the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The Impact of Hurricane Disturbances on a Tropical Forest: Implementing a Palm Plant Functional Type and Hurricane Disturbance Module in ED2-HuDi V1.0
- 1School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA, United States
- 2Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, United States
- 3Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, United States
- 4USDA Forest Service, International Institute of Tropical Forestry, Río Piedras, PR, United States
- 1School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA, United States
- 2Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, United States
- 3Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, United States
- 4USDA Forest Service, International Institute of Tropical Forestry, Río Piedras, PR, United States
Abstract. Hurricanes commonly disturb and damage tropical forests. It is predicted that changes in climate will result in changes in hurricane frequency and intensity. Modeling is needed to investigate the potential response of forests to future disturbances. Unfortunately, existing models of forests dynamics are not presently able to account for hurricane disturbances. We implement the Hurricane Disturbance in the Ecosystem Demography model (ED2) (ED2-HuDi). The hurricane disturbance includes hurricane-induced immediate mortality and subsequent recovery modules. The parameterizations are based on observations at the Bisley Experimental Watersheds (BEW) in the Luquillo Experimental Forest in Puerto Rico. We add one new plant functional type (PFT) to the model—Palm, as palms cannot be categorized into one of the current existing PFTs and are known to be an abundant component of tropical forests worldwide. The model is calibrated with observations at BEW using the generalized likelihood uncertainty estimates (GLUE) approach. The optimal simulation obtained from GLUE has a mean relative error of −21 %, −12 %, and −15 % for stem density, basal area, and aboveground biomass, respectively. The optimal simulation also agrees well with the observation in terms of PFT composition (+1%, −8 %, −2 %, and +9 % differences in the percentages of Early, Mid, Late, and Palm PFTs, respectively) and size structure of the forest (+0.8 % differences in the percentage of large stems). Lastly, using the optimal parameter set, we study the impact of forest initial condition on the recovery of the forest from a single hurricane disturbance. The results indicate that, compared to a no-hurricane scenario, a single hurricane disturbance has little impact on forest structure (+1 % change in the percentage of large stems) and composition (< 1 % change in the percentage of each of the four PFTs) but leads to 5 % higher aboveground biomass after 80 years of succession. The assumption of a less severe hurricane disturbance leads to a 4 % increase in aboveground biomass.
- Preprint
(1541 KB) -
Supplement
(2495 KB) - BibTeX
- EndNote
Jiaying Zhang et al.
Status: closed
-
RC1: 'Comment on gmd-2021-410', Anonymous Referee #1, 15 Feb 2022
This paper uses a forest dynamics model (ED2) to assess how hurricane disturbance effects composition and structure of a tropical forest. Key innovations in the work are (1) the development of a plant functional type specifically to represent palms, and (2) the implementation of hurricane disturbance in the ED2 model.Overall, the paper is heavily focused on development of the model (warranted given the venue). However, I did think the biological context and implications could be more thoroughly presented, especially in the intro and discussion. I also felt there could be more reference to previous work on forest dynamics in the Luquillo Mountains (see detailed comments).
Although the manuscript is mainly well-written, there is room for improvement with respect to grammar / language. Please see detailed comments for some of the most important parts. There are numerous references on key points to papers that are in review. I do not know the policy of this journal but some things of the things being referenced are really critical to undestand this paper properly (see detailed comments but, e.g., which species were classified into which PFTs, and on what basis?). I would think that some of these details should be included here in supplementary tables. At minimum, the papers in review should be posted to a pre-print server so the information cited is actually available.
Please note that many of the details on ED2 implementation / testing / calibration are outside of my expertise. Some of my detailed comments below do address these points, especially parts that I found could be more clear. Overall, however, I focus my comments on the general presentation of information in the paper and the links to the biology of the study system.
Thanks for the opportunity to review this manuscript, I hope the authors find my comments useful.
INTRODUCTIONGeneral: Overall, I think this section could be improved by making a more clear introduction to the fact that this study uses a vegetation model to understand hurricane effects on a tropical forest. As it stands, much of the text focuses on effects of hurricanes and then it jumps into PFTs and there is little/no text giving the context / background / rationale / etc. for using vegetation models, including using them to study hurricane effects. This is particularly surprising because there has been other work (including in wet forests of Puerto Rico) on modeling hurricane effects (see, e.g., Uriarte et al. 2009; Natural disturbance and human land use as determinants of tropical forest dynamics: results from a forest simulator). Although not investigating hurricane impacts, another study using ED2 for wet forests in Puerto Rico did include palms as a 'late' PFT (Feng et al. 2017; Improving predictions of tropical forest response to climate change through integration of field studies and ecosystem modeling). It seems strange that this literature is not reviewed in the introduction.
L 35-42: Exposure (to hurricane winds) seems to be an important missing factor here?
L 45: It is not clear what is meant by "faster resprouting" - does it refer to sprouts being generated vs. time since disturbance? It seems to mean something different because the next sentence mentions time since disturbance as a separate point. Please clarify.
L 47: Not clear what is meant by "higher recovery equilibrium". Equilibrium of what? Equilibrium in what sense?
