Supplement information 1 Open burning of rice, corn and wheat straw: primary emissions, 2 photochemical aging, and secondary organic aerosol formation

Abstract. Agricultural residues are among the most abundant biomass burned globally, especially in China. However, there is little information on primary emissions and photochemical evolution of agricultural residue burning. In this study, indoor chamber experiments were conducted to investigate primary emissions from open burning of rice, corn and wheat straws and their photochemical aging as well. Emission factors of NOx, NH3, SO2, 67 non-methane hydrocarbons (NMHCs), particulate matter (PM), organic aerosol (OA) and black carbon (BC) under ambient dilution conditions were determined. Olefins accounted for  >  50 % of the total speciated NMHCs emission (2.47 to 5.04 g kg−1), indicating high ozone formation potential of straw burning emissions. Emission factors of PM (3.73 to 6.36 g kg−1) and primary organic carbon (POC, 2.05 to 4.11 gC kg−1), measured at dilution ratios of 1300 to 4000, were lower than those reported in previous studies at low dilution ratios, probably due to the evaporation of semi-volatile organic compounds under high dilution conditions. After photochemical aging with an OH exposure range of (1.97–4.97)  ×  1010 molecule cm−3 s in the chamber, large amounts of secondary organic aerosol (SOA) were produced with OA mass enhancement ratios (the mass ratio of total OA to primary OA) of 2.4–7.6. The 20 known precursors could only explain 5.0–27.3 % of the observed SOA mass, suggesting that the major precursors of SOA formed from open straw burning remain unidentified. Aerosol mass spectrometry (AMS) signaled that the aged OA contained less hydrocarbons but more oxygen- and nitrogen-containing compounds than primary OA, and carbon oxidation state (OSc) calculated with AMS resolved O ∕ C and H ∕ C ratios increased linearly (p

