Multiscale Spatial Associations Between the Built Environment and Urban Vitality: Evidence from Changchun, China

Urban vitality matters for urban regeneration, yet built-environment–vitality associations vary spatially and by variable. We examined 2621 regular 500 m grid cells across Changchun’s built-up area. PCA combined 2024 population, 2024 nighttime light, and 2018 mean building height into a composite Urban Vitality index. Five built-environment indicators—Functional Density, POI Diversity, Mean NDVI, Transit Proximity, and Commercial Proximity—were analyzed using Global Moran’s I, ordinary least squares (OLS), geographically weighted regression (GWR), and multiscale geographically weighted regression (MGWR). PC1 explained 82.15% of the variance, and Urban Vitality showed strong spatial autocorrelation (Moran’s I = 0.8332). Model fit increased from OLS (R2 = 0.7767) to GWR (R2 = 0.9225) and MGWR (R2 = 0.9296). MGWR bandwidths were localized for Functional Density (61), POI Diversity (62), and Mean NDVI (71), but broader for Transit Proximity (2548) and Commercial Proximity (2619). Adjusted local inference identified significant positive associations for Functional Density, POI Diversity, and Transit Proximity, significant negative associations for Mean NDVI, and no significant local association for Commercial Proximity. Although sensitivity analyses generally preserved fit and median directions, excluding building height changed the Transit Proximity bandwidth from 2548 to 70 neighbors, indicating that its estimated association scale was sensitive to vitality-index specification. Commercial Proximity also remained sensitive to indicator operationalization. The findings identify where local diagnostic follow-up is warranted, while intervention effects remain outside the scope of the analysis.