L 51-54: It is a bit odd that this last part of the paragraph highlights a single article (Wang and Eltahir 2000) rather than providing some kind of summary / conclusion point about the preceding paragraph. If the Wang and Eltahir paper is very important to mention specifically then please provide more context.
L 55: The start to this paragraph is abrupt and could be more smoothly linked to the preceeding text. I think the jump is mainly in the fact that all of a sudden you are talking about terrestrial biosphere models and PFTs but none of this has been introduced. A better link would help the overall flow here. The introduction is pretty short as it stands so there is space to develop this more.
L 57-63: More citations for the justification of the 3 PFTs mentioned would be useful.
L 63: I would advocate to start a new paragraph when introducing the palm. This is a focal point of the paper but it is kind of buried in this paragraph.
L 64-65: Also reference:
Uriarte, María, Jill Thompson, and Jess K. Zimmerman. 2019. “Hurricane María Tripled Stem Breaks and Doubled Tree Mortality Relative to Other Major Storms.” Nature Communications 10 (1): 1362. https://www.nature.com/articles/s41467-019-09319-2L 69-70: "To account for these unique characteristics [in what? for what reason?], we define a Palm PFT." As other comments above, I think there needs to be an expansion of the introduction about vegetation modeling - building up to the overall aims of this study.
L 74: Seems strange to say "The results indicate that a single hurricane disturbance has little impact on forest structure" when much of the introduction was spent discussing the various impacts hurricanes have on forest structure / composition. Does this really mean "long-term" forest structure? Please clarify.
METHODS
L 92: "Since there is little knowledge about the traits of Palm." This is not a complete sentence and seems like it should be merged with the previous statement or otherwise revised.L 93: Not all palms have this low "wood density" and I think, in general, you should couch the statements about palms with something like "many palms" since it is such a diverse group and we do not actually know the degree to which these statements might be true or contradicted by some palms. In fact, the range of wood density for palms (Arecaceae) in the global wood density database (Chave et al. 2009, Zanne et al. 2009) is 0.180 - 0.883 (median = 0.54). (checked with the 'wdData' in the 'BIOMASS' R package, v. 2.1.5).
L94-96: It is difficult to assess this decision because there are no details on the other traits used in the model. If palms "grow fast in open canopies like early tropical trees" then what is the reason to assume they "have the same probability distributions as those of late tropical trees"? I am guessing that wood density is strongly related to growth rate in high light conditions in the model. But isn't it also related to mortality rates (including in shade)? Since the introduction of this palm PFT is such a big part of this paper I think it should be explained in more detail here.
L 104-105: Palms can be shorter than other trees, given the same DBH but I am a bit skeptical of these allometric relationships. For one thing, when I plot them, palms can never reach more than about 13 m height at the maximum diameter (20 cm), which is too short but other trees are predicted to reach unreasonably tall heights for this forest (~60 m for early PFT at the maximum 90 cm diameter). The only justification for these fitted parameters is from a paper by the authors 'in revision'.
L 107: This sentence requires a citation.
L 108: This sentence needs revision for clarity / grammar.
L 112-114: I am questioning the ramifications of these 'tricks' implemented in the model to help allow palms to survive despite their allometry.
L 115-117: Here the authors use default allometry of Early PFT for Palms but this seems inconsistent with the statement in L 94-96 about "...we assume that the traits of Palm have the same probability distributions as those of late tropical trees..." Please clarify.
L 124: A bit confused by "(sc) is the ratio of the cohort density that survived to the cohort density *before* the disturbance,"... should this not be the proportion that survives after the disturbance? Since sc=1-λc (L 127), and that λc "varies with hurricane strength, ...", it makes sense that sc would be post-hurricane survival... Please clarify / revise.
L 135: It seems like "Given mortality, the rate of each cohort (λc)" should be revised to "Given the mortality rate of each cohort (λc)"?
L 138-145: It is really not clear what is showing on the x-axis of these figures (x: proportion of large stems). How can this be the same during a given hurricane event for all PFTs? And why is mortality lower for all PFTs from hurricane Maria compared to Hugo? This section needs clarification.
L 146: Again some very relevant references seem to be missing. In particular Uriarte et al. 2009 (Natural disturbance and human land use as determinants of tropical forest dynamics: results from a forest simulator) and Uriarte et al. 2012 (Multidimensional trade-offs in species responses to disturbance: implications for diversity in a subtropical forest).
L 150-160: Please clarify the data upon which these functions are based.
L 170: I am missing details on the basis by which species were assigned to PFTS... this seems too important to have only as cited in a work 'in review'.
L 161-173: It seems that perhaps this description of the census data should go earlier in the text? But more importantly, it is a bit problematic to have such important references to work that is 'in review' (not to mention there are two Zhang et al. 'in review' papers so we don't know for sure which one is being cited here). Perhaps now this paper is published? What is the plan if this manuscript is accepted before the outcome of the one 'in review'? It seems like posting a pre-print of the other work could be at least a partial solution.
L 172: Does the Scatena et al. (1993) biomass allometry apply to palms? Should a caveat be included here?
L 193-194: Clarify specifically the reason why the dark respiration factor from Feng et al. 2018 has "too wide a range". I am not familiar with the paper cited, which seems relevant but from a completely different study system.
