How do scientists identify where Malaysia’s haze came from?
Scientists trace a haze episode by comparing several independent kinds of evidence: where fires may be burning, what satellites see in the atmosphere, which way air was moving, what the ground monitors measured, and—when available—the chemical properties of the particles. A back-trajectory model such as NOAA’s HYSPLIT can estimate where an air parcel previously travelled. But a trajectory is not proof that one individual hotspot produced all the particles detected at a Malaysian station.
What is the fastest credible source-attribution workflow? Identify the exact Malaysian district and observation time, inspect ASMC’s regional fire and smoke products, compare meteorological winds, run or consult an appropriate transport model, then test the hypothesis against measured pollution. NOAA’s HYSPLIT reference explains how atmospheric models calculate back trajectories and dispersion. NASA FIRMS warns that its fire markers are limited satellite thermal detections, not direct proof of a particular fire’s full extent or cause.
This guide focuses on source attribution and evidence standards. It helps readers distinguish plausible smoke movement from legal responsibility and teaches original worked examples that respect uncertainty. For today’s public-health decision, use Malaysia APIMS, regardless of whether the source has yet been fully identified.
Source-attribution evidence in one minute
- Fire detection: NASA FIRMS or ASMC thermal anomalies show possible active burning at particular locations and times.
- Smoke imagery: Satellite optical observations show broad plume shapes where visibility and sensing conditions allow.
- Wind analysis: Meteorological observations and models estimate the path of moving air at several heights.
- Back trajectory: A model estimates where air reaching a selected point may have travelled previously.
- Forward dispersion: A model estimates where released material may move and spread.
- Ground measurements: Malaysia DOE monitoring stations report the actual receiving-area pollution index.
- Chemistry: Biomass-burning tracers, carbon components or other measured species may help evaluate source types.
- Uncertainty: Weather-model error, satellite gaps, emissions estimates, chemical processing and local emissions all matter.
- Legal evidence: A scientifically plausible source region is not the same as proof that a specific company violated a law.
First decide what the attribution question actually is
There are several questions that sound similar but demand different evidence. ‘Did smoke from Kalimantan reach western Sarawak?’ is an atmospheric transport question. ‘What percentage of Kuala Lumpur’s measured PM2.5 came from peat fires?’ is a quantitative source-apportionment question. ‘Which landowner illegally lit a fire?’ is a legal responsibility question.
The first may be investigated with satellites and trajectories. The second can require detailed concentration and source-emission modelling or chemical analysis. The third needs verified conduct, land records, jurisdiction and appropriate legal process.
A good article states which question it is answering before offering a conclusion. Otherwise evidence sufficient for a broad plume route may be overused to accuse an individual.
Start with the receiving town, not an alarming fire photograph
A scientifically sound investigation begins by naming the affected Malaysia station and its observation time. A station near Port Klang and one near Serian sample very different geographical contexts. Their likely source regions and wind paths may not be the same.
Record whether the quantity reported is the Malaysian API or a pollutant concentration in µg/m³. Air-quality indexes may use different averaging periods and dominant pollutants, so not every apparent spike means an equal increase in smoke particles.
Only then look upstream. The Malaysia API and PM2.5 guide explains why a clearly defined receiving measurement is essential.
Why a satellite fire hotspot is a clue, not a fire map of exact area
NASA’s FIRMS active-fire products detect unusually hot locations in satellite pixels. The NASA FIRMS map guidance explains that a highlighted pixel does not mean the entire pixel burned and may include hot material from different sources.
The system offers MODIS and VIIRS detections with different approximate pixel sizes. The indicator is valuable for locating possible activity but not a direct measurement of hectares burned, smoke emitted or who ignited the fuel.
Cloud, heavy smoke, canopy cover and orbit timing may hide fires. A blank area on the map is not proof that no burning exists.
MODIS and VIIRS: different spatial resolutions
NASA FIRMS provides MODIS detections with an approximate 1-kilometre pixel scale and VIIRS fire detections associated with a 375-metre pixel scale. The finer nominal grid can improve detection of some smaller fires, but its coloured square is not an assertion that a 375-by-375-metre patch was completely burned.
