Trends in Annual, Seasonal, and Daily Temperature and Its Relation to Climate Change in Puerto Rico

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Small island studies such as these are necessary to support appropriate climate actions and ensure small island climate resilience.

The following are suggestions for improvement of the manuscript.

1. Conduct a thorough literature review and include the relevant studies in the introduction.
2. In the results and discussion section, (i) convert to SI units, (ii) reword captions to describe the figures. Figures must stand on their own, (iii) provide explanations for the legends, (iv) Figures are not mentioned in the body of the text such as Figures 3, 4 & 6, (iv) comparison of your results with similar studies is lacking therefore has to be included in the discussion, there are similar studies for Puerto Rico as well, (v) 
3. Certain statements require references and support for such statements such as lines 33, 77, 118, 200
4. Line 122 requires a description and a statement giving the standard.
5. Relocate the limitations from the conclusions to a separate subsection.
6. Remove strong statements as they are not justified.
7. The conclusion should have the main message(s) provided by your study. Rewrite the conclusion accordingly.

Author Response

Please find the responses to reviewer 1 attached. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript, “Trends in Annual, Seasonal and Daily Temperature and its relation to Climate Change in Puerto Rico” examines the recent changes in annual, seasonal and daily temperature in Puerto Rico and identify their key climate predictors. The methodology is reasonable and well-described, and the paper is well-written. However, major modifications are required before publication.

Major Comments:

1. The stepwise regression approach evaluates correlations but does not establish causation. For example, in L393-395: “For TAVG, the most frequently selected predictors were SST, SAT, and CO₂, appearing in multiple models and indicating their strong influence on annual mean temperatures.”. The high correlation between SST, SAT, CO₂, and station temperatures does not necessarily imply that regional SST or SAT changes are causing the observed changes at the stations. It is possible that other factors drive overall temperature variations in the region, resulting in these high correlations.

2.   The description of temperature trend changes is lengthy and repetitive. For instance, L221-245 describe annual temperature trends: the first paragraph states the overall trend, while the second paragraph details the specific numerical changes. I suggest combining these two paragraphs into one and removing repetitive content to enhance conciseness. There are inconsistencies between the figure citations in the text and the actual figure numbering. A thorough proofreading is recommended to ensure consistency between the text and figures.

Minor Comments:

1. L14: "over recent times" → "over the recent decades"

2. L24: "driving both mean temperature trends…" → "for both mean temperature trends…"

3. L85: 1.5°F (~0.8°C)

4. L87-88: What does it mean “suggesting that warming began early in the observational record”？

5. L178: Revise to "at annual and seasonal scales"

6. L184: "best climate predictors for TAVG…"

7. L200 and L505: Revise to "64-70°W"

8. Table 1: Correct "sight" to "sites"

Author Response

Please find the responses to reviewer 2 attached. 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript entitled “Trends in Annual…” authored by Hernandez Ayala et al. proposes a description of air temperature trends in Puerto Rico based on 12 (relatively) long weather data series, which is enforced by a regression analysis aimed to identify the drivers supporting the observed trends. The study could have some local/regional relevance, but it is built on a manner that diminishes its potential international visibility. Basically, the trend analysis is correct, but classical, while the input data used in the proposed stepwise regression seem to be randomly selected. These features of the paper needs to be improved. Also, the paper is to close to previous authors work, and therefore, it is needed to be underline what the current study brings new in the field. As well, the discussion part is very weak, with almost no comparison with other similar works. In this line, the relevance of the study could be emphasized by comparing Puerto Rico with other similar tropical regions. Additionally, I have listed other comments below that need to be solved when revising the manuscript.

1. The authors should explain what the study brings new when compared with the literature in the field: https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=temperature+trends+in+puerto+rico&btnG=
2. Each of the 12 weather stations should be presented briefly in their physico-geographical characteristics. Altitude, local environment, land use and other similar features could be helpful in better understanding the results.
3. Please follow the journal requirements related to units measures that should be used in the paper. You use °F which is not common in international scientific journals.
4. The Introduction is too long. It could be easily subject to shortening.
5. Section 3.1 should not begin with figure, but with text. Also for the other sections.
6. The values of Adj R² in Table 1-4 should be converted in maps which would be easier to understand and to interpret.

