Enhancing the applicability of satellite remote sensing for PM2.5 estimation using MODIS deep blue AOD and land use regression in California, United States

Abstract

We estimated daily ground-level PM2.5 concentrations combining Collection 6 deep blue (DB) Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth (AOD) data (10 km resolution) with land use regression in California, United States, for the period 2006-2012. The Collection 6 DB method for AOD provided more reliable data retrievals over California's bright surface areas than previous data sets. Our DB AOD and PM2.5 data suggested that the PM2.5 predictability could be enhanced by temporally varying PM2.5 and AOD relations at least at a seasonal scale. In this study, we used a mixed effects model that allowed daily variations in DB AOD-PM2.5 relations. Because DB AOD might less effectively represent local source emissions compared to regional ones, we added geographic information system (GIS) predictors into the mixed effects model to further explain PM2.5 concentrations influenced by local sources. A cross validation (CV) mixed effects model revealed reasonably high predictive power for PM2.5 concentrations with R-2 = 0.66. The relations between DB AOD and PM2.5 considerably varied by day, and seasonally varying effects of GIS predictors on PM2.5 suggest season-specific source emissions and atmospheric conditions. This study indicates that DB AOD in combination with land use regression can be particularly useful to generate spatially resolved PM2.5 estimates. This may reduce exposure errors for health effect studies in California. We expect that more detailed PM2.5 concentration patterns can help air quality management plan to meet air quality standards more effectively.