Fig. 1
From: MetaFlux: Meta-learning global carbon fluxes from sparse spatiotemporal observations
Schematic diagram of our MetaFlux methodology from the meta-learning phase that meta-trains GPP and Reco from FLUXNET2015 eddy covariance flux tower data using station-level ERA5 reanalysis and RS-based MODIS predictors. The meta-trained and validated ensemble is then used to upscale a global 0.25-degree GPP and Reco at daily and monthly timescale between the years 2001 and 2021. Finally, the ensemble global mean and uncertainty estimates are validated and compared with evidences from the literature.
