the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Nov 2022
The Permafrost and Organic LayEr module for Forest Models (POLE-FM) 1.0
Winslow D. Hansen
Adrianna Foster
Bejamin Gaglioti
Rupert Seidl
Werner Rammer
Abstract. Climate change and increased fire are eroding the resilience of boreal forests. This is problematic because boreal vegetation and the cold soils underneath store approximately 30 % of all terrestrial carbon. Society urgently needs projections of where, when, and why boreal forests are likely to change. Permafrost (i.e., subsurface material that remains frozen for at least two consecutive years) and the thick soil-surface organic layers (SOLs) that insulate permafrost are important controls of boreal forest dynamics and carbon cycling. However, both are rarely included in process-based vegetation models used to simulate future ecosystem trajectories. To address this challenge, we developed a computationally efficient permafrost and SOL module that operates at fine spatial (1 ha) and temporal (daily) resolutions. The module mechanistically simulates daily changes in depth to permafrost, annual SOL accumulation, and their complex effects on boreal forest structure and functions. We coupled the module to an established forest landscape model, iLand, and benchmarked the model in interior Alaska at spatial scales of stands (1 ha) to landscapes (61,000 ha) and over temporal scales of days to centuries. The coupled model could generate intra- and inter-annual patterns of snow accumulation and active layer depth (portion of soil column that thaws throughout the year) consistent with independent observations in 17 instrumented forest stands. The model was also skilled at representing the distribution of near-surface permafrost presence in a topographically complex landscape. We simulated 34.6 % of forested area in the landscape as underlain by permafrost; a close match to the estimated 33.4 % from the benchmarking product. We further determined that the model could accurately simulate moss biomass, SOL accumulation, fire activity, tree-species composition, and stand structure at the landscape scale. Modular and flexible representations of key biophysical processes that underpin 21st-century ecological change are an essential next step in vegetation simulation to reduce uncertainty in future projections and to support innovative environmental decision making. We show that coupling a new permafrost and SOL module to an existing forest landscape model increases the model’s utility for projecting forest futures at high latitudes. Process-based models that represent relevant dynamics will catalyze opportunities to address previously intractable questions about boreal forest resilience, biogeochemical cycling, and feedbacks to regional and global climate.
-
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
(2439 KB)
-
Supplement
(23 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2439 KB) -
Supplement
(23 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Winslow D. Hansen et al.
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-1062', Anonymous Referee #1, 10 Jan 2023
General comments
Hansen et al. present a computationally efficient permafrost and soil organic layer module and its coupling to an established forest landscape model. The new module can be used to simulate the annual soil-surface organic layer accumulation and the interannual and seasonal patterns of snow accumulation and active layer depth. Coupled with iLand, the model is used to simulate moss biomass, fire activity, forest composition, stand structure, the soil-surface organic layer accumulation, and the permafrost distribution in a complex landscape in interior Alaska. The computational efficiency of the new module offers great opportunities for the simulation of large spatial extents – also demonstrated here.
The manuscript in its current form is well-written and well-structured but requires improvements. It will then be an exciting addition to a growing body of work concerned with the complex interactions between boreal forests and permafrost. As such, some of the model results fit well with observations (e.g. the permafrost distribution). Still, it is unclearly described where exactly these observations were made and how the data was selected from the previous studies cited. The authors should include the most important data and study site descriptions rather than pointing the readers at former studies. To me, some of the figures seem to be suggesting relationships that aren’t necessarily statistically proven. Adding to this, the time periods for the comparison between observed and simulated data are different between different values (e.g. annual fire probability is compared for the model output years 201-300, and near-surface permafrost for the years 271-300). The reason for this should be discussed.
The modelling approach is reasonable and I don’t see any major issues with the new modules. Nevertheless, the modelled active layer thaw and freezing are delayed compared to observations, and the snow cover height does not fit the observations very well (from what is shown). This raises the question if the modelling results regarding the snow phenology are meaningful. While this can be explained by the fact that the snow parameterization is simple and does not include canopy interception, snow compaction or redistribution, the snow depth is one of the key factors impacting the thermal and hydrological permafrost regime. Similarly, shading and the reduction of wind underneath the forest canopy have been found to severely impact the thermal and hydrological regime of the ground, dampening the below-ground temperatures and changing the available plant water. I understand, that this is not the main focus here but should be mentioned and “used” to explain the found differences between model simulations and observations.
