An instantaneous spatiotemporal model for predicting traffic-related ultrafine particle concentration through mobile noise measurements

Highlights: • A GAM model was developed to predict UFP concentration from mobile measurements. • This is one of the first models using noise level to predict areal UFP concentration. • The model covariates include noise, street canyon index and meteorological data. • The model could explain 80% deviance of the UFP number concentration. People living near roadways are exposed to high concentrations of ultrafine particles (UFP, diameter < 100 nm). This can result in adverse health effects such as respiratory illness and cardiovascular diseases. However, accurately characterizing the UFP number concentration requires expensive sets of instruments. The development of an UFP surrogate with cheap and convenient measures is needed. In this study, we used a mobile measurement platform with a Fast Mobility Particle Sizer (FMPS) and sound level meter to investigate the spatiotemporal relations of noise and UFP and identify the hotspots of UFP. UFP concentration levels were significantly influenced by temporal and spatial variations (p < 0.001). We proposed a Generalized Additive Models to predict UFP number concentration in the study area. The model uses noise and meteorological covariates to predict the UFP number concentrations at an industrial site in Taichung, Taiwan. During the one year sampling campaign from fall 2013 to summer 2014, mobile measurements were performed at least one week for each season, both on weekdays and weekends. The proposed model can explain 80% of deviance and has coefficient of determination (R²) of 0.77. Moreover, the developed UFP model was able to adequately predict UFP concentrations, and can provide people with a convenient way to determine UFP levels. Finally, the results from this study could help facilitate the future development of noise mobile measurement.