[Objective] By monitoring the long-term dynamic changes in the characteristics of plant communities in large forest plots, the spatial patterns and maintenance mechanisms of species diversity are revealed, providing a scientific basis for biodiversity conservation in the region. [Methods] This study takes the typical forest ecosystem as the research object in the Qinghai area of Qilian Mountain National Park. Furthermore, we used the adjacent grid method to conduct a survey of each tree in a 24 hm2 large sample plot. [Results] There was a total of 35 835 trees, of which Picea crassifolia and Juniperus przewalskii accounted for 57.84% and 23.82%, respectively. Species richness and plant height are 3 species and 10.7 m, respectively. Shannon-Wiener and Simpson index of spruce forest are 0.74 and 0.43, respectively, Shannon-Wiener index was relatively low. The Shannon-Wiener index is significantly influenced by tree height, species richness, and Simpson index. As the tree height increases, the Shannon-Wiener decreases, while the species richness and Simpson index increase significantly. The coefficients of determination for the training and testing sets of the machine learning model were 0.95 and 0.93, respectively, with root mean square errors of only 0.06 and 0.08. This indicates that the model has a high explanatory power and prediction accuracy for the Shannon-Wiener data. [Conclusion] In conclusion, increasing species richness and optimizing tree species structure could effectively enhance the biodiversity of the area.