Storing time-integrated variables

[AlexisRenchon]

Storing time-integrated variables (e.g., daily/annual sums)

Some parameterisations require time-integrated quantities — for example, the Optimal LAI model uses the annual sum of potential GPP and precipitation. Currently, these accumulated variables are stored in the model cache (p.biomass).

Should we instead promote them to prognostic state variables in Y? They could live either within their respective model's namespace or in a new shared namespace, making them accessible across models.

Handling climatological inputs (e.g., mean annual precipitation)

Some model inputs represent long-term climatologies that are assumed constant but should, in principle, evolve under a changing climate. How should we handle their time evolution?

• Keep them constant: simple, but ignores climate change.
• Update yearly: captures trends, but introduces artificial step changes at year boundaries.
• Use a moving average: provides smooth evolution, but adds complexity and requires choosing a window size.