the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Multi-site evaluation of modelled methane emissions over northern wetlands by the JULES land surface model coupled with the HIMMELI peatland methane emission model
Abstract. Northern peatland stores a large amount of organic soil carbon and is considered to be one of the most significant CH4 sources among wetlands. The default wetland CH4 emission scheme in JULES (land surface model of the UK Earth System model) only takes into account the CH4 emissions from inundated areas in a simple way. However, it is known that the processes for peatland CH4 emission are complex. In this work, we coupled the process-based peatland CH4 emission model HIMMELI (HelsinkI Model of MEthane buiLd-up and emIssion for peatlands) with JULES (JULES-HIMMELI) by taking the HIMMELI input data from JULES simulations. Firstly, the soil temperature, water table depth (WTD) and soil carbon simulated by JULES, as well as the prescribed maximum leaf area index (LAI) in JULES were evaluated against available datasets at the studied northern wetland sites. Then, the simulated CH4 emissions from JULES and JULES-HIMMELI simulations were compared against the observed CH4 emissions at these sites. Moreover, sensitivities of CH4 emissions to the rate of anoxic soil respiration (anoxic Rs), surface soil temperature and WTD were investigated. Results show that JULES can well represent the magnitude and seasonality of surface (5–10 cm) and relatively deep (34–50 cm) soil temperatures, whereas the simulated WTD and soil carbon density profiles show large deviations from the site observations. The prescribed maximum LAI in JULES was within one standard deviation of the maximum LAIs derived from the Sentinel-2 satellite data for Siikaneva, Kopytkowo and Degerö sites, but lower for the other three sites. The simulated CH4 emissions by JULES have much smaller inter-annual variability than the observations. However, no specific simulation setup of the coupled model can lead to consistent improvements in the simulated CH4 emissions for all the sites. When using observed WTD or modified soil decomposition rate, there were only improvements in simulated CH4 fluxes at certain sites or years. Both simulated and observed CH4 emissions at sites strongly depend on the rate of anoxic Rs, which is the basis of CH4 emission estimates in HIMMELI. By excluding the effect from the rate of anoxic Rs on CH4 emissions, it is found that the Rs-log-normalized CH4 emissions (log normalization of the ratio of CH4 emission to anoxic Rs rate) show similar increasing trends with increased surface soil temperature from both observations and simulations, but different trends with raised WTD which may due to the uncertainty in simulated O2 concentration in HIMMELI. In general, we consider the JULES-HIMMELI model is more appropriate in simulating the wetland CH4 emissions than the default wetland CH4 emission scheme in JULES. Nevertheless, in order to improve the accuracy of simulated wetland CH4 emissions with the JULES-HIMMELI model, it is still necessary to better represent the peat soil carbon and hydrologic processes in JULES and the CH4 production and transportation processes in HIMMELI, such as plant transportation of gases, seasonality of parameters controlling oxidation and production, and adding microbial activities.
- Preprint
(2790 KB) - Metadata XML
-
Supplement
(1201 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on bg-2022-229', Anonymous Referee #1, 11 Jan 2023
Dear Authors,
This study simulated CH4 emissions from six northern wetlands based on a state-of-the-art land surface model (JULES) coupled with a CH4 module (HIMMELI). The manuscript is generally well written, and I was able to follow the main conclusion. Authors evaluated model performance in terms of CH4 emissions with several model settings, and found that neither settings could not “reasonably” represent the observed CH4 emissions. I felt that the analysis still has room for improvement, and thus recommend a resubmission after a major revision.
Major limitation of this study is that the model lacked ability to reproduce observed CH4 dynamics. It is not possible to say that a model that can correctly represent an observation is the correct model, but at the very least, a model that cannot reproduce the observation to a reasonable degree is not likely to be the correct model. I recommend that the model be modified and parameterized so that the model can, at least to a reasonable degree, reproduce the observations. Susiluoto et al. (2018) provided a sophisticated method for parameterizing the HIMMELI module; such exercise could improve the representation of the model in each site.
The manuscript showed the model performance in terms of CH4 emissions, soil temperature, WTD, LAI, and soil carbon. The advantage of using eddy covariance data for model validation is that the various factors, including CO2 and water vapor exchanges, involved in CH4 dynamics can be validated simultaneously. Did the model have the ability to reproduce CO2 fluxes (GPP, ecosystem respiration, NEP), evapotranspiration, and net radiation in each site? Describing their performance is necessary to interpret the model performance for CH4 dynamics.
This study seems to assume that there was a general parameter set for simulating CH4 dynamics across the wetlands. Since the type and origin of the wetlands were diverse (Table 1), the parameter set could be site specific. I understand that such site specific treatments are difficult on a global scale, but this is a site-specific study; so, careful consideration of site specific parameterization would be required. Furthermore, please show the list of important parameters for representing CH4 dynamics in the manuscript.
The descriptions regarding accuracy/precision are not quantitative and thus vague. Please provide quantitative descriptions when describing accuracy/precision (mean bias, RMSE, R2). Please show R2 and p-value when describing the presence or absence of correlation.
Specific comments
Line 200: For Taylor diagram of soil temperature, how the authors standardized RMSE and SD (i.e., using mean temperature in Kelvin or Celsius)? Furthermore, if the authors use Taylor diagrams for evaluating precision, be consistent using Taylor diagrams for WTD, and CH4 fluxes.
References
Susiluoto et al. (2018) Geosci. Model Dev., 11, 1199-1228.
Citation: https://doi.org/10.5194/bg-2022-229-RC1 -
AC1: 'Reply on RC1', Yao Gao, 03 Mar 2023
We deeply appreciate all the reviewers comments for improving the scientific significance and quality of this manuscript. Our point-to-point response to reviewer 1’ comments are listed below.
Response to Referee comments 1:
Dear Authors,
This study simulated CH4 emissions from six northern wetlands based on a state-of-the-art land surface model (JULES) coupled with a CH4 module (HIMMELI). The manuscript is generally well written, and I was able to follow the main conclusion. Authors evaluated model performance in terms of CH4 emissions with several model settings, and found that neither settings could not “reasonably” represent the observed CH4 emissions. I felt that the analysis still has room for improvement, and thus recommend a resubmission after a major revision.
Major limitation of this study is that the model lacked ability to reproduce observed CH4 dynamics. It is not possible to say that a model that can correctly represent an observation is the correct model, but at the very least, a model that cannot reproduce the observation to a reasonable degree is not likely to be the correct model. I recommend that the model be modified and parameterized so that the model can, at least to a reasonable degree, reproduce the observations. Susiluoto et al. (2018) provided a sophisticated method for parameterizing the HIMMELI module; such exercise could improve the representation of the model in each site.
Authors Response (AR): We will add model parameter optimization for all the sites to improve the model results for CH4 fluxes.
