the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Nov 2025
Global parameter sensitivity analysis of modelling water, energy and carbon dynamics in a temperate swamp
Abstract. Forested peatlands cover a land area of 7 x 105 km2 and store ~77 Pg C in Canada. However, the carbon (C) cycling of forested peatlands, particularly swamps, has been understudied. Few modelling studies have been done on temperate swamp C cycling partly because of the scarcity of field measurements in this ecosystem. These gaps create uncertainties in modelling the C dynamics of temperate swamps and consequently limit our understanding of this ecosystem. To improve our understanding of the processes, interactions and feedbacks that mediate temperate swamp C cycling, we simulated the long-term (40 years) plant processes, energy, water and C fluxes of Beverly Swamp, a well-preserved swamp in Southern Ontario using a process-based model (CoupModel). CoupModel v6 was systematically calibrated for Beverly Swamp using the Generalized Likelihood Uncertainty Estimate (GLUE) method and validated with field measurements. The GLUE approach and its multicriteria constraints reduced the uncertainties associated with the modelling process and reasonably improved some of the simulation outcomes when compared to the initial single run and prior uniform distribution. Global sensitivity analysis of the parameters identified the important parameters that greatly influence temperate swamp C flux simulations and the interconnections that exist between simulated variables and parameters. Plant-related processes and hydrological variables exerted the strongest control on soil respiration simulation. However, these dynamics may be altered as climate continues to warm in coming decades. Results from this study provide valuable knowledge for predicting the fate of swamp C cycle in the region under a changing climate.
- Preprint
(2016 KB) - Metadata XML
-
Supplement
(821 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-1368', Anonymous Referee #1, 22 Mar 2026
-
AC1: 'Reply on RC1', Oluwabamise Afolabi, 25 Jul 2026
General comments
We thank the reviewers for the constructive feedback on our manuscript. Even though we have responded to the reviewer’s specific comments, we would like to address some of the general comments here. Both reviewers raised concerns about the overall language of the manuscript and the poor presentation of the manuscript’s contribution to scientific advancement. We do agree with the reviewers that the study’s linkage to scientific contexts can be further improved, as the current version is heavily focused on presenting the modeling details of the studied temperate swamp peatland. We initially focused on this aspect because the ecosystem is understudied and our manuscript (in addition to Afolabi et al., 2025) is one of the first process-based modeling studies on temperate swamp peatlands where very few data exist. We felt it was an avenue to present some of these foundational details for setting the modeling context of the ecosystem.
To address the general concerns highlighted above, we propose to do the following in the revision:
- We will better present the scientific importance of the paper;
- Include additional discussion sections on the uniqueness of temperate swamp peatlands systems, compared to other wetlands and peatlands. These discussions will cover the importance of parameters and processes that regulate seasonal surface and groundwater exchanges and its impacts on soil carbon (C) fluxes;
- Compare parameter posterior distributions and process interactions to previous peatland modeling studies by CoupModel and other existing swamp and wetland models. This would put current study into context and better highlight the modeling advancements of this study.
- We will revise the general language of the entire manuscript and thoroughly proofread it before resubmission.
We also observed that the reviewers raised concerns about the selection of the study’s calibration and validation periods and data. The separation of our calibration and validation periods was largely constrained by data availability. We acknowledged in the manuscript and Afolabi et al. (2025) that we relied on data from the 1980s, short studies from 1998 – 2000 and resumed measurements from 2022 to 2023. These data were mostly not continuous and were done by different groups with varying methods (excluding the validation period). Thus, there were uncertainties associated with these measurements, which we have discussed in the current manuscript and Afolabi et al. (2025).
To further evaluate the model and show the robustness of our model calibration, we will undertake additional experiments in the revision by using all (both calibration and validation, and/or swap the calibration and validation period) data for calibration and constraining the soil respiration with new RMSE threshold to show that our key results are not influenced by the choice of calibration and validation periods or subjectiveness of the defined criteria. We anticipate these additional experiments will not change the key findings of the paper because we initially conducted some of these experiments and analyzed the results before reporting them in the current version.
We should also mention that our model evaluation manuscript is a follow-up work to Afolabi et al. (2025), which was a single-run experiment with the CoupModel to simulate the long-term carbon dynamics of the temperate swamp peatland. The initial single run set-up of Afolabi et al. (2025) in CoupModel for the Beverly Swamp was not without considerable uncertainties (see section 4 of Afolabi et al., 2025 for details). To address the highlighted issues in Afolabi et al. (2025), we conducted an uncertainty analysis of the initial model set-up with the GLUE approach in this manuscript. This approach assisted in identifying the errors associated with the different aspects (e.g., measurement, parameter and model structure) of the modelling exercise and also analyzed the extent that the model data discrepancies were explained by the measured data, model structure and parameters.
The findings of our study are important to the selection of model structure, parameter distribution and processes for modeling hydrology and carbon related processes in temperate swamp peatlands. Since this is the first attempt to evaluate the performance of CoupModel (version 6 and GLUE calibrated) in a temperate swamp peatland, where the hydroperiods, vegetation cover and other biophysical conditions are different from other peatland types (e.g., bogs and fens), we anticipate that the analysis of our manuscript will assist the model structure selection and parameterization of large-scale ecological models (e.g., CLASSIC and CaMP) when simulating swamp peatland carbon flux at regional and global scales, as the swamp category is currently missing in these models for Canadian studies. Furthermore, some of the important parameters and soil respiration constraining variables (e.g., WTL, VMC and LAI) that were identified in this study are expected to help inform the choice of variables to be measured in the field for carbon related studies. In addition, the behavioral models identified by the manuscript can be used for climate change assessment studies of the swamp peatland.
