Milena Stracquadanio, Ettore Petralia , Massimo Berico, Teresa M.G. La Torretta, Antonella Malaguti, Mihaela Mircea, Maurizio Gualtieri, Luisella Ciancarella

Laboratory of Atmospheric Pollution, National Agency for New Technologies, Energy and Sustainable Economic Development, ENEA, 40129 Bologna, Italy


Received: May 23, 2018
Revised: September 21, 2018
Accepted: November 12, 2018

Download Citation: ||https://doi.org/10.4209/aaqr.2018.05.0186  


Cite this article:

Stracquadanio, M., Petralia, E., Berico, M., La Torretta, T.M., Malaguti, A., Mircea, M., Gualtieri, M. and Ciancarella, L. (2019). Source Apportionment and Macro Tracer: Integration of Independent Methods for Quantification of Woody Biomass Burning Contribution to PM10. Aerosol Air Qual. Res. 19: 711-723. https://doi.org/10.4209/aaqr.2018.05.0186


HIGHLIGHTS

  • Source Apportionment and Macro Tracer approaches for BB determination are compared.
  • The two independent methods resulted in good agreement.
  • Determination of a representative conversion factor for olive hardwood burning.
  • Methods combination for a reliable assessment of BB contribution.
  • Methods combination useful for site-specific conversion factors assessment.
 

ABSTRACT


During the last few years, the rise in woody biomass burning (BB) for household heating has caused an increase in PM mass concentrations, particularly for the fine fraction, in Europe, as reported by the European Environmental Agency. Estimating the contribution from biomass combustion to airborne particulate matter is therefore an important issue in air quality governance, due to its potential health and environmental impacts.

Wood burning’s contribution to PM10 was estimated in winter at a rural site in southern Italy by means of two independent methods: source apportionment analysis with Positive Matrix Factorization (BBPMF) and the macro tracer approach, based on levoglucosan concentrations (BBLevo). PM10 and PM2.5 samples were collected every 24 h and every 8 h, respectively, and analyzed to determine the organic and elemental carbon, levoglucosan, inorganic ions and elements.

The results obtained via these methods showed good agreement (r = 0.85), with a linear correlation slope of about 1, and provide a reliable assessment of the BB contribution.

Woody biomass combustion contributed significantly to the PM10 (on average, slightly less than 30% of the total mass) during winter. The combination of the independent methods proposed here may be used as a methodology for refining the BB contribution to air pollution.


Keywords: Biomass combustion; Particulate matter; Levoglucosan; PMF; Emission sources.


INTRODUCTION


National and international environmental policies point out the need for identifying and quantifying the contribution of emission sources to atmospheric concentrations, distinguishing also the contribution of the local and regional sources from the long-range transported pollutants. Air Quality Directives (Council Directive 2008/50/EC; Council Directive 2004/107/EC) reinforce the importance of having information on pollution sources for the development of efficient air quality plans and for the evaluation of human risk-exposure.

Identification and quantification of the air pollution sources have been performed with various methodologies such as explorative methods, emission inventories, photochemical and dispersion models (Lagrangian, Gaussian and Eulerian models) and receptor models (Viana et al., 2008; Belis et al., 2014). The receptor models (RMs, e.g., Principal Component Analysis, Multilinear Engine, Chemical Mass Balance) are mathematical or statistical procedures that are able to provide information on the contribution of emitting sources at a specific receptor site using in-situ measured experimental data and solving a mass balance equation, some of which may run also without the knowledge of the pollution sources.

One of the most commonly RMs used is the Positive Matrix Factorization (PMF, P. Paatero, Helsinki, Finland; Paatero and Tapper, 1994) that explains the correlation between the different measured species (carbonaceous fractions, inorganic ions, trace metals, etc.) through their linear combinations. These combinations, resulting in clustered “factors” (sources), originate from hidden common trends of species.

In the last years the international research community, e.g., European Commission with the Forum for Air quality Modelling (FAIRMODE; http://fairmode.jrc.ec.europa.eu/) and United States Environmental Protection Agency (https://www.epa.gov/scram/air-pollutant-receptor-modeling) have dealt with source apportionment matter and some basic guidelines and procedures (Belis et al., 2014; Norris et al., 2014; Pernigotti et al., 2016) are in continuous improvement for joint assessment protocols among all the scientific subject involved in air quality evaluation (Karagulian et al., 2012; Belis et al., 2015a, b).

Source apportionment is gaining relevance in highly polluted areas to identify the main emission sources affecting air quality in order to propose/develop/implement protocols and strategies aiming at the reduction of air pollution.

