Estimating the Statistical Significance of Cross–Correlations between Hydroclimatic Processes in the Presence of Long–Range Dependence
, Ilias Deligiannis, Panayiotis Dimitriadis *, Theano Iliopoulou, Nikos Mamassis and Demetris KoutsoyiannisRound 1
Reviewer 1 Report
Major comments
1) Abstracts must be standalone. What is the difference between cross-correlations and an innovative stochastic test? The potential readers cannot decipher just by reading the abstract.
2) Line 23: “innovative statistical test is constructed using a stochastic approach “what makes the statistical test used in this study innovative? It requires a detailed comparison among already available methods.
3) Lines 9 & 23-25: I think the terms” Renewable energy management systems” and “cleaner forms of energy” are irrelevant in your manuscript. I suggest removing or replacing these terms.
4) Line 33: the term “these hydrological processes” should be defined to make your work understandable for potential readers.
5) Line 168: “…year contains less than 300 days of values is considered null…), what is the criteria/reference for it?
6) Lines168:” …. most likely…” should be supported with evidence/rigorous analysis and discussed in detail.
7) The proposed “stochastic test” or “Monte-Carlo analysis”? The expression should be consistent throughout the manuscript.
8) Lines 73 & 196: The data analysis should be supported with long-term data (at least 30 years). Only 20 years is not enough to understand the natural variability of hydroclimatic processes.
9) The significance, aim and scientific contribution of this study were not presented. Clearly, describe these key points.
Minor comments
Line 66: ‘many studies in literature these studies should be cited or mentioned.
Line 78: delete ‘R/S’
Figures 2-4: ‘Gen Gaussian’ should be defined in the text of the figure description.
Author Response
Dear Reviewer,
We thank you for your suggestions and comments. Please find detailed responses and our revisions to the manuscript below.
Abstracts must be standalone. What is the difference between cross-correlations and an innovative stochastic test? The potential readers cannot decipher just by reading the abstract.
Response: Added explanatory text to abstract, see primarily lines 18-21.
- Line 23: “innovative statistical test is constructed using a stochastic approach “what makes the statistical test used in this study innovative? It requires a detailed comparison among already available methods.
Response: Added explanatory text to abstract as above, also see the comparison in text to the classical t-test for statistical significance. The lack of pre-whitening as described in the text is another innovative factor of the proposed method.
- Lines 9 & 23-25: I think the terms” Renewable energy management systems” and “cleaner forms of energy” are irrelevantin your manuscript. I suggest removing or replacing these terms.
Response: Removed these terms where possible, however the results of the study could be linked to managing renewable systems, and thus tie-in to the scientific contribution of this study.
- Line 33: the term “these hydrological processes” should be defined to make your work understandable for potential readers.
Response: Defined these processes, see new Line 35-36.
- Line 168: “…year contains less than 300 days of values is considered null…), what is the criteria/reference for it?
Response: It is related to uncertainty in timeseries when there are more than a certain amount of missing data, added citation from the existing References.
- Lines168:” …. most likely…” should be supported with evidence/rigorous analysis and discussed in detail.
Response: Our study does not focus on specific examples and locations, only an overall global approach to timeseries. Microclimates are left as a possible future topic of research in the Discussion section. We studied a few particular outliers while conducting our own research, but could not find any conclusive results to present in this article.
- The proposed “stochastic test” or “Monte-Carlo analysis”? The expression should be consistent throughout
Response: The stochastic test uses Monte Carlo analysis. The term “stochastic test” is more broad, and has been replaced in the text where relevant.
- Lines 73 & 196: The data analysis should be supported with long-term data (at least 30 years). Only 20 years is not enough to understand the natural variability of hydroclimatic processes.
Response: The study is limited by whichever of the two cross-correlated timeseries has the shortest length, and wind speed has only been calculated fairly recently (last 20-40 years) in most studied stations. That being said, few of the stations included in the analysis actually had only 20 to 30 years of data. In the short future, even more data will be available, and the study could be repeated to further validate the results.
- The significance, aim and scientific contribution of this study were not presented. Clearly, describe these key points.
Response: The aim of the study is to showcase the uncertainty of hydroclimatic processes, how this impacts any statistical significance of analysed results, and present an innovative method to tackle these issues. A key factor is that it avoids pitfalls related to assumed independency of the studied processes, and this has been explained thoroughly in the text. Further clarification has been added in the Abstract, see above response.
This concludes our responses to comments. We would again like to thank the Reviewer for their time, and are happy to return to address any further comments that they may have.
Sincerely,
The Authors
Reviewer 2 Report
Please find attached.
Comments for author File: 
Comments.pdf
Author Response
Dear Reviewer,
We thank you for your suggestions and comments. Please find detailed responses and our revisions to the manuscript below.
- The presented description assumes that the reader knows the described statistical methods well. People unfamiliar with the described methods, after reading the introduction, will not know whether the described long-range dependence relate to time or space and how to understand them. Too much remains a matter of guesswork.
Response: Further text has been added to explain long-range dependence in simple terms. See new Lines 77-82.
- Hydroclimatic processes is probably not the best name for air temperature, dew point, wind speed and precipitation. The term temperature itself is too ambiguous.
Response: It is possibly a generalised term, but we could not find a more suitable alternative.
- The performed numerical experiment was well described, but as I wrote earlier, there was no introduction to the described issues that was easier to understand. This applies, for example, to Hurst phenomenon.
Response: Further text has been added to explain long-range dependence in simple terms. See for example new Lines 39-40 and 77-82.
- The explanation of the problem of aggregation of meteorological data was completely omitted. In the case of wind speed, this is not a trivial problem.
Response: See amended Lines 181-189, and new text in the Discussion and Conclusions, Line 310-319.
- The figures are very uneven. Some of them are fine, some lack descriptions of the axles, and some of the captions do not fully explain what the graph shows. The algorithm for selecting the precipitation station seems quite subjective (fuzzy). I would love to see some simple statistics (box plot?) related to the described implementation.
Response: We have clarified the text in some captions, see also new Figure 6 for the requested box plots related to precipitation station selection.
- The statistical part should be written a bit more accessible, so that the average reader does not feel lost at the very beginning. The introduction (as a section) is just too short.
Response: Further text has been added to explain long-range dependence in simple terms. See for example new Lines 39-40 and 77-82.
- For which station on Figure 6 cross-correlation between annual mean temperature and dewpoint is negative?
Response: There is a misunderstanding here, there is no negative cross-correlation of note in Norway, these were spotted in different locations on a global scale, see amended Lines 231-232.
- I missed an in-depth discussion with regard to the tested parameters. The authors limit themselves to general statements about the influence of the microclimate. This does not sound very convincing.
Response: Microclimates are left as a possible future topic of research in the Discussion section. We studied a few particular outliers while conducting our own research, but could not find any conclusive results to present in this article.
- The caption of Figure 10 without context in the text is incomprehensible. There was also no reference to these results in the discussion. Overall, the discussion seems to be quite brief.
Response: The Figure has been moved to Figure 7, and is now part of Materials and Methods section. Caption has been amended to further clarify the thought process. See new text in the Discussion, for example Lines 310-319.
- Again, I would consider writing conclusions in such a way that this section can be read by the average reader without having to refer to other parts or cited publications.
Response: As per above item, we have expanded the Discussions and Conclusions section.
This concludes our responses to comments. We would again like to thank the Reviewer for their time, and are happy to return to address any further comments that they may have.
Sincerely,
The Authors
Round 2
Reviewer 1 Report
The authors addressed the minor edits I pointed out but did not attempt to address some of the MAJOR issues commented in my review (comments #3,6 & 8 should be addressed satisfactorily). Please go back to my initial review and address these.
Author Response
We again Thank the Reviewer for the feedback on our revision. Please find below our replies to these comments:
Response to Comment 3: These terms have been now removed.
Response to Comment 6:
See amended text in Discussion and Conclusions, lines 310-317. To keep the section standalone, for specific references and analysis see amended further analysis and link to case study in Australia in lines 236-241.
Response to Comment 8:
See amended text in lines 181-195 and 241-246. Few stations actually kept in the data had less than thirty years of data. We keep some newer built stations (above 20 years) in order to have a wider spectrum of results from around the globe, which gave cross-correlations similar to thirty-year stations, see Results section. A sanity check on the results shows that timeseries with large gaps were much more likely to cause erroneous outliers, but these stations have not been included in the analysis due to the applied filters.
Sincerely,
The Authors
