the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Improvement of the soil drainage simulation based on observations from lysimeters
Bertrand Decharme
Florence Habets
Christine Delire
Noële Enjelvin
Paul-Olivier Redon
Pierre Faure-Catteloin
Patrick Le Moigne
Abstract. Soil drainage is the main source of groundwater recharge and river flow. It is therefore a key process for water resource management. In this study, we evaluate the soil drainage simulated by the Interaction-Soil-Biosphere-Atmosphere (ISBA) land surface model currently used for hydrological applications from the watershed scale to the global scale. This validation is done using seven lysimeters from two long term experiment sites measuring hourly water dynamics between 2009 and 2019 in northeastern France. These 2-meter deep lysimeters are filled with different soil types and are either maintained bare soil or covered with vegetation. The commonly used closed-form equations describing soil-water retention and conductivity curves from Brooks and Corey (1966) and van Genuchten (1980) are tested. The results indicate a good performance by the different experiments in terms of soil volumetric water content and water mass. The drained flow at the bottom of the lysimeter is well modeled using Brooks and Corey (1966) while some weaknesses appears with van Genuchten (1980) due to the complexity of its hydraulic conductivity function. Combining the soil-water curve of van Genuchten (1980) with the hydraulic conductivity function of Brooks and Corey (1966) allow to solve this problem and even to improve the simulation of the drainage dynamic, especially for intense drainage events. The study highlights the importance of the vertical heterogeneity of the soil hydrodynamic parameters to correctly simulate the drainage dynamic, as well as the primary influence of the n and b parameters which characterize the shape of the soil-water retention curve.
- Preprint
(2180 KB) - Metadata XML
- BibTeX
- EndNote
Antoine Sobaga et al.
Status: closed
-
RC1: 'Comment on egusphere-2022-274', Anonymous Referee #1, 07 Jun 2022
The manuscript focuses on the use of a Richards-based solver to reproduce hydrological observations from multiple lysimeters in France. The case study is also used to compare three soil hydraulic models (i.e., Brooks-Corey, van Genuchten-Mualem, a combination of both). The aim is relevant for HESS and somehow interesting, however the manuscript possesses multiple methodological weaknesses:
- The choice to use the ISBA LSM model, which was conceived to operate on larger scales, to investigate a process at the lysimeter level (and prove a soil physics point: Brooks vs van Genuchten) is questionable. The model solves the Richards equation using a Crank-Nicolson scheme but there are no details about the spatial discretization, boundary conditions, etc. By reading this (https://doi.org/10.1029/2018MS001545), the model seems to use a multi-layer approach based on the finite difference. Widely used vadose zone hydrological model such as HYDRUS or SWAP use schemes that comply with the mass conservative approach proposed by Celia et al. (1990). These models have been widely tested, and would be a more rational choice to investigate processes at the lysimeter level and compare multiple soil hydraulic models.
- The whole methodology on the comparison between model predictions and observations is cumbersome to read, not novel, and weak.
- No error metric is reported to compare multiple soil hydraulic models. Besides fitting (which should be quantified), other metrics should be used to compare also the complexity of the models (e.g., at least Akaike Information Criterion)
- The calibration procedure should compare time series of modeled and observed soil water quantities (e.g., water contents). An objective function or a likelihood (e.g., NSE, Gaussian, etc) should be selected, and a numerical algorithm should be used to perform the model calibration. Further, parameters uncertainty should be assessed to see how informative are data, and whether the choice of a more complex model is justified. Only after having performed a statistically robust analysis, it is possible to try to explain why BC+VG is better and when. As they are, methods don’t support enough the conclusions, and neither represent a novel contribution to the field.
Specific comments:
L15-20 Not really. Drainage is the amount of water that bypasses the root zone.
L47-50 Nonlinearity cannot be a source of criticism, otherwise an endless number of equations used in environmental modeling should be “criticized”. I would remove this part. Richards equation is not perfect, but we are still far from finding a viable, widely used, and extensively validated alternative.
L56 BC66 has that sharp singular point near the air-entry pressure that makes it not very stable. (https://doi.org/10.1029/93WR03238). Authors indeed discuss this point later. However, more specific references are needed to prove your point that BC66 is more numerically stable than VG80.
Data: Please add details about TDR sensors (e.g., type, accuracy, calibration type) and tipping bucket resolution
L124-125 Is the heat transport included in the numerical simulation of lysimeters? If yes, key equations should be provided. Otherwise, it should be removed from the text.
L130-140 This part should be moved after the Richards equation, and should describe how it is connected to the sink term S(z). Key equations should provided. Citing refs is good, but the manuscript should stand by itself.
L147-148 what is the discretization of the soil profile? What are the boundary conditions used?
