Next Article in Journal
Spatial Distribution and Management of Trace Elements in Arid Agricultural Systems: A Geostatistical Assessment of the Jordan Valley
Previous Article in Journal
A Study on the Optimization of Ecological Spatial Structure Based on Landscape Risk Assessment: A Case Study of Wensu County, Xinjiang, China
Previous Article in Special Issue
Landscape–Ecological Problems Resulting from Spatial Conflicts of Interest in the Poľana Biosphere Reserve
 
 
Article
Peer-Review Record

Tabonuco and Plantation Forests at Higher Elevations Are More Vulnerable to Hurricane Damage and Slower to Recover in Southeastern Puerto Rico

Land 2025, 14(7), 1324; https://doi.org/10.3390/land14071324
by Michael W. Caslin 1, Madhusudan Katti 1,*, Stacy A. C. Nelson 1 and Thrity Vakil 2
Reviewer 1:
Reviewer 2: Anonymous
Land 2025, 14(7), 1324; https://doi.org/10.3390/land14071324
Submission received: 8 May 2025 / Revised: 17 June 2025 / Accepted: 18 June 2025 / Published: 21 June 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript studies the damages of a large hurricane on forest structure in Puerto Rico based on the 3DVR measurements taken 7 years after the hurricane. Introduction to the background of this study and the two research questions is sufficient. However, the Study Area, Methods, Materials, and Results of this manuscript are not sufficiently described. Particularly, spatial analysis should be applied in order to consider spatial autocorrelation and detect relationships between damages and geographic variables. The current result depends on classifications of damage ranks and independence of sample points. Below are major concerns for this manuscript.

First, details of the tree community compositions, dominant species for canopy, stand density, and disturbance regimes in the study area are not fully described in Section 2. 

Second, the sampling distance is 40 m (Line 130-131), but the photo range of the 3DVR equipment is not provided, which makes it difficult to assess the sampling frequency of this study. 

Third, the authors should try Simpson index and/or effective number of species to repeat the analysis for VHDI (Lines 164-165). 

Fourth, the transformation for the class data to meet the normality requirement can be replaced by the application of generalized linear models, which also allows consideration for spatial autocorrelation among the 75 sample points. 

Fifth, no detail of elevation, slope and aspect data provided. It is crucial to use high-resolution geographic data because the sampling distance is 40 m (the resolution should be 10 m or 3 m ideally). 

Sixth, statistical analyses for spatial processes should be applied to the sample data in order to consider spatial autocorrelation likely occurring to the 75 sample points.

Seventh, the manuscript applies several single factor analyses (like Figures 13-15), but actually a model consisting multiple factors can be applied to analyze how these factors are associated with canopy closure and damage rank.

Eighth, some terms used and arguments made in Discussion are not perfectly appropriate. Lines 239 - 244 should be mentioned in Introduction to justify the study of the damages of this hurricane on Puerto Rico forests. Lines 255-256 need references. Interaction is not considered in the results, but it is mentioned in Lines 261-264. In Lines 270-275, the arguments made by personal observations are not academically sound. In Lines 279-281, the interpretation for the statistical difference based on mean elevation for the two types of sites is not accurate.

Ninth, no references at all in sections 5.3 and 5.4, which significantly reduce the scientific soundness of this manuscript. Please find supporting evidence for these discussions.

Tenth, Conclusion should be revised accordingly after spatial analyses are conducted on these data. The current Conclusion is too strong without sufficient support from the Results of this manuscript.

 

Minor comments:

(1) A histogram or a table of categories is needed to justify the definition of the damage key (Lines 138-140).

(2) The 24 points not sampled should be shown in Figure 4.

(3) The three Figure 7s can be provided in Appendix.

(4) Definition of the 8 categories of ground cover should be clearly defined in 3.3.2

(5) Why are the degrees of freedom for the F test in Figure 11 1 and 68, instead of 1 and 71?

(6) If the t test modified for spatial processes (Dutilleul et al., 1993) is applied to Figure 12, the significance likely disappears.

(7) Figure 15 should be organized in the same layout as Figure 14.

(8) 7 years are not particularly a long-term period (Line 230). The 75 sample points are not large enough for a landscape level study (Line 231). The authors should use the terms more scholarly and specific.

