A geographically and temporally weighted regression model for assessing intra-urban variability of volatile organic compounds (VOCs) in Yangpu district, Shanghai

The adverse health effects of volatile organic compounds (VOCs) highlight the need of spatiotemporal models to assess exposure for epidemiological studies. Herein, a geographically and temporally weighted regression (GTWR) model was developed to estimate the spatial variability of the VOCs, including acetone, benzene, toluene, and m/p-xylene in Yangpu (YP) District, Shanghai, China. The GTWR model shows a good prediction performance, with cross-validation (CV) R² = 0.60, 0.75, 0.64, and 0.62 for acetone, benzene, toluene, m/p-xylene, respectively. Moreover, the GTWR model outperforms the ordinary least squares (OLS), geographically weighted regression (GWR), and temporally weighted regression (TWR) model in terms of higher CV R² and lower CV root mean square error (RMSE) values. The 100 m × 100 m gridded concentrations of acetone, benzene, toluene and m/p-xylene estimated from the GTWR model range from 1.57 to 2.57 ppbv, 0.24–1.06 ppbv, 0.40–1.56 ppbv, and 0.31–1.05 ppbv, respectively. Among them, the concentrations of acetone, toluene and m/p-xylene are far below the thresholds of the United States Environmental Protection Agency (USEPA), whereas the benzene concentrations are higher than that of the USEPA's standard by 2–10 times. The VOC emissions are closely associated with transportation, industrial activities, commercial activities, and the release from recreational areas. The lifetime cancer risk (LCR) of benzene for juveniles, adults and elderly are in the range of 2.3 × 10⁻⁵- 0.98 × 10⁻⁵, 3.1 × 10⁻⁵-1.4 × 10⁻⁴, and 3.7 × 10⁻⁵-1.6 × 10⁻⁴, respectively, indicating potential and high cancer risks to human health. The integrated non-cancer risks (HQ) of four VOC species are far below the safe level (=1), suggesting a negligible non-cancer risk for residents living in YP District.