Sensor characteristics, viewing geometry and observation times also differ. A plot mixing detections from both without a proper methodology can produce misleading totals.
Students should report which instrument produced a hotspot count, and avoid summing overlapping detections as though every dot represented a unique ignition.
Smoke images observe plumes, not individual legal offences
A corrected-reflectance satellite image may show smoke travelling away from detected burning. Such visual alignment can support a plausible source-pathway story, especially when repeated over time.
However, smoke can obscure underlying fires, cloud can resemble or hide plumes, and different source areas may feed one broad regional haze layer. Analysts use spectral and temporal information rather than relying on visual impression alone.
A photograph of smoke emerging from a map location is not evidence that a named owner intentionally lit it. Scientific observation and legal attribution must remain separate.
The 2026 ASMC data provide a genuine regional example
The ASMC July 2026 review documents smoke from West Kalimantan drifting northwest towards parts of western Sarawak on 15 July 2026. This links a possible source region, observed plume and receiving geography at a specific historical time.
ASMC subsequently raised its southern ASEAN alert to Level 3 on 26 August in response to heightened hotspot activity, dense smoke and dry conditions. The alert history explains the regional risk categories.
These reports are useful starting evidence. They cannot establish precise proportions for every local Malaysian station or prove that each detected fire was illegal.
What is a backward trajectory?
A backward trajectory calculates the estimated previous positions of an air parcel arriving at a selected place and time. In effect, a model asks where the air came from under the meteorological fields it uses.
NOAA’s HYSPLIT model is widely used for local and long-range atmospheric transport studies, including fire smoke. Analysts can specify different initial arrival heights and examine how routes differ.
A parcel’s route does not prove that it contained smoke, and meteorological uncertainty grows when following a parcel farther backwards. It is a modelled pathway hypothesis to compare with other evidence.
What is a forward trajectory?
A forward trajectory follows an estimated air parcel from a selected starting point into future positions or a historical simulation period. It can help test whether observed winds make transport from a suspected fire source towards a receiving region physically plausible.
But a simple trajectory carries no guarantee about actual particle concentration. The real plume spreads, mixes vertically, experiences deposition and may receive new emissions along the route.
A dispersion simulation attempts more than a single path, using additional physical assumptions. The eduKate forecasting guide explains those differences.
Why models must use real atmospheric height
Wind speed and direction vary with height. An air parcel reaching a ground station may have travelled aloft before mixing down. A model run beginning at 100 metres can produce a different history from one beginning at 500 or 1,500 metres, depending on atmospheric conditions.
Analysts therefore often examine multiple arrival heights and alternative meteorological inputs. A map showing just one line can hide important uncertainty about the actual transport history.
Do not interpret a satellite plume high in the atmosphere as an exact representation of air being inhaled beside a Malaysian school gate.
The meaning of 2,500-foot winds on ASMC maps
The ASMC regional haze situation map displays winds at approximately 2,500 feet, based on model analysis. These winds help explain regional transport but do not necessarily match the surface breeze at a beach or apartment block.
When comparing a back trajectory with an ASMC chart, check the time, height and geographical representation. A change in altitude may alter the probable pathway.
This is why an arrow over the Strait of Malacca is useful science context but not a personal exposure measurement.
Advection and dispersion are not interchangeable
Advection means material is transported by moving air. Dispersion also considers spreading and mixing within that flow. A line along which an air parcel may move cannot by itself represent the width, depth or changing concentration of a smoky region.
Rainfall, turbulent mixing, differences in emission rates and surface processes alter the pollutant field. A concentration at the destination depends on more than whether the trajectory line passes over a fire.
A valid source-attribution study may use both trajectories and a dispersion model, reporting what each can and cannot explain.
Source strength is a missing link in many online claims
Two detected fires can release very different quantities of smoke because of fuel, area, moisture, combustion mode and duration. A deep peat fire may behave differently from short-lived surface vegetation burning.
A back trajectory passing over ten hotspots cannot conclude that each contributed one-tenth of the receiving city’s pollution. The source strengths and their timing would need to be estimated.
