Land-Surface Sensitivity and Climate Variability Shape Global Wet-Dry Regime Transitions

Faiz, Muhammad Abrar; Zhang, Liangliang; Liu, Dong; Fu, Qiang; Li, Mo; Qi, Xiaochen; Li, Tianxiao; Cui, Song

doi:10.5194/egusphere-2026-4696

Preprints

https://doi.org/10.5194/egusphere-2026-4696

Preprints

18 Aug 2026

| 18 Aug 2026

Land-Surface Sensitivity and Climate Variability Shape Global Wet-Dry Regime Transitions

Muhammad Abrar Faiz, Liangliang Zhang, Dong Liu, Qiang Fu, Mo Li, Xiaochen Qi, Tianxiao Li, and Song Cui

Abstract. Hydroclimatic change is commonly assessed using gradual wetting or drying trends, yet many regions may shift among persistent wet, persistent dry, and transitional wet-dry states. The global geography of these regime transitions and their associations with land-surface conditions and climate variability remain insufficiently quantified. This study aimed to identify global wet-dry regime transitions across historical and future periods and to evaluate how these transitions relate to land-surface characteristics and large-scale climate variability. A stage-focused standardized index was developed to estimate terrestrial water-storage anomalies from monthly water-balance components, including precipitation, evapotranspiration, and runoff. The index was applied to global gridded observations from 1979 to 2023 and to CMIP6 projections. Four hydroclimatic regimes were classified: wet-gets-wetter, dry-gets-drier, wet-gets-dry, and dry-gets-wet. Land-cover diagnostics and large-scale climate indices were used to examine associations among observed transitions, land-surface conditions, and ocean-atmosphere variability.

Drying covered 52.9 % of the global land area during 1979–2000, with persistent drying as the dominant regime. From 1979–2000 to 2001–2023, persistent drying remained the largest category but declined to 28.7 %, followed by wet-gets-dry, wet-gets-wetter, and dry-gets-wet regimes, covering 22.7 %, 24.4 %, and 24.3 % of global land, respectively. Land-surface diagnostics showed stronger drying in croplands and urban areas, larger dry-to-wet transitions in wetlands, and elevated wet-to-dry transitions in barren regions. Climate diagnostics indicated that internal variability was linked to large-scale ocean-atmosphere patterns, with the strongest association for the Arctic Oscillation and weaker but spatially coherent associations with the Pacific Decadal Oscillation, North Atlantic Oscillation, and Southern Oscillation Index. Under SSP585, the CMIP6 ensemble mean indicated an increase in wet-gets-wetter regimes to 35.2 % by 2081–2100 and a reduction in transitional regimes. These findings provide a global diagnostic framework for identifying wet-dry regime transitions and highlight regions where land-surface sensitivity and climate variability may amplify persistent drying, hydroclimatic instability, and risks to future adaptation.

Received: 04 Aug 2026 – Discussion started: 18 Aug 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2332 KB)

Supplement (5432 KB)

Download & links

Muhammad Abrar Faiz, Liangliang Zhang, Dong Liu, Qiang Fu, Mo Li, Xiaochen Qi, Tianxiao Li, and Song Cui

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-4696', Anonymous Referee #1, 24 Sep 2026
This manuscript presents a global assessment of hydroclimatic transitions using a DD, WW, WD, and DW framework based on historical and future changes in SPEI-SR. By incorporating a storage residual into the conventional SPEI framework, the study examines hydroclimatic transitions and their relationships with precipitation, evaporative demand, hydrological memory, land-surface characteristics, irrigation, and large-scale climate variability. The direct–lagged decomposition and response-time analysis further examine how climate anomalies propagate into runoff and soil moisture, while the extension to future SSP scenarios provides insights into possible future hydroclimatic changes. However, several aspects of the methodology, physical interpretation, and presentation require further clarification and validation before the main conclusions can be fully supported. I recommend major revision before the manuscript can be considered for publication.

