Satellite measurements as a constraint on estimates of vegetation carbon budget

A methodology for coupling prognostic models of vegetation and remote sensing measurements is presented. This approach, based on control theory, postulates that errors in model parameters or initial states can be corrected, by minimizing the differences between satellite observations and modelled reflectances. These reflectances are computed from model state variables, (carbon in the leaves) with a direct “scene” model. The differences between this approach and previous studies dealing with the coupling of satellite data and vegetation modelling are briefly discussed. A synthetic experiment with a generic vegetation model, designed for carbon cycle investigations, shows that processes such as allocation of carbon to the leaves and phenological behavior of a deciduous-forest canopy can be controlled, and that carbon exchanges estimation can then be improved. Furthermore, we show that assimilation of real satellite observations in a crop model leads to improved simulation of vegetation functioning, as ground-truth information confirms that a good retrieval of the sowing date, the control parameter, is achieved. Finally, we point out the interests and requisites for application to natural vegetation carbon fluxes estimations at the global scale.