The objective of this study is to analyzes the scale variation and the spatial patterns of Picea crassifolia populations in Qilian Mountains, using box-counting dimension, information dimension and correlation dimension, and to identify the relationships between the spatial pattern and some selected environmental factors. Sixteen plots (20m×20m) with different elevation in Picea crassifolia forest of QiLian Mountains were investigated to calculate fractal dimensions, and some selected environmental factors (elevation, slope degree, slope direction, soil depth , soil bulk density , soil organic carbon, total nitrogen, soil water content) were measured. The results shows that the values of box-counting dimension, information dimension and correlation dimension of pattern of Picea crassifolia populations ranged from 0.992 to 1.938, 0996 to 1.860 and 1.418 to 1.759, and its average value is 1.716,1.632 and 1.568, respectively. Box-counting dimension is negatively correlated with elevation and soil water content, while positively correlated with slope direction. Information dimension negatively correlated with elevation, but positively correlated with gradient. Correlation dimension negatively correlated with slope direction. The principal component analysis shows that elevation and soil water content are the first principal component affecting fractal dimension, followed by slope direction and slope degree, which accounts for 83.8% of the variance of fractal dimension.