Dear Authors,

I have now received two reviews of your manuscript, both reviewers are positive about your manuscript. I think it is a very interesting study on an important topic and I am recommending minor revisions before publication, many congratulations.

I have read through your responses document and I'm happy with your suggested revisions, so look forward to reading your revised manuscript.

A few minor comments:

- In the abstract, when you say "between 1970 and 2099 for the RCP4.5 and 8.5 scenarios", can you please indicate this is based on a single GCM? I would take this sentence to imply aDGVM was forced by the entire CMIP ensemble, for example. I don't mean this comment in any negative sense to be clear, I'm fine with your experimental set up.
- Also, when you talk about lags in the abstract, could you indicate in brackets a couple of examples of what you mean by lags (causes) to capture interested readers? You currently use the word lag(s) seven times, but you don't really explain what is meant. I note you do mention fire at the end, but I still think it would be useful.
- When you state: "For example, CO2 fertilization effects may be reduced by increased drought due to water limitation effects on plant growth (Temme et al., 2019).". I'm struggling a little with this assertion, firstly because drought is typical short-term for many/most ecosystems, so what evidence is there that it would significantly alter the plants' capacity to exploit a higher CO2 concentration over some multi-year period? Also, if CO2 = greater non-structural carbohydrates or reduces evaporation, both could reduce/delay drought impacts. It is of course also true, that this may make little difference, particularly if the drought is long-lasting, but still, I think more care is needed with this sentence.
- Finally, I wonder whether you'd consider adding further to your discussion. You're under no obligation here, just a thought. I think it would be useful to discuss (briefly) key process lags that the model doesn't capture, but in reality may be important. For example, drought legacy/recovery (I presume it is largely instantaneous in the model - i.e. the model doesn't simulate embolism). You also talk about tree cover decline by the end of the century, I wonder how much acclimation capacity the model has? Can (or does) the model project greater root investment under eCO2? Are greater roots then linked to greater water uptake potential in the model? I guess overall what I'm wondering is how much capacity our models (or your specific model) has to dynamically respond to the projected climate and how much of what we learn is still limited by our model process capacity?

Best wishes,

Martin De Kauwe

Dear Authors,

I'm happy to recommend publication based on your revisions. Thank you also for taking the time to respond to my queries.

Best wishes,

Martin De Kauwe