the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies
David M. Medvigy
Benjamin Smith
Jeffrey S. Dukes
Claus Beier
Mikhail Mishurov
Xiangtao Xu
Jeremy W. Lichstein
Craig D. Allen
Klaus S. Larsen
Cari Ficken
William T. Pockman
William R. L. Anderegg
Anja Rammig
Abstract. Climatic extreme events are expected to occur more frequently in the future, increasing the likelihood of unprecedented climate extremes (UCEs), or record-breaking events. UCEs, such as extreme heatwaves and droughts, substantially affect ecosystem stability and carbon cycling by increasing plant mortality and delaying ecosystem recovery. Quantitative knowledge of such effects is limited due to the paucity of experiments focusing on extreme climatic events beyond the range of historical experience. Here, we use two dynamic vegetation demographic models (VDMs), ED2 and LPJ-GUESS, to investigate the hypothesis that ecosystem responses to UCEs (e.g., unprecedented droughts) differ qualitatively from ecosystem responses to milder extremes, as a result of non-linear ecosystem responses. Additionally, we explore how unprecedented droughts in combination with increasing atmospheric CO2 and/or temperature may affect ecosystem stability and carbon cycling. We explored these questions using simulations of pre-drought and post-drought conditions at well-studied forest sites in Australia and Costa Rica. Both models produced nonlinear responses to UCEs. Due to the two models having different but plausible representations of processes and interactions, they diverge in sensitivity of biomass loss due to drought duration or intensity, and differ between each site. Biomass losses are most sensitive to drought duration in ED2, but to drought intensity in LPJ-GUESS. Elevated atmospheric CO2 concentrations (eCO2) alone did not completely buffer the ecosystems from carbon losses during UCEs in the majority of our simulations. Our findings highlight contrasting differences in process formulations and uncertainties in models, notably related to availability in plant carbohydrate storage and the diversity of plant hydraulic schemes, in projecting potential ecosystem responses to UCEs. The different hypotheses of plant responses to UCEs existing in models reflect knowledge gaps, which should be tested with targeted field experiments. This iterative modeling-experimental framework would help improve predictions of terrestrial ecosystem responses and climate feedbacks.
- Preprint
(2208 KB) -
Supplement
(737 KB) - BibTeX
- EndNote
Jennifer A. Holm et al.
Status: closed
-
RC1: 'Comment on bg-2022-65', Anonymous Referee #1, 06 Jun 2022
Manuscript by Holm et al. titled "Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies" explores the responses of two dynamic vegetation demographic models (VDMs) to a series of unprecedented climate extremes (UCEs) scenarios that are designed to test two high-level hypotheses on the nature of ecosystem responses to UCEs (mainly in the form of droughts, with additional effects of increasing atmospheric CO2 [eCO2] and temperature). The study finds strong nonlinearities to extreme droughts in the responses of the two VDMs with different sensitivities to drought intensity and duration scenarios at the two test sites, and attempts to interpret the underlying mechanisms of such responses.General comments:The manuscript is in itself coherent, well-written, the structure is easy to follow with the aid of straightforward visualizations and tables. The experimental framework is well-thought-out in general although it is missing a couple of important aspects (more on that soon). It was interesting to see such different responses of models so clearly which I believe would lead to novel follow-up studies.My main concern about this manuscript is how much novelty it brings in itself. If I may put it this way, one could compile Table 3 (the crux of the study) from the literature as authors themselves are partly doing here. Surely, the framework provides a systematic structure to highlight contrasting differences in model process formulations and bring these suggestions (Table 3) together, but after reading the manuscript I was left with a feeling that what genuinely new was learnt from this exercise was not clear. In other words, authors seem to be exposing mainly the known unknowns, and the link between some of the suggestions and simulations in this study seems rather weak. It was still a very informative read if it was meant to be a literature review paper, but as a research paper I wonder if it is missing the opportunity of putting its potential [L703: to drive progress in improving our understanding of terrestrial ecosystem responses to UCEs] into action.To be more specific, while I acknowledge the fact that one paper can only tackle so much, I think this study would benefit from additional simulation experiments to explore one or more of the raised questions, and try to disentangle some of the potential driving mechanisms further. Below are examples of such "low-hanging fruits":i) Additional no drought (truly) eCO2-only simulations to inspect its role in mortality overshoot some more (L496-498)ii) Running ED2 with less PFTs at Palo Verde (L550-551)iii) Imposing additional mortality to emulate post-drought secondary stressors' effects (L613-615) - I believe at least ED2 has a stop-restart functionality and the disturbance interval in LPJ-GUESS can be played with.iv) Testing different mortality thresholds from conductivity loss in ED2 (L653-657)I am sure that the authors are capable of devising much more clever ones than these. Even if some of these configurations are unrealistic and outcome from such exercises are still inconclusive, the findings could be interesting and it would add a valuable dimension to this manuscript.That said, I think there is no fundamental flaw in the study, and it is of interest for the Biogeosciences community. I would be happy to reevaluate after the revisions.Specific comments:As mentioned above, the presented framework is practical and logical but I was wondering whether the following points are obscuring the results:1) While no model-data comparisons can be performed on UCE responses due to lack of data (L228), there is no (mention of) validation as to whether these models were able to capture the modern vegetation at the test sites. Please cite the studies if this has been done elsewhere.2) Similarly, authors don't use any site-level observations to inform model parameterization (L290 no site-level tuning was conducted). At the least, this needs to be justified more.Without some reliability that the model simulations are representative of the sites, it is not clear how much of the responses could be due to misconfiguration, even if the scenarios are being evaluated with respect to a baseline.3) The uncertainties in parameters and models' sensitivities to them are not accounted for. I am sure the authors are aware of such uncertainties and sensitivities (e.g. sensitivity to water- and mortality-related parameters in LPJ-GUESS, Oberpriller et al., 2022 doi: 10.5194/gmd-2021-287 and in ED2, Raczka et al. 2018 doi: 10.1029/2018JG004504). I believe the findings would have been more robust if the models were run in ensembles and results were reported using the ensemble means at least.Please find more line-by-line comments below:L62 - frequency: While this study has enough to investigate under the existing framework, one