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.