Combining remote sensing data at different spatial, temporal and spectral resolutions to characterise semi-natural grassland habitats for large herbivores in a heterogeneous landscape
by Christoph Benjamin Raab
Date of Examination:2019-07-04
Date of issue:2019-11-22
Advisor:Prof. Dr. Johannes Isselstein
Referee:Prof. Dr. Johannes Isselstein
Referee:Prof. Dr. Niko Balkenhol
Referee:Prof. Dr. Hannes Feilhauer
Files in this item
Name:raab_diss_07052019_web.pdf
Size:45.9Mb
Format:PDF
Description:Dissertation
Abstract
English
Semi-natural grasslands are ecosystems with high biodiversity. In Europe, such open and half-open areas are a fundamental characteristic of the cultural landscape, originating from and depending on management activities. The possibilities that grazing with wildlife can provide for sustaining these open-land ecosystems are subject to current research activities, because only a small proportion of grasslands protected under the EU Habitats Directive has a favourable conservation status. For an active grazing management, spatial information about the landscape structure and forage quality and quantity is required, as they can affect the spatial distribution and activities of free-ranging herbivores and thus their influence on the ecosystem e.g. by grazing. The collection of field data, however, is labour-intensive, time-consuming and often limited to a particular location. Therefore, this thesis is concerned with techniques and concepts offered by satellite remote sensing technology to characterise a heterogeneous landscape dominated by semi-natural grassland. After a general introduction to the wider research context in Chapter 1, Chapter 2 illustrates how Tasselled-Cap-transformed multi- temporal RapidEye remote sensing data can be successfully used to derive a classification map for a heterogeneous landscape. The results suggest that the RapidEye Tasselled Cap Transformation, which was designed for agricultural applications, can be an effective data compression tool, suitable to map heterogeneous landscapes with no measurable negative impact on classification accuracy. Chapter 3 presents a framework on mapping semi-natural grasslands at community level using multi-temporal RapidEye remote sensing imagery. For this, an automated training data selection was successfully implemented based on the Random Forest proximity measure. This strategy can support the reporting obligations under Art.-17 of the EU Habitats Directive in the future. Chapter 4 discusses how semi-natural grassland forage quantity and quality indicators can be predicted using combined optical and radar satellite remote sensing data. A permutation-based variable importance measure indicated a strong contribution of simple-ratio-based optical indices to the model performance. The final Chapter 5 summarises and discusses the results of this work with reference to the current research context. The findings of this thesis can help to understand and manage the grazing behaviour of free- ranging large herbivores and thus, support the conservation of semi-natural grassland in the future.
Keywords: semi-natural grassland; European Habitats Directive; remote sensing; monitoring; Random Forest; RapidEye; Sentinel-2; Sentinel-1