Investigation of agricultural residue burning and wildfire impacts on air quality via satellite retrievals in Southern Turkey

Abstract

Air pollution is one of the major global problems due to its environmental, health and climate impacts. Respiratory and cardiovascular diseases caused by air pollution are important concern. Therefore, air quality monitoring is a significant necessity to control air pollution in addition to emission estimation and air quality modeling. Ground measurements and satellite retrievals are used to monitor air quality and understand the impacts of sources such as point and mobile sources, and episodic air pollution events like wildfires. Remote sensing is more advantageous compared with ground-based measurements to monitor air quality with wide spatial coverage and increasing spatial resolution with new instruments. In addition, it shows the effects of sources better in places that are in higher elevation by measuring total tropospheric column such as emissions caused by industrial activities with high stack height or buoyant plumes from fires. Thus, it provides a great advantage while monitoring air quality in fire periods. Developments in remote sensing instruments and retrieval algorithms still continue and spatial resolution is increasing. In addition, it will be possible to monitor air quality with higher temporal resolution by geostationary satellites in the near future. Agricultural residue burning can be defined as burning of agricultural waste after harvesting, and these activities are prohibited in Turkey similar to many other countries. In spite of the regulations, and being aware of harms caused by agricultural residue burning, farmers may still prefer to burn residues after harvesting to cultivate the area easier, and kill insects and weeds. Farmers perform burning activities from June to October period depending on the climate and region in Turkey. These residue burning activities have significant impacts on environment, soil productivity and also air quality. Even though wildfires are natural periodic events, there is a significant increase in number and duration of wildfires in the recent years caused by climate change. Turkey was one of the countries that impacted by these recent intense wildfire events. Southwest Turkey struggled with wildfires in 2021 summer period that took several weeks to control, and these fires also captured a global attention. Forests were destroyed, animals were killed, and local people were harmed caused by these wildfires. According to the Global Forest Watch, fire counts did not exceed 125 in the last 4 years in Antalya and Mugla. However, in July-August 2021, Antalya fire counts were highest with 673, followed by 368 in Mugla. Another important impact of these fires was decreased air quality in the surrounding regions. The aim of this thesis is investigating the impacts of agricultural residue burning activities and wildfires on air quality in Southeastern Turkey using CO, HCHO, and NO2 satellite retrievals, ground measurements, and meteorological parameters. Visible Infrared Imaging Radiometer Suite (VIIRS) Fire Radiative Power (FRP) product was downloaded from National Aeronautics and Space Administration (NASA) Fire Information for Resource Management System (FIRMS) archive and monthly summation of FRP was spatially joined with 1×1 km2 gridded domain. Study areas and periods of agricultural residue burning and wildfires were determined based on the spatial distribution of FRP signals. Adana, Osmaniye, Mersin, Gaziantep, Diyarbakir, Sanliurfa, and Mardin provinces were selected to investigate residue burning effects on air quality. Antalya, Mugla, and Mersin provinces were selected for wildfires. The TROPOspheric Monitoring Instrument (TROPOMI) CO, HCHO and NO2 retrievals and ground observations of CO and NO2 were used to monitor air quality for both cases. TROPOMI Level-2 CO, HCHO, and NO2 retrievals were downloaded from NASA Goddard Earth Sciences Data and Information Services Center (GES DISC) then, processed spatially and temporarily. Ground measurements were assessed for both cases to understand whether these stations were capable of capturing the pollution from the emissions caused by residue burning and fires. Meteorological parameters were used for both investigations in order to understand the transport of pollutants. Statistical and spatiotemporal analyses were performed in both cases. For wildfires, June-September 2021 period was selected that includes one month before the start of the fires, fire period and one month after the end of the fires. This period provided a time range to compare pollution levels with wider period for non-fire days. Eight wildfire regions were selected based on the total FRP signals with wide impact areas and high FRP values. TROPOMI CO, HCHO and NO2 retrievals average total columns were calculated with retrievals within 3 km around for the selected fire regions and within 6 km around the close-by air quality monitoring stations (AQMS). Results were categorized as fire region (FR) and ground station region (GSS). Pollution level changes on fire days and non-fire days were compared spatially and temporally. Satellite retrieved pollution levels over fire regions and AQMS locations along with ground observations at AQMSs were compared for fire days and non-fire days. According to FRP time series, most intense fires were observed in Antalya (An-1) followed by Mugla (Mu-1) region. Fires started at the end of July in Antalya and Mersin, and at the beginning of August in Mugla provinces. Accordingly, in spatial distribution of FRP, largest signals were observed in Antalya-1 in July, and Mugla-1, Mugla-2 (Mu-2), and Mugla-4 (Mu-4) in August. FRP signals were not observed in June and September. Wind speed, temperature and FRP were investigated together to understand the meteorological effects. Time series showed wind speed