Figure1. Very confusing. It is difficult to appreciate differences. What are the dashed lines? Figure+Caption should be self-explanatory
L176 There is not a single error metric to support the conclusion that one formulation is better than the other. It is really puzzling to see that.
L192 VG80 not stable for n<1.3?! Never experienced something like this. Indeed, I agree with the Authors that n>1.1 is a good constraint.
L197 Having a highly negative tortuosity is not recommended. Actually Schaap suggests a value of -1 (https://doi.org/10.2136/sssaj2000.643843x)
Citation: https://doi.org/10.5194/egusphere-2022-274-RC1 - AC1: 'Reply on RC1', antoine sobaga, 18 Sep 2022
-
RC2: 'Comment on egusphere-2022-274', Anonymous Referee #2, 05 Aug 2022
This manuscript presents the results of an LSM application using two different approaches according to BC and VG to reproduce soil water mass, volumetric water content, and drainage water flux volume observed in seven lysimeters over a period of more than five years. Furthermore, approaches by Braud et al. (1995) and Valiantzas (2011) are tested to simulate the soil hydrology of these lysimeters. They derived hydrodynamic parameters directly from the observation and compare them with several pedotransfer functions commonly used by LSMs. The LSM used has a multilayer diffusion approach of the Interaction-Soil-Biosphere-Atmosphere (ISBA) model, which solves a variant of the Richards equation.
- The term "drainage" in this context means the transit of a liquid through a porous medium. In the present case, it is water through the upper soil layers. It is neither a quantity nor a volume. Therefore, if the amount of water is to be addressed this must be explicitly stated as drainage water.
- For the lysimeters, the experimental setup is sufficiently described, but the lower boundary condition is not mentioned in detail as a special feature of the lysimeter. Since drainage in particular is considered as a special aspect, this has to be described in detail for the lysimeters. Especially the consequences/impacts of the chosen design on the drainage amount of water must be discussed. Otherwise, it is assumed here that the lower boundary layer of the lysimeter corresponds to a naturally layered soil, and this is de facto not the case.
- The methods section on the comparison between the model predictions and the lysimeter observations is very unclearly written and needs a more comprehensible description.
- Why these applied models were selected is not convincingly presented, especially since there are more current modeling approaches that promise better simulation of processes and results.
- Lysimeters provide "point" information compared to LSM. Here, indications are missing how this discrepancy is addressed or how lysimeter results could be scaled.
Line 107: A specification of the measurement resolution is missing here.
Line 181 ff: Is this the case? It is often stated that in zero-tension lysimeters, the seepage water formation takes place under water-saturated conditions. I am not aware of any study that has decisively investigated this. Of course, small-scale saturated structures are also conceivable with corresponding fingering. Is this the case? It is often stated that in zero-tension lysimeters, leachate formation takes place under water-saturated conditions. I am not aware of any study that has decisively investigated this. Of course, small-scale saturated structures are also conceivable with corresponding fingering. Especially before the background of a very heterogeneous material of a former industrial site, which was filled manually into lysimeters. Hydrophobic structures are also conceivable.
Line 200 ff: Should be discussed later, because it is manual filling with disturbed profiles. This has an impact on the parameter estimation.
Line 233 ff: “At the bottom of the soil“ What do you mean by this?This wording is very unclear or does not make sense.
Line 242: Masses of what? Water?
Line 266f: I do not understand this argumentation. The temporal resolution is criticized as limiting and therefore I reduce the temporal resolution even more or aggregate the data?
Line 265ff: In order to be able to classify the different seepage water quantities, a distinction must be made between vegetated and unvegetated lysimeters. This has been done. But to be able to investigate or classify the differences between the vegetated lysimeters, measurements of the crop development (LAI) or the crop yield (harvest amount), etc. are absolutely needed. Only with this information different ETp results can be classified.
Line 289 ff: Also, non rainfall water like dew, hoar frost, etc.
Table 1: Regarding the contents of the table: am I correct in assuming that the remainder is 100% silt? If not, what then? But should still be presented in more detail for clarity. To standardize the presentation, the number of decimal places should be the same for all data.
Figure 3, 5-7: Measurement units of the Y-axes are missing
Furthermore, I share the remarks of the reviewer 1.
Citation: https://doi.org/10.5194/egusphere-2022-274-RC2 - AC2: 'Reply on RC2', antoine sobaga, 18 Sep 2022
-
RC3: 'Comment on egusphere-2022-274', Wolfgang Durner, 16 Aug 2022
Please see my comments in the attached pdf document.
Wolfgang Durner
-
AC3: 'Reply on RC3', antoine sobaga, 18 Sep 2022
We would like to thank Wolfgang Durner for its very useful review.
Especially, thank you for pointing out the improved version of the Van Genuchten proposed by Iden et al., 2015 which we were not aware of.
