So2Sat: Big Data for 4D Global Urban Mapping - 1016 Bytes from Social Media to Earth Observation Satellites

Project Leader
Xiaoxiang Zhu

Contact Person
Xiaoxiang Zhu

Cooperation Partners
German Aerospace Center- Remote Sensing Technology Institute (DLR-IMF)
German Aerospace Center - German Remote Sensing Data Center (DLR-DFD)
Technical University of Munich (TUM)


In 2050, it is expected that three fourths of the world's population will live in metropolises. This transition will inherently alter the physical dimensions and configurations of cities at all scales, which is a fact presenting an enormous challenge to urban planners and logisticians. Yet, our understanding of urbanization at these scales is primarily based on United Nations population figures, but these statistics do not provide information on the distribution, pattern, and evolution of the built environment. For example, the knowledge of the spatial evolution as well as the posulation density of informal settlements such as **slums or refugee settlements** in many mega-cities is far from sufficient for a sustainable planning.

So2Sat is an ambitious European Research Council (ERC) starting grant project. In the project, we will use revolutionary mapmaking methods to close these knowledge gaps and investigate how human settlements grow in global scale. We have the privilege to access the data supplied by several German and European earth observation satellites, such as TanDEM-X and EnMAP, which are equipped with innovative sensor technologies. We will develop new algorithms for the derivation of geo-information from these measurements. This makes it possible to create high-resolution 3D/4D maps of the cities up to individual building. For the first time, this information will also be combined with data from social media: crowdsourcing platforms such as OpenStreetMap providing up-to-date map material; photos posted to the social media providing authentic and current images where building facades can be seen with much higher resolution than the earth observation data or where, for example, the extent of damage caused by a natural disaster can be revealed. The major challenge here is consolidating this information and evaluating it automatically in a global scale. By exploiting the recent advances in signal processing and computer vision, the aim of So2Sat is to - for the first time - systematically fuse the remote sensing data acquired by sensors mounted on diverse platforms, in particular from Earth observation satellites, with the massive data available from GIS and social media to map 3D urban infrastructures and their evolution over time, i.e. 4D, in high resolution and on a global scale. The outcome of this project will be a first and unique global and consistent 3D/4D spatial data set on the urban morphology of settlements, and a multidisciplinary application derivate assessing population density. This is seen as a giant leap for urban geography research as well as for formation of opinions for stakeholders based on resilient data.

Among the various objectives in So2Sat, the following pilot applications are of great social impact to the humanity:

3D/4D urban infrastructure model

To date, global 3D urban information is inexistent, not even to mention its temporal elvolution, i.e. 4D. The state-of-the-art global digital elevation model (DEM) from the German TanDEM-X mission delivers a DEM of high quality (12 m posting) close to HRTI-3 standard. However, when it comes to urban mapping, layover effects caused by the side-looking nature of the radar satellites handicap the use of TanDEM-X data alone for a precise 3D reconstruction in urban areas. The So2Sat project will develop algorithms and methodology to fuse and combine TanDEM-X data and ground-level data from social networks on a global scale. The Figure shows a 3D model of Berlin created by *fusing the result of SAR interferometry and optical multiview stereo matching*. Behind each of the SAR points, there is also the deformation history with an accuracy up to millimeter.

High resolution population density map

Almost every study dealing with population data from national to global scales relies on the data published by the United Nations or from LandScan. The database is globally available, but they are only at a spatial resolution of about 1km, and with few temporal samples which is not suitable for constant monitoring. So2Sat will seek for significant breakthroughs in high-resolution and high-quality population estimation. The figure to the right provides the estimated building-level population in a slum area in Mumbai. With the So2Sat project, we aim to extend such analysis to a wide range of cities and building types.

Localization of informal settlements

Currently, the spatial layout of informal settlements within the wider urban context is generally subject to descriptive studies in geography. So2Sat will develop a formal ontology to describe the physical characteristics of informal settlements across the globe. With the capability to globally locate and characterize such informal layouts in cities, a unique spatial knowledge will be available to tackle the more specific question of the amount of slum dwellers, where extrapolation methods will be developed. The image depicts a slum area in Dharavi, Mumbai mapped by remote sensing data. Left, the classification of space, and right, the estimated population per building.