The manuscript showed the model performance in terms of CH4 emissions, soil temperature, WTD, LAI, and soil carbon. The advantage of using eddy covariance data for model validation is that the various factors, including CO2 and water vapor exchanges, involved in CH4 dynamics can be validated simultaneously. Did the model have the ability to reproduce CO2 fluxes (GPP, ecosystem respiration, NEP), evapotranspiration, and net radiation in each site? Describing their performance is necessary to interpret the model performance for CH4 dynamics.
AR: We will check the model performance on CO2 fluxes, evapotranspiration in the revised manuscript.
This study seems to assume that there was a general parameter set for simulating CH4 dynamics across the wetlands. Since the type and origin of the wetlands were diverse (Table 1), the parameter set could be site specific. I understand that such site specific treatments are difficult on a global scale, but this is a site-specific study; so, careful consideration of site specific parameterization would be required. Furthermore, please show the list of important parameters for representing CH4 dynamics in the manuscript.
AR: We will perform site-specific parameter optimization for HIMMELI and identify a list of important parameters.
The descriptions regarding accuracy/precision are not quantitative and thus vague. Please provide quantitative descriptions when describing accuracy/precision (mean bias, RMSE, R2). Please show R2 and p-value when describing the presence or absence of correlation.
AR: We will add the quantitative descriptions to our results, including accuracy metrics (mean bias, RMSE, R2) and p-value to describe the correlation between variables. By including these quantitative descriptions, we will provide a more thorough and objective analysis of our results.
Specific comments:
Line 200: For Taylor diagram of soil temperature, how the authors standardized RMSE and SD (i.e., using mean temperature in Kelvin or Celsius)? Furthermore, if the authors use Taylor diagrams for evaluating precision, be consistent using Taylor diagrams for WTD, and CH4 fluxes.
AR: The ratio of standard deviation is calculated by dividing the standard deviation of model results by the standard deviation of the corresponding measurements. The RMSE wasnot standardized. The unit of temperature doesnot lead to differences in those metrics. We will keep the presentation format (metrics and visualization) of our evaluation consistent across all variables.
References
Susiluoto et al. (2018) Geosci. Model Dev., 11, 1199-1228.
Citation: https://doi.org/10.5194/bg-2022-229-AC1
-
AC1: 'Reply on RC1', Yao Gao, 03 Mar 2023
-
RC2: 'Comment on bg-2022-229', Anonymous Referee #2, 03 Feb 2023
Gao and colleagues took the JULES land surface model and coupled it to the HIMMELI peatland CH4 emission model. They then evaluated the coupled model over five high-latitude peatland sites using an assortment of different model setups. Several of the different model setups were used to test the influence of using observed water table depths (WTD), prescribed LAI, modified soil carbon decomposition rate, etc. The paper is generally clearly written, but strangely contains no equations. This makes it challenging to understand how the coupled models work. The authors give references but it would be much more preferable to give high-level equations so that the basics of the models could be seen. This is mostly relevant to HIMMELI as it would have been nice to more easily see how the inputs from JULES can be traced to estimated CH4 emissions.
One of my most significant issues with this paper can be found in statements like this (line 400, emphasis mine): "It is found that the coupled model does not lead to significant improvements in the simulated CH4 emissions in comparison to the default wetland CH4 emission scheme in JULES. Nevertheless the coupled model is considered to be more realistic in simulating wetland CH4 emissions as HIMMELI simulates CH4 production, oxidation and transportation by vascular plants, ebullition and diffusion in a vertical soil column, while the default wetland methane emission scheme in JULES only takes into account CH4 emission from saturated wetlands in a simple way." - How is this statement justifiable in any way? This is like saying even though we don't show any improved skill with our model, it is better because it is more complicated or because it has lots of equations! This is absolutely the wrong conclusion from this work. If the model does not add skill then the parameters associated with it and the processes simulated are unconstrained and the model is not useful. The authors even note on line 416, that there are no O2 observations. So then how is the O2 aspect of HIMMELI going to be properly parameterized if measurements for it are not available at any site? It could be then a tuning parameter, but what is the value then of the added complexity if it just adds another unconstrained parameter to contribute to model equifinality? I find other major issues with the work. How do you simulate peatlands with no moss, shrubs, or sedges? All sites (Table 1) have one of those vascular plants along with moss and are not C3 grass dominant. How do you build peat when there is no simulation of moss? These aren't tropical peatlands, most of the peat is coming from the highly recalcitrant litter from the moss. This makes me wonder why JULES was chosen since it seems like an unsuitable land surface model to couple with HIMMELI (although the most recent JULES might be a good choice). I recommend rejection with possible resubmission after major revisions. I think the authors need to clearly demonstrate what scientific or technical advance this work represents. At the present I have trouble seeing what that advancement would be considering the models do not demonstrate an improvement in skill over the basic land surface model. The work also offers little insight into the processes represented.
Smaller comments:
Line 33: The WTD and soil C show large deviations from the obs. So then how could the simulated CH4 emissions be expected to get the right answer for the right reason? The paper would be improved if there was some deeper insights into how to correct these problems.
L 95: TOPMODEL (all caps)
L 159: I don't understand the point about the Song et al. paper. Why mention this?
L 193: Don't ref a webpage that will change over time. Put the configuration used in the appendix so it will be pinned to this work.
L 196: As mentioned above, I don't know how you can simulate peatlands when they are set up to be grasslands. This seems like a fundamentally flawed starting point.
L 198: How were the soil params adjusted? What params?
L 200: 'hundreds of times'? How was it checked whether the model was in equilibrium?
L 203: Show equations please. It is very hard to understand how this all ties in without equations. They are needed in a modelling paper.
L 207: SW is the typical abbreviation for shortwave. I suggest using something like SM for soil moisture.
L 234: I find the soil T correlation to the observed surprisingly good considering how poorly the WTD was simulated. Are these correlations for the runs using simulated or observed WTD? If with the simulated, isn't that surprising that the WTD can be so off but the soil T so good? The latent heat impacts of the water has a strong impact on the soil T so I would expect the soil T would be pretty poor if the soil was too dry/wet.
L 243: It would be good to see mean bias values too.
L 265: Mer Bleue has about 25 cm diff between its hummocks and hollows. What is the model compared to here?
L 273: Why was no run done with the observed LAI? It seems like that would be easy and a natural way to check the influence of the LAI.
L 288: I have no idea what to make of the comparison between what would be a simulated grassland soil and each site's peat soil. If the simulated soil C is in-line with the peatland, doesn't that indicate that the model is actually doing some pretty weird stuff? This is C3 grass after all.
L 323: The biases don't really say what is written. Lompo gets worse while Degero changes less than Siikaneva (which was said to be smaller change)
L 381: Quantify the correlation here, and elsewhere should also be properly quantified. There is a lot of vague wording that could be replaced by stats.
Data and code availability does not follow proper open science standards. It is unreasonable to ask people to email for code or data. Please modernize your practices here.
Fig 1: What are the soil layer thicknesses in the model(s)? 5 cm or thicker? If thicker, then how do you compare at 5 cm?
Fig 5: Doesn't take into account any uncertainty in the observations, which I would assume to be considerable given they aren't the site obs.
Citation: https://doi.org/10.5194/bg-2022-229-RC2 -
AC2: 'Reply on RC2', Yao Gao, 03 Mar 2023
We deeply appreciate all the reviewers comments for improving the scientific significance and quality of this manuscript. Our point-to-point response to all the reviewer2’ comments are listed below.
Response to referee comments 2:
Gao and colleagues took the JULES land surface model and coupled it to the HIMMELI peatland CH4 emission model. They then evaluated the coupled model over five high-latitude peatland sites using an assortment of different model setups. Several of the different model setups were used to test the influence of using observed water table depths (WTD), prescribed LAI, modified soil carbon decomposition rate, etc. The paper is generally clearly written, but strangely contains no equations. This makes it challenging to understand how the coupled models work. The authors give references but it would be much more preferable to give high-level equations so that the basics of the models could be seen. This is mostly relevant to HIMMELI as it would have been nice to more easily see how the inputs from JULES can be traced to estimated CH4 emissions.
AR:Including HIMMELI equations in this paper would be redundant. Instead, we will add references to the equations in HIMMELI paper to demonstrate the link between JULES and HIMMELI.
One of my most significant issues with this paper can be found in statements like this (line 400, emphasis mine): "It is found that the coupled model does not lead to significant improvements in the simulated CH4 emissions in comparison to the default wetland CH4 emission scheme in JULES. Nevertheless the coupled model is considered to be more realistic in simulating wetland CH4 emissions as HIMMELI simulates CH4 production, oxidation and transportation by vascular plants, ebullition and diffusion in a vertical soil column, while the default wetland methane emission scheme in JULES only takes into account CH4 emission from saturated wetlands in a simple way." - How is this statement justifiable in any way? This is like saying even though we don't show any improved skill with our model, it is better because it is more complicated or because it has lots of equations! This is absolutely the wrong conclusion from this work. If the model does not add skill then the parameters associated with it and the processes simulated are unconstrained and the model is not useful. The authors even note on line 416, that there are no O2 observations. So then how is the O2 aspect of HIMMELI going to be properly parameterized if measurements for it are not available at any site? It could be then a tuning parameter, but what is the value then of the added complexity if it just adds another unconstrained parameter to contribute to model equifinality?
AR: We will perform site-specific parameter optimization for HIMMELI and identify a list of important parameters. We also take into account the O2 aspect by optimizing parameters that are related to oxidation of methane which is dominantly not co-located with production in the soil column, and has different temperature dependence. So, indirectly, we are also take O2-related processes into account although there are no direct observations of O2.
I find other major issues with the work. How do you simulate peatlands with no moss, shrubs, or sedges? All sites (Table 1) have one of those vascular plants along with moss and are not C3 grass dominant. How do you build peat when there is no simulation of moss? These aren't tropical peatlands, most of the peat is coming from the highly recalcitrant litter from the moss. This makes me wonder why JULES was chosen since it seems like an unsuitable land surface model to couple with HIMMELI (although the most recent JULES might be a good choice).
AR: Currently, the JULES PFT includes tropical broadleaf evergreen trees, temperate broadleaf evergreen trees, broad-leaf deciduous trees, needle-leaf evergreen trees, needle-leaf deciduous trees, C3 grass, C4 grass, arctic grass, evergreen shrubs, deciduous shrubs. Peat is described with peat soil parameter values in this version of JULES used in this manuscript. JULES has a dynamic vegetation model, however, the simulated vegetation types by the dynamic vegetation model at our studied sites were not realistic. In land surface modelling, the most suitable PFT is often chose from the available PFTs with modification according to the need. C3 grass has been shown to work well to represent the productive sedge fens common in high northern latitudes (Qiu et al., 2018). Due to the lack of representation of moss in JULES, we applied C3 grass to represent the average vegetation growing in northern peatlands and the maximum LAI for C3 grass was decreased from 3 m2/m2 to 1.3 m2/m2 according to measurement at Lompölöjankka site. We acknowledge that inappropriate PFT will introduce uncertainties in simulated peat, as recognized in Chadburn et al. (2022), where a new peat module was developed for JULES and realistic peat soil carbon profile were archived. However, a moss PFT with its unique hydraulic and ecological functions in JULES was also suggested to be considered when simulating peat. In general, wetland related ecology, hydrology, and soil carbon processes of JULES are need further development. The original CH4 emission module in JULES is very simple. Nevertheless, JULES with its methane module has been used to estimate the methane emissions from wetlands, and to investigate the influence of climate variability on methane emissions (Gedney et al., 2019). In this work, we implemented a process-based wetland CH4 emission model and evaluated the modelled performance over the studied sites. Our aim was to gain more insight on the model deficiencies and to improve the simulated CH4 fluxes. Parameter optimization will be added to the revised manuscript to improve the simulated CH4 flux and gain more insight into the model processes.
Chadburn, S. E., Burke, E. J., Gallego-Sala, A. V., Smith, N. D., Bret-Harte, M. S., Charman, D. J., Drewer, J., Edgar, C. W., Euskirchen, E. S., Fortuniak, K., Gao, Y., Nakhavali, M., Pawlak, W., Schuur, E. A. G., and Westermann, S.: A new approach to simulate peat accumulation, degradation and stability in a global land surface scheme (JULES vn5.8_accumulate_soil) for northern and temperate peatlands, Geosci. Model Dev., 15, 1633–1657, https://doi.org/10.5194/gmd-15-1633-2022, 2022.
Gedney, N., Huntingford, C., Comyn-Platt, E. and Wiltshire, A.: Significant feedbacks of wetland methane release on climate change and the causes of their uncertainty, Environmental Research Letters, 2019, 14(8): 084027.
Qiu, C., Zhu, D., Ciais, P., Guenet, B., Krinner, G., Peng, S., Aurela, M., Bernhofer, C., Brümmer, C., Bret-Harte, S., Chu, H., Chen, J., Desai, A. R., Dušek, J., Euskirchen, E. S., Fortuniak, K., Flanagan, L. B., Friborg, T., Grygoruk, M., Gogo, S., Grünwald, T., Hansen, B. U., Holl, D., Humphreys, E., Hurkuck, M., Kiely, G., Klatt, J., Kutzbach, L., Largeron, C., Laggoun-Défarge, F., Lund, M., Lafleur, P. M., Li, X., Mammarella, I., Merbold, L., Nilsson, M. B., Olejnik, J., Ottosson-Löfvenius, M., Oechel, W., Parmentier, F.-J. W., Peichl, M., Pirk, N., Peltola, O., Pawlak, W., Rasse, D., Rinne, J., Shaver, G., Schmid, H. P., Sottocornola, M., Steinbrecher, R., Sachs, T., Urbaniak, M., Zona, D., and Ziemblinska, K.: ORCHIDEE-PEAT (revision 4596), a model for northern peatland CO2, water, and energy fluxes on daily to annual scales, Geosci. Model Dev., 11, 497–519, https://doi.org/10.5194/gmd-11-497-2018, 2018.
I recommend rejection with possible resubmission after major revisions. I think the authors need to clearly demonstrate what scientific or technical advance this work represents. At the present I have trouble seeing what that advancement would be considering the models do not demonstrate an improvement in skill over the basic land surface model. The work also offers little insight into the processes represented.
Smaller comments:
Line 33: The WTD and soil C show large deviations from the obs. So then how could the simulated CH4 emissions be expected to get the right answer for the right reason? The paper would be improved if there was some deeper insights into how to correct these problems.
AR: Soil respiration and WTD simulated by JULES are used to calculate anoxic soil respiration, which is an input variable for HIMMELI. In HIMMELI, the anoxic soil respiration is then multiplied with fm, which is the fraction of the anoxic respiration used to produce methane. These two, anoxic respiration and fm, are the most uncertain parameters of the model system, they are strongly correlated in optimization and they are most important determinants of the level of methane emissions. Because lack of measured anoxic soil respiration and soil respiration, we compared the simulated and observed WTD and soil carbon in the manuscript as the background information. Nevertheless, the correct estimation of WTD and soil C doesnot really mean correct anoxic soil respiration. Therefore, we will use parameter optimization of the model to improve the model results.
L 95: TOPMODEL (all caps)
AR: We will correct this.
L 159: I don't understand the point about the Song et al. paper. Why mention this?
AR: We adopted the soil carbon density from the IGBP-DIS dataset to compare with the JULES simulated soil carbon density, due to the scarcity of in-situ measured soil carbon density. Song et al.’s paper adopted IGBP dataset as the initial soil carbon condition for CH4 emission modelling at multiple sites. We adopted Song et al.’s paper here just to show that the the gridded soil carbon data from IGBP-DIS has been used for the site study. However, we now think it is not so meaningful to show this comparison in the manuscript. Thus, we will delete the relevant text and figure regarding to this data.
L 193: Don't ref a webpage that will change over time. Put the configuration used in the appendix so it will be pinned to this work.
AR: This will be changed according to reviewer’s suggestion.
L 196: As mentioned above, I don't know how you can simulate peatlands when they are set up to be grasslands. This seems like a fundamentally flawed starting point.
AR: The main problem with not simulating peat correctly in the JULES version used in this study is the lack of proper peat accumulation and degradation processes. In Chadburn et al. (2022), a new approach to simulate peat accumulation, degradation and stability in JULES was developed. This new approach showed realistic profiles of soil organic carbon for peatlands. To improve the simulation of peat, a moss PFT and its unique functions were suggested to be included in JULES. In land surface modelling, the most suitable PFT is often chose from the available PFTs with modification according to the need. C3 grass has been shown to work well to represent the productive sedge fens common in high northern latitudes (Qiu et al., 2018). Due to the lack of representation of moss in JULES, we applied C3 grass with a lower LAI to represent the average vegetation growing in northern peatlands.
L 198: How were the soil params adjusted? What params?
AR: The model reads a list of soil parameters (see http://jules-lsm.github.io/vn4.9/namelists/ancillaries.nml.html?highlight=soil%20parameters#list-of-soil-params) from an ancillary file. In order to improve the simulated WTD, we manually adjusted the soil parameters impacting soil hydrology within their parameter range. These adjustments were made to the following parameters: b - exponent in soil hydraulic characteristics; satcon - hydraulic conductivity; sm_crit - volumetric soil moisture content at the critical point; sm_sat – volumetric soil moisture content at saturation; sm_wilt – volumetric soil moisture content at the wilting point. We will list the turned soil parameters in the text.
L 200: 'hundreds of times'? How was it checked whether the model was in equilibrium?
AR: We repeated the forcing data for 400 cycles to archieve soil carbon spin up. The spin up was checked in the way that the soil moisture, soil temperature and soil carbon reach a state of equilibrium under the applied forcing.
L 203: Show equations please. It is very hard to understand how this all ties in without equations. They are needed in a modelling paper.
AR: We will show the equations in appendix.
L 207: SW is the typical abbreviation for shortwave. I suggest using something like SM for soil moisture.
AR: This will be changed according to reviewer’s comments.
L 234: I find the soil T correlation to the observed surprisingly good considering how poorly the WTD was simulated. Are these correlations for the runs using simulated or observed WTD? If with the simulated, isn't that surprising that the WTD can be so off but the soil T so good? The latent heat impacts of the water has a strong impact on the soil T so I would expect the soil T would be pretty poor if the soil was too dry/wet.
AR: The WTD produced by JULES is the grid box mean water table depth, which is calculated by the TOPMODEL approach (Gedney and Cox, 2003). The soil temperature and soil moisture are simulated in a vertical layered soil column (Burke et al., 2017). This is the main reason that the simulated WTD showed large biases to the observed WTD, despite the simulated soil temperature performing well.
Burke, E. J., Chadburn, S. E., and Ekici, A.: A vertical representation of soil carbon in the JULES land surface scheme (vn4.3_permafrost) with a focus on permafrost regions, Geoscientific Model Development, 10, 959– 975, doi:10.5194/gmd-10-959-2017,http://www.geosci-model-dev.net/10/959/2017/, 2017a.
Gedney N, Cox P M.: The sensitivity of global climate model simulations to the representation of soil moisture heterogeneity, Journal of Hydrometeorology, 4(6), 1265-1275, 2003.
L 243: It would be good to see mean bias values too.
AR: This will be modified according to reviewer’s comments.
L 265: Mer Bleue has about 25 cm diff between its hummocks and hollows. What is the model compared to here?
AR: This need to be further checked with the data provider. The provided data includes soil temperature measured from hollows and hummocks but only one type of water table depth.
L 273: Why was no run done with the observed LAI? It seems like that would be easy and a natural way to check the influence of the LAI.
AR: We donot have observed LAI in daily. We have the run with Sentinel-2 maximum LAI for Kopytkowo site.
L 288: I have no idea what to make of the comparison between what would be a simulated grassland soil and each site's peat soil. If the simulated soil C is in-line with the peatland, doesn't that indicate that the model is actually doing some pretty weird stuff? This is C3 grass after all.
AR: A new approach to simulate peat accumulation, degradation and stability in JULES has been developed and demonstrating realistic profiles of soil organic carbon for peatlands (Chadburn et al., 2022). However, it has been suggested that the inclusion of a moss PFT and its unique functions would future enhance the accuracy of JULES simulations. While the ecological and hydraulic behavior of mosses is significant, it is important to note that peat accumulation and degradation processes play a more crucial role in determining the carbon profile of peat soil. For the soil in the model it does not matter where the C is coming, as long as there is correct amount of it.
L 323: The biases don't really say what is written. Lompo gets worse while Degero changes less than Siikaneva (which was said to be smaller change)
AR: The overestimated CH4 flux changed to be underestimated. Degero changes more, which is from 0.18 to -0.13. Siikaneva changes from 0.52 to 0.42.
