Remote sensing for site selection in vegetation survey along a successional gradient in post-industrial vegetation

Abstract

Vegetation characteristics are an important proxy to measure the outcome of ecological restoration and monitor vegetation changes. Similarly, the classification of remotely sensed images is a prerequisite for many field ecological studies. We have a limited understanding of how the remote sensing approach can be utilized to classify spontaneous vegetation in post-industrial spoil heaps that dominate urban areas. We aimed to assess whether an objective a priori classification of vegetation using remotely sensed data allows for ecologically interpretable division. We hypothesized that remote sensing-based vegetation clusters will differ in alpha diversity, species, and functional composition; thereby providing ecologically interpretable division of study sites for further analyses. We acquired remote-sensing data from Sentinel 2A for each studied heap from July to September 2020. We recorded vascular plant species and their abundance across 400 plots on a post-coal mine in Upper Silesia, Poland. We assessed differences in alpha diversity indices and community-weighted means (CWMs) among remote sensing-based vegetation units. Analysis of remotely sensed characteristics revealed five clusters that reflected transition in vegetation across successional gradients. Analysis of species composition showed that the 1st (early-succession), 3rd (late-succession), and 5th (mid-succession) clusters had 13, 10, and 12 exclusive indicator species, respectively, however, the 2nd and 4th clusters had only one species. While the 1st, 2nd, and 4th can be combined into a single cluster (early-succession), we found the lowest species richness in the 3rd cluster (late-succession) and the highest in the 5th cluster (mid-succession). Shannon's diversity index revealed a similar trend. In contrast, the 3rd cluster (late-succession) had significantly higher phylogenetic diversity. The 3rd cluster (late-succession) had the lowest functional richness and the highest functional dispersion. Our approach underscored the significance of a priori classification of vegetation using remote sensing for vegetation surveys. It also highlighted differences between vegetation types along a successional gradient in post-mining spoil heaps.

Keywords: functional diversity; indicator species; phylogenetic diversity; post‐mining sites; species composition.

FIGURE 1

Example of site and plot selection using the background of remotely sensed classes and their distribution on chosen post‐industrial objects (black numbers).

FIGURE 2

Distribution of study plots (n = 400) in Upper Silesia. The study design shows plots in the north (N), south (S), east (E), and west (W) directions at 50 m from the central plot (C).

FIGURE 3

Result of principal components analysis (PCA) of remotely sensed characteristics of pixels (Table 1), colored according to k‐means clustering.

FIGURE 4

Result of nonmetric multidimensional scaling (NMDS, stress = 0.1631) of vegetation in study plots (points), colored according to k‐means clustering (Figure 2).

FIGURE 5

Mean (+SE) values of the community‐weighted mean (CWM) of ecological indicator values (EIV) and functional traits describing studied vegetation among remote sensing‐based clusters, assessed using linear mixed‐effects models (Table 4). The same letters denote groups that did not differ at the confidence level α = .05 after multiple hypotheses adjustment, according to a Tukey posteriori test.

FIGURE 6

Mean (+SE) values for alpha diversity indices of studied vegetation among remote sensing‐based clusters, assessed using linear mixed‐effects models and a generalized linear mixed‐effect model (Table 5). The same letters denote groups that did not differ at the confidence level α = .05 after multiple hypotheses adjustment, according to a Tukey posteriori test.