Z. Fang et al.: Open burning of rice, corn and wheat straws The increase in tropospheric ozone (O 3 ) in aged biomass burning plumes could last for days and even months (Thompson et al., 2001;Duncan et al., 2003;Real et al., 2007) with complex atmospheric chemistry (Arnold et al., 2015;Müller et al., 2016).Moreover, biomass and biofuel burning could contribute up to 70 % of the global secondary organic aerosol (SOA) burden (Shrivastava et al., 2015) and hence influence the seasonal variation of global SOA (Tsigaridis et al., 2014).Since it produces large amounts of primary and secondary pollutants, it is essential to characterize primary emissions and photochemical evolution of biomass burning in order to better understand its impacts on air quality (Huang et al., 2014), human health (Alves et al., 2015) and climate change (Andreae et al., 2004;Koren et al., 2004;Laskin et al., 2015;X. Huang et al., 2016).
Open burning of agricultural residues, a convenient and inexpensive way to prepare for the next crop planting, could induce severe regional haze events (Cheng et al., 2013;Tariq et al., 2016).Among all the biomass burning types, agricultural residue burning in the field is estimated to contribute ∼ 10 % of the total mass burned globally (Andreae and Merlet, 2001), and its relative contribution is even larger in Asia (∼ 34 %), and especially in China (> 60 %) (Streets et al., 2003), where > 600 million people live in the countryside (NBSPRC, 2015).Agricultural residues burned in China were estimated to be up to 160 million ton in 2012, accounting for ∼ 40 % of the global agricultural residues burned (J.Li et al., 2016).As estimated by Tian et al. (2011), agricultural residue burning contributed 70-80 % of non-methane hydrocarbons (NMHCs) and particulate matter (PM) emitted by biomass burning in China during 2000-2007.A better understanding of the role agricultural residual burning plays in air pollution in China and elsewhere requires better characterization of primary emission and atmospheric aging of emitted trace gases and particles for different types of agricultural residues under different burning conditions.
In the past 2 decades, there have been increasing numbers of characterizations of biomass burning emissions.Andreae and Merlet (2001) summarized emission factors (EFs) for both gaseous and particulate compounds from seven types of biomass burning.Akagi et al. (2011) updated the emission data for 14 types of biomass burning, and newly identified species were included.Since biomass types and combustion conditions may differ in different studies, reported emission factors are highly variable, especially for agricultural residue burning (Li et al., 2007(Li et al., , 2009(Li et al., , 2017;;Cao et al., 2008;Zhang et al., 2008;Yokelson et al., 2011;Brassard et al., 2014;Sanchis et al., 2014;Wang et al., 2014;Ni et al., 2015;Kim Oanh et al., 2015;Stockwell et al., 2016;Bruns et al., 2017;Tkacik et al., 2017).Moreover, previous studies on agricultural residue burning were mostly carried out near fire spots or in chambers with low dilution ratios.Since biomass burning organic aerosols (BBOAs) are typically semi-volatile (Grieshop et al., 2009b;May et al., 2013), it is expected that measured BBOA emission factors would be affected by dilution processes (Lipsky and Robinson, 2006), and BBOA emission factors under ambient dilution conditions are still unclear.Furthermore, knowledge on NMOGs emitted from agricultural residue burning is very limited.As reported by Stockwell et al. (2015), ∼ 21 % (in weight) of NMOGs in biomass burning plumes have not been identified yet.Therefore, comprehensive measurement and characterization of gaseous and particulate species emitted by agricultural residue burning under ambient dilution conditions are urgently needed.
Great attention has been drawn to SOA formation and transformation in biomass burning plumes recently, since a significant increase in mass and apparent change in physicochemical characteristics of aerosols have been observed during atmospheric aging of biomass burning plumes in both field and laboratory studies (Grieshop et al., 2009a, b;Hennigan et al., 2011;Heringa et al., 2011;Lambe et al., 2011;Jolleys et al., 2012;Giordano et al., 2013;Martin et al., 2013;Ortega et al., 2013;Ding et al., 2016aDing et al., , b, 2017)).For agricultural residue burning, evolution processes have not been well characterized yet.To our knowledge, up to now there has only been a chamber study (Li et al., 2015) which has investigated the evolution of aerosol particles emitted by wheat straw burning under dark conditions.Although field studies (Adler et al., 2011;X. X. Liu et al., 2016) witnessed the evolution in mass concentrations, size distribution, oxidation state and optical properties of aerosol particles emitted by agricultural residue burning, these changes could be also influenced by other emission sources and meteorological conditions as well.Since NMOGs emitted by agricultural residue burning are not fully quantified, it is still challenging to predict the concentration and physicochemical properties of SOA that resulted from biomass burning (Spracklen et al., 2011;Jathar et al., 2014;Shrivastava et al., 2015;Hatch et al., 2017).Bruns et al. (2016) suggested that the 22 major NMOGs identified in residential wood combustion could explain the majority of observed SOA, but it remains unclear whether identified NMOGs emitted by agricultural residue burning could fully (or at least largely) explain the SOA formed.In addition, aerosol mass spectrometry (AMS) has been widely used to characterize sources and evolution of ambient OA (Q.Zhang et al., 2011).Although agricultural residue burning is an important type of biomass burning in Asia and especially in China, the lack of AMS spectra for primary and aged OA from agricultural residue burning significantly limits further application of AMS in BBOA research.
In this study, plumes from agricultural residue open burning were directly introduced into a large indoor chamber to firstly characterize primary emissions and then investigate their photochemical evolution under ∼ 25 • C and ∼ 50 % relative humidity.Corn, rice and wheat straws, which account for more than 90 % of the crop residues burned in China (FAO, 2017), were chosen.A suite of advanced online and offline techniques were utilized to measure gaseous and particulate species, enabling comprehensive measurements of emission factors of gaseous and particulate compounds for burning of each type of straw under ambient dilution conditions.In addition, the corresponding formation and transformation of SOA during photochemical aging were investigated using a large indoor smog chamber.This work would help improve our understanding of primary emission, SOA formation and thus environmental impacts of agricultural residue burning.

