About Us

Ambient air pollution causes ~4% of total deaths in the United States, more than three times the number caused by motor vehicle crashes. Our group assesses air pollution-related effects of human activity, focusing on mechanistic modeling of outdoor air pollution and its health impacts, quantifying inequities in the distribution of those impacts, and proposing and testing solutions. We study the relationships between emissions, human activities that cause them, and the resulting health impacts, and we develop modeling capabilities to enable these types of analyses.

Approach

How do we science?

Atmospheric modeling

Modeling the emission, transport, transformation, and removal of chemicals in the atmosphere

Impact assessment

Modeling the impact of behavior and policy scenarios on human health and health equity

Machine learning

Leveraging machine learning and artificial intelligence to improve existing models and allow new insights

Projects

Air-quality-related environmental equity impacts of decarbonization scenarios

The U.S. and other nations are at the beginning of a major technological shift from a fossil fuel-based economy to one driven by non-climate-forcing energy generation. The speed and success of this technology shift (“decarbonization”) depend in part upon whether it is perceived to be implemented equitably with benefits that accrue broadly across society.

An Observation-Driven Framework for Air Pollution Equity and Justice Intervention Modeling

From highways to industry, construction to residential energy use, Black people and other people of color in America are disproportionately affected by nearly every major source of air pollution. This exposure shortens life expectancy, increases incidence of disease, and compounds historic injustices and economic disadvantage.

Interpretable machine learning for high-speed, high- fidelity GEOS-Chem model simulations with uncertainty quantification

For many full-physics Earth Science models—including the GEOS-Chem model—there are two main causes of their high computational cost: 1) the models represent phenomena that occur at disparate spatial and temporal scales, which result in stiff differential equations requiring small timesteps for integration; and 2) representing (part of) the Earth System requires a large number of state variables, resulting in a large memory footprint and computational burden for processes that operate on each state variable.

Machine-learned atmospheric chemical mechanisms

Atmospheric chemical reactions are responsible for a number of important phenomena, including smog and some of the drivers of global climate change. However, these reactions are computationally intensive to simulate using traditional methods, resulting in models of the atmosphere that are slow and expensive to operate.