Advances in Emissions, Modeling, and Source Apportionment

Air pollution is a major environmental problem leading to impacts on human and ecosystem health. To effectively manage the problem, it is essential that the origins of the pollutants be identified and quantified so they can be targeted for policy actions. Time-resolved, spatially complete, high-resolution concentration and emission data sets for major air pollutants are now fundamental for meaningful health impact assessments, climate change analysis, and environmental policy evaluations. This Viewpoint considers how advances in emissions, modeling, and source apportionment have evolved over the past 60 years with Environmental Science & Technology (ES&T) publishing some of the seminal papers.

Source Apportionment

The earliest efforts to apportion air pollution data were in the 1960s just before and just after 1967 when ES&T was first published. − Specific elemental tracers were applied to obtain quantitative estimates of source contributions. , However, with the advent of multiple species analysis methods like instrumental neutron activation, X-ray fluorescence, ion beam analysis, and inductively coupled plasma mass spectrometry, Friedlander recognized that any given element could be emitted by multiple source types, and thus, a least-squares regression was needed. This approach was applied in several papers until Kowalczyk et al. , recognized that it was necessary to perform a weighted least-squares fit. Subsequently, Watson et al. extended the analysis by incorporating the uncertainties in the measurement of the species concentrations in the profiles of the particulate emissions from the various sources. This effective variance least-squares approach has been incorporated into the U.S. Environmental Protection Agency’s Chemical Mass Balance software (USEPA CMB Model version 8.2).

However, the CMB model requires information about the profile of chemical species in the emissions from the likely sources in an air shed and has problems dealing with secondary species formed in the atmosphere from gaseous emissions. Thus, another class of models based on factor analysis has been developed beginning in the late 1970s. , These methods initially apportioned variance and not species variation so that a modified approach called target transformation factor analysis was developed and applied in a number of studies. Subsequently, it was recognized that these eigenvector-based methods were inappropriate given the nature of the error structure of environmental data giving rise to an explicit least-squares approach to the factor analysis problem, positive matrix factorization (PMF). , This methodology has increasingly been used particularly after the release of the U.S. Environmental Protection Agency’s version of PMF (USEPA PMF version 5.14) that has now become the most widely used software for source apportionment for a variety of problems in air, − water, − and soils. , Thus, the combination of multiple species analytical methods and advanced data analysis tools has enabled development and application of source apportionment to support management strategies leading to environmental improvements and better human and ecological health, with ES&T publishing multiple seminal papers supporting these developments such as the addition of dispersion normalization for preprocessing atmospheric data for PMF analyses , and uncertainty analyses for the resolved source-specific concentrations. ,

Emissions Inventory

Beyond receptor-based source apportionment, comprehensive air quality assessment requires detailed bottom-up emission inventories to drive atmospheric models. Over the past 60 years, ES&T has documented substantial advances in emission inventories, with spatial and temporal resolutions progressing from coarse annual country totals to finely gridded, near-real-time maps at subkilometer and hourly scales. This enhanced granularity supports more accurate air quality simulations through better model inputs.

Sectoral details have also deepened. Specialized inventories now target specific sources that were previously embedded in aggregate totals, exemplified by dedicated emission assessments, for example, for road dust, aviation, , and the iron and steel industry. Long time-series reconstructions of key pollutants (e.g., global NO _x , SO₂, and PM from 1960 onward) have provided an empirical backbone for attributing observed air quality changes to technology shifts and policy interventions across regions and decades. ,− The chemical scope of inventories has expanded beyond classical pollutants (SO₂, NO _x , and PM) to encompass hazardous species such as polycyclic aromatic hydrocarbons and trace metals, as well as the full volatility spectrum of organic compounds. ,− Taken together, advances in resolution, temporal detail, speciation, sectoral coverage, and integration with geospatial big data have helped transform emission inventories into dynamic, high-detail tools that underpin more effective air quality management and atmospheric chemistry research.

A good example of this broadening inventory includes reactive halogens. Early studies have mainly focused on halogen sources and chemistry in polar or marine regions. However, facilitated by the development of measuring technologies (e.g., chemical ionization mass spectrometry (CIMS)), intensive field campaigns have been conducted in continental regions, − detecting ClNO₂, Cl₂, Br₂, BrCl, etc. Unexpectedly high concentrations and different diurnal variations cannot be explained by the traditional halogen sources and chemistry mechanism established in marine or polar areas. In addition, based on field evidence and laboratory experiments, potential new mechanisms are proposed, ,, such as the production of Cl₂ and Br₂ from halogen activation via aerosol iron or nitrate photochemistry. However, additional measurements of the halogen emission characteristics from anthropogenic sources and more laboratory experiments are required to identify kinetic parameters, especially under nearly ambient conditions.

