lable at ScienceDirect

Atmospheric Environment 120 (2015) 15e27

Contents lists avai

Atmospheric Environment

journal homepage: www.elsevier .com/locate/atmosenv

Particulate and gaseous emissions from the combustion of differentbiofuels in a pellet stove

E.D. Vicente a, M.A. Duarte a, L.A.C. Tarelho a, T.F. Nunes a, F. Amato b, X. Querol b,C. Colombi c, V. Gianelle c, C.A. Alves a, *

a Centre for Environmental and Marine Studies, Department of Environment and Planning, University of Aveiro, 3810-193 Aveiro, Portugalb Institute of Environmental Assessment and Water Research, IDAEA, Spanish Research Council (CSIC), C/Jordi Girona 18-26, 08034 Barcelona, Spainc Regional Centre for Air Quality Monitoring, Environmental Monitoring Sector ARPA Lombardia, 20129 Milan, Italy

h i g h l i g h t s

� The highest emission factors were obtained for agro-fuels.� Organic carbon contributed no more than 30% of the PM10 mass.� Mannosan and galactosan were not detected in almost all samples.� Treated wood in pellets generated high contents of Pb, Zn and As in PM10.

a r t i c l e i n f o

Article history:Received 22 January 2015Received in revised form18 July 2015Accepted 22 August 2015Available online 28 August 2015

Keywords:Pellet stovePM10

OC/ECAnhydrosugarsInorganic species

* Corresponding author.E-mail address: celia.alves@ua.pt (C.A. Alves).

http://dx.doi.org/10.1016/j.atmosenv.2015.08.0671352-2310/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

Seven fuels (four types of wood pellets and three agro-fuels) were tested in an automatic pellet stove(9.5 kWth) in order to determine emission factors (EFs) of gaseous compounds, such as carbon monoxide(CO), methane (CH4), formaldehyde (HCHO), and total organic carbon (TOC). Particulate matter (PM10)EFs and the corresponding chemical compositions for each fuel were also obtained. Samples wereanalysed for organic carbon (OC) and elemental carbon (EC), anhydrosugars and 57 chemical elements.The fuel type clearly affected the gaseous and particulate emissions. The CO EFs ranged from 90.9 ± 19.3(pellets type IV) to 1480 ± 125 mg MJ�1 (olive pit). Wood pellets presented the lowest TOC emissionfactor among all fuels. HCHO and CH4 EFs ranged from 1.01 ± 0.11 to 36.9 ± 6.3 mg MJ�1 and from0.23 ± 0.03 to 28.7 ± 5.7 mg MJ�1, respectively. Olive pit was the fuel with highest emissions of thesevolatile organic compounds. The PM10 EFs ranged from 26.6 ± 3.14 to 169 ± 23.6 mg MJ�1. The lowestPM10 emission factor was found for wood pellets type I (fuel with low ash content), whist the highest wasobserved during the combustion of an agricultural fuel (olive pit). The OC content of PM10 ranged from8 wt.% (pellets type III) to 29 wt.% (olive pit). Variable EC particle mass fractions, ranging from 3 wt.%(olive pit) to 47 wt.% (shell of pine nuts), were also observed. The carbonaceous content of particulatematter was lower than that reported previously during the combustion of several wood fuels in tradi-tional woodstoves and fireplaces. Levoglucosan was the most abundant anhydrosugar, comprising 0.02e3.03 wt.% of the particle mass. Mannosan and galactosan were not detected in almost all samples.Elements represented 11e32 wt.% of the PM10 mass emitted, showing great variability depending on thetype of biofuel used.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Biomass combustion has been encouraged aiming at reducingfossil fuel consumption. However, high emissions from incomplete

fuel combustion in small-scale appliances like woodstoves andfireplaces have been reported in several countries (Fine et al., 2004;Gonçalves et al., 2010; Schmidl et al., 2008). Other technologies areavailable for domestic heating purposes with advantages from theemission point of view. Small scale pellet heating systems areinstalled in rising tendency. The wood pellet market has experi-enced a large growth in recent years as a result of the EU objective

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e2716

to increase the share of renewable energy (Goh et al., 2012). The useof pellets as fuel in small scale appliances for heating purposes hasbeen pointed out as suitable in order to reduce the emissions fromthis sector (Pettersson et al., 2010). Between 2008 and 2010, theproduction of wood pellets in EU increased by 20.5%, reaching 9.2million tons in 2010, representing 61% of the global production. Inthe same period, the EU wood pellet consumption increased by43.5% to reach over 11.4 million tons in 2010, representing nearly85% of the global wood pellet demand. In the segment of residentialheating, the main drivers for market expansion are often indirectsupport measures for the installation of pellet stoves and boilers, aswell as the relative cost competitiveness of wood pellets comparedto traditional fuels, such as LPG heating oil and natural gas, espe-cially in rural areas that are not yet supplied by gas grids (Cocchiet al., 2011). For example, in Spain, the Ministry of Economy hasencourage the installation of boilers as part of the 2004e2012Energy Plan, which aimed to promote the use of biomass, such aspellets, olive pit and almond shell, as an energy source. In Portugal,Spain and other southern European countries, the cork and olive oilsectors generate large amounts of residues that can be used as fuelfor heating in small scale appliances (Garcia-Maraver et al., 2014).In Portugal, the potential for producing pellets through the use ofagricultural residues is recognised. The energy potential derivedfrom almond residues, for example, was estimated to be93 kton y�1. Cereal straw and pruning residues are the agro-residues with more energy potential (Monteiro et al., 2012). Agri-cultural fuels present different physical and chemical characteris-tics compared with woody fuels (Verma et al., 2011; Obernbergerand Thek, 2004). For that reason the combustion of these fuels insmall scale appliances can be a challenge (Carvalho et al., 2008).Pellet appliances are sensitive to variations of the ash-forming el-ements in the fuel due to slag formation on the burner, which isproblematic with respect to combustion efficiency (€Ohman et al.,2004), due to critical inorganic elements, such as alkaline metals,that give rise to increased particulate emissions (Carvalho et al.,2013). Ash accumulation and slag formation can even lead to shutdown of the appliance (Carvalho et al., 2008). The chemicalcomposition of the fuels has also influence on the ash meltingbehaviour, deposit formation and corrosion (Obernberger andThek, 2004).

Although good combustion conditions lead to lowest particulateemissions, several studies have reported highest oxidative stress,inflammatory, cytotoxic and genotoxic activities and decreasedcellular metabolic activity from particles generated under efficientcombustion conditions rather than particles resulting fromsmouldering combustion (Happo et al., 2013; Uski et al., 2014).Besides the health effects, biomass combustion particles are effi-cient cloud condensation nuclei and can influence the formation ofprecipitation (P€oschl, 2005; Rose et al., 2010). In spite of theincreasing use of pellet stoves, their emissions are poorly typified.The variability in properties of biomass is great and may signifi-cantly influence the efficiency and environmental impacts associ-ated with their use, constituting an issue of great research interest.

The aim of this research is to characterise the emissionsresulting from a Portuguese model of pellet stove with growingmarket share in the residential sector, where different solid biofuels(four types of pellets, olive pit, almond shell and shell of pine nuts)have been burnt. The work comprises extensive quantitative andqualitative data for gaseous compounds and particulate matter(PM10). Particulate samples were analysed for organic (OC) andelemental carbon (EC), anhydrosugars and 57 elements. Acomparative analysis of particulate phase emissions of parentpolycyclic aromatic hydrocarbons (PAHs) and their derivatives(alkyl-PAHs, oxygenated-PAHs, azaarenes and nitrated-PAHs) fromthe combustion of these biofuels in the pellet stovewith those from

amanually fired appliance can be found elsewhere (Vicente et al., inpress).

2. Experimental work

2.1. Combustion equipment, fuels and experimental procedure

The combustion experiments were carried out using a top-feedpellet stove with a nominal output of 9.5 kW, manufactured inPortugal by Solzaima, model Alpes (Fig.1). The stove has an internalpellet storage tank with 20 kg capacity and the fuel is supplied byan auger screw to the top feed burner. The primary air is suppliedthrough holes in the bottom of the grate, and secondary air is feedabove the grate through three holes. Both primary and secondaryair are driven by an electric fan located downstream the combus-tion chamber. The primary air flow rate was monitored continu-ously during the combustion cycle using a mass flow meter. Thetemperature was measured continuously using K-type thermo-couples located at several points along the combustion and exhaustsystem (combustion chamber, chimney and dilution tunnel). Thestove can be operated at five levels of power output by automati-cally modifying the fuel feed rate and the exhaust gas fan speed. Inorder to cover different behaviours by users, the emission factors(EFs) of distinct gaseous components and particulate matter (PM10)were determined for three levels of power output (lowest, mediumand highest). Since the differences in emissions between differentlevels of power output for the same fuel were not statistically sig-nificant, average EFs were obtained for each compound. The igni-tion of the fuel is made through an electrical resistance located onthe grate of the stove. The ignition phase was not included in theresults.

It should be pointed out that a short cleaning period of the grateis programmed to occur. During the cleaning process, the fuelsupply decreases and the combustion air supply increases for a fewminutes in order to remove the bottom ashes from the grate; thus,the lighter ash fractions are then carried out with flue gas.

Each of the combustion experiments was performed after apreheating period of the pellet stove. When the level of poweroutput was changed, it took about 40 min to start the experimentsto ensure that the combustion process had already attained a newsteady operation condition.

