Primary production responses to extreme changes in North American Monsoon precipitation move up an elevation gradient through time by Munson, Seth; Bradford, John; Butterfield, Bradley; Gremer, Jennifer
This is an interesting and relevant study. From 2016-2020 the authors experimentally manipulated warm-season precipitation in five dryland communities along an elevation gradient (ca. elevation from 1550-2600 m) in Arizona, southwestern USA. They either excluded ca. 50% of the natural Monsoon summer precipitations or added ca. 30% more water to ambient summer rainfalls, simulating extreme drought, heavy rainfall and control treatments. The study analyzed how primary production responded to these treatments across main plant functional groups (C4 and C3 perennial grasses, annuals, forbs and shrubs), time (2016-2020) and communities or sites along an elevation and climatic conditions gradient.
The experimental setup is impressive and I some of their approximations to obtain reliable response data were smart (e.g., how they managed to perform non-destructive estimations of ANPP in the experimental plots or analyzing sensitivity and responsive ratios to water treatments compared to controls). Congratulations. Moreover, despite having large amounts of results and some complex interactions among potential factors affecting ANPP productivity, the authors were able to highlight the most salient results and built a clear and relevant story.
I just have one general concern related with the statistical analyses. The authors analyzed ANPP in 1m2 plots that contain plant communities where different plant functional groups coexist. There was a separate analysis of the responses of each (but coexisting) functional group to precipitation treatments, sites and years since the onset of the experiment. Although I understand the reasons for dividing the analyses (potentially clearer results), responses of coexisting functional groups may not be independent from responses of and interactions with the other functional groups. This potential lack of independence has both an ecological and statistical modeling nature and should be incorporated in, at least, some preliminary analyses. I am not declaring/opining the analysis are wrong, but it would be desirable that apart from performing single group response analyses, the authors present either multivariate analyses including responses of all groups or considering the functional group an additional fixed factor in the analyses. It would be desirable that this, at least, be performed for ANPP data. If these analyses support responses at functional group level, then they can be presented as supporting information.
The rest of my comments are minor. I hope they will help to clarify some parts of this interesting paper.

Specific comments
Title. I am not convinced that the title reflects the main or most relevant results of your study. You showed that the effects of monsoon precipitation extremes on ANPP varied by plant functional type and, yes, across sites/along the elevation gradient and that some functional types require some years to be responsive to treatments. Overall, ongoing climate change may change community composition and productivity along this gradient. These results are not well represented by current title that may be too general.
Keywords. The studied communities are drylands, aren’t they (see abstract)? Shouldn’t this term be included here or in the title?
Intro. The first two pages seem to be more related with general patterns of dryland ANPP to different abiotic factors and climate change. The authors then focus the introduction on trends in south-west USA drylands.
Some sentences require appropriate citations (e.g., L. 64-66, 68-69) and, overall, citation in this first part of the introduction can improve. I was particularly surprised that you did not cite any of the seminal works of the special issue on drylands in Oecologia 2004, for example, or even paradigmatic works on dryland ANPP and functioning like Noy-Meir 1973, AREES (but include cites from 1965).
Material and Methods
Lines 210-213. As plots contain some or all of the functional groups distinguished, should not be your analyses accompanied (or controlled) with some multivariate analyses including as response variable ANPP of all groups present at each plot (separated by functional group) or include an additional fixed factor in the analyses – functional group? Plant recruitment, productivity and population fate may be affected by interactions with (and productivity of) accompanying functional groups within a plot (e.g., see your discussion L. 368-370). Moreover, productivity response of each functional group ANPP is not independent (statistically speaking) from responses of coexisting groups, even whether they did or did not share temporal or spatial resource acquisitions niches, are they?
L. 213. What correlation structure (https://stat.ethz.ch/R-manual/R-devel/library/nlme/html/corClasses.html) did you include in the RM-ANOVA? Or did you include plot as a random factor?
L. 222-225 What variable/s were the covariates in the ANCOVA models? Was it warm-season precipitation (working here as the independent variable/ regressor) and site or treatment were the fixed factors?
L 226-228 Very clever.
L. 226 Just a suggestion. If you use typical indices that compare control and treatment responses (such as effect size indices used in meta-analysis, LnRR or similar in interaction studies), this response variable might have a normal distribution. Corrections (modeling) for heteroscedasticity are nowadays available in LMM (e.g., varident, varexp, https://stat.ethz.ch/R-manual/R-devel/library/nlme/html/varClasses.html). Anyhow, plain ratios are not usually the best functions to be included as responses variables in LMM statistical models. It is just a comment for future work. Your results on this regard were quite clear, although see my comments on including the statistical results for these regressions.
L. 232 Are associate errors raw (a-priori) errors? Standard errors of transformed and modeled variables cannot be back-transformed.
Statistical analyses. What statistical program did you use?

Results
Please check Table 1 caption and variable names in columns. Either there is no WWP or there is no MWP, and it is not clear what MSP is (MAP?).
L. 252-253. I do not understand this sentence. Depending on the plant functional group, you have quite different responses (Table 2). Site seems the main factor here (note F maximum values irrespective of the functional group analyzed), but in almost all cases you have significant two-way statistical interactions (in importance/effect order): site x year; then site x treatment (except for shrubs, only responsive to treatment and site); treatment x year for some groups, and only one single 3-way interaction site x treatment x year for perennial forbs). Thus, this sentence seems to be too vague.
All regression results (panels on the right side of Figs. 3, 5 and Figs 4,6). It would be useful to include the statistical results of such analyses (either in appendices or Figs.) and not just including some general trends in the Figures captions. The selection of dashed lines for non-significant trends and (presumably) sites in Figs. 4-6 is not fortunate.... If trends were not significant, you could eliminate those lines in order to avoid confusion.
Lines 265-268, Fig. 4 and similar ones. Did you analyze the differences “in sensitivity” across treatments x site via regression analyses with dummies? How did you analyze the differences across slopes (L. 267)?
Abstract (L.37) and the lack of significant results for shrubs (L. 293-294) and discussion on this regard. Table 2 shows a significant effect of treatment on shrubs ANPP (F=3.4, p=0.04) irrespective of site, year or interactions with these other fixed factors. Is it not worth mentioning what the differences across treatments were?

Discussion (and conclusions).
L. 407-419. Results of C3-grasses to either drought or water addition treatments are somehow unexpected, aren’t they? Could you please expand on the possible mechanisms behind the C3 responsiveness to water additions/shortage during the summer season?
L441-443 and 450-451. Tone down this conclusion. You did not demonstrate that these communities are prone to (alien) invasion during large or extreme drought events.

The authors have overall addressed all my concerns. Thank you.

Still there are some details that can improve readability of graphs. I would suggest deleting trends that are not significant (e.g. Fig. 5) and select other colors or line frames to distinguish between water and drought treatments in right panels in Figs. 3, 4 and 6 (particularly the later two).