the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Sep 2023
Inferring the tree regeneration niche from inventory data using a dynamic forest model
Abstract. The regeneration niche of trees is governed by many processes and factors that are challenging to determine. Besides species distribution, which determine if seeds are available, complex local dynamics in forest ecosystems (e.g., competition, pathogens) exert fundamental influence on tree regeneration. Consequently, the representation of tree regeneration in dynamic forest models is a notoriously complicated process which often involves many subprocesses. The ForClim forest gap model described regeneration mainly by species traits and the ecological knowledge linking these traits to regeneration properties. However, this regeneration module was never validated with large-scale data. Here, we compare this trait-based approach with an alternative, namely an inverse calibration approach, where we estimate regeneration parameters from a large observational dataset of unmanaged European forests. The model inversion was done using Bayesian inference with a simple and complex model variant without and with competition during regeneration. In this approach, we estimate shade and drought tolerance as well as the temperature requirements for 11 common tree species along with the intensity of regeneration (i.e., the maximum regeneration rate). We find that the parameters determining species’ light niche (i.e., light requirements) are similar between the trait based and calibrated values for both model variants, but only the complex model led to plausible estimates of the drought niche. The temperature niche as defined in by traits could not be recovered from the data by either model variant using inverse calibration. The parameter estimates differed between the complex and the simple model, with the complex model performing better. In both model variants, the calibration strongly changed the parameters that determine regeneration intensity compared to the default.
We conclude that the regeneration niche of the tree species in this large European dataset can be recovered in terms of the stand-level parameters light availability and regeneration intensity, while abiotic drivers (temperature and drought) are more elusive. The higher performance of the inversely calibrated models underpins the importance of informing dynamic models by real-world observations. Future research should focus on an even larger environmental coverage of observations of demographic processes in unmanaged forests to verify our findings at species range limits under extreme climatic conditions.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(3232 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(3232 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2114', Robert Schlicht, 09 Nov 2023
General comments
Within the framework of the ForClim forest model, the authors present a computationally intensive study that derives, for two regeneration models (“simple” and “complex”), based on recruitment rates observed in extensive forest inventory data, Bayesian estimates and credible intervals of parameters representing the regeneration intensity (kEstDens or kTrMax, depending on the model) and species-specific regeneration (or “young tree”) parameters representing shade, temperature and drought tolerance (kLy, kDDMiny, kDrToly, respectively). They then compare these estimates – and their effect on the performance of subsequent simulations – (1) to earlier estimates based on ecological knowledge and (2) between the two models, and they provide a detailed discussion of the results.
The general approach is sound and well-thought-out, and the manuscript appears to have been carefully prepared, with useful and readable appendices. The full code and data are provided, although, due to the complexity of the analysis (several thousand files) I cannot, in the scope of this referee comment, assess the quality or correctness of the implementation. Also, the study is relevant because the justification of parameter values chosen in simulation models is, aside the model structure, often one of their weakest points, and a careful analysis such as the one presented here can help shed light on these questions.
A difficulty I see for many potential readers is the high level of technicality in the presentation, which, together with some smaller issues (like wording that is not always very concise or clear, and long lines) makes the manuscript a bit hard to read. This is probably in part due to the complexity of the model, but given this complexity I believe there is still room for improvement.
To make the manuscript accessible to a wider audience, I therefore suggest that the authors try to read the manuscript the from the point of view of an uninitiated reader and revise it so as to present important arguments in the main text in a concise and not too technical language. Some diagrams or tables might be of help here. In fact, the appendices already contain much of that material. For example, it would probably be a good idea to present the table with the relevant parameters (see Table B1) early on in the main text to help readers see through the various k... variables, which otherwise, on a first reading, are just overwhelming.
There are also a number of smaller comments and questions that I list below.
Specific commentsLine 12: “determine”: Replace with “determines”.
Line 23: “in”: Remove.
Lines 44-45: This is confusing as it gives the impression that the regeneration niche affects the entire lifespan of a tree. My guess is that “one adult” refers to a dying tree and the “new adult” refers to a tree that just reached maturity, but this distinction is not clear from the way it is written. It also suggests there are well-defined generations.
Line 49: “niches of young trees and adults”: This distinction is important, but isn’t the niche of a young tree already an oversimplification since, clearly, the niche very much depends on the life stage of the young tree?
Line 64: I would assume that mortality is one of the key determinants of regeneration because most trees do not survive a very young age.
