Jia et al. (2026) Projected droughts readily trigger vegetation decline

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-10-01
• Authors: Zixu Jia, Tao Lin, Xiangzhong Guo, Faith Ka Shun Chan, Lichang Yin, Hongkai Geng, Yicheng Zheng, Junmao Zhang, Yuan Chen, Meixia Lin, Jing Lin, Yuqin Liu
• DOI: 10.1016/j.jhydrol.2026.136545

Research Groups

• State Key Laboratory of Regional and Urban Ecology, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen, China
• University of Chinese Academy of Sciences, Beijing, China
• Key Laboratory of Urban Environment and Health, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen, China
• Fujian Key Laboratory of Digital Technology for Territorial Space Analysis and Simulation, Fuzhou, China
• Xiamen Key Laboratory of Smart Management on the Urban Environment, Xiamen, China
• School of Geographical Sciences, Faculty of Science and Engineering, University of Nottingham Ningbo China, Ningbo, China
• Key Laboratory of Ecological Safety and Sustainable Development in Arid Lands, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi, China

Short Summary

This study provides a global assessment of vegetation responses to terrestrial water storage drought under historical and future climate scenarios, revealing that projected droughts will trigger vegetation decline under milder conditions due to increased vulnerability, particularly for severe vegetation loss.

Objective

• To investigate how global vegetation responds to variations in the Terrestrial Water Storage Drought Severity Index (TWS_DSI) under historical and future climate change scenarios.
• To determine if vegetation responses exhibit specific drought trigger thresholds across different drought intensities.
• To assess how these drought trigger thresholds might evolve under future climate conditions.

Study Configuration

• Spatial Scale: Global, with analyses conducted at pixel scales ranging from 0.05° to 0.5° resolution. Plant functional types (PFTs) were reclassified from 1 km to 0.5° resolution.
• Temporal Scale: Historical period (1982-2022, with some datasets covering 1982-2014 or 1982-2020) and future climate scenarios (2015-2100, with late-century analysis for 2060-2100). Data were primarily monthly or semi-monthly, with vegetation lag responses quantified up to 24 months.

Methodology and Data

• Models used:
  • Distributed Lag Nonlinear Model (DLNM) combined with Generalized Additive Model (GAM) for lagged response analysis.
  • Bivariate copula functions (Clayton, Frank, Gumbel, Gaussian, t-copula) for joint distribution modeling and conditional probability calculation.
  • Machine learning algorithms (Random Forest, XGBoost, LightGBM) for predicting future Terrestrial Water Storage Anomaly (TWSA).
  • Terrestrial Ecosystem Models (TEMs) from ISIMIP3b: CLASSIC and VISIT.
  • Global Circulation Models (GCMs) from ISIMIP3b: GFDL-ESM4, MPI-ESM1-2-HR, MRI-ESM2-0, IPSL-CM6A-LR, and UKESM1-0-LL.
  • Geographically Weighted Regression (GWR) for identifying dominant influencing factors.
• Data sources:
  • Vegetation Indicators:
    • Leaf Area Index (LAI): GIMMS LAI4g (1982-2020), GLASS LAI (1982-2018), GLOBMAP LAI (1982-2020).
    • Fraction of Photosynthetically Active Radiation (FPAR): GIMMS FPAR4g (1982-2022).
    • Normalized Difference Vegetation Index (NDVI): GIMMS NDVI3g+ (1982-2022).
    • kernel Normalized Difference Vegetation Index (kNDVI).
    • Gross Primary Productivity (GPP): NIRv-based (1982-2018).
  • Drought Index: Terrestrial Water Storage Drought Severity Index (TWS_DSI) derived from GTWS-MLrec (1940-2022), a machine-learning reconstructed global monthly terrestrial water storage anomaly (TWSA) dataset.
  • Climate Forcing Data: ISIMIP3b bias-corrected CMIP6 climate forcing data (precipitation, air temperature, surface pressure, wind speed, specific humidity, downward shortwave radiation) for historical, SSP126, SSP370, and SSP585 scenarios.
  • Hydrological Variables: Monthly evapotranspiration and runoff data from ISIMIP3b TEMs (CLASSIC, VISIT).
  • Ancillary Data: Plant Functional Types (PFTs) data (global 1-km resolution land projection dataset based on SSP-RCP scenarios).
  • Validation Data: GLEAM evapotranspiration, GRUN runoff, GIMMS LAI4g.

Main Results

• Historical Lag Responses: Global vegetation exhibited pronounced drought-memory effects, with mean optimal lag times ranging from 8.20 months for LAI to 10.59 months for NDVI. Forests consistently showed the longest mean lag times across all five vegetation indicators.
• Future Lag Responses: Under future climate scenarios (SSP126, SSP370, SSP585), mean optimal LAI lag times varied from 5.15 to 10.48 months across different TEM-GCM combinations, with inter-model differences being more significant than inter-scenario changes.
• Drought Trigger Thresholds (Historical): Drought trigger thresholds became progressively more negative with increasing vegetation loss severity. Mean thresholds ranged from -1.15 to -1.26 for ≤40th percentile loss, to -2.83 to -3.14 for ≤10th percentile loss. Shrublands generally exhibited the highest (least negative) thresholds, while forests showed the lowest (most negative) thresholds.
• Drought Trigger Thresholds (Future): Compared to the historical baseline, average drought trigger thresholds increased by 30.0%, 36.3%, and 35.3% under the SSP126, SSP370, and SSP585 scenarios, respectively. This indicates that milder drought conditions in the future could trigger vegetation decline. The projected upward shift was concentrated at severe vegetation loss levels (e.g., ≤10th percentile threshold increased by 1.00-1.29), while the threshold for moderate loss (≤40th percentile) showed little change or a slight shift toward stronger drought.
• Dominant Factors: Historically, precipitation primarily controlled thresholds for moderate vegetation loss. For severe vegetation loss, the influence of CO2, temperature, and solar radiation became more balanced with precipitation. Under future scenarios, as radiative forcing and vegetation loss severity increased, the dominance of precipitation weakened, and temperature and evapotranspiration became more influential.

Contributions

• Provides the first global assessment of vegetation responses to Terrestrial Water Storage Drought Severity Index (TWS_DSI) under both historical and future climate scenarios, using five diverse vegetation indicators.
• Quantifies the temporal lags and identifies specific drought trigger thresholds for different levels of vegetation loss, revealing a nonlinear and hierarchical response.
• Systematically projects the evolution of these drought trigger thresholds under future climate conditions, highlighting increased vulnerability to milder droughts for severe vegetation decline.
• Characterizes the spatially varying associations between drought trigger thresholds and key climatic and environmental factors at a global scale.
• Offers a scientific basis for assessing ecosystem vulnerability and informing targeted ecological conservation and climate adaptation strategies, including a conceptual ecological early-warning and restoration framework.

Funding

• National Key R&D Program of China (Grant No. 2023YFF1304700)

Citation

@article{Jia2026Projected,
  author = {Jia, Zixu and Lin, Tao and Guo, Xiangzhong and Chan, Faith Ka Shun and Yin, Lichang and Geng, Hongkai and Zheng, Yicheng and Zhang, Junmao and Chen, Yuan and Lin, Meixia and Lin, Jing and Liu, Yuqin},
  title = {Projected droughts readily trigger vegetation decline},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.136545},
  url = {https://doi.org/10.1016/j.jhydrol.2026.136545}
}