CFD Model of the Density-Driven Bidirectional Flows through the West Crack Breach in the Great Salt Lake Causeway

Round 1

Reviewer 1 Report

In this manuscript, the authors develop a computational fluid dynamics model, using Flow3D software, to simulate the bi-directional flows through the West Crack Breach in the Great Salt Lake Causeway.  The unique flow patterns present in the WC Breach and its impact on GSL ecosystem are well detailed and justify the CFD modelling efforts.  The model was found to simulate the complex flow patterns with reasonable accuracy when compared to field measurements.  The model is well described and mesh refinement steps demonstrate solution independence.  In particular, I found the acknowledgement of potential sources of error from field measurements and extent in which the authors accommodated these uncertainties (e.g., location of the ADVM, density vs. conductivity curve, initial density difference) to be well described and very informative.  Overall, the manuscript was well written, but I recommend the authors closely review the manuscript to correct many small editorial issues and ensure terminology and acronyms are consistent throughout the paper.  Regardless, the topic would be of interest to the journal audience, and I recommend the manuscript for publication following minor revisions.  A detailed list of line by line comments are provided below:

Line 32: Add “the” before “Caspian Sea”.

Line 54: Consider revising text to state “…we discuss the unique features of the GSL…”

Line 55: Remove “about”.

Line 61: Throughout the entire manuscript acronyms such as “GSL” are used without adding “the” before.

Line 71: Insert “was” between “breach” and “created”

Line 74: Revise text to state “During the past 3-years” and use parentheses around “WSE” instead of commas.

Line 117: Use consistent punctuation when using the term “bidirectional”.  The authors use “bidirectional” and “bi-directional” interchangeably throughout the manuscript.

Line 135: Consider “hydraulic structure of the breach”  

Line 151: Write USGS out prior to first use of the acronym.

Line 197: Define “SC” at first use of the acronym.

Line 199 & 203: Add “the” before “USGS”.

Figure 4:  Is there any explanation for the incredibly wide range of density measurements at the higher conductivity in the HDR samples?  This seems to introduce the most variability into the data set.

Figure 6:  What is the justification of the seemingly random distances selected for the ADVM profiles?  Why not just look a profiles evenly spaced away from the pier?

Figure 7:  Consider using a different naming convention for the vertical slice locations as their reference in Figure 9 are incredibly confusing. For example “…at profile location a, (a), at location c, (b), at location e (c), and at location i (d)”.

Line 317: “differences”

Figure 9: Why is the density difference provided in slugs/ft^3 here?

Figure 12 & 13: Consider adding north arrows and scale to help readers orientate to the two different views.

Line 389: Check spelling of “densimetric”.

Line 399: Check spelling of “While”.

Line 400: Replace “that” with “than”.

Line 400: Consider revising the text “Though by the time the flow comes our of the breach…”

Line 413: Remove dash in “relatively”.

Line 415: Remove dash in “getting”.

Line 426: Delete “in” before “North”.

Line 434: Add “the” after “through”.

Line 439: Remove additional “0.” in R-value.

Line 442-444: This is the first mention of a hydraulic structure rating curve in the entire text.  What justification is there for this development and why isn’t it included in this manuscript?  Furthermore, it would be worthwhile for the authors to comment on the necessity of a simplified rating curve over use of the CFD model for future hydraulic modelling of the Great Salt Lake.

Author Response

REVIEWER 1

 

In this manuscript, the authors develop a computational fluid dynamics model, using Flow3D software, to simulate the bi-directional flows through the West Crack Breach in the Great Salt Lake Causeway.  The unique flow patterns present in the WC Breach and its impact on GSL ecosystem are well detailed and justify the CFD modelling efforts.  The model was found to simulate the complex flow patterns with reasonable accuracy when compared to field measurements.  The model is well described and mesh refinement steps demonstrate solution independence.  In particular, I found the acknowledgement of potential sources of error from field measurements and extent in which the authors accommodated these uncertainties (e.g., location of the ADVM, density vs. conductivity curve, initial density difference) to be well described and very informative.  Overall, the manuscript was well written, but I recommend the authors closely review the manuscript to correct many small editorial issues and ensure terminology and acronyms are consistent throughout the paper.  Regardless, the topic would be of interest to the journal audience, and I recommend the manuscript for publication following minor revisions.  A detailed list of line by line comments are provided below:

 

Response: Thank you for your comments and suggested revisions as they have helped us improve the manuscript and make many corrections.  Please see our response to your comments and suggestions below, and in the manuscript. 

