Aerosol Pollutants during Agricultural Biomass Burning : A Case Study in Ba Vi Region in Hanoi , Vietnam

Impact of traditional biomass burning activities on regional air quality is a major environmental concern. Measurement campaigns were performed during post-harvesting activities in the Ba Vi region in Hanoi in May–June of 2015 and 2016. To quantify the source for regional haze the sampling of rice straw burning emissions was performed on fields. Carbonaceous (OC, EC, BrC) fractions, heavy metals, organic and ionic composition, and microstructure were characterized. A set of functionalities (hydroxyl, aliphatic, carbonyl, carboxylate, and nitro groups) revealed a functional marker of pile combustion. Optical, microstructural, and chemical analyses of environmentally-dangerous pollutants from traffic and cooking sources provided characteristics and functional markers of different pollution sources. Chemical features of rice straw burning were identified on the Ba Vi site during the haze episode of 2015, when PM10 mass approached the high smoke intensity, up to 167 μg m. Small-scale meteorology affected PM10, OC and EC, and ion mass in days of highest relative humidity and fogs. In days of highest smoke OC dominated PM10 mass by up to 42%, the OC/EC ratio approached 20, in line with observations of mainly smoldering emissions across the fields. Spectral features of regional haze smoke demonstrated the absorption of rice straw burning whereas the impact of biogenic, traffic, and cooking sources were significantly lower. Individual particle analyses showed carbonaceous particles internally/externally mixed with inorganic fly ash and dust. Smoke micromarkers revealed the microstructure of regional aerosols representative for Southeast Asia in BB periods. Significantly lower PM10 mass concentrations and strong difference in aerosol composition before post-harvesting activities suggested that agricultural burning represents a large contribution to air quality degradation in the rural area of Vietnam.


INTRODUCTION
Combustion has been increasingly recognized as the most important source of environmentally-active aerosols, containing black carbon (BC), organic carbon (OC), and inorganic fly ash.They impact directly and indirectly on the Earth's radiation balance, cloudiness, and subsequently affect the regional and global climate (Ramanathan and Carmichael, 2008).BC contribute significantly to the atmospheric warming due to absorption of visible radiation (Bond et al., 2013).Strong evidence has been provided for light absorption of OC, also known as brown carbon (BrC) (Alexander et al., 2008;Kirchstetter and Thatcher, 2012) which can change the direct aerosol radiative forcing from cooling to warming.Atmospheric Asian brown clouds formed from internally-mixed BC induce multiple effects on regional air temperature and agriculture in China and South Asia (Engling and Gelencser, 2010).
Emission and speciation of biomass burning (BB) aerosols is highly source-dependent; it varies considerably between urban and rural regions, populated and remote areas (Reid et al., 2005;Yan et al., 2006).Aerosol composition relates to combustion phase (open flaming vs. smoldering) and type of biomass (Engling et al., 2009;Popovicheva et al., 2015a).Distinct regional characteristics in BB smoke have been observed, suggesting that local emissions dominate aerosol chemistry (Engling et al., 2011;Lim et al., 2012), while transport from wildfire sources impacts the composition of aged aerosols (Diapouli et al., 2014;Engling et al., 2014;Popovicheva et al., 2016).Specifically, BB may profoundly affect air quality and public health in urban areas, indicating smoke as an indicator of harmful pollution (Amiridis et al., 2012;Popovicheva et al., 2014a).Based on composition and morphology, distinct types of aerosols such as soot, tar balls, and organic particles with inorganic inclusions dominated by potassium and sulfates were identified in fire emissions (Posfai et al., 2003;Hand et al., 2005).
BB activities in northern Southeast (SE) Asia comprise a variety of combustion processes, ranging from residential burning, domestic heating, and cooking while burning of agricultural residuals is a particularly important source during dry seasons (Engling et al., 2009;Tsay et al., 2013).In Vietnam rice straw burning deserves attention as the most important pollution practices in agricultural provinces (Le et al., 2013;Сo et al., 2014).The biggest Vietnamese cities face severe levels of urban pollution, with critical concentrations of particulate matter (PM) significantly exceeding air quality standards (Oanh et al., 2006).Presently, the main contribution to the high pollution in Hanoi is road traffic due to the exponential increase in number of vehicles during last decades (MONRE, 2011).Various components of traffic-emitted aerosols are considered as dangerous pollutants with respect to human health, causing respiratory and cardiovascular diseases (Hien et al., 2007;Steiner et al., 2013).Residential cooking emissions are a key source of PM especially in winter when urban emissions are dominant (Hai and Oanh, 2013).Surrounding agricultural areas of Hanoi can act as a seasonal source of pollution during intense rice straw burning in the weeks following harvest periods, [kenh14.vn].However, a lack of knowledge concerning the properties of source-dependent aerosols hinders assessments of BB impact on air quality: adverse air pollution and impacts of chemically -active aerosols are remaining underestimated.
Initiated in 2007, the Seven South-East Asian Studies (7-SEAS) campaign seeks to perform interdisciplinary research in the field of aerosol-environment interactions, particularly focusing on impacts of BB (Huang et al., 2013;Tsay et al., 2013).Major agricultural and domestic emissions from remote sources were assessed in near-source and ambient campaigns in the Son La province, northwestern Vietnam (Popovicheva et al., 2016).Measurements of sizesegregated OC, EC, organic/inorganic composition, molecular markers, and microstructure revealed the major features of BB in ambient aerosols.
This paper is based on continued work following the 7-SEAS campaign and focuses on rural area in the Ba Vi region, a particularly important rice -producing area in Vietnam, which is close to the urban area of Hanoi.Realtime measurements and sampling on the Ba Vi site during haze episodes yield optical (BC) and chemical (OC, EC, ions, functionalities, elements) characteristics for evaluation of PM constitutes and impacts of environmentallydangerous pollutants.Individual particle analyses provide the major groups comprising the smoke microstructure and micromarkers of pollution.Because the Ba Vi region is a highly populated area with various potential sources of emissions, the characterization of on-field emissions as well as traffic and residential cooking in the Hanoi city is intentionally conducted, in order to emphasize the specific atmospheric pollutants from the major local sources and then to compare them with those identified in the Ba Vi region during a haze episode.Analyses of carbon fractions, organic and inorganic functionalities, and individual particle grouping is performed assuming that ambient aerosols during haze period are highly affected by emissions from smoldering rice straw on surrounding fields.The assessment of the aerosol background before harvesting supports the assessment of the pollution sources.

