Rödenbeck two-step scheme
Before implemented the "background" (in the Rodenbeck meaning) in the CIF, I need to understand better the two-step approach. My objective is to fully implement it within the CIF, so it becomes possible to do 2-step with any combination of global/regional model.
Please correct me if I am not correct:
- Step 1: compute a regular global inversion with whatever model with all possible observations, making sure that the spin-up period is sufficient
- Intermediate step: run two forwards with the global model: one regular, and another with all fluxes at 0 with the exception of fluxes within the regional domain
- Step 2: Run the regional inversion by accounting for the background as computed in the intermediate step.
Although there is no conceptual difficulty in this recipe, I have a strong concern about it. In the intermediate step, regional fluxes are transported in the global domain, hence allowing them to leave the regional domain and re-enter it later. On the contrary, with the regional model, once air masses have left the domain, they cannot re-enter. In the end, there will be a misfit between the background computed with the global model and what really sees the regional model. Typically, some mass is lost and the regional inversion will try to compensate for it by increasing artificially regional fluxes.
Although this should not be a real issue for short periods of inversions, I think it is critical if one computes e.g., one or more years of inversion with the two-step scheme, due to the re-circulation of air masses in the hemisphere.
Did I miss something? Or there is a critical limitation here (that is probably known by the two-step users?)
Thanks in advance