A Bayesian Downscaler Model to Estimate Daily PM2.5 Levels in the Conterminous US

• Authors: Yikai Wang, Emory University; Xuefei Hu, Emory University; Howard Chang, Emory University; Lance Waller, Emory University; Jessica H. Belle, Emory University; Yang Liu, Emory University
• Journal Title: International Journal of Environmental Research and Public Health
• Volume: Volume 15, Number 9
• Publisher: MDPI | 2018-09-01
• Type of Work: Article | Final Publisher PDF
• Publisher DOI: http://doi.org/10.3390/ijerph15091999
• Subjects: Environmental Sciences; Health Sciences, Public Health; Biology, Biostatistics
• Keywords: Science & Technology; Life Sciences & Biomedicine; Environmental Sciences; Public, Environmental & Occupational Health; Environmental Sciences & Ecology; PM2; 5; Bayesian downscaler; exposure modeling; aerosol optical depth; MODIS; AEROSOL OPTICAL DEPTH; REGRESSION

Abstract

There has been growing interest in extending the coverage of ground particulate matter with aerodynamic diameter ≤ 2.5 µm (PM2.5) monitoring networks based on satellite remote sensing data. With broad spatial and temporal coverage, a satellite-based monitoring network has a strong potential to complement the ground monitor system in terms of the spatiotemporal availability of the air quality data. However, most existing calibration models focus on a relatively small spatial domain and cannot be generalized to a national study. In this paper, we proposed a statistically reliable and interpretable national modeling framework based on Bayesian downscaling methods to be applied to the calibration of the daily ground PM2.5concentrations across the conterminous United States using satellite-retrieved aerosol optical depth (AOD) and other ancillary predictors in 2011. Our approach flexibly models the PM2.5versus AOD and the potential related geographical factors varying across the climate regions and yields spatial-and temporal-specific parameters to enhance model interpretability. Moreover, our model accurately predicted the national PM2.5with an R2at 70% and generated reliable annual and seasonal PM2.5concentration maps with its SD. Overall, this modeling framework can be applied to national-scale PM2.5exposure assessments and can also quantify the prediction errors.