Measurement Campaigns
In late May to mid June 2015 intensive on-field burnings were observed nearby Hanoi city.Measurement campaigns were carried out during a haze period from 27 May to 14 June 2015 in the mountain region of Ba Vi National Park located 48 km west of Hanoi, Fig. 1.The site was located 400 meters above sea level (a.s.l.) inside a natural forest containing lower mountain evergreens, mixed coniferous and broadleaf vegetation.Additional sampling campaign was conducted the following year, from 1 to 6 June 2016.That time was the pre-harvest period when the fires on fields had not started yet, thus allowing assessing the aerosol properties in non-BB period.
For characterization of traditional rice straw burning, near-source sampling was performed on fields before and after measurement campaign.To investigate atmospheric pollutants from sources different from BB, additional measurements were performed in the Hanoi city.Traffic and cooking sources were sampled on city streets during rush hour on work day and during the weekend in March of 2015.Sampling sites were set up near the Kim-Lien tunnel and on the Nghia Do street close to a traditional small restaurant with a typical cooking stove, using coal Long-range transport to Ba Vi site from the wider regions of South East Asia was examined by means of 120 hours air mass back trajectories (BWT) at starting heights of 500, 1000 and 1500 meter a.s.l. which were calculated for every 24 hour during the sampling period.The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (NOAA ARL READY Website) was used.To evaluate the level of regional pollution by BB emissions the days of air mass transport to the Hanoi city through the Ba Vi region were identified.During these days the data obtained on the SPARTAN site situated on roof of a big building of the VAST institute in the Caugiay District in the center of the Hanoi city (21°0'N, 105°8'E) (Snider et al., 2016) were evaluated as well.

Instrumentation and Sampling
A mobile platform was created for performing the measurement campaigns.Aerosol equivalent black carbon (EBC) concentrations were determined using custom made portable aethalometer which measures the attenuation of light transmitted through particles that are deposited on a quartz fiber filter (Hansen and Rosen, 1985).The light where A is the exposed filter area, and V is the volume of air sampled.ATN is the light attenuation defined as where I o and I are the light intensity incident and transmitted through the filter, respectively.Light attenuation is measured ever minutes until I/I o decreases to half the initial value, at three wavelengths of 450, 550, and 650 nm.Intercalibration with the AE33 aethalometer (Magee Scientific) was performed at the GAW station of N.S.C.R. Demokritos (Athens), simultaneously with long-term measurements of ambient aerosols, as described elsewhere (Popovicheva et al., 2017).The level of uncertainty (1-sigma) of EBC measurements is 30 ng m -3 for 6-minute integrated sampling.MiniVol ™ TAS samplers with PM 10 and PM 2.5 impactors (Airmetrics) were used for aerosol sampling at a flow rate of 5 L min -1 on quartz fiber (Pall Science) and Teflon (Zeflour) filters of a diameter 47 mm.A custom-made two-stage cascade impactor with aerodynamic cut-off sizes of 10 µm and 630 nm at the first and second stage, respectively, was utilized to collect particles on metal (Ti) substrates.Instrumentation was installed at the site Ba Vi on a forest-fire watch tower (20 m above ground).24 h samples were collected by MiniVol samplers in the period from 27 May to 14 June of 2015, excluding the time of the lower PM mass on 5 and 6 June, when the sample was collected during two days.Impactor collection was performed during a period of one hour, resulting in visible spots on the Ti substrates.Meteorological data (temperature (T), relative humidity (RH), wind direction (WD), wind speed (WS), and precipitation) was provided by Vietnam Metservices.
During on-field campaigns, the samples were taken in the smoke plume at a distance of 2 meters downwind from the crop residue piles produced by farmers according to typical agricultural burning practices.The smoldering samples were collected during the smoldering phase of the pile burn.Collection of complete samples has encompassed all phases of combustion, from fire initiation through open flame to smoldering, in accordance with visual observations.Because of the variability of smoke plume from the piles repeated sampling was done.A list of representative samples collected on quartz filters is presented in Table 1.
The SPARTAN site was equipped with a nephelometer continuously running concurrent with an automated PM 2.5 sampler, operated according to SPARTAN standard operation procedure (SOP) (Snider et al., 2015).For each filter, the sampling was timed at regular 24h intervals through out a 9-day period.Dry (35% RH) hourly PM 2.5 estimates are provided by combining scatter at ambient RH with filter mass and chemical composition information.

