A large sample analysis of European rivers on seasonal river flow correlation and its physical drivers

Iliopoulou, Theano; Aguilar, Cristina; Arheimer, Berit; Bermúdez, María; Bezak, Nejc; Ficchì, Andrea; Koutsoyiannis, Demetris; Parajka, Juraj; Polo, María José; Thirel, Guillaume; Montanari, Alberto

doi:https://doi.org/10.5194/hess-23-73-2019

Articles | Volume 23, issue 1

https://doi.org/10.5194/hess-23-73-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-23-73-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 23, issue 1

Research article

|

07 Jan 2019

Research article |

| 07 Jan 2019

A large sample analysis of European rivers on seasonal river flow correlation and its physical drivers

Theano Iliopoulou, Cristina Aguilar, Berit Arheimer, María Bermúdez, Nejc Bezak, Andrea Ficchì, Demetris Koutsoyiannis, Juraj Parajka, María José Polo, Guillaume Thirel, and Alberto Montanari

Download

Final revised paper (published on 07 Jan 2019)
Supplement to the final revised paper
Preprint (discussion started on 03 Apr 2018)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review of paper: A large sample analysis of seasonal river flow correlation and its physical drivers by Theano Iliopoulou et al.', Anonymous Referee #1, 27 Apr 2018
- AC1: 'Reply to Reviewer1', Theano Iliopoulou, 29 May 2018
- AC4: 'Final response on behalf of all co-authors', Theano Iliopoulou, 06 Jul 2018
RC2: 'Review of "A large sample analysis of seasonal river flow correlation and its physical drivers" by Iliopoulou et al.', Anonymous Referee #2, 30 Apr 2018
- AC2: 'Reply to Reviewer2', Theano Iliopoulou, 29 May 2018
- AC3: 'Final response on behalf of all co-authors', Theano Iliopoulou, 06 Jul 2018

Peer-review completion

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

ED: Publish subject to revisions (further review by editor and referees) (16 Jul 2018) by Louise Slater

AR by Theano Iliopoulou on behalf of the Authors (27 Aug 2018) Author's response Manuscript

ED: Referee Nomination & Report Request started (09 Sep 2018) by Louise Slater

RR by Anonymous Referee #1 (19 Sep 2018)

RR by Anonymous Referee #3 (24 Oct 2018)

