Some typical roadside tree-belts were selected for this study in Shanxi Province. Concentration of heavy metals (Pb、Cr、Cu、Zn) in roadside surface soils along the tree-belts were measured using by ICP-AES. Pollution dispersion patterns of the four heavy metals were studied by pollution index which was the ratio of heavy metal concentrations in polluted soils against that in non-polluted background soil. Protective effects of different tree-belt structure types on soil heavy metals pollution were compared. The results showed: (1)The pollution dispersion patterns of the four heavy metals (Pb、Cr、Cu、Zn)were unimodal with the major pollution distance ranging from 10 to 70m. The peak values of pollution index and the peak positions of four heavy metal concentration were different. The peak value of pollution index of the four heavy metals was 2.297, 1.000, 2.019 and 0827. All the peak positions of the pollution values for the four heavy metals were not at road shoulders with the peak position of Cu at 20m from roads and that of Pb, Cr and Zn at 40m. (2) The roadside tree-belts had significant protective effects on the four heavy metal pollution. The pollution dispersion pattern of the four heavy metals had significant changes due to the influence of roadside tree-belt. T test showed significant difference (P＜0.001) in heavy metal concentration between roadside soils with tree-belts and without tree-belts (Pb、Cr、Cu、Zn).The pollution distance was limited mainly within 50m from the road. The concentration and pollution index of Pb, Cr and Cu was reduced with the peak value positions of Cr and Zn being shifted beforehand. The peak value positions of Cr and Zn were at 20m, with peak values of pollution index being 0.580 and 0.599, and peak values of concentration being 93945mg/kg and 108.549mg/kg. The peak value position of Cu was at 20m with the peak value of pollution index being 0.605 and the peak value of concentration being 54.511mg/kg. The peak value position of Pb was at 40m with the peak value of pollution index being 1.046 and the peak value of concentration being 46.940mg/kg. (3) Tree-belt width，shelterbelt porosity and species composition showed significant influence on heavy metal pollution protective effects of roadside tree-belts. The optimum tree-belt width was between 40 and 70m to effectively prevent the heavy metal pollution. Roadside tree-belts of mono-species with canopy density ≥0.6 could limit the main pollution ranges and reduce the pollution index of the 100m position to 0.139-0.977. Mixed tree-belt with canopy density ≥0.6 could be more effective to reduce roadside soil heavy metal pollution, limiting the peak value of pollution index within 20m. Moveover, the tree-belt with the mean height of main tree species ≥10m, could also prevent atmospheric fine particles spread, especially for Cu which often adheres to larger particles.