Specific Comments
- Section 5.1 should be moved to the Discussion section. Its current content focuses on interpretation and mechanism explanation rather than presenting new results. Integrating it into the Discussion would improve the logical coherence of the manuscript.
Response: We thank the Reviewer for identifying this. We will ensure that the manuscript is thoroughly reviewed to capture the suggestions in our next revision.
- The data basis raises concerns regarding robustness. Key calibration and validation datasets are sparse and temporally inconsistent (e.g., combining 1998–2000 and 2022–2023 observations). The authors should better justify this strategy and discuss its implications for model reliability and uncertainty.
Response: Our study site, Beverly Swamp included data from the 1980s, short studies from 1998 – 2000 by Davidson et al (2019), McCarter et al. (2024) and resumed measurements from 2022. Thus, Beverly Swamp to our knowledge, represents one of the most well-studied temperate swamp peatlands. The long-term coverage of empirical data makes long term modeling study as ours possible, however, these data collections were in some cases not continuous and were done by different groups with varying methods over a 40+ year period. This sparsity of continuous field data influenced our selection of the calibration and validation periods to coincide with periods of data availability. Most of the measurements used for the study are on daily time resolution with the exception of soil respiration (bi-weekly), LAI (seasonal) and snow depth (seasonal). Please see Table 2 of manuscript for details
Because we are aware that considerable uncertainties are associated with the measurements, we discussed this measurement uncertainty in Afolabi et al. (2025) and also elaborated the extent that the model data discrepancies can be explained by the measured data, model structure and parameters in our manuscript. Nevertheless, we will elaborate this in the revision. Furthermore, we are open to undertake additional experiments such as using all (both calibration and validation) the data for calibration and constraining the soil respiration with new RMSE threshold to show that our key results are not influenced by the choice of calibration and validation periods.
- The treatment of lateral water flow using proxy data is a critical assumption. Although acknowledged, it is not sufficiently quantified. A more rigorous uncertainty assessment or sensitivity discussion specific to this assumption is needed.
Response: Our study was constrained to use stream discharge records from Westover gauging station at the exit of swamp (south of the swamp) as proxy for lateral flow input into the swamp because of the unavailability of this information. We were confident to rely on this proxy data at the exit of the swamp because Woo and Valverde (1981) and McCarter et al., 2024 had identified in their water balance studies of the swamp that lateral inflow and outflow from the swamp were roughly the same. Nevertheless, our previous study (Section 4.1 of Afolabi et al., 2025) hypothesized that some of the mismatch between the simulated and observed WTL during the summer and autumn seasons may be associated with surface runoff shortage during this period. The outflow data may not have sufficiently captured the water release timing of the dam, which eventually contributed to the time lag of lateral water input into the swamp and some mismatch in the overall water balance simulation.
To quantify the inconsistencies associated with the lateral flow proxy data assumption, an uncertainty analysis was undertaken for this input variable as part of the GLUE process. We introduced the variable into the model as average flow input (0.86 mm day-1) defined by the parameter of Gwsourceflow, qsof (see section 2.4.2 for details). Furthermore, we ascertained the implications of the lateral flow proxy data assumption on our C flux simulations by analyzing the sensitivity of the swamp C fluxes to changes in lateral water inputs (section 3.4). Our experiment results in section 3.4 showed that the swamp’s C flux displayed minimal sensitivity to changes in lateral water input, so we do not expect this proxy data assumption to significantly influence the outcome of our modeling experiment. Nevertheless, we will elaborate this in the next revision.
- The Results section is overly descriptive, particularly in terms of parameter sensitivity rankings. The authors should better synthesize the findings, highlight dominant controls, and link them more explicitly to physical processes rather than listing parameter importance.
Response: Many thanks for the comment. As suggested, we will ensure that the findings are better synthesized with physically processes and not just the parameter importance in the next revision.
- The Discussion is currently insufficient in depth. It should be strengthened by (i) systematic comparison with existing studies (especially bog and fen systems), (ii) clearer interpretation of process interactions, and (iii) explicit discussion of model limitations and applicability.
Response: Many thanks for the comment. We will ensure that the discussion section is better strengthened by incorporating the feedback above.
- Model structural limitations need to be more explicitly addressed. The use of a one-dimensional model for a spatially heterogeneous swamp system may constrain interpretation, and this should be critically discussed.
Response: Many thanks for the comment. Based on the evaluation results of this manuscript and those of Afolabi et al. (2025), we are confident that the one-dimensional model (CoupModel) captured the critical biophysical and biochemical interactions and feedback of the swamp. However, we agree with the Reviewer that one-dimensional models like CoupModel are not without limitations. This concurrence informed our discussions (section 4.1 of Afolabi et al., 2025) on the implications of using a one-dimensional model for a spatially heterogenous swamp system like Beverly swamp. For instance, we mentioned in Afolabi et al. (2025) that snow disappearance in the open canopy section of the swamp during winter season (Smith & Woo, 1986) may not have been properly accounted for in the simulation because one-dimensional designs are targeted to mostly capture vertical and not spatially distributed transport and complicated phase change processes. This limitation may have affected energy partitioning in winter and contributed to the overestimation of sensible and ground heat fluxes during this season, explaining some of the mismatch in snow accumulation simulations and soil thermal processes in winter. We can also elaborate this aspect in the manuscript.
