Spatiotemporal differentiation and associated factors of modern and contemporary garden heritage in Fuzhou based on XGBoost-SHAP

Within the global heritage conservation framework, analyzing spatiotemporal evolution and associated factors of urban heritage is significant for its protection. This study focuses on modern and contemporary garden heritage in Fuzhou, encompassing gardens built or renovated during 1840–1990 and pre-1840 gardens transformed in 1840–1990. Integrating spatial analysis with XGBoost-SHAP machine learning across four phases (1840–1911, 1912–1949, 1949–1990, 1991–2025), we reveal spatiotemporal differentiation patterns, nonlinear factor associations, and resilience-based conservation pathways. Results demonstrate Fuzhou’s garden heritage exhibits “single-core clustering with axial extension” evolution. The heritage centroid migration trajectory from northwest to southeast corresponds to the city’s “crossing the river, expanding eastward” development, with extant gardens forming a “one core, one belt” configuration. Cultural factors show the highest explanatory power with approximately linear positive association, while road network density Ո GDP demonstrate the strongest synergistic interaction. This study provides methodological and empirical references for garden heritage conservation in high-density urban areas.