Abstract: With the combination of airborne Lidar and panchromatic images in 1981 and 2021, we investigated the canopy height structure of tropical forests in Menglun sub-reserve in the Xishuangbanna National Nature Reserve of Yunnan Province, and analyzed its relationship with environmental factors by using multiple regression tree (MRT) method. The results showed that forests in the Menglun sub-reserve could be clustered into seven types based on canopy height structures, with tropical rainforest, monsoon evergreen broad-leaved forest, secondary forest, and flood plain forest as the main types. The potential solar radiation, altitude, terrain profile curvature, slope and the brightness value of imageries in 1981 and 2021 were main factors that drove the classification. The tropical seasonal rainforest dominated by Pometia pinnata occupied the largest area in valley and low-land. The monsoon evergreen broad-leaved forest dominated by Castanopsis echinocarpa mainly distributed in the ridge and disturbed areas. The secondary forests had homogeneous canopy surface, which was significantly different from the primary forests. The activities of swidden agriculture about three decades ago had legacy impacts on the physiognomy of secondary forests.

Key words: tropical seasonal rainforest, canopy height, swidden agriculture, LiDAR, multiple regression tree