Impact of Mobile Monitoring Network Design on Air Pollution Exposure Assessment Models

Abstract

Short-term mobile monitoring campaigns are increasingly used to assess long-term air pollution exposure in epidemiology. Little is known about how monitoring network design features, including the number of stops and sampling temporality, impacts exposure assessment models. We address this gap by leveraging an extensive mobile monitoring campaign conducted in the greater Seattle area over the course of a year during all days of the week and most hours. The campaign measured total particle number concentration (PNC; sheds light on ultrafine particulate (UFP) number concentration), black carbon (BC), nitrogen dioxide (NO2), fine particulate matter (PM2.5), and carbon dioxide (CO2). In Monte Carlo sampling of 7327 total stops (278 sites × 26 visits each), we restricted the number of sites and visits used to estimate annual averages. Predictions from the all-data campaign performed well, with cross-validated R 2s of 0.51–0.77. We found similar model performances (85% of the all-data campaign R 2) with ∼1000 to 3000 randomly selected stops for NO2, PNC, and BC, and ∼4000 to 5000 stops for PM2.5 and CO2. Campaigns with additional temporal restrictions (e.g., business hours, rush hours, weekdays, or fewer seasons) had reduced model performances and different spatial surfaces. Mobile monitoring campaigns wanting to assess long-term exposure should carefully consider their monitoring designs.

Impact of Roadside Mobile Monitoring Design on Epidemiologic Inference – A Case Study of Ultrafine Particles and Cognitive Function

“…We used the annual average site PNCs from each sampling campaign to develop universal kriging -partial least squares (UK-PLS) exposure prediction models. PNC was log-transformed and regressed against the first two PLS components, which summarized 188 geographic covariates predictive of TRAP (e.g., land use, roadway proximity, population density), as previously detailed (Blanco et al, 2023;Blanco, Gassett, et al, 2022). We evaluated each model by comparing the five-fold cross-validated site predictions against the annual averages from the all-data campaign (our best estimates).…”

“…We evaluated each model by comparing the five-fold cross-validated site predictions against the annual averages from the all-data campaign (our best estimates). We and others have shown the importance of validating model predictions against unbiased estimates, and how comparisons against biased, unstable campaign measurements (e.g., from restricted sampling designs) produces noisy and misleading conclusions Blanco et al, 2023;Kerckhoffs et al, 2016;Messier et al, 2018). We evaluated the performance of each model using mean-square error (MSE) -based R 2 (R 2 MSE), which evaluates whether pairs of predictions and observations are the same (i.e., along the one-to-one line) rather than simply linearly associated, like traditional regression-based R 2 , and is thus better suited to evaluate predictive performance.…”

“…Most campaigns produce poor to moderately performing exposure assessment models. Limited attention has been paid to the impact of monitoring design on exposure assessment models (Apte et al, 2017;Blanco et al, 2023;Blanco, Gassett, et al, 2022;Hatzopoulou et al, 2017;Messier et al, 2018;Saha et al, 2019), and no studies have assessed how these design features impact exposure assessment models or subsequent epidemiologic health inferences.…”

“…Many epidemiologic studies require long-term exposures, which is unique from other applications such as commuter or high exposure studies where shorter-term, on-road, and/or weekday business or rush hour exposures may be most relevant. We have previously shown that common, restricted sampling designs produce biased annual-average exposure predictions Blanco et al, 2023). The objective of this study was to investigate the degree to which stationary (temporary roadside stop) monitoring design choices impact subsequent epidemiologic inferences.…”

Characterizing Ultrafine Particle Mobile Monitoring Data for Epidemiology

“…To achieve unbiased annual average estimates, the design of drive passes is important. Mobile campaigns should strive for designs that are temporally balanced by the time of day, day of the week, and season of the year. , However, the expense and logistical challenges of mobile monitoring studies increase with the number of drive passes and the attention paid to a temporally balanced sampling design.…”