This study examines spatial variability in conservative and non-conservative tracers across 50 “kettle lakes” in northern Ontario. Authors employ a combination of correlation, ordination, and simple water balance analyses to develop a “lake typology.” The authors argue that landscape position is dominant driver of lake hydrology, suggesting lakes at higher elevations are “recharge” lakes (ie they contribute to local groundwater aquifer) and lakes at lower elevation are “discharge” lakes (ie they receive water from local groundwater). Finally, authors suggest recharge lakes are more sensitive to short term changes in hydroclimatic conditions than discharge lakes.
While this manuscript is in revision, I am reviewing this manuscript for the first time. In general, I found the manuscript quite interesting and worthy of publication. However, unfortunately, I believe both the narrative structure and analyses should be developed further before publication. In particular, more information about analyses is needed in the methods section, results associated with landscape position/morphology are overstated, and conclusions about inter-annual variability are confusing/unconvincing.
Below I provide both general and specific comments in an effort to help the authors improve their manuscript.
General Comments:
1) The authors should work to further develop their narrative. In particular, authors should work to make their main points more easily accessible to readers. One approach is to structure the paper to target three different types of readers: (1) readers who will only review the abstract/conclusion/figures, (2) readers who will lightly skim discussion, and finally, (3) readers who will thoroughly read the manuscript. Because the vast majority of readers will fall into the first category, it is imperative that authors tell a coherent story with the abstract/conclusions/figures.
2) The abstract should be streamlined. (Refer to comment #1 above.) The current abstract reads like a list of facts, and provides minimal synthesis of those facts. While this is a stylistic decision, I prefer abstracts that mirror the structure of the paper [eg intro, broad objective statement, study design, 2-3 major results, and a brief conclusion that links to objective statement]. This helps the reader extract needed information, and for readers who are taking a deeper dive, it gives them a road map to refer back to if needed. Also, in general, the abstract should be constrained to one paragraph.
3) The methods are incomplete. There is no description of the ordination analysis, comparative statistics, correlation, or breakpoint analyses. While it is often acceptable to omit simple analyses from a methods section, I do not believe it is appropriate in this case. Further, this will help guide readers as they pick through the results section.
4) The description of the sampling design is ambiguous. For the synoptic sampling, 50 lakes were sampled across three sampling campaigns. Were all lakes sampled during each campaign, or were 50 total lakes sampled and only some sampled multiple times?
5) Authors should discuss the limitations of the temporal component of their sampling design. Throughout the document, the authors suggest the groundwater and precipitation observations are too similar to differentiate. I’m not sure if I agree with this statement, as it vastly oversimplifies associated isotopic fractionation and solute mixing processes. (Specifically for water isotopes, see recent review paper by Sprenger et al., 2016 in Reviews of Geophysics. Note, the detail presented in the Sprenger review is admittedly beyond the scope of this study.) While I think the sampling design is adequate for the given inference, it may be worth discussing the benefits of greater temporal resolution sampling.
6) I am concerned about the assumption dV=0 in the water balance analysis. In large lakes with consistent inflow and outflow channels, this assumption is likely appropriate. However, in small lakes without surface connections, this assumption is absolutely not valid. [As noted in the authors discussion, ephemeral lakes are actually defined by their seasonal variation in volume.] It appears the lakes sampled in this study are in-between these two endmembers. Therefore, the authors should caution readers about the assumption dV=0, and further, cite other papers that use a similar approach to develop I/E ratios to justify their methods.
7) I encourage the authors to use non-parametric statistics. (eg Use a Wilcoxon rank-sum test instead of an ANOVA.) I would wager that their data likely violate normality assumptions.
8) I would like more information about the NMDS analysis.
In the methods section, the following information should be provided [at a minimum]: What was the input data? [Was it from across all 50 sites and all three synoptic sampling events?] Was the data scaled before running the analysis? What was the dissimilatory measure? How many iterations were used to fit the model? Were the final results rotated so the dominant gradient varies along the primary axis? What statistical package did the authors use (R, SAS, PC-ORD?).
In the results section: How many axes were utilized, and why was that decision made? Did the model converge, and if so, after how many iterations? What was the resulting stress?
9) I appreciate the authors attempt to classify lakes in Figure 7. This is one of the more compelling parts of the manuscript. To further clarify the discussion surrounding the differences between higher level groups, I encourage the authors to add an additional plot that displays differences between the higher level groups. Personally, I love boxplots because of the amount of information they provide! Also, authors may consider showing differences across groups in either ordination or dataspace. (A multivariate, nonparametric test like MRPP or NP-MANOVA may be appropriate.) Further, cluster or CART analyses could be used to differentiate groups. However, just to be clear, the listed analyses are simply suggestions for improvement, and authors should not feel obligated to use them!
10) The authors’ conclusions about landscape vs local drivers of hydrology are likely overstated given the sampling design. The authors develop a compelling argument that landscape position (and associated geologic setting) drive lake hydrology. However, their analysis of hydrogeomorphic features at each site is very limited due to data availability. Important site-specific variables like network order (ie how many lakes drain into the lake in question), contributing watershed area, and soil characteristics like specific yield can all be used to explain observed variability in hydrologic data. While it is understandable/acceptable that authors did not collect these variables, I think they are over reaching by saying landscape position is more important than local hydrogeomorphic characteristics. I would encourage the authors to add a caveat to this statement, highlighting measures that could be used to better explore local geomorphic drivers.
11) The results presented in figure 8 are quite confusing. The discussion should restructured in such a way that clearly delineates annual and inter-annual variability in ET and E/I ratios. As presented, the conclusions (L708-712) that recharge wetlands are more sensitive to climate extremes that discharge are unconvincing. I would encourage authors to present a conceptual model of annual variability for both recharge and discharge wetlands, highlighting how results presented in Figure 8 support the proposed model. Also, consider visualizing this information using another format.
12) Finally, I would encourage authors to explicitly link their results to management activities in northern Ontario. What information from this study will be useful for managers? Maybe even write this section for them. Authors begin to do this when describing mining activities. However, there is no discussion how the “cottage development” industry could use the information derived from this study.
Specific line-by-line comments:
L33 Abstracts should be 1 paragraph.
L72 Typically, anions such as Cl and Br are considered conservative tracers. While they may not exhibit conservative behavior in all settings, they are much less reactive than tracers like DOM, N, and P.
L102 This is an odd place for a comma
L103 Lake typology is a new term for me, and potentially for other readers as well. Maybe consider defining it here?
L130 Please rephrase. “The esker are” is somewhat ambiguous, especially since the term eskers has not been defined.
L164. How were features digitalized in ArcGIS 10.3 from google earth? I think I know what you mean, but please be more explicit here.
L171-173 I’m not sure what this is referring too. Please rephrase to be more clear.
L233 How far away were these samples taken from the study area. Note, it’s acceptable that precipitation samples were taken “off-site” given the level of detail/inference in this paper. However, this is still an important detail.
L251 Please define hydrologic and isotopic steady state, and then use previous studies to confirm the validity (and limitations) of those assumptions.
L326 Please add a parenthetical designation of “E/I” ratio
L361 Please describe the breakpoint analysis in more detail. This detail should likely go in the methods section.
L522 Again, please describe logistic regression in the methods section.
Figure 1. Maybe add a line to map to indicate the potential location of conceptualized cross section presented in b
Figure 4 and 5. Authors should consider flipping their axis.
Figure 6. How were these grouping boundaries defined? Are you connecting points at the fringes, or are these polygons the result of an analysis? |