It would additionally be interesting to show the differences and especially the improvements that the new module adds to the overall model performance in terms of stand structure and tree species composition. This could be done by showing an additional map.
Specific comments
l.34: Here it should be noted that while air temperatures are rising (warming) another important aspect are changes in the precipitation patterns (leading to droughts -> increasing fire/pest risks)
l.56: This is the case for Alaska/Canada but not in Eastern Siberia. Specify the geographical focus here already.
l.58: Also, hydrology (strongly interlinked with topography).
l.59: Furthermore, boreal forests are important in protecting permafrost (shading, snow cover interception, lowering of turbulent heat fluxes, litter layer):
- Chang et al., 2015, Arctic Antarct. Alpine Res. 47 267–79, https://doi.org/10.1657/AAAR00C-14-016
- Fisher et al., 2016, Glob. Change Biol. 22 3127–40, https://doi.org/10.1111/gcb.13248
- Stuenzi et al., 2021, Environ. Res. Lett. 16 084045, https://doi.org/10.1088/1748-9326/ac153d
l.60/61: active-layer vs. active layer
l.70: It’s unclear to me what’s meant here in regard to Kruse et al. 2022
l.73: This is true but there are many newer models available:
- Karra, S., Painter, S. L., and Lichtner, P. C.: Three-phase numerical model for subsurface hydrology in permafrost-affected regions (PFLOTRAN-ICE v1.0), The Cryosphere, 8, 1935–1950, https://doi.org/10.5194/tc-8-1935-2014, 2014.
- Perreault et al. (2021) Numerical modelling of permafrost dynamics under climate change and evolving ground surface conditions: application to an instrumented permafrost mound at Umiujaq, Nunavik (Québec), Canada, Écoscience, 28:3-4, 377-397, https://doi.org/10.1080/11956860.2021.1949819
- Yokohata et al.: Model improvement and future projection of permafrost processes in a global land surface model, Progress in Earth and Planetary Science (2020): https://doi.org/10.1186/s40645-020-00380-w
- Westermann et al.: Simulating the thermal regime and thaw processes of ice-rich permafrost ground with the land-surface model CryoGrid 3, Geosci. Model Dev., 9, 523–546, https://doi.org/10.5194/gmd-9-523-2016, 2016.
l.98: Is this the density of falling snow or the snowpack below the canopy? What about the representation of transient internal snow properties (such as e.g. snow compaction)?
l.124: Air temperature at what height? Above the canopy? There are no canopy density-specific parameterizations/factors for shading of the ground, precipitation interception, or changes in the turbulent surface fluxes, correct? While these are very important factors for permafrost conditions under boreal forest canopies, this is not the focus here. I would, nevertheless, suggest discussing this in more detail in the conclusions section (and possibly restructuring it into a discussion and a short conclusion section).
l.147: Why 0.7?
l.149: Where is the SLA value for moss from?
l.172: I wasn’t able to find this value in the given reference.
l.176: full full
l.180: I’d propose mentioning the reference here.
l.236: Where were these measurements conducted? I’d suggest describing the validation studies in more detail here rather than having the readers go through all the individual studies cited.
l.246: This is unclear: How many are located in mid-elevations, and what is the bias in low- or high-elevation locations?
l.340: The model struggles to simulate the timing of thawing and freezing, 15 days is a rather long period compared to the overall short unfrozen season. I suppose this is mostly due to the o
l.353: close
l.398: While this is one of the major advances of the presented model, I’d suggest adding this to the main discussion and omit from adding an extra subchapter for one sentence only.
l.390: This conclusion reads more like a discussion section, and could maybe be split up into a more extensive discussion and a short conclusion.
l.406: Also, because there is no mechanistic snow parameterization included.
Figure 1: Would you need any references here for the vectorized trees or is this a modified photograph?
Figure 2: The shown relationships (solid lines) between simulated and observed values are misleading. The simulated snow depth seems to be overestimated on average. Instead of the solid lines, it would be interesting to see actual correlation statistics here.
Figure 3: Here, only one of seven forest stands is shown which leaves the reader wondering whether this one shows the best fit out of the seven validation sites.
Figure 4: Is the dark grey area the “sampled landscapes”?