Analytical Methodology
Off-line examination of light attenuation of particles deposited on quartz filter samples was performed using the multiple wavelength light transmission instrument (transmissometer) based on the methodology of Kirchstetter et al. (2004).The intensity of light transmitted through quartz filters was measured at seven wavelengths from the near-ultraviolet to near-infrared spectral region.The dependence of the attenuation (ATN) on the wavelength λ is parameterized using a power law relationship: where the Absorption Angstrom Exponent (AAE) is a measure of strength of the spectral variation of aerosol light absorption.BC produced by high-temperature combustion sources fit within the Rayleigh scattering regime for nearvisible wavelengths with theoretical λ -1 relationship (Bond and Bergstrom, 2006).Gravimetrical analyses were performed using Standard Operation Procedure (SOP).Teflon and quartz filter samples were conditioned for about 24 h in a temperature and humidity controlled room (T= 20 ± 5°C, RH = 40 ± 10%) and weighted before and after sampling by a 5-digit analytical balance.
Organic (OC) and elemental carbon (EC) were measured by thermo-optical transmittance (TOT) instrument (Lab OC-EC Aerosol Analyzer, Sunset Laboratory, Inc.).Quartz fiber filters were pre-heated at 550°C for 6 hours to remove possible OC contaminations.The samples were heated first in He and then in an mixture of 2% O 2 in He, using controlled heating ramps of the EUSAAR_2 thermal protocol (Cavalli et al., 2010).Charring correction was applied by monitoring the sample transmittance throughout the heating process.Total carbon (TC) was determined as TC = EC + OC.The analytical limit of detection (LOD) for the TOT instrument was 0.2 µg C cm -2 .
Water soluble components were measured by capillary electrophoresis at the Capel 103 system (Lumex, Russia) with UF detection.Filter samples were extracted in 5 mL of distillated water by ultrasonic agitation for 45 min and then filtered.Inorganic anions (SO 4 2-NO 3 -, Cl -, F -, and PO 4 3-) and cations (Na + , NH 4 + , K + , Mg 2+ , and Ca 2+ ) as well as organic ions (HCOO -, CH 3 COO -, (COO -) 2 ) were measured in aqueous extracts with the relative standard deviation of 10%.A mixture of benzimidazole, tartaric acid, and 18crown-6-ether was used as an electrolyte for cation measurements.Anions were analysed in a chromate buffer prepared from chromium oxide (VI), diethanolamine, and cetyltrimethylammonium hydroxide solution.The LOD for ion concentrations was in the range of 0.1-0.5 mg L -1 .
Fourier Transform Infrared (FTIR) spectroscopy measures the absorbance associated with frequency of the vibrational mode of the bond, thus providing the functional groups representing the various classes of organic compounds in the entire aerosol composition.FTIR spectra of filter samples were collected with an IRPrestige-21 spectrometer (Shimadzu, Japan) in a diffuse reflection mode, as described elsewhere (Popovicheva et al., 2015a, b).Spectra were recorded in the range from 450 to 4000 cm -1 with resolution of 4 cm -1 , with the number of scans being 100.IR Solution software was applied to subtract the FTIR spectrum of blank substrates, to correct the baseline absorbance as well as to perform the Kubelka-Munk (K-M) conversion of the reflectance spectrum into one which is proportional to the sample concentration.The FTIR spectra were presented in K-M normalized on the maximum band intensity.Identification of absorption bands was carried out according to Shimadzu FTIR database and Coates' practical approach (Coates, 2000), as well as using authentic chemical standards.
Round punches of filter samples (about 13 mm in diameter) were prepared with SMART STORE (XearPro) to be measured by total reflection X-ray fluorescence (TXRF).
Measurements were performed for 600 s with a Bruker S2 PICOFOX spectrometer equipped with molybdenum X-rays tube operating at 50 kV and 750 µA and Silicon-Drift Detector (SDD) (SDD) (Borgese et al. 2011(Borgese et al. , 2012)).Detection limits were 1 mg for K, Ca, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, As, Sr, Ba, Tl, Pb and Bi.Lighter elements, like Si, cannot be identified due to absorption effects.Semiquantitative analysis was performed on the measured spectrum, by calculating each element peak area after background subtraction.Data were processed to obtain the relative mass percentage of identified elements.
The SPARTAN network supplies the methodology for characterization of filter samples collected to infer PM 2.5 composition, described elsewhere (Snider et al., 2016).They analyzed the water-soluble ions NO 3 -, SO 4 2-, NH 4 + , K + , and Na + by ion chromatography (IC) with a Dionex ICS-1100 (anions) and a Dionex ICS-1000 (cations) instrument (Thermo Scientific).Equivalent black carbon (EBC) was defined as the black carbon content of PTFE filters derived via surface reflectance using the Diffusion Systems EEL43M smoke reflectometer.
For individual particle analyses LEO 1430-vp (Karl Zeiss) field emission scanning electron microscope (SEM) with a spatial resolution of 7 nm, equipped with an Oxford energy dispersive detector (INCA), was used.Energy dispersion X-ray (EDX) spectra for Z elements (Z ≥ 5) were recorded.Samples were studied in the high vacuum mode at 10 kV acceleration voltage and a beam current of 1 nA, as described elsewhere (Popovicheva et al., 2012).Briefly, 500-1000 individual particles with a diameter from 0.1 to 3 µm were measured.This number was considered to be sufficient for obtaining a representative overview of groups and types of particles (Liu et al., 2000;Popovicheva et al., 2014a, b).EDX analysis yielded a data matrix containing elements at weight concentrations above the detection limit 0.1 wt%.Cluster analysis separated individual particles into characteristic groups of similar chemical composition and morphology, as close as possible to physico-chemically identifiable particles.Naming of particle groups is based on both morphological features and most abundant elements after C and O.

