Integrated assessment of volatile organic compounds from industrial biomass boilers in China: emission characteristics, influencing factors, and ozone formation potential

Industrial biomass boilers (IBBs) are widely promoted in China as a type of clean energy. However, they emit large amount of volatile organic compounds (VOCs) and the emission characteristics and the underlying factors are largely unknown due to the sampling difficulties. In this study, three wood pellet-fueled and two wood residue-fueled IBBs were selected to investigate the characteristics of VOC emissions and to discover their underlying impacting factors. The emission factor of VOCs varied from 21.6 ± 2.8 mg/kg to 286.2 ± 10.8 mg/kg for the IBBs. Oxygenated VOCs (OVOCs) were the largest group, contributing to 30.3 − 73.6% of the VOC emissions. Significant differences were revealed in the VOC source profiles between wood pellet-fueled and wood residue-fueled IBBs. Operating load, excess air, furnace temperature, and fuel type were identified as the primary factors influencing VOC emissions. The excess air coefficient should be limited below 3.5, roughly corresponding to the operating load of 62% and furnace temperature of 630 °C, to effectively reduce VOC emissions. VOC emissions also showed great differences in different combustion phases, with the ignition phase having much greater VOC emissions than the stable combustion and the ember phases. The ozone formation potential (OFP) ranged from 4.3 to 31.2 mg/m3 for the IBBs, and the wood residue-fueled IBBs yielded higher OFP than the wood pellet-fueled ones. This study underscored the importance of OVOCs in IBB emissions, and reducing OVOC emissions should be prioritized in formulating control measures to mitigate their impacts on the atmospheric environment and human health.


Introduction
Biomass is considered as a type of clean and renewable energy source due to its low emissions of sulfur dioxide (SO 2 ) and nitrogen oxide (NO x ) during combustion (Vassilev et al. 2015). However, biomass burning is second only to biogenic sources in terms of volatile organic compound (VOC) emissions globally (Yokelson et al. 2013;Brilli et al. 2014;Vassilev et al. 2015). VOCs not only pose adverse impact on human health by containing a great number of toxic and harmful constituents (Boeglin et al. 2006;Chen et al. 2017;Ballesteros-González et al. 2020), but also act as important precursors in the formation of ozone and secondary organic aerosol (SOA) in the atmosphere (Crutzen and Andreae 1990;Johnson et al. 2006;Yuan et al. 2013;Alvarado et al. 2015;Zhang et al. 2017). In order to formulate effective control strategies to negate the environmental and human health impacts of VOCs, characterization of VOCs from biomass burning is urgently needed.
Biomass burning refers to a large class of activities involving combustion of biomass. Over the past few decades, extensive researches have been conducted to investigate VOC emissions from open biomass burning of specific fuels (e.g., Christian 2003;Yang et al. 2020); open biomass burning of crop residues, forest, shrub land, and grassland (e.g., Karl et al. 2007;Kudo et al. 2014;Sirithian et al. 2018;Mor et al. 2021); biomass fuels in residential boilers and stoves (e.g., Zhang and Smith 1999;Huy et al. 2021;Zhang et al. 2021); and biomass fuels in domestic fireplaces (e.g., McDonald et al. 2000;Evtyugina et al. 2014). However, studies on VOC emissions from biomass fuels in industrial biomass boilers (IBBs) are relatively scarce (Yan et al. 2016;Geng et al. 2019;Zhang et al. 2020).
