The aim of this study was to develop a method to improve the accuracy of identifying mountain forests from multi-source remote sensing data. The Sunan County section of Qilian Mountains National Park was used as the test area. Data from the ICESat/GLAS （geoscience laser altimeter system） space borne LiDAR， Landsat OLI images， Google Earth high-resolution images， digital elevation model data， and field inventory data were used in the classification process. The vertical structure and the spectral， seasonal， and topographic characteristics identified from these data sources were integrated to extract forest information hierarchically. The results show that， after using a physically based terrain correction model to eliminate the influence of terrain on GLAS waveforms， the vertical structure derived from GLAS data combined with spectral information accurately identified forests at medium spatial resolution from remote sensing images of mountainous areas with complex terrain conditions. Compared with the classification based only on spectral characteristics， the classification incorporating vertical structure characteristics showed significantly improved accuracy， with overall accuracy improved by 10.67%. These results show that the vertical structure information provided by GLAS data can enhance the separability between different land cover types， and make forest range identification more effective. Secondly， for different forest types with similar spectral and vertical structure characteristics， the addition of terrain information reduced the impact of different objects with the same spectrum and the same objects with different spectra caused by terrain shadows， thereby significantly improving the classification accuracy of forest types. The addition of aspect information had a more significant effect than the addition of elevation information on improving the classification accuracy in this region. In addition， seasonal characteristics provided by multi-source and multi-temporal remote sensing images improved the classification of different forest types， while different band combinations had little effect on classification accuracy. The results of this study provide a reference for exploring mountain forest identification methods with low cost， high time efficiency， convenient operation， and a certain guarantee of accuracy.