The feeding rate for each fuel was evaluated by prior calibrationof the screw feeding system for the three levels of power output.Also, in all experiments, pre-weighed fuel was poured into the fuelhopper and after combustion the remaining fuel was re-weighed, inorder to verify the fuel consumption rate. The feed rate showedgreat variations among fuels (Table 2). This can be related to thedifferences in the physico-mechanical properties of the fuels, suchas the diameter and length, the bulk density, the fine content andthe mechanical durability (Carvalho et al., 2013; Verma et al., 2011).

To investigate the influence of fuel quality on emissions, fourtypes of pellets were selected. These biofuels were selected basedon their commercial availability in order to represent the widerange of pellets on the Portuguese market. Pellets type I werecommercial wood pellets made of golden wattle, cedar and pine.Pellets type II were composed of 75% of lignocellulosic residuesand 25% of dust from the furniture manufacturing industry.Pellets type III were composed of 65% of lignocellulosic residuesand 35% of dust from the furniture manufacturing industry.Pellets type IV were made with a mixture of 50% of wastewoodchips (several woods from construction and demolition,pine wood pallets, forest biomass, paper and paperboard) and50% of dust from the furniture manufacturing industry. In addi-tion, EFs for the combustion of olive pit, almond shell and shell ofpine nuts were obtained. This characterisation was carried out at

G

A

C

E

F

H

I

B

E

A D

N

M

L

K

J

Fig. 1. Schematic representation of the experimental installation. A e Stove; B e Combustion chamber; C e Grate of the stove; D e Air flow meter; E e Exhaust duct (Chimney); F e

Gas sampling and analysis system; G e Water-cooled gas sampling probe; He Gas sampling pump; I e Gas condensation unit; J e Pitot tube; Ke Dilution tunnel; L,M e TECORAPM10 sampling system; N e Fan.

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e27 17

the request of the Madrid City Council, because in the Autono-mous Region these agro-fuels have shown a rapidly increasingpenetration in the residential heating sector due to the imple-mentation of the national Energy Plan. Fuel properties, includingmoisture, C, H, N, O, S and ash content, and also lower heatingvalue (LHV), are listed in Table 1. Moisture and ash content ofwood pellets for non-industrial use are two of the parameters

established by the European norm EN 14961-2. Pellets type III, IV,olive pit and shell of pine nuts presented moisture contenthigher than the limit of 10% established by the norm. The ashcontent of pellets type II and III was higher than 3%, themaximum allowed by the norm. The nitrogen content of pelletstype II, III and IV was also higher than the maximum valueallowed by the norm (1 wt. %). The content of major and trace

Table 1Characteristics of the biofuels used in the experiments.

Pellets type I Pellets type II Pellets type III Pellets type IV Olive pit Shell of pine nuts Almond shell

Proximate analysis (wt.%, as received) Moisture 8.4 8.8 10.9 10.7 12.9 12.9 9.5Ultimate analysis (wt.%, dry basis) Ash 0.73 3.2 3.8 2.0 0.66 1.3 1.4

C 49.7 47.4 48.3 47.4 50.9 49.8 49.3H 6.9 6.58 6.53 6.79 6.59 6.59 6.76N 0.16 2.31 2.06 2.11 0.21 0.30 0.34S <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01O (by difference) 42.5 40.5 39.3 41.7 41.6 42.0 42.

LHV (MJ kg�1) 18.3 17.5 17.4 17.8 18.5 18.7 18.4

wt.% oxide Al2O3 0.02 0.1 0.1 0.1 <0.01 0.04 0.03CaO 0.2 0.7 0.7 0.3 0.1 0.2 0.3Fe2O3 0.02 0.1 0.2 0.1 0.01 0.1 0.1K2O 0.1 0.1 0.1 0.1 0.3 0.2 0.3MgO 0.03 0.1 0.1 0.1 0.02 0.1 0.1Na2O 0.01 0.1 0.1 0.1 <0.01 0.02 0.1P2O5 0.01 0.04 0.04 0.1 0.03 0.03 0.1SO3 0.03 0.2 0.2 0.2 0.03 0.1 0.05

mg kg�1 of dry fuel Zn 92 181 304 76 3.90 31 57As 1.8 3.8 7.3 1.9 16 <0.5 1Pb 2.3 119 207 19 4.8 8.6 4.4Cr 4.2 4.9 7 4.8 <0.5 2.6 3.7Cu 1.5 10 17 7.1 5.0 8.7 8.6

Table 2Operating conditions, PM10 emission factors and carbonaceous content of particulate matter.

Fuel Number ofexperiments

Fuel feedrate (kg h�1)

Dilutionratio

O2 concentration (% v, drygases) in the exit flue gas

Flue gas temperature (�C) inthe combustion chamber

PM10

(mg MJ�1)OC(mg MJ�1)

EC(mg MJ�1)

OC/EC

Pellets type I 4 1.18 ± 0.17 35.2 ± 9.2 16.9 ± 1.3 683 ± 48.9 26.6 ± 3.14 7.40 ± 1.86 1.77 ± 0.44 4.2Pellets type II 9 1.40 ± 0.23 29.9 ± 3.4 16.6 ± 0.39 681 ± 42.9 86.4 ± 13.6 14.3 ± 8.46 13.9 ± 8.32 1.0Pellets type III 7 1.12 ± 0.17 26.0 ± 3.4 18.2 ± 0.26 743 ± 10.1 102 ± 8.63 8.38 ± 1.21 9.67 ± 2.81 0.9Pellets type IV 6 1.44 ± 0.10 29.7 ± 6.6 15.7 ± 0.53 800 ± 3.10 75.6 ± 9.39 15.2 ± 6.28 8.89 ± 2.07 1.7Olive pit 6 0.88 ± 0.07 31.0 ± 4.9 17.0 ± 0.33 828 ± 57.2 169 ± 23.6 48.7 ± 30.9 5.18 ± 3.51 9.4Shell of pine nuts 8 0.96 ± 0.18 30.4 ± 12.1 18.4 ± 0.17 623 ± 48.8 117 ± 33.9 18.7 ± 3.12 54.5 ± 23.5 0.3Almond shell 6 1.36 ± 0.29 25.4 ± 3.1 17.2 ± 0.42 876 ± 34.4 112 ± 4.05 12.1 ± 3.00 7.00 ± 0.94 1.7

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e2718

elements in the fuels were determined by Inductively-CoupledPlasma Atomic-Emission Spectrometry (ICP-AES) and byInductively-Coupled Plasma Mass Spectrometry (ICP-MS),respectively, following the same methodology described in Sec-tion 2.3 for the particulate matter filters. Pellets (made up ofwood and wood wastes) are characterised by high Ca and K, dueto the occurrence of wewellite, K salts and organic-bearing K. Theuse of wood preservatives, such as chromated copper arsenate(CCA), give rise to relatively high and soluble concentrations ofCr, Cu and As. Olive pit, almond shell and shell of pine nuts showhigher K than Ca concentrations and relatively low content ofmost trace elements with the exception of As in olive pit(Table 1).

2.2. Flue gas measurements

The combustion flue gas was sampled at the exit of the chimneyby means of a heated (at 180 �C) sampling line and conducted to anonline Fourier Transform Infrared Gas analyser (FTIR Gasmet,CX4000). This equipment has a multicomponent measurementcapability, which enables, among others, the real-time monitoringof CO, TOC, CH4 and HCHO. TOC includes CH4, ethane (C2H6), ethane(C2H4), propane (C3H8), hexane (C6H14) and HCHO. The sample cellheating ensures that high water vapour concentrations or corrosivegases will not pose a problem. Since the water content of thesample gas is continuously measured with the Gasmet™, the re-sults can be reported on either a “wet” or “dry” basis. The con-centration of several gaseous compounds was recorded as wet gas,and then corrected to dry basis based on the H2O concentrationmeasured.

2.3. Particulate matter sampling and analysis

Particulate matter (PM10) was collected after the dilution of theexhaust with atmospheric air in a tunnel under isokinetic condi-tions. The sampling point was located 10 m downstream the dilu-tion tunnel entrance. The dilution systemwas described in previousstudies (Alves et al., 2011; Gonçalves et al., 2010; Calvo et al., 2014).The dilution ratio (DR) was defined as the ratio between the totaldiluted flow (in the dilution tunnel) and the undiluted flue gas flow(in the chimney) and ranged from 25.4 to 35.2 (Table 2). Thesampling train included a PM10 inlet head, a pump, and a controland data acquisition system, all part of a TCR TECORA (model2.004.01) instrument. The equipment has been operated at a flowof 2.3 m3 h�1. The particulate matter sampling was made understeady state operating conditions, which have been evaluated bycontinuous monitoring of flue gas composition. A cascade impactorTCR Tecora (model MSSI) was used to obtain size segregated par-ticulate matter samples. For each fuel, sampling with cascadeimpactor was carried out at each power of level output. Thesesamples were collected in parallel with the PM10 filters that havebeen used to perform the chemical characterisation reported in thepresent study. The multistage cascade impactor, designed accord-ing to ISO23210-2009 norm for the nozzle dimensioning, allowsseparation into 3 particulate size fractions: >10 mm, between 2.5and 10 mm and below 2.5 mm. The sampling flow rate was 3 m3 h�1.The particulatematter (PM10) samples for gravimetric and chemicalanalyses were collected on 47 mm diameter quartz fibre filters pre-baked at 500 �C for 6 h, to remove organic contaminants. The filterswere kept in a desiccator to stabilise without hydration orcontamination. The gravimetric quantification was performed with

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e27 19

a microbalance (RADWAG 5/2Y/F with an accuracy of 1 mg). Filterweight was obtained from the average of six measurements, whenthe variations were less than 0.02%.

