the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 May 2022
Improved representation of groundwater-dominated catchment using SWAT+gwflow and modifications to the gwflow module
Abstract. Recent water availability and scarcity problems have highlighted the importance of surface water-groundwater interactions. Thus, groundwater models such as Modular ground-water Flow (MODFLOW) were coupled to the Soil and Water Assessment Tool (SWAT (+)). However, this solution is complex, needing code modifications, complex coupling, and high computation time. Lately, a new groundwater module (gwflow) was developed directly inside the SWAT+ code to tackle those issues. This research assesses gwflow’s capabilities in representing surface – groundwater systems interactions in the Dijle catchment, Belgium. A hydrological model was set up using the standalone SWAT+ and SWAT+gwflow. In addition, the interaction between the soil and the groundwater is not represented in the new module, hence, it was modified to account for such exchanges. Finally, pumping is also included in the module to enable the modeling of transient state conditions. Model comparison is made using Nash-Sutcliffe efficiency (NSE) for the calibration period (1986 to 1996) and two validation periods (1975 to 1983 and 1997 to 2002).
It is found that the SWAT+gwflow model is better representative (NSE of 0.6) than the standalone SWAT+ (NSE of 0.4). This is signified during two validation periods where the standalone scored negative NSE while the new model’s NSE was 0.7 and 0.5. This shows that, in a highly groundwater-driven catchment of this type, the simplistic representation of groundwater systems by the standalone SWAT+ model has pitfalls. In addition, the modification we made on the gwflow module has improved the model performance as groundwater-soil interaction is inevitable whenever the water table reaches the soil profile. In conclusion, groundwater-surface water processes need to be appropriately designated in hydrological models; hence, the modification we made to the gwflow module (groundwater–soil interaction) is found to be critical. This novel modification can also have an implication for other distributed hydrological models to consider such exchanges in their modeling scheme. This paves a road towards refining coupled ground-surface water hydrogeological models. Finally, the modified SWAT+gwflow model is more appropriate for assessing the Dijle catchment hydrology than the standalone model.
- Preprint
(1030 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on hess-2022-169', Anonymous Referee #1, 27 May 2022
The manuscript "Improved representation of groundwater-dominated catchment using SWAT+gwflow and modifications to the gwflow module" by Yimer et al. is about the development of a new groundwater module for SWAT+ to provide an opportunity to consider a more process based groundwater interaction with the soil and consequently with discharge. This study tries comparing the basic SWAT+ and the newly developed SWAT+gwflow. For this, both model versions are applied on the Dijle catchment, which is non for groundwater dominance.
In general, a more process-based representation of groundwater within hydrological models and additionally a more easy to handle and fast model application should be interesting to the hydrological community. The current manuscript has a string focus on the SWAT model and is mostly written in a form of a technical note. Especially the beginning of the introduction deals with technical details of SWAT, which might be not relevant for all hydrolgical modellers. Consequently it might be worth to rethink the form of presenting this study.
In the following I am going to summarise the most critical shortcomings of the mansucript:
L. 33 - 58: This seems to be a technical summary about SWAT. It is unclear why such detailed information of SWAT is provided at this position. What is the demand for further research regarding the general hydrological community?
L. 59: There are several efforts to improve groundwater representation in hydrological models. Since the authors are focused on SWAT, it would be scientificly sound to mention and discuss these efforts at least for current and previous SWAT versions and to work out the necessity to consider other approaches (e. g. the current study).
L. 60: This seems to be the motivation of this study?
L. 93: Limitations of SWAT were also shown in Luo et al. (2012), Pfannerstill et al. (2013), Nguyen and Dietrich (2018), Shao et al. (2019). Additionally, these studies provided solutions to improve the representation of groundwater within SWAT.
L. 97: Please provide a clear structure: what are the gaps? which problems need to be solved? how are the problems solved? The introduction ends up with a rather technical information, which is not appropriate to guide readers through this study. It would be more helpful to provide the aim of this study at the end of the introduction.
L. 151: The introduction ends up mentioning the integration of pumping. This part is not mentioned here?
L. 215: Please provide references.
L. 274: It is unclear which parameters and which ranges were used for the sensitivity analysis. This is very crucial to reproduce the results of this study and to check if this method is appropriate.
L. 285: The whole section about sensitivity is unclear due to missing information (selected parameters, parameter ranges).
L. 286: It would be better to provide a table with calibrated parameters.
L. 302: It is impossible to follow the author's opinion and conclusion about poor/good model performance and the comparison between the different model structures.
1. Please provide a table with calibrated parameter values for both model setups.
2. Please provide appropriate figures for simulated/observed discharge. Figue 5b does not show any meaningful information since data is very compressed. A flow duration curve would be helpful to show the deviation between simulated and observed discharge magnitude. Furthermore, it seems that several flow events are not adequately reproduced by timing and magnitude. This may be due to the fact that a weather station inside the catchment is missing! What about other performance measures? Poor model performance, especially for the basic SWAT, might be related to the selection of inappropriate performance measures or imbalanced weighting.
L. 355: The evaluation based on monthly discharge should play a minor role. Of course, groundwater processes are more time-delayed and could be checked at a monthly time scale. However, other crucial hydrological processes happen immediately on a precipitation event. For a consistent model behaviuor, all relevant processes need to be represented and of course they need to be evaluated. Consequently, it is necessary to focus to an appropriate extent on daily discharge.
L. 343: Are these model results realistic? Are there any options to evaluate this?
L. 356: Please provide discharge for the validation periods with an additional figure.
L. 377: It is impossible to reproduce this conclusion since necessary information was not provided (parameters, parameter ranges, appropriate figures for daily discharge).
L. 389: This conclusion cannot be confirmed due to missing information.
L. 392: This aspect is fully new and was not mentioned before.
Due to the mentioned shortcomings I see the necessity for a major revision.
Citation: https://doi.org/10.5194/hess-2022-169-RC1 -
AC1: 'Comment on hess-2022-169', Estifanos Addisu Yimer, 27 Jun 2022
First of all, we would like to express our sincere appreciation to Reviewer 1 for the positive overall assessment of our manuscript and for the critical questions that have led to an improved and clearer manuscript.
Please find below the detailed answers to the question:
- 33 - 58: This seems to be a technical summary about SWAT. It is unclear why such detailed information of SWAT is provided at this position. What is the demand for further research regarding the general hydrological community?
The gwflow module is designed to replace the groundwater component's conceptual representation in SWAT+, and we focused on improving this new module. To provide the reader with the reason behind, we went from the basics so that every detail can be understood. It is also with an intention to take this article as a reference on how the coupling of SWAT progressed in time and its limitations and advancements. As you have stated in your subsequent question (question number 2), our focus is to deal with SWAT.
The major outcome of our research entails an important implication to the hydrological community, where not accounting for groundwater-soil interaction can be an essential factor for unsatisfactory hydrological simulations in hydrological models. Hence, other hydrological models should consider such interactions to pave the road for effective ground-surface water coupling.
- 59: There are several efforts to improve groundwater representation in hydrological models. Since the authors are focused on SWAT, it would be scientifically sound to mention and discuss these efforts at least for current and previous SWAT versions and to work out the necessity to consider other approaches (e. g. the current study).
The current and previous SWAT versions have a similar approach when it comes to groundwater hydrology representation. They have lumped conceptual representation, which has only been changed after coupling techniques came into practice. However, since previous coupling techniques have significant limitations (as stated in lines 82-83), gwflow module is developed. Previous coupling techniques are discussed between lines 66 to 80.
- This seems to be the motivation of this study?
Yes, it is one of the main motivations behind this research where holistic approaches are needed to understand the overall geohydrology of a given catchment. In addition, a comparison of the standalone SWAT+ and SWAT+gwflow is made to show the limitation of the conceptual representation of groundwater hydrology by the standalone model. Moreover, soil-groundwater interactions, which are neglected in the previous version of gwflow are accounted which is the major contribution of this article.
- 93: Limitations of SWAT were also shown in Luo et al. (2012), Pfannerstill et al. (2013), Nguyen and Dietrich (2018), Shao et al. (2019). Additionally, these studies provided solutions to improve the representation of groundwater within SWAT.
Thank you for suggesting more references which we have not come across. The references you have suggested have not been implemented in the SWAT+ model so far, but in the future, we will take them as an input to modify the module or the standalone model even further. We would like to affirm that the main intention of this research is to modify the gwflow module and make a comparison with the standalone SWAT+ model. Hence, you may please take it as an extension of the gwflow module developed by Ryan T.Bailey (Bailey et.al 2020).
- 97: Please provide a clear structure: what are the gaps? which problems need to be solved? how are the problems solved? The introduction ends up with a rather technical information, which is not appropriate to guide readers through this study. It would be more helpful to provide the aim of this study at the end of the introduction.
We have modified the end of the introduction with the suggestions you have made. You may find them in the edited version of the manuscript from line 91 to 97.