Biomass fuels in IBBs are mostly in the forms of molded pellets, including wood pellets and straw pellets, as well as raw materials, including wood residues such as eucalyptus wood chips, bark, sawdust, and wood shavings (Guan et al. 2020;Santana et al. 2021). Geng et al. (2019) characterized 107 VOC species from two IBBs fueled by wood pellets and straw pellets, respectively. They found that oxygenated VOCs (OVOCs) and aromatics accounted for the major part of VOCs, and the emission factors (EFs) of straw pellets were higher than that of wood pellets. Zhang et al. (2020) measured 86 VOC species from five bagasse-fueled IBBs and found alkenes and OVOCs contributed the most to VOCs. Still in the initial stage, the studies of IBB emissions have two major limitations. One is the lack of VOC characterization in different burning phases. Burning phase-specific VOC compositions have been measured for residential biomass combustion and small-scale pellet boiler (Bhattu et al. 2019;Czech et al. 2017), but not for IBBs. In addition, as VOC emissions from IBBs are strongly dependent upon operating conditions, the operational factors influencing VOC emissions from IBBs needed to be well investigated to provide scientific support in formulating of control measures. Influencing factors of VOC emissions have been studied in detail for residential stoves (Fachinger et al. 2017;Bhattu et al. 2019;Sun et al. 2019; Thompson et al. 2019), but not for IBBs either.
It is noted that the number of samples collected in the published IBB source characterization studies are relatively fewer, mostly no more than five (Yan et al. 2016;Geng et al. 2019;Zhang et al. 2020). This is because IBB sampling is technically challenging. Compared with other types of biomass burning such as residential stove and crop residual, the IBB sampling platform is narrow and the sampling port is high from the ground, as illustrated in Fig. S1 in the supporting information (SI), thus posing great challenges to the safety of sampling personnel and the placement of sampling system. In addition, IBB sampling is largely restricted by the operational time of the factory, significantly reducing the flexibility of the sample collection. These technical difficulties make IBB probably the least studied category in the biomass burning emission characterization.
Considering the scientific limitations of the current studies and the challenges in IBB sampling, we conducted measurement of VOC emissions from five IBBs in this study, determined their EFs and source profiles, evaluated the chemical reactivity of VOC emissions, and estimated the impacts of a set of parameters including operating load, excess air coefficient, furnace temperature, fuel type, and different combustion phases on the VOC emissions from IBBs. The results of this study would deepen understanding on VOC emissions from IBBs in China, and provide a useful reference for optimizing the operating conditions of IBBs so as to reduce VOC emissions and alleviate their contributions to the ambient ozone and particulate matter pollution.

Information of IBBs and fuel
China's biomass feedstock reached 102.57 million cubic meters in 2021, with Guangdong province contributing about 10% of the production (http:// www. stats. gov. cn/ tjsj/ ndsj/ 2021/ index ch. htm). Wood pellet-fuel manufacturers are mainly concentrated in southern China due to the relatively extensive forest area in these regions (Guan et al. 2020). There are also large quantities of wood residues generated from the wood processing industry in these regions. Wood pellet-fuel has commonly been used in IBBs to generate steam, while there are also some IBBs that use wood residues as fuel to generate steam in Guangdong province. In this regard, wood pellets and wood residues are representative IBB fuels and selected in this study.
Three wood pellet-fueled IBBs (BB1, BB2, and BB3) and two wood residue-fueled IBBs (BB4 and BB5) in Zhaoqing City, Guangdong Province, China, were selected for VOC sampling and analysis. The average operating load of BB1 − BB5 were 52%, 59%, 77%, 35%, and 71%, respectively. Detailed information of the selected IBBs is listed in Table S1 in the SI, and the sampling photos are presented in Fig. S1 in the SI. An electric heating silicon nitride ceramic igniter was used to ignite the IBBs. Since electric heating does not cause additional emissions of VOCs, it would not affect the collection of VOCs. All IBBs were equipped with dust removal devices, while no desulfurization and denitrification unit was installed. Flue gas treatment systems were operating normally during sampling period. The oxygen concentrations of flue gas were monitored by a portable infrared flue gas analyzer (Photon, Madur, Austria). Detailed information on the fuels of wood pellets and wood residues, including elemental composition and heating value, is provided in Table S2 in the SI, and the photos of biomass fuels used in this study are shown in Fig. S2 in the SI. For both kinds of fuels, the percentages of volatile matter contents by weight were greater than 74.6%, and the moisture contents of fuels on an air-dried basis ranged from 6.2 to 10.7%.

