Analysis of soil salinization and land use change under water conservation retrofit in the Hetao irrigation district

This study focuses on the Shenwu Irrigation Area, situated in the upstream section of the Hetao Irrigation District, within the context of water conservation retrofitting. We establish soil sampling sites in this region, employ Google Earth Engine algorithms to develop models for soil salinity inversion and land use classification, and analyze the change patterns of land salinization and land use types. The results showed that the Blue, Green, Red, NIR, SWIR1 and SWIR2 bands and the SI, SI1, SI2, SI3 and SI4 spectral indices were the ones that had correlation coefficients greater than 0.55 between the various parameters and soil salinity in 2015 and 2022. The soil salinity inversion model constructed using Random Forest demonstrates higher R 2 values and lower MAE and RMSE compared to the Support Vector Machine and Gradient Boosting Tree, establishing it as the optimal model for soil salinity inversion. After the water conservation retrofit, non-saline soil increased by 326.5 km 2 , indicating improved soil conditions. The overall accuracy of the Gradient Boosting Tree classification model is 95.00% and the Kappa coefficient is 0.93, which is the optimal model for land use type classification. After water conservation retrofits, the water body and wasteland area decreases and the building land and cropland area increases. This research offers a theoretical basis for alleviating soil salinization in irrigated area and improving the ecological circumstances of these locales.