The accurate estimation of forest volume is the basis for improving the level of sustainable forest management.Based on the airborne laser point cloud data,49 laser point cloud feature variables were generated.Combined with the ground survey sample data,three methods of fixed parameter,Pearson screening,and stepwise regression screening were used to screen out the independent variables used for modeling,and then linear and nonlinear regression fitting was used to establish the accumulation models of the three main tree species in Hainan Tropical Rainforest National Park.The results show that:1)Among the linear and nonlinear models of the three tree species(Acacia confusa,Hevea brasiliensis, Eucalyptus robusta),the accuracy of the optimal models was above 0.83 with A.confusa and E.robusta having better nonlinear models,and H.brasiliensis having a better linear model.2)The height class variable has the greatest influence on the accumulation model.The intensity and density of the laser point cloud have a greater influence on the accumulation model,and the coverage class variable also has some influence.The structural parameters obtained from airborne LiDAR data play an important role in modeling,and can be popularized and used in related operational work in the future.