Michael
Priestley
ac,
Xiangrui
Kong
a,
Xiangyu
Pei‡
a,
Julia
Hammes§
a,
Daniel
Bäckström
b,
Ravi K.
Pathak
a,
Jan B. C.
Pettersson
a and
Mattias
Hallquist
*a
aDepartment of Chemistry and Molecular Biology, University of Gothenburg, 412 96, Gothenburg, Sweden. E-mail: hallq@chem.gu.se
bResearch Institutes Sweden (RISE), Brinellgatan 4, 504 62 Borås, Sweden
cIVL Swedish Environmental Research Institute, Aschebergsgatan 44, 411 33, Gothenburg, Sweden
First published on 16th February 2023
Biomass burning is a growing alternative to fossil fuels for power generation. Small-scale residential wood combustion introduces air pollutants to local populated areas compared to remote, large-scale facilities. Pellet fuel appliances are an alternative to log burning as they are considered more efficient; however, as emissions are not as well characterized, comparative studies to establish the potential benefits of increased pellet stove usage is required. Here we describe a distinction in optical and chemical properties of emissions from a residential pellet and a log-burning stove using state-of-the-art online measurements. Specifically, we report the first online simultaneously phase-resolved semi-volatile organic measurements from such appliances, using a Time of Flight Chemical Ionisation Mass Spectrometer with a Filter Inlet for Gas and AEROsols (FIGAERO-ToF-CIMS). Pellet particle emissions were 90% brown carbon-containing substituted mono-aromatic compounds (SMAs) whereas wood emissions contained equal black carbon (BC) and organic carbon (OC) and contained 3–55 times more poly-aromatic hydrocarbons (PAHs). Lowering pellet fuel loadings increased OC and SMA emission factors (EFOC and EFSMA), therefore increasing particle mass as well as optical absorption, i.e. “brownness”. The consequence of these EF differences is illustrated for a hypothetical national burden assuming a base case 10:1 log to pellet stove energy demand usage and a ‘swapped’ demand to simulate increased pellet stove usage. This results in net decreases of PM and BC burdens by 12% and 42% respectively but is somewhat offset by a 57% increase in OC burden. Changes in phase resolved speciated organic EFs suggest the reduction in particle burdens is somewhat offset by increases in gas burdens for some organic species, which could contribute to delayed particle burdens through secondary aerosol formation.
Environmental significanceAn alternative to fossil fuels considered sustainable, and thus a viable replacement, is biomass combustion. Residential wood combustion has grown in popularity, increasing air pollution in populated areas. The composition and quantity of emissions can vary substantially depending on factors including appliance type and efficiency. We quantify and contrast optical and chemical properties of gaseous and particle emissions from a modern wood burning and a pellet burning stove. While still generally less polluting than the wood-burning stove in terms of particulate emissions, here we show the pellet stove emitted similar quantities of organics in the form of brown carbon, which can be exacerbated if appliance efficiencies are reduced. As a result, a hypothetical national-scale scenario based on results from the two stoves tested in this study indicated particulate matter reductions effected by swapping appliances may be modestly offset. Additionally, the pellet appliance emitted emit larger quantities of SVOCs as gases, likely increasing SOA burdens even though the POA burden is reduced. |
Emissions from RWC can vary substantially from several factors to an order of magnitude, as they are highly dependent on combustion conditions, which are often dictated by fuel loadings, log size, moisture content, wood type and air supply.11 A recent technological solution for reducing RWC emissions is the use of wood pellets as fuel rather than logs.12,13 Pellet stoves autonomously control oxygen and fuel supplies to optimise burning conditions,14 contrasting with less efficient, manually loaded wood log burners.15,16 The difference in emissions between appliances is large enough that emissions are treated separately in emission inventories.16,17 However, detailed chemical and optical measurements of emissions from these appliances are lacking, especially concerning the composition of organic carbon (OC), and is an area recommended for further investigation.11
POA absorbing light at 300–400 nm is called brown carbon (BrC) with similar optical and radiative properties to BC,18 but a comprehensive understanding of its impact on climate is lacking.19–21 The largest source of BrC in many European countries is RWC, which is an especially acute problem during the winter months and during the night.22–24 BrC originates from lignin pyrolysis and is typically associated with less efficient combustion,25 although secondary BrC sources are also known.26,27 Even if a full understanding of BrC impacts on the climate are not yet understood, recent work shows the direct radiative effect of BrC is comparable, and in some instances, potentially greater than BC, e.g. in the tropical mid and upper troposphere.28 Whilst global modelling efforts to include BrC for improved radiative forcing estimates are ongoing,29 the treatment of BrC as a pollutant and contributor to poor air quality at shorter times scales and at local or regional levels is not currently considered, or specifically targeted by legislation.
