City planners who want to understand urban emissions have traditionally worked from static inventories and sparse monitoring. Cornell engineers have demonstrated an alternative: a Sustainable Urban Digital Twin, a near-real-time virtual model of urban carbon dioxide piloted in Manhattan. The framework, described in Environmental Modelling and Software, lets officials track conditions, locate hotspots and trial interventions virtually before committing to them on the street.
The prototype is organised in four modular layers. A physical layer gathers data from several large databases, including city sensors recording carbon dioxide concentration, temperature and humidity every five minutes. A digital layer organises and models those feeds, a brain layer applies Bayesian modelling and machine learning to make predictions with quantified uncertainty, and a service layer turns the results into maps, visualisations and suggested actions for decision-makers.
The team, led by H. Oliver Gao of Cornell's Systems Engineering Program and the Center for Transportation, Environment, and Community Health, stresses that the platform is a research prototype: it handles carbon dioxide but not yet particulate matter, nitrogen oxides or ozone. The next steps are to add models for those pollutants, build AI-assisted decision support and simplified community dashboards, and test the approach in other cities with local partners.