Exploring the accuracy and completeness patterns of global land-cover/land-use data in OpenStreetMap

Abstract

OpenStreetMap (OSM) can supply useful information to improve land-cover/land-use (LCLU) mapping. However, many concerns have been paid attention to OSM data quality, because the data were edited by global volunteers and they have only been assessed at a city/regional scale rather than at a global scale. This study assesses the quality of OSM-LCLU data for 168 countries worldwide. OSM-based LC datasets are firstly produced for different countries by referring to global open LC data, and these dataset are then compared in terms of accuracy and completeness. Moreover, a number of variables and three regression models (OLS, SLM and SEM) are used to understand the accuracy and completeness patterns at a global scale. We found that: 1) although most countries are characterized by a low completeness, they have a relatively high accuracy. 2) Both socio-economic variables and the area percentages of various OSM-LCLU types have been found to be significantly correlated with these patterns. 3) The SLM and SEM models are preferred, and most of countries with a relatively high completeness are spatially aggregated in Europe. Not only the global pattern of the OSM-LCLU data quality have been recovered, but also the analytical method can be applied to different countries and regions.