Tammy Yuan: How Causal Inference and Machine Learning Advance Understanding and Modeling Wetland CH4 Emissions

Abstract:

Methane (CH4) is the second most important greenhouse gas, with a global warming potential 28-34 times greater than that of CO2 over a 100-year time horizon. Wetlands are the largest natural source of global CH4 emissions and can be linked to major climate feedback. However, the magnitude, temporal dynamics, and dominant drivers of wetland CH4 emissions remain uncertain from regional to global scale, partially due to limitations in understanding and modeling of wetland CH4 emission processes, and limited ground observations.In this talk, Dr. Kunxiaojia (Tammy) Yuan, will first introduce how to use causality inference to understand the dependencies between wetland CH4 emissions and their drivers, and how to model wetland CH4 emissions across FLUXNET-CH4 sites with physically interpretable and causality-guided machine learning. Then she will leverage causality-guided machine learning, multi-source CH4 observations (including eddy covariance towers in FLUXNET-CH4 and chambers), and remote sensing, to better estimate wetland CH4 emissions in the Boreal-Arctic area. Furthermore, she will demonstrate how climate change has significantly increased wetland CH4 emissions in the Boreal-Arctic area during the past two decades. 

Bio:

Dr. Kunxiaojia (Tammy) Yuan is a postdoctoral scholar at the Climate & Ecosystem Sciences Division of Lawrence Berkeley National Laboratory (LBNL). Her research interests include greenhouse gas (e.g., CH4 and CO2) emissions from natural ecosystems, biosphere-atmosphere interactions, disturbances (e.g., wildfires and deforestation), and advanced techniques including interpretable machine/deep learning, causality inference, remote sensing, and land surface modeling