VGI in production of global human settlements layer from multi-temporal Landsat satellite imagery

Global human settlement information is required by a number of institutions operating globally, and to derive standardised, action-oriented indicators. The new Landsat-based Global Human Settlement Layer provides such information baseline. The dataset describes the spatial evolution of human settlements in the past 40 years. The automatic processing workflow has made large use of already available spatial information describing human settlements, gathered from various sources at different scales, thematic definition and completeness or accuracy. In general, the data were collected and made available by open collaborative projects or, more structured, by academic or governmental organizations. Notwithstanding well-known quality issues of Volunteered Geographic Information (VGI), this source also was used to derive training sets, as it offers a unique opportunity to exploit local knowledge at global scale. These fine-scale training sets consist of information extracted from sources with comparable or better scale with respect to the image data. They include Open Street Map data (settlement places, and urban cover), the GeoNames settlements location points, and the settlement areas extracted from other fine-scale land cover information. The broad-scale training sets are made by the MODIS global urban extents, the Meris GlobCover, and the population density grid LandScan. Some fine-scale training set sources, and especially VGI data, are arbitrarily available in some locations with unknown completeness and update characteristics. Therefore, relying only on these sources would introduce instability in the overall system reliability performances. In this work, we show how the training sets have been created, combined and used in the production of the global product. The main problems and their mitigation are described. Also, the remarks on current trends in VGI are discussed, which may bring an interesting input to the VGI community.

FLORCZYK Aneta;  CARNEIRO FREIRE Sergio Manuel;  KEMPER Thomas;  PESARESI Martino; 

2016-10-17

International Society for Digital Earth

JRC101306