A 40-year dataset for tracking soil salinization dynamic (1985–2024) at 100 m resolution in the Western Songnen Plain, China

In our study, we identified and classified soil salinization degrees in the Western Songnen Plain at 100 m spatial resolution using ground surveys data and remote sensing imagery, combined with machine learning algorithms over the period 1985 to 2024. The publicly available data used in this paper are shown below: 1) The remote sensing satellite data used in this study are freely available on GEE platform (available at https://code.earthengine.google.com/). 2) Land cover data can be accessed at https://doi.org/10.5281/zenodo.4417810. 3) SSSG dataset is available at https://files.isric.org/public/global_soil_salinity. 4) LDSS data are available at https://doi.org/10.6084/m9.figshare.13295918.v1. 5) The NLS are available at https://gtdc.mnr.gov.cn/Share#/secondSurvey. All source scripts used for saline soil identification model, soil EC prediction model and prediction model input parameters are publicly available on GitHub at: https://github.com/mercyxinian/Code.git. The presented data file contains: 1) Soil sampling metadata Filename: Soil_EC_sampling_points.csv Format: CSV (UTF-8 encoded) Description: Contains georeferenced field observations of soil electrical conductivity (EC), used for prediction model training and validation. Column Description ID Unique identifier for each sample point Lon Longitude coordinate of each sample point Lat Latitude coordinate of each sample point Municipal Administrative region (city) where the point is located LnEC Natural logarithm of observed soil EC1:5 (in dS m⁻¹), measured in lab TIR Thermal infrared reflectance value from Landsat imagery SIT Salinity Index based on Red, NIR, and SWIR bands PDI Perpendicular Drought Index, used as a proxy for surface soil moisture 2) Model files TIRSITPDI_predicted.mat Format: Matlab .mat file Description: Contains the trained Neural Network Fitting (NNF) model for soil EC prediction. This model was optimized using 14,000 iterations and parameter tuning (e.g., number of hidden layers, learning rate, activation function). Soil_EC_prediction_model.m Format: MATLAB script Description: Implements the prediction process. It reads spectral input parameters (TIR, SIT, PDI), applies the trained model, and outputs predicted soil EC values. 3) Annual Salinity Mapping Outputs (1985–2024) These folders contain annual gridded maps and summary statistics derived from the soil EC prediction model. 📁 Statistical_results_by_year/ Contents: CSV tables and a .png summarizing the area (in km²) of saline soils in each salinity class per year. Classes: Slightly saline (2–4 dS m⁻¹), Moderately saline (4–8), Highly saline (8–16), Extremely saline (>16) 📁 Salinization_degree_maps/ Contents: Raster maps (GeoTIFF, EPSG:4326, 100 m resolution) of classified salinity zones for each year (1985–2024) based on U.S. Salinity Laboratory classification. The .png contains year-by-year classified salinity zones. 📁 Saline_soil_identification/ Contents: Binary maps (GeoTIFF) showing annual identification of saline vs. non-saline soils from 1985 to 2024. The .png contains year-by-year identification results.