Among the most relevant anthropogenic emission sources (such as vehicular traffic, industrial and agriculture activities), an important contribution is given by biomass-based household heating appliances during wintertime. In the last years, woody biomass burning (BB) for domestic or residential heating has increased its diffusion driven by climate and energy policy and the higher costs of other fuels (EC Commission Stuff Working Document, 2014; EC Renewable Energy Progress Report, 2015).

The incomplete or inefficient combustion of wood biomass in conditions typical of small and medium domestic heating device use has been recognized as an important source of particulate matter (PM10, PM2.5, PM1) and other air pollutants harmful to human health (EC Commission Staff Working Document, 2014; WHO, 2015, www.eea.europa.eu/publications/air-quality-in-europe-2015/download).

Quantification of BB contribution is still challenging due to mixing of sources, heating device type, species of wood and combustion conditions, and meteorology. BB contribution to PM is usually estimated by using RMs (Bernardoni et al., 2011; Waked et al., 2014; Petralia et al., 2017) or specific macro tracer approaches (Sdizat et al., 2006; Schmidl, 2008; Piazzalunga et al., 2011; Shahid et al., 2019).

The macro tracer approach is based on the identification of markers (such as levoglucosan or 14C) that are directly related to biomass combustion emissions allowing thus the estimation of BB contribution to the mass concentration of PM. Levoglucosan and its stereoisomers, mannosan and galactosan, are monosaccharide anhydrides produced by pyrolysis processes at temperatures higher than 300°C of cellulose and hemicellulose, respectively, and are considered specific markers of BB (Simoneit et al., 1999; Fine et al., 2002; Gelencsér et al., 2007; Yttri et al., 2011; Nielsen et al., 2017). Levoglucosan may undergo degradation processes in the atmosphere due to oxidation reactions with radicals present in the ambient air; in particular, the reactions with the OH radical have been widely studied (Hennigan et al., 2010; Hoffman et al., 2010; Lai et al., 2014). In fact, levoglucosan atmospheric lifetime could vary widely between 1.2 and 3.9 days (Lai et al., 2014) under different environmental condition (temperature [T] and relative humidity [RH]), OH concentration, other atmospheric radical concentrations and levoglucosan mixing state together with inorganic salt and soot (Hennigan et al., 2010; Hoffman et al., 2010; Lai et al., 2014). As a consequence, the use of levoglucosan for estimating the contribution of BB to PM is under debate since the possible underestimation due to the marker degradation, particularly in summer atmospheric conditions and/or sampling aged PM (Busby et al., 2016). The macro tracer approach proposed by Schmidl et al. (2008) estimates the BB contribution to PM10 depending on the levoglucosan atmospheric concentration through a specific conversion factor related to the wood species burnt and BB fuel type (logs or pellets). This method, already applied in Italy (Piazzalunga et al., 2011), results in more precise site specific conversion factor assessment and so for BB contribution estimation, since it considers the species and fuel type actually burnt at certain area.

Despite the wide application of these methods, both show advantages and limits. PMF allows to obtain information on the contribution of emission sources in a specific receptor site without knowing a priori the different types of sources persisting at a site; on the contrary, the contribution, in terms of absolute concentration value, could be underestimated. Indeed, the analytical speciation of the chemical composition of PM is never complete, because of the difficulties in determining the water content, in establishing a complete and specific geochemical characterization of various metal oxide, and in assessing the best organic matter (OM) quantification (Sciare et al., 2005; Bae et al., 2006; Terzi et al., 2010; Perrino et al., 2013; Perrino et al., 2014; Genga et al., 2017). The lack of this information affects considerably PM10 mass reconstruction from the characterized species (Belis et al., 2014). Macro tracer approach allows to estimate the BB contribution to PM from a single tracer (levoglucosan) but this evaluation is highly dependent on the knowledge of local type of solid biomass fuel (logs, chips, pellets, bark and sawdust) and wood species consumption inventory as well as on levoglucosan degradation processes; both factors lead to the uncertainty or underestimation of the real BB contribution.

This study estimates the BB contribution to PM10 samples collected at a rural site located in the south of Italy during winter by applying PMF and macro tracer methods in an independent and complementary way. The simultaneous application of these methods may improve our possibility to correctly evaluate the BB contribution to air pollution leading to a more accurate and “realistic” quantification and to obtain more reliable site specific BB evaluation.


