Processing of biomass burning aerosol in the Eastern Mediterranean during summertime

Introduction Conclusions References


Conclusions References
Tables Figures

Introduction
Organic species represent a significant and often major mass fraction of submicron aerosol.At continental midlatitudes, organic material contributes ∼ 20-50 % of the total fine particulate mass while in tropical forested areas this contribution can be as high as 90 % (Kanakidou et al., 2000;Zhang et al., 2007).During the last few years, factor analysis of measured organic mass spectra from aerosol mass spectrometry (AMS) data has enabled the deconvolution of organic aerosol (OA) to different components Introduction

Conclusions References
Tables Figures

Back Close
Full based on composition (Zhang et al., 2011).This has provided valuable insights on the source and transformation processes of OA in the atmosphere (Lanz et al., 2008(Lanz et al., , 2010;;Hildebrandt et al., 2010Hildebrandt et al., , 2011;;Cubison et al., 2011;Aiken et al., 2011;He et al., 2011;Ulbrich et al., 2009).Different components exhibit distinct mass spectra, with the two most common ones being the hydrocarbon-like OA (HOA) and oxygenated OA (OOA) (Zhang et al., 2007).In some cases the OOA component is further resolved into two subtypes that have different degrees of volatility, namely the low-volatility OOA (LV-OOA) and semi-volatile OOA (SV-OOA) (Jimenez et al., 2009;Ng et al., 2010).Other components are sometimes identified, e.g.biomass burning OA (BBOA) (Aiken et al., 2009;Cotrell et al., 2008;He et al., 2010;Huang et al., 2011) and cooking OA (COA) (He et al., 2010;Mohr et al., 2009;Allan et al., 2010).The differentiation of the factors is based on the mass spectra of the components, mainly in terms of two mass fragments (CO + 2 at m/z 44 and C 2 H 3 O + at m/z 43) (Ng et al., 2010) (Schneider et al., 2006;Alfarra et al., 2007).The enhanced signal at m/z 60 is associated with levoglucosan and other similar species (e.g.mannosan and galactosan).Nevertheless, the sum of levoglucosan, mannosan and galactosan alone can account only for around 1/10 or less of the total signal of m/z 60 in BBOA.This indicates that the majority of it originates from different molecules with the same fragmentation pattern as levoglucosan in the AMS (Aiken et al., 2009;Lee et al., 2010b).Given that levoglucosan is a monomer that results from the thermochemical decomposition of cellulose at elevated temperatures in the absence of oxygen, such species may include dimers and trimmers of similar molecules (Cubison et al., 2011).Laboratory-based studies (Hennigan et al., 2010;Hoffmann et al., 2010;Kessler et al., 2010) have estimated an atmospheric oxidation lifetime for levoglucosan in typical conditions that ranges from 15 h to a few days.Nevertheless, the extent of degradation of the AMS tracer signals f 60 and f 73 after emission is not well known.f 60 has been reported to decrease during photochemical processing of BBOA and from addition of Introduction

Conclusions References
Tables Figures

Back Close
Full secondary organic aerosol (Huffman et al., 2009).It is unclear whether commonly-used tracers persist long enough in aged, atmospheric processed BBOA, to serve as effective markers; f 60 is currently thought to be a robust tracer for BB for aging timescales of at least one day (Cubison et al., 2011).The Eastern Mediterranean is considered an area of importance for atmospheric aerosol research where aerosols play a significant role in radiative cooling up to five times greater than the warming induced by greenhouse gases (Vrekoussis et al., 2005).
During summertime, when enhanced anthropogenic aerosol inflow is combined with limited precipitation and persistent high relative humidity, the total light scattering may be twice of that during wintertime (Kalivitis et al., 2011).The organic fraction during summer accounts for one third of the dry submicron aerosol mass, with organics being highly oxidized (Hildebrandt et al., 2010a).Based on long-term measurements of carbonaceous aerosols in the area, Sciare et al. (2008) demonstrated that every year during two periods (March-April and July-September) there is a high biomass burning contribution to the aerosol from long-range transport from Southern Europe and European countries surrounding the Black Sea.This study presents high temporal resolution measurements in a remote background site in the Eastern Mediterranean during late summer (August-September 2012).With the concurrent use of an Aerosol Chemical Speciation Monitor (ACSM, Aerodyne Inc.) and a 7-wavelenght aethalometer we identified and studied different biomass burninginfluenced air masses that arrived in the area.The measurements are analyzed to gain insights into the atmospheric processing of BBOA.Introduction

