the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Mar 2023
The Ability of Hydrologic-Land Surface Models to Concurrently Simulate Permafrost and Hydrology
Abstract. Hydrologic-land surface models (H-LSMs) provide physically-based understanding and predictions of the current and future states of the world’s vast high-latitude permafrost regions. Two major challenges, however, hamper their parametrization and validation when concurrently representing hydrology and permafrost. One is the high computational complexity, exacerbated by the need to include a deep soil profile to adequately capture the freeze/thaw cycles and heat storage. The other is that soil-temperature data are severely limited, and traditional model validation, based on streamflow, can show the right fit to these data for the wrong reasons. There are few observational sites for such vast, heterogeneous regions, and remote sensing provides only limited support. In light of these challenges, we develop 16 parametrizations of a Canadian H-LSM, MESH, for the sub-arctic Liard River Basin and validate them using three data sources: streamflows at multiple gauges, soil temperature profiles from few available boreholes, and multiple permafrost maps. The different parametrizations favor different sources of data and it is challenging to configure a model faithful to all three data sources, which are at times inconsistent with each other. Overall, the results show that: (1) surface insulation through snow cover primarily regulates permafrost dynamics after model initialization effects decay over, relatively long time and (2) different parametrizations yield different partitioning patterns of solid-vs-liquid soil-water and produce different low-flow but similar high-flow regimes. We conclude that, given data scarcity, an ensemble of model parametrizations is essential to provide a reliable picture of the current states and future spatio-temporal co-evolution of permafrost and hydrology.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(2294 KB)
-
Supplement
(4681 KB)
-
This preprint has been withdrawn.
- Preprint
(2294 KB) - Metadata XML
-
Supplement
(4681 KB) - BibTeX
- EndNote
Interactive discussion
Status: closed
-
RC1: 'Comment on tc-2023-20', Anonymous Referee #1, 21 Apr 2023
Review of the manuscript “The Ability of Hydrologic-Land Surface Models to Concurrently Simulate Permafrost and Hydrology” submitted for publication in The Cryosphere by Mohamed S. Abdelhamed, Mohamed E. Elshamy, Saman Razavi, and Howard S. Wheater.
This reviewer has expertise in permafrost field observations and in numerical modelling of heat and water transfer in frozen soil.
The paper compares model results generated with an ensemble of different parameterizations for the model MESH with observations of ground temperature and river discharge in the Liard River Basin in Canada. Testing and improving the quality of concurrent simulation of permafrost dynamics and river discharge is an important research area and well within the scope of The Cryosphere. The tools and data used are relevent for this topic. The present study, however, clearly falls short of the quality required for this journal.
My overall impression is that this manuscript is not yet ready for publication and should be resubmitted after considerable revision. The title suggests generic insight (“ability of Hydrologic-Land Surface Models”, the running title suggests insight about challenges (“H-LSM challenges to model…”), whereas the actual manuscript compares one set of ensemble results from one single models with observations. The main research question is only presented near the end of the manuscript: "[...] is it possible to concurrently simulate permafrost dynamics and hydrology over a large domain using an H-LSM?”. It is unclear what threshold needs to be met for this to be answered affirmatively, neither is the state of the art of existing models and their success clarified nor the exact gap in knowledge identified. The abstract then again states something different: “…the study results show that: (1) […] snow-cover primarily regulates permafrost […] and (2) different parametrizations yield different partitioning patterns […].” The former is textbook knowledge and was not the question addressed in this study. Why is it shown as one of two results worth communicating in the abstract? The latter needs further elaboration to become valuable insight. Then, the authors conclude that “an ensemble of model parametrizations is essential to provide a reliable picture of the current states and future spatio-temporal co-evolution of permafrost and hydrology.”. The appropriateness of ensembles has not been the question addressed in the paper, nor does it easily follow from the results presented. For resubmission, a clear and relevant research question needs to be identified, investigated, and adressed in the conclusion.
Several of the details in this study also merit more detailed attention. For example, with permafrost and long-term ground temperature being long-term average responses to surface conditions, why is spinup done with single years, only?
Citation: https://doi.org/10.5194/tc-2023-20-RC1 -
RC2: 'Comment on tc-2023-20 - Complete Review', Anonymous Referee #2, 07 May 2023
The authors present the implementation of the MESH hydrological/land surface model in the Liard River Basin (275000 km2) Canada in the period 1979-2019, to simulate permafrost and streamflow dynamics, which are then compared with ground temperature and active layer observations, distributed observation-based maps of permafrost and 4 streamgauge observations. The main scope of the manuscript is to assess the model capability to reproduce concurrently local and gridded permafrost dynamics and streamflow and to test 16 different model configurations (Table 3), where the initialization (spin-up years) and 2 parameters, related to minimum snow depth to consider a grid cell snow covered and vertical distribution of organic content of the soil, are modified.
The article is largely technically sounds, but it is completely uninspired, spending large parts just describing what we can see in the figures and tables (section 3.1.2 is almost unreadable, it uses even bullet points), but it does not provide insights on cryospheric or hydrological processes, on how to improve the models or the modeling approach, beyond the classic conclusions “we need more and better data”, “permafrost maps are uncertain”. The entire manuscript reads as a model application (by the way embedded in two other manuscripts in preparation (LL 617-622)) and not even one of the most comprehensive applications as some choice on the downscaling of the input data is debatable (see minor comments), and only few outputs and variables are analyzed. Why are authors not exploring and comparing also surface temperature and snow cover from remote sensing? The most critical comment, we learn very little from this study, to the point that the authors themselves conclude: “it does not provide a concrete answer to our main research question,” (LL 592), or “this result highlights the difficulty that modellers may encounter when configuring models to simultaneously simulate permafrost and hydrology”. This is repeated even twice (LL 591-592, LL 697-698), and many other repetitions are present. The entire conclusion - section 4 - is largely a repetition of the result section, sometime even with new result introduced (LL 676-689). The discussion is very limited, likely because there is not much to discuss. In summary, I am sure the manuscript has value for MESH users, and it would be fine as a project report, or technical report. For a research article, however, it is not fine. I am ok, if the authors use one model in one catchment only, but they need to use it to address a compelling question, carrying out dedicated numerical experiments, or learn from the model application something that might interest the scientific community, beyond saying “modelers might encounter challenges”. These aims are not achieved here.
Minor Comments
LL 41-42. I think the choice of the literature to introduce state-of-the-art Land-Surface-Models could be quite different and definitely more updated.
LL 92. Is a large suite of input data an issue nowadays? Are these data more than what a LSM needs? I do not think so.
LL 134-143. I think it would be good to state at this stage already how many streamgauges are available.
LL 150-153. The use of brackets is a bit confusing.
LL158. I would suggest referring to “Soil organic matter” rather than “organic matter”.
LL 163. It is not clear what “retention capacity” refers to here. Do you mean water content at field capacity?
LL 185-190. How many prognostic surface temperatures are solved in the model? Do different land covers share the same soil hydrology and soil temperature? How are the tiles interacting with each other if at all? I think these model aspects should be reported here as well.
LL 205. Why longwave radiation is distributed uniformly over the day? Usually, longwave radiation is assumed to depend on the air temperature at power of 4. The diurnal cycle is relevant. This is an incorrect assumption.
LL 211. Why did you assume a constant relative humidity? What is known to be fairly constant at daily scale is specific humidity or equivalently vapor pressure, relative humidity has a clear diurnal variability, at least in the warm season. This might be an important incorrect assumption.
LL 202-213. More generally, in this part, why not using the normalized diurnal hourly patterns of ERA5 and the daily value from the W5E5 dataset?
LL 290. Is Figure 4 really needed in the article? It can be placed in the supplementary material.
