SI figures

[beckybanbury]

@teixeirak

What do you think to something like this for the multi panel plots? This is for the six climate variables that we present in the paper, and for the four major fluxes presented in our current figure 5. Plots are not scaled

[beckybanbury]

Helene also suggested a scatterplot matrix for the climate variables to be plotted against each other - I've plotted out this (which obviously has each pair against each other twice so isn't ideal); how do you think it would be best to present this data?

[teixeirak]

You could do that figure with just the half above the diagonal.

[teixeirak]

It looks good. I’m pretty ambivalent as to whether it’s better to use this or the scaled version that you currently have. Up to you.

If you go with this, you could condense by labeling common axes just along the left side and bottom (and use the same x-axis scale for all clime variables).

[beckybanbury]

It looks good. I’m pretty ambivalent as to whether it’s better to use this or the scaled version that you currently have. Up to you.

I was thinking we could keep the original one in the main manuscript and that this could just go in the SI, so that people could find the unscaled data if they were interested?

[teixeirak]

Okay, sounds good.

[beckybanbury]

[beckybanbury]

SI to add:

• [x] table/graphs of effects of climate within growing season
• [x] table/graphs describing lack of major trends in allocation between variable pairs tested
• [ ] table with details of %variation explained for each flux (summary of results, r-sq, p-values etc; similar to table previously included)

[beckybanbury]

SI alternative to fig. 2; unscaled results for latitude

[teixeirak]

I found our most recent correspondence on this:

First, I think it's a good point about figure 2 being hard to read with the lines all plotted on one graph. It could be worth splitting it into panels (even if only 2 or 3), and that would also have the advantage of allowing us to plot confidence intervals. It probably is also re-considering the scaling of the lines. Because they don't allow direct comparison, the main benefit of it is really bringing all of the lines into scale with each other, for ease of plotting on one graph. Helene's comments have highlighted that, for readers, it might be more interesting to be able to read actual values off the graph - this might be more informative than the scaled values. What do you think?

For Fig. 2, breaking them up into 2-3 panels would be very repetitive with Fig. 3. Having reviewed our references to figures in the manuscript, I don’t know that we really need Fig. 2. Are there any variables there that are not included in Fig. 3? I do think ti would be good to generate nice figures with confidence intervals for each variable independently for the SI.

[teixeirak]

All of the fluxes in Fig. 2 are also in Fig. 3, so I let's just drop Fig. 3.

[beckybanbury]

Graphs for effect of climate within growing season

[teixeirak]

Graphs for effect of climate within growing season

Looks good!

[teixeirak]

It looks good. I’m pretty ambivalent as to whether it’s better to use this or the scaled version that you currently have. Up to you.

I was thinking we could keep the original one in the main manuscript and that this could just go in the SI, so that people could find the unscaled data if they were interested?

[teixeirak]

Graphs for effect of climate within growing season

@beckybanbury, Did you run this analysis on respiration?

[beckybanbury]

@teixeirak none of the climate variables are significant for R_auto (but sample size is relatively small in comparison to the other C fluxes)

[teixeirak]

Thanks! Let's keep it in this figure.

[teixeirak]

Actually, thinking a bit more... We have the 9 variables. This figure is still missing R_auto_root and BNPP_root-fine. We need a consistent criteria for which are plotted. I think that since it's SI, it's good to show all of them. I don't know if 9 can fit, though. The other criteria that could make sense is to just plot the variables with at least one significant correlation, although in a way that's presenting a bit of a biased picture.

[beckybanbury]

@teixeirak if we plot the above plot with all 9 variables, should we also plot the first plot in this thread (flux against climate variables) with all 9? Currently it is only 4 fluxes

[teixeirak]

I think that would be good if we can do so feasibly. It's SI, so no need to be too selective with what gets into the figures.

[beckybanbury]

[teixeirak]

@beckybanbury, could you do the 9-map plot without the legend (not needed because its in the title), and perhaps just 2 maps per row? Since this will be in the SI, we have plenty of space for a full-page figure.

[beckybanbury]

like this?

[teixeirak]

Looks great!

[teixeirak]

@beckybanbury, looking at your checklist above, it seems we have the first two covered with figures. I'd recommend adding the p-value or dAIC, R2, and perhaps equations to each subplot so that we have those presented without having to make tables. I think we can also refer readers to a .csv table in the results folder, just need to make sure its clear which one is relevant.

I think this is still on your list: "table with details of %variation explained for each flux (summary of results, r-sq, p-values etc; similar to table previously included)". I wonder if that could be solved with the same approach as above?

[beckybanbury]

I think that this works fine for the growing season data (above), but it is a little harder to read for the climate and seasonality plots (see below) - what do you think?

[beckybanbury]

@teixeirak In the manuscript draft someone has suggested including a table of the coefficients and stats for the MATxMAP interaction models - this seems like a good idea.

I also think it would be worth making a table of the outputs for each variable, as the graphs we've plotted don't plot every variable we tested. E.g. from the manuscript: "In addition to MAT, temperature seasonality, annual temperature range, and annual frost days were consistently identified as strong univariate predictors of FACF. {cite some table/ SI?}" - we haven't graphed annual temperature range or annual frost days as they correlated strongly with temp seasonality, but we should include the results somewhere if we've mentioned them (or alternatively drop them entirely). Similarly "Of the climate variables tested, annual wet days, aridity, cloud cover, precipitation seasonality, maximum vapour pressure deficit and water stress months were poor or non-significant explainers of variation in FACF, explaining less than 20% of the variation in each of the carbon fluxes."

Obviously that will be a lot of information to include!

[teixeirak]

I think that this works fine for the growing season data (above), but it is a little harder to read for the climate and seasonality plots (see below) - what do you think?

I'd say it's borderline. Since it's the SI we don't need to be too picky-- perhaps leave them in for now? (can always remove if coauthors or reviewers complain)

For both this and the comment above, I think it would make sense to make sure that the tables with model outputs are easy to find within GitHub and clearly labeled. Then, tables that are small enough could easily be pulled into the appendix, whereas larger ones could simply be referenced, and perhaps loaded on the journal webpage as separate appendix files (in .csv format).

Does that sound good to you?

[beckybanbury]

Yes, great! I'll try and get a simple table to add to the appendix, and the rest can be referenced like you say.

I'm happy to leave the R2 values on the plots