 

L95: Which infrastructure?

L118: 15% of missing data represents a consistent amount. Please explain how you gap filled these missing data in final series?

L121: The NOAA acronym should be delivered somewhere before in the article.

In Figure 1, please use a classical and standardized color scale for digital elevation model (commonly from green for lowlands to brown for mountain tops). Also, please organize better the figure. The color and map scale could be inside a single plot with the map.

L136-151: Extensive explanations on Mann-Kendal test are not needed. This test is common in trend analysis.

L198: CO₂ instead of CO2.

L196-198: The selection of the parameters looks random.

L370: Why this is “interestingly”? Some important explanations have to be given exactly where they have been observed.

Author Response

Please find the responses to reviewer 3 attached. 

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

GENERAL COMMENTS

The study analyses observations of daily mean, maximum and minimum temperature at 12 stations in Puerto Rico in the years 1970-2024. The significance of trends is assessed for the annual and seasonal means of all three parameters, as well as for the frequency of their high extremes, exceeding the 90^th, 95^th or 99^th percentile. In addition, stepwise forward multiple regression is used to identify the large-scale predictors that best explain the interannual variations in all these parameters. The results reveal a widespread warming in all the studied temperature indicators, though with some local exceptions for some of them. The regression analysis shows, perhaps unsurprisingly, that the regional mean surface air temperature, sea surface temperature and the atmospheric carbon dioxide concentration are important for temperature variability even on the local (station) scale.

Although there are earlier studies of temperature variability and trends in Puerto Rico, most of them have focused on the capital (San Juan) area of the island and/or included a smaller number of parameters. Therefore, this comprehensive analysis provides valuable new information on temperature trends and variability in Puerto Rico. Nevertheless, the manuscript still requires several changes to serve its readers in a satisfactory manner.

First, the presentation of the trend analysis results is far from optimal. The maps in Figures 2-7 only show the Mann-Kendall Tau coefficients (a form or rank correlation between year and the studied parameter), apparently with the motivation that Tau directly determines the statistical significance of the trends. However, the magnitude of the trends is not shown at all, even though this would be more important information than their statistical significance. Curiously, even the statistical significance of the trends is not explicitly reported, and the reader is apparently assumed to infer this indirectly from the Tau values.

My primary recommendation is to replace the Tau values in the maps of Figs. 2-7 with the corresponding trend slopes, calculated either by the least squares of Sen’s method. The statistical significance (or lack thereof) of the trends could be indicated for example by using solid (dashed) outlines for the station circles where the trends are (are not) statistically significant.

Second, even though Fahrenheit may be the default temperature unit in Puerto Rico, it is non-standard in an international context. I have nearly never encountered the use of Fahrenheit in scientific articles, including the journals of the American Meteorological Society, although Fahrenheit is also prevalent in everyday use in the United States. Therefore, I would strongly recommend that you express the trends in Celsius or Kelvin instead of Fahrenheit. At the very least, add the corresponding values in Celsius or Kelvin to parentheses if you think that the local needs in Puerto Rico make it necessary to report the trends on the Fahrenheit scale.

Third, although the stepwise multiple linear regression analysis is innovative, the interpretation of its results may be severely complicated by the mutual dependence of the predictors. For example, the interannual variations of surface air temperature and sea surface temperature are probably strongly correlated, and they both are expected to be correlated with the carbon dioxide concentration because of the warming trend. I am not asking you to remove this analysis but adding a table that reports the pairwise correlations of the predictors (at least, between their annual mean values) would give important background information for the interpretation of the results.

