The manuscript presents an effort to characterize the physical drivers of marine heatwaves (MHWs) using the ECCO state estimate framework. The authors should be commended for undertaking the challenge of quantifying the upper ocean heat budget globally and attributing MHWs to specific dynamical processes. However, there are several important concerns that limit confidence in the presented results. First, there is a notable mismatch between the ECCO product and observations, raising questions about the representativeness and reliability of the diagnosed drivers. Second, there are questions regarding the heat budget closure and identification of dominant processes, complicating the interpretation of the contributions from different terms. These issues, along with several concerns related to the interpretation and presentation of results, warrant further consideration and clarification.

Major concerns:

1. Comparisons between ECCO and OISST reveal clear mismatches in the number, duration, and intensity of MHWs globally (Figures 3–5). This raises the question: how representative are MHWs in ECCO of those in the real ocean? The title of the manuscript is ambitious, yet the results are evidently dependent on the fidelity of the ECCO product.
2. The calculation of heat budget terms, which is key to the manuscript, is not clearly described. Offline calculations of the heat budget often fail to close due to the use of temporally averaged velocity and temperature fields, which do not account for nonlinear covariance terms like ∇∙(θ′T′). Depending on the region, these covariance terms can be significant. The authors do not report how well the budget closes or the magnitude of the residual term, which raises questions about the accuracy of the diagnosed contributions from different processes.
3. The partitioning of contributions into advective and diffusive terms is inherently tied to the model’s resolution. Coarse-resolution products may misattribute unresolved advective processes to subgrid-scale diffusion. Without acknowledging this limitation, the attribution results risk being potentially misleading.
4. The identification of leading terms is based on predefined thresholds (30%, 16%) that are not rationalized. The conclusions drawn in Sections 4.2 and 4.3 are thus sensitive to these subjective choices, which undermines their robustness.

Other comments:

Page 7, The claim that "ECCO provides an overall good representation..." is somewhat overstated. ECCO underestimates the magnitude of the trend in the Northwest Pacific, Northwest Atlantic, and Southwest Atlantic. The spatial patterns in MHW metrics also show clear mismatches in Figures 3–6.

Page 9, The statement that "intensity is slightly underestimated..." is problematic. ECCO upper ocean heat content and OISST surface temperatures are not directly comparable.

Page 10, “In these regions” should be “in other regions”?

Page 11, “Figure 8a d” should be “Figure 7a d”.

Page 11, “consistent with the onset phase being on average longer than the decline phase (Figure 4).” Why longer duration means overall pattern being similar to that of onset, especially the counter-argument is made later in the same paragraph? Besides, Figure 4 doesn’t show onset vs decline.

Page 14, “How often each budget terms is "the" leading term vs "a" leading term (excluding overlap with "the") across NEP, SWP, and TASMAN regions”. The definition of “the” and “a” leading term needs to be clearly explained.

Pages 15-18, Figures 11-13: 1. the percentage contribution should be based on the total temperature change, i.e., , where  is the temperature change during onset or decline, so the contribution from one term can exceed 100%, if there is one negative term or more. 2. the numbers are only meaningful if the heat budget is closed, that is,  equals . Otherwise, if there are residuals in the budget, these numbers are less meaningful.

Pages 15-18: all the numbers depend on how the percentage of the heat budget terms are calculated, and the threshold used for ranking contributors.

The authors address the fundamental mechanisms driving surface marine heat waves in the global ocean using a combination of observational products and data assimilative model state estimates. The introduction does an excellent job of motivating the work. The methods and budget analyses are clearly articulated. They identify the key regions where atmospheric drivers and oceanic dynamics dominate the generation and decay of MHWs. They have already astutely responded to a bevy of comments from another referee and have indicated their responses to these criticisms, which are well supported in my opinion.  Their conclusions are justified and substantiated by their analyses. The paper is eminently suitable for publication.  

Main comments:

This study presents a comprehensive analysis of the leading dynamical processes of global MHWs during the onset and decline periods, based on ECCO and OISST datasets. While the paper is generally clear and the results could contribute to our understanding of MHW evolution, significant shortcomings preclude recommending it for publication in its current form.

Firstly, the authors used daily and monthly datasets to calculate MHWs after removing linear trends. The spatial patterns of the MHW metrics differ significantly between the OISST and ECCO datasets (see Figures 3 and 4). This suggests that ECCO does not capture MHW characteristics in the same way as observations. Therefore, I suggest that the authors use more ocean reanalysis datasets, such as GLORYS or BRAN.

Secondly, in the section discussing the mechanism of global MHWs, the authors conducted a heat budget analysis, but did not present the residual term. Does this mean that ECCO can perform a close heat budget analysis?

Finally, I agree with RC1 that the identification of leading terms is based on predefined thresholds (30%, 16%) that are not explained. The authors should demonstrate this more clearly.