Review of hess-2024-1503

Title: Novel Statistical Analysis Illustrates Importance of Flow Source for Extreme Variation in Dissolved Organic Carbon in a Eutrophic Reservoir in the Great Plains, by Baron et al.

Baron et al. present long-term (1990-2019) chemical and hydrological data from the Buffalo Pound Lake, a drinking water source lake in the Canadian prairie region. By using novel statistical analyses, they aimed to find drivers of DOC concentrations in the lake at various temporal scales. Upstream regulated flow and several chemical parameters accounted for most of the variation in lake DOC concentration. They conclude that both flow regulation and natural processes in the face of a changing climate pose important challenges for drinking water treatability.

This was an interesting read. Investigating drivers of DOC concentrations at short- and long-term scales is a recurrent but a relevant topic, more so in atypical areas such as the prairies. The manuscript is well-written, and appropriate and novel statistical methods have been used, which are well presented and justified. Yet, a few concerns and quite a few specific comments, which are probably not major overall, but would require some work before the manuscript can be accepted for publication. I therefore suggest the authors to consider my comments and amend the text accordingly or rebut.

General comments

The hypothesis that “changes in lake water chemistry would impact DOC at shorter timescales” is ambiguous. Changes in lake water chemistry can relate to processes happening in the catchment, which would be the ones driving both overall lake water chemistry and lake DOC concentrations (these processes would relate more to allochthonous sources of DOC). But changes in lake water chemistry can also relate to internal processes in the lake, which in turn can drive DOC concentrations (these processes would relate more to autochthonous sources of DOC). I would like to see more explicit hypothesis considering whether both or one of the group of processes are expected to be important. I would also like to see this differentiation more explicitly made throughout the discussion.

In relation to that and as much as I would think it should be the case, the lack of relevance of the local catchment flow (Q-LC) to explain DOC concentrations in the lake appears to imply that in-lake processes are more important (?). The authors should reconcile this observation with the explanations they provide that argue that catchment processes drive DOC concentrations in the lake under certain conditions.

In relation to that, I am left unconvinced of the mechanisms/processes/situations that relate to high DOC concentrations in the lake. Indeed, the statistical methods that the authors use generally fail at the upper range of DOC concentrations. On one hand, I would like to see a more explicit explanation of the circumstances that lead to higher DOC concentrations, and on the other hand, the authors should acknowledge at this upper range their analyses did not provide a satisfactory answer.

The statistical methods were well-presented and justified. Yet, their results are difficult to follow at times. I would appreciate if the authors provide more analogies to how results would be presented when using more common methods. For example, how do predictors of DOC relate to DOC? Are they “positively” related, “negatively” related, something else? This is not clear in the text. Maybe an extra column specifying this in Table 2 would help?

In some parts, the connection of the discussion to the actual results was not fully clear. Please, consider making clearer links in this regard.

Specific comments

Title

Shouldn’t it be “Novel Statistical Analysis Illustrates the Importance of Flow Source for Extreme Variation in Dissolved Organic Carbon in a Eutrophic Reservoir in the Great Plains”. That is, please include “the” in front importance.

Abstract

L. 9. It would make sense to clarify that these trends have been overwhelmingly positive trends.

L. 10. They might not be universal, but I think there is little doubt that the prevailing driver was the decline in sulfur deposition and consequent increase in organic matter solubility.

1 Introduction

L. 25-38. When describing “the debate over the factors that govern DOC concentrations”, one must consider that such drivers operate on varying temporal and spatial scales (see e.g. Clark et al., 2010, doi: 10.1016/j.scitotenv.2010.02.046). Thus, drivers are not necessarily exclusive, they might just be dominant at different temporal and spatial scales. Elaborating on my previous point, there is little doubt that, at the long-term scale, increasing DOC trends observed across vast regions in the Northern hemisphere affected by acid deposition were driven by, indeed, reductions in sulfur emissions. This is especially true in smaller, forest headwater catchments. Areas affected by varying chloride or nitrogen deposition (mentioned in the paragraph) would behave similarly as they would trigger the same chemical effect on organic matter solubility and I therefore would consider them as analogous, not differentiated, drivers. Areas less affected by acid deposition of any kind where other drivers might come into play might of course show other patterns.

