Font Size: 

In recent years, an emerging trend in information community is the growing use of web applications to collect and share geographic information. Such initiatives have reduced the inaccessibility of geodata. Collaborative mapping platforms such as OpenStreetMap (OSM) have become an important source and potential complementary source of geodata for any Spatial Data Infrastructure (SDI) initiatives. However, as volunteered geodata were generated from people with various skill levels randomly, quality issues such as vandalism, missing details, and incomplete content are inevitable in this approach.

In this study, we assessed the completeness and the geometric accuracy of OSM building footprints in two major metropolitan areas of Taiwan: Taipei City and Taichung City. We compared OSM quality with a reference dataset from authority. The completeness assessment was computed in different scales by computing the total number and the total area. The assessment shows mixed results. The completeness of total area (C_Area) is 17.4% in Taipei and 12.8% in Taichung respectively. The highest complete location is Central district of Taichung (C_Area = 74.3%). Generally, the central business districts have higher completeness than the low-density areas. The consistency of the completeness is a significant issue in two cities. An interesting finding is that the resolution of OSM building footprints in several districts of Taichung is higher than the reference dataset.

Then, we used an overlap method to identify OSM building footprints with a 1:1 relation to the reference dataset. 100 buildings were sampled randomly. The completeness of the 100 corresponding OSM building footprints reaches 86.4%. Using a turning function, the geometric accuracy assessment identifies that 10% is very similar to the reference building yet 12% is very dissimilar. However, the number of vertices in OSM building footprints is only 35% as compared with the reference dataset. Thus, we conclude the authoritative data is more complex in building representation.

Furthermore, as the Taiwanese OSM community intended to tag the building footprint for evacuation, we tried to identify the completeness of evacuation building in the two cities. The result shows that 47.1% of evacuation building can be identified on OSM in a precision tolerance of 20-meter.

These results indicate that the completeness of OSM building footprints is not consistent, and the OSM building representation is not as complex as the reference dataset. However, the OSM building footprints in high resolution show a great potential for research use, particularly in a scenario of disaster management. We also found that the high-resolution footprints are due to the promotion of the university education. Therefore, we suggest that the education system should consider using OSM as a tool for enabling a resilient society. Such action also improves the completeness of OSM building footprints and as a result, it would improve a better source for a large-scale SDI platform as well.

VGI; data quality; disaster management