A Bayesian Downscaler Model to Estimate Daily PM2.5 levels in the Continental US

TL;DR

This paper introduces a Bayesian downscaling model that accurately estimates daily PM2.5 levels across the US by integrating satellite data and geographical factors, improving national air quality assessments.

Contribution

The study develops a flexible, interpretable Bayesian framework for national-scale PM2.5 estimation using satellite AOD data, addressing limitations of previous small-scale models.

Findings

Achieved 70% R2 in national PM2.5 prediction.

Generated reliable seasonal and annual PM2.5 maps.

Model effectively captures spatial and temporal variability.

Abstract

There has been growing interest in extending the coverage of ground PM2.5 monitoring networks based on satellite remote sensing data. With broad spatial and temporal coverage, satellite based monitoring network has a strong potential to complement the ground monitor system in terms of the spatial-temporal availability of the air quality data. However, most existing calibration models focused on a relatively small spatial domain and cannot be generalized to national-wise study. In this paper, we proposed a statistically reliable and interpretable national modeling framework based on Bayesian downscaling methods with the application to the calibration of the daily ground PM2.5 concentrations across the Continental U.S. using satellite-retrieved aerosol optical depth (AOD) and other ancillary predictors in 2011. Our approach flexibly models the PM2.5 versus AOD and the potential related geographical factors varying across the climate regions and yields spatial and temporal specific parameters to enhance the model interpretability. Moreover, our model accurately predicted the national PM2.5 with a R2 at 70% and generates reliable annual and seasonal PM2.5 concentration maps with its SD. Overall, this modeling framework can be applied to the national scale PM2.5 exposure assessments and also quantify the prediction errors.