Challenges in developing a global gradient-based groundwater model (G<sup>3</sup>M v1.0) for the integration into a global hydrological model

Reinecke, Robert; Foglia, Laura; Mehl, Steffen; Trautmann, Tim; Cáceres, Denise; Döll, Petra

doi:https://doi.org/10.5194/gmd-12-2401-2019

Articles | Volume 12, issue 6

https://doi.org/10.5194/gmd-12-2401-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-12-2401-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 12, issue 6

Model description paper

|

18 Jun 2019

Model description paper |

| 18 Jun 2019

Challenges in developing a global gradient-based groundwater model (G³M v1.0) for the integration into a global hydrological model

Robert Reinecke, Laura Foglia, Steffen Mehl, Tim Trautmann, Denise Cáceres, and Petra Döll

Download

Final revised paper (published on 18 Jun 2019)
Supplement to the final revised paper
Preprint (discussion started on 15 May 2018)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Reinecke et al', Anonymous Referee #1, 10 Jun 2018
- AC1: 'Response to RC 1 and 2', R. Reinecke, 24 Jul 2018
SC1: 'Code and Data availability', Lutz Gross, 22 Jun 2018
- AC2: 'Response to SC1', R. Reinecke, 24 Jul 2018
RC2: 'Review of “Beyond the bucket …” by Reinecke et al', Anonymous Referee #2, 29 Jun 2018
- AC1: 'Response to RC 1 and 2', R. Reinecke, 24 Jul 2018
AC1: 'Response to RC 1 and 2', R. Reinecke, 24 Jul 2018
- AC3: 'Fixed citations of markup manuscript', R. Reinecke, 24 Jul 2018

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Robert Reinecke on behalf of the Authors (06 Aug 2018) Author's response Manuscript

ED: Referee Nomination & Report Request started (23 Aug 2018) by Wolfgang Kurtz

ED: Reconsider after major revisions (02 Sep 2018) by Wolfgang Kurtz

Dear authors,

the two referees of your manuscript raised a couple of important concerns regarding the applied modelling strategy and the validation of the proposed global gradient-based groundwater model. Therefore, I ask for substantial revisions of the manuscript. Some important points to consider are given below:

- Both reviewers question the choice of a confined aquifer for the simulation of river-aquifer exchange. Evidence needs to be provided that the chosen assumption does not bias model results, e.g. by comparison with an unconfined representation of the upper model layer (as used in other available large-scale groundwater models), or by conducting a sensitivity study as suggested by referee 2.

- The verification of the model shows considerable deficiencies. There is a significant mismatch between modelled and observed water table depths which is mainly justified by scale effects. Referee 1 rightfully points out that these discrepancies undermine the credibility of the modelling approach, also with respect to future model extensions and asks for a regional-scale validation of the model (which also avoids the problems of comparison of hydraulic heads over large ranges). The provided comparison with the Central Valley model only provides a rough qualitative comparison whereas a more detailed analysis would be needed for validation. If scale is a general obstacle for a proper model validation, then a logical consequence would be to increase model resolution to an appropriate level.

- At the current stage, it is not clear what was learned from the modelling exercise and what are the advantages and differences compared to already existing model set-ups and software packages. As advertised in the manuscript, a main purpose of this study is to use a steady state groundwater model to gain first insights into the credibility of the model set-up for future coupled transient simulations. However, this is not streamlined at the moment, i.e. it is not clear what experience you gained from the steady state simulations to move forward to transient coupled simulations. The problems outlined by the referees regarding model set-up and model verification will prevail for transient coupled simulations and therefore need to be properly addressed at the current stage. It needs to be pointed out more clearly, how this study serves as a basis for the development of the desired coupled WaterGAP model.

- You should more clearly explain the advantages and differences of your modelling approach compared to already existing infrastructure as requested by both referees. For example it is still not clear why G3M would be technically more suitable for in-memory coupling than already existing groundwater models like Modflow, i.e. why is it easier to access and exchange model variables with G3M. As pointed out by referee 1, the global-scale model set-up itself could be implemented with already existing groundwater modelling software. Following referee 2 also a more in-depth analysis on the differences between the proposed model and the model of de Graaf et al. is needed.

- The manuscript quite often refers to implementation that are planned in the future (i.e. coupling to WaterGAP, transient simulations). While it is ok to outline these future plans in order to justify the proposed modelling approach for the current global steady-state model, the frequency of references to future work is often misleading (as also acknowledged by referee 2), because it is often not clear which kind of feature is really implemented and which one is intended to be implemented. Therefore, I would ask you to limit references to future work to the necessary minimum.