MATERIALS AND METHODS



PM Sampling and Measurement Techniques

The aerosol was sampled in a coastal zone in Apulia (Italy) (Fig. 1(a)) at three sites: Cisternino (40.74278°N; 17.41583°E), Torchiarolo-Lendinuso (40.5175°N; 18.07889°E) and Torchiarolo-Fanin (40.48945°N; 18.04723°E). Cisternino (Ci) site is defined as rural site, while Torchiarolo-Lendinuso (TL) and Torchiarolo-Fanin (TF) are classified as suburban sites (according to the criteria in Italian Decr. Leg. 13 agosto 2010, n. 155). Ci is located 11 km from the sea and at an elevation of 400 m, TL and TF are located 0.7 and 5 km away from the sea respectively at an elevation of 30 m a.s.l. The area around the urban sites is moderately urbanized but highly agricultural, characterized by olive groves, vineyards, non-irrigated arable areas and complex crops. All these land-cover typologies determine soils exposure by the wind erosion with eventual re-suspension phenomena of soil and dust material. The distances of Torchiarolo from larger cities in the region are 16 km NNW and 16 km SSE for Brindisi and Lecce respectively, while Taranto is located 65 km WSW. The nearest medium size towns are San Pietro Vernotico (4 km W) and Squinzano (5 km S). The four lane motorway Strada Statale 613 is located around 600 meters W-SW. The prevalent directions of wind during the sampling period were WSW and S-SSE and the wind velocities vary from weak (up to 2–4 m s–1) to moderate (up to 6–8 m s–1) (Fig. 1(b)).

 
Fig. 1. Sampling sites: (a) Torchiarolo (Torchiarolo-Fanin [TF]), Cisternino (Cisternino [Ci]), Lendinuso (Torchiarolo-Lendinuso [TL]) (elab. from https://www.google.com/earth/). (b) Wind rose.Fig. 1. Sampling sites: (a) Torchiarolo (Torchiarolo-Fanin [TF]), Cisternino (Cisternino [Ci]), Lendinuso (Torchiarolo-Lendinuso [TL]) (elab. from https://www.google.com/earth/). (b) Wind rose.

PM10 was sampled in all three sites with FAI Hydra Dual Sampler. At TF, the measurements were carried out with the following instruments: Sunset Laboratory Model-4 Semi-Continuous OCEC Field Analyzer, Aerodyne Aerosol Chemical Speciation Monitor (ACSM), FAI Hydra Dual Sampler (equipped with PM2.5 sampling head) and VAISALA MAWS 100 weather station.

24-hour PM10 samples were collected on polytetrafluoroethylene (PTFE) membranes and quartz fiber filters for 30 consecutive days, from 23 January to 21 February 2016; 29 parameters were quantified: organic carbon (OC), elemental carbon (EC), inorganic ions (Cl, NO3, SO42–, Na+, NH4+, K+, Mg2+, Ca2+), elements (Al, Si, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Ba, W, Pb, Sc, Ce, Rb) and gravimetric PM10. The chemical characterization of PM10 was used as input for PMF (PMF 5.0, USEPA) to identify the emission sources. Only at TF was possible to investigate the macro tracer contribution of biomass combustion using a set of additional parameters such as mass concentration of levoglucosan, PM2.5 with 8-h time resolution and on-line EC and OC with 2-h time resolution.

ACSM on-line measurements were conducted from 21 January to 23 February 2016, with 30-min time-resolution. Data were then pooled and mediated over 8 h or 24 h in post processing operations, to make species time series data homogeneous in order to perform right comparisons and correlations. ACSM measures the major non-refractory (at 600°C) components of PM1 (NR-PM1) such as Organic Matter (OM), nitrate (NO3), sulfate (SO42–), ammonium (NH4+), chloride (Cl), providing loadings for 100 different macro-molecules (individuated by their mass-to-charge, m/z): In particular, the m/z 60 is associated to levoglucosan (Alfarra et al., 2007) and so used for comparison with the laboratory levoglucosan measurements.

Gravimetric determination of PM10 and PM2.5 was carried out according to UNI EN 12341 and UNI EN 14907 methods. PM gravimetric mass was determined using an electronic balance (0.01 mg resolution; model AX205, Mettler Toledo).

Water soluble inorganic ions concentrations (Cl, NO3, SO42–, Na+, K+, NH4+, Ca2+, Mg2+) in 24-h PM10 samples were determined according to CEN/TR16269 method. Teflon filters were extracted in 18-MW Milli-Q® water in ultrasonic bath (32 min) and analyzed by ion chromatography (IC). The detail of the analytical procedure can be found in Malaguti et al. (2015a).