- 151: The introduction ends up mentioning the integration of pumping. This part is not mentioned here?
That is correct, but the issue here is that we have not included pumping in the modelling scheme of the Dijle catchment due to data limitations, and we intend to include it in another paper to show the impact of pumping in the general hydrology. Nevertheless, we have tried it with a dummy number, and it worked well. The pumping term is included in equation 3 (Q pump), where the groundwater change in volume with time is solved by including this term. Recently, Bailey et.al 2021 published a paper by including pumping in a catchment located in the USA to investigate the impact of tile drain in the hydrology of the catchment.
- 215: Please provide references.
We have added the reference accordingly.
- 274, L.285 and L.286: It is unclear which parameters and which ranges were used for the sensitivity analysis. This is very crucial to reproduce the results of this study and to check if this method is appropriate.
We have included the parameters used (in Table format) for calibration and the sensitivity analysis results (in plot format) in the supplementary material. In addition, the calibrated parameter sets is also included in the supplementary material.
Note that we had a master’s student who tried to calibrate the SWAT+ model for the Dijle catchment for almost five months, and it was impossible to achieve good agreement between modelled and observed streamflow values, which is the primary motivation for this research. The best parameter set we achieved was based on several simulations made on that thesis (the student is also a co-author of this article - Lise Leda Piepers). The main conclusion of the thesis was that the standalone model could not represent the hydrology effectively due to limitations in the groundwater hydrology representation.
Moreover, to ensure transparency in our work, we are willing to share the model we developed for the Dijle catchment. It is also a catchment we used to prepare the tutorials for SWAT+gwflow model (previous version of the model) and recently we replaced it with the case study we have from Tuscany, Italy – Ombrone catchment. Link for tutorial: https://swat.tamu.edu/software/plus/gwflow/
- 302
- Please provide a table with calibrated parameter values for both model setups.
- You may check the answer we provided above (#8).
- Please provide appropriate figures for simulated/observed discharge. Figue 5b does not show any meaningful information since data is very compressed. A flow duration curve would be helpful to show the deviation between simulated and observed discharge magnitude.
- Thank you for your suggestion, and we have included a flow duration curve to show the deviation. As the flow duration curve has indicated, the differences are significant during peak flows, but those events are few, with flow rate below 30 m3/s in most events. The major player in this catchment is the groundwater hydrology, which is also indicated in the baseflow filtering discussed in question number 11 below.
- 355: The evaluation based on monthly discharge should play a minor role. Of course, groundwater processes are more time-delayed and could be checked at a monthly time scale. However, other crucial hydrological processes happen immediately on a precipitation event. For a consistent model behavior, all relevant processes need to be represented and of course they need to be evaluated. Consequently, it is necessary to focus to an appropriate extent on daily discharge.
It is correct that daily time step can be appropriate, and due to that, our assessment does not only rely on using monthly but also daily model outputs to investigate further. For instance, you may check Table 1, where objective functions are estimated for both daily and monthly time steps.
- 343: Are these model results realistic? Are there any options to evaluate this?
Yes they are realistic, and we can evaluate this by assessing the baseflow component of the streamflow using baseflow filtering techniques (e.g. WETSPRO - Willems 2009). This will allow us to understand how much of the total flow is accounted for by baseflow, interflow, and overland flow. We can infer from the filtering that baseflow accounts for the majority of the streamflow (around 80% of the total streamflow), indicating the prevalence of subsurface than surface processes. You may please check the WETSPRO result below.
Figure: The baseflow filtering parameters (top) where w is the difference between 1 and the percentage of baseflow. The bottom plot shows the filtered baseflow for the catchment outlet.
For a detailed description of the baseflow filtering methodology, you may refer to the link: https://bwk.kuleuven.be/hydr/pwtools.htm#Wetspro. We have also included in the .zip file the filtering excel workbook.
- 356: Please provide discharge for the validation periods with an additional figure.
Thank you for your suggestion, we have included plots for monthly and flow duration curve for daily timestep in the supplementary material for the catchment outlet and additional gauging station inside the catchment.
- 392: This aspect is fully new and was not mentioned before.
One limitation of the SWAT hydrological model is the nonrepresentation of wetlands in the modelling scheme, which is clearly stated in several literature (Golden et.al 2014, Wellen et.al 2015). Especially during drought periods, wetlands rely on their interaction and the water they will receive from the groundwater and/or surface water bodies. On top of this recommendation, here in Belgium, the amount of drainage water from agricultural lands is not well known, which is a concern of the government, and we were asked to assess the amount and the impact it will have on water availability. This is the reason why we put this recommendation on the inclusion of tile drains in hydrological simulations to capture the actual hydrological behavior of a given catchment. We believe that including wetlands and tile drains (ditches) in the modelling scheme of the Dijle catchment can improve the hydrological simulation even further (on top of the improvement we made). To summarize, the points we raised in line 392 and further are recommendations.
-
AC1: 'Comment on hess-2022-169', Estifanos Addisu Yimer, 27 Jun 2022
-
AC1: 'Comment on hess-2022-169', Estifanos Addisu Yimer, 27 Jun 2022
First of all, we would like to express our sincere appreciation to Reviewer 1 for the positive overall assessment of our manuscript and for the critical questions that have led to an improved and clearer manuscript.
Please find below the detailed answers to the question:
- 33 - 58: This seems to be a technical summary about SWAT. It is unclear why such detailed information of SWAT is provided at this position. What is the demand for further research regarding the general hydrological community?
The gwflow module is designed to replace the groundwater component's conceptual representation in SWAT+, and we focused on improving this new module. To provide the reader with the reason behind, we went from the basics so that every detail can be understood. It is also with an intention to take this article as a reference on how the coupling of SWAT progressed in time and its limitations and advancements. As you have stated in your subsequent question (question number 2), our focus is to deal with SWAT.
The major outcome of our research entails an important implication to the hydrological community, where not accounting for groundwater-soil interaction can be an essential factor for unsatisfactory hydrological simulations in hydrological models. Hence, other hydrological models should consider such interactions to pave the road for effective ground-surface water coupling.
- 59: There are several efforts to improve groundwater representation in hydrological models. Since the authors are focused on SWAT, it would be scientifically sound to mention and discuss these efforts at least for current and previous SWAT versions and to work out the necessity to consider other approaches (e. g. the current study).
The current and previous SWAT versions have a similar approach when it comes to groundwater hydrology representation. They have lumped conceptual representation, which has only been changed after coupling techniques came into practice. However, since previous coupling techniques have significant limitations (as stated in lines 82-83), gwflow module is developed. Previous coupling techniques are discussed between lines 66 to 80.
- This seems to be the motivation of this study?
Yes, it is one of the main motivations behind this research where holistic approaches are needed to understand the overall geohydrology of a given catchment. In addition, a comparison of the standalone SWAT+ and SWAT+gwflow is made to show the limitation of the conceptual representation of groundwater hydrology by the standalone model. Moreover, soil-groundwater interactions, which are neglected in the previous version of gwflow are accounted which is the major contribution of this article.
- 93: Limitations of SWAT were also shown in Luo et al. (2012), Pfannerstill et al. (2013), Nguyen and Dietrich (2018), Shao et al. (2019). Additionally, these studies provided solutions to improve the representation of groundwater within SWAT.
Thank you for suggesting more references which we have not come across. The references you have suggested have not been implemented in the SWAT+ model so far, but in the future, we will take them as an input to modify the module or the standalone model even further. We would like to affirm that the main intention of this research is to modify the gwflow module and make a comparison with the standalone SWAT+ model. Hence, you may please take it as an extension of the gwflow module developed by Ryan T.Bailey (Bailey et.al 2020).
- 97: Please provide a clear structure: what are the gaps? which problems need to be solved? how are the problems solved? The introduction ends up with a rather technical information, which is not appropriate to guide readers through this study. It would be more helpful to provide the aim of this study at the end of the introduction.
We have modified the end of the introduction with the suggestions you have made. You may find them in the edited version of the manuscript from line 91 to 97.
- 151: The introduction ends up mentioning the integration of pumping. This part is not mentioned here?
That is correct, but the issue here is that we have not included pumping in the modelling scheme of the Dijle catchment due to data limitations, and we intend to include it in another paper to show the impact of pumping in the general hydrology. Nevertheless, we have tried it with a dummy number, and it worked well. The pumping term is included in equation 3 (Q pump), where the groundwater change in volume with time is solved by including this term. Recently, Bailey et.al 2021 published a paper by including pumping in a catchment located in the USA to investigate the impact of tile drain in the hydrology of the catchment.
- 215: Please provide references.
We have added the reference accordingly.
- 274, L.285 and L.286: It is unclear which parameters and which ranges were used for the sensitivity analysis. This is very crucial to reproduce the results of this study and to check if this method is appropriate.
We have included the parameters used (in Table format) for calibration and the sensitivity analysis results (in plot format) in the supplementary material. In addition, the calibrated parameter sets is also included in the supplementary material.
