Due to the ecological fragility and high risk of salinization, desert soils have become a focal point in soil evolution studies. This study focuses on the No. 12 Regiment Farm in Aral City, Xinjiang, and systematically analyzes the impacts of fifty years of cultivation on key soil properties based on a comprehensive investigation of 74 cultivated profiles and 6 native desert profiles (1 m depth). Results showed that long-term cultivation significantly improved soil texture: sand content decreased by an average of 387.2-460.6 g/kg, while silt and clay contents increased by an average of 330.6-400.9 g/kg and 55.7-67.1 g/kg, respectively. Soil texture transformed from sandy soil to silty loam. Soil salinity was markedly enriched along the profile, with the electrical conductivity in the 0-20 cm layer more than doubling compared to the desert control, showing typical surface salt accumulation. Calcium carbonate content also increased significantly, with an average increase of up to 97.1 g/kg. To further quantify the three-dimensional spatial distribution of soil properties, several depth function models were employed to fit vertical variation patterns. The binomial exponential function yielded the highest fitting accuracy (R²>0.85 for all properties). For spatial prediction, the random forest (RF) model outperformed boosted regression trees (BRT) and multiple linear regression (MLR), achieving excellent performance (R²>0.90). In conclusion, long-term cultivation has improved soil texture in desert areas, promoted the accumulation of clay and silt, salts, and calcium carbonate, and accelerated pedogenic processes, while also increasing the risk of salinization. The index function with a constant term best assesses the depth variation patterns of soil properties, and the RF method shows the optimal performance in predicting all soil properties from spatial scale in the desert area.