This is a really interesting study. I love your clear hypotheses, and the ideas tested are really nice. The data collection was impressive. However, I felt that the weak alignment of the hypotheses to the findings presented, and the amount of repetition of information in the discussion are letting you down a bit. I have provided suggestions that I hope will help you to improve the paper.
MAJOR POINTS
1. You “hypothesized that PTNs will be less connected and complex in communities in more arid environments, with lower phylogenetic diversity and functional richness and productivity, and, by contrast, PTNs will be more connected and complex in cooler, moister environments, with higher phylogenetic diversity, functional richness and productivity”. [lovely clear hypotheses!] Your results were interpreted as “Confirming hypotheses”. However, you found the expected relationships for aridity, but not for phylogenetic diversity, functional richness or net primary productivity. That is, you seem to have minimised all the important results that were counter to your hypotheses. I think you should rephrase throughout to give equal weight to both expected and unexpected (even if they are null) results.
2. Given the mechanism you suggest for the relationship between aridity and connectivity/complexity, the fact that productivity isn’t correlated seems actually very important to me. I think this should be discussed, not swept under the carpet. I even wonder if this lack of relationship should be mentioned in the title (I don’t insist).
3. I think the paper would be stronger if you aligned the results, analyses and figures with the hypotheses. For example:
a) None of your hypotheses require a pooled climate variable. Thus, I think the climate PCA needs to go – it actively obscures the answer to your main questions.
b) Your results start out by discussing variation in traits, rather than by addressing your first hypothesis. This is more than an ordering problem though - I don’t think these ANOVAs are relevant for the hypotheses you set out.
c) The relationships between individual traits and climate (lines 350-360) are not relevant for the questions you pose here. Just because you CAN run analyses with your dataset doesn’t mean that you SHOULD. Including unrelated analyses increases your chances of false discoveries – and also waters down your findings as your novel results are hidden behind a bunch of analyses that have been done elsewhere. I recommend that you remove all the analyses that do not relate to your hypotheses (and thus delete the whole first section of the results, which is not relevant).
d) Fig 3 seems to not quite address your hypotheses about the relationships between network connectivity and complexity and aridity, PD, functional richness and productivity. Starting with the X axes - Instead of “Climate – PC1” (which includes all sorts of variables that are not included in the hypothesis), we actually need to see the results for aridity. We also have no figures here for productivity, or for phylogenetic diversity, despite these being in your initial hypothesis. Similarly, on the Y axes, we have a series of network parameters that don’t obviously relate to “connectivity and complexity” (see below).
e) Your hypotheses are about complexity and connectivity, but your analyses are about ED, AC, AL, and Q (D should be included too – not differing across sites with different aridity etc is still an answer, even if it is one you don’t particularly love!). If you don’t have a way to test the variables listed in your hypothesis, you probably need to change the phrasing in the hypothesis to something that you can actually test – OR clearly highlight in both analyses and figures which variables describe connectivity and which describe complexity – that is, we need a clear connection between your hypotheses and your analyses.
f) Line 367 – this seems to be the answer to your main question, but it is presented without any supporting stats.
g) Your second hypothesis was “that traits with greatest connectivity within the PTN, being involved in mediating multiple functions, would tend to show lower variation across species relative to other traits less connected in the PTN”. However, the first two paragraphs in the relevant section of the results (beginning on line 383) focus only on the traits with high connectedness. I don’t think the text from line 384-397 relates to any of your hypotheses, and thus I think it should be deleted. Figure 4 could also be labelled so that it is much clearer to the casual browser of your paper what it actually shows.
h) Table 1 could be relegated to the SI.
i) Table 3 is very nice – the network complexity explainers are very very helpful. However, this table only lists hypotheses for three of the 4 variables you mention in the introduction (phylo diversity is missing, and should be added if you keep this part – see below). I also don’t think the site value for each variable needs to be in the main text (so split the table after the rationale and move the right hand half to the SI).
j) You have given your predictions in the text, in table 3, and in figure 1 (and again in the figure legend). This is too much repetition. I recommend you make sure the rationale and predictions are really clear in the text, and drop either figure 1, or the prediction part of table 3; my vote is to drop fig 1)
k) Figure 1 is missing PD as a variable, and the networks here aren’t particularly helpful for helping readers understand your hypotheses. If you do keep this figure, I would just keep the top bit (a) – I don’t think part b (Hypoth 2) is particularly necessary or helpful.
4. The discussion spends far too much time repeating results and findings from previous studies (which is information you already gave in the introduction). Instead, your discussion needs to bring out the important new information, and show your readers how this helps the field to advance.
MINOR COMMENTS
1. I would avoid the use of acronyms (especially PTN, which is not widely known) throughout – acronyms make it that much harder for your readers to grasp your message, and they really don’t save that much space. I would definitely write out the variable names in full on your figures to make it easier for your readers to understand what you found (e.g. Fig 2, 3 and 4), and in your results. I found the results section particularly difficult to read with all the variables reduced to letters/acronyms instead of words.
2. Line 20 you say: “We hypothesized that given that network complexity would be associated with specialization of species within communities to greater available niches, trait network connectivity and complexity would be higher in communities with greater water availability, phylogenetic diversity, functional richness and productivity and lower in communities adapted to higher aridity”. This sentence is very complicated, and confounds the part that you are testing with the assumptions that underpin your putative mechanism. I recommend that you pull the mechanism and the clean hypothesis apart here.
3. When you list your results in the abstract, you use different words, relationship directions, and orders of information than in your hypotheses. If you align these, you will make your abstract much easier for your readers to understand.
4. I absolutely approve of the selection of the most abundant species at each site – but some people will likely worry about this – maybe explain why this is a good design choice.
5. Line 255 needs a full stop (or a period if you are American?).
6. Calling soil pH “climate” is a real stretch (also, why is soil pH listed twice both in the methods and the figure 3 legend?). You also seem to include both Aridity index and MAT and MAP in the PCA – you never explain what the aridity index actually is, but I wonder if it is calculated from MAT and MAP – which would mean that you included these variables twice.
7. Give the actual p values rather than asterisks or <0.05/>0.05please – this takes the same amount of space, but gives the reader more information – and is particularly important for readers who will be trying to assess the real significance of each p value given the number of tests run.
8. With only 6 sites, don’t you risk over-parameterising if you fit non-linear analyses? (this takes your remaining degrees of freedom down pretty low, and running each analysis for both linear and non-linear fits doubles the number of tests done – which makes me worry about false discoveries).
9. I am not at all sure about this, but the data distribution on the CVs (Fig 4) looks pretty non-normal. Might it be worth log transforming them for analysis?
10. Why Gross primary productivity rather than NPP?
11. Claiming that “our paper indicates the centrality of stress in driving simplifying of trait networks” is going beyond the scope of your data – you don’t really have information about stress.
12. You state “Single trait climate relationship analyses are often contradictory and challenging to interpret across multiple traits. Here we overcame this limitation by exploring PTNs”. I don’t think this is right – the PTN analyses are asking wildly different questions to the single trait climate-relationship analyses, they’re not just different approaches to the same question.
I hope at least some of these comments are helpful
Angela Moles