Li et al. (2026) Vegetation increasingly controlled by atmospheric dryness rather than soil moisture

Identification

• Journal: Nature Geoscience
• Year: 2026
• Date: 2026-10-28
• Authors: Lingcheng Li, Yilin Fang, Jiafu Mao, Shiqin Xu, Zhonghua Zheng, Wantong Li, Mingjie Shi, Charles D. Koven, Jeffrey Quintin Chambers, L. Ruby Leung, Nate G. McDowell
• DOI: 10.1038/s41561-026-02107-w

Research Groups

• Atmospheric, Climate, and Earth Sciences Division, Pacific Northwest National Laboratory, Richland, WA, USA
• Energy and Environment Division, Pacific Northwest National Laboratory, Richland, WA, USA
• Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN, USA
• Division of Biological and Environmental Science and Engineering, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia
• Department of Earth and Environmental Sciences, The University of Manchester, Manchester, UK
• Department of Environmental Science, Policy and Management, University of California, Berkeley, Berkeley, CA, USA
• Climate & Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
• Department of Geography, University of California, Berkeley, Berkeley, CA, USA
• School of the Environment, Washington State University, Pullman, WA, USA

Short Summary

This study investigates the relative dominance of soil moisture (SM) versus atmospheric dryness (vapour pressure deficit, VPD) on global vegetation responses and how this influence has changed over the past four decades. Using an explainable artificial intelligence framework, the authors find that while SM currently dominates vegetation variability in most regions, VPD influence has significantly strengthened, and SM influence has weakened, leading to an asymmetric shift towards atmospheric control, particularly in SM-dominated ecosystems.

Objective

• To determine whether soil moisture (SM) or atmospheric dryness (vapour pressure deficit, VPD) dominates terrestrial vegetation responses globally, and to quantify how their relative influence has changed over the past four decades.

Study Configuration

• Spatial Scale: Global, with data harmonized to a 0.5° × 0.5° spatial resolution, covering 43 IPCC climate reference regions.
• Temporal Scale: Four decades, from 1982 to 2022, analyzed in two epochs: pre-2001 (1982–2001) and post-2001 (2002–2022).

Methodology and Data

• Models used:
  • Automated Machine Learning (AutoML) framework (specifically FLAML algorithm)
  • Explainable Artificial Intelligence (XAI) using SHapley Additive exPlanations (SHAP) with TreeExplainer
  • Seasonal-Trend Decomposition Procedure based on Loess (STL)
• Data sources:
  • Satellite-derived vegetation indices: Normalized Difference Vegetation Index (NDVI) from GIMMS-3G+NDVI and PKU GIMMS NDVI products.
  • Reanalysis data: ERA5-Land hourly reanalysis dataset for root zone soil moisture (m³ m⁻³), shortwave solar radiation (W m⁻²), 2-meter air temperature (°C), and dew point temperature (°C) to calculate vapour pressure deficit (Pa).
  • Ancillary datasets: MODIS Land Cover Type product (MCD12C1) and IPCC climate reference regions.

Main Results

• Soil moisture (SM) dominates vegetation greenness variability in 57.2% of global vegetated areas, particularly in arid and semi-arid regions.
• Over the past four decades (1982–2022), VPD influence has strengthened across 62.0% of vegetated areas, while SM influence has decreased in 55.8% of regions.
• An asymmetric transition is observed: 65.7% of SM-dominated ecosystems have shifted towards stronger VPD control, whereas VPD-dominated regions show more balanced transitions (20.9% towards SM, 21.9% towards VPD).
• Changes in climatological mean VPD (ΔVPD, up to 23.8% relative importance) and SM (ΔSM, up to 26.3% relative importance), along with their interannual variabilities (ΔCVVPD, ΔCVSM), are the primary environmental controls driving these temporal shifts.
• Rising VPD increases its importance by directly increasing the hydraulic cost of carbon gain, forcing stomatal closure, even when SM is adequate.

Contributions

• This study provides the first global assessment of how the relative importance of atmospheric dryness (VPD) and soil moisture (SM) in controlling vegetation greenness has shifted over the past four decades.
• It introduces an AutoML-enhanced explainable artificial intelligence (XAI) framework to robustly disentangle nonlinear hydroclimatic driver contributions to vegetation dynamics.
• The research identifies a novel asymmetric transition pattern, where SM-dominated ecosystems predominantly shift towards enhanced atmospheric control, contrasting with more balanced transitions in VPD-dominated regions.
• The findings offer critical implications for understanding global carbon and water cycle projections and provide guidance for improving Earth system model representations of vegetation responses to hydroclimatic drivers under climate change.

Funding

• Next-Generation Ecosystem Experiments–Tropics (US Department of Energy, Office of Science, Office of Biological and Environmental Research)
• Oak Ridge National Laboratory (ORNL) Reducing Uncertainties in Biogeochemical Interactions through Synthesis and Computing (RUBISCO) Scientific Focus Area
• Terrestrial Ecosystem Science Scientific Focus Area (Earth and Environmental Systems Sciences Division of the Office of Biological and Environmental Research in the DOE Office of Science)

Citation

@article{Li2026Vegetation,
  author = {Li, Lingcheng and Fang, Yilin and Mao, Jiafu and Xu, Shiqin and Zheng, Zhonghua and Li, Wantong and Shi, Mingjie and Koven, Charles D. and Chambers, Jeffrey Quintin and Leung, L. Ruby and McDowell, Nate G.},
  title = {Vegetation increasingly controlled by atmospheric dryness rather than soil moisture},
  journal = {Nature Geoscience},
  year = {2026},
  doi = {10.1038/s41561-026-02107-w},
  url = {https://doi.org/10.1038/s41561-026-02107-w}
}