On-Field Burning, Traffic and Cooking Sources
High PM 10 mass concentrations and carbon content were found in smoke from rice straw burning collected downwind from a pile burn (Table 1).In the complete phase, PM 10 mass of 16.0 ± 1.5 mg m -3 was observed.TC contributed 41 ± 11% to PM 10 , in line with carbon content reported for various BB sources (Reid et al., 2005;Oanh et al., 2011;Popovicheva et al., 2014a).Smoke mass was dominated by OC and ionic species such as HCOO -, K + , NH 4 + , and Cl -, with contributions from certain elements, as shown by the average mass balance in Fig. 2. Organic matter (OM), taking into account heteroatoms in organic compounds, was determined by choosing an OM/OC conversion factor of 1.4 (Engling et al., 2014).OC was found to comprise 40 ±11% of PM 10 .OC contribution to TC was as high as 97% for complete phase and even higher (99%) for smoldering phase combustion, indicating dominance of smoldering during combustion of rice straw piles.Prevailing ambient conditions such as high RH up to 87% and T ~32°C (Table 1), influenced the combustion to proceed preferably in the smoldering phase.Consequently, EC accounted for only 1.5 ± 0.3% of PM 10 mass.In both phases, OC/EC ratios Fig. 2. Average mass balance of carbon fraction, ions, and elements in PM 10 smoke from rice straw burning in the complete phase on fields in the Ba Vi region.Percentage of components is indicated.of 46 ± 19 (complete) and 145 (smoldering) were obtained, indicating preferable condensation of volatilized organics as mechanism of particle formation at low-temperature combustion (Reid et al., 2005;Popovicheva et al., 2015c).During controlled field burning of rice straw somewhat smaller OC/EC ratios of 10 ± 2 and 5.8 ± 3.2 were obtained by Engling et al. (2009) and Oanh et al. (2011), respectively.
Examination of OC and EC data from traffic and cooking emissions shows a significant difference in comparison with rice straw burning (Table 1).At the traffic site, TC averaged over the sampling period was 106 µg m -3 .Around 220-240 motorbikes and 20 cars per minutes passed the site, clearly marking it as a traffic-related source of particulate emissions.OC contributed 92% to TC, while EC/TC and OC/EC ratios were 0.08 and 11, respectively, similar to those observed in megacities (Chen et al., 2014;Popovicheva et al., 2014a), indicating higher contribution of EC from fossil fuel combustion.TC averaged over the sampling period in a typical restaurant was 51.2 µg m -3 .Coal combustion periodically changed from smoldering to flaming phase.OC constituted 96% of TC, and the OC/EC ratio was 22, meaning that organic species were more prevalent in cooking than traffic emissions, yet still less than in rice-straw burning.
Rice straw smoke samples appeared brown rather than black, rendering them as absorbers of solar radiation with stronger wavelength selectivity in the blue and ultraviolet regions.Spectral absorption of BB shows the combined impact of both BC absorbing from 670 nm down to 500 nm and BrC which increases the absorption below 500 nm (Kirchstetter et al., 2004).Spectral dependence of the light attenuation (ATN) for rice straw smoke in the complete burning phase is shown in Fig. 3. On logarithmic scale, it well approximated by a power law relationship (3).The Absorption Angstrom Exponent (AAE) is obtained as the slope of the linear regression with R 2 around 0.9.AAE was found to be 3.25, demonstrating a strong wavelength dependence of the predominant BrC absorbing component in rice straw burning emissions.Kirchstetter et al. (2004) obtained a similar high value of 2.5 for smoke of savanna fires.
Weak spectral dependence of ATN with AAE around 1 was found for diesel soot and urban aerosols (Kirchstetter et al., 2004) due to dominant BC derived from hightemperature combustion of fossil fuels.Similar results for traffic and cooking emissions in Hanoi city were obtained, with AAE equal to 1.3 and 0.96, respectively, meaning that both traffic and cooking sources produce less BrC and more EC than rice straw burning.
High mass concentrations of EBC were recorded at the traffic site (4.8 ± 0.9 µg m -3 ), associated with intensive pollution from diesel and gasoline vehicle emissions.Even higher EBC, 6.2 ± 0.9 µg m -3 , was obtained for cooking emissions.At the same time EC measured on the filter samples from both sources were found to be 1.5 times higher than EBC, similar to values obtained in other megacities (Popovicheva et al., 2014a).
Classes of organic/inorganic compounds in aerosol composition were inferred from analyses of FTIR absorbance of rice straw burning emissions (Fig. 4(a)).The most intensive feature is a wide band in the region of 3550-3130 cm -1 relating to hydroxyl O-H stretches in alcohols, phenols, and acids.As O-H groups are hydrophilic functionalities responsible for effective water uptake (Kuznetsov et al., 2003), the rice straw burning produces a more hygroscopic aerosol than other types of biomass (Petters et al., 2009).This wide band can contain the vibrations of ammonium NH 4 + around 3200 cm -1 which are not well resolved.Prominent bands relate to aliphatic C-C-H Fig. 3. Spectral dependence of light attenuation of smoke at the Ba Vi site in June 2015, of on-field rice straw burning in complete phase, traffic, and cooking sources.Absorption Angstrom Exponent (AAE) is a slope of the linear regression in logarithmic scale of a power law regression, shown in Table 1.stretches (2922-2850 cm -1 ) and bends (1454 cm -1 ) in alkanes, acid carboxyl C(O)OH (1707 cm -1 ) and carboxylate RC(O)O (1596 cm -1 ) groups in carboxylic acids, ketones/ aldehydes, and RCOO -ions in carboxylic acid salts.Slightly prominent asymmetric (1505 cm -1 ) stretches in aromatic nitro compounds Ar-NO 2 should be noted as well.Inorganic compound vibrations were also identified such as sulfates SO 4 2-(1047 and 624 cm -1 ) and carbonates CO 3 2-in various salts.The whole functional pattern of rice straw burning aerosol obtained on fields in the Ba Vi region demonstrates similar features to emissions from on-field cassava and corn bush burning in Son La province (Popovicheva et al., 2016) due to dominant impact of similar burning practices used by farmers on agricultural fields.It is noteworthy that intensive hydrophilic O-H functionalities are explicitly characteristic of rice straw burning emissions.
Formate HCOO -and inorganic ions comprised 10.7 and 6.5% of PM 10 mass of rice straw burning emissions, respectively (Fig. 2).K + , Cl -, and NH 4 + were the main inorganic ion species, well in accordance with the ion composition obtained for rice straw field burning (Engling et al., 2009).Relative concentrations of organic and inorganic ions with respect to average total water-soluble ions (TWSI) are shown in Fig. 5. K + ions are widely used as a marker of BB, with PM fractions between 0.5% and 6% depending on biomass type (Reid et al., 2005;Alves et al., 2010).In complete phase of rice straw burning K + ions comprised 2.4% and 14% of PM 10 and TWSI, respectively.Relatively high concentrations of NH 4 + are also noteworthy, being a unique feature observed at particulate characterization of experimental rice straw fires (Engling et al., 2009), that were not obtained in on-field emissions in Son La province where wood, corn, and leaves were burned after harvesting.
TWSI in traffic and cooking emissions was found to be 120 and 140 µg m -3 , respectively (Table 1).Sulfate ions were predominant, comprising 24 and 32% of TWSI, respectively (Fig. 5), as expected from fossil fuel combustion and especially in the traffic samples collected road-side in large cities (Hai and Oanh, 2013).Together with аmmonium and nitrates, sulfate salts dominate the regional pollution spectrum of large cities.Na + and Cl -account for a significant part of TWSI in both emissions, while Ca 2+ , Mg 2+ , and K + are representative of road and soil dust.High concentrations of K + and Ca 2+ found in traffic emissions indicate the impact of lubrication oil used in vehicles (Popovicheva et al., 2014b).Finally, acetate ions dominate among organic ions in both emissions, in contrast to rice straw burning emissions in which formate is the most abundant carboxylate ion.
The remaining unidentified aerosol components in the mass closure of rice straw burning emissions, around 25% may be comprised of water and elements as part of dust and soil minerals evolved by hot convection during fires.The biggest relative abundance among the elements was observed for Fe (86%), while other elements such as Ni, Zn, and Mn comprised 4, 7, and 1.8%, respectively (Fig. 6).In traffic and cooking source emissions, the composition of heavy metals was found to be similar, with additional Ti and Cr, accounting for 3 and 4%, respectively (Fig. 6).