Line 78: How does this definition of the regeneration niche fit to the abstract definition in lines 44-45? Which relevant aspects are possibly missing in this more specific definition? Also, I do not understand “among others”, which also appears to be in the wrong place.
Line 81: “Indeed ... DFMs:”: This sentence is not clear; maybe it is not needed.
Lines 110-111: What is the reason for the implausible measurements? How can we be sure the 696 plots not removed (of a total of 865) have correct measurements?
Line 192: Remove “(”.
Line 225: P(...) should probably be removed here.
Lines 235-237: There appears to be some confusion here. Assuming the recruitment rates for all species and plots are independent random variables (which might need some justification), the likelihood, being the probability density function evaluated at the observed data, is multiplicative. Taking logarithms to transform multiplication into addition we arrive at the log-likelihood, which, consequently, is additive. So taking sums of log-likelihoods results in a joint log-likelihood, not likelihood. In fact, I would suggest to explicitly include “log” in all expressions that are supposed to represent log-likelihoods (i.e. write log P(...)). Also, while the representation of the densities determines a multiplicative constant in the likelihood and hence an additive constant in the log-likelihood, a rescaled log-likelihood is no longer a log-likelihood, although it can, of course, be used in the way the authors indicate. (Rescaling the log-likelihood corresponds to changing the base of the logarithm, but in statistical theory log usually means the natural logarithm.) In line 235 the authors need to decide whether this is about the intermediate sums over the plots for each species or about the total sum.
Line 325: “following”: This seems to be the wrong word.
Line 362: Presumably the p-values are not meaningful here because the implied test of independence of the TBA and ICA features rests on the assumption that the (TBA, ICA) pairs for different species are exchangeable (for example, resulting from independent repetitions of the same experiment), which is clearly not satisfied here.
Line 491: “estimates”: Does this mean the credible intervals?
Line 584: “likelihood”: This needs some clarification because likelihoods exist in any statistical model. Possibly the authors here refer to an exact computation in a part of their analysis as opposed to some approximate procedure.
Line 608: Remove “.”.
Line 910: “affects”: Remove “s”.
Line 926: eq. A1: Having two variables PEst s and PEsts whose names differ by a space in the index only is probably not a good idea. (Also, “×” might be nicer than “*”.)
Citation: https://doi.org/10.5194/egusphere-2023-2114-RC1 -
RC2: 'Comment on egusphere-2023-2114', Anonymous Referee #2, 18 Dec 2023
Comments on “Inferring the tree regeneration niche from inventory data using a dynamic forest model.”
The aim of this study was to validate and compare two regeneration modules (simple and complex) of the ForClim model using an inverse calibration and trait-based approach. They found that parameters determining species' light requirements are consistent between trait-based and calibrated values for both model variants. Parameter values were consistent between the two models but changed during the calibration process, with temperature and drought exhibiting the highest uncertainties.
The methodology and results appear robust, and I do not see any methodological issues. The methodology is innovative, and model validation is crucial. However, more effort is needed to better understand the work. The methodology is quite complex for readers in ecology, and the introduction does not adequately introduce the concepts necessary for understanding the methodology. A simplified conceptual diagram of the methodology could greatly enhance the main text. For instance, including Figure B1 in the main text, along with both calibration methods, would be a valuable addition. The results and discussion were well-written and understandable.
Introduction: More work is needed for the introduction; it does not allow us to understand correctly and lacks the keys to comprehend the objectives. For instance, more details should be provided to understand inverse calibration. The initial paragraphs focusing on niches do not contribute to understanding; on the contrary, they distract the reader from more crucial matters. As your article leans more towards methodology, the introduction should present the philosophy behind the methodology. This is especially important in the context of a changing climate. Additionally, elucidate more on how DFMs model regeneration. The sentences are often too complex for effective reading.
Line 37: Why is it particularly important?
Line 44: The regeneration niche is defined as the set of factors for successful replacement. It's odd to have an increase when discussing the niche.
Line 49: I wouldn't use affirmations such as: "Differences between the niches of young trees and adults are evident."
Line 51: Ontogenetic niche shifts are differences between the niches of young trees and adults. One cannot explain the other if the entities are the same.
Line 53: "Together" refers to what? The same goes for line 45.
Line 57: What are the issues?
Lines 58-59: I didn't understand the second part of the sentence: "which puts the quantified effects in the context of specific processes."
Lines 58: Replace "represent" with "model"?