Note that we had a master’s student who tried to calibrate the SWAT+ model for the Dijle catchment for almost five months, and it was impossible to achieve good agreement between modelled and observed streamflow values, which is the primary motivation for this research. The best parameter set we achieved was based on several simulations made on that thesis (the student is also a co-author of this article - Lise Leda Piepers). The main conclusion of the thesis was that the standalone model could not represent the hydrology effectively due to limitations in the groundwater hydrology representation.
Moreover, to ensure transparency in our work, we are willing to share the model we developed for the Dijle catchment. It is also a catchment we used to prepare the tutorials for SWAT+gwflow model (previous version of the model) and recently we replaced it with the case study we have from Tuscany, Italy – Ombrone catchment. Link for tutorial: https://swat.tamu.edu/software/plus/gwflow/
- 302
- Please provide a table with calibrated parameter values for both model setups.
- You may check the answer we provided above (#8).
- Please provide appropriate figures for simulated/observed discharge. Figue 5b does not show any meaningful information since data is very compressed. A flow duration curve would be helpful to show the deviation between simulated and observed discharge magnitude.
- Thank you for your suggestion, and we have included a flow duration curve to show the deviation. As the flow duration curve has indicated, the differences are significant during peak flows, but those events are few, with flow rate below 30 m3/s in most events. The major player in this catchment is the groundwater hydrology, which is also indicated in the baseflow filtering discussed in question number 11 below.
- 355: The evaluation based on monthly discharge should play a minor role. Of course, groundwater processes are more time-delayed and could be checked at a monthly time scale. However, other crucial hydrological processes happen immediately on a precipitation event. For a consistent model behavior, all relevant processes need to be represented and of course they need to be evaluated. Consequently, it is necessary to focus to an appropriate extent on daily discharge.
It is correct that daily time step can be appropriate, and due to that, our assessment does not only rely on using monthly but also daily model outputs to investigate further. For instance, you may check Table 1, where objective functions are estimated for both daily and monthly time steps.
- 343: Are these model results realistic? Are there any options to evaluate this?
Yes they are realistic, and we can evaluate this by assessing the baseflow component of the streamflow using baseflow filtering techniques (e.g. WETSPRO - Willems 2009). This will allow us to understand how much of the total flow is accounted for by baseflow, interflow, and overland flow. We can infer from the filtering that baseflow accounts for the majority of the streamflow (around 80% of the total streamflow), indicating the prevalence of subsurface than surface processes. You may please check the WETSPRO result below.
Figure: The baseflow filtering parameters (top) where w is the difference between 1 and the percentage of baseflow. The bottom plot shows the filtered baseflow for the catchment outlet.
For a detailed description of the baseflow filtering methodology, you may refer to the link: https://bwk.kuleuven.be/hydr/pwtools.htm#Wetspro. We have also included in the .zip file the filtering excel workbook.
- 356: Please provide discharge for the validation periods with an additional figure.
Thank you for your suggestion, we have included plots for monthly and flow duration curve for daily timestep in the supplementary material for the catchment outlet and additional gauging station inside the catchment.
- 392: This aspect is fully new and was not mentioned before.
One limitation of the SWAT hydrological model is the nonrepresentation of wetlands in the modelling scheme, which is clearly stated in several literature (Golden et.al 2014, Wellen et.al 2015). Especially during drought periods, wetlands rely on their interaction and the water they will receive from the groundwater and/or surface water bodies. On top of this recommendation, here in Belgium, the amount of drainage water from agricultural lands is not well known, which is a concern of the government, and we were asked to assess the amount and the impact it will have on water availability. This is the reason why we put this recommendation on the inclusion of tile drains in hydrological simulations to capture the actual hydrological behavior of a given catchment. We believe that including wetlands and tile drains (ditches) in the modelling scheme of the Dijle catchment can improve the hydrological simulation even further (on top of the improvement we made). To summarize, the points we raised in line 392 and further are recommendations.
-
RC2: 'Comment on hess-2022-169', Anonymous Referee #2, 05 Aug 2022
The interactions between surface water-soil-groundwater is critical for the terrestrial hydrological processes especially over humid regions that have low water table depth. Estifanos et al. presented a modeling work over a groundwater-influenced catchment named Dijle by using SWAT and SWAT+gwflow model, and showed improvements of streamflow modeling after adding the gwflow. Although the topic is important, the current manuscript is more a technical report than a scientific research. Detailed comments are below.
- The authors declare that the primary novel side of this research article is representations of groundwater-soil interactions in the introduction. However, I did not see any details on the parameterization of groundwater-soil interactions in the model description. Moreover, detailed analysis on the influence of this process on the modeling result.
- Again, the current analysis simply compared two modeling results in modeling streamflow and simply showed the modeled groundwater balance. A scientific analysis on why the SWAT+gwflow model shows improvement in modeling streamflow (e.g., you can analyze the runoff components, the difference between two modeling results in water table depth, recharge) is needed.
- In addition, the current work also include the pumping in the SWAT+gwflow model. However, how much influence does the pumping process have? The relative importance between pumping and groundwater-soil interactions? Which factors contribute to the improved streamflow modeling (e.g., the original groundwater module, the groundwater-soil interactions, and the pumping) are not analyzed.
- The authors also emphasize that, compared with the precious study that used MODFLOW and SWAT, the current SWAT+gwflow model does not require any additional hydrological model. However, as the precious work obtained “nearly similar results”, I think the current point is relatively weak and is technical. Instead, it is more scientific to analyze the influence of direct coupling of groundwater module and SWAT on hydrological processes. For example, whether the SWAT model presents improved simulation of soil moisture, evapotranspiration after coupling the groundwater module and considering the groundwater-soil water interactions.
- The manuscript also needs improvement. For example,
- I did not find any introductions to the streamflow data (including its record length, location) in the “Study area and data”.
- Very detailed information is given in the study catchment in section 2, including the tributaries. However, the locations of tributaries are not given in Figure 1, which makes the reader confused.
- What does the different colors in Figure 2 mean?
- What does the “uncalibrated zone”mean in Figure 3? Why it is not calibrated?
- Whose simulation results is shown in Figure? SWAT or SWAT+gwflow? Moreover, both SWAT and SWAT+gwflow should been shown in Figure 5 to help us better understand the improvement in SWAT+gwflow.
- The unit of PBIAS should be given in Table 1.
- I can not see the yellow, light blue and blue regions in the right panel of Figure 7. And what does the white area in the right panel of Figure 7 mean is not declared.
- L385 “MDOFLOW”should be “MODEFLOW”
Citation: https://doi.org/10.5194/hess-2022-169-RC2 -
AC2: 'Reply on RC2', Estifanos Addisu Yimer, 30 Aug 2022
First of all, we would like to express our sincere appreciation to Reviewer 2 for the positive overall assessment of our manuscript and for the critical questions that have led to an improved and clearer manuscript.
Please find below the detailed answers to the question:
- The authors declare that the primary novel side of this research article is representations of groundwater-soil interactions in the introduction. However, I did not see any details on the parameterization of groundwater-soil interactions in the model description. Moreover, detailed analysis on the influence of this process on the modeling result.
In the original water balance equation of the gwflow module (Bailey et al., 2020), the interaction between SWAT+ HRU soil profiles and gwflow grid cells occurs only through the passing of deep percolation from the soil profile to groundwater, i.e. the water table elevation, and corresponding groundwater storage, simulated for each grid cell does not interact with nor affect the hydrological processes of the soil profile. This could lead to situations where the gwflow module simulates a water table within the soil profile, with all inherent groundwater calculations such as lateral flow, while at the same time, the HRU in the same vicinity as the grid cell is simulating percolation and soil lateral flow. This leads to duplicate processes, but by different equations. To remedy this flaw, in this study, we modify the gwflow module to allow the transfer of groundwater to the soil profile for grid cells where the water table rises into the soil profile. This is particularly important for groundwater-dominated catchments (i.e., high baseflow fractions) where the water table is shallow. For each cell, for each time step, the simulated water table elevation (i.e., groundwater head) is compared to the elevation of the soil profile bottom of the HRU to which it is connected spatially. If the water table elevation wtelev (m) is above the elevation of the soil profile bottom soilelev (m), then the volume of groundwater to transfer to the soil profile Qgw-->soil is calculated as:
Qgw-->soil = (wtelev - soilelev)*Ahru*Sy
where Ahru is the area (m2) of the HRU within the spatial area of the grid cell, and Sy is specific yield of the aquifer. This volume of water is added to the HRU soil layers that are at or below the elevation of the water table. Including this transfer term can yield higher rates of recharge during a model simulation due to added water to the soil profile that can subsequently move downward to the water table. Therefore, when presenting and discussing model results or estimating recharge rates for a watershed or portion of a watershed, the “net” recharge should be estimated by subtracting the transfer volume from the simulated recharge volume, i.e., true recharge = Qrech – Qgw-->soil.
As for the result before and after accounting gwsoil interaction can be found in Table 1 inside the revised manuscript (also can be found below). It can be seen that when the groundwater – soil interaction is neglected, there is a lot of saturation excess flow (164.2 mm/yr), because the groundwater is allowed to rise to the ground surface and then discharge directly to nearby streams. When gwsoil is accounted for, then the groundwater is transferred to the soil, leading to higher ET, surface runoff, and lateral flow. Also, there is still recharge (net = 837 - 733 = ~104 mm/yr). This implies the high groundwater will be routed to the streams via soil lateral flow instead of direct excess flow that probably leads to a significant amount of water reaching the river.
Table 1. The water balance components (in mm) with and without applying groundwater – soil interaction.
Inputs to Watershed
Without gwsoil
With gwsoil
Precipitation
838.9
838.9
Boundary Inflow
-5.5
-26.5
Lake seepage to groundwater
0.4
0.3
Outputs from Watershed
Surface ET
544.4
559
Surface runoff
66.4
83.6
Soil lateral flow
10.6
43.2
Stream seepage to groundwater
-4.5
-4.4
Saturation excess flow
164.2
0
Groundwater ET
0.2
0
Internal Flow
Recharge to the water table
204.8
837.2
Groundwater transfer to soil
0
733.9
On top of this, the streamflow simulation deteriorated when gwsoil interaction is not considered (NSE of 0.6 when accounting gwsoil interaction and NSE of 0.1 when it's neglected). All changes we made (the comparison between before and after applying the gwsoil interaction) can be found inside the revised document. - Again, the current analysis simply compared two modeling results in modeling streamflow and simply showed the modeled groundwater balance. A scientific analysis on why the SWAT+gwflow model shows improvement in modeling streamflow (e.g., you can analyze the runoff components, the difference between two modeling results in water table depth, recharge) is needed.
The standalone SWAT+ model approach regarding groundwater hydrology representation has a shortcoming. It has lumped conceptual representation, which has only been changed after coupling techniques came into practice. This lumped representation limited us from having output such as the groundwater table depth or recharge. Due to this, we are not able to compare the results from SWAT+ and SWAT+gwflow based on recharge and water table depth. However, a detailed analysis of before and after accounting gwsoil interaction is now included inside the revised document and the comparison is also discussed below (on top of our answer for your first question).
Daily hydrologic fluxes of watershed inputs and outputs (Figure 1) show seasonal patterns of ET, runoff, and soil lateral flow. The components of water yield (surface runoff, soil lateral flow, saturation excess flow) are shown in Figure 7A, with saturation excess flow being zero, due to the inclusion of the gwsoil term, and hence the groundwater head as simulated by the gwflow module not allowed to rise to the ground surface and induce saturation excess flow, but instead groundwater is incorporated into the soil profile and managed by the HRU soil processes of runoff and soil lateral flow. For this condition, therefore, soil lateral flow acts as a pathway for groundwater to reach the stream network, and hence can be considered as baseflow. This allows an accurate depiction of baseflow, as shown by the streamflow comparison in Figure 2A. The inclusion of the gwsoil mechanisms results in net recharge (Figure 2C) that, when averaged over the entire watershed, is usually positive, but can be negative as groundwater is transferred to the soil profile but no recharge occurs.
Figure 1. Daily hydrologic fluxes (mm/day) for watershed inputs (precipitation, boundary inflow; positive value) and watershed outputs (negative values), for the period 1993-1996.
The average annual hydrologic fluxes (mm/yr) within the watershed system are listed in Table 1 above. For the simulation period, on average ET accounts for 67% of precipitation, and surface runoff, soil lateral flow, and net recharge account for 10%, 5%, and 12%, respectively. When neglecting the gwsoil mechanism (i.e., turning off this feature in the SWAT+ code), hydrologic flux rates change dramatically (Table 1). This condition yields a very high rate of saturation excess flow (164 mm), as groundwater head is allowed to rise to the ground surface, i.e. there is no interaction with the soil profiles of the HRUs. This high rate of saturation excess flow, when combined with surface runoff (66.4 mm) and soil lateral flow (10.6 mm) and accounting for stream seepage to groundwater (4.5 mm), results in a total water yield of 237 mm, which greatly overestimates streamflow at the gaging site. Net recharge (Figure 2D) is also much higher in this condition (204 mm), as groundwater is not removed from the aquifer and placed in the soil profile. The high recharge rates result in high groundwater head, which reaches the ground surface in many places of the watershed, resulting in saturation excess flow.
Figure 2. Water yield and net recharge for the condition of including the gwsoil mechanism (A and C) and not including the gwsoil mechanism (B and D).
- In addition, the current work also include the pumping in the SWAT+gwflow model. However, how much influence does the pumping process have? The relative importance between pumping and groundwater-soil interactions? Which factors contribute to the improved streamflow modeling (e.g., the original groundwater module, the groundwater-soil interactions, and the pumping) are not analyzed.
It is correct that we have included pumping inside the gwflow module, but the issue here is that we have not included pumping in the modelling scheme of the Dijle catchment due to data limitations, and we intend to include it in another study area to show the impact of pumping in the general hydrology. Nevertheless, we have tried it with a dummy number, and it worked well. The pumping term is included in equation 3 (Q pump), where the groundwater change in volume with time is solved by including this term. Moreover, recently, Bailey et.al 2022 published a paper by including pumping while modelling (using SWAT+gwflow) a catchment located in the USA to investigate the impact of tile drain in the hydrology of the catchment. However, without the inclusion of pumping, the comparison made before and after including gwsoil interaction in our modelling scheme showed the importance of accounting such interactions (without gwsoil interaction, the streamflow simulation resulted in unsatisfactory NSE (0.1) while the inclusion resulted in NSE of 0.6). - The authors also emphasize that, compared with the precious study that used MODFLOW and SWAT, the current SWAT+gwflow model does not require any additional hydrological model. However, as the precious work obtained “nearly similar results”, I think the current point is relatively weak and is technical. Instead, it is more scientific to analyze the influence of direct coupling of groundwater module and SWAT on hydrological processes. For example, whether the SWAT model presents improved simulation of soil moisture, evapotranspiration after coupling the groundwater module and considering the groundwater-soil water interactions.
Thank you for suggesting further research points. The point we were inferring to was that the effort and time needed to build coupled surface water model and MODFLOW model is much higher than that of SWAT+gwflow model development. This is attributed to the need for building surface water model (e.g WetSpaSS) to get recharge that forces MODFLOW. Contrarily, the SWAT+gwflow model does not require recharge from another hydrological model. Moreover, since we don’t have the model result of MODFLOW model (our comparison was based on the result presented in their paper), we can’t compare the soil moisture, ET, and groundwater-soil interactions. Furthermore, as described in lines 301 to 304, their model setup is not SWAT+MODFLOW; instead, its MODFLOW model forced with recharge from WetSpaSS model.
I can assure you that this will be an essential part of a research we are currently working on where we are trying to compare SWAT+gwflow model setup versus SWAT+MODFLOW model setup for a catchment in Ethiopia. Hence, we would like to thank you for pointing out these fundamental research approaches.- The manuscript also needs improvement. For example,
- I did not find any introductions to the streamflow data (including its record length, location) in the “Study area and data”.
It is inside the methodology part, which is indicated in lines 247 to 253. Thank you for the note, and we have added the information about the streamflow data in the “Study area and data” section in line 127.
- Very detailed information is given in the study catchment in section 2, including the tributaries. However, the locations of tributaries are not given in Figure 1, which makes the reader confused.
Since the plot contains already congested information, we are forced not to include the tributaries inside the plot.
- What does the different colors in Figure 2 mean?
They are different geological units (layers), however, our main interest is the unconfined part of the geological unit. Hence, we indicated the impermeable layer along with the “Brussels sand” aquifer above it. We have noted that it can be misleading, thus, we have included the explanation in the caption.
- What does the “uncalibrated zone”mean in Figure 3? Why it is not calibrated?
The uncalibrated zones are explained from lines 212 to 221. The number of zones identified from the global dataset is 24, nevertheless, if we took all zones and assigned one hydraulic conductivity and specific yield parameters, then it will be difficult to calibrate the model (over-fitting will also be an issue). Due to this, we abstain from including those minor zones during calibration.
- Whose simulation results is shown in Figure? SWAT or SWAT+gwflow? Moreover, both SWAT and SWAT+gwflow should been shown in Figure 5 to help us better understand the improvement in SWAT+gwflow.
The results shown are for SWAT+gwflow. The result from the standalone SWAT+ model gave unsatisfactory result (NSE less than 0.5), hence, we decided to show result from only the coupled model. I.e., if we plot them together since the standalone model is giving results that are worse than the mean streamflow value, it will distort the view, thus, we removed it from the plot.
- The unit of PBIAS should be given in Table 1.
