We don't currently output any of the biophysical variables resolved by land use type, we just have a few and they are mainly carbon cycle variables. I suspect that including a small set of the biophysical ones would be of interest to include when land use change is active, i.e something like: veg temperature, 2m air temperature, sensible heat flux, latent heat flux, shortwave absorbed and net longwave.

One issue is how to actually do this. The relevant data is all outside of FATES, but the classic way of outputting sub-gridscale data via the ELM and CLM vector output format probably wouldn't work, because the patch land use labels aren't constant in time, so it wouldn't be straightforward to map from the patch vector to a (lat x lon x landuse) array in postprocessing of history files.

The simplest way to do this I think would be to pass the relevant variables as a set of patch-resolved bc_in vectors and then let FATES put them on the fates_levlanduse dimension. It would have to be at high frequency, since these are necessarily all at that timestep.

A more complicated solution might be to be to pass the land-use labels to ELM and CLM as bc_out vectors, and then do the remapping onto a land-use dimension outside of FATES. That might be more extensible, but I'm not sure it would be worth the additional complexity required to add that. I suspect that the computational cost of each would be similar, but not really sure.

A further complication on any of this is the bare-ground stuff. Some fraction of bare-ground fluxes and temperatures would need to be averaged in with the vegetated patches of each land use type to do it properly.