When we measure air pollution in a city, we are looking at a complex, well-blended "soup" of gases and particulate matter. A sensor might tell us that the concentration of fine particulate matter ($\text{PM}_{2.5}$) or carbon dioxide ($\text{CO}_2$) is dangerously high, but it cannot inherently tell us where those molecules originated. Did a local coal-fired power plant emit them, did they drift from a wildfire hundreds of miles away, or are they the product of morning commuter traffic?
To design effective environmental policies, researchers must solve this puzzle, a field known as source apportionment. Traditional methods often struggle because the atmosphere is highly dynamic and chemically non-linear.
Two complementary paradigms have emerged to map atmospheric pollutants back to their origins: tagged tracers within numerical models and chemical ratio fingerprints within physical observations. Together, these techniques allow scientists to preserve the identity of an emission source from the smokestack to the receptor.
The Non-Linearity Trap of Brute-Force Modeling
To understand why advanced source attribution is necessary, consider how a traditional Chemical Transport Model (CTM) estimates a sector's impact. The intuitive approach is "brute-force" sensitivity testing: run a baseline simulation of the atmosphere, then run a second simulation with the emissions from a specific sector (e.g., transportation) completely zeroed out. The difference between the two runs is assumed to be that sector's contribution.
However, atmospheric chemistry does not follow simple addition and subtraction. When precursor gases like nitrogen oxides ($\text{NO}_x$), sulfur dioxide ($\text{SO}_2$), and volatile organic compounds (VOCs) react to form secondary aerosols, their reaction rates depend heavily on the background chemical environment.
If you completely remove $\text{NO}_x$ from a simulated city, you alter the oxidant balance (such as hydroxyl radicals), which inadvertently speeds up or slows down the chemical transformation of other sectors' emissions. This non-linearity introduces severe mathematical artifacts, meaning the sum of all isolated sectors rarely equals 100% of the observed pollution.
Mechanism 1: Tagged Tracers in Forward Models
To bypass the non-linearity trap, researchers use source-oriented chemical transport modeling. Instead of altering emission inventories, these models inject virtual "tagged tracers" alongside the real emissions.
[Emission Source: Traffic] ----> Inject Primary PM + Tagged Tracer (Non-reactive) ----\
----> [Atmospheric Grid] -> Track Mass Directly
[Emission Source: Biomass] ----> Inject Precursor Gas + Tagged Precursor (Reactive) --/
Tracking Primary Particulate Matter
For primary pollutants—particles emitted directly into the air like black carbon or dust—the mechanism is straightforward. As explained by Jianlin Hu et al. in their study of ultrafine organic aerosols in California (Long-term particulate matter modeling for health effect studies in California – Part 2: Concentrations and sources of ultrafine organic aerosols), the model assigns an artificial, non-reactive tracer to each emission category (e.g., wood smoke, meat cooking, vehicle exhaust).
These tracers undergo identical physical transport processes—such as advection, diffusion, and dry/wet deposition—as the bulk pollutant mass. Because the tracers do not react chemically, they act as a linear accounting system, allowing researchers to directly quantify exactly how much mass at any given grid cell belongs to a specific economic sector or geographic region without disturbing the underlying chemistry (Hao Guo et al.; Xue Qiao et al.).
Tracking Secondary Aerosols
For secondary pollutants, which form via chemical reactions in the air, the tagging mechanism must be more sophisticated. As demonstrated by Jinlong Ma et al., reactive tracers can be appended to precursor gases like $\text{NO}_x$, $\text{SO}_2$, and $\text{NH}_3$.
When the model simulates the chemical transformation of these gases into secondary inorganic aerosols (like nitrate and sulfate), the tags are mathematically carried through the chemical kinetic equations. This preserves the source identity across complex chemical phases, allowing for clean attribution even during highly volatile atmospheric events like the COVID-19 lockdowns (Jinlong Ma et al.).
By maintaining these explicit source-receptor pathways, scientists can even use machine learning algorithms to isolate whether model-data discrepancies stem from errors in meteorology or uncertainties in specific sector inventories (Shuai Wang et al.).
Mechanism 2: Chemical Fingerprints and Stoichiometric Ratios
While models track virtual tags forward in time, field researchers use physical observations to trace pollution backward. The core challenge of observational source apportionment is that absolute concentrations fluctuate wildly based on weather conditions; a windy day can dilute a pollution plume to a fraction of its original density.
To overcome this, researchers rely on stoichiometric and chemical ratios. While absolute concentrations change with dilution, the relative ratio between co-emitted species remains remarkably stable as a plume travels through space.
