Master Thesis

A closer look from space: The potential of Sentinel-2 data for beta-diversity measurements on small scale in a tundra alpine ecosystem

Edvinas Rommel (06/2018-05/2018)

Support: Carl Beierkuhnlein, Duccio Rocchini

Spectral heterogeneity of remote sensing imagery can be used efficiently for biodiversity assessment and presents a valuable source of ecological information, especially when free available data is used. This approach, known as the spectral variation hypothesis (SVH) is based on environmental heterogeneity, which is both related to spectral variability and to biodiversity. Little attention has been paid so far to the question whether the SVH can be applied on small scale and on free available data. The present study aims to fill this gap by firstly investigating the positive influence of topographic complexity on beta-diversity on community level scale in Hardangervidda National Park. Secondly the performance of spectral heterogeneity, derived from Sentinel-2 data, for beta-diversity prediction is quantified. A vegetation survey was conducted in eight 100 x 100 m sites, each containing 20 systematic-randomly placed plots of 2 x 2 m. Four sites were located in high topographic complexity areas (HTC) and four in low topographic complexity (LTC). Jaccard and Bray-Curtis dissimilarity were calculated for each site and correlated with eight different spectral heterogeneity measurements all based on the mean Euclidean distance to the spectral centroid using different spatial resolution and band numbers. High topographic complexity was related to high beta-diversity. The same pattern could be observed in spectral heterogeneity. Spectral heterogeneity performed better with frequency based dissimilarity than with abundance based dissimilarity, probably because the latter was biased due to vegetation patchiness. The best performing heterogeneity measurement was NDVI heterogeneity with a spatial resolution of 10 m. An increase in spectral resolution could increase the performance of SVH but could not compensate the lack in spatial resolution. The results of this study show that free available remote sensing data can provide reasonable estimates of beta-diversity patterns on small scale. Moreover, as theoretically demonstrated, spectral heterogeneity could even be used to improve efficiency of field surveys as well as mitigating location mismatch between remote sensing and in-situ data.