Abstract: Soil samples were collected with two sampling scales in Chongqing and their As，Hg，Cd，Pb and Cr levels measured to study the feasibility and adaptability of geostatistics in analyzing soil spatial variability of heavy metals. Results indicate that As，Cd and Cr on the small scales show normal symmetrical distribution，but Pb，Hg on the small scales and As，Hg，Cd，Pb and Cr on the large scales show skewed distribution. The CV of different heavy metals are between 23.2%～24.1%，showing a significant spatial variability. High spatial variability exists under large scales，while low spatial variability occurs under small scales. Typical semivarigram structure observation of samples on the two scales show that the geostatistical analytical method can correctly analyze soil heavy metal spatial variability，regardless of the size of the sampling scale. On small scales，heavy metal has strong spatial or medium auto-correlation with nugget values between 0.112～0.736. On large scales，heavy metal has medium auto-correlation with nugget values between 0.301～0.570. Spatial interpolations and cross validations show that the spatial forecast results of As，Cr，Hg and Cd on small scales precisely reflect their spatial variability；and on large scales，the estimated results somewhat warp. The estimated result for Pb deviates on all scales.