Abstract: Dove tree (Davidia involucrata) is one of the national first-class protective plants in China. In order to understand its variation and adaptation mechanism of leaf phenotypic traits, 11 leaf phenotypic traits of 15 populations were investigated in Sichuan Province. The variation degrees of 11 leaf phenotypic traits were tested by nested analysis of variance, correlation analysis, and cluster analysis. Significant and abundant variations of 11 phenotypic traits were observed at inter and intra population levels. The mean phenotypic differentiation coefficient among populations was 37%, which was less than that within populations (63%), indicating that the variation within populations was the main source of the phenotypic variation of dove tree. The mean coefficient of variation (CV) of 11 phenotypic indices was 16.22%, ranging from 4.32% to 29.10%. The mean dispersion coefficient of populations was 53.21%, ranging from 39.38% to 87.14%. The results of correlation analysis showed that there were significant correlations among leaf length, leaf width, left leaf width, and petiole length, implying that some leaf phenotypic traits of dove tree coordinate each other in the 15 dove tree populations. Leaf size, leaf thickness and petiole size were negatively correlated with latitude and altitude, with the leaves being gradually smaller and thinner with elevated latitude and altitude. The mean values of Shannon index for leaf phenotypic traits and for dove tree populations were 2.360 and 2.331 respectively, and were well-distributed as well. Results from the clustering analysis showed that the 15 dove tree populations in Sichuan Province could be classified into three groups, with a lack of continuous geographical variations. In conclusion, the dove tree populations in Sichuan Province had higher phenotypic variations, which were related with the heredity and habitats. Our results provided scientific basis for conservation and utilization of dove tree resources.