Linking Forest Biomass and Carbon Flux Using a Machine Learning Model of Aircraft Fluxes and Forest Canopy Structure

Forests play a critical role in climate and potential mitigation efforts by processing significant amounts of carbon from the atmosphere. A comprehensive understanding of vegetation processes is therefore required for linking forest biomass with integrated carbon flux. However, large uncertainties persist in terrestrial biosphere models that project carbon exchange, and discrepancies persist with atmosphere-based methods to attribute forest carbon fluxes. Here, we describe a machine learning approach that combines direct carbon flux observations from NASA's Carbon Airborne Flux Experiment (CARAFE) with remote sensing forest canopy and ecosystem information over forested sites in Maryland. This methodology provides a fully empirical constraint on the relationship between forested biomass and net ecosystem exchange (NEE) for forested sites across the state. We further evaluate regional-scale carbon flux estimates from the high-resolution Ecosystem Demography (ED) biophysical process model. These results enable development of a framework synthesizing multiple approaches to improve characterization of carbon flux estimations and uncertainties intrinsic to policy and decision-making.