Investigating the performance of models estimating NO2 and PM2.5 in the US metropolitan areas and their environmental inequality and vulnerable people areas

Abstract

ABSTRACT The study explores the effects of incorporating satellite-based estimates and kriging based on land use regression (LUR) on the ability to predict NO2 and PM2.5 concentrations in the urban cores of the Metropolitan Statistical Areas (MSAs) located on the continental United States (US). We determined that satellite-based estimates and kriging combined with LUR resulted in the lowest averaged error values and spatially unbiased error distributions for predicting NO2, while kriging combined with LUR, but without satellite-based estimates, resulted in the lowest averaged error values and spatially unbiased error distributions for predicting PM2.5. Satellite-based estimates have a marginal effect on discerning PM2.5 concentrations. Using predicted data, we identified the relationship between various demographic factors and concentrations of these pollutants. While median income and percentage of whites are only associated with PM2.5, the percentage of children is associated with both NO2 and PM2.5.