Source Microstructure
Analysis of individual particles reveals the complex morphological and chemical composition of carbonaceous particles internally and externally mixed with inorganic fly ash and dust, Fig. 7 Particle-phase compounds collected inside the smoke plume from a rice straw pile burn produced a quasi-liquid substance with 80 and 20 wt% of C and O, respectively, a composition typical for tar.The abundance of these quasi-liquid particles is likely due to condensation of semi volatile organics in dense smoke.Groups of particles obtained by cluster analyses are presented in Fig. 8.The most abundant group, 45.1%, contains mainly C (87%) and O (12%) with negligible amounts of other elements (< 3 wt%)."Tar ball" particles appeared with roughly spherical morphology (Fig. 7.1), formed by bimolecular homogeneous nucleation of polar organic matter with water vapor (Hand et al., 2005); therefore, the tar balls are assigned to Group Organic.This group is representative of the smoldering phase, and was found to comprise 60% of smoke particles in a combustion chamber (Popovicheva et al., 2015c) and can serve as micromarker of smoldering BB.In on-field emissions in Son La province the abundance of Group Organic was 58% (Popovicheva et al., 2016).Other particles produced during rice straw burning are rich in soil minerals, which are evolved by hot convection during fires (Kavouras et al., 2012).Group Si, Ca, and Fe-rich of irregularly shaped particles containing aluminosilicates, calcium oxides, and iron oxides, respectively, was found (Fig. 8).K was distributed over 10% of smoke particles (Fig. 9), and half of K-containing particles also contain chlorine and sulfur in form of potassium chloride and sulfates, in accordance with measured K + , Cl -, and SO 4 2-ions.

Fig. 7.
Representative EDX spectra and micrographs of 1) tar ball of Group Organic in rice straw burning; 2) soot agglomerate of Group Soot and 3) internal mixed aerosol in traffic emission; 4) soot particles encapsulated with organic species of Group Soot/Organic in cooking emission; 5) tar ball in Group Organic and 6) potassium chloride in Group K-rich at the Ba Vi site.Ti and Cu is substrate artifact.
Traffic emissions in the Hanoi city revealed significantly different microstructure (Fig. 8).The Group Soot, 13.8%, containing mainly C and O, is composed of agglomerates of small primary particles (Fig. 7(b)), with typical morphology of soot emitted by vehicle engines (Popovicheva et al., 2014b(Popovicheva et al., , 2015a)).In urban environments of megacities approximately 17% of all ambient particles were found to belong to Group Soot, being mainly of traffic origin (Diapouli et al., 2014;Popovicheva et al., 2014b).The most abundant is Group Fe-rich, 38.5%, and Ca-rich, 18.5%, of iron oxides and calcium sulfates, respectively, originating from street dust and construction.The distinctive feature of traffic emissions is enhanced abundance of Ca, Fe, and Si (Fig. 9), and high concentrations of complex externally mixed aerosols (Fig. 7.3).In cooking emissions, Group Soot/Organic, 20.6%, present in form of soot particles encapsulated with organic species (Fig. 7.4), is representative, in agreement with higher OC/EC ratios found for this emission type.S dominates among all elements (Fig. 9) in sulfate of Group S-rich, that is well in accordance with the high concentrations of SO 4 2-ions obtained in cooking emission (Fig. 5).

