The potential of OpenStreetMap for land use/land cover mapping

Land use and land cover (LULC) maps are important tools for the management of the environment, in particular for the modelling of biogeochemical cycles and climate. LULC maps are typically produced using supervised classification of satellite images supported by expert knowledge. Nowadays, the advance of the crowdsourced project OpenStreetMap (OSM) offers an alternative to official LULC maps. In OSM, LULC is usually mapped by manual digitalisation of satellite imagery and/or by the import of existing databases. We explored the potential of OSM to contribute to the improvement of an existing LULC map, Corine Land Cover (CLC), by comparing forest cover in Southern Belgium and Luxembourg. We focused on three forest types that were common to the two databases: coniferous, deciduous and mixed forest. Confusion matrices showed that the two databases differed in terms of classes over about one quarter of the forest area. Moreover, 9.9% and 8.0% of the forest area in OSM and CLC, respectively, were unmapped as forest area compared to the other database. Compared to CLC data, small patches of forest areas are mapped in OSM data. This study is a first attempt to show the potential of a crowdsourced project, i.e., OSM, to contribute to an official geodatabase. While current drawbacks of OSM such as the lack of completeness may limit its application in some operational applications, we claim that its large accessibility and its potentially high reactiveness makes OSM a serious alternative to existing geodatabase.