L 197: Please clarify: you say, "clumping factor is defined as the projected area of leaves per unit ground area" but then the following details about ranging from 0-1 is more about the relative clumping of leaves over a given unit area
L 237: Why do background mortality rate for large stems? Treefall disturbance rate between small and large stems seems very similar - what data is this based on?
L 250-260: I find this part on palm recruitment to be confusing, in part, because there seems to be a disconnect between the model and data/biology of the system regarding palm "seedlings" and recruitment. Indeed, these palms produce abundant seeds and seedling density could perhaps be considered similar to early successional species. However, the decay of palm seedling abundance with time since disturbance is less dramatic than for early successional species, to my knowledge (there is relevant data available on this at least from the LFDP and prior studies). But how does this really relate to the 'observed recruitment of palms' in the data? The text is not clear about what "recruitment" of palms actually means in the data (at what height / diamter do they enter the census?). These are typically not at all "seedlings" since palms produce robust diameter stems prior to growing taller. If the model considers newly recruited individuals as those represented in the data, then it may not really be reasonable to assume that the "seedling" density of palms is similar to that of the early successional PFT. I think some additional details and work revising this section would be valuable.
L 314-317: It is somewhat difficult to assess this with knowledge of the study system without knowing which species are included in each PFT. I am missing a table showing this. L 170 says this information is in Zhang et al. (in review) but seems to important to simple be cited in another paper, especially when that paper is not published yet.
L 319: Is 25% underestimation and 38% overestimation considered "consistent with observations"?
RESULTS
L 327: It would be good to include units and more informative labels on the figure itself. The legend does not seem to define the red line, which it should.L 350: Instead of referring to wood density of Prestoea decurrens, the authors could cite measurements of wood density for the study species directly (0.31 g cm^3), which is available here:
https://datadryad.org/stash/dataset/doi:10.5061/dryad.j2r53L 359-362: Is it also possible that the posterior PDFs do not change much from the priors because of some characteristics / amount of data going into the models? Attributing this fact to some reason seems like more of a discussion point than a result.
L 449: 2 cm yr-1 increment in DBH is extremely high and I don't know where this number comes from? The abstract for the paper cited (Brandeis 2009) says, "...growth rate averaged... 0.36 cm/year in subtropical wet/rain forests, and 0.20 cm/year in lower montane forests." (https://www.fs.usda.gov/treesearch/pubs/34208). This value is also more than 2x higher than the maximum DBH growth rate shown in Figure 10. Something here seems to need clarification.
L 455: Change "the ones" to "the experiments"
L 465-467: The growth rates shown here for late PFT trees are quite a bit higher than what is typical in these forests (see comment about L 449).
DISCUSSIONIn general, this section is very short and it feels like there is a lot of work to be done in terms of putting the pieces together for a robust interpretation of the study results. Also the discussion focuses almost entirely on the modeling exercise but extremely little points back to the biology of the system.
L 486: I would like to see a brief introduction to the discussion section that quickly summarizes the key findings and provides a structure to what we can expect to read in the rest of the section.
L 510-517: RE: clumping factor: It is not clear what are the implications for the clumping factor. The value controls LAI but what does that mean for the simulated dynamics?
L 529: "...vegetation dynamics."
-
RC2: 'Comment on gmd-2021-410', Anonymous Referee #2, 20 Feb 2022
This study examined the impact of hurricane disturbances on tropical forests by use of modeling. Therefore, authors extended the ED2 model by a new disturbance component on hurricanes and an additional PFT of palms, and calibrated the model with the GLUE approach for a forest site in Puerto Rico. With a sensitivity and scenario analysis, authors discussed the uncertainty of their model calibration and demonstrated the impact of forest state and structure before a hurricane event on the recovery of forests. The study is comprehensively conducted and described and the manuscript clearly written. I have a few points to recommend for minor improvement of the manuscript.
General comments:
The relevance of studying hurricane impacts on tropical forests and why modelling is an important tool besides observations should be emphasized more clearly in the introduction, abstract and conclusion. What is your motivation of extending the ED2 model by hurricane disturbances? E.g. in the conclusion (page 20, lines 528-529) you state that no model has implemented hurricane disturbances so far. Please write in more detail about the relevance of such applications. Which benefits can models provide in this context (besides observations)? Few points are mentioned in the conclusion, but the relevance of your study should also be emphasized in the abstract and introduction.
Further, why did you choose the ED2 model? How is it related to other models studying disturbance impacts on tropical forests in general. Please shortly relate your work to the current scientific literature on modeling tropical forests and disturbances in general.
Minor points:
1) page 2, line 31: Can you provide few numbers? How often do they occur on average?
2) section 2.1: I think the general model description of ED2 could benefit from a summarized description of its basic structure. Although you refer to literature references, it is important to have general information on the main processes (recruitment, growth, competition, mortality) also in your manuscript. Especially in section 2.3.2 you mention different mortality sources in the absence of hurricane disturbances and to understand this, already more information in the section on the model description is required.
3) page 5 line 122: Is lambda_d affecting the entire patch or a fraction of a patch?