possible extension that comes to mind is to include the increase in frequency. Authors consider magnitude and duration but not the frequency of back-to-back UCEs. Please consider coming back to it in the discussion, potentially in an explicit "limitations of the current study and future steps" sub-section.L77: Could the authors expand on the definition of 'moderate' quantitatively here? I.e. what were the treatments like in the mentioned studies, e.g. Beier et al X months, Kayler et al. Y amountL86-87: Authors are not coming back to these studies that document ecosystem responses to extreme droughts later in the text. Please consider including discussion on how your findings compare to these studies (even if they are post-hoc and limited).L152-154: While there is some observational support for linear response as cited by the authors, I wonder if expecting a linear response from these highly non-linear VDMs forms a plausible null-hypothesis as it sounds rather easy to refute (Sitch et al. 2008 GCB, McDowel et al., 2013 New Phyt, Rollinson et al. 2017 GCB, Bastos et al. 2020 Phil. Trans. R. Soc. B, Oberpriller et al. 2022 GMD) Could the authors elaborate with additional justifications in the text for not choosing a more specific null hypothesis?L207: Again, it would be great if authors could consider giving numbers as to what makes these studies 'moderate' to motivate the readers.L243: Can't one specify more PFTs or even species in LPJ-GUESS?L288: Just curious, were there no droughts during these historical site-specific climate periods?L291: Will the parameters sets be also made available in the Dryad repository? (the data availability statement says all model simulation data will be made available but I wanted to make sure all model configurations are also included)L305: On table S1 - Are the "drought + temp + 200 ppm eCO2" rows missing?L352: Looking at Fig 2 LPJ-GUESS seems to show some differences to drought duration, maybe consider saying "very little" instead of "no sensitivity".L404-406: This is an example of what I meant above by robustness of the results. Here authors identify it as a potentially anomalous signal because it was unexpected, but some of the other effects we are seeing could also be "anomalous", and vice versa, we may not be able to see certain effects because parameters were not varied in their potential ranges, limiting the overall impact of the study.L457-459: While authors mention that these swings in LPJ-GUESS LAI could be contributing to mortality response they do not explain the reason for these swings in LPJ-GUESS. A historical validation/assessment of the model behaviour at the site could have been useful here.L459-463: The reasoning behind how the model simulations lead to the suggestion of better representing morphological and physiological characteristics relevant to plant-water relations (e.g. leaf age) is not very obvious. Could the authors add some connecting and specific thoughts to the text?L480-483: Again the transition from simulations to the suggestion feels rather abrupt and generic. This recommendation could easily have come from the cited literature without performing this study at all, what else was learnt from the simulations?In sub-section 4.1.3 there is no referral to the process implementations of the models used in the study specifically. The only connection back to the results is the last sentence of the first paragraph, the rest of this subsection mostly reads like an introduction. Proposal to investigate allometric partitioning theory is not explicitly motivated by the simulations but rather by literature studies. Please consider re-wording this section and making the connections to this study more specific.L532-547: Could the authors say anything about which model implementation is more realistic according to their simulations? Can there be some more concrete conclusions/suggestions than "the need to better understand NSC dynamics"?L551-553: As mentioned in the beginning, it would have been nice to see some additional experiment as an attempt to disentangle whether ED2's resistance is due to functional trait diversity (or due to e.g. NSC implementation).L596: Improved how? What are specifically lacking in ED2 and LPJ-GUESS at the moment?L643: The thought process seems incomplete in this section. Was there going to be a suggestion?L654: Again, it feels like a missed opportunity that there were no additional simulation experiments with different thresholds to investigate this further.Table 3 - last row: Not sure if I'm following right but is the authors' recommendation here is to collect data on UCEs which they identified, by definition, to be unavailable? Please consider rewording.L708: As far as I can tell, the LPJ-GUESS code is not publicly available to download via this link, but available upon request. Please check and revise the statement accordingly if needed.Citation: https://doi.org/
10.5194/bg-2022-65-RC1 -
AC1: 'Reply on RC1', Jennifer A. Holm, 31 Jul 2022
Dear Biogeosciences Reviewer,
We appreciate your review and comments on the following paper “Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies”. Your time spent on this peer-review process is appreciated. There are valid points raised, and others that we help to clarify. Below is a point-by-point response to the reviews.
In general, we think clearly stating at the beginning of the manuscript that this study was not intended to solely be an original research paper, but rather a review of how models currently capture unprecedented climate extremes on forest ecosystems by using a handful of novel model results, will alleviate many of the reviewers' concerns. In order to fully investigate responses to such novel climate extremes that have not been experienced, we had to generate original model results which we believe is a strong novelty of this paper.
R1 has concerns about:
1. R1: The novelty of the paper; and whether this paper was a review paper or a research paper.
Answer: We believe the strong novelty of our study is the assessment of unprecedented climate extremes (UCEs) and their impacts. In the revised manuscript, we will improve the text about tighter linkages between concepts, hypotheses, and model outcomes. For example, we want to emphasize that a goal of this paper is to demonstrate how to use these vegetation demographic models (VDMs) to help generate future hypotheses about UCEs. Therefore, we used the models and sites as conceptual “experimental” tools to investigate the given hypotheses.
We will also revise the manuscript to state more clearly at the beginning that this is more of a review and “guidebook” manuscript, but with a combination of original research in order to address novel forest responses to climate extremes. We believe this combination of original research with review of current limitations of models is a strong novelty in and of itself.
We would like to clarify that the original “research” aspects of this study were the new UCE model results, and specifically the integrated C-loss with sensitivity to different climate change treatments which has not been done with these models, and is a knowledge gap that was filled by this study. The discussion section and detailed tables are designed to be a “review” and to help guide future research.
2. R1: While one paper can only tackle so much, R1 thinks this study would benefit from additional simulation experiments.
