The project addresses the important question of where and how fast biomes are changing over time and space, including a quantitative description of the most relevant parameters characterizing this change.
Landscape heterogeneity has recently been used as an indicator of biome boundaries in several studies. Changes over time in the heterogeneity indicate biome transition zones that have been subject to potentially abiotic changes. Such heterogeneities are best observed using remote sensing data. While the temporal variation of the NDVI has been studied almost exclusively on a pixel level we address the spatial structure of the temporal variation in this project. As a result we quantify and model the changes in the biome boundaries and the possibly resulting tipping points.
The project is structured into three milestones.
This first milestone has two purposes. First, it allows the PhD student to get acquainted with spatio-temporal statistics and the handling of remote sensing data. We intend to use NDVI data at different resolutions (sources AHVRR, MODIS, possibly up to Landsat-8 OLI or Sentinel-2 MSI) and apply several decomposition techniques to separate the data into different roughness spaces. We explore different computing environments to handle and process data (e.g., Google Earth Engine, R). In the second step we determine the optimal spatial scale that best detects the transition zones. For this scale we develop a parametric description of the ecotones.
The second milestone addresses changes of ecotones. We estimate changes over time of the ecotones boundaries. The resulting gradient is then again smoothed and predicted using different (observable) proxies.
The third milestones may address the use of these proxies as indicators for “tipping points”, i.e., massive nonlinear shifts in biomes. Based on the results of the first two milestones we establish rules that indicate where tipping points are (in some phase space).