Combination of Using Pairwise Comparisons and Composite Reference Series: A New Approach in the Homogenization of Climatic Time Series with ACMANT
Round 1
Reviewer 1 Report
The article is a excellent contribution to the problem of homogenization of time series in Climatology.
It states well the problems that we face and searches for solutions.
It has a big potential to address many scientists.
But the main (and at the same time the only) problem is discrepancy between its beginning and end.
Something has to be changed, other words has to be applied.
The title states:
"Combination of using pairwise comparisons and composite reference series: a new approach in the homogenization of climatic time series"
The Abstract is then in accordance with the Title, and in the end, there is a statement:
"This time series comparison method is embedded into the ACMANT homogenization method, and tested in large, commonly available monthly temperature test datasets."
This is fine. So far so good.
But then, if one look at “5. Concluding remarks”, it starts with:
"This paper reports about the most recent developments of the ACMANT homogenization method."
and whole pages before the conclusions (together with Conclusions tiself) are solely about ACMANT. For instance comparison of the last versions.
To be in accordance, either the Title and Abstract,
or the end of the article (chapters 4 and and 5 - Conlusions),
has to be changed,
The article is either about ACMANT, or generally about reference series.
Specific comments:
line 101:
“may seem a contradiction, which, however, only a seeming contradiction”.
There is probably missing "is" before “only”, and with repeating "seem contradiction" the sentence looks strange … please revise it
line 113:
“inhomogeneities do not seem significant in”
There is probably missing "to be" after “seem”
line 142
"(" is not ended up with ")", the sentence seems not to be finished ...
line 183
"that two consecutive breaks with the same sign shifts easily can be mixed up with trend-like inhomogeneities."
"same sign shifts easily" - please revise the sentence, it does not make too much sense
line 216
I would not put sentences in the brackets as stand-alone, but as a part of the previous sentence.
I suggest either to put away the brackets, or to include the brackets into the previous sentence.
line 225
"break points often not only the long-term 225 means change,"
some word is missing here
line 280
other experiments - could some references be given?
line 285
"correlating time series"
maybe "correlated" should be here?
line 316
"... , the combined time series comparison method in ACMANT"
some word is missing here
line 495 states: "No adjustment is applied between step 1 and step 2"
but later, line 502 states: "therefore the detected semi-synchronous breaks of step 1 must be fixed before the detection with composite reference series".
these statements seem to be in contradiction
line 569:
"test datasets are denoted by Y1…Y6 (U1…U6) in 569 each figure."
It would be fine to explain briefly what Y and U series mean. Without finding and reading the given reference, it is not clear at all.
Figures could also indicate what Y and U means (at least a brief reference that it means benchmark datasets). If one would start with exploring figures, he would be lost.
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
In this manuscript a combined approach of pairwise comparison and composite reference series is uesed to produce a homogenizated climatic times series. The content is relevant and of interest for a wider research community. Climatic time series based on observational data still have the problem that they show jumps in the time series due to changes in the observational station position or other inhomogeneities due to technical problems or processing. But these times series are used in many research communities to calibrate their models or run model simulations. Therefore, a good approach to homogenize the climatic time series is needed.
Although I have difficulties to accept this manuscript as a research manuscript. First of all, no climatic time series and examples of the method are shown. Second, only the perfomance measures of the old to the new ACMANT version is shown. It reads like a part of a manual about the better performance. The science and the application of the method with examples to real world data is missing. I suggest to improve the manuscript and send it in again.
Some further comments are listed in the following:
- Please write our ACMANT the first time this phrase is mentioned.
- Lines 73 to 76: Please add references for both methods.
- Much of the explanations of the pros and cons written in the method secction should be moved to the introduction. For example, lines 87-164
- The use of ACMANT lacks behind the development. Currently, hourly to sub-hourly observational data are needed. Maybe the author can describe if ACMANT will also be developed for hourly and sub-hourly data in the future.
- Examples of time series at different locations are missing with showing the times series before and after ACMANT is applied.
Author Response
Thank you for the review
In this manuscript a combined approach of pairwise comparison and composite reference series is uesed to produce a homogenizated climatic times series. The content is relevant and of interest for a wider research community. Climatic time series based on observational data still have the problem that they show jumps in the time series due to changes in the observational station position or other inhomogeneities due to technical problems or processing. But these times series are used in many research communities to calibrate their models or run model simulations. Therefore, a good approach to homogenize the climatic time series is needed.
Although I have difficulties to accept this manuscript as a research manuscript.
Unfortunately, I have not found any objective and relevant argument in your comments which would support your unfavourable opinion about the manuscript.
The goal of my paper is to describe a new time series comparison method providing improved accuracy when it is appropriately embedded to a homogenization method. I report also about the other favourable characteristics of my homogenization method (ACMANT). I provide comparisons with other methodologies, theoretical analysis and experimental justification with high quality synthetic and surrogated test datasets.
First of all, no climatic time series and examples of the method are shown.
More exactly: synthtic and surrogated time series are used, and it is the best tool for testing the efficiency / accuracy of homogenization methods.
Second, only the perfomance measures of the old to the new ACMANT version is shown.
I think that it is fine and sufficient for the objectives of the paper.
It reads like a part of a manual about the better performance.
I do not think so.
The science ... is missing.
A correct review is missing.
...the application of the method with examples to real world data is missing.
Methodological development can be combined with real data homogenization, but I do not know any scientific reason why such combination of topics should be obligatory in a study.
I suggest to improve the manuscript and send it in again.
Some further comments are listed in the following:
- Please write our ACMANT the first time this phrase is mentioned.
- I do not understand this comment.
- Lines 73 to 76: Please add references for both methods.
- I give references where I consider them the most useful.
- Much of the explanations of the pros and cons written in the method secction should be moved to the introduction. For example, lines 87-164
- Although I admit that the paper could have a longer Introduction part, I believe that the present structure is clear and helpful for readers.
- The use of ACMANT lacks behind the development.
- The references include such examples.
- Currently, hourly to sub-hourly observational data are needed.
- Without inserting "sometimes", your statement is false.
- Maybe the author can describe if ACMANT will also be developed for hourly and sub-hourly data in the future.
- The presented methodological development contributes to the improvement of data accuracy in all time scales.
- Examples of time series at different locations are missing with showing the times series before and after ACMANT is applied.
- I hope that the climatological community will use more frequently ACMANT than did it in the past.
Reviewer 3 Report
General Remarks: Handling time series data and getting some informative information is a challenge in the community. Lots of uncertainties are related to the time series analysis and it is good to see the new methodology, that may minimise these errors/biases in the outcomes. However there are some issues, which need to be addressed by the author.
- In its current form manuscript is quite descriptive. What is the main development in the new ACMANT version compared to the earlier ? Please clarify this in your abstract.
- What is the difference in the method described here and the method as of "nearest neighbour"?
- what is the robustness of this method? will that be sensitive to the region where it is applied?
- Will be useful for the trend analysis?
- Line 31-35: how you are differentiating between noise, weather and climate signal? and station effect? weather its contribute to noise? and what does station effect mean, is this related to the instrumental issues ? or something? Can you describe with real time example, these three components?
Author Response
Thank you for the review.
- Now I have put "ACMANT" into the title in a way that the connection between the actual methodological development and ACMANT is clearer.
- When the "nearest neighbour" (in singular) is applied in homogenization, a principal problem is that a break of the difference series can be either for the break of the candidate series or the break of the neighbour series. Therefore this approach is rarely used in modern time series homogenization.
- Robustness: proven by its mathematical basis and the experimental results shown in the study. Sensitivity to regions: there is no sufficient knowledge about this.
- Yes, it is demonstarted in the paper.
- The components of observed climatic time series are climate signal, station effect and noise. Weather is the main component of noise, as the temporal variation of weather is generally fast (in comparison with the other components) and irregular. In synthetic / surrogate time series the components are exactly known, as we generate the data. In real data we usually know approximately the typical ratios of the components, and one purpose of the homogenization is to learn them for stations and networks as accurately as possible.
Round 2
Reviewer 1 Report
No further comments
Author Response
Thank you very much.