Experimental setup
Photochemical aging was investigated in a smog chamber at the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIG-CAS).The GIG-CAS smog chamber is a ∼ 30 m 3 fluorinated ethylene propylene (FEP) reactor housed in a temperature-controlled room.Details of the chamber setup and associated facilities are provided elsewhere (Wang et al., 2014;Liu et al., 2015;Deng et al., 2017).Briefly, 135 black lamps (1.2 m long, 60 W Philips, Royal Dutch Philips Electronics Ltd, the Netherlands) are used as light sources, giving a NO 2 photolysis rate of approximately 0.25 min −1 .Two Teflon-coated fans are installed inside the reactor to ensure introduced gaseous and particulate species mixed well within 2 min.Prior to each experiment, the reactor was flushed with the purified dry air at a rate of 100 L min −1 for at least 48 h.The compressed indoor air is forced through an air dryer (FXe1; Atlas Copco; Sweden) and a series of gas scrubbers containing activated carbon, Purafil, Hopcalite and allochroic silica gel, followed by a PTFE filter to provide the source of the purified air.The purified dry air contains < 1 ppb NO x , O 3 and carbonyl compounds, < 5 ppb NMHCs and no detectable particles with relative humidity < 5 %.
Corn, rice and wheat straws were collected from Henan, Hunan and Guangdong provinces, respectively.Since moisture content in straws would affect emission factors of atmospheric pollutants (Sanchis et al., 2014;Ni et al., 2015), all the agricultural residues used in this study were dried in a stove at 80 • C for 24 h before being burned.After baking, the water content in the crop residues was less than 1 %.The water content of crop residues was measured by using the method recommended by Liao et al. (2004).Straws were weighed before and after baking in a stove at 105 • for 24 h, and the difference in weights was calculated to be the weight of the water in the crop residues.Water content was the quotient of the water weight and the whole weight of the straws.In each experiment, ∼ 300 g straws were burned and the burning typically lasted for 3-5 min.Straws were ignited by a butane-fueled lighter and burned under open field burning conditions.The resulting smoke was collected by an inverted funnel and introduced into the chamber using an oilfree pump (Gast Manufacturing, Inc, USA) at a flow rate of ∼ 15 L min −1 through a 5.5 m long copper tube (inner di-ameter: 3/8 inch), and the residence time in the tube was estimated to be < 2 s.Before each experiment, the transfer tube was pre-flushed for 15 min with ambient air and 2 min with smokes (not introduced into the chamber reactor).During the whole process, the tube was heated at 80 • C to reduce the losses of organic vapors.Based on the volumes of the smoke introduced and the chamber reactor, the dilution ratios were estimated to be 1300-4000, falling into the typical range (1000-10000) under ambient dilution conditions (Robinson et al., 2007).After being characterized in the dark for > 20 min, black lamps were turned on and the diluted smokes were photochemically aged for 5 h.At the end, the black lamps were switched off and the aged aerosols were characterized in the next 1 h to determine the particle wall loss.The particle size evolved through the course of photooxidation, and the differences in particle wall-loss rates during photoreaction and after the lamps were off brought about by the size evolving are estimated to be within ±9 % (Fig. S1 in the Supplement).
In total 20 experiments were conducted (9 for rice straw, 6 for corn straw and 5 for wheat straw), among which 14 experiments were conducted only in the dark to measure primary emissions and 6 experiments were carried out both in the dark and under irradiation to investigate photochemical evolution of open straw burning emissions.Tables 1 and 2 summarize important experimental conditions and key results for all the experiments.

