The Southern edge of Mu Us Sandy Land is the frontier of land desertification. Conservation and sustainable use of the vegetation in this sandy land-grassland transition area is crucial for preventing vegetation deterioration in the sandy land and its bordering areas. Investigating spatial heterogeneity of soil nutrients and relevant ecological processes in this critical area is useful to understand the processes and mechanisms of land desertification, and helpful for sustainable land use planning. Based on the field survey along a sandy land-grassland transect, we analyzed the heterogeneity of soil nutrients using the geostatistical techniques such as semivariogram, Kriging interpolation, and spatial distribution maps. Results showed that (1) The semivariogram of soil available potassium (K) was best described by the spherical model and that of soil total nitrogen (N) and soil available phosphorus (P) were best described by the exponential model. Soil available K showed a characteristic of strong spatial autocorrelation, and 88.2% of the variation was found within the distance of 511m. Soil total N and available P showed a modest degree of autocorrelation; and 50.1% and 51.4% of the variation were found within the distance of 511m and 143 m respectively. Geostatistical analysis indicated that spatial variation in soil available P occurred in a small scale; whereas ecological processes determining soil total N and available K contents exerted their effects in a larger scale. (2) Regular spatial patterns of all the 3 soil nutrients were identified. The content of soil available K and P decreased at first and then increased along the transect from Lespedeza potaninii community to Artemisia ordosica community; and this trend was approximately consistent with the changing process of the ecological boundary in the study area. While the change of soil total N along the transect was quite different, decreasing gradually from Lespedeza potaninii community to Artemisia ordosica community, which corresponded with the decrease in vegetation cover. (3) The spatial distribution of the standard deviations of the contents of all these soil nutrients showed very lower values, confirmed the accuracy of Kriging interpolation in this study. (4) The spatial pattern of soil nutrients was closely related with distribution of vegetation types. Distribution of soil nutrients was also related with soil type, micro-topography, plant litter composition, soil animal activity and grazing. Existence of aeolian sandy soil and arenosol in the study area resulted in some variation in soil nutrients content. The wavy topography also affected the movement of soil nutrients and consequently their distribution. Vegetation changes along the sandy land-grassland transect had different species composition and productivity, and different plant litter composition and activities of soil invertebrates and microbes. Also, grazing had profound effects of soil nutrients through excreta deposition. More intense grazing in the area of Lespedeza potaninii community and L. potaninii-Artemisia ordosica community boundary area (good quality forages) than in the area of Artemisia ordosica community (which contained poisonous plants such as Cynanchum komarovii) caused different nutrient cycling rate in these areas. (5)The strong wind in the study area also had a big effect on the process of desertification. The strong wind, combined with the effect of "islands of fertility" led to the rapid process of desertification of the semiarid grassland. These research findings provided the basis for understanding vegetation degradation and desertification processes and useful for land use planning in this fragile transitional region of the sandy land and grassland.