The organic (OC) and elemental carbon (EC) content of PM10 wasanalysed by a thermal optical transmission technique. This methodallows the differentiation of various particulate carbon fractionsthrough volatilisation and oxidation to CO2 under controlledheating. The CO2 produced is then quantified in a non-dispersiveinfrared (NDIR) analyser. The blackening of the filter is monitoredusing a laser beam and a photodetector, which enables separatingthe EC formed by pyrolysis. A more detailed description of themethod can be found elsewhere (Gonçalves et al., 2014).

The analysis of major and trace elements was performed by ICP-AES and ICP-MS. For each biofuel, punches from the various repli-cate filter samples obtained for the three levels of power outputwere combined and analysed together in order to obtain meanvalues for each combustion condition. As proposed by Querol et al.(2001), each set of filter punches was subjected to acid digestion(1.25 mL HNO3: 2.5 mL HF: 1.25 mL HClO4). Three multi-elementalsolutions Spec® 1 (rare earth elements, REE), Spec® 2 (alkalis, earthalkalis, and metals) and Spec® 4 (Nb) were used to constructexternal calibration curves. The mean precision and accuracy werecontrolled by repeated analysis of 0.025 mg of NBS-1633b (fly ash)reference material (NIST, Gaithersburg, MD, USA). They fall belowtypical analytical errors and are in the ranges of 3e5% and <10%, forICP-AES and ICP-MS, respectively. The detection limits were 0.01 ngm�3 for most of the trace elements analysed.

In order to quantify levoglucosan and its isomers, punches ofvarious replicate filter samples obtained for each level of poweroutput were extracted with milliQ water during 1 h in an ultrasonicbath. Determinations of levoglucosan were performed using aMetrohm 881.0020 Compact IC Pro ion cromatographic systemequipped with the amperometric 896.0030 Professional DetectorMetrohm (Metrohm IC Application Work, 2008, 2009). The Met-rosep Carb 1 e 150/4.0 column was used, with another columnASupp15 above, allowing a better separation of mannosan andgalactosan. The columns are thermostated at 40 �C. As eluent it wasused, NaOH 70mM, prior outgassing before the application in orderto remove CO2. The elution flow was 0.6 mL min�1. The systemwascalibrated using a solid standard of levoglucosan (ACROS OR-GANICS, >99%) diluted in milliQ water. Different concentrations ofstandard, from 0.01 to 10 ppm, were prepared.

3. Results and discussion

3.1. Gaseous emissions

Gaseous emissions are greatly affected by the biofuel type(Fig. 2). Excepting pellets type I, higher gaseous emissions foragricultural fuels than for wood pellets were observed. Ashes ofpellets type I presented a much higher C content (27%) comparedwith that of pellets type II, III, and IV (1.5e3.7%). The t-test(a ¼ 0.05) revealed that the difference in the CO EFs from pelletstype I and the other three types of wood pellets is statisticallysignificant (p < 0.0001). On the other hand, the differences betweenemissions from pellets type I and two of the three agricultural fuelswere found to be statistically insignificant (p ¼ 0.9631 andp ¼ 0.1883 for shell of pine nuts and almond shell, respectively).

The CO EFs ranged from 90.9 ± 19.3 (pellets type IV) to1480 ± 125 mg MJ�1 (olive pit). The high CO EFs observed foragricultural fuels indicate poor combustion conditions. The CO EFsobtained in the present study for agricultural fuels are in the rangeof values obtained for conventional batch combustion(830e2700 mg MJ�1) (Lamberg et al., 2011a). Wood pellets presentlower bulk densities than agricultural fuels allowing favourable air

excess in the combustion chamber, since lower density fuel deliversless energy into the combustion chamber (Verma et al., 2012). TheCO EF was about 16 higher for olive pit than for pellets type IV.Carvalho et al. (2013) reported eleven to fourteen times higher COemissions from several agricultural fuels than from wood pellets.Although the difference between agricultural and wood fuels isclose to the difference documented by Carvalho et al. (2013), the COEFs are much higher in the present study (Table 3). The differencesbetween the results of the present study and literature data can beassociated with the pelletisation of the agricultural fuels, which cancontribute to improved feeding into the stove and, consequently, tomore favourable operating conditions. Fuel blending can poten-tially contribute to reduce emissions from the combustion of agri-cultural fuels (Miranda et al., 2012). In the present study, thechemical composition of the fuel was found to have influence onthe CO EFs. The CO EFs were negatively correlated with the AlO3(R2 ¼ 0.88) and the SO3 (R2 ¼ 0.79) and positively correlated withthe K2O (R2 ¼ 0.71) contents in the fuels.

The difference in the gaseous emissions between agriculturaland wood pellets can be related to the formation of ash agglom-erates in the grate of the stove, creating an irregular air combustionflow rate through the fuel bed, which has a negative effect in theoxidation of the gaseous compounds. The difference between olivepit CO EF and wood pellets (type I, II, III and IV) is statistically sig-nificant (p < 0.0001). The difference in the CO EFs from almondshell and shell of pine nuts is not statistically significant(p ¼ 0.2302), in contrast to the statistically significant difference inthe EFs from olive pit and the other two agricultural fuels(p ¼ 0.0001 and p ¼ 0.0003 for shell of pine nuts and almond shell,respectively). The higher moisture content of the agricultural fuelsmay also have led to poor combustion conditions and consequentlyto higher CO EFs. Verma et al. (2012) used a multi-fuel boiler(40 kW) to test peat pellets, apple pellets, wheat straw pellets withhydrated lime as additive and sun flower husk pellets. The authorsobserved that the CO emissions for the combustion of peat pelletswere 50.0, 6.1 and 2.5 times higher than those of apple, straw andsun flower husk, respectively. The results obtained in the presentand previous studies highlight the importance of the fuel incontinuous combustion appliances. These devices are designed forwood pellets with a specific size and shape and for that reason thefeeding system and combustion chamber may not be prepared foralternative fuels. Such fuels are chemically different than woodpellets presenting higher concentrations of ash forming elementsand therefore causing operating problem that maximise CO emis-sions concomitantly with slagging on the grate of the stove.Carvalho et al. (2013) investigated the performance of a pelletboiler fired with agricultural fuels and found that the combustionchamber was not optimal for these fuels. Furthermore, the authorsalso reported that the boiler was incapable to adjust an optimalexcess air ratio and the efficiency was gradually reduced due to ashaccumulation on the heat exchangers in consecutive combustionexperiments.

With respect to TOC (Fig. 2), wood pellets generated lower EFsthan other fuels. The emission values ranged from 2.2 ± 0.2 (pelletstype I) to 68.7 ± 11.4 (olive pit) mg MJ�1. Olive pit was the fuel withhighest EF, which is possibly related to the fact that olive fruit has astony pit containing 30% lipids and 20% carbohydrates. Johanssonet al. (2004) compared the emissions from different types ofcombustion appliances. The authors observed that the operationand combustion appliance exert a great influence on TOC emis-sions. These variations are reflected in the discrepancies betweenresults. In our study, it was observed that the fuel characteristicsalso play a major role in TOC emissions. As observed by Johanssonet al. (2004), the TOC emissions correlated with the CO emissions(R2 ¼ 0.86). Similar to CO EFs, the difference in the TOC EFs from

Fig. 2. CO, TOC, HCHO and CH4 emission factors for the combustion of the distinct biofuels.

Table 3Comparison between CO EFs from this study and literature values.

Appliance Fuel CO EF (mg MJ�1) Reference

Pellet stove (8 kW) Low-quality cheap pellets and high quality pellets with DIN-PLUS certification 88 Ozgen et al. (2014)Pellet boiler (25 kW) 350Pellet stove (8 kW) Commercial pine stem and spruce stem pellets 101e201 Sippula et al. (2007)

Alder, birch, and willow stem pellets 207e355Pellet boiler (25 kW) Pine steam wood pellets 3.55e335a Lamberg et al. (2011b)Pellet boiler (15 kW) Hay pellets 280 Carvalho et al. (2013)

Wheat bran pellets 224Straw pellets 223

Pellet stove (9.5 kW) Wood pellets 90.9e740 This studyAgro fuels 732e1480

a Different operating conditions (boiler load, primary and secondary air supplies).

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e2720

pellets type I and the other three types of wood pellets is statisti-cally significant (p < 0.0001), as well as the difference between theTOC EFs for olive pit and wood pellets (p < 0.0001). The higheremissions of TOC from agricultural fuels are, at least in part, due tofuel characteristics, which impair the combustion efficiency, asdiscussed above for CO. Differences statistically insignificant werefound when comparing the TOC EFs from pellets type I and almondshell (p¼ 0.1981), pellets type IV and pellets type II (p¼ 0.1190) andalmond shell and shell of pine nuts (p ¼ 0.2664).

