In this study, we investigated the phenotypic diversity of 26 characters of five main natural populations of Calanthe arcuata distributed at the north and south slopes on the main peak of the Qinling Mountains in China. The statistical methods used included Shannon-Wiener index, nested variance analysis, coefficient of variation analysis, principal component analysis (PCA), correlation analysis and cluster analysis. The results showed that: 22 phenotypic traits had rich coefficient of variation of 10.95% ~ 46.05% within and among the populations in C. arcuata, and the intra-population variation was greater than the inter-population variation. The petal color in this study had a relatively higher diversity index (1.12) among the four quality traits, indicating that the percentage of petal color varied greatly. The average coefficient of variation of leaf length-to-width ratio was the largest, and that of the middle sepal width was the smallest. The PCA analysis indicated that the length of middle sepal, width of middle sepal, length of the lateral sepal, width of synsepal, width of leaves and number of flowers were the main sources of phenotypic variation. The correlation of phenotypic traits was mainly concentrated among the flowers, and highly correlated with the altitude among the geographical factors. The five populations are clustered into two groups, with the Dadian (DD) population, which was at a much higher altitude than the other four populations, clustered in a branch of itself, and the 22 phenotypic traits were clustered in R into three groups. This study provides a theoretical basis for the strategy of in-situ conservation of the wild populations and for establishment of germplasm bank, breeding, development and utilization of these plants.