Liu et al. (2026) Soil salinity inversion in croplands of the Yellow River Delta: a feature-enhanced stacking framework integrating multi-source remote sensing and environmental covariates

Identification

• Journal: CATENA
• Year: 2026
• Date: 2026-09-09
• Authors: Xuelong Liu, L. G. Chen, Changle Liu, Canting Zhang, Hongjia Wang, Tingting Zhang, Ailing Wang
• DOI: 10.1016/j.catena.2026.110587

Research Groups

College of Resources and Environment, Shandong Agricultural University (SDAU), National Engineering Research Center for Efficient Utilization of Soil and Fertilizer Resources (NERCSEUSFR), Guangrao County Agriculture and Rural Affairs Bureau.

Short Summary

This study developed a feature-enhanced Stacking ensemble framework to invert soil salinity in croplands of the Yellow River Delta, integrating Sentinel-1 SAR data, Sentinel-2 multispectral data, and environmental covariates. The framework achieved high accuracy (R2 = 0.802) and outperformed traditional methods.

Objective

• Investigate the feasibility of using a feature-enhanced Stacking ensemble framework for soil salinity inversion in complex coastal environments.

Study Configuration

• Spatial Scale: Regional scale, focusing on croplands in Guangrao County, Yellow River Delta.
• Temporal Scale: Not specified.

Methodology and Data

• Models used: Stacking ensemble learning with four base learners (RF, XGBoost, SVR, 1D-CNN) and two meta-learners (Ridge regression).
• Data sources: Sentinel-1 synthetic aperture radar (SAR) data, Sentinel-2 multispectral data, multi-source environmental covariates.

Main Results

• The feature-enhanced 1D-CNN-Stacking model achieved the best independent validation performance (R2 = 0.802, RMSE = 0.437 g/kg).
• Distance to the sea, salinity index 1 (SI1), and soil sand content consistently exhibited high importance under both RF-based and permutation importance assessments.

Contributions

• The study provides an effective strategy for ensemble optimization by coordinating meta-feature-space construction and meta-learner selection.
• The feature-enhanced Stacking framework improves regional soil salinity mapping supported by multi-source data.

Funding

• This research was funded by the National Engineering Research Center for Efficient Utilization of Soil and Fertilizer Resources (NERCSEUSFR).

Citation

@article{Liu2026Soil,
  author = {Liu, Xuelong and Chen, L. G. and Liu, Changle and Zhang, Canting and Wang, Hongjia and Zhang, Tingting and Wang, Ailing},
  title = {Soil salinity inversion in croplands of the Yellow River Delta: a feature-enhanced stacking framework integrating multi-source remote sensing and environmental covariates},
  journal = {CATENA},
  year = {2026},
  doi = {10.1016/j.catena.2026.110587},
  url = {https://doi.org/10.1016/j.catena.2026.110587}
}