EC and OC mass concentrations were determined in 24-h PM10 samples by means of a Thermal/Optical Carbon Analyzer (Sunset Laboratory Inc.) according to CEN/TR16243:2011 method. On-line EC-OC mass concentrations were measured with Semi-Continuous OC-EC Field Analyzer (Model-4, Sunset Laboratory Inc.) every 2 h. The NIOSH-like (RT-QUARTZ-840) thermal optical transmittance protocol was used both for off-line and on-line analysis. Detailed information about the analytical procedures are reported in Malaguti et al. (2015b).

Metals and trace elements (Na, Mg, S, Al, Si, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Ba, W, Pb, Sb, Sc, Ce, Rb) were estimated using energy-dispersive X-ray fluorescence (ED-XRF; NEX CG, Rigaku). The calibration was carried out by means of Standard Reference Material® NIST 2783 Air Particulate on Filter Media. The Method Detection Limit (MDL) and combined relative standard uncertainty are reported in Supplementary Material (Table S1).

Anhydrosugars (levoglucosan, mannosan and galactosan) were analyzed using High-Performance Anion Exchange Chromatography (HPAEC; Dionex™ ICS-5000, Thermo Scientific™) equipped with Thermo Scientific™ Dionex™ CarboPac™ MA1 Analytical (4 × 250 mm) & Guard (4 × 50 mm) Columns and Pulsed Amperometric Detection (PAD; Dionex™ ED50A Stand-Alone Electrochemical Detector & Electrochemical Detector Conductivity Cells, Thermo Scientific™). Four squares (1 cm2 each) portions of quartz fiber filter from 8 h PM2.5 samples were extracted with 2 mL of 18-MW Milli-Q® water in ultrasonic bath (32 min). Extracted water was filtered to remove debris with a 0.22 µm pore size nylon syringe filter (0.2 µm, Acrodisc® Nylon membrane, Pall®).

A good resolution between anhydrosugars (levoglucosan, mannosan and galactosan) and sugar alcohols (Arabitol, Sorbitol) was achieved (Fig. 2) developing an analytical method starting from Iinuma et al. (2009) consisting of a gradient elution of a 1 M aqueous sodium hydroxide (NaOH) solution at a flow rate of 0.4 mL min–1. Method Detection Limit (MDL) for levoglucosan was evaluated equal to 9.2 ng m–3 by EPA method (MDL = t(n1,1α=0.99) × SD where SD SD is the standard deviation of eight replicate standard solution), while the combined relative standard uncertainty (εR) was 0.07.


Fig. 2. Anhydrosugars HPAEC-PAD chromatogram.Fig. 2. Anhydrosugars HPAEC-PAD chromatogram.


Biomass Burning Identification Methods
: Positive Matrix Factorization (PMF) and Macro Tracer Approaches


PMF

PMF (USEPA PMF 5.0, Norris et al., 2014; PMF2, Paatero, 2004) was applied to chemical data obtained from PM10 samples collected at the three sites investigated.

PMF performs the decomposition of the measured concentrations dataset (X matrix) into three different matrices containing the sources, following the mass balance formula:

 

where G is the sources’ contribution (time variations) matrix, F is the sources’ profiles (such as fingerprint) matrix, and E is the residual (mass not explained by the model) matrix.

PMF can be run on a number of cases (each sampling occasion) which must be minimum three times the number of variables (in our study, 19 different parameters/species) (Belis et al., 2014). The PMF was applied to a dataset (n = 1767) of 19 measured parameters in 93 daily samples collected at TF, Ci and TL.

The selected data followed the quality criteria defined by Belis et al. (2014) and Norris et al. (2014) for each parameter: Only measurements parameters having both 50% of values above detection limit (ADL) and a signal-to-noise ratio (S/N) value > 1 were considered. The parameters so used for source profile identification with PMF analysis were EC, OC, Cl, NO3, SO42–, Na+, NH4+, K+, Mg2+, Ca2+, Al, Si, Ti, V, Mn, Fe, Ni, Cu, Zn, all considered as “strong” according to Norris et al. (2014); total-gravimetric PM10 was considered as total variable and set as “default to weak” (Norris et al., 2014). This assumption distributed the PM mass among the different sources leading to an almost realistic estimation of PM10 sources. The retrieval of the entire PM10 mass is particularly required for calculation of biomass combustion factor as explained in Section “Estimations of BB Contribution to PM10-Comparison between Macro Tracer Method and PMF Method”. Levoglucosan filter concentrations were not included in the PMF input dataset in order to avoid its eventual influence on source profile definitions and on BB factor calculation. The “Results” section (“Estimations of BB Contribution to PM10-PMF and Source Apportionment”) shows two reconstructed PM10 mass: The sum of the masses of all the speciated elements reconstructed by PMF (Σspec) and as the sole component PM10 reconstructed mass coming from PMF output (PM10_comp).