Formaldehyde (HCHO) is known to induce severe poisoning,irritation and other immune toxic effects (Tang et al., 2009). Themeta-analysis by Zhang et al. (2009) support an association be-tween formaldehyde and leukemia. Cerqueira et al. (2013) con-ducted experiments using five different types of common woodspecies growing in Portugal in order to characterise formaldehydeand acetaldehyde emissions from residential combustion in awoodstove. Assuming a heating value of 18 MJ kg�1 of dry biofuel(Telmo et al., 2010), EFs for formaldehyde were estimated to be

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e27 21

between 36.3 and 98.4 mg MJ�1. In the present study, the formal-dehyde EFs ranged from 1.01 ± 0.1 (pellets type II) to 36.9 ± 6.3(olive pit) mgMJ�1 (Fig. 2). Formaldehyde emissions have shown todepend on the combustion appliance (Tissari et al., 2007). In thepresent study, it was observed that the type of fuel also plays amajor role on the emissions of this compound. Formaldehyde is animportant product in biomass pyrolysis at temperatures higherthan around 200 �C (Zhang et al., 2011). The differences betweenformaldehyde EFs may be due to dissimilar cellulose and lignincontents in the fuels (Gonz�alez et al., 2009). Furthermore, variablecombustion efficiencies also influence these emissions (Yokelsonet al., 1997). The difference in the HCHO EFs from pellets type Iand the other three types of wood pellets is statistically significant(p < 0.0001), as well as the difference between the HCHO EF forolive pit and the other fuels. No significant differences were foundbetween the EFs from almond shell and shell of pine nuts(p ¼ 0.2067) or between pellets type I and shell of pine nuts(p ¼ 0.6324). The HCHO EFs correlate linearly with the TOC EFs(R2 ¼ 0.97) and with CO EFs (R2 ¼ 0.84).

One of the most important volatile organic compound (VOC)from biomass combustion is methane (CH4), which is an effectivegreenhouse gas (Obaidullah et al., 2012). Johansson et al. (2004)compared the emissions from commercial residential boilers firedwith wood logs and wood pellets. In their study, it was observedthat, in all cases, methane was the VOC with highest concentration.Pettersson et al. (2011) characterised the emissions from a wood-stove with variations in fuel, appliance and operating conditions.The authors also found that methane was the predominant VOCwith EFs ranging from 9 to 1600 mg MJ�1. Tissari et al. (2008a)characterised the effect of two different combustion conditionson particulate and gaseous emissions from a conventional masonryheater, namely normal combustion and smouldering combustion.The authors found that the emission of methane from smoulderingcombustion was 11-fold those from normal combustion. Thus, CH4emissions strongly depend on the combustion conditions. Olssonand Kj€allstrand (2006) studied the emissions of organic com-pounds from wood burning in a modern eco labelled residentialboiler (30 kW). The authors verified that CH4 emissions decreasewith increasing combustion efficiency. Leskinen et al. (2014) alsoobserved differences among the CH4 EFs from distinct combustionconditions, reporting average values of 0.3, 4.3 and 115 mgMJ�1 forefficient, intermediate and smouldering combustion, respectively.In the present work, the CH4 EFs ranged from 0.23 ± 0.03 (pelletstype V) to 28.7 ± 5.7 (olive pit) mg MJ�1 (Fig. 2). The statisticalanalysis revealed that the difference between the olive pit CH4 EFand all the other fuels is statistically significant (p < 0.0001), as wellas the difference between the pellets type I EF and the other woodpellets (p < 0.0001). A clear linear correlation was found betweenthe CH4 and TOC EFs (R2 ¼ 0.99) and CO EFs (R2 ¼ 0.87).

3.2. PM10 emission factors

Particulate matter emissions from residential wood combustionfor heating purposes is one topic of great concern. With theexception of pellets type II and shell of pine nuts, the size-segregated samples indicated that a mass fraction higher than80% of the total particulate matter emitted is concentrated in thefinest range. The very fine particles in smoke can go deep into thelungs. The technical characteristics of residential combustion ap-pliances play a major role on emissions. Appliances may be dividedinto two groups: manually fed batch-combustion systems andautomatically fed systems. The fuel properties may also play asignificant influence on the amount and composition of particleemissions (Fernandes et al., 2011; Schmidl et al., 2008; Lamberget al., 2011a,b; Sippula et al., 2007). The transient combustion

phases were not included in the results of the present study.However, it is important to take into account that the unsteadyphases can significantly affect the emissions. Toscano et al. (2014)reported that unsteady state phases, such as the startup phase,can significantly affect the emission factors. In their study, the PMconcentration was up to 72% more than those measured in steadystate condition.

PM10 EFs from the combustion experiments are listed in Table 2.The EFs ranged from 26.6 ± 3.14 (pellets type I) to169 ± 23.6 mg MJ�1 (olive pit). The difference in the PM10 EFs frompellets type I and the other three types of wood pellets is statisti-cally significant (p < 0.0001), as well as the difference betweenPM10 EFs for olive pit and wood pellets (p < 0.0001). In general,higher EFs were obtained for biomass fuels other than pellets,which may be related to the fuel physical characteristics, such asdimensions (length and diameter), fine content and particle den-sity. These properties are important as regards the fuel supply tothe burner and affect the combustion behaviour (Carvalho et al.,2013; Obernberger and Thek, 2004; Verma et al., 2011), which, inturn, influence the equipment performance and its emissions(Table 2). The pelletisation of agricultural fuels and the use ofblended pellet fuels can contribute to optimise the combustionprocess and subsequently reduce the emissions (Mediavilla et al.,2009; Heschel et al., 1999). Insignificant differences were onlyfound between the PM10 EF from pellets type II and IV (p ¼ 0.1166)and from almond shell and shell of pine nuts (p ¼ 0.7282). Thechemical composition of the fuel can significantly influence theparticulate emissions (e.g. Obernberger and Thek, 2004). Inorganicaerosols formation occurs through nucleation and condensation ofthe ash forming vapours like alkali metals and easily volatile heavymetals. These compounds are released during combustion from thefuel to the gas phase. They have a high vapour pressure and, as soonas the vapour pressure exceeds the saturation pressure, particlesare formed by nucleation or they condense on surfaces of existingparticles (Skotland, 2009; Ingwald and Gerold, 2006). The high EFsobserved for the agricultural biomass fuels may be related to thefact that these fuels present higher K content than woody pellets(Table 1). As observed in previous studies (e.g. Sippula et al., 2007),in the present study the PM10 EFs seems to correlate with the fuelK2O content (R2 ¼ 0.57). Excluding pellets type I, a clear linearcorrelation was found between the PM10 EFs from the other fuelsand the CO EF (R2 ¼ 0.91), the CH4 EFs (R2 ¼ 0.95), the HCHO EFs(R2 ¼ 0.99) and the TOC EFs (R2 ¼ 0.94), indicating that high par-ticulate emissions are associated with high emissions of productsfrom incomplete combustion. Although Southern European coun-tries have not yet imposed standards, stringent particulate matteremission limits for solid fuel heating appliances are currently beingdiscussed in other countries. For example, in Germany, fromJanuary 2015 onwards, stoves have to comply with more stringentemission limit values, ranging between 13 and 27 mg MJ�1,depending on the type of fuel. Except for pellets type I, the emis-sions from all other biofuels far exceed the limits set by the Ger-mans. However, the PM10 EFs observed in the present study are8e10 times lower than those obtained in a traditional fireplace(Alves et al., 2011; Fernandes et al., 2011). Fig. 4 presents a com-parison between the results obtained in previous studies usingdistinct appliances and the results recorded in the present work.Open fireplaces present the highest particulate EFs. The low tem-peratures in these appliances contribute to very inefficient com-bustions. Smouldering is the prevalent combustion phase. Thecontinuous improvement of combustion technologies hascontributed to dramatic reductions in emissions.

Fernandes et al. (2011) found PM2.5 EFs of 699 ± 327, 447 ± 169and 103 ± 50.6 mg MJ�1 for the combustion in a fireplace, wood-stove and eco-labelled woodstove, respectively (Fig. 3). Although

Fig. 4. Chemical composition of PM10 (P

chemical constituents ¼ OC � 1.8 þ EC þOxides), for the combustion of the distinct biofuels.

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e2722

the emissions resulting from the combustion in automatically firedappliances are significantly lower than those from log woodstovesand fireplaces, variations can be observed. These variations may beassociated either with the fuel properties or with the burningmode(full or part load operation) (Schmidl et al., 2008; Garcia-Maraveret al., 2014; Riva et al., 2011; Lamberg et al., 2011b). Garcia-Maraver et al. (2014) found that gaseous and PM emissions aresignificantly affected by the fuel type rather than the boiler thermalload. Ozgen et al. (2014) tested two types of pellets (low qualitycheap pellets and high quality pellets with DIN-PLUS certification)in a pellet stove (8 kW) and boiler (25 kW). The authors reported anaverage emission factor of 109 mg MJ�1 for the pellet stove and of61 mg MJ�1 for the pellet boiler.

Automatically fired appliances allow operating under morestable and efficient conditions. Particulate matter resulting fromthe combustion in these appliances is mainly composed of alkalisalts. The fuel ash content and composition have a great influenceon emissions. Thus, the variability in the PM10 EFs observed in thepresent study for the different fuels may be explained by thedifferent ash contents. Pellets type I showed the lowest emissionfactor and the lowest ash content (0.73 wt.%), whilst pellets type IIIexhibited the highest emission factor and ash content (3.8 wt.%).Sippula et al. (2007) found that the fuel ash content correlatedlinearly with the PM1 emission.