DETAILED COMMENTS

1. L22-23. regional mean surface air temperature? The correlation between the local site-specific surface air temperature with itself is always one.
2. L83 and later. For temperature changes given in Fahrenheit, please also indicate the corresponding change in Celsius or Kelvin, which are far more universally understood units. For example: 2.24 F (~1.24 C)
3. L110-111. temperature trends ... and their relation
4. Section 2.2 and Figures 2-7. The Mann-Kendall Tau statistics, which is just a rank correlation coefficient, tells nothing about the absolute magnitude of the trends, which is at least as important as their statistical significance. For example, if one time series has larger interannual variability than another, it may have a larger warming trend even if its Tau value is smaller. Therefore, I would strongly urge you to include information on trend magnitude in Figures 2-7 (see also the General comments).
5. L146-147. Is this p-value for a one- or two-sided statistical test? What is the corresponding value of the Mann-Kendall Tau coefficient?
6. How did you determine these percentile thresholds? Are they for the whole 54-year period (when, for example, the 90th percentile would be the 0.1 * 365 * 54 = 1971st highest value, assuming no missing data)?
7. L210-212. Even the adjusted R-squared values will give an optimistic estimate of the skill of the regression models, due to a selection bias that arises when choosing the best set of predictors among the six candidates. Reference: DelSole, T., and J. Shukla, 2009: Artificial Skill due to Predictor Screening. J. Climate, 22, 331–345.
8. Figures 2a-b. Please use colours with larger mutual contrasts, to allow for an easier identification between the three stations.
9. How large are the warming trends in absolute units?
10. L235-238. Please report the results in a way that is comparable between the three parameters, e.g. the means for the 12 stations.
11. L265-270. Same as the previous comment.
12. L277-278. The last three maps (d-f) are for SON.
13. Tmin at Juncos (a) and Tmax at Ponce (b)
14. L322 and 324. On the average, there should be 36.5 days per year above the 90th percentile. Based on your numbers, the 54-year average for Arecibo is 31.9, whereas that for TAVG at Juncos is 39.6. Is this variation just a rounding artefact, due to the practice of reporting the temperatures in full Fahrenheit degrees?
15. Two decimals is too much.
16. L331-332. Is this interpretation true? Based on Fig. 5c, the increase in the TMIN 90th percentile events in San Juan does not appear exceptional, as there are at least four stations with larger Tau coefficients.
17. L342-343. This may be a premature interpretation. Because the total number of days above the 95th percentile is only half of that above the 90th percentile, the trends for the 95th percentile exceedances are inherently noisier than those for the 90th percentile exceedances.
18. L347-348. from the 1970-1997 to the 1998-2024 period.
19. L348-349. L322 reported nearly the same numbers for the 90th percentile. Are the 90th and 95th percentile identical at this station?
20. Tmax in Aquirre (a) and Ponce (b)
21. Tables 1-4. Please always use the same number of decimals (two?).
22. Heading of Table 1, L2. By site?
23. Table 1. L15-16 are not in decreasing order.
24. Add a cross-correlation matrix reporting the pairwise correlations of the six predictors (cf. Ceneral comments).
25. L407-408. Move the reference to Table 2 to the end of the sentence.
26. The "Conclusions" section should not include new results that were not reported earlier in the manuscript. So please report this result earlier in the relevant place or omit it.

Author Response

Please find the responses to reviewer 4 attached. 

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The revised manuscript titled “Trends in Annual, Seasonal and Daily Temperature and its Relation to Climate Change in Puerto Rico” has strengthened the scientific rigor and clarity of the study. The authors have addressed the previously raised concerns. However, minor modifications are still required before publication.

1. Lines 380-383: Please clarify the statement “the forward stepwise approach reduces the likelihood of including redundant independent variables.” What is the underlying rationale for this claim?
2. Table 2: The word “sight” should be corrected to “site”.

Author Response

Comments 1: The revised manuscript titled “Trends in Annual, Seasonal and Daily Temperature and its Relation to Climate Change in Puerto Rico” has strengthened the scientific rigor and clarity of the study. The authors have addressed the previously raised concerns. However, minor modifications are still required before publication.

Response 1: We appreciate the reviewer’s suggestions in the previous revision process; they have made this a stronger and better manuscript. We will now address the minor comments in this round.

Comments 2: Lines 380-383: Please clarify the statement “the forward stepwise approach reduces the likelihood of including redundant independent variables.” What is the underlying rationale for this claim?

