Yang et al. (2026) High-resolution mapping of saturated soil hydraulic conductivity across China’s drylands

Identification

• Journal: International Journal of Applied Earth Observation and Geoinformation
• Year: 2026
• Date: 2026-03-03
• Authors: Ting Yang, Xiaoxu Jia, Liantao Niu, Huang Laiming, Chunlei Zhao, Xiangdong Li, Xiang Ren, Ruofan Wang, Shao Ming
• DOI: 10.1016/j.jag.2026.105176

Research Groups

• Modern Agricultural Engineering Laboratory, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences
• Shandong Dongying Institute of Geographic Sciences
• Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences
• College of Resources and Environment, University of Chinese Academy of Sciences
• College of Life Sciences, Yan’an University
• Key Laboratory of Low Altitude Geographic Information and Air Route, Civil Aviation Administration of China

Short Summary

This study developed a novel machine learning approach integrating multi-sensor Sentinel-1/2 remote sensing data and environmental covariates to generate high-resolution (90 m) saturated soil hydraulic conductivity (Ks) maps across China's drylands, demonstrating superior accuracy and spatial detail compared to existing datasets.

Objective

• To develop a novel, high-resolution (90 m) mapping approach for saturated soil hydraulic conductivity (Ks) in China's drylands by integrating multi-sensor Sentinel-1/2 remote sensing data with environmental covariates using a random forest regression model on the Google Earth Engine platform.

Study Configuration

• Spatial Scale: China's drylands (73.57°E to 134.57°E and 25.99°N to 53.54°N) at a 90-meter resolution.
• Temporal Scale: Remote sensing data averaged from 2020 to 2022; lab-based Ks samples collected from 2011 to 2023.

Methodology and Data

• Models used: Random Forest (RF) regression model.
• Data sources:
  • Remote Sensing: Sentinel-1 (C-band Synthetic Aperture Radar, VV and VH polarization, 10 m resolution), Sentinel-2 (multispectral imagery, 10-20 m resolution, 10 selected bands), MODIS MCD12Q1 (IGBP land classification, 500 m).
  • Environmental Covariates: Climate (precipitation, actual evapotranspiration, land surface temperature), Vegetation (Fraction of Absorbed Photosynthetically Active Radiation (FPAR), Leaf Area Index (LAI)), Soil (bulk density, clay content, sand content from a 90 m global soil dataset), Topography (elevation, slope, aspect).
  • Ground Truth: Over 5,000 lab-based Ks samples (0–10 cm and 10–30 cm depths) from China's drylands.
  • Platform: Google Earth Engine (GEE).

Main Results

• High-resolution remote sensing data significantly improved Ks prediction accuracy for both surface (0–10 cm) and subsurface (10–30 cm) layers.
• Model performance for ln(Ks/(cm min⁻¹)) on the testing set:
  • Surface (0–10 cm): Coefficient of determination (R²) = 0.61, Root Mean Square Error (RMSE) = 1.61.
  • Subsurface (10–30 cm): R² = 0.61, RMSE = 1.24.
• The generated 90-m Ks map outperformed existing global and regional datasets in terms of spatial detail and statistical accuracy, with RMSE reduced by 9.30%–21.28% and standard deviation (STD) reduced by 13.61%–22.81%.
• Fine-scale heterogeneity, including high-Ks zones in the Loess Plateau and Taklamakan Desert, was successfully resolved, which was not visible in coarser products.
• Soil factors, particularly clay content, were the most influential in controlling Ks variability, followed by climate and topographic variables. Sentinel-1 VV polarization and Sentinel-2 B8A (near-infrared) were identified as highly important remote sensing bands.

Contributions

• Proposed a novel Ks mapping approach integrating high-resolution Sentinel-1/2 data, environmental covariates, and machine learning on the Google Earth Engine platform, advancing beyond traditional pedo-transfer functions (PTFs) and coarse digital soil mapping (DSM).
• Systematically explored the predictive potential of individual Sentinel-1/2 bands and derived indices, establishing clear linkages between sensor-specific spectral/radar properties and soil hydraulic properties.
• Generated a 90-meter resolution Ks product that accurately resolves local variations and fine-scale heterogeneity, overcoming the limitations of coarse-resolution or point-based approaches.
• Provided quantitative evidence for the potential linkage between remote-sensing observables and soil physical processes, enhancing the understanding of Ks spatial dynamics.
• Established a critical foundation for precision land management and process-based hydrological modeling at a regional scale.

Funding

• National Key R&D Program of China (Grant No. 2025YFD1500204)
• National Natural Science Foundation of China (Grant No. U2444217)
• Third Xinjiang Scientific Expedition of the Ministry of Science and Technology of the PRC (Grant No. 2022xjkk0904)
• Shandong Provincial Natural Science Foundation (Grant No. ZR2024QD048)

Citation

@article{Yang2026Highresolution,
  author = {Yang, Ting and Jia, Xiaoxu and Niu, Liantao and Laiming, Huang and Zhao, Chunlei and Li, Xiangdong and Ren, Xiang and Wang, Ruofan and Ming, Shao},
  title = {High-resolution mapping of saturated soil hydraulic conductivity across China’s drylands},
  journal = {International Journal of Applied Earth Observation and Geoinformation},
  year = {2026},
  doi = {10.1016/j.jag.2026.105176},
  url = {https://doi.org/10.1016/j.jag.2026.105176}
}