Sampling system
The VOC sampling system for IBBs is illustrated in Fig. 1. The sampling outlet was located between the flue gas treatment device and the chimney vent. The flue gas was pumped for several minutes to assure the air in the pipeline was drained away before sampling. The sampling system consisted of three modules. The first module was a portable flowmeter (Series 2440, Kurz Instruments Inc., USA) which detected the flue gas flow in the chimney, and the sampling probe was inserted at the end of the flow tube. The second module was a portable infrared flue gas analyzer (Photon, Madur., Austria) that measured the oxygen content of the flue gas. The third module was the VOC sampling channel composed of the heated pipelines and filters to remove water and particulate matter within the flue gas, respectively. The heated pipelines were maintained at a temperature about 120 °C to prevent condensation of the VOCs, and the Teflon pipe was used behind the stainless steel tube to reduce adsorption. The flue gas was then separated into two branches. The first branch connected a SUMMA canister (3.2 L, Entech Instruments., USA) to collect non-methane hydrocarbons (NMHCs), and the sampling time was 20 min. The SUMMA canister was rinsed by Nutech 2101DS cleaner for eight cycles using high-purity nitrogen before sampling to ensure no residual contamination. The SUMMA canisters were blank analyzed to determine whether or not they were clean. One blank SUMMA canister filled with high-purity nitrogen from each SUMMA canister batch was analyzed, and the results for each target compound were below the method detection limit (0.01-0.75 µg/m 3 ), which comply with the blank standard of US Environmental Protection Agency (USEPA). The second branch connected in serial two silica cartridges coated with 2,4-dinitrophenylhydrazine (DNPH) (200 mg/6 mL, Bonna-Agela., China) to collect OVOC samples. The OVOC concentration was obtained by adding the concentrations in the two tubes. The flow rate was approximately 200 mL/min, and the sampling time was 20 min. A KI oxidant scrubber (Bonna-Agela., China) was used before the DNPH cartridges to eliminate the interference from ozone. After sampling, DNPH cartridges were packed tightly with tin papers or aluminum foils, stored at temperature lower than 4 °C, and analyzed within 7 days after sampling to minimize sample degradation during cold storage. VOC samples collected from SUMMA canisters were also analyzed within a week. To increase the representativeness, we collected three VOC samples consecutively for each IBB.
Samples were collected when the flue gas was almost a constant flow through the chimney. In this regard, the impact of the flow rate fluctuation during the sampling period would be minimized. In addition, as highlighted in Fig. 1, a flowlimiting valve (Entech Instruments, CS1200E Constant flow sampler) was used to ensure that flue gas was collected at a constant rate throughout the sampling period. Meanwhile, the VOC sampling system was positioned on the sampling platform upstream from the curved structure more than 3 times the chimney diameter to ensure that there were no flow disturbances from any air supply system (PRC MEE 2007).
