[Objective] The spatial heterogeneity of wind-eroded surface particles in Xilamuren desert steppe area was studied to estimate the surface-erosion-caused degradation in order to guide the protection and utilization of grassland resources in this area. [Methods] Digital image processing technique was used to obtain the composition information of soil particles in a desert steppe slope. The spatial heterogeneity of three particles as highly erodible particles(< 0.42 mm), semi-erodible particles(0.42~0.84 mm), and non-erodible particles(> 0.84 mm) was analyzed with classical statistical and geostatistical methods. [Results] The results showed the percentage rank of the three kinds of soil particle was non-erodible particles > semi-erodible particles > highly erodible particles. The coefficients of the variation ranged from 11.23% to 38.17%, indicating that soil particles had a moderate degree of variability. The optimal semivariance models were exponential, spherical and spherical models respectively. The results of Kriging demonstrated that the most soil particle on the top of slope was coarse, and displayed an decreasing tread from the top to the bottom in the slope profile. The most severity of soil coarsening was found at the south and southwest of the slope, the rank of coarse degree was as follows: South > Southwest > Northwest > West > Southeast > North > Northeast > East. The results can provide reference to grassland desertification control in the future. [Objictive] Geostatistical analysis displayed strong spatial correlation, indicating that the spatial variability of soil particles was mainly caused by structural factors. Human activities also has its effect on it.