Assess citizen science based land cover maps with remote sensing products: the Ground Truth 2.0 data quality tool

摘要: One of the main concerns in adopting citizen science is data quality. Derived products inherit intrinsic limitations capture methodology as well uncertainties observations. OpenStreetMap tools are designed to minimize positional accuracy by ensuring a good co-registration observations with imagery or direct use GPS. When thematically annotating objects contributed citizens, uncertainty increases. During H2020 GroundTruth 2.0 project two land-cover derived from OSM were analyzed; one created University Heidelberg (http://osmlanduse.org) and another elaborated Coimbra (https://vgi.uc.pt/vgi/osm/osm2lulc/). To be able assess quality both maps, third product remote sensing was introduced reference map. In tool show compare maps part MiraMon Map Browser developed. The objective allow final users auto-evaluate their region interest. confusion matrix has been used method derive overall commission omission estimators Kappa coefficient. Most discrepancies between (RS) related different approaches during capturing. assesses individual exposed using OGC standard describes an interoperable approach based on QualityML.