Abstract

Urban green spaces (UGS) are increasingly planned as connected networks, yet many connectivity assessments reduce cost variation to a static threshold and provide limited guidance on where to intervene. We develop a percolation-based framework for translating dynamic UGS connectivity into ecological management priorities. The framework constructs a least-cost, cost-weighted UGS network, activates edges in ascending cost-weighted distance to identify critical transitions and merger hierarchies, and converts the full percolation history into management zones, a three-level edge hierarchy, and a four-type node typology. Applied to Nanjing, China, the analysis shows that system-level connectivity emerges within a narrow high-cost critical band triggered by a small set of cross-barrier links. The merger history reveals three persistent spatial subsystems, which are further resolved into five management zones. Robustness experiments indicate that protecting map-prioritized edges and nodes slows the decline of weighted global efficiency under the tested removal rules. The framework links process-based connectivity diagnosis with spatially explicit planning actions, including corridor safeguarding, within-zone reinforcement, and stepping-stone protection, while remaining transferable with local calibration.