Machine learning unveils temperature and mobility as critical predictors of urban crime in New York City (2008–2022): insights for sustainable urban systems

Understanding the factors shaping urban crime is crucial for building safe and sustainable cities. Yet quantifying their impacts remains challenging due to the complex interplay of socio-environmental influences. We use machine learning to identify key contributors to crime rates in New York City. We identify temperature and mobility as the most consistent and dominant predictors at both citywide and borough levels. Warmer temperatures, increased public transportation usage, and reduced taxi usage show the strongest positive associations with crime, controlling for precipitation, COVID−19 mortality, and noise complaints, which contribute modestly. Socio-economic and demographic variables indicate more spatially heterogeneous impacts. These findings suggest a layered theoretical framework that integrates General Aggression Theory, Routine Activity Theory, and structural vulnerability to explain crime patterns. These co-occurring mechanisms underscore the need for multifaceted interventions. We propose three actionable pathways for sustainable urban systems: 1) Climate-responsive infrastructure: Prioritize heat mitigation (context-sensitive greening initiatives, cooling corridors). 2) Mobility-centered safety planning: Embed crime-deterrent features into transit hubs. 3) Community-driven environmental governance: Address hyperlocal risks through noise-sensitive zoning and equity-focused placemaking in racially segregated neighborhoods. This study advances urban crime research by linking theory and data-driven modeling to inform climate-resilient, mobility-aware urban safety planning.