Abstract: The remote sensing classification method can be used to quickly identify the dominant tree species in Pu’er City, further improve the accuracy of tree species coverage classification, and provide a reference for forest monitoring in the region. Based on the global-scale remote sensing cloud computing platform（Google Earth Engine, GEE）, the multi-temporal Sentinel-2 data corrected by the atmosphere and terrain were integrated to identify the spectral information of tree species, extract texture, phenology, terrain and other characteristic factors, and conduct different combination, hierarchical classification and random forest（RF） methods were used to classify the five dominant tree species in Pu’er City: Simao pine, tea tree, oak, rubber and Eucalyptus urophylla. The results showed that when multi-temporal images were combined with multi-features for classification, the classification accuracy of terrain features in forest and non-forest, coniferous and broad forest, and dominant tree species was higher than the introduction of phenology and texture features. The overall accuracy of forest and non-forest classification was 99.5%（ Kappa=0.98）, F₁ of user accuracy and mapping accuracy was 98.48%; the overall accuracy of coniferous forest and broadleaf forest classification was 98.7%（Kappa=0.96）, F₁ was 97.64%; the overall accuracy of dominant tree species classification was 85.83 %（Kappa=0.80）, F₁ was 85.19%. Dominant tree species were mainly distributed on steep slopes in the west, southwest and south slopes at an altitude of 1300-1700m. Extracting multiple features from multi-temporal images for classification can effectively improve the classification accuracy of dominant tree species in Pu’er City, and can provide a large-area, high-precision forest cover classification map more accurately.