the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A re-examination of the dry gets drier and wet gets wetter paradigm over global land: insight from terrestrial water storage changes
- 1State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan, 430072, China
- 2School of Environment and Society, Tokyo Institute of Technology, Yokohama 226-8503, Japan
- 1State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan, 430072, China
- 2School of Environment and Society, Tokyo Institute of Technology, Yokohama 226-8503, Japan
Abstract. The “dry gets drier and wet gets wetter” (DDWW) paradigm has been widely used to summarise the expected trends of the global hydrologic cycle under climate change. However, the paradigm is challenged over land due to the choice of different metrics and datasets used and is still unexplored from the perspective of terrestrial water storage anomaly (TWSA). Considering the essential role of TWSA in wetting and drying of the land system, here we built upon a large ensemble of TWSA datasets, including satellite-based products, global hydrological models, land surface models, and global climate models to evaluate the DDWW hypothesis during the historical (1985–2014) and future (2071–2100) periods under various scenarios with a 0.05 significance level. We find that 28.1 % of global land confirms the DDWW paradigm, while 23.3 % of the area shows the opposite pattern during the historical period. In the future, the DDWW paradigm is still challenged with the percentage supporting the pattern lower than 20 %, and both the DDWW-validated and DDWW-opposed proportion increase along with the intensification of emission scenarios. The different choices of data sources and varying significance levels (0.01–0.1) have subtle influences on the evaluation results of the DDWW paradigm. Our findings will provide insights and implications for global wetting and drying trends from the perspective of TWSA under climate change.
- Preprint
(10550 KB) -
Supplement
(11875 KB) - BibTeX
- EndNote
Jinghua Xiong et al.
Status: final response (author comments only)
-
RC1: 'Comment on hess-2022-190', Anonymous Referee #1, 18 Jun 2022
General comments:
This is a resubmitted manuscript and this is my second review. This study re-examine the “dry gets drier and wet gets wetter” (DDWW) paradigm using the terrestrial water storage anomaly (TWSA) derived from GRACE observational products, land surface models, and GCMs. The results showed the global patterns of dryness/wetness trends in both history (1985-2014) and future (2071-2100).
In this version, the authors have improved the text, added discussion, and provided more uncertainty analyses. I am happy with the authors’ efforts. However, there are substantial issues which need to be addressed. The authors should set out to solve scientific problem rather than analyzing data. At present, I did not feel the new knowledge and new (and convincing) methods provided by this paper. At least, the authors have not fully express the innovation and significance of this study.
- Concept: The authors shouldrecall the original meaning of "dry gets drier and wet gets wetter" paradigm from existing studies, because the title/the authors intend to perform a “re-examine” work. I think the authors acknowledge that the DDWW rule is used to explain the changing trend of surface dryness/wetness or climate condition, while this study explains the DDWW rule from a TWS perspective which includes groundwater/glacier changes. The GRACE observation contains the signal of changes in groundwater/glacier. As the climate warms, ice/glaciers are degrading with an increase in runoff/soil moisture (moisten the land surface). Meanwhile, as the mass decreases (water flows away), what GRACE observes is a decrease trend in gravity (drying). There are processes in the opposite direction. As such, the TWSA trends can be opposite with the previous studies focusing on the land-surface conditions (soil moisture/runoff and ET) (Wang et al., 2021: Long-term relative decline in evapotranspiration with increasing runoff on fractional land surfaces; Yang et al., 2019: Combined use of multiple drought indices for global assessment of dry gets drier and wet gets wetter paradigm). Rather than a new perspective, I would also think of this study as a simulation or an application of GRACE, land surface models, and climate models.
- Method: In the discussion, the authors need to justify why it is necessary to assess the changes in dryness/wetness from a perspective of terrestrial water storage change? What are the advantages of the methodology used in this method? This study has many redundant operations (e.g., the use of GRACE to correct GCMs). I feel if the study directly using P-ET is more convincing than using partial outputs (soil moisture, snow water...). While changes in TWSA do not equal to changes in surface dryness/wetness, there should have "bridges" to connect the integrated TWS and various land-surface processes (runoff, soil moisture and ET) (Trautmann et al, 2022: The importance of vegetation in understanding terrestrial water storage variations). It is a pity that this study did not find such “bridges” as it leaned toward analyzing data. The use of TWS retrieved by the GRACE to correct GCM simulations is not convincing. Not only are there many uncertainties in the GRACE retrieval product, but also what GRACE observes is completely different from what GCMs simulate (Table S2). Since these models express different objects, how can these outputs ensemble? What will happen if the study do not use GRACE to correct the GCM simulations as most climatologists do? There are still have a prediction result from GCM, right? What are the differences? One way is to show that the corrected results are more reliable than the previous one, which may involve using in-situ observed data. Moreover, the authors criticize the use of P-ET as an indicator to identify dry/wet changes, but I think P-ET is closer to changes in TWS because various hydrological models and GCMs appear to do not account for surface water storage. I suggest the authors provide a technical route.
