated with respiratory and cardiovascular effects. Epidemiological studies of air pollution link particulate exposure with increased illness and mortality at the population level. Smoke episodes can also cause eye and throat irritation and worsen existing respiratory conditions. Children, older adults, pregnant people, and people with certain pre-existing conditions are among the groups considered more vulnerable in public-health assessments.
Exposure is highly uneven. People living close to fires may experience prolonged concentrations well above those recorded in distant cities. Agricultural workers, firefighters, transport workers, and others whose activities take place outdoors can have limited separation from polluted air. Housing quality, building ventilation, occupation, and access to cleaner indoor environments influence individual exposure. A regional average consequently conceals substantial differences between communities, even where they are affected by the same broad smoke plume.
Health measurement requires distinguishing observed events from model-based estimates. Recorded clinic visits, hospital admissions, and deaths provide direct administrative evidence, but they may not capture all effects, particularly where access to care or reporting is limited. Estimates of premature mortality commonly combine modeled smoke concentrations, population distributions, baseline mortality, and exposure–response relationships. Their results depend on the time period, geographical coverage, and assumptions used. Such estimates are not equivalent to a count of individually identified deaths certified as caused by haze.
Air-quality indices translate measurements into categories intended for public communication. Singapore’s Pollutant Standards Index and Malaysia’s Air Pollutant Index use their own calculation and reporting conventions. Other services may report an air-quality index based on different standards or averaging periods. Index values therefore cannot always be compared directly across countries or platforms. The World Health Organization revised its air-quality guidelines in 2021, setting guideline levels for PM2.5 of as an annual mean and for the 24-hour mean. These are health-based guideline values, not a boundary below which all risk disappears.
Economic effects extend beyond medical expenditure. Poor visibility can interrupt flights and navigation; schools may close; construction, agriculture, and other outdoor activities may be curtailed. Tourism, retail activity, and worker productivity can also be affected. Estimates differ according to whether they include only immediate financial losses or also ecosystem damage and the economic valuation of illness and premature death. Figures from separate assessments are therefore not necessarily additive, and the country generating much of the smoke may itself bear a large share of the losses.
Forests, wildlife, and the carbon cycle
Fire damages tropical forests through several pathways. Flames and heat kill seedlings and trees, while burning peat can destroy roots and undermine larger stems. Some trees die only after the initial fire, leaving additional dead wood that may fuel later events. Repeated burning can prevent canopy recovery and favor ferns, grasses, or other open-land vegetation. The outcome depends on burn severity, surviving seed sources, drainage, and the interval before the next fire, rather than following one inevitable course.
These changes reduce and divide wildlife habitat. Orangutans, which occur naturally on Borneo and Sumatra, can lose feeding trees, nesting sites, and routes through the canopy. Fire and clearance may force animals into plantations, roadsides, or settlements, increasing contact with people. Smoke adds a separate exposure pathway, but the direct respiratory effects on free-ranging wildlife are harder to quantify than burned area or forest loss. Claims about numbers of animals killed specifically by smoke require evidence beyond observations that an area burned or that animals were rescued nearby.
Fragmentation also alters the conditions under which surviving populations persist. Loss of forest connections can restrict dispersal, while damaged forest edges often experience hotter and drier conditions. These processes interact with habitat fragmentation caused by roads and agriculture. Species differ in their ability to cross open land, tolerate degraded habitat, or exploit secondary vegetation. Consequently, post-fire changes in biodiversity cannot be represented solely by the number of hectares burned.
Peat fires have particular importance for the carbon cycle. Burning vegetation releases carbon accumulated during plant growth, whereas burning peat consumes a much older belowground store. Subsequent aerobic decomposition of drained peat continues to produce emissions even in years without visible fires. Carbon losses therefore include immediate combustion, longer-term soil oxidation, and reduced storage in damaged vegetation. Forest regrowth can recover some aboveground carbon, but replacing a deep peat deposit requires far longer than restoring a tree canopy.
The climatic effects of smoke are more complex than those of carbon dioxide alone. Fires emit several greenhouse gases, and their particles influence radiation and clouds. Black carbon absorbs sunlight, while many organic particles scatter it; interactions with clouds depend on particle properties and atmospheric conditions. These short-lived effects do not cancel the long-lived consequences of releasing stored carbon. A separate hydrological consequence is peat subsidence: as land elevation falls through oxidation, compaction, and burning, low-lying landscapes may become more vulnerable to flooding, especially where drainage is increasingly constrained by river or sea levels.