(9) Please provide the coordinates where the three photos in Figure 16 were taken.

(10) Because no assessment was taken before Hurricane Maria, it is not solid to make absolute argument on recovery of canopy in Lines 297-299.

Author Response

Thank you for all your thoughtful and constructive comments.

Comment 1: First, details of the tree community compositions, dominant species for canopy, stand density, and disturbance regimes in the study area are not fully described in Section 2. 

Response: The dominant tree species of the forest and the spacing of the tree plantings have been provided on lines 113-117. The hurricane disturbance regime can be found on lines 117-120. We have also added a sentence about previous land use history of the study area on lines 106-108.  

Comment 2: Second, the sampling distance is 40 m (Line 130-131), but the photo range of the 3DVR equipment is not provided, which makes it difficult to assess the sampling frequency of this study.

Response: The photo range of the Ricoh Theta V 360° camera has been added to line 193-194. Links to the product specifications for the Ricoh Theta V 360° camera and the Oculus Go VR Headset can now be found at the end of the document before the references. Based on your comment, it looks like you may be concerned about the potential for trees and other vegetation overlapping between photos and we have addressed this by adding additional information in subsection 3.1 starting at the end of line 151-153. Basically, the density of vegetation meant that the photos only show plants in about a 10-15m radius around each sampling location. We verified this by carefully examining the trees in adjacent photos to ensure there was no overlap while in the field, and In the few cases where we could see some overlap, we dropped the point. This is explained on lines 141-145.

Comment 3: Third, the authors should try Simpson index and/or effective number of species to repeat the analysis for VHDI (Lines 164-165).

Response: We repeated our analysis using the Simpson Diversity Index for VHDI (lines 179-182)  and found no changes in our results. The new result has been incorporated into the manuscript as figure 7b on page 10. Figure 7 is also new and reorganized. 

Comment 4: Fourth, the transformation for the class data to meet the normality requirement can be replaced by the application of generalized linear models, which also allows consideration for spatial autocorrelation among the 75 sample points. 

Response: We have added more information to both the methods and results section to address this concern about spatial autocorrelation. Overall, the relatively small spatial extent of our study area, and therefore the smaller sample size, limits our ability to conduct multivariate spatial statistical analyses without the risk of overfitting statistical relationships and complicating interpretation of results. We believe our results and interpretations remain valid under these constraints. See details in our response to Comment 6 below.

Comment 5: Fifth, no detail of elevation, slope and aspect data provided. It is crucial to use high-resolution geographic data because the sampling distance is 40 m (the resolution should be 10 m or 3 m ideally). 

Response: We have added a new section in the methodology to address this. This new Section 3.4 can be found on lines 214-219.

Comment 6: Sixth, statistical analyses for spatial processes should be applied to the sample data in order to consider spatial autocorrelation likely occurring to the 75 sample points.

Response: We have run spatial analyses using the Spatial Autocorrelation tool in ArcGIS Pro to our vegetation variables to compute Global Moran’s I, which is now reported in Table 1 on page 8. Analysis indicated no spatial clustering in VHDI and Grass/Herbaceous groundcover. However, for % canopy closure, we found it to have statistically significant clustering. We conducted further analysis within ArcGIS Pro utilizing the Local Bivariate Relationships tool and present this in the completely rewritten statistical analysis section 3.5 (lines 221-239), in results section 4.1 (lines 241-251), and in the discussion (lines 312-326). We have also added a new Figure 6 with a map illustrating the spatial variation in the elevation–canopy closure relationship.

Comment 7: Seventh, the manuscript applies several single factor analyses (like Figures 13-15), but actually a model consisting multiple factors can be applied to analyze how these factors are associated with canopy closure and damage rank.

Response: We think that multivariate spatial analyses (like MGWR) are not warranted or feasible given the ecological nature of the variables and the study area’s relatively small spatial extent, where some clustering is not unexpected.

