Pan-European open building footprints: analysis and comparison in selected countries

Abstract

This paper presents a comprehensive analysis of four non-governmental open building datasets available at the European Union (EU) level, namely OpenStreetMap (OSM), EUBUCCO, Digital Building Stock Model (DBSM) and Microsoft’s Global ML Building Footprints (MS). The objective is to perform a geometrical comparison and identify similarities and differences between them, across five EU countries (Belgium, Denmark, Greece, Malta and Sweden) and various degrees of urbanisation from rural to urban. This is done in a two-step process: first, by comparing the total number and the total areas of building polygons for each dataset and country; second, by intersecting the building polygons and calculating the fraction of the area of each dataset represented by the intersection. Results highlight the influence of urbanisation on the dataset coverage (with increasing completeness when moving from rural to urban areas) and the varying degrees of overlap between the datasets based on a number of factors, including: the amount and up-to-dateness of the input sources used to produce the dataset; the presence of an active OSM community (for OSM and the datasets based on OSM); and the accuracy of Machine Learning algorithms for MS. Based on these findings, we provide insights into the strengths and limitations of each dataset and some recommendations on their use.