General comments:

Anderegg et al. have investigated drought-induced stress, insects, and fire in the US based on empirical data. They used statistical models to predict disturbance risk for each agent until the end of the 21st century. According to their analysis, the risk of fire might increase the most (by a factor of 4) among all investigated disturbance agents. The combination of multiple sources is quite impressive and must have been a major effort. The statistical analysis is sound, although I assume predictions went far outside the observed data room which should be clarified (e.g., by supplementary figures). The text has been written overall well. I have a few general concerns that justify a moderate revision of the study: 1. The Introduction focuses strongly on carbon, but the study is not about carbon at all. I wonder if it either needs a somewhat different focus of the introduction or, even better, an additional analysis that compares disturbance with a biomass distribution map. 2. The methods section describes the modeling approach in all details, but a description of the data can only be found in the supplement. I know Ecology Letters is restrictive in the manuscript length, but I personally would have balanced data and methods some more, writing a summary about both in the main text and have a supplement explaining the details. But maybe other people see this different and I am too picky about it. 3. The results are mixed with discussion. I wonder if it is possible in Ecology Letters to merge the results and discussion sections avoiding the dichotomy entirely? 4. There is no figure provided that compares the future risk of disturbance agents. A comparison is difficult, because disturbance agents are on different scales in all figures.

Specific comments:

Introduction: The introduction points out the importance of future disturbance and climate stress which is good, I just wonder why the study does not predict future carbon loss, if the introduction makes such a strong point about it.
L 61 Somehow unclear whether treatment and baseline should be measured over the same time period
L 74 At this point it remains unclear whether climate stress and drought is the same. What is meant by climate stress should be defined early on.

Methods: The data and methods are well-suited for addressing the study objectives. A lot of different data sources were considered and the statistical methods are complex, although not overly complicated in each individual step. Here and there I have a question to better understand the approach. The data description is extremely short, while the model description is extensive.
L 106-108 This is a bit confusing. I don’t know if this is well understandable without reading the supplement.
L 109 I assume SI should not be abbreviated the first time you are referring to the supplement to clarify the information can be found there.
L 118 Based on the previous sentence, this sounds as if time-invariant climate variables account for heterogeneity in vegetation
L 123 Okay, but why this threshold?
L 143-144 Why not using an ordinary drought index that considers both temperature and precipitation as in the climate stress model below?
L 171 Fig or Figure (should be consistent)
L 215 I assume it should read ‘stats’ instead of ‘glm’

Results: I like the result figures in the main text and supplement and I also think they are very useful for teaching besides their scientific value. They nicely visualize disturbance hotspots and temporal developments. I only wonder if it is possible to add at least one figure or table where all disturbance agents are scaled in the same way to visualize the difference among agents? Moreover, I don’t want to be very picky about it, but I have barely seen so many citations in a results section. This already adds discussion to the results. Maybe results and discussion sections can be merged into one, if the journal allows it. Apart from that, I found the results section compact and informative.
L 319 predominantly drought stress – what else? Heat?

Discussion: The discussion does not sound groundbreaking to me, but it is a nice read. Yet, I don’t have any particular suggestions to improve the section’s content.
L 386-387 Maybe not in CMIP6 projections, but there are examples of insect-driven mortality modeled based on first principles in process-based modeling studies (e.g., https://link.springer.com/article/10.1007/s10980-016-0396-4).
L 390-391 dampening factors should be mentioned

I have reviewed all comments and edits of the revision. The authors have carefully revised their manuscript and addressed all my points satisfactorily. The extension of the supplementary materials is very useful to get a better idea where carbon stocks are threatened by disturbance. At the same time, the authors do not overemphasize carbon in their manuscript anymore. Information about data and methods is better balanced after the revision, and the results and discussion sections are well distinguished now. The manuscript reads well throughout. I do not have any further comments and congratulate the authors to their study.

Best regards,
Dominik Thom