The current state of carbon-cycle data assimilation

Model–data fusion (MDF), also called ‘data assimilation’, describes the approach of combining as much data as possible with explicit dynamical models to constrain the evolution of any system. The paradigm is weather prediction. MDF holds the promise of combining the plethora of data available on the terrestrial carbon cycle but to date there have been few successful applications using multiple streams of data. The two most likely reasons for this are a lack of statistical sophistication in the attempts and weaknesses in current terrestrial biosphere models; these failures do not represent a failure of the paradigm. Solutions lie in a more rigorous consideration of the statistics of the problem and the inclusion of more elements of the model within the MDF framework.