Linking Climate and Air Quality: Insights from Machine Learning, Chemical Transport Modeling, and Field Measurements
Author(s)
Yin, Lifei
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Climate change and air pollution are two of the most significant global challenges of the industrial era, both exerting adverse impacts on human health. Their complex interactions make this an urgent area of research. Climate and air quality are interconnected through multiple pathways: many anthropogenic sources emit both air pollutants and greenhouse gases, while climate change affects air quality by altering meteorological conditions that govern the emissions, formation, transport, and removal of air pollutants. In turn, air pollutants influence climate directly through interactions with solar radiation and indirectly by modifying cloud properties. To improve future air quality and climate projections and support effective policymaking, it is of vital importance to understand underlying processes linking climate and air pollution and to accurately represent the relationships in regional and global models.
This dissertation investigates the impact of rising temperature on surface PM2.5 (particulate matter with aerodynamic diameters less than 2.5 μm) and ozone (O3) levels across the contiguous United States (CONUS), reproduces and explains these observed sensitivities using a state-of-the-art chemical transport model, and characterizes the mixing state and optical properties of light absorbing particles—black carbon (BC) —in different environments in the Southeast US. The research integrates high resolution datasets derived from machine learning models, nationwide observational networks, chemical transport modeling, and in situ field measurements to address key gaps in our understanding of the temperature dependence of air pollution and the climate implications of aerosol mixing state.
Chapter 2 examines the sensitivity of surface-level PM2.5 and O3 to summer temperature using long-term, high-resolution datasets generated by ensemble machine learning models. Across the eastern US, stringent emission control strategies over the past two decades have significantly reduced the positive temperature responses of PM2.5 and O3, thereby lowering the population exposure to air pollution during heat events. In contrast, the western US has experienced an increase in PM2.5–temperature sensitivity, probably driven by the growing temperature-sensitivity of wildfires. These findings underscore the effectiveness of emission reductions in mitigating climate-induced air quality deterioration in some regions, while highlighting new challenges in areas facing intensified climate-driven pollution sources such as wildfire.
Chapter 3 improves the representation of temperature-dependent PM2.5 responses in the GEOS-Chem chemical transport model through targeted updates to emission inventories and secondary organic aerosol (SOA) formation from biogenic emissions. Simulations for 2000–2022 reveal that chemical production processes—particularly isoprene-derived SOA and sulfate formation—dominate PM2.5 sensitivity in the eastern US, with a declining trend driven by reductions in anthropogenic SO2 emissions. In the western US, primary emissions from wildfires, increasingly drive the observed sensitivity. Transport processes are major contributors to interannual variations nationwide. For the first time, this work quantitatively attributes PM2.5–temperature sensitivity to individual processes, demonstrating how the interplay between chemistry, emissions, and meteorology determines regional air quality responses to warming.
Chapter 4 presents field measurements of the mixing state of refractory BC (rBC)-containing particles from two campaigns conducted at an urban site in Atlanta, GA (ATL) and a rural high-elevation site in Boone, NC (APP). At ATL site, rBC core size increases linearly with particle diameter, indicating more uniform internal mixing. In contrast, APP site exhibits a bimodal distribution of rBC core sizes in larger particles, suggesting the influence of local and long-range transported aerosols with distinct aging histories. The absorption enhancement factor (Eabs), calculated based on realistic mixing states, deviates significantly from estimates assuming fully internal mixing—particularly at the APP site, where mixing state heterogeneity leads to overestimation of Eabs by over 40%. These findings emphasize the importance of accurately resolving aerosol mixing states in optical properties simulations, especially in regions influenced by both local emissions and long-range transport.
In summary, this dissertation: (1) quantifies the temperature dependence of PM2.5 and O3, and demonstrates the effectiveness of emission control policies in mitigating climate-induced air quality degradation across the CONUS; (2) improves the performance of GEOS-Chem in simulating PM2.5–temperature relationships and identifies the dominant temperature-sensitive processes that drive regional and temporal variability; and (3) characterizes the mixing state of black carbon in contrasting environments and demonstrates the impact of mixing state heterogeneity on modeled optical properties. These findings have several implications: (1) stringent emission control strategies can effectively reduce air pollution sensitivity to warming and thereby lowering associated public health risks; (2) wildfire in the western US is becoming increasingly temperature sensitive, underscoring the need for urgent attention and regulation action; and (3) resolving the mixing state of light-absorbing aerosols is critical for reliable estimation of aerosol radiative effects, especially in regions with more heterogeneously mixed aerosol population. Together, these studies enhance our understanding of key aspects of interactions between climate and air pollution and support the development of science-based strategies for air quality and climate management.
Sponsor
Date
2025-08-20
Extent
Resource Type
Text
Resource Subtype
Dissertation (PhD)