3.3. PM10 chemical composition

Particles resulting from residential wood combustion arecomposed of soot, organic matter and fine fly ash (Obaidullah et al.,2012; Orasche et al., 2012; Bølling et al., 2009; Tissari, 2008). Par-ticles from manually operated small appliances are, in general,dominated by organic matter since the combustion conditions aremore incomplete. Table 2 displays the EFs of carbonaceous con-stituents for the distinct combustion experiments. The OC EFsranged from 7.40 ± 1.86 (pellets type I) to 48.7 ± 30.9 (olive pit)mg MJ�1

. The differences between the OC EFs from pellets type Iand pellets type II and III are not statistically significant (p ¼ 0.1430and p ¼ 0.3118, respectively), and neither are the differences

Fig. 3. Comparison between particulate emission factors (mg MJ�1) from differenttypes of small-scale combustion appliances and the present study.

between pellets type II, III and IV. Statistically significant differ-ences were observed between the EF from pellets type I and IV(p ¼ 0.0451). The difference between the olive pit OC EF and all theother fuels was found to be statistically significant. The EC EFsranged from 1.77 ± 0.44 (pellets type I) to 54.5 ± 23.5 (shell of pinenuts). The shell of pine nuts EC EF is significantly higher than the ECEF from the other agricultural fuels (p ¼ 0.0003 and p ¼ 0.0004, forolive pit and almond shell, respectively). The EC EF from pelletstype I was statistically different from almost all the other fuels. Nostatistically significant difference was found between the pelletstype I and olive pit EFs (p ¼ 0.0947).

Total carbon represented from 17.0 (almond shell) to 62.6 (shellof pine nuts) wt.% of the PM10 mass. The OC EFs were found tocorrelate with the TOC EFs (R2 ¼ 0.84), the HCHO EFs (R2 ¼ 0.78),and the CH4 EFs (R2 ¼ 0.82). Although the carbonaceous content ofparticulate matter in the present study was lower than that ob-tained during the combustion in traditional woodstoves and fire-places (Alves et al., 2011), high OC and EC emissions were observedduring the combustion of several fuels. The EC content in samplesfrom the combustion of shell of pine nuts was much higher thanthat in PM10 from other fuels. Primary soot particles are formedmainly in the flame from hydrocarbons. Depending on the tem-perature and oxidative conditions, primary particles may be burnedor remain in the particulate phase. As a consequence of the insuf-ficient mixing of combustion gases and air, the flame zone alwayscontains fuel-rich areas providing the conditions for the formationof soot particles (Torvela et al., 2014; Obaidullah et al., 2012; Tissari,2008). This mechanism may be the explanation for the high ECcontent in the particulate matter from the combustion of shell ofpine nuts. The difference in the EC content between emissions fromshell of pine nuts and the other two agricultural fuels may also berelated to the alkali salts content of the fuels. Almond shell andolive pit contain much more Na and K. Since alkali metals can havea catalytic effect in the soot oxidation process and thereby increasethe soot burnout rate (Wiinikka, 2015), this may be the reason forthe lower EC EFs compared with shell of pine nuts. It has beendescribed that the composition of the fuel burned plays animportant role on soot formation (Baeza-Romero et al., 2010). Frommodelling results, Fitzpatrick et al. (2008) reported that the amountof soot is dependent on the cellulose/lignin ratio. One of thedecomposition products of lignin is eugenol, which has been foundto greatly contribute to soot formation during wood combustion(Fitzpatrick et al., 2008). The combustion of shell of pine nuts

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e27 23

generated higher amounts of EC in the particulate matter than OC,whereas the opposite was observed for olive pit. It should berecalled that this latter fuel was also the one with highest CO andTOC EFs.

Improved combustion efficiency with higher combustion tem-peratures and flaming combustion contribute to higher EC emis-sions than those resulting from the combustion in appliances withlower efficiency like fireplaces and traditional woodstoves(Fernandes et al., 2011). Gonçalves et al. (2010) tested several fuelsin an eco-labelled log woodstove (6 kW) and found that total car-bon represented 43.9e63.2 wt.% of the PM2.5 mass. Schmidl et al.(2011) observed higher total carbon content in the particulatemass emitted from traditional manually fired appliances with ECcontributing to about 30% and OC to 30e40% of the PM mass.Leskinen et al. (2014) studied the physico-chemical properties ofparticles emitted under three different combustion conditions(efficient, intermediate and smouldering combustion) and reportedthat, during efficient combustion, particles were composed of ash-related material, while during smouldering conditions, they con-tained more EC than OC. The PM1 from intermediate conditionsconsisted mainly of OC and EC in almost equal fractions.

In the present study, the OC/EC ratio ranged from 0.9 to 4.2 andfrom 0.3 to 9.4 for pellets and agro fuels combustion, respectively(Table 2). These values are within the ranges reported in the liter-ature (Table 4), since combustion parameters such as fuel type,temperatures in the firebox, fuel feed rate, etc., are believed tosignificantly influence the emission composition. The burningconditions have great influence on the OC to EC ratios, as can beseen in Table 4. The types of fuel burned, as well as the kind ofappliance, are critical factors affecting the carbonaceous emissions.Zhang et al. (2013) have evaluated prescribed burning versuscontrolled burning in a stove. The authors reported lower OC/ECratios when burning wooden logs compared to combustion ofgreener fuel types and observed lower OC/EC ratios in woodstoveburns compared to prescribed burns (Table 3). Gonçalves et al.(2010) reported OC to EC ratios in the range of 0.9e4.4 in parti-cles emitted from an eco-labelled woodstove and different types ofPortuguese woods (Table 3).

In order to obtain a more precise mass balance, the OCmass wasconverted to total mass of organic matter (OM) using a factor thataccounts for oxygen, hydrogen, and some other atoms present inthe organic material. The adopted OM/OC ratio was 1.8, which istypical for wood smoke aerosols (Turpin and Lim, 2001). Thisconversion allows to reconstruct the PM10 mass taking into accountthe gravimetric data and the mass of all quantified species. The OMfraction varied from 14.8 (pellets type III) to 51.9 (olive pit) wt.% ofthe PM10 mass (Fig. 4).

As displayed in Table 5, levoglucosan, mannosan, and gal-actosan contents in particulate matter samples showed variations,

Table 4Comparison between OC/EC ratios from this study and literature values.

Appliance Fuel

Fireplace Portuguese woods and briquettesWoodstove Wooden logs (mainly pine)

Leaves/duff (Green foliage and branches from twodominant shrubs: manzanita and bitterbrush)

Log woodstove (8 kW) Spruce wood logsLog woodstove (6 kW) Portuguese woodsPellet stove (13 kW) Spruce pelletsPellet boiler (25 kW) Spruce pelletsPellet stove (9.5 kW) Wood pellets

Agro fuels

a Cold start.b Nominal load.

probably due to differences in the relative content of cellulose andhemicellulose in diverse biofuels tested in the present study(Engling et al., 2006). Levoglucosan was the most abundantanhydrosugar, encompassing 0.02e3.03 wt.% of the particle mass.This sugar was detected in almost all samples. For pellets type IIand III, levoglucosanwas only detected during the operation at thelower level of power output. Mannosan and galactosan were ab-sent from almost all samples. The yields of mannosan and gal-actosan decreased more rapidly than those of levoglucosan underincreasing combustion severity (temperature and duration) (Kuoet al., 2011).

Levoglucosan, mannosan and galactosan have been used astracers for residential wood combustion and wildfires (Fine et al.,2004; Engling et al., 2006; Vicente et al., 2012). The levoglucosanto mannosan ratio has been described to be wood type specific,with low ratios for softwoods and higher values for hardwoods(Engling et al., 2006; Schmidl et al., 2008). However, the use oflevoglucosan as a quantitative tracer may be associatedwith large uncertainty since laboratory measurements of levo-glucosan emissions have shown large variations depending onthe type of stove, biofuel quality, and operator's behaviour(Hedberg et al., 2006). Schmidl et al. (2011) studied the influenceof system (manually and automatically fired appliances) andwood type on anhydrosugars emissions. The authors observedthat automatically fired appliances did not emit detectableamounts of anhydrosugars in full- and part-load operation. Lev-oglucosan and mannosan were only detected during the start-upphase.

Recent studies of levoglucosan suggest that the trans-glycosylation process (cellulose degradation pathway) occurs atlower temperatures than previously assumed, between 150 and350 �C (Kuo et al., 2008), withmaximumyields at 250 �C, regardlessof plant species. Therefore, levoglucosan is not a suitable tracer forsophisticated appliances with automatically fired wood combus-tion in which high temperatures are reached.

Inorganic elements accounted for about 10.6e31.7 wt.% of thePM10 mass (Table 6). The dominant inorganic species in the PM10samples from pellets type I and II, olive pit, and almond shell werepotassium and sulphate. The combustion of pellets type III and IVand shell of pine nuts generated higher amounts of phosphorous. Ahigh phosphorous content in particulate matter is usually typicalfor cereal fuel combustion (Tissari et al., 2008b). The proportion ofpotassium in PM10 inorganic fraction was 66.0% for pellets type I,40.6% for pellets type II, 86.8 for olive pit and 88.4% for almondshell. It was not detected in samples from pellets type II and III. Inaddition to these compounds, almost all samples contained zinc,lead, magnesium and sodium and, to a lesser extent, other ele-ments. For pellets type II, III and IV, high contents of zinc(14.6e19.9 wt.% of total inorganic mass), lead (7.6e17.6 wt.% of total

Fraction sampled OC/EC Reference

PM2.5 2.2e35.5 Gonçalves et al. (2011)PM2.5 2.8 ± 1.3 Zhang et al. (2013)

12.1 ± 3.3

PMtot 11.5ae1.2b Orasche et al. (2012)PM2.5 0.9e4.4 Gonçalves et al. (2010)PMtot 4.4ae5.9b Orasche et al. (2012)

8.5ae71.0b

PM10 0.9e4.2 This study0.3e9.4

Table 5Weight percentage (wt.%) of anhydrosugars in PM10.

