Mapping global exposure from space: A review of existing products and comparison of two new layers of global urban extent

Abstract

In order to assess and quantify earthquake risk over large spatial extents on a global or regional scale spatial data localizing human assets is on strong demand. However, due to the large-scale extent of human activities on our planet, the scientific community has been lacking tools and methodologies to capture the entity of elements at risk, especially with enhanced thematic and geometric detail on global scale. First generation of global land cover datasets and maps of urban extent produced since the millenium relied on coarse resolution satellite sensors such as MODIS or DMSP-OLS. Despite the availability of finer scale imagery no other product of global coverage has been made available until today. However, global human exposure mapping from remote sensing is now entering a new era and new products currently in the making such as the Global Urban Footprint and the Global Human Settlement Layer will deliver spatial information on human settlements at unprecedented spatial resolutions. Once available on a global scale, these layers will improve significantly the knowledge base for the localization of human assets with regard to georisks. This work reviews past mapping products of coarse geometric resolution and further presents first validation efforts of the two new layers based on standard accuracy measures with regard to a geometrically and thematically highly resolved building reference. This first assessment reveals that the high resolution settlement layers produce comparable outputs that are capable of mapping even smaller settlements in low-density rural areas although significant over-classification due to the specific characteristics of each classifier is obvious. To gain a deeper understanding of these layers further validations efforts using pattern-based regression analysis and measures of inter-map agreement are currently underway.