Global model-data-fusion estimates of ecosystem carbon fluxes

Large uncertainties preside over terrestrial carbon flux estimates on a global scale. In order to gain an improved understanding of ecosystem C fluxes, we implement a Monte Carlo based model-data-fusion approach: we assimilate MODIS LAI, plant-trait data, and the Harmonized World Soil Database (HWSD) into the Data Assimilation Linked Ecosystem Carbon (DALEC) Model, and we implement our approach on an 8-day timestep 1 x 1 degree resolution for the period 2001-2010. In addition to observational constraints, we implemented a novel bayesian parameter inter-dependence network in order to impose ecological and dynamic constraints on DALEC parameter values. We determined the spatial and temporal dynamics of major terrestrial C fluxes and model parameter values on a global scale (GPP = 123 +/- 8 Pg C yr^-1 & NEE = -1.8 +/- 2.7 Pg C yr^-1). In order to validate our approach, we also implemented our model-data-fusion setup at flux-tower scale, and compared DALEC NEE fluxes against in-situ NEE measurements (AMERIFLUX network) across multiple biomes and plant-functional types (NEE daily bias = +/- 0.83 gC m^-2 day^-1). In anticipation of the BIOMASS mission, we examine the additional uncertainty reduction resulting from above-ground biomass data assimilation. We anticipate that our global model-data-fusion approach will be an important step towards bridging the gap between globally spanning remotely-sensed biometric data and the full ecosystem C cycle.