Answer: The scope of our paper was not to do many, in-depth, detailed experiments. But rather focus on how models currently capture UCEs in extreme durations/lengths and interactions with climate change. Therefore, we are not fully sure if there would be a strong benefit to adding in additional simulations experiments, which would be a large, costly endeavor and make the already very long manuscript even longer.
Probably the strongest point to not adding more simulations, is that additional simulations as suggested by R1, by itself, may not lead to the novelty and clarity that R1 seeks. We plan to re-write the manuscript to state that outstanding modeling perturbations and experiments are for subsequent studies. The initial modeling investigation here was to highlight how VDMs (as opposed to typical LSMs) could be used to answer hypotheses and guide future studies.
3. R1: Lack of model validation at the two sites, and site-level observations to inform model parameterization.
Answer: We agree that model comparison to some site-level observations is a worthy improvement to the paper. Upon revision of the manuscript, we will include site level benchmarking observations such as biomass, stand density, and leaf area index to do basic level model validation. We will also add a sentence that both models have already been run and validated at these sites in previously published papers (Xu et al. (2016); Medlyn et al. 2016), thus making model application possible with a “built in” reasonable degree of validation. These references were included in the original manuscript, but we failed to make a clear connection that they were publications that validated the models at the two sites.
4. R1: Reporting uncertainties in model parameters.
Answer: The ED2 and LPJ-GUESS models are well documented and investigated VDMs, with many previous studies that have looked into parameter uncertainty. Upon revision of this manuscript we will include a general description of the model parameters with the largest uncertainties. For example we know that parameters related to plant hydraulics and non-structural carbohydrate storage have large uncertainties and thus included these parameters in Table 1.
However, the aim of the paper was not to do in-depth, detailed experiments with tuned parameters specific to each site. But rather set up a general modeling framework so that hypotheses about unprecedented climate extremes could be investigated, and provide understanding on how model behavior of physiological and ecological processes might be lacking in state-of-the-srt ecosystem models in order to capture extremes.
We were glad to hear that R1 thought the paper was a “very informative read” and that “the presented framework is practical and logical”. As well as “the manuscript is in itself coherent, well-written, the structure is easy to follow with the aid of straightforward visualizations and tables.”
Citation: https://doi.org/10.5194/bg-2022-65-AC1 -
AC3: 'Reply on AC1', Jennifer A. Holm, 05 Jan 2023
Dear Reviewer,
We realized that some reviewer comments were not fully addressed from the last review period, and we would like to address these outstanding comments here. Thanks, and sorry for this oversight. The last manuscript version has been updated to include these new changes based on the reviewer comments.
RC: While this study has enough to investigate under the existing framework, one possible extension that comes to mind is to include the increase in frequency. Authors consider magnitude and duration but not the frequency of back-to-back UCEs. Please consider coming back to it in the discussion, potentially in an explicit "limitations of the current study and future steps" sub-section.
Author Answer: We agree it is important to include that back-to-back UCEs were not explored in this manuscript, and should be considered in future studies. In the summary section we included the following sentence:
“For example, we begin to investigate duration of droughts but we did not consider frequency of back-to-back UCEs.”
RC: While there is some observational support for linear response as cited by the authors, I wonder if expecting a linear response from these highly non-linear VDMs forms a plausible null-hypothesis as it sounds rather easy to refute.
Author Answer: We agree that complex biological life systems are not linear, and that VDMs are also highly nonlinear. The correct hypotheses that we were trying to test was comparing different degrees, or amplitudes, of nonlinearities, with carbon loss becoming more strongly nonlinear with increasing UCEs. Therefore, we removed text that described the null hypothesis as a linear relationship between carbon stock and drought, and instead are describing the null hypothesis and near-linear, and alternative hypothesis as different degrees of non-linearities.
RC: Could the authors expand on the definition of 'moderate' quantitatively here?
Author Answer: In the revised manuscript we have updated the text to say: “However, the majority of such experiments apply moderate treatments based on a historical sense, which are mostly weaker in intensity and/or shorter in duration than potential future UCEs”
“…take into account only low to moderate drought intensities (such as 50% rain excluded) or single events, or combine drought with moderate effects of temperature change. Where there has been 100% rain exclusion, it was on very small plots of 1.5 m2 (Meir et al., 2015).”
RC: Looking at Fig 2 LPJ-GUESS seems to show some differences to drought duration, maybe consider saying "very little" instead of "no sensitivity".
Author Answer: We have updated the text to say LPJ-GUESS very little sensitivity to drought duration, instead of no sensitivity.
RC: The reasoning behind how the model simulations lead to the suggestion of better representing morphological and physiological characteristics relevant to plant-water relations (e.g. leaf age) is not very obvious. Could the authors add some connecting and specific thoughts to the text?
Author Answer: In the discussion section of the revised manuscript we included a sentence that describes that leaf age classifications could be used in models to provide distinctions in the variations of leaf productivity and turnover time, which help to determine the balance between carbon gains and losses.
Citation: https://doi.org/10.5194/bg-2022-65-AC3
-
AC3: 'Reply on AC1', Jennifer A. Holm, 05 Jan 2023
-
AC1: 'Reply on RC1', Jennifer A. Holm, 31 Jul 2022
-
RC2: 'Comment on bg-2022-65', Anonymous Referee #2, 17 Jun 2022
“Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies” by Holm et al.
This study investigates carbon dynamics in response to unprecedented climate extremes (UCEs), specifically droughts as well as in combination with warming and CO2 fertilization, using two dynamic vegetation demographic models (VDMs), ED2 and LPJ-GUESS. It illustrates 1) conceptual frameworks and hypotheses on potential ecosystem response to UCEs, 2) VDM simulations to present potential ecosystem response, and finally 3) the limits of the current understanding of these responses. It is clearly written and will be an interesting topic to the readers of Biogeosciences. However, I fear that some important components are missing to make this study complete or original. For example, it is not clear whether the main focus of this study is to introduce the newest model development or review what is missing in the current VDMs and suggest future directions, or even the combination of both. In any case, model simulations need to be compared with the observation/data, at least partially, to describe the limitation of the current VDMs.
