Xu et al. (2026) Time-lag and cumulative drought effects decouple vegetation sensitivity from damage risk in the upper Yangtze River basin

Identification

• Journal: Ecological Indicators
• Year: 2026
• Date: 2026-03-07
• Authors: Xiaoxiang Xu, Quanzhi Yuan, Pan Zhao, Luping Jia, Ping Ren
• DOI: 10.1016/j.ecolind.2026.114739

Research Groups

• Institute of Geography and Resources Science, Sichuan Normal University, Chengdu, China
• Sustainable Development Research Center of Resource and Environment of Western Sichuan, Sichuan Normal University, Chengdu, China
• Key Lab of Land Resources Evaluation and Monitoring in Southwest, Ministry of Education, Sichuan Normal University, Chengdu, China

Short Summary

This study analyzed vegetation response to drought in the upper Yangtze River basin (1990-2022) using NDVI and multi-scale SPEI, developing a composite drought sensitivity index and quantifying loss risk with a Copula-Bayes framework, revealing that drought sensitivity does not always align with actual vegetation loss probability.

Objective

• To characterize the spatiotemporal patterns of drought time-lag and cumulative effects on vegetation.
• To assess the drought sensitivity of major vegetation types along key hydroclimatic gradients.
• To quantify the probability of drought-induced vegetation loss under different drought intensities to assess vegetation vulnerability.

Study Configuration

• Spatial Scale: Upper Yangtze River Basin (UYRB), southwestern China (90°32′ E to 111°27′ E longitude, 24°27′ N to 35°45′ N latitude). Data resampled to a 5 km spatial resolution.
• Temporal Scale: 1990 to 2022 (33 years), using monthly data.

Methodology and Data

• Models used:
  • Pearson correlation coefficients (for lag and cumulative effects).
  • Composite Drought Sensitivity (DS) index.
  • Copula-Bayes framework for vegetation loss probability.
  • Six Copula functions (Gaussian, Clayton, Frank, Gumbel, Student-t, Joe) were tested.
  • Marginal distributions: Generalized Extreme Value (GEV), Normal (NORM), Gamma (GAM), and Log-Normal (LOGN) for NDVI; Normal for SPEI.
  • Jenks natural breaks method for classifying drought sensitivity.
• Data sources:
  • Normalized Difference Vegetation Index (NDVI): GIMMS 3G+ (1982–2022), original 1/12° spatial, 15-day temporal resolution, aggregated to monthly and resampled to 5 km.
  • Standardized Precipitation Evapotranspiration Index (SPEI): Recalculated at 5 km resolution using monthly precipitation and potential evapotranspiration data (1990–2022 subset from 1901–2024 products), calculated at 1-12 month timescales.
  • Meteorological data: 1 km resolution monthly mean precipitation and monthly potential evapotranspiration (PET) for China (1901–2024), subset 1990-2022 used, resampled to 5 km.
  • Aridity Index (AI): Calculated from multi-year mean annual precipitation to PET, resampled to 5 km.
  • Land use data: (1990–2022), original 30 m spatial resolution, aggregated to 5 km.
  • Digital Elevation Model (DEM) data: 30 m spatial resolution.

Main Results

• Vegetation responses to drought are primarily dominated by cumulative effects (65.4% of vegetated regions), with average response periods ranging from 5 to 8 months. Lag effects are less extensive (28.4% of area) and characterized by shorter response times (1-3 months dominant).
• The duration of vegetation response is linked to water availability, with arid regions responding more rapidly (average cumulative response time of 4.3 ± 2.9 months) and humid regions showing prolonged responses.
• Grasslands (mean cumulative response time: 5.3 ± 4.1 months) and shrublands (6.3 ± 4.2 months) exhibit the highest drought sensitivity, while croplands respond slowest (8.1 ± 3.8 months).
• Drought sensitivity shows spatial heterogeneity, with higher sensitivity in arid and semi-arid zones (37% and 34% high sensitivity, respectively), but also unexpectedly high sensitivity in some humid regions, particularly for forests and grasslands.
• The probability and spatial extent of vegetation damage increase with drought severity. Slight vegetation loss probability ranges from 34% to 52% under mild drought and can exceed 80% in some regions under severe/extreme drought. Severe vegetation loss probability remains relatively low overall (average 6.5% under mild drought, 19.1% under extreme drought).
• A "decoupling" phenomenon exists where high-sensitivity regions (e.g., Hengduan Mountains) exhibit relatively low vegetation loss probability, while some moderately/lowly sensitive regions (e.g., Eastern Sichuan Basin, Guizhou Plateau) show significantly higher loss probability. This is attributed to ecological adaptations, microclimatic refuges, compound drought-heat events, and the tail dependence structure of Copula models.

Contributions

• Developed an integrated drought sensitivity index that combines both response magnitude and temporal scales (lag and cumulative effects) for a more holistic assessment.
• Quantified probabilistic vegetation loss risk under varying drought intensities using a Copula-Bayes framework, incorporating temporally optimized drought indices.
• Revealed the nonlinear responses of ecosystems to extreme drought and provided a quantitative framework for assessing vegetation sensitivity and loss risk in regions with atypical drought characteristics.
• Identified and explained a "decoupling" phenomenon between drought sensitivity and actual vegetation loss risk, offering new insights into complex vegetation-drought interactions.
• Provided methodological and conceptual insights for future research on vegetation–climate interactions, drought risk assessment, and ecosystem resilience under changing climate conditions.

Funding

• Projects of National Natural Science Foundation of China (No. 41930651)
• Sichuan Science and Technology Program (No. 2023NSFSC1979)

Citation

@article{Xu2026Timelag,
  author = {Xu, Xiaoxiang and Yuan, Quanzhi and Zhao, Pan and Jia, Luping and Ren, Ping},
  title = {Time-lag and cumulative drought effects decouple vegetation sensitivity from damage risk in the upper Yangtze River basin},
  journal = {Ecological Indicators},
  year = {2026},
  doi = {10.1016/j.ecolind.2026.114739},
  url = {https://doi.org/10.1016/j.ecolind.2026.114739}
}