Citation: https://doi.org/10.5194/egusphere-2022-1062-RC1 - AC1: 'Reply on RC1', Winslow Hansen, 08 Mar 2023
-
RC2: 'Comment on egusphere-2022-1062', Anonymous Referee #2, 30 Jan 2023
This paper presents a useful addition to the land modeling community, with a well-described model of permafrost that can be coupled to other land models and has code and documentation available. The need for such a model I felt was well-justified and the paper provides detailed model equations and background, describes and demonstrates the model capabilities, and compares a variety of model outputs against observational data. Overall the model appears to produce reasonable results, and, with some modifications, I believe this paper will make an excellent addition to GMD.
My primary concerns are with the performance claims that I believe need additional support from model validation and with the minimal discussion of model limitations. While the model is compared to observational data, several aspects of this could use further clarification. First, it is unclear to me why the years selected are necessarily comparable to the model results and why different years are chosen for different comparisons, which may just speak to a need for some additional justification for this in the text. Additionally, differences between model results and data are quantified frequently in terms of RMSE but the reader has no real perspective to understand the relative significance of these RMSE values. This could be addressed by including the range or uncertainty of the observational and simulated values and by giving some measure of statistical significance. Also, some discrepancies between model results and observations seem to not really be discussed or have only limited discussion, such as the model’s near-surface permafrost extent by aspect being considerably worse in the north aspect than others, later maximum annual active-layer depth or earlier freeze depth. It would be helpful to address these more and help the reader understand potential causes and the implications of these for interpreting other model results.
Model limitations (such as critical assumptions or missing processes and their implications for results) are only lightly touched on in the conclusions section, and I believe the paper would benefit from further elaboration on these points, to help the reader put this model in perspective and better understand how to consider its results. In the interest of putting this model in perspective, it would additionally be useful to see some comparison of iLand’s performance with and without this new module to some of the observational datasets to demonstrate the value of this module beyond providing additional outputs.
Specific comments:
Equation 9 - Would be helpful to add units to the constant to make this conversion clearer to the reader
L.280 Unclear to me why we expect these model simulation years to correspond to these data years. I suggest adding some clarification on this, here and in other places where model results are compared to time-specific data.
L.353-354 Somewhat unclear. Would be helpful to elaborate more on the importance of aspect and mention the differences in results between north versus other aspects.
L.375 Would be useful to add comparison here to iLand results without the permafrost module
L.397-398 The claim “Benchmarking results demonstrate the model recreates temporal and spatial patterns consistent with observations” feels not completely supported by the text and I believe needs more justification or caveats.
L.398-399 The claim “Our model will contribute to improving 21st -century projections of boreal forest change.” is not really supported by the current results, but would be by the addition of a figure or data showing improvements to iLand’s projections with the addition of the permafrost module.
Figure 3. The representation of frozen and unfrozen soil in this figure is somewhat confusing, as it seems like there should be two different fills - one for observed, one for simulated - but instead there is only the one. Can this be separated out or otherwise clarified?
Figure 5. Maybe add the years for the observed dataset in this figure caption, since the simulation years are given.
Figure 7. Some clarification on the caption for Panel B would be helpful, as it took me a minute to figure out what was meant here.
Citation: https://doi.org/10.5194/egusphere-2022-1062-RC2 - AC2: 'Reply on RC2', Winslow Hansen, 08 Mar 2023
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-1062', Anonymous Referee #1, 10 Jan 2023
General comments
Hansen et al. present a computationally efficient permafrost and soil organic layer module and its coupling to an established forest landscape model. The new module can be used to simulate the annual soil-surface organic layer accumulation and the interannual and seasonal patterns of snow accumulation and active layer depth. Coupled with iLand, the model is used to simulate moss biomass, fire activity, forest composition, stand structure, the soil-surface organic layer accumulation, and the permafrost distribution in a complex landscape in interior Alaska. The computational efficiency of the new module offers great opportunities for the simulation of large spatial extents – also demonstrated here.
The manuscript in its current form is well-written and well-structured but requires improvements. It will then be an exciting addition to a growing body of work concerned with the complex interactions between boreal forests and permafrost. As such, some of the model results fit well with observations (e.g. the permafrost distribution). Still, it is unclearly described where exactly these observations were made and how the data was selected from the previous studies cited. The authors should include the most important data and study site descriptions rather than pointing the readers at former studies. To me, some of the figures seem to be suggesting relationships that aren’t necessarily statistically proven. Adding to this, the time periods for the comparison between observed and simulated data are different between different values (e.g. annual fire probability is compared for the model output years 201-300, and near-surface permafrost for the years 271-300). The reason for this should be discussed.
