The kernel density estimation method of species abundance distribution was used to simulate the species abundance distribution of two communities which were dominated respectively by rare species Tsoongiodendron odorum Chun and by vulnerable species Tsuga longibracteata Cheng. The results show that the species abundance distribution in tree and shrub layers and all woody plants of the two forests all can well be described by using nonparametric kernel method based on the sample data without assuming the species abundance distribution．The nonparamefic kernel density estimation is suggested to be a useful method for species abundance distribution simulation．