Forest Structure and Composition Are Critical to Hurricane Mortality

Round 1

Reviewer 1 Report

Overall, I think this manuscript presents an important contribution to the understanding of the effects of forest state and hurricane strength on forest dynamics. I thought it was well-written and very interesting. I also think that the results support the conclusion that the combined effects of storm factors and forest structure and composition lead to differences we see in patterns of hurricane damage and mortality. I was very happy to see the comparison between two forest sites (the focal location of Bisley and Luquillo) and between two hurricanes (Hugo and Maria) because these comparisons start to move us away from specific analyses of particular forest sites in a particular hurricane. Instead, we may be able to start to make more generalizations.

A similar study comparing the effects of Hurricanes Hugo and Maria on these rainforests of Puerto Rico was published a few years ago by Uriarte et al. (2019), but that study only used the Luquillo Experimental Forest and focused more on the relationships between tree diameter and wood density on damage and mortality. In addition, that study was only for trees >= 10 cm while this study includes smaller stems (between 2.5 and 10 cm diameter). With their additional state, they are able to show relative importance of hurricane strength, forest state (forest structure and composition), and Accumulated Cyclone Energy (ACE, local winds) with forest state mattering more than hurricane strength or ACE. They are only able to conclude this because they have two sites and two hurricanes that passed over sites that differed greatly in the 30 or so years between the hurricanes.

I had 4 “major” criticisms/suggestions and then just a few minor suggestions.

• I highly recommend using a continuous diameter rather than using categories. Using these categories for diameter may obscure the overall patterns and also makes the text cumbersome because each category is discussed. Is there a reason why categories were used? If the authors can’t justify the use of categories sufficiently, I would recommend changing the analyses to be continuous. Then Fig. 1b could be made into a series of graphs as in Uriarte et al. (2019).
• It would have been better if the damage categories matched those of other studies (Uriarte et al. 2019, Zimmerman et al. 1994, Curran et al. 2008, Bellingham et al. 1995). I also wasn’t sure I understood these damage categories – why weren’t the “traditional” damage categories of major branch damage, stem break/snap, and uprooting used? It looks like Damage III includes what other studies define (and usually separate) as stem break and uproot. Is that correct? If so, is there any way to separate stem breaks from uproots because that would greatly help. Not much is done in this manuscript with the different levels of damage, so I think it would be better to remove them or change them to categories used by others.
• I would like more information on how species were categorized into those PFT – what evidence was used to designate a species as early vs. mid- vs. late successional? I also suggest using “fast, intermediate, slow growing” as PFT categories rather than “early, mid, and late successional species” to focus on demographic rates rather than when it establishes in succession (which may be rather jumbled).
• I also would suggest removing the tree ferns from the analysis or give them their own PFT as was done for Prestoea because ferns really aren’t comparable to the trees in this study. The authors make the argument that Prestoea should be separated into its own PFT because it is a monocot – for this reason alone the tree ferns should at least be separated if not excluded (palms are at least a type of seed plant!). Also, other studies of hurricane effects have not included tree ferns, even from Puerto Rico (e.g., Zimmerman et al. 1996, Uriarte et al. 2019), making it hard to compare patterns between this study and others. It looks like the high mortality due to Hurricane Hugo at Bisley might have been due in large part to the 70% mortality of Cyathea portoricensis in that storm. What happens if the two tree fern species are removed from these analyses? At the very least, I would recommend adding an analysis that removes them. Were the calculations in Fig. 5 conducted without these species?

Fig. 1 – I think the data here would be easier to digest as a table (like Table 2 from Curran et al. 2008). The lower part of each panel is particularly hard to read/understand. In particular, Damage III is the most interesting one, and the dots for this category aren’t even visible in c. I would just change all of these to a series of tables.

Fig. 2 – I really like this figure! Would it be possible to create one just with the stems >= 10 cm?

Fig. 3 – I like these graphs, but it would make more sense to code Hurricane Maria in blue and Hugo in red, following Uriarte et al. (2019).

Fig. 5 – I would shade the two hurricanes in blue and red as in Fig. 3 (hopefully changed to be like Uriarte et al. 2019).

L123: Change to “was likely due”

L170: Change to “likely partly explain”

L193-227: Change present tense verbs to past tense.