Conclusions References
Tables Figures

Back Close
Full Aerosols, Clouds, and Trace gases Research Infrastructure Network (ACTRIS).A detailed description of the site can be found in Mihalopoulos et al. (1997) and Sciare et al. (2003).We focus on periods of intense biomass burning events that occurred at the Greek islands of Chios, Euboea and Andros, one along the Dalmatian Coast of Croatia that impacted a majority of the Balkan region and one in Sicily.All events took place during August and September of 2012.Air mass origin and plume arrival time were estimated by HYSPLIT backtrajectory analysis (www.arl.noaa.gov/ready/hysplit4.html).
Locations for each fire are shown in Fig. 1 together with a MODIS image of the Chios fire.

Instruments and methods
High temporal resolution measurements were performed with an Aerodyne Research ACSM (Ng et al., 2011a), which measures aerosol mass and chemical composition (ammonium, sulfate, nitrate, chloride and organics) of non-refractory submicron aerosol particles in real-time.The ACSM is designed and built around the same technology as the Aerosol Mass Spectrometer (AMS), in which an aerodynamic particle focusing lens is combined with high vacuum thermal particle vaporization, hard electron impact ionization and final detection of the resulting ions with a commercial quadrupole mass spectrometer.During the measurement period ambient air was drawn into the ACSM by a PM 10 aerosol inlet without prior drying.The corresponding measurements had a temporal resolution of 30 min.Mass concentrations are calculated with the recommended collection efficiency (CE) of 0.5 for all constituents, apart from sulfate (0.25).The resulting concentrations were compared to those from filter measurements and are discussed in subsequent sections and the Supplement.Black carbon (BC) absorption measurements were conducted with a 7-wavelength aethalometer (Magee Scientific, AE31).Ambient air was drawn into the aethalometer by the same PM 10 aerosol inlet as mentioned above, but after passing through a diffusional silica dryer.Samples were obtained every 5 min and were subsequently averaged to 30 min in order to match with the ACSM time intervals.

Conclusions References
Tables Figures

Back Close
Full During the whole measurement period daily PM 1 filter samples were collected on pre-combusted quartz fiber filters.Samples were analyzed for organic and elemental carbon using an OC/EC analyzer (SUNSET Laboratory Inc.) as well as for watersoluble ions after extraction in ultrasonic bath with nanopure water.The solutions obtained were analyzed by ion chromatography (IC) for anions (Cl − , Br − , NO − 3 , SO 2− 4 , C 2 O 2− 4 ) and cations (K + , Na + , NH + 4 , Mg 2+ , Ca 2+ ) using the procedure described by Bardouki et al. (2003).Results for organic carbon (OC), sulfate, ammonium and nitrate were compared to ACSM-derived concentrations.As the ACSM provides organic matter (OM) concentrations, the filter-derived OM was calculated by multiplying OC concentrations with a conversion factor (in our case 2.4; Sect.3.3), which is the ratio of the average organic molecular mass to the carbon mass for the organic aerosol.Particle number distribution was monitored using a scanning mobility particle sizer (SMPS) in 5 min time resolution, O 3 concentrations and meteorological parameters (wind speed and direction, pressure, temperature, relative humidity) were also continuously monitored.

PMF analysis
The measured mass spectra from the ACSM were deconvolved into different components using the positive matrix factorization (PMF) analysis exploiting the Multilinear solver (ME-2) (Paatero and Tapper, 1994;Paatero, 1999).The interface for initiating and controlling the multilinear engine algorithm (ME-2, P. Paatero, University of Helsinki), as well as analyzing the results, was developed at the Paul Scherrer Institute (Canonaco et al., 2013) within the software package Igor Pro Version 6.22A (Wavemetrics Inc., Lake Oswego, OR, USA).The interface is intended to facilitate the analysis of the source apportionment of measured data (Ng et al., 2011b).Briefly, PMF uses a least squares algorithm to minimize the squared difference between modeled and the measured data, i.e. the residual weighted by the uncertainty.Here, the measurement error of the ACSM represents the uncertainty.The input organics and organics error Introduction

Conclusions References
Tables Figures

Back Close
Full matrices for the analysis are derived automatically from the ACSM data analysis software, using a simple automated procedure (Ulbrich et al., 2009).For ME-2 and PMF the entries in the model solution are fit using a least squares algorithm that minimizes iteratively the model uncertainty.For the ACSM data analysis the signals at m/z 18 and 44 are equal in the fragmentation table (Allan et al., 2003(Allan et al., , 2004)).

PM 1 Composition
ACSM sulfate and ammonium concentrations were on average 19 and 15 % lower than the corresponding PM 1 filter measurements (R 2 = 0.8 and 0.68 respectively) (Fig. SI-2).These discrepancies can be attributed to errors and fluctuations in the collection efficiency of these species.Using the parameterization proposed by Middlebrook et al. (2012) for the effect of high aerosol acidity the NH 4 /NH 4,neutr is equal to 0.803 (R 2 = 0.67), where NH 4,neutr represents the theoretical ammonium concentration needed to neutralize the inorganic anion mass concentrations.Similar acidic behavior of the aerosol has been observed for the site in previous measurements (Hildebrandt et al., 2010).Organic matter concentrations calculated from OC analysis (POM = 2.4OC, as derived from our dataset, using the estimate of Aiken et al. (2008), Sect.3.3) and organic concentrations from the ACSM are in good agreement with an average slope of 0.91 and R 2 = 0.57.Slightly higher OC concentrations from filter analysis may be due to the adsorption of gaseous organic compounds on the quartz-fiber filters.The time series of the high temporal resolution measurements is shown in Fig. 2a.We compared the PM 1 mass concentrations by adding the ACSM and BC masses with the mass estimated by the SMPS volume, using an average organic aerosol density equal to 1.35 g cm −3 as determined by Lee et al. (2010a) for the same site during late spring 2008 (Fig. 2b).The ACSM concentrations for ammonium and sulfate are corrected based on the filter concentrations.The mass derived from gravimetric mea-Introduction

Conclusions References
Tables Figures

Back Close
Full surements of the concurrent daily PM 1 filters is also given.The ACSM + BC mass concentrations have an average value of 9.2 ± 4.8 µg m −3 which is in good agreement with the 10.9 ± 4.2 µg m −3 calculated from the SMPS.The difference may be attributed to components that cannot be measured by the ACSM.The ACSM measurements were then divided to those corresponding to fire events and the rest, based on black carbon (BC) measurements from the aethalometer, the HYSPLIT analysis and available satellite pictures.Figure 2c and d represent the relative contribution of each constituent to the PM 1 mass (sum of the derived constituents) during non-fire and fire events, respectively.During fire events, the contribution of organics and BC increase substantially at the expense of sulfate.

Identification of biomass burning events
Biomass burning events were identified by an enhanced BC concentration (Fig. 3a).
We assume that 250 ng m −3 corresponds to the background concentration for Finokalia during periods where the site is not affected by biomass burning.This limit is consistent with the 213 ± 34 ng m −3 found by Cristofanelli et al. (2013) in Mt.Cimone, Italy and comparable to average background levels reported for European continental background sites (Van Dingenen et al., 2005;Putaud et al., 2004).The average BC concentration during the measurement period was of 387 ± 149 ng m −3 , with lower values being close the background value (250 ng m −3 ).Possible biomass burning incidents are identified by having BC concentrations at least 2.5 times higher than the Finokalia background (higher than 625 ng m −3 ).This is approximately equivalent to two standard deviations above the average background value.As a second step, locations of possible fire events were identified by the Fire Information for Resource Management System (FIRMS).The exact location and plume age was calculated by the HYSPLIT backtrajectory analysis.
Five distinct wildfire events affecting the site were identified using the approach described above during the sampling period: a large fire that burned most of the island of Chios (19-21 August); the extensive wildfire at the Dalmatian Coast in Croatia and Introduction

Conclusions References
Tables Figures

Back Close
Full southern Bosnia-Herzegovina which resulted in smoke plumes which spread all across the Balkans during 28-30 August; the smaller fires on the islands of Euboea on 5 September and Andros on 13 September (backtrajectories and MODIS images for each can be found in the Supplement); and a fire that took place in the island of Sicily during 18-21 September and influenced the site on 21 September.
The above periods were characterized by strong biomass burning ACSM tracers at m/z 60 and 73 (Fig. 3).The signal was considerably weaker for the Sicily fire emissions that travelled the longest before reaching Finokalia.BC concentrations were also relatively high towards the end of September, with an average concentration of 535 ± 109 ng m −3 , which is close to 2.5 higher than the background level for Finokalia.
The difference is that during this period BC concentrations build up gradually and the ACSM tracers do not reveal any noticeable variability.In addition, FIRMS does not identify any significant events in the area during 25-30 September, therefore this is not regarded as a fire event.Using the approach of Sandradewi et al. (2008) we also calculated the contribution of the biomass burning to the total BC concentrations (see Supplement).

Elemental composition of organic aerosol
The ACSM dataset was used to calculate the elemental composition and ratio of organic mass to organic carbon (OM / OC) based on the estimates proposed by Aiken et al. (2008)  corresponding to an average f 44 of 18.2 %.The higher O / C value found in this study can be attributed to the higher contribution of f 44 (23.3 ± 3.7 %) The H / C ratio varies in a range of 1.12-1.34,with a mean value of 1.23 ± 0.03.The O / C ratio is a proxy for the oxidation state and the photochemical age of organic aerosols (Ng et al., 2011c;Huang et al., 2011) and thus its variability during the measurement period reflects variations in OA age at Finokalia.Combining information from both ratios, for most of the cases, lower O / C coincides with higher H / C and all fire events can be clearly identified in the time series.During the identified fire events both O / C and H / C have average values of 0.77 ± 0.09 and 1.27 ± 0.02, respectively.These values are consistent with the presence of less oxidized aerosol than the study average.There are lower-thanaverage O / C ratios around the 21st of September, when biomass-burning-influenced air masses from Sicily arrived at the station after a travel time of more than 30 h.The average OM / OC was 2.4 ± 0.14 which is characteristic of a remote background site such as Finokalia and which is comparable to the value of 2.2 found by Hildebrandt et al. (2010b) and the value of 2.1 used by Sciare et al. (2005) to convert OC to POM for mass reconstruction during the MINOS summer campaign.These very high values of OM / OC reflect the highly oxidized nature of organic Eastern Mediterranean aerosol, which during the middle of the summer is in fact even more oxidized than found by Hildebrandt et al. (2010a) for the same site during late spring and early summer.Introduction