Fuel Level of power output Levoglucosan Mannosan Galactosan

Pellets type I Lowest 3.03 <ld <ldMedium 1.39 0.44 0.37Highest 2.88 1.55 0.73

Pellets type II Lowest 0.14 0.07 <ldMedium <ld <ld <ldHighest <ld <ld <ld

Pellets type III Lowest 0.02 <ld <ldMedium <ld <ld <ldHighest <ld <ld <ld

Pellets type IV Lowest 0.25 <ld <ldMedium 0.23 <ld <ldHighest 0.16 <ld <ld

Olive pit Lowest 0.41 0.31 <ldMedium 0.27 <ld <ldHighest 0.05 <ld <ld

Shell of pine nuts Lowest 0.27 <ld <ldMedium 0.14 <ld <ldHighest 0.43 <ld <ld

Almond shell Lowest 0.37 <ld <ldMedium <ld <ld <ldHighest 0.11 <ld <ld

ld e limit of detection (0.075 mg cm�2).

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e2724

inorganic mass) and arsenic (0.079e0.030 wt.% of total inorganicmass) were recorded. The use of CCA-treated wood in pellets hasbeen already reported to be problematic from the point of view ofheavy metals emissions (Wasson et al., 2005). The high concen-trations of Pb in samples could be due to Pb-based paint on oldwood or uptake of Pb from soil contaminated by lead arsenatepesticide (Sander, 1997).

Leskinen et al. (2014) reported particulate mass contents of Znand K of 5.5, 1.0 and 0.4 and 28.0, 6.7 and 1.5% for efficient, inter-mediate and smouldering combustion, respectively. The authorsfound that the ash emissions were mainly composed of potassium,sulphate and chloride at almost equal concentrations. However, thefraction of sulphate was higher in samples from efficient combus-tion than in the other two conditions.

Taking into account that elements exist at the highest oxidationstates, it is recommended to convert the measured element con-centrations into the respective mass concentrations of the mostcommon oxides for PM10 mass balances. When accounting for theunmeasured oxygen, elements in their oxide form (e.g. Al2O3, CaO,Fe2O3, K2O, MgO, Na2O, P2O5, etc.) represented 21.6 (shell of pinenuts) to 42.2 (almond shell) wt.% (Fig. 4) of PM10.

4. Conclusions

This paper constitutes an attempt to quantify particle andgaseous emissions from the combustion in a pellet stove ofdifferent biofuels. It was found that the fuel type and its propertiescan significantly influence the gaseous and particulate EFs fromresidential combustion appliances. In addition, through theorganic and elemental carbon mass fractions, it was possible toverify that the chemical composition of particles can also varynoticeably. Combustion in modern appliances can significantlyreduce PM emissions when compared to old type wood combus-tion systems.

The highest EFs were obtained for biomass fuels other thanpellets. The ash accumulation and slag formation on the grate ofthe stove, as well as the physical characteristics of the agricul-tural fuels that difficult the feeding to the burner, impair thecombustion conditions and generate high gaseous and particu-late emissions. The pelletisation of agricultural fuels and the useof blended pellet fuels can contribute to optimise the combustion

process and subsequently reduce the emissions. It was possibleto verify that the enhanced emissions of CO were accompaniedby emissions of other unoxidised components like TOC, HCHOand CH4 during agro-fuels combustion. Olive pit presented thehighest EFs for these compounds. CH4, which have much higherglobal warming potential than CO2, accounted for a significantfraction of TOC in all experiments. The lowest PM10 EFs wereobtained for the combustion of pellets with lower ash content,whilst the highest were registered for agro fuels. The fuelcomposition, namely its ash content, was found to be a param-eter with great influence on particulate emissions. Furthermore,the higher alkali metal content in agricultural biomass fuels incomparison with woody pellets also contributed to the higherPM10 EFs. Pellets made up of wood and wood wastes werecharacterised by high K and ash content. The particulate emis-sions were found to be correlated with the gaseous emissions,i.e., with the products of incomplete combustion. Combustion ofdifferent fuels generated differences in the OC and EC contents ofsmoke particles. Levoglucosan was the most abundant anhy-drosugar in particulate matter emissions. Mannosan and gal-actosan were absent from almost all samples, indicating that hightemperature biomass combustion contributes little to anhy-drosugar formation. The dominant inorganic species in PM10samples showed wide fluctuations depending on the fuel burned.Potassium, usually pointed out as a biomass combustion tracer,was not detected in some samples. The use of treated wood tomanufacture pellets was found to be problematic due to highemissions of heavy metals, such as Zn, Pb and As.

The comparison of emissions from this study with data from theliterature showed striking differences, either in particulate emis-sion levels or in chemical composition, between old-type appli-ances and modern stoves or boilers. Fuel composition andcharacteristics also play a major role in emissions. The detailedcharacterisation of smoke particles from appliances with high ef-ficiency is wanted since emissions can vary remarkably.

On the legislative level, national and European regulations withmore stringent particle emission limits for solid fuel heating ap-pliances must be adopted. European-wide regulations comprisinggeneral performance requirements to encourage further techno-logical innovation are required. Measures leading to a strict qualitycontrol of biofuels should be envisaged.

Table 6Weight percentage (wt.%) of several elements in PM10.

Pellets type I Pellets type II Pellets type III Pellets type IV Olive pit Shell of pine nuts Almond shell

Li 0.002 0.002 0.001 0.001 <ld <ld 0.001Ti 0.015 0.008 0.007 0.044 0.003 0.002 0.010V 0.001 0.001 <ld 0.002 <ld 0.001 0.001Cr 0.034 0.012 0.003 0.054 0.001 <ld 0.015Mn 0.063 0.005 0.002 0.025 0.003 <ld 0.007Cu 0.031 0.050 0.032 0.128 0.009 0.001 0.035Zn 0.810 5.98 3.46 3.26 0.050 0.134 0.064As 0.001 0.079 0.030 0.022 0.001 <ld <ldRb 0.048 0.039 0.018 0.075 0.018 0.010 0.035Sr 0.009 0.002 <ld 0.005 0.002 <ld 0.003Zr 0.040 0.036 0.016 0.010 0.001 <ld <ldNb 0.002 0.001 0.001 0.001 <ld <ld 0.001Mo 0.034 0.014 0.006 0.493 0.001 0.016 0.125Sn 0.003 0.047 0.017 0.029 <ld <ld <ldSb 0.002 0.046 0.017 0.028 <ld <ld <ldBa 0.011 0.007 0.004 0.026 0.001 <ld 0.007La <ld <ld <ld 0.001 <ld <ld <ldCe <ld <ld <ld 0.002 <ld <ld 0.001Nd <ld <ld <ld 0.001 <ld <ld <ldSm <ld <ld <ld 0.001 <ld <ld <ldBi <ld 0.006 0.003 0.015 <ld 0.001 <ldPb 0.169 5.29 2.91 1.70 0.027 0.005 0.011Th <ld <ld <ld 0.001 <ld <ld <ldAl 0.132 0.090 0.016 0.007 0.072 <ld 0.047Ca 1.34 0.143 <ld <ld 0.175 <ld 0.154Fe 0.128 0.054 1.697 5.096 0.012 0.248 <ldK 17.3 12.2 <ld <ld 20.6 <ld 28.1S 4.16 1.03 1.72 1.96 2.27 1.22 2.59Mg 0.368 0.017 5.61 3.78 0.037 4.07 0.044Na 1.41 4.75 0.092 0.115 0.373 0.099 0.362P 0.072 0.098 6.69 5.29 0.065 4.82 0.130Ni 0.006 0.001 <ld 0.100 <ld <ld 0.025Ga <ld 0.001 <ld 0.002 <ld <ld <ldGe <ld <ld <ld 0.002 <ld <ld 0.002Y <ld <ld <ld 0.007 <ld <ld 0.002Cs 0.002 0.002 0.001 0.003 0.001 <ld 0.001Gd <ld <ld <ld 0.001 <ld <ld <ldDy <ld <ld <ld 0.001 <ld <ld <ldEr <ld <ld <ld 0.001 <ld <ld <ldHf 0.002 0.001 0.001 <ld <ld <ld <ldW 0.003 0.002 0.001 0.003 <ld <ld <ldU <ld <ld <ld 0.003 <ld <ld 0.001B 0.001 0.010 0.004 0.088 <ld <ld <ldCd 0.003 0.010 0.005 0.006 <ld <ld <ldTI <ld 0.001 <ld 0.001 <ld <ld <ld

Note: Be, Co, Pr, Eu, Ho, Tm, Lu, Ta, Tb, Yb, Sc and Se were all above the limit of detection (ld).