Observation, forecasting, and attribution
Modern haze monitoring combines ground measurements, satellites, meteorological observations, and atmospheric models. Instruments such as MODIS and VIIRS identify active fires by detecting thermal anomalies. These observations are commonly displayed as hotspots. They provide valuable information about the location and timing of combustion, but a hotspot is not a standardized unit of burned area, fuel consumption, or smoke production. A single fire may be detected repeatedly, while another may be missed because of cloud cover, overpass timing, low temperature, or its small size.
Remote sensing of burned area uses changes in surface reflectance and other properties to map the land affected after a fire. This answers a different question from active-fire detection. Underground peat combustion introduces further difficulty because surface observations do not directly reveal how deeply the soil burned. Emissions inventories therefore combine mapped activity with estimates of fuel availability, combustion completeness, peat burn depth, and pollutant-specific emission factors. Differences in these assumptions can produce substantially different emission totals from the same general event.
Satellites also measure aerosol optical depth, which describes how strongly the atmospheric column attenuates light. It is useful for following smoke plumes but is not a direct measurement of surface PM2.5. A thick smoke layer aloft may have limited immediate effect at breathing level, while a shallow near-surface layer can create severe exposure without an exceptionally large total-column signal. Cloud screening, humidity, and the assumed optical properties of particles introduce additional complications. Surface monitoring stations remain essential for relating regional imagery to local air quality.
Transport models use winds, mixing, deposition, and chemical processes to estimate where smoke travels and how concentrations change. Back trajectories can indicate the likely pathways of air arriving at a receptor, while source-tagging or sensitivity experiments estimate contributions from particular regions. These techniques support forecasting and scientific attribution but depend on the quality of both emissions and weather data. Forecasting fire weather is also different from forecasting actual burning: drought can be predicted with some skill while the timing of human ignition remains uncertain.
Environmental attribution must finally be distinguished from legal attribution. Evidence that smoke originated in a particular district does not automatically identify the person or organization legally responsible. Investigations may need accurate land boundaries, ownership and management records, field inspections, and evidence of conduct. Measurement uncertainty is not an absence of knowledge; it describes the limits within which conclusions can be supported. Clear separation of observed hotspots, modeled emissions, downwind exposure, and legal responsibility is central to interpreting haze maps and public claims.
Regional agreements, national regulation, and landscape restoration
The principal regional legal instrument is the ASEAN Agreement on Transboundary Haze Pollution, signed in 2002 and entering into force in 2003. It establishes a framework for monitoring, prevention, preparedness, information exchange, and cooperation among the member states of the Association of Southeast Asian Nations. Indonesia ratified the agreement in 2014. Its operation reflects the wider ASEAN emphasis on cooperation and national implementation, rather than a regional enforcement system that directly prosecutes individual burners.
National measures address different parts of the problem. Indonesia has prohibited many forms of land-clearing by fire, regulated peatland management, and pursued administrative, civil, and criminal action in fire-related cases. Singapore enacted its Transboundary Haze Pollution Act in 2014, creating a basis for action against entities whose conduct causes or contributes to haze pollution in Singapore under the conditions specified by the law. Cross-border enforcement nevertheless depends on access to evidence, identification of responsible entities, and cooperation across jurisdictions. The existence of a legal prohibition does not by itself resolve unclear tenure or prevent ignition.
After the 2015 crisis, Indonesia established the Peatland Restoration Agency in 2016. Its mandate initially focused on restoring about two million hectares of peatland across seven provinces. Restoration approaches include canal blocking, water-level management, revegetation, and support for livelihoods compatible with wetter conditions. Ecological restoration on peat differs from simply planting trees: seedlings may fail or remain vulnerable if drainage continues to expose the soil to oxidation and fire. The relevant hydrological unit can extend beyond a single concession or village boundary.
Rewetting also creates practical trade-offs. Raising water tables can conflict with crops and infrastructure established for drained conditions, while canal structures require maintenance and coordination among upstream and downstream users. Paludiculture, the cultivation of crops suited to wet peatland conditions, represents one approach to retaining productive use without intensive drainage. Its feasibility depends on suitable species, local knowledge, processing facilities, transport, and viable markets. Fire prevention programs likewise depend on sustained local participation rather than equipment distribution alone.
Commodity governance provides another set of mechanisms. The Roundtable on Sustainable Palm Oil and corporate no-deforestation and no-peat commitments seek to influence land management through certification and purchasing requirements. Their effectiveness depends on traceability, credible monitoring, supplier inclusion, and responses to non-compliance. A certified estate can still be exposed to smoke or fire originating outside its boundaries, making landscape-level cooperation important. Monitoring therefore increasingly combines fire records with peat-water levels, restoration maintenance, land-tenure information, and supply-chain data, since a quiet fire season may reflect favorable rainfall as well as improved management.