Let me know, if you need any further guidance for revising the manuscript.

Kind regards,
Wolfgang Kurtz

Hide

AR by Robert Reinecke on behalf of the Authors (08 Oct 2018) Author's response Manuscript

ED: Referee Nomination & Report Request started (22 Oct 2018) by Wolfgang Kurtz

RR by Anonymous Referee #3 (14 Nov 2018)

Suggestions for revision or reasons for rejection

Authors have formulated a gradient based groundwater modeling approach for implementation in global hydrologic models. This is an important area of research and the manuscript can potentially help addressing this important problem. However, authors have not fully addressed reviewers’ and editor’s comments, and the manuscript still has some deficiencies that need to be addressed. For example, response to comments 3 and 5 of the Editor is still not convincing after revising the manuscript.

I concur with the editor’s comment (3) that it is still not clear what is learned from this modeling exercise and comparison with existing methods are required to show the gradient based approach indeed is superior to the existing approaches.

Evaluation of the model performance against existing observations and model simulations such as CVHM and ParFlow models are qualitative. A more robust quantitative analysis using various statistical measures are required to show where the new approach works and where it does not.

Authors indicate lack of enough observations to validate their modeling approach. Perhaps, they can run the model for a higher resolution and evaluate the impact of coarse simulation on simulated hydrologic fluxes. This could provide valuable insights regarding model development and implementation for the coupling approach as well.

The main contribution of this paper is not entirely clear. Authors have highlighted a number of improvements that the use of a gradient-based groundwater model could have on simulating global hydrological processes but many of those improvements have not been made in this version of the manuscript including the coupling with the global hydrologic model. As pointed out by the reviewers, I suggest authors to focus this paper on the sensitivity analysis of model parameterization and conceptual formulations. This can certainly help with the development of the transient model as well as the coupling method.

Could you please further explain how lateral connectivity between neighboring groundwater cells are calculated?

Page 1 – L18: “as simulation of unsaturated flow and SW body elevation” is not clear. Please describe.

Page 1 - L31-34 : It is not clear.

Page 2- L5: replace recharge from soil with “precipitation”

Page 2- L27: replace “This flow direction” with “losing streams”

Page 3- L16-17: Not clear. Please explain

Figure 1. Add further details to the caption. Describe P30 of 30” DEM and why average of 30” DEM is compared with P30?

Page 4- Final paragraph needs revision to improve consistency with the rest of the text.

Page 6 – L6: replace “loosing” with losing

Page 6 – L28: Replace equation 5 with 6

Section 2.2. If the coupling is not done yet, why it is discussed in this paper? You could devote this section to assess the impacts of simplifications/improvements you made compared to the linear reservoir GW model. Or this section should come last (future work) if you plan to include the coupling approach in this paper.

Page 8 – L4: Could you please justify the choice of ocean conductance value.
Section 3.1. Why P30 has been considered as the surface elevation of surface water bodies in the model?

Section 3.2. L18 – “the amount of river water that recharges GW is only about a 40th of the drainage to GW,” not clear

Section 3.2 – L20 – Why outflow from SW body is not limited by water availability? River stage in equation 5 should control this.

Section 3.2- Could you verify estimated global water budget against other existing global models?

Section 3.3 – No attempt has been made to verify SW-GW exchange rate. This is an important contribution of this work and model validation is required. Perhaps, use ParFlow simulations over CONUS to check these fluxes at steady-state.

Section 3.6. Comparison with the Central Valley hydrologic model is purely qualitative. It does not seem this qualitative comparison adds any value to assess G3M model performance.

Hide

RR by Anonymous Referee #4 (02 Dec 2018)

Suggestions for revision or reasons for rejection

The paper presents an application of the G3M model for steady-state groundwater flow using output for the WGHM as input. The longer term objective is to fully integrate G3M into WaterGap2 for transient flow simulations, with feedback between G3M and WaterGap2. The paper heavily focuses on this long-term objective, which brings confusion because the paper does not present any results from the fully-coupled model. Any comment or observation on the fully-coupled model therefore remains purely speculative, since the work has yet to be done. I recommend to change the focus to what is actually presented in the paper, which is the steady-state uncoupled application. It would greatly clarify the paper. Section 2.2 should therefore be deleted. The long-term objective could be mentioned in the discussion, as a perspective for future work. It would then be quite acceptable for the authors to justify the choice of using G3M instead of MODFLOW (as raised by reviewers), for tighter integration with WaterGap2.