This is one of the most important limits of an otherwise convincing map: knowing where emissions may exist is not the same as knowing how much pollutant entered the air.
Rain and atmospheric removal change the plume
Particles can be removed by deposition and rainfall along their transport path. A plume passing through sustained showers may evolve differently from one carried through prolonged dry air.
At the same time, smouldering sources may continue emitting and feed new material into the atmosphere. This can maintain poor air quality even when a particular satellite image suggests that an older plume has dispersed.
Transport and removal should therefore be considered together. A trajectory with no emission or removal model cannot predict the exact downwind concentration.
Aerosol optical depth is not surface PM2.5
Satellite aerosol optical depth describes how aerosols affect light along an atmospheric column. It is useful for understanding broad aerosol burden, but it is not a direct measurement of the near-surface PM2.5 mass a resident breathes.
A large column signal may include smoke aloft; cloud can prevent successful retrieval. Ground concentration is a different observation made under a different method.
The NASA aerosols explainer describes the broader science, while APIMS addresses Malaysia’s official ground-based monitoring and public index.
Why chemical composition can make the source story stronger
Samples of fine particles can contain organic carbon, elemental carbon, inorganic ions and distinctive organic compounds. Some constituents are associated with biomass burning, while others are common in traffic, industry or natural processes.
Chemical observations may help establish whether a receiving-area pollution episode is consistent with biomass burning. But no single molecule necessarily names a particular provincial fire or identifies the operator responsible.
A defensible source-apportionment study combines appropriate chemical markers with meteorological evidence, emission inventories, multiple samples and uncertainty estimates.
Levoglucosan: a useful but limited biomass-burning marker
Levoglucosan is an anhydrosugar produced during combustion of cellulose-rich material and is frequently studied as a biomass-burning marker in aerosol samples. Its abundance can provide evidence of burning-related particles when interpreted with supporting chemistry.
A Southeast Asian molecular-marker study investigated levoglucosan and potassium with PM2.5 in northern Southeast Asia. It demonstrates the analytical approach, not a direct measurement of Malaysia’s October 2026 air.
Chemical processing, different fuels and other sources complicate interpretation. One marker should never be used alone to establish a precise ignition location or criminal responsibility.
Potassium and elemental carbon are not exclusive fingerprints
Potassium-containing particles can be associated with certain biomass-burning emissions, but other environmental sources also contribute potassium. Elemental carbon and black carbon are related measurement concepts that may include soot from traffic and other combustion sources.
The relative abundance of components can be informative, especially across repeated samples, yet neither ‘potassium detected’ nor ‘black carbon present’ proves a particular peat fire source.
Researchers use multiple components, chemical ratios and appropriate statistical or physical models rather than treating one element as a universal fingerprint.
What is a source profile?
A source profile describes the composition of pollutants associated with a given type of emission under specified conditions. For example, vegetation burning, vehicle exhaust and industrial processes may have distinct typical mixtures.
Those profiles vary with fuel and operating conditions. The age of a plume also changes its chemical makeup, so an old or geographically inappropriate profile may mislead.
A credible receptor analysis states where profiles came from and whether they are suitable for the local Malaysian and Southeast Asian atmosphere.
What receptor modelling tries to estimate
Receptor models use pollutant measurements at an affected location to estimate contributions from different classes of sources, based on observed composition and model assumptions. Positive matrix factorisation and chemical mass-balance approaches are examples.
Such results can estimate source categories under suitable data conditions, but are not independent proof of which named person lit a fire. A modelled contribution has uncertainty and may combine chemically similar sources.
Describing a category as ‘regional biomass burning’ can be scientifically justified without pretending that all particles came from a single land concession.
Why matching timestamps is fundamental
A satellite might record a hotspot in the afternoon, a wind analysis may represent the evening, and an API reading may use a 24-hour averaging method. Combining them without aligning their observation times produces an apparently precise but logically weak explanation.
Air parcels require time to travel. A reading before the supposed source ignition cannot be explained by that later ignition alone. Time zones, UTC stamps and instrument latency need attention.
A defensible map places each observation on a common timeline and explicitly identifies any gaps.