4) page 6 line 165: Please add the size of a plot. You mention it later in the manuscript (page 9, line 261), but it should already appear here.
5) page 9 line 247: How would this assumption affect your model simulations of long-term studies (e.g. longer than 100 years)?
6) page 17, Fig. 9f: The stem proportion of Mid PFTs seems to still decline (if simulating longer than 112 years; similarly basal area, Fig. 9c) in some scenarios. Nevertheless, you state that the forest reaches a steady state after 80 years (page 18, line 439). How did you determine its steady state? Further, do you have an explanation why Mid PFTs are still declining (in comparison to the other PFTs) and “mostly have small stems“ (page 18, line 443-444)?
Specific comments:
1) page 3, lines 51-54: Difficult to understand in relation to the previous sentences. Can you rephrase? (e.g. what is “the initial vegetation condition”?)
2) check some spelling and grammar in your manuscript, e.g. page 4 (line 92, “Since …”), page 4 (line 108, “They then were to use …”), page 10 (line 285-286), page 10 (line 300, “of” is missing after “impact”), page 12 (line 341, “compated”), page 19 (line 479, “AB”), page 19 (line 499, “utlized”), page 20 (line 518)
3) define and describe variables the first time you mention them in the manuscript (e.g. page 4, line 98, H and DBH should be defined including their units)
-
RC3: 'Comment on gmd-2021-410', Anonymous Referee #3, 22 Feb 2022
Summary
Cyclonic storms are one of the major natural disturbances in tropical forests, and the intensity of tropical cyclones has been projected to increase over this century. Characterizing hurricane damage and post-hurricane recovery is critical for estimating forest resilience and the fate of tropical forests. This study implements a new hurricane module in a dynamic vegetation model, the Ecosystem Demography model (ED), to account for hurricane-caused tree mortality and post-hurricane recovery, which is primarily driven by wind speed, forest structure, and functional diversity. The study also added a new plant functional type for Palms, which can differ from other dicot tropical tree species in terms of ecophysiology and responses to hurricanes. The study performs some model sensitivity tests using GLUE and provides much detailed information on the methodology and results. Altogether, the study highlights the importance of representing the hurricane effect in terrestrial biosphere models.
Comments:
The manuscript provides a comprehensive model calibration and sensitivity analysis within the framework of GLUE. The materials and methodology are clear. Major comments are listed below.
First, the hurricane module is way less discussed in the study compared with functional diversity, and the Palm PFT despite the title focusing more on hurricanes. The method section describes a general framework to include hurricane module (i.e. link hurricane damage to hurricane intensity, forest structure, and species diversity). However, it is not clear what is the uncertainty/biases associated with the framework, which I believe can be large. For example,
- the key relationship in the hurricane module is parameterized by only two points (Fig. 1) and the low hurricane mortality for early successional big trees (Fig. 1b) is somewhat suspicious when the large tree fraction is small.
- Shouldn’t Palms have generally lower mortality compared with other PFTs under hurricanes?
- It is also mentioned that partial crown damage is prevalent under hurricanes, which is not included in this framework and not even discussed.
- What are the key hurricane-related parameters that make the model capture changes in stem density and composition? (Fig. 4)
Given the title, readers would expect some in-depth exploration/discussion of the hurricane module and parameterization. Therefore, I would recommend including more sensitivity tests for the hurricane module or changing the title and intro to focus on Palm PFT.
Second, the GLUE trait optimization seems to be quite sensitive to light-related parameters. For example, the equilibrium clumping factor has a rather low value (< 0.4 while reported values are >0.6 over tropical forests). Quantum efficiency and dark respiration are dominating the variance (Fig.8). I think this might be because the canopy structure and light environment of the model are highly biased. Fig.S2 shows the initial LAI can exceed 8 (constrained by observed demography I guess?), which is rather high. This might explain why optimal Clf is so low and can be caused by biases in allometry (in fact, the allometric parameters can have huge effects but are not tested in the study). Meanwhile, this model does not consider acclimation to understory light. It is understandable that fully addressing these issues is challenging but they need to be acknowledged and discussed.
Third, the hurricane impact and recovery simulations are interesting but are underexplored. Why only look at the impacts on equilibrium forest structure? Shouldn’t the time scale be the average return interval of hurricanes in Puerto Rico? What about using additional initial conditions by sub-sampling different plots?
Minor comments:
Line 55 : the transition from hurricane impact (the previous paragraph) to functional diversity/PFT (this paragraph) seems somewhat abrupt. Some elaboration about why palm is unique, or why we need to incorporate this particular PFT in the context of hurricane disturbance will be helpful, e.g., the relative abundance of palm in hurricane-prone sites. And this information about palm should probably come before the explanation about early and late-successional species (line 58).
Line 69-71: we define a Palm PFT --> there is a need for a separate Palm PFT.
Line 85-86: maybe specify the version of ED2? ED-2.2 if citing Longo et al. 2019.
Line 159: Fig.2 uses time since disturbance to modify external seed rain rate (not seedling density). This assumes the recovery time scale is a constant. Why not use total LAI/BA? Early PFT seed rain can be high when LAI/BA is low but decreases when LAI/BA is high. This would be more ecologically meaningful.
