You are here

Evaluation of alternative sensor-based exposure assessment method

Principal Investigator: 
,

University of Washington

This study is comparing exposure estimates from wearable personal exposure monitors with estimates derived from other exposure assessment methods.

Funded under
Status: 
In review
Abstract

Abstract for the 2017 HEI Annual Conference

Evaluation of Alternative Sensor-Based Exposure Assessment Methods
Edmund Seto1, Elena Austin1, Graeme Carvlin1, Jeffery Shirai1, Alan Hubbard2, Katharine Hammond2, Ying-Ying Meng3, Michael Jerrett3, Ronald Cohen2
1University of Washington, Seattle, WA, USA; 2 University of California, Berkeley, CA, USA; 3 University of California, Los Angeles, CA, USA.
 
Background The Bay Area Near Roadway Sensor (BANRS) Study evaluated a “next generation” air quality monitoring system that utilizes lower-cost electrochemical gas and optical PM sensors. Sensor data have the potential to be incorporated into new spatial-temporal models for use in exposure assignments for traffic-related air pollution health effects studies.
 
Methods Two monitors, including of both gas and particle sensors, were deployed at a regulatory near roadway monitoring site in Oakland, CA for approximately 1 year. Gas sensors (CO, NO, NO2 and O3) were manufactured by Alphasense (Essex, UK). Particles were measured using a Shinyei (New York, NY) PPD42NS optical sensor. Sensor measurements were compared to regulatory monitors on the 1-hour time scale. The relationship between sensor response and reference instruments was examined using correlation plots and multivariate regression. An iterative method was used to calibrate sensor results to regulatory instruments using the regression model. Additionally, regulatory monitor and sensor data were compared to hourly traffic count data, and sensor, ultrafine PM (UFP), black carbon (BC), and noise level data were collected along three transects downwind of the 880 freeway to characterize pollutant decays downwind of roadways.
 
Results After calibration, of the various sensors, the highest R2 values relating sensor data to regulatory monitoring data were observed for hourly NO that ranged from 0.64 and 0.94 for different months. For hourly CO, the R2 values ranged from 0.50 and 0.95. The R2 values for hourly NOx were lower, ranging from 0.43 to 0.82. The correlations were improved for longer 24-hour averaging periods. The sensors performed well in identifying consistent diurnal patterns in pollutant levels as those observed from regulatory instruments. The PM sensor data were correlated with reference PM2.5 measurements during non-winter months, but not well correlated during the winter. When air quality data were compared to hourly traffic data, neither the reference regulatory PM2.5 monitor nor the PM sensor measurements were sensitive to variations in hourly traffic. However, UFP and BC, as well as the NO/CO ratio were found to be sensitive indicators of hourly traffic.  For the downwind transects, pollutant decays were observed for NO, NO2, CO, BC, UFP, and noise with increasing downwind distance from the freeway. However, there was some evidence of potential local sources observed in the transect closest to the Port of Oakland. Except for one site that was near train traffic, ambient noise along the transects demonstrated a decay with increasing distance from freeway.
 
Conclusions The low-cost sensors performed well, in some cases over several months, but some sensors failed, suggesting the need for routine sensor maintenance in future deployments. When the sensors operated properly, they produced data that were correlated with regulatory near roadway monitoring instruments, and that were correlated with hourly traffic counts. As the availability of next-generation sensor data increases, and their calibrations against trusted instruments are documented, we may find new spatial-temporal exposure models begin to use these data in health effects studies.
 
POSTER