Answers are grounded in this paper and cite their sources.
Demonstrates improved accuracy in identifying urban agricultural land in Nanjing, indicating better urban planning outcomes.
Urban agriculture (UA) provides ecological, social, and economic co-benefits in compact cities, but accurately identifying UA land remains a challenge due to small patch sizes, fragmented patterns, and mixing with other urban land covers. We develop a multi-modal geospatial-imagery fusion framework that integrates high-resolution satellite imagery with multi-scale geospatial indicators, including points of interest (POIs), population, transportation, hydrology, and nighttime light data to improve UA land identification. Two imagery-only baselines are extended into fusion variants via skip-level and mid-layer integration. Using Nanjing, China, as the testbed, our experiments show that fusion significantly improves overall accuracy (OA) and mean Intersection-over-Union (mIoU) relative to imagery-only baselines. Channel-wise zero-out ablation further reveals the importance of transport-related POIs (especially at 1000 m), large-scale population distribution (1500 m), and neighborhood-scale public services (500 m). Applying the optimized fusion model to a 50 km strip across the urban-suburban-rural gradient uncovers a spatial transition: from fragmented, embedded UA patches in the core city to increasingly continuous and stable belts toward rural peripheries. The approach enhances both accuracy and interpretability, providing actionable evidence for UA spatial optimization in territorial planning and urban renewal.
No takes yet. Share an insight, caveat, or question.
黄彩琼 et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: