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

Abstract

Abstract. Long-term exposure to particulate matter (PM) with aerodynamic diameters < 10 (PM10) and 2.5 µm (PM2.5) has negative effects on human health. Although station-based PM monitoring has been conducted around the world, it is still challenging to provide spatially continuous PM information for vast areas at high spatial resolution. Satellite-derived aerosol information such as aerosol optical depth (AOD) has been frequently used to investigate ground-level PM concentrations. In this study, we combined multiple satellite-derived products including AOD with model-based meteorological parameters (i.e., dew-point temperature, wind speed, surface pressure, planetary boundary layer height, and relative humidity) and emission parameters (i.e., NO, NH3, SO2, primary organic aerosol (POA), and HCHO) to estimate surface PM concentrations over South Korea. Random forest (RF) machine learning was used to estimate both PM10 and PM2.5 concentrations with a total of 32 parameters for 2015–2016. The results show that the RF-based models produced good performance resulting in R2 values of 0.78 and 0.73 and root mean square errors (RMSEs) of 17.08 and 8.25 µg m−3 for PM10 and PM2.5, respectively. In particular, the proposed models successfully estimated high PM concentrations. AOD was identified as the most significant for estimating ground-level PM concentrations, followed by wind speed, solar radiation, and dew-point temperature. The use of aerosol information derived from a geostationary satellite sensor (i.e., Geostationary Ocean Color Imager, GOCI) resulted in slightly higher accuracy for estimating PM concentrations than that from a polar-orbiting sensor system (i.e., the Moderate Resolution Imaging Spectroradiometer, MODIS). The proposed RF models yielded better performance than the process-based approaches, particularly in improving on the underestimation of the process-based models (i.e., GEOS-Chem and the Community Multiscale Air Quality Modeling System, CMAQ).

Mapping and Understanding Patterns of Air Quality Using Satellite Data and Machine Learning

“…The relatively high importance of AOD and the good model performance emphasize the suitability of AOD to infer on PM10 concentrations when additional parameters are taken into account. A comparison to similar studies, for example, by Grange et al (2018) and Park et al (2019), reveals comparable relative feature importances of AOD, solar radiation (Park et al, 2019), DOY, and BLH (Grange et al, 2018;Park et al, 2019). The high importance of the 3-day mean of the east-west wind component found in this study aligns with the high importance of the back trajectory clusters in the study by Grange et al (2018).…”

“…Here, DOY is used as seasonal proxy. To mirror the seasonal cycle, DOY was converted to a sine curve with +1 representing summer solstice and 1 representing winter solstice (Park et al, ; Stolwijk et al, ). To further approximate variability in emission strengths based on human activity, day of the week is included.…”

“…This suggests that wind information needs to be extended to a longer time scale to influence PM predictions.As shown in Grange et al (2018), wind speed aggregated for a daily period can substantially influence PM10 concentrations, with lower speeds causing higher concentrations. Park et al (2019) also found the maximum wind speed of previous 3 hr to be of importance for their statistical predictions of PM10. The north-south wind component PDs (instantaneous and 3 days) do not provide clear trends.…”

“…While numerous air pollution studies apply statistical models mainly to accurately predict PM concentrations (Hu et al, ; Kloog et al, ; Stafoggia et al, ; van Donkelaar et al, ; Zhang et al, ), recent studies additionally analyzed feature importances and the information content of explanatory variables in the statistical models to infer processes. For example, Park et al () used a RF model to predict PM10 in South Korea and found large influence of AOD, day of the year (DOY), wind speed, and solar radiation on the modeled PM concentrations. Similarly, Grange et al () obtained high feature importances for wind speed, back trajectory cluster, DOY, and air temperature, using the RF approach for PM10 measurements in Switzerland.…”

Validation, comparison, and integration of GOCI, AHI, MODIS, MISR, and VIIRS aerosol optical depth over East Asia during the 2016 KORUS-AQ campaign

“…Currently operating low Earth orbit (LEO) satellite sensors, such as the Moderate Resolution Imaging Spectroradiometer (MODIS), the Multi-angle Imaging Spectroradiometer (MISR), and the Visible/Infrared Imager Radiometer Suite (VIIRS), provide global aerosol information but at a temporal resolution that is limited to once per day at least and, typically, to once every 2-3 d due to cloud cover Hsu et al, 2013;Jackson et al, 2013;Levy et al, 2013). Most satellite-based aerosol retrieval techniques and algorithms have been developed for these LEO sensors (Diner et al, 1998;Higurashi and Nakajima, 1999;Hsu et al, 2004;Kaufman et al, 1997;Remer et al, 2005;Torres et al, 1998). To overcome temporal resolution limitations, there were several attempts to retrieve AOD using first-generation meteorological geostationary satellites such as the Geostationary Operational Environmental Satellite (GOES), the Geostationary Meteorological Satellite (GMS), and the Multifunction Transport Satellite (MTSAT), but they showed worse accuracy than those of LEO sensors due to the wider and fewer visible channels with coarser spatial resolution, which make it difficult to distinguish aerosol types (Kim et al, 2008;Knapp et al, 2002;Urm and Sohn, 2005;Wang et al, 2003;Yoon et al, 2007).…”