Spatiotemporal Variation Characteristics of Precipitation in the Huaihe River Basin, China, as a Result of Climate Change

Round 1

Reviewer 1 Report

The concept of the study is very interesting and the use of such a large data set (1960-2020) of 233 meteorological stations in the Huauhe River Basin in China is impressive.

However, I have methodological doubts: how to justify the use of a non-parametric model for data that are normally distributed (air temperature)? The authors used Pearson's correlation coefficient to relate the data, rather than Spearman's.

The description of Pearson's correlation coefficient (subsection 3.2.2.) as a basic statistical procedure, while correct, is so basic that I would recommend skipping it or clearly limiting it to the core information related to the objective of the study. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This work is very interesting; in my opinion, it deserves publication, but not in this form. A review of many parts of the work is needed, some of them are very challenging.

Section 2

10 lines only for the description of the study area are too few. It is necessary to deepen the geomorphology of the study area, not only from the climatological point of view but also to report the references of the works published on the geology of the Huaihe river basin. For example, it would be very important to describe the water environmental supporting capacity of the various areas of the basin and to combine this information with your study (also spatial) on the rainfall trend.

Figures 1 and 2 are very similar. If you don't want to add more information, Figure 1 is redundant and you can leave only Figure 2. Alternatively, you can add the course of the Huaihe River, its tributaries and larger lakes to Figure 1. There is a typo in the legend: "Altitude" and not "Alittude"

Section 3.1

Also, the dataset employed needs to be described with much more care. Do all 233 stations have 60-year time series with no gaps? The adjustments you had to make should be described in more detail: what percentage of the stations needed to integrate the data? how long were the time gaps? How was the annual cumulative rainfall evaluated in cases of missing rainfall data for many days? Why the data are not public?

Section 3.2

What does it mean that the Mann-Kendall test was used to test for significance? The significance is a parameter chosen before carrying out the test (usually, the confidence level is chosen...). The use of the Pearson coefficient requires that certain hypotheses be tested, in particular the Gaussianity of the samples whose correlation is to be observed. If this assumption is not valid, other tests for correlation can be used, such as Spearman's correlation coefficient. Finally, for the correlation of time series that exhibit a trend, a precise statistical tool should be used, which is the detrended cross-correlation analysis (DCCA). There is no trace of these in the 5 lines devoted to the exposition of the statistical methods chosen.

Section 3.2.1 In lines 144-145, it is not clear what the alternative hypothesis H1 is. Once it has been defined, one can choose to carry out a unilateral or bilateral test, explaining the reasons for the choice. For the Mann-Kendall method, in addition to the description, references should also be cited. The final part of the description, in which UB_k is defined, in particular, the equation 3-9 in line 166, is not clear. In line 169, the confidence level should be 0.95 (1-alpha), while alpha is usually called significance.

Section 3.2.2

This section also does not list the assumptions under which Pearson's correlation coefficient can be used.

Section 4

All captions write "distribution of the significance level". As I have already said, this level is established at the beginning of the statistical analysis, and it is fixed. I think the colourmaps report the values of the statistic (S or Z, it's not clear...). By themselves, these values mean nothing. It would be preferable to report the p-value. it is not clear how the spatial interpolation was carried out, which is quite useless for this type of analysis. It would be enough to transform each of the symbols indicating the positions of the stations so that they show the result corresponding to one of the four described in lines 170-171. This goes for all colourmaps, even those for each season.

Section 4.3 Table 4.3 shows a rather poor correlation. Is it not that, by changing the statistical technique for the study of correlation, this result can change?

In figures 7 and 8 the correlation coefficient scale is not between -1 and 1, as the definition in line 188 would suggest.

In the bibliography there are both Roman and Arabic numerals, I think only one is enough. References 33 to 40 are only templates.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

After the first review, your article has eliminated some defects; unfortunately, there are others left, also because you replied to some observations only in the cover letter, and not by changing the text of the article, as it would have been more correct to do. For this reason, I will resume the list of remarks I had already drawn up in the first review, inviting you to transfer some of the comments you reserved only for me to the article.

Points 1 and 2 are fine, you have made the changes following the cover letter comments.

Point 3. The sentence you wrote in the cover letter

"Among the 233 stations in the Huaihe River basin, the missing rate of precipitation stations is about 8%, while the missing rate of temperature stations is about 12%. The missing data are mainly missing for one day or several continuous days. The missing data are replaced by the average value of neighbouring stations, and then the monthly and yearly data are calculated. By mutual agreement, the day-by-day series data are only used for scientific research and are not allowed to be made public." (or a variant thereof) must be placed somewhere in the article, such as after the word "study." in line 131. It should be specified whether at the end of this procedure we obtain time series with no missing years (it is well known that, in such cases, it is necessary to use a variant of the Mann-Kendall test).

Point 4. It would be useful to better clarify in the paper the results obtained with the gaussianity tests (for example: do all the time series pass the Kolmogorov test, with 95% confidence?). It is well known that the statistical concept of correlation refers to a joint probability distribution from which two variables are drawn. With non-stationary time series, successive values don't come from the same distribution, so there's no concept of correlation. Your series appear to be almost all non-stationary (you claim that there is a trend), how do you justify the use of the correlation coefficient? This must be written in the paper!!

Point 5. Now the meaning of UB_k is clear, even if we don't understand how it is possible that the subscript k could be the same in equation 3-9 on line 174 (what does m represent? perhaps you meant n?). With the notation you use, it seems that the order of the terms in the time series is the same. Perhaps you meant to write UB_k=-UF_(n-1-k). I also believe that it is correct to specify, in the paper, what kind of results we intend to show using this statistical tool. In the current version of your article, it doesn't seem to me that results regarding mutation time and mutation area are shown. Then you have to justify in the paper why you have included these equations in the description of the statistical methods used.

Now, reading line 177, we understand that the significance level is 0.05 and is fixed (that is, it is the same for all the data you analyse, both in time and space). It is therefore not clear how it is possible that you present results in the following figures which show the variation of the significance in the area.

Point 6. You have now written the validity hypotheses using the Pearson correlation coefficient. It would be essential to include in the paper, after lines 184-187, the results you obtained using the control tests you listed.

Point 7. The observations I made in the previous revision did not prompt you to make changes to the paper. I'll try to be clearer. The level of significance is fixed once and for all, it makes no sense to write "spatial distribution of the level of significant values" in the captions of figures 3, 4, 5, and 6. What is represented is another quantity, which has a critical value of 1.96 (as you write in line 178 of the paper). I believe (if I understand correctly) that in the figures the value of Z calculated according to equation 3-4 (on line 161) is represented. In any case, the colour scales show the presence of values greater than 1, so it's certainly not a significant level. You have to correct the captions in the paper!!

Point 8

Without needing to quote all the references you presented in the cover letter, please introduce in the paper some justification for the fact that in essence, the results presented do not depend on the choice of statistical method.

Point 9 Now the figure showing the distribution of the correlation coefficient looks fine.

Point 10 Now, in the list of the cited works, the numbers in Arabic numerals are present twice. One number is enough. Please change the format of the bibliography.

Author Response

Please see the attachment.

Author Response File: Author Response.docx