L. 45-46. Here you use both, catchment (rather UK English) and watershed (rather American English). Just use one of the two here and throughout the manuscript (it appears that you mainly use catchment so use that at every instance).

L. 47. Are you referring to DOC exports or to concentrations here? You already mentioned before that “DOC export is highly correlated with precipitation and annual runoff”, which is true and rather uninteresting because it is self-evident given that export = runoff x concentration, and runoff generally varies across a much wider range of values than concentration does.

L. 51-53. This is a very important point that I was eager to see. Do you have a reference to back this up? My perception of this system is that most of the area is hydrologically non-effective, i.e. I find half to be a low estimate.

L. 70-71. But changes in lake water chemistry are concomitant to changes in DOC and therefore not necessarily drivers of DOC in the lake, i.e. they also depend on hydrological connectivity with the landscape and upstream sources, on processes occurring in the catchment, etc. Or you mean that in-lake processes are important for driving DOC concentrations?

2 Methods

L. 78. This is just out of curiosity for my own understanding. Can the climate of a region that receives only about 300 mm of annual precipitation still be classified as “subhumid”. I would consider that in the range of arid or semi-arid regions. But probably the evapotranspiration is very low too despite the warm summers?

L. 86-87. Interesting and important remark. However, I find the sentence oddly constructed (“contributes flow in 1:2 runoff years”?). Can you rephrase?

L. 95-97. Can Lake Diefenbaker keep up with the demands from Buffalo Pound Lake under all circumstances?

L. 100. In Figure 1, I assume Ridge Creek is a small tributary into the Qu’Appelle River (you also describe it in the text as such). It would therefore be helpful to represent it in the figure as a lotic water system the same as e.g. Iskwao Creek, i.e. with a blue line.

L. 109-123. I assume water samples are filtered before they are chemically analysed. What is the pore size of the filter?

L. 125-126. Required a complete record at what temporal scale? Monthly, as implicitly suggested? Please, specify.

L. 137-174. I very much appreciate the effort to get the hydrology right and the consideration of water mass balances and catchment (effective) contributing areas. There is just one thing I am not sure I understand. How come Q-BP (the inflow to the lake) that is very much influenced by Q-LD (the outflow from Lake Diefenbaker, which is outside the catchment area of BP) is included in equation 4 that attempts to estimate only the local catchment flows? Shouldn’t Q-BP be Q-U (the ungauged portion) in this equation?

L. 176-179. Perhaps, remind the reader here that, for this analysis, you are using the monthly values that you estimated earlier.

3 Results

L. 265. Aren’t both lakes covered with ice?

L. 263-272. Are typical peaks across the three Q generally associated with snowmelt events, or also with rainfall events?

L. 302-303. Are all these significantly coherent relationships found analogous to positive correlations or there are any negative correlations too? Is this something that can be said at all at this point? Either way, I think it is important to specify this for the reader.

4 Discussion

L. 359-360. How is climate having an overriding influence on DOC concentration? You have not analysed any climatic variable.

L. 355-388. This section makes an interesting description of the general context of BPL, but how does it relate to your results? I fail to see the connection.

L. 390-405. Let me see if I understand this correctly. Q-LD would have a “negative” relationship with DOC concentrations at BPL, meaning that when it is the prevailing source of water to the lake (that is when the catchment is generally hydrologically disconnected), DOC concentrations are generally low. By contrast, when the catchment does hydrologically connect to the lake via the activation of e.g. Ridge Creek and Iskwao Creek, you would expect to have higher DOC concentrations from organic matter-rich catchment sources. However, you were not able to see this through your analyses. Is my interpretation correct? And if so, how is all of this reconciled?

L. 406-412. I would appreciate here if you’d explicitly mention whether these synchronous or lagging patterns imply, in each specific case, that DOC and the corresponding chemical parameter both increase, both decrease, or they go in opposite directions at the time scales considered.

L. 411-412. Still, local catchment flow was not a predictor of DOC concentrations in BPL.

L. 415-416. Following a previous comment, it can be that DOC concentration in BPL is linked to other chemical constituents in the lake, but is it really driven by in-lake chemistry itself? On the other hand, you did not find a relationship with Q-LC. All to say that I am having difficulties reconciling all these results so I would appreciate if you can make it clearer.