LL 308. Well, the low-flow bias is likely in the order of 50-100%, judging from Figure 5, I would not conclude that low-flows are simulated reasonably well. This is clearly an overstatement.
LL 530-539. This is likely the only discussion of the entire result section. I would suggest separating results from discussion, this might help to improve the quality of the narrative.
LL 545. Please note that in catchments with a very strong snow-driven seasonality, the NSE and other performance metrics are always very high, this is likely why variability is not significant among the various model configurations. A metric that discounts for seasonality would be more informative.
LL 565. Fig. 10. Please note that the model overestimates consistently spring discharge and underestimates autumn discharge, more evident in the last 2 stations, therefore, there should be something with regards to water storage and release in the catchment that is quite different from reality. I know that this is likely difficult to capture in the model, but it should not be hidden in the presentation of the results.
LL 598-600. What is this conclusion based on? What is supporting it? I don’t see a part of the manuscript, emphasizing streamflow on the rest.
Citation: https://doi.org/10.5194/tc-2023-20-RC2 -
RC3: 'Comment on tc-2023-20', Anonymous Referee #3, 08 May 2023
The Ability of Hydrologic-Land Surface Models to Concurrently Simulate Permafrost and Hydrology
by Mohamed S. Abdelhamed, Mohamed E. Elshamy, Saman Razavi, Howard S. Wheater
Summary of Paper
The authors present a paper by applying the Hydrologic-land surface models (H-LSM) to better understand the current and future high latitude permafrost distribution and the related hydrology within the permafrost environment. Their investigation region is the Liard River basin in Arctic Canada, where the authors develop different parametrizations of the Canadian H-LSM. Their validation data consists of three data sources using runoff at different gauging stations, some available ground temperature profiles, and different calculated permafrost maps of the investigated area, where three of them are based on a statistical approach, one is a thermal approach and one on machine learning approach.
This paper is quite comprehensive and very clearly structured. Overall, the science is sound and very well explained. The questions addressed at the beginning of the paper are well covered the research and is of great interest to the permafrost community. I listed some concerns, which I see mainly in the simulation of the permafrost temperature profiles, and which are summarized below. However, I appreciated very much that the authors seem to be aware about most of the problems and do address them in this paper in a very open way.
General:
- Line 178-188: I am not specialists in LSM approaches; however, the CLASS version 2.6 seems to explain the coupled water and energy balances for pre-defined soil columns with three layers of 0.1, 0.25 and 3.75 m in the model basing. The results are mainly shown in Figure 3 and the supplement. If we compare now the results of figure S8 to S16 and tables S5 to S13 between simulated and observed temperature profiles and its variability, it is clearly visible that most temperature profiles show currently in a large part of their temperature profiles 0° conditions. Therefore, one has to consider, particularly in strongly ice-bearing permafrost conditions, that in such cases temperature changes are significantly reduced by latent heat uptake during ground ice melt. Additional measurements that are sensitive to changes in ground ice and unfrozen water content are required to observe changes within the permafrost until the frozen material has thawed completely. In my view, the CLASS approach here seems in my view not reliable to detect these types of situations. Please explain how you could improve these conditions?
- Line 226 to 232: A big problem when taking PE or PA is that you do not consider the most important amount of ice content in the ground. The ice content value is extremely important as it influences how fast permafrost can be changed at a certain grid point or not. If you have a high content of ice, then the change from permafrost to permafrost-free conditions will take much longer time. As explained above your measured permafrost profiles seem to have at some places a very high ice content and it seems that the model is not really recognizing these conditions reasonably. Therefore, the most important question arise, how will the authors change the extent of the permafrost profiles in depth with time? Also, the time series shown in Figure 7 of the simulated ALT seems unfortunately not very promising. Therefore, how should the long-term reaction of permafrost with high ice content be realistically simulated? It would be nice if the authors could present some possible future changes to include this problem within their modelling approach?
- Specific Comments:
- Line 25: please also explain abbreviations in the abstract, such as H-LSM and MESH.
