OSM land use data quality and its association with city-level socioeconomic conditions in China

Abstract

OpenStreetMap (OSM) can serve as an alternative to map the built environment in data-sparse areas such as China. However, the quality of OSM land use data and its relationship with city-level socioeconomic conditions (e.g., income level) remain understudied. Using a sample of 332 Chinese cities and the time-series China Urban Land Use Mapping dataset as the reference, we investigated inter- and intra-city variations in the completeness and accuracy of OSM land use data as of the year 2022, and their associations with city-level socioeconomic conditions. Across the 332 cities, we found that the median completeness and overall accuracy of citywide OSM land use data was 21.90% and 79.57%, respectively. After adjusting for population density and province-fixed effects, a one-standard deviation (1-SD) increase in GDP per capita, education level, vehicle ownership, and urbanization rate were associated with 2.54%, 2.48%, 1.92%, and 2.48% higher completeness of OSM land use data in a given city. However, a 1-SD increase in vehicle ownership was associated with 0.97% lower overall accuracy. The accuracy of some land uses in OSM was positively associated with city-level socioeconomic conditions. These findings highlight the complexity of land use mapping and potential limitations of using OSM land use data in Chinese cities.