I agree with the previous reviewers that the lessons learned from the steady-state application are not clear, compared to previous work. Since this paper does not present any results from a coupled model, arguments about coupling strategy as outcome are not valid. A future paper that actually presents results with the coupled model could address this coupling and assess challenges and issues. The other novel elements mentioned relate to scale challenges and equation solved but they are not unique to G3M. The other gradient-based GW flow models mentioned in the paper also face similar issues. The main outcomes of this steady-state application should be much more clearly highlighted and justified, in the context of previous studies.

The previous reviewers raised concerns about using the fully-saturated groundwater flow equation, as opposed to an unconfined flow equation, for the upper layer. The justification given in response to these concerns are rather confusing and difficult to understand. I think that a short description of the way WCHM treats the soil “compartment” is needed and could help justify the approach. Right now, the reader has to guess what “soil” refers to.

The paper is, in general, not clearly written. The style is often unnecessarily complicated, with very long sentences and repetition. The paper also does not focus on the essential and there is a lot of unnecessary detail. On the other hand, some important information is not presented clearly, such as a short description of how soils are treated in WCHM. Another example is Line 5 on page 8 that states : “It is assumed that there is exchange of water between GW and one river stretch in each 5' grid cell”. If I understand correctly, there is a river for every top cell in the model. If it is the case, it is an important assumption and should be stated much more clearly and earlier, and not “buried” on page 8.

The terminology used is also very confusing for a reader with a hydrogeology background. For example, groundwater is used to refer to both the contained (the “groundwater” or subsurface compartment) and the content (groundwater that flows according to the governing equation). The paper also uses the term “drainage” to represent fluid exchange between the various compartments (subsurface, rivers, soil, wetlands, etc.). The use of drainage is extremely confusing and a better terminology would greatly clarify the paper. Also, hydrogeologists do not use hydraulic head for surface water or surface water table.

