SUMMARY
The aerodynamic roughness length (z0m) is a crucial parameter for reliably simulating turbulent exchanges between the land surface and the atmosphere. Due to the large number of input variables related to vegetation growth and aerodynamic conditions near the surface, estimating z0m precisely is difficult and, to date, no universal model has been established. Understanding the z0m changes in time series data and the relative contributions of vegetation indices and meteorological factors is important to providing a basis for modelling z0m. In this paper, the main meteorological factors that influence z0m in different seasons are presented based on data from three automatic weather stations (AWSs) that represent various land surface patterns in the Heihe river basin. A correlation analysis identified the dominant factors that influence z0m changes at half-hour and daily scales; then, a factor analysis was performed to identify the different contributions of vegetation indices and meteorological factors to z0m at different time scales. The results show that meteorological factors (wind speed, wind direction and atmospheric stability) are the main driving factors for z0m at the Arou and Guantan sites, which are situated in grassland and forest mountain areas, respectively, and that the vegetation indices have no impact on the z0m variations in these areas. In contrast, for the Daman site, situated in flat farmland, the vegetation indices are the primary driving factors, while meteorological factors such as wind speed and atmospheric stability are secondary factors, and wind direction has no significant influence. Finally, a detailed analysis was conducted to detect the relationships between half-hourly z0m measurements and three dominant meteorological factors.