Abstract: To understand the distribution mode of soil texture for enhanced precision fertilization in tobacco fields, spatial variability of soil texture in Pingdingshan tobacco field was analyzed via geostatistics method on Geographic Information System (GIS) platform. The spatial distributions of clay, sand and silt were then quantitatively built by ordinary Kriging — a linear unbiased interpolation method. The results show that soil granules have relatively large auto-correlation lengths, and with spatial heterogeneity scales of 609 m, 657 m and 429 m respectively for clay, sand and silt. About 80% of spatial variability of clay and sand is caused by structural factors. Directional semi-variogram analysis indicates that clay and sand have a relatively high spatial auto-correlation. Further analysis shows that while clay and sand are highly anisotropic, silt is feebly anisotropic. Our experiment demonstrates that the relatively strong spatial correlation and spatial variability of the soil texture may be the major factor responsible for spatial variations in soil properties, especially initial distributions of soil texture. It is therefore suggested that spatial variations in soil texture should be incorporated in fertilization scheme designs.