[Objective] We studied the variation pattern of soil salinity at different layer of GSM(gravel-sand mulched field) and CK in order to provide basis for improvement and utilization of soil salinization in the northwest arid area in China.[Methods] Based on the statistics at different layer of GSM and CK, classical statistics and geo-statistics were used to analyze the statistical characteristic value, semi-variance simulation model and its fitting parameters, and the isocline distribution map of soil salinity in Jingtai County of Gansu Province.[Results] Mean soil salinity of GSM was lower than that of CK, and the result showed:CK > OGM(old gravel-sand mulched field) > NGM(new gravel-sand mulched field) > MGM(middle gravel-sand mulched field). There were significant differences in soil salinity among different layers of GSM and CK, and the result showed:layer Ⅳ(30-50 cm) > layer Ⅲ(20-30 cm) > layerⅡ(10-20 cm) > layerⅠ(0-10 cm). The variations of soil salinity in layerⅡ, Ⅲ, Ⅳ(CK) and MGM were weak, while variations in other layers were moderate. The best semi-variance theory model of layer Ⅱ(CK) was linear model, the optimal model for layerⅠ, Ⅲ(CK), layer Ⅲ(NGM), layer Ⅰ(MGM) and layer Ⅰ, Ⅱ, Ⅲ(OGM) was exponential model, while the best model for other layers was spherical model. The spatial auto-correlation of soil salinity at layer Ⅱ(CK) was lower, while it was high at other layers. The soil salinity in northern area was higher than that in southern area.[Conclusion] Gravel sand on soil surface can reduce soil salt accumulation and control soil secondary salinization, and there is spatial variability and correlation exists in soil salinity among different layer of GSM and CK.