- Results and mechanism: This study does not involve mechanism analysis, and does not analyze why some typical places are getting drier or wetter. Fig. 2 and Fig. 4 make no sense as they are another displays of the same results in Fig. 1 and Fig. 3. Although this division method was used in the IPCC6 and even considered popular by the authors, it did not bring any innovative insights to this study. Moreover, they are difficult to interpret. Instead, the readers are more care about how dryness/wetness changes in time and why there are changes happen.
- Innovation and significance: The authors need to rethink and justify what are the new results or developments reported in this study? Why are these new results or developments significant?
Specific comments:
- Line 9-10 and Line 17-18: These statements are contradicted. You are saying the DDWW is challenged due to the choice of different metrics and datasets used, but you also stated the different data sources have subtle influences on the evaluation results.
- Line 21: “The hydrological conditions of the land surface have experienced...”. The first sentence of this manuscript is talking about land surface condition. This is contrary to the author’s argument that they are not concerned with the surface dryness/wetness, but with the entire land system.
- Line 37: What are oceanic records?
- Line 45: P-ET is the amount of water remaining in the land system, but the components in the GCMs (soil moisture and snow water) and VIC (moisture), Noah (soil moisture, snow, and canopy water) are parts of the water stored in the land system (Table S2), and thus the models lack some components of the terrestrial water storage.
- Line 70: Is long-term P-ET approximately equal to the change in terrestrial water storage (TWS)? Why the authors do not use P-ET to construct an index and to perform the prediction of TWSA? Instead, this study uses partial outputs of soil moisture/snow data in the GCMs.
- Table 1: What are the differences between GRACE reconstructions and GRACE masons solutions?
- Figure 2 makes no sense and it is hard to interpret. Instead, the manuscript can present dryness/wetness changes in some key regions here.
- Line 273-276: Why did the authors use AI derived from CRU data to define wet and dry zones rather than TWS-DSI? The following DDWW analyses are based on the changes of TWS-DSI.
- Line 281: “We compare AI and TWSA derived from DATASET and CMIP6 between 1985 and 2014 in Figure S6”. Is it a result comparison between Figure S5 and Figure S6 here?
- Line 273-283: These contents about how to operate should be adjusted to the method section?
- Line 284: “Figure 3 illustrates the test results...”. This is not a good way to express the content of figures.
- Line 324-325: “Greve and Senevirtne (2015) used climate projections from CMIP5 to establish the measure for assessment of the DDWW paradigm...”. The method used by Greve (2015) is more acceptable to peers.
- The presentation of Figure 4 makes no sense and it is hard to interpret. It may be more interesting to modify it to temporal changes over key areas.
- Line 355-358: “...resulting in the lack of certain TWSA components.”. Why the authors do not use P-ET derived from the models to represent TWSA? In this case, none of the flaws discussed here exist.
- Line 417-419: Please explain what are the advantages of the developed TWS-DI?
- Line 447-462: The authors need to refine the conclusions. These do not look like conclusions, at least not serving for the purpose of this study.
- The authors need to recheck and simplify the expression andlogic of the entire manuscript, as many expressions seem redundant and use uncommon words, making it difficult to read.
- AC1: 'Reply on RC1', Shenglian Guo, 19 Jul 2022
-
RC2: 'Comment on hess-2022-190', Yannis Markonis, 22 Jun 2022
General Comments
The manuscript examines the dry gets drier and wet gets wetter (DDWW) paradigm from a water storage perspective over the terrestrial fraction of the water cycle. The topic is contemplated within the scope of Hydrology and Earth System Sciences via the study of the spatial and temporal characteristics of global water resources and would be of interest to its readership. It is a topic that remains under debate in the last decade and studies that bring new evidence, such as this, can substantially impact the ongoing research. The analysis based on the terrestrial water storage drought severity index is reasonable and aims to quantify wet/dry regime trends under different warming scenarios and assess the agreement or rebuttal of the DDWW paradigm. The study has clear research hypothesis and objectives, that are reflected in the results, and help us increase our understanding of the changes in the global water cycle. However, there are some important aspects that should be addressed before being considered for publication.
Before moving to the specific revision comments, it is important to note the lack of consistency across the hydrologic and climatic communities in the use of the terms “wet/wetter” and “dry/drier”, which in some cases can be misleading (Roth et al. 2021). Thus, it is easy to misinterpret whether a variable is truly appropriate for describing wetting or drying over a region. In this study, a variable (terrestrial water storage) which is not directly involved in the formulation of DDWW paradigm (precipitation/evaporation) is used to validate the paradigm itself. This raises concerns about the applicability of terrestrial water storage as a metric that can confirm or falsify the DDWW hypothesis. The study needs to convincingly prove this. A feasible way to achieve it would be to compare the current findings with P – E, taken from the models (GHMs, LHMs, and CMIP6). This also holds the opportunity to highlight the mechanisms involved in the observed changes and/or pinpoint the biases in the models. Some caution should be taken here since the comparison should also consider surface runoff to satisfy the budget closure. In any case, though, this can help to bridge the different methodologies and explore their complementarity.
Specific Comments
Lines 1-2: The term “re-examination” should be reconsidered and perhaps be replaced with something like “alternative/complementary examination”.
Lines 25-27: It would be more appropriate to reference the original study of Held and Soden (2006).
Lines 34-42: You might want to look at the work of Roderick et al. (2014).
Lines 43-74: This paragraph could be split into two (line 56 perhaps?) to improve readability.
Line 84 (Table 1): Since the three GRACE reconstructions come from two papers it could help the readers if they were also somehow distinguished in the table.
Lines 84-85: Averaging the datasets always comes with certain challenges. For example, in this case we would expect that the three reconstructions of GRACE are strongly correlated and thus their impact to the estimation of the mean would be stronger. A cross-correlation matrix between the datasets would help to assess the magnitude of the impact and comment on it in the manuscript.
Lines 87-88: Can you please provide some information about the percentage of the missing values, as well as their distribution among the number of consecutive missing values?
Lines 107-108: “kinds” might not be necessary here.
Lines 167-168: Please also cite the original paper of Hempel et al. (2013) which has been used by Xiong et al. (2022).
Lines 177-178: Before applying linear interpolation and assessing the statistical significance of the slope the auto-correlation structure of the time series should be investigated. High values of auto-correlation coefficient could result to biased estimates of t, so if this is the case, alternative methods of slope significance should be applied (Hamed and Rao, 1998; Yue et al. 2002). In addition, the reference to the work of Greve et al. (2014) should be revisited as a different statistical test is applied at that study than the t-test.
Line 180: Should the region be considered as “uncertain” or should it be considered a region with no or non-significant long-term change?
Lines 183-184: It is preferable to keep a single tense for the whole manuscript (past/present). This comment also applies for other lines.
Line 185: Citation of the report is missing.
Lines 190-208: My understanding after reading the bias-correction method of Xiong et al. (2022) is that that the CMIP6 ensemble was bias-corrected using GRACE TWSA. If this holds true, then the evaluation of the TWSA derived from the ensemble mean of CMIP6 raises some questions about its validity and therefore NRMSE is lower compared to DATASET.
Lines 214-216: Is there any likely explanation about the increase in the range of DATASET after 2010? Perhaps it can be linked to the decline of TWSA of a specific dataset.
Line 219: It would be helpful to remind the readers that for the historical period DATASET is used, and not only the CMIP6 data that are used for the scenarios.
Line 221: Since you are referring to the slope “/a” is redundant. Still if you would like to keep it, you could consider replacing it with “/yr”.
Line 228: I think “increasing” is the correct word here.
Lines 228-229: There is no study related to S. Europe in the references. Most importantly, in Figure 1b it appears that SE. Europe is wet and the rest of the south not significantly drier. This contradicts older studies reporting a drying trend over the Mediterranean (e.g., Hoerling et al. 2012) and could shed new light to the ongoing discussion about the current and future conditions of S. Europe, so I would recommend elaborating more.
Lines 233-235: Do you mean that your results for these regions disagree with the previous studies? If yes, you could clarify a bit and discuss potential reasons for the disagreement. Also, you might want to replace “alternatively” with “on the contrary”.
