Soil organic carbon, the main part of terrestrial carbon pool, plays an important role in terrestrial carbon cycle, and is one of the most important components in the research of global change. This paper aims to analyze soil organic carbon content (SOC) distribution patterns of the Picea crassifolia forest patches in Qilian Mountain and to determine which factors affect those patterns. We have investigated the relationships between edaphic environmental factors (soil, forest floor, topography, and canopy) and understory vegetation (composition and biomass) among 67 plots representing four major patch types which include: (1) big forest patch in the shady slope; (2) small forest patches in the semi-shady slope; (3) forest gaps; and (4) grassland in the sunny slope. The area of each plot is 10m×10m, and the data of elevation, slope grade, plant cover, biomass, canopy density and diameter at breast height of Picea crassifolia in the each plot have been estimated. Soil samples have been taken and analyzed in the laboratory for SOC content, nitrogen, moisture, and bulk density. The results indicate that the average SOC content is (84.9±26.7)g/kg in the surface layer (020cm depth) of soil under Picea crassifolia forest patches. The SOC content is significantly greater under shade forest patch ((101.6±31.8)g/kg) compared to the semi-shade forest patches ((64.2±10.3)g/kg) and sunny slope grassland ((23.2±5.6)g/kg). Correlation analysis shows that SOC content is positively correlated with elevation, soil water content, soil bulk density and shrub biomass, and negatively correlated with canopy density of Picea crassifolia forest; the patch areas of forest gaps (0.020.12hm²) and smaller forest patches (0.170.89hm2) in the semi-shady slope have not significant effect on SOC content. Further principal component analysis shows that elevation and soil water content are the first principal component affecting SOC content, followed by the canopy density of Picea crassifolia forest and shrub biomass, which account for 83.8% of the variance of SOC content.