Regional Aerosol in Ba Vi Region PM 10 Evolution
A view from the Ba Vi site during the measurement campaign demonstrates the many fires in the region and the resulting degradation of visibility (Fig. 1).Evolution of daily mass concentrations from 28 May to 14 June 2015 is shown in Fig. 10(a It exceeded the WHO (World Health Organization) 24h standard guideline value of 50 µg m -3 on 11 of total 13 measurement days, indicating that the PM level was highly impacted by pollution.The average PM 10 concentration exceeded that 67 ± 32 µg m -3 reported for a rural site in Vinh Phuc province, northwest of Hanoi city, which was not affected by BB activities (Co et al., 2014).
According to classification of smoke during BB periods (Popovicheva et al., 2016), PM 2.5 mass ranged from low levels ≤ 40 µg m -3 to moderate ~80 µg m -3 , and subsequent higher smoke intensity.Since PM 2.5 mass concentrations are lower than PM 10 with typical ratios of 0.72-0.74(Co et al., 2014), it can be concluded that PM 10 at the Ba Vi site ranged from low levels of ≤ 54 µg m -3 to high smoke intensity above 109 µg m -3 .
The weather during the sampling period was characterized by high temperatures, varying from 27 to 38°C (Table 1).Wind speeds were low and quite stable, ranging from 1.5 to 2.5 m s -1 .Analysis of daily wind roses shows that predominant wind direction between 27 May and 14 June 2015 was northeast and southeast, typical for the period of monsoon season changing during this time of the year (Hai and Oanh, 2013).There was no correlation between daily PM 10 and wind speed observed.Unlike wind speed/direction, the relative humidity (RH) during the sampling period appeared to be inversely proportional to daily PM, varying significantly from 75 to 90% (Fig. 10).Intensive fogs with high RH impacted smoke, leading to significant decrease of PM 10 levels on 5 and 6 June.In addition, precipitation caused the PM 10 levels to decrease significantly on certain days (Fig. 10).After a few days of continuous precipitation, from 1 to 6 June, the smoke intensity dropped down to the lowest value with PM 10 mass concentrations of ~39 µg m -3 .120 h BWTs demonstrate the long-range transport of air masses from Indo-China (Burma, Thailand, Laos) and west of Vietnam during the sampling period.According to Fig. 9. Abundance of elements over particles of on-field rice straw burning, traffic and cooking emissions, and regional smoke at the Ba Vi site.FIRMS WEB Fire Mapper and satellite image data source UMD MCD14ML (C6), biggest fires were observed on the border of Laos and Vietnam, south of the Ba Vi region, from 24 May to 31 May.On 1 June fires nearly disappeared until 3-4 June and increased again on 8 June.Fire evolution combined with air mass trajectories allows the postulation that PM 10 maximum observed on 3 and 11 June was influenced by fire-polluted air mass transported from the south of Vietnam and Laos.
Background PM 10 level averaged over the sampling period from 1 to 6 June 2016, was found to be near 30 µg m -3 .It was time when post-harvesting fires had not started yet.Since the aerosol loading during the haze episode of 2015 even on the days with lower PM 10 concentrations was higher than 30 µg m -3 , it can be suggested that the regional aerosol burden in the Ba Vi region in spring of 2015 was significantly influenced by agricultural fires in the surrounding area.