Line 64: "Regeneration in most DFMs is captured relatively simply." Be more specific.
Lines 73-75: "Specifically, trade-offs between meaningful observations for key small-scale processes and the coverage of wide abiotic gradients impede a comprehensive analysis across the stages of tree regeneration (Clark et al., 1999)." I didn't understand; give examples instead.
Lines 79 to 84: I did not understand the end of this long sentence: "which is explicitly modeled are expected to have more predictive power than traits that act on multiple processes which are not considered individually." Keep your sentences short and comprehensible.
Line 86: I didn't understand this sentence: "In ForClim, the interplay between the traits of multiple species is implemented in a simple variant of the regeneration model that does not consider competition and a more complex model variant that includes it."
Line 90: The regeneration process cannot be synthesized by only seedling and sapling growth rates; the mortality and density of seedlings are also important components of regeneration. In line 90, "it remains unclear whether the simulated regeneration processes (i.e., ingrowth rates) are actually in agreement with observational data," but in line 92, "we evaluate possible reasons for mismatches between process formulations and observations." You acknowledge the uncertainty regarding whether the simulated processes align with observations, and then you evaluate the reasons for the mismatches. However, these two paragraphs do not establish a clear connection.
Lines 147-148: I did not understand this sentence. And it's unfortunate because it is central to the understanding of the article.
Results:
Lines 321-322: This kind of sentence is unnecessary.
Line 354: The explanation for the grey rectangle regarding priors is unclear.
Citation: https://doi.org/10.5194/egusphere-2023-2114-RC2 -
AC1: 'Response to reviewer comments', Yannek Käber, 31 Jan 2024
Dear Editors and Reviewers, We are grateful for the constructive and positive feedback, which has significantly contributed to the improvement of our manuscript. Both reviewers raised concerns about the readability and clarity of the manuscript, and we agree. In particular, the introduction needed improvement to make the text more accessible to a wider audience. At the same time, we acknowledge that there was a lack of coherence in various parts of the text. Thank you for your valuable comments and suggestions to resolve these issues. We thoroughly reviewed each comment and made appropriate revisions to the manuscript. The text is now easier to understand and the introduction is more coherent. The main improvements made are - Extensive improvements to the introduction by removing much of the text on concepts and ideas not directly related to our study. We have also placed more emphasis on explaining the context and motivation behind our methodological approach. - We have moved Figure B1 and Table B1 to the methods section as we believe they provide important information for understanding our approach. - We resolved an issue with the definition and use of the terminology regarding the likelihood. - We believe that these changes have significantly improved the accessibility and clarity of the manuscript. A point-by-point response to the reviewers' comments, is attached. The revised manuscript will be uploaded for your consideration. We appreciate the opportunity to refine our work and look forward to any further comments you may have. Thank you for your attention to our submission and for the valuable insights you have provided. Best regards, Yannek Käber on behalf of all co-authors
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2114', Robert Schlicht, 09 Nov 2023
General comments
Within the framework of the ForClim forest model, the authors present a computationally intensive study that derives, for two regeneration models (“simple” and “complex”), based on recruitment rates observed in extensive forest inventory data, Bayesian estimates and credible intervals of parameters representing the regeneration intensity (kEstDens or kTrMax, depending on the model) and species-specific regeneration (or “young tree”) parameters representing shade, temperature and drought tolerance (kLy, kDDMiny, kDrToly, respectively). They then compare these estimates – and their effect on the performance of subsequent simulations – (1) to earlier estimates based on ecological knowledge and (2) between the two models, and they provide a detailed discussion of the results.
The general approach is sound and well-thought-out, and the manuscript appears to have been carefully prepared, with useful and readable appendices. The full code and data are provided, although, due to the complexity of the analysis (several thousand files) I cannot, in the scope of this referee comment, assess the quality or correctness of the implementation. Also, the study is relevant because the justification of parameter values chosen in simulation models is, aside the model structure, often one of their weakest points, and a careful analysis such as the one presented here can help shed light on these questions.
A difficulty I see for many potential readers is the high level of technicality in the presentation, which, together with some smaller issues (like wording that is not always very concise or clear, and long lines) makes the manuscript a bit hard to read. This is probably in part due to the complexity of the model, but given this complexity I believe there is still room for improvement.
