the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Sep 2024
Global Assessment of Socio-Economic Impacts of Subnational Droughts: A Comparative Analysis of Combined Versus Single Drought Indicators
Abstract. The accurate assessment of the propagation of drought hazards to socio-economic impacts poses a significant challenge and is still less explored. To address this, we analyzed a sub-national disaster dataset called the Geocoded Disaster (GDIS) and evaluated the skills of multiple drought indices to pinpoint drought areas identified by GDIS. For the comparative analysis, a widely used Standardized Precipitation Index (SPI), Normalized Difference Vegetation Index (NDVI), Standardized Soil Moisture Index (SSI), and Standardized Temperature Index (STI) were globally computed at the subnational scale. In addition, we developed a novel Combined Drought Indicator (CDI), which was generated by a weighted average of meteorological and agricultural anomalies. Out of 2142 drought events in 2001–2021 recorded by GDIS, NDVI, SSI, SPI, and STI identified 1867, 1770, 1740, and 1680 drought events, respectively. In terms of the skill to cover GDIS-documented drought events, CDI outperformed the other single-input-based drought indices and identified 1885 events. This emphasizes the importance of using CDI to evaluate socio-economic drought risks and prioritize areas of greater concern.
- Preprint
(2500 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on hess-2024-245', Jasmin Heilemann, 28 Dec 2024
Manuscript title: Global Assessment of Socio-Economic Impacts of Subnational Droughts: A Comparative Analysis of Combined Versus Single Drought Indicators
General comments:
In general, it is an interesting paper within the scope of the journal that addresses the highly relevant issue of detecting socio-economic impacts of drought using biophysical indicators on a global scale. The paper proposes a Combined Drought Indicator (CDI), which is constructed by using single-input based drought indices (SPI, NDVI, SSI, STI) and performing a Principal Component Analysis (PCA). The authors show that the CDI outperforms the single-input drought indices in its ability to capture drought events observed in the global GDIS dataset.
While the paper presents a very relevant analysis with noteworthy results, what is currently missing from the paper is a discussion of the benefits of using the PCA method to construct the CDI. This discussion should include details of the benefits of PCA for the CDI, as well as a discussion of the applicability of the CDI for (regionalized) drought impact monitoring and prediction. Including this in the manuscript would significantly enhance the paper and give greater significance to the implications, with potential applications of the CDI beyond this paper. My specific comments are listed below.
Specific comments:
- Title: Global analysis of sub-national droughts: “Sub-national” droughts sound misleading, as droughts are not constrained by national borders. If possible, rephrase as “droughts at sub-national scale”, or similar.
- Abstract: “Out of 2142 drought events in 2001-2021 recorded by GDIS, NDVI, SSI, SPI, and STI identified 1867, 1770, 1740, and 1680 drought events, respectively. […] CDI outperformed the other single-input-based drought indices and identified 1885 events.” Consider adding percentages or otherwise present convincing quantitative results that show the superiority of the CDI more directly.
- Fig. 1: It is a bit confusing to frame the “wet” conditions as drought categories. Would it be possible to find a different notion, e.g. moisture? Or Drought/Wetness category?
- Lines 116-125: The topic of the paper are socio-economic impacts of droughts. However, socioeconomic impacts manifest very differently in different sectors. E.g., in the introduction, you mention urban areas/water shortages in dams (lines 25-36). The drought indices you chose (SPI, NDVI, SSI, STI) are mostly useful for the ag sector. Please elaborate on how this affects the results, and if/how the CDI can capture socio-economic impacts in non-ag sectors, e.g. urban areas.
- Section 3.2: Here, I miss a description of the reasons why the PCA method was chosen to construct the PCA. This is the main innovation of the paper, and should therefore be featured more prominently, also in the introduction. E.g., explain what the added value of the PCA is compared to other techniques to compute a CDI. Why are regression-based approaches not used? (e.g. is it an advantage that the PCA does not have a dependent variable?)
- Lines 227-229: What is the total number of observations for the single-based drought indices used in the PCA? Does this number meet the requirements of the no. of observations usually applied in PCA? Please specify.
- Lines 258: “…has been widely accepted in previous work.” Which previous work? Please provide citations. Please extend this to the other text when you mention previous work without giving references.
- Lines 258-265: The thresholds of the drought indices used for detecting the drought impacts listed in the GDIS dataset are very crucial, though the explanation remains too vague (it’s a simple process, but I had to read over the section several times to understand this). Please make this process more explicit, e.g. via adding a table. Also, I miss a clear explanation of how the spatial scales between the gridded drought indices and the sub-national GDIS events are matched for the detection of drought impacts (is it counted as drought event if more than half of the pixels in the GDIS area show a deviation below the drought threshold? Or do you first calculate the average of the drought indices across all grid points and then compare it with the thresholds?)
