With respect to changing ecosystems, a much-debated scenario is the reduction or loss of the biological or economic productivity and complexity of the land and its features, often referred to as land degradation. In this context, vegetation dynamics and especially trends in vegetation productivity over a time series of remote sensing data have been studied extensively in the past. However, underlying drivers of such changes remain widely unknown.
This thesis contributes to an improved assessments of vegetation and vegetation dynamics in global tropical drylands by 1) supplementing the knowledge about the extent of woody vegetation, 2) moving the analysis from pure vegetation productivity indicators to a measure of vegetation sensitivity to changes in rainfall as a simple proxy for vegetation functioning and 3) examining underlying potential drivers of change in vegetation productivity and functioning. Different remote-sensing based tools and methods within a multi-scale and data-rich approach have been applied and the results are presented in four research papers (two first- and two co-author publications). The first research paper introduced a method to monitor changes in vegetation sensitivity to rainfall, which is a major aspect of vegetation functioning in dryland areas. The method is based on linear regression between vegetation productivity and rainfall that calculated in a short temporal window, sequentially moved along a time series of remote sensing data (SeRGS, sequential linear regression slopes). Building upon that, the second research paper applied the method on global tropical dryland areas for the period 2000 – 2015. Moreover, this paper adapted and applied a method to estimate the relative importance of various potential driving factors on the observed changes. Complementing these studies, research paper three investigated a time series of field-measured woody foliage and herbaceous mass from Senegal and its relation to rainfall; and extrapolated the results to a larger scale using a wide range of satellite data. Finally, research paper four mapped individual non-forest trees in the western Sahara and Sahel based on state-of-the-art deep learning methods and very high resolution (0.5m spatial resolution) satellite data.