polishing figures, tables, & analyses

teixeirak commented 5 years ago

analyses

[x] be consistent in the set of variables presented (although not every variable has to appear in every figure/ table). Currently, Table 1, Fig. 1, Fig. 2, etc. have different sets of variables. See issue #26.

tables

figures

general:

[x] it would be helpful to have a single color for each variable that is consistent across all figures.
[x] it would also be helpful to have different shapes for each variable, again consistent across all figures. This would be particularly valuable when people print it in black & white.

Fig. 1 (map):

[x] display consistent set of variables (#26)

Fig. 2 (all variables scaled):

[x] display consistent set of variables (#26)

Fig. 3 (stacked plots):

[x] as mentioned under general comments, use the same color/symbol for each variable. This will be particularly valuable here in that it will make it easy to identify variables repeated across plots, and therefore how the figure shows subsidiary components of various fluxes
[x] reorder: (a) GPP = NPP + R_auto, (b) NPP = ANPP + BNPP, (c) ANPP = ANPP_foliage+ ANPP_woody_stem, (d) BNPP_root, BNPP_root_fine, R_root, TBCF (total belowground C flux; no data-- just summed line)= BNPP_root+R_root
[x] remove numbers on variables, just list as e.g., "ANPP + BNPP" (rather than "sum of (1) + (2)")
[x] add x-axis labels

Fig. 4 (responses to climate variables):

[x] reconsider which climate variables to display: Temp seasonality and temp annual range should be quite closely correlated (right?), so I don't think it makes sense to include both. Also, I'd stay away from vapour pressure. I think that vapor pressure deficit would be a much better choice. It may be good to show either precipitation or precipitation seasonality, whichever explains more variation (even though they're not the strongest predictors, I think most readers will be very interested to see those relationships.)

Fig. 5 (ratio tests):

[x] drop ANPP_woody (issue #22 )
[x] adjust y-axis so that max is 2 instead of 3.

beckybanbury commented 5 years ago

@teixeirak in terms of figure 4, I'm unsure about the best precipitation variable to include. Precipitation seasonality is not a good predictor, so I don't think we want to put that in (not significant for most variables). Annual precipitation from WorldClim and from ForC show completely different patterns (see below), so I don't really trust either of them (!). The results using data from WorldClim certainly seem a lot less reliable. I don't know if it's worth trying to download another set of precipitation data from a different source to compare..? Effect_of_AnnualPre_MATURE_only_poly_all Effect_of_map_MATURE_only_poly_all

I'll exclude vapour pressure. VPD doesn't come out as a great predictor for any of the variables, but I could include it; other variables that come out as reasonably good predictors that could be included are annual frost days or potential evapotranspiration.

teixeirak commented 5 years ago

Please use the map variable (data from ForC). The big difference is that these data range up to higher values, and C cycling tends to be reduced in those very wet ecosystems. Very high precipitation is generally more localized (i.e., you wouldn't get 6000 mm yr-1 over an entire grid cell, but you get it in places where topography is conducive to high ppt). Most global gridded products, WorldClim included, have spatial resolution that is too course to capture these areas with very high precipitation.

beckybanbury commented 5 years ago

combined_plots @teixeirak what do you think of this combination of climate variables on the graph? I've only included four response variables because it was getting hard to read the graphs with the quantity of points if more than that were plotted, but I can change the set/add more if you think others would be more meaningful.

beckybanbury commented 5 years ago

@teixeirak shall I remove vapour pressure from the analysis entirely, and not bother including it in the paper? I've included variables that we tested which turned out to be bad predictors, but if you think that vapour pressure is scientifically a bit meaningless, shall I just leave it out?

teixeirak commented 5 years ago

Please use the map variable (data from ForC). The big difference is that these data range up to higher values, and C cycling tends to be reduced in those very wet ecosystems. Very high precipitation is generally more localized (i.e., you wouldn't get 6000 mm yr-1 over an entire grid cell, but you get it in places where topography is conducive to high ppt). Most global gridded products, WorldClim included, have spatial resolution that is too course to capture these areas with very high precipitation.

@beckybanbury, I should add... we also want to give precedence to the mat variable.

beckybanbury commented 5 years ago

@teixeirak I had already prioritised that!

teixeirak commented 5 years ago

Great! Just wanted to make sure.

teixeirak commented 5 years ago

@teixeirak shall I remove vapour pressure from the analysis entirely, and not bother including it in the paper? I've included variables that we tested which turned out to be bad predictors, but if you think that vapour pressure is scientifically a bit meaningless, shall I just leave it out?

Yeah, I think I'd leave it out.

teixeirak commented 5 years ago

@teixeirak what do you think of this combination of climate variables on the graph? I've only included four response variables because it was getting hard to read the graphs with the quantity of points if more than that were plotted, but I can change the set/add more if you think others would be more meaningful.

Sorry for the delayed reply on this. A few thoughts:

I like showing precipitation. Even if its not a top predictor, its something that's commonly reported (hence better local data), and readers will be curious.
Let's show either temperature seasonality or temperature annual range, not both.
Solar radiation is kind of interesting, albeit not a great predictor.
PET is a nice one.

teixeirak commented 5 years ago

I guess what I'm saying is that the set of variables you showed above was great!

teixeirak commented 5 years ago

NEW ROUND:

General

[ ] ensure compliance with journal guidelines
[x] use subscripts for variable names (e.g., $BNPP_{fine root}$ -- I think that annotation style should work)
[ ] for latitude, use "absolute latitude (°N or S)"
[ ] carefully check all axis labels -- I noticed a few missing units/ parentheses around units,

Figure 1 (map)

[x] lat and long labels are small and probably unnecessary.
[x] better to put legend on top of map, if space permits. The text may also be a bit too small.

Figure 2 (flux closure) - looks good!

Figure 3 (latitude)

[x] add y-axis label
[x] add x-axis title

Figure 4 (MAT-MAP)

[x] add y-axis label
[x] add x-axis title

Figure 5 (multiple climate variables

[x] update caption

Table 1

[ ] confirm that I was correct to switch references between 1.1 and 1.1.alt
[x] knitting the manuscript on my (or other) computer breaks because of the file path that's specific to your computer.
[ ] I don't think the use of () is very intuitive, and it also overlaps in meaning with no. Consider replacing with no or n.s.
[ ] add legend

forc-db / Global_Productivity