All five IBBs were measured under the stable combustion phase. To examine the differences of VOC emissions in different combustion phases, we also collected VOC samples during the ignition phase and the ember phase separately for BB1. Flue gas temperature is the essential parameter to determine the combustion phases. The ignition phase was defined as the period after ignition until the flue gas temperature stabilized, the stable combustion phase followed with high and relatively stable flue gas temperature, and the ember phase was the period with continuously decreased temperature after the shut-down procedure was initiated (Fachinger et al. 2017). Figure S3 in the SI shows the change of flue gas temperature with elapsed time and the separation of three phases. Due to the short duration of the ignition phase, only one sample was collected. An 8-L Tedlar sampling bag was used to collected VOC sample for 2 min, and the sample was immediately transferred to the SUMMA

VOC analysis
C 2 − C 12 NMHC sampled by the SUMMA canister were analyzed according to the TO-15 protocol recommended by USEPA. The high concentrations of the source samples were beyond the range of GC/MS analysis, so the source samples were diluted to reach the GC/MS detection range before analysis. The source samples were pre-analyzed, and the appropriate dilution ratio was determined according to the ratio of the highest response of the pre-analyzed compound to the highest concentration point (100 ppb for source samples) of the standard curve. Then, according to the determined dilution ratio, the source samples were diluted with a high precision diluter for qualitative and quantitative analysis. The dilution of the source samples in this study ranges from 3 to 100 times. The diluted samples were concentrated using a threestage channel cryogenic pre-concentrator (Model 7200, Entech Instruments Inc., USA), and the moisture and CO 2 were removed by a water management trap and a soda asbestos tube, respectively. VOC species were characterized by a gas chromatography (7890B, Agilent Technologies Inc., USA) coupled with a quadrupole mass spectrometry/flame ionization detection system (5977B, Agilent Technologies Inc., USA). The OVOC analysis was conducted according to the TO-11A protocol recommended by USEPA. OVOCs collected in the DNPH cartridges were eluted with 5 mL of acetonitrile in a pollution-free environment, and the eluent was extracted to a 3-mL volumetric flask for testing and finally analyzed by the high performance liquid chromatography (1260, Agilent Technologies Inc., USA). Detailed analysis methods were described in the previous studies Yan et al. 2016). In brief, the VOC samples were cryogenically concentrated and removed H 2 O and CO 2 through a liquid nitrogen-cooled cryogenic trap (0.32 cm × 20 cm) at − 160 °C, a trap (0.32 cm × 20 cm) at − 40 °C with TenaxTA as adsorbents, and cryofocus trap (0.08 cm × 5 cm) at − 170 °C in sequence. After that, the temperature of the trap will rise rapidly and the VOCs were transferred to the GC-MSD/FID system. The GC equipped with a DB-1 capillary column (60 m × 0.32 mm × 1.0 μm, Agilent Technologies, USA) helium was used as the carrier gas. The initial GC oven temperature was 10 °C for 3 min, and the temperature was raised to 120 °C at 5 °C/ min; then, the temperature was raised to 250 °C at 10 °C/ min, and finally held at 250 °C for 10 min. The MSD was performed under the selective ion mode, and the ion source of mass spectrometry was electron ionization (EI) with electron impacting of 70 eV.

Calculation of emission factors
The EFs of IBBs were calculated on a mass basis by Eq. (1) (Yan et al. 2016).
where EFs is the emission factor (mg/kg fuel); C i is the converted mass concentration at a reference oxygen level of the total VOCs (mg/m 3 ); Q is the volume flow rate of flue gas in the chimney (m 3 /h); M is fuel consumption per hour (kg/h).
In order to avoid the impact of different levels of excess air in different IBBs, the measured concentrations of each VOC species were converted by Eq. (2).
where C i is the converted mass concentration at a reference oxygen; C i,M is the measured concentration in the flue gas; (O 2 ) A is the oxygen content in the ambient air, which is 21%; (O 2 ) M is the measured oxygen content in the flue gas; (O 2 ) R is the reference oxygen content. A value of 9% was used for (O 2 ) R in this calculation, as suggested by the Chinese standard GB/T 13,271-2014 for the coal-fired and biomass-fired boilers (PRC MEE 2014).
In general, combustion processes occur with an amount of excess air. The excess air coefficient, , can be estimated based on the measured oxygen content by Eq. (3).

Estimation of ozone formation potential
The ozone formation potential (OFP) of a VOC species was calculated using the maximum incremental reactivity (MIR) value by Eq. (4).
