From the world generator’s perspective, I think we would want to have both the ‘potential biome’ without regard to ecosystem and the fully classified ‘ecosystem biome’ and either of them would be used in different situations.
So, the worldgen would work by partitioning the world into ‘potential biome’ regions, which would then have species propagate across them over time, and the ‘ecological biome’ would be dynamically changing over time, and would be categorized based on the static ‘potential biome’ and the dynamic species populating it.
The ‘potential biome’ regions themselves would be static in the initial implementation (as they are right now) but they could also be changed over time to add in environmental changes. But I think we will wait to have that implemented until later; I at least rank that as a lower priority.
The conditions can also vary drastically over the course of smaller timescales of a year, like for example the rainfall over Africa:
It would be easiest to handle this by just assigning based on average for the entire year. I am not sure if on the timescales of Macroscopic and Aware that we could easily distinguish the differences between seasonal variation and variation over eons without having each game cycle represent an entire year and have the player have to survive through all seasons. This is above however we conceive of implementing the day/night cycle, which would surely be much longer than it currently is in Microbe/Multicellular.
This is somewhat of a tangent, but I think in the 3D world we will need a day/night cycle of around 20 minutes, as any shorter than that in a 3D game would probably be too disorienting, and perhaps nauseating to have the planet rotate much faster.
I split off further discussion of the day/night cycle length here: 3D Day/Night Cycle Length