Questions with answers:

Just a list of few questions we might want to discuss about while in Seattle:

• Core lengths! Fixed!
• At what point should we be worried about the short-term scale being pushed to at/under 1 year? Answer: As long as the $\rho$ and $\gamma$ (scale and intensity) are not strongly correlated and $\gamma$ not pushing towards 0 (which would mean you're heading towards white noise.
• If we want to add an allometric (size-related) mechanistic relationship, what would be the best way to implement it? (we mentioned two different possibilities in Banff) Answer: Find the notes on this! But seems like an extra component we may not need right now because the long-term is likely doing this for us.
• How do we want to model different sites? Should we include their geographical distances (but knowing that sometimes that is less relevant, for example two sites close together but far apart on an elevation gradient could be pretty different .... we could make a metric of how far apart they are in climate space but that seems hard to make and harder to interpret). Answer: This would be a nightmare. If we want some of this general information, the way you should do it is to check whether different stands have the same functional behaviour.
• Is there any way to tell how much the model is getting its climate responses (e.g., $\beta_\text{GDD}$) from growth responses across space, versus within a single tree? Answer: We discussed this the first day, it seemed like no direct way, but that the model should integrate over these and definitely will use some of the inter-annual variation in climate responses. Mike says: "A lot of the evolution of this [stuff I am asking for] is in the climate variables themselves and we're not modeling that. We're looking conditionally on how the growth of trees correlates with local climate behaviour. The modeling is made easier that the growth only depends on the climate and its current behaviour and how the climate got that way. We decouple the local conditional relationship of climate and growth and how the climate got that way." This is okay, but we need to make sure get good coverage of forecasts in future (we're really relying on our assumptions here). "A more subtle issue can arise if there is confounding. Like sunspots drive climate and increase light for trees [okay, we are so not worried about this].'
• If we want to add a correlation matrix for plant species that spans angiosperms and gymnosperms, how could we do this robustly? Answer: Just add them as two separate phylogenetic groups (two separate matrices, instead of one). _
• Can we decompose long (species) and short (stand) terms? Answer Victor thinks the answer is yes, once we divide out the core lengths.
• How do we interpret the short (stand) and long (species) term $\rho$? Answer As stand and species GP dynamics; species level captures allometry, disease and stuff on a longer scale, while stand capture shorter-term possibly climatic stuff. And consider now changing the priors on them (see to do).
• How to add species phylogeny to the short (stand) and long (species) components? Answer We answered this without Mike: basically be sure we want it.
• VPD and soil moisture interaction -- how would we model it? Answer Add a linear interaction for now and leave it in even if it is 0.

• What would be the modeling strategy to test for/implement climate legacy patterns? (if we believe they matter?) Answer We didn't really ask this as our model already feels like it has a lot ... maybe include this as a specific separate project?

Questions that don't exactly have answers:

• How do we want to implement different species response to climate? ($\kappa$, $\beta_\text{sm}$...)
• How to find changes over time in recent year? Answer so far Mike suggested looking at the retrodictive checks ...

To do from meeting

• Put lower bound of 1 on $\rho$ for short-term; put an upper bound on $\rho$ of the longest data we have.
• Plot the visualization of the exp. quadratic instead of the posterior of the terms! This would be very nice! And show the decay better.
• Try wider priors on rho for stand and species GP (especially as we are now thinking of them as 'stand and species GP' and not short and long term GP).
• Add a linear interaction for VPD and SM.