Estimation of ground-level particulate matter concentrations based on synergistic use of MODIS, MERRA-2 and AERONET AODs over a coastal site in the Eastern Mediterranean


Tuna Tuygun G., Gündoğdu S., Elbir T.

Atmospheric Environment, vol.261, 2021 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 261
  • Publication Date: 2021
  • Doi Number: 10.1016/j.atmosenv.2021.118562
  • Journal Name: Atmospheric Environment
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Applied Science & Technology Source, Aqualine, Aquatic Science & Fisheries Abstracts (ASFA), Artic & Antarctic Regions, BIOSIS, CAB Abstracts, Chemical Abstracts Core, Chimica, Communication Abstracts, Compendex, Computer & Applied Sciences, EMBASE, Environment Index, Geobase, Greenfile, INSPEC, Metadex, Pollution Abstracts, Public Affairs Index, Veterinary Science Database, Civil Engineering Abstracts
  • Keywords: Aerosol optical depth (AOD), Gap-filled AOD, Artificial neural network, PM estimation, AEROSOL OPTICAL DEPTH, BOUNDARY-LAYER HEIGHT, PM2.5 CONCENTRATIONS, PM10 CONCENTRATIONS, SATELLITE DATA, AIR-QUALITY, LAND, PREDICTION, VARIABLES, COLUMNAR
  • Dokuz Eylül University Affiliated: Yes

Abstract

© 2021 Elsevier LtdSatellite-derived aerosol optical depth (AOD) products are widely used to estimate the spatial and temporal characteristics of ground-level particulate matters (PM). Satellite-based approaches in the operational use of PM estimation models have limitations such as missing values in satellite-based variables due to the impact of cloud cover, relative humidity, and aerosol vertical distribution. Therefore, the spatial-temporal resolution of the estimation models is forced to be improved by using AOD products having different temporal resolutions from different platforms. In this study, ground-based PM10 concentrations were aimed to be estimated using different gap-filled AOD datasets between 2008 and 2016 over a coastal site in the Eastern Mediterranean. For the estimation of PM10 concentrations, daily AOD data were mainly used from MODIS. Then, two different gap-filled MODIS AOD datasets with both AERONET and MERRA-2 AODs were also examined separately on both multi-annual and multi-seasonal basis since the number of MODIS AOD data was temporally limited in the region. A pattern recognition neural network (PRNN) model was used for the estimation of PM10 concentrations. AOD with several meteorological parameters from an on-site meteorological station and aerosol diagnostic products from MERRA-2 were used as inputs to the estimation model. The most significant variables for the model were identified in 23 independent variables using the multiple linear regression (MLR). The results indicated that the best estimation (R = 0.74) was obtained with the gap-filled AODMODIS+MERRA dataset for the period covering all years whereas the AODMODIS dataset alone showed the poorest performance (R = 0.62). However, the performance of the gap-filled datasets varied with the seasons. For example, the AODMODIS dataset alone showed the best estimation performance (R = 0.67) in the summer whereas the gap-filled AODMODIS+AERONET dataset had the best performance (R = 0.59) in the winter. Overall results suggest that for estimating ground-level PM concentrations, an approach of gap-filled AOD dataset usage for estimation models is useful especially in rainy seasons such as the winter and autumn where MODIS AOD retrievals are limited.