Conclusions References
Tables Figures

Back Close
Full

Organic aerosol source apportionment using PMF
The PMF analysis of the ACSM mass spectra (MS) acquired during the measurement period was performed investigating solutions using 1 to 4 factors.The 1-factor solution was not satisfactory.The 2-factor solution resulted in a mixed processed BBOA profile and an OOA profile.The mixed processed BBOA factor had characteristics of both BBOA, such as signal at m/z 57, 60 and 73, but also of OOA, with pronounced signal at m/z 28 and 44.It correlates well with both BC and nitrate (R 2 = 0.6 and 0.74, respectively) but poorly with sulfate.Its diurnal variability coincides mostly with the arrival time of biomass burning-laden air masses.Its O / C was 0.49 suggesting oxygenated BBOA.When compared to reference mass spectra it correlated relatively well with BBOA but mostly with OOA.In order to separate the mixed behavior of this processed BBOA profile we used the 3-factor solution.The PMF solutions with 4 factors resulted in splitting behavior of the 3 factors, without identifying a new factor profile.The splitting behavior was affirmed by poor correlation with tracer species and by mass spectra that appear also to be split between factors.
The 3-factor solution was selected for having good correlation with tracer species while displaying distinct diurnal variation, factors have MS features that compare well to reference mass spectra.The stability of the solution was tested by initializing the algorithm of the analysis from 20 different starting points ("seeds") and results showed very little variability, with no unexplained mass observed.Similar results were also obtained by rotating the obtained solution of the model using f peak runs.Details about the PMF for all cases can be found in the Supplement.
The three factors correspond to fresh BBOA, an OOA component derived from biomass burning (OOA-BB) and a more oxidized OOA. Figure 5 shows the mass spectra of the three factors and the time series of each factor, associated with external reference time series.The BBOA, OOA-BB and OOA accounted on average for 18, 30 and 52 % of the organic mass, respectively.The average diurnal profile of each factor is also depicted in Fig. 5, where the peak of the BBOA and OOA-BB factor coincide with Introduction

Conclusions References
Tables Figures

Back Close
Full arrival times of smoke-influenced air masses during early morning hours.The identified factors were also examined for their mass spectra signatures, their correlation with reference mass spectra and also with other characteristic time series which justify their behavior.
The BBOA component had the characteristic markings of biomass burning aerosol, which are enhanced signals at m/z 60 and 73 probably from levoglucosan and related compounds (Schneider et al., 2006;Cottrell et al., 2008).It also had prominent peaks at m/z 27, 29, 41, 43, 55, 57, 77 and 91 and significant contributions at high m/z's which are indicative of freshly-emitted organic aerosol.Apart from levoglucosan fragments, m/z 57 can also be a good marker for fresh aerosol, as it is a typical fragment of saturated hydrocarbon compounds or of long alkyl chains (C 4 H + 9 ) (Alfarra et al., 2004;Capes et al., 2008) or can also originate from levoglucosan (most probably C 3 H 5 O + ), fragment which is pronounced in our spectrum.The fresh BBOA factor in our study had an O / C of 0.19 and an H / C of 1.35 consistent with other observed BBOA (Fig. 4c).The BBOA factor time series showed good correlation with both BC (R 2 = 0.62) as well as with nitrate (R 2 = 0.75) and no correlation with sulfate or ammonium.
The OOA-BB mass spectrum displayed characteristics of oxidized OA with enhanced signals at m/z 18, 29 and 44.It also had a small signal at m/z 43, implying its potential origin from the fresh BBOA.Its O / C was 0.96 and its H / C 1.3, indicating its processed nature.OOA-BB correlated with nitrate (R 2 = 0.63) and BC (R 2 = 0.48) and displayed lower correlation with sulfate and ammonium (R 2 = 0.29 and 0.38 respectively).OOA-BB peaked during the fire periods and exhibited similar diurnal variability with the BBOA factor (Fig. 5) supporting its classification as processed biomass-burning OA.
The OOA component had the characteristics of oxidized organic aerosol, like OOA-BB, but the relative intensity of m/z 18 and 44 is even more pronounced, indicating more oxidized aerosol, which is also supported by its elevated O / C (1.23).The intensity of other fragments in its spectrum is low.This factor correlates well with both sulfate and ammonium (R 2 = 0.59 and 0.62, respectively).Introduction