where OFP is the ozone formation potential (mg/m 3 ); C i is mass concentration of species i (mg/m 3 ); MIR i is the MIR value of species i extracted from Carter. (2010). (1)
By normalizing concentrations of VOC species with respect to the total VOC mass concentration, we obtained the VOC source profiles from the five IBBs, as shown in Fig. 2. The top 10 species in abundance are marked blue, and their contribution percentages were shown in Table 1. They accounted for 71.4 ± 0.9%, 80.9 ± 6.3%, 75.7 ± 1.5%, 74.2 ± 1.1%, and 74.8 ± 2.5% of the total VOCs in BB1 − BB5, respectively. Although the total proportions of the dominant species were similar among the five IBBs, large differences existed in the VOC source profiles. In general, wood pellet-fueled IBBs emitted higher amounts of formaldehyde, acetaldehyde, benzene, and ethyne, whereas wood residue-fueled IBBs emitted much greater proportions of methyl ethyl ketone (MEK), ethane and acrolein but significantly lower benzene and ethyne. Chloromethane, a tracer for VOC emissions from biomass burning (Wang et al. 2009;Sun et al. 2019), was detected from all five IBBs. As OVOCs was the dominant group, we further analyzed the relative contributions of 18 OVOC species in the five IBBs, as shown in Fig. 3. Formaldehyde and acetaldehyde were the most abundant species, accounting for 42.9 − 61.0% of the total OVOCs. This finding was consistent with those from previous biomass combustion studies. For example, Hedberg et al. (2002) found that formaldehyde and acetaldehyde accounted for 49.8% of the total OVOCs emitted from the residential stoves. Mo et al. (2016) reported their total proportions of 68.3%, 43.7%, and 52.4% from the burning of rice stalks, corn stalks and wheat straw, respectively. We also found that acrolein and acetone accounted for 5.9 − 33.8% of the total OVOCs. The predominance of OVOC species with relatively lower molecular weights may be due to the incomplete oxidation combustion of alkanes and alkenes in the furnace with a high temperature. Apart from the lowmolecule OVOCs, some high-molecule OVOCs, including aldehydes, alcohols, and lipids, were also observed in the IBB emissions. For example, benzaldehyde accounted for 7.6% and 3.8% in BB1 and BB5, and vinyl acetate accounted for 11.2% and 7.5% in BB2 and BB5, respectively.

Emission factors of VOCs
The total and group-specific EFs are shown in Fig. 4a. Among the five IBBs, the total EFs ranged from 21.6 ± 2.8 mg/kg to 286.2 ± 10.8 mg/kg of fuel. Significant differences in EFs were revealed among IBBs, which cannot be solely explained by the difference in fuel types. Studies have shown that a series of factors, including fuel type, combustion mode, boiler type, and smoke treatment, can impact VOC emissions (Burling et al. 2010;Achaw and Afriyie 2015;Sefidari et al. 2014;Szyszlak-Bargłowicz et al. 2015;Pognant et al. 2018;Geng et al. 2019). Influences of the  Fig. 3 Percentages of OVOC species emitted from the five IBBs factors on VOC emissions were analyzed in detail in "Factors influencing VOC emissions" section. We further examined EFs in different VOC groups. OVOCs were consistently the largest contributor to the total VOC emissions. Their EFs ranged from 6.5 ± 3.3 mg/kg to 120.1 ± 11.5 mg/kg, accounting for 30.3-73.6% of the total VOCs, as shown in Fig. 4b. Zhang et al. (2020) found that OVOCs were the second largest group from IBBs, accounting for 13.4 − 42.2% of the total VOCs. Their relatively lower OVOC contribution was mainly because they did not measure formaldehyde. The largest contribution of OVOCs highlighted the importance of including OVOCs in the emission characterization of the IBBs.
