Wang et al. (2025) A 40-year dataset of soil salinity dynamics (1985–2024) at 100 m resolution in the Western Songnen Plain, China

Identification

• Journal: Scientific Data
• Year: 2025
• Date: 2025-11-13
• Authors: Bin Wang, Xiaojie Li, Zeyu Gao, Zhongjun Jia, Zhengwei Liang
• DOI: 10.1038/s41597-025-06057-7

Research Groups

• State Key Laboratory of Black Soils Conservation and Utilization, Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun, China
• University of Chinese Academy of Sciences, Beijing, China

Short Summary

This study developed and validated a 40-year (1985–2024) high-resolution (100 m) dataset of soil salinity dynamics for the Western Songnen Plain, China, using remote sensing, extensive field data, and machine learning, revealing significant spatiotemporal variability in salinization trends. The resulting dataset provides crucial information for improved monitoring and sustainable land management in salinized regions.

Objective

• To generate a high-resolution (100 m), long-term (1985–2024) dataset of soil salinity dynamics in the Western Songnen Plain, China, to overcome limitations of existing coarse-resolution and infrequent products, thereby supporting accurate assessment and sustainable land management.

Study Configuration

• Spatial Scale: Western Songnen Plain, China (approximately 46,985 km²), mapped at 100 m resolution.
• Temporal Scale: Annual, spanning 40 years from 1985 to 2024.

Methodology and Data

• Models used:
  • Saline soil identification: Random Forest (RF) (overall accuracy = 0.893, Kappa = 0.782), K-Nearest Neighbour (KNN), Classification and Regression Tree (CART), Support Vector Machines (SVM).
  • Soil EC prediction: Neural Network Fitting (NNF) (R² = 0.467, RMSE = 0.729 dS m⁻¹, MAE = 0.556 dS m⁻¹), Gaussian Process Regression (GPR), Least-Squares Boosting (LSBoost), Kernel Partial Least Squares Regression (KPLS), Tree, SVM, Linear Regression (LR), SVM Kernel.
• Data sources:
  • Field surveys: 942 georeferenced soil samples collected in 2012, 2014, 2018, and 2019 for soil electrical conductivity (EC) measurement (1:5 soil-to-water suspension). 3,487 reference points for saline/non-saline identification derived from visual interpretation of high-resolution Google Earth Pro imagery.
  • Satellite imagery: Landsat-5 Thematic Mapper (TM) (1984–2011) and Landsat-8 Operational Land Imager/Thermal Infrared Sensor (OLI/TIRS) (2013–present) for annual 100 m resolution composites. Landsat-7 Enhanced Thematic Mapper Plus (ETM+) used for reflectance matching.
  • Derived indices: Salinity Index (SIT), Perpendicular Drought Index (PDI) as a soil moisture proxy.
  • Ancillary data: China Land Cover Dataset (CLCD), FAO90 soil classification system.

Main Results

• Achieved high accuracy in saline soil identification (RF model: Overall Accuracy = 0.893, Kappa = 0.782) and reliable soil EC prediction (NNF model: R² = 0.467, RMSE = 0.729 dS m⁻¹).
• Generated a comprehensive 40-year (1985–2024) annual dataset of soil salinity dynamics at 100 m resolution for the Western Songnen Plain.
• The estimated total saline soil area for 2009 (5,068.43 km²) showed a minimal deviation of 2.78% from the National Land Survey data (5,213.50 km²).
• Revealed distinct spatiotemporal patterns in soil salinization:
  • 1985–1998 (Pre-intensification): Stable soil salinity with median EC values around 0.5 dS m⁻¹.
  • 1999–2006 (Salinization phase): Gradual increase in EC, with medians often exceeding 0.6 dS m⁻¹, indicating intensification likely due to expanded irrigation.
  • 2007–2024 (Desalinization/stabilization stage): EC declined and stabilized, with most medians below 0.5 dS m⁻¹, possibly reflecting the impact of restoration efforts.
• The dataset and associated model files are publicly available at Zenodo (https://doi.org/10.5281/zenodo.17044260).

Contributions

• Provides an unprecedented high-resolution (100 m) and long-term (annual from 1985–2024) soil salinity dataset, significantly improving upon existing coarser (250 m to 1 km) and less temporally frequent global/regional products.
• Developed a robust and transferable methodological framework that integrates extensive multiyear field observations, consistent Landsat time-series imagery, and advanced machine learning algorithms (NNF, RF) for accurate soil salinity mapping.
• Enhanced prediction reliability by explicitly incorporating soil moisture effects through the Perpendicular Drought Index (PDI), particularly effective in bare and sparsely vegetated areas.
• Offers critical data and a validated approach for improved soil salinity monitoring, precision agriculture, and informed land management strategies in arid and semi-arid regions facing similar salinization challenges.

Funding

• National Key Research and Development Program (No. 2022YFD1500505)
• Natural Science Foundation of Jilin Province (No. YDZJ202201ZYTS550)
• Cropland Degradation Monitoring

Citation

@article{Wang202540year,
  author = {Wang, Bin and Li, Xiaojie and Gao, Zeyu and Jia, Zhongjun and Liang, Zhengwei},
  title = {A 40-year dataset of soil salinity dynamics (1985–2024) at 100 m resolution in the Western Songnen Plain, China},
  journal = {Scientific Data},
  year = {2025},
  doi = {10.1038/s41597-025-06057-7},
  url = {https://doi.org/10.1038/s41597-025-06057-7}
}