L271: Remove the comma after [28]

Table S1: Family names should not be in italics.

Table S2: change mortality from probabilities to percentages and just report one decimal point

Curran, T. J., R. L. Brown, E. Edwards, K. Hopkins, C. Kelley, E. McCarthy, E. Pounds, R. Solan, and J. Wolf. 2008. Plant functional traits explain interspecific differences in immediate cyclone damage to trees of an endangered rainforest community in north Queensland. Austral Ecology 33:451-461.

Author Response

We thank the reviewer for the helpful comments. The point-by-point response is attached.

Author Response File: 

Response to Reviewer 1 Comments

Comments and Suggestions for Authors

Overall, I think this manuscript presents an important contribution to the understanding of the effects of forest state and hurricane strength on forest dynamics. I thought it was well-written and very interesting. I also think that the results support the conclusion that the combined effects of storm factors and forest structure and composition lead to differences we see in patterns of hurricane damage and mortality. I was very happy to see the comparison between two forest sites (the focal location of Bisley and Luquillo) and between two hurricanes (Hugo and Maria) because these comparisons start to move us away from specific analyses of particular forest sites in a particular hurricane. Instead, we may be able to start to make more generalizations.

A similar study comparing the effects of Hurricanes Hugo and Maria on these rainforests of Puerto Rico was published a few years ago by Uriarte et al. (2019), but that study only used the Luquillo Experimental Forest and focused more on the relationships between tree diameter and wood density on damage and mortality. In addition, that study was only for trees >= 10 cm while this study includes smaller stems (between 2.5 and 10 cm diameter). With their additional state, they are able to show relative importance of hurricane strength, forest state (forest structure and composition), and Accumulated Cyclone Energy (ACE, local winds) with forest state mattering more than hurricane strength or ACE. They are only able to conclude this because they have two sites and two hurricanes that passed over sites that differed greatly in the 30 or so years between the hurricanes.

I had 4 “major” criticisms/suggestions and then just a few minor suggestions.

• I highly recommend using a continuous diameter rather than using categories. Using these categories for diameter may obscure the overall patterns and also makes the text cumbersome because each category is discussed. Is there a reason why categories were used? If the authors can’t justify the use of categories sufficiently, I would recommend changing the analyses to be continuous. Then Fig. 1b could be made into a series of graphs as in Uriarte et al. (2019).

Re: Even though the variable DBH is continuous theoretically, DBH measurements are discrete, with a precision of 0.1 cm. Continuous DBH analyses on census data are impossible if without any assumptions or smoothing. In fact, those continuous-DBH graphs in Uriarte et al. (2019) are not from pure census data, but from models with certain parameters and assumptions. Categorical DBH analyses, on the other hand, don’t require any assumptions or smoothing and are easy to interpret and understand. Moreover, many studies cited in this manuscript (e.g., Crow 1980, Heartsill Scalley et al. 2010, Heartsill Scalley 2017) and even the ones the reviewer referred in the following comments (Zimmerman et al. 1994, Curran et al. 2008, Bellingham et al. 1995) used diameter categories instead of continuous diameters. Therefore, we are keeping categorical DBH analyses in this manuscript.

• It would have been better if the damage categories matched those of other studies (Uriarte et al. 2019, Zimmerman et al. 1994, Curran et al. 2008, Bellingham et al. 1995). I also wasn’t sure I understood these damage categories – why weren’t the “traditional” damage categories of major branch damage, stem break/snap, and uprooting used? It looks like Damage III includes what other studies define (and usually separate) as stem break and uproot. Is that correct? If so, is there any way to separate stem breaks from uproots because that would greatly help. Not much is done in this manuscript with the different levels of damage, so I think it would be better to remove them or change them to categories used by others.

Re: Damage categories include intact (no damage), categories I, II, and III (corresponding to light, intermediate, and severe damages, respectively). The damage categories are pretty much similar to those in the papers pointed out (Zimmerman et al. 1994, Curran et al. 2008, Bellingham et al. 1995), except that those papers focused on the severe damage category and had more classifications on this category (separating stem breaks and uproots), while our manuscript combines stem breaks and uproots as severe damages and also includes light and intermediate damages as well. Stem breaks and uproots in the Damage III category (severe damage) in our manuscript are grouped together in the census observations and cannot be separated in this dataset. Hurricanes cause not only severe damages to trees but also light and intermediate damages (such as defoliation and branches breaking), those damages are as important as severe damages (such as stems breaking and uprooting) by opening the canopy and allowing succession after a hurricane disturbance. Therefore, we are keeping the damage analyses in this manuscript.