Whilst many BrC constituents are harmful to human health, such as poly-aromatic hydrocarbons (PAHs) and substituted mono-aromatic compounds (SMAs), e.g. nitrophenol,30,31 the full extent of BrC toxicity is unknown. These compounds are also responsible for the particle ‘brownness’ as they contain chromophores, e.g. aromatic, nitrated, multifunctional carbonyl or unsaturated moieties. Key BrC markers are known to have both gaseous and particle phase routes to formation and many of these organic constituents are semi-volatile with significant gas and particle components.
Detailed analysis of BrC constituents is commonly performed offline by analysing filter samples using separation techniques such as gas chromatography (GC)32 and high performance liquid chromatography (HPLC).33 Online measurements of residential wood burning emissions reported using aerosol mass spectrometer (AMS) give bulk aerosol properties,15 and ambient measurements frequently return biomass burning factors when used in conjunction with positive matrix factorisation (PMF),34 although little information on the chemical composition of the BrC is derivable.
Trace compounds from biomass burning can be measured online in the gas phase with mass spectrometric techniques, e.g. proton-transfer mass spectrometry (PTR).35,36 Iodide chemical ionisation mass spectrometry (CIMS) has been used to quantify low molecular weight BrC components such as N containing aromatic compounds.37–40 Time of Flight Chemical Ionisation Mass Spectrometer with a Filter Inlet for Gas and AEROsols (FIGAERO-ToF-CIMS)41 has been demonstrated to measure gas and particle phase BrC components from ambient biomass burning in China.39,42 Chamber studies of BrC formed from nitration of unsaturated heterocycles, which combined optical, HPLC and iodide-CIMS measurements, showed that complementary optical, gas and particle phase measurements are a suitable combination to characterise BrC.43
Here, for the first time, we present simultaneous, online measurements of a range of organic compounds, including SMAs, in both particle and gas phases from a wood burning stove and a pellet burning stove, using a FIGAERO-ToF-CIMS.41,44 We describe differences in gaseous and particle emission factors (EFs) and optical and chemical properties of particle emissions from these two RWC appliances using state-of-the-art, online instrumentation. Contrasting the appliances under different operating conditions to represent a range of combustion efficiencies, we compare the effect of OC composition on optical absorption and use a case scenario analysis to evaluate potential differences in a hypothetical annual emission burden on a generalised, national scale. The phase partitioning of the OC measured here by FIGAERO-ToF-CIMS will be presented in a future publication.
The ten-year-old pellet stove used in this study had a claimed nominal energy output of 3–10 kW with an 83% heating efficiency, according to the appliance manual. This model is produced by a leading European manufacturer and was operated according to factory settings to simulate real-world usage. The wood pellet fuel (6 mm, Scandbio45) is manufactured from logging industry waste wood without any additives. The pellets have an ash content ≤0.5% (w/w) and moisture content of 6–8% (w/w, dry basis) and are designed for use in commercial and residential buildings. Pellets were introduced to the combustion chamber automatically and a constant burn condition was held throughout the experiment which typically lasted 6 to 8 hours. Two load settings were investigated. The low load fuel consumption rate of 0.69 kg per hour is designed to maintain a temperature of 100 °C (as measured at the chimney bottom) and the high (full) load rate of 1.16 kg per hour represents maximized appliance usage.
The pellet stove was compared with a popular, contemporary wood log stove of similar size with a claimed nominal energy output of 6 kW, and an 86% heating efficiency, according to the appliance manual. Two wood fuels were used; spruce, with 19.6% moisture content (dry basis); and birch with the bark removed, with 17.0% moisture content (dry basis). The wood logs were approximately 27 cm long and weighed 0.7 kg each. Their placement in the combustion chamber was standardized by stacking two logs in a T shape one on top of the other, according to the manufacturer's recommendations. A 2.1 kg load was used for the first batch of the day and then 1.4 kg for the batches thereafter. The air supply to the log stove was fully opened after loading and then reduced to 50% after three minutes, in line with the manufacturer's recommendations. The cycle was repeated approximately every 50 minutes for a total of 6 to 8 cycles.
The FIGAERO-ToF-CIMS particle sample from the spruce fuel experiment occurred at a lower modified combustion efficiency (MCE) (0.9519 ± 0.0374) than the average cycle, whereas the birch FIGAERO-ToF-CIMS sample occurred at a period of higher MCE (0.9938 ± 0.0035) than average (Fig. S1†). As combustion condition is more important than wood fuel type for emission profiles,16 we use the two FIGAERO samples to constrain upper and lower limits of wood emissions which provide a good comparison with the low load (MCE = 0.9674 ± 0.0106) and high load (MCE = 0.9951 ± 0.0009) pellet combustion efficiencies. FIGAERO derived EFs reflect this distinction. As the pellet MCEs are constant, the timing of the particle phase FIGAERO sampling was not crucial (Fig. S1†). See ESI† for further details of sensitivities, sampling, calibrations, background and FIGAERO cycle information.