For identification and confirmation, the resulting factors profiles were compared to emissive profiles in SPECIEUROPE database (Pernigotti et al., 2016) and other literature reference-profiles (cit. lit. in Section “Estimations of BB Contribution to PM10-PMF and Source Apportionment”). 


Macro Tracer Method

BB contribution to PM was estimated by macro tracer method proposed by Schmidl et al. (2008) on the basis of wood species and woody fuel type (firewood, pellets or logs).

According to this approach, PM10 mass contribution from biomass burning can be calculated from levoglucosan concentration (BBLevo) as follows:

 

where FW is a conversion factor calculated considering both the woody species burnt in the area of study and the relative spread of pellet burning appliances:

 

where ci is the weight of each woody species and woody fuel type in the examined site.

This formula takes into account endemic woody species for Austrian and alpine regions where this approach has been developed and applied (Schmidl et al., 2008; Piazzalunga et al., 2011). The woody species mainly burned in Apulia are about 70% olive, approximately 10% beech and oak, 10% other local species (e.g., almond tree) according to Spagnolo et al. (2014) and the remaining 10% is associated to pellet combustion.

The emission factors for olive and for the other endemic species burnt in the study area were not available; therefore, three different hypotheses were considered in the formula proposed by Schmidl. Olive woody biomass emission factor was assumed to be the equal to that reported for oak wood combustion (1) (as in equation FW1), beech wood combustion (2) (FW2) and a combination of 50% oak and 50% beech (3) (in FW3). Oak and beech were selected as representative of the upper and lower limit of hardwood species PM10 emission factors (Kistler et al., 2012):

  


RESULTS AND DISCUSSION



Samples Characterization

Table 1 shows the mean, median, standard deviation, minimum and maximum values, 25th and 75th percentile concentrations calculated for 24-h samples at TF site for PM10, PM2.5, OC, EC and levoglucosan. Descriptive statistics for EC, OC, Cl, NO3, SO42–, Na+, NH4+, K+, Mg2+, Ca2+, Al, Si, Ti, V, Mn, Fe, Ni, Cu, Zn of samples from the three examined sites are reported in Supplementary Materials (Table S2).


Table 1. Main statistical parameters values calculated for 24-h time resolution (PM10, PM2.5, OC, EC, levoglucosan) and 8-h time resolution (PM2.5, OC, EC, levoglucosan) concentrations.

PM10 shows a mean concentration value of 32 µg m–3 with a minimum value of 17 µg m–3 and a maximum value 58 µg m–3. Daily PM10 limit value of 50 µg m–3 set by AQ Directive 50/2008 was exceeded four times during the sampling campaign, on 28 January and 2, 15 and 16 February with 56 µg m–3, 54 µg m–3, 58 µg m–3 and 52 µg m–3, respectively. On the 15 and the 16 of February a Saharan dust intrusion event occurred (Fig. S1), based on Al concentration value > 1.0 µg m–3 and Ca/Al ratio < 2 (according to Koçak et al. (2012)).

Daily PM2.5 concentrations reached the highest values on 28 January with 52 µg m–3, almost equal to PM10 and the mean value over the sampling campaign was ca. 22 µg m–3. During the campaign, the fine fraction represented on average about 70% of PM10 daily concentrations.

Levoglucosan was detected in all the collected samples, confirming the presence of woody biomass combustion as PM emission source in the sampling site. Daily levoglucosan concentrations vary from a minimum value of 0.15 µg m–3 on 18 February to a maximum value of 1.9 µg m–3 on 30 January, with a mean value of 0.75 µg m–3.

At higher time resolution (8 h), PM2.5 concentrations vary from 2 µg m–3 to 84 µg m–3 and levoglucosan from 0.053 µg m–3 to 4.1 µg m–3. As expected ACSM’s m/z 60 and levoglucosan time series showed a good agreement, both showing temporal patterns similar to that of PM2.5 with higher concentrations from 4:00 p.m. to 8:00 a.m. (Fig. 3). More likely these high values are due to house-heating/cooking activities. A similar trend observed for the EC/OC data support this hypothesis. This daily cycle of biomass burning indicators suggests the presence of significant local source rather than long-range transport. Generally, the major emission sources of fine OC and EC at urban-suburban areas are traffic, domestic heating, cooking and industrial activities (Piazzalunga et al., 2011). The high correlation coefficients between the woody biomass combustion tracers (levoglucosan and m/z 60) and PM2.5 and EC/OC (Table 2) suggest that the biomass burning (domestic heating and cooking activities) is one of the most relevant sources of fine PM at the sampling site (Tzompa-Sosa et al., 2016).


