Long-range Transport of Aerosols from Biomass Burning over Southeastern South America and their Implications on Air Quality

The long-range transport of aerosols is a global issue since it may significantly affect the air quality of regions without high concentrations of fine particulate matter (PM2.5). Two significant occurrences of long-range transport of aerosols over the state of Parana in Brazil, which occurred during the periods Aug. 16–18 and Sep. 10–14 of 2016, were studied in order to characterize the atmospheric synoptic conditions of these events and to estimate their contribution to the air quality conditions in the northern region of the state. The South American Low Level Jet (SALLJ) was the key meteorological component used to define the origin of the air mass trajectories over the region. In the first event, the SALLJ lost its configuration, bringing air masses from the western part of São Paulo (state), while in the second event, the SALLJ formed over southern Brazil and brought air masses from the northern and central parts of the nation. The significant number of fires from biomass burning in central Brazil associated with synoptic conditions contributed to the increase in PM2.5 concentrations by approximately 70–87% in the region. The transport of aerosols was a determining factor in PM2.5 exceeding the air quality standard in the region. Therefore, to minimize this problem, it is imperative to control biomass burning in Brazil.


INTRODUCTION
The long-range transport of pollutants can influence the air quality of several places worldwide.For instance, aerosols produced in Asia can be transported to locations as far as North America, and substantially contribute to local levels of air pollutants (Lin et al., 2012;Cooper et al., 2015;Verstraeten et al., 2015).Even in remote areas of the Amazon and oceans, where the aerosol concentrations are usually very low, long-distance transport may account for most of the contribution of certain components of the aerosol concentration, as evidenced a long-time ago (e.g., Andreae, 1983;Betzer et al., 1988;Swap et al., 1992).However, the sources associated with long-distance transport are quite varied and involve both natural and anthropogenic aerosols.Many different techniques have been applied to confirm long-range transport as an important source of the local concentration of pollutants (Cuspilici et al., 2017;Yadav et al., 2017).This includes satellite-based estimates, and aircraft and surface observations, as well as numerical models (Freitas et al., 2009;Rosário et al., 2013;Uno et al., 2016;Wang et al., 2017).
The long-range transport of pollutants is already recognized as a significant source of pollutants in Asia (e.g., Lin et al., 2005;Qu et al., 2016;Kanaya et al., 2017;Tatsuta et al., 2017;Yeh et al., 2017) and other places around the world (Karaca et al., 2009;Kaneyasu et al., 2014;Prospero et al., 2014;Vorkamp and Rigét, 2014;Ancellet et al., 2016;Salvador et al., 2016).In addition, it can also be harmful to people's health (Chen et al., 2017).Oh et al. (2015), for example, found that aerosol emission in China plays an important role in the occurrence, during the cold season, of strong pollution episodes.They attributed these episodes to meteorological conditions with the presence of a strong high-pressure system over the Eastern China-Korea region.
The air quality in large urban areas is usually more impacted by local human activities.However, the long-range transport of aerosol can be the main source of episodes of elevated particle concentrations in smaller urban areas.According to Naeher et al. (2007), when considering pollutants released by biomass burning, particulate matter is considered the most harmful to public health.For example, the effects of long-range transport of PM 2.5 from vegetation fires on daily mortality and hospital admissions, in the Helsinki metropolitan area, were investigated by Kollanus et al. (2016).Their results presented evidence of an association between long-range transport of PM 2.5 from vegetation fires and an increase in cardiovascular mortality in Helsinki.
Large projects, such as the South American Biomass Burning Analysis (SAMBBA), have addressed biomass burning in South America.Several studies were published with an emphasis on the Amazon (Pickering et al., 1996;Evangelista et al., 2007;Freitas et al., 2009;Pereira et al., 2011) and São Paulo regions (Silveira et al., 2013;Pereira et al., 2017;Vasconcellos et al., 2007).Long-range transport of biomass burning aerosols, especially those originating in the Amazon and central-west region of Brazil, are transported to the southern and southeastern regions of South America (Pickering et al., 1996;Hart and Spinfaiiae, 1999;Freitas et al., 2005;Silva Dias, 2006;Freitas et al., 2009;Rosário et al., 2013;De Oliveira et al., 2016).Ulke et al. (2007) associated the South American Low Level Jet (SALLJ) with the smoke plume pattern found in the southeastern region of South America, and De Oliveira et al. (2016) characterized the aerosol particles from the Amazon region and their transport to southern Brazil during austral winters.However, the meteorological conditions of the events were not completely addressed in these studies, and, until now, the contribution of long-range transport to local aerosol concentrations has not been quantified.Therefore, the influence of the biomass burning, originating in the Amazon and central-west regions of Brazil, on the air quality and health in southern Brazil remains an open question and is surrounded by great uncertainty.
Local and regional studies are fundamental in filling these gaps and supporting future global analysis and political decisions concerning the long-distance transport of pollutants.Thus, this work analyzed two events of longrange transport of aerosols over southeastern South America in order to characterize the atmospheric synoptic conditions of the events and quantify their contribution to PM 2.5 concentrations.In addition, this work analyzes the impact of long-range transport on the air quality of the region.

