Crowdsourced Validation and Updating of Dynamic Features in OpenStreetMap - An analysis of Shelter Mapping after the 2015 Nepal Earthquake

The paper presents results from a validation process of OpenStreetMap (OSM) rapid mapping activities using crowdsourcing technology in the aftermath of the Gorkha earthquake 2015 in Nepal. We present a framework and tool to iteratively validate and update OSM objects. Two main objectives are addressed: first, analyzing the accuracy of the volunteered geographic information (VGI) generated by the OSM community; second, investigating the spatio-temporal dynamics of spontaneous shelter camps in Kathmandu. Results from three independent validation iterations show that only 10 % of the OSM objects are false positives (no shelter camps). Unexpectedly, previous mapping experience only had a minor influence on mapping accuracy. The results further show that it is critical to monitor the temporal dynamics. Out of 4,893 identified shelter camps, 54% were already empty/closed six days after the first mapping. So far, updating geographical features during humanitarian crisis is not properly addressed by the existing crowdsourcing approaches.