- Line 327: Please specify why you chose April as the month for displaying the PCA results. Does it represent the yearly average best? How important are intra-annual fluctuations? April is not a typical drought month in the northern or southern hemisphere.
- Table 2: You show the false-negative (when a GDIS drought event existed, but the drought index did not indicate a drought event) in the table as “not observed”. Likewise, what is the rate of false-positive cases (how often did the drought index indicate a drought event not reported in the GDIS?)? You discuss this in the text (lines 390ff), but it would be beneficial for the reader to understand the magnitude of these cases in numbers.
- Discussion: In the discussion, an important point would be how the CDI could be used/applied for drought impact forecasting and/or policy-making. Could the CDI (computed via PCA) help to improve drought impact forecasting? How does the regionalization of the CDI affect the capacity to be used for that purpose?
- Lines 570ff: You could additionally mention that text mining is a research field potentially providing alternative impact databases for droughts next to the GDIS.
- Line 550: “despite experiencing higher climatic anomalies, developed nations are less likely to be socio-economically affected …”. This statement needs to be specified. It needs to become clear that the higher climatic anomalies relate to the local climate, and are not compared in absolute terms. A small anomaly in an already dry climate can provoke much more negative drought impacts compared to a larger anomaly in a wetter climate. Otherwise, this suggests that climate/drought impacts in developed nations are higher than in developing countries, which is not the case.
- Lines 583-584: “Moreover, there are other methods and techniques that could be used to compute weights in CDI …” Like which methods? Please specify and give a short reason why they could be apt.
Technical corrections:
- Line 121: “combine drought indicator” -> correct to “combined”
- 3a: The legend in this panel is missing.
- 9: Second panel: “CDI vs. TCI”, please correct to “CDI vs. STI”
- Appendix 2: This figure shows four times the same plot. This should be corrected.
Citation: https://doi.org/10.5194/hess-2024-245-RC1 - AC1: 'Reply on RC1', Sneha Kulkarni, 10 Apr 2025
-
RC2: 'Comment on hess-2024-245', Anonymous Referee #2, 20 Feb 2025
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2024-245/hess-2024-245-RC2-supplement.pdf
- AC2: 'Reply on RC2', Sneha Kulkarni, 10 Apr 2025
-
RC3: 'Comment on hess-2024-245', Anonymous Referee #3, 27 Feb 2025
Title: Global assessment of socio-economic impacts of subnational droughts: A comparative analysis of combined versus single drought indicators
Authors: Kulkarni et al.
Summary
This study presents a global assessment of combined drought indicator (CDI) in identifying socio-economic drought impacts at the subnational level and compares it with single hydro-meteorological drought indices, such as SPI, STI, NDVI, and SSI. The GDIS dataset was utilized to represent observed socio-economic drought impacts. The authors found that CDI outperforms any single drought index in detecting drought events. Its performance is even higher when short drought events lasting 2 months or less are excluded in the analysis. This study highlights the importance of using CDI to evaluate socio-economic drought impacts.
Assessment
This paper analyzes the performance of CDI in identifying drought events recorded in the GDIS database. Furthermore, the study compares the performance of CDI with individual hydro-meteorological drought indices in detecting droughts. I find the work and its findings interesting, especially when I look at the title, which suggests an assessment of socio-economic drought impacts. However, the aim and title of the manuscript seems misleading. Additionally, the manuscript contains some typos and formatting issues. Below, I provide four general comments, aimed at improvement and clarification. I suggest the authors consider these comments in their revised manuscript.
General Comments
I have four general comments regarding the manuscript:
- The aim of this paper is to evaluate the performance of multiple drought indices, including CDI to pinpoint drought areas recorded in the GDIS database. This objective is also stated in the abstract. However, the title suggests an assessment of socio-economic drought impacts using both single and combined drought indicators. When I read the paper, I do not find any results related to drought impacts; instead, the analysis focusses solely on drought occurrences. I expected to see socio-economic impacts, such as economic loses, human mortality, etc. Therefore, I suggest either incorporating an analysis of socio-economic impacts or modifying the title to better reflect the study’s aim and findings. I personally make a clear definition between event and impact.
- The introduction section could be better structured for improved readability and logical flow. Currently, the introduction negins with a general statement about droughts (paragraph 1), followed by socio-economic drought impacts (paragraph 2), GDIS database (paragraph 3), and then go back again to drought types and indicators (paragraph 4), and finally discusses the CDI (paragraph 5). To enhance the coherence, I suggest reorganizing the introduction so that the discussion on droughts (paragraphs 1, 4, 5) comes first and then followed by impacts and the impact database (paragraph 2 and 3). By doing this, it would create a more logical and improve the overall flow of the paper.
- If the GDIS database includes different types of drought impacts, such as economic loses, human mortality, forest fire, yield loses, etc, I suggest establishing a link between drought indicators and their corresponding impacts. For example, if NDVI is the selected indicator for a certain region, I expect that agricultural impact is the dominant impact in that region. If the selected indicator is SPI, then maybe the dominant impact on that region might be related to meteorological drought, and so on. Or at least, I recommend discussing these relationship in the paper.
