Shifting of Meteorological to Hydrological Drought Risk at Regional Scale

Round 1

Reviewer 1 Report

The manuscript "Shifting of meteorological to hydrological drought risk at regional scale" presented a study of statistical analysis of meteorological drought (quantified by SPI) and hydrological drought (quantified by SDI) in the Soan River basin, Pakistan. The authors used statistical methods including trend analysis, linear regression and moving average to detect time series characteristics and inter-relationship between the two types of droughts.

 

While the overall objective and approaches of the study is quite clear, there seem to be several places where the conclusions are not supported by, or contradictory to, the results. In other places, the approach is poorly described or inconsistent. Therefore, I would not recommend the manuscript as in its current form to be published.

 

Some examples where the conclusions are not supported by the results:

1. Lines 204-205 state that "[Figure 2 illustrates] ... a comparison between the 6-month and 9-month time scale gave similar results". However, Figure 2 doesn't show 9-month results (but only shows 1-, 3-, 6- and 12-month results)?
2. Lines 219-221 state "[Figure 3 illustrates] ... Based on the analysis, it resulted that the years from 1991 to 2001 were the driest years of the selected time series." - This doesn't seem to be the case (at least to me visually) from Figure 3.
3. Line 230: "Overall analysis showed that there was an increasing trend from 1983 to 2015..." - is SPI increasing? Or the negative of SPI (a.k.a. drought) increasing?
4. Lines 289 - 290: "... it was found that the difference between SPI and SDI had increased over the past 3 decades" - where is the result that supports this conclusion?
5. Lines 295 - 297: "This decline in precipitation was larger in Sub-Basin 1, as compared to Sub-Basin 2" - but the right panel in Figure 10a shows increasing trend for precipitation for Sub-Basin 2?
6. Lines 299 - 200: "the runoff coefficient also had a declining trend at both Sub-Basins ..." - but the left panel in Figure 10c shows increasing trend in runoff coefficient in Sub-Basin 1?

 

Places where I find the approach is not clearly described / inconsistent:

1. Figure 1: Unclear where exactly the right-side basin is located in the left-side map? Also, unclear where's the exact boundary of sub-basin 2 in the right panel.
2. Line 119 states "... provided daily precipitation data from two meteorological stations within the basin)", but 4 meteorological stations shown in Figure 1? Which ones were used for each sub-basin exactly?
3. Lines 155 - 163 describe the definition of SPI-3/6/9/12, but the description here seems inconsistent with Equations 1 - 4?
4. Equations 6 - 8 not understandable - needed better denotation of variables and clarification. Also, unclear from the text/equation in this section how the moving average calculation can be used for lag time analysis?
5. Figure 4 (and same for Figure 7): What is the dotted line on the plot?
6. Section 3.2.2 - might be related to comment 4 above - unclear how the lag time is derived

 

Other comments:

1. Line 138 - while SPI is most likely to be within the range of -3 to 3, mathematically there isn't a hard bound
2. Line 274: Should be Fig. 9?

Author Response

Please see the attachment 

Author Response File: Author Response.pdf

Reviewer 2 Report

The submitted paper deals with the issues of meteorological and hydrological droughts, their mutual relations and changes over time. It covers an area that is poorly described in the literature, mainly due to the lack of data. I my opinion the research raises very important issues. This becomes even more important when referring to climate change. Although the method used is traditional and the amount of data analyzed is really modest, the article could get attention of many Applied Sciences readers. However, in my opinion there are few drawbacks in the paper, which can be eliminated by carrying out some major revisions following the comments/suggestions below.

1. The methodology is simple and does not require extensive explanation. Nevertheless, the authors took this section very seriously. As a result, the section “Material and Methods” is disproportionately long compared to those containing results and discussion.

2. The Results section requires upgrading/re-writing. The presented results does not give the authors right to draw such categorical conclusions as they did. I am reluctant to agree with some definite statement of the authors, for example: 

“The temporal variations of hydrological drought periods in sub-basins 1 and 2 at different time scales are shown in Figure 5, which indicates an obvious increasing trend in the occurrence of hydrological drought over the past three decades.” (lines 238-240). In my opinion, this cannot be said about sub-basin 1 referring to Fig. 5. 

“The coefficient of determination (R2) and regression coefficient (a) extracted from linear regression equations (Table 1) showed that a significant correlation existed between SDI and SPI, and it increased with the increase in time scale.” (lines 258-260). Such a general statement is not true for the 3- and 12-month time scale. The coefficient of determination (R2) for these periods is relatively poor.  

3. The figures should be chosen more selectively. Some of them, such as Figures 4 and 7, should be completely omitted. They are a repetition of Figures 3 and 6, respectively. Figure 9d is incorrect. This is a duplicate of Fig. 9f.

4. Trend analyses of precipitation, runoff depth and runoff coefficient should be found in the Results section. Using only the linear regression to derive the tendencies/trends in these parameters seems to be at least not enough or missed sometimes (especially in relation to precipitation – see Fig. 10). I believe that this description requires other justification than just questionable regression fits.

5. Discussion and Conclusion sections are extremely short. I suggest combining these sections and enriching them with the results/conclusions of similar studies.

6. Finally, the paper consists minor editing errors (especially lack of spaces before or inside parentheses) and some linguistic clumsiness, which could be avoided in the next, revised version of the paper.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The study analyzes the meteorological and hydrological two types of droughts in a significant area of Pakistan. The SDI (Streamflow Drought Index) and SPI (Standardized Precipitation Index) are used to make a spatiotemporal assessment of drought. To this end, the authors used: regression analysis, Mann-Kendall trend test, Moving average approach.