Instrumentation
Commercial instruments were used for online monitoring of NO x (EC9841T, Ecotech, Australia), NH 3 (Model 911-0016, Los Gatos Research, USA) and SO 2 (Model 43i, Thermo Scientific, USA).CH 4 and CO were analyzed offline using gas chromatography (Agilent 6980GC, USA) coupled with a flame ionization detector and a packed column (5A molecular sieve 60/80 mesh, 3 m × 1/8 inch) (Zhang et al., 2012), and CO 2 was analyzed using a HP 4890D gas chromatograph (Yi et al., 2007).The detection limits were all less than 30 ppbv for CH 4 , CO and CO 2 .The relative standard deviations (RSDs) of CO and CO 2 measurements were both less than 3 % based on seven duplicate injection of 1.0 ppmv standards (Spectra Gases Inc, USA).Volatile organic compounds (VOCs) were continuously measured using a proton-transferreaction time-of-flight mass spectrometer (PTR-TOF-MS; Model 2000, Ionicon Analytik GmbH, Austria).Calibration of the PTR-TOF-MS was performed every few weeks using a certified custom-made standard mixture of VOCs (Ionicon Analytik Gmbh, Austria) that were dynamically diluted to six levels (2, 5, 10, 20, 50 and 100 ppbv).Methanol, acetonitrile, acetaldehyde, acrolein, acetone, isoprene, crotonaldehyde, 2-butanone, benzene, toluene, o-xylene, chlorobenzene and α-pinene were included in the calibration mixture.Their sensitivities, indicated by the ratio of the normalized counts per second to the concentration levels of the VOCs Z. Fang et al.: Open burning of rice, corn and wheat straws in ppbv, were used to convert the raw PTR-TOF-MS signal to concentration (Z.Huang et al., 2016).Quantification of the compounds that were not included in the mixture was performed by using calculated mass-dependent sensitivities based on the measured sensitivities (Stockwell et al., 2015).Mass-dependent sensitivities were linearly fitted for oxygencontaining compounds and the remaining compounds separately.The decay of toluene measured by PTR-TOF-MS was used to derive the OH radical concentrations for every 2 min during each experiment, and the OH exposure was calculated as the product of the OH concentration and the time interval.Continuous monitoring of 20 SOA precursors (including 9 NMHCs and 11 oxygen-containing VOCs) from PTR-TOF-MS provided us with data to do the SOA prediction discussed in the Sects.2.3.5 and 3.3.2.Air samples were also collected from the chamber reactor using 2 L electropolished stainless-steel canisters before and after smoke injection.In total 67 C 2 -C 12 NMHCs were measured (Table S1 in the Supplement) using an Agilent 5973N gas chromatography mass-selective detector/flame ionization detector (GC-MSD/FID; Agilent Technologies, USA) coupled to a Preconcentrator (Model 7100, Entech Instruments Inc., USA), and analytical procedures have been detailed elsewhere (Wang and Wu, 2008;Zhang et al., 2010Zhang et al., , 2012)).Results from GC-MSD/FID were used to quantify the emission factors of 67 NMHCs discussed in the Sect.3.1.
Particle number/volume concentrations and size distribution were measured with a scanning mobility particle sizer (SMPS; Classifier model 3080, CPC model 3775, TSI Incorporated, USA).The SMPS was operated with a sheath flow of 3.0 L min −1 and a sampling flow of 0.3 L min −1 , allowing for a size scanning range of 14 to 760 nm within 255 s.A high-resolution time-of-flight aerosol mass spectrometer (HR-TOF-AMS; Aerodyne Research Incorporated, USA) was used to measure chemical compositions of nonrefractory aerosols (DeCarlo et al., 2006).The HR-ToF-AMS was operated by alternating every other minute between the high sensitivity V mode and the high-resolution W mode. Toolkit Squirrel 1.57I was used to obtain real-time concentration variations of sulfate, nitrate, ammonium, chloride and organics, and toolkit Pika 1.16I was used to determine the detailed compositions of OA (Aiken et al., 2007(Aiken et al., , 2008;;Canagaratna et al., 2015).The AMS signal at m/z 44 was corrected for the contribution from gaseous CO 2 .The ionization efficiency of the AMS was calibrated routinely by measuring 300 nm monodisperse ammonium nitrate aerosols.Considering the underestimation of particulate matter by the AMS, aerosol mass measured by AMS was corrected with the data from the SMPS and the aethalometer.Conductive silicon tubes were used for aerosol sampling to reduce electrostatic losses of particles.
BC was measured with a seven-channel aethalometer (Model AE-31, Magee Scientific, USA).Cheng et al. (2016) measured the mass absorption efficiency (MAE) of BC from biomass burning at wavelengths of 532 and 1047 nm, respec-tively, and the absorption Ängström exponents (AAEs) were estimated to be in the range of 0.9-1.1.Based on the relationship between MAE and wavelength, a MAE value of 4.7 m 2 g −1 was calculated for 880 nm by assuming the AAE to be 1.0.The MAE value was then applied to convert absorption data in 880 nm to BC mass concentrations.Aethalometer attenuation measurements were corrected for particle loading effects and the scattering of filter fibers using the method developed by Kirchstetter and Novakov (2007) and Schmid et al. (2006).

Particle effective density
Assuming that particles are spherical and non-porous, the effective density (ρ eff ) can be estimated by Eq. ( 1) (DeCarlo et al., 2004;Schmid et al., 2007): where ρ 0 is the standard density (1.0 g cm −3 ), and d va and d m are the AMS-measured vacuum aerodynamic diameter and SMPS-measured mobility diameter.The input diameters to this equation were determined by comparing distributions of vacuum aerodynamic and electric mobility diameters, using the AMS and SMPS, respectively.Derived ρ eff was used to convert volume concentrations of aerosol particles measured by the SMPS to mass concentrations.