We integrated these new closing equations and run new simulations. The results are very good, and we will incorporate them into a new version of the article.
As we mentioned in our article, although Van Genuchten (1980) proposed an improvement in the closure equation compared to Brooks & Corey (1966) , the meteorological and climate modeling community that uses Land Surface Models (LSMs) usually use the Brooks & Corey closure equations, as they are more numerically stable and necessitate less parameters. The introduction of a proposed Iden et al. (2105) air-entry suction allows a major advance for LSM modeling.
This progress is made possible by your involvement in this peer review process and we are very grateful
In the following attached file, we provide details answers to your questions and comments.
-
AC3: 'Reply on RC3', antoine sobaga, 18 Sep 2022
Status: closed
-
RC1: 'Comment on egusphere-2022-274', Anonymous Referee #1, 07 Jun 2022
The manuscript focuses on the use of a Richards-based solver to reproduce hydrological observations from multiple lysimeters in France. The case study is also used to compare three soil hydraulic models (i.e., Brooks-Corey, van Genuchten-Mualem, a combination of both). The aim is relevant for HESS and somehow interesting, however the manuscript possesses multiple methodological weaknesses:
- The choice to use the ISBA LSM model, which was conceived to operate on larger scales, to investigate a process at the lysimeter level (and prove a soil physics point: Brooks vs van Genuchten) is questionable. The model solves the Richards equation using a Crank-Nicolson scheme but there are no details about the spatial discretization, boundary conditions, etc. By reading this (https://doi.org/10.1029/2018MS001545), the model seems to use a multi-layer approach based on the finite difference. Widely used vadose zone hydrological model such as HYDRUS or SWAP use schemes that comply with the mass conservative approach proposed by Celia et al. (1990). These models have been widely tested, and would be a more rational choice to investigate processes at the lysimeter level and compare multiple soil hydraulic models.
- The whole methodology on the comparison between model predictions and observations is cumbersome to read, not novel, and weak.
- No error metric is reported to compare multiple soil hydraulic models. Besides fitting (which should be quantified), other metrics should be used to compare also the complexity of the models (e.g., at least Akaike Information Criterion)
- The calibration procedure should compare time series of modeled and observed soil water quantities (e.g., water contents). An objective function or a likelihood (e.g., NSE, Gaussian, etc) should be selected, and a numerical algorithm should be used to perform the model calibration. Further, parameters uncertainty should be assessed to see how informative are data, and whether the choice of a more complex model is justified. Only after having performed a statistically robust analysis, it is possible to try to explain why BC+VG is better and when. As they are, methods don’t support enough the conclusions, and neither represent a novel contribution to the field.
Specific comments:
L15-20 Not really. Drainage is the amount of water that bypasses the root zone.
L47-50 Nonlinearity cannot be a source of criticism, otherwise an endless number of equations used in environmental modeling should be “criticized”. I would remove this part. Richards equation is not perfect, but we are still far from finding a viable, widely used, and extensively validated alternative.
L56 BC66 has that sharp singular point near the air-entry pressure that makes it not very stable. (https://doi.org/10.1029/93WR03238). Authors indeed discuss this point later. However, more specific references are needed to prove your point that BC66 is more numerically stable than VG80.
Data: Please add details about TDR sensors (e.g., type, accuracy, calibration type) and tipping bucket resolution
L124-125 Is the heat transport included in the numerical simulation of lysimeters? If yes, key equations should be provided. Otherwise, it should be removed from the text.
L130-140 This part should be moved after the Richards equation, and should describe how it is connected to the sink term S(z). Key equations should provided. Citing refs is good, but the manuscript should stand by itself.
L147-148 what is the discretization of the soil profile? What are the boundary conditions used?
Figure1. Very confusing. It is difficult to appreciate differences. What are the dashed lines? Figure+Caption should be self-explanatory
L176 There is not a single error metric to support the conclusion that one formulation is better than the other. It is really puzzling to see that.
L192 VG80 not stable for n<1.3?! Never experienced something like this. Indeed, I agree with the Authors that n>1.1 is a good constraint.
L197 Having a highly negative tortuosity is not recommended. Actually Schaap suggests a value of -1 (https://doi.org/10.2136/sssaj2000.643843x)
Citation: https://doi.org/10.5194/egusphere-2022-274-RC1 - AC1: 'Reply on RC1', antoine sobaga, 18 Sep 2022
-
RC2: 'Comment on egusphere-2022-274', Anonymous Referee #2, 05 Aug 2022
This manuscript presents the results of an LSM application using two different approaches according to BC and VG to reproduce soil water mass, volumetric water content, and drainage water flux volume observed in seven lysimeters over a period of more than five years. Furthermore, approaches by Braud et al. (1995) and Valiantzas (2011) are tested to simulate the soil hydrology of these lysimeters. They derived hydrodynamic parameters directly from the observation and compare them with several pedotransfer functions commonly used by LSMs. The LSM used has a multilayer diffusion approach of the Interaction-Soil-Biosphere-Atmosphere (ISBA) model, which solves a variant of the Richards equation.