The modelling approach is reasonable and I don’t see any major issues with the new modules. Nevertheless, the modelled active layer thaw and freezing are delayed compared to observations, and the snow cover height does not fit the observations very well (from what is shown). This raises the question if the modelling results regarding the snow phenology are meaningful. While this can be explained by the fact that the snow parameterization is simple and does not include canopy interception, snow compaction or redistribution, the snow depth is one of the key factors impacting the thermal and hydrological permafrost regime. Similarly, shading and the reduction of wind underneath the forest canopy have been found to severely impact the thermal and hydrological regime of the ground, dampening the below-ground temperatures and changing the available plant water. I understand, that this is not the main focus here but should be mentioned and “used” to explain the found differences between model simulations and observations.
It would additionally be interesting to show the differences and especially the improvements that the new module adds to the overall model performance in terms of stand structure and tree species composition. This could be done by showing an additional map.
Specific comments
l.34: Here it should be noted that while air temperatures are rising (warming) another important aspect are changes in the precipitation patterns (leading to droughts -> increasing fire/pest risks)
l.56: This is the case for Alaska/Canada but not in Eastern Siberia. Specify the geographical focus here already.
l.58: Also, hydrology (strongly interlinked with topography).
l.59: Furthermore, boreal forests are important in protecting permafrost (shading, snow cover interception, lowering of turbulent heat fluxes, litter layer):
- Chang et al., 2015, Arctic Antarct. Alpine Res. 47 267–79, https://doi.org/10.1657/AAAR00C-14-016
- Fisher et al., 2016, Glob. Change Biol. 22 3127–40, https://doi.org/10.1111/gcb.13248
- Stuenzi et al., 2021, Environ. Res. Lett. 16 084045, https://doi.org/10.1088/1748-9326/ac153d
l.60/61: active-layer vs. active layer
l.70: It’s unclear to me what’s meant here in regard to Kruse et al. 2022
l.73: This is true but there are many newer models available:
- Karra, S., Painter, S. L., and Lichtner, P. C.: Three-phase numerical model for subsurface hydrology in permafrost-affected regions (PFLOTRAN-ICE v1.0), The Cryosphere, 8, 1935–1950, https://doi.org/10.5194/tc-8-1935-2014, 2014.
- Perreault et al. (2021) Numerical modelling of permafrost dynamics under climate change and evolving ground surface conditions: application to an instrumented permafrost mound at Umiujaq, Nunavik (Québec), Canada, Écoscience, 28:3-4, 377-397, https://doi.org/10.1080/11956860.2021.1949819
- Yokohata et al.: Model improvement and future projection of permafrost processes in a global land surface model, Progress in Earth and Planetary Science (2020): https://doi.org/10.1186/s40645-020-00380-w
- Westermann et al.: Simulating the thermal regime and thaw processes of ice-rich permafrost ground with the land-surface model CryoGrid 3, Geosci. Model Dev., 9, 523–546, https://doi.org/10.5194/gmd-9-523-2016, 2016.
l.98: Is this the density of falling snow or the snowpack below the canopy? What about the representation of transient internal snow properties (such as e.g. snow compaction)?
l.124: Air temperature at what height? Above the canopy? There are no canopy density-specific parameterizations/factors for shading of the ground, precipitation interception, or changes in the turbulent surface fluxes, correct? While these are very important factors for permafrost conditions under boreal forest canopies, this is not the focus here. I would, nevertheless, suggest discussing this in more detail in the conclusions section (and possibly restructuring it into a discussion and a short conclusion section).
l.147: Why 0.7?
l.149: Where is the SLA value for moss from?
l.172: I wasn’t able to find this value in the given reference.
l.176: full full
l.180: I’d propose mentioning the reference here.
l.236: Where were these measurements conducted? I’d suggest describing the validation studies in more detail here rather than having the readers go through all the individual studies cited.
l.246: This is unclear: How many are located in mid-elevations, and what is the bias in low- or high-elevation locations?
l.340: The model struggles to simulate the timing of thawing and freezing, 15 days is a rather long period compared to the overall short unfrozen season. I suppose this is mostly due to the o
l.353: close
l.398: While this is one of the major advances of the presented model, I’d suggest adding this to the main discussion and omit from adding an extra subchapter for one sentence only.
l.390: This conclusion reads more like a discussion section, and could maybe be split up into a more extensive discussion and a short conclusion.
l.406: Also, because there is no mechanistic snow parameterization included.
Figure 1: Would you need any references here for the vectorized trees or is this a modified photograph?