Fusion of Complementary Data Sources

Emission inventories drive chemical transport models to simulate pollutant concentrations, but model uncertainties have motivated fusion approaches that combine simulations with space- and ground-based observational data. Over the past several decades, ES&T has recognized the key importance of the evolution of methodologies that fuse complementary data sources together to generate high-accuracy, high-resolution concentration fields, initially focused on PM_2.5 , and then extended to O₃, , NO₂, and, more recently, PM_2.5 composition. −

This fusion of complementary data sources has unlocked great potential. Using satellite-retrieved aerosol optical depth and trace-gas columns, previous studies have established statistical and hybrid statistical–physical models that link satellite observations with ground measurements and auxiliary predictors (e.g., meteorology and land use) to derive gridded surface concentrations. ,, Recently, machine-learning and deep-learning approaches have been applied to this problem, − substantially improving the spatial and temporal resolution and accuracy of satellite-based estimates, with spatial grids refined to 1 km or even hundreds of meters and temporal resolution reaching hourly scales. In parallel, satellite-derived information about aerosol size distributions and absorbing properties, as well as model-based speciation, has opened pathways to infer PM_2.5 chemical composition from space. − ,

Air Quality Modeling

Air quality modeling has evolved continuously over the past 60 years from traditional forward chemical transport simulations toward a suite of sensitivity-capable modeling systems that can diagnose, attribute, and optimize. Advances in forward sensitivity methods, such as implementation of the decoupled direct method (DDM) in three-dimensional models, have enabled efficient, first-principles quantification of how concentration metrics respond to emissions and parameter perturbations. , Methodological extensions to higher-order direct sensitivities and the characterization of nonlinear responses (notably for ozone) have clarified when linear approximations break down and why control strategies need to be regime-aware. , The development of adjoint frameworks, exemplified by the CMAQ adjoint, has further shifted the field from “what happens if” experiments toward gradient-informed attribution and optimization, enabling source–receptor analysis at scale and laying groundwork for policy-relevant benefit-per-ton and control design applications. ,

As our knowledge of atmospheric processes and chemistry improves along with computational power, the models can continue to improve. As an example, atmospheric models, including box models and chemical transport models, have been used to investigate the impacts of reactive halogens in continental regions. − Recent work has highlighted the role of hydroxymethanesulfonate in ambient aerosol particularly in colder, more humid conditions and provided information that is needed to be able to effectively model its atmospheric behavior. −

Linking Exposure to Air Pollution with Health Outcomes

A major milestone in air quality research has been the quantitative linking of air pollution to exposure and health, supported by rapid progress in spatiotemporal exposure fields and impact assessment. Satellite-enabled and data fusion approaches have steadily improved the resolution and coverage of air pollutant concentration fieldsfrom regional satellite-based estimation to global, long-term reconstructions and near-real-time productsproviding a consistent basis for epidemiology, inequality analysis, and accountability studies. ,,,,, These data fusion advances have, in turn, enabled more comprehensive burden estimation frameworks for global assessments, ,,, while integrated modeling has made it increasingly feasible to connect marginal precursor emissions to population health outcomes, both globally and within fast-changing regions. − Taken together, these developments appear to have moved ES&T’s air chemistry corpus toward a full chain of evidence: emissions → chemistry/transport → exposure → health impact.

Future Research

Looking ahead, a key research need is to develop unified multipollutant, multicomponent frameworks that simultaneously estimate PM_2.5, O₃, NO₂, SO₂, and speciated PM_2.5 at fine spatial and temporal scales, while providing robust and transparent uncertainty characterization. The rapid emergence of new satellite platforms, including geostationary missions, multiangle imagers, and hyperspectral sensors, creates opportunities and challenges for integrating heterogeneous data streams into consistent long-term records that remain stable under changing emission patterns and climate conditions. Capturing extreme events such as wildfires, dust storms, and heat waves is particularly important, as these episodes are becoming dominant drivers of short-term health risks but their impacts on air pollution are often difficult to quantify with existing data sets and methods. Finally, there is a need to embed these high-resolution pollution data sets more deeply into health, climate, and socioeconomic analyses, so that satellite-based products not only map exposure but also support source attribution, policy evaluation, and the design of equitable, evidence-based mitigation strategies, an area where ES&T will continue to play a leading role.

Another leap is likely to come from making each link of that chain dynamic, multisource, and uncertainty-aware. For emissions, a central need is a new generation of inventories that fuse bottom-up activity and technology data with satellite retrievals and ground observations using AI/ML, while preserving physical and chemical consistency. ES&T’s recent demonstrations of ML-enabled top-down inference of NO _x emissions from satellite NO₂ highlight the feasibility of near-real-time updates and scalable deployment, and work on observation-constrained, physically informed fusion for surface concentrations illustrates the importance of hybrid constraints (rather than purely “black-box” prediction) for operational products. , At the modeling layer, a key research priority is not to replace CTMs, but to build hybrid systems: CTMs for causal fidelity, sensitivity-enabled methods (DDM/adjoint) for attribution and optimization, and emulators/surrogates for speed, ensemble uncertainty quantification, and scenario exploration. ,,,

Finally, if air chemistry is to remain policy-relevant under rapid decarbonization, tighter coupling of climate–air pollution–socioeconomic dynamics in integrated assessment will be required, explicitly valuing health cobenefits and distributional outcomes across pathways and regions. − These frontiers collectively suggest a discipline that is becoming increasingly predictive, actionable, and equity-aware, while retaining the mechanistic rigor that has characterized ES&T’s contributions for six decades.

Biography

Professor Philip K. Hopke has been studying atmospheric aerosol properties for more than 55 years and has been a leader in the development of mathematical methods to identify and quantify air pollution sources. Prof. Hopke has held faculty positions at the SUNY College at Fredonia, the University of Illinois at Urbana-Champaign, Clarkson University, and the University of Rochester. He has published 72 papers in Environmental Science & Technology.

The authors declare no competing financial interest.

Published as part of Environmental Science & Technology special issue “60th Anniversary of Environmental Science and Technology”.