STUDY REGION AND MEASUREMENTS
Londrina, a medium-size city located in the north of Paraná, was selected for this case study (Fig. 1).It is the most important city in the region and located about 450 km from São Paulo, a mega-city.Its metropolitan area is the home of about one million people, for which the main economic activities are agribusiness, commerce, and services.The fleet running in the region comprises almost 400,000 vehicles, with 49% of them burning gasohol (25% anhydrous ethanol and 75% gasoline).The remaining half of the fleet Fig. 1.Location of the study region, with emphasis on the distribution of the urban areas, as well as the Brazilian states which potentially contribute to long-range transport of pollutants: Paraná (PR), São Paulo (SP), Mato Grosso do Sul (MS), Goiás (GO), Mato Grosso (MT), Rondônia (RO), Amazonas (AM), Pará (PA), and Acre (AC).are flex-fuel (35.4%), diesel (6.8%), and other types (8.8%) (IBGE, 2017).Flex-fuel vehicles burn gasohol or ethanol, depending on the driver.Diesel sold at petrol stations in the region has 8% biodiesel.
Londrina represents one of the few locations outside big cities in South America where pollutant measurements have been regularly made (Pelicho et al., 2006;Freitas et al., 2012;Pinto et al., 2014;Martins et al., 2016;Targino and Krecl, 2016;Beal et al., 2017;Targino et al., 2017;Almeida et al., 2018).The city is in a strategic position to evaluate the impact of long-range transport, especially from the biomass burning activities in the Amazon forest.The population, for a long time, has witnessed events of bad air quality associated with smoke.In addition, Londrina is not as big as São Paulo, Curitiba, or Porto Alegre, all of which show high levels of air pollutants associated with local sources.The strong contribution of local emissions makes it difficult to identify the real contribution from remote sources.For example, Pereira et al. (2017) found that remote sources also contribute to concentrations of aerosol pollutants in São Paulo but that the local sources have a strong influence on the concentrations of particles.
The dataset used in this work is a compilation of several sources, including short-and long-term measurements of particles in the region of Londrina: PM 2.5 ; black carbon (BC), trace elements, and ozone (O 3 ).Table 1 summarizes the type of data, the local field measurements campaigns, and the additional information compiled from different agencies, as well as the periods for which they are available.
The local measurements of O 3 concentrations started in March of 2015, using an ultraviolet photometry monitor, model 49i (Thermo Scientific, MA, USA), that operates at a wavelength of 254 nm.The degree of ultraviolet light absorption is directly related to O 3 concentrations, as described by the Beer-Lambert Law, and is recorded every minute.BC concentrations were measured every two minutes using an aethalometer, model AE42 (Magee Scientific, Berkeley, CA, USA), at 7 wavelengths (), at 5 L min -1 , and with a selected inlet of 2.5 µm.Both measurements were performed at Federal University of Technology, Paraná (UTFPR), Campus of Londrina, located in a semi-urbanized area (see Fig. 1).More information about the setups and the principle operation of the aethalometer is available in the instruction manual on the website (www.mageesci.com)and in Targino and Krecl (2016).BC analysis was concentrated on two wavelengths, 370 nm and 880 nm.The shorter wavelength of 370 nm is more sensitive to the presence of organic compounds originating in biomass burning, while those from vehicular exhaust have weak absorption at this wavelength (Kirchstetter et al., 2004;Targino and Krecl, 2016).The wavelength of 880 nm is the standard wavelength for measuring BC.BC concentrations were used together with other data to evidence the events of long-range transport of aerosols.The hourly and daily data were considered valid only with 75% availability or more.
PM 2.5 and elemental compositions were available only for a short-term period during measurements.This type of measurement was performed under specific conditions in order to study personal exposure under the population's normal working conditions in an urban environment and, more specifically, to study the association of short-term exposure to PM with respiratory symptoms in healthy adults.Therefore, measurements are available only for the periods and temporal resolutions displayed in Table 1.It is important to note that the official monitoring of the air quality in Londrina (by one station) started only in 2017.