Property | Measurement | Unit | Pellets high load | Pellets low load | Birch | Spruce | ||
---|---|---|---|---|---|---|---|---|
FIGAERO sample | Whole wood cycle | FIGAERO sample | Whole wood cycle | |||||
Combustion | Temperature (chimney bottom) | °C | 152 ± 8 | 98 ± 1 | 265 ± 5 | 261 ± 14 | 236 ± 15 | 284 ± 13 |
MCE | — | 0.9951 ± 0.0009 | 0.9674 ± 0.0106 | 0.9938 ± 0.0035 | 0.9758 ± 0.026 | 0.9519 ± 0.0374 | 0.9615 ± 0.0218 | |
Dilution factor (tunnel) | — | 9.4 | 9.4 | 24.5 ± 0.1 | 24.9 ± 2.8 | 18.8 ± 3.0 | 22.3 ± 2.3 | |
Optical | AAE | — | 3.5 ± 0.4 | 3.1 ± 0.1 | 1.0 ± 0.1 | 0.9 ± 0.1 | 2.1 ± 0.8 | 1.0 ± 0.6 |
MAC405 | m2 g−1 | 2.7 ± 0.9 | 11.5 ± 6.6 | 2.1 ± 1.0 | 6.2 ± 11.0 | 67.0 ± 67.4 | 33.8 ± 27.4 | |
MAC781 | m2 g−1 | 0.3 ± 0.1 | 1.4 ± 0.7 | 1.1 ± 0.5 | 3.1 ± 5.1 | 35.3 ± 32.4 | 11.3 ± 10.3 | |
Emission factors | CO2 | kg kg−1 | 1.81 ± 0.19 | 1.77 ± 0.27 | 1.82 ± 0.06 | 1.81 ± 0.62 | 1.73 ± 0.65 | 1.79 ± 0.52 |
mg MJ−1 | 126519 ± 13516 | 123065 ± 18510 | 95562 ± 3413 | 110068 ± 37713 | 105556 ± 39621 | 108815 ± 31951 | ||
CO | mg kg−1 | 10853 ± 2455 | 39259 ± 7621 | 5437 ± 3788 | 15759 ± 14364 | 55453 ± 19523 | 28579 ± 18342 | |
mg MJ−1 | 757 ± 171 | 2737 ± 531 | 331 ± 231 | 960 ± 875 | 3379 ± 1190 | 1742 ± 1118 | ||
THC | mg kg−1 | 812 ± 204 | 4778 ± 1216 | 534 ± 248 | 1470 ± 2959 | 11237 ± 9123 | 1742 ± 1118 | |
mg MJ−1 | 57 ± 14 | 333 ± 85 | 33 ± 15 | 90 ± 180 | 685 ± 556 | 121 ± 244 | ||
PM | mg kg−1 | 462 ± 92 | 1411 ± 759 | 515 ± 226 | 592 ± 749 | 2512 ± 4856 | 484 ± 7547 | |
mg MJ−1 | 32 ± 6 | 98 ± 53 | 31 ± 14 | 36 ± 46 | 153 ± 296 | 30 ± 460 | ||
OC | mg kg−1 | 220 ± 56 | 638 ± 394 | 149 ± 77 | 170 ± 237 | 689 ± 1391 | 196 ± 7547 | |
mg MJ−1 | 15 ± 4 | 44 ± 53 | 9 ± 5 | 10 ± 14 | 42 ± 85 | 12 ± 460 | ||
BC | mg kg−1 | 42 ± 60 | 221 ± 239 | 239 ± 90 | 284 ± 326 | 1260 ± 2352 | 206 ± 1748 | |
mg MJ−1 | 3 ± 4 | 15 ± 17 | 15 ± 5 | 17 ± 20 | 77 ± 143 | 13 ± 107 | ||
Sum PAHs | mg kg−1 | 4 ± 0 | 7 ± 0 | — | 17 ± 0 | — | 172 ± 0 | |
mg MJ−1 | 0 ± 0 | 1 ± 0 | — | 1 ± 0 | — | 10 ± 0 | ||
Emission factors (FIGAERO) | Levoglucosan, C6H10O5, particle (gas) | mg kg−1 | 16.46 ± 4.94 (0.20 ± 0.06) | 24.6 ± 7.40 (3.50 ± 1.10) | 26.9 ± 8.1 (0.10 ± 0.00) | — | 49.4 ± 14.8 (0.00 ± 0.00) | — |
mg MJ−1 | 1.15 ± 0.34 (0.01 ± 0.00) | 1.72 ± 0.52 (0.24 ± 0.07) | 1.64 ± 0.49 (0.01 ± 0.0) | — | 3.01 ± 0.90 (0.00 ± 0.00) | — | ||
Sum FIGAERO organics, particle (gas) | mg kg−1 | 29.0 ± 8.70 (20.9 ± 6.30) | 65.7 ± 19.7 (84.4 ± 25.3) | 46.1 ± 13.8 (8.20 ± 2.50) | — | 81.5 ± 24.5 (14.7 ± 4.4) | — | |
mg MJ−1 | 2.02 ± 0.61 (1.46 ± 0.44) | 4.58 ± 1.37 (5.88 ± 1.76) | 2.81 ± 0.84 (0.50 ± 0.15) | — | 4.97 ± 1.49 (0.90 ± 0.27) | — | ||
Sum substituted aromatics (literature definition, n = 19), particle (gas) | mg kg−1 | 3.12 ± 0.94 (4.19 ± 1.26) | 11.8 ± 3.60 (15.9 ± 4.80) | 0.61 ± 0.18 (0.51 ± 0.15) | — | 2.28 ± 0.69 (0.44 ± 0.13) | — | |
mg MJ−1 | 0.22 ± 0.07 (0.29 ± 0.09) | 0.83 ± 0.25 (1.11 ± 0.33) | 0.04 ± 0.01 (0.03 ± 0.01) | — | 0.14 ± 0.04 (0.03 ± 0.01) | — |
Measurements of CO and CO2 allow for the calculation of the modified combustion efficiency (MCE, eqn (1)).50 This is commonly calculated for ambient biomass burning measurements to describe the phase of burning and to reconcile with emission factors.
(1) |
Here we calculate MCE to compare efficiencies of different appliances used under different settings. As these appliances are designed to combust fuel efficiently, it is expected that the MCE values will be very high. Flue temperature was measured at the chimney bottom and chimney top.
(2) |
(3) |
