An Ensemble Approach to Improve Surface PM2.5 Estimate from Space

Abstract:

We propose an ensemble approach to estimate the surface PM2.5 mass concentrations using multiple chemistry transport models (CTMs) and aerosol optical depths (AODs) retrieved from multiple satellite sensors. The hypothesis is that each satellite AOD product or CTM has its unique strengths and weaknesses, and a combination of them can yield a better estimation of surface PM2.5. We conduct an experiment study for a one-year (i.e., 2012) period to test this hypothesis. First, two sets of model-based surface PM2.5 are obtained from WRF-Chem and GEOS-Chem. Second, four sets of satellite-estimated surface PM2.5 are generated from MODIS (Aqua) and VIIRS AODs by multiplying the ratio of each model's collocated PM2.5 to AODs; these are MODIS/GEOS-Chem, MODIS/WRF-Chem, VIIRS/GEOS-Chem and VIIRS/WRF-Chem. We evaluate the satellite-estimated PM2.5 of each set and their ensemble averages against EPA’s hourly in-situ measurements at 650 sites within the contiguous United States. The results show that the ensemble averages are overall in better agreement with the EPA’s measurements, although some ensemble members could perform better in some regions.