Line 236: In SSP126 scenario, S. Europe also has a strong wetting trend.
Lines 236-245: This paragraph discusses only the SSP126 scenario, while the other two more probable scenarios remain uncommented. It would be nice to discuss the differences between each projection scenario and highlight the regions that all scenarios agree. Another striking difference appears in spatial clustering between the historical period and the model results in terms. It is evident that in the historical period there is stronger spatial homogeneity, while the models replicate this behavior only for SSP126 scenario. Any idea why this is happening?
Line 245: You could consider rephrasing to “a pattern also considered”.
Lines 246-259: It is not very clear what the investigation of the changes over the SREX regions offers to the study.
Line 260: The legend is not very clear (some spaces between the numbers would help). Also, the scale order should be from higher to lower. Since the stippling marks are not very clear, could you please remove them and reproduce this map with only the statistically significant slopes in the supplementary material?
Line 266: Same concerns here for the legend as the previous comment on Figure 1. The bar plot needs also revising as D should be above D (p<0.05). Additionally, I am not certain that the pie plots help the readers and are not discussed in the manuscript. You might want to consider removing them and adding them as a separate figure in the supplementary material.
Lines 273-276: These lines would fit better to the Methods section.
Line 274: Please elaborate about transitional regions.
Line 290: You might want to remove “Under climate change” since you mention the SSP126 scenario.
Lines 295-304: Again, I have similar concerns about SREX regions as the ones for lines 246-259. If you decide to keep them and justify the added value they offer in the analysis, please consider presenting these results in an individual paragraph.
Line 319: “In climate model projections”, would be more appropriate than “Under climate change”.
Line 336: Please see the comment about Figure 2 (Line 266) regarding the pie charts.
Line 343: I would recommend using “Non-significant” instead of “Uncertain” here.
Line 348: It would be helpful to the readers to link the limitations with some suggestions for future research, especially for the first two paragraphs.
Lines 379-385: Similarly to lines 190-208, bias-correction comes with certain limitations which need to be mentioned here.
Line 382: A minor typo here “bias correction”.
Lines 403-407: It would be preferable to present the differences to the 0.05 threshold both in text and Figure S13.
Lines 421-424: A quite strong statement appears here. Are there any other studies that support it or is it derived only by the results of this study?
Line 457: Another minor typo “is still challenged”.
Line 463: It would be very beneficial to the community to share the data used in the manuscript figures, as well as the DATASET and bias-corrected CMIP6 members. This will have a positive impact on the study itself, as it will improve its reproducibility.
References
Hamed, K. H., & Rao, A. R. (1998). A modified Mann-Kendall trend test for autocorrelated data. Journal of hydrology, 204(1-4), 182-196.
Hempel, S., Frieler, K., Warszawski, L., Schewe, J., & Piontek, F. (2013). A trend-preserving bias correction–the ISI-MIP approach. Earth System Dynamics, 4(2), 219-236.
Held, I. M., & Soden, B. J. (2006). Robust responses of the hydrological cycle to global warming. Journal of climate, 19(21), 5686-5699.
Hoerling, M., Eischeid, J., Perlwitz, J., Quan, X., Zhang, T., & Pegion, P. (2012). On the increased frequency of Mediterranean drought. Journal of climate, 25(6), 2146-2161.
Roderick, M. L., Sun, F., Lim, W. H., & Farquhar, G. D. (2014). A general framework for understanding the response of the water cycle to global warming over land and ocean. Hydrology and Earth System Sciences, 18(5), 1575-1589.
Roth, N., Jaramillo, F., Wang-Erlandsson, L., Zamora, D., Palomino-Ángel, S., & Cousins, S. A. (2021). A call for consistency with the terms ‘wetter’and ‘drier’in climate change studies. Environmental Evidence, 10(1), 1-7.
Yue, S., Pilon, P., Phinney, B., & Cavadias, G. (2002). The influence of autocorrelation on the ability to detect trend in hydrological series. Hydrological processes, 16(9), 1807-1829.
- AC2: 'Reply on RC2', Shenglian Guo, 19 Jul 2022
-
RC3: 'Comment on hess-2022-190', Anonymous Referee #3, 22 Jun 2022
This is my second review of the manuscript, which has been resubmitted after the previous discussion round.
The authors present a re-examination of the dry gets drier, and wet gets wetter paradigm over global land, based on terrestrial water storage estimates from different sources. They make use of GRACE reconstructions, global hydrological models, and land surface models, as well as CMIP6 models for the future perspective. They conclude that the DDWW paradigm is challenged both in the historical period but also in the future.
