Ecology Guided Optimization of Block Scale Urban Morphology for Low Carbon Renewal: A Machine Learning and Multi Objective Framework
We present an ecology guided, block scale framework that links urban morphology (UM) to carbon emissions (CE) mitigation through explicit ecological strategies. In Wuhan, China, we delineate 708 blocks by road segmentation and zoning, then build 17 indicators across planar and vertical morphology, blue green ratios, sky view factor (SVF), and activity intensity. We benchmark common state of the art machine learning models with spatial cross validation. XGBoost performs best (R² 0.85) and GRF reveals spatial nonstationarity. Key drivers are building area density (BAD), floor area ratio (FAR) and average nighttime lights (ANL), while higher water and green ratios and SVF mitigate CE. We embed the XGBoost predictor in NSGA-II to jointly optimize three objectives under zoning integrity, ecological red lines, and blue green infrastructure (BGI) protection: minimize CE, reduce morphological adjustment cost, and lessen inter block inequality. Pareto solutions cut CE by 15%, shrink high emission clusters by 21%, and reduce inequality by 18%. Ecological levers include moderate cuts to FAR and density in core hotspots, expansion of BGI by 3–6% in the periphery, SVF above 0.65, and improved street connectivity. Keywords: Urban Morphology; Low Carbon Renewal; Ecological Strategy; Machine Learning; Multi-objective optimization.