 

The aim of the analysis is somehow clear, but it is not clear how the authors found the results. The Introduction is written based on some works that do not support what is stated in the premise. For example, some references are linked to local results (both in time and space) that, in my opinion, are not appropriate to support general statements. In general, several references are out of context also in other parts of the papers.

Besides, the methods are not presented clearly.  For instance, the set of the dependent variables in the regression analysis is not explained. I understand that the meteorological drought index is the dependent variable, but it is not clear to me whether the authors tested the linear correlation of this dependent variable at a shorter scale, say 3, against a set of drought indexes at the remaining time scales or at the same time scale. Also, the definition of the SPI time scales is not clear because, in one situation, it is computed by using the cumulate precipitation of the following months, in one another, by using the cumulate precipitation of the preceding months. Furthermore, several equations should be revised. Therefore, it is hard for the reader to comprehend the nature of the relationships that the authors intend to explore.

Consequently, the presentation of the results is quite hard to follow. In addition, several results are reported without an associated explanation in the methodological section. These drawbacks make the evaluation of results difficult to do.

 

To summarize, a more high-level re-thinking of the scientific design and a focus on quantitative findings, which I am sure have been obtained throughout this research study, would give merit to this paper. Furthermore, several critical issues have been highlighted that need to be addressed and that hopefully will make this paper more impactful. 

 

1. Introduction: Not very clear, Ref [7] is out of scope.

 

1. Data:

 

- The zero point station is roughly 140 km far from the Dock Pathan streamflow station. Is this distance appropriate for the scale of the analysis? Please, include some discussion about it in the Section.

- Sec 2.2.1, it is explained how to compute SPI starting from monthly precipitation. However, a description of the original data is missing. Are data from rain gauges recorded daily? If yes, missing data have been evaluated before monthly aggregation? Please, comment on that.

 

1. Methods:

 

- Sec 2.2.2, Eq. 2 is not clearly stated because Rik refers to a single year. How do the authors compute the standard deviation and average value of the precipitation? The same applies to Eq. 4.

- Sec 2.2.1 and Sec 2.2.2 are not presented clearly since there are repetitions of the same concepts. The authors should consider that SDI is built following the SPI method. Therefore, the Section would be more readable if the SPI method was presented exhaustively (including the procedure for the standardization onto the Gamma distribution) and, then recall this method when presenting the SDI. Moreover, in Sec. 2.2.1 SPI-3 of October refers to the cumulate precipitation of October-November-December while in Section 2.2.2 SPI-3 of October refers to the cumulate precipitation of August-September-October. This makes the comprehension of the approach quite hard, for this reason, I kindly suggest revising these two sections.

- Sec 2.3 Did the authors take into account the multicollinearity issue in their regression analysis? Since drought indexes for different scales are computed by using a common set of monthly data, it might exist a strong correlation between dependent variables.

- Sec 2.3 R2 is not appropriate to discriminate between dependent variables in regression analysis. To this end, other methods are available, for example, the stepwise regression analysis.

- Eq. 7 requires an increasing index for the average to apply. The entire block of Equations 6-8 needs to be revised.

 

1. Results

 

- the assertion that is given in line 200 “there was an increase in the frequency of drought specifically after 1999” holds for sub-basin 1 but not for sub-basin 2. How do the authors explain this difference? Are there any other factors influencing the drought regime it is worth including in the analysis?

-the results about drought occurrences, which are given between lines 202 and 217 are not verifiable anywhere in the text. What is the method used to count the drought events? Did the authors adopt specific thresholds of severity?

-the authors drew some conclusions about the trend of drought severity from Figure 3. Did they make some statistical analysis to support such a conclusion?

- I understand from the text from Line 228 to 233 that a nonparametric Mann-Kendall test has been applied on the time series of the number of months identified as drought months (which threshold?) per year. However, the remainder of Figure 4 is a bit confusing because Figure 4 reports the time series of SPI as differences between the two basins.

- the same comments given above apply to Section 3.1.2

-Sec 3.2.1 Coefficient of determination should not be used to compare different models. Consequently, comments on results associated with Table 1, are not supported by a robust methodology.

 

1. Discussion

 

-the statement of decreasing precipitation trend because of graphs in Figure 10 is not supported by robust statistical analysis.

-it is not explained how the trend coefficients for a brief period of 5- or 10-years have been computed.

 

 

 

Specific comments

 

 

1. Ref [21] is not appropriate neither in the Introduction nor in Section 2.2.2.
2. Figure 1: the green color for mountains is misleading. Please, consider the possibility of inverting the color scale. Besides, the Soan River is not visible on the map.
3. Line 138: The standardization process maps the data onto a Gaussian distribution; therefore, SPI values are not restricted to -3, +3 interval being a small mass of probability in the distribution tails. Please, revise this assertion accordingly. 
4. Line 163: Ref [22] does not support the statement in the text.
5. Please, correct the spelling of “Metrological” here and there in the text.
6. Refs given on line 202 are not appropriate.
7. Figure 3: the double scale of SPI for the two basins is unneeded because the SPI range is the same for construction. The adding of horizontal lines would allow to better identify drought severity.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

I thank the authors for their efforts in revising the paper. I am satisfied with their responses to my queries.