An investigation of anthropogenic influences on hydrologic connectivity using stress tests

Herzog, Amelie; Hellwig, Jost; Stahl, Kerstin

doi:https://doi.org/10.5194/hess-2023-243

Preprints

https://doi.org/10.5194/hess-2023-243

Preprints

18 Oct 2023

| 18 Oct 2023

Status: a revised version of this preprint is currently under review for the journal HESS.

An investigation of anthropogenic influences on hydrologic connectivity using stress tests

Amelie Herzog, Jost Hellwig, and Kerstin Stahl

Abstract. Worldwide, human influences directly and indirectly threaten environmental flows through groundwater (GW) abstraction. This highlights the need to consider GW withdrawals together with climatic changes in future water management plans to maintain water availability in river networks. In alluvial geological contexts, such as the Dreisam River in SouthGermany, contributions from GW often sustain streamflow in the summer months. In the specific case of the Dreisam however, several hydrological drought events between 2015 and 2022 lead to interruptions of longitudinal connectivity in the stream network. This raises the question on where and when the stream network is gaining or losing water from the GW and how these vertical connectivity changes influence streambed drying. This study therefore aims to analyse both, how changes in longitudinal and vertical connectivity in the Dreisam valley respond to stresses from climatic variations of recharge and to anthropogenic water abstractions from the hydrological system. As GW-SW interaction is difficult to measure, numerical groundwater modeling was used to obtain a spatial distribution of the exchange flow between GW and SW. The results show in which stream reaches, the connection between GW and SW ceases during dry conditions. Changes of vertical connectivity due to GW abstraction were found to be stronger than due to recharge stress on short timescales. A combined analysis of vertical and longitudinal connectivity depicts local points along the stream network, where the effect of GW abstraction on temporal drying dynamics is likely particularly strong. These results have to be interpreted within the limits of model reality and uncertainty. Simulated zero water levels were in good agreement with measured zero water levels at 50 % of measurement locations, whereof the majority was in the valley bottom. Future work needs to improve coupling to upstream contributions and bedrock aquifers along the hillslopes of the valley. Overall, our findings highlight the value of a combined analysis of different dimensions of hydrologic connectivity for the evaluation of model results. Approaches that better distinguish locations affected by natural and anthropogenic drivers of hydrologic drought and streamflow intermittency deserve further development and are needed for application on different spatial and temporal scales.

Received: 16 Oct 2023 – Discussion started: 18 Oct 2023

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.

Download & links

Amelie Herzog, Jost Hellwig, and Kerstin Stahl

Status: final response (author comments only)

RC1:
'Comment on hess-2023-243', Anonymous Referee #1, 06 Nov 2023

Herzog et al. Study the longitudinal, vertical, and lateral connectivity of GW and SW in the Dreisam valley. The study is nicely done, generally well described and scientifically interesting. Mainly, I suggest adding some more explanation and discussion to better guide the reader and thus better help her to understand the storyline presented. I generally like the study and have no major comments beyond the constructive suggestions made below.
[1] There are a few editorial changes needed regarding the English. An example from the abstract: “This raises the question on” should be “question of”. Or “By reason of the physically-based” would better read as “Due to the phy…”. Please have another read through the document for these instances.
[2] In the abstract, “model reality” should probably be “model realism” to be more in line with the language used in other papers (e.g. Gharari et al. 2014 HESS; Hrachowitz et al. 2014 WRR; Wagener 2003 HP).
[3] I like the abstract, but could you quantify terms a bit more. E.g. what is a short time scale in the context of this study? How many measurement locations did you consider? Etc. This is all discussed in the manuscript but might make the contributions of the study clearer right away if mentioned in the abstract.
[4] It might be interesting to connect the metrics discussed, developed, and estimated in this study to the metrics (often called signatures) used in other studies. E.g. the recent study on reservoir impact in the UK by Salwey et al. (2023, WRR).
[5] The authors state in lines 69ff: “While such parameter uncertainties are relevant when it comes to obtaining the best model results, they are less relevant if the focus is on process understanding.” I do not agree with this statement. Understanding which parameter dominate system responses, and what preferred values they take when they do so, has long been part of assessing models regarding their physical realism (e.g. Reusser et al., 2009, HESS). So dismissing parameter uncertainty as a simple problem of model performance is really understating the problem. I therefore would expect a discussion of the potential influence of parameter uncertainty on the study outcomes in the conclusions or discussions sections. Even if a more detailed analysis is not feasible in this study.
Late in the paper, the authors stated “The third research question addressed the sensitivities and changes of modelled connectivities in response to the applied stress test scenarios.” Is this question really completely unrelated to parameter uncertainty? I can accept if the authors cannot add this element to this study, but a basic discussion of the potential influence would be good.
[6] I am afraid that I am a bit lost when looking at Figure 5. The super short caption is hardly helping me to understand what I am looking at here. The text discusses gaining and loosing conditions. Maybe making those explicit in the figure would be a start? More info please.
[7] A general comment after looking at the next figure. Can you please make the captions more extensive. It is a bit annoying to have to go through much of the paper to look for abbreviations, variable names, location details etc. to understand figures. Please make the captions much more detailed so that the reader does not have to go through the text to understand the figure content. Or at least tell the reader exactly where to find the info needed to interpret the figure.
[8] Regarding the conclusions. I understand that the authors discuss what they specifically learn about their study region. However, it might be nice to add a short paragraph on what innovations, understanding or questions might be transferrable to other studies. What outcomes are general?
[9] As future work, the authors might want to consider a broader sensitivity analysis which could include both the stress test to the system as well as uncertainty in parameters or other model inputs. That would create a generic framework for analyses of the type presented here.

