FBE- Geomatik Mühendisliği Lisansüstü Programı - Doktora
Bu koleksiyon için kalıcı URI
Gözat
Yazar "Aldabash, Midyan" ile FBE- Geomatik Mühendisliği Lisansüstü Programı - Doktora'a göz atma
Sayfa başına sonuç
Sıralama Seçenekleri
-
ÖgeParticulate matter 2.5 – PM2.5 mapping using MODIS satellite data and multivariate non-linear regression analysis over Marmara Region – Turkey(Lisansüstü Eğitim Enstitüsü, 2021) Aldabash, Midyan ; Bektaş Balçık, Filiz ; 692447 ; Geomatik MühendisliğiThis study estimated and mapped the dry-mass concentrations of PM2.5 (Particulate Matter 2.5) on the ground-level for the Marmara Region, Turkey, during 2013-2017, using multivariate nonlinear regression analysis. The study was conducted using AOD550 (Aerosol Optical Depth at 550 nm) derived from the collection C6.1 of MODIS (Moderate Resolution Imaging Spectroradiometer) Aqua and Terra of the NASA (National Aeronautics and Space Administration), meteorological variables of the ERA-5 assimilation of the ECMWF (European Centre for Medium-Range Weather Forecasts), and PM2.5 ground measurements of the UHKİA/NAQMS (National Air Quality Monitoring System) ground-stations network. MODIS AOD data were validated against AERONET (AErosol RObotic NETwork) ground-based sunphotometer AOD obtained from three sites around Turkey. In addition, MODIS AOD was compared to MERRA-2 (Modern-ERA Retrospective Analysis for Research and Application) Version 2 of AOD 550nm. The study aims mainly to map the PM2.5 dry-mass concentrations using MODIS C6.1 AOD550 gridded data over the Marmara Regions, Turkey, for the period between 2013 and 2017. Since a wide variety of health issues are attributed to PM2.5, we aim to assess human exposure to high PM2.5 concentrations in the region due to analyzing the spatiotemporal variability of PM2.5. Besides, an objective is to evaluate the efficiency of the MODIS sensor to retrieve AOD at 550 over the study area using both sunphotometer and gridded data. The fourth objective is to examine the associations between meteorological conditions and PM2.5. Finally, Turkey is a candidate country for a European Union membership. Therefore, we aim to assess the vulnerability of humans in the Marmara region to excessive PM2.5 concentrations based on the E.U.s' air quality regulations. Datasets of AOD obtained from three sites (METU-ERDEMLI, ATHENS-NOA, and Cyprus CUT-TEPAK) of AERONET ground-based sunphotometer were used to validate MODIS AOD. Collocated AOD pixels from the Terra and Aqua daily rasters were averaged and combined to improve the availability AODs. However, MODIS Terra, Aqua, and combined Terra/Aqua AOD datasets were validated independently against AERONET. Also, the three MODIS datasets were compared to the MERRA-2 AOD datasets of Morning, Noon, combined Morning/Noon, respectively. Furthermore, the statistics of both dataset validation versus AERONET were investigated under different R.H. conditions. According to the statistics of seasonal and daily regression analysis, MODIS AOD is better than MERRA-2 by the mean of R2, MAE, and RMSErel. Combined AODs from MODIS Terra/Aqua and MERRA-2 Morning/Noon exhibited better validation results than the individual datasets against AERONET AOD. Apparent annual cycling was demonstrated by the three datasets of AOD over Turkey. Though, MODIS underestimated while MERRA-2 overestimated AERONET. In addition, MODIS was found more efficient than MERRA-2 in detecting extreme episodes of AOD over Turkey. Particles' hygroscopic growth by humidity changes aerosols' microphysical characteristics. It was found that the R2 of the regression analysis increases in low RH conditions. However, no significant changes were found in the MAE/RMSErel. Therefore, MODIS combined Terra/Aqua AOD dataset has been chosen to estimate PM2.5 concentrations over the Marmara Region. The linear correlation between ambient MODIS AOD and PM2.5 has been found weak (R2=0.152). The hygroscopic growth effects of R.H. (Relative Humidity) on the particles were corrected by the nonlinear extinction coefficient 𝑓(𝑅𝐻). Also, BLH (Boundary Layer Height) has been utilized to estimate the AOD at the ground level (AODh0). The MODIS combined Terra/Aqua AOD outcomes were divided by the BLH values obtained from ERA-5 meteorological gridded datasets. Daily mean pixel values of relative humidity, boundary layer height, temperature, wind speed, precipitation, and sea level pressure, were calculated from averaging every days' provided ERA-5 rasters. However, daily pixel values of resultant wind speed have been calculated from the vertical and horizontal wind speed rasters. The meteorological variables were added to the MODIS AOD in a multivariate additive model to improve the PM2.5–AOD relationship. Meteorological variables along with the 𝑓(𝑅𝐻) and BLH corrections have been found increasing the additive models' R2 into 0.492. Collection window of 3x3 pixels (30x30 km2) of the MODIS AOD was used to calculate the daily values of AOD over every PM2.5 measuring site. Thus, to increase the daily AOD values' representativity, only the collected number of daily matchups resulted from averaging more than 5 pixels were considered in the model. Since R.H. changes the aerosols' microphysical properties, daily matchups with extreme humidity events (R.H.>90%) were excluded from the model. Besides, minimal values of PM2.5 (PM2.5<5 μg/m3) are hard to be detected by the statistical models; thus, they were excluded to reduce bias and improve R2. Although 7% of the data were excluded, the models' R2 increased to 0.602. At the last stage of the model, monthly and seasonal predictors were added to optimize the models' efficiency and statistics. The last stage of the model caused a 14% inclination for the models'R2. Clear annual cycling of PM2.5 was detected by both the model and datasets of ground stations. The monthly average PM2.5 was found to decrease in summer and reach its maximum in the winter season. The concentrations reached their maximum in January (39 μg/m3) and reached their minimum in July (13 μg/m3). This could be explained by burning low-quality coal and fossil fuels for heating in the cold months. Also, the model tends to underestimate and overestimate PM2.5 in the months of winter and summer, respectively. Daily maps of PM2.5 were averaged for every month to reduce the missing pixels and produce monthly- seasonal- and annual-average maps of PM2.5. Availability of pixels in every raster ranged between 33% and 66% for winter and summer, respectively. Compared to other cities of the region, the monthly average map of PM2.5 has shown high concentrations over Edirne, Bursa, and Kocaeli. This is due to the intensive burning of coal and fossil fuels for heating. According to the E.U.s' (European Union) limit of PM2.5 (25 μg/m3), the seasonal and annual maps were classified to produce the E.U.s' exceedance maps. The seasonal exceedance maps showed that the PM2.5 exceeds the limit in most of the Marmara Region in winter. However, more than 90% of the region is lower than the E.U.s' limit for the other three seasons. The annual exceedance maps clearly showed that the area of exceedance declined from 11.8% to 2.1% between 2013 and 2017, respectively. Finally, the model outcomes of PM2.5 were validated against ground measured PM2.5. Since PM2.5 seasonal and annual variability were found in the Marmara Region, leave-one-out CV Cross-Validation (CV) was conducted on a seasonal and annual basis. The number of available matchups of the leave-season-out for winter was found the least among the four seasons. However, a good agreement is found between the model-estimated and ground-measured PM2.5. The R2 ranged between 0.51 and 0.65 for spring and summer, respectively. The final R2 and RMSErel were 0.687 and 27%, respectively. However, the model underestimated the extreme events of PM2.5, especially in winter.