Well noted, and we now have included the unit inside the table.
- I cannot see the yellow, light blue and blue regions in the right panel of Figure 7. And what does the white area in the right panel of Figure 7 mean is not declared.
The yellow, light blue and blue regions correspond to few pixels and if you zoom in, you will find them. Most values are below 70 mm (yellow). The white area stands for areas with no groundwater-soil interaction. The red parts are for low flux values but due to significance level, they were plotted as zero. Now, since this can be misleading, we have replaced it with a new plot.
- L385 “MDOFLOW”should be “MODEFLOW”
Thank you, and it has been modified accordingly.
Citation: https://doi.org/10.5194/hess-2022-169-AC2 - The authors declare that the primary novel side of this research article is representations of groundwater-soil interactions in the introduction. However, I did not see any details on the parameterization of groundwater-soil interactions in the model description. Moreover, detailed analysis on the influence of this process on the modeling result.
Status: closed
-
RC1: 'Comment on hess-2022-169', Anonymous Referee #1, 27 May 2022
The manuscript "Improved representation of groundwater-dominated catchment using SWAT+gwflow and modifications to the gwflow module" by Yimer et al. is about the development of a new groundwater module for SWAT+ to provide an opportunity to consider a more process based groundwater interaction with the soil and consequently with discharge. This study tries comparing the basic SWAT+ and the newly developed SWAT+gwflow. For this, both model versions are applied on the Dijle catchment, which is non for groundwater dominance.
In general, a more process-based representation of groundwater within hydrological models and additionally a more easy to handle and fast model application should be interesting to the hydrological community. The current manuscript has a string focus on the SWAT model and is mostly written in a form of a technical note. Especially the beginning of the introduction deals with technical details of SWAT, which might be not relevant for all hydrolgical modellers. Consequently it might be worth to rethink the form of presenting this study.
In the following I am going to summarise the most critical shortcomings of the mansucript:
L. 33 - 58: This seems to be a technical summary about SWAT. It is unclear why such detailed information of SWAT is provided at this position. What is the demand for further research regarding the general hydrological community?
L. 59: There are several efforts to improve groundwater representation in hydrological models. Since the authors are focused on SWAT, it would be scientificly sound to mention and discuss these efforts at least for current and previous SWAT versions and to work out the necessity to consider other approaches (e. g. the current study).
L. 60: This seems to be the motivation of this study?
L. 93: Limitations of SWAT were also shown in Luo et al. (2012), Pfannerstill et al. (2013), Nguyen and Dietrich (2018), Shao et al. (2019). Additionally, these studies provided solutions to improve the representation of groundwater within SWAT.
L. 97: Please provide a clear structure: what are the gaps? which problems need to be solved? how are the problems solved? The introduction ends up with a rather technical information, which is not appropriate to guide readers through this study. It would be more helpful to provide the aim of this study at the end of the introduction.
L. 151: The introduction ends up mentioning the integration of pumping. This part is not mentioned here?
L. 215: Please provide references.
L. 274: It is unclear which parameters and which ranges were used for the sensitivity analysis. This is very crucial to reproduce the results of this study and to check if this method is appropriate.
L. 285: The whole section about sensitivity is unclear due to missing information (selected parameters, parameter ranges).
L. 286: It would be better to provide a table with calibrated parameters.
L. 302: It is impossible to follow the author's opinion and conclusion about poor/good model performance and the comparison between the different model structures.
1. Please provide a table with calibrated parameter values for both model setups.
2. Please provide appropriate figures for simulated/observed discharge. Figue 5b does not show any meaningful information since data is very compressed. A flow duration curve would be helpful to show the deviation between simulated and observed discharge magnitude. Furthermore, it seems that several flow events are not adequately reproduced by timing and magnitude. This may be due to the fact that a weather station inside the catchment is missing! What about other performance measures? Poor model performance, especially for the basic SWAT, might be related to the selection of inappropriate performance measures or imbalanced weighting.
L. 355: The evaluation based on monthly discharge should play a minor role. Of course, groundwater processes are more time-delayed and could be checked at a monthly time scale. However, other crucial hydrological processes happen immediately on a precipitation event. For a consistent model behaviuor, all relevant processes need to be represented and of course they need to be evaluated. Consequently, it is necessary to focus to an appropriate extent on daily discharge.
L. 343: Are these model results realistic? Are there any options to evaluate this?
L. 356: Please provide discharge for the validation periods with an additional figure.
L. 377: It is impossible to reproduce this conclusion since necessary information was not provided (parameters, parameter ranges, appropriate figures for daily discharge).
L. 389: This conclusion cannot be confirmed due to missing information.
L. 392: This aspect is fully new and was not mentioned before.
Due to the mentioned shortcomings I see the necessity for a major revision.
Citation: https://doi.org/10.5194/hess-2022-169-RC1 -
AC1: 'Comment on hess-2022-169', Estifanos Addisu Yimer, 27 Jun 2022
First of all, we would like to express our sincere appreciation to Reviewer 1 for the positive overall assessment of our manuscript and for the critical questions that have led to an improved and clearer manuscript.
Please find below the detailed answers to the question:
- 33 - 58: This seems to be a technical summary about SWAT. It is unclear why such detailed information of SWAT is provided at this position. What is the demand for further research regarding the general hydrological community?
The gwflow module is designed to replace the groundwater component's conceptual representation in SWAT+, and we focused on improving this new module. To provide the reader with the reason behind, we went from the basics so that every detail can be understood. It is also with an intention to take this article as a reference on how the coupling of SWAT progressed in time and its limitations and advancements. As you have stated in your subsequent question (question number 2), our focus is to deal with SWAT.
The major outcome of our research entails an important implication to the hydrological community, where not accounting for groundwater-soil interaction can be an essential factor for unsatisfactory hydrological simulations in hydrological models. Hence, other hydrological models should consider such interactions to pave the road for effective ground-surface water coupling.
- 59: There are several efforts to improve groundwater representation in hydrological models. Since the authors are focused on SWAT, it would be scientifically sound to mention and discuss these efforts at least for current and previous SWAT versions and to work out the necessity to consider other approaches (e. g. the current study).
The current and previous SWAT versions have a similar approach when it comes to groundwater hydrology representation. They have lumped conceptual representation, which has only been changed after coupling techniques came into practice. However, since previous coupling techniques have significant limitations (as stated in lines 82-83), gwflow module is developed. Previous coupling techniques are discussed between lines 66 to 80.
- This seems to be the motivation of this study?
Yes, it is one of the main motivations behind this research where holistic approaches are needed to understand the overall geohydrology of a given catchment. In addition, a comparison of the standalone SWAT+ and SWAT+gwflow is made to show the limitation of the conceptual representation of groundwater hydrology by the standalone model. Moreover, soil-groundwater interactions, which are neglected in the previous version of gwflow are accounted which is the major contribution of this article.
- 93: Limitations of SWAT were also shown in Luo et al. (2012), Pfannerstill et al. (2013), Nguyen and Dietrich (2018), Shao et al. (2019). Additionally, these studies provided solutions to improve the representation of groundwater within SWAT.
Thank you for suggesting more references which we have not come across. The references you have suggested have not been implemented in the SWAT+ model so far, but in the future, we will take them as an input to modify the module or the standalone model even further. We would like to affirm that the main intention of this research is to modify the gwflow module and make a comparison with the standalone SWAT+ model. Hence, you may please take it as an extension of the gwflow module developed by Ryan T.Bailey (Bailey et.al 2020).
- 97: Please provide a clear structure: what are the gaps? which problems need to be solved? how are the problems solved? The introduction ends up with a rather technical information, which is not appropriate to guide readers through this study. It would be more helpful to provide the aim of this study at the end of the introduction.
We have modified the end of the introduction with the suggestions you have made. You may find them in the edited version of the manuscript from line 91 to 97.
- 151: The introduction ends up mentioning the integration of pumping. This part is not mentioned here?
That is correct, but the issue here is that we have not included pumping in the modelling scheme of the Dijle catchment due to data limitations, and we intend to include it in another paper to show the impact of pumping in the general hydrology. Nevertheless, we have tried it with a dummy number, and it worked well. The pumping term is included in equation 3 (Q pump), where the groundwater change in volume with time is solved by including this term. Recently, Bailey et.al 2021 published a paper by including pumping in a catchment located in the USA to investigate the impact of tile drain in the hydrology of the catchment.
- 215: Please provide references.
We have added the reference accordingly.
- 274, L.285 and L.286: It is unclear which parameters and which ranges were used for the sensitivity analysis. This is very crucial to reproduce the results of this study and to check if this method is appropriate.
We have included the parameters used (in Table format) for calibration and the sensitivity analysis results (in plot format) in the supplementary material. In addition, the calibrated parameter sets is also included in the supplementary material.
Note that we had a master’s student who tried to calibrate the SWAT+ model for the Dijle catchment for almost five months, and it was impossible to achieve good agreement between modelled and observed streamflow values, which is the primary motivation for this research. The best parameter set we achieved was based on several simulations made on that thesis (the student is also a co-author of this article - Lise Leda Piepers). The main conclusion of the thesis was that the standalone model could not represent the hydrology effectively due to limitations in the groundwater hydrology representation.
Moreover, to ensure transparency in our work, we are willing to share the model we developed for the Dijle catchment. It is also a catchment we used to prepare the tutorials for SWAT+gwflow model (previous version of the model) and recently we replaced it with the case study we have from Tuscany, Italy – Ombrone catchment. Link for tutorial: https://swat.tamu.edu/software/plus/gwflow/
- 302
- Please provide a table with calibrated parameter values for both model setups.
- You may check the answer we provided above (#8).
- Please provide appropriate figures for simulated/observed discharge. Figue 5b does not show any meaningful information since data is very compressed. A flow duration curve would be helpful to show the deviation between simulated and observed discharge magnitude.
- Thank you for your suggestion, and we have included a flow duration curve to show the deviation. As the flow duration curve has indicated, the differences are significant during peak flows, but those events are few, with flow rate below 30 m3/s in most events. The major player in this catchment is the groundwater hydrology, which is also indicated in the baseflow filtering discussed in question number 11 below.
- 355: The evaluation based on monthly discharge should play a minor role. Of course, groundwater processes are more time-delayed and could be checked at a monthly time scale. However, other crucial hydrological processes happen immediately on a precipitation event. For a consistent model behavior, all relevant processes need to be represented and of course they need to be evaluated. Consequently, it is necessary to focus to an appropriate extent on daily discharge.
It is correct that daily time step can be appropriate, and due to that, our assessment does not only rely on using monthly but also daily model outputs to investigate further. For instance, you may check Table 1, where objective functions are estimated for both daily and monthly time steps.
- 343: Are these model results realistic? Are there any options to evaluate this?
Yes they are realistic, and we can evaluate this by assessing the baseflow component of the streamflow using baseflow filtering techniques (e.g. WETSPRO - Willems 2009). This will allow us to understand how much of the total flow is accounted for by baseflow, interflow, and overland flow. We can infer from the filtering that baseflow accounts for the majority of the streamflow (around 80% of the total streamflow), indicating the prevalence of subsurface than surface processes. You may please check the WETSPRO result below.
Figure: The baseflow filtering parameters (top) where w is the difference between 1 and the percentage of baseflow. The bottom plot shows the filtered baseflow for the catchment outlet.
For a detailed description of the baseflow filtering methodology, you may refer to the link: https://bwk.kuleuven.be/hydr/pwtools.htm#Wetspro. We have also included in the .zip file the filtering excel workbook.
- 356: Please provide discharge for the validation periods with an additional figure.
Thank you for your suggestion, we have included plots for monthly and flow duration curve for daily timestep in the supplementary material for the catchment outlet and additional gauging station inside the catchment.
- 392: This aspect is fully new and was not mentioned before.
One limitation of the SWAT hydrological model is the nonrepresentation of wetlands in the modelling scheme, which is clearly stated in several literature (Golden et.al 2014, Wellen et.al 2015). Especially during drought periods, wetlands rely on their interaction and the water they will receive from the groundwater and/or surface water bodies. On top of this recommendation, here in Belgium, the amount of drainage water from agricultural lands is not well known, which is a concern of the government, and we were asked to assess the amount and the impact it will have on water availability. This is the reason why we put this recommendation on the inclusion of tile drains in hydrological simulations to capture the actual hydrological behavior of a given catchment. We believe that including wetlands and tile drains (ditches) in the modelling scheme of the Dijle catchment can improve the hydrological simulation even further (on top of the improvement we made). To summarize, the points we raised in line 392 and further are recommendations.
-
AC1: 'Comment on hess-2022-169', Estifanos Addisu Yimer, 27 Jun 2022
-
AC1: 'Comment on hess-2022-169', Estifanos Addisu Yimer, 27 Jun 2022
First of all, we would like to express our sincere appreciation to Reviewer 1 for the positive overall assessment of our manuscript and for the critical questions that have led to an improved and clearer manuscript.
Please find below the detailed answers to the question:
- 33 - 58: This seems to be a technical summary about SWAT. It is unclear why such detailed information of SWAT is provided at this position. What is the demand for further research regarding the general hydrological community?
The gwflow module is designed to replace the groundwater component's conceptual representation in SWAT+, and we focused on improving this new module. To provide the reader with the reason behind, we went from the basics so that every detail can be understood. It is also with an intention to take this article as a reference on how the coupling of SWAT progressed in time and its limitations and advancements. As you have stated in your subsequent question (question number 2), our focus is to deal with SWAT.
The major outcome of our research entails an important implication to the hydrological community, where not accounting for groundwater-soil interaction can be an essential factor for unsatisfactory hydrological simulations in hydrological models. Hence, other hydrological models should consider such interactions to pave the road for effective ground-surface water coupling.
- 59: There are several efforts to improve groundwater representation in hydrological models. Since the authors are focused on SWAT, it would be scientifically sound to mention and discuss these efforts at least for current and previous SWAT versions and to work out the necessity to consider other approaches (e. g. the current study).
The current and previous SWAT versions have a similar approach when it comes to groundwater hydrology representation. They have lumped conceptual representation, which has only been changed after coupling techniques came into practice. However, since previous coupling techniques have significant limitations (as stated in lines 82-83), gwflow module is developed. Previous coupling techniques are discussed between lines 66 to 80.
- This seems to be the motivation of this study?
Yes, it is one of the main motivations behind this research where holistic approaches are needed to understand the overall geohydrology of a given catchment. In addition, a comparison of the standalone SWAT+ and SWAT+gwflow is made to show the limitation of the conceptual representation of groundwater hydrology by the standalone model. Moreover, soil-groundwater interactions, which are neglected in the previous version of gwflow are accounted which is the major contribution of this article.
- 93: Limitations of SWAT were also shown in Luo et al. (2012), Pfannerstill et al. (2013), Nguyen and Dietrich (2018), Shao et al. (2019). Additionally, these studies provided solutions to improve the representation of groundwater within SWAT.
Thank you for suggesting more references which we have not come across. The references you have suggested have not been implemented in the SWAT+ model so far, but in the future, we will take them as an input to modify the module or the standalone model even further. We would like to affirm that the main intention of this research is to modify the gwflow module and make a comparison with the standalone SWAT+ model. Hence, you may please take it as an extension of the gwflow module developed by Ryan T.Bailey (Bailey et.al 2020).
- 97: Please provide a clear structure: what are the gaps? which problems need to be solved? how are the problems solved? The introduction ends up with a rather technical information, which is not appropriate to guide readers through this study. It would be more helpful to provide the aim of this study at the end of the introduction.
We have modified the end of the introduction with the suggestions you have made. You may find them in the edited version of the manuscript from line 91 to 97.
- 151: The introduction ends up mentioning the integration of pumping. This part is not mentioned here?
That is correct, but the issue here is that we have not included pumping in the modelling scheme of the Dijle catchment due to data limitations, and we intend to include it in another paper to show the impact of pumping in the general hydrology. Nevertheless, we have tried it with a dummy number, and it worked well. The pumping term is included in equation 3 (Q pump), where the groundwater change in volume with time is solved by including this term. Recently, Bailey et.al 2021 published a paper by including pumping in a catchment located in the USA to investigate the impact of tile drain in the hydrology of the catchment.
- 215: Please provide references.
We have added the reference accordingly.
- 274, L.285 and L.286: It is unclear which parameters and which ranges were used for the sensitivity analysis. This is very crucial to reproduce the results of this study and to check if this method is appropriate.
We have included the parameters used (in Table format) for calibration and the sensitivity analysis results (in plot format) in the supplementary material. In addition, the calibrated parameter sets is also included in the supplementary material.
Note that we had a master’s student who tried to calibrate the SWAT+ model for the Dijle catchment for almost five months, and it was impossible to achieve good agreement between modelled and observed streamflow values, which is the primary motivation for this research. The best parameter set we achieved was based on several simulations made on that thesis (the student is also a co-author of this article - Lise Leda Piepers). The main conclusion of the thesis was that the standalone model could not represent the hydrology effectively due to limitations in the groundwater hydrology representation.
Moreover, to ensure transparency in our work, we are willing to share the model we developed for the Dijle catchment. It is also a catchment we used to prepare the tutorials for SWAT+gwflow model (previous version of the model) and recently we replaced it with the case study we have from Tuscany, Italy – Ombrone catchment. Link for tutorial: https://swat.tamu.edu/software/plus/gwflow/
- 302
- Please provide a table with calibrated parameter values for both model setups.
- You may check the answer we provided above (#8).
- Please provide appropriate figures for simulated/observed discharge. Figue 5b does not show any meaningful information since data is very compressed. A flow duration curve would be helpful to show the deviation between simulated and observed discharge magnitude.
