Wang et al. (2026) Spatial Prediction of Soil Texture at the Field Scale Using Synthetic Images and Partitioning Strategies

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-01-14
• Authors: Yiang Wang, J Li, Shuai Bao, Yuxin Ma, Yan Zhang, Dianyao Wang, Yihan Ma, HuanJun Liu
• DOI: 10.3390/rs18020279

Research Groups

Not explicitly stated in the provided text.

Short Summary

This study developed and evaluated a comprehensive remote sensing strategy on Google Earth Engine to improve field-scale soil texture prediction, finding that a specific multi-year mean compositing approach combined with spectral indices significantly enhances accuracy for precision agriculture.

Objective

• To evaluate the effects of different compositing time windows, compositing modes, and compositing data types on the prediction accuracy of field-scale soil texture using the random forest algorithm, and to assess three local partitioning regression strategies based on crop growth, soil synthetic image brightness, and soil type.

Study Configuration

• Spatial Scale: Field scale, applied to Northeastern China.
• Temporal Scale: Multi-year (2021-2024) May images; evaluation of different compositing time windows and intervals.

Methodology and Data

• Models used: Random forest algorithm, global regression, local partitioning regression (based on crop growth, soil synthetic image brightness, and soil type).
• Data sources: Google Earth Engine (GEE) platform, soil synthetic images generated using mean and median compositing modes, extracted image bands, and introduced spectral indices (including moisture spectral indices).

Main Results

• The use of mean compositing of multi-year May images (2021-2024) combined with the "band reflectance + spectral indices" dataset improved the average R² for clay, silt, and sand particles by 8.89%, 9.50%, and 2.48%, respectively, compared to other compositing methods.
• The introduction of spectral indices significantly improved prediction accuracy, increasing the average R² for clay, silt, and sand particles by 4.58%, 3.43%, and 4.59%, respectively, compared to using only image band data.
• Global regression generally outperformed local partitioning regression; however, the local partitioning regression strategy based on soil type showed good accuracy, with its average R² for soil particles only 1.08% lower than that of global regression under the optimal compositing method.

Contributions

• Innovatively constructs a comprehensive strategy for soil texture prediction combining moisture spectral indices, a specific compositing time window, a specific compositing mode, and soil type partitioning.
• Provides a new paradigm for field-scale soil texture prediction, particularly for Northeastern China.
• Lays the foundation for data-driven water and fertilizer decision-making in smart agriculture.

Funding

Not explicitly stated in the provided text.

Citation

@article{Wang2026Spatial,
  author = {Wang, Yiang and Li, J and Bao, Shuai and Ma, Yuxin and Zhang, Yan and Wang, Dianyao and Ma, Yihan and Liu, HuanJun},
  title = {Spatial Prediction of Soil Texture at the Field Scale Using Synthetic Images and Partitioning Strategies},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18020279},
  url = {https://doi.org/10.3390/rs18020279}
}