If process-based modeling is a great tool for studying the response of ecosystems to extreme drought events, which cannot be easily conducted with in-situ manipulation experiment (line90), I expect optimized or calibrated model simulations for specific study sites using historic drought events for the justification of the projections and to define the limitations of the current VDMs. Even if UCEs haven’t occurred at the study site (line228), validating the model with some historic drought events (that the models are sensitive to drought events, and the directions and magnitudes of the responses are reasonable) will be necessary to support the conclusion. Of course, it is not easy to track the effects of a certain drought event in reality because not only single event occurs during a certain time period, while controlling other factors. In addition, ecosystem responses to extreme droughts will be different from those to less intense droughts as the responses can be nonlinear. Yet, providing the evidence that the model performance is reasonable is critical. As one example, LPJ-GUESS showed large swings in LAI at EucFACE site (line476) – is it reasonable? If not, is it because of some parameters too sensitive specifically for this site or the model structure that needs to be modified? In this regard, another question arises: why two specific study sites are selected? Models were not tuned specifically for each site (line290), and comparison between the model simulation and observation was not made because data lack (line228; no observations in C change or no drought events?). If the model is not tuned for each site because of the lacking data, how about selecting other sites, where observations are available, at least for aboveground C? Or can any observations from FACE experiment be used for validating the model, such as the response to CO2 fertilization? Adding this information and validating the model simulations will make the results more plausible and this study very unique.
In the discussion, what is missing in the current VDMs are thoroughly reviewed. It can be nicer if there was a tighter connection between the simulation results and discussion by providing more detailed descriptions of the two models. Some parts are discussed based on the simulation results but mostly shortcomings are written based on literatures, which can be done without modeling exercises. Providing more details will make the story more coherent, what the current VDMs can do based on the simulation results, what are still missing, and what needs to be done. Another important point to address is, despite multiple missing processes in the current VDMs: how the results of the VDM simulations are different from conventional ESM simulations?
Some other comments:
Line77: ‘moderate treatments’ is vague. How can you differentiate moderate drought from unprecedented drought? Also, where can you draw the line between moderate and unprecedented droughts in the figures?
Line213: can you provide climatology of the sites, e.g., what are the mean and interannual variability of temperature and precipitation?
Line434: ‘multiple mechanisms that vary among ecosystem types’ – if the model was not tuned for each site, what drives the variations between the two sites, climate or the number of PFTs?
Line483: ‘further explored’ can be more specific or explored in this study
Citation: https://doi.org/10.5194/bg-2022-65-RC2 -
AC2: 'Reply on RC2', Jennifer A. Holm, 31 Jul 2022
Dear Biogeosciences Reviewer,
We appreciate your review and comments on the following paper “Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies”. Your time spent on this peer-review process is appreciated. There are valid points raised, and others that we help to clarify. Below is a point-by-point response to the reviews.
In general, we think clearly stating at the beginning of the manuscript that this study was not intended to solely be an original research paper, but rather a review of how models currently capture unprecedented climate extremes on forest ecosystems by using a handful of novel model results, will alleviate many of the reviewers' concerns. In order to fully investigate responses to such novel climate extremes that have not been experienced, we had to generate original model results which we believe is a strong novelty of this paper.
R2 has concerns about:
- R2: It is not clear whether the main focus of this study is to introduce the newest model development or review what is missing in the current VDMs and suggest future directions, or even the combination of both.
Answer: We will rewrite the manuscript to make it clearer that the goal of this paper is to highlight what is missing in current VDMs, even after taking into account cutting edge model developments. In order to review what is lacking in current VDMs, we needed to describe the existing model frameworks and latest model developments (which might have been a little confusing). We will make this more clear in the revision.
We will also revise the manuscript to state more clearly at the beginning that this is more of a review and “guidebook” manuscript, but with a combination of an original research in order to address novel forest responses to climate extremes. We believe this combination of original research with review of current limitations of models is a strong novelty in and of itself.
2. R2: Lack of model validation and comparisons with the observation/data, and validating against some historic drought events at these sites.
Answer: We agree that model comparison to some site-level observations is a worthy improvement to the paper. Upon revision of the manuscript, we will include site level benchmarking observations such as biomass, stand density, and leaf area index to do basic level model validation. With regards to validating the models against historic drought events at these sites, as also pointed out by reviewer #2, it is not easy to track the effects of a drought event in reality when these events can be rare, field campaigns didn’t occur during the event, all while controlling for other factors. The goal of this paper was to investigate ecosystem responses to extreme droughts that will likely be occurring in the future, and how they are different from less intense droughts as the responses can be nonlinear.
3. R2: LPJ-GUESS showed large swings in LAI at EucFACE site, and R2 is wondering if this is reasonable?
Answer: In the revised manuscript, we will work to provide observational data of leaf area index (LAI) at the EucFACE site to compare against the large swings from the LPJ-GUESS model.
4. R2: Question about why these two specific study sites are selected? (And why no site site level tuning was done).
Answer: We choose these two sites (the Australian and Costa Rica sites) mainly because the models have already been run at these sites in previously published papers (Xu et. al., 2016; Medlyn et. al., 2016), thus making model application possible with a “built in” reasonable degree of validation. These sites also contain multiple measurements and information that allowed for previous papers to validate the models. These sites were also chosen because they span an interesting range of vegetation types (a temperate-subtropical transitional forest in EucFACE and a seasonally dry tropical forest at PaloVerd), and are in warm, seasonally dry climates that are more likely to experience droughts in the future. In the revised manuscript we will include a description of why the sites were chosen, reference the publications where model validation has already occurred, and include this in Section 2.2 where we already initially described the sites. We will also emphasize that the purpose of this paper wasn’t to do site comparisons between many different sites, but just select a few for hypothesis testing.
5. R2: Would like to see a tighter connection between the simulation results and discussion section which describes what is lacking in VDMs, by providing more detailed descriptions of the two models.
