Global Classification of Human Settlement with TanDEM-X Data
Marconcini, Mattia; Esch, Thomas; Felbier, Andreas; Heldens, Wieke
German Aerospace Center (DLR), GERMANY

Since the beginning of the 21st century, more than half of the global human population is living in urban environments and the dynamic trend of urbanization is expected to grow incredibly fast, with the number of urban dwellers currently increasing by about 180.000 people every day (as an example, in China and India the number of new cities with more than one million inhabitants will come up to 30 and 26, respectively, just within the next 20 years) [1]. With this perspective, the actual decade marks the start of an "urban century".
It is approximated that urban areas cover about 2% to 3% of the Earth's surface. However, despite this rather marginal significance in terms of spatial coverage, metropolitan areas represent the focal points of human activity. Therefore, the impacts of urbanization on the natural and human environment are much more far-reaching at all geographic and socioeconomic scales than the purely area-related perspective might imply. In this framework, an effective monitoring of urban sprawl represents a key issue to analyse and understand the complexity, cross-linking and increasing dynamics of urban environments in order to ensure a sustainable development of urban and peri-urban areas.
To this purpose, in the last decades satellite Earth observation (EO) has proved to be a promising tool in combination with widely automated methods of data processing and image analysis for providing up-to date geo-information on urban settlements at global scale [2],[3]. Nevertheless, the geometric resolution of the current EO-based geo-information products is limited to 300-1000 m, thus often resulting in poor accuracy to support decision makers and urban planners (nowadays, the MODIS 500 at 463 m spatial resolution [4] and GlobCover at 309 m spatial resolution [5] are considered to be the most accurate state-of-the-art data sets provided on a global level).
TanDEM-X (TerraSAR-X add-on for Digital Elevation Measurement, TDM) is a German radar satellite mission candidate for ESA's Third Party Mission scheme, which aim at the provision of a global digital elevation model (DEM) at 12 m spatial resolution [6]. Besides this primary goal, the global coverage with very high resolution (VHR) TerraSAR-X (TSX) and TanDEM-X (TDX) imagery collected from 2011 to 2013 can be used to characterize settlement patterns world-wide in a so far unique spatial detail. Accordingly, the German Remote Sensing Data Center (DFD) of the German Aerospace Center (DLR) has implemented a fully-automated pixel based image analysis procedure that detects built-up areas from the global TSX/TDX imagery acquired in the context of the TDM. In particular, a texture feature (i.e., the so called speckle divergence [7]) is first extracted from the original SAR backscattering amplitude image and an unsupervised approach exploiting the Jensen-Shannon divergence [8] has been developed to automatically identify an optimal thresholds for it in order to derive reliable candidate labeled samples of the urban class accordingly. These are then used to effectively train a support vector data description (SVDD) one-class classifier which is employed to detect built-up areas [9]. In addition, specific mosaicking and masking operations are applied in an automated post-editing phase to further improve the quality of the final products.
The output of this approach will be a global binary mask of human settlements derived from ~180.000 TSX/TDX images outlining urban and non-urban areas at the unprecedented spatial resolution of ~12 m. The intended, world-wide data set is referred to as Global Urban Footprint (GUF) and a public domain version of it will be made available at three arcsec (~50-75m) spatial resolution.
With its global coverage and the enormous spatial detail, this initiative represents a promising contribution to global analyses of urban and peri-urban areas and it is expected to definitely improve the characterization of human settlements even in remote regions where no other sources of information are now available.
It is worth noting that as primary follow-on analysis we are currently investigating the possibility of adapting the proposed methodology to ERS and ENVISAT data by accounting for short- and/or long-term coherence. This would represent a powerful tool for effectively mapping urban growth at high spatial resolution from 1991 to 2012 and could be then futher extended to Sentinel-1 data in order to set-up an operational system for a global urban monitoring on a regular basis.

[1] United Nations, "World Population Prospects: The 2010 Revision": (15.02.13);
[2] P. Gamba, and M. Herold, "Global Mapping of Human Settlement - Experiences, Datasets, and Prospects, "CRC Press, 2009;
[3] D. Potere, A. Schneider, S. Angel, and D.L. Civco, "Mapping urban areas on a global scale: which of the eight maps now available is more accurate?," International Journal of Remote Sensing, vol. 30, no. 24, pp. 6531-6558, 2009;
[4] A. Schneider, M.A. Friedl, and D. Potere, "Mapping global urban areas using MODIS 500-m data: New methods and datasets based on urban ecoregions," Remote Sensing of Environment, vol. 114, pp. 1733-1746, 2010;
[5] European Space Agency (ESA), "GLOBCOVER 2009: Products description and validation report". Available URL: http:// (15.02.13);
[6] G. Krieger, A. Moreira, H. Fiedler, I. Hajnsek, M. Werner, M. Younis, and M. Zink, "TanDEM-X: A Satellite Formation for High Resolution SAR Interferometry," IEEE Transactions on Geoscience and Remote Sensing, vol. 45, no. 11, pp. 3317-3341, 2007;
[7] T. Esch, M. Thiel, A. Schenk, A. Roth, A. Müller, and S. Dech, "Delineation of Urban Footprints From TerraSAR-X Data by Analyzing Speckle Characteristics and Intensity Information," IEEE Transactions on Geoscience and Remote Sensing, vol. 48, no. 2, pp. 905-916, 2010;
[8] J. Lin, "Divergence measures based on the Shannon entropy," IEEE Transactions on Information Theory, vol. 37, no. 1, pp.145-151, 1991;
[9] D. M. J.Tax, and R. P. W. Duin, "Support vector data description," Machine Learning, vol. 54, no. 1, pp.45-66, 2004.