Local Data Analysis
Hourly average concentrations of O 3 and BC at 370 nm and 880 nm are shown in Fig. 2(A).The O 3 profile presents a very well defined diurnal cycle, with peak concentrations around 15:00 local time, which is typical for O 3 .However, BC concentrations presented three peaks in the morning, midday, and evening, which were associated with different traffic intensity.The expressive peak of BC at 880 nm, at midday, coincides with the end of classes in the morning period at UTFPR, at lunchtime.Vehicular activities and emissions increase at this time since the students and members of the staff usually go home or go have lunch outside the university, and they use buses or cars as transport modes.However, homes and restaurants are not usually   close to people's workplaces in Brazil, increasing vehicular activity in the city during this period.On the other hand, the peak in the evening was higher for BC at 370 nm than 880 nm, which is usually the period when local fires of biomass burning are observed around the university, as well as waste burning, as evidenced by Beal et al. (2017).Note that sweeping the yard and burning branches and leaves, as well as some other types of waste produced by homes, are common in the outskirts of Brazilian cities. Associated with higher emission of BC by vehicular activity and biomass burning in the evening, stable atmospheric conditions and the reduction of the boundary layer height contribute to the higher concentrations observed at that time.
The boxplot in Fig. 2 (B) shows the average concentrations of O 3 and BC recorded in the years of 2014, 2015, and 2016 at UTFPR.Results show large variability as well as non-normal distributions of concentrations for both pollutants.Concentration analyses during the week (WE) and the weekend (WK) showed higher average concentrations for BC, at both wavelengths, during the WE than WK.This is an expected result since the vehicular activity is higher during the WE; however, local biomass burning was observed every day, mainly during the winter season.For O 3 , average and median concentrations were similar for the WE and the WK, with no weekend-effect.

