Evaluating Smoke Transport Between Canada and NYC Using a Graph Neural Network

Project Objectives

•What caused the smoke transport in the June 2023 Canadian wildfires?

•How can we use a graph neural network to emulate wildfire smoke transport?

•Smoke plumes degrade air quality, which is particularly important in heavily populated areas such as New York City.

Data and Results

•Data from GEOS-CF, a high-resolution model which combines the GEOS meteorological forecasting model and the GEOS-Chem atmospheric chemistry module.

•Identified and analyzed a couple fires of interest using the GEOS-CF data.

•Preliminary training and evaluation of a graph neural network.

Significance

•Machine learning is quickly becoming the future of modeling and is much more accessible to a broader community than traditional numerical modeling.

•Next steps include continuing to train and test the GNN, as well as improving it to take inputs from both real-world and model data.

Publication or Conference

Attending and presenting at AGU 2025.

------------------------------------------------------------------------------------------------

Center for Geological and Climate Hazards

Interdisciplinary Research Experience for Undergraduates in Atmospheric and Geological Hazards