Detailed comments
The abstract provides a good example of the writing style. First, it is much too long. The abstract should be short, precise and to the point. It should not try to explain everything, such as the difference between reservoir and gradient-based GW models. It also contains sentences that are either unclear of very complicated. For example, this excerpt from lines 19-20 : “We identify challenges linked to the coarse resolution, which necessitates the deviation from established processes in regional groundwater modelling as simulation of unsaturated flow and SW body elevation”. That sentence is both complicated and unclear, and does not inform the reader.
P2. The first paragraph in the introduction (lines 1-20) presents only generalities and should be deleted. The paper should focus on the model right from the start.
P2. Line 34 : what are “macro-scale models”?
P2, Lines 37-38 : what is “the condition of SW”? Be more specific.
P2. Lines 38-39 : the excerpt “Miguez-Macho et al. (2007) linked a land surface model with a two-dimensional gradient-based GW model and computed, with a daily time step, gradient-based GW flow” is one example of unnecessary repetition that does not help the reader. There is no need to repeat that Miguez-Macho et al. used their gradient-based model to compute gradient-based GW flow. Another example of repetition is on page 3, lines 18-19 : “In this study, we present the Global Gradient-based Groundwater Model (G3M) that is to be integrated into the GHM WaterGAP 2” and just a bit further, line 29 is : “G3M is to replace this linear reservoir model in WGHM”. Actually, that last repetition is even more confusing because GHM WaterGAP 2 and WGHM are not even the same model. I have noted several such repetitions that I will not list but that the authors should identify and eliminate.
P3. Line 3 : “GW above the land surface”. Check terminology for more clarity. GW above land surface is no longer GW. Should probably write instead something like : groundwater exfiltration.
P3. Line 12 : The difference with the de Graaf et al paper is that they added “an additional drainage flus to GW drainage”. That explanation is given several times in the paper but it is never clearly described and it is difficult to understand what de Graaf et al. did exactly.
P3, Line 30 : The paper mentions the G3M model and now the G3M-f framework. First, what is the difference between G3M and G3M-f? Second, is it relevant to mention anything else that the model, G3M, since the application here is totally decoupled from a global hydrologic model?
P3, Lines 33-34 : The formulation “We want to find out whether we can use gradient-based groundwater modelling at the global scale, …, to improve estimation of flows between SW and GW … and capillary rise” is rather surprising. Isn’t the working hypothesis that gradient-based models will improve simulations?
P3, lines 36-40 : This is an example of a very complicated and unclear sentence : “Steady-state simulations are a well-established first step in groundwater model development to understand the basic model behaviour limiting model complexity and degrees of freedom, thus providing insights into dominant processes and uncovering possible model-inherent characteristics impossible to observe in a fully coupled transient model.” For example, what is “the basic model behaviour limiting model complexity and degrees of freedom”? What are the “model-inherent characteristics” that can’t be observed in a fully coupled transient model?
P3, lines 40-41 : I don’t know what is meant by “A transient model might obfuscate model inherent trends due to the slow changing nature of groundwater processes.”
P3, line 41 : The following statement is quite bold and I am not sure that I agree : “A fully coupled model furthermore adds complexity and uncertainty to the model outcome”. If it is the case, what do the authors want to develop a fully-coupled model if its outcome will be more uncertain?
P4. Section 2 on Model description. I suggest to reorganize that section because it is not clear. I suggest to start by presenting the governing equations (equations 2 to 7) and then present the global-scale components. All simplifying hypothesis should be clearly stated. The exact input data originating from the global hydrologic model should also be clearly presented.
P4, lines 10-11 : “G³M differs from traditional local and regional GW models”. Is it really the case? I think that the main difference is the scale of application and the use of WCHM output as input.
P4, line 17 : “At this scale, information listed above is poor or non-existing”. It should be reworded. The information contained for smaller-scale (as mentioned just above) is still available at the large (global) scale. You probably mean something else.
P4, line 23 : Not clear what is meant by : Due to the lack of the distribution of hydrogeological properties.
P4, line 33 : I assume that “groundwater boxes” are actually “groundwater cells”. If it’s the case, then “cells” should be only consistently.
P5. Figure 1 could be improved because it is not clear what is shown exactly. Also, what are “virtual layers”?
P5. Equation 2 is not the correct partial differential equation (PDE). The cell volumes (delta_x, delta_y, delta_z) only appear when the PDE is integrated over a 3D cell. Also, writing that the partial differential equation is “a function of hydraulic head gradients” is not rigorously correct. The PDE is derived from applying mass conservation to a representative elementary volume, where groundwater flow is described by its mass flux. The hydraulic head gradients appear because the mass flux is expressed with Darcy’s Law.
P5, lines 20-21 : It is confusing to write that “Inflows in the groundwater are accounted for as…” because the equation is for both inflow and outflow.
P7, line 35 : in the exponential, what is m? what is the value of f?
P8, line 4 : the value of c_ocean is set to 100 m2/day. It appears to be several orders of magnitude greater than other conductances. Based on equation (4), I suspect that this large value is similar to specifying a first-type boundary condition for all cells located on the ocean boundaries. Is it the case?
P9, lines 11-15. The paragraph is not clear.
P12, line 15 : “High conductance values are reached in the tropical zone due to a higher GW recharge”. Is it really the case and not the opposite, i.e. because the conductance is large, groundwater recharge is larger?
P14, line 11 : what is meant by : comparison to local studies suggested a unit conversion error?
P15, lines 18-19 : I don’t understand the sentence : Plotting hydraulic head instead of depth to GW has the disadvantage that the goodness of fit is dominated by the topography as the observed heads are calculated based on the surface elevation of the model.
P17, lines 29-30. I don’t understand the reference to “model decision”. What does it mean?
Table 2 : It is incorrect to write that the first 3 models solve the 3D Darcy equation. They solve a 3D mass conservation equation where the fluid flux is expressed with Darcy’s law. It is also the case for ParFlow, which uses Darcy’s Law to represent fluid fluxes in Richards’ equation.

Hide

ED: Reconsider after major revisions (17 Dec 2018) by Wolfgang Kurtz

AR by Robert Reinecke on behalf of the Authors (09 Feb 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (22 Feb 2019) by Wolfgang Kurtz

RR by Anonymous Referee #3 (16 Mar 2019)

RR by Edwin Sutanudjaja (23 Apr 2019)

ED: Publish subject to minor revisions (review by editor) (26 Apr 2019) by Wolfgang Kurtz

AR by Robert Reinecke on behalf of the Authors (20 May 2019) Author's response Manuscript

ED: Publish as is (29 May 2019) by Wolfgang Kurtz

AR by Robert Reinecke on behalf of the Authors (30 May 2019) Author's response Manuscript

Short summary

G³M is a new global groundwater model (http://globalgroundwatermodel.org) that simulates lateral and vertical flows as well as exchanges with surface water bodies like rivers, lakes, and wetlands for the whole globe except Antarctica and Greenland. The newly developed model framework enables an efficient integration into established global hydrological models. This paper presents the G³M concept and specific model design decisions together with first results under a naturalized equilibrium.

Challenges in developing a global gradient-based groundwater model (G3M v1.0) for the integration into a global hydrological model

Download

Interactive discussion

Peer-review completion

Challenges in developing a global gradient-based groundwater model (G³M v1.0) for the integration into a global hydrological model