Figure 3 Location of vegetation plots.

2.4 Data Processing

2.4.1 Image Interpretation

Standard false-color images are useful to highlight vegetation features as these are sometimes better suited for identifying certain vegetation than true-color images. To obtain more accurate visual interpretation results, we conducted this study based on standard false-color and true-color images, and the characteristics of vegetation morphology and distribution (Figure 4). We visually interpreted orthophoto images from vegetation plots and selected over 100 training samples for each plant species. Three pixel-based supervised classifiers (Support Vector Machine, Maximum Likelihood, and Artificial Neural Network) in ENVI were used to identify the main vegetation species. Classification results were verified, modified, and interpreted, and their accuracy was validated using synchronous field data. Two accuracy indices were selected: overall accuracy and Kappa coefficient.
Figure 4 Typical vegetation in the study area .

2.4.2 Diversity Assessment

Alpha-diversity is commonly used to assess species richness and relative dominance within a target community (Rocchini et al., 2016). Alpha-diversity encompasses different aspects, including species richness, evenness, and diversity (Peet, 1974). Species richness measures the abundance of species in a community, often quantified using the Margalef index (Clifford and Stephenson, 1975). Evenness quantifies the distribution of species within a community, often assessed by the Pielou index (Pielou, 1966); Diversity reflects the overall species richness and evenness and can be evaluated by Simpson’s index and Shannon-Wiener index (Simpson, 1948; Shannon, 1949). Grids ranging from 1 x 1m to 100 x 100m were generated using ArcGIS as research units, and vegetation diversity indices were calculated for each grid.
Moran’s I analysis was used to explore spatial correlation and clustering patterns of plant diversity (Lozada & Bertin, 2022). Global Moran’s I analysis, drawing upon geostatistical theory, is employed to assess the overall spatial clustering of plant diversity (Moran, 1950). Local Moran’s I analysis was applied to identify specific local clusters, thereby supplementing the limitations of global spatial autocorrelation in delineating precise clustering regions.

2.4.3 Spectral Vegetation Index

Considering the influence of soil reflectance on diversity assessment within arid and semi-arid regions, the following vegetation indices were selected (Kacic & Kuenzer, 2022): Normalized Difference Vegetation Index (NDVI) (Tucker, 1979), Difference Vegetation Index (DVI) (Tucker, 1979), Visible-Band Difference Vegetation Index (VDVI) (Wang et al., 2015), Soil Adjusted Vegetation Index(SAVI) (Huete, 1988), Modified Soil Adjusted Vegetation Index(MSAVI) (Qi et al., 1994), and Excess Green - Excess Red (EXG - EXR) (Meyer and Neto, 2008). Subsequently, correlation and regression analyses were conducted using the R4.3.0 software to explore the relationship between these vegetation indices and diversity indices.