Absrtact: 【Objective】 The accurate identification of mountainous tree species is the basis of remote sensing mapping, and it is very important for choosing the optimal temporal. Sensitive spectral index was introduced into SVM tree species identification method to improve the accuracy of tree species identification. 【Method】 ZY-1 02C and ZY-3 multispectral remote sensing images were selected to establish and screen sensitive spectral indices, and then the indices were input into support vector machine as conditional attribute to select the remote sensing identification optimal phase of Quercus acutissima and Robinia pseudoacacia in Mount Tai, meanwhile, every temporal sensitive bands were input into support vector machine as conditional attribute as control. 【Result】 The results showed that sensitive bands of three temporals were concentrated in three or four bands. Compared with the sensitive band SVM, the recognition accuracy of SVM constructed by sensitive spectral indices was improved, in which the Quercus acutissima was averagely increased by 0.24% than fourth band in May 12, by 1.23% than fourth band in September 29, by 0.63% and 1.57% respectively than the third and fourth bands in December 7; the Robinia pseudoacacia was averagely increased by 3.40% than fourth band in May 12, by 13.97% than fourth band in September 29, by 43.86% and 8.81% respectively than the third and fourth bands in December 7; the overall recognition accuracy of Quercus acutissima and Robinia pseudoacacia was averagely increased by 2.81% than fourth band in May 12, by 7.10% than fourth band in September 29, by 22.22% and 5.19% respectively than the third and fourth bands in December 7. The sensitive spectral indices SVM optimal recognition accuracy of three temporals were as follows: September 29th Quercus acutissima was 82.12%, Robinia Pseudoacacia was 57.37%; December 7th Quercus acutissima was 85.94%, Robinia Pseudoacacia was 83.78%; May 12th Quercus acutissima was 93.59% and Robinia Pseudoacacia was 85.44%. The highest overall recognition accuracy was 89.24% in May 12th. 【Conclusion】 Compared with the sensitive band, the recognition accuracy of SVM constructed by sensitive spectral indices can be improved effectively, especially for Robinia pseudoacacia. May 12 (spring) is the optimal temporal to identify the Quercus acutissima and Robinia pseudoacacia. This study provides a technical support for precise identification and forestry management of Taishan tree species.