A Bayesian ensemble approach to combine PM2.5 estimates from statistical models using satellite imagery and numerical model simulation

Abstract

Ambient fine particulate matter less than 2.5 μm in aerodynamic diameter (PM 2.5) has been linked to various adverse health outcomes. PM 2.5 arises from both natural and anthropogenic sources, and PM 2.5 concentrations can vary over space and time. However, the sparsity of existing air quality monitors greatly restricts the spatial-temporal coverage of PM 2.5 measurements, potentially limiting the accuracy of PM 2.5-related health studies. Various methods exist to address these limitations by supplementing air quality monitoring measurements with additional data. We develop a method to combine PM 2.5 estimated from satellite-retrieved aerosol optical depth (AOD) and chemical transport model (CTM) simulations using statistical models. While most previous methods utilize AOD or CTM separately, we aim to leverage advantages offered by both data sources in terms of resolution and coverage using Bayesian ensemble averaging. Our approach differs from previous ensemble approaches in its ability to not only incorporate uncertainties in PM 2.5 estimates from individual models but also to provide uncertainties for the resulting ensemble estimates. In an application of estimating daily PM 2.5 in the Southeastern US, the ensemble approach outperforms previously developed spatial-temporal statistical models that use either AOD or bias-corrected CTM simulations in cross-validation (CV) analyses. More specifically, in spatially clustered CV experiments, the ensemble approach reduced the AOD-only and CTM-only model's root mean squared error (RMSE) by at least 13%. Similar improvements were seen in R 2. The enhanced prediction performance that the ensemble technique provides at fine-scale spatial resolution, as well as the availability of prediction uncertainty, can be further used in health effect analyses of air pollution exposure.

Satellite‐Based Daily PM2.5 Estimates During Fire Seasons in Colorado

“…The details of this modeling framework are described in Murray et al (2018); thus, we only provide a brief summary here. The first stage involves two statistical downscalers to calibrate the PM 2.5 -AOD relationship varying in both time and space (i.e., the AOD downscaler) and calibrate CMAQ PM 2.5 simulations (i.e., the CMAQ downscaler), respectively.…”

“…Therefore, our study focused on the fire seasons (April to September) over 2011–2014. We applied a Bayesian ensemble model (Murray et al, 2018), which takes advantage of both the high-resolution of MAIAC AOD and full coverage of model simulations to predict daily PM 2.5 . To our knowledge, this is the first study focusing on the PM 2.5 estimation using 1 km high-resolution AOD data in a Mountain State over fire seasons.…”

Random forest models for PM2.5 speciation concentrations using MISR fractional AODs

“…This process was also repeated for 10 times. Finally, a spatialclustered CV was performed following Murray et al [35] to test the model's robustness when a large spatial group of data were missing instead of a single location. A hierarchical clustering method by proximity of monitoring locations was used to define 10 clusters of all the monitors.…”

Prediction and assessment of change in PM2.5 during COVID-19 lockdown using remote sensing and deep learning approach: A case study of Kanpur city

“…(Van Donkelaar et al, 2006) used a global chemical transport model (GEOS-CHEM) and simulate the factors effecting AOD and PM2.5 for better prediction of PM2.5. (Murray et al 2019) has estimated PM2.5 by combining both the AOD values and CTM simulation using statistical method and results shows improvement in R 2 value. (Sorek-Hamer et al 2015) used the AOD product, mixed-effect model, and the daily calibration approach to predict PM2.5 and observed a considerable improvement in prediction of PM2.5 for high re ectance regions.…”