Driving Behaviour as a Missing Control Lever in Urban NOx Mitigation: A Network-Level Digital Twin of Spatial–Temporal Hotspot Migration

Air pollution remains a significant global issue. Understanding the drivers of air pollution and reducing associated premature deaths have become one of the United Nations’ Sustainable Development Goals. In recent years, traffic emissions have become the main source of urban air pollution in the UK and globally. In the EU, road transport has been identified as the primary source of total NO_x (comprising NO₂ and NO), contributing 39% of total emissions. In densely populated regions, large-scale events and daily congestion further aggravated the problems. Although traffic flow and temporal character are proven to play an important part in the high-resolution NO_x prediction, a systematic component of NO_x variability remains unexplained. Thus, current policies based solely on traffic flow control and fleet turnover can hardly reduce the whole network emission effectively in real-world. The observed persistence and spatial coherence of the residual errors also suggest that missing information should not be identified as random noise, but comes from the network dynamics at the microscale driven by individual driving behaviour.

Our study addresses this gap by examining network-level driving behaviour as a candidate mechanism underlying this unexplained variability, and quantifying its impact on urban NO_x emissions. By developing a driving style–based urban digital twin emission simulation framework, we investigate how behavioural heterogeneity influences the spatial and temporal distribution of NO_x emissions across the Greater Manchester road network. The empirical driving behaviour parameters were localized from 16,897,293 records. Apart from that, realistic hourly traffic flows from Department for Transport (DfT) were employed as constraints in the HBEFA-linked emission modelling within SUMO for spatial analysis based on a ring-based segmentation. Results show that behavioural heterogeneity not only affects the magnitude of NO_x emissions, but also the spatial distribution.Contrary to the widely held belief that aggressive driving uniformly increases emissions across the network, we find that aggressive driving reshapes emission patterns by relocating hotspots rather than simply amplifying network-wide totals.

Therefore, emission mitigation benefits are highly context-dependent: policies promoting smoother driving are likely to be most effective in suburban and inter-ring corridors, while targeted restrictions on aggressive driving may be necessary in high-density urban cores during vulnerable periods. Moreover, behavioural interventions aligned with travel demand patterns may outperform static, network-wide emission policies. By constructing a scalable network-level digital twin, this study establishes driving style as a controllable and policy-relevant parameter. The developed framework supports scenario testing, spatial sensitivity evaluation, and behavioural–emission inference at the network scale, contributing to more effective and spatially equitable transport emission management.