- Thank you for your suggestion, and we have included a flow duration curve to show the deviation. As the flow duration curve has indicated, the differences are significant during peak flows, but those events are few, with flow rate below 30 m3/s in most events. The major player in this catchment is the groundwater hydrology, which is also indicated in the baseflow filtering discussed in question number 11 below.
- 355: The evaluation based on monthly discharge should play a minor role. Of course, groundwater processes are more time-delayed and could be checked at a monthly time scale. However, other crucial hydrological processes happen immediately on a precipitation event. For a consistent model behavior, all relevant processes need to be represented and of course they need to be evaluated. Consequently, it is necessary to focus to an appropriate extent on daily discharge.
It is correct that daily time step can be appropriate, and due to that, our assessment does not only rely on using monthly but also daily model outputs to investigate further. For instance, you may check Table 1, where objective functions are estimated for both daily and monthly time steps.
- 343: Are these model results realistic? Are there any options to evaluate this?
Yes they are realistic, and we can evaluate this by assessing the baseflow component of the streamflow using baseflow filtering techniques (e.g. WETSPRO - Willems 2009). This will allow us to understand how much of the total flow is accounted for by baseflow, interflow, and overland flow. We can infer from the filtering that baseflow accounts for the majority of the streamflow (around 80% of the total streamflow), indicating the prevalence of subsurface than surface processes. You may please check the WETSPRO result below.
Figure: The baseflow filtering parameters (top) where w is the difference between 1 and the percentage of baseflow. The bottom plot shows the filtered baseflow for the catchment outlet.
For a detailed description of the baseflow filtering methodology, you may refer to the link: https://bwk.kuleuven.be/hydr/pwtools.htm#Wetspro. We have also included in the .zip file the filtering excel workbook.
- 356: Please provide discharge for the validation periods with an additional figure.
Thank you for your suggestion, we have included plots for monthly and flow duration curve for daily timestep in the supplementary material for the catchment outlet and additional gauging station inside the catchment.
- 392: This aspect is fully new and was not mentioned before.
One limitation of the SWAT hydrological model is the nonrepresentation of wetlands in the modelling scheme, which is clearly stated in several literature (Golden et.al 2014, Wellen et.al 2015). Especially during drought periods, wetlands rely on their interaction and the water they will receive from the groundwater and/or surface water bodies. On top of this recommendation, here in Belgium, the amount of drainage water from agricultural lands is not well known, which is a concern of the government, and we were asked to assess the amount and the impact it will have on water availability. This is the reason why we put this recommendation on the inclusion of tile drains in hydrological simulations to capture the actual hydrological behavior of a given catchment. We believe that including wetlands and tile drains (ditches) in the modelling scheme of the Dijle catchment can improve the hydrological simulation even further (on top of the improvement we made). To summarize, the points we raised in line 392 and further are recommendations.
-
RC2: 'Comment on hess-2022-169', Anonymous Referee #2, 05 Aug 2022
The interactions between surface water-soil-groundwater is critical for the terrestrial hydrological processes especially over humid regions that have low water table depth. Estifanos et al. presented a modeling work over a groundwater-influenced catchment named Dijle by using SWAT and SWAT+gwflow model, and showed improvements of streamflow modeling after adding the gwflow. Although the topic is important, the current manuscript is more a technical report than a scientific research. Detailed comments are below.
- The authors declare that the primary novel side of this research article is representations of groundwater-soil interactions in the introduction. However, I did not see any details on the parameterization of groundwater-soil interactions in the model description. Moreover, detailed analysis on the influence of this process on the modeling result.
- Again, the current analysis simply compared two modeling results in modeling streamflow and simply showed the modeled groundwater balance. A scientific analysis on why the SWAT+gwflow model shows improvement in modeling streamflow (e.g., you can analyze the runoff components, the difference between two modeling results in water table depth, recharge) is needed.
- In addition, the current work also include the pumping in the SWAT+gwflow model. However, how much influence does the pumping process have? The relative importance between pumping and groundwater-soil interactions? Which factors contribute to the improved streamflow modeling (e.g., the original groundwater module, the groundwater-soil interactions, and the pumping) are not analyzed.
- The authors also emphasize that, compared with the precious study that used MODFLOW and SWAT, the current SWAT+gwflow model does not require any additional hydrological model. However, as the precious work obtained “nearly similar results”, I think the current point is relatively weak and is technical. Instead, it is more scientific to analyze the influence of direct coupling of groundwater module and SWAT on hydrological processes. For example, whether the SWAT model presents improved simulation of soil moisture, evapotranspiration after coupling the groundwater module and considering the groundwater-soil water interactions.
- The manuscript also needs improvement. For example,
- I did not find any introductions to the streamflow data (including its record length, location) in the “Study area and data”.
- Very detailed information is given in the study catchment in section 2, including the tributaries. However, the locations of tributaries are not given in Figure 1, which makes the reader confused.
- What does the different colors in Figure 2 mean?
- What does the “uncalibrated zone”mean in Figure 3? Why it is not calibrated?
- Whose simulation results is shown in Figure? SWAT or SWAT+gwflow? Moreover, both SWAT and SWAT+gwflow should been shown in Figure 5 to help us better understand the improvement in SWAT+gwflow.
- The unit of PBIAS should be given in Table 1.
- I can not see the yellow, light blue and blue regions in the right panel of Figure 7. And what does the white area in the right panel of Figure 7 mean is not declared.
- L385 “MDOFLOW”should be “MODEFLOW”
Citation: https://doi.org/10.5194/hess-2022-169-RC2 -
AC2: 'Reply on RC2', Estifanos Addisu Yimer, 30 Aug 2022
First of all, we would like to express our sincere appreciation to Reviewer 2 for the positive overall assessment of our manuscript and for the critical questions that have led to an improved and clearer manuscript.
Please find below the detailed answers to the question:
- The authors declare that the primary novel side of this research article is representations of groundwater-soil interactions in the introduction. However, I did not see any details on the parameterization of groundwater-soil interactions in the model description. Moreover, detailed analysis on the influence of this process on the modeling result.
In the original water balance equation of the gwflow module (Bailey et al., 2020), the interaction between SWAT+ HRU soil profiles and gwflow grid cells occurs only through the passing of deep percolation from the soil profile to groundwater, i.e. the water table elevation, and corresponding groundwater storage, simulated for each grid cell does not interact with nor affect the hydrological processes of the soil profile. This could lead to situations where the gwflow module simulates a water table within the soil profile, with all inherent groundwater calculations such as lateral flow, while at the same time, the HRU in the same vicinity as the grid cell is simulating percolation and soil lateral flow. This leads to duplicate processes, but by different equations. To remedy this flaw, in this study, we modify the gwflow module to allow the transfer of groundwater to the soil profile for grid cells where the water table rises into the soil profile. This is particularly important for groundwater-dominated catchments (i.e., high baseflow fractions) where the water table is shallow. For each cell, for each time step, the simulated water table elevation (i.e., groundwater head) is compared to the elevation of the soil profile bottom of the HRU to which it is connected spatially. If the water table elevation wtelev (m) is above the elevation of the soil profile bottom soilelev (m), then the volume of groundwater to transfer to the soil profile Qgw-->soil is calculated as:
Qgw-->soil = (wtelev - soilelev)*Ahru*Sy
where Ahru is the area (m2) of the HRU within the spatial area of the grid cell, and Sy is specific yield of the aquifer. This volume of water is added to the HRU soil layers that are at or below the elevation of the water table. Including this transfer term can yield higher rates of recharge during a model simulation due to added water to the soil profile that can subsequently move downward to the water table. Therefore, when presenting and discussing model results or estimating recharge rates for a watershed or portion of a watershed, the “net” recharge should be estimated by subtracting the transfer volume from the simulated recharge volume, i.e., true recharge = Qrech – Qgw-->soil.
As for the result before and after accounting gwsoil interaction can be found in Table 1 inside the revised manuscript (also can be found below). It can be seen that when the groundwater – soil interaction is neglected, there is a lot of saturation excess flow (164.2 mm/yr), because the groundwater is allowed to rise to the ground surface and then discharge directly to nearby streams. When gwsoil is accounted for, then the groundwater is transferred to the soil, leading to higher ET, surface runoff, and lateral flow. Also, there is still recharge (net = 837 - 733 = ~104 mm/yr). This implies the high groundwater will be routed to the streams via soil lateral flow instead of direct excess flow that probably leads to a significant amount of water reaching the river.
Table 1. The water balance components (in mm) with and without applying groundwater – soil interaction.
Inputs to Watershed
Without gwsoil
With gwsoil
Precipitation
838.9
838.9
Boundary Inflow
-5.5
-26.5
Lake seepage to groundwater
0.4
0.3
Outputs from Watershed
Surface ET
544.4
559
Surface runoff
66.4
83.6
Soil lateral flow
10.6
43.2
Stream seepage to groundwater
-4.5
-4.4
Saturation excess flow
164.2
0
Groundwater ET
0.2
0
Internal Flow
Recharge to the water table
204.8
837.2
Groundwater transfer to soil
0
733.9
On top of this, the streamflow simulation deteriorated when gwsoil interaction is not considered (NSE of 0.6 when accounting gwsoil interaction and NSE of 0.1 when it's neglected). All changes we made (the comparison between before and after applying the gwsoil interaction) can be found inside the revised document. - Again, the current analysis simply compared two modeling results in modeling streamflow and simply showed the modeled groundwater balance. A scientific analysis on why the SWAT+gwflow model shows improvement in modeling streamflow (e.g., you can analyze the runoff components, the difference between two modeling results in water table depth, recharge) is needed.
The standalone SWAT+ model approach regarding groundwater hydrology representation has a shortcoming. It has lumped conceptual representation, which has only been changed after coupling techniques came into practice. This lumped representation limited us from having output such as the groundwater table depth or recharge. Due to this, we are not able to compare the results from SWAT+ and SWAT+gwflow based on recharge and water table depth. However, a detailed analysis of before and after accounting gwsoil interaction is now included inside the revised document and the comparison is also discussed below (on top of our answer for your first question).
Daily hydrologic fluxes of watershed inputs and outputs (Figure 1) show seasonal patterns of ET, runoff, and soil lateral flow. The components of water yield (surface runoff, soil lateral flow, saturation excess flow) are shown in Figure 7A, with saturation excess flow being zero, due to the inclusion of the gwsoil term, and hence the groundwater head as simulated by the gwflow module not allowed to rise to the ground surface and induce saturation excess flow, but instead groundwater is incorporated into the soil profile and managed by the HRU soil processes of runoff and soil lateral flow. For this condition, therefore, soil lateral flow acts as a pathway for groundwater to reach the stream network, and hence can be considered as baseflow. This allows an accurate depiction of baseflow, as shown by the streamflow comparison in Figure 2A. The inclusion of the gwsoil mechanisms results in net recharge (Figure 2C) that, when averaged over the entire watershed, is usually positive, but can be negative as groundwater is transferred to the soil profile but no recharge occurs.
Figure 1. Daily hydrologic fluxes (mm/day) for watershed inputs (precipitation, boundary inflow; positive value) and watershed outputs (negative values), for the period 1993-1996.
The average annual hydrologic fluxes (mm/yr) within the watershed system are listed in Table 1 above. For the simulation period, on average ET accounts for 67% of precipitation, and surface runoff, soil lateral flow, and net recharge account for 10%, 5%, and 12%, respectively. When neglecting the gwsoil mechanism (i.e., turning off this feature in the SWAT+ code), hydrologic flux rates change dramatically (Table 1). This condition yields a very high rate of saturation excess flow (164 mm), as groundwater head is allowed to rise to the ground surface, i.e. there is no interaction with the soil profiles of the HRUs. This high rate of saturation excess flow, when combined with surface runoff (66.4 mm) and soil lateral flow (10.6 mm) and accounting for stream seepage to groundwater (4.5 mm), results in a total water yield of 237 mm, which greatly overestimates streamflow at the gaging site. Net recharge (Figure 2D) is also much higher in this condition (204 mm), as groundwater is not removed from the aquifer and placed in the soil profile. The high recharge rates result in high groundwater head, which reaches the ground surface in many places of the watershed, resulting in saturation excess flow.
Figure 2. Water yield and net recharge for the condition of including the gwsoil mechanism (A and C) and not including the gwsoil mechanism (B and D).
- In addition, the current work also include the pumping in the SWAT+gwflow model. However, how much influence does the pumping process have? The relative importance between pumping and groundwater-soil interactions? Which factors contribute to the improved streamflow modeling (e.g., the original groundwater module, the groundwater-soil interactions, and the pumping) are not analyzed.
It is correct that we have included pumping inside the gwflow module, but the issue here is that we have not included pumping in the modelling scheme of the Dijle catchment due to data limitations, and we intend to include it in another study area to show the impact of pumping in the general hydrology. Nevertheless, we have tried it with a dummy number, and it worked well. The pumping term is included in equation 3 (Q pump), where the groundwater change in volume with time is solved by including this term. Moreover, recently, Bailey et.al 2022 published a paper by including pumping while modelling (using SWAT+gwflow) a catchment located in the USA to investigate the impact of tile drain in the hydrology of the catchment. However, without the inclusion of pumping, the comparison made before and after including gwsoil interaction in our modelling scheme showed the importance of accounting such interactions (without gwsoil interaction, the streamflow simulation resulted in unsatisfactory NSE (0.1) while the inclusion resulted in NSE of 0.6). - The authors also emphasize that, compared with the precious study that used MODFLOW and SWAT, the current SWAT+gwflow model does not require any additional hydrological model. However, as the precious work obtained “nearly similar results”, I think the current point is relatively weak and is technical. Instead, it is more scientific to analyze the influence of direct coupling of groundwater module and SWAT on hydrological processes. For example, whether the SWAT model presents improved simulation of soil moisture, evapotranspiration after coupling the groundwater module and considering the groundwater-soil water interactions.
Thank you for suggesting further research points. The point we were inferring to was that the effort and time needed to build coupled surface water model and MODFLOW model is much higher than that of SWAT+gwflow model development. This is attributed to the need for building surface water model (e.g WetSpaSS) to get recharge that forces MODFLOW. Contrarily, the SWAT+gwflow model does not require recharge from another hydrological model. Moreover, since we don’t have the model result of MODFLOW model (our comparison was based on the result presented in their paper), we can’t compare the soil moisture, ET, and groundwater-soil interactions. Furthermore, as described in lines 301 to 304, their model setup is not SWAT+MODFLOW; instead, its MODFLOW model forced with recharge from WetSpaSS model.
I can assure you that this will be an essential part of a research we are currently working on where we are trying to compare SWAT+gwflow model setup versus SWAT+MODFLOW model setup for a catchment in Ethiopia. Hence, we would like to thank you for pointing out these fundamental research approaches.- The manuscript also needs improvement. For example,
- I did not find any introductions to the streamflow data (including its record length, location) in the “Study area and data”.
It is inside the methodology part, which is indicated in lines 247 to 253. Thank you for the note, and we have added the information about the streamflow data in the “Study area and data” section in line 127.
- Very detailed information is given in the study catchment in section 2, including the tributaries. However, the locations of tributaries are not given in Figure 1, which makes the reader confused.
Since the plot contains already congested information, we are forced not to include the tributaries inside the plot.
- What does the different colors in Figure 2 mean?
They are different geological units (layers), however, our main interest is the unconfined part of the geological unit. Hence, we indicated the impermeable layer along with the “Brussels sand” aquifer above it. We have noted that it can be misleading, thus, we have included the explanation in the caption.
- What does the “uncalibrated zone”mean in Figure 3? Why it is not calibrated?
The uncalibrated zones are explained from lines 212 to 221. The number of zones identified from the global dataset is 24, nevertheless, if we took all zones and assigned one hydraulic conductivity and specific yield parameters, then it will be difficult to calibrate the model (over-fitting will also be an issue). Due to this, we abstain from including those minor zones during calibration.
- Whose simulation results is shown in Figure? SWAT or SWAT+gwflow? Moreover, both SWAT and SWAT+gwflow should been shown in Figure 5 to help us better understand the improvement in SWAT+gwflow.
The results shown are for SWAT+gwflow. The result from the standalone SWAT+ model gave unsatisfactory result (NSE less than 0.5), hence, we decided to show result from only the coupled model. I.e., if we plot them together since the standalone model is giving results that are worse than the mean streamflow value, it will distort the view, thus, we removed it from the plot.
- The unit of PBIAS should be given in Table 1.
Well noted, and we now have included the unit inside the table.
- I cannot see the yellow, light blue and blue regions in the right panel of Figure 7. And what does the white area in the right panel of Figure 7 mean is not declared.
The yellow, light blue and blue regions correspond to few pixels and if you zoom in, you will find them. Most values are below 70 mm (yellow). The white area stands for areas with no groundwater-soil interaction. The red parts are for low flux values but due to significance level, they were plotted as zero. Now, since this can be misleading, we have replaced it with a new plot.
- L385 “MDOFLOW”should be “MODEFLOW”
Thank you, and it has been modified accordingly.
Citation: https://doi.org/10.5194/hess-2022-169-AC2 - The authors declare that the primary novel side of this research article is representations of groundwater-soil interactions in the introduction. However, I did not see any details on the parameterization of groundwater-soil interactions in the model description. Moreover, detailed analysis on the influence of this process on the modeling result.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 915 | 540 | 45 | 1,500 | 42 | 36 |
- HTML: 915
- PDF: 540
- XML: 45
- Total: 1,500
- BibTeX: 42
- EndNote: 36
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