Events of Long-range Transport of Particles
Significant events of long-range transport of aerosols are often observed in Londrina, especially during the dry season in the winter.Two of these events coincided with local campaigns of measurements in 2016, the first occurring Aug. 16-18 and the second, Sep.10-14.The reader can access local pictures in the supplementary material and check how these plumes affect visibility when crossing the region (Fig. 1S).Fig. 3 shows BC concentrations at both wavelengths recorded every two minutes, for both events, making it possible to see peaks of BC concentrations at 370 nm.For example, on Aug. 17 and 18, the BC concentrations at 370 nm were higher than at 880 nm, indicating the influence of biomass burning sources on BC concentrations.
During the two events, PM 2.5 measurements were recorded every minute.Analyzing the profile presented in Fig. 4, the increase in PM 2.5 concentrations on these days is very clear when compared with other days of sampling.In the first event, the average concentration of measurements performed in the afternoon on Aug. 17 was 5.6 µg m -3 , increasing to 25 µg m -3 on Aug. 18 but with concentrations around 40 µg m -3 during the last 3 hours of measurements in the afternoon (top panel in Fig. 4).During the second event, the effect of long-range transport was more intense on Sep. 13 as well as on the morning of Sep. 14 (bottom panel in Fig. 4).The average concentrations were 28 µg m -3 in the afternoon of Sep. 13 and 21.7 µg m -3 and 3.4 µg m -3 on Sep. 14 in the morning and afternoon periods, respectively.Table 2 shows a summary of the average daily PM 2.5 for the first event and BC concentrations for both events.Based on the average daily PM 2.5 concentrations observed before (Aug.16) and after (Aug.19 and 20) the days of the event, we estimated an increase of about 70% in the PM 2.5 concentration in relation to the observation on Aug. 18 (peak of the first event).For the second event, PM 2.5 concentrations are not available for a full 24-h.However, based on the PM 2.5 concentrations recorded during the same hours of the days for the period, the possible contribution of long-range transport was estimated around 87% for PM 2.5 , taking into account the average concentrations for mornings (28 µg m -3 on Sep. 13 and 21.7 µg m on Sep.14) and afternoons (3.4 µg m -3 on Sep.14).
Potassium (K + ) has been recognized as a good tracer for the biomass burning contribution (Falkovich et al., 2005;Caumo et al., 2016).Analyses of the traces' elemental composition in PM 2.5 filters revealed that the highest K + concentrations were observed on Aug. 18 and Sep.13, compared to other samples in the period.The K + concentrations ranged from 288 ng m -3 to 696 ng m -3 in filters influenced (some hours of sampling) by long-range transport of aerosols, which were 1.25 to 3.77 times higher   than the average for all filters sampled in the whole period (Aug. 1 till Sep. 29, 2016).Therefore, this particular pattern associated with K + is more evidence of long-range transport of aerosols in the region.Images taken by the Moderate Resolution Imaging Spectroradiometer (MODIS) product, from the Aqua and Terra satellite, for the periods, were also analyzed to better evidence the events.For example, Fig. 5 shows the composition of images from the MODIS product (satellite Aqua and Terra) for the region, taken on Sep. 12, 2016.It is possible to see the plume of biomass burning and several spots of fire in the central part of Brazil (State of Mato Grosso [MT]).Considering that local fires were not observed, the main source of additional aerosols in the region was the long-range transport of aerosols from biomass burning in northern South America.Additional MODIS images for the periods are available as supplemental material (Fig. 2S).
To identify the predominant regions from where air masses are coming, backward trajectory analyses were performed using the HYSPLIT model (Hybrid Single-Particle Lagrangian Integrated Trajectory) (Stein et al., 2015).The model is available on the website (http://ready.arl.noaa.gov/HYSPLIT.php)and was run for 120 h and at heights of 100, 500, and 800 m.Fig. 6 shows the fire outbreaks and the back trajectories arriving in Londrina at 00:00, 12:00 and 18:00 for both events.The analysis showed air masses coming from the state of São Paulo and central part of Brazil.The MODIS recorded a large number of fires in regions through which the trajectories of the air masses passed during the sampling period, with emphasis on the states of São Paulo (southeastern region) and Mato Grosso (western region) (Figs.6(A) and 6(B), which, together, contributed more than 50% of the fires in this period.A clear line of fires can be observed in the central and northern parts of Brazil (the states of Mato Grosso, Tocantins, Acre, and Rondônia).Backward trajectories suggest that such fires contributed to the concentration and composition of the PM 2.5 collected in both events in Londrina.
For the first event, it is important to note that backward trajectories indicate air masses coming from the state of São Paulo, a region of intense sugarcane production, which is also associated with the burning process during harvesting.The aerosols originating in sugar cane burning in São Paulo were related to several health problems (Arbex et al., 2007;Silveira et al., 2013).For the second event analyzed, the air mass coming from the central and northern parts of Brazil presented a high frequency and intensity of biomass burning.

ATMOSPHERIC CONDITIONS OF LONG-RANGE TRANSPORT EVENTS
In order to characterize the synoptic conditions  predominant during both events, a diagnostic analysis of meteorological fields was carried out using the Global Forecast System (GFS) dataset, which contains 0.5° gridded data of several meteorological variables at 31 vertical pressure levels, from the National Centers for Environmental Information (NCEI), a National Oceanic and Atmospheric Administration (NOAA) division website.The infrared temperature highlighted Geostationary Satellite Imagery (GOES-13) satellite imagery complement the large-scale data, while specific surface station data were used to characterize local changes during the events.