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e27 25

Acknowledgements

This work was financially supported by AIRUSE e Testing anddevelopment of air quality mitigation measures in SouthernEurope, LIFE 11 ENV/ES/000584, and project PTDC/AAC-AMB/116568/2010 (FCOMP-01-0124-FEDER-019346) BiomAshTech e

Ash impacts during thermo-chemical conversion of biomass, sup-ported by the Portuguese Foundation for Science and Technology.Fulvio Amato is beneficiary of the Juan de la Cierva postdoctoralGrant (JCI-2012-13473) from the Spanish Ministry of Economy andCompetitiveness.

References

Alves, C., Gonçalves, C., Fernandes, A.P., Tarelho, L., Pio, C., 2011. Fireplace andwoodstove fine particle emissions from combustion of western Mediterraneanwood types. Atmos. Res. 10, 692e700.

Baeza-Romero, M.T., Wilson, J.M., Fitzpatrick, E.M., Jones, J.M., Williams, A., 2010.In situ study of soot from the combustion of a biomass pyrolysis intermediate;eugenol; and n-decane using aerosol time of flight mass spectrometry. EnergyFuels 24, 439e445.

Bølling, A.K., Pagels, J., Yttri, K.E., Barregard, L., Sallsten, G., Schwarze, P.E., Boman, C.,2009. Health effects of residential wood smoke particles: the importance ofcombustion conditions and physicochemical particle properties. Part. Fibre

Toxicol. 6, 29. http://dx.doi.org/10.1186/1743-8977-6-29.Calvo, A.I., Tarelho, L.A.C., Alves, C.A., Duarte, M., Nunes, T., 2014. Characterization of

operating conditions of two residential wood combustion appliances. FuelProcess. Technol. 126, 222e232.

Carvalho, L., Wopienka, E., Lundgren, J., 2008. Challenges in small-scale combustionof agricultural biomass fuels. Int. J. Energy Clean Environ. 9, 127e142.

Carvalho, L., Wopienka, E., Pointner, C., Lundgren, J., Verma, V.K., Haslinger, W.,Schmidl, C., 2013. Performance of a pellet boiler fired with agricultural fuels.Appl. Energy 104, 286e296.

Cerqueira, M., Gomes, L., Tarelho, L., Pio, C., 2013. Formaldehyde and acetaldehydeemissions from residential wood combustion in Portugal. Atmos. Environ. 72,171e176.

Cocchi, M., Nikolaisen, L., Junginger, M., Goh, C.S., Hess, R., Jacobson, J., Ovard, L.P.,Thr€an, D., Hennig, C., Deutmeyer, M., Schouwenberg, P.P., Marchal, D., 2011.Global wood pellet industry. Market and trade study. In: IEA Bioenergy. Dec.2011.

Engling, G., Carrico, C.M., Kreidenweis, S.M., Collett Jr., J.L., Day, D.E., Malm, W.C.,Lincoln, E., Min Hao, W., Iinuma, Y., Herrmann, H., 2006. Determination oflevoglucosan in biomass combustion aerosol by high-performance anion-ex-change chromatography with pulsed amperometric detection. Atmos. Environ.40, 299e311.

Fernandes, A.P., Alves, C.A., Gonçalves, C., Tarelho, L., Pio, C., Schimdl, C., Bauer, H.,2011. Emission factors from residential combustion appliances burning Portu-guese biomass fuels. J. Environ. Monit. 13, 3196e3206.

Fine, P.M., Cass, G.R., Simoneit, B.R.T., 2004. Chemical characterization of fine par-ticle emissions from the wood stove combustion of prevalent United States treespecies. Environ. Eng. Sci. 21, 705e721.

Fitzpatrick, E.M., Jones, J.M., Pourkashanian, M., Ross, A.B., Williams, A., Bartle, K.D.,

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e2726

2008. Mechanistic aspects of soot formation from the combustion of pine wood.Energy Fuels 22, 3771e3778.

Garcia-Maraver, A., Zamorano, M., Fernandes, U., Rabaçal, M., Costa, M., 2014.Relationship between fuel quality and gaseous and particulate matter emissionsin a domestic pellet-fired boiler. Fuel 119, 141e152.

Goh, C.S., Junginger, M., Cocchi, M., Marchal, D., Thr€an, D., Hennig, C., Heinim€o, J.,Nikolaisen, L., Schouwenberg, P.-P., Bradley, D., Hess, R., Jacobson, J., Ovard, L.,Deutmeyer, M., 2012. Wood pellet market and trade: a global perspective.Biofuels Bioprod. Biorefining 7, 24e42.

Gonçalves, C., Alves, C., Evtyugina, M., Mirante, F., Pio, C., Caseiro, A., Schmidl, C.,Bauer, H., Carvalho, F., 2010. Characterisation of PM10 emissions from wood-stove combustion of common woods grown in Portugal. Atmos. Environ. 44,4474e4480.

Gonçalves, C., Alves, C., Fernandes, A.P., Monteiro, C., Tarelho, L., Evtyugina, M.,Pio, C., 2011. Organic compounds in PM2.5 emitted from fireplace and wood-stove combustion of typical Portuguese wood species. Atmos. Environ. 45,4533e4545.

Gonçalves, C., Alves, C., Nunes, T., Rocha, S., Cardoso, J., Cerqueira, M., Cerqueira, M.,Pio, C., Almeida, S.M., Hillamo, R., Teinil€a, K., 2014. Organic characterisation ofPM10 in Cape Verde under Saharan dust influxes. Atmos. Environ. 89, 425e432.

Gonz�alez, J.F., Rom�ana, S., Encinarb, J.M., Martínez, G., 2009. Pyrolysis of variousbiomass residues and char utilization for the production of activated carbons.J. Anal. Appl. Pyrolysis 85, 134e141.

Happo, M.S., Uski, O., Jalava, P.I., Kelz, J., Brunner, T., Hakulinen, P., M€aki-Paakkanen, J., Kosma, V.-M., Jokiniemi, J., Obernberger, I., Hirvonen, M.-R., 2013.Pulmonary inflammation and tissue damage in the mouse lung after exposureto PM samples from biomass heating appliances of old and modern technolo-gies. Sci. Total Environ. 443, 256e266.

Hedberg, E., Johansson, C., Johansson, L., Swietlicki, E., Brorstr€om-Lund�en, E., 2006.Is levoglucosan a suitable quantitative tracer for wood burning? comparisonwith receptor modeling on trace elements in Lycksele, Sweden. J. Air WasteManag. Assoc. 56, 1669e1678.

Heschel, W., Rweyemamu, L., Scheibner, T., Meyer, B., 1999. Abatement of emissionsin small-scale combustors through utilisation of blended pellet fuels. FuelProcess. Technol. 61, 223e242.

Ingwald, O., Gerold, T., 2006. Recent developments concerning pellet combustiontechnologies e a review of Austrian developments. In: Swedish Bioenergy As-sociation (Ed.), Proceedings of the 2nd World Conference on Pellets, pp. 31e40.J€onk€oping, Sweden.

Johansson, L.S., Leckner, B., Gustavsson, L., Cooper, D., Tullin, C., Potter, A., 2004.Emission characteristics of modern and old-type residential boilers fired withwood logs and wood pellets. Atmos. Environ. 38, 4183e4195.

Kuo, L.-J., Herbert, B.E., Louchouarn, P., 2008. Can levoglucosan be used to charac-terize and quantify char/charcoal black carbon in environmental media? Org.Geochem. 39, 1466e1478.

Kuo, L.-J., Louchouarn, P., Herbert, B.E., 2011. Influence of combustion conditions onyields of solvent-extractable anhydrosugars and lignin phenols in chars: im-plications for characterizations of biomass combustion residues. Chemosphere85, 797e805.

Lamberg, H., Nuutinen, K., Tissari, J., Ruusunen, J., Yli-Piril€a, P., Sippula, O.,Tapanainen, M., Jalava, P., Makkonen, U., Teinil€a, K., Saarnio, K., Hillamo, R.,Hirvonen, M.-R., Jokiniemi, J., 2011a. Physicochemical characterization of fineparticles from small-scale wood combustion. Atmos. Environ. 45, 7635e7643.

Lamberg, H., Sippula, O., Tissari, J., Jokiniemi, J., 2011b. Effects of air staging and loadon fine-particle and gaseous emissions from a small-scale pellet boiler. EnergyFuels 25, 4952e4960.

Leskinen, J., Tissari, J., Uski, O., Vir�en, a., Torvela, T., Kaivosoja, T., Tapanainen, M.,Jalava, P., Makkonen, U., Teinil€a, K., Saarnio, K., Hillamo, R., Hirvonen, M.-R.,Jokiniemi, J., 2014. Fine particle emissions in three different combustion con-ditions of a wood chip-fired appliance e particulate physico-chemical proper-ties and induced cell death. Atmos. Environ. 86, 129e139.

Mediavilla, I., Fern�andez, M.J., Esteban, L.S., 2009. Optimization of pelletisation andcombustion in a boiler of 17.5 kWth for vine shoots and industrial cork residue.Fuel Process. Technol. 90, 621e628.

Miranda, T., Arranz, J.I., Montero, I., Rom�an, S., Rojas, C.V., Nogales, S., 2012. Char-acterization and combustion of olive pomace and forest residue pellets. FuelProcess. Technol. 103, 91e96.

Monteiro, E., Mantha, V., Rouboa, A., 2012. Portuguese pellets market: analysis ofthe production and utilization constrains. Energy Policy 42, 129e135.

Obaidullah, M., Bram, S., Verma, V.K., Ruyck, J. De, 2012. A review on particleemissions from small scale biomass combustion. Int. J. Renew. Energy Res. 2 (1).

