Soil salinity mapping of oasis farmland using hybrid deep learning and texture normalization

Soil salinization in arid oases reduces farmland productivity and requires high-precision remote sensing monitoring. However, texture-induced scattering hinders accurate soil salinity mapping. This dataset supports a hybrid deep learning framework that combines a texture normalization strategy (TNS) with multi-temporal, multi-source remote sensing imagery and environmental covariates to map soil salinity in oasis farmland.Contents. The repository provides (i) field observations of soil salinity (and related soil properties where available), (ii) multi-temporal remote sensing predictors derived from satellite imagery, (iii) environmental covariates (e.g., terrain, climate, land surface conditions), and (iv) code/metadata needed to reproduce data preprocessing and model training/evaluation.Data organization. Files are organized into folders including data/ (tabular samples and/or raster predictors), metadata/ (variable definitions, units, coordinate reference system, missing-value rules), code/ (scripts for preprocessing and modeling), and docs/ (README and usage notes). A README file documents variable meanings, units, spatial reference (EPSG/projection), spatial resolution, temporal coverage, and the recommended workflow for reproduction.Usage. The dataset can be used to (1) reproduce the experiments reported in the associated study, (2) benchmark alternative machine learning/deep learning models under consistent inputs, and (3) support follow-up research on soil salinity mapping in arid agro-ecosystems.Licence and notes. The data are released under CC BY 4.0. If point locations are included, any sensitive information has been anonymized or spatially generalized to protect privacy and field-site security.