August, 2016, Event 1
The first event occurred between Aug. 16 and 19, 2016.According to the GFS analysis, since Aug. 15, 1800 UTC, the Paraná state had been under the influence of a closed surface low pressure center of 1008 hPa.This small lowpressure center lasted for about 18 hours more, turning into a surface at 1200 UTC over southern Brazil.While a previous cold front was moving east at the same latitudes but over the Atlantic Ocean, with the associated extratropical surface, a cyclone with a central pressure of about 990 hPa was distant from continent at about 40°S, 30°W.However, this synoptic condition was accompanied by deep convection systems in Paraná, as seen by the GOES-13 infrared imagery.The Londrina's surface station registered moderate 1-hour accumulated rain only at 0800 UTC on Aug. 16 (17.4mm).The associated surface high-pressure system following the cold front was not so strong over the continent, showing a center of only 1020 hPa, which is a relatively low value for August.
A second extratropical cyclone coming from the Pacific Ocean passed the Andes Range Mountains at the latitude of 45° S and organized a new cold front over the Atlantic Ocean that extended its influence over the continent at latitudes of the southern region of Brazil, including the eastern part of Paraná, on Aug. 17 and 18.At 850 hPa (about 1500 m agl), the GFS analyses show that the wind flow was from the northwest on Aug. 17, which can be associated with the SALLJ (the South American Low Level Jet, discussed below in Section 4.3), extending from Bolivia to Paraná, as shown in Fig. 7 at 12:00 UTC on Aug. 17.Most of the SALLJ ended exactly over the state of Paraná at that time.Only a few millimeters of rain were registered by the Londrina's surface station between 19:00 and 21:00 UTC on Aug. 18.The cold front, together with its cloudiness, moved quickly to the east, and after that, the SALLJ lost its configuration in the following hours.Behind the cold front, a relatively stronger high-pressure system started to advance from the southwest to the east-northeast on Aug. 18.At this moment and later, winds below 700 hPa became weak, and the western part of the surface highpressure system dominated the synoptic condition in Paraná.
A few millimeters of rain were registered by the Londrina's surface station during the first hours of Aug. 19, while the high-pressure center reached 1022 hPa at the coastline of Fig. 7. GFS analysis at 12:00 UTC of Aug. 17, 2016 for wind (m s -1 ) at 850 hPa and surface pressure (hPa) fields.The pressure field contour intervals are 4 hPa each.The red ellipse highlights the Low Level Jet position.
the entire southern region and São Paulo.All analyses are supported by both GFS analyses and GOES-13 infrared imagery, which show that this synoptic condition forced surface winds to blow from the ocean to the continent in Paraná on Aug. 19.The high-pressure center continued to move east from 12:00 UTC on Aug. 19.

September, 2016, Event 2
The predominant synoptic situation during the second event, which occurred between Sep. 12 and 13, 2016, is less complex than that of the first event, due to a typical cyclogenesis observed in its fast development from 12:00 UTC on Sept. 12.During the development and after reaching its mature stage at 12:00 UTC on Sept. 13, the northwestern part of the system, associated with the extratropical cyclone seen in the infrared satellite image in Fig. 8 (left panel), formed the SALLJ.This brought air parcels from the central part of the continent, including the northern region of the Amazon basin, to higher latitudes.While the synoptic cyclone developed into its final stages of life, the SALLJ continued to influence the entire area of Paraná.The SALLJ can be easily seen in the GFS analysis of Fig. 8 (right panel), which shows the wind field at 850 hPa and the surface pressure field.The surface pressure field enhances the extratropical system, and the 850 hPa wind field shows the SALLJ, which can be detected in the South American continent at the latitudes of southern Brazil.