L 381: Quantify the correlation here, and elsewhere should also be properly quantified. There is a lot of vague wording that could be replaced by stats.
AR: We will improve those vague wording parts with statistics.
Data and code availability does not follow proper open science standards. It is unreasonable to ask people to email for code or data. Please modernize your practices here.
Fig 1: What are the soil layer thicknesses in the model(s)? 5 cm or thicker? If thicker, then how do you compare at 5 cm?
AR: The soil layer thickness in JULES model was set to be 0.05m, 0.134m, 0.248m, 0.389m, 0.557m, 0.748m, 0.964m, 1.201m, 1.461m, 1.742m, 2.044m, 2.367m, 2.709m, 3.07m. So there is no problem in comparing soil temperature at 5cm depth.
Fig 5: Doesn't take into account any uncertainty in the observations, which I would assume to be considerable given they aren't the site obs.
AR: In our manuscript, we cited Song et al. (2020) to demonstrate the use of gridded soil carbon data from IGBP-DIS dataset for the site study, which was due to the limited availability of measured soil carbon density at the sites. There is no information of the uncertainties of IGBP-DIS soil carbon density data. We now think it is not so meaningful to show this comparison in the manuscript. Thus, we will delete the relevant text and figure regarding to this data.
Citation: https://doi.org/10.5194/bg-2022-229-AC2
-
AC2: 'Reply on RC2', Yao Gao, 03 Mar 2023
Status: closed
-
RC1: 'Comment on bg-2022-229', Anonymous Referee #1, 11 Jan 2023
Dear Authors,
This study simulated CH4 emissions from six northern wetlands based on a state-of-the-art land surface model (JULES) coupled with a CH4 module (HIMMELI). The manuscript is generally well written, and I was able to follow the main conclusion. Authors evaluated model performance in terms of CH4 emissions with several model settings, and found that neither settings could not “reasonably” represent the observed CH4 emissions. I felt that the analysis still has room for improvement, and thus recommend a resubmission after a major revision.
Major limitation of this study is that the model lacked ability to reproduce observed CH4 dynamics. It is not possible to say that a model that can correctly represent an observation is the correct model, but at the very least, a model that cannot reproduce the observation to a reasonable degree is not likely to be the correct model. I recommend that the model be modified and parameterized so that the model can, at least to a reasonable degree, reproduce the observations. Susiluoto et al. (2018) provided a sophisticated method for parameterizing the HIMMELI module; such exercise could improve the representation of the model in each site.
The manuscript showed the model performance in terms of CH4 emissions, soil temperature, WTD, LAI, and soil carbon. The advantage of using eddy covariance data for model validation is that the various factors, including CO2 and water vapor exchanges, involved in CH4 dynamics can be validated simultaneously. Did the model have the ability to reproduce CO2 fluxes (GPP, ecosystem respiration, NEP), evapotranspiration, and net radiation in each site? Describing their performance is necessary to interpret the model performance for CH4 dynamics.
This study seems to assume that there was a general parameter set for simulating CH4 dynamics across the wetlands. Since the type and origin of the wetlands were diverse (Table 1), the parameter set could be site specific. I understand that such site specific treatments are difficult on a global scale, but this is a site-specific study; so, careful consideration of site specific parameterization would be required. Furthermore, please show the list of important parameters for representing CH4 dynamics in the manuscript.
The descriptions regarding accuracy/precision are not quantitative and thus vague. Please provide quantitative descriptions when describing accuracy/precision (mean bias, RMSE, R2). Please show R2 and p-value when describing the presence or absence of correlation.
Specific comments
Line 200: For Taylor diagram of soil temperature, how the authors standardized RMSE and SD (i.e., using mean temperature in Kelvin or Celsius)? Furthermore, if the authors use Taylor diagrams for evaluating precision, be consistent using Taylor diagrams for WTD, and CH4 fluxes.
References
Susiluoto et al. (2018) Geosci. Model Dev., 11, 1199-1228.
Citation: https://doi.org/10.5194/bg-2022-229-RC1 -
AC1: 'Reply on RC1', Yao Gao, 03 Mar 2023
We deeply appreciate all the reviewers comments for improving the scientific significance and quality of this manuscript. Our point-to-point response to reviewer 1’ comments are listed below.
Response to Referee comments 1:
Dear Authors,
This study simulated CH4 emissions from six northern wetlands based on a state-of-the-art land surface model (JULES) coupled with a CH4 module (HIMMELI). The manuscript is generally well written, and I was able to follow the main conclusion. Authors evaluated model performance in terms of CH4 emissions with several model settings, and found that neither settings could not “reasonably” represent the observed CH4 emissions. I felt that the analysis still has room for improvement, and thus recommend a resubmission after a major revision.
Major limitation of this study is that the model lacked ability to reproduce observed CH4 dynamics. It is not possible to say that a model that can correctly represent an observation is the correct model, but at the very least, a model that cannot reproduce the observation to a reasonable degree is not likely to be the correct model. I recommend that the model be modified and parameterized so that the model can, at least to a reasonable degree, reproduce the observations. Susiluoto et al. (2018) provided a sophisticated method for parameterizing the HIMMELI module; such exercise could improve the representation of the model in each site.
Authors Response (AR): We will add model parameter optimization for all the sites to improve the model results for CH4 fluxes.
The manuscript showed the model performance in terms of CH4 emissions, soil temperature, WTD, LAI, and soil carbon. The advantage of using eddy covariance data for model validation is that the various factors, including CO2 and water vapor exchanges, involved in CH4 dynamics can be validated simultaneously. Did the model have the ability to reproduce CO2 fluxes (GPP, ecosystem respiration, NEP), evapotranspiration, and net radiation in each site? Describing their performance is necessary to interpret the model performance for CH4 dynamics.
AR: We will check the model performance on CO2 fluxes, evapotranspiration in the revised manuscript.
This study seems to assume that there was a general parameter set for simulating CH4 dynamics across the wetlands. Since the type and origin of the wetlands were diverse (Table 1), the parameter set could be site specific. I understand that such site specific treatments are difficult on a global scale, but this is a site-specific study; so, careful consideration of site specific parameterization would be required. Furthermore, please show the list of important parameters for representing CH4 dynamics in the manuscript.
AR: We will perform site-specific parameter optimization for HIMMELI and identify a list of important parameters.
The descriptions regarding accuracy/precision are not quantitative and thus vague. Please provide quantitative descriptions when describing accuracy/precision (mean bias, RMSE, R2). Please show R2 and p-value when describing the presence or absence of correlation.
AR: We will add the quantitative descriptions to our results, including accuracy metrics (mean bias, RMSE, R2) and p-value to describe the correlation between variables. By including these quantitative descriptions, we will provide a more thorough and objective analysis of our results.
Specific comments:
Line 200: For Taylor diagram of soil temperature, how the authors standardized RMSE and SD (i.e., using mean temperature in Kelvin or Celsius)? Furthermore, if the authors use Taylor diagrams for evaluating precision, be consistent using Taylor diagrams for WTD, and CH4 fluxes.