- The manuscript would benefit from improved integration between sections. Currently, the linkage between objectives, methods, results, and conclusions is somewhat loose. The authors should ensure that each objective is clearly addressed and revisited in the results and discussion.
Response: Many thanks for the comment. We will ensure that the sections are better harmonized in the revision.
- Overall, the manuscript has a solid foundation, but requires deeper analysis, clearer positioning of its contribution, and improved discussion to meet publication standards.
Response: Many thanks for the comments on the manuscript. We will ensure to incorporate the recommendations you highlighted above and revise the manuscript accordingly.
Citation: https://doi.org/10.5194/egusphere-2025-1368-AC1 -
AC3: 'Reply on RC1', Oluwabamise Afolabi, 25 Jul 2026
General comments
We thank the reviewers for the constructive feedback on our manuscript. Even though we have responded to the reviewer’s specific comments, we would like to address some of the general comments here. Both reviewers raised concerns about the overall language of the manuscript and the poor presentation of the manuscript’s contribution to scientific advancement. We do agree with the reviewers that the study’s linkage to scientific contexts can be further improved, as the current version is heavily focused on presenting the modeling details of the studied temperate swamp peatland. We initially focused on this aspect because the ecosystem is understudied and our manuscript (in addition to Afolabi et al., 2025) is one of the first process-based modeling studies on temperate swamp peatlands where very few data exist. We felt it was an avenue to present some of these foundational details for setting the modeling context of the ecosystem.
To address the general concerns highlighted above, we propose to do the following in the revision:
- We will better present the scientific importance of the paper;
- Include additional discussion sections on the uniqueness of temperate swamp peatlands systems, compared to other wetlands and peatlands. These discussions will cover the importance of parameters and processes that regulate seasonal surface and groundwater exchanges and its impacts on soil carbon (C) fluxes;
- Compare parameter posterior distributions and process interactions to previous peatland modeling studies by CoupModel and other existing swamp and wetland models. This would put current study into context and better highlight the modeling advancements of this study.
- We will revise the general language of the entire manuscript and thoroughly proofread it before resubmission.
We also observed that the reviewers raised concerns about the selection of the study’s calibration and validation periods and data. The separation of our calibration and validation periods was largely constrained by data availability. We acknowledged in the manuscript and Afolabi et al. (2025) that we relied on data from the 1980s, short studies from 1998 – 2000 and resumed measurements from 2022 to 2023. These data were mostly not continuous and were done by different groups with varying methods (excluding the validation period). Thus, there were uncertainties associated with these measurements, which we have discussed in the current manuscript and Afolabi et al. (2025).
To further evaluate the model and show the robustness of our model calibration, we will undertake additional experiments in the revision by using all (both calibration and validation, and/or swap the calibration and validation period) data for calibration and constraining the soil respiration with new RMSE threshold to show that our key results are not influenced by the choice of calibration and validation periods or subjectiveness of the defined criteria. We anticipate these additional experiments will not change the key findings of the paper because we initially conducted some of these experiments and analyzed the results before reporting them in the current version.
We should also mention that our model evaluation manuscript is a follow-up work to Afolabi et al. (2025), which was a single-run experiment with the CoupModel to simulate the long-term carbon dynamics of the temperate swamp peatland. The initial single run set-up of Afolabi et al. (2025) in CoupModel for the Beverly Swamp was not without considerable uncertainties (see section 4 of Afolabi et al., 2025 for details). To address the highlighted issues in Afolabi et al. (2025), we conducted an uncertainty analysis of the initial model set-up with the GLUE approach in this manuscript. This approach assisted in identifying the errors associated with the different aspects (e.g., measurement, parameter and model structure) of the modelling exercise and also analyzed the extent that the model data discrepancies were explained by the measured data, model structure and parameters.
The findings of our study are important to the selection of model structure, parameter distribution and processes for modeling hydrology and carbon related processes in temperate swamp peatlands. Since this is the first attempt to evaluate the performance of CoupModel (version 6 and GLUE calibrated) in a temperate swamp peatland, where the hydroperiods, vegetation cover and other biophysical conditions are different from other peatland types (e.g., bogs and fens), we anticipate that the analysis of our manuscript will assist the model structure selection and parameterization of large-scale ecological models (e.g., CLASSIC and CaMP) when simulating swamp peatland carbon flux at regional and global scales, as the swamp category is currently missing in these models for Canadian studies. Furthermore, some of the important parameters and soil respiration constraining variables (e.g., WTL, VMC and LAI) that were identified in this study are expected to help inform the choice of variables to be measured in the field for carbon related studies. In addition, the behavioral models identified by the manuscript can be used for climate change assessment studies of the swamp peatland.
Specific Comments
1. Section 5.1 should be moved to the Discussion section. Its current content focuses on interpretation and mechanism explanation rather than presenting new results. Integrating it into the Discussion would improve the logical coherence of the manuscript.