Figure 2: The shown relationships (solid lines) between simulated and observed values are misleading. The simulated snow depth seems to be overestimated on average. Instead of the solid lines, it would be interesting to see actual correlation statistics here.
Figure 3: Here, only one of seven forest stands is shown which leaves the reader wondering whether this one shows the best fit out of the seven validation sites.
Figure 4: Is the dark grey area the “sampled landscapes”?
Citation: https://doi.org/10.5194/egusphere-2022-1062-RC1 - AC1: 'Reply on RC1', Winslow Hansen, 08 Mar 2023
-
RC2: 'Comment on egusphere-2022-1062', Anonymous Referee #2, 30 Jan 2023
This paper presents a useful addition to the land modeling community, with a well-described model of permafrost that can be coupled to other land models and has code and documentation available. The need for such a model I felt was well-justified and the paper provides detailed model equations and background, describes and demonstrates the model capabilities, and compares a variety of model outputs against observational data. Overall the model appears to produce reasonable results, and, with some modifications, I believe this paper will make an excellent addition to GMD.
My primary concerns are with the performance claims that I believe need additional support from model validation and with the minimal discussion of model limitations. While the model is compared to observational data, several aspects of this could use further clarification. First, it is unclear to me why the years selected are necessarily comparable to the model results and why different years are chosen for different comparisons, which may just speak to a need for some additional justification for this in the text. Additionally, differences between model results and data are quantified frequently in terms of RMSE but the reader has no real perspective to understand the relative significance of these RMSE values. This could be addressed by including the range or uncertainty of the observational and simulated values and by giving some measure of statistical significance. Also, some discrepancies between model results and observations seem to not really be discussed or have only limited discussion, such as the model’s near-surface permafrost extent by aspect being considerably worse in the north aspect than others, later maximum annual active-layer depth or earlier freeze depth. It would be helpful to address these more and help the reader understand potential causes and the implications of these for interpreting other model results.
Model limitations (such as critical assumptions or missing processes and their implications for results) are only lightly touched on in the conclusions section, and I believe the paper would benefit from further elaboration on these points, to help the reader put this model in perspective and better understand how to consider its results. In the interest of putting this model in perspective, it would additionally be useful to see some comparison of iLand’s performance with and without this new module to some of the observational datasets to demonstrate the value of this module beyond providing additional outputs.
Specific comments:
Equation 9 - Would be helpful to add units to the constant to make this conversion clearer to the reader
L.280 Unclear to me why we expect these model simulation years to correspond to these data years. I suggest adding some clarification on this, here and in other places where model results are compared to time-specific data.
L.353-354 Somewhat unclear. Would be helpful to elaborate more on the importance of aspect and mention the differences in results between north versus other aspects.
L.375 Would be useful to add comparison here to iLand results without the permafrost module
L.397-398 The claim “Benchmarking results demonstrate the model recreates temporal and spatial patterns consistent with observations” feels not completely supported by the text and I believe needs more justification or caveats.
L.398-399 The claim “Our model will contribute to improving 21st -century projections of boreal forest change.” is not really supported by the current results, but would be by the addition of a figure or data showing improvements to iLand’s projections with the addition of the permafrost module.
Figure 3. The representation of frozen and unfrozen soil in this figure is somewhat confusing, as it seems like there should be two different fills - one for observed, one for simulated - but instead there is only the one. Can this be separated out or otherwise clarified?
Figure 5. Maybe add the years for the observed dataset in this figure caption, since the simulation years are given.
Figure 7. Some clarification on the caption for Panel B would be helpful, as it took me a minute to figure out what was meant here.
Citation: https://doi.org/10.5194/egusphere-2022-1062-RC2 - AC2: 'Reply on RC2', Winslow Hansen, 08 Mar 2023
Peer review completion
Journal article(s) based on this preprint
Winslow D. Hansen et al.
Data sets
Inputs and outputs for iLand simulations Winslow Hansen https://doi.org/10.25390/caryinstitute.21339090
Model code and software
iLand source code Werner Rammer Winslow Hansen https://doi.org/10.25390/caryinstitute.21339090
Winslow D. Hansen et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 353 | 95 | 15 | 463 | 22 | 4 | 4 |
- HTML: 353
- PDF: 95
- XML: 15
- Total: 463
- Supplement: 22
- BibTeX: 4
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2439 KB) - Metadata XML
-
Supplement
(23 KB) - BibTeX
- EndNote
- Final revised paper