Using these optical properties, it is possible to estimate the aerosol mass fraction that originates from the organic fraction (OC) and refractory fraction (BC). We assume all absorption at 781 nm (babs781) is due to BC only and thus the fraction of babs405 caused by BC is calculated by rearranging eqn (2) as:
(4) |
Absorption due to the organic fraction is equal to the remainder of absorption at 405 nm.
(5) |
(6) |
(7) |
OM = PM − BC | (8) |
Total PM is estimated using the SMPS measurement of volume and then multiplied by an assumed effective density of 1.2 g cm−3. This was considered a reasonable value as BB particles at low mobility diameters have been measured as 1.15 ± 0.23 g cm−3, attributed to fractal black carbon57 and effective densities of 1.19 ± 0.05 g cm−3 for BrC particles have been reported elsewhere.58 However, as the morphology and composition of the emissions are distinct, this assumption is a limitation. The OC:OM ratio of 1.8 for biomass burning aerosol59 was used to derive OC from OM.
It is difficult to verify these OC mass concentration estimations, as no other online measurements of OC were made. However, comparisons of mass concentrations derived from SMPS measurements with other methods of quantification have been demonstrated to show good agreement where accurate effective densities were used. Bougiatioti et al.60 compared Aerosol Chemical Speciation Monitor (ACSM) and BC measurements with SMPS derived mass concentrations for an ambient wild fire study and found good agreement between measurements, although this treatment included ammonium sulfate in the effective density calculation and a significant sulfate fraction of 19% was measured by the ACSM. Bau et al.61 showed mass concentrations derived from SMPS data were in good agreement (R2 ∼ 0.9) compared to gravimetric and chemical analyses when analysing mass loadings of 20–1000 μg m−3 of elemental carbon generated using carbon electrodes from a spark-discharge generator.
(9) |
EFs were converted from g kg−1 to g MJ−1, according to eqn (9), using the energy contents of the fuels, ECF, and heating efficiencies of the appliances, EA. Fc, ECF and EA were not measured directly but assumed. The variation in Fc, ECF and EA for the fuels use here are on the order of fractions to a few percent (see associated references), which is much lower than the measurement variability. ECF of the wood logs was assumed to be 5.3 kW h kg−1 (ref. 67) and 4.7 kW h kg−1 was assumed for the pellets,45 accounting for differences in reported moisture contents. EA values were not measured, they were instead taken from the appliance manufacturers manuals. This was 86% for the wood stove and 83% for pellet stove. Of the assumed variables in eqn (9) (rather than measured variables), EA is probably the largest source of error and is likely over estimated. This overestimation would yield lower EFs and so provides a low estimate of EFXi.
The emissions from the pellet stove were constant for a given loading. For both loadings we used a three-hour time period over which the EFs were integrated, which is more than enough time to provide an accurate, representative measurement. The emissions from the log stove displayed emissions cycling in which there is large variability. In order to reduce uncertainties from this variability, EFs were calculated using average values from six or seven cycles for the spruce and birch cases respectively. For the wood log cases, these averaged values are reported in Table 1 under the sub heading whole wood cycle. For the adjacent columns subtitled FIGAERO sample, see the description of sampling in the Iodide FIGAERO-ToF-CIMS section. Emission factors are shown in Fig. 1.
