Buses are an important component of public transportation but are also major contributors to urban noise pollution, particularly conventional internal combustion engine models. While electric buses offer quieter alternatives, limited research exists on how operational conditions influence noise levels across both bus types. This study investigates the impact of multiple operational factors including, bus age, average speed, trip distance, occupancy levels, road conditions, time and day of bus operation, and road network, on noise emissions in conventional and electric buses. Using data from 31 conventional and 12 electric buses (1465 trips total), noise levels were recorded via validated smartphone-based sound meters and analyzed using interpretable machine learning models, specifically Extra Gradient Boosting (XGBoost) and Shapley Additive Explanations (SHAP). The results show that electric buses are consistently quieter and less acoustically sensitive to operational variations, whereas conventional buses exhibit greater noise variability, especially at higher average speeds and bus ages. This study offers novel insights by approaching bus noise generation from a multi-operational factor approach using interpretable machine learning models. Findings support targeted operational strategies such as speed regulation, fleet renewal based on bus age, and schedule optimization as effective noise-mitigation tools. Future studies can focus on adding more operational factors, and perform inter-city analysis to improve knowledge in the domain of bus noise generation.
Davies et al. (Tue,) studied this question.