How far backwards should a trajectory be run?
The useful back-trajectory period depends on the question, distance, meteorological reliability and study design. A very short path may miss relevant sources; a much longer one accumulates model uncertainty and encounters many possible emission regions.
There is no universal number of hours that proves a source. NOAA HYSPLIT supports flexible settings, but the analyst needs to justify the choice and compare different heights or meteorological inputs where appropriate.
A research report should state the chosen arrival time, altitude, duration and source of weather data rather than display only the prettiest route.
Worked maths: two hypothetical air-parcel routes
Invented model example, not a real plume: A parcel moves at a constant average 18 km/h for eight hours in one direction. The simple distance is 18 × 8 = 144 km.
A second hypothetical parcel travels at 12 km/h for the same time and covers 96 km, a 48-kilometre difference. Neither arithmetic result tells you whether smoke was present or how much PM2.5 was deposited at a destination.
Real HYSPLIT paths curve and change speed and height, making these calculations conceptual rather than forecasts. Their value is teaching how wind uncertainty changes a possible source region.
Worked exercise: source fractions need matched quantities
Entirely synthetic emission scenario: Suppose a validated analysis estimated 30 µg/m³ of the observed PM2.5 came from regional burning and 20 µg/m³ from other sources during a specified comparable period. The total of those categories would be 50 µg/m³, and the burning share 30 ÷ 50 = 60%.
This arithmetic is valid only under the assumed validated source-apportionment method. It cannot be performed using hotspot counts or API digits in place of pollutant concentrations. Real methods have uncertainty and source categories may not be perfectly distinguishable.
The exercise shows why a percentage needs a stated measurement basis and a defensible modelling procedure.
A realistic Sarawak attribution question
ASMC documented smoke transport from West Kalimantan towards western Sarawak in July 2026. A scientist could combine the dated regional image with matching winds, possible fire detections and local observations to evaluate that transport pathway.
But this does not establish that every Sarawak station received the same amount or that a particular landholder caused the emissions. The finding applies to a specified date and observed regions.
The Malaysia fire-source map explains the provincial geography behind this scientific case.
A realistic Klang Valley attribution question
A poor-air event around Port Klang may coincide with fires on Sumatra and suitable winds across the Strait of Malacca. That would support investigation of a transboundary contribution.
Yet the Klang Valley also includes traffic, port-related activities and other local sources. A station’s API cannot apportion the relative share from each source by itself.
Appropriate chemical measurements and transport analysis may strengthen attribution; they should not be replaced by assumptions about the nearest geographical fire.
Where model uncertainty comes from
Wind fields, rainfall forecasts, emission estimates, plume injection height, vertical mixing, land-surface conditions and model resolution all influence a trajectory or dispersion result. Different credible configurations can yield different plausible paths.
Researchers may run several starting heights or ensemble simulations to inspect the range. A single sharply drawn line is visually persuasive but can convey false certainty if its limitations are hidden.
An honest scientific answer reports uncertainty as part of the result, not as a disclaimer added after a categorical accusation.
Why satellite pixels cannot decide legal responsibility
NASA FIRMS itself warns that active-fire thermal anomalies have limited accuracy and should not be used alone for protecting life or property. A map may guide a field team, but legal attribution requires independent evidence about land use, conduct and the applicable law.
A satellite pixel can overlap a concession boundary or show activity near several holdings. It does not independently prove who started or maintained the fire.
The Malaysia haze law and ASEAN article explores jurisdiction and accountability separately from atmospheric source science.
What would count as a stronger attribution conclusion?
A conclusion such as ‘the observations are consistent with transport from southern Sumatra during the stated period’ is appropriate when fire detections, smoke imagery and winds align but precise contribution remains uncertain.
A stronger quantitative statement requires measured pollutant data, a well-supported methodology and uncertainty estimates. A claim that one identified company caused a certain hospital admission count requires entirely different legal and epidemiological evidence.
Writing a modest supported conclusion is not weakness. It is how scientific confidence can be distinguished from speculation.