The EFs were combined to reflect a more realistic case assuming 90% ‘good’ combustion and 10% ‘bad’ combustion as elsewhere.68 The 10% bad combustion criteria referenced here is based on an assumption of user behaviour rather than on data and is termed “expected behaviour” by the authors of that study. Additionally, a worst case 20% bad combustion scenario termed “worse than expected” and best case 0% bad combustion scenario termed “good combustion” were used to produce a range of realistic results. Here we use the 10% bad combustion or, “expected behaviour” scenario as a basis for our national burden scenario. Here, instead of weighting ‘typical EFs’, we use the efficient high pellet load EFs for the ‘good’ case and less efficient low pellet load for the ‘bad’ case. For the wood stove, birch is assumed the ‘good’ case as it was the most efficient, and spruce the ‘bad’ case, as it was less efficient.
This analysis uses emission factors presented in this manuscript only, i.e., derived from these two individual appliances. Additionally, the energy swap scenario may represent an unrealistically large transition to pellet use from wood log combustion. As such, this analysis is not meant to provide evidence for a realistic policy implementation but is an exercise in contextualising our results and demonstrating hypothetical impacts, rather than presenting definitive burdens or burdens for any specific country. A more detailed study using a more accurate representation of RWC appliances, and a more realistic scenario would be beneficial to appreciate the true impact of the emission factors derived here, however such a comprehensive study is beyond the scope of this manuscript.
Typical concentrations of OC during sampling were ∼1000 μg m−3 during the high MCE birch and pellet high load experiments (Table S3†). This increased to ∼2500 μg m−3 for the low pellet loading and ∼5700 μg m−3 for the fresh spruce experiment. The wood log measurements were made at higher dilution rates than the pellet measurement, as described in the dilution tunnel section. The FIGAERO sampling time of 5 minutes at 2 slm then provided between 20–50 μg deposited onto the filter for analysis.
The EFBC for the low load pellet case and wood logs (206 ± 1750 mg kg−1 to 284 ± 326 mg kg−1) are comparable to literature values59,63,73 whilst the high load pellet case is five to seven times lower (42 ± 60 mg kg−1). Pellet EFBC of these magnitudes have been reported elsewhere for both pellet and wood log burning stoves.73,74 High BC EFs indicate fuel rich, oxygen poor environments. This is counter-intuitive for the low loading pellet case, as it is expected that when the fuel loading is decreased, the fuel to oxygen ratio decreases and the environment is more oxygen rich. However, for these appliances, air injection is controlled autonomously. This suggests that at lower loadings, the air injection into the combustion chamber was not optimal, thus creating a fuel rich environment, even at lowered fuel loadings.
EFsOC also broadly agree with literature values59,63,75 for all appliances, although here, pellet EFOC of 220 ± 56 mg kg−1 for the high loading and 638 ± 394 mg kg−1 for the low loading, are generally greater than the birch EFOC (170 ± 237 mg kg−1) and spruce (196 ± 7550 mg kg−1) cases. The low pellet load EFOC is noticeably larger than all other EFOC, although the wood log EFsOC demonstrate extremely high variabilities of between four to 38 times their average values. The low pellet load EFOC is also much larger than those used elsewhere to calculate RWC impacts.68 As these EFs are derived from optical, online measurements rather than the gravimetric, offline filter-based measurements, there is the potential that volatile compounds that would otherwise evaporate from the filter before measurement may instead be captured here, providing larger than typical values. The variabilities reported for the wood logs EFs are high which reflects the entire combustion cycle.
Offline GC-MS detects only three of 16 PAHs (Table S2†) in the pellets and birch cases and at relatively low abundances, i.e. 0.006% of total PM. In contrast, all 16 are identified in the spruce case and at much higher magnitudes, 0.04% of PM.
An AAE of 0.9 ± 0.1 for the birch emission particles and 1.0 ± 0.6 for the spruce particles indicate a lack of BrC compounds modifying typical BC absorption; however, brief periods of AAE >4.0 are observed when pollutant emissions are high, e.g., on refuelling, suggesting the presence of BrC at those times when temperatures are lower and combustion is least efficient. The birch particle MAC405 of 6.2 ± 11.0 m2 g−1 and MAC781 of 3.1 ± 5.1 m2 g−1 are more similar to the pellet MACs, suggesting a similar level of absorption, but are an order of magnitude lower than for spruce particles, where MAC405 = 33.8 ± 27.4 m2 g−1 and MAC781 = 11.3 ± 10.3 m2 g−1 indicating more intense absorption. It is also clear the wood log MACs are much more variable, again in line with combustion cycles. PAHs form under similar conditions to BC, can be effectively absorbed by BC and contribute to BC formation through condensation56,76 thus the concomitance of high EFBC and EFPAH in the spruce case is expected.