 

Line 32: Add “the” before “Caspian Sea”.

Response: Revised as suggested.

 

Line 54: Consider revising text to state “…we discuss the unique features of the GSL…”

Response: Revised as suggested.

 

Line 55: Remove “about”.

Response: Revised as suggested.

 

Line 61: Throughout the entire manuscript acronyms such as “GSL” are used without adding “the” before.

Response: Revised as suggested.

 

Line 71: Insert “was” between “breach” and “created”

Response: Revised as suggested.

 

Line 74: Revise text to state “During the past 3-years” and use parentheses around “WSE” instead of commas.

Response: Revised as suggested.

 

Line 117: Use consistent punctuation when using the term “bidirectional”.  The authors use “bidirectional” and “bi-directional” interchangeably throughout the manuscript.

Response: Revised as suggested.

 

Line 135: Consider “hydraulic structure of the breach”  

Response: Revised as suggested.

 

Line 151: Write USGS out prior to first use of the acronym.

Response: Revised as suggested.

 

Line 197: Define “SC” at first use of the acronym.

Response: Revised as suggested.

 

Line 199 & 203: Add “the” before “USGS”.

Response: Revised as suggested.

 

Figure 4:  Is there any explanation for the incredibly wide range of density measurements at the higher conductivity in the HDR samples?  This seems to introduce the most variability into the data set.

 

Response: We thank the reviewer for making this interesting observation. The HDR samples were not collected by our team, and data was made available to us through publicly available reports. Thus, it is difficult for us to comment with any surety. In general, the higher conductivity measurements are the ones that HDR reported from North of the breach, whereas the lower conductivity measurements are from South of the breach. Within the higher conductivity measurements, there is variation in density without change in conductivity. The lower density but high conductivity measurements are from couple of months in the summer. We are not sure what could result in this behavior in a hypersaline environment, but it is a question we will dig a little deeper in the future.

 

Figure 6:  What is the justification of the seemingly random distances selected for the ADVM profiles?  Why not just look a profiles evenly spaced away from the pier?

 

Response: We agree that one could have compared profiles evenly spaced from the pier. We chose the locations on the basis of an approximate location provided to us by our collaborators at USGS, while trying to keep the number of profiles to a manageable number. 

 

Figure 7:  Consider using a different naming convention for the vertical slice locations as their reference in Figure 9 are incredibly confusing. For example “…at profile location a, (a), at location c, (b), at location e (c), and at location i (d)”.

Response: The vertical slices have been renamed as 1, 2, 3 …

 

Line 317: “differences”

Response: Revised as suggested.

 

Figure 9: Why is the density difference provided in slugs/ft^3 here?

Response: Revised to metric, this was an oversight on our part.

 

Figure 12 & 13: Consider adding north arrows and scale to help readers orientate to the two different views.

Response: We have added an arrow pointing to the North.

 

Line 389: Check spelling of “densimetric”.

Response: Revised as suggested.

 

Line 399: Check spelling of “While”.

Response: Revised as suggested.

 

Line 400: Replace “that” with “than”.

Response: Revised as suggested.

 

Line 400: Consider revising the text “Though by the time the flow comes our of the breach…”

Response: Revised as suggested.

 

Line 413: Remove dash in “relatively”.

Response: Revised as suggested.

 

Line 415: Remove dash in “getting”.

Response: Revised as suggested.

 

Line 426: Delete “in” before “North”.

Response: Revised as suggested.

 

Line 434: Add “the” after “through”.

Response: Revised as suggested.

 

Line 439: Remove additional “0.” in R-value.

Response: Revised as suggested.