As for spatial autocorrelation and multivariate spatial modeling, we considered the application of GWR. However, GWR and its multiscale variants require strong assumptions about variable stationarity and data density. Given our study site’s limited spatial extent and relatively uniform topographic conditions, we concluded that applying GWR or MGWR would likely introduce more complexity than analytical benefit, and may risk overfitting due to insufficient spatial resolution. Furthermore, elevation, slope, and aspect operate at comparable spatial scales within our site's limited spatial extent. This reduces the likelihood of scale mismatches among predictors. We therefore opted not to pursue GWR or MGWR in the final analysis, but do believe our approach adequately accounts for spatial structure within the constraints of our dataset and respectfully suggest that future studies be conducted across broader or more heterogeneous landscapes.

Comment 8: Eighth, some terms used and arguments made in Discussion are not perfectly appropriate. Lines 239 - 244 should be mentioned in Introduction to justify the study of the damages of this hurricane on Puerto Rico forests. Lines 255-256 need references. Interaction is not considered in the results, but it is mentioned in Lines 261-264. In Lines 270-275, the arguments made by personal observations are not academically sound. In Lines 279-281, the interpretation for the statistical difference based on mean elevation for the two types of sites is not accurate.

Response: Thank you for your comment. We have decided to remove the sentence from line 239-244 because this information was covered in the introduction. We have added a relevant reference to lines 255-256 (now lines 301-303). The interaction mentioned in lines 261-264 (now lines 264-311) refers to the complex relationship between hurricanes and forests that have been studied extensively, and not to any specific statistical interaction between variables included in our study. We have edited the text to clarify the context for the previous personal observations made (including by two of the co-authors) because they are relevant and provide a broader context to our study. We agree that it is not academically sound to include unpublished personal observations in datasets for analyses or results, which we have not done, but do believe it is important to include them in the Discussion. We do not understand how stating the difference in mean elevations is not accurate. Nevertheless we have deleted the sentence on lines 279-281.

Comment 9: Ninth, no references at all in sections 5.3 and 5.4, which significantly reduce the scientific soundness of this manuscript. Please find supporting evidence for these discussions.

Response: We have added some reference(s) where relevant to support our discussion. Nevertheless, some of our conclusions are drawn based on our results which we are presenting here because they are novel and not previously published. Furthermore, it is not unusual or scientifically unsound for authors to present novel conclusions in the Discussion section of papers without citing references when they are not available in the published literature.

Comment 10: Tenth, Conclusion should be revised accordingly after spatial analyses are conducted on these data. The current Conclusion is too strong without sufficient support from the Results of this manuscript.

Response: We have substantially revised the Results and Conclusion sections in numerous places to incorporate the new spatial analyses and figures.

 

Minor Comments:

Comment 1: A histogram or a table of categories is needed to justify the definition of the damage key (Lines 138-140).

Response: We are unable to create such a histogram because of the manner in which the data were collected in 3D VR, using these categories apriori, i.e., not counting exact number of trees in each image to record them. We had determined the categories based on a preliminary visual assessment of a subset of the images.

Comment 2: The 24 points not sampled should be shown in Figure 4. 

Response: We have added the additional survey points not sampled to the map.

Comment 3: The three Figure 7s can be provided in Appendix. 

Response: We have reorganized all the figures and reduced the overall number to 9 instead of the previous 16.

Comment 4: Definition of the 8 categories of ground cover should be clearly defined in 3.3.2

Response: The categories are labeled distinctly using common names/words, so we don’t know how to define them more clearly.

Comment 5: Why are the degrees of freedom for the F test in Figure 11 1 and 68, instead of 1 and 71?

Response: Unfortunately, a few of the 360° photos taken in the field were completely foggy due to the inside of the camera lens being completely fogged up. As a result, several photos had to be excluded from VHDI analysis.

Comment 6: (6) If the t test modified for spatial processes (Dutilleul et al., 1993) is applied to Figure 12, the significance likely disappears.

Response: Since we didn’t find evidence of spatial clustering or autocorrelation in Grass/Herbaceous ground cover (see Table 1), we don’t think the modified t-test is applicable in this case.

Comment 7: (7) Figure 15 should be organized in the same layout as Figure 14.

Response: Figure 15 (now Figure 8) is organized differently because on the graph elevation is the independent variable (continuous X-axis) and damage rank is the dependent variable (categorical Y axis). In all the other similar graphs, now consolidated in Figure 7, damage rank is the independent variable and therefore on the X-axis.

Comment 8: 7 years are not particularly a long-term period (Line 230). The 75 sample points are not large enough for a landscape level study (Line 231). The authors should use the terms more scholarly and specific.

Response: While we agree that 7 years is not long-term, in the context of other studies where much of the literature is focused on the immediate and short-term implications after hurricane impact, we think it is reasonable to say that this study is more long-term than most studies.

Comment 9: Please provide the coordinates where the three photos in Figure 16 were taken.

Response: The coordinates have been added to the figure (now figure 9).

Comment 10: Because no assessment was taken before Hurricane Maria, it is not solid to make absolute argument on recovery of canopy in Lines 297-299.

Response: We have made a revision which can be seen on lines 356-358.

Reviewer 2 Report

Comments and Suggestions for Authors

Review of the manuscript land-3661176

 

Title: Running up the Hill and into a Hurricane: Forests are More Vulnerable to Hurricane Damage and Slower to Recover at Higher Elevations in Southeastern Puerto Rico

 

 

General comments:

This work studies the impact of hurricane Maria in 2017 on the forests of Las Casas de la Selva, a sustainable plantation in Puerto Rico. The authors used 360° photographs taken at 75 sites in 2024, seven years after the hurricane, to document the state of the forest structure. These images allowed for an assessment of the vertical structure of the forest, the canopy cover and the herbaceous ground cover, and for comparing the different degrees of hurricane impact on the forest. In addition, the authors analysed the effects of site elevation, slope, and aspect on the extent of hurricane damage using a vertical habitat diversity index (VHDI) calculated from the hemispherical photographs and considering the strata of herbaceous, shrub, understory and canopy. The main result was that canopy cover decreased and herbaceous ground cover increased with increasing elevation, and that the VHDI did not show any significant differences between contrasting hurricane-damaged sites. The authors have shown that the elevation has a significant impact on hurricane damage to both mixed plantations and native forests in Las Casas de la Selva, with higher elevations being more vulnerable to hurricane damage. The authors stressed that these results had implications for the restoration of plantations and the management of forests in hurricane-prone areas where higher intensity hurricane patterns are predicted. I think this work is a valuable starting point for future long-term research to better understand the impact of hurricanes on the different components of the forest. I consider the manuscript is well structured, well written, the statistical methods used are sound and reliable, and the bibliographic references are detailed and up-to-date. However, I believe that the work could be improved by considering some suggestions detailed in the specific remarks below.

 

Specific remarks:

- Lines 2-4. Title. I propose a shorter and more focused title, such as: “Caribbean Forests at Higher Elevations are More Vulnerable to Hurricane Damage and Slower to Recover in Southeastern Puerto Rico”.

 

- Lines 19-20. To avoid repetition and to make the information clearer and more structured, I suggest moving this sentence to line 15 and rewording both sentences in part as follows: “We studied forest structure variation across 75 sites at Las Casas de la Selva, a sustainable forest plantation in Patillas, Puerto Rico, seven years after Maria hit the property. At each site, we analyzed 360° photos in a 3DVR headset to quantify the vertical structure and transformed them into hemispherical images to quantify canopy closure and groundcover”.

 

- Line 122. The current Figure 3 should be Figure 2?

 

- Line 140. Figures 1 and 4 may be combined into a single figure concerning the location of the study area and the study design. This will help to reduce the excessive number of figures in the work.

 

- Lines 132, 141, 203, 207, 213. Please remove unnecessary “:” at the end of these sub-section titles.

 

- Lines 157-158. The three pictures of Figures 7 may be combined to form one Figure X, named Fig. Xa, Xb and Xc, relating to field equipment used in the work.

 

- Lines 157, 186 and 195. The three images of Figures 8, 9 and 10 can be combined into a single Figure and then reducing the excessive number of figures in the manuscript. This figure will be related to images of canopy and ground obtained in the field and how these were analyzed in the work.

 

- Lines 206, 212 and 222. The three graphs of Figures 11, 12 and 14 have the same format and then can be combined into a single Figure X, named as Fig. Xa, Xb, Xc. This would reduce the excessive number of figures in the work.