• I would like more information on how species were categorized into those PFT – what evidence was used to designate a species as early vs. mid- vs. late successional? I also suggest using “fast, intermediate, slow growing” as PFT categories rather than “early, mid, and late successional species” to focus on demographic rates rather than when it establishes in succession (which may be rather jumbled).

Re: The classifications of PFT have important applications for predictive ecology and modeling, and this Early, Mid, and Late PFT classification is in consistent with vegetation demography models, such as Ecosystem Demography model (ED2) (Moorcroft et al. 2001, Medvigy et al. 2009, and Longo et al. 2019). Early PFT includes species that establish and recruit in open canopy and gaps formed after disturbances and grow rapidly in the high light environment. Mid PFT includes species that have intermediate growth rates and are somewhat shade tolerant. Late PFT includes species that have low growth rates and are shade tolerant. Many previous studies have classified species into Early, Mid, and Late PFTs, and the classification in this study follows previous studies (Walker 1991, Schowalter and Ganio 1999, Uriarte et al. 2005, Muscarella et al. 2013, Heartsill Scalley 2017, Feng et al. 2018). We have added the references in the revised manuscript in response to reviewer comment.

• I also would suggest removing the tree ferns from the analysis or give them their own PFT as was done for Prestoea because ferns really aren’t comparable to the trees in this study. The authors make the argument that Prestoea should be separated into its own PFT because it is a monocot – for this reason alone the tree ferns should at least be separated if not excluded (palms are at least a type of seed plant!). Also, other studies of hurricane effects have not included tree ferns, even from Puerto Rico (e.g., Zimmerman et al. 1996, Uriarte et al. 2019), making it hard to compare patterns between this study and others. It looks like the high mortality due to Hurricane Hugo at Bisley might have been due in large part to the 70% mortality of Cyathea portoricensis in that storm. What happens if the two tree fern species are removed from these analyses? At the very least, I would recommend adding an analysis that removes them. Were the calculations in Fig. 5 conducted without these species?

Re: We agree that tree ferns are different from trees and deserve a unique PFT in future studies. However, either removing them from analyses or categorizing them as a unique PFT is not practical in this study. Tree ferns are a member of the forest, and they should not be removed from analyses. The abundance of the tree ferns (40 stems before Hugo and 63 stems before Maria) is not enough to categorize them as a unique PFT, as was the case with other PFTs (more than 100 stems for each census before either hurricane).

It is true that tree fern Cyathea portoricensis (CYAPOR) had high mortality during hurricane Hugo (72%), but it is not the reason for the high overall mortality of the forest (58%) because the relative abundance of the species in the forest is less than 5% of the total abundance (39 stems of CYAPOR vs 936 stems of all stems) (Table S2). The other tree fern CYAARB has only 1 stem before Hugo. Therefore, tree ferns contribute only a small portion to the total stem abundance in the forest. The mortality shown in Figure 5 is for stems with DBH ≥ 10 cm. As we know, tree ferns do not have large DBHs, and the tree ferns in this study barely exceed 10 cm in DBH. As shown in the newly added figure Fig. S4, neither of the two tree fern species have showed up in the graph, which means that the abundance of the tree ferns is less than 10 stems. Therefore, the results in Fig. 5 would not change much even if the tree ferns were removed from analyses.

Fig. 1 – I think the data here would be easier to digest as a table (like Table 2 from Curran et al. 2008). The lower part of each panel is particularly hard to read/understand. In particular, Damage III is the most interesting one, and the dots for this category aren’t even visible in c. I would just change all of these to a series of tables.

Re: The lower part of each panel in the figure has the same information as in the upper panel, but for a different scale (absolute numbers in the upper panel and relative numbers in the lower panel). Figure 1 contains too much information to be included in a single Table. It is much clearer and easier to digest than in a series of tables. Nevertheless, as suggested by the reviewer we have added the data in supplementary tables S3, S4, and S5.