Response: We thank the Reviewer for identifying this. We will ensure that the manuscript is thoroughly reviewed to capture the suggestions in our next revision.
2. The data basis raises concerns regarding robustness. Key calibration and validation datasets are sparse and temporally inconsistent (e.g., combining 1998–2000 and 2022–2023 observations). The authors should better justify this strategy and discuss its implications for model reliability and uncertainty.
Response: Our study site, Beverly Swamp included data from the 1980s, short studies from 1998 – 2000 by Davidson et al (2019), McCarter et al. (2024) and resumed measurements from 2022. Thus, Beverly Swamp to our knowledge, represents one of the most well-studied temperate swamp peatlands. The long-term coverage of empirical data makes long term modeling study as ours possible, however, these data collections were in some cases not continuous and were done by different groups with varying methods over a 40+ year period. This sparsity of continuous field data influenced our selection of the calibration and validation periods to coincide with periods of data availability. Most of the measurements used for the study are on daily time resolution with the exception of soil respiration (bi-weekly), LAI (seasonal) and snow depth (seasonal). Please see Table 2 of manuscript for details
Because we are aware that considerable uncertainties are associated with the measurements, we discussed this measurement uncertainty in Afolabi et al. (2025) and also elaborated the extent that the model data discrepancies can be explained by the measured data, model structure and parameters in our manuscript. Nevertheless, we will elaborate this in the revision. Furthermore, we are open to undertake additional experiments such as using all (both calibration and validation) the data for calibration and constraining the soil respiration with new RMSE threshold to show that our key results are not influenced by the choice of calibration and validation periods.
3. The treatment of lateral water flow using proxy data is a critical assumption. Although acknowledged, it is not sufficiently quantified. A more rigorous uncertainty assessment or sensitivity discussion specific to this assumption is needed.
Response: Our study was constrained to use stream discharge records from Westover gauging station at the exit of swamp (south of the swamp) as proxy for lateral flow input into the swamp because of the unavailability of this information. We were confident to rely on this proxy data at the exit of the swamp because Woo and Valverde (1981) and McCarter et al., 2024 had identified in their water balance studies of the swamp that lateral inflow and outflow from the swamp were roughly the same. Nevertheless, our previous study (Section 4.1 of Afolabi et al., 2025) hypothesized that some of the mismatch between the simulated and observed WTL during the summer and autumn seasons may be associated with surface runoff shortage during this period. The outflow data may not have sufficiently captured the water release timing of the dam, which eventually contributed to the time lag of lateral water input into the swamp and some mismatch in the overall water balance simulation.
To quantify the inconsistencies associated with the lateral flow proxy data assumption, an uncertainty analysis was undertaken for this input variable as part of the GLUE process. We introduced the variable into the model as average flow input (0.86 mm day-1) defined by the parameter of Gwsourceflow, qsof (see section 2.4.2 for details). Furthermore, we ascertained the implications of the lateral flow proxy data assumption on our C flux simulations by analyzing the sensitivity of the swamp C fluxes to changes in lateral water inputs (section 3.4). Our experiment results in section 3.4 showed that the swamp’s C flux displayed minimal sensitivity to changes in lateral water input, so we do not expect this proxy data assumption to significantly influence the outcome of our modeling experiment. Nevertheless, we will elaborate this in the next revision.
4. The Results section is overly descriptive, particularly in terms of parameter sensitivity rankings. The authors should better synthesize the findings, highlight dominant controls, and link them more explicitly to physical processes rather than listing parameter importance.
Response: Many thanks for the comment. As suggested, we will ensure that the findings are better synthesized with physically processes and not just the parameter importance in the next revision.
5. The Discussion is currently insufficient in depth. It should be strengthened by (i) systematic comparison with existing studies (especially bog and fen systems), (ii) clearer interpretation of process interactions, and (iii) explicit discussion of model limitations and applicability.
Response: Many thanks for the comment. We will ensure that the discussion section is better strengthened by incorporating the feedback above.
6. Model structural limitations need to be more explicitly addressed. The use of a one-dimensional model for a spatially heterogeneous swamp system may constrain interpretation, and this should be critically discussed.
Response: Many thanks for the comment. Based on the evaluation results of this manuscript and those of Afolabi et al. (2025), we are confident that the one-dimensional model (CoupModel) captured the critical biophysical and biochemical interactions and feedback of the swamp. However, we agree with the Reviewer that one-dimensional models like CoupModel are not without limitations. This concurrence informed our discussions (section 4.1 of Afolabi et al., 2025) on the implications of using a one-dimensional model for a spatially heterogenous swamp system like Beverly swamp. For instance, we mentioned in Afolabi et al. (2025) that snow disappearance in the open canopy section of the swamp during winter season (Smith & Woo, 1986) may not have been properly accounted for in the simulation because one-dimensional designs are targeted to mostly capture vertical and not spatially distributed transport and complicated phase change processes. This limitation may have affected energy partitioning in winter and contributed to the overestimation of sensible and ground heat fluxes during this season, explaining some of the mismatch in snow accumulation simulations and soil thermal processes in winter. We can also elaborate this aspect in the manuscript.