Obernberger, I., Thek, G., 2004. Physical characterisation and chemical compositionof densified biomass fuels with regard to their combustion behavior. BiomassBioenergy 27, 653e669.

€Ohman, M., Nordin, A., Hedman, H., Jirjis, R., 2004. Reasons for slagging duringstemwood pellet combustion and some measures for prevention. BiomassBioenergy 27, 597e605.

Olsson, M., Kj€allstrand, J., 2006. Low emissions fromwood burning in an ecolabelledresidential boiler. Atmos. Environ. 40, 1148e1158.

Orasche, J., Seidel, T., Hartmann, H., Schnelle-kreis, J., Chow, J.C., Ruppert, H.,Zimmermann, R., 2012. Comparison of emissions from wood combustion. Part1: emission factors and characteristics from different small-scale residentialheating appliances considering particulate fine particle and gaseous emissionsfrom normal and smouldering wood combustion in a conventional masonryheater matter and polycyclic aromatic hydrocarbon (PAH)-related toxicological

potential. Energy Fuels 26, 6695e6704.Ozgen, S., Caserini, S., Galante, S., Giugliano, M., Angelino, E., Marongiu, A.,

Hugony, F., Morreale, C., 2014. Emission factors from small scale appliancesburning wood and pellets. Atmos. Environ. 94, 144e153.

Pettersson, E., Lindmark, F., €Ohman, M., Nordin, A., Westerholm, R., Boman, C., 2010.Design changes in a fixed-bed pellet combustion device: effects of temperatureand residence time on emission performance. Energy Fuels 24, 1333e1340.

Pettersson, E., Boman, C., Westerholm, R., Bostr€om, D., Nordin, A., 2011. Stove per-formance and emission characteristics in residential wood log and pelletcombustion, part 2: wood stove. Energy Fuels 25, 315e323.

P€oschl, U., 2005. Atmospheric aerosols: composition, transformation, climate andhealth effects. Angew. Chem. Int. Ed. Engl. 44, 7520e7540.

Querol, X., Alastuey, A., Rodriguez, S., Plana, F., Ruiz, C.R., Cots, N., 2001. PM10 andPM2.5 source apportionment in the Barcelona Metropolitan area, Catalonia,Spain. Atmos. Environ. 35, 6407e6419.

Riva, G., Pedretti, E.F., Toscano, G., Duca, D., Pizzi, A., 2011. Determination of poly-cyclic aromatic hydrocarbons in domestic pellet stove emissions. BiomassBioenergy 35, 4261e4267.

Rose, D., Nowak, A., Achtert, P., Wiedensohler, A., Hu, M., Shao, M., Zhang, Y.,Andreae, M.O., P€oschl, U., 2010. Cloud condensation nuclei in polluted air andbiomass burning smoke near the mega-city Guangzhou, Chinadpart 1: size-resolved measurements and implications for the modeling of aerosol particlehygroscopicity and CCN activity. Atmos. Chem. Phys. 10, 3365e3383.

Sander, B., 1997. Properties of Danish biofuels and the requirements for powerproduction. Biomass Bioenergy 12, 177e183.

Schmidl, C., Luisser, M., Padouvas, E., Lasselsberger, L., Rzaca, M., Ramirez-SantaCruz, C., Handler, M., Peng, G., Bauer, H., Puxbaum, H., 2011. Particulate andgaseous emissions from manually and automatically fired small scale com-bustion systems. Atmos. Environ. 45, 7443e7454.

Schmidl, C., Marr, I.L., Caseiro, A., Kotianov�a, P., Berner, A., Bauer, H., Kasper-Giebl, A.,Puxbaum, H., 2008. Chemical characterisation of fine particle emissions fromwood stove combustion of common woods growing in mid-European Alpineregions. Atmos. Environ. 42, 126e141.

Sippula, O., Hyto, K., Tissari, J., Raunemaa, T., Jokiniemi, J., 2007. Effect of wood fuelon the emissions from a top-feed pellet stove. Energy Fuels 21, 1151e1160.

Skotland, C.H., 2009. Measurement of Temperature Conditions in Grate Zone of a 1MW Wood-pellets Boiler Fired with High Ash Content Wood-pellets (Master ofScience in Energy and Environment). Norwegian University of Science andTechnology.

Tang, X., Bai, Y., Duong, A., Smith, M.T., Li, L., Zhang, L., 2009. Formaldehyde inChina: production, consumption, exposure levels, and health effects. Environ.Int. 35, 1210e1224.

Telmo, C., Lousada, J., Moreira, N., 2010. Proximate analysis, backwards stepwiseregression between gross calorific value, ultimate and chemical analysis ofwood. Bioresour. Technol. 101, 3808e3815.

Tissari, J., Hyt€onena, K., Lyyr€anen, J., Jokiniemi, J., 2007. A novel field measurementmethod for determining fine particle and gas emissions from residential woodcombustion. Atmos. Environ. 41, 8330e8344.

Tissari, J., Lyyr€anen, J., Hyt€onen, K., Sippula, O., Tapper, U., Frey, A., Saarnio, K.,Pennanen, A.S., Hillamo, R., Salonen, R.O., Hirvonen, M.-R., Jokiniemi, J.,2008a. Fine particle and gaseous emissions from normal and smoulderingwood combustion in a conventional masonry heater. Atmos. Environ. 42,7862e7873.

Tissari, J., Sippula, O., Kouki, J., Vuorio, K., Jokiniemi, J., 2008b. Fine particle and gasemissions from the combustion of agricultural fuels fired in a 20 kW burner.Energy Fuels 22, 2033e2042.

Tissari, J., 2008. Fine Particle Emissions from Residential Wood Combustion(Doctoral thesis). Department of Environmental Science. University of Kuopio.

Toscano, G., Duca, D., Amato, A., Pizzi, A., 2014. Emission from realistic utilization ofwood pellet stove. Energy 68, 644e650.

Torvela, T., Tissari, J., Sippula, O., Kaivosoja, T., Leskinen, J., Vir�en, A., L€ahde, A.,Jokiniemi, J., 2014. Effect of wood combustion conditions on the morphology offreshly emitted fine particles. Atmos. Environ. 87, 65e76.

Turpin, B.J., Lim, H.-J., 2001. Species contributions to PM2.5 mass concentrations:revisiting common assumptions for estimating organic mass. Aerosol Sci.Technol. 35, 602e610.

Uski, O., Jalava, P.I., Happo, M.S., Leskinen, J., Sippula, O., Tissari, J., M€aki-Paakkanen, J., Jokiniemi, J., Hirvonen, M.-R., 2014. Different toxic mechanismsare activated by emission PM depending on combustion efficiency. Atmos.Environ. 89, 623e632.

Verma, V.K., Bram, S., Gauthier, G., De Ruyck, J., 2011. Evaluation of the performanceof a multi-fuel domestic boiler with respect to the existing European standardand quality labels: part-1. Biomass Bioenergy 35, 80e89.

Verma, V.K., Bram, S., Delattin, F., Laha, P., Vandendael, I., Hubin, A., De Ruyck, J.,2012. Agro-pellets for domestic heating boilers: standard laboratory and reallife performance. Appl. Energy 90, 17e23.

Vicente, A., Alves, C., Monteiro, C., Nunes, T., Mirante, F., Cerqueira, M., Calvo, A.,Pio, C., 2012. Organic speciation of aerosols from wildfires in central Portugalduring summer 2009. Atmos. Environ. 57, 186e196.

Vicente, E.D., Vicente, A.M., Bandowe, B.A.M., Alves, C.A., 2015. Particulate phaseemission of parent polycyclic aromatic hydrocarbons (PAHs) and their de-rivatives (alkyl-PAHs, oxygenated-PAHs, azaarenes and nitrated-PAHs) frommanually and automatically fired combustion appliances. Air Qual. Atmos.Health. http://dx.doi.org/10.1007/s11869-015-0364-1 (in press).

Wasson, S.J., Linak, W.P., Gullett, B.K., King, C.J., Touati, A., Huggins, F.E., Chen, Y.,

E.D. Vicente et al. / Atmospheric Environment 120 (2015) 15e27 27

Shah, N., Huffman, G.P., 2005. Emissions of chromium, copper, arsenic, andPCDDs/Fs from open burning of CCA treated wood. Environ. Sci. Technol. 39,8865e8876.

Wiinikka, H., 2015. High Temperature Aerosol Formation and Emission Mini-misation during Combustion of Wood Pellets (Doctoral thesis).

Yokelson, R.J., Susott, R., Ward, D.E., Reardon, J., Griffith, D.W.T., 1997. Emissionsfrom smoldering combustion of biomass measured by open-path Fouriertransform infrared spectroscopy. J. Geophys. Res. 102 (D15), 18865e18877.

Zhang, L., Steinmaus, C., Eastmond, D.A., Xin, X.K., Smith, M.T., 2009. Formaldehydeexposure and leukemia: a newmeta-analysis and potential mechanisms. Mutat.Res. 681, 150e168.

Zhang, X., Li, J., Yang, W., Blasiak, W., 2011. Formation mechanism of levoglucosanand formaldehyde during cellulose pyrolysis. Energy Fuels 25, 3739e3746.

Zhang, Y., Obrist, D., Zielinska, B., Gertler, A., 2013. Particulate emissions fromdifferent types of biomass burning. Atmos. Environ. 72, 27e35.