A Data-Driven Future for Atmospheric Chemistry, Wildfires, Climate, and Society
Makoto Kelp (Stanford University)
Colloquium
Thursday, February 15, 2024, 3:30 pm
Abstract
Modeling and measuring atmospheric composition are essential for studying the impacts of human activity and predicting future changes in a warming climate. However, our observational systems suffer from spatiotemporal blind spots and inequitable coverage. These measurements inform computationally expensive atmospheric models that have limitations in their ability to accurately represent the complexity of physical processes. With the escalating impacts of climate change and increasing wildfires in the Western US, we are left with insufficient data streams and tools to construct effective mitigation solutions. However, recent advances in machine learning (ML) and data-driven approaches applied to atmosphere models and air quality datasets can open avenues for new research perspectives. I will discuss how 1) ML models can replace the computational bottleneck within a global atmospheric chemistry model with a stable, faster emulator, 2) computational sensing can diagnose the optimal and equitable design of air pollution sensor networks at national and urban scales, and 3) data-informed modeling of prescribed fires may dampen the risks of future megafires. My research program aims to bring insights from atmospheric science to enhance the efficiency and interpretability of ML/data science practices while shedding new light on environmental justice issues.
Bio
Makoto Kelp is a NOAA Climate & Global Change postdoctoral fellow in the Climate and Earth System Dynamics Group at Stanford University. He holds a Ph.D. in Atmospheric Chemistry from Harvard University and a B.A. in Chemistry from Reed College. Makoto leverages data-driven methods, including machine learning and computational sensing, to uncover new perspectives in atmospheric chemistry and land-climate-human interactions. He places a special focus on how machine learning integration can facilitate the next generation of global atmosphere models and how data-driven science can improve our understanding of the interplay among fires, climate, and society in the Western United States.