 

Line 442-444: This is the first mention of a hydraulic structure rating curve in the entire text.  What justification is there for this development and why isn’t it included in this manuscript?  Furthermore, it would be worthwhile for the authors to comment on the necessity of a simplified rating curve over use of the CFD model for future hydraulic modelling of the Great Salt Lake.

 

Response: We thank the reviewer for bringing up a very pertinent question about a rating curve for the breach structure. For purpose of developing an accurate water and salt balance model of the lake, it is important to correctly estimate the flow going through the breach under different conditions. Even though a CFD model of the breach cannot be directly coupled with a water/salt balance model of the lake, it can be used to simulate different conditions (e.g. lake level, density difference, etc.) and the results then can be used to develop a rating curve for the breach. This rating curve then can be used within the water/slat balance model. One could try to develop a simple rating curve for the structure from measurements of the flow, water level etc. at the breach. Though, this rating curve would only be valid for the range of observed conditions at the lake. Thus, a validated CFD model is more appropriate for developing a rating curve for conditions that have not been observed, but can happen in the future (e.g. drop in lake level due to sustained drought).

            In the current paper we haven’t discussed about the rating curve, as the paper is primarily about development of a CFD model that can capture the complicated flow at the breach. Also, developing the rating curve will require an extensive set of simulations, which is outside the ambit of the current study.  

 

Reviewer 2 Report

The article presents an interesting study of the density-affeteced currents between two parts of a lake with different salinities. Measurements of velocity profiles are given, together with results from a CFD model study. The Great Salt Lake and its barrier provide a unique possibility for studying stratified flow. I am glad to see that scientists are taking advantage of this particular local phenomenon to study stratified flow in more detail.

 

1. The main measured result from the current study is the vertical velocity profile in the breach/channel between the two parts of the lake. The vertical profile shows water flowing in one direction at the surface and in another direction close to the bed. One weakness of the study is that we don’t know the exact location where the measurements were taken. Measuring a profile in a channel with an ADV is usually a fairly straightforward task. Especially since there is a bridge over the channel which is possible to base the measurements from. I would encourage the authors to carry out a new measurement campaign. An ADV also gives a profile of the turbulence. This is not included in the current article. It would have been very interesting to compare measured vs. computed turbulence profiles in the channel. Turbulence and its damping is very important in stratified flow.

 

2. The numerical modelling has been done with the Flow-3D software. Being a commercial product, it is more tested and less prone to bugs compared with in-house software. The downside is that the authors and readers of the article can not be completely sure which equations are solved and how the numerical algorithms work. The current article describes some of the algorithms used by the software, but not the important ones: How is the density difference computed in the Navier-Stokes equations? Which equations are solved to compute the density and the mixing of the water from the two sides of the lake? Probably, there is also an equation for the salinity solved? These equations have to be given in the paper, together with details on their solution methods.

 

3. The authors tested several turbulence models initially on a 2D grid. What was the cell size of this grid compared with the final 3D grid?

 

4. Turbulence modelling is especially important for stratified flow, as the turbulence is dampened by the density gradients. How was the turbulence damping modelled in the current study? This has to be stated in the paper.

 

5. If the authors did not use any turbulence damping by stratification, the turbulence models would overpredict the eddy-viscosity and thereby give poor results. LES would normally underpredict turbulence on a normal grid, but because the stratification is damping the physical turbulence, then the LES model gives the best results. Could this be an explanation of the results in the current case?

 

6. What was the water depth in the breach/channel? Fig. 9 suggest around 5 meters. The vertical scale in Fig. 9 should be given more decimals, or the zero level should be set to the bed.

 

7. What was the y+ value close to the bed in the LES computations? This should be given in the paper together with a justification of its value. A reference should be given for the normal range of y+ values in LES studies. LES usually requires the y+ value to be in the lamninar sublayer, This is most often not possible, except for laboratory scales. The k-omega model does not have this problem. In any case, the authors should present the 2D results with the different turbulence models in the article.

 

8. How was the friction between the water and the lake bed computed? Which roughness coefficient was used? Was any calibration of the friction coefficients done?

