GSFLOW–GRASS v1.0.0: GIS-enabled hydrologic modeling of coupled groundwater–surface-water systems

Ng, G.-H. Crystal; Wickert, Andrew D.; Somers, Lauren D.; Saberi, Leila; Cronkite-Ratcliff, Collin; Niswonger, Richard G.; McKenzie, Jeffrey M.

doi:https://doi.org/10.5194/gmd-11-4755-2018

Articles | Volume 11, issue 12

https://doi.org/10.5194/gmd-11-4755-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-11-4755-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 11, issue 12

Model description paper

|

30 Nov 2018

Model description paper |

| 30 Nov 2018

GSFLOW–GRASS v1.0.0: GIS-enabled hydrologic modeling of coupled groundwater–surface-water systems

G.-H. Crystal Ng, Andrew D. Wickert, Lauren D. Somers, Leila Saberi, Collin Cronkite-Ratcliff, Richard G. Niswonger, and Jeffrey M. McKenzie

Download

Final revised paper (published on 30 Nov 2018)
Preprint (discussion started on 30 Jan 2018)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'review comments on the paper by Ng et al.', Anonymous Referee #1, 26 Feb 2018
- AC1: 'initial reply to referee 1's comments', G.-H. Crystal Ng, 07 Mar 2018
  - RC4: 'Few clarifications on the first round of comments', Anonymous Referee #1, 12 Mar 2018
    - AC2: 'Final response to Reviewer 1', G.-H. Crystal Ng, 26 Apr 2018
RC2: 'Review comments on Ng et al.', Anonymous Referee #2, 07 Mar 2018
- AC3: 'Response to Reviewer 2', G.-H. Crystal Ng, 26 Apr 2018
RC3: 'review of Ng. et al', Anonymous Referee #3, 08 Mar 2018
- AC4: 'Response to Reviewer 3', G.-H. Crystal Ng, 26 Apr 2018

Peer-review completion

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

AR by Gene-Hua Crystal Ng on behalf of the Authors (26 Apr 2018) Author's response Manuscript

ED: Referee Nomination & Report Request started (02 May 2018) by Min-Hui Lo

RR by Anonymous Referee #1 (17 May 2018)

Suggestions for revision or reasons for rejection

I would like to thank the authors for their effort in improving the manuscript according to the reviewers’ comments. I see clear changes in several parts of the manuscript, most notably in the introduction part and in the description (including also a new feature) of the proposed toolkit. All of these changes are beneficial for the overall quality of the work. However, I must also admit that one of my major remarks concerning the Results section still persists in this revised version. Here my opinion is that the effort made by the authors did not really go to the ‘right’ direction. The discussion still contains many vague statements and it is not supported by illustrative figures. That is, the outcome of the results section appears disconnected from those statements (made in the Introduction) about the novelty of the toolkit. This is, in my opinion, a major issue for a GMD paper where the reader expects to see a more in-depth discussion on the technical advancement of the proposed toolkit. I provide below some specific comments that I hope will be useful to address my major concern with the work.

1. Authors state that the scope of the three examples is to demonstrate the robustness of the developed GRASS GIS domain builder across diverse topographic settings. I have problems in assessing this ‘robustness’ from Figures 5-6-7. In my opinion robust means that the output of the model does not change ‘much’ with small perturbations of the input. According to this I would frame the exercise in a different way. For each test case you identify a major challenge and then you exploit the toolkit to automatically generate a suite of model grid configurations that are tested in terms of different model output. This exercise will provide the opportunity to demonstrate the validity of the approach in guiding users during preliminary (i.e., grid definition) still critical modeling steps.

2. In the first example it seems that one of the most important features of the toolkit is the combination of flow paths extracted using high-resolution topography with the ‘fewest’ possible computational cells for MODFLOW. Could the authors identify (using their pre- and post-processing tool) the ‘exact’ value of the fewest possible computational cells?

3. In the second example, it seems that one of the most challenging task for the toolkit is to appropriately handle NULL value cells. I am not sure about it.

4. Again in the second example, I am really confused by those statements connecting the numerical convergence of MODFLOW with the merit of the proposed toolkit. Specifically, I do not see how the suite of tools can avoid numerical convergence issues when MODFLOW cells run dry. Overall, I think that numerical convergence problems may still appear in MODFLOW and this won’t make the proposed tool useless.

5. In the third example, authors state that under smooth topographic gradients simple downslope flow-routing algorithms fail and they use this as an example to demonstrate the ability of their toolkit under such low-relief conditions. However, it is not clear to which extent this ability is due to a new feature of the proposed toolkit or to an already existing GRASS (r.watershed) functionality.

6. Finally, all applications presented in this work are at relatively small catchment scale. Can the authors proof the utility of their tool at large, i.e. continental, scale using for instance the CONUS domain?

According to the aforementioned points I would encourage authors to make a substantial effort to augment the insights from the results section, which will bring the overall quality of the manuscript suitable for GMD standard.

