Closed teixeirak closed 3 years ago
write a review on what we've learned 50 years since Odum...here's my chance (a couple years late)!)
Wow, I feel like that's an idea worth keeping. It is time for something like that.
I very much like your ideas above K!
Updates:
Before proceeding with editing the age trend figures, I'd like to get coauthor comment (especially @ValentineHerr) on the options:
This is the easiest, and sufficient.
This would be along the lines of the second panel of the schematic (current version below--imagine this with NPP broken up into several components). I think this would be super cool!
Critically, doing this would require that, at a minimum, we fit a different functional form to age trends in NEP (something that's probably good to do regardless, unless overly complicated, @ValentineHerr ). I think it would be pretty easy to import the equations we have generated into my Matlab script for the schematic figure, so wouldn't be a terrible amount of work.
If we do this, we'd want to make another figure showing biome differences across mature forests (currently in age trend figures).
I really like those stacked area plots!
I'm going to play a little more with this to see if it's feasible. Besides the fact that we're currently fitting the wrong functional form to NEP (and probably others, but NEP is most egregious), we're also missing data for some fluxes (especially for tropical forests), so I'm not sure how well it could work.
@teixeirak these figures look great and think it is feasible to re-fit equations. I don't have much time to dig into this (I have a couple meetings for my project with Bill), but I can certainly give it a try. What functional form are you thinking of? for now, I believe (if I dug up the right info), all variables for young biomes are fitted with a linear mixed model that looks like this: lmer(mean ~ -1 + log10(stand.age) + Biome + (1|geographic.area/plot.name), data = df.young_model) and we add an interaction between Biome and log10(stand.age) if we have enough data and if log10(stand.age) is significant on its own.
Great, @ValentineHerr ! I'll need to think a bit about what functional form we'll want. I'll open a new issue with that.
Here's a very rough draft of the stacked area results plots I'm hoping to make:
@bpbond , inspired by your example-- thanks!
Here's an example rough first draft age trends figure with real results (for age trends, not mature forests):
I like these and am almost certain that I'd like to include them (probably in addition to modified version of current age trends figures).
However, there are a few caveats: (1) There are, of course, multiple different variable combination options (e.g., include total biomass or AGB + root?), and in some cases there is significant lack of closure. It will require going through each flux to see which are most trustworthy. I'll probably include some dashed lines comparing different ways of estimating (e.g., dashed blue and red lines are GPP and Reco, respectively). (2) some data are missing age trends (e.g., Rauto), so for those I just calculated based on relationships of other variables. (3) We fit the same functional form to all variables (y~ log10[age]), but this is clearly not the best fit for all (especially NEP; issue #62 ), and it is mathematically inconsistent with expected trends. We may find a good solution to issue #62, but my memory from 3-4 years ago is that @ValentineHerr looked into this but we determined that our data wouldn't support fitting models with more parameters, which would be necessary to give more realistic curves. Because of this, we'll probably end up with a situation where these figures are a bit inconsistent with the schematic, and with what we understand as the reality. I think that just ends up being discussion--we currently have insufficient data to resolve the shapes of all these trends (and within-biome variation is large enough that this isn't the best way to resolve those trends).
@bpbond , it would be great to hear what you (or others) think of this, particularly (3) above.
đź‘Ź nice! What are the white, black, and blue lines in the figures?
Re (3) I agree that's a really interesting point for the discussion—we don't have the data to resolve the functional form unambiguously. An interesting parallel can be drawn with Sulman et al. (2018) I think.
Thanks, @bpbond . I added these figures (still rough drafts) to the manuscript.
The solid and dashed lines show estimates for some key fluxes as regression models vs sum of regression models for components. (Don't worry, I'll be sure to get all that in the legend/caption.)
I think it's still good to show current figs 7-8 (minus the maps), right?
Yes, I think so.
Ok, I think we can call this issue resolved.
R3 makes a lot of comments about the figures and proposes rearrangement:
Third, I found the presentation of the results made it difficult to see clearly the major differences in C fluxes and pools across biomes and age classes. The illustrated C budgets (the majority of the figures, 8 in total) are visually very appealing, but the reader has to do a lot of flipping back and forth to see how any particular flux or pool varies across biomes and age class. Figure 6 is more synthetic, but each panel is very small and the differences from one group to the next are hard to see. It is also difficult to compare results for young forests (as scatterplots) with the box plots for the mature forests. My advice is to move some of the budgets to SI, and include in the main manuscript more figures that clearly illustrate the most interesting trends with biome, and to allow an expansion of the results in Figure 6 (for example, by first showing scatterplots of all forests, young and old, as a function of stand age), and then perhaps another that is a box plot comparison of mature versus old forests in each biome.
Note: @bpbond said in an email that he likes the proposed rearrangement.
Here are my initial thoughts: