I revised the paper entitled “Pervasive impacts of climate change on the woodiness and ecological generalism of dry forest plant assemblages” by Moura et al. In this paper the authors used environmental niche models in 2800 species of the Caatinga region to devise the potential plant assemblages of the future and elaborate on how the change in composition (species richness, beta diversity, % of narrow-range species and % of woody species) may occur with climate change. The paper is abundantly clear, well written and benefits from a strong “by the book” building of ENMs. Because the Caatinga is an important area known to be exhibiting increasing risk of changes with climate change I think this work is a very nice addition to literature highlighting the problematics of the area. That said, there are a couple of things I’d like to highlight, that I think can be improved.
Major comments:
Error assessment. The authors are using a total of around 2800 (one per species) x 6 (types of model) models, each of which would have a given uncertainty associated. Key here is how the process of error propagation of each model is handled when starting to stack predictions to calculate diversity indices. The authors did not comment on this issue thoroughly. I understand it can be a very tricky issue but ensembles should give proper estimates of errors, especially is some tests are going to be performed afterwards. My concern is that, for a given place, this error accumulation may be huge, preventing to state that a given pattern is significant.
Authors should also dedicate more work to highlight limitations of their study. There are many limitations in ENMs models that need to be at least mentioned (e.g., inference from space for time substitution approaches). Most importantly, if the main message of the paper is of such “catastrophic” nature.
Suggestions:
The message of the paper is in general simple. That is perfectly fine (very appreciated actually), but there are a couple of things that I feel could have been investigated and there is room in the paper for its inclusion. Those are mere suggestions.
Although I understand why to separate woody from non woody plants; I wonder whether it may result interesting to also make some room for looking at shrubs separately. Shrubs are considered key growth forms in drylands because they have access to deep water layers without exhibiting the high water demands of trees. Also they are announced to increase importantly with increasing aridity. Thus I think the patterns that authors may find for trees cf. shrubs may not be in the same direction and separating them may lead to more clear results and interesting tradeoffs.
Finally, PCAs approach used is robust for doing ENMs, however they prevent to conclude anything about the major drivers of this species lost. It would be good to have a look at which are the major determinants of species composition variation (e.g., by looking which are the major factors defining niches).
Minor comments:
Line 43: has altered or is altering, rather than have been altered
Line 53: climate emergency is a higher threat: higher than what?
Lines 70-73: on the other hand one may argue that woody species have higher resistance to drought and aridity because they can access deep water resources, right? This is something important to consider, because this process may not be well reflected in ENMs. This also is important when discussing hypotheses in lines (79-81).
In introduction I miss a couple of definitions such as: narrow-ranged species (respect to what in particular?; I’d try to give the reader here a clearer perspective on what are authors addressing exactly). Also biotic homogenization-heterogenization: ecosystems may homogenize by loosing species, but also by making all of them equally abundant; defining the scope and what is going to be measured in introduction may help the reader to have a clearer expectation from the very beginning.
General in Methods: Very thorough and “by the book” methodology on niche building leads to strong base for conclussions. The only exception is lack of citations for the binarisation process (lines 154-160). I find it reasonable in any case.
Line 149: what is the baseline period? I guess it is current climate? This has not been described yet, right?
Figure 4 is not highly understandable because it’s hard to grasp the joint variation of both features only with a color gradient (e.g., it is hard to distinguish if white colors are associated to high % of woody or to low). I think separating both processes may help even if it leads to a 6 pannel figure.

I have read the revision made by the author's and I find my questions and comments have been adequately adressed. I have no further comments and I therefore recommend publication of this article.