3.3 Microhabitat differences between marked and control sites
We found a total of 157 scent marked trees in the sample line survey, but due to topography and slope, etc., not every location of the marked trees was suitable for sample squares, so we made a total of 144 marked samples, corresponding to 114 control samples. Compared to the control sample, bamboo density (U = 5474.00, P < 0.001), bamboo cover (U = 5905.50, P < 0.001), and slope (U = 6328.50, P < 0.001) were lower in the marked sample; however, tree density (U = 4731.00, P< 0.001) and tree cover (U = 5729.00, P< 0.001) were higher; shrub density (U =7523.50,P = 0.226) and shrub cover (U = 7500.00, P = 0.130) were not significantly different between both (Table S3). In order to detect which variables had the greatest influence on the selection of this sample site by pandas, we performed stepwise logistic regression analysis on the remaining variables after Pearson Correlation Analysis to remove the significantly correlated variables. Finally, only tree density, bamboo density, slope, tree cover and shrub cover entered into the stepwise logistic regression equation. where tree cover was the most explained variable for whether giant pandas chose the location for scent marking or not (Table 3), and giant pandas preferred to scent mark at sites with higher tree cover.