How does the Spatial Distribution of Three-Dimensional Green Volume Evolve in Urbanization? Evidence from Nanjing, China

Abstract

Conventional assessments of urban green space usually rely on planar indicators, which cannot fully capture changes in vegetation height, canopy volume, and layered structure during urbanization. To address this limitation, this study estimated three-dimensional green volume (3DGV) and compared its spatiotemporal evolution with two-dimensional green space (2DGS) in the main city of Nanjing from 1990 to 2020. Field-based reference samples, Landsat-derived vegetation indices, and machine-learning regression were integrated to reconstruct decadal-scale 3DGV, while 2DGS was extracted from high-resolution remote-sensing images using support vector machine classification. Spatial autocorrelation analysis and spatial regression models were then used to examine temporal changes, clustering patterns, and associations with socioeconomic, land-use, and natural-environment variables. The results show that both 3DGV and 2DGS experienced a decline–recovery trajectory and exhibited significant positive spatial autocorrelation, but their recovery magnitudes and local clustering patterns differed across phases. The two indicators shared several associated variables, yet differed in coefficient direction, statistical significance, and phase-specific patterns. Land-use variables were more closely associated with changes in 2DGS, whereas socioeconomic and natural-environment variables showed more frequent or indicator-specific associations with 3DGV in several phases. These findings indicate that 3DGV adds a structural dimension to planar green-space assessment by revealing spatially uneven changes in vegetation volume, canopy development, and layered planting structure. The combined 2DGS–3DGV framework provides an additional and complementary basis for diagnosing urban green-space change and supporting structure-oriented greening strategies in high-density cities.