Trace Gas Ratios ($\text{CO}/\text{CO}_2$ and $\text{NO}_x/\text{SO}_2$)
Different combustion processes burn fuel at varying temperatures and efficiencies, leaving behind distinct gaseous signatures:
- Combustion Efficiency ($\text{CO}/\text{CO}_2$): Incomplete combustion (like biomass burning or older vehicle fleets) produces a high ratio of carbon monoxide ($\text{CO}$) relative to carbon dioxide ($\text{CO}_2$). Highly efficient industrial processes produce very little $\text{CO}$ per unit of $\text{CO}_2$. As shown by Jean-Daniel Paris et al., when biogenic uptake is low, this ratio acts as a powerful tracer to distinguish between European and Asian anthropogenic plumes crossing Siberia.
- Fuel Type ($\text{NO}_x/\text{SO}_2$ and $\text{CO}/\text{SO}_2$): Vehicles burn petroleum, emitting high levels of $\text{NO}_x$ but negligible sulfur. Coal-fired power plants and heavy industries emit massive amounts of $\text{SO}_2$. Therefore, a high $\text{NO}_x/\text{SO}_2$ or $\text{CO}/\text{SO}_2$ ratio points directly to a mobile (traffic) source, while a low ratio indicates a stationary (industrial) source (Dongsheng Ji et al.; Guiqian Tang et al.).
Particulate and Molecular Markers
For particulate matter, source apportionment relies on identifying unique chemical "fingerprints" using receptor models like Chemical Mass Balance (CMB) or high-resolution mass spectrometry:
$$\text{Source Contribution} \propto \frac{\text{Concentration of Marker Species}}{\text{Known Emission Profile Ratio}}$$
- Ion Ratios: The ratio of nitrate to sulfate ions ($\text{NO}_3^-/\text{SO}_4^{2-}$) in $\text{PM}_{2.5}$ serves as a proxy for urban evolution. An increasing nitrate-to-sulfate ratio across a region signals a structural shift from coal-dominated industrial pollution to traffic-dominated urban emissions (Zirui Liu et al.; Yi Huang et al.).
- Organic Molecular Markers: Certain organic compounds are highly specific to their parent fuels. For example, levoglucosan and lignin-like compounds are unique thermal breakdown products of cellulose, serving as definitive indicators of wood and biomass burning (Zheng Li et al.; Jianzhong Xu et al.). Conversely, hopanes are organic compounds derived from engine lubricating oil, making them unambiguous markers for fossil-fuel vehicle exhaust (Wing Sze Chow et al.; Yifan Wen et al.).
Closing the Loop: Inverse Modeling and Joint Assimilation
The true power of this research direction emerges when forward-tagged models and observational chemical ratios are combined via inverse modeling.
This is particularly vital for greenhouse gas monitoring. Measuring $\text{CO}_2$ alone cannot distinguish between human fossil fuel combustion and the natural breathing of the biosphere (photosynthesis and respiration). To isolate the human signal, researchers use a multi-tracer approach.
[Satellite/Ground Sensors]
|
|--> Measures CO2 Gradient (Total Carbon)
|--> Measures Proxy Species (CO or NO2)
|
v
[Inverse Transport Model] ---> Uses Theoretical Proxy/CO2 Ratios ---> Isolates Fossil Fuel Plume
Because fossil fuels co-emit $\text{CO}_2$ alongside proxy species like $\text{CO}$ or $\text{NO}_2$, while the biosphere does not, the theoretical ratio between these concentration gradients allows transport models to mathematically filter out the biogenic background (Irène Xueref-Rémy et al.; Alex Boon).
By jointly assimilating satellite or ground-based measurements of these proxy species into high-resolution Lagrangian transport models, authors can perform "top-down" optimization of emission inventories (K. Nalini et al.; I. B. Konovalov et al.). If a city's observed $\text{NO}_2$ or $\text{CO}$ plume is smaller than expected relative to the simulated $\text{CO}_2$ plume, the inverse model automatically corrects overestimates in the underlying fossil fuel inventories (Ke Che et al.).
This tight coupling of physical signatures and numerical tracking transforms the atmospheric chemical soup into a transparent, traceable map of human and natural activity.
Key Foundational Work
Global budget of methanol: Constraints from atmospheric observations (2005) — contributed 1 concept elements to this direction.
On evaluating ocean models with atmospheric potential oxygen (2007) — contributed 1 concept elements to this direction.