Evaluating global land degradation using ground-based measurements and remote sensing

Abstract

Understanding the impacts of land degradation is, at least in part, limited by our ability to accurately characterize those impacts in space and time. While in recent decades remote sensing has offered unprecedented coverage of the land surface, the evaluation of remote sensing products is often limited or lacking altogether. In this chapter we use a survey-based approach to evaluate how well already existing remotely sensed datasets depict areas of land degradation. Ground-based surveys are compared to existing maps of land degradation and independent remote sensing datasets. This provides a metric of evaluation by using the commonly understood confusion-matrix. A representative set of case study countries was chosen after all countries were grouped using a k-means clustering approach (see Chap. 2). Survey sites within each country were sampled according to the intersection of agro-ecological zones, land cover, and the dataset to be evaluated. This two-tiered approach to sampling ensured a diversity of ground-truth surveys and therefore a robustness of results. Although ground-based surveys are resource and time-intensive, they provide information on both the evolution of the land cover and the drivers of land-cover change. Land degradation is a very complex process where diverse data are often needed for interpretation.