As described previously, for the wood log cycle averages, the MCE was lower for the spruce combustion compared to the birch. For the transient sampling, this distinction is amplified. The birch sampling occurs at an even greater MCE of 0.9938 ± 0.0035 whereas the spruce sampling occurs at an even lower MCE of 0.9615 ± 0.0218 when compared to the cycle averages. Correspondingly, the EFOC of the transient measurement during FIGAERO sampling is lower for the higher MCE birch case at 149 ± 77 mg kg−1 or approximately 85% that of the average EFOC. Further, the transient EFOC of 689 ± 1390 mg kg−1 for the lower MCE spruce case is approximately 3.5 times greater than the cycle average. See Table 1 for comparisons of transient vs. average EFs.
For the pellet emission particles, the sum of speciated particulate OM measured by the FIGAERO-ToF-CIMS (EFOM_FIG_par) was 29.0 ± 8.70 mg kg−1 for the high loading and 65.7 ± 19.7 mg kg−1 for the low loading (Table 1). EFOM_FIG_par for the wood log emissions was 46.1 ± 13.8 mg kg−1 for the birch case and 81.5 ± 24.5 mg kg−1 for the spruce case. These are of a comparable order of magnitude to the those of the pellet cases, although broadly, the lower MCE low pellet loading and spruce cases exhibit the higher EFOM_FIG_par compared to the higher MCE high pellet loading and birch cases.
For the pellet cases and the spruce case, approximately 10–13% of the OM is quantified by FIGAERO-ToF-CIMS but for the birch case this increases to 31%. This suggests the majority of the OM is comprised of compounds not readily detectable by FIGAERO-ToF-CIMS. This may be for several reasons. For example, some compounds are not detectable by iodide ionisation e.g. resin acids,77 aliphatic compounds or other evaporites that have not combusted in the low temperature pellet cases. Additionally, some larger molecular weight compounds have lower volatilities than those accessible through filter evaporation as employed by the FIGAERO and thus remain unquantified. It may be that the high MCE and high temperatures of the birch combustion produces the greatest quantity of detectable compounds e.g., more oxygenated and of lower molecular weights. Additionally, the combustion conditions between the birch and high pellet loading cases are different: the temperature of the log fire is greater, and the heterogeneity of the fuel in the chamber is greater, e.g. producing more pyrolytic conditions and thus more pyrolysis products.
As levoglucosan is a pyrolysis product and well-characterised BB tracer,78 we describe its emission characteristics here as an example for the behaviour of an important semi-volatile species. Particle phase EFLevoglucosan for the high pellet load were 16.5 ± 4.9 mg kg−1 and 24.6 ± 7.4 mg kg−1 for the low pellet load, consistent with literature values.75,78–81 This corresponds to 5% and 2% of PM respectively, which again agrees with the literature.81–83 A particle phase EFLevoglucosan of 26.9 ± 8.1 mg kg−1 for the birch case is comparable in magnitude to the low pellet loading. This is consistent with the descriptions of (a) combustion heterogeneity in the high MCE, high temperature birch case, and (b), the lower MCE, high EFBC description of the lower pellet loading case, both suggesting or indicating more pyrolytic conditions (see chemical composition of emission section). Moreover, the high temperature, low MCE spruce case presents the highest EFLevoglucosan of 49.4 ± 14.8 mg kg−1, indicating these conditions are the most conducive for production of pyrolysis products. This is confirmed by the very high transient spruce case EFBC of 1260 ± 1716 mg kg−1. This is further reflected in the wood cycle average EFPAH where 172 mg kg−1 is measured for the spruce case compared to 4 mg kg−1 for the high pellet loading, 7 mg kg−1 for the low pellet loading and 17 mg kg−1 for the birch case.
Levoglucosan is a pyrolysis product requiring high temperatures and low oxygen content for its formation. While some SMAs are formed in a similar manner, e.g. nitrocatechol, others, like vanillin, require an environment to which the pellet stove is optimised to ensure they are not combusted, i.e., oxidative conditions and lower temperatures. The 19 SMAs (see Materials and methods) described in the literature (Table S1†) as BrC components32,33,37,40,48,84 that were identified in this work contributed 1.4% by mass of OM for the high pellet loading and 1.8% for the low pellet loading. Sum particle phase EFs (EFSMA_par) were 3.12 ± 0.94 mg kg−1 for the high pellet loading and 11.8 ± 3.60 mg kg−1 for the low pellet loading. When compared with their equivalent MCE wood log counterparts, these are an order of magnitude greater, cf. the high pellet loading EFSMA_par with the birch case, where EFSMA_par = 0.61 ± 0.18 mg kg−1, and the low pellet loading with EFSMA_par = 2.28 ± 0.69 mg kg−1 for the spruce case. One explanation may be a greater proportion of SMAs do not combust in the pellet stove and so can survive to be emitted. This anecdotally agrees with lower pellet stove chimney bottom temperatures of 100 °C–150 °C compared to the wood burning stove. This explanation is predicated on the assumption that chimney bottom temperature is an accurate proxy for combustion chamber temperature, which may not be the case if heat transfer from the combustion chamber is very efficient and minimal heat is lost through the chimney.