Line 197: the definition of clumping factor is wrong. Should be effective LAI divided by total LAI.
Line 208-210 Any explanation for choosing stem density/DBH growth/BA as target state variables? Why not include mortality? There are large discrepancies between simulated mortality (almost constant across years) and observed mortality (large inter-annual variability) in Fig. S5
Line 328. Fig. 3. Are black dots observations or simulation? No information is provided in the caption. If they are observations like Fig. 4, what are the error bars, cross-plot variance? Also, I wonder how sensitive the results are to the length of training years. What about using half of the period as training?
Citation: Some of the most important information in methodology (such as allometric parameters, line 101) cite studies that are in review or in revision, and the paper only provides minimal information about them. There should be at least a brief description.
-
AC1: 'Author response to referee comments', Jiaying Zhang, 08 Apr 2022
The comment was uploaded in the form of a supplement: https://gmd.copernicus.org/preprints/gmd-2021-410/gmd-2021-410-AC1-supplement.pdf
Status: closed
-
RC1: 'Comment on gmd-2021-410', Anonymous Referee #1, 15 Feb 2022
This paper uses a forest dynamics model (ED2) to assess how hurricane disturbance effects composition and structure of a tropical forest. Key innovations in the work are (1) the development of a plant functional type specifically to represent palms, and (2) the implementation of hurricane disturbance in the ED2 model.Overall, the paper is heavily focused on development of the model (warranted given the venue). However, I did think the biological context and implications could be more thoroughly presented, especially in the intro and discussion. I also felt there could be more reference to previous work on forest dynamics in the Luquillo Mountains (see detailed comments).
Although the manuscript is mainly well-written, there is room for improvement with respect to grammar / language. Please see detailed comments for some of the most important parts. There are numerous references on key points to papers that are in review. I do not know the policy of this journal but some things of the things being referenced are really critical to undestand this paper properly (see detailed comments but, e.g., which species were classified into which PFTs, and on what basis?). I would think that some of these details should be included here in supplementary tables. At minimum, the papers in review should be posted to a pre-print server so the information cited is actually available.
Please note that many of the details on ED2 implementation / testing / calibration are outside of my expertise. Some of my detailed comments below do address these points, especially parts that I found could be more clear. Overall, however, I focus my comments on the general presentation of information in the paper and the links to the biology of the study system.
Thanks for the opportunity to review this manuscript, I hope the authors find my comments useful.
INTRODUCTIONGeneral: Overall, I think this section could be improved by making a more clear introduction to the fact that this study uses a vegetation model to understand hurricane effects on a tropical forest. As it stands, much of the text focuses on effects of hurricanes and then it jumps into PFTs and there is little/no text giving the context / background / rationale / etc. for using vegetation models, including using them to study hurricane effects. This is particularly surprising because there has been other work (including in wet forests of Puerto Rico) on modeling hurricane effects (see, e.g., Uriarte et al. 2009; Natural disturbance and human land use as determinants of tropical forest dynamics: results from a forest simulator). Although not investigating hurricane impacts, another study using ED2 for wet forests in Puerto Rico did include palms as a 'late' PFT (Feng et al. 2017; Improving predictions of tropical forest response to climate change through integration of field studies and ecosystem modeling). It seems strange that this literature is not reviewed in the introduction.
L 35-42: Exposure (to hurricane winds) seems to be an important missing factor here?
L 45: It is not clear what is meant by "faster resprouting" - does it refer to sprouts being generated vs. time since disturbance? It seems to mean something different because the next sentence mentions time since disturbance as a separate point. Please clarify.
L 47: Not clear what is meant by "higher recovery equilibrium". Equilibrium of what? Equilibrium in what sense?
L 51-54: It is a bit odd that this last part of the paragraph highlights a single article (Wang and Eltahir 2000) rather than providing some kind of summary / conclusion point about the preceding paragraph. If the Wang and Eltahir paper is very important to mention specifically then please provide more context.
L 55: The start to this paragraph is abrupt and could be more smoothly linked to the preceeding text. I think the jump is mainly in the fact that all of a sudden you are talking about terrestrial biosphere models and PFTs but none of this has been introduced. A better link would help the overall flow here. The introduction is pretty short as it stands so there is space to develop this more.
L 57-63: More citations for the justification of the 3 PFTs mentioned would be useful.
L 63: I would advocate to start a new paragraph when introducing the palm. This is a focal point of the paper but it is kind of buried in this paragraph.
L 64-65: Also reference:
Uriarte, María, Jill Thompson, and Jess K. Zimmerman. 2019. “Hurricane María Tripled Stem Breaks and Doubled Tree Mortality Relative to Other Major Storms.” Nature Communications 10 (1): 1362. https://www.nature.com/articles/s41467-019-09319-2L 69-70: "To account for these unique characteristics [in what? for what reason?], we define a Palm PFT." As other comments above, I think there needs to be an expansion of the introduction about vegetation modeling - building up to the overall aims of this study.