Response 2: Sure. The forward stepwise approach reduces the likelihood of including redundant independent variables because it adds predictors incrementally based on their unique contribution to explaining variance, excluding those that offer little additional explanatory power. We edited the sentence to add this rationale.

Comments 3: Table 2: The word “sight” should be corrected to “site”.

Response 3: Thanks for pointing that out; it has been corrected.

Reviewer 3 Report

Comments and Suggestions for Authors

Symbol of Fahrenheit degree is not needed in the manuscript.

The changes indicated for Figure 1 as suggested in the revision is not perceivable on the figure. 

Author Response

Comments 1: Symbol of Fahrenheit degree is not needed in the manuscript.

Response 1: Thanks for the suggestion, but we will keep it throughout the manuscript for the benefit to the reader.

Comments 2: The changes indicated for Figure 1, as suggested in the revision, are not perceivable in the figure. 

Response 2: Thanks for pointing that out, the latest version of the manuscript now shows the suggested changes to Figure 1.

Reviewer 4 Report

Comments and Suggestions for Authors

GENERAL COMMENTS

The authors have satisfactorily addressed all the main concerns that I reported in my previous review. Therefore, I am now pleased to recommend publication after some minor text edits.

DETAILED COMMENTS

1. L33-34. consecutive extremely hot days
2. DJF, MAM, JJA and SON
3. L243-244, L273-274 and L302-303. temperature change from 1970-1997 to 1998-2024.
4. the weakest?
5. Check the figure numbers. Figure 2c represents annual mean trends in TAVG, not DJF trends in TMAX.
6. L277-278. changes of 1.5 C (2.7 F) or higher between the 1970-1997 and 1998-2024 periods and yearly Sen's slope increases .... Sen's slopes represent the whole 1970-2024 period, not the difference between its two halves.
7. L287-288. SON TAVG (d) ... SON TMIN (e), ...... SON TMAX (f)
8. the strongest?
9. were overall weaker
10. five, according to Figure 6f.
11. L304-305. increase ... shows
12. L315-316. positive trends in TAVG and TMIN
13. 0.09 or 0.9?
14. TAVG and TMIN? In fact, in both JJA and SON, all 12 stations have increasing trends in both TAVG and TMIN (Figure 6).
15. L468-469. Delete "from the".

Author Response

Comments 1: L33-34. consecutive extremely hot days

Response 1: It has been edited in the latest version. 

Comments 2: DJF, MAM, JJA, and SON

Response 2: It has been edited in the latest version.

Comments 3: L243-244, L273-274 and L302-303. temperature change from 1970-1997 to 1998-2024.

Response 3: It has been edited in the latest version.

Comments 4: the weakest?

Response 4: It has been edited in the latest version.

Comments 5: Check the figure numbers. Figure 2c represents annual mean trends in TAVG, not DJF trends in TMAX.

Response 5: Thanks for pointing that out, it has been fixed.

Comments 6: L277-278. changes of 1.5 C (2.7 F) or higher between the 1970-1997 and 1998-2024 periods and yearly Sen's slope increases .... Sen's slopes represent the whole 1970-2024 period, not the difference between its two halves.

Response 6: Noted, and it has been corrected.

Comments 7: L287-288. SON TAVG (d) ... SON TMIN (e), ...... SON TMAX (f)

Response 7: It has been edited in the latest version.

Comments 8: the strongest?

Response 8: It has been edited in the latest version.

Comments 9: Were overall weaker. 

Response 9: It has been edited in the latest version.

Comments 10: Five, according to Figure 6f.

Response 10: It has been edited in the latest version.

Comments 11: L304-305. increase ... shows

Response 11: It has been edited in the latest version.

Comments 12: L315-316. Positive trends in TAVG and TMIN

Response 12: It has been edited in the latest version.

Comments 13: 0.09 or 0.9?

Response 13: It was 0.9, thanks for pointing that out. It has been corrected now.

Comment 14: TAVG and TMIN? In fact, in both JJA and SON, all 12 stations have increasing trends in both TAVG and TMIN (Figure 6).

Response 14: Thanks for the observation. The sentence has been modified.

Comments 15: L468-469. Delete "from the".

Response 15: It has been edited in the latest version.