Estimation of ground level particulate matter concentrations through the synergistic use of satellite observations and process-based models over South Korea

Abstract. Long-term exposure to particulate matter (PM) with aerodynamicdiameters 10 (PM 10 ) and 2.5 µ m (PM 2.5 ) hasnegative effects on human health. Although station-based PM monitoring hasbeen conducted around the world, it is still challenging to provide spatiallycontinuous PM information for vast areas at high spatial resolution.Satellite-derived aerosol information such as aerosol optical depth (AOD) hasbeen frequently used to investigate ground-level PM concentrations. In thisstudy, we combined multiple satellite-derived products including AOD withmodel-based meteorological parameters (i.e., dew-point temperature, windspeed, surface pressure, planetary boundary layer height, and relativehumidity) and emission parameters (i.e., NO, NH 3 , SO 2 ,primary organic aerosol (POA), and HCHO) to estimate surface PM concentrations over South Korea. Randomforest (RF) machine learning was used to estimate both PM 10 andPM 2.5 concentrations with a total of 32 parameters for 2015–2016. Theresults show that the RF-based models produced good performance resulting R 2 values of 0.78 and 0.73 and root mean square errors (RMSEs) of 17.08 and8.25 µ g m −3 for PM 10 andPM 2.5 , respectively. In particular, the proposed models successfullyestimated high PM concentrations. AOD was identified as the most significantfor estimating ground-level PM concentrations, followed by wind speed, solarradiation, and dew-point temperature. The use of aerosol information derivedfrom a geostationary satellite sensor (i.e., Geostationary Ocean Color Imager, GOCI) resulted in slightlyhigher accuracy for estimating PM concentrations than that from apolar-orbiting sensor system (i.e., the Moderate ResolutionImaging Spectroradiometer, MODIS). The proposed RF models yieldedbetter performance than the process-based approaches, particularly inimproving on the underestimation of the process-based models (i.e., GEOS-Chemand the Community Multiscale Air Quality Modeling System, CMAQ).