Long-Term Spatiotemporal Variation of Droughts in the Amazon River Basin

Round 1

Reviewer 1 Report

Overview and general recommendation:

Authors presented a study about the spatiotemporal distribution of droughts in Amazonia by statistically analysing the Standardized Precipitation-Evapotranspiration Index at multiple time scales.

The topic is interesting and the paper is well written. The introduction well describes the background of the research and the aims are clearly stated. I particularly appreciate the scheme in fig. 2 which allows the reader to follow the proposed methodology notwithstanding the large amount of data sets.  

Minor changes to the manuscript are suggested and annotated in the attached pdf file. Some concerns are shown below.

Main comments:

1. “any coupling”. It would be better to replace with “any relationship”.
2. “the extreme drought events tended to be more frequent since 1960 onward”. According to me this finding is biased by a low number and/or low quality of available data prior to this period.
3. L347: Pearson correlation. The rainfall in the considered period is a Gaussian variable? If data are not normally distributed, a non-parametric correlation coefficient (e.g. rho of Spearman) is more suitable.
4. Discussion: Given the high variability of the study area topography, some considerations on the strong control of surface orography on rainfall and droughts stability should be added.
5. Fig.1. Please, check the legend of fig.1d. The maximum elevation shown in the figure is 360 m but it should be 3600 m.
6. Fig.6. There is a sort of overlapping in Longitude that makes the coordinates not easy to be read. I suggest to use a lag of 10° in labelling for the W-coordinate to improve the readability of the figure.

Comments for author File: Comments.pdf

Author Response

COVER LETTER

The authors would like to thank both reviewers for their very constructive comments that we believe has made a significant positive impact on the quality of the paper. Attempts were made to incorporate all suggestions. If the suggestions were not incorporated, an explanation is provided.  Please see detailed responses to each comment below.  Thank you again for taking the time to review this article.

Reviewer: 1

Main comments:

1. “Any coupling”. It would be better to replace with “any relationship”.

Thank you. We agree with the reviewer in this suggestion. The suggested change has been made. Please refer to the line 244.

2. “The extreme drought events tended to be more frequent since 1960 onward”. According to me this finding is biased by a low number and/or low quality of available data prior to this period.

Thanks for this comment. This is a good point that needs clarification. We agree with the reviewer in that the CRU TS-based SPEI dataset during the first half of the 20^th century is based on a relatively low number of ground-based climatic data in the Amazon River basin; so one might expect that extreme drought events are not well captured during this period. Therefore, a caution has been added to this sentence. Please refer to the lines 336-340.

3. L347: Pearson correlation. The rainfall in the considered period is a Gaussian variable? If data are not normally distributed, a non-parametric correlation coefficient (e.g. rho of Spearman) is more suitable.

Thank you for this suggestion. For each variable, we applied a Shapiro-Wilk normality test to verify the normality assumption. Most of these time series (i.e. SPEI, EVI, T2M, and rainfall) and their anomalies showed a non-Gaussian distribution. Therefore, the Pearson correlation coefficient was replaced by the rho of Spearman in Table 3. Figure 2 was also updated with new non-parametric correlation analysis. Please refer to the lines 381-386 and Figure 2.     

4. Discussion: Given the high variability of the study area topography, some considerations on the strong control of surface orography on rainfall and droughts stability should be added.

The authors agree with the reviewer in that the drought conditions may be controlled by terrain characteristics in the ARB. To explore this relationship in detail, we compared the number of dry spells (NDS) and the average duration of the dry spells (DDS) (see Figure 9) against the elevation, land cover and type of climate during 1975–2018. To do this, for each SPEI pixel, the elevation was derived from SRTM, the land cover was derived from the European Space Agency Climate Initiative’s Land Cover Project (ESA CCI Land Cover), and the Köppen-Geiger climate classes was derived from a gridded climate dataset developed by Beck et al. [2018][1]; these datasets are briefly defined in lines 198 – 212, while the methodology is mentioned in lines 256-273. All datasets (SPEI at 0.5 degree; ESA CCI Land Cover at 300 m; SRTM elevation at 250 m; type of climate at 1 km) were resampled using the nearest neighbour technique to match the 0.5 degree latitude by 0.5 degree longitude grid of SPEI. Then, we used cluster analysis on these time series and the coordinates each SPEI pixel in order to identify homogeneous groups in terms of NDS, DDS, elevation, land cover, and type of climate. Please refer to the lines 512-545 and Figure 10, where we addressed this part of the question.     

5. Fig.1. Please, check the legend of fig.1d. The maximum elevation shown in the figure is 360 m but it should be 3600 m.

Thanks for the catch.  The value of 360 in the legend of Figure 1d refers to the lower limit of the range. This range has been modified (from ‘> 360’ to ‘360 – 5800’) in order to avoid the confusion with the maximum elevation. Please refer to the legend of Fig. 1d.      

