Abstract: The canopy width is a necessary variable for the growth of desert vegetation. Predicting the crown width of desert plant in different habitats can provide reference for the scientific management of desert vegetation. Taking the species of Haloxylon ammodendron in the Gurbantunggut Desert as the research object, select 6 commonly（M1, M2, M3, M4, M5, M6） used regression models, using tree height and ground diameter as survey factors, establish a canopy model of H. ammodendron in different habitats of highlands, gentle slopes, and flat lands. Perform regression analysis on the model using Origin and Anaconda software, select the optimal model based on four fitting accuracy standards（R2, RMSE, MAE, and MAPE）. The model fitting results indicate that: in the tree height-shrub crown width and shrub ground diameter models, the R2 values of the six regression models are all greater than 0. 5, the model fitting effect is good. M5（cubic polynomial model） is the optimal model for tree height crown width and ground diameter crown width in three different habitats. All six models have statistical significance（Sig. ＜0. 001）, which has important reference value for predicting the crown size of H. ammodendron in the Gurbantunggut Desert, and provides a certain scientific basis for maintaining the ecology and vegetation restoration of the region.