L 74: Seems strange to say "The results indicate that a single hurricane disturbance has little impact on forest structure" when much of the introduction was spent discussing the various impacts hurricanes have on forest structure / composition. Does this really mean "long-term" forest structure? Please clarify.
METHODS
L 92: "Since there is little knowledge about the traits of Palm." This is not a complete sentence and seems like it should be merged with the previous statement or otherwise revised.L 93: Not all palms have this low "wood density" and I think, in general, you should couch the statements about palms with something like "many palms" since it is such a diverse group and we do not actually know the degree to which these statements might be true or contradicted by some palms. In fact, the range of wood density for palms (Arecaceae) in the global wood density database (Chave et al. 2009, Zanne et al. 2009) is 0.180 - 0.883 (median = 0.54). (checked with the 'wdData' in the 'BIOMASS' R package, v. 2.1.5).
L94-96: It is difficult to assess this decision because there are no details on the other traits used in the model. If palms "grow fast in open canopies like early tropical trees" then what is the reason to assume they "have the same probability distributions as those of late tropical trees"? I am guessing that wood density is strongly related to growth rate in high light conditions in the model. But isn't it also related to mortality rates (including in shade)? Since the introduction of this palm PFT is such a big part of this paper I think it should be explained in more detail here.
L 104-105: Palms can be shorter than other trees, given the same DBH but I am a bit skeptical of these allometric relationships. For one thing, when I plot them, palms can never reach more than about 13 m height at the maximum diameter (20 cm), which is too short but other trees are predicted to reach unreasonably tall heights for this forest (~60 m for early PFT at the maximum 90 cm diameter). The only justification for these fitted parameters is from a paper by the authors 'in revision'.
L 107: This sentence requires a citation.
L 108: This sentence needs revision for clarity / grammar.
L 112-114: I am questioning the ramifications of these 'tricks' implemented in the model to help allow palms to survive despite their allometry.
L 115-117: Here the authors use default allometry of Early PFT for Palms but this seems inconsistent with the statement in L 94-96 about "...we assume that the traits of Palm have the same probability distributions as those of late tropical trees..." Please clarify.
L 124: A bit confused by "(sc) is the ratio of the cohort density that survived to the cohort density *before* the disturbance,"... should this not be the proportion that survives after the disturbance? Since sc=1-λc (L 127), and that λc "varies with hurricane strength, ...", it makes sense that sc would be post-hurricane survival... Please clarify / revise.
L 135: It seems like "Given mortality, the rate of each cohort (λc)" should be revised to "Given the mortality rate of each cohort (λc)"?
L 138-145: It is really not clear what is showing on the x-axis of these figures (x: proportion of large stems). How can this be the same during a given hurricane event for all PFTs? And why is mortality lower for all PFTs from hurricane Maria compared to Hugo? This section needs clarification.
L 146: Again some very relevant references seem to be missing. In particular Uriarte et al. 2009 (Natural disturbance and human land use as determinants of tropical forest dynamics: results from a forest simulator) and Uriarte et al. 2012 (Multidimensional trade-offs in species responses to disturbance: implications for diversity in a subtropical forest).
L 150-160: Please clarify the data upon which these functions are based.
L 170: I am missing details on the basis by which species were assigned to PFTS... this seems too important to have only as cited in a work 'in review'.
L 161-173: It seems that perhaps this description of the census data should go earlier in the text? But more importantly, it is a bit problematic to have such important references to work that is 'in review' (not to mention there are two Zhang et al. 'in review' papers so we don't know for sure which one is being cited here). Perhaps now this paper is published? What is the plan if this manuscript is accepted before the outcome of the one 'in review'? It seems like posting a pre-print of the other work could be at least a partial solution.
L 172: Does the Scatena et al. (1993) biomass allometry apply to palms? Should a caveat be included here?
L 193-194: Clarify specifically the reason why the dark respiration factor from Feng et al. 2018 has "too wide a range". I am not familiar with the paper cited, which seems relevant but from a completely different study system.
L 197: Please clarify: you say, "clumping factor is defined as the projected area of leaves per unit ground area" but then the following details about ranging from 0-1 is more about the relative clumping of leaves over a given unit area
L 237: Why do background mortality rate for large stems? Treefall disturbance rate between small and large stems seems very similar - what data is this based on?
L 250-260: I find this part on palm recruitment to be confusing, in part, because there seems to be a disconnect between the model and data/biology of the system regarding palm "seedlings" and recruitment. Indeed, these palms produce abundant seeds and seedling density could perhaps be considered similar to early successional species. However, the decay of palm seedling abundance with time since disturbance is less dramatic than for early successional species, to my knowledge (there is relevant data available on this at least from the LFDP and prior studies). But how does this really relate to the 'observed recruitment of palms' in the data? The text is not clear about what "recruitment" of palms actually means in the data (at what height / diamter do they enter the census?). These are typically not at all "seedlings" since palms produce robust diameter stems prior to growing taller. If the model considers newly recruited individuals as those represented in the data, then it may not really be reasonable to assume that the "seedling" density of palms is similar to that of the early successional PFT. I think some additional details and work revising this section would be valuable.
