Dear Authors,

The reviewers raised several fundamental points, particularly regarding model evaluation and uncertainty, and requested clarification on a few methodological aspects.
The authors' response is convincing; therefore, I suggest proceeding with the revision and reassessment by the reviewers.

Best regards

Dear Authors,
We have received feedback from two out of three reviewers. Unfortunately, one reviewer has not responded despite multiple requests. The reviewers' assessments are contrasting. While both reviewers acknowledge the manuscript's improvement, one is satisfied with the modifications, while the other has more concerns.
I agree that the manuscript has improved significantly since the first submission and is well-written. However, I also agree with Reviewer 2 that further analysis is needed, particularly regarding the idea that a sensitivity analysis of the model runs can help to better understand the role of the different drivers than the machine learning itself. The analysis with the machine learning is interesting in my opinion, but should be expanded to the GFED BA data to understand if CLM can reproduce the relationship between drivers and BA. I see two possible solutions, the first is to run the sensitivity analysis, and the second to apply (for current climate) the machine learning approach to both the model and GFED (see below more details)
After carefully reading the manuscript and reviews, I suggest a substantial revision.
Below are some additional comments.
Line 54: Add a reference for this statement from GFED5.
Line 118: "of key carbonaceous species to provide": Please clarify which species.
Line 130: It is unclear why the authors did not conduct a test involving changes in land cover according to available land use change scenarios. This should be clarified.
Validate the model against burned area. I suggest moving the validation to the results section.
Lines 225–227: Why don't the authors report immediately the average BA calculated over the same time period? This value is reported below, and I would move it up. For the recent time period, it seems that the CLM substantially underestimates the BA (3.9 vs. 5.2 million km²). Therefore, I would not call them aligned, but rather, I would try to explain the differences. The authors attempt to demonstrate regional differences below, but a more in-depth discussion of this bias and the significant seasonal variations (Figure 1) would be beneficial to readers.

Section 2.5: I agree with Reviewer 2 that a sensitivity analysis can be less “black boc”, however I see value in the analysis of the authors. My suggestion is to run the same analysis on GFED as they did on the model. This would allow us to determine whether the model and GFED "observations" share the same climate and vegetation drivers. Variables such as aboveground biomass, climate reanalysis, and surface soil moisture can be used as predictors of GFED biomass (BA) for the current climate period.
Line 512: The statement is correct, considering the CLM model structure. However, considering the model structure, the role of biomass itself can be overestimated, while the fine biomass (deadwood, understory, and fine flammable material) is one of the main drivers, as determined by data analysis. This is also encapsulated in the intermediate fire productivity hypothesis. Analyzing the drivers using GFED can also help determine if the role of total biomass is overestimated in the current analysis.
I suggest splitting the discussion and conclusions.

Best regards

Dear authors,
Thank you for the work on the article and for the new analyses.
There are a few aspects that need improvement.
Figure 8 is not entirely readable.
The references list is not well formatted. Please improve the format.
Best regards