Li et al. (2026) Soil moisture determines the maximum drought loss to vegetation in Central Asia

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-04-01
• Authors: Zheng Li, Zhonghao Fu, Jiongchang Zhao, Bo Liu, David Makowski, Shijia Li, Lu Tan, Akylbek Kazhigulovich Kurishbayev, Duman Imanmadi, Ainura Aldiyarova, Wenfeng Liu
• DOI: 10.1016/j.ejrh.2026.103389

Research Groups

• State Key Laboratory of Efficient Utilization of Agricultural Water Resources, China Agricultural University, Beijing, China
• National Field Scientific Observation and Research Station on Efficient Water Use of Oasis Agriculture in Wuwei of Gansu Province, Wuwei, China
• Center for Agricultural Water Research in China, College of Water Resources and Civil Engineering, China Agricultural University, Beijing, China
• International Economic & Technical Cooperation and Exchange Center, Ministry of Water Resources, Beijing, China
• UMR Applied Mathematics and computer Science (MIA518), INRAe AgroParistech, Université Paris-Saclay, Palaiseau, France
• College of Humanities and Development, China Agricultural University, Beijing, China
• International Office, China Agricultural University, Beijing, China
• Kazakh National Agrarian Research University, Almaty, Kazakhstan
• National Academy of Sciences of the Republic of Kazakhstan, Almaty, Kazakhstan

Short Summary

This study quantifies the spatiotemporal patterns and driving mechanisms of maximum drought-induced vegetation loss (MDVL) in Central Asia from 1982 to 2022, revealing a significant intensification of vegetation loss since 1992, primarily driven by soil moisture during the resistance period.

Objective

• To calculate monthly-scale maximum drought-induced vegetation loss (MDVL) based on multi-source vegetation data.
• To systematically analyze the driving mechanisms of multiple factors—including temperature (T), soil moisture (SM), downward shortwave radiation (SSRD), vapor pressure deficit (VPD), drought duration (Duration), and drought intensity (Intensity)—on MDVL.
• To identify the primary influencing factors of MDVL and their spatial contribution patterns.

Study Configuration

• Spatial Scale: Central Asia (Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Uzbekistan, and China’s Xinjiang Uygur Autonomous Region). Data were harmonized to a spatial resolution of approximately 9.25 km (5 arcminutes).
• Temporal Scale: Multi-decadal analysis, primarily covering 1982–2022, with specific datasets spanning different periods (e.g., SIF from 2000–2022, LAI from 1982–2020).

Methodology and Data

• Models used:
  • Random Forest regression model
  • SHapley Additive exPlanations (SHAP) analysis
• Data sources:
  • Vegetation:
    • Kernel Normalized Difference Vegetation Index (kNDVI) derived from Global Inventory Monitoring and Modelling System (GIMMS) 3gv1 NDVI dataset (1982–2022, 15-day temporal resolution).
    • Leaf Area Index (LAI) from GIMMS LAI4g dataset (1982–2020, 15-day temporal resolution).
    • Solar-Induced Chlorophyll Fluorescence (SIF) from globally spatially continuous SIF dataset (CSIF) (2000–2022, 4-day temporal resolution).
  • Climate & Soil (ERA5-Land): Monthly averaged reanalysis data from Copernicus Climate Data Store (CDS) by ECMWF (1982–2022, approximately 11.1 km spatial resolution). Includes:
    • 2 m surface air temperature (Ta)
    • 2 m dew point temperature (Td)
    • Downward shortwave solar radiation (SSRD)
    • Soil moisture (SM) (0–100 cm depth, calculated from 0–7 cm, 7–28 cm, 28–100 cm layers).
  • Drought Index: Standardized Precipitation Evapotranspiration Index (SPEI) from SPEIbase v2.9 (1982–2022, approximately 55.5 km spatial resolution, monthly temporal resolution).
  • Environmental & Topographic:
    • Climate zone data (ClimateZ) from Köppen-Geiger climate classification (1 km resolution).
    • Soil data (sand and clay content) from Harmonized World Soil Database (HWSD) (1 km resolution).
    • Elevation data (DEM) from MERIT DEM (90 m spatial resolution).
    • Land cover type data (LandC) from MOD12Q1 (500 m resolution).
    • Maximum plant root depth (mroot) from EartH2Observe (1 km resolution).
    • Burned area from Global Fire Emissions Database (GFED5) (1997–2020, monthly).
  • Derived variables: Vapor Pressure Deficit (VPD) calculated from Ta, Td, and elevation. Resistance period variables (ResSM, ResT, ResVPD, ResSSRD) calculated as cumulative sums or means during the resistance time.

Main Results

• Maximum drought-induced vegetation loss (MDVL) has exhibited a significant decreasing trend (indicating intensifying vegetation loss) across Central Asia annually since 1992.
• Northern Kazakhstan and southeastern Central Asia experienced more severe vegetation drought-induced loss, with MDVL values generally below −0.04 (dimensionless) for kNDVI, −0.2 square meters per square meter (m²/m²) for LAI, and −4.0 x 10⁻⁵ watts per square meter per nanometer per steradian (W m⁻² nm⁻¹ sr⁻¹) for SIF.
• Regions with lower MDVL (less severe loss), accounting for approximately 71% of the total study area, are mainly distributed in central Central Asia, dominated by grassland vegetation.
• Soil moisture during the resistance period (ResSM) is the dominant factor driving MDVL variations across Central Asia, accounting for 54–70% of the primary contribution. Higher soil moisture (volumetric water content, m³/m³) significantly alleviates the severity of vegetation loss.
• Temperature during the resistance period (ResT, in Kelvin) exhibits a non-monotonic relationship with MDVL, where extreme high or low values exacerbate vegetation loss.
• Higher downward shortwave solar radiation (ResSSRD) intensified vegetation water consumption and increased the severity of vegetation loss.
• Higher vapor pressure deficit (ResVPD, in kilopascals) corresponded to a drier atmosphere and more severe vegetation loss, particularly in northern Central Asia.
• The Random Forest regression models explained 68.8% of the variance for kNDVI-based MDVL, 62.7% for LAI-based MDVL, and 72.9% for SIF-based MDVL.

Contributions

• This study systematically reveals the spatiotemporal distribution patterns of maximum drought-induced vegetation loss (MDVL) and its underlying driving mechanisms across Central Asia using multi-decadal satellite data and advanced machine learning techniques (Random Forest and SHAP analysis).
• It innovatively identifies soil moisture as the most important environmental factor affecting MDVL in Central Asia, also dominating the largest spatial area (53–69%) among all drivers.
• The findings provide new theoretical perspectives for assessing ecosystem resilience in arid regions and constructing cross-border ecological barriers, thereby aiding in mitigating the impacts of drought on vegetation in Central Asia.

Funding

• National Key Research and Development Program of China (No. 2024YFC3213700)
• National Natural Science Foundation of China (No. 52239002 and 32361143871)
• Pinduoduo-China Agricultural University Research Fund (No. PC2023A02002)

Citation

@article{Li2026Soil,
  author = {Li, Zheng and Fu, Zhonghao and Zhao, Jiongchang and Liu, Bo and Makowski, David and Li, Shijia and Tan, Lu and Kurishbayev, Akylbek Kazhigulovich and Imanmadi, Duman and Aldiyarova, Ainura and Liu, Wenfeng},
  title = {Soil moisture determines the maximum drought loss to vegetation in Central Asia},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.103389},
  url = {https://doi.org/10.1016/j.ejrh.2026.103389}
}