The authors interpret the storage residual as a proxy for terrestrial water storage changes and use it as an important component of the proposed SPEI-SR framework. However, SR represents a water-balance residual rather than a direct observation of terrestrial water storage. The authors should evaluate whether SR-derived storage variability is consistent with GRACE-derived TWS anomalies. Such a comparison would provide important validation for the proposed SPEI-SR framework.

The current DD–WW classification based solely on the signs of linear trends may overinterpret weak changes as meaningful hydroclimatic transitions. Grid cells with trends close to zero but opposite signs between the two periods can still be classified as WD or DW, although such changes may simply reflect sign changes around zero rather than robust transitions. This issue is particularly relevant because both statistically significant and nonsignificant trends are retained in the classification. The authors should consider a magnitude threshold or uncertainty range, or provide additional analyses of trend magnitudes (e.g., Trend1–Trend2 scatter plots), to distinguish substantial trend reversals from weak variations around zero.

The direct–lagged decomposition and response-time analysis in Fig. 2 provide useful information on hydroclimatic memory. However, their links with the DD–WW patterns are supported mainly by spatial correspondence and regional interpretation. Since the direct/lagged components and response times have already been quantified at the grid-cell level, a quantitative comparison of these metrics among the DD, WW, DW, and WD classes would further strengthen the interpretation.

The large-scale climate analysis remains relatively limited. Most correlations between the global transition-state fractions and PDO, AO, and NOI are modest, and the differences between positive and negative phases are also relatively small. Moreover, these climate indices characterize large-scale modes of variability but do not directly diagnose the associated atmospheric circulation processes. The authors should avoid interpreting these associations as evidence of specific circulation mechanisms unless additional diagnostics are provided. Analyses of geopotential height, winds, moisture transport, or related circulation variables would provide stronger physical support.

The supplementary material presents future projections from individual CMIP6 models, but the robustness of the DD–WW classifications across models is not quantitatively evaluated. Since the transition categories depend on trend signs, relatively small inter-model differences may lead to different classifications. A simple measure of model agreement, such as the fraction of models agreeing on the dominant transition class, would help clarify the uncertainty in the projected DD–WW patterns.

The presentation of Fig. 3c–d could be improved. Since DD, WW, DW, and WD are mutually exclusive categories whose proportions sum to 100% within each land-cover or irrigation class, 100% stacked bar charts may provide a more intuitive comparison among land-cover types and irrigation levels.

In line 100, “We used the CMIP6 models under seven SSPs” confuses the number of CMIP6 models with the number of SSP scenarios. Please correct this statement and carefully check the manuscript for similar conceptual or terminological errors.
Citation: https://doi.org/10.5194/egusphere-2026-4696-RC1
- AC1: 'Reply on RC1', Muhammad Abrar Faiz, 28 Sep 2026
  
  Thank you, referee, for the constructive assessment. The suggested analyses address important aspects of validation and classification robustness. The revised manuscript will therefore include (1) independent GRACE/GRACE-FO evaluation of the storage-change residual; (2) trend-magnitude and uncertainty-based robustness assessment of the DD, WW, WD, and DW classification; (3) quantitative comparison of direct, lagged, and response-time metrics among transition classes; (4) spatially explicit CMIP6 inter-model agreement; (5) revised 100% stacked presentations for Figs. 3c-d; and (6) corrected and standardized CMIP6/SSP terminology. The corresponding methods, results, discussion, figures, and supplement will be revised accordingly, and the detailed revised response will identify the exact manuscript changes and line numbers.
  
  Citation: https://doi.org/10.5194/egusphere-2026-4696-AC1
RC2:
'Comment on egusphere-2026-4696', Anonymous Referee #2, 03 Oct 2026
The paper builds an index (SPEI-SR) from P − AET − Q, sorts land into DD, WW, WD and DW from trend signs in two periods, and links these classes to land cover, irrigation, climate indices and CMIP6 scenarios. The question is useful, but I have concerns about the core of the analysis. Several problems below concern the core of the analysis.

