Remote sensing of the terrestrial carbon cycle has been a long sought after goal of the Earth Science community. Understanding vegetation feedback on climate change is key to predicting future climate scenarios and atmospheric CO₂ levels. In theory, vegetation carbon uptake could be inferred from remote sensing inputs as the product of photosynthetically active radiation (PAR) incident upon the canopy, the fraction of it being absorbed by photosynthetically active vegetation elements (f_PAR  ) and the efficiency (ε) with which plants can use this absorbed radiation energy to produce biomass. While f_PAR and PAR have long been available from remote sensing observations globally, direct inference of ε has not yet been possible. Here, we report an innovative multi-angle imaging spectrometer design and analysis approach  designed to quantify ε from space. Validation so far indicates that the remote sensing algorithms and models are robust across a range of diverse forested ecosystem types.  These spectral measurements together with a new ecosystem-atmosphere carbon and water exchange model can infer the photosynthetic rate, net primary production (NPP) and evapotranspiration. Our multi-angle sensor design and analysis approach is grounded in recent theoretical and experimental developments that have been extensively validated using tower-based radiometer measurements and scaled to space using multi-angle spectrometer images from the European Satellite CHRIS/PROBA.

Based on these results, we developed a data assimilation scheme to modeling ε from data assimilation of remotely sensed observations and meteorological constraints. In the forward mode, the model reduces a spatially explicit maximum convergence efficiency (ε_opt) using a non-linear, multivariate response function of meteorological constraints. This multivariate function accounts for the variability in ε due to the highly dynamic changes in the environment in between satellite overpasses. However, since ε can also be observed directly during a satellite overpass, these spaceborne observations can be used to invert our model thereby inferring a pixel specific ε_opt every few satellite overpasses. As a result, ε_opt can be used to describe slower changes in the conditions affecting ɛ (such as soil nutrient status and edaphic water stresses for instance)

With this type of sensor and approach we have been able to accurately quantify interannual variations in NPP, evapotranspiration and autotrophic respiration at fine temporal and spatial scales across a broad range of forested biomes in North America and Australia.