You are here

Scalable multi-pollution exposure assessment using routine mobile monitoring platforms

Principal Investigator: 

University of Texas–Austin

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.

Funded under

Abstract for the 2019 HEI Annual Conference

Mapping Air Pollution at High Spatial Resolution: Comparing Mobile and Fixed Sensing Approaches

Sarah E. Chambliss and Joshua S. Apte

1University of Texas at Austin, Austin, TX, USA

Background. Urban air pollution concentrations can vary sharply over short distances. High spatial resolution surfaces of urban air quality data are needed for a variety of purposes, including for exposure assessment in health studies. Two recent advances in measurement technology enable high resolution pollution mapping: on one hand, lower-cost sensors suitable for fixed deployment, and on the other, mobile sampling platforms suitable for extensive deployment on city streets.

Objectives. This poster utilizes measurements from an unusually rich field experiment to compare and contrast the insights about fine-scale air pollution variability that can be gained from sampling using mobile and fixed sampling techniques. We investigate the following research questions: (i) can repeated mobile air sampling reproduce the same time-averaged spatial patterns that are observed by fixed-site observations, and (ii) can intermittent mobile sampling recover similar time-resolved pollutant concentration data as fixed-site monitoring?

Study Design. For 100 days in summer 2017, a dense network of 100 custom-built black carbon (BC) sensors (ABCD, Aerosol Black Carbon Detector) was deployed in West Oakland, California. The ABCD sensor network provided BC information at 1 minute time resolution for each site. In parallel, we repeatedly sampled on-road air quality in the same neighborhood using two Google Street View cars custom-equipped with photoacoustic extinctiometer instrumentation for measuring BC at 1 Hz time resolution. The normal sampling program consisted of repeatedly driving every street in the ~5 km2 domain, resulting in ~300 daytime hours of on-road BC measurements. Approximately 2% of the time-resolved in-motion BC measurements were obtained within 25 m of one of the 100 ABCD sampling sites. On average, the Google cars passed each ABCD monitor ~33 times over the course of the campaign, accumulating an average of 6 seconds of measurements while passing each monitor. In addition, we parked the Google cars within close proximity to 30 of the ABCD monitors for 10-15 minute sessions to create stationary “colocation” events.

Results. The overall spatial patterns detected by both the mobile and fixed sampling approaches aligned well. We observed no meaningful systematic difference in BC concentration between the fixed ABCD sensors and the in-motion mobile measurements. The average on-road concentration collected within 25 m of each ABCD monitor was moderately well correlated with the daytime average concentration reported by the BC sensor, with an R2 of 0.57. In contrast, the ~15-minute BC concentration resulting from a small number (median = 2-3) of parked colocations at each of 30 sites was only weakly correlated (R2 ~ 0.2) between mobile and fixed sensors. This comparison suggests that highly dynamic fine-scale transport of conserved pollutants may cause substantial disagreement in real-time concentrations among even closely spaced monitors, although these concentration differences may substantially be reduced at longer time averaging scales. Results from this unique experiment can inform the design of future studies that wish to compare mobile and fixed observations.




Messier KP, Chambliss SE, Gani S, Alvarez R, Brauer M, Choi JJ, Hamburg SP, Kerckhoffs J, LaFranchi B, Lunden MM, Marshall JD, Portier CJ, Roy A, Szpiro AA, Vermeulen RCH, Apte JS. Mapping Air Pollution with Google Street View Cars: Efficient Approaches with Mobile Monitoring and Land Use Regression. Environ Sci Technol. 2018 Oct 24. doi: 10.1021/acs.est.8b03395