Aerosol Chemistry
The average TC concentration from 28 May to 14 June 2015 was 18.5 µg m -3 .The highest TC level of 42 µg m -3 was recorded on 4 June.During the entire sampling period OC contributed between 88 and 95% to TC, in line with observations of predominant smoldering fire emissions on agricultural fields.The evolution of aerosol constituents (OC and EC) was found to be in relation to smoke intensity (Fig. 10(b)), except on 12-13 June when precipitation was significant.OC comprised around 28% of PM 10 on 4 June and decreased to 7.5%, suggesting almost no BB impact on 5 and 6 June in comparison with other days of the haze.EC levels during the BB season were lower than OC levels, yet correlated with PM 10 (Fig. 10(b)).The highest EC value was 4.9 µg m -3 on 4 June, comprising 7.8% of PM 10 .Low EC levels indicate that smoldering emissions were dominant during the entire sampling period.Low EC/TC ratios from 0.045 to 0.12 revealed EC to be a minor carbonaceous atmospheric pollutant in the smoky atmosphere (Fig. 10(c)).The lowest EC level occurred on 7 June when the lowest PM 10 mass was observed, probably because of almost all fires being extinguished on the surrounding fields following the sustained rain period in the previous days.The OC/EC ratios in PM 10 were above 7.5 and as high as 20 during sampling period, with the highest value observed on the same day (7 June) when EC emissions were radically suppressed (Fig. 10(b)).On average, the OC/EC ratio was 13.2 ± 3.8.Remembering that high OC/EC ratios are characteristic for on-field emissions (Table 1), it can indicate that smoldering fires, dominated by high OC, prevailed during the smoke period in the Ba Vi region.Additionally, meteorological conditions (low wind speed, high temperatures and strong solar radiation) could favor formation of secondary organic aerosols due to condensation of heavier hydrocarbons generated by acid-catalyzed oxidation and photochemical reactions in the smoke plume (Alves et al., 2010), that could have influenced high daily OC/EC ratios as well.
Real-time EBC on selected days of June of 2015 are shown in Fig. 11, varying from 0.5 to 8 µg m -3 .The highest concentration was recorded on 4 June, in good agreement with the highest EC level observed on the same day.For comparison, daily BC levels in PM 10 averaged at 4.1 ± 1.4 µg m -3 in Vinh Phuc province, not affected by BB (Co et al., 2014).AAE, obtained from attenuation spectral dependence, was 1.4 ± 0.2 for regional aerosols (Fig. 3), well in accordance with lower OC/EC ratios compared to emissions on fields.The most prominent FTIR absorption bands on days with the lowest PM 10 levels of 40-60 µg m -3 (5-7 June) were of N-H (1602 cm -1 ) and C-N (1308 cm -1 ) groups in aromatic amino compounds (Fig. 4(b)).They indicate the abundance of biogenic groups in bioaerosols (Coury and Dilner, 2009), which is a reasonable feature of ambient environments after and during rain periods.The presence of aromatics in low smoke aerosols is confirmed by C = C-H (810 cm -1 ) bending and C = C-H stretching (3027 cm -1 ) vibrations, probably due to the regional pollution by aromatic compounds.Since aromatic stretch absorption was identified only in a functional pattern of coal cooking activities (Fig. 4(a)), it allows the identification of cooking (or other coal combustion) emissions of particulate matter in that period.The bands at 3209 and 1194 cm -1 in low smoke identify sulfate ammonium, also indicating the possible impact of cooking emissions and secondary aerosol formation at regional level.
Fire emissions during the haze episode impact aerosol chemistry significantly.The spectrum collected on 2 June at a PM 10 level of 165 µg m -3 demonstrates strong bands of hydroxyls (3383 cm -1 ), carboxylates (1595 cm -1 ), and aromatic nitro groups (1489 cm -1 , Fig. 4(b)), similar to the characteristic absorption in the functional pattern of rice straw burning (Fig. 4(a)), whereas the bands of C-N and C = C-H groups, presumably from biogenic and cooking sources, were significantly decreased.
Mass concentrations of three organic and nine inorganic ions during the haze episode were summarized in Fig. 12.On days of highest smoke, TWSI comprised 47% of PM 10 ,  and correlated well with the smoke evolution on other days.The highest levels, (> 5 µm m -3 ), are found for CH 3 COO -, Ca 2+ , Na + , and Cl -ions.The dominant abundance of organic acid ions confirms the FTIR identification of carboxylate ions in the pattern of rice straw burning and ambient aerosols (Fig. 4).High levels of Ca 2+ ions were observed preferably in coarse particles in Son La province (Popovicheva et al., 2016), probably associated with soil dust resuspension due to long-lasting smoldering on the fields.Na + and Cl -could be linked to aged sea salt as the Ba Vi site is located about 100 km west of the Bac Bo (Tonkin) Gulf (Fig. 1).Cl -ions were also prominent in the pattern of smoke emissions from rice straw burning, and therefore they could impact the regional abundance due to intensive fires.The relative contribution of K + to PM 10 was high, on average from 31 May to 9 June as high as 4%, and up to 6.9% on days of high smoke (Fig. 12), similar to observations during wildfire smoke events (Alves et al., 2010;Huang et al., 2013).
Observations of Indonesian BB episodes showed all inorganic ions at higher levels than in the clear period (Engling et al., 2014).In this study, the analyses of the ion abundance in dependence on the level of smoke intensity showed that K + , Mg 2+ , Cl -, SO 4 2 , (COO) 2 -, and CH3COO - were at highest concentrations on days with highest smoke, confirming the strong influence of BB activities on the local aerosol burden.Although it should be noted that secondary inorganic components, SO 4 2-, NO 3 -, and NH 4 + , were not always related to the smoke level, because only 30-40% of sulfur in the Vietnam atmosphere originated from domestic sources, with the rest contributed mainly from southern China and Thailand (Huang et al., 2013).
Heavy metals Fe, Ni, Ti, Cr, Zn, Cu, Mn, and Pb were found in PM 10 , and their relative concentrations were summarized in Fig. 6.Fe, Mn, Ti, and Zn were measured as the most abundant elements in north-east of Hanoi (Gatari et al., 2006).Since Fe, Mn, and Ti are mainly distributed in the coarse fraction of soil-derived particles, their main source in the Ba Vi region could be soil dust remobilized by fires.The abundance of hazardous trace metals such as Ni, Cr, and Zn (Hai and Oanh, 2013) were enhanced by anthropogenic dust, and their variability was measured at the Ba Vi site to be higher than in any other source analyzed.The presence of elements such as Fe, Cr, Pb, and Mn may also be linked to coal combustion and industrial activities (Hai and Oanh, 2013).
Individual particle analyses of regional aerosols exhibit a large variability of sizes and types, including spherical, chain, and crystalline mineral particles.C and O-containing particles are found in roughly spherical or liquid-like shape (Fig. 8.5), separated in Group Organic, thus constituting a micromarker of rice straw burning in regional smoke.Soot chain agglomerates were rarely observed.The abundance of Group Organic varied from 20 to 39% with smoke intensity, similar to observations in Son La province where it was found in the range 12-67% (Popovicheva et al., 2016).Therefore, it can be confirmed that tar balls, ubiquitously observed in Asian outflow with spectral absorption features of BrC (Alexander et al., 2008), originate from intensive burning practices such as those observed in this study.
Grouping of particles sampled on 4 June 2015 with the highest abundance of carbonaceous aerosols was shown in Fig. 8. Fly ash and dust are observed in Groups Si-, Ca-, Fe-rich with solid irregular morphology, similar to those found in on-field emission in the Ba Vi region and Son La province during BB periods.Group Si-rich of various K, Ca, Mg-aluminosilicates of soil origin was found to be most abundant.Groups Ca-, Na-, and Fe-rich containing calcium oxides, sulfates as well as chlorides, iron oxides, and Fealuminosilicates were less abundant.Group K-rich is appeared in regional smoke where K is either condensed on carbonaceous particles or externally mixed potassium sulfate and chloride (Fig. 7.6).Since potassium in ionic form (K + ) is acting as a BB marker, Group K-rich is considered as a micromarker of smoke during periods of intensive fires.
During sampling days from 1 to 6 June 2016 OC and EC were found at levels of 12.6 ± 4.2 µg m -3 and 0.59 ± 0.58 µg m -3 , respectively, being the lowest compared with the haze episode of 2015.OC/EC and EC/TC ratios were found to be 12.5 ± 2 and 0.07 ± 0.02, respectively, closer to the traffic emissions than rice straw burning in the period of high smoke (Table 1).This indicates a contribution of EC to regional aerosol preferably from traffic, when no wildfires were observed.Organic compounds at that time were dominated by aromatic nitro compounds and alkanes, according to absorption of Ar-NO 2 (1542 cm -1 ) and C-C-H (2914-2854 cm -1 ) groups (Fig. 4(b)).The characteristic bands in background spectra were different from functional patterns of rice straw burning, additionally indicating the absence of the impact of on-field burning at that period.A large band of sulfates in Ca, Na, and Mg salts at 678 cm -1 emphasize secondary inorganic compounds.Background ion concentrations during sampling days from 1 to 6 June 2016 were relatively low in comparison with haze period of 2015 (Fig. 12).

