Life cycle thinking and machine learning for urban metabolism assessment and prediction

The real-world urban systems represent nonlinear, dynamical, and interconnected urban processes that require better management of their complexity. Thereby, we need to understand, measure, and assess the structure and functioning of the urban processes. We propose an innovative and novel evidence-based methodology to manage the complexity of urban processes, that can enhance their resilience as part of the concept of smart and regenerative urban metabolism with the overarching intention to better achieve sustainability. We couple Life Cycle Thinking and Machine Learning to measure and assess the metabolic processes of the urban core of Lisbon's functional urban area using multidimensional indicators and measures incorporating urban ecosystem services dynamics. We built and trained a multilayer perceptron (MLP) network to identify the metabolic drivers and predict the metabolic changes for the near future (2025). The prediction model's performance was validated using the standard deviations of the prediction errors of the data subsets and the network's training graph. The simulated results show that the urban processes related to employment and unemployment rates (17%), energy systems (10%), sewage and waste management/treatment/recycling, demography & migration, hard/soft cultural assets, and air pollution (7%), education and training, welfare, cultural participation, and habitat ecosystems (5%), urban safety, water systems, economy, housing quality, urban void, urban fabric, and health services and infrastructure (2%), consists the salient drivers for the urban metabolic changes. The proposed research framework acts as a knowledge-based tool to support effective urban metabolism policies ensuring sustainable and resilient urban development.