Beyond spatiotemporal modeling: a review of applications of machine learning for traffic-related air pollution toward non-exhaust emissions

Traffic-related air pollution (TRAP), increasingly shaped by non-exhaust emissions, remains a major urban health concern. This review provides a structured synthesis of over 50 studies (2020–2024) applying machine learning (ML) to TRAP, focusing on spatial modeling, contributing factor identification, non-exhaust emission characterization, and source apportionment. Key challenges include data sparsity, inconsistent features, and limited interpretability. Advancing ML integration and transparency is essential for improving exposure assessment and environmental health.