Investigating the Use of Historical Node Location Data as a Source to Improve OpenStreetMap Position Quality

Abstract

Past years have witnessed the rapid use of participatory mapping and the establishment of open-source mapping infrastructures. The collection of user-generated geographic content has proven to provide highly effective means of acquiring useful and detailed global environmental data. OpenStreetMap is perhaps the best-established and reliable example that makes use of crowdsourced maps, constantly and rapidly evolving. Since data is heterogeneous in nature, the position quality of the objects edited and mapped into OpenStreetMap varies. The current OpenStreetMap practice is that the last node location edit of a map-feature is presented in the map, possibly inferring that it is the most accurate representation. Accordingly, OpenStreetMap presents the recent version of the accumulated edited and mapped data, disregarding the earlier versions. This chapter investigates the positional quality of historical node versions, evaluating the current OpenStreetMap practice alongside alternative spatiotemporal data models for improved node location calculation. With the proper incremental use of the historical node versions, the more accurate location for close to 60% of all the analyzed nodes was automatically selected between the last version location and the location calculated by the best model. This statistic increases to 76% for nodes that show larger discrepancies when compared to the reference locations. These preliminary results challenge the current OpenStreetMap practice by suggesting an alternative data model that use historical versions. The outcome validates and extends the existing patchwork approach associated with the crowdsourcing of user-generated geographic data and information that are the basis of OpenStreetMap.