Unveiling spatial heterogeneity and nonlinearity of OpenStreetMap building completeness in China: an assessment using authoritative benchmarks

While OpenStreetMap (OSM) serves as a critical infrastructure for urban management and emergency response, the completeness of its building data remains spatially uneven. The current research lacks nationwide empirical evaluations grounded in authoritative, high-precision reference data, particularly regarding how socioeconomic factors shape completeness through complex nonlinear and spatially nonstationary mechanisms. To address this gap, this study utilized the 2023 MAPWORLD dataset China’s official authoritative geographic information source—as a benchmark to conduct a high-precision evaluation of OSM building completeness across 320 Chinese cities. In this study, an integrated GWR-SHAP analytical framework is developed. Geographically Weighted Regression (GWR) identifies spatial non-stationarity by mapping the spatial variation in local regression coefficients, whereas the Shapley Additive Explanations (SHAP) method captures complex nonlinear relationships and threshold effects that are often overlooked in traditional linear spatial models. The results indicated a markedly low overall completeness of 1.425%, characterized by pronounced spatial disparities. Volunteer activity shows the strongest association with completeness, exhibiting stage-dependent spatial heterogeneity: positive effects are substantial in the western and central regions (“completion effects”) but diminish in eastern coastal areas (“saturation effects”). Furthermore, factors such as economic development and innovation capacity demonstrate significant threshold effects, delineating three distinct urban agglomeration regimes: the “mature” Yangtze River Delta, the “core–periphery” structures of the Chengdu–Chongqing and Middle Yangtze regions, and the “mixed” Central Plains. These findings advance the theoretical understanding of the socio-spatial mechanisms underlying Volunteered Geographic Information (VGI) evolution and provide precise empirical evidence for formulating differentiated crowdsourcing incentives and collaborative data improvement strategies.