Abstract: Based on the base of Midong district agricultural products as an example, by the methods of geostatistics combined with ArcGIS, the statistical analysis module, analysis of spatial variability of 7 kinds of heavy metals in soil. Through the comparison of forecasting errors of different interpolation methods, the optimal interpolation model selection of spatial variability of soil heavy metals in the region, the spatial distribution characteristics of heavy metals in the soil map fitting.Using Geographical information system(GIS )and geostatistics to analyze the spatial variability of heavy metal(Hg，Cu，Zn，Pb，Ni，Cd and Cr)contents of farmland soil in Midong, Urumqi. The results show that: In addition to Cu, Zn, Pb and Hg are more than the background value of soil, the average content of 7 kinds of heavy metals in Midong District farmland soil did not exceed the " national soil environmental quality standard " ( GB156182-1995 ) level two standard limit, compared with the Urumqi farmland soil heavy metal background values, Cr, Ni, Cd average content is not more than the background value, while the average content of Cu, Zn, Pb and Hg were 1.39 times, 1.49 times the background value, 1.04 times and 2 times.Analysis shows that the spatial structure, 7 kinds of heavy metals have spatial variability structure better, can use the index model, spherical model and Gauss model, and has a nugget effect in different degree. Spatial variability of soil heavy metals in farmland of variable substrate effect: Hg>Cu>Zn>Pb>Ni>Cd>Cr. The base effect of Cd and Cr <25%, have strong spatial correlation, substrate effect the rest of 4 heavy metals are between 25%~75%, belongs to the medium intensity spatial correlation. Variation of 7 kinds of heavy metals was low ( range 0.010~1.851 km ), indicating that they are related in a smaller range. The prediction error in the 0.981~1.023 between 7 kinds of heavy metal standard RMS, shows that the fit of the model is high, strong adaptability. The Zn standard root mean square error of prediction is 0.998, close to 1, the highest degree of fitting. Pb, Ni, Cu, Zn and Hg have moderate intensity spatial correlation, it shows that their contents are affected by exogenous pollution. The map of spatial distribution of soil heavy metal content in the study area was obtained using ordinary Kriging interpolation method. In addition to inconspicuouslyCd spatial distribution law is not clear while the other six heavy metals display significant spatial distribution.The map of spatial distribution of soil heavy metal content in the study area was obtained using ordinary Kriging interpolation method. In addition to inconspicuously Cd spatial distribution, there were significant spatial distribution of other 6 kinds of heavy metals.

Key words: Midong Urumqi, soil, Heavy metal, Spatial distribution, Geographical information system, Geostatistics