Citation: https://doi.org/10.5194/hess-2023-243-RC1
- AC1: 'Reply on RC1', Amelie Herzog, 22 Dec 2023
  
  We would like to thank you for your constructive comments and feedback on this manuscript. We think that the suggested revisions based on the Referee’s comments will certainly improve the article. Please find our responses to the main points in the attached document.
  
  Citation: https://doi.org/10.5194/hess-2023-243-AC1
RC2:
'Comment on hess-2023-243', Anonymous Referee #2, 06 Dec 2023

Review Herzog et al (HESS-D).
I have read this discussion paper with great interest. I do not doubt the science that went into this study. However, the way it is presented is rather unsatisfying. This feeling covers all aspects of the paper, specifically how the methodology and results are presented as well as the interpretation of results and conclusion.
Overall I am a bit puzzled about how I should interpret this study. It starts off with a larger context. But, much more than a ‘very general, first attempt to model SW-GW interactions in the Dreisam valley (L422-424)’ it is not. I am questioning how much we learned ‘to help gain insights into factors to consider in future modeling studies’ (L424). What are these insights specifically and to what extent are the findings from this study transferable to other studies? (and is that not already done, for example, in the larger scale groundwater-surface water models available?).
I regret there is no proper sensitivity analysis done of at least the most important parameter settings that impact the groundwater-surface water interactions. However, I do understand that this is probably not something the authors want to add to their analysis at this stage. I found all conclusions on uncertainties of the simulated leakage, because of the very limited model evaluation, somewhat hard to interpret. I have several points of concern:
[1] Model concept. There are several aspects that I cannot fully understand. (L118) ‘We used a combination of..’; I am not sure what this means. Are the models coupled in a way, or does one model use the outputs of the other, or did you use a combination of model results of the analysis? Not clear. (L123-124) ‘Surface water fluxes, well extractions, and recharge are added in the form of boundary conditions’. How does this work in Modflow6? As far as I know, you have to translate surface water fluxes to heads and use it as a general head boundary in Modflow. To be able to calculate heads from fluxes there needs to be some calculations using assumptions on river depth and width. What was used here? For wells in Modflow, you can define this as a specific pumping rate or specific head. Recharge is probably used as an upper boundary and as an input flux. Is capillary rise considered as well? Not enough information is provided to understand how the Modflow model is built and how boundary conditions are implemented. (L129-136): ‘the model’ is this referring to the smaller Dreisam model or the bigger model? ‘requires spatially distributed parameters determining’ what kind of parameters? Determining surface and subsurface flows so I am assuming soil parameterization, elevation etc but this is not mentioned anywhere. (L134): as the aquifers are unconfined (at least I did not read otherwise) does it make sense to use specific storage (neglectable of unconfined conditions) compared to specific yields? (L156-157) ‘based on Manning’: I think this needs a bit more explanation (or reference to the original code and/or coupling between surface and groundwater).
(L142-158) A general remark for this whole section is that there is not a clear structure, especially not in this last part. It might help to show a small model conceptualization or flow chart: what goes in (parameterization, forcing) and what comes out (discharge, which is then used to calculate river heads), etc. I truly recommend going through this section again and making sure everything is very clear: which input data and where did you get it from, what is calculated in which model component, and how does it feed into the next model component? A proper understanding of the model setup is crucial to understanding and interpreting results as a reader (also, specifically as no sensitivity analysis is performed on any parameter settings. Then at least be clear and open on how you construct and parameterize your models).
Some additional questions: the RoGeR model was at daily resolution, how about the groundwater model and the river routing? Same for spatial resolution? And vertical discretization of both the hydrological and groundwater models is also not that clear to me. Human interactions are implemented by groundwater extraction only. And is there a return flow to surface water and/or groundwater I cannot follow that either.
(L142-150) More generally, I have some trouble understanding the term ‘leakage’ in this study. As I understand it now, it is used here to describe groundwater drainage and river infiltration. For me, leakage would mean more the unintended movement or loss of groundwater from its natural subsurface flow and not the dynamic interaction between groundwater and surface water that changes due to groundwater pumping. Maybe the terminology is something to verify with a groundwater expert as well.
[2] Evaluation of model results. Similar to the previous I have many questions. More in general I find it hard to understand what ZWL represents. A conceptual figure just showing a connected groundwater-surface water system and a disconnected system might already help.
Another more general comment, in the writing you write ‘validation’ while the heading reads ‘evaluation’. Reading this section I am wondering if validation should not be replaced by evaluation in this section. L166: ‘We preselect a set of stations’ How did this pre-selection go, or is that what is described next? (if so then use something like ‘to this end’ instead of ‘hereby’). (L167-170) I cannot follow this. L172: ‘calibrated’: there was nothing on a calibration before.
[3] Stresstest scenario definition It is not completely clear what goes into the first stress ‘changes in groundwater recharge’. Is this driven by climate input only or also by varying soil parameterization? I found the use of scenario somewhat misleading as ‘scenario’ refers to projections and potential future and the analysis is on the recent past (or current climate). So analysis or assessment might be a better term.
Check throughout the manuscript how you refer to these tests. I also read the stress test (without scenario) this heading can then also read the stress test definition.
[4] validation of zero water levels. The first finding is that simulated groundwater levels (heads or depths?) are underestimated (too deep or too shallow?) but that this does not impact simulated discharge. The underestimation has to do with the parameterization of the river bed. It is not clear which part of the parameterization is impacting the results. Probably the drainage level (aka the boundary condition) is meant here as the river bed resistance (for example) does not impact groundwater heads that much. Results are not well explained here. More in general in the result section (here and other pasts) there are several parts that are redundant and describe methods instead of results, for example :L216, L223-226. L279-271.
[5] L250: ‘leakage’ is the same as ‘leakage flow’ that was used before in this Alinea. And can you explain why simulated leakage flow is highly uncertain for areas with increasing flows? How did you evaluate the level of uncertainty? L252 “Which conditions they may change’ only considering limited model choices of groundwater pumping and recharge right?
[6] Figure 4: I do not fully understand this figure. What do the violins represent? Green is the reference (and not a green dot) and orange is the well scenario (and not an orange box) (see legend). What are the percentiles representing? What is the threshold? Where do ‘losing’ and ‘gaining’ come from? I can understand this but throughout the whole methodology and result this terminology was not used (but positive and negative leakage).
[7] L335-344: I disagree with the argument for not doing a sensitivity analysis. Also for a general interpretation of the result a sensitivity analysis, and varying parameter settings of a few key parameters, would have been useful to better understand what we learn from this modeling experiment (for example look at what is done in large-scale modeling studies). Also, I am a bit skeptical about the ‘reduce calculation times’ argument. How long does your model run the calculation times do not increase in when you re-run your model for different parameter settings, it is just more work. I would strongly recommend to rewrite your argumentation for not doing a sensitivity analysis or simply don’t bother to explain.
Minor comments
[8] L105-107. In this sentence, it is not clear if ‘baseflow’ is the groundwater that is released to the stream or the constant flow in a stream that comes from the gradual release of groundwater. Also, what follows is confusing ‘the degree of connectivity’; connectivity of what? Also, this suggests you discuss the groundwater discharge to the stream that contributes to rivers’ baseflow. I recommend being as clear as possible on the terminology, and baseflow is a difficult term.
[9] writing in general. I recommend an English language check to correct grammar and common writing mistakes. For example, linking sentences together with ‘and’ where the start of a new sentence would be preferred; referring to previously mentioned aspects with ‘this’ or ‘these’ where it is not always clear where it refers to; misuse of commas and often no use of comma’s where comma’s are needed for the readability. Sometimes the meaning of the sentence also completely changes when a comma is not placed (I had the read some sentences several times to understand a comma was missing). The level of the manuscript will increase significantly if the writing is improved.
L128: To represent both the surface and surface systems
L123: well extractions à groundwater extractions
L127: “The extension of the model domain”à the first part is confusing: the model domain covers.
L129 ‘< comma> as well as’
L131-133: ‘The percolation … ‘: a clear example where commas should be placed or new sentence should start as now it reads as if you sum up recharge and the sum of interflow and overland runoff.
L133: ‘time invariant’: constant or static is a more common way of writing.
L160-161: leave this out.
L162: A direct validation is not possible
L164: If I understand it correctly you compared observed water table heads and streamflow to simulated values. Thus, ‘negative outliers of stream stages …. For gw heads far below the surface only concerns the simulated values? Not clear from the writing.
L245 is slightly more positive (meaning leakage in the natural system is lightly higher)
L255-256: to stay consistent with the unit provided for the area. Also, a unit for leakage flow and specific leakage should be given.
L257 (setting the min…. ); I do not understand this. Is L271 similar (I can understand the latter).
L274 modelled à simulated.
L318: is it, not the other way around? Because of the resolution of your modflow model, you are not able to represent your drainage at a level of detail needed to accurately estimate the infiltration of losing streams. Which results in an underestimation of groundwater heads (or overestimation of depths).
L321-327: this section hints at a hillslope effect but does not explain anything. (it turns in circles).
L328: are weirs and bridges impacting the river bed to such an extent that it will impact your modeling more than, for example, the parameterization you use for riverbed conductance and the uncertainty related to that?
L332: not ‘would be’: a sensitivity analysis will be meaningful (and will be needed to properly interpret the modeling exercise.
L427 “modeled leakage’
L430: I have read this before, but here it says ‘Due to the uncertainties we are not able to investigate the change in the magnitude’. Of course, you can study the changes in magnitude. You are not able to fully interpret the results, not because of the uncertainties but because of the lack of sensitivity analysis and/or observed or more reliable modeled data.

Citation: https://doi.org/10.5194/hess-2023-243-RC2
- AC2: 'Reply on RC2', Amelie Herzog, 22 Dec 2023
  
  We would like to thank you for your constructive comments and feedback on this manuscript. We think that the suggested revisions based on the Referee’s comments will certainly improve the article. Please find our responses to the main points in the attached document.
  
  Citation: https://doi.org/10.5194/hess-2023-243-AC2

Amelie Herzog, Jost Hellwig, and Kerstin Stahl

Viewed

Total article views: 542 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
393	114	35	542	21	24

HTML: 393
PDF: 114
XML: 35
Total: 542
BibTeX: 21
EndNote: 24

Views and downloads (calculated since 18 Oct 2023)

Month	HTML	PDF	XML	Total
Oct 2023	166	39	6	211
Nov 2023	40	9	3	52
Dec 2023	33	18	9	60
Jan 2024	24	8	1	33
Feb 2024	17	7	2	26
Mar 2024	29	14	4	47
Apr 2024	30	7	4	41
May 2024	13	6	3	22
Jun 2024	41	6	3	50

Cumulative views and downloads (calculated since 18 Oct 2023)

Month	HTML	PDF	XML	Total
Oct 2023	166	39	6	211
Nov 2023	40	9	3	52
Dec 2023	33	18	9	60
Jan 2024	24	8	1	33
Feb 2024	17	7	2	26
Mar 2024	29	14	4	47
Apr 2024	30	7	4	41
May 2024	13	6	3	22
Jun 2024	41	6	3	50

Viewed (geographical distribution)

Total article views: 523 (including HTML, PDF, and XML) Thereof 523 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 29 Jun 2024

Short summary

We investigated the influence of anthropogenic water withdrawals and climatic changes on streambed drying the Dreisam valley (24 km²) in Germany by means of physical modeling. The model was able to reproduce streambed drying at 50 % of locations. Our results show that groundwater withdrawals intensifiy riverbed drying on short timescales during phases with low groundwater levels at specific locations. At some upstream locations however, climatic preconditions are the primary influencing factor.


Total:	0
HTML:	0
PDF:	0
XML:	0