Review comments on “Sea-surface temperature variability and climate drivers in Cuba’s Jardines de la Reina National Park (2003–2022)” by Castillo-Alvarez et al.

The manuscript analyzes the seasonal-to-decadal sea-surface temperature (SST) variability and its climate drivers within Cuba's Jardines de la Reina National Park (JRNP) from 2003 to 2022, a region facing increased marine heatwaves (MHWs) and rapid warming. The seasonal cycle is primarily governed by net air-sea heat exchange. Superimposed on this is a warming trend, strongest in winter and transition months, alongside a notable step-like shift towards a persistently warmer state between 2011 and 2013. MHWs intensified during the second decade, with the mean maximum intensity being higher inside the gulfs, while upper categories occurred more frequently offshore. Empirical Orthogonal Function analysis reveals that the dominant, basin-wide warming mode (EOF1) is linked to interannual variability of ENSO/WHWP and latent-heat flux, and to low-frequency variations of the NAO; a secondary shelf-offshore dipole mode (EOF2) is tied to the Tropical North Atlantic (TNA) index. The findings suggest that monitoring net heat flux, mixed-layer depth, and large-scale indices like ENSO/WHWP and NAO can aid in providing early warnings for MHW risk and guiding conservation efforts in the JRNP. This is a well-written manuscript with well-prepared graphs to support the research. 

Below are my major and minor comments

Major comments

1. Data Uncertainty and Consistency: The analysis relies on data from diverse sources (remote sensing, reanalysis datasets), which are known to have significant uncertainties/errors, particularly in coastal regions. Crucially, the lack of in situ measurements for validation makes an uncertainties analysis vital. A more detailed comparison of data consistency is necessary. For example, while a correlation between remote sensing Sea Surface Temperature (SST) and reanalysis SST (ERA5, GLORYS12) is mentioned, more in-depth quantitative details of this comparison are required to address the lack of in situ validation.
2. Sensitivity of Box Selection: The selection of the two analysis boxes in the GAM and CS appears arbitrary. A sensitivity analysis should be performed to evaluate how the results change based on the location and size selection of these analysis boxes.
3. Repetitiveness and Discussion Depth: Sections 4.1 and 4.2 largely reiterate information already presented in Section 3. These sections should either be combined, or the discussion should be substantially deepened. For instance, the discussion could benefit from comparing the findings to previous studies or similar studies conducted in other regions.
4. Mechanism for Regime Shift: The observed regime shift between 2011 and 2013 is a significant and interesting finding. However, the current paper does not present a mechanism to explain this shift. A more in-depth discussion is needed to explore the potential causes and implications of this regime shift, which would substantially enhance the paper's contribution.
5. Climatology section for MHW: The climatology used for Sea Surface Temperature (SST) analysis, which spans approximately 20 years, deviates from the standard practice of using a 30-year climate normal in Marine Heat Wave (MHW) studies. This choice of a shorter baseline may hinder the comparability of the results with other studies that adhere to the climate normal. It is suggested that the authors either adopt a 30-year climate normal from alternative datasets or, at minimum, quantify the influence of the chosen baseline period on the determination of MHWs.

Minor Comments

• Line 46: A period is required after the word "progresses."
• Lines 164-165: The Access statement should be moved to the Acknowledgement section.
• Lines 172-173: This information is redundant as it has already been presented in Section 2.2.
• Line 179: More detail is needed regarding the determination of the effective sampling size and effective degree of freedom.
• Lines 197-199: In the equation, the terms Delta_h, u_{-h}, and w_{-h} must be defined.
• Figure 10c: The line is labeled as PC-1, but the legend incorrectly mentions PC-2. PC-2 appears to be an error.

This work examines the physical drivers of sea surface temperature (SST) trends and variability in an area encompassing the Gulf of Ana Maria, the Jardines de la Reina National Park, and the Caribbean Sea off the southern coast of Cuba. The authors find differences between the shallow gulf waters and the much deeper Caribbean Sea, which appears to be primarily a result of the mixed layer depth. They correlate modes of local SST variability with large-scale climate modes, finding (potentially) significant correlations with the North Atlantic Oscillation and the El Niño-Southern Oscillation. The authors note that understanding the mechanisms that cause SST trends and anomalies (and hence marine heatwaves) can be used to better predict the occurrence of marine heatwaves for coral ecosystems in this region.

The paper is well written and the analysis is presented in a thorough manner such that the authors’ conclusions are supported by the data (with a few relatively minor exceptions; see below). The analysis is well done, although perhaps not particularly novel and the conclusions reached may not be broadly applicable outside of the study region. Still, I would recommend the paper for publication if the comments below can be addressed in a satisfactory manner.

Line 246-247: The results of this sensitivity test should be shown in supplementary material.

Lines 270-271: It would be nice to see a supplementary figure with the cross-correlation vs. lag to support this sentence.

Lines 298-299 and Fig. 6: In the text the significance test is written as a p-value (“p<0.05”) and in the figure it is as the equivalent percentage level (“95% level”). I recommend using the percentage level for both as it seems to be more common in our field.

Lines 308-313 and Fig. 7: Has the determination of the transition points of the piecewise/step fits been done empirically? If so, by what method? See Reeves, et al. (2007). A Review and Comparison of Changepoint Detection Techniques for Climate Data. Journal of Applied Meteorology and Climatology, 46(6), 900–915. https://doi.org/10.1175/JAM2493.1

Lines 353-366 and Fig. 10: The results of the significance testing of the correlation coefficients appear to be missing despite the mention of the method of significance testing. Additionally, are the effective degrees of freedom calculated before or after low-pass filtering. adjusted for the low-pass filtering? This is especially important for the PC1-NAO correlation because after low-pass filtering both time series have very few degrees of freedom left (perhaps around N=5 judging by the number of peaks and valleys).

Fig. 10c: The text in the correlation box should be “PC1-NAO,” not “PC2-NAO.”

General comment: It is a matter of personal taste, but I think that a more standard first-person voice (e.g., “We diagnosed…”) may improve readability compared with third-person voice (e.g., “The authors diagnosed…”).