- The term "drainage" in this context means the transit of a liquid through a porous medium. In the present case, it is water through the upper soil layers. It is neither a quantity nor a volume. Therefore, if the amount of water is to be addressed this must be explicitly stated as drainage water.
- For the lysimeters, the experimental setup is sufficiently described, but the lower boundary condition is not mentioned in detail as a special feature of the lysimeter. Since drainage in particular is considered as a special aspect, this has to be described in detail for the lysimeters. Especially the consequences/impacts of the chosen design on the drainage amount of water must be discussed. Otherwise, it is assumed here that the lower boundary layer of the lysimeter corresponds to a naturally layered soil, and this is de facto not the case.
- The methods section on the comparison between the model predictions and the lysimeter observations is very unclearly written and needs a more comprehensible description.
- Why these applied models were selected is not convincingly presented, especially since there are more current modeling approaches that promise better simulation of processes and results.
- Lysimeters provide "point" information compared to LSM. Here, indications are missing how this discrepancy is addressed or how lysimeter results could be scaled.
Line 107: A specification of the measurement resolution is missing here.
Line 181 ff: Is this the case? It is often stated that in zero-tension lysimeters, the seepage water formation takes place under water-saturated conditions. I am not aware of any study that has decisively investigated this. Of course, small-scale saturated structures are also conceivable with corresponding fingering. Is this the case? It is often stated that in zero-tension lysimeters, leachate formation takes place under water-saturated conditions. I am not aware of any study that has decisively investigated this. Of course, small-scale saturated structures are also conceivable with corresponding fingering. Especially before the background of a very heterogeneous material of a former industrial site, which was filled manually into lysimeters. Hydrophobic structures are also conceivable.
Line 200 ff: Should be discussed later, because it is manual filling with disturbed profiles. This has an impact on the parameter estimation.
Line 233 ff: “At the bottom of the soil“ What do you mean by this?This wording is very unclear or does not make sense.
Line 242: Masses of what? Water?
Line 266f: I do not understand this argumentation. The temporal resolution is criticized as limiting and therefore I reduce the temporal resolution even more or aggregate the data?
Line 265ff: In order to be able to classify the different seepage water quantities, a distinction must be made between vegetated and unvegetated lysimeters. This has been done. But to be able to investigate or classify the differences between the vegetated lysimeters, measurements of the crop development (LAI) or the crop yield (harvest amount), etc. are absolutely needed. Only with this information different ETp results can be classified.
Line 289 ff: Also, non rainfall water like dew, hoar frost, etc.
Table 1: Regarding the contents of the table: am I correct in assuming that the remainder is 100% silt? If not, what then? But should still be presented in more detail for clarity. To standardize the presentation, the number of decimal places should be the same for all data.
Figure 3, 5-7: Measurement units of the Y-axes are missing
Furthermore, I share the remarks of the reviewer 1.
Citation: https://doi.org/10.5194/egusphere-2022-274-RC2 - AC2: 'Reply on RC2', antoine sobaga, 18 Sep 2022
-
RC3: 'Comment on egusphere-2022-274', Wolfgang Durner, 16 Aug 2022
Please see my comments in the attached pdf document.
Wolfgang Durner
-
AC3: 'Reply on RC3', antoine sobaga, 18 Sep 2022
We would like to thank Wolfgang Durner for its very useful review.
Especially, thank you for pointing out the improved version of the Van Genuchten proposed by Iden et al., 2015 which we were not aware of.
We integrated these new closing equations and run new simulations. The results are very good, and we will incorporate them into a new version of the article.
As we mentioned in our article, although Van Genuchten (1980) proposed an improvement in the closure equation compared to Brooks & Corey (1966) , the meteorological and climate modeling community that uses Land Surface Models (LSMs) usually use the Brooks & Corey closure equations, as they are more numerically stable and necessitate less parameters. The introduction of a proposed Iden et al. (2105) air-entry suction allows a major advance for LSM modeling.
This progress is made possible by your involvement in this peer review process and we are very grateful
In the following attached file, we provide details answers to your questions and comments.
-
AC3: 'Reply on RC3', antoine sobaga, 18 Sep 2022
Antoine Sobaga et al.
Antoine Sobaga et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 508 | 217 | 23 | 748 | 11 | 10 |
- HTML: 508
- PDF: 217
- XML: 23
- Total: 748
- BibTeX: 11
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1