A student’s evidence table for one haze event
Create columns titled observation date, location, instrument, quantity, unit, what it supports, and what it cannot prove. Enter one hotspot observation, one smoke image, one wind analysis and one official Malaysian API reading.
The student’s final paragraph should state a plausible source-pathway relationship and at least two limitations. This makes an excellent applied geography and science lesson without any smoke-generation experiment.
Use a historical date and trustworthy sources so it does not become a false live warning in a parent group.
Why the exact source does not have to be settled before protection
Air-quality authorities can provide protective advice based on measured local pollution while scientists investigate the sources. Waiting for complete source attribution could delay sensible actions during an evolving event.
The upstream objective remains fire prevention and evidence-based enforcement. The downstream objective is reducing avoidable exposure for schools, families and outdoor workers.
Both tasks matter, but their information thresholds differ. A responsible article does not confuse a public hazard notice with a legal finding.
Frequently asked questions about haze source attribution
Can HYSPLIT prove which fire caused Malaysia haze? It estimates air transport pathways; proof of a specific source needs other evidence and uncertainty assessment.
Is a back trajectory a recorded flight path of smoke? No. It is a modelled air-parcel path based on meteorological information.
What is the difference between trajectory and dispersion? A trajectory describes a possible parcel path; dispersion also addresses spreading and transport of pollutants across an area or volume.
Does a satellite hotspot mean a whole pixel burned? No. A pixel can contain a smaller hot area or other thermal anomalies.
Is VIIRS 375 m the size of a fire? No. It refers to a nominal detection-pixel scale, not exact burned area.
Are MODIS and VIIRS detections interchangeable in one count? Not without a suitable method accounting for overlap, resolution and observation differences.
Can heavy cloud hide a real fire? Yes. Satellite algorithms may miss activity under clouds, canopy or thick smoke.
Can a satellite measure the API at one Malaysian house? No. Aerosol imagery and official ground air-quality measurements are different products.
What does aerosol optical depth describe? The effect of aerosols on light through an atmospheric column, not directly a person’s PM2.5 dose.
Why should trajectories be modelled at multiple heights? Winds vary with altitude and smoke can move aloft before reaching ground level.
Is levoglucosan proof of a particular plantation fire? No. It can indicate biomass-burning contributions but does not identify an individual ignition by itself.
Does potassium prove every particle is from peat? No. It has multiple potential sources and must be interpreted with other evidence.
Does regional smoke stop changing after leaving a fire? No. Atmospheric chemistry, humidity, dilution and deposition can alter it.
Can an API index be used directly in chemical mass balance? No. Pollutant concentrations, not generic index digits, are required under an appropriate model.
Are matching map arrows enough to blame one country? No. Source emission evidence, timing and other contributors also matter.
What is a source profile? A measured or modelled mixture of substances associated with a particular emission type under stated conditions.
What is a receptor model? A method estimating source categories from measurements at a receiving location with explicit assumptions.
Did ASMC document West Kalimantan smoke reaching Sarawak in July 2026? Yes, its monthly review reports such transport on a specified July date.
Is there one universally correct back-trajectory duration? No. The duration must fit the research question and meteorological uncertainty.
Can a source contribution be 60% because 60% of hotspots are on one island? No. Detection share and measured particle mass contribution are different quantities.
Does a regional alert prescribe a Malaysian school closure? No. Actual institutional instructions come from the responsible Malaysian authorities.
Is the closest fire always responsible for the worst pollution? No. Emission strength, winds, plume height and removal matter.
Does finding smoke automatically reveal illegal conduct? No. Legal responsibility needs evidence under the applicable law and procedures.
Why do several instruments strengthen a conclusion? Independent fire, plume, wind and ground observations test different parts of the causal chain.
Where do I check current information? Use ASMC for regional conditions and Malaysia DOE APIMS for official local air quality.
Continue through the linked eduKate Malaysia haze library
Official tools and research sources
Editorial review: 18 October 2026. All trajectory and concentration examples are explicitly imaginary educational data; no actual HYSPLIT run or chemical source-apportionment analysis is claimed. Historical NASA imagery is dated to 2005 and does not demonstrate today’s Malaysian API.