Estimating Size-Resolved Surface Particulate Matter Concentrations Using MISR High-Resolution Size-Fractionated Aerosol Optical Depth

Abstract:

There is significant public health interest in gaining a better understanding of the health effects associated with particulate matter (PM) of different composition and size, yet ground-based monitoring data for such PM species is extremely limited. Due to their spatial and temporal coverage, satellite observations of total column aerosol optical depth (AOD) have increasingly been used to estimate surface concentrations of PM. While techniques for using satellite observations of AOD to predict surface concentrations of PM_2.5 have been established, predicting surface concentrations of different particle sizes and species is more challenging. The Multi-angle Imaging SpectroRadiometer (MISR) instrument has the unique capability of estimating both total column AOD as well as total column size fractionated (small, medium and large) AOD. Using MISR AOD and AOD size fractionated products derived from high-resolution (275 m) observations reported at a spatial scale of 4.4 km in combination with national Air Quality System (AQS) monitoring data over the 2008-2009 period, we examine the association between size-fractionated MISR AOD and surface measurements of PM at different sizes (PM_2.5 and PM₁₀) and PM_2.5 species (EC, OC, SO₄^2-, NH₄⁺) over the greater Los Angeles area. While there was a limited sample size of speciated PM data, the strongest univariate association found was between AOD and PM_2.5 SO₄^2- (R²=0.76). Incorporating meteorological data from weather stations in the area resulted in improvements to the models associating AOD with PM_2.5 and PM₁₀ mass. We found that PM_2.5 was best predicted by a spatio-temporal model of AOD that also included dew point temperature and wind speed (R²=0.61), and that PM₁₀ was best predicted by a spatio-temporal model of large fraction AOD that also included atmospheric pressure and wind speed (R²=0.65). These flexibly specified spatio-temporal models enabled reliable predictions of surface PM_2.5 and PM₁₀ concentrations at a 4.4km resolution over the study region from MISR AOD and large fraction AOD, respectively.