6. Fig.6. There is a sort of overlapping in Longitude that makes the coordinates not easy to be read. I suggest to use a lag of 10° in labelling for the W-coordinate to improve the readability of the figure.

Thanks for the suggestion. A lag of 10° in labelling for the W-coordinate has been used. Please refer to Fig. 6.      

Minor comments:

1. Line 123 ‘…the CRU TS dataset to delineate those areas...’

Thanks for the suggestion. The suggested change has been made. Please refer to the line 123.

2. Why in fig. 1b a different time span is showed?

Thanks. Fig 1b shows the mean annual rainfall derived from the Climate Hazards Group InfraRed Precipitation with Stations (CHIRPS; temporal coverage: 2000-2018), whereas lines 157-160 refers to the timespan of the CRU TS-based SPEI dataset (temporal coverage: 1901-2018). That is, they are different variables and datasets.

3. Line 253, Fig3: I suggest to add the years of reference.

Thanks for the suggestion. For clarity, the reference period used for the calculation of SPEI has been added. Please see the caption in Figure 3, line 288; and Figure 4, line 310.

Other minor changes suggested by the reviewer 1 have been highlighted in yellow.

Additionally, others changes closely related with your suggestions have been done. They are also highlighted in yellow. Thank you very much for helping us to improve the overall quality of our manuscript. We hope our revision meet your approval.

References:

[1] Beck, H. E., Zimmermann, N. E., McVicar, T. R., Vergopolan, N., Berg, A., & Wood, E. F. (2018). Present and future Köppen-Geiger climate classification maps at 1-km resolution. Scientific data, 5, 180214.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this paper, authors tried to assess long-term spatiotemporal variation of droughts over the Amazon River basin but there are some points that need to be considered as follows: 1. It’s not clear how the authors dealt with uncertainty caused by using different geospatial datasets with different spatial and temporal scales. And if any homogeneity test has been done before using different datasets. 2. Base on the content the ground-measured data are collected over some stat 13 streamflow gauge stations and 5 rainfall gauge stations, but it is not clear that how the authors could extend only 18 point-based data over the whole of the study area. And how it’s possible to extend a few point-based data to the whole of such a big area? 3. It’s not clear why the authors tried to combine different parameters and then compared the results with SPI as one commonly used drought index. Furthermore, SPI is a meteorological drought index that can be used to reveal drought occurrence and measure drought intensity over only rainfed areas, but it cannot be used over the irrigated regions. Therefore, this point should be clearly indicated in the title, abstract, and the content of the paper that the main topic of this paper is only “meteorological drought”. 4. For better results, I suggest that authors consider cluster analysis to show if those trend analyses have better results over the group of stations that may have a similar condition in terms of climatological conditions. In overall I recommend that the authors make a major revision to this manuscript considering all the above-mentioned points.

Author Response

COVER LETTER

The authors would like to thank both reviewers for their very constructive comments that we believe has made a significant positive impact on the quality of the paper. Attempts were made to incorporate all suggestions. If the suggestions were not incorporated, an explanation is provided.  Please see detailed responses to each comment below.  Thank you again for taking the time to review this article.

Reviewer: 2

Main comments:

1. In this paper, authors tried to assess long-term spatiotemporal variation of droughts over the Amazon River basin but there are some points that need to be considered as follows: 1. It’s not clear how the authors dealt with uncertainty caused by using different geospatial datasets with different spatial and temporal scales. And if any homogeneity test has been done before using different datasets.

Thanks, this is a good question that needs clarification. All gridded datasets were clipped using a shapefile of the ARB as a mask while conserving their native spatial resolution to avoid the spatial uncertainty inherent in the interpolation techniques (see lines 214-216). For the applied wavelet coherence analysis (WCA), the area-averaged values of SPEI, EVI, T2M, and TRMM-based rainfall were paired during their common periods (i.e. 2001-2018, 1980-2018, and 1998-2018, respectively) instead of the period 2001-2018 to avoid the loss of information and capture also the underlying modes of variability with low frequency in SPEI, EVI, T2M, and TRMM-based rainfall (see line 248). The same criterion was applied when the coupling between the drought conditions and streamflow was examined (see lines 253-255). Regarding the homogeneity of datasets; the CRU climate dataset developed by the Climatic Research Unit (University of East Anglia) is based on a large number of stations with good quality control and homogeneity check, so the SPEI derived from this dataset is a reliable data for drought studies (see lines 166-171). Besides, details on the validation, calibration, and quality control of other datasets can be found in the links shown (see lines 197-212; and lines 186-188).