Answer: In the revised manuscript, we will improve the text about tighter linkages between concepts, hypotheses, and model outcomes. For example, we want to make better connections that large losses in carbon were likely linked to uncertainties in how we currently represent plant hydraulics, non-structural carbohydrate storage, or phenology diversity, which we then reviewed in the discussion section. We will also revise the manuscript to emphasize that a goal of this paper is to demonstrate how to use these VDMs in order to help generate future hypotheses about UCEs. Therefore, we used the models and sites as conceptual “experimental” tools to investigate the given hypotheses. In the supplemental material we will do a better job of referencing previous papers that have done more through investigations of specific model processes within each of the models, and listing which processes or mechanisms were evaluated and why.
We are glad to hear that R2 thought the paper was “clearly written and will be an interesting topic to the readers of Biogeosciences.” As well as, “In the discussion, what is missing in the current VDMs are thoroughly reviewed”, which as we listed above in our answer to concern #1 is the goal of this paper.
Citation: https://doi.org/10.5194/bg-2022-65-AC2
-
AC2: 'Reply on RC2', Jennifer A. Holm, 31 Jul 2022
Status: closed
-
RC1: 'Comment on bg-2022-65', Anonymous Referee #1, 06 Jun 2022
Manuscript by Holm et al. titled "Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies" explores the responses of two dynamic vegetation demographic models (VDMs) to a series of unprecedented climate extremes (UCEs) scenarios that are designed to test two high-level hypotheses on the nature of ecosystem responses to UCEs (mainly in the form of droughts, with additional effects of increasing atmospheric CO2 [eCO2] and temperature). The study finds strong nonlinearities to extreme droughts in the responses of the two VDMs with different sensitivities to drought intensity and duration scenarios at the two test sites, and attempts to interpret the underlying mechanisms of such responses.General comments:The manuscript is in itself coherent, well-written, the structure is easy to follow with the aid of straightforward visualizations and tables. The experimental framework is well-thought-out in general although it is missing a couple of important aspects (more on that soon). It was interesting to see such different responses of models so clearly which I believe would lead to novel follow-up studies.My main concern about this manuscript is how much novelty it brings in itself. If I may put it this way, one could compile Table 3 (the crux of the study) from the literature as authors themselves are partly doing here. Surely, the framework provides a systematic structure to highlight contrasting differences in model process formulations and bring these suggestions (Table 3) together, but after reading the manuscript I was left with a feeling that what genuinely new was learnt from this exercise was not clear. In other words, authors seem to be exposing mainly the known unknowns, and the link between some of the suggestions and simulations in this study seems rather weak. It was still a very informative read if it was meant to be a literature review paper, but as a research paper I wonder if it is missing the opportunity of putting its potential [L703: to drive progress in improving our understanding of terrestrial ecosystem responses to UCEs] into action.To be more specific, while I acknowledge the fact that one paper can only tackle so much, I think this study would benefit from additional simulation experiments to explore one or more of the raised questions, and try to disentangle some of the potential driving mechanisms further. Below are examples of such "low-hanging fruits":i) Additional no drought (truly) eCO2-only simulations to inspect its role in mortality overshoot some more (L496-498)ii) Running ED2 with less PFTs at Palo Verde (L550-551)iii) Imposing additional mortality to emulate post-drought secondary stressors' effects (L613-615) - I believe at least ED2 has a stop-restart functionality and the disturbance interval in LPJ-GUESS can be played with.iv) Testing different mortality thresholds from conductivity loss in ED2 (L653-657)I am sure that the authors are capable of devising much more clever ones than these. Even if some of these configurations are unrealistic and outcome from such exercises are still inconclusive, the findings could be interesting and it would add a valuable dimension to this manuscript.That said, I think there is no fundamental flaw in the study, and it is of interest for the Biogeosciences community. I would be happy to reevaluate after the revisions.Specific comments:As mentioned above, the presented framework is practical and logical but I was wondering whether the following points are obscuring the results:1) While no model-data comparisons can be performed on UCE responses due to lack of data (L228), there is no (mention of) validation as to whether these models were able to capture the modern vegetation at the test sites. Please cite the studies if this has been done elsewhere.2) Similarly, authors don't use any site-level observations to inform model parameterization (L290 no site-level tuning was conducted). At the least, this needs to be justified more.Without some reliability that the model simulations are representative of the sites, it is not clear how much of the responses could be due to misconfiguration, even if the scenarios are being evaluated with respect to a baseline.3) The uncertainties in parameters and models' sensitivities to them are not accounted for. I am sure the authors are aware of such uncertainties and sensitivities (e.g. sensitivity to water- and mortality-related parameters in LPJ-GUESS, Oberpriller et al., 2022 doi: 10.5194/gmd-2021-287 and in ED2, Raczka et al. 2018 doi: 10.1029/2018JG004504). I believe the findings would have been more robust if the models were run in ensembles and results were reported using the ensemble means at least.Please find more line-by-line comments below:L62 - frequency: While this study has enough to investigate under the existing framework, one possible extension that comes to mind is to include the increase in frequency. Authors consider magnitude and duration but not the frequency of back-to-back UCEs. Please consider coming back to it in the discussion, potentially in an explicit "limitations of the current study and future steps" sub-section.L77: Could the authors expand on the definition of 'moderate' quantitatively here? I.e. what were the treatments like in the mentioned studies, e.g. Beier et al X months, Kayler et al. Y amountL86-87: Authors are not coming back to these studies that document ecosystem responses to extreme droughts later in the text. Please consider including discussion on how your findings compare to these studies (even if they are post-hoc and limited).L152-154: While there is some observational support for linear response as cited by the authors, I wonder if expecting a linear response from these highly non-linear VDMs forms a plausible null-hypothesis as it sounds rather easy to refute (Sitch et al. 2008 GCB, McDowel et al., 2013 New Phyt, Rollinson et al. 2017 GCB, Bastos et al. 2020 Phil. Trans. R. Soc. B, Oberpriller et al. 2022 GMD) Could the authors elaborate with additional justifications in the text for not choosing a more specific null hypothesis?L207: Again, it would be great if authors could consider giving numbers as to what makes these studies 'moderate' to motivate the readers.L243: Can't one specify more PFTs or even