The South American Low Level Jet (SALLJ)
As shown in the previous sections, the SALLJ seems to be an important factor in the long-range transport of pollutants in the state of the Paraná, when coupled with biomass burnings.The SALLJ has been studied for a long time, mainly its role in bringing heat and moisture to feed convective systems observed in the La Plata Basin (Argentina), and the South and Southeast Regions of Brazil (e.g., Virji, 1981).Such thermodynamic characteristics inspired the South American Low Level Jet Experiment (SALLJEX), as described in Vera et al. (2006).The same criteria defined by Bonner (1968) to characterize the LLJ observed in the North America Great Plains (east of the Rock Mountains) has been used by the SALLJ.Generally, the Bonner's criteria can be summarized as follows: the magnitude of the wind must present a vertical profile, in low levels of the troposphere, in a way that the wind maximum shows a jet-like profile.Climatology based on the NCEP/NCAR reanalysis data (Marengo et al., 2004) indicates Northern Paraguay as the main area of maximum frequency of occurrence of SALLJ, as well as the level of 850 hPa as its maximum wind magnitude level.
As an example, Fig. 9 shows the 850 hPa wind magnitude field at 12:00 UTC of 13, in Sep.2016,where the wind magnitude over Londrina is about 20 m s -1 at that level, with the SALLJ core shown in the Southern Paraná and Northeastern of Paraguay, which means that the SALLJ marginally affected Londrina in that case.This statement, supported by Fig. 10, shows the vertical profiles of wind magnitude for a point at Northeast Paraguay (25.0°S, 55.0°W) (left panel), and a point next to Londrina (23.5°S, 51.0°W) (right panel).These figures show that the wind magnitude profile for Paraguay follows Bonner's criterion between 1000 and 700 hPa strictly, showing a classic LLJ profile, while for Londrina, the wind magnitude profile does not fall above the 850 hPa level, as expected, according to Bonner's criterion.However, Londrina's wind magnitude shows at least a like-jet profile with a strong maximum of 850 hPa.The second and major maximum wind magnitude seen next to 300 hPa level in both profiles are associated with the upper level subtropical jet, a common feature observed over the continent during cold front passages in South America, which has little importance for the transport of particles in the low atmosphere, in the cases studied here.

CONCLUSIONS
The combination of synoptic conditions and the longrange transport of biomass burning aerosols increases PM 2.5 concentrations in the northern region of Paraná, as well as in other areas of southern South America.The SALLJ is a key component in defining meteorological conditions for the transport of aerosols.Depending on its configuration, the SALLJ can bring air masses from western São Paulo (state), as well as from the Amazon and central parts of Brazil, where fires are frequent during the austral winterspring transition.
The long-range transport of aerosols was evidenced by the increase in PM 2.5 , BC 370 nm and the potassium concentrations measured in the region.The contribution varied from 70% to 87%, depending on the plume event crossing the region.Long-range transport events are responsible for causing PM 2.5 concentrations to exceed the recommendations issued by the World Health Organization, which does not commonly occur when contributions originate Finally, it is imperative to eliminate the anthropogenic biomass burning in Brazil since it affects not only the regions where it occurs but also contributes to the deterioration of atmospheric air quality in other regions.

g
NASA and NOAA a UTFPR = Federal University of Technology, Parana; b period of 2016; c measurements performed during some hours in the morning and afternoon; d USP = University of São Paulo; e SIMEPAR = Parana Meteorological System; INMET = National Institute of Meteorology; NCEI = National Centers for Environmental Information; NOAA= National Oceanic and Atmospheric Administration; f MODIS = Moderate Resolution Imaging Spectroradiometer; GOES-13 = Geostationary Satellite Imagery, channel 13; g NASA = National Aeronautics and Space Administration.

Fig. 2 .
Fig. 2. O 3 and BC concentrations measured at UTFPR in 370 nm and 880 nm in the years 2014-2016.(A) Average hourly concentrations; (B) Distributions of all recorded concentrations at week (WE) and weekend (WK).

Fig. 4 .
Fig. 4. PM 2.5 concentrations measured in the city of Londrina during the events.Top panel, first event on Aug. 18, 2016.Below panel, second event on Sep. 13, 2016.

Fig. 6 .
Fig. 6.Trajectory of air masses arriving in Londrina, calculated by HYSPLIT, where figures A and B represent the trajectories of the air masses and fire outbreaks on Aug. 18 and Sep.13, 2016, respectively.

Fig. 8 .
Fig. 8. Infrared image from the GOES-13 geostationary satellite at 12:00 UTC on Sep. 13, 2016.Color bar indicates temperature ranges in °C with colder temperatures indicating higher altitudes.South America countries and Brazilian states geopolitical divisions are indicated (left panel).GFS analysis at 12:00 UTC of Sep. 13, 2016 for wind (m s -1 ) at 850 hPa and surface pressure (hPa) fields.The pressure field contour intervals are each 4 hPa.The red ellipse highlights the Low Level Jet position (right panel).

Fig. 9 .
Fig. 9. GFS analysis at 12:00 UTC of Sep. 13, 2016 for wind magnitude at 850 hPa shaded according to the color bar in m s -1 .

Table 1 .
Variables measured and compiled in their respective periods and sources.

Table 2 .
Average 24-h PM 2.5 and BC concentrations for the first and second events.