Modeling the spatiotemporal distribution of ambient ultrafine particles in Shanghai, China
Year: 2025
Type: article
Abstract: Background The interpretation of epidemiological evidence on ambient ultrafine particles (UFPs) is limited by uncertainties in exposures exposure due to unaccounted spatiotemporal variability. Objective To develop models for estimating the spatiotemporal distribution of ambient UFPs in Shanghai, China. Methods We established a study-specific monitoring network and repeatedly measured on-road UFP across three typical seasons. Using traffic, land use, air pollutants, and meteorological data, we constructed five prediction models based on both linear and nonlinear approaches, in single- and two-stage frameworks. Results Mean (± standard deviation) of on-road UFP concentrations was 26,313 (± 8,932) N∙cm −3 , showing pronounced spatial and temporal variations. The Random Forest (RF) one-stage model (R 2 = 0.63, root mean squared error (RMSE) = 5463 N∙cm −3 ) and RF-based two-stage model (R 2 = 0.62, RMSE = 5521 N∙cm −3 ) performed best, followed by Support Vector Regression (SVR) one-stage model (R 2 = 0.57, RMSE = 5830 N∙cm −3 ), SVR-based two-stage model (R 2 = 0.53, RMSE = 6146 N∙cm −3 ) and multiple linear regression model (R 2 = 0.39, RMSE = 6975 N∙cm −3 ). External validation confirmed the superiority of the RF two-stage model. Conclusions The models effectively captured spatiotemporal variation in UFP concentrations, supporting improved exposure assessment and health effect estimation.
Source: Environment International
Institution: Fudan University
Language: English
Subfield: Health, Toxicology and Mutagenesis
Field: Environmental Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $4,750
OpenAlex ID: W4416442291