species in LPJ-GUESS?L288: Just curious, were there no droughts during these historical site-specific climate periods?L291: Will the parameters sets be also made available in the Dryad repository? (the data availability statement says all model simulation data will be made available but I wanted to make sure all model configurations are also included)L305: On table S1 - Are the "drought + temp + 200 ppm eCO2" rows missing?L352: Looking at Fig 2 LPJ-GUESS seems to show some differences to drought duration, maybe consider saying "very little" instead of "no sensitivity".L404-406: This is an example of what I meant above by robustness of the results. Here authors identify it as a potentially anomalous signal because it was unexpected, but some of the other effects we are seeing could also be "anomalous", and vice versa, we may not be able to see certain effects because parameters were not varied in their potential ranges, limiting the overall impact of the study.L457-459: While authors mention that these swings in LPJ-GUESS LAI could be contributing to mortality response they do not explain the reason for these swings in LPJ-GUESS. A historical validation/assessment of the model behaviour at the site could have been useful here.L459-463: The reasoning behind how the model simulations lead to the suggestion of better representing morphological and physiological characteristics relevant to plant-water relations (e.g. leaf age) is not very obvious. Could the authors add some connecting and specific thoughts to the text?L480-483: Again the transition from simulations to the suggestion feels rather abrupt and generic. This recommendation could easily have come from the cited literature without performing this study at all, what else was learnt from the simulations?In sub-section 4.1.3 there is no referral to the process implementations of the models used in the study specifically. The only connection back to the results is the last sentence of the first paragraph, the rest of this subsection mostly reads like an introduction. Proposal to investigate allometric partitioning theory is not explicitly motivated by the simulations but rather by literature studies. Please consider re-wording this section and making the connections to this study more specific.L532-547: Could the authors say anything about which model implementation is more realistic according to their simulations? Can there be some more concrete conclusions/suggestions than "the need to better understand NSC dynamics"?L551-553: As mentioned in the beginning, it would have been nice to see some additional experiment as an attempt to disentangle whether ED2's resistance is due to functional trait diversity (or due to e.g. NSC implementation).L596: Improved how? What are specifically lacking in ED2 and LPJ-GUESS at the moment?L643: The thought process seems incomplete in this section. Was there going to be a suggestion?L654: Again, it feels like a missed opportunity that there were no additional simulation experiments with different thresholds to investigate this further.Table 3 - last row: Not sure if I'm following right but is the authors' recommendation here is to collect data on UCEs which they identified, by definition, to be unavailable? Please consider rewording.L708: As far as I can tell, the LPJ-GUESS code is not publicly available to download via this link, but available upon request. Please check and revise the statement accordingly if needed.Citation: https://doi.org/
10.5194/bg-2022-65-RC1 -
AC1: 'Reply on RC1', Jennifer A. Holm, 31 Jul 2022
Dear Biogeosciences Reviewer,
We appreciate your review and comments on the following paper “Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies”. Your time spent on this peer-review process is appreciated. There are valid points raised, and others that we help to clarify. Below is a point-by-point response to the reviews.
In general, we think clearly stating at the beginning of the manuscript that this study was not intended to solely be an original research paper, but rather a review of how models currently capture unprecedented climate extremes on forest ecosystems by using a handful of novel model results, will alleviate many of the reviewers' concerns. In order to fully investigate responses to such novel climate extremes that have not been experienced, we had to generate original model results which we believe is a strong novelty of this paper.
R1 has concerns about:
1. R1: The novelty of the paper; and whether this paper was a review paper or a research paper.
Answer: We believe the strong novelty of our study is the assessment of unprecedented climate extremes (UCEs) and their impacts. In the revised manuscript, we will improve the text about tighter linkages between concepts, hypotheses, and model outcomes. For example, we want to emphasize that a goal of this paper is to demonstrate how to use these vegetation demographic models (VDMs) to help generate future hypotheses about UCEs. Therefore, we used the models and sites as conceptual “experimental” tools to investigate the given hypotheses.
We will also revise the manuscript to state more clearly at the beginning that this is more of a review and “guidebook” manuscript, but with a combination of original research in order to address novel forest responses to climate extremes. We believe this combination of original research with review of current limitations of models is a strong novelty in and of itself.
We would like to clarify that the original “research” aspects of this study were the new UCE model results, and specifically the integrated C-loss with sensitivity to different climate change treatments which has not been done with these models, and is a knowledge gap that was filled by this study. The discussion section and detailed tables are designed to be a “review” and to help guide future research.
2. R1: While one paper can only tackle so much, R1 thinks this study would benefit from additional simulation experiments.
Answer: The scope of our paper was not to do many, in-depth, detailed experiments. But rather focus on how models currently capture UCEs in extreme durations/lengths and interactions with climate change. Therefore, we are not fully sure if there would be a strong benefit to adding in additional simulations experiments, which would be a large, costly endeavor and make the already very long manuscript even longer.
Probably the strongest point to not adding more simulations, is that additional simulations as suggested by R1, by itself, may not lead to the novelty and clarity that R1 seeks. We plan to re-write the manuscript to state that outstanding modeling perturbations and experiments are for subsequent studies. The initial modeling investigation here was to highlight how VDMs (as opposed to typical LSMs) could be used to answer hypotheses and guide future studies.
3. R1: Lack of model validation at the two sites, and site-level observations to inform model parameterization.
Answer: We agree that model comparison to some site-level observations is a worthy improvement to the paper. Upon revision of the manuscript, we will include site level benchmarking observations such as biomass, stand density, and leaf area index to do basic level model validation. We will also add a sentence that both models have already been run and validated at these sites in previously published papers (Xu et al. (2016); Medlyn et al. 2016), thus making model application possible with a “built in” reasonable degree of validation. These references were included in the original manuscript, but we failed to make a clear connection that they were publications that validated the models at the two sites.
