Traffic Emissions and Air Quality Variations in Urban Corridors: A Spatial Analysis of Bengaluru City

Abstract

Aims: This study aims to examine the spatial and seasonal variability of urban air pollution across major traffic corridors in Bengaluru by integrating traffic, meteorological, and air quality data. It also evaluates the performance of machine learning models for predicting PM2.5 concentrations and identifies the key factors influencing air pollution. Study Design: Quantitative observational study based on environmental data analysis and machine learning. Place and Duration of Study: The study was conducted across seven major traffic corridors in Bengaluru, India, using traffic, meteorological, and air quality data collected over different seasons. Methodology: Traffic, air quality, and meteorological datasets from seven traffic corridors were integrated into a unified dataset. The data were preprocessed and analyzed to evaluate spatial and seasonal pollution patterns. Linear Regression, Random Forest, Extreme Gradient Boosting (XGBoost), and Artificial Neural Network (ANN) models were developed to predict PM2.5 concentrations. Model performance was evaluated using the coefficient of determination (R²). Results: The analysis revealed considerable spatial variation in PM2.5 concentrations across the selected corridors. Mysore Road and Tumakuru Road recorded the highest pollution levels (>90–100 µg/m³), whereas Jayadeva Junction showed comparatively lower concentrations. Seasonal analysis indicated lower pollution during the monsoon due to rainfall and higher pollution during the post-monsoon and winter seasons because of reduced atmospheric dispersion. Among the evaluated models, Linear Regression achieved the highest predictive performance (R² = 0.45), followed closely by ANN (R² = 0.44). Feature importance analysis indicated that traffic congestion contributed more to PM₂.₅ concentration prediction than the evaluated meteorological variables. Conclusion: The proposed framework demonstrates the feasibility of predicting PM₂.₅ concentrations using integrated traffic and meteorological data and supports evidence-based traffic management and air quality control strategies for sustainable urban development.