Estimating Black Carbon Concentration from Urban Traffic Using Vision-Based Machine Learning

TL;DR

The paper tackles the lack of street-level black carbon data by leveraging ubiquitous traffic video and weather information to predict BC concentrations near roadways. It introduces a vision-based pipeline that extracts vehicle covariates via CV methods and fuses them with environmental data, training an XGBoost regressor that achieves an R^2 of 0.72 and RMSE of 129.42 ng/m^3 on Downtown Toronto data. Key findings highlight the predictive importance of near-source vehicle behaviors, especially stop-and-go dynamics and proximity to the measurement site, as well as wind speed in determining BC dispersion. The approach demonstrates the feasibility of using existing urban infrastructure to generate actionable BC estimates for pollution reduction, urban planning, and environmental justice initiatives, while acknowledging the need for longer-term, seasonally diverse data collections.

Abstract

Black carbon (BC) emissions in urban areas are primarily driven by traffic, with hotspots near major roads disproportionately affecting marginalized communities. Because BC monitoring is typically performed using costly and specialized instruments. there is little to no available data on BC from local traffic sources that could help inform policy interventions targeting local factors. By contrast, traffic monitoring systems are widely deployed in cities around the world, highlighting the imbalance between what we know about traffic conditions and what do not know about their environmental consequences. To bridge this gap, we propose a machine learning-driven system that extracts visual information from traffic video to capture vehicles behaviors and conditions. Combining these features with weather data, our model estimates BC at street level, achieving an R-squared value of 0.72 and RMSE of 129.42 ng/m3 (nanogram per cubic meter). From a sustainability perspective, this work leverages resources already supported by urban infrastructure and established modeling techniques to generate information relevant to traffic emission. Obtaining BC concentration data provides actionable insights to support pollution reduction, urban planning, public health, and environmental justice at the local municipal level.

Figures (22)

• Figure 1: End-to-end system overview for BC estimation with computer vision and machine learning techniques.
• Figure 2: Result of applying YOLO and Hough Transform to determine the type of vehicle and distance from the microaethalometer.
• Figure 3: Cosine similarity between Fourier transformed BC vector and Fourier transformed TotalVehicle vector. The corrected BC signal is determined by the shift that maximizes cosine similarity (160 seconds).
• Figure 4: Original BC signal (gray) plotted alongside vehicle counts (dotted line) reveals a temporal misalignment. The corrected BC signal (black), shifted by 160 seconds, shows alignment with vehicle activity.
• Figure 5: Correlation matrix. We excluded highly correlated features above 0.70.