Emission factors and modified combustion efficiency
The carbon mass balance approach (Ward et al., 1992;Andreae and Merlet, 2001) was used to calculate fuel-based emission factors (EFs) for each compound (g kg −1 dry fuel).
The emission factor for the ith species, EF i , is calculated by Eq. ( 2): where m i is the concentration (g m −3 ) of the ith species; [CO 2 ], [CO], and [HC] are the background-corrected carbon mass concentration (g C m −3 ) of the CO 2 , CO, and speciated hydrocarbons, respectively; [PM C ] is the background-corrected carbon in the particle phase (g C m −3 ); and EF C is the emission factor of carbon into the air determined by elemental and gravitational analyses, given by Eq. (3): where ω fuel and ω ash are mass fractions of carbon in the dry fuel and its ash, and m fuel and m ash are the mass of dry fuel and its ash.The modified combustion efficiency (MCE) is defined by Eq. ( 4) (Heringa et al., 2011;Hennigan et al., 2011;Ni et al., 2015): . (4)

Ozone formation potential
The ozone formation potential (OFP) of the speciated NMHCs was calculated from the emission factor and maximum incremental reactivity (MIR) of each individual NMHC, using Eq. ( 5): where OFP is the ozone formation potential of NMHCs emitted per unit of biomass (unit: g kg −1 ), and MIR i is the MIR of the ith NMHC (unit: g O 3 g NHMC −1 ) (Carter, 2008).

Wall-loss corrections
Due to the loss of particles and vapors to chamber walls, measured data in chamber studies need to be corrected for wall loss.For this purpose, in our study 1 h dark decay of aged aerosols was undertaken after photochemical aging was terminated.The loss of particles on the chamber wall is a first-order process (McMurry and Grosjean, 1985).The wallloss rates of AMS-measured organics, sulfate, nitrate, chloride and ammonium were determined using the dark decay data and were applied to wall-loss correction for the entire experiment.By assuming that the condensed materials on the wall remain completely in equilibrium with the gas phase, we used the ω = 1 case to correct the OA mass, where ω is a proportionality factor of organic vapor partitioning to chamber walls and suspended particles (Weitkamp et al., 2007;Henry et al., 2012).For SMPS measurements, the number concentration in each size channel (110 channels in total) was corrected for wall loss separately, since wall-loss rates of aerosol particles are size-dependent (Takekawa et al., 2003).for open burning of rice, corn and wheat straw.
Assuming that primary OA (POA) levels stayed constant during aging processes, the mass concentration of SOA formed could be estimated as the difference in OA mass concentrations before and after photochemical aging.It should be noted that POA would decrease during aging processes (Tiitta et al., 2016), probably leading to the underestimation of the formed SOA.In papers where those SOA yields were borrowed from, no organic vapor wall loss was accounted for when calculating the mass concentration of the formed SOA, so the same wall-loss correction method was used when comparing the predicted SOA and the formed SOA.

Emissions of gaseous pollutants
Table 1 compares emission factors of gaseous and particulate species measured in our and previous studies.In our study, the emission factors of NO x were 1.47 ± 0.61, 5.00 ± 3.94, and 3.08 ± 0.93 g kg −1 for rice, corn, and wheat straw, and NO accounted for 84 ± 11 % of NO x primary emission for all experiments.Emission factors of NH 3 were measured to be 0.45 ± 0.15, 0.63 ± 0.30 and 0.22 ± 0.19 g kg −1 for rice, corn and wheat straw.Our measured emission factors of reactive nitrogen species were comparable to those reported by previous studies (Li et al., 2007;Tian et al., 2011).Emission factors of SO 2 were 0.07 ± 0.07, 0.99 ± 1.53 and 0.72 ± 0.34 g kg −1 for rice, corn and wheat straw.Our measured emission factors of SO 2 were lower than those reported by Cao et al. (2008) and Kim Oanh et al. ( 2015) for rice straw, but higher than those reported by Cao et al. (2008) for corn and wheat straw.Due to low sulfur contents in crop straws, the SO 2 emission factors for open burning of crop residues were much lower than those for domestic coal combustion, which were determined to be 2.43-5.36g kg −1 for raw bituminous coal (Du et al., 2016).S1, in which a comprehensive dataset of emission factