Near-Surface PM2.5 Concentrations Derived from Satellites, Simulation and Ground Monitors

Abstract:

Exposure to fine particulate matter (PM_2.5) is globally associated with 3.2 million premature deaths annually. Satellite retrievals of total column aerosol optical depth (AOD) from instruments such as MODIS, MISR and SeaWiFS are related to PM_2.5 through local aerosol vertical profiles and optical properties. A globally applicable and geophysically-based AOD to PM_2.5 relationship can be calculated from chemical transport model (CTM) simulations. This approach, while effective, ignores the wealth of ground monitoring data that exist in some regions of the world. We therefore use ground monitors to develop a geographically weighted regression (GWR) that predicts the residual bias in geophysically-based satellite-derived PM_2.5. Predictors such as the AOD to PM_2.5 relationship resolution, land cover type, and chemical composition are used to predict this bias, which can then be used to improve the initial PM_2.5 estimates.

This approach not only allows for direct bias correction, but also provides insight into factors biasing the initial CTM-derived AOD to PM_2.5 relationship. Over North America, we find significant improvement in bias-corrected PM_2.5 (r²=0.82 versus r²=0.62), with evidence that fine-scale variability in surface elevation and urban factors are major sources of error in the CTM-derived relationships. Agreement remains high (r²=0.78) even when a large fraction of ground monitors (70%) are withheld from the GWR, suggesting this technique may add value in regions with even sparse ground monitoring networks, and potentially worldwide.