Header image  
  HOME ::

How good is crowd-sourced PPGIS/PGIS/VGI Data?


Spatial data generated through crowd-sourcing methods such as PPGIS/PGIS/VGI can vary greatly in data quality. How does one measure the quality of spatial data? Conventional GIS data is assessed for data quality using the criteria of positional accuracy, attribute accuracy, logical consistency, and completeness. The quality of VGI data has been examined in multiple studies by comparing the digital contributions of volunteers to the OpenStreetMap (OSM) project to authoritative GIS data. These studies have found that the positional accuracy of OSM data is comparable to geographical data maintained by national mapping agencies and commercial providers. More specifically, Haklay et al. (2010) found between 80% and 86% spatial accuracy in a test case in England. In 2014, we examined the quality of PPGIS/VGI data generated from a statewide study in Victoria, Australia, whose purpose was to identify public land values and preferences. We examined the quality of PPGIS/VGI data for conservation planning (where the public mapped biological values) by comparing the data with an expert model of ecological values called NaturePrint. Specifically, we examined the positional accuracy and completeness of the PPGIS data using the NaturePrint model as an accuracy benchmark. We found that about 70% of PPGIS points that identified biological/conservation values were spatially coincident (position accurate) with the NaturePrint model. The PPGIS process did not comprehensively identify all the largest, contiguous areas of high conservation importance in the state, missing about 20% of areas, primarily on small public land units in less densely populated regions of the state. The full publication is available here.

Figure above shows the spatial accuracy of PPGIS data compared to NaturePrint expert data for (a) mapped public land locations, and (b) non-public land locations. Green colors indicate greater PPGIS spatial accuracy (dark green is best) with orange/red colors indicating less spatial accuracy (red is worst). The least accurate PPGIS data was mapped in the greater Melbourne urban area (shown by arrow) where PPGIS participants perceived these locations to have high conservation value, but the expert model did not consider these locations to have high conservation value.

For more information about this study, contact Greg Brown.