Figure 6 Main misclassified areas in detail of three supervised methods.

3.1.2 Distribution of Vegetation Species

Vegetation distribution varies greatly across different regions. Poplar and Caragana are predominantly found in the east, while Poplar and Artemisia are abundant in the central, and Artemisia and Salix are mostly present in the west. Poplar was the primary tree species and formed a vertical community structure of trees-shrubs-herbs. Shrubs like Salix, Artemisia, and Caragana had a patchy distribution, and some Salix and Caragana were planted in rows, whereas Artemisia tended to exhibit a clustered distribution. The clustering of the same species is due to their common habitat requirements, which creates a competitive advantage for the group as a whole, forming a suitable growth environment (Amarasekare, 2003). This may enhance resilience against adverse conditions, species invasion, facilitates resource competition and promotes growth.
Shrubs were the most widely distributed vegetation class, covering 42.5% of the area, followed by Grass and Bare classes (Figure 5-c). Grasslands are common in semi-arid regions, but in the study area, the vegetation is predominantly composed of shrubs, which can be considered a form of shrub-encroached grasslands. However, the ecological succession dynamics and vegetation patterns resulting from shrubification have been extensively discussed, with ongoing debates regarding the stability of this new ecological state and its potential implications for ecosystem functioning and biodiversity conservation (Suding et al., 2004; Pierce et al., 2019; Ding and Eldridge, 2023).

3.2 Vegetation Diversity

3.2.1 Species Accumulation Curves

As the area of a plot increases, there is typically a greater number of species present in the SVM-classified UAV data. We designated the study area into sampling plots of varying scales by ArcGIS and quantified the average species count within each plot area., ultimately creating a species-area curve (Figure 7). The curve exhibits a clear inflection point at the 10m x 10m scale, covering around 65% of the surveyed species. This corresponds to the standard plot size for vegetation surveys. The 30m × 30m scale covers at least 95% of the total number of species, meeting the criteria for ”minimum area for optimal communities” (Barbour et al., 1980). Additionally, a 30m x 30m plot size is typically used for local vegetation surveys in the semi-arid region of Inner Mongolia, where species numbers are relatively low. Therefore, vegetation diversity was analyzed at 10m x 10m and 30m x 30m scales.