Nonlinear Transitions in Urban Sustainability: A Hybrid Assessment of Resilience Tipping Points Using Emergy, GIS, Carbon Metrics, and Neural Networks

Facing the intensifying challenges of global climate change, research on urban ecosystem sustainability and resilience has become increasingly urgent. This study constructs a hybrid assessment framework integrating emergy analysis, GIS, carbon accounting, and neural networks to systematically evaluate the sustainability evolution of Zhenjiang City from 2000 to 2020 and project its resilience trajectory to 2050. The results show that Zhenjiang experienced a resilience tipping point around 2015. From 2000 to 2015, the built-up area expanded by 17.5 times, droving PM 2.5 concentration to exceed 200 μg/m 3 and the emergy sustainability index to decline by 59%. From 2015 to 2020, under strong policy interventions, PM 2.5 dropped sharply by 85.6%, forest area expanded by 3.5 times, and wetland emerged from none, with an accumulation–release asymmetry index greater than 2, validating the characteristics of slow accumulation and rapid recovery. Rainwater emergy accounts for more than 94% of total emergy, making it the dominant factor in the sustainability of small and medium-sized cities. Built-up areas have achieved relative carbon–emergy decoupling, while farmland faces a high carbon emission dilemma. SHAP analysis shows that the contribution of spatial ecological indicators is 3.5 times that of population density, and policy benefits exhibit a time lag of approximately 6 years. Neural network projections indicate that the comprehensive sustainability score can improve by approximately 58% by 2050, though this improvement depends on the continuous optimization of ecological space. This study provides a quantifiable assessment tool and decision-making basis for resilience management in small and medium-sized cities.