 

- Line 206. Rewrite by changing “2” in R2 to a superscript value and change “F = 1,68” by “F1,68”.

 

- Figure 11. In the ANOVA statistics of the figure, change “F = 1,68” by “F1,68”.

 

- Line 210. Rewrite by changing “2” in R2 to a superscript value and change “F = 1,71” by “F1,71”.

 

- Figure 12. In the ANOVA statistics of the figure, change “F = 1,71” by “F1,71”.

 

- Line 216. Rewrite by changing “2” in R2 to a superscript value.

 

- Line 219. Rewrite by changing “2” in R2 to a superscript value and change “F = 1,71” by “F1,71”.

 

- Figure 14. In the ANOVA statistics of the figure, change “F = 1,71” by “F1,71”.

 

- Figure 12. In the ANOVA statistics of the figure, change “F = 1,71” by “F1,71”.

 

- Figure 15. In the Logistic Regression statistics of the figure, change “F = 1,73” by “F1,73”.

 

- Line 342. Change “tabonuco” by “Tabonuco”.

 

Author Response

We are grateful for your thoughtful comments and suggestions.

Comment 1: Lines 2-4. Title. I propose a shorter and more focused title, such as: “Caribbean Forests at Higher Elevations are More Vulnerable to Hurricane Damage and Slower to Recover in Southeastern Puerto Rico”.

Response: We have reworded the title to make it more specific and focused: “Tabonuco and Plantation Forests at Higher Elevations are More Vulnerable to Hurricane Damage and Slower to Recover in Southeastern Puerto Rico”. We opted for this title because our study site was located in a mixed Tabonuco and Plantation forest and because Tabonuco forest type is native to Puerto Rico and the Lesser Antilles.

Comment 2: Lines 19-20. To avoid repetition and to make the information clearer and more structured, I suggest moving this sentence to line 15 and rewording both sentences in part as follows: “ . At each site, we analyzed 360° photos in a 3DVR headset to quantify the vertical structure and transformed them into hemispherical images to quantify canopy closure and groundcover”.

Response: Thank you for your great advice. We have made the recommended changes moving lines 19-20 to line 15 and re-worded both sentences starting at the end of line 12 and ending on line 16 as “We studied forest structure variation across 75 sites at Las Casas de la Selva, a sustainable forest plantation in Patillas, Puerto Rico, seven years after Hurricane Maria hit the property. At each site we analyzed 360° photos in a 3DVR headset to quantify the vertical structure and transformed them into hemispherical images to quantify canopy closure and ground cover.”

Comment 3: Line 122. The current Figure 3 should be Figure 2?

Response: Thank you so much for spotting this! On line 122, figure 2 was hidden behind figure 3 which was something we failed to notice. We have fixed this mistake and figure 2 (now figure 1) can be seen on page 4.

Comment 4: Line 140. Figures 1 and 4 may be combined into a single figure concerning the location of the study area and the study design. This will help to reduce the excessive number of figures in the work.

Response: Figures 1 and 4 have now been combined into one figure (now Figure 3) which can be found on page 5.

Comment 5:  Lines 132, 141, 203, 207, 213. Please remove unnecessary “:” at the end of these sub-section titles.

Response: All the unnecessary “:” have been removed from the subsection titles.

Comment 6: - Lines 157-158. The three pictures of Figures 7 may be combined to form one Figure X, named Fig. Xa, Xb and Xc, relating to field equipment used in the work. 

Response: We have combined the three figure 7’s (now Figure 4) and it is now in the specified format you recommended and can be found on page 6. The accessories for the camera have also been removed so the focus would be on the camera.

Comment 7:- Lines 157, 186 and 195. The three images of Figures 8, 9 and 10 can be combined into a single Figure and then reducing the excessive number of figures in the manuscript. This figure will be related to images of canopy and ground obtained in the field and how these were analyzed in the work.

Response: We have made the changes you suggested combining figures 8, 9, and 10 (now figure 5) into one using the format you suggested. You can find the new figure on page 7.

Comment 8: - Lines 206, 212 and 222. The three graphs of Figures 11, 12 and 14 have the same format and then can be combined into a single Figure X, named as Fig. Xa, Xb, Xc. This would reduce the excessive number of figures in the work.