7. The manuscript would benefit from improved integration between sections. Currently, the linkage between objectives, methods, results, and conclusions is somewhat loose. The authors should ensure that each objective is clearly addressed and revisited in the results and discussion.
Response: Many thanks for the comment. We will ensure that the sections are better harmonized in the revision.
8. Overall, the manuscript has a solid foundation, but requires deeper analysis, clearer positioning of its contribution, and improved discussion to meet publication standards.
Response: Many thanks for the comments on the manuscript. We will ensure to incorporate the recommendations you highlighted above and revise the manuscript accordingly.
Citation: https://doi.org/10.5194/egusphere-2025-1368-AC3
-
AC1: 'Reply on RC1', Oluwabamise Afolabi, 25 Jul 2026
-
RC2: 'Comment on egusphere-2025-1368', Anonymous Referee #2, 17 Jun 2026
This manuscript uses the GLUE method to calibrate CoupModel and investigate parameter uncertainty, equifinality, and processes controlling soil respiration in a temperate swamp ecosystem. The topic is relevant because temperate swamps remain relatively understudied compared with other peatland ecosystems. However, several issues need to be addressed before it can be considered for publication:
- The abstract is too general to be informative. For example, the statement "the GLUE approach and its multicriteria constraints reduced the uncertainties associated with the modelling process and reasonably improved some of the simulation outcomes" should specify which model variables or processes showed improved performance and provide quantitative evidence to support this claim. The magnitude of uncertainty reduction and the corresponding changes in model performance should be reported using relevant metrics, such as changes in RMSE, R², bias, or other appropriate evaluation metrics. The statement “Global sensitivity analysis of the parameters identified the important parameters that greatly influence temperate swamp C flux simulations and the interconnections that exist between simulated variables and parameters. Plant-related processes and hydrological variables exerted the strongest control on soil respiration simulation.” provides no information on the actual model parameters, processes, and variables involved.
- The manuscript would benefit from a careful review of the language and overall presentation by the senior authors. Too often the text is difficult to follow and fails to communicate substantive information (it is not necessarily wrong, but it is imprecise, vague, or insufficiently informative). For example, the opening statement “A recent modelling study indicated that many aspects of a swamp’s thermal, hydrological and biogeochemical conditions could be adequately modelled, gaining insight into the ecosystem’s response to disturbance” provides little concrete information regarding what was modeled or what specific insights were obtained. As written, the sentence serves primarily as a citation rather than a synthesis of previous findings. The subsequent statement “however, parameter estimation remained difficult given the scarcity of previous studies in swamps”, is also unclear. The connection between this statement and the findings of Afolabi et al. (2025) is not adequately explained. It is unclear whether the difficulty in parameter estimation arises from limited observational data for calibration and validation, insufficient understanding of key processes, a lack of previous modelling studies to constrain parameter ranges, or some combination of these factors. In fact, the remainder of the paragraph appears to suggest that the primary issue is the scarcity of observational data rather than the scarcity of previous studies. The authors should clarify the source of the uncertainty and present the motivation for the study more clearly and logically. The statement “Modelling studies have shown that there are interrelationships between biogeochemical and biophysical processes in peatlands” is misleading. The existence of interactions between biogeochemical and biophysical processes is a fundamental characteristic of peatland ecosystems and has been demonstrated through observational, experimental, and modelling studies, not something that modelling studies uniquely "showed." I will not list all such instances here. I encourage the authors to carefully review the text and improve the precision and clarity.
- Objective ii: “present acceptable model structure and parameter distribution for simulating temperate swamp C”. How is “acceptable” defined? Are there quantitative criteria used to determine whether a model structure or parameter distribution is acceptable?
- Method: The criteria used to define behavioral simulations require further justification. The manuscript states that the thresholds were inspired by measurement uncertainty and previous modelling experience, but it is unclear how the specific R2 or ME thresholds for each variable were determined. Please provide a more rigorous justification for the selected thresholds and evaluate the sensitivity of the posterior parameter distributions to these choices.
- I am not fully convinced that differences between prior and posterior parameter distributions provide a robust measure of parameter sensitivity. Such differences may reflect parameter constraint or identifiability under the chosen observations and behavioral thresholds, and depend on the choice of prior parameter ranges. Please clarify the distinction between parameter sensitivity and parameter constraint/identifiability within the GLUE framework and discuss the implications for the interpretation of the results.
- The rationale for selecting 1998–2000 as the calibration period and 2022–2023 as the validation period is unclear. Please explain why these years were chosen, if observations between 2000-2022 are available, it would be helpful to justify why they were not used for calibration or validation.
- Figure 4: The temporal resolution of the validation data (2022–2023) appears substantially coarser than that of the calibration data (1998–2000), with the validation observations appearing to be monthly averages. Please clarify the temporal resolution of the observations and simulations shown in this figure. If different temporal aggregations were used for calibration and validation, please justify this choice and discuss how it may affect the comparison of model performance between the two periods.
- The scientific contribution of the study remains somewhat unclear. What new insights are obtained that were not already shown in Afolabi et al., Metzger et al. studies? Are the dominant controls, parameter sensitivities, equifinality patterns, or hydrological influences in temperate swamps fundamentally different from those reported for bogs or fens?
