Two Ways to Quantify Korean Drought Frequency: Partial Duration Series and Bivariate Exponential Distribution, and Application to Climate Change
Yixin Mao
Round 1
Reviewer 1 Report
The manuscript "Two ways of quantify drought frequency: partial duration series and bivariate exponential distribution, and application to climate change" implemented two different statistical analysis based only on precipitation data at four stations in Korea, with goal of quantify the characteristics of drought severity and duration. On top of these two sets of results, the authors further showcased how these two analyses can be used to study the evolution of a certain drought event, and the future climage change impact on drought characteristics.
Overall, the methods in the manuscript are simple but useful, which makes it an interesting addition to both the research community on drought and also to local stakeholders. My major comments on the manuscript include 1) the manuscript can be more succint with main points better highlighted, and 2) some part of the methodology needs more clarification. I list detailed comments as follows, and would like to see them addressed before publication at Atmosphere.
Major comments:
1. Equation 1 (and similarly, Equation 4) needs more clarification. Intuitively, it is understandable that lambda is included in the denominator of the formula for return level, since the more frequent an event is, the shorter the return should be. What's the intuition that the return level also includes the term {1-F(s)}' in its denominator? Does the prime (') mean derivative? Additionally, for Equation 4, what's the difference between Equation 4 versus Equation 5 and 6 - isn't the definition of F(s,d) the integrals in Equation 6?
2. Some figures and corresponding explanations in the manuscript seems to merely be a verification of methology choices. Some examples I see:
- Figure 5 seems to just be validating the Gumbel curve fit;
- not sure how much additional info Figure 4 is adding to the story given Figure 6 is presented;
- Figures 7 and 8 seem to be more of a display of model fitting quality without adding much to the story;
- Figure 9 again is sort of a raw result display without much quantification or discussion. Is it also redundant to Figure 11?
- Hard to see the information provided in Figure 10 - better to use a 2D heatmap. Also, is the point of this figure to show the fitting of the probability function to real data? Then should add real data points too, with quantitative summary;
- Not sure what Figure 12 is showing
These content may be more suitable for supplemental material. For the ones that were shown for the purpose of model validation, perhaps a quantitative summary of goodness-of-fit can be included and discussed in the main manuscript while the detailed figures can go into supplemental material. If the authors included some of these figures for additional purpose that would actually add insights to the main story, then more explanation is needed in the text to show how so. I think this would make the paper more succinct and better bring out the main story - right now the main messages are kind of burried in secondary details.
3. While the authors mentioned that they implemented two approaches for drought analyses, it wasn't until the Discussion section did it become clear to me that these two approaches were implemented completely separately, with potential inconsistent results. It would be good to make it clear from the beginning (from introduction to methods to results sections, as well as in abstract) that these are two independant approaches and one of the goals of the study is to compare the pros/cons/suitable situations of them. I think this would harden the main points of the study.
Minor comments:
1. Give a brief definition of SPI
2. Figure 2 and Figure 15d: Use actual year and month as x labels, currently hard to correspond points on plot to a year or season
3. Caption of several plots need more details - Figures should be understandable on its own w/o manuscript text. For example, In Figure 2, what is the red line (this only gets mentioned later in the text)? Figure 4 - explanation of x axis; Figure 5 - explanation of axes; Figure 6 - what are x and + (again, this appears only in text); Figure 7 - what's the difinition of "severity" and "relative frequency" here? etc.
4. Equation 4: Should be F(s,d)?
5. Line 231: spring of 2001?
6. Figure 5: would it make more sense to call the y-axis something like "3-mon SPI"?
7. Table 3: Unit of duration mean and standard deviation?
8. Lines 307-311: Where did these numbers come from (Table 4, as referenced, is bivariate analysis, not univariate?)
9. Lines 311-313: Not sure what this sentence mean - do you mean conclusions based on severity versus based on duration can be different? Or different results based on different statistical approaches?
10. Lines 327-328: Not obvious to me that the correlation is strong - need quantitative results
11. Lines 403-405: The threshold of SPI=-1 used in the bivariate approach should have a direct effect on BED results. What about looking at the sensitivity of BED results to this threshold?
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Review of atmosphere-788569: “Two ways to quantify drought frequency: partial duration series and bivariate exponential distribution, and application to climate change” By Won et al.
Review to authors:
The authors quantify the frequency of drought by using univariate analysis and bivariate frequency analysis, in which the latter represents the dependency of drought severity and duration. The paper well explains the methods and why to use them. The authors provide discussion of caveats using two approaches, which is fairly appreciated. The final discussion of two approaches based on RCM outputs is a plus. Overall, the paper is well written and deserves publication with some minor revisions, which are listed below:
- Fig. 2: Please specify the red line. It would be useful if the x-axis denotes the date instead of the number of months.
- Line 197: ‘was’ is not correct in grammar. Authors just delete it.
- Eq (4): F(s,s) should be F(s,d).
- Line 231: most severe in the spring of 2005, ‘2005’ should be 2001, right?
- Line 270: ‘average of 33 drought events occurred (average of five sites)’: I am not clear by this statement and authors need to clarify and/or correct it.
- Fig. 9: The cross correlation between duration and -SPI6 is not large at Seosan. The statement in lines 326–328 is not convincing. In Fig. 9, it is hard to see a good linear relationship between duration and -SPI6.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Very interesting and well-written study on the assessment of present and future drought frequency and severity in Korea. My suggestion is the manuscript to be accepted for publication after considering the following minor points.
I know it is too much for the present study, but in a follow-up, it would be very interesting if the authors investigate also the effect of temperature on droughts by calculating the SPEI index (Vicente-Serrano et al. 2010). This would be of great interest for future changes where in a warmer world, evapotranspiration is expected to play a greater role.
A very recent study on global drought projections that could be cited in the introduction is (Spinoni et al. 2020).
In the title, abstract, and/or keywords the authors should add the word “Korea” for search and indexing purposes.
On line 138 the authors should provide a reference of the L-moments method for parameters estimation.
I would move sections 4.1-4.3 to results.
References:
Spinoni J, Barbosa P, Bucchignani E, et al (2020) Future Global Meteorological Drought Hot Spots: A Study Based on CORDEX Data. J Clim 33:3635–3661. doi: 10.1175/jcli-d-19-0084.1
Vicente-Serrano SM, Beguería S, López-Moreno JI, et al (2010) A Multiscalar Drought Index Sensitive to Global Warming: The Standardized Precipitation Evapotranspiration Index. J Clim 23:1696–1718. doi: 10.1175/2009JCLI2909.1
Author Response
Please see the attachment.
Author Response File: Author Response.docx
