Methane, a greenhouse gas that has a much higher heat-trapping capacity than carbon dioxide over short time frames, is a powerful driver of climate change. Wetlands are known to be a significant source of natural methane emissions, but the factors underlying specific methane emission behaviors are poorly understood. Improving our understanding and monitoring of wetland methane emissions is essential for accurate climate modeling, as it helps assess the overall impact of these emissions on global warming and informs strategies for mitigating climate change. The goal of CH4cast is to help in understanding methane fluxes within wetland ecosystems.
Flux towers are the measurement equipment used to understand terrestrial (including wetland) methane behavior. While they are the gold-standard for observing methane fluxes, they can be expensive and difficult to maintain, and there is minimal coverage worldwide. With these data-sparsity challenges in mind, this project has three goals: 1) To train and compare machine learning models using globally available variables that can give us a ballpark estimate of the CH4 fluxes at wetlands without flux towers, 2) To develop models that are interpretable enough that they might point researchers to important physical processes driving methane emissions in wetlands, and 3) To visualize our predicted methane fluxes at a number of identified wetlands without flux towers, alongside the existing fluxnet data.