THE CHALLENGE OF SPATIAL RESOLUTION: MACHINE LEARNING LIKELY OVERESTIMATED GLOBAL TERRESTRIAL CARBON UPTAKE BY 30%

Speaker: Songyan Zhu, University of Southampton, United KingdomGlobal terrestrial ecosystems absorb about one-third of annual anthropogenic carbon dioxide emissions, yet the exact uptake remains uncertain. Gross primary production (GPP) estimates range widely from 100 to 200 Pg C, with significant uncertainty in interannual variability. Upscaling GPP from eddy covariance networks using machine learning and Earth observation data, such as MODIS satellites, has advanced global GPP estimates to approximately 140 Pg C in 2010. However, the impact of satellite spatial resolution on GPP upscaling remains underexplored.