2. Based on the content the ground-measured data are collected over some stat 13 streamflow gauge stations and 5 rainfall gauge stations, but it is not clear that how the authors could extend only 18 point-based data over the whole of the study area. And how it’s possible to extend a few point-based data to the whole of such a big area?

Thanks for the question. The reviewer makes a very important point that involves making clear distinction between a meteorological drought episode and a hydrological drought episode. The main impetus behind looking at monthly rainfall anomalies from 5 rainfall gauge stations was to show their changes moderately concomitant with the temporal persistence of drought conditions reflected by the SPEI12 at the gauge station scale (i.e. a pixel-station comparison); this motivation is mentioned in lines 467-470. When we analyzed monthly streamflow anomalies from 13 streamflow gauge stations, our attention was focused on the spatial coverage and persistence of drought conditions within the sub-basins upstream of the streamflow gauge (that is, the red areas located upstream of the streamflow gauge in figures 8c to 8h), due to that the propagation of one meteorological drought event into one hydrological drought requires this condition (please see lines 478-482). As expected, a strong coherence of streamflow anomalies was associated with the persistence of drought conditions in sub-basins upstream of the streamflow gauge (please see Figure 8).

3. It’s not clear why the authors tried to combine different parameters and then compared the results with SPI as one commonly used drought index. Furthermore, SPI is a meteorological drought index that can be used to reveal drought occurrence and measure drought intensity over only rainfed areas, but it cannot be used over the irrigated regions. Therefore, this point should be clearly indicated in the title, abstract, and the content of the paper that the main topic of this paper is only “meteorological drought”.

The focus of the current paper is to evaluate the long-term spatiotemporal patterns of drought events under global warming (see lines 58-65), the results of which can be used by future studies to focus on the mechanisms responsible for the signals identified in particular regions of the Amazon River Basin. The SPEI takes into account not only rainfall but also the temperature in its calculation, so this is more appropriate than SPI for drought monitoring under global warming (see lines 80-86). For this reason, the SPEI was chosen in this study. As the SPI, the SPEI has multi-scalar features, so the ability to identify different drought types. In a huge diversity of studies[1], usually a time scale of 3 months or less is adequate for agricultural drought monitoring (also termed as short-term drought), while 12 months or more to capture reasonably well the hydrological drought (also termed as long-term drought) (see lines 162-164). On the other hand, the analysis of EVI against SPEI at different time scales was used for assessing vegetation response to drought conditions (see lines 379-386), while the visual comparison of SPEI with streamflow anomalies at 13 streamflow gauge stations is a classical approach to assess hydrological droughts (see lines 456-466). Note that the above mentioned approaches addressed different type of drought in the ARB (i.e. meteorological, agriculture, and hydrological). We think that the main topic of this paper goes beyond the meteorological drought, so the major change of title, abstract, and the content of paper, which as the reviewer has alluded to, does not make sense.  

Regarding irrigated regions, according to the FAO[2], the irrigated area of Brazil was 5,066,000 ha in 2016, which is equivalent to 0.83% of the entire basin. From this estimation, one can infer that the percentage of area occupied by irrigated agriculture in the entire ARB is insignificant. Thus, the uncertainty in the detection of drought conditions on irrigated areas is also insignificant.   

4. For better results, I suggest that authors consider cluster analysis to show if those trend analyses have better results over the group of stations that may have a similar condition in terms of climatological conditions. In overall I recommend that the authors make a major revision to this manuscript considering all the above-mentioned points.

Many thanks for the suggestion. For this revised version of manuscript, we considered cluster analysis to explore in more detail the association between the droughts (in terms of frequency and duration) and different local factors (please see lines 512-545). This analysis was not applied to the monthly rainfall and streamflow anomalies from station gauges due to the aforementioned points in the second question. Additionally, others changes closely related with your suggestions and first reviewer have been done. They are highlighted in yellow. Thank you very much for helping us to improve the overall quality of our manuscript. We hope our revision meet your approval.

References:

[1] Pei, Z., Fang, S., Wang, L., & Yang, W. (2020). Comparative analysis of drought indicated by the SPI and SPEI at various timescales in inner Mongolia, China. Water, 12(7), 1925

Mathbout, S., Lopez-Bustins, J. A., Martin-Vide, J., Bech, J., & Rodrigo, F. S. (2018). Spatial and temporal analysis of drought variability at several time scales in Syria during 1961–2012. Atmospheric Research, 200, 153-168

Sun, Y., Liu, S., Dong, Y., Dong, S., & Shi, F. (2020). Effects of multi-time scales drought on vegetation dynamics in Qaidam River Basin, Qinghai-Tibet Plateau from 1998 to 2015. Theoretical and Applied Climatology.

[2] FAO AQUASTAT at https://bit.ly/38EPs40

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

I recommend this paper for publishing in the present form.