Aerosol Evolution and Composition in Hanoi
Hanoi city has much higher air pollution levels than other cities in SE Asia (Huang et al., 2013).K + ion concentrations in Hoa Binh (rural city nearby) were almost a factor of 2.5 lower than those in Hanoi, although the two sites are close to each other.PM 2.5 composition from the SPARTAN site in Hanoi from 22 May to 25 July of 2015 is shown in Fig. 13.The trend of PM 2.5 decreasing can be a seasonal reduction of total emissions (Hai and Oanh, 2013).The average PM 2.5 from 22 May to 19 June of 2015 was 45 ± 18 µg m -3 , comparable to that for the haze episode of our study in Ba Vi province.The average OC/EC ratio (estimated as ratio of carbon residual material to EBC) at that time was 4.6, being 3 times less than during the haze episode at the Ba Vi site.K + ion fractions in PM 2.5 from 31 May to 9 June were 2%, while they were 5.5% during the same period at the Ba Vi site.
The Ba Vi region could be considered as background for the Hanoi area, although during BB seasons this region is affected by local sources.It is obvious that the sources for the haze episode in the Ba Vi region were different from localized smoke occurring in the urban area due to the trapping of anthropogenic emissions.However, the spread of smoke haze resulting from BB could lead to the increase of regional aerosol level.Dense smoke from agricultural fires could be transported from nearby areas to the Hanoi city center, causing the atmosphere over Hanoi to be polluted by smoke haze to some extent.According to air mass trajectories, Hanoi may have been affected by BB from the Ba Vi region on 11 June 2015.

CONCLUSIONS
Chemical characterization of ambient aerosols in the Ba Vi region during a haze period in the post-harvest burning season shows a predominant impact of biomass burning on the ambient PM 10 mass and its constituents.PM 10 levels evolved from low to high smoke intensity, influenced by local meteorology and fire-polluted mass transported from the south of Vietnam and Laos.High OC/TC and OC/EC ratios indicate the impact of rice straw smoldering, supported by distinguishable absorption of BrC.A functional pattern (hydroxyl, aliphatic, carbonyl, carboxylate, and nitro groups) is introduced as a functional marker of rice straw burning.At high smoke its spectral features become pronounced, being an effective indication of BB as the major regional source for the haze at the molecular level.High concentrations of organic and inorganic ions were dominated by CH 3 COO -, Ca 2+ , Na + , and Cl -ions, where the relative contribution of the BB marker K + to PM 10 mass is similar to wildfire smoke observed in other regions across the world.The main source of heavy metals such as Fe, Mn, and Ti was soil dust remobilized by fires; hazardous Ni, Cr, Zn, and Pb were abundant as well.Microstructure of regional aerosols demonstrates smoke-influenced types of particles at microscopic level, indicating the micromarkers of rice straw burning by extended tar ball and K-polluted groups.Evaluating the multipollutant regional aerosol in the Ba Vi region and possible effects on Hanoi city, revealed that agricultural fires from the surrounding fields produce environmentally-dangerous compounds as an alarming signal for human health and air quality.These findings indicate a clear need for establishing a network of longterm air quality monitoring stations in the Ba Vi region.
Fig. 1. a) Rice-straw biomass burning activities on fields of Ba Vi region, b) and c) location of the Ba Vi site (21°3'N, 105°21'E) in Ba Vi region in Hanoi, Vietnam.

Fig. 4 .
Fig. 4. FTIR spectra of a) on-field rice straw burning in smoldering phase, cooking and traffic emissions; b) low (black) and high (red) smoke at the Ba Vi site in 2015 and 2016.Absorption bands are indicated.

Fig. 5 .
Fig. 5. Relative concentrations of water-soluble ions with respect to total ion concentration in percentages in a) rice straw burning, b) traffic, and c) cooking emissions.

Fig. 6 .
Fig. 6.Relative concentrations of heavy metals at the Ba Vi site in May-June 2015, in rice straw burning, traffic and cooking emissions.

Fig. 8 .
Fig. 8. Grouping of particles from a) on-field rice straw burning, b) traffic and c) cooking emissions, and d) regional smoke at the Ba Vi site.Naming, relative abundance, and averaged weight percentage of elements in each group are indicated.

Fig. 10 .
Fig. 10.24 h PM 10 mass concentrations, a) precipitation and relative humidity, b) OC and EC, c) OC/EC and EC/TC ratios during the measurement campaign at the Ba Vi site in 2015.

Fig. 11 .
Fig. 11.Equivalent black carbon at the Ba Vi site in June 2015.

Fig. 12 .
Fig. 12. 24 h PM 10 mass concentrations, organic and inorganic ions on the Ba Vi site in 2015 and 2016.

Fig. 13 .
Fig. 13.PM 2.5 mass composition from the SPARTAN site in Hanoi city.Time duration of haze episodes observed in Bavi province is indicated by shadow.

Table 1 .
Sampling of on-field rice straw burning and urban emissions, and at Ba Vi site in June 2016: date, temperature (T), relative humidity (RH), PM mass, organic (OC) and elemental carbon (EC), OC/EC and EC/TC ratios, absorption Angstrom exponent (AAE), and total water-soluble ions (TWSI).