Professor Songhua Hu’s Team, in Collaboration with the Massachusetts Institute of Technology, Publishes Urban Big Data Research in Nature Sustainability

This study presents the first city-scale microscopic traffic emission model built entirely on urban crowdsourced data, including traffic camera feeds, mobile phone location data, and floating car data. By integrating multiple artificial intelligence techniques, the framework systematically extracts key elements such as fine-grained vehicle-type traffic flows, traffic signal control states, multimodal origin–destination (OD) patterns, and dynamic link-level speeds. These heterogeneous data sources are aligned within a unified simulation–optimization framework, enabling high-resolution reconstruction of vehicle operations and emission levels across the entire urban network. Building upon this foundation, the study further conducts simulations of various representative transportation policies and unexpected events, demonstrating the framework’s strong inference capability and near real-time responsiveness in complex urban systems. For example, in analyzing the congestion pricing policy in Manhattan, New York, the model estimates a reduction of approximately 16–22% in traffic-related emissions, highlighting the critical value of fine-grained, data-driven approaches for policy evaluation.

The study is titled Ubiquitous data-driven framework for traffic emission estimation and policy evaluation. It has been featured by MIT News and selected for the Nature Sustainability Research Briefing, which highlights only around ten representative studies worldwide each year.