Other points:

1. The discussion in the Introduction connecting surface-subsurface coupling approaches to model input parameters is misleading. You may have that a fully coupled and a sequentially coupled approach requires the same amount of input information.
2. I would avoid statements about the potential of the tool in developing countries where computational resources may be limited. I do not see how the proposed tool may help with respect to this limitation.
3. Usually captions do not contain discussion of the presented results. Please check Figures 5-6-7.

Hide

RR by Anonymous Referee #2 (20 May 2018)

ED: Reconsider after major revisions (20 May 2018) by Min-Hui Lo

AR by Gene-Hua Crystal Ng on behalf of the Authors (30 May 2018) Manuscript

ED: Referee Nomination & Report Request started (07 Jun 2018) by Min-Hui Lo

RR by Anonymous Referee #1 (03 Jul 2018)

Suggestions for revision or reasons for rejection

I would like to thank the authors for their effort in clarifying several points made in my last round of review comments. In acknowledging such effort, I must recognize that my major concerns, expressed during the first and second iteration of this review process, still persist. These points cannot be addressed just reformulating some sentences in the text but they need, in my opinion, additional substantial work by the authors. Indeed, as previously stated, the Results section fails to offer a clear view of the technical advancements proposed with the GSFLOW-GRASS toolkit. This is, in my opinion, a major issue for a GMD paper. Here I must also highlight that despite this central concern, Figures 5-6-7 (and their information content) have remained nearly unchanged through the review process. I will try to re-iterate on these major issues starting from the newly included table (i.e., Table 2 of the revised manuscript) that summarizes the tests of GSFLOW-GRASS capabilities/technical challenges for the three selected test cases.

The technical challenges identified for the first test case (i.e., Shullcas River, Peru) are the (i) steep topography and (ii) seasonal rainfall.

(i) As concerns the first technical challenge authors argue that a steep topography requires the definition of the surface flow path using high-resolution topography with the option of resolving the subsurface at relatively coarse resolution. Therefore, the technical challenge consists in handling a surface discretization not automatically aligned with subsurface flow path. This technical challenge reflects the sketch of Figure 1d, if I am right. Therefore, I think that a good way to test the toolkit would be to generate different subsurface configurations (in terms of grid resolution) with the same number of HRUs obtained with the high-resolution DEM (30 m for this test case). Here you could show the great potential of the automated workflow in generating consistent modeling scenarios that potentially subtend different level of information. Some concrete numbers on the computational gain would substantiate some too-vague statements still present in the manuscript (“…steep topography and narrow canyons of Shullcas is the need for impractically high resolution and computational expense if using a standard gridded model domain…”)

(ii) The aim of this work is not to present any scientific validation of the results obtained by implementing the GSFLOW-GRASS workflow, as also stated by authors. Therefore, I do not see any technical challenge implied in the term “Seasonal rainfall” included by authors in this table.

The technical challenges identified for the second test case (i.e., Water Canyon, USA) are the (i) NULL cells defining irregular coastline, (ii) small basin, and (iii) small number of coarse-resolution cells.

(i) I must admit that I still do not see the challenge of handling NULL cells. Authors show the modification of the code but this remains for me a basic feature of any terrain analysis tool.
(ii) Why a small basin is a challenging case for GSFLOW-GRASS?
(iii) What’s the technical challenge associated to a small number of coarse-resolution cells? What’s the meaning of small?

According to point (i), (ii), and (iii) I have problems in identifying the technical capabilities of the proposed workflow from this second test case. Again, there are confusing statements in the discussion of the results that do not bring much insight:

“The GSFLOW-GRASS implementation does simulate low streamflow as expected (Figure 6B), but the domain-builder generated sufficiently thick and hydrologically corrected MODFLOW cells that they maintain water and avoid any computational problems.”

What do you mean with “sufficiently thick” and “hydrologically corrected” MODFLOW cells?

The technical challenges identified for the third test case are (i) two-layer soil hydraulic conductivity, (ii) poorly integrated drainage, (iii) low relief, and (iv) large basin.

(i) The tasks associated to the first point (i.e., two-layer soil hydraulic conductivity) are not discussed in the main text.

(ii) and (iii) In my previous review I highlighted the problem that all the discussion for this test case is about an already existing GRASS feature (r.watershed). Therefore, one could argue that nothing new related to GSFLOW-GRASS has been discussed/presented in this test case. This issue will persist as long as authors do not expand the focus of their exercise. Authors also argue that “Most GIS tools will artificially fill these pits, but r.watershed's least-cost-path algorithm is able to route flow across them in a more realistic way“. This is a quite interesting technical aspect to show with a practical example where you substantiate on relevant modeling issues using GSFLOW-GRASS. As it is right now the reader has no concrete results to get a better insight on what the authors state.