Overall, the authors took into account my points previously made and the manuscript considerably improved compared to the initial submission. In particular, the authors corrected the calculation of the percentage values by area-weighing the grid boxes and added more discussion on the uncertainties inherent in their analysis.
As such, I'm happy with the changes made. However, there are still some open methodological points that need to be addressed (see specific comments). Also, in parts the manuscript might need to be checked for grammar and wording by a native speaker.
Specific comments:
Line 81: I suggest changing the title to “Data pre-processing”
Line 84: Change to “(see Table 1 and next sections)”
Line 86: Change to “resampled to 1° x 1° resolution to compare against the average”
Line 89: Change to “As for DATASET, the members of the CMIP6 ensemble have been resampled to …”
Line 106: “We have implemented an ensemble …” It’s not clear in this sentence that you take the mean of the reconstructions derived from the three forcing datasets. Please rephrase.
Line 107/108: Why only listing a subset of the variables used for the derivation of the CSR reconstruction and not all?
Line 109: Change to “these three GRACE reconstructions”
Line 154: Change to “for which TWSA outputs are”
Line 158: “and future periods (Krishnan and Bhaskaran, 2020)” Why citing this specific paper about wind speed in the Bay of Bengal? This does not appear to be the standard reference for CMIP6. Please use Eyring et al. 2016 instead.
Line 162/163: “the sum of total soil moisture and snow water, which has been proven reliable to assess the TWS changes”. This is already mentioned a few lines above, please merge and rephrase.
Line 192/193: Change to “… (NRMSE) between the mean GRACE TWSA and the ensemble means of DATASET and CMIP6 data after bias correction during the period April 2002-December 2014, with the NRMSE calculated as the ratio of RMSE …”
Line 193: What is meant with "change range" of TWSA? The range of the TWSA values (i.e., max. minus min.) from DATASET and CMIP6 respectively? Please specify.
Line 201: Change to “uncertainties in the CMIP6 simulations that remain even after undergoing the bias correction”
Line 210: Change to “The GRACE TWSA ranges from roughly −20 to 20 mm and shows …”
Line 211: Change to “A similar temporal pattern is captured …”
Line 214/215: Change to “Moreover, the fluctuation range of DATASET is generally greater than the CMIP6 range before 2010. After 2010, DATASET tends to underestimate TWSA compared to CMIP6 and GRACE, and shows an increase in range.”
Line 217/218: Please rephrase, simplify, or merge with the following sentence. You look at long-term trends here.
Line 219: “the historical period 1985-2014” Please clarify that the trend estimate for the historical period is based on DATASET (both in the text and in the figure captions). It would be interesting to see how the CMIP6 historical trends compare with the ones from DATASET.
Line 237: Change to “become wetter because”
Line 261, Figure 1: Please clarify in the caption what data source (i.e., DATASET or CMIP6) is used for the individual temporal subset.
Line 269, Figure 2: Change to “the bar plot displays the global percentage.”
Line 284: “Figure 3 illustrates the test results of DDWW paradigm at a 5% significance level (p=0.05)” Based on what test? Does the mentioned significance level relate to the test results for long-term trends? Please clarify how you derive these results in the methods section.
Line 332, Figure 3: Please clarify the data source (i.e., DATASET or CMIP6) used for the different temporal subsets. I would be interesting to see how the results for DDWW based on the two data sources compare during the historical period. This could help to shed more light on the applicability of the CMIP6 ensemble for investigating the DDWW paradigm also in future periods.
Line 340, Figure 4: ““D” and “W” indicate regions with drying and wetting trends, respectively.” I guess this does not belong to this figure caption?
Line 364: “reported to show underestimation or overestimation” -> variable-specific biases?
Line 421/422: “Despite the magnitude bias from satellite products, simulations of LSMs and GHMs, and GCMs projections, …” Not sure what is meant here? Please rephrase.
Line 453: “significance levels from 0.01 to 0.1” For the test on long-term trends?
- AC3: 'Reply on RC3', Shenglian Guo, 19 Jul 2022
Jinghua Xiong et al.
Jinghua Xiong et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 713 | 155 | 15 | 883 | 43 | 1 | 1 |
- HTML: 713
- PDF: 155
- XML: 15
- Total: 883
- Supplement: 43
- BibTeX: 1
- EndNote: 1
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1