To make the manuscript accessible to a wider audience, I therefore suggest that the authors try to read the manuscript the from the point of view of an uninitiated reader and revise it so as to present important arguments in the main text in a concise and not too technical language. Some diagrams or tables might be of help here. In fact, the appendices already contain much of that material. For example, it would probably be a good idea to present the table with the relevant parameters (see Table B1) early on in the main text to help readers see through the various k... variables, which otherwise, on a first reading, are just overwhelming.
There are also a number of smaller comments and questions that I list below.
Specific commentsLine 12: “determine”: Replace with “determines”.
Line 23: “in”: Remove.
Lines 44-45: This is confusing as it gives the impression that the regeneration niche affects the entire lifespan of a tree. My guess is that “one adult” refers to a dying tree and the “new adult” refers to a tree that just reached maturity, but this distinction is not clear from the way it is written. It also suggests there are well-defined generations.
Line 49: “niches of young trees and adults”: This distinction is important, but isn’t the niche of a young tree already an oversimplification since, clearly, the niche very much depends on the life stage of the young tree?
Line 64: I would assume that mortality is one of the key determinants of regeneration because most trees do not survive a very young age.
Line 78: How does this definition of the regeneration niche fit to the abstract definition in lines 44-45? Which relevant aspects are possibly missing in this more specific definition? Also, I do not understand “among others”, which also appears to be in the wrong place.
Line 81: “Indeed ... DFMs:”: This sentence is not clear; maybe it is not needed.
Lines 110-111: What is the reason for the implausible measurements? How can we be sure the 696 plots not removed (of a total of 865) have correct measurements?
Line 192: Remove “(”.
Line 225: P(...) should probably be removed here.
Lines 235-237: There appears to be some confusion here. Assuming the recruitment rates for all species and plots are independent random variables (which might need some justification), the likelihood, being the probability density function evaluated at the observed data, is multiplicative. Taking logarithms to transform multiplication into addition we arrive at the log-likelihood, which, consequently, is additive. So taking sums of log-likelihoods results in a joint log-likelihood, not likelihood. In fact, I would suggest to explicitly include “log” in all expressions that are supposed to represent log-likelihoods (i.e. write log P(...)). Also, while the representation of the densities determines a multiplicative constant in the likelihood and hence an additive constant in the log-likelihood, a rescaled log-likelihood is no longer a log-likelihood, although it can, of course, be used in the way the authors indicate. (Rescaling the log-likelihood corresponds to changing the base of the logarithm, but in statistical theory log usually means the natural logarithm.) In line 235 the authors need to decide whether this is about the intermediate sums over the plots for each species or about the total sum.
Line 325: “following”: This seems to be the wrong word.
Line 362: Presumably the p-values are not meaningful here because the implied test of independence of the TBA and ICA features rests on the assumption that the (TBA, ICA) pairs for different species are exchangeable (for example, resulting from independent repetitions of the same experiment), which is clearly not satisfied here.
Line 491: “estimates”: Does this mean the credible intervals?
Line 584: “likelihood”: This needs some clarification because likelihoods exist in any statistical model. Possibly the authors here refer to an exact computation in a part of their analysis as opposed to some approximate procedure.
Line 608: Remove “.”.
Line 910: “affects”: Remove “s”.
Line 926: eq. A1: Having two variables PEst s and PEsts whose names differ by a space in the index only is probably not a good idea. (Also, “×” might be nicer than “*”.)
Citation: https://doi.org/10.5194/egusphere-2023-2114-RC1 -
RC2: 'Comment on egusphere-2023-2114', Anonymous Referee #2, 18 Dec 2023
Comments on “Inferring the tree regeneration niche from inventory data using a dynamic forest model.”
The aim of this study was to validate and compare two regeneration modules (simple and complex) of the ForClim model using an inverse calibration and trait-based approach. They found that parameters determining species' light requirements are consistent between trait-based and calibrated values for both model variants. Parameter values were consistent between the two models but changed during the calibration process, with temperature and drought exhibiting the highest uncertainties.
The methodology and results appear robust, and I do not see any methodological issues. The methodology is innovative, and model validation is crucial. However, more effort is needed to better understand the work. The methodology is quite complex for readers in ecology, and the introduction does not adequately introduce the concepts necessary for understanding the methodology. A simplified conceptual diagram of the methodology could greatly enhance the main text. For instance, including Figure B1 in the main text, along with both calibration methods, would be a valuable addition. The results and discussion were well-written and understandable.
