Assimilating Earth Observation Data into Land Surface Models

Data assimilation techniques such as the ensemble Kalman filter and the sequential Metropolis-Hastings algorithm provide a means of integrating satellite data with ecosystem models to optimally adjust their temporal trajectory. To some extent these methods can compensate for poor model parameterisations but a preferable scenario is to calibrate the model well in the first instance. This paper explores how a site specific model calibration can be adapted to a different site using only MODIS reflectance data. Results show that, using reflectance data only, estimates of the net carbon budget of a field site can be extended to a nearby site, but that this best facilitated by re-calibration rather than sequential data assimilation.