The second largest group was aromatics, with the EFs ranging from 5.9 ± 0.2 mg/kg to 64.1 ± 8.5 mg/kg and accounting for 16.1 − 27.3% of the total VOCs. This portion was close to 27.9 − 29.2% reported by Geng et al. (2019). Benzene and some other aromatics are mainly produced by the degradation of lignin, so the high content of lignin in the biomass fuel may be responsible for the high emissions of aromatics (Greenberg et al. 2006;Evtyugina et al. 2014). Emissions of alkanes and alkenes were relatively lower from all five IBBs, accounting for 3.6 − 14.6% and 0.9 − 9.2% of the total VOCs, respectively. Alkanes and alkenes are mainly produced by the thermal degradation of cellulose and hemicellulose in 240 − 350 °C and 200 − 260 °C, respectively (Greenberg et al. 2006). The low emissions of alkanes and alkenes from the IBBs may be due to the much higher furnace temperature (> 500 °C) which further oxidized alkanes and alkenes into OVOCs. Table 2 compares the VOC EFs in this study with those reported in previous studies and assessed the differences between them. VOC EFs from IBBs in this study were generally within the range of reported values (Geng et al. 2019;Zhang et al. 2020), but much lower than the official recommended value for IBBs in China (1130 mg/ kg) (PRC MEE 2014). VOC emission inventories using the recommended EFs may markedly overestimate the actual emission from IBBs. EFs of VOCs from IBBs in this study were about three times those from power plants boilers (Yan et al. 2016). This is mainly due to the fact that OVOCs, accounting for up to 30.3%-72.1% in this study, had not been detected in power plant boilers. VOC EFs from IBBs in this study were significantly lower than that of residential crop residue briquettes stove . Apart from the difference in fuel, the equipment of a special combustion chamber with an air supply system and a bag filter as the flue gas treatment device are the primary factors for the low EFs from IBBs. In comparison, the residential biomass stove usually only has a simple combustion chamber without an air supply system, leading to lower combustion efficiency and higher emissions from incomplete fuel combustion (Evtyugina et al. 2014;Thompson et al. 2019). Flue gas is often emitted directly through the chimney without flue gas treatment device or even without the chimney Weyant et al. 2019). Therefore, lower combustion efficiency and the absence of flue gas treatment devices collectively
To examine the influence of fuel properties and operating conditions on VOC emissions from the IBBs, correlations between EFs and a set of parameters were calculated, as shown in Table S4 in the SI. A first glance revealed that furnace temperature, operating load, heating value, fixed carbon, and element carbon all showed negative correlations with the EFs. We shall analyze their impacts on VOC emissions in detail in the following sections.

Operating load, excess air coefficient, and furnace temperature
Operating load, excess air coefficient, and furnace temperature are three parameters that are interrelated with each other. Operating load refers to the percentage of the rated load of the boiler during operation, and it is one of the key factors to govern the completeness of fuel combustion (Pognant et al. 2018). Excess air coefficient refers to the ratio of the actual air supply to the theoretical air supply for fuel combustion. Since excess air brings the heat from the combustion chamber to outside, high excess air coefficient would not only reduce the temperature of the furnace, but decrease the combustion efficiency and release more pollutants from incomplete combustion (Szyszlak-Bargłowicz et al. 2015;Geng et al. 2019). As shown in Table S5 in the SI, the oxygen contents in the flue gases of all five IBBs were greater than 13%. The excess air coefficient of IBBs ranged from 3.0 to 4.5. Figure 5 shows the relationship between operating load, excess air coefficient, and furnace temperature with the EFs of the five IBBs. As expected, the operating load and furnace temperature were negatively correlated with the EFs, while the excess air coefficient was positively correlated with the EFs. BB4 with the lowest operating load (35%) and furnace temperature (510 °C) and the highest excess air coefficient (4.5) had the highest EFs of 293.1 mg/kg, whereas BB3 with the highest operating load (78%) and furnace temperature (850 °C) and the lowest excess air coefficient (3.0) showed the lowest EFs of 19.6 mg/kg.
In general, when the boiler is in the low operating load, it would not only increase the heat loss of exhaust gas and decrease the boiler operation efficiency, but also increase the emissions of pollutants (Szyszlak-Bargłowicz et al. 2015). Due to the lower operating load of the boiler (BB4), excessive air entered the boiler furnace, elevating the excessive oxygen content in the flue gas and excessive air coefficient of the boiler, reducing the furnace temperature, and increasing VOC EFs. In addition, VOCs are released when furnace temperature rises above 200 °C, and many more VOCs can be emitted at about 500 °C in exothermic pyrolysis phases but they are further oxidized to CO 2 and H 2 O over 500 °C (Greenberg et al. 2006). The highest EFs of BB4 may be due to the furnace temperature not high enough to ignite the volatiles released during combustion, leading to more incomplete combustion of the biomass and less chance of organic compounds being oxidized into CO 2 and H 2 O. If excess air coefficient is too high, the products of incomplete combustion increase because of the low temperature (Quintero-Marquez et al. 2014). The higher VOC EFs of BB4 may attribute to higher excess air coefficient (4.5) bringing the heat from the combustion chamber to outside, reducing the temperature of the furnace and increasing VOC emissions.