Introduction: More work is needed for the introduction; it does not allow us to understand correctly and lacks the keys to comprehend the objectives. For instance, more details should be provided to understand inverse calibration. The initial paragraphs focusing on niches do not contribute to understanding; on the contrary, they distract the reader from more crucial matters. As your article leans more towards methodology, the introduction should present the philosophy behind the methodology. This is especially important in the context of a changing climate. Additionally, elucidate more on how DFMs model regeneration. The sentences are often too complex for effective reading.
Line 37: Why is it particularly important?
Line 44: The regeneration niche is defined as the set of factors for successful replacement. It's odd to have an increase when discussing the niche.
Line 49: I wouldn't use affirmations such as: "Differences between the niches of young trees and adults are evident."
Line 51: Ontogenetic niche shifts are differences between the niches of young trees and adults. One cannot explain the other if the entities are the same.
Line 53: "Together" refers to what? The same goes for line 45.
Line 57: What are the issues?
Lines 58-59: I didn't understand the second part of the sentence: "which puts the quantified effects in the context of specific processes."
Lines 58: Replace "represent" with "model"?
Line 64: "Regeneration in most DFMs is captured relatively simply." Be more specific.
Lines 73-75: "Specifically, trade-offs between meaningful observations for key small-scale processes and the coverage of wide abiotic gradients impede a comprehensive analysis across the stages of tree regeneration (Clark et al., 1999)." I didn't understand; give examples instead.
Lines 79 to 84: I did not understand the end of this long sentence: "which is explicitly modeled are expected to have more predictive power than traits that act on multiple processes which are not considered individually." Keep your sentences short and comprehensible.
Line 86: I didn't understand this sentence: "In ForClim, the interplay between the traits of multiple species is implemented in a simple variant of the regeneration model that does not consider competition and a more complex model variant that includes it."
Line 90: The regeneration process cannot be synthesized by only seedling and sapling growth rates; the mortality and density of seedlings are also important components of regeneration. In line 90, "it remains unclear whether the simulated regeneration processes (i.e., ingrowth rates) are actually in agreement with observational data," but in line 92, "we evaluate possible reasons for mismatches between process formulations and observations." You acknowledge the uncertainty regarding whether the simulated processes align with observations, and then you evaluate the reasons for the mismatches. However, these two paragraphs do not establish a clear connection.
Lines 147-148: I did not understand this sentence. And it's unfortunate because it is central to the understanding of the article.
Results:
Lines 321-322: This kind of sentence is unnecessary.
Line 354: The explanation for the grey rectangle regarding priors is unclear.
Citation: https://doi.org/10.5194/egusphere-2023-2114-RC2 -
AC1: 'Response to reviewer comments', Yannek Käber, 31 Jan 2024
Dear Editors and Reviewers, We are grateful for the constructive and positive feedback, which has significantly contributed to the improvement of our manuscript. Both reviewers raised concerns about the readability and clarity of the manuscript, and we agree. In particular, the introduction needed improvement to make the text more accessible to a wider audience. At the same time, we acknowledge that there was a lack of coherence in various parts of the text. Thank you for your valuable comments and suggestions to resolve these issues. We thoroughly reviewed each comment and made appropriate revisions to the manuscript. The text is now easier to understand and the introduction is more coherent. The main improvements made are - Extensive improvements to the introduction by removing much of the text on concepts and ideas not directly related to our study. We have also placed more emphasis on explaining the context and motivation behind our methodological approach. - We have moved Figure B1 and Table B1 to the methods section as we believe they provide important information for understanding our approach. - We resolved an issue with the definition and use of the terminology regarding the likelihood. - We believe that these changes have significantly improved the accessibility and clarity of the manuscript. A point-by-point response to the reviewers' comments, is attached. The revised manuscript will be uploaded for your consideration. We appreciate the opportunity to refine our work and look forward to any further comments you may have. Thank you for your attention to our submission and for the valuable insights you have provided. Best regards, Yannek Käber on behalf of all co-authors
Peer review completion
Journal article(s) based on this preprint
Data sets
Supplementary material for Käber et al. "Inferring the regeneration niche from forest inventory data using a dynamic forest model". In Geoscientific Model Development (1.0) Yannek Käber, Florian Hartig, and Harald Bugmann https://doi.org/10.5281/zenodo.8334092
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 223 | 133 | 14 | 370 | 11 | 13 |
- HTML: 223
- PDF: 133
- XML: 14
- Total: 370
- BibTeX: 11
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Florian Hartig
Harald Bugmann
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(3232 KB) - Metadata XML