Citation: https://doi.org/10.5194/egusphere-2025-1368-RC2 -
AC2: 'Reply on RC2', Oluwabamise Afolabi, 25 Jul 2026
General comments
We thank the reviewers for the constructive feedback on our manuscript. Even though we have responded to the reviewer’s specific comments, we would like to address some of the general comments here. Both reviewers raised concerns about the overall language of the manuscript and the poor presentation of the manuscript’s contribution to scientific advancement. We do agree with the reviewers that the study’s linkage to scientific contexts can be further improved, as the current version is heavily focused on presenting the modeling details of the studied temperate swamp peatland. We initially focused on this aspect because the ecosystem is understudied and our manuscript (in addition to Afolabi et al., 2025) is one of the first process-based modeling studies on temperate swamp peatlands where very few data exist. We felt it was an avenue to present some of these foundational details for setting the modeling context of the ecosystem.
To address the general concerns highlighted above, we propose to do the following in the revision:
- We will better present the scientific importance of the paper;
- Include additional discussion sections on the uniqueness of temperate swamp peatlands systems, compared to other wetlands and peatlands. These discussions will cover the importance of parameters and processes that regulate seasonal surface and groundwater exchanges and its impacts on soil carbon (C) fluxes;
- Compare parameter posterior distributions and process interactions to previous peatland modeling studies by CoupModel and other existing swamp and wetland models. This would put current study into context and better highlight the modeling advancements of this study;
- We will revise the general language of the entire manuscript and thoroughly proofread it before resubmission.
We also observed that the reviewers raised concerns about the selection of the study’s calibration and validation periods and data. The separation of our calibration and validation periods was largely constrained by data availability. We acknowledged in the manuscript and Afolabi et al. (2025) that we relied on data from the 1980s, short studies from 1998 – 2000 and resumed measurements from 2022 to 2023. These data were mostly not continuous and were done by different groups with varying methods (excluding the validation period). Thus, there were uncertainties associated with these measurements, which we have discussed in the current manuscript and Afolabi et al. (2025).
To further evaluate the model and show the robustness of our model calibration, we will undertake additional experiments in the revision by using all (both calibration and validation, and/or swap the calibration and validation period) data for calibration and constraining the soil respiration with new RMSE threshold to show that our key results are not influenced by the choice of calibration and validation periods or subjectiveness of the defined criteria. We anticipate these additional experiments will not change the key findings of the paper because we initially conducted some of these experiments and analyed the results before reporting them in the current version.
We should also mention that our model evaluation manuscript is a follow-up work to Afolabi et al. (2025), which was a single-run experiment with the CoupModel to simulate the long-term carbon dynamics of the temperate swamp peatland. The initial single run set-up of Afolabi et al. (2025) in CoupModel for the Beverly Swamp was not without considerable uncertainties (see section 4 of Afolabi et al., 2025 for details). To address the highlighted issues in Afolabi et al. (2025), we conducted an uncertainty analysis of the initial model set-up with the GLUE approach in this manuscript. This approach assisted in identifying the errors associated with the different aspects (e.g., measurement, parameter and model structure) of the modelling exercise and also analyzed the extent that the model data discrepancies were explained by the measured data, model structure and parameters.
The findings of our study are important to the selection of model structure, parameter distribution and processes for modeling hydrology and carbon related processes in temperate swamp peatlands. Since this is the first attempt to evaluate the performance of CoupModel (version 6 and GLUE calibrated) in a temperate swamp peatland, where the hydroperiods, vegetation cover and other biophysical conditions are different from other peatland types (e.g., bogs and fens), we anticipate that the analysis of our manuscript will assist the model structure selection and parameterization of large-scale ecological models (e.g., CLASSIC and CaMP) when simulating swamp peatland carbon flux at regional and global scales, as the swamp category is currently missing in these models for Canadian studies. Furthermore, some of the important parameters and soil respiration constraining variables (e.g., WTL, VMC and LAI) that were identified in this study are expected to help inform the choice of variables to be measured in the field for carbon related studies. In addition, the behavioral models identified by the manuscript can be used for climate change assessment studies of the swamp peatland.
Specific Comments
1. The abstract is too general to be informative.
Response: We thank the Reviewer for the feedback. We will carefully update these changes in our revision.
2. The manuscript would benefit from a careful review of the language and overall presentation by the senior authors.
Response: We thank the Reviewer for the feedback. We will carefully revise the general language of the entire manuscript and thoroughly proofread it before resubmission.
3. Objective ii: “present acceptable model structure and parameter distribution for simulating temperate swamp C”. How is “acceptable” defined? Are there quantitative criteria used to determine whether a model structure or parameter distribution is acceptable?
Response: Our manuscript adopted the definition of Beven (2006); Beven and Lane (2022) that described “acceptable or behavioural” models as set of representations (model inputs, model structures, model parameter sets, model errors) that are acceptably consistent with the observations. These acceptable or behavioural models were generated by setting the plausibility thresholds or limits of acceptability presented in section 2.4 to identify ensembles (combination of model structure and parameter sets) that produce acceptable solutions that fit well with the multiple observed variables described in section 2.2. Simulations and parameter sets are assessed in GLUE as being acceptable based on the degree to which they match the observed data. In situations where the modeled variables are unrealistic, we categorized them as unacceptable solution and it is rejected. Ultimately, the model structure and parameter combinations and parameter value ranges that were used to generate the behavioral models were adjudged to be acceptable for simulating the swamp’s C flux. This analysis was also covered in the result sections (3.1 and 3.2). We can also elaborate this further in the revision
4. Method: The criteria used to define behavioral simulations require further justification. The manuscript states that the thresholds were inspired by measurement uncertainty and previous modelling experience, but it is unclear how the specific R2 or ME thresholds for each variable were determined. Please provide a more rigorous justification for the selected thresholds and evaluate the sensitivity of the posterior parameter distributions to these choices.
