Identifying priority urban ecological restoration areas using an integrated machine-learning and multi-dimensional...

Rapid urbanization has intensified anthropogenic pressure on urban ecosystems, underscoring the need for robust methods to identify priority areas for ecological restoration. To address the disconnect between static assessments and spatially heterogeneous ecological pressures, and the separation of structural optimization from functional recovery in previous studies, we develop an integrated framework that combines machine-learning-based resistance modeling, identification of ecological units, and five-dimensional ecological diagnosis to support restoration prioritization and typology classification. Applied in Shanghai, we identify key ecological restoration areas exhibiting a pronounced core-periphery pattern along urban, suburban, and exurban gradients, but do not simply coincide with areas of lowest ecosystem services or highest resistance. Instead, they are systematically concentrated in transitional zones where ecosystem service provision remains important while ecological resistance is relatively high. The five-dimensional diagnostic framework classifies all ecological units into four restoration types: quality improvement, connectivity enhancement, stability enhancement, and ecosystem function improvement. High and very high priority units account for 44% of ecological space, while 26.1% are classified as medium priority, providing a structured basis for prioritizing intervention intensity and sequencing restoration actions. The proposed framework provides a mechanism-driven, spatially explicit, and transferable tool for guiding differentiated ecological restoration strategies in high-density urban regions.