Research Seminars @ Illinois

Advancing Understanding of Aerosol Impacts over North America through Atmospheric

Chemistry Modeling and Satellite Remote Sensing

Advisor: Assistant Professor Hannah Horowitz

Abstract

Ambient fine particulate matter (diameter <2.5 microns, PM2.5) poses a major risk to human health,

with chemical composition playing an important role in determining toxicity. Ground-based

monitoring networks are spatially sparse, and measurements of PM2.5 chemical components are

even more limited. Atmospheric chemical transport modeling provides broad-scale estimates of

PM2.5 mass and composition but remains sensitive to uncertainties in emission inventories and

chemical formation mechanisms. Satellite aerosol optical depth (AOD) observations are widely

used to estimate surface PM2.5 and to constrain model simulations, but most applications rely on

Standard products with 8 to 40 hour latencies. Near-real-time (NRT) satellite AOD observations

(1 to 3 hour latency) offer the potential to improve the timeliness of satellite-constrained PM2.5

estimates. In addition to health impacts, PM2.5 influences the climate through its radiative impacts,

with implications for urban heat. This dissertation examines uncertainties in modeling PM2.5,

evaluates the application of NRT satellite AOD to constrain modeled PM2.5, and investigates

interdecadal trends and relationships between patterns of air pollution and urban heat in North

America.

First, the GEOS-Chem atmospheric model is used to evaluate how uncertainties in biomass

burning (BB) emissions, anthropogenic emissions, and secondary organic aerosol (SOA)

production mechanisms affect PM2.5 predictions over the contiguous U.S. during wildfire and nonwildfire

periods in 2021. Simulations driven by two estimates of wildfire emissions, two estimates

of anthropogenic emission, and two representations of organic aerosol formation are compared. In

August, simulations driven by the Global Fire Emissions Database (GFED) biomass burning

inventory produce PM2.5 concentrations >38% higher than that driven by the Global Fire

Assimilation System (GFAS), overestimating PM2.5 by more than 80%. Compared to groundbased

observations, the GFAS-driven simulation has the highest spatial correlation and lowest bias

in August. In December, performance is similar across scenarios, with the simple SOA scheme

yielding lower bias than the complex scheme. These findings demonstrate that BB inventories

dominate uncertainty in simulated PM2.5 during wildfire-impacted periods, whereas SOA

chemistry exerts greater influence when wildfire activity is limited. Accounting for these

uncertainties can guide the choice of model parameters and inform future model developments to

improve their policy or source attribution applications.

Second, modeled PM2.5 concentrations and chemical composition over the contiguous U.S. from

July 2021 to June 2022 is adjusted using satellite-derived AOD from both Standard and NRT

products. NRT-adjusted PM2.5 is broadly consistent with Standard-adjusted results. Satellite AOD

adjustment generally improves the modeled spatial variability but increases the bias against

observed PM2.5 concentrations, particularly during summer over the western U.S. Summertime

wildfire smoke introduces further uncertainty in AOD-PM2.5 scaling framework by altering the

relationship between AOD and surface PM2.5. Across all regions during summer, organic aerosol

is the most consistently overestimated PM2.5 component. Overall, the performance of satelliteadjusted

PM2.5 varies by species, region, and season, but does not depend on the latency of the

satellite AOD product.

Third, interdecadal changes (2002-2011 vs. 2012-2021) in urban-rural temperature differences

(ΔT), AOD, and surface PM2.5 concentrations are evaluated across cities in the U.S., Canada, and

Mexico. Air quality improved across much of the eastern U.S., while increasing wildfire activity

contributed to rising PM2.5 concentrations in parts of western U.S. and Canada. Relationships

between urban-rural differences in aerosol burden (ΔAOD and ΔPM2.5) and ΔT vary by region and

by diurnal period. Tropical regions show a positive relationship between ΔAOD and daytime ΔT,

whereas there is a negative relationship between ΔPM2.5 and nighttime ΔT. Arid regions

commonly exhibit a daytime urban cool island and nighttime urban heat island, with positive

associations between daytime ΔT and aerosol changes. Temperate regions display opposing

daytime and nighttime ΔPM2.5-ΔT relationships, whereas ΔAOD-ΔT only showed daytime

associations. No consistent relationship is observed in cold/continental climates between aerosols

and ΔT. The U.S. is the only country exhibiting a statistically significant positive nighttime

ΔAOD-ΔT relationship. These findings demonstrate that urban aerosol-temperature interactions

are regionally complex and do not translate uniformly across North America.

This dissertation advances understanding of key drivers of uncertainties in PM2.5 modeling,

evaluates the performance of NRT satellite-constrained PM2.5 estimates, and characterizes

evolving aerosol-urban heat interactions across North America. By integrating chemical transport

modeling, satellite remote sensing, and urban-scale analysis, this work provides insight into PM2.5

mass and composition across spatial and temporal scales. Continued integration of observational

and modeling approaches such as these is essential for reducing uncertainty in air pollution

estimates and informing effective mitigation strategies.