 

9. The authors use some numerical expressions that most of the readers of the journal will not understand. These expressions therefore have to be explained in the paper: “GMRES subspace”, “interblock boundary coefficient” and “volume fraction cleanup”.

 

10. It seems that the density in the current study was only assumed to be a function of salinity. Was there a temperature gradient in the vertical direction in the lake? Or a temperature difference between the two sides of the lake?

 

11. What was the boundary conditions at the sides marked P in Fig. 7? Was there an inflow/outflow of water? If not, then the CFD model did not compute a steady situation, as the water levels would eventually be the same on both sides of the breach. If a steady solution was not obtained, more information about the transient nature of the computations should be given. For example, the computed water levels as a function of time should be given (similar to the measurements in Fig. 5, but for a shorter time)

 

12. Line 294: The word “veracity” is probably not in the vocabulary of most readers of the journal. I had to google it. The article would be more readable by replacing the word with “accuracy”.

 

13. The density profile in Fig. 12 looks interesting. Was any procedure used to sharpen vertical gradients in the salinity/density? False diffusion might easily smear the density gradients. Some diffusion of salinity/density would take place naturally. A question is how well the CFD model reproduces this process. This should be discussed in the article.

 

14. Line 399: “Whiel”.

 

15. Line 401 “our of the breach”

 

16. Fig. 15 needs color scales.

 

17. The thickness and velocity of the deep brine layer (DBL) was also computed with the CFD model. However, these computations were not verified against measurements. Although the results look plausible, the quantitative accuracy of the results are uncertain. This should be pointed out in the article.

 

Line 443: “a hydraulic structure rating curve can be successfully developed for hydrologic modeling of the Great Salt Lake”. I am not sure about this statement. Although the results are qualitatively plausible, the input data are uncertain and many choices have been made to obtain the computed velocity profile. The velocities are “calibrated” by the choice of the salinity used as boundary conditions. Earlier in the conclusion, the velocities are said to be within 20 – 50 %. The accuracy of a rating curve produced by the CFD model is therefore uncertain. I suggest that this sentence about the rating curve be removed from the article. Or rephrased to “a hydraulic structure rating curve may be developed for hydrologic modeling of the Great Salt Lake”

 

To have a high-quality paper, I would recommend that the field measurements are done over again, so that the location is known. And turbulence is also measured. I also recommend that the damping of the turbulence from the stratification is included in the CFD model, and the results recomputed. Then the measured and computed turbulence can also be compared. However, if the Associate Editor thinks this is not required, the authors should at least include in the article a discussion of the points above.

Author Response

REVIEWER 2

 

The article presents an interesting study of the density-affeteced currents between two parts of a lake with different salinities. Measurements of velocity profiles are given, together with results from a CFD model study. The Great Salt Lake and its barrier provide a unique possibility for studying stratified flow. I am glad to see that scientists are taking advantage of this particular local phenomenon to study stratified flow in more detail.

 

Response: We thank Reviewer 2 for the time and effort spent reviewing our manuscript and for the thorough and detailed comments and suggested revisions, which we have carefully considered.  The conditions common to field work often do not facilitate the same high precision exhaustive measurements that we are able to perform in the laboratory environment.  We have expounded on this in detail for corresponding comments that target such limitations and restrictions, and we also note that despite the challenges of the field, the field measurements and the model results have good agreement evidenced by the level of care taken both during the field campaign and during modeling.

 

 

1. The main measured result from the current study is the vertical velocity profile in the breach/channel between the two parts of the lake. The vertical profile shows water flowing in one direction at the surface and in another direction close to the bed. One weakness of the study is that we don’t know the exact location where the measurements were taken. Measuring a profile in a channel with an ADV is usually a fairly straightforward task. Especially since there is a bridge over the channel which is possible to base the measurements from. I would encourage the authors to carry out a new measurement campaign. An ADV also gives a profile of the turbulence. This is not included in the current article. It would have been very interesting to compare measured vs. computed turbulence profiles in the channel. Turbulence and its damping is very important in stratified flow.