Fig. 3. 8-h concentration of PM2.5, levoglucosan, EC, OC, and m/z 60 during field campaign.Fig. 3. 8-h concentration of PM2.5, levoglucosan, EC, OC, and m/z 60 during field campaign.


Table 2. Levoglucosan, EC, OC, and PM1 BB tracers (m/z 60) linear correlation coefficients.


Estimations of BB Contribution to PM10


PMF and Source Apportionment

PMF analysis identified 9 factors (sources): fresh marine aerosol, soil dust, biomass combustion (BB), vehicular traffic emission, industrial/port activity, re-suspended dust, nitrate-rich secondary aerosol, sulfate-rich secondary aerosol and aged marine aerosol. In this section, the data for the biomass combustion source are shown and discussed; the other sources are reported in Supplementary Material (Fig. S2).

BB factor (Fig. 4) is well isolated, since it shows a profile characterized almost exclusively by K+, typically used as indicator of biomass burning in European source apportionment studies (Rodriguez et al., 2004; Andersen et al., 2007; Petralia et al., 2017), together with OC and EC with a relative contribution of 48.9%, 41.6% and 40.4%, respectively. Furthermore, the EC/OC ratio (equal to 0.12) is in accordance with data previously reported on emissions from residential BB by Szidat et al. (2006); it is noteworthy the elevated correlation (r = 0.85) between 24-h levoglucosan and 24-h OC apportioned to BB by PMF.


Fig. 4. Biomass burning factor profile obtained from PMF analysis; vertical bars (principal y-axis): mass concentration (µg m−3) of the single element in the factor; black squares (secondary y-axis): % contribution of the single element in the factor over the totality of factors.Fig. 4. Biomass burning factor profile obtained from PMF analysis; vertical bars (principal y-axis): mass concentration (µg m3) of the single element in the factor; black squares (secondary y-axis): % contribution of the single element in the factor over the totality of factors.

Fig. 5 reports the mass contribution to PM10 (µg m–3) for Σspec and for PM10_comp related to each factor. Fig. 6 shows the daily sources apportionment derived from the sum of all the speciated elements (Σspec) and for parameter PM10_comp; for a more immediate interpretation, similar sources have been clustered: secondary inorganic aerosol, sea salt and soil dust come from clustering respectively nitrate-rich secondary aerosol and sulfate-rich secondary aerosol, fresh marine aerosol and aged marine aerosol, and soil dust and re-suspended dust.


Fig. 5. Mass contribution (µg m−3) of factors to Σspec (grey bars) and mass contribution (µg m−3) of PM10_comp (black bars) attributable to each factor. N_Expl: unexplained mass.Fig. 5. Mass contribution (µg m3) of factors to Σspec (grey bars) and mass contribution (µg m3) of PM10_comp (black bars) attributable to each factor. N_Expl: unexplained mass.


Fig. 6. Daily sources apportionment derived from the sum of all the speciated elements, (a) Σspec and (b) for parameter PM10_comp. Secondary inorganic aerosol, sea salt and soil dust come from clustering respectively nitrate-rich secondary aerosol and sulfate-rich secondary aerosol, fresh marine aerosol and aged marine aerosol, soil dust and re-suspended dust.Fig. 6. Daily sources apportionment derived from the sum of all the speciated elements, (a) Σspec and (b) for parameter PM10_comp. Secondary inorganic aerosol, sea salt and soil dust come from clustering respectively nitrate-rich secondary aerosol and sulfate-rich secondary aerosol, fresh marine aerosol and aged marine aerosol, soil dust and re-suspended dust.

Σspec is not sufficient to reconstruct the daily gravimetric PM10 mass concentration (Fig. 6(a)), while if PM10_comp is considered in the PMF (Fig. 6(b)), the agreement between source reconstructed mass concentration and measured daily gravimetric PM10 mass concentration increases (Table 3). In particular, in Table 4 the single source contributions to PM10 are reported for the clustered sites. Even though the sources mass contribution to PM10 may differ among the two mass reconstructions (for BB Σspec is 5.0 µg m–3 while considering PM10_comp apportioned mass is 7.8 µg m–3), the relative contribution (Table S3) does not vary significantly (for BB the relative contribution is 20.2% and 20.3% for Σspec and PM10_comp, respectively).


