A monitoring station on a city rooftop records a winter morning spike in fine particulate matter, with PM2.5 well above health guidelines. The filter that captured those particles holds a dark smudge, a few hundred micrograms of material collected over twenty-four hours. That smudge contains the answer to a question regulators badly need answered: where did this come from?

The candidates are many. Diesel trucks on the ring road. Wood stoves in nearby suburbs. A coal-fired plant two hundred kilometres upwind. Ammonia drifting from farmland and reacting in the atmosphere to form new particles. Each source has its advocates and its defenders, and each control measure carries real costs.

Source apportionment is the scientific discipline that works backward from what arrives at a receptor (a monitor, a neighbourhood, a pair of lungs) to the sources that sent it there. It treats the air sample as evidence and the atmosphere as a crime scene. Understanding how it works shows why some pollution policies succeed while others spend money without clearing the air.

Every combustion process, industrial operation, and natural emission leaves behind a characteristic mixture of chemical species. Scientists call this a source profile: the relative proportions of elements, ions, and organic compounds that a source emits. A source profile works less like a single fingerprint and more like a recipe. It is defined by ratios rather than by any one ingredient.

Some ingredients are especially telling. Vanadium and nickel point to the combustion of residual fuel oil, the heavy fuel historically burned by ships and some power plants. Levoglucosan, an anhydrosugar formed when cellulose burns, signals wood smoke. Copper, antimony, and barium come from brake pads. Silicon, aluminium, calcium, and iron reflect mineral dust from soil and roads. Hopanes and steranes, molecules found in lubricating oil, mark motor vehicle exhaust.

These marker species are called tracers, and the ideal tracer has three properties. It is emitted mostly by one source type. It survives transport without reacting away. It can be measured reliably at low concentrations. Few tracers are perfect. Potassium comes from wood smoke but also from soil and fireworks, so analysts rely on combinations of species rather than single markers.

The underlying logic is conservation of mass. If the material arriving at a receptor is a mixture of contributions from several sources, then the measured concentration of each species equals the sum of what each source contributed. With enough species measured across enough samples, that sum can be untangled mathematically. One complication is that some particles were never emitted at all. Sulfate, nitrate, and much organic aerosol form in the atmosphere from precursor gases, so they carry only a faint trace of their origins. These secondary particles often make up a large share of urban PM2.5, and attributing them requires extra detective work.

Takeaway

A pollution source is identified by its chemical ratios, not by any single compound. The recipe matters more than any one ingredient, which is why measuring many species beats measuring one with great precision.

Chemical mass balance (CMB) is the oldest and most intuitive receptor model. The analyst supplies a library of source profiles measured directly at emission points, such as tailpipe tests, stack samples, and controlled smoke chamber burns. The model then finds the combination of source contributions that best reproduces the ambient sample. In effect, it solves a set of simultaneous equations weighted by the uncertainty in each measurement. The US Environmental Protection Agency has distributed CMB software for decades, and it remains widely used.

CMB's strength is that it can work with a single sample, and its results map directly onto named sources. Its weakness is its dependence on those profiles. If a local source is missing from the library, or if real-world emissions differ from test conditions, the model silently misattributes mass. Sources with similar profiles, such as gasoline and diesel vehicles, can become collinear, which makes the mathematics unstable.

Positive matrix factorization (PMF), developed in the 1990s, takes the opposite approach. Given a long time series of samples with dozens of measured species, it searches for a small number of recurring patterns, or factors, that together explain the variation in the data. Crucially, it constrains every source contribution and profile to be non-negative, since a source cannot emit negative mass. The analyst then interprets each factor. For example, a factor rich in levoglucosan and potassium that peaks on cold evenings is likely residential wood burning.

PMF needs no prior profiles and can reveal sources nobody knew to look for. In exchange, it demands large datasets and careful judgement about how many factors to keep. Other tools fill the remaining gaps. UNMIX and principal component analysis offer alternative factorizations. Radiocarbon measurements separate fossil carbon from modern biological carbon. Wind direction and back-trajectory analyses link factors to geographic origins. Strong studies combine several methods and check whether they agree.

Takeaway

Every receptor model trades one kind of knowledge for another. CMB borrows certainty from measured source profiles, PMF borrows it from patterns in the data, and confidence comes when independent methods converge.

Air quality managers face a practical problem: limited budgets, many possible controls, and a legal obligation to meet health-based standards. Without source apportionment, decisions tend to default to the most visible or politically convenient targets. With it, agencies can rank sources by their actual contribution to concentrations where people breathe. A tonne of emissions cut at one source can deliver far more clean air than a tonne cut elsewhere, depending on proximity, timing, and atmospheric chemistry.

The results are often surprising. Cities that assumed traffic dominated their winter particulate problem have found residential wood burning contributing a large share on cold, still nights. Those findings have prompted no-burn day programmes and stove replacement schemes. Coastal studies have tracked vanadium and nickel factors falling after rules forced ships to burn cleaner, low-sulfur fuel. In those cases, the apportionment result doubles as evidence that a policy worked.

Apportionment also answers the question of jurisdiction. If most of a city's PM2.5 is secondary sulfate and nitrate formed from emissions hundreds of kilometres away, local traffic restrictions will disappoint. Ammonium nitrate frequently traces back to agricultural ammonia combining with nitrogen oxides from combustion. Effective control may therefore require reaching beyond city limits into farm policy. Separating local from regional contributions determines whether action belongs at the municipal, national, or international level.

Increasingly, apportionment is being linked to health outcomes. Not all particles are equally harmful per microgram. Epidemiological studies have examined whether combustion-derived particles, rich in metals and organic compounds, carry greater risk than mineral dust. If they do, the policy goal shifts from reducing total mass to reducing the most toxic fractions. Source apportionment provides the map that makes that targeting possible, connecting what leaves a chimney to what settles in an airway.

Takeaway

The most effective pollution control targets the sources that actually reach people, not the ones that are easiest to see. Apportionment turns emission inventories into exposure priorities.

Source apportionment rests on a simple premise: pollution carries the signature of its origin. By measuring enough chemical species and applying models like CMB and PMF, scientists can trace those signatures from a filter back to a tailpipe, a stove, or a ship's funnel.

The methods carry real uncertainty. Profiles vary, secondary particles blur origins, and factors require interpretation. Robust conclusions come from combining approaches and testing whether they agree, not from trusting any single model output.

For environmental protection, the practical lesson is clear. Effective air quality management starts at the receptor, where people breathe, and works backward. Controlling the sources that actually drive exposure, rather than those that merely look responsible, is how limited resources become measurable health gains.