Enhancing Soil Salinity Mapping in Cropland by Integrating Agronomic Features With Remote Sensing

Soil salinization in cropland, particularly in arid and semi-arid regions, is a major form of land degradation. Although remote sensing allows for the soil salinity mapping by analyzing the spectral responses of soil and vegetation, a common limitation is the oversight of field management practices, which are known to significantly influence soil moisture dynamics and salinity accumulation. This study addresses this gap by integrating novel agronomic features—crop planting structure and irrigation indicator—into salinity prediction models. Crop planting structure was a quantitative representation of crop classification results using target encoding. The irrigation indicator was derived from the Temperature Vegetation Dryness Index (TVDI) to capture seasonal soil moisture variations associated with irrigation during the growing season. Our results demonstrate that incorporating these features significantly improves model performance compared to conventional models using only environmental and spectral variables. The coefficient of determination (R²) increased by 0.11 reaching 0.74. Notably, crop planting structure emerged as the most influential predictor, underscoring the value of field management information in cropland salinity mapping. Future research will explore advanced spatial representations of management practices, such as the distribution of irrigation infrastructure and continuous cropping years, to further enhance predictive reliability in soil salinity mapping.