Suggestions for revision or reasons for rejection

I have been asked to look at a new version of this paper, but for independence I have not looked at previous versions, comments, or replies. This work contains some merit, but I can’t in good consciousness recommend publication in its current form. In the author’s own words, the results are often ‘expected’ and the discussion section mostly ‘confirms’ previous work and understanding of what is controlling catchment streamflow. Based on reading this paper, it is my feeling that we risk being more confused about catchment hydrology by doing this type of analysis, which I think is justified by the considerable difficulty the authors have in providing even qualitative links to actual hydrological processes in the discussion. I hope the authors take this as trying to be upfront with the concerns, and are able to engage with or use my main reasons for this decision below:
1. The following aspects of the methodology are unclear: For the HFS, the max daily discharge in the 3-month HFS is chosen. Is this value distinct from the max yearly discharge? In most cases, I suspect not. If the max discharge is in the second month of the HFS, does the lag-1 represent the correlation with the previous months mean discharge (also technically in the 3 month HFS), or with the last month before the onset of the HFS? If it is the latter, then the analysis is no longer technically a lag-1 analysis, and the study could be a big mix of lag-1, lag-2 and even lag-3 analyses that are all confused as representing a lag-1 value. Of course, it could be argued that you wish to be outside the HFS season for the correlation analysis, but what is the value of having inconsistent time periods in your lag analysis, especially given how sensitive the correlation will be to changing lag lengths? Moreover, what if a single catchment has max discharge always moving between the first and third month of the HFS over all the years of record? This will have a large impact on the ‘lag-1’ correlation even before any hydrological interpretations are involved. Some clarification on the mechanics of this analysis would really help.
2. The authors do not consider how the design of their study may have conspired to control the reported results before referring to a myriad of hydrological explanations. The core issue is one of signal vs noise. The LFS lag analysis uses a correlation between mean values that are by definition weighted by the central tendency of the data being considered, whereas the HFS uses a correlation between a max value and a mean, which is by design a far noisier signal, and hence displays little to no correlation with other variable throughout the study. Can the authors image a scenario where this would not be an expected result?
3. Related to point 2, the authors use a suite of metrics, many of which (P, SR, BFI) have a natural correlation with HFS and LFS since they are either derived from the same data or help generate it. The HFS analysis produces such a noisy signal that no result can be found, and this is hardly a surprising result (as mentioned above). LFS is not as noisy, and so displays better correlations. The heart of the paper is then to say that the correlations are better with hydrological processes that will also natural reduced the noise, e.g. higher groundwater flow subsidies and snowmelt, and worse correlations with processes and drivers that have increased noise. Again, can the authors think of a situation where this would not be an expected result?
4. Given these factors, it is unclear what processes or understanding can be revealed by such an analysis, since the study is producing most of the results by design, rather than by hydrological insight. In this sense the analysis in this manuscript obscures the actual hydrology, for example if you just plot actual baseflow on the maps in Figures 7 and 8 a clearer pattern of the hydrological controls on low flows would be revealed (or indeed baseflow against elevation, as documented by a lot of previous work). Surely, the LFS lag analysis only obscures these key hydrological drivers rather than making them clearer or easier to understand? I think this is also clearly shown in the discussion section, which is highly speculative about general processes and mostly confirms the results of previous workers rather than adding new understanding.
5. I don’t see the value or utility of section 7, it is incredibly short and not at all mentioned in the discussion section, therefore its completely unclear what we have learnt from this exercise, or in what context it’s results should be considered. This asymmetry is considerable given it has more length devoted to describing the methodology than anything else in the paper (section 2.3). However, after reading it a couple of times I found this to be the most interesting part of the paper, since it asks an interesting question about how you would expect HFS or LFS to change based on obtaining the new average discharge for the previous month (an update). However, this seems to have already been published and discussed in detail by Aguilar et al (2017), so what is the utility of the very brief repetition of the same work on a single river in this study? Given the results and methodology are far closer to Aguilar et al. (2017) than the rest of the submitted manuscript, it seems entirely out of place and only confirms their previous work.
Minor comments:
Figure 10 c, no colour scale provided
131: I don’t understand the basis of correlating LFS with the mean flow of the previous month on the expectation this is a robust proxy for storage. If you define the LFS as the month with the lowest flow, then by definition the previous months will have higher flow, so how will this be a robust proxy for storage? In fact, you will be correlating against months that could also be included in the definition of HFS, which we would not suggest are a good indicator of storage.
154: “SR (m3 s–1 km–2) is computed as the mean daily flow of the river standardized by the size of its basin area. It may be an important physical driver as it is an indicator of the catchment’s wetness” – so this basically says that runoff can be considered as an indicator of how wet a catchment is. This is like saying rainfall can be considered an indicator of how much water is falling from the sky, hopefully the authors can see the silliness of such a statement without further explanation.
479: “This result was expected since the correlation refers to average flow while the HFS correlation is related to rapidly occurring events” See major points 2 -4 , the design of the study is a major control on the results reported here rather than actual hydrological processes.

Hide

ED: Publish subject to revisions (further review by editor and referees) (31 Oct 2018) by Louise Slater

AR by Theano Iliopoulou on behalf of the Authors (16 Nov 2018) Author's response

ED: Publish as is (06 Dec 2018) by Louise Slater

AR by Theano Iliopoulou on behalf of the Authors (14 Dec 2018) Author's response Manuscript

Short summary

We investigate the seasonal memory properties of a large sample of European rivers in terms of high and low flows. We compute seasonal correlations between peak and low flows and average flows in the previous seasons and explore the links with various physiographic and hydro-climatic catchment descriptors. Our findings suggest that there is a traceable physical basis for river memory which in turn can be employed to reduce uncertainty and improve probabilistic predictions of floods and droughts.