Fig. 2 – I really like this figure! Would it be possible to create one just with the stems >= 10 cm?

Re: We do not attempt to repeat what has been done in Uriarte et al. (2019). Nevertheless, we have added Figure S4 in supplements for stems ≥ 10 cm.

Fig. 3 – I like these graphs, but it would make more sense to code Hurricane Maria in blue and Hugo in red, following Uriarte et al. (2019).

Re: The color for each hurricane event is arbitrary, and we do not attempt to use the same color as in Uriarte et al. (2019), after all, this paper and Uriarte et al. (2019) are independent.

Fig. 5 – I would shade the two hurricanes in blue and red as in Fig. 3 (hopefully changed to be like Uriarte et al. 2019).

Re: Figure 5 is clear now and it contains four colors representing four groups of comparison. Adding more colors to the figure would make the figure more complex. We prefer to keep as it is.

L123: Change to “was likely due”

Re: revised as suggested.

L170: Change to “likely partly explain”

Re: revised as suggested.

L193-227: Change present tense verbs to past tense.

Re: revised as suggested.

L271: Remove the comma after [28]

Re: revised as suggested.

Table S1: Family names should not be in italics.

Re: revised as suggested.

Table S2: change mortality from probabilities to percentages and just report one decimal point

Re: revised as suggested.

Curran, T. J., R. L. Brown, E. Edwards, K. Hopkins, C. Kelley, E. McCarthy, E. Pounds, R. Solan, and J. Wolf. 2008. Plant functional traits explain interspecific differences in immediate cyclone damage to trees of an endangered rainforest community in north Queensland. Austral Ecology 33:451-461.

Reviewer 2 Report

Great job.  I question the location of the Methods section at the end of the paper and not before Results.  Also wonder about the capitalization of the word hurricane.  I thought if used in talking of specific hurricane it was capitalized as in Hurricane Hugo.  when used to refer to more than one hurricane it was not capitalized as in hurricanes Hugo and Maria.  Check with editor on that.

Here are my suggested edits: 

line 10 – Hurricanes can cause severe….

line 18 – substitute describe for study

line 30 – delete “the” before long-term

line 53 – replace complexes with complicates or influences

lines 61-62 - Knowledge of tropical cyclone ecology from this forest system could lead to advances in our understanding of fundamental forest dynamics and responses to hurricane disturbances.

line 65 – replace reported with showed

lines 67-73 – suggested rewording “Early successional species like Cecropia schreberiana and Miconia tetrandra had the highest mortality (51.85% and 30.77%, respectively), while the mortality of other species was relatively low (less than 20%) (Figure 1a; Table S2). Late successional species (e.g., Sloanea berteriana and Dacryodes excelsa), mid successional species (e.g., Casearia arborea and Inga laurina), and palm (Prestoea montana) were more prone to defoliation than mortality or snapping (Figure 1a, Table S2).

line 75 – comma after defoliated

line 79 - ….chance of being hit by falling branches from large trees.

line 94 - ….C. schreberiana, which experienced ~50% mortality during both hurricanes

line 135 – I suggest a comma before after P. montana

lines 151-154 – suggested rewording “Species composition and size structure present before hurricanes Hugo and María at BEW. (a) Stem counts of the most abundant species and (b) distribution of each DBH size.”

line 160 – delete “away”

lines 162-164 – suggested wording “Hurricane Huo impacted the BEW forest (within 150 km) for 24 hours (September 18 00:00 UTC – 162 September 19 00:00 UTC, 1989) as did Hurricane María (September 20 163 00:00 UTC – September 21 00:00 UTC, 2017).

line 167 – was significantly higher

line 178 – delete “The” at the end of the line

line 179 – capitalize Small

line 181 – delete “the” before local

line 184 – delete “the forest at”

line 192 – delete “the” before local, topographic, and accumulated

line 193 – was much lower; delete “that before during

line 194 – delete “the” before size

line 175 – Hugo was more

line 197 – that was more

line 211 – delete “at site

line 214 – was significantly higher than mortality

line 219 – comma after species

line 221 – delete “the” before late

line 223 – delete “at site”

line 226 – insert “and” before forest state

lines 229-232 – suggested rewording “Comparison of mortality and impact factors at Luquillo Forest Dynamics Plot (LFDP) and Bisley Experimental Watersheds (BEW) during hurricanes Hugo and María. (a) The mortality at the two 230 sites during the two events. The mortality and impact factors at (b) BEW during the two events 231 and (c) LFDP during the two events.”