- Line 31: It is in my view an assumption that a model ensemble is essential to provide in view of a reliable picture of the current states and future spatio-temporal co-evolution of permafrost and hydrology. This is not predefined given. Please be more careful with such statements in the abstract.
- Line 160-164: How the authors will include ground layers with high amount of air. This means how they include coarse ground layers containing not so much water in the pores by high amounts of air. Please also describe this ground layers.
- Line 251: Please show also the depth of the boreholes.
Citation: https://doi.org/10.5194/tc-2023-20-RC3
Interactive discussion
Status: closed
-
RC1: 'Comment on tc-2023-20', Anonymous Referee #1, 21 Apr 2023
Review of the manuscript “The Ability of Hydrologic-Land Surface Models to Concurrently Simulate Permafrost and Hydrology” submitted for publication in The Cryosphere by Mohamed S. Abdelhamed, Mohamed E. Elshamy, Saman Razavi, and Howard S. Wheater.
This reviewer has expertise in permafrost field observations and in numerical modelling of heat and water transfer in frozen soil.
The paper compares model results generated with an ensemble of different parameterizations for the model MESH with observations of ground temperature and river discharge in the Liard River Basin in Canada. Testing and improving the quality of concurrent simulation of permafrost dynamics and river discharge is an important research area and well within the scope of The Cryosphere. The tools and data used are relevent for this topic. The present study, however, clearly falls short of the quality required for this journal.
My overall impression is that this manuscript is not yet ready for publication and should be resubmitted after considerable revision. The title suggests generic insight (“ability of Hydrologic-Land Surface Models”, the running title suggests insight about challenges (“H-LSM challenges to model…”), whereas the actual manuscript compares one set of ensemble results from one single models with observations. The main research question is only presented near the end of the manuscript: "[...] is it possible to concurrently simulate permafrost dynamics and hydrology over a large domain using an H-LSM?”. It is unclear what threshold needs to be met for this to be answered affirmatively, neither is the state of the art of existing models and their success clarified nor the exact gap in knowledge identified. The abstract then again states something different: “…the study results show that: (1) […] snow-cover primarily regulates permafrost […] and (2) different parametrizations yield different partitioning patterns […].” The former is textbook knowledge and was not the question addressed in this study. Why is it shown as one of two results worth communicating in the abstract? The latter needs further elaboration to become valuable insight. Then, the authors conclude that “an ensemble of model parametrizations is essential to provide a reliable picture of the current states and future spatio-temporal co-evolution of permafrost and hydrology.”. The appropriateness of ensembles has not been the question addressed in the paper, nor does it easily follow from the results presented. For resubmission, a clear and relevant research question needs to be identified, investigated, and adressed in the conclusion.
Several of the details in this study also merit more detailed attention. For example, with permafrost and long-term ground temperature being long-term average responses to surface conditions, why is spinup done with single years, only?
Citation: https://doi.org/10.5194/tc-2023-20-RC1 -
RC2: 'Comment on tc-2023-20 - Complete Review', Anonymous Referee #2, 07 May 2023
The authors present the implementation of the MESH hydrological/land surface model in the Liard River Basin (275000 km2) Canada in the period 1979-2019, to simulate permafrost and streamflow dynamics, which are then compared with ground temperature and active layer observations, distributed observation-based maps of permafrost and 4 streamgauge observations. The main scope of the manuscript is to assess the model capability to reproduce concurrently local and gridded permafrost dynamics and streamflow and to test 16 different model configurations (Table 3), where the initialization (spin-up years) and 2 parameters, related to minimum snow depth to consider a grid cell snow covered and vertical distribution of organic content of the soil, are modified.