Table 3. Mean, median, standard deviation, minimum and maximum values for PM10 and Σspec and PM10_comp mass reconstruction.


Table 4. Σspec and PM10_comp sources mass concentration data (NE: unexplained mass).

The comparison of sources mass concentration obtained with Σspec and PM10_comp shows remarkable difference for specific sources and in particular for BB, for which the difference is of 40%. This discrepancy is related almost completely to the difference between OM and OC, since the organic fraction enters in PMF analysis as OC. To support this conclusion, we compared the BB OC-to-OM conversion factor, fBB equal to 1.86, reported in Genga et al. (2017) and Philip et al. (2014), with the one calculated here (fBB_PMF):

 

where OCBB is the OC apportioned by PMF to biomass burning factor. The obtained fBB_PMF value (equal to 1.70) is in good agreement with the literature value, therefore supporting our conclusion that the difference between PM10_comp and Σspec is related to the uncertainties in the quantification of OM starting from OC data.

As reported in “Biomass Burning Identification Methods: Positive Matrix Factorization (PMF) and Macro Tracer Approaches” Section, PM10_comp is considered more representative of gravimetric measured PM10 (Fig. S3) since it reproduces almost completely the PM10 mass, associated to all the sources; for clearness and for comparison with the macro tracer approach from now on we will refer just to BB obtained from PM10_comp PMF (BBPMF).


BB 
Macro Tracer Approach

In order to calculate the BBLevo contribution to PM10, 8-h levoglucosan concentrations were averaged every 24 h and multiplied for the conversion factors (FW1, FW2, and FW3) calculated, as reported in “Biomass Burning Identification Methods: Positive Matrix Factorization (PMF) and Macro Tracer Approaches” Section.

Although PM10 and BBLevo contributions show the same trend, their maximum and minimum concentration values differ in occurring days. The highest BBLevo mass concentration value was recorded on 30 January, with 15, 21 and 36 µg m–3 depending on the specific FW considered while the lowest occurred on 18 February, with 1.2, 1.7, and 2.9 µg m–3. The relative BBLevo contribution to PM10 shows maximum value on 12 February (35, 49, and 84%) and minimum on 18 February (6, 8, and 14%) (Table 5).


Table 5. Relative BB contribution to PM10 calculated by means of FW1, FW2 and FW3.

The BBLevo daily contribution to PM10 estimated with macro tracer method ranges from 18 to 44%, depending on the specific FW value (Table 5) and on the conversion factor used in Eq. (1). In order to evaluate the FW more representative of the specific conditions of the studied area the comparison of the independent results from PMF and macro tracer has been performed as reported in the following section.


Comparison between Macro Tracer Method and PMF
 Method

Since levoglucosan concentrations are available just for TF site, only the data from this site were extracted from the PMF output dataset and compared with the macro tracer results. BB contribution to PM10 calculated by means of the two independent methods (BBPMF and BBLevo) were analyzed by linear correlation. BBLevo calculated with FW3 best fitted the BBPMF values with r value equal to 0.85 and a slope value equal to 0.9 (Fig. 7): The slope value close to 1 indicates a tight correspondence between the data series obtained with the two independent methods. BBPMF varies from 1.0 µg m–3 to 24 µg m–3 with a mean value of 7.8 µg m–3, while BBLevo, calculated with FW3, from 1.7 µg m–3 to 21 µg m–3 with a mean value of 8.5 µg m–3. The relative contribution to gravimetric PM10 is 25% and 26% for BBPMF and BBLevo respectively.


Fig. 7. BBPMF and BBLevo FW (11, 19, 8) linear correlation.Fig. 7. BBPMF and BBLevo FW (11, 19, 8) linear correlation.

BB is the main source of PM in the studied area, accounting for about 30% (range from a minimum of 10% to a maximum of 50%). Nava et al. (2015) reported a comparable value of BB contribution to PM10 (37%) for a Centre of Italy site, confirming the relevant contribution of BB in suburban areas of Italy.

Even though levoglucosan may undergo atmospheric degradation processes, in our case we can speculate a relative chemical stability considering the agreement with the calculated daily BB mass contribution by the two methods.