The speciation of the organic fraction is substantially different between appliances, at similar OC emissions, likely caused by the differences in MCE, combustion temperature and fuel heterogeneities in the combustion chamber. The low temperature pellet emissions preserve a greater quantity of SMAs compared to the high temperature wood log cases. Lowering the pellet loading causes more pyrolysis, increasing BC, PAHs and levoglucosan, but also reduces temperatures, preserving more OC as well as SMAs. The high MCE birch case is still susceptible to fuel and temperature heterogeneities in the combustion chamber causing an increase in BC, PAH and levoglucosan, exacerbated by higher temperatures, which also combusts the SMAs. The spruce case had high temperatures and low MCEs again with heterogeneities in the combustion chamber, which produces the largest quantities of BC, PAH and levoglucosan.
For the wood cases, AAEs equal to one indicate a lack of wavelength dependence on absorption and thus a lack of BrC, which reconciles with low EFSMA. Additionally, MACs for the birch emissions were of a comparable order to the pellet emissions, indicating weak absorption. Any absorption is then attributed to the BC, of which, little was measured during FIGAERO sampling. MACs for the spruce emissions were an order of magnitude greater than for the birch, suggesting much stronger absorption, which is attributable to the order of magnitude increase in EFBC. Although PAHs were measured for the wood emissions, they represent cycle averages and are difficult to compare with the transient OC, BC and SMA EFs and transient optical measurements discussed here. However, PAHs emissions were clearly greater from the wood cases compared to the pellets, but not so great as to modify the dominance of BC on the optical absorption.
Scenario | Appliance | Energy consumption (PJ) | BC (kt) | OC (kt) | PM (kt) | THC (kt) | PAHs (t) | FIGAERO organics particle (gas) (t) | Levoglucosan particle (gas) (t) | Literature. SMAs particle (gas) (t) |
---|---|---|---|---|---|---|---|---|---|---|
Current | Pellet | 0.7 | 0.003 | 0.013 | 0.027 | 0.059 | 0.07 | 1.59 (1.33) | 0.84 (0.02) | 0.20 (0.26) |
Wood | 7.0 | 0.116 | 0.071 | 0.302 | 0.652 | 13.3 | 21.2 (3.78) | 12.4 (0.06) | 0.35 (0.21) | |
Total | 7.7 | 0.119 | 0.084 | 0.329 | 0.711 | 13.4 | 22.8 (5.11) | 13.3 (0.09) | 0.55 (0.47) | |
Swapped | Pellet | 7.0 | 0.029 | 0.125 | 0.270 | 0.592 | 0.70 | 15.9 (13.3) | 8.45 (0.23) | 1.97 (2.6) |
Wood | 0.7 | 0.012 | 0.007 | 0.030 | 0.065 | 1.33 | 2.12 (0.38) | 1.24 (0.01) | 0.04 (0.02) | |
Total | 7.7 | 0.041 | 0.132 | 0.300 | 0.657 | 2.03 | 18.01 (13.7) | 9.69 (0.24) | 2.00 (2.63) | |
Total | Current-swapped | −0.078 | 0.049 | −0.029 | −0.054 | −11.3 | −4.73 (8.58) | −3.59 (0.15) | 1.45 (2.16) |
Given the assumed energy demand distribution here, the PM burden was 0.329 kt, the BC burden was 0.119 kt and the OC burden was 0.084 kt. Swapping demand resulted in a PM reduction of 12% to 0.300 kt and a BC reduction of 42% to 0.041 kt. Additionally, the particulate PAH burden reduced considerably by 84% from 13.37 t to 2.03 t. In contrast was an increase of OC to 0.132 kt, or 57%. However, the sum of OC speciated compounds (particle phase FIGAERO organics) reduce by 20% from 22.8 t to 18.1 t. For the gas phase FIGAERO organics, there was instead an increase from 5.1 t to 13.7 t (268%), shifting much of the particle mass savings to the gas phase. In line with the decrease in particle phase FIGAERO organics is the decrease in the levoglucosan burden by 27% from 13.3 t to 9.7 t. Once again there is an increase in corresponding gaseous emissions, from 0.09 t to 0.24 t (267%).
