10th International Aerosol Conference
September 2 - September 7, 2018
America's Center Convention Complex
St. Louis, Missouri, USA

Abstract View


Modeling On-road Fine and Ultrafine Particle Concentrations in Los Angeles

Nu Yu, Shi Shu, Lu Zhang, Yan Lin, Jun Wu, YIFANG ZHU, UCLA

     Abstract Number: 311
     Working Group: Aerosol Modeling

Abstract
Air pollution is among the top threats to public health in Los Angeles. Traffic related particulate matter (PM) exposure has been linked to different adverse health effects. In this study, fine PM (PM2.5) and ultrafine particle (UFP) concentrations were measured on roadways outside moving taxi vehicles in the Greater Los Angeles area from April to November, 2013. The total distance driven is approximately 11,000 kilometers and the total hours of field measurement is approximately 500 hours. We developed and compared four types of models to estimate UFP and PM2.5 concentrations on freeways, arterial roads, and local surface streets: multiple linear regression models without and with temporal variance structures (MLR and sMLR models), and generalized additive models without and with temporal variance structures (GAMs and sGAMs). A backward selection procedure was used to select predictive variables from meteorological condition panel, traffic condition panel, and spatial feature panel. The modeling results show that GAMs explained the highest percentages of the total UFP and PM2.5 variance. The MLR models generated the best cross validation (CV) results with least discrepancy from the general R2s. The models performed the best on the data collected on arterial roads comparing with the data collected on freeways and local streets. The arterial road models generated general R2s equivalent or above 0.40 and CV R2s ranging from 0.34 to 0.73. All arterial road models and the local street PM2.5 sMLR model and GAM generated both general and CV R2s greater than 0.30. Temperature, relative humidity, vehicle speed, annual average daily traffic (AADT) and time of day variables were consistently selected by these models. These predictors, except for time of day, show positive associations with UFP or PM2.5 concentrations, which are consistent with previous studies, and the physical and chemical properties of the traffic related UFP and PM2.5.