L 314-317: It is somewhat difficult to assess this with knowledge of the study system without knowing which species are included in each PFT. I am missing a table showing this. L 170 says this information is in Zhang et al. (in review) but seems to important to simple be cited in another paper, especially when that paper is not published yet.
L 319: Is 25% underestimation and 38% overestimation considered "consistent with observations"?
RESULTS
L 327: It would be good to include units and more informative labels on the figure itself. The legend does not seem to define the red line, which it should.L 350: Instead of referring to wood density of Prestoea decurrens, the authors could cite measurements of wood density for the study species directly (0.31 g cm^3), which is available here:
https://datadryad.org/stash/dataset/doi:10.5061/dryad.j2r53L 359-362: Is it also possible that the posterior PDFs do not change much from the priors because of some characteristics / amount of data going into the models? Attributing this fact to some reason seems like more of a discussion point than a result.
L 449: 2 cm yr-1 increment in DBH is extremely high and I don't know where this number comes from? The abstract for the paper cited (Brandeis 2009) says, "...growth rate averaged... 0.36 cm/year in subtropical wet/rain forests, and 0.20 cm/year in lower montane forests." (https://www.fs.usda.gov/treesearch/pubs/34208). This value is also more than 2x higher than the maximum DBH growth rate shown in Figure 10. Something here seems to need clarification.
L 455: Change "the ones" to "the experiments"
L 465-467: The growth rates shown here for late PFT trees are quite a bit higher than what is typical in these forests (see comment about L 449).
DISCUSSIONIn general, this section is very short and it feels like there is a lot of work to be done in terms of putting the pieces together for a robust interpretation of the study results. Also the discussion focuses almost entirely on the modeling exercise but extremely little points back to the biology of the system.
L 486: I would like to see a brief introduction to the discussion section that quickly summarizes the key findings and provides a structure to what we can expect to read in the rest of the section.
L 510-517: RE: clumping factor: It is not clear what are the implications for the clumping factor. The value controls LAI but what does that mean for the simulated dynamics?
L 529: "...vegetation dynamics."
-
RC2: 'Comment on gmd-2021-410', Anonymous Referee #2, 20 Feb 2022
This study examined the impact of hurricane disturbances on tropical forests by use of modeling. Therefore, authors extended the ED2 model by a new disturbance component on hurricanes and an additional PFT of palms, and calibrated the model with the GLUE approach for a forest site in Puerto Rico. With a sensitivity and scenario analysis, authors discussed the uncertainty of their model calibration and demonstrated the impact of forest state and structure before a hurricane event on the recovery of forests. The study is comprehensively conducted and described and the manuscript clearly written. I have a few points to recommend for minor improvement of the manuscript.
General comments:
The relevance of studying hurricane impacts on tropical forests and why modelling is an important tool besides observations should be emphasized more clearly in the introduction, abstract and conclusion. What is your motivation of extending the ED2 model by hurricane disturbances? E.g. in the conclusion (page 20, lines 528-529) you state that no model has implemented hurricane disturbances so far. Please write in more detail about the relevance of such applications. Which benefits can models provide in this context (besides observations)? Few points are mentioned in the conclusion, but the relevance of your study should also be emphasized in the abstract and introduction.
Further, why did you choose the ED2 model? How is it related to other models studying disturbance impacts on tropical forests in general. Please shortly relate your work to the current scientific literature on modeling tropical forests and disturbances in general.
Minor points:
1) page 2, line 31: Can you provide few numbers? How often do they occur on average?
2) section 2.1: I think the general model description of ED2 could benefit from a summarized description of its basic structure. Although you refer to literature references, it is important to have general information on the main processes (recruitment, growth, competition, mortality) also in your manuscript. Especially in section 2.3.2 you mention different mortality sources in the absence of hurricane disturbances and to understand this, already more information in the section on the model description is required.
3) page 5 line 122: Is lambda_d affecting the entire patch or a fraction of a patch?
4) page 6 line 165: Please add the size of a plot. You mention it later in the manuscript (page 9, line 261), but it should already appear here.
5) page 9 line 247: How would this assumption affect your model simulations of long-term studies (e.g. longer than 100 years)?
6) page 17, Fig. 9f: The stem proportion of Mid PFTs seems to still decline (if simulating longer than 112 years; similarly basal area, Fig. 9c) in some scenarios. Nevertheless, you state that the forest reaches a steady state after 80 years (page 18, line 439). How did you determine its steady state? Further, do you have an explanation why Mid PFTs are still declining (in comparison to the other PFTs) and “mostly have small stems“ (page 18, line 443-444)?
Specific comments:
1) page 3, lines 51-54: Difficult to understand in relation to the previous sentences. Can you rephrase? (e.g. what is “the initial vegetation condition”?)
2) check some spelling and grammar in your manuscript, e.g. page 4 (line 92, “Since …”), page 4 (line 108, “They then were to use …”), page 10 (line 285-286), page 10 (line 300, “of” is missing after “impact”), page 12 (line 341, “compated”), page 19 (line 479, “AB”), page 19 (line 499, “utlized”), page 20 (line 518)
3) define and describe variables the first time you mention them in the manuscript (e.g. page 4, line 98, H and DBH should be defined including their units)
-
RC3: 'Comment on gmd-2021-410', Anonymous Referee #3, 22 Feb 2022
Summary
Cyclonic storms are one of the major natural disturbances in tropical forests, and the intensity of tropical cyclones has been projected to increase over this century. Characterizing hurricane damage and post-hurricane recovery is critical for estimating forest resilience and the fate of tropical forests. This study implements a new hurricane module in a dynamic vegetation model, the Ecosystem Demography model (ED), to account for hurricane-caused tree mortality and post-hurricane recovery, which is primarily driven by wind speed, forest structure, and functional diversity. The study also added a new plant functional type for Palms, which can differ from other dicot tropical tree species in terms of ecophysiology and responses to hurricanes. The study performs some model sensitivity tests using GLUE and provides much detailed information on the methodology and results. Altogether, the study highlights the importance of representing the hurricane effect in terrestrial biosphere models.
Comments:
The manuscript provides a comprehensive model calibration and sensitivity analysis within the framework of GLUE. The materials and methodology are clear. Major comments are listed below.
First, the hurricane module is way less discussed in the study compared with functional diversity, and the Palm PFT despite the title focusing more on hurricanes. The method section describes a general framework to include hurricane module (i.e. link hurricane damage to hurricane intensity, forest structure, and species diversity). However, it is not clear what is the uncertainty/biases associated with the framework, which I believe can be large. For example,
- the key relationship in the hurricane module is parameterized by only two points (Fig. 1) and the low hurricane mortality for early successional big trees (Fig. 1b) is somewhat suspicious when the large tree fraction is small.
- Shouldn’t Palms have generally lower mortality compared with other PFTs under hurricanes?
- It is also mentioned that partial crown damage is prevalent under hurricanes, which is not included in this framework and not even discussed.
- What are the key hurricane-related parameters that make the model capture changes in stem density and composition? (Fig. 4)
Given the title, readers would expect some in-depth exploration/discussion of the hurricane module and parameterization. Therefore, I would recommend including more sensitivity tests for the hurricane module or changing the title and intro to focus on Palm PFT.
Second, the GLUE trait optimization seems to be quite sensitive to light-related parameters. For example, the equilibrium clumping factor has a rather low value (< 0.4 while reported values are >0.6 over tropical forests). Quantum efficiency and dark respiration are dominating the variance (Fig.8). I think this might be because the canopy structure and light environment of the model are highly biased. Fig.S2 shows the initial LAI can exceed 8 (constrained by observed demography I guess?), which is rather high. This might explain why optimal Clf is so low and can be caused by biases in allometry (in fact, the allometric parameters can have huge effects but are not tested in the study). Meanwhile, this model does not consider acclimation to understory light. It is understandable that fully addressing these issues is challenging but they need to be acknowledged and discussed.
Third, the hurricane impact and recovery simulations are interesting but are underexplored. Why only look at the impacts on equilibrium forest structure? Shouldn’t the time scale be the average return interval of hurricanes in Puerto Rico? What about using additional initial conditions by sub-sampling different plots?
Minor comments:
Line 55 : the transition from hurricane impact (the previous paragraph) to functional diversity/PFT (this paragraph) seems somewhat abrupt. Some elaboration about why palm is unique, or why we need to incorporate this particular PFT in the context of hurricane disturbance will be helpful, e.g., the relative abundance of palm in hurricane-prone sites. And this information about palm should probably come before the explanation about early and late-successional species (line 58).
Line 69-71: we define a Palm PFT --> there is a need for a separate Palm PFT.
Line 85-86: maybe specify the version of ED2? ED-2.2 if citing Longo et al. 2019.
Line 159: Fig.2 uses time since disturbance to modify external seed rain rate (not seedling density). This assumes the recovery time scale is a constant. Why not use total LAI/BA? Early PFT seed rain can be high when LAI/BA is low but decreases when LAI/BA is high. This would be more ecologically meaningful.
Line 197: the definition of clumping factor is wrong. Should be effective LAI divided by total LAI.
Line 208-210 Any explanation for choosing stem density/DBH growth/BA as target state variables? Why not include mortality? There are large discrepancies between simulated mortality (almost constant across years) and observed mortality (large inter-annual variability) in Fig. S5
Line 328. Fig. 3. Are black dots observations or simulation? No information is provided in the caption. If they are observations like Fig. 4, what are the error bars, cross-plot variance? Also, I wonder how sensitive the results are to the length of training years. What about using half of the period as training?
Citation: Some of the most important information in methodology (such as allometric parameters, line 101) cite studies that are in review or in revision, and the paper only provides minimal information about them. There should be at least a brief description.
-
AC1: 'Author response to referee comments', Jiaying Zhang, 08 Apr 2022
The comment was uploaded in the form of a supplement: https://gmd.copernicus.org/preprints/gmd-2021-410/gmd-2021-410-AC1-supplement.pdf
Jiaying Zhang et al.
Jiaying Zhang et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 676 | 107 | 21 | 804 | 29 | 7 | 5 |
- HTML: 676
- PDF: 107
- XML: 21
- Total: 804
- Supplement: 29
- BibTeX: 7
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1