Response: Many thanks for the comment. We will make the justification for selecting the criteria for behavioral and non-behavioural simulations more explicit in the revision. We should highlight that the combination of R2 and ME were selected as performance indices because previous studies that used GLUE in CoupModel (e.g. Metzger et al., 2016a; Wang et al., 2022; Wu et al., 2019) have shown the 2 to be effective in capturing temporal variabilities (R2) and magnitude of error (ME). Secondly, our experience from the single run in Afolabi et al. (2025) and uncertainty estimation of field measurement in section in Table 2 of manuscript helped identify the minimum threshold for each of the variable. This is not uncommon because the choice of performance criteria for selecting acceptable models in the GLUE framework comes with some subjectiveness. We are open to conducting additional experiment to evaluate the sensitivity of the posterior parameter distributions to different criteria choices (e.g. RMSE constrain)
5. I am not fully convinced that differences between prior and posterior parameter distributions provide a robust measure of parameter sensitivity. Such differences may reflect parameter constraint or identifiability under the chosen observations and behavioral thresholds, and depend on the choice of prior parameter ranges. Please clarify the distinction between parameter sensitivity and parameter constraint/identifiability within the GLUE framework and discuss the implications for the interpretation of the results.
Response: Even though parameter sensitivity and parameters constraint are related, they are two distinct processes in the GLUE framework. The GLUE approach is able to complete global sensitivity analysis simultaneously with calibration process (parameter constraint). Unlike the parameter sensitivity analysis described in section 2.4.1, the parameter calibration process was achieved by the multicriteria constraint of the posterior. The difference between the prior and posterior distribution of the experiment mirrored both the parameter constraint and sensitivity. The narrow difference in the prior and posterior distribution in some of the parameters reflected the uniqueness of the processes in swamps relative to what was reported for other peatland types (e.g. bogs and Fens by Metzger et al., 2016). We agree with the author that is important to elaborate the difference between the parameter sensitivity and constraint and the implications for result interpretation in the next revision.
6. The rationale for selecting 1998–2000 as the calibration period and 2022–2023 as the validation period is unclear. Please explain why these years were chosen, if observations between 2000-2022 are available, it would be helpful to justify why they were not used for calibration or validation.
Response: The rationale for selecting the calibration and validation periods were majorly based on field data availability. Bi-weekly soil CO2 flux measurements were only available for 1998-2000 (calibration period) and 2022 -2023 (validation period). This means the data measurement for soil respiration in Beverly swamp was not continuous (No CO2 fluxes between 2000 and 2022). This sparsity of continuous field data influenced our selection of the calibration and validation periods to coincide with periods of data availability. We will make this selection rationale more explicit in our revision.
7. Figure 4: The temporal resolution of the validation data (2022–2023) appears substantially coarser than that of the calibration data (1998–2000), with the validation observations appearing to be monthly averages. Please clarify the temporal resolution of the observations and simulations shown in this figure. If different temporal aggregations were used for calibration and validation, please justify this choice and discuss how it may affect the comparison of model performance between the two periods.
Response: We thank the reviewers for the important comment above. We should add that our study explicitly described the differences in the measurement techniques, frequency and length of measurements, and spatial variability of soil respiration data measurements that were used for model initialization, calibration (1998 – 2000) and validation (2022 – 2023) in Afolabi et al. (2025) and Table 2 and section 3.2.1 of this manuscript.
Bi-weekly soil CO2 flux measurements obtained by closed opaque chamber method and analyzed with gas chromatograph (1998-2000) or measured directly with LI-COR portable greenhouse gas analyzer (LI-7810) (2022 -2023) were sourced from Davidson et al., 2019 and Schmidt & Strack, 2024, respectively. This means the data measurement for soil respiration in Beverly swamp was not continuous and it was conducted by different research groups with varying methods over the study period. This sparsity of continuous field data influenced our selection of the calibration and validation periods to coincide with periods of data availability. Even though more measurements were made for the calibration period (1998 – 2000) than the validation period (2022 -2023), the two periods have a bi-weekly temporal scale for the soil CO2 flux measurements, and our simulations were validated against this time scale. We highlighted the uncertainties associated with these measurements and the implications on our simulations in Afolabi et al. (2025) and this manuscript. Nevertheless, we will elaborate this in our next revision.
8. The scientific contribution of the study remains somewhat unclear. What new insights are obtained that were not already shown in Afolabi et al., Metzger et al. studies? Are the dominant controls, parameter sensitivities, equifinality patterns, or hydrological influences in temperate swamps fundamentally different from those reported for bogs or fens?
Response: We thank the reviewers for the important comment above. We agree that the manuscript needs to better present the scientific importance of the study.