4. R1: Reporting uncertainties in model parameters.
Answer: The ED2 and LPJ-GUESS models are well documented and investigated VDMs, with many previous studies that have looked into parameter uncertainty. Upon revision of this manuscript we will include a general description of the model parameters with the largest uncertainties. For example we know that parameters related to plant hydraulics and non-structural carbohydrate storage have large uncertainties and thus included these parameters in Table 1.
However, the aim of the paper was not to do in-depth, detailed experiments with tuned parameters specific to each site. But rather set up a general modeling framework so that hypotheses about unprecedented climate extremes could be investigated, and provide understanding on how model behavior of physiological and ecological processes might be lacking in state-of-the-srt ecosystem models in order to capture extremes.
We were glad to hear that R1 thought the paper was a “very informative read” and that “the presented framework is practical and logical”. As well as “the manuscript is in itself coherent, well-written, the structure is easy to follow with the aid of straightforward visualizations and tables.”
Citation: https://doi.org/10.5194/bg-2022-65-AC1 -
AC3: 'Reply on AC1', Jennifer A. Holm, 05 Jan 2023
Dear Reviewer,
We realized that some reviewer comments were not fully addressed from the last review period, and we would like to address these outstanding comments here. Thanks, and sorry for this oversight. The last manuscript version has been updated to include these new changes based on the reviewer comments.
RC: While this study has enough to investigate under the existing framework, one possible extension that comes to mind is to include the increase in frequency. Authors consider magnitude and duration but not the frequency of back-to-back UCEs. Please consider coming back to it in the discussion, potentially in an explicit "limitations of the current study and future steps" sub-section.
Author Answer: We agree it is important to include that back-to-back UCEs were not explored in this manuscript, and should be considered in future studies. In the summary section we included the following sentence:
“For example, we begin to investigate duration of droughts but we did not consider frequency of back-to-back UCEs.”
RC: While there is some observational support for linear response as cited by the authors, I wonder if expecting a linear response from these highly non-linear VDMs forms a plausible null-hypothesis as it sounds rather easy to refute.
Author Answer: We agree that complex biological life systems are not linear, and that VDMs are also highly nonlinear. The correct hypotheses that we were trying to test was comparing different degrees, or amplitudes, of nonlinearities, with carbon loss becoming more strongly nonlinear with increasing UCEs. Therefore, we removed text that described the null hypothesis as a linear relationship between carbon stock and drought, and instead are describing the null hypothesis and near-linear, and alternative hypothesis as different degrees of non-linearities.
RC: Could the authors expand on the definition of 'moderate' quantitatively here?
Author Answer: In the revised manuscript we have updated the text to say: “However, the majority of such experiments apply moderate treatments based on a historical sense, which are mostly weaker in intensity and/or shorter in duration than potential future UCEs”
“…take into account only low to moderate drought intensities (such as 50% rain excluded) or single events, or combine drought with moderate effects of temperature change. Where there has been 100% rain exclusion, it was on very small plots of 1.5 m2 (Meir et al., 2015).”
RC: Looking at Fig 2 LPJ-GUESS seems to show some differences to drought duration, maybe consider saying "very little" instead of "no sensitivity".
Author Answer: We have updated the text to say LPJ-GUESS very little sensitivity to drought duration, instead of no sensitivity.
RC: The reasoning behind how the model simulations lead to the suggestion of better representing morphological and physiological characteristics relevant to plant-water relations (e.g. leaf age) is not very obvious. Could the authors add some connecting and specific thoughts to the text?
Author Answer: In the discussion section of the revised manuscript we included a sentence that describes that leaf age classifications could be used in models to provide distinctions in the variations of leaf productivity and turnover time, which help to determine the balance between carbon gains and losses.
Citation: https://doi.org/10.5194/bg-2022-65-AC3
-
AC3: 'Reply on AC1', Jennifer A. Holm, 05 Jan 2023
-
AC1: 'Reply on RC1', Jennifer A. Holm, 31 Jul 2022
-
RC2: 'Comment on bg-2022-65', Anonymous Referee #2, 17 Jun 2022
“Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies” by Holm et al.
This study investigates carbon dynamics in response to unprecedented climate extremes (UCEs), specifically droughts as well as in combination with warming and CO2 fertilization, using two dynamic vegetation demographic models (VDMs), ED2 and LPJ-GUESS. It illustrates 1) conceptual frameworks and hypotheses on potential ecosystem response to UCEs, 2) VDM simulations to present potential ecosystem response, and finally 3) the limits of the current understanding of these responses. It is clearly written and will be an interesting topic to the readers of Biogeosciences. However, I fear that some important components are missing to make this study complete or original. For example, it is not clear whether the main focus of this study is to introduce the newest model development or review what is missing in the current VDMs and suggest future directions, or even the combination of both. In any case, model simulations need to be compared with the observation/data, at least partially, to describe the limitation of the current VDMs.
If process-based modeling is a great tool for studying the response of ecosystems to extreme drought events, which cannot be easily conducted with in-situ manipulation experiment (line90), I expect optimized or calibrated model simulations for specific study sites using historic drought events for the justification of the projections and to define the limitations of the current VDMs. Even if UCEs haven’t occurred at the study site (line228), validating the model with some historic drought events (that the models are sensitive to drought events, and the directions and magnitudes of the responses are reasonable) will be necessary to support the conclusion. Of course, it is not easy to track the effects of a certain drought event in reality because not only single event occurs during a certain time period, while controlling other factors. In addition, ecosystem responses to extreme droughts will be different from those to less intense droughts as the responses can be nonlinear. Yet, providing the evidence that the model performance is reasonable is critical. As one example, LPJ-GUESS showed large swings in LAI at EucFACE site (line476) – is it reasonable? If not, is it because of some parameters too sensitive specifically for this site or the model structure that needs to be modified? In this regard, another question arises: why two specific study sites are selected? Models were not tuned specifically for each site (line290), and comparison between the model simulation and observation was not made because data lack (line228; no observations in C change or no drought events?). If the model is not tuned for each site because of the lacking data, how about selecting other sites, where observations are available, at least for aboveground C? Or can any observations from FACE experiment be used for validating the model, such as the response to CO2 fertilization? Adding this information and validating the model simulations will make the results more plausible and this study very unique.