Conclusions References
Tables Figures

Back Close
Full Figure 6 shows the similarity (expressed as the angle θ between the spectra; Kostenidou et al., 2009) between the identified factors in this study and other relative organic aerosol spectra found in the AMS mass spectral database (http://cires.colorado.edu/jimenez-group/AMSsd/).Our BBOA spectrum correlates well with the average BBOA of Ng et al. (2011), obtained by averaging together BBOA mass spectra from two sites (Mexico City and Houston data sets) and has some similarity with the spectra of OA from laboratory burns of pine, oak and spruce (Schneider et al., 2006).The derived BBOA had comparable features with ambient BBOA spectra from Mexico (Aiken et al., 2009), Po Valley (Saarikoski et al., 2012) and Barcelona (Mohr et al., 2012) (θ between 25 and 30 • ).The OOA-BB factor correlates well with organic aerosol from the Po Valley and Barcelona, as well as with less oxidized OOA reported for the same sampling site (Finokalia) in past studies during spring and fall (Hildebrandt et al., 2010a(Hildebrandt et al., , 2011)).It does not correlate well with OOA associated with wood burning in fireplaces in Paris during the winter (Crippa et al., 2013) nor with semi-volatile OA from Pasadena (Hersey et al., 2011).Finally, OOA correlates well with the more oxidized OOA found for the same site in previous campaigns and resembles to the OOA found in Barcelona, Zurich during winter and OOA Po Valley.
The contribution of each factor to the OA and the characteristic fragments to the OA mass spectrum of the different fire events is summarized in Table 1.These results are based on separate PMF analysis for each event by selecting the corresponding 3-factor solution.OOA-BB and OOA mass spectra during each event correlated well with the respective OOA-BB and OOA spectra for the whole measurement period (shown in Fig. 5b) but also with the corresponding spectra of the rest of the events (θ < 15 and 9 degrees respectively).f 60 and f 73 diminish and f 44 increases as the transport time of the plume increases.According to Cubison et al. (2011), f 60 has consistent background level values around 0.3 % in OA with negligible BB influence, with plumes exhibiting a trend toward lower f 60 with time.This is also the case in our study where the Sicily and Croatia fire events exhibit the lowest values, close to these background levels.We estimated an e-folding time of 12 h for m/z 73 and 28 h for m/z 60 for these fire events.Introduction

Conclusions References
Tables Figures

Back Close
Full The BBOA fraction in the OA during these events decreases from approximately 40 % for 7 h of processing to around 15 % for 16 h.The OOA-BB fraction remains relatively constant, while the OOA increases from around 30 % to around 60 % as the aerosol ages.These are consistent with a transformation of the BBOA to OOA-BB and then to OOA during the chemical aging of the OA.
Based on these results, the biomass burning-influenced air masses arriving from Sicily after 30 h of travel lacked any characteristic BBOA marker.This is an indication that these markers have a finite lifetime, especially during summertime Mediterranean conditions, which combine high temperatures and high insolation periods.f 60 seems to be a robust BB-tracer only for timescales up to one day from emission.Even with a few hours of atmospheric processing, and mostly during nighttime, the majority of the freshly-emitted BBOA is processed and transformed to more oxidized OA, with OOA-BB accounting for 30 % on average of the total organic mass.This is consistent with the observations of aging of fresh biomass burning OA in plumes over Canada within a few hours after emission (Lathem et al., 2013).This aging was attributed to possible gas-phase oxidation of semivolatile species and also to evaporation of primary semivolatile species (Cubison et al., 2011).Therefore, this aging can be attributed to substantial evaporation of semi-volatile species upon dilution of the BB plume but also to heterogeneous or homogeneous reactions that take place during the transport time.This suggests that the contribution of biomass-burning-influenced aerosol can be largely underestimated as a significant part of it may be misidentified as OOA, but not taking into account that it is derived from BBOA.Based on the results it can be seen that biomass burning may contribute almost half of the organic aerosol in the area during summertime.

