Liu et al. (2025) High-resolution soil salinity mapping and driving factor analysis at regional scale using multi-source remote sensing data

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-11-16
• Authors: Yannan Liu, Yan Zhu, Yingzhi Qian, Wanli Xu, Guanghui Wei, Jiesheng Huang
• DOI: 10.1016/j.jhydrol.2025.134604

Research Groups

• State Key Laboratory of Water Resources Engineering and Management, Wuhan University, Wuhan, Hubei 430072, China
• Xinjiang Academy of Agricultural Sciences, Urumqi, Xinjiang Uygur Autonomous Region, 830091, China
• Xinjiang Tarim River Basin Authority, Korla, Xinjiang Uygur Autonomous Region, 841000, China

Short Summary

This study mapped high-resolution soil salinity and analyzed driving factors across 15 oasis irrigation districts in southern Xinjiang, evaluating five machine learning models with diverse predictor sets and identifying quantile random forest as the best performer, with wind-related variables being crucial.

Objective

• To systematically evaluate the performance of five soil salinity mapping models (multiple linear regression, random forest, Cubist, quantile random forest, and quantile neural network) across five nested predictor sets.
• To explore the influence of input variable combinations on model performance for accurate and efficient soil salinity prediction frameworks.
• To identify dominant driving factors of soil salinization at regional scales.
• To provide a practical and scalable framework for high-resolution soil property prediction at regional scales.

Study Configuration

• Spatial Scale: 15 major oasis irrigation districts in southern Xinjiang, covering a total area of 2.1 × 10^7 hectares (210,000 km²).
• Temporal Scale: Snapshot/Campaign-based data collection for soil salinity samples.

Methodology and Data

• Models used: Multiple linear regression, random forest, Cubist, quantile random forest, quantile neural network.
• Data sources: 6,045 soil salinity samples (from three soil depths), multi-source remote sensing data, soil properties, climate data, topography data, remote sensing indices, soil degradation data, and hydrogeological conditions data.

Main Results

• Soil salinity exhibited strong spatial variability, with coefficients of variation ranging from 161 % to 293 %.
• Quantile random forest consistently achieved the best performance, with root mean square error (RMSE) of 2.64–4.29 g/kg and coefficient of determination (R²) values of 0.64–0.79.
• All models performed worst when excluding wind-related variables, underscoring the crucial role of wind in salt transport and redistribution in southern Xinjiang.
• Adding more variables beyond a key subset yielded a negligible impact on model accuracy.
• Dominant driving factors of soil salinization varied by region and scale, indicating that no single predictor set is universally optimal.
• A nested variable modeling approach that balances prediction accuracy and computational efficiency is recommended.

Contributions

• Provides a systematic evaluation of multiple machine learning models for high-resolution soil salinity mapping at a regional scale.
• Highlights the critical importance of wind-related variables in soil salinization processes in arid and semi-arid regions like southern Xinjiang.
• Demonstrates that an optimal subset of variables can achieve high accuracy without excessive computational burden, recommending a nested variable modeling approach.
• Offers a practical and scalable framework for high-resolution soil property prediction, crucial for land management and agricultural sustainability.

Funding

No specific funding projects, programs, or reference codes were explicitly mentioned in the provided paper text.

Citation

@article{Liu2025Highresolution,
  author = {Liu, Yannan and Zhu, Yan and Qian, Yingzhi and Xu, Wanli and Wei, Guanghui and Huang, Jiesheng},
  title = {High-resolution soil salinity mapping and driving factor analysis at regional scale using multi-source remote sensing data},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134604},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134604}
}