(iv)What’s the challenge associated to a large basin? Technically a small basin resolved at high resolution could be the same of a coarse-resolution large basin.

Other points:

1. Captions still contain extensive discussion of the results that should be moved in the main text.

2. I disagree with this statement:

“Unlike PIHM, GSFLOW-GRASS employs regular groundwater grid cells that are distinct from the irregular surface units, which makes the integrated domain building more complicated but allows for more complex representation of the surface-water and aquifer systems.”

PIHM implements a fully unstructured grid that allows for a detailed representation of surface-subsurface water interactions. See reference below:

D. Wang, Y. Liu, and M. Kumar (2018). Using nested discretization for a detailed yet computationally efficient simulation of local hydrology in a distributed hydrologic model. Scientific Reports.

Hide

ED: Reconsider after major revisions (03 Jul 2018) by Min-Hui Lo

AR by Gene-Hua Crystal Ng on behalf of the Authors (15 Aug 2018) Manuscript

ED: Referee Nomination & Report Request started (21 Aug 2018) by Min-Hui Lo

RR by Anonymous Referee #1 (06 Sep 2018)

Suggestions for revision or reasons for rejection

I would like to thank the authors for their very detailed replies to my last round of comments. My overall impression is that their effort adds to the clarity and quality of some parts of manuscript (see new section 3.2.1 – section 3.2.2) while keeping partially solved some other major concerns that I had with the manuscript. In particular, as already expressed in my very first review and re-iterated over the two successive steps, the information content of Fig. 5-6-7 does not clearly reflect the capabilities and claimed features (that I do not doubt exist) of the proposed toolkit. In expressing my opinion I highlighted that such issues would have required some substantial work (including improvements of the figures) going beyond some pure text modification/clarification.

1) In the current form, the results demonstrate that the proposed automated workflow is able to generate a set of consistent input and output files starting from different physiographic settings. This is clearly an important and necessary achievement for the toolbox. However, this is not sufficient to demonstrate the robustness of the toolkit to handle the technical challenges identified for each test case. A rigorous approach to test the technical capabilities (as I have already suggested) would consist in the generation of different configurations to be ingested in the hydrological model. Indeed, in the implementation of the automated workflow the user has to define different “a priori” parameters values (e.g., thresholds value for the definition of the HRU, subsurface coarsening, etc.) that result in different configurations that eventually produce different results of the integrated hydrological model. This is in my opinion the way to show how the proposed toolkit could really guide users through important modeling steps. I enforce my opinion highlighting that statements like “…79 HRU cells were needed to capture nearly all the essential flow-routing information…”; “…GSFLOW-GRASS’ hydrological correction to enforced integrated subsurface proved essential for preventing unrealistic results”, “Early model tests for Shullcas showed that simple grid coarsening using the mean value of the elevations…” etc… are still rather vague and do not reflect the information content of Fig5-6-7.

2) In the discussion of each test case you should remove all the text (including also “Applications” column in Table 3) providing a generic description of the “potential” scientific issues (e.g., erosion) associated to the case study and “demonstrated” with GSFLOW-GRASS. These issues are not truly addressed in this work and all of these argumentations appear disconnected from the main objective of the work. Note here that the head paragraph in the “Examples” section contains already an exhaustive and general description of the three test cases.

3) In the last revised version authors cite the work of Gardner et al. (2018), which proposes a series of technical (pre-processing) solutions for the same integrated hydrological model (i.e., GSFLOW-Arcpy). As the work of Gardner et al. is now published and available online (it was not during the first, second, and third review iteration), I think it is quite important to:

- Better highlight the contribution (need) of GSFLOW-GRASS in the introduction.
- Have a dedicated section presenting the differences and similarities between the solutions implemented in both toolkit (i.e., GSFLOW-GRASS and GSFLOW-Arcpy).

Hide

ED: Reconsider after major revisions (06 Sep 2018) by Min-Hui Lo

AR by Gene-Hua Crystal Ng on behalf of the Authors (14 Oct 2018) Author's response Manuscript

ED: Referee Nomination & Report Request started (17 Oct 2018) by Min-Hui Lo

RR by Anonymous Referee #1 (07 Nov 2018)

ED: Publish subject to minor revisions (review by editor) (07 Nov 2018) by Min-Hui Lo

AR by Gene-Hua Crystal Ng on behalf of the Authors (09 Nov 2018) Author's response Manuscript

ED: Publish as is (12 Nov 2018) by Min-Hui Lo

AR by Gene-Hua Crystal Ng on behalf of the Authors (15 Nov 2018)

Short summary

The profound importance of water has led to the development of increasingly complex hydrological models. However, implementing these models is usually time-consuming and requires specialized expertise, stymieing their widespread use to support science-driven decision-making. In response, we have developed GSFLOW–GRASS, a robust and comprehensive set of software tools that can be readily used to set up and execute GSFLOW, the U.S. Geological Survey's coupled groundwater–surface-water flow model.