Response: We have made the changes you suggested and combined the previous figures 11, 12, and 14 into the new Figure 7, which also includes a new graph for Simpson VHDI (as recommended by another reviewer) and can be found on page 10.

Comment 8:- Line 206. Rewrite by changing “2” in R2 to a superscript value and change “F = 1,68” by “F1,68”.

Response: We have made the recommended changes.

Comment 9: - Figure 11. In the ANOVA statistics of the figure, change “F = 1,68” by “F1,68”.

Response: We have made the recommended changes in the new figure 7. 

Comment 10: - Line 210. Rewrite by changing “2” in R2 to a superscript value and change “F = 1,71” by “F1,71”. 

Response: We have made the changes you suggested.

Comment 11: - Figure 12. In the ANOVA statistics of the figure, change “F = 1,71” by “F1,71”.

Response: We have made the recommended changes.

Comment 12: - Line 216. Rewrite by changing “2” in R2 to a superscript value. 

Response: The result in the previous line 216 was replaced with new spatial analysis.

Comment 13: - Line 219. Rewrite by changing “2” in R2 to a superscript value and change “F = 1,71” by “F1,71”.

Response: We have adjusted font sizes to ensure the superscripts appear correctly, and subscripted the degrees of freedom for the F values throughout the manuscript.

Comment 14: - Figure 14. In the ANOVA statistics of the figure, change “F = 1,71” by “F1,71”.

Response: Done. 

Comment 15: - Figure 15. In the Logistic Regression statistics of the figure, change “F = 1,73” by “F1,73”.

Response: Done.

Comment 16: - Line 342. Change “tabonuco” by “Tabonuco”.

Response: Done.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for improving the manuscript with additional details of the study area, appropriate analyses, further discussion on the results, and a revised conclusion.

Just three minor comments on the text added to the revision.

  1. In Table 1, please rewrite the p-value for canopy closure as 2x10-6
  2. Section 4.4 and figure 7d should be presented using the Local Bivariate Relationships analysis as canopy closure is spatially autocorrelated. 
  3. In Lines 311-349, and Lines 361-388, the authors should provide supporting evidence for their discussion for the possible causes and factors associated with their findings in the study area, especially evidence with biological and/or ecological understanding of trees and forests.

Other than these, I believe the manuscript has been well improved for publication in Land.

Author Response

Comments 1: In Table 1, please rewrite the p-value for canopy closure as 2x10-6

Response 1: Thank you for pointing this out. We have changed the p-value to 2x10-6.

Comments 2: Section 4.4 and figure 7d should be presented using the Local Bivariate Relationships analysis as canopy closure is spatially autocorrelated. 

Response 2: We have redone the analysis of the relationship between canopy closure and damage by applying Local Bivariate Relationships analysis and replaced the earlier figure 7d with a new figure 7 showing those results. This became possible when we reexamined the original raw data file for the damage variable, and realized that Michael had recorded the actual number of dead/recovering trees at each site, not just the Damage Rank grouping. This allowed us to use Damage as a continuous rather than a categorical variable, making it possible to apply the LBR tool in ArcGIS Pro (the tool requires both variables to be continuous).

Since we now have Damage as a continuous variable, we redid the other analyses for Damage also using LBR, which led to additional changes to the manuscript: 1) we removed the previous Figure 7 and 8 entirely and replaced them with two new figures showing significant LBR results; 2) the difference we had found in % Grass/Herbaceous Cover lost statistical significance when accounting for spatial autocorrelation (something you had also predicted earlier); and 3) we removed all the ANOVA/Logistic Regression results since they are no longer needed, and replaced them with the relevant LBR results. The main findings remain the same as before, but now they are supported by spatially explicit statistical analyses, making them more robust. Your insights have been invaluable, therefore we want to thank you again for helping us make this paper stronger.

Comments 3: In Lines 311-349, and Lines 361-388, the authors should provide supporting evidence for their discussion for the possible causes and factors associated with their findings in the study area, especially evidence with biological and/or ecological understanding of trees and forests.

Response 3: We have added 7 new citations (including 1 new paper added to the reference list) in these sections, pointing to other studies for supporting evidence.

Back to TopTop