Description
For fluxes it may be advantageous to assimilate using the mean flux and uncertainty over a specified lag, rather than just have the option to assimilate a single timestep estimate. We already do this for wood stock changes relating to mortality, production and increment. However, we could do the same for other fluxes e.g. GPP, NEE, Reco. This change would increase the flexibility with which we assimilate these datasets, for example matching the weekly or monthly GPP when running at finer timescales. For single-timestep constraints, the lag would default to one timestep, matching the current implementation.
However it is a non-trivial change to do properly due to the restructuring of the OBS object and repeated changes to the likelihood functions. Luke and I have chatted already, but leaving this as a note for a reference/reminder