I suspect that the second approach will be more computationally effective. Most trophic relationships are done by resource type, not functional type. This makes it easier to sort through, say, a dictionary of fallen seeds in a grid cell to find the appropriate seeds than sorting through a list of plant functional types to see if each has seeds and then if you can even eat those specific seeds.

At the moment, the PlantsModel isn't exposing very much at the plant functional type level through Data.

• The two propagule arrays do have a PFT dimension - they potentially have different sizes.
• layer_leaf_mass is aggregated across PFTs, although we could disaggregate by PFT
• The subcanopy vegetation does not have PFTs at the moment - sosubcanopy_vegetation_biomass and subcanopy_seedbank_biomass are not structured that way.

The main worry I have right now is the stoichiometry. Each cohort will have stochiometric ratios. We can take the mass weighted average across resources to give "average" leaves/fruit - either across all PFTs like in the layer_leaf_mass or by PFTs like in the propagules. That's certainly the easiest solution! Ultimately though, animals might want to selectively feed on the most nutritious trees/fruit. That would force us to move to a per cohort resource rather than per PFT. The ugly thing there is that once you're at a cohort level, the data arrays are ragged which is all sorts of not very xarray but that may be something we just have to solve.

I don't have strong feelings about the two alternative implementations but I think your intuition about the integrated one is right. We're going to have to have that logic somewhere and I think it is going to be cleaner to set up resources where there is an established mapping between the existing animal foraging requirements and the available resources.

Minor things:

• I don't think you have to handle trapping missing variable names. If they aren't defined, the PlantsModel is screwed up.
• IIRC, the values in the data layers are per cell not per m2, so you don't have to scale by cell area.

Ah interesting, so for now I can ignore PFTs for everything but propagules?

I am happy to do the easiest and simplest version right now and worry about more guided foraging patterns in the future.

I will go with the resource-type integrated classes for now then.