The main scientific contribution of this manuscript is to improve our understanding of the long-term processes, interactions and feedback that mediate carbon cycling in temperate swamp peatland, where very few data exist on the ecosystem and it is understudied. To the best of our knowledge, this is one of the first process-based modeling studies on temperate swamp peatland carbon cycling and it is also a first attempt to evaluate the performance of CoupModel (version 6 and GLUE calibrated) in a temperate swamp peatland, where the hydroperiods, vegetation cover and other biophysical conditions are different from other peatland types (e.g., bogs and fens).Even though this manuscript is a follow-up to Afolabi et al. (2025a), it is very distinct because it is firstly a model evaluation of the initial model set-up (single-run) of Afolabi et. al. (2025) in CoupModel, which was embedded with uncertainties, while simulating the long-term C dynamics of the temperate swamp peatland. Hence, there was the need for a follow-up publication to undertake uncertainty analysis of the initial model set-up and to discuss the extent that the model data discrepancies can be explained by the measured data, model structure and parameters. Consequently, we undertook a systematic uncertainty analysis by using the Generalized Likelihood Uncertainty Estimation (GLUE) approach to identify the errors associated with the different aspects (e.g., measurement, parameter and model structure) of the modelling exercise because this determines the usefulness of the model outcomes.
The findings of our study are important to the selection of model structure, parameter distribution and unique processes for simulating hydrology and carbon related processes in temperate swamp peatlands that are currently lacking. Since this is the first attempt to evaluate the performance of CoupModel (version 6 and GLUE calibrated) in a temperate swamp peatland, where the hydroperiods, vegetation cover and other biophysical conditions are different from other peatland types (e.g., bogs and fens), we anticipate that the analysis of our manuscript will assist the model structure selection and parameterization of large-scale ecological models (e.g., CLASSIC and CaMP) when simulating swamp peatland carbon flux at regional and global scales, as the swamp category is currently missing in these models for Canadian studies. Furthermore, some of the important parameters and soil respiration constraining variables (e.g., WTL, VMC and LAI) that were identified in this study are expected to help inform the choice of variables to be measured in the field for carbon related studies. In addition, the behavioural models identified by the manuscript can be used for climate change assessment studies of the swamp peatland.
Unlike Metzger et al. (2015 & 2016a) which relied on older CoupModel versions (e.g. v4 and v5) to simulate the biophysical conditions and biogeochemical cycles of ecosystems such as fens and bogs peatlands (but not temperate swamp peatlands), our study systematically calibrated the CoupModel (v6) using the GLUE method to reproduce critical interactions and feedbacks that prevails between biophysical processes and the carbon cycle in a temperate swamp over a space of 40 years. Please see Table 3 for specific differences. Furthermore, we briefly compared the results of our modeling study in a temperate swamp peatland to that of bogs or fens (e.g. Metzger et al.) in some sections of the manuscript (e.g., Line 572to Line 574). However, we agree with the reviewer that it is important to further elaborate the differences between the dominant controls, parameter sensitivities, equifinality patterns, or hydrological influences of our study and those reported for bogs or fens in the next revision of this manuscript. Many thanks.
Citation: https://doi.org/10.5194/egusphere-2025-1368-AC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,445 | 808 | 139 | 2,392 | 288 | 153 | 151 |
- HTML: 1,445
- PDF: 808
- XML: 139
- Total: 2,392
- Supplement: 288
- BibTeX: 153
- EndNote: 151
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript addresses a relevant topic and applies established modelling and uncertainty analysis approaches; however, several substantive issues need to be addressed before it can be considered for publication. A moderate revision is recommended.
The scientific contribution remains somewhat limited. The study mainly applies existing methods (CoupModel + GLUE + GSA) to a swamp system, but the advancement beyond previous studies is not sufficiently articulated. The authors should clearly define the key scientific questions and explicitly state what new insights are gained at the process or system level.
Section 5.1 should be moved to the Discussion section. Its current content focuses on interpretation and mechanism explanation rather than presenting new results. Integrating it into the Discussion would improve the logical coherence of the manuscript.
The data basis raises concerns regarding robustness. Key calibration and validation datasets are sparse and temporally inconsistent (e.g., combining 1998–2000 and 2022–2023 observations). The authors should better justify this strategy and discuss its implications for model reliability and uncertainty.
The treatment of lateral water flow using proxy data is a critical assumption. Although acknowledged, it is not sufficiently quantified. A more rigorous uncertainty assessment or sensitivity discussion specific to this assumption is needed.
The Results section is overly descriptive, particularly in terms of parameter sensitivity rankings. The authors should better synthesize the findings, highlight dominant controls, and link them more explicitly to physical processes rather than listing parameter importance.
The Discussion is currently insufficient in depth. It should be strengthened by (i) systematic comparison with existing studies (especially bog and fen systems), (ii) clearer interpretation of process interactions, and (iii) explicit discussion of model limitations and applicability.
Model structural limitations need to be more explicitly addressed. The use of a one-dimensional model for a spatially heterogeneous swamp system may constrain interpretation, and this should be critically discussed.
The manuscript would benefit from improved integration between sections. Currently, the linkage between objectives, methods, results, and conclusions is somewhat loose. The authors should ensure that each objective is clearly addressed and revisited in the results and discussion.
Overall, the manuscript has a solid foundation, but requires deeper analysis, clearer positioning of its contribution, and improved discussion to meet publication standards.