P − AET − Q is the change in soil water and snow. Summed over k months, it is just storage now minus storage k months ago. A steady drying trend in storage does not produce a trend in this index. So the SPEI-SR trend does not tell us whether a place is drying or wetting. The authors need to show it against the soil moisture and snow trends in TerraClimate.
The four classes look like random sorting. About 52.9% of land dries in the first period and 51.4% in the second. If the periods were independent, DD would be 27%. The paper reports 28.7%. The other classes are also close to chance. The classes use only the sign of the trend, so a tiny trend counts the same as a strong one. I do not see any evidence here of real regime transitions. Please test against a null model and map only cells where both trends are significant.
TerraClimate has no land-cover change, no irrigation and no impervious surfaces. So Fig. 3 only shows where each land type sits in climate space. The differences are also small (DD is 31.7–35.7%) and have no test. The stories about urban runoff and irrigation buffering are not backed by this data, and the title depends on them. Eq. 4 has a similar problem. Q and SM are computed from P and PET, so regressing them on P and PET only recovers the model. It is not external validation.
The future result is likely a mistake. In L383–387, "WW from 28.8% to 35.2%" compares Fig. 4a with Fig. 4c, which is SPEI-3 against SPEI-12 for the same years. For the same timescale, the change is 28.8% to 29.1%. So the abstract claim looks like a timescale effect, not a future change. The main text also does not say which models were used, whether bias was corrected, or what baseline was used for the future series.

The numbers contradict each other. The abstract and Fig. 1 give 52.9% drying and DD of 28.7%. L239–244 give 46.9% and 30.7%. L442 calls 47.1% "historical WW", but that is the share of wetting trends in the first period. "Dry" also means a trend sign in Fig. 1 but an aridity split in Fig. 2b. The authors need to fix these.
The climate-index section gives no coefficients, p-values or checks for autocorrelation. It is not clear how annual class shares are obtained when the classes come from two fixed periods. The abstract says North Atlantic Oscillation but the methods use the Northern Oscillation Index. The circulation explanations have no circulation data supports.
minors:
L100 says "seven SSPs" and lists two. L99 says ERA5, but the data are TerraClimate. The future period ends in 2099 in some places and 2100 in others.

References 9 and 13 are the same paper, and reference 17 is incomplete. The acknowledgements thank reviewers for improving the paper, which does not fit a first submission.

Fig. 3 is hard to read, and the boxplots in Fig. 4e–f look the same for every region. The choice of hotspots and the clustering in L207 are not explained in the Methods.

4.Code is available only "upon reasonable request". It should be available in a public repository according to Copernicus policy.
Citation: https://doi.org/10.5194/egusphere-2026-4696-RC2

Muhammad Abrar Faiz, Liangliang Zhang, Dong Liu, Qiang Fu, Mo Li, Xiaochen Qi, Tianxiao Li, and Song Cui

Supplement

https://doi.org/10.5194/egusphere-2026-4696-supplement

Muhammad Abrar Faiz, Liangliang Zhang, Dong Liu, Qiang Fu, Mo Li, Xiaochen Qi, Tianxiao Li, and Song Cui

Viewed

Total article views: 257 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
140	55	62	257	61	59	50

HTML: 140
PDF: 55
XML: 62
Total: 257
Supplement: 61
BibTeX: 59
EndNote: 50

Views and downloads (calculated since 18 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	21	9	8	38
Sep 2026	100	35	47	182
Oct 2026	19	11	7	37

Cumulative views and downloads (calculated since 18 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	21	9	8	38
Sep 2026	100	35	47	182
Oct 2026	19	11	7	37

Viewed (geographical distribution)

Total article views: 253 (including HTML, PDF, and XML) Thereof 253 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 05 Oct 2026

Short summary

Water availability is changing in ways that are not always gradual. This study examined whether places are becoming wetter, drier, or shifting between the two by tracking changes in water gained from rain and lost through evaporation and river flow. The results show that drying still affects many regions, but many areas are also shifting between wet and dry conditions. This helps identify where future water risks may grow, especially in drylands and northern regions.


Total:	0
HTML:	0
PDF:	0
XML:	0