AR: The ratio of standard deviation is calculated by dividing the standard deviation of model results by the standard deviation of the corresponding measurements. The RMSE wasnot standardized. The unit of temperature doesnot lead to differences in those metrics. We will keep the presentation format (metrics and visualization) of our evaluation consistent across all variables.
References
Susiluoto et al. (2018) Geosci. Model Dev., 11, 1199-1228.
Citation: https://doi.org/10.5194/bg-2022-229-AC1
-
AC1: 'Reply on RC1', Yao Gao, 03 Mar 2023
-
RC2: 'Comment on bg-2022-229', Anonymous Referee #2, 03 Feb 2023
Gao and colleagues took the JULES land surface model and coupled it to the HIMMELI peatland CH4 emission model. They then evaluated the coupled model over five high-latitude peatland sites using an assortment of different model setups. Several of the different model setups were used to test the influence of using observed water table depths (WTD), prescribed LAI, modified soil carbon decomposition rate, etc. The paper is generally clearly written, but strangely contains no equations. This makes it challenging to understand how the coupled models work. The authors give references but it would be much more preferable to give high-level equations so that the basics of the models could be seen. This is mostly relevant to HIMMELI as it would have been nice to more easily see how the inputs from JULES can be traced to estimated CH4 emissions.
One of my most significant issues with this paper can be found in statements like this (line 400, emphasis mine): "It is found that the coupled model does not lead to significant improvements in the simulated CH4 emissions in comparison to the default wetland CH4 emission scheme in JULES. Nevertheless the coupled model is considered to be more realistic in simulating wetland CH4 emissions as HIMMELI simulates CH4 production, oxidation and transportation by vascular plants, ebullition and diffusion in a vertical soil column, while the default wetland methane emission scheme in JULES only takes into account CH4 emission from saturated wetlands in a simple way." - How is this statement justifiable in any way? This is like saying even though we don't show any improved skill with our model, it is better because it is more complicated or because it has lots of equations! This is absolutely the wrong conclusion from this work. If the model does not add skill then the parameters associated with it and the processes simulated are unconstrained and the model is not useful. The authors even note on line 416, that there are no O2 observations. So then how is the O2 aspect of HIMMELI going to be properly parameterized if measurements for it are not available at any site? It could be then a tuning parameter, but what is the value then of the added complexity if it just adds another unconstrained parameter to contribute to model equifinality? I find other major issues with the work. How do you simulate peatlands with no moss, shrubs, or sedges? All sites (Table 1) have one of those vascular plants along with moss and are not C3 grass dominant. How do you build peat when there is no simulation of moss? These aren't tropical peatlands, most of the peat is coming from the highly recalcitrant litter from the moss. This makes me wonder why JULES was chosen since it seems like an unsuitable land surface model to couple with HIMMELI (although the most recent JULES might be a good choice). I recommend rejection with possible resubmission after major revisions. I think the authors need to clearly demonstrate what scientific or technical advance this work represents. At the present I have trouble seeing what that advancement would be considering the models do not demonstrate an improvement in skill over the basic land surface model. The work also offers little insight into the processes represented.
Smaller comments:
Line 33: The WTD and soil C show large deviations from the obs. So then how could the simulated CH4 emissions be expected to get the right answer for the right reason? The paper would be improved if there was some deeper insights into how to correct these problems.
L 95: TOPMODEL (all caps)
L 159: I don't understand the point about the Song et al. paper. Why mention this?
L 193: Don't ref a webpage that will change over time. Put the configuration used in the appendix so it will be pinned to this work.
L 196: As mentioned above, I don't know how you can simulate peatlands when they are set up to be grasslands. This seems like a fundamentally flawed starting point.
L 198: How were the soil params adjusted? What params?
L 200: 'hundreds of times'? How was it checked whether the model was in equilibrium?
L 203: Show equations please. It is very hard to understand how this all ties in without equations. They are needed in a modelling paper.
L 207: SW is the typical abbreviation for shortwave. I suggest using something like SM for soil moisture.
L 234: I find the soil T correlation to the observed surprisingly good considering how poorly the WTD was simulated. Are these correlations for the runs using simulated or observed WTD? If with the simulated, isn't that surprising that the WTD can be so off but the soil T so good? The latent heat impacts of the water has a strong impact on the soil T so I would expect the soil T would be pretty poor if the soil was too dry/wet.
L 243: It would be good to see mean bias values too.
L 265: Mer Bleue has about 25 cm diff between its hummocks and hollows. What is the model compared to here?
L 273: Why was no run done with the observed LAI? It seems like that would be easy and a natural way to check the influence of the LAI.
L 288: I have no idea what to make of the comparison between what would be a simulated grassland soil and each site's peat soil. If the simulated soil C is in-line with the peatland, doesn't that indicate that the model is actually doing some pretty weird stuff? This is C3 grass after all.
L 323: The biases don't really say what is written. Lompo gets worse while Degero changes less than Siikaneva (which was said to be smaller change)
L 381: Quantify the correlation here, and elsewhere should also be properly quantified. There is a lot of vague wording that could be replaced by stats.
Data and code availability does not follow proper open science standards. It is unreasonable to ask people to email for code or data. Please modernize your practices here.
Fig 1: What are the soil layer thicknesses in the model(s)? 5 cm or thicker? If thicker, then how do you compare at 5 cm?
Fig 5: Doesn't take into account any uncertainty in the observations, which I would assume to be considerable given they aren't the site obs.
Citation: https://doi.org/10.5194/bg-2022-229-RC2 -
AC2: 'Reply on RC2', Yao Gao, 03 Mar 2023
We deeply appreciate all the reviewers comments for improving the scientific significance and quality of this manuscript. Our point-to-point response to all the reviewer2’ comments are listed below.
Response to referee comments 2:
Gao and colleagues took the JULES land surface model and coupled it to the HIMMELI peatland CH4 emission model. They then evaluated the coupled model over five high-latitude peatland sites using an assortment of different model setups. Several of the different model setups were used to test the influence of using observed water table depths (WTD), prescribed LAI, modified soil carbon decomposition rate, etc. The paper is generally clearly written, but strangely contains no equations. This makes it challenging to understand how the coupled models work. The authors give references but it would be much more preferable to give high-level equations so that the basics of the models could be seen. This is mostly relevant to HIMMELI as it would have been nice to more easily see how the inputs from JULES can be traced to estimated CH4 emissions.
AR:Including HIMMELI equations in this paper would be redundant. Instead, we will add references to the equations in HIMMELI paper to demonstrate the link between JULES and HIMMELI.
One of my most significant issues with this paper can be found in statements like this (line 400, emphasis mine): "It is found that the coupled model does not lead to significant improvements in the simulated CH4 emissions in comparison to the default wetland CH4 emission scheme in JULES. Nevertheless the coupled model is considered to be more realistic in simulating wetland CH4 emissions as HIMMELI simulates CH4 production, oxidation and transportation by vascular plants, ebullition and diffusion in a vertical soil column, while the default wetland methane emission scheme in JULES only takes into account CH4 emission from saturated wetlands in a simple way." - How is this statement justifiable in any way? This is like saying even though we don't show any improved skill with our model, it is better because it is more complicated or because it has lots of equations! This is absolutely the wrong conclusion from this work. If the model does not add skill then the parameters associated with it and the processes simulated are unconstrained and the model is not useful. The authors even note on line 416, that there are no O2 observations. So then how is the O2 aspect of HIMMELI going to be properly parameterized if measurements for it are not available at any site? It could be then a tuning parameter, but what is the value then of the added complexity if it just adds another unconstrained parameter to contribute to model equifinality?
AR: We will perform site-specific parameter optimization for HIMMELI and identify a list of important parameters. We also take into account the O2 aspect by optimizing parameters that are related to oxidation of methane which is dominantly not co-located with production in the soil column, and has different temperature dependence. So, indirectly, we are also take O2-related processes into account although there are no direct observations of O2.
I find other major issues with the work. How do you simulate peatlands with no moss, shrubs, or sedges? All sites (Table 1) have one of those vascular plants along with moss and are not C3 grass dominant. How do you build peat when there is no simulation of moss? These aren't tropical peatlands, most of the peat is coming from the highly recalcitrant litter from the moss. This makes me wonder why JULES was chosen since it seems like an unsuitable land surface model to couple with HIMMELI (although the most recent JULES might be a good choice).
AR: Currently, the JULES PFT includes tropical broadleaf evergreen trees, temperate broadleaf evergreen trees, broad-leaf deciduous trees, needle-leaf evergreen trees, needle-leaf deciduous trees, C3 grass, C4 grass, arctic grass, evergreen shrubs, deciduous shrubs. Peat is described with peat soil parameter values in this version of JULES used in this manuscript. JULES has a dynamic vegetation model, however, the simulated vegetation types by the dynamic vegetation model at our studied sites were not realistic. In land surface modelling, the most suitable PFT is often chose from the available PFTs with modification according to the need. C3 grass has been shown to work well to represent the productive sedge fens common in high northern latitudes (Qiu et al., 2018). Due to the lack of representation of moss in JULES, we applied C3 grass to represent the average vegetation growing in northern peatlands and the maximum LAI for C3 grass was decreased from 3 m2/m2 to 1.3 m2/m2 according to measurement at Lompölöjankka site. We acknowledge that inappropriate PFT will introduce uncertainties in simulated peat, as recognized in Chadburn et al. (2022), where a new peat module was developed for JULES and realistic peat soil carbon profile were archived. However, a moss PFT with its unique hydraulic and ecological functions in JULES was also suggested to be considered when simulating peat. In general, wetland related ecology, hydrology, and soil carbon processes of JULES are need further development. The original CH4 emission module in JULES is very simple. Nevertheless, JULES with its methane module has been used to estimate the methane emissions from wetlands, and to investigate the influence of climate variability on methane emissions (Gedney et al., 2019). In this work, we implemented a process-based wetland CH4 emission model and evaluated the modelled performance over the studied sites. Our aim was to gain more insight on the model deficiencies and to improve the simulated CH4 fluxes. Parameter optimization will be added to the revised manuscript to improve the simulated CH4 flux and gain more insight into the model processes.
Chadburn, S. E., Burke, E. J., Gallego-Sala, A. V., Smith, N. D., Bret-Harte, M. S., Charman, D. J., Drewer, J., Edgar, C. W., Euskirchen, E. S., Fortuniak, K., Gao, Y., Nakhavali, M., Pawlak, W., Schuur, E. A. G., and Westermann, S.: A new approach to simulate peat accumulation, degradation and stability in a global land surface scheme (JULES vn5.8_accumulate_soil) for northern and temperate peatlands, Geosci. Model Dev., 15, 1633–1657, https://doi.org/10.5194/gmd-15-1633-2022, 2022.
Gedney, N., Huntingford, C., Comyn-Platt, E. and Wiltshire, A.: Significant feedbacks of wetland methane release on climate change and the causes of their uncertainty, Environmental Research Letters, 2019, 14(8): 084027.
Qiu, C., Zhu, D., Ciais, P., Guenet, B., Krinner, G., Peng, S., Aurela, M., Bernhofer, C., Brümmer, C., Bret-Harte, S., Chu, H., Chen, J., Desai, A. R., Dušek, J., Euskirchen, E. S., Fortuniak, K., Flanagan, L. B., Friborg, T., Grygoruk, M., Gogo, S., Grünwald, T., Hansen, B. U., Holl, D., Humphreys, E., Hurkuck, M., Kiely, G., Klatt, J., Kutzbach, L., Largeron, C., Laggoun-Défarge, F., Lund, M., Lafleur, P. M., Li, X., Mammarella, I., Merbold, L., Nilsson, M. B., Olejnik, J., Ottosson-Löfvenius, M., Oechel, W., Parmentier, F.-J. W., Peichl, M., Pirk, N., Peltola, O., Pawlak, W., Rasse, D., Rinne, J., Shaver, G., Schmid, H. P., Sottocornola, M., Steinbrecher, R., Sachs, T., Urbaniak, M., Zona, D., and Ziemblinska, K.: ORCHIDEE-PEAT (revision 4596), a model for northern peatland CO2, water, and energy fluxes on daily to annual scales, Geosci. Model Dev., 11, 497–519, https://doi.org/10.5194/gmd-11-497-2018, 2018.
I recommend rejection with possible resubmission after major revisions. I think the authors need to clearly demonstrate what scientific or technical advance this work represents. At the present I have trouble seeing what that advancement would be considering the models do not demonstrate an improvement in skill over the basic land surface model. The work also offers little insight into the processes represented.
Smaller comments:
Line 33: The WTD and soil C show large deviations from the obs. So then how could the simulated CH4 emissions be expected to get the right answer for the right reason? The paper would be improved if there was some deeper insights into how to correct these problems.
AR: Soil respiration and WTD simulated by JULES are used to calculate anoxic soil respiration, which is an input variable for HIMMELI. In HIMMELI, the anoxic soil respiration is then multiplied with fm, which is the fraction of the anoxic respiration used to produce methane. These two, anoxic respiration and fm, are the most uncertain parameters of the model system, they are strongly correlated in optimization and they are most important determinants of the level of methane emissions. Because lack of measured anoxic soil respiration and soil respiration, we compared the simulated and observed WTD and soil carbon in the manuscript as the background information. Nevertheless, the correct estimation of WTD and soil C doesnot really mean correct anoxic soil respiration. Therefore, we will use parameter optimization of the model to improve the model results.
L 95: TOPMODEL (all caps)
AR: We will correct this.
L 159: I don't understand the point about the Song et al. paper. Why mention this?
AR: We adopted the soil carbon density from the IGBP-DIS dataset to compare with the JULES simulated soil carbon density, due to the scarcity of in-situ measured soil carbon density. Song et al.’s paper adopted IGBP dataset as the initial soil carbon condition for CH4 emission modelling at multiple sites. We adopted Song et al.’s paper here just to show that the the gridded soil carbon data from IGBP-DIS has been used for the site study. However, we now think it is not so meaningful to show this comparison in the manuscript. Thus, we will delete the relevant text and figure regarding to this data.
L 193: Don't ref a webpage that will change over time. Put the configuration used in the appendix so it will be pinned to this work.
AR: This will be changed according to reviewer’s suggestion.
L 196: As mentioned above, I don't know how you can simulate peatlands when they are set up to be grasslands. This seems like a fundamentally flawed starting point.
AR: The main problem with not simulating peat correctly in the JULES version used in this study is the lack of proper peat accumulation and degradation processes. In Chadburn et al. (2022), a new approach to simulate peat accumulation, degradation and stability in JULES was developed. This new approach showed realistic profiles of soil organic carbon for peatlands. To improve the simulation of peat, a moss PFT and its unique functions were suggested to be included in JULES. In land surface modelling, the most suitable PFT is often chose from the available PFTs with modification according to the need. C3 grass has been shown to work well to represent the productive sedge fens common in high northern latitudes (Qiu et al., 2018). Due to the lack of representation of moss in JULES, we applied C3 grass with a lower LAI to represent the average vegetation growing in northern peatlands.
L 198: How were the soil params adjusted? What params?
AR: The model reads a list of soil parameters (see http://jules-lsm.github.io/vn4.9/namelists/ancillaries.nml.html?highlight=soil%20parameters#list-of-soil-params) from an ancillary file. In order to improve the simulated WTD, we manually adjusted the soil parameters impacting soil hydrology within their parameter range. These adjustments were made to the following parameters: b - exponent in soil hydraulic characteristics; satcon - hydraulic conductivity; sm_crit - volumetric soil moisture content at the critical point; sm_sat – volumetric soil moisture content at saturation; sm_wilt – volumetric soil moisture content at the wilting point. We will list the turned soil parameters in the text.
L 200: 'hundreds of times'? How was it checked whether the model was in equilibrium?
AR: We repeated the forcing data for 400 cycles to archieve soil carbon spin up. The spin up was checked in the way that the soil moisture, soil temperature and soil carbon reach a state of equilibrium under the applied forcing.
L 203: Show equations please. It is very hard to understand how this all ties in without equations. They are needed in a modelling paper.
AR: We will show the equations in appendix.
L 207: SW is the typical abbreviation for shortwave. I suggest using something like SM for soil moisture.
AR: This will be changed according to reviewer’s comments.
L 234: I find the soil T correlation to the observed surprisingly good considering how poorly the WTD was simulated. Are these correlations for the runs using simulated or observed WTD? If with the simulated, isn't that surprising that the WTD can be so off but the soil T so good? The latent heat impacts of the water has a strong impact on the soil T so I would expect the soil T would be pretty poor if the soil was too dry/wet.
AR: The WTD produced by JULES is the grid box mean water table depth, which is calculated by the TOPMODEL approach (Gedney and Cox, 2003). The soil temperature and soil moisture are simulated in a vertical layered soil column (Burke et al., 2017). This is the main reason that the simulated WTD showed large biases to the observed WTD, despite the simulated soil temperature performing well.
Burke, E. J., Chadburn, S. E., and Ekici, A.: A vertical representation of soil carbon in the JULES land surface scheme (vn4.3_permafrost) with a focus on permafrost regions, Geoscientific Model Development, 10, 959– 975, doi:10.5194/gmd-10-959-2017,http://www.geosci-model-dev.net/10/959/2017/, 2017a.
Gedney N, Cox P M.: The sensitivity of global climate model simulations to the representation of soil moisture heterogeneity, Journal of Hydrometeorology, 4(6), 1265-1275, 2003.
L 243: It would be good to see mean bias values too.
AR: This will be modified according to reviewer’s comments.
L 265: Mer Bleue has about 25 cm diff between its hummocks and hollows. What is the model compared to here?
AR: This need to be further checked with the data provider. The provided data includes soil temperature measured from hollows and hummocks but only one type of water table depth.
L 273: Why was no run done with the observed LAI? It seems like that would be easy and a natural way to check the influence of the LAI.
AR: We donot have observed LAI in daily. We have the run with Sentinel-2 maximum LAI for Kopytkowo site.
L 288: I have no idea what to make of the comparison between what would be a simulated grassland soil and each site's peat soil. If the simulated soil C is in-line with the peatland, doesn't that indicate that the model is actually doing some pretty weird stuff? This is C3 grass after all.
AR: A new approach to simulate peat accumulation, degradation and stability in JULES has been developed and demonstrating realistic profiles of soil organic carbon for peatlands (Chadburn et al., 2022). However, it has been suggested that the inclusion of a moss PFT and its unique functions would future enhance the accuracy of JULES simulations. While the ecological and hydraulic behavior of mosses is significant, it is important to note that peat accumulation and degradation processes play a more crucial role in determining the carbon profile of peat soil. For the soil in the model it does not matter where the C is coming, as long as there is correct amount of it.
L 323: The biases don't really say what is written. Lompo gets worse while Degero changes less than Siikaneva (which was said to be smaller change)
AR: The overestimated CH4 flux changed to be underestimated. Degero changes more, which is from 0.18 to -0.13. Siikaneva changes from 0.52 to 0.42.
L 381: Quantify the correlation here, and elsewhere should also be properly quantified. There is a lot of vague wording that could be replaced by stats.
AR: We will improve those vague wording parts with statistics.
Data and code availability does not follow proper open science standards. It is unreasonable to ask people to email for code or data. Please modernize your practices here.
Fig 1: What are the soil layer thicknesses in the model(s)? 5 cm or thicker? If thicker, then how do you compare at 5 cm?
AR: The soil layer thickness in JULES model was set to be 0.05m, 0.134m, 0.248m, 0.389m, 0.557m, 0.748m, 0.964m, 1.201m, 1.461m, 1.742m, 2.044m, 2.367m, 2.709m, 3.07m. So there is no problem in comparing soil temperature at 5cm depth.
Fig 5: Doesn't take into account any uncertainty in the observations, which I would assume to be considerable given they aren't the site obs.
AR: In our manuscript, we cited Song et al. (2020) to demonstrate the use of gridded soil carbon data from IGBP-DIS dataset for the site study, which was due to the limited availability of measured soil carbon density at the sites. There is no information of the uncertainties of IGBP-DIS soil carbon density data. We now think it is not so meaningful to show this comparison in the manuscript. Thus, we will delete the relevant text and figure regarding to this data.
Citation: https://doi.org/10.5194/bg-2022-229-AC2
-
AC2: 'Reply on RC2', Yao Gao, 03 Mar 2023
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
590 | 234 | 45 | 869 | 87 | 33 | 33 |
- HTML: 590
- PDF: 234
- XML: 45
- Total: 869
- Supplement: 87
- BibTeX: 33
- EndNote: 33
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1