Wang et al. (2025) Estimating soil moisture at farm scale with high spatial resolution: integrating remote sensing data, and machine learning

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-11-30
• Authors: Di Wang, Shaohang Xu, Shuai Wang, Wenhui Liu, Naiquan Zheng, Zhen Wang, Yao Rong, Chenglong Zhang, Chaozi Wang, Nigenare Amantai, Zailin Huo
• DOI: 10.1016/j.jhydrol.2025.134707

Research Groups

• Center for Agricultural Water Research in China, China Agricultural University, Beijing, China.
• State Key Laboratory of Efficient Utilization of Agricultural Water Resources, Beijing, China.
• School of Civil and Hydraulic Engineering, Ningxia University, Yinchuan, China.
• Institute of Ecology, College of Urban and Environmental Sciences and Key Laboratory for Earth Surface Processes of Ministry of Education, Peking University, Beijing, China.

Short Summary

This study develops a machine learning-based downscaling framework that integrates evapotranspiration and groundwater depth to estimate surface soil moisture at a 30 m resolution from 9 km coarse data. The approach significantly improves soil moisture monitoring in complex agricultural environments by accounting for both upper and lower boundary conditions.

Objective

• To develop and validate a high-resolution (30 m) surface soil moisture (SSM) downscaling framework using machine learning that incorporates actual evapotranspiration (ET) and groundwater depth (GWD) to better capture field-scale dynamics.

Study Configuration

• Spatial Scale: Farm scale (30 m resolution) downscaled from a 9 km resolution; validated across 59 regional sampling points.
• Temporal Scale: Continuous monitoring and long-term estimation (specific duration not specified in the provided text, but validated against continuous monitoring data).

Methodology and Data

• Models used: Gradient Boosting Decision Tree (GBDT) for knowledge transfer from coarse to high resolution.
• Data sources: Remote sensing data (Land Surface Temperature (LST), NDVI, albedo), reanalysis/modeled products (Actual ET, GWD, soil texture), and ground-based validation (59 regional sampling points and 3 continuous monitoring sites).

Main Results

• Temporal Performance: The model achieved a mean R of 0.658, MAE of 0.014 cm³/cm³, and RMSE of 0.021 cm³/cm³.
• Spatial Performance: The model achieved a mean R of 0.672, MAE of 0.029 cm³/cm³, and RMSE of 0.061 cm³/cm³.
• Boundary Condition Impact: Incorporating ET (upper boundary) and GWD (lower boundary) improved estimation accuracy by up to 70.3%, especially in irrigated areas where shallow groundwater influences soil moisture.

Contributions

• Proposes a novel SSM downscaling framework that transfers learned relationships from 9 km resolution to 30 m resolution.
• Integrates both upper (ET) and lower (GWD) boundary conditions into machine learning models, effectively capturing field-scale dynamics in areas where irrigation inputs are difficult to quantify.
• Demonstrates strong extrapolation capability across different datasets and potential for application to other high-resolution products like CLDAS.

Funding

• Not specified in the provided text excerpt.

Citation

@article{Wang2025Estimating,
  author = {Wang, Di and Xu, Shaohang and Wang, Shuai and Liu, Wenhui and Zheng, Naiquan and Wang, Zhen and Rong, Yao and Zhang, Chenglong and Wang, Chaozi and Amantai, Nigenare and Huo, Zailin},
  title = {Estimating soil moisture at farm scale with high spatial resolution: integrating remote sensing data, and machine learning},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134707},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134707}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)