In the discussion, what is missing in the current VDMs are thoroughly reviewed. It can be nicer if there was a tighter connection between the simulation results and discussion by providing more detailed descriptions of the two models. Some parts are discussed based on the simulation results but mostly shortcomings are written based on literatures, which can be done without modeling exercises. Providing more details will make the story more coherent, what the current VDMs can do based on the simulation results, what are still missing, and what needs to be done. Another important point to address is, despite multiple missing processes in the current VDMs: how the results of the VDM simulations are different from conventional ESM simulations?
Some other comments:
Line77: ‘moderate treatments’ is vague. How can you differentiate moderate drought from unprecedented drought? Also, where can you draw the line between moderate and unprecedented droughts in the figures?
Line213: can you provide climatology of the sites, e.g., what are the mean and interannual variability of temperature and precipitation?
Line434: ‘multiple mechanisms that vary among ecosystem types’ – if the model was not tuned for each site, what drives the variations between the two sites, climate or the number of PFTs?
Line483: ‘further explored’ can be more specific or explored in this study
Citation: https://doi.org/10.5194/bg-2022-65-RC2 -
AC2: 'Reply on RC2', Jennifer A. Holm, 31 Jul 2022
Dear Biogeosciences Reviewer,
We appreciate your review and comments on the following paper “Exploring the impacts of unprecedented climate extremes on forest ecosystems: hypotheses to guide modeling and experimental studies”. Your time spent on this peer-review process is appreciated. There are valid points raised, and others that we help to clarify. Below is a point-by-point response to the reviews.
In general, we think clearly stating at the beginning of the manuscript that this study was not intended to solely be an original research paper, but rather a review of how models currently capture unprecedented climate extremes on forest ecosystems by using a handful of novel model results, will alleviate many of the reviewers' concerns. In order to fully investigate responses to such novel climate extremes that have not been experienced, we had to generate original model results which we believe is a strong novelty of this paper.
R2 has concerns about:
- R2: It is not clear whether the main focus of this study is to introduce the newest model development or review what is missing in the current VDMs and suggest future directions, or even the combination of both.
Answer: We will rewrite the manuscript to make it clearer that the goal of this paper is to highlight what is missing in current VDMs, even after taking into account cutting edge model developments. In order to review what is lacking in current VDMs, we needed to describe the existing model frameworks and latest model developments (which might have been a little confusing). We will make this more clear in the revision.
We will also revise the manuscript to state more clearly at the beginning that this is more of a review and “guidebook” manuscript, but with a combination of an original research in order to address novel forest responses to climate extremes. We believe this combination of original research with review of current limitations of models is a strong novelty in and of itself.
2. R2: Lack of model validation and comparisons with the observation/data, and validating against some historic drought events at these sites.
Answer: We agree that model comparison to some site-level observations is a worthy improvement to the paper. Upon revision of the manuscript, we will include site level benchmarking observations such as biomass, stand density, and leaf area index to do basic level model validation. With regards to validating the models against historic drought events at these sites, as also pointed out by reviewer #2, it is not easy to track the effects of a drought event in reality when these events can be rare, field campaigns didn’t occur during the event, all while controlling for other factors. The goal of this paper was to investigate ecosystem responses to extreme droughts that will likely be occurring in the future, and how they are different from less intense droughts as the responses can be nonlinear.
3. R2: LPJ-GUESS showed large swings in LAI at EucFACE site, and R2 is wondering if this is reasonable?
Answer: In the revised manuscript, we will work to provide observational data of leaf area index (LAI) at the EucFACE site to compare against the large swings from the LPJ-GUESS model.
4. R2: Question about why these two specific study sites are selected? (And why no site site level tuning was done).
Answer: We choose these two sites (the Australian and Costa Rica sites) mainly because the models have already been run at these sites in previously published papers (Xu et. al., 2016; Medlyn et. al., 2016), thus making model application possible with a “built in” reasonable degree of validation. These sites also contain multiple measurements and information that allowed for previous papers to validate the models. These sites were also chosen because they span an interesting range of vegetation types (a temperate-subtropical transitional forest in EucFACE and a seasonally dry tropical forest at PaloVerd), and are in warm, seasonally dry climates that are more likely to experience droughts in the future. In the revised manuscript we will include a description of why the sites were chosen, reference the publications where model validation has already occurred, and include this in Section 2.2 where we already initially described the sites. We will also emphasize that the purpose of this paper wasn’t to do site comparisons between many different sites, but just select a few for hypothesis testing.
5. R2: Would like to see a tighter connection between the simulation results and discussion section which describes what is lacking in VDMs, by providing more detailed descriptions of the two models.
Answer: In the revised manuscript, we will improve the text about tighter linkages between concepts, hypotheses, and model outcomes. For example, we want to make better connections that large losses in carbon were likely linked to uncertainties in how we currently represent plant hydraulics, non-structural carbohydrate storage, or phenology diversity, which we then reviewed in the discussion section. We will also revise the manuscript to emphasize that a goal of this paper is to demonstrate how to use these VDMs in order to help generate future hypotheses about UCEs. Therefore, we used the models and sites as conceptual “experimental” tools to investigate the given hypotheses. In the supplemental material we will do a better job of referencing previous papers that have done more through investigations of specific model processes within each of the models, and listing which processes or mechanisms were evaluated and why.
We are glad to hear that R2 thought the paper was “clearly written and will be an interesting topic to the readers of Biogeosciences.” As well as, “In the discussion, what is missing in the current VDMs are thoroughly reviewed”, which as we listed above in our answer to concern #1 is the goal of this paper.
Citation: https://doi.org/10.5194/bg-2022-65-AC2
-
AC2: 'Reply on RC2', Jennifer A. Holm, 31 Jul 2022
Jennifer A. Holm et al.
Jennifer A. Holm et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 660 | 221 | 24 | 905 | 56 | 13 | 19 |
- HTML: 660
- PDF: 221
- XML: 24
- Total: 905
- Supplement: 56
- BibTeX: 13
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
