Spatial patterns of soil moisture for a grassland and an arable land in eastern Kerqin sandy land were studied by means of geostatistics. The results showed that soil rnoisture could be well expressed by a spherical model in grassland at both 0～10?cm and 10～20?cm layers. The spatial corelation coefficient, range and fractal dimension were 0.979 and 0.999, 11.57 m and 19.41 m, and 1.571 and 1.535 at 0～10 cm and 10～20 cm layers, respectively. In arable land, spherical and linier model could be expressed by semivariogram for soil moisture at 0～10 cm and 10～20 cm layers. The spatial correlation coefficient, range and fractal dimension were 0.344 and 0.914, 7.67 m and 9.31 m, 1.571 and 1.535 at 0～10 cm and 10～20 cm layers, respectively. Moran's Is and Kriging maps expressed the spatial structure characters in detail. Grassland had higher soil moisture content and spatial heterogeneity, closer spatial autorelation and dependence, and regular pattern, while arable land had lower soil moisture contents and spatial heterogeneity, looser spatial autorelation and dependence, and serious fragmentation (especially at 0～10 cm layer). Analysis indicated that agricultural cultivation might cause wind erosion, desertification and loss of productivity of sandy land. Therefore, the arable sandy land should be returned into grassland to conserve and restore the health of sandy land ecosystem.