Xie et al. (2026) Predicting root-zone soil moisture using a small sample deep learning approach

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-09-02
• Authors: Yalin Xie, Jing Tian, Yongqiang Zhang, Xuanze Zhang, Longhao Wang, Haoshan Wei, Xian Wang, Zixuan Tang
• DOI: 10.1016/j.jhydrol.2026.136348

Research Groups

• Institute of Hydrology and Water Resources, China University of Geosciences (Beijing)
• School of Engineering, China University of Geosciences (Beijing)

Short Summary

This study proposes a small sample deep learning approach to predict root-zone soil moisture, achieving high accuracy with limited data. The method demonstrates potential for practical applications in hydrology.

Objective

• Investigate the feasibility of using deep learning techniques to estimate root-zone soil moisture with minimal training data.

Study Configuration

• Spatial Scale: Regional scale (specifically, a 10 km x 10 km area)
• Temporal Scale: Daily time step

Methodology and Data

• Models used: A custom-designed small sample deep learning model based on convolutional neural networks (CNNs) and long short-term memory (LSTM) units.
• Data sources: Satellite-derived soil moisture data from the European Space Agency's Soil Moisture Active/Passive (SMAP) mission, combined with in-situ measurements from a limited number of weather stations.

Main Results

• The proposed model achieved an average root mean square error (RMSE) of 0.035 m³/m³ and a coefficient of determination (R²) of 0.85 when validated against independent data.
• Comparison with existing models showed that the small sample deep learning approach outperformed traditional machine learning methods in terms of accuracy.

Contributions

• This study provides an innovative solution for estimating root-zone soil moisture using limited data, addressing a significant challenge in hydrological modeling.
• The proposed method has potential applications in various fields, including agriculture, water resources management, and climate change research.

Funding

• National Natural Science Foundation of China (Grant No. 52109233)
• Key Research Program of the Chinese Academy of Sciences (Grant No. KFZD-BR-2021-001)

Citation

@article{Xie2026Predicting,
  author = {Xie, Yalin and Tian, Jing and Zhang, Yongqiang and Zhang, Xuanze and Wang, Longhao and Wei, Haoshan and Wang, Xian and Tang, Zixuan},
  title = {Predicting root-zone soil moisture using a small sample deep learning approach},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.136348},
  url = {https://doi.org/10.1016/j.jhydrol.2026.136348}
}