Unlike the FIGAERO organics and levoglucosan, the SMA burden increases in both particle (0.55 t to 2.00 t, 364%) and gas phases (0.47 t to 2.63 t, 560%). Despite these large relative increases, the absolute total mass of the SMAs is still small compared to other classes of organic compounds, however the health impacts of this magnitude increase are uncertain. An additional uncertainty is how important this increase might be, considering the reduction of other deleterious material such as the PAHs.
From this exercise, two points become clear. Firstly, the comparison of emissions from the two stoves in this study shows that the pellet stove was a greater source of at least some classes of organic carbon that are relevant to climate processes and health, i.e. the SMAs. This is also evident from the reported EFs (on both a per fuel mass and per useful energy delivered basis). Secondly, even if particle phase emissions were decreased, corresponding gas phase emissions could increase, in part, offsetting the total reduction of a pollutant. Shifting the distribution of a pollutant from the particle to gas phase is likely to increase the propensity for secondary chemistry later on, which could further impact climatic and health effects. This highlights the issue of shifting a primary pollution issue into a secondary pollution issue and the need for further studies to quantify the particle burdens from POA and precursors of SOA.
This study also highlights the importance of simultaneous gas and particle measurements to accurately quantify total, rather than just particle phase, organic emissions from RWC appliances. Furthermore, for an extended study covering a larger range of available appliances, e.g., linked to emission inventories, there may be many benefits to applying this methodology to assess the effectiveness of, for example, appliance swap schemes, or policies that aim to transition to technologies where the gas phase component of emission might be significant.
For both appliances, periods of high MCE were characterised by high emissions of CO2 and low emissions of particles and other gases. Average particle mass loadings from the pellet stove were lower than the wood-burning stove, although, under low loadings, the pellet stove emitted a larger concentration of particle mass than efficient wood burning, including a significant OC fraction.
Although mass concentrations were highly variable between the operational modes of the two appliances, the physicochemical properties of the particles were distinct to each appliance. The pellet particles were spherical, approximately 10% BC and approximately 90% optically absorbing OM (BrC) by mass. Contrastingly, the wood log particles were larger and mainly amorphous BC accounting for approximately 50% of PM mass.
The optically active components of the OC rich pellet particle emissions were a mixture of PAHs and SMAs. Sum EFSMA were up to an order of magnitude greater in the pellet cases than for the wood log emissions. Decreasing the fuel loading of the pellet stove decreased the MCE. Correspondingly, BC EFs increased by a factor of six, OM EFs by a factor of five as well as the optical absorption (MAC) of the material by a factor of four to five, caused by an increase in EFs of SMAs by a factor of four and PAHs by a factor of two. Thus running the pellet boiler under reduced fuel loadings produced more particles and ‘browner’ particles.
Optical absorption from birch emission particles at a high MCE was of a comparable magnitude to the pellets, although no BrC was present. The spruce particle emissions measured at low MCE indicate much stronger absorption than the high MCE birch particle emissions. This absorption was likely dominated by high emissions of BC. PAH emissions from the low MCE spruce particle emissions were ten times greater than the next highest emission from the high MCE birch particles and 100 to 200 times greater than those measured from the pellet stove.
To contextualise the impact of the EFs, the impact of swapping energy demands between the wood log stove and the pellet stove tested in this study was estimated at a national scale using a hypothetical pollution burden analysis scenario. This hypothetical pollution burden analysis is specific to the individual appliances studied here, as well as the conditions under which they were operated, meaning these results are not necessarily representative of the appliance types in general, or their true usage in the field. Reductions of 12% for PM and 42% for BC were somewhat offset by an increase in OC of 57%. The effect on speciated organic compounds were more nuanced. For example, total concentrations of key compounds decreased, such as the major product levoglucosan, whereas others, such as the SMAs, increased. Irrespective of how the total burden of speciated organics change, gas phase burdens always increased. This illustrates one of the major challenges in emissions characterisation of organic compounds: that shifting particle phase material into the gas phase via phase partitioning can obscure true reductions in total material emitted, and also provide delayed particulate pollution if the gas phase emissions are precursors to subsequent SOA formation. As such, future work should investigate representations of phase partitioned semi-volatile organic compounds from polluting emission sources. Additionally, more accurate assessments of national emissions considering gas and particle phase emissions are needed to better understand the true effectiveness of abatement policies on pollutant reduction. Investigating the SOA formation potential from shifting emissions into the gas phase should also be investigated in greater detail, as this might have ramifications for transporting delayed particle pollution away from source regions.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2ea00022a |
‡ Now at College of Environmental and Resource Sciences, Zhejiang University, 310058 Hangzhou, China. |
§ Now at AWA Sweden AB, Södra Hamngatan 37-41, 41106 Göteborg, Sweden. |
This journal is © The Royal Society of Chemistry 2023 |