line 237 – hurricane strength at landfall (in knots)

lines 241-242 – suggested rewording “Orange font color indicates that the factor cannot explain mortality while the green font color indicates that the factor is consistent with mortality.”

line 244 – delete “The” and start sentence with “Relative”

line 247 – delete “Our results demonstrate” and Start sentence with “Hurricane”

line 252 - suggested rewording “graphic exposure, and forest structure and composition prior to disturbance are all”

line 258 – delete “as” before clearly

line 264 – delete “will”

line 272 – delete “the” before hurricanes

line 278 – shouldn’t this section be before Results??

lines 280-281 - The Bisley Experimental Watersheds (BEW; 18°20’N, 65°50’W) is located on the northeastern side of the Luquillo Mountains in Puerto Rico.

line 285 – delete “continuously”

line 290 – suggested rewording “cm in each plot recorded every five years beginning in 1989.”

line 293 – refer to them

line 298 – suggested “their successional status - early, mid, or late successional trees and palm (Table S1).”

line 299 – comma before and after Prestoea montana

line 302 – delete “the” before early

line 303 – delete “the” before late

line 305-307 – suggested wording “Damages in the post-María census were classified as Damage I if a stem was defoliated less than 50%, Damage II if a stem had broken branches or heavy defoliation (>50%), and Damage III if the trunk or roots were broken. Trees were declared dead if there was no evidence of”

lines 309-314 – delete “the” and start sentence with “Crowns were classified as Dominant if the percentage of the crown under direct sunlight was more than 75%, Co-dominant if 50-75%, Intermediate if 25-50%, and Suppressed if less than 25%. Mortality from hurricane María was calculated as the ratio of the number of 312 stems recorded dead to the total number of stems recorded in the post-María census.  Since trees that appeared dead right after a hurricane may resprout and be alive later and we do not”

line 316 – hurricane Hugo was

line 319 – information from later censuses

line 324 – local wind was estimated using a modified

line 327 – calculated the likelihood of exposure in terms of topographic angle (ø), which was calculated

line 337 – were reconstructed

line 340 – parameters were set

line 355 – were calculated

line 358 – The difference in mortality between the two events was tested

line 379 – Species composition and ……

line 383 – are on a logarithmic scale

line 405 – is accessible

line 412 – do not put volume number in bold

lines 417-418 – do not capitalize all the words in the article title

lines 495-496 - do not capitalize all the words in the article title

Supplementary Material

line 28 – are given in parentheses

line 32 – Species composition and size structure of BEW and LFDP sites at times of hurricanes Maria and Hugo.

line 45 – Mortality of each species from hurricanes Hugo and Maria.

Author Response

We thank the reviewer for the useful comments. We put the Methods section at the end because the most important part of the manuscript is not about the methods but the results. The Methods section is very long and putting it before Results would distract the main topic of the manuscript. The journal Forests accepts free format, so we prefer this style and are keeping the current order of the sections.

As for whether capitalizing hurricane before a hurricane name, we got feedbacks from multiple experts, and some prefer to capitalize it while some prefer not to. To our knowledge, as long as it is consistent throughout the manuscript, it should be fine. Therefore, we are keeping the current form without capitalization.

All of the rest suggested edits have been addressed in the revised version of the manuscript.

Reviewer 3 Report

Ideally, hurricane strength would have been randomly assigned to replicated plots, but this is obviously impossible (!) and the comparative approach taken here is the best that can be done when reporting the impacts of rare extreme events. Puerto Rico is an odd little island, both ecologically and biogeographically, but nowhere else has this data this good and the presentation is excellent throughout. I visited Luquillo a few years after Hugo, so I have an advantage over most readers, but I would be happy to see this published as it is. For readers who do not know the study area, there is a vast and easily accessible literature, well-cited here. 

Author Response

Thanks!