Linear fitting was used between operating load, excess air coefficient, and EFs, whereas exponential fitting was adopted between furnace temperature and EFs . When the operating load increased by 10%, the excess air coefficient decreased by 1.0 and the EFs decreased by 64.0 mg/kg and 200.4 mg/kg, respectively. Therefore, in order to reduce VOC emissions from the IBBs, it is essential to control the appropriate operating load and excess air coefficient to increase the combustion efficiency. In the National Standard GB 13,271-2001 of China, the recommended value of excess air coefficient for the coal-fired and biomass-fired boilers is 1.8 (PRC MEE = .

Fig. 5
Regression between operating load, excess air coefficient, and furnace temperature of the IBBs with the VOC EFs 2001). Considering the air leakage from the flue and auxiliary boilers, the measured oxygen content at the end of flue may rise to some extent. According to the results from this study, it is suggested that the excess air coefficient should be limited below 3.5, roughly corresponding to operating load of greater than 62% and furnace temperature higher than 630 °C.

Fuel type
We have demonstrated that operating load, excess air, and furnace temperature are the leading factors governing VOC emissions in "Operating load, excess air coefficient, and furnace temperature" section. In this section, in order to examine the impact of fuel type on VOC emissions, we selected the wood pellet-fueled BB1 and the wood residuefueled BB4 for analysis. These two IBBs shared similar operating load, excess air, and furnace temperature, and had the same rated evaporation capacity of 10 t/h. The EFs and relative contributions of the VOC groups from BB1 and BB4 are shown in Fig. 6. The EF for BB1 and BB4 was 163.9 ± 18.4 mg/kg and 286.2 ± 10.8 mg/kg, respectively. The higher EFs of wood residue-fueled BB4 is due to its lower heating value, fixed carbon, and element carbon content (Table S2 in the SI). Similar results have been reported in the literatures. Sun et al. (2019) found that heating value, fixed carbon, and element carbon showed negative correlations with VOC EFs for domestic heating appliances. Yao et al. (2015) also found that the EFs of wood pellets were smaller than those of the direct firewood combustion.

Combustion phase
Apart from measuring VOC emissions in the stable combustion phase, we also carried out VOC sampling for BB1 in the ignition phase and the ember phase separately. In this section, we analyzed VOC emissions during different phases, and discussed the impacts of the combustion phase on VOC emissions. Figure 7 shows the VOC EFs during the ignition, stable combustion and ember phases. Obviously, VOC emissions were mostly concentrated in the ignition phase. The total EFs in the ignition phase was 2,272.0 mg/kg, about 14 times that of the stable combustion phase (163.9 mg/kg) and 5 times that of the ember phase (458.1 mg/kg). The dominant percentage of 75.8% in the ignition phase was similar to 60 − 75% of VOCs from the start phase in the log wood stove combustion (Bhattu et al. 2019). EFs of six groups (alkanes, alkenes, alkyne, aromatics, halocarbons, and OVOCs) show the similar trend as the total EFs in different phases; the main reason is that the EFs in the three combustion phases differ substantially. The EFs of alkanes, alkenes, alkyne, aromatics, and halocarbons are even in different orders of magnitudes. Compared to stable combustion phase, furnace temperature was much lower during ignition and ember phases, leading to non-ideal combustion conditions and less efficient combustion of volatilized organics (Fachinger et al. 2017). Incomplete combustion of the biomass is more likely to occur in the ignition and ember phases. This trend in emissions at different combustion phases was also reported in residential biomass combustion (Bhattu et al. 2019).

Assessment of ozone formation potential
In order to assess the OFP of VOC emissions, we firstly calculate VOC concentrations released from the five IBBs. Figure 8a shows the VOC concentrations from the five IBBs and their percentages from different VOC groups. The total VOC concentration ranged from 1.1 to 7.7 mg/m 3 for the five IBBs. VOC concentration seemed to be more fuel type-dependent that the wood residue-fueled IBBs had higher concentration than the wood pellet-fueled ones. The structure of wood pellets is relatively dense so that the volatiles on the surface are firstly released and those inside are released slowly. The slow release of the volatiles increases the completeness of combustion, leading to relatively lower VOC concentration (Li et al. 2011). Figure 8b shows the OFPs of VOC emissions from the five IBBs and their percentages from different VOC groups. The OFP ranged from 4.3 to 31.2 mg/m 3 for the five IBBs. The order of the five IBBs in terms of OFP (the shape of blue curve in Fig. 8a) was similar to that in terms of VOC concentration (the shape of red curve in Fig. 8b), indicating that VOC concentration was the leading factor for OFP and the impact of chemical composition was rather trivial. The relatively lower OFP for the wood pellet-fueled IBBs indicated that wood pellet is a relatively cleaner type of biomass fuel in comparison with wood residues. Wherever possible, replacement of the wood residues by the cleaner wood pellet is recommended to reduce VOC emissions and their potential to form ozone and other secondary pollutants.
OVOCs accounted for 56.3 − 80.7% of the total OFP, even greater than its mass contribution of 30.3 − 73.6%, further underscoring their importance in the IBB emissions. Lowmolecule OVOCs, especially formaldehyde, acetaldehyde, and acrolein, were consistently within the top 10 VOC species contributing to OFP for all IBBs, as shown in Fig. S4 in the SI. With a contribution of 50.4 − 69.7% to the VOC emissions, the top 10 species contributed to 83.3 − 89.9% of the OFP for the five IBBs. Therefore, reducing OVOC emissions should be prioritized in formulation of future control measures to mitigate emissions from the IBBs and their impacts on the atmospheric environment and human health.

Conclusions
In this study, five representative IBBs with two fuel types were selected to characterized VOC emissions and their impacting factors. The total VOC EFs ranged from 21.6 ± 2.8 mg/kg to 286.2 ± 10.8 mg/kg. OVOCs were the largest contributor to the total VOC emissions, accounting for 30.3 − 73.6% of the total VOCs. Although large differences existed in the VOC source profiles among the five IBBs, formaldehyde and acetaldehyde were consistently the most abundant species, accounting for 42.9 − 61.0% of the total OVOCs.
Operating load, excess air, and furnace temperature were the leading factors affecting VOC emissions. Operating load and furnace temperature were negatively, whereas excess air was positively correlated with the EFs. We suggested that the excess air coefficient should be limited below 3.5, roughly corresponding to the operating load of 62% and furnace temperature of 630 °C, to effectively reduce VOC emissions from the IBBs. Apart from the operating conditions, fuel type also had a significant impact on VOC emissions. The EF in the ignition phase was 14 times that in the stable combustion and 5 times that in the ember phase. The total OFP of five IBBs ranged from 4.3 to 31.2 mg/m 3 , and OVOCs contributed 56.3 − 80.7% to the total OFP, much greater than their contributions in VOC emissions. Wood residue-fueled IBBs yielded higher OFP than wood pelletfueled ones, indicating wood pellet is a relatively cleaner types of biomass fuel in comparison with wood residues.
This study highlighted the importance of OVOC emissions from IBBs. Reducing OVOC emissions should be prioritized in formulating control measures to mitigate their impacts on the atmospheric environment and human health. Although with great challenges in sample collection, it is suggested to carry out a more detailed study to give a more Fig. 8 a Total VOC concentrations (blue squares) and the contribution percentages from different VOC groups for the five IBBs and b total OFP (red dots) and the contribution percentages from different VOC groups for the five IBBs  Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.