peaked on the same days that FRP peaked in Antalya-1 region. In addition, temperature was relatively higher during the fire period. Fire days were selected from 28th July to 7th August and spatial distribution of CO, HCHO and NO2 retrievals were examined during selected fire days. Strong signals were observed for all air pollutants. CO signals were observed over the fire regions and transporting to the south with wide path. HCHO signals were south direction of fire regions and NO2 was mostly close to the fire regions. Wind roses were created for each region and it was resulted that different signal locations can be related with transport of the pollutants with wind, formation mechanisms, and lifetime of pollutants. Spatial distribution of monthly total column averages and ratio to June averages were calculated and assessed spatially. Strong signals of CO and NO2 above 3.5 in Antalya and Mugla in July and August were observed, but HCHO signals were not as clear as CO and NO2. Strong signals were diminished in September/June ratios since there was no fire event in September. FRP, fire region (FR), and ground station region (GSS) were compared with time series and boxplots for fire days (FD) and non-fire days (NFD). In time series, peak total columns were higher around AQMSs than fire region with all pollutants when the FRP peaked in Antalya-1. However, in boxplots for comparison of pollutant retrievals on fire days and non-fire days, fire region upper and median values were higher than AQMS locations in terms of CO and NO2 on fire days. However, HCHO around AQMS locations on fire days upper and median value was higher than over fire region in Antalya-1. Differences between fire days and non-fire days were significant for all pollutants. Correlations of FRP, fire region and AQMS location pollutant retrievals, and ground observations were performed. Highest correlation with FRP was found with fire region CO (R=0.79) in Antalya-1. Correlations of CO and NO2 were lower around AQMS location than fire region, yet it was the opposite for HCHO. Lowest correlation with FRP was AQMS NO2 observations (R=0.24). Moreover, correlation between AQMS location NO2 and AQMS NO2 ground observations was also low even though they represent the same regions. Finally, difference between pollutant averages on fire days and non-fire days were calculated. Highest change was in Antalya-2 (An-2) region NO2 with 314.3% increase. NO2 percentage changes were higher compared to other pollutants. Furthermore, fire region CO and NO2 increase was higher in most of the regions, unlike the HCHO. This difference can be related to HCHO formation mechanism. Antalya-1 and Antalya-2 mean NO2 ground observations showed increases during fire period. However, Mugla-1 and Mersin-2 regions showed decreases, thus, these stations were not representative for capturing the wildfire impacts. For agricultural residue burning, temporal and spatial changes of fires in 2018-2021 summer periods were examined with total monthly fire counts and their spatial distribution. Regions with frequently observed fire counts and CORINE Land Cover database were used to determine whether the burned areas were agricultural or not. According spatial distribution of monthly summed FRP, five fire regions were selected and change of pollutant concentrations in fire days and non-fire days were examined for 2018 June – October period. Daily and monthly average total column of CO, HCHO and NO2 were calculated around the fire regions and AQMS locations with similar methodology previously used in wildfires. Spatial distribution of total FRP showed strong signals in Sanliurfa–Mardin border in June, and Adana–Osmaniye and Adana–Mersin borders in August 2018. Sanliurfa and Diyarbakir signals were mostly in sparsely distributed areas. Spatial distribution of monthly averages of pollutant retrievals showed that the strongest signals were in Adana–Osmaniye border in August for CO and HCHO. Another signal was observed in Sanliurfa–Mardin border with the same pollutants. NO2 signals were mostly in urban areas, and significant NO2 changes in agricultural fire regions were not observed. According to FRP times series for the selected fire regions, highest FRP values were in Adana–Osmaniye border. When the time series of CO and HCHO were examined, the peaks during fire periods were observed in Adana–Osmaniye border for CO and HCHO. NO2 total columns did not change significantly in any fire region during the fire periods. Moreover, when the fire days and non-fire days were compared, CO and HCHO median and upper values increased on fire days, but NO2 did not change. Correlations were calculated FRP with CO, HCHO and NO2 retrievals over fire regions and AQMS locations and AQMS ground CO and NO2 measurements. Strongest correlations with FRP were fire region CO and fire region HCHO. Correlations were moderate around AQMS locations when compared with fire regions. In Adana–Osmaniye border, no correlation was found for AQMS ground CO and NO2 measurements, but correlation was moderate with FRP for AQMS ground NO2 measurements in Adana–Mersin border. AQMS location CO and NO2 were not correlated with AQMS ground CO and NO2 measurements. Wind roses for the fire regions and wind and pollution roses for AQMS locations were prepared. According to wind roses, wind direction was not from fire regions to the AQMS locations during fire episodes. Additionally, pollution roses showed that the increase in pollution concentrations were not related with residue burning sources. As a result, wildfires caused decrease in air quality with increasing pollution levels. Agricultural fire signals were not strong as wildfires considering FRP and pollutants. However, according to consistency of fire regions in terms of VIIRS FRP product signals, it can be resulted that agricultural residue burning practices still continue in southeast part of Turkey.