Summary and conclusions
High temporal resolution measurements were conducted at a remote background site in the Eastern Mediterranean with an ACSM and an aethalometer during August-Introduction

Conclusions References
Tables Figures

Back Close
Full September 2012.During the measurement period PM 1 concentrations ranged from 3.2 to 34.2 µg m −3 with an average value of 12.2 ± 4.7 µg m −3 with organics and sulfate being the most abundant species, each accounting for ∼ 40 % of the PM 1 .Five discrete biomass burning events were identified from characteristic tracers in the organics mass spectrum as well as in the BC time series, and origin of the air masses was verified by backward trajectory analysis.For all the cases the smoke was transported for several hundreds of kilometers.In three events the smoke arrived at the site early in the morning, after around 8 h of night-time atmospheric processing.One additional fire event involved air masses with an approximate travel time of 30 h, coming from Sicily.During the fire events the contribution of organics to the total mass increased to almost 50 %, with BC also showing an increase in contribution to almost 10 % of the PM 1 .By performing PMF analysis three factors were derived: one having characteristics of biomass burning, one with oxygenated BBOA, and one with highly oxygenated organic aerosol.The BBOA component showed reasonable correlation with reference average BBOA mass spectra while more oxidized OOA correlated well with reference OOA mass spectra.BBOA correlated well with BC and nitrate, processed BBOA (OOA-BB) also had high correlation with nitrate and more oxidized OOA correlated well with sulfate and ammonium.More than 80 % of the total organic mass is accounted for by the two oxygenated OOA factors.The OA as a result was highly oxygenated with an overall O / C ratio that had a mean value of 0.97 ± 0.13.Even within 8-9 h of travel, and mostly during nighttime, most of the freshly-emitted BBOA had been processed and transformed to more oxidized OOA, with processed-BBOA accounting for 30 % on average of the total organic mass.This suggests that the biomass-burning contribution to OA can be misidentified as OOA contribution, therefore underestimating the importance of BBOA.
The biomass burning-influenced air masses arriving from Sicily after 30 h of travel, lacked any characteristic BBOA marker.This is an indication that these markers have a finite lifetime, especially during summertime Mediterranean conditions, which combine Introduction

Conclusions References
Tables Figures

Back Close
Full  Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | and Ng et al. (2011c), as follows: O / C = (3.82± 0.005)f 44 + (0.0794 ± 0.007) (1) OM / OC = (1.26± 0.02)O/C + (1.18 ± 0.001) (2) H / C = (1.01 ± 0.08) + (6.07 ± 1.39)f 43 + (−16.01 ± 5.34)f Discussion Paper | Discussion Paper | Discussion Paper | where f 43 and f 44 are calculated as f 43 = [m/z 43] (µg m −3 )/C OA and f 44 = [m/z 44] (µg m −3 )/C OA with C OA being the total OA mass concentration in µg m −3 .Figure 4 shows the variation of the atomic ratios of O / C and H / C and the mass ratio of OM / OC of the OA during the measurement period, averaging every 2 h, as well as f 43 vs. f 44 compared to other studies.According to Herringa et al. (2011) the presence of wood-burning POA and its aging SOA are found at the left edge of the triangle.The O / C ratio varies between 0.52 and 1.34, with a mean value of 0.97 ± 0.13, values which are within the observed values of low-volatility and semi-volatile OOA (0.3-1.1; Ng et al., 2011c) and dicarboxylic acids (0.44-2.00;Kuwata et al., 2012).For the same region, Hildebrand et al. (2010a) estimated an O / C of 0.8 for a sunny May period, Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | high temperatures and high insolation periods.f 60 seems to be a robust BB-tracer only for timescales up to one day from emission.Discussion Paper | Discussion Paper | Discussion Paper | various types of combustion-related aerosol particles, Int.J. Mass.Spectrom., 258, 37-49, 2006.Sciare, J., Bardouki, H., Moulin, C., and Mihalopoulos, N.: Aerosol sources and their contribution to the chemical composition of aerosols in the Eastern Mediterranean Sea during summertime, Atmos.Chem.Phys., 3, 291-302, doi:10.5194/acp-3-291-2003,2003Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 3 .Fig. 6 .
Fig. 3.The time series of the biomass tracer loadings of m/z 60 and m/z 73 and black carbon.The shaded areas represent the identified fire events and locations.

Table 1 .
Relative contribution of each one of the three identified PMF factors to the OA and relative intensity of characteristic markers in each fire event.