Mobility-driven estimate reveals elevated air pollution exposure and socioeconomic disparities beyond residence-based approaches in Boston

Nail F. Bashan, Yang Zhang, Michelle L. Bell, Qi WangJournal of Exposure Science & Environmental Epidemiology 2025 · DOI

TL;DR

This study reveals that traditional air pollution exposure assessments, which rely solely on home addresses, significantly underestimate actual exposure levels and mask crucial socioeconomic disparities. By integrating detailed human mobility data with hyperlocal air quality monitoring in Boston, researchers found that people experience higher pollution exposure when they are out and about, especially during peak travel times, and these elevated exposures are not evenly distributed across demographic groups.

What Problem Does This Paper Address?

Accurately assessing individual exposure to air pollution is critical for understanding its health impacts and designing effective public health interventions. However, traditional methods often rely on static residence-based estimates, assuming people spend most of their time at home. This approach overlooks the significant amount of time individuals spend in other locations (e.g., workplaces, transit, recreational areas) where air pollution levels can differ substantially. This is particularly problematic in urban environments with high spatiotemporal variability in air quality, driven by traffic, industrial activity, and complex urban topography. Consequently, residence-based estimates can lead to misclassification of exposure, potentially masking or underestimating the true burden of air pollution, especially among vulnerable populations and exacerbating existing environmental injustices. This paper aims to quantify this underestimation and reveal how mobility patterns contribute to socioeconomic disparities in air pollution exposure.

Methods

The researchers employed a novel approach that integrates two key data streams:

  • High-resolution Human Mobility Data: Data was collected from 990 participants in the BostonWalks study using a smartphone GPS application (Catch-My-Day). This provided approximately 155,000 trip records and stay locations over a seven-month period. The app automatically recorded trips, modes of transportation, and stay locations, with participants also providing detailed sociodemographic information.
  • Hyperlocal Air Quality Data: PM2.5 measurements were obtained from a network of 294 publicly accessible PurpleAir sensors, collecting data at 2-minute intervals. These sensors were calibrated and cleaned to ensure data accuracy.

The core of their methodology involved:

  1. Mobility-Informed Exposure Calculation: For each participant's "stay location" (defined as a location where they remained for at least 30 minutes), they calculated a daily adjusted exposure. This involved averaging PM2.5 readings from PurpleAir sensors within a 4 km buffer around the stay location and then subtracting the region's daily average PM2.5 to normalize for day-to-day fluctuations. This "relative exposure metric" highlights pollution levels at a specific location compared to the regional baseline.
  2. Comparison with Residence-Based Estimates: They compared these mobility-informed exposures to traditional home-based estimates. This was done by calculating the average PM2.5 within a 4 km radius of the participant's home location for the same time intervals. The difference between stay location exposure and home location exposure indicated whether traveling away from home led to higher or lower pollution.
  3. Statistical Analysis: They used spatial autocorrelation (Moran's I) to assess the clustering of mobility exposures. Weighted Least Squares (WLS) regressions were used to evaluate differences in exposure across various sociodemographic groups (race, income, education, age, occupation), correcting for potential sample biases. T-tests were used to compare exposure differences between weekdays and weekends, and across different trip purposes.

Datasets

  • BostonWalks Study Data: This dataset comprises approximately 155,000 trip records and associated stay locations from 990 participants in the Boston metropolitan area, collected between June 1 and December 31, 2023. Participants were recruited to ensure balanced representation across age, gender, income, race, ethnicity, and education.
  • PurpleAir Sensor Data: PM2.5 measurements from 294 calibrated PurpleAir sensors across the Boston metropolitan area, collected at 2-minute intervals during the same June-December 2023 period.

Key Insights and Ideas

  • Mobility is a Game-Changer for Exposure Assessment: The study unequivocally demonstrates that ignoring human mobility leads to a significant underestimation of actual air pollution exposure. People are exposed to more pollution when they are away from home, not less.
  • Disparities are Hidden in Static Data: Residence-based approaches mask important environmental injustices. The elevated exposures experienced by certain groups when mobile are completely obscured if only home locations are considered.
  • Peak Travel Times are High-Risk Zones: Weekday peak travel hours, coinciding with increased traffic and commuting, are particularly critical periods for elevated exposure, especially for employed individuals and those with higher incomes.
  • Urban Design Matters: The downtown-centric design of cities like Boston, which necessitates longer commutes, contributes to concentrated pollution exposure in specific areas and during specific times.
  • Sociodemographic Factors Amplify Disparities: The study highlights how race, income, and education intersect with mobility patterns to create unequal exposure burdens. For instance, employed individuals and those with higher education experience the most significant discrepancies between mobility-informed and residence-based estimates.

Key Results

  • Underestimation of Exposure: Mobility-informed exposures averaged 0.10 µg/m³ higher than residence-based estimates on weekdays, with this difference reaching up to 0.45 µg/m³ on high-pollution days.
  • Peak Hour Exposure: Employed and higher-income individuals experienced significantly elevated exposures during peak travel hours (up to +0.30 µg/m³).
  • Spatial Clustering: Mobility exposures showed stronger spatial clustering on weekdays (Moran's I = 0.4) compared to weekends (I = 0.2), indicating more predictable and concentrated exposure patterns during the work week.
  • Sociodemographic Differences:
    • Employed individuals showed the largest difference, with residence-based estimates averaging 0.12 µg/m³ lower than mobility-based exposure.
  • Individuals with higher education levels experienced elevated exposures during peak hours.
  • On weekday mornings, White individuals experienced higher exposures than Black/African-American and Asian/Asian-American groups.
  • Trip Purpose Matters: Leisure, dining, and "other" trip purposes showed significant weekday vs. weekend exposure differences, with weekdays being higher, unlike work-related trips which showed less variability.

Limitations and Future Work

  • Indoor vs. Outdoor: The study relied on outdoor air quality measurements as a proxy for all stays, not accounting for the variability of indoor air quality, which can differ significantly.
  • Seasonal Variation: The data was collected from June to December 2023, potentially missing important seasonal variations in mobility and pollution levels during winter and spring.
  • Low-Cost Sensor Limitations: While calibrated, PurpleAir sensors may still introduce systematic biases compared to regulatory-grade monitors.
  • Spatial Buffering: The 4 km buffer, while enhancing neighborhood representativeness, might smooth over hyperlocal pollution gradients that could influence individual exposure.

Future work should aim to:

  • Integrate indoor/outdoor pollution infiltration ratios for more precise individual exposure assessment.
  • Conduct travel surveys that capture periods of fluctuating pollution and shifting travel behavior across all seasons.
  • Explore these patterns in diverse cities with different urban designs.
  • Link these refined, mobility-based exposure estimates with health outcome data to better quantify public health and economic impacts.

Summary updated

Abstract

Integrating dynamic mobility data with hyperlocal air quality monitoring provides a refined framework for accurate exposure assessment, informing equitable public health policies and targeted interventions.
BibTeX
@article{bashan2025mobility,
  author = {Nail F. Bashan and Yang Zhang and Michelle L. Bell and Qi Wang},
  title = {{Mobility-driven estimate reveals elevated air pollution exposure and socioeconomic disparities beyond residence-based approaches in Boston}},
  journal = {Journal of Exposure Science \& Environmental Epidemiology},
  year = {2025},
  doi = {10.1038/s41370-025-00820-z},
}
Figure