The article is largely technically sounds, but it is completely uninspired, spending large parts just describing what we can see in the figures and tables (section 3.1.2 is almost unreadable, it uses even bullet points), but it does not provide insights on cryospheric or hydrological processes, on how to improve the models or the modeling approach, beyond the classic conclusions “we need more and better data”, “permafrost maps are uncertain”. The entire manuscript reads as a model application (by the way embedded in two other manuscripts in preparation (LL 617-622)) and not even one of the most comprehensive applications as some choice on the downscaling of the input data is debatable (see minor comments), and only few outputs and variables are analyzed. Why are authors not exploring and comparing also surface temperature and snow cover from remote sensing? The most critical comment, we learn very little from this study, to the point that the authors themselves conclude: “it does not provide a concrete answer to our main research question,” (LL 592), or “this result highlights the difficulty that modellers may encounter when configuring models to simultaneously simulate permafrost and hydrology”. This is repeated even twice (LL 591-592, LL 697-698), and many other repetitions are present. The entire conclusion - section 4 - is largely a repetition of the result section, sometime even with new result introduced (LL 676-689). The discussion is very limited, likely because there is not much to discuss. In summary, I am sure the manuscript has value for MESH users, and it would be fine as a project report, or technical report. For a research article, however, it is not fine. I am ok, if the authors use one model in one catchment only, but they need to use it to address a compelling question, carrying out dedicated numerical experiments, or learn from the model application something that might interest the scientific community, beyond saying “modelers might encounter challenges”. These aims are not achieved here.
Minor Comments
LL 41-42. I think the choice of the literature to introduce state-of-the-art Land-Surface-Models could be quite different and definitely more updated.
LL 92. Is a large suite of input data an issue nowadays? Are these data more than what a LSM needs? I do not think so.
LL 134-143. I think it would be good to state at this stage already how many streamgauges are available.
LL 150-153. The use of brackets is a bit confusing.
LL158. I would suggest referring to “Soil organic matter” rather than “organic matter”.
LL 163. It is not clear what “retention capacity” refers to here. Do you mean water content at field capacity?
LL 185-190. How many prognostic surface temperatures are solved in the model? Do different land covers share the same soil hydrology and soil temperature? How are the tiles interacting with each other if at all? I think these model aspects should be reported here as well.
LL 205. Why longwave radiation is distributed uniformly over the day? Usually, longwave radiation is assumed to depend on the air temperature at power of 4. The diurnal cycle is relevant. This is an incorrect assumption.
LL 211. Why did you assume a constant relative humidity? What is known to be fairly constant at daily scale is specific humidity or equivalently vapor pressure, relative humidity has a clear diurnal variability, at least in the warm season. This might be an important incorrect assumption.
LL 202-213. More generally, in this part, why not using the normalized diurnal hourly patterns of ERA5 and the daily value from the W5E5 dataset?
LL 290. Is Figure 4 really needed in the article? It can be placed in the supplementary material.
LL 308. Well, the low-flow bias is likely in the order of 50-100%, judging from Figure 5, I would not conclude that low-flows are simulated reasonably well. This is clearly an overstatement.
LL 530-539. This is likely the only discussion of the entire result section. I would suggest separating results from discussion, this might help to improve the quality of the narrative.
LL 545. Please note that in catchments with a very strong snow-driven seasonality, the NSE and other performance metrics are always very high, this is likely why variability is not significant among the various model configurations. A metric that discounts for seasonality would be more informative.
LL 565. Fig. 10. Please note that the model overestimates consistently spring discharge and underestimates autumn discharge, more evident in the last 2 stations, therefore, there should be something with regards to water storage and release in the catchment that is quite different from reality. I know that this is likely difficult to capture in the model, but it should not be hidden in the presentation of the results.
LL 598-600. What is this conclusion based on? What is supporting it? I don’t see a part of the manuscript, emphasizing streamflow on the rest.
Citation: https://doi.org/10.5194/tc-2023-20-RC2 -
RC3: 'Comment on tc-2023-20', Anonymous Referee #3, 08 May 2023
The Ability of Hydrologic-Land Surface Models to Concurrently Simulate Permafrost and Hydrology
by Mohamed S. Abdelhamed, Mohamed E. Elshamy, Saman Razavi, Howard S. Wheater
Summary of Paper
The authors present a paper by applying the Hydrologic-land surface models (H-LSM) to better understand the current and future high latitude permafrost distribution and the related hydrology within the permafrost environment. Their investigation region is the Liard River basin in Arctic Canada, where the authors develop different parametrizations of the Canadian H-LSM. Their validation data consists of three data sources using runoff at different gauging stations, some available ground temperature profiles, and different calculated permafrost maps of the investigated area, where three of them are based on a statistical approach, one is a thermal approach and one on machine learning approach.
This paper is quite comprehensive and very clearly structured. Overall, the science is sound and very well explained. The questions addressed at the beginning of the paper are well covered the research and is of great interest to the permafrost community. I listed some concerns, which I see mainly in the simulation of the permafrost temperature profiles, and which are summarized below. However, I appreciated very much that the authors seem to be aware about most of the problems and do address them in this paper in a very open way.
General:
- Line 178-188: I am not specialists in LSM approaches; however, the CLASS version 2.6 seems to explain the coupled water and energy balances for pre-defined soil columns with three layers of 0.1, 0.25 and 3.75 m in the model basing. The results are mainly shown in Figure 3 and the supplement. If we compare now the results of figure S8 to S16 and tables S5 to S13 between simulated and observed temperature profiles and its variability, it is clearly visible that most temperature profiles show currently in a large part of their temperature profiles 0° conditions. Therefore, one has to consider, particularly in strongly ice-bearing permafrost conditions, that in such cases temperature changes are significantly reduced by latent heat uptake during ground ice melt. Additional measurements that are sensitive to changes in ground ice and unfrozen water content are required to observe changes within the permafrost until the frozen material has thawed completely. In my view, the CLASS approach here seems in my view not reliable to detect these types of situations. Please explain how you could improve these conditions?
- Line 226 to 232: A big problem when taking PE or PA is that you do not consider the most important amount of ice content in the ground. The ice content value is extremely important as it influences how fast permafrost can be changed at a certain grid point or not. If you have a high content of ice, then the change from permafrost to permafrost-free conditions will take much longer time. As explained above your measured permafrost profiles seem to have at some places a very high ice content and it seems that the model is not really recognizing these conditions reasonably. Therefore, the most important question arise, how will the authors change the extent of the permafrost profiles in depth with time? Also, the time series shown in Figure 7 of the simulated ALT seems unfortunately not very promising. Therefore, how should the long-term reaction of permafrost with high ice content be realistically simulated? It would be nice if the authors could present some possible future changes to include this problem within their modelling approach?
- Specific Comments:
- Line 25: please also explain abbreviations in the abstract, such as H-LSM and MESH.
- Line 31: It is in my view an assumption that a model ensemble is essential to provide in view of a reliable picture of the current states and future spatio-temporal co-evolution of permafrost and hydrology. This is not predefined given. Please be more careful with such statements in the abstract.
- Line 160-164: How the authors will include ground layers with high amount of air. This means how they include coarse ground layers containing not so much water in the pores by high amounts of air. Please also describe this ground layers.
- Line 251: Please show also the depth of the boreholes.
Citation: https://doi.org/10.5194/tc-2023-20-RC3
Model code and software
MESH-Model Environment and Climate Change Canada https://github.com/MESH-Model/MESH-Releases/releases/tag/SA_MESH_1.4%2FSA_MESH_1.4.1813
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 542 | 223 | 53 | 818 | 95 | 52 | 56 |
- HTML: 542
- PDF: 223
- XML: 53
- Total: 818
- Supplement: 95
- BibTeX: 52
- EndNote: 56
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Cited
1 citations as recorded by crossref.
Mohamed S. Abdelhamed
Mohamed E. Elshamy
Saman Razavi
Howard S. Wheater
This preprint has been withdrawn.
- Preprint
(2294 KB) - Metadata XML
-
Supplement
(4681 KB) - BibTeX
- EndNote