Pinus yunnanensis Franch is a major forest species in southwest China as a source of timber and industrial raw materials. The genetic quality of the species is declining and the differentiation of offspring is strong as affected by environmental change and improper management measures. To assess the phenotypic diversity of natural populations, the evaluation of twelve phenotypic traits in nine populations from its whole distribution was performed. Studies revealed plentiful phenotypic variations within and among populations. The phenotypic variation within the population was 4.03%, and was lower than that among populations (21.04%), indicating that the phenotypic variation among populations was the main source. The mean differentiation coefficient was 91.23%, and the mean coefficient of variation of twelve traits was 28.27%, ranging from 14.18% (length of needles) to 70.11% (height under the branches). No significant correlation between plant height and environmental factors was found. Mean annual temperature, mean temperature of the driest quarter, mean temperature of the wettest quarter, and latitude were significantly correlated to diameter breast height, respectively. Temperature is the most important factor affecting the diameter of breast height. Three principal components that represent plant shape, needle, and lateral branch trait, respectively, were obtained while the cumulative contribution rate reached 74.40%. According to the unweighted pair-group method with arithmetic means (UPGMA) cluster analysis, nine populations were divided into three clusters. However, populations were not clustered strictly according to geographic distance, implying that there is a discontinuity in the variation of phenotypic traits. Compared with other populations, the Lufeng population contains obvious advantages in plant height, diameter breast height, crown diameter, and needle length and width, whereas the Yongren population has the worst performance in plant height, crown diameter, and the number of lateral branches. Moreover, for selecting superior families, both the comprehensive scoring method and principal component analysis were combined. By comparing trait values from 258 families, eleven superior families with an actual gain of each trait ranging from 0.02% to 32.23% were successfully screened out. This study provides a certain reference significance for the breeding of improved varieties and plantation management of P. yunnanensis.