Scalable multi-pollution exposure assessment using routine mobile monitoring platforms

This New Investigator Award study will measure air pollutants in intensive campaigns with Google Street View cars in Oakland and Delhi and compare exposure estimates to conventional methods.

Poster  abstract for HEI Annual Conference 2023

Scalable Multi-Pollutant Exposure Assessment Using Routine Mobile Monitoring Platforms

Joshua S. Apte

University of California, Berkeley, CA, USA

Background and   Objectives. The absence of spatially resolved air pollution measurements remains a major gap for health studies of air pollution, especially in disadvantaged communities in the United States and in lower-income countries. Many urban air pollutants vary over short spatial scales. Primary air pollutants from traffic, such as black carbon (BC), nitrogen oxides (NO + NO₂ = NO_x), and ultrafine particles (UFP), have especially sharp spatial gradients. In the U.S. context, these gradients have major consequences for environmental justice. Conventional fixed-site pollution monitoring methods lack the spatial resolution needed to characterize these heterogeneous human exposures and localized pollution hotspots. I report on my HEI Rosenblith Award project, where we conducted a series of validation and scaling investigations to explore the potential for mobile air quality measurements to provide high-resolution pollution exposure estimates.

Methods. The core approach of the study revolved around repeated mobile monitoring to develop time-stable estimates of central-tendency air pollution exposures at sub-city-block spatial resolution. In the San Francisco Bay Area, we collected ~ 3y of data using two Google Street View cars that were custom equipped to measure NO, NO₂, BC, and UFP.  We supplemented mobile measurements with both regulatory-fixed monitoring and, during a 100-day experiment, an exceptionally dense grid of \100 stationary black carbon monitors. Using these complementary approaches, we compared patterns and insights between mobile and fixed-site monitoring. We then scaled up the multi-pollutant mobile monitoring approach to 13 different neighborhoods with nearly 500,000 inhabitants, evaluated how the within- and between-neighborhood heterogeneity in concentrations affected population exposure and environmental disparities, and compared our insights with those from a widely used empirical exposure model. Next, we evaluated the advantages and tradeoffs for coupling mobile monitoring with statistical land-use-regression models to estimate intraurban variation in air pollution in a data-efficient manner. Finally, we reproduced our mobile monitoring approach in a pilot study in Bangalore, India.

Results. We found a moderate to high concordance in the time-averaged spatial patterns between mobile monitoring and dense fixed-site observations of black carbon, and observed seasonal multi-pollutant dynamics that were consistent among with regulatory monitors, enhancing our confidence in the mobile data. After scaling up measurements, we found fine-scale spatial variation was responsible for a large fraction of the overall heterogeneity in population exposures, and led to substantial racial/ethnic inequality in air pollution exposures across the Bay Area. To reduce sampling effort, we demonstrated that empirical modeling via land-use regression (LUR) could dramatically reduce data requirements for building high resolution air quality maps, at the cost of a model precision. Finally, our pilot in Bangalore showed preliminary evidence that our mapping approach was also feasible in an Indian city.

Conclusion. We demonstrated that mobile monitoring can produce insights about air pollution exposure that are externally validated against multiple other analysis approaches, while adding complementary information about spatial patterns and exposure heterogeneity and inequity that is not readily obtained with other methods.