- Regarding drought indicators, the accumulation periods used for analyzing SPI, STI, and SSI are not clearly stated. I suspect that the authors have only used 1 month accumulation period. I suggest reconsidering the use of the term SSI for soil moisture drought. In many publications, SSI refers to the Standardized Streamflow Index, which is used to identify streamflow drought. The standardized soil moisture index is commonly referred as SMI or SSMI.
Line by line comments
L refers to line and P refers to page.
P1: Abstract: I suggest to expand the abstract in order to capture the summary of your study. I missed an explanation about PCA and results on long and short drought durations. In the EGU journal, you can have longer abstract.
P1L12: In this sentence, the study area should be mentioned, which is global. I only found this in Line 15.
P2L36: Here the authors can see an example of reference typo.
P2L40: In the last sentence of paragraph 1, the authors mention about drought propagation from meteorological drought to socio-economic drought. However, I miss an explanation on drought types. Later in paragraph 4, I read meteorological drought, agricultural drought, and hydrological drought.
P2L48: Again, I see typo on references -> Tiwari and Mishra (2019) “and” Wu et al. (2018).
P2L52: Please provide references for DIR and EDII.
P2L54: Again, missing comma between references.
P2L59-63: I suggest the authors to explain more about GDIS. Or at least describe GDIS in detail in the method section.
P3L77: Missing word “and” between Sandeep et al. (2021) and Tao et al. (2021).
P3L91: Missing space between parameter and (Jiao et al., 2019a). Starting from here, I will not mention one by one the typo regarding missing space or in references. Please check carefully throughout the manuscript.
P3L95-97: European Drought Observatory (EDO) also utilizes combined drought indices.
P4L122: Here, it is clearly stated GDIS drought events and not impacts.
P5L139: Better use x for indicating resolution.
P5L141: In this sentence, the authors mention: to explore the linkage between drought indicators and “socio-economic impacts”. It is drought impacts or drought events? It is very confusing. I make a clear distinction between event and impact (see my general comment).
P5L152: The authors may remove the word “originally” since ERA5 Land already has spatial resolution of 0.1 degree.
P6L170: Define what is socio-economic impact of drought in the GDIS data.
P6L175: Disaster no or disaster number?
P8L232-233: I think ith and jth should be in italic.
P9L253-254: How about if drought indices indicate drought but GDIS not, so is it false alarm?
P13: Figure 4, please provide alphabets a, b, c, and d for each figure in Figure 4.
P14L350: Same as Figure 4, please provide alphabets for Figure 5. Here the authors refer to Figure 5a but there is no figure 5a.
P14L355: What is AEP?
P14L360: The authors can also explain about false alarm. Furthermore, Table 2 is summarizing the findings and it is worth detailed explanation, such as the results with and without short droughts and the shifting.
P15L376: Beside the number, I think it is more meaningful to also present the percentage. How many percent drought was detected.
P16L409: Lower threshold here means index below 0 or long drought duration? I think the latter.
P17L426-433: In this paragraph, the authors could highlight which shifting month yields highest performance and false alarm.
P19: Figure 8. The authors could make a bold line for threshold -1 in order to improve the readability.
P19L471-472: Where can I see the figures showing the correlation between CDI and hydro-meteorological indices? The authors could provide these figures in Appendix.
P21L501-502: Make it clear that SPI-1 and SPI-3 are effective to detect meteorological and agricultural droughts while longer time scales are better for detecting hydrological droughts.
P21L530: Or maybe the impact is not related to meteorological drought impact?
P21L533: Same, maybe the impact is not related to soil moisture drought impact.
P23L571: Do the authors mean takeaway message?
P23L580-581: I think the statement here is very weak saying that lack of availability of high-res hydrological data. There are some high-res hydrological models that provide hydrological data. The Lisflood hydrological model of Europe provide 1x1 km spatial data for streamflow and soil moisture. Or do the authors mean in situ observation?
P23L592-593: Mention again the agricultural and meteorological drought indicators.
P24L601-602: Does GDIS provide drought socio-economic impacts or just drought events? (see my general comment). Please make it clear.
P32: Figure Appendix 1: missing figure g.
P33: Provide alphabetical figures.
Citation: https://doi.org/10.5194/hess-2024-245-RC3 - AC3: 'Reply on RC3', Sneha Kulkarni, 10 Apr 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 332 | 94 | 135 | 561 | 18 | 15 |
- HTML: 332
- PDF: 94
- XML: 135
- Total: 561
- BibTeX: 18
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 174 | 32 |
| China | 2 | 49 | 9 |
| Germany | 3 | 38 | 6 |
| Netherlands | 4 | 19 | 3 |
| Romania | 5 | 19 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 174