As already explained in Section “Biomass Burning Identification Methods: Positive Matrix Factorization (PMF) and Macro Tracer Approaches”, the FW3 assumes an olive emission coefficient intermediate between oak and beech ones. Since in the studied area olive is the most burned wood species (about 70%; Spagnolo et al., 2014) we tried to determine a hypothetical olive emission factor x from the data of the daily BBPMF contributions:

 

Calculated olive emission factor results on average equal to 12. This value, derived from mathematical calculations and theoretical approximations, should be confirmed by further experimental emission tests on olive wood but represents to our knowledge the first proposed conversion factor for this wood species. Interestingly Fourtziou et al. (2017) reported a relation factor between levoglucosan and night PM1 equal to 11.25 as representative for samplings in Greece where olive wood biomass is used for domestic heating. However, Maenhaut et al. (2016) reported a conversion factor for Flanders region equal to 22.6. This value is higher than the values here considered representative for olive biomass burning but is in line with other conversion factors reported for other northern Europe regions: In Finland, Saarnio et al. (2012) reported a factor equal to 24, equal to median value reported in Switzerland (Gianini et al., 2013). Higher conversion factors have been also reported, for example by Ducret-Stich et al. (2013) for a Swiss Alpine valley (factor equal to 37).


CONCLUSIONS


Gravimetric measurements and detailed chemical characterizations of PM were performed in the study area to identify the main emission sources, with particular attention to BB. The daily values for PM10 varied between 9 µg m–3 and 51 µg m–3, with the fine fraction representing, on average, about 70% of the concentration.

PMF was shown to be a good method for quantifying the emission sources, although the difference between the sum of the reconstructed speciation parameters (Σspec) and the experimentally measured gravimetric PM10 remains an issue to be solved. On the other side, the macro tracer approach, developed and applied mostly in the Alpine region, easily facilitated the evaluation of BB, although the available conversion factors (FW) do not cover the entire variety of woody biomass burned in Europe.

These two methods (PMF and the macro tracer approach) showed good agreement, as confirmed by the correlation coefficient (r = 0.85) and the slope (close to 1). The values for the mean concentration as well as the minimum and maximum values from the 24-h BBPMF and BBLevo series were, in fact, comparable: 7.8 vs. 8.5 µg m–3 for the average, 1.0 vs. 1.7 µg m–3 for the minimum and 24 vs. 21 µg m–3 for the maximum. BB was identified as one of the most significant sources of PM10 in the study area, contributing, on average, slightly less than 30%, with a maximum of 50% and a minimum of 10%.

Comparing the results from these two independent methods also allowed us to determine a possible conversion factor that is specific to the studied area and potentially representative of olive wood burning. Furthermore, the good agreement between the methods led us to identify and exclude an underestimation of the BB contribution in the macro tracer approach due to levoglucosan degradation. Levoglucosan is thus confirmed as a valid tracer for BB during winter in the examined area, exhibiting a daily-cycle (over 8 h) that corresponds to the expected house-heating/cooking activities, with higher concentrations during the late afternoon and night.

Integrating these two methods therefore results in a good methodology for realistically and independently quantifying BB’s contribution to PM. Applying this approach to other case studies may further confirm the utility of this dual evaluation and of identifying specific conversion factors (Fw). These conversion factors, mathematically identified and demonstrated with ad hoc tests, can also be used in simple and cheap procedures to routinely evaluate BB in the long term, increasing our knowledge of BB’s impact on air quality in every region of Europe. Furthermore, establishing a database of Fw values that are representative of different geographical areas may enable the development of automated instruments for the real time quantification of BB.

The two methods, each with its specific strengths and weaknesses, converge when evaluating BB in the studied area. We recommend that this integrated approach be considered a state of the art procedure for the quantification of BB.


ACKNOWLEDGEMENT


This work was supported by ENEL s.p.a. The authors wish to thank Dr. Carmela Tortorella, Dr. Silvana Iacobellis, Dr. Aldo Giove, Mr. Renato Nacci and Dr. Daniele Marseglia for the precious support throughout the study.



Share this article with your colleagues 

 

Subscribe to our Newsletter 

Aerosol and Air Quality Research has published over 2,000 peer-reviewed articles. Enter your email address to receive latest updates and research articles to your inbox every second week.

5.9
2020CiteScore
 
 
81st percentile
Powered by
Scopus

2020 Impact Factor: 3.063
5-Year Impact Factor: 2.857

Aerosol and Air Quality Research (AAQR) is an independently-run non-profit journal that promotes submissions of high-quality research and strives to be one of the leading aerosol and air quality open-access journals in the world. We use cookies on this website to personalize content to improve your user experience and analyze our traffic. By using this site you agree to its use of cookies.