 

Response: We agree that velocity profiles can be straight forward to conduct.  However, we ask the reviewer to reflect on the shear size of the structure and on the magnitude of velocities passing through the west crack breach.  Simply lowering an ADV for profiling is not possible due to such challenges as accurate measurements would not be obtained.  Furthermore, the extreme salinity values present a particular challenge to instrumentation including the supporting algorithms used to interpret signals into velocity and turbulence values.  Although we agree that it would be best if USGS had specifically recorded the uplooker location and if we were able to take additional flow measurements, this is a limitation due to the field conditions and we cannot include this request in our study.  Finally, such an effort would from our perspective require more time than allotted by the journal to revise this manuscript. We would also like to point out that the continuous measurement using the ADVM uplooker is being done by a group of scientists independent from us. We are collaborating with them and for this research we use the publicly available data they generate, and any campaign to get more measurements will require more time and effort, which they are not able to do due to limited government funding and thus is outside the ambit of the current study.

 

 

2. The numerical modelling has been done with the Flow-3D software. Being a commercial product, it is more tested and less prone to bugs compared with in-house software. The downside is that the authors and readers of the article can not be completely sure which equations are solved and how the numerical algorithms work. The current article describes some of the algorithms used by the software, but not the important ones: How is the density difference computed in the Navier-Stokes equations? Which equations are solved to compute the density and the mixing of the water from the two sides of the lake? Probably, there is also an equation for the salinity solved? These equations have to be given in the paper, together with details on their solution methods.

 

Response: We have revised the methodology section to note many of the equations, etc. used by the solver. To provide a brief overview, the buoyancy effects are accounted for in the momentum balance (Navier-Stokes) equation through the body force term that is dependent on the excess density relative to the density of water south of the breach. Salt transport between the two sides of the breach is modeled using an advection-diffusion equation.

 

3. The authors tested several turbulence models initially on a 2D grid. What was the cell size of this grid compared with the final 3D grid?

Response: Eight cell sizes were studied from 0.03 m to 0.3 m.  This is very similar to the 3D case.

 

4. Turbulence modelling is especially important for stratified flow, as the turbulence is dampened by the density gradients. How was the turbulence damping modelled in the current study? This has to be stated in the paper.

 

Response: Thank you for the suggested revisions.  We have revised the manuscript with more discussion of turbulence modeling and other aspects of the model for clarity. We would like to clarify that none of the turbulence closures we used or tested accounted for the effect of stratification. Effect of stratification on turbulence is important to account for in phenomena where high Richardson number stably-stratified flow is present. Though, under these kinds of conditions, even RANS models that account for stratification effects (e.g. Mellor-Yamada) have been found to fail (Ye et al., 2013, JGRE). On the other hand, for gravity current where the primary effect of buoyancy is to accelerate the current, simulations done with traditional RANS turbulence closures have been fairly successful at the laboratory scale (An et al., 2012, Env. Fluid Mechanics; Stancanelli et al., 2018, Water). Though we completely agree with the reviewer that for simulating long-term evolution of gravity driven flows in hypersaline environments, e.g. If we want to simulate the spatio-temporal evolution of the deep brine layer, the model will have to account for the effect of stratification on turbulence induced mixing at the interface of DBL and the ambient water. For the current simulations, as the gravity evolve across a relatively short distance, the effect of stratification induced modulation of turbulence might not be dominant.

 

 

5. If the authors did not use any turbulence damping by stratification, the turbulence models would overpredict the eddy-viscosity and thereby give poor results. LES would normally underpredict turbulence on a normal grid, but because the stratification is damping the physical turbulence, then the LES model gives the best results. Could this be an explanation of the results in the current case?

 

Response: This comment from the reviewer is unclear to us regarding a specific suggested portion of text for revision to the manuscript.  It seems to be a conjecture on the LES model results.  We have however added additional text to the results discussing about the details of the LES model implemented in Flow-3D. As per our understanding of the turbulence closures in Flow-3D, all of them (including the LES) adopt different ways to calculate the turbulent eddy viscosity for the momentum balance and transport equations.

Reviewer’s hypothesis is, the reason the CFD model shows a good match with the field-data is that potential over-prediction of turbulent eddy viscosity due to lack of stratification-effect term in the turbulence closure, is being nullified by the under-prediction of turbulent eddy-viscosity due to lower than required resolution of the computational mesh. On the surface this looks as a plausible explanation. Though, the detailed exploration of this is beyond the scope of the current paper. Additionally, we would like to point out that using the k-omega turbulence closure also resulted in equally good results, albeit at a higher computational cost. 

 

6. What was the water depth in the breach/channel? Fig. 9 suggest around 5 meters. The vertical scale in Fig. 9 should be given more decimals, or the zero level should be set to the bed.

Response: We thank the reviewers for the suggestion, though It is common to report field measurements of lake levels in elevations, not in depths and a reader can easily see the depth from these elevations.

 

7. What was the y+ value close to the bed in the LES computations? This should be given in the paper together with a justification of its value. A reference should be given for the normal range of y+ values in LES studies. LES usually requires the y+ value to be in the lamninar sublayer, This is most often not possible, except for laboratory scales. The k-omega model does not have this problem. In any case, the authors should present the 2D results with the different turbulence models in the article.

 

Response: Based on the mesh size used, y+ close to the bed is around 5000, which is much higher than those used by boundary resolved LES simulations (where y+ < 10). Though these simulations are not boundary resolved LES and like RANS, the solver uses a wall model to estimate the velocity profile near the bottom. Also, as mentioned earlier the LES model in Flow-3D isn’t a traditional dynamic LES model, and is one similar to what Smagorisnky (1963) suggested for general circulation models. The simulations conducted for the 2D setup showed equally good results for the LES and the k-omega model, and we decided to use the LES model due to 30 percent lower computational cost. We thank the reviewer for this comment.  We have revised the text to discuss the 2D results with different turbulence models to compliment the 3D discussion of turbulence and corresponding details.  Please refer to the revised manuscript for details.

 

8. How was the friction between the water and the lake bed computed? Which roughness coefficient was used? Was any calibration of the friction coefficients done?

 

Response: Calibration of the friction coefficients was not done, because the simulated steady-state WSE at the breach matched well with the observed WSE, additionally the predicted velocity profile matched the observed profile. Also, the calibration using the friction factor is usually required in large river or lake models, as the bathymetry data is often not granular enough to capture all the features that can cause change in the flow profile. For the current study, the domain size is relatively small and we had access to high-resolution bathymetry data of the breach. Additionally, flow through the breach does not involve moving bedload, so we did not have any information about the bedload sediment size distribution, which we could have used to inform the friction factor in the model. Thus, we used the default value in the model.

 

9. The authors use some numerical expressions that most of the readers of the journal will not understand. These expressions therefore have to be explained in the paper: “GMRES subspace”, “interblock boundary coefficient” and “volume fraction cleanup”.

 

Response: We have revised the text to define any terms that we expect not to be known to CFD practitioners. Also, we expect that readership of this journal to include those with understanding of these common CFD terms, especially for this special issue which would attract such numerical modeling experts.

 

10. It seems that the density in the current study was only assumed to be a function of salinity. Was there a temperature gradient in the vertical direction in the lake? Or a temperature difference between the two sides of the lake?

 

Response: Thank you for this comment.  The effect of temperature on water density is quite minor but we did note any such temperature differences in this study. Sensitivity of density to salinity and temperature was checked using the equation of state for the south side of the lake produced by Naftz et al. (2011). It clearly showed that the primary source for variation in density is due to salinity. The text has been revised accordingly.

 

11. What was the boundary conditions at the sides marked P in Fig. 7? Was there an inflow/outflow of water? If not, then the CFD model did not compute a steady situation, as the water levels would eventually be the same on both sides of the breach. If a steady solution was not obtained, more information about the transient nature of the computations should be given. For example, the computed water levels as a function of time should be given (similar to the measurements in Fig. 5, but for a shorter time)

 

Response: For clarification a boundary condition specified with a water surface elevation, and allows free inflow/outflow of water. During the simulations we monitor the velocity of the water going through the breach and the water surface elevation at the breach, and the simulation is run until both these parameters reach a steady state. The WSE at the breach at steady stat was compared against measured WSE.  We have clarified this in the text.

 

12. Line 294: The word “veracity” is probably not in the vocabulary of most readers of the journal. I had to google it. The article would be more readable by replacing the word with “accuracy”.

Response: Revised as suggested.

 

13. The density profile in Fig. 12 looks interesting. Was any procedure used to sharpen vertical gradients in the salinity/density? False diffusion might easily smear the density gradients. Some diffusion of salinity/density would take place naturally. A question is how well the CFD model reproduces this process. This should be discussed in the article.

 

Response: No explicit procedure was used to sharpen the density profile. False diffusion can definitely smear the sharp density interface. This false diffusion might be induced from an unsuitable turbulence model, or from numerical errors induced due to lack of mesh-resolution or numerical diffusion due to the time-stepping method adopted. This is the reason why we tested multiple turbulence models, of which two of them (k-epsilon and RNG) were found to be too diffusive. Additionally, we also used a 2^nd order monotonicity preserving scheme for time-stepping of the scalar/density transport equation. We did it because we found that using a lower order method was resulting in smearing of the density interface.

 

 

14. Line 399: “Whiel”.

Response: Revised as suggested.

 

15. Line 401 “our of the breach”

Response: Revised as suggested.

 

Fig. 15 needs color scales.

 Response: Thank you for catching this oversight on our part.  We have revised as requested.

17. The thickness and velocity of the deep brine layer (DBL) was also computed with the CFD model. However, these computations were not verified against measurements. Although the results look plausible, the quantitative accuracy of the results are uncertain. This should be pointed out in the article.

 

Response: We agree with the reviewer that DBL thickness predicted has not been compared against measurements, but we have compared well against the velocity profile recorded by the ADVM uplooker. This gives us a fair amount of confidence about the CFD results, including the thickness of DBL just south to the breach. We have added these comments to the manuscript as suggested.

 

Line 443: “a hydraulic structure rating curve can be successfully developed for hydrologic modeling of the Great Salt Lake”. I am not sure about this statement. Although the results are qualitatively plausible, the input data are uncertain and many choices have been made to obtain the computed velocity profile. The velocities are “calibrated” by the choice of the salinity used as boundary conditions. Earlier in the conclusion, the velocities are said to be within 20 – 50 %. The accuracy of a rating curve produced by the CFD model is therefore uncertain. I suggest that this sentence about the rating curve be removed from the article. Or rephrased to “a hydraulic structure rating curve may be developed for hydrologic modeling of the Great Salt Lake”

Response: We have made some changes to text to convey an important point about uncertainty in development of the rating curve in the future. We agree that developing a rating curve using the CFD model is possible, but we have to be cognizant about the uncertainties. We would also like to politely point out to the reviewer that an R² of about 0.95 between the measured and simulated velocities for a hydraulic structure is relatively high. Also, the different choices of density difference used in the simulation isn’t for the purpose of calibration, but to account for the uncertainty in estimating densities from specific conductance data. So, a rating curve developed by varying the water surface elevation of the lake and density difference between two sides of the lake will be useful.

To have a high-quality paper, I would recommend that the field measurements are done over again, so that the location is known. And turbulence is also measured. I also recommend that the damping of the turbulence from the stratification is included in the CFD model, and the results recomputed. Then the measured and computed turbulence can also be compared. However, if the Associate Editor thinks this is not required, the authors should at least include in the article a discussion of the points above.

Response: Again, we thank Reviewer 2 for the time and effort spent providing this detailed review.  We have endeavored to address all comments where possible and have added clarifications to both the paper and also the response to Reviewer 2 to provide justification for the methodology and results of this study.  We hope that the improvements will be found satisfactory and look forward to future communications with the Editor and Guest Editors.

Round 2

Reviewer 2 Report

The authors have improved the article on may points since the last review, and also answered most of my questions. However, the newest PDF file with the manuscript contains a large number of errors in language and equations. Please read through it once more. The equations and text are repeated multiple times. It makes it difficult to read this section of the text.

 

Remaining points to improve in the article:

 

7. The authors have given an answer to my question about the y+ value in the Response to Reviewers Comments. However, this text should also be given in the article.

 

8. There is no explanation of how the bed shear stress was computed. A formula should be given. The authors state that the default friction factor was used. What was the default friction factor? This should be given in the paper.

 

11. The authors state that inflow and outflow on the sides of the WSE were computed (line 509). How was this computed? This should be stated in the article.

Author Response

REVIEWER 2

 

The authors have improved the article on may points since the last review, and also answered most of my questions.

Response: Thank you for your second review and returning so quickly (<24 hrs).  We appreciate the positive feedback after extensive revisions.

 

However, the newest PDF file with the manuscript contains a large number of errors in language and equations. Please read through it once more. The equations and text are repeated multiple times. It makes it difficult to read this section of the text.

Response: We submitted a clean word file and a markup pdf file.  The pdf would be difficult to follow due to the extensive revisions; we checked the word version from what we were able to download and it appears that this file includes the pdf track changes, so our apologies for any difficulty in review. 

Response: We do not see in the clean version the issues you noted, only in tracked changes, but we have done as requested and made a careful review of the entire paper once more.

 

Remaining points to improve in the article:

 

The authors have given an answer to my question about the y+ value in the Response to Reviewers Comments. However, this text should also be given in the article.

Response: This has been added from the response to reviewer comments.  Please see the tracked changes version for details.

 

There is no explanation of how the bed shear stress was computed. A formula should be given. The authors state that the default friction factor was used. What was the default friction factor? This should be given in the paper.

Response: This has been added to the paper also.  In Flow-3D, the surface condition may be described as no-slip, partial slip, or free-slip.  In this research the no-slip option was selected as it represents nearly all fluid-solid interfaces.  Shear stresses at solids or wall shear is modeled y assuming a zero tangiential velocity at the wall and for the first cell adjacent to the wall a logarithmic velocity profile is applied (log-law region). Flow-3D guidance prefers that the first cell include the viscous sublayer and also be well within the turbulent region of the boundary layer.  However, in this solver roughness is considered in three steps.  First smooth and rough wall logarithmic velocity expressions are merged by introducing a modified viscosity at a wall that reflects roughness and any enhanced momentum exchange caused by the roughness height.  This is then used to compute local at the wall shear stresses and these values are then used to define transport variables at the boundary for k-epsilon and RNG k-epsilon, etc.  These shear stresses are also included in the momentum equations used by FLOW-3D for mean flow velocities.  The software does make a number of checks using Re, etc. to see if the flow at the boundary is in the laminar sublayer, is a smooth or rough wall boundary, etc.  From field observations the roughness height was considered negligible due to the global geometric features adjacent and in the breach and therefore in the results this is confirmed by model sensitivity also being negligible to roughness height and good agreement shown with our final results. 

 

The authors state that inflow and outflow on the sides of the WSE were computed (line 509). How was this computed? This should be stated in the article.

Response: We have also included this in the paper.  In Flow-3D there are two types of pressure conditions, referred to as static or stagnation pressure conditions. In a static condition, the pressure is more or less continuous across the boundary and the velocity at the boundary is assigned a value based on a zero normal-derivative condition across the boundary.  When a fixed pressure boundary is requested, the mesh generator creates one additional layer of boundary mesh cells in which the pressures are defined. Three types of specified pressures are possible. If the pressure is the local static pressure, then across this extra layer of cells, zero velocity-derivative conditions are imposed similar to the continuative condition described next. If the pressure is an upstream stagnation pressure, then a zero normal velocity is imposed at the mesh boundary to approximate stagnation conditions.  However, for this study with variable density we specified a ‘regular type’ of pressure boundary, where fluid can enter or leave the domain and corresponding densities are tracked. Fluid density and all the advecting scalars are prescribed at the boundary. Hydrostatic pressure distribution at the boundary is code-calculated using the prescribed density.