L. 423-444. First, these are very high concentrations of sulfate compared to what I am used to in other natural environments. I assume this can only be explained by the geological settings of the region containing large amounts of gypsum and pyrite, i.e. the ultimate source of sulfate in the catchment and the lake should primarily be mineral weathering. If this is correct, please make it more explicitly clear in the text. Second, if I understood it right from section 3.3 and Figure 4d, sulfate has a complex relationship with DOC, where above certain threshold sulfate and DOC are “negatively” related and below this threshold they are “positively” related (excluding the upper sulfate concentration range where the model did not perform well). Is this correct and if so, how do you interpret it? I miss this explanation in this discussion.

L. 441-444. This might be the case, but how would this drive DOC concentrations in the lake? You need to provide support for in-lake control of DOC concentrations, if that’s one of your lines of argumentation, which is still not clear to me.

L. 449-450. But yet again, Q-LC was not related to DOC concentrations in BPL.

L. 451-473. Maybe these in-lake mechanisms are of greater importance than the catchment input mechanisms given the lack of relevance of Q-LC? I don’t know, I am sceptical about that, but I am worried about the lack of explanatory power of Q-LC. In any case, you should be more explicit in differentiating catchment processes that can drive DOC concentration in the lake via allochthonous sources, and in-lake processes that can drive DOC concentrations via autochthonous sources. And once you make that differentiation clear, it would be best if you argue for either one of them with more conviction.

L. 480. This was not entirely clear to me according to your results.

L. 475-510. What I take from here is that this is a very challenging system in which no scenario is easy to manage. Would you provide a more explicit description of the conditions that would be best for both ecology and industry, even if they are not “natural” and potentially infrequent?

5 Conclusions

L. 5113-514. I would very much agree with this a priori, but given the lack of explanatory power of Q-LC, can you still claim that “pulses of allochthonous DOC from the local catchment during wet periods are linked to higher DOC”?

General comments

This manuscript evaluates a long time series of lake water quality to learn about controls of short- and long-term changes in DOC concentrations. This is a good fit for HESS and of interest to a broader water science community.

I like the methodological approach and find results convincing. I have, however, some major concerns: The introduction does not clearly lead to the objectives, the results need improvement on their presentations and the discussion often lacks the clear relation to the results. This needs improvement.

Moreover, consider this thought:

The model describes the dependencies of a set of water quality variables but is partly looking at a hen-egg problem. Are high DOC concentrations triggered by high TP concentrations within the lake or are they responding to the same water source coming into the lake? The GAM cannot clarify cause-effect relationships – so it can be used for system understanding but not for predictions. This can be made more clearer in the manuscript.

For more specific results see my comments below.

Specific comments

Abstract

L9: Better be specific and name it “concentration“ trends or “flux” trends, if this is the better fit.

L14ff: While the title puts stress on the novelty of the statistical approach, the abstract is not doing this. Would it make sense to add this aspect here?

Introduction

I am not fully convinced by the line of argumentation in the introduction that does not define the lack of knowledge that is addressed by the objectives. Similarly the methodological approach is not clearly motivated by the problem that is addressed here. Finally, the hypothesis is not grounded in the state-of-the-art knowledge described before.

L29: I am not sure what “across systems” means here? Across regions? Across water compartments?

L32f: If “across regions” is meant, Temnerud and Winterdahl are maybe not the best fit as a reference as they look at Sweden only.

L35ff: For hydrological changes this would be a potential additional reference: 10.1002/2017GB005749, for Nitrogen deposition this one: 10.1111/gcb.13758

L47: Is landscape complexity the best term here? I understand landscape complexity more as the complexity and heterogeneity of a given landscape and not of complexity across landscapes. This may be termed more clearly.

L59ff: DOM (why not DOC?) as a “master variable” and “limnological behavior” deserves more explanation.

Methods

L84: What is the dominant land use here and is that related to the nutrient rich soils? Are there people living in the catchment and what happens to their wastewaters?

L85ff: Does the non-contributing areas are totally disconnected or connected via a groundwater pathway?

L90ff: Is there any regulation of water level/ flow at the dam of Buffalo Pond or is everything managed upstream only? Additionally, Eyebrow Lake is not mentioned in the text. Does this lake play a role in this complex lake system?

L97f: This last sentence does not make sense for me when not underpinned with facts and referenced.

L109ff: I miss information how samples where taken and how often. How many samples are averaged on the mentioned monthly base and how was averaging done?

Fig. 1: No need to write where the middle of the lake is when coordinates are given on the axis. Maybe mention the yellow intake point in the captions and define its abbreviation.

L138ff: I am a bit puzzled by the flow reconstruction. Why is BP inflow the reference point since the catchment contributing to the water in the lake seems to be larger (Fig. 1, areas contributing downstream of BP inflow). This needs further explanation. I note that in lines 166-174 there is a section on ungauged flows downstream of BP inflow but it is not explained how this QLC was used in the analysis. I note that this is part of chapter 2.4. … for the sake of understanding I suggest to first describe what is needed for the analysis and then to describe how this data is constructed.

L184f: This text would profit from an explanation why GAM has been used. I note that this is described later but you also justify the use of wavelet analysis here – no reason not to do it for GAM as well.

Results

L262: Consider a different header here. “Temporal parameters” sounds not too good for me.

Fig. 2: At this multi-annual scale it is hard to see the timing of the seasonal dynamics. Panels a and b are, to my understanding, managed flows to meet the water demand while panel c is a natural seasonal dynamic. Any idea how to show these differences? Maybe by a plot as day of the year in the SI? Consider to use the same y-scale for all discharge plots.

L273: DOC concentration is described after discharge in the text but shown in Fig. 3 after showing all other constituents. Maybe show DOC (as the master variable here) earlier?

L300-301 and 303-304: Two sentences saying the same thing here. Maybe combine both.

L325ff: The text reads as if it is given that there is a clear driver-response relationship between predictors and DOC concentrations. However, for the constituent you partly look at a set of potentially connected variables. E.g. TP and Chl a can both describe algae biomass. NH4 may occur because algae break down. DOC may be excreted by algae. TP may decrease when flow increases due to a dilution effect of wastewater sources… All these interactions mean that predictors are not independent and you partly look at a hen-egg problem. This is more part of the discussion but I suggest to spent effort in this text to avoid this clear driver-response style of writing.

L339: What is the <~7 mg/L referring to? Root mean squared error?

Fig. 5: Consider to keep same colors for same constituents across the figures. Impressive fit of the observation by the way.

Discussion

L355ff: This chapter 4.1 reads like a summary and conclusions. You make statements that are justified in later chapters. I find it more appealing for the reader to first argue and discuss and then make statements. From this text alone, the reader does not know the basis of your statements (which analysis the statements are referring to and how they were interpreted).

L369: Not a good sign when the discussion is the first time when the reader learns that there is agriculture in the catchment.

L384: Same is true for the lake residence time – this needs to come earlier.

L392: This needs to be also part of the site description.

L394: That DOC can be allochthonous and autochthonous in lakes should be more explicitly part of the introduction.

L395ff: Have you tried a DOC mass balance? This could strengthen the argumentation here? How high need concentration in the QLC be to be visible in the lake when most of the water is coming from LD?

L403ff: It would be helpful to learn about correlation among the predictors. You state that water inflow to the BP lake comes often from LD. So, how is QBP correlated with QLD and what are cumulative fractions of the lake water balance (eq. 1)?

L427ff: This section reveals a large number of formerly undescribed data. I see that this data is not your own result but it may be helpful if there is an overview on average water quality from these unpublished sources as a table somewhere above and/ or the SI.

L445ff: This section describes relationships between DOC and TP as source-driven mainly. Are there additional in-lake mechanisms as well? Joint release of DOC and TP from lake sediments under iron-reducing conditions? Or is this what you mean in L450f?

L480ff: Brining in new data here in the late part of the discussion is not good. Your discussion should be based on everything you describe in the results plus literature.

L475ff: The whole chapter 4.3 seems to be very detached from the results and discussion above. What of this information is directly related to your findings. I suggest to shift part of this to the problem definition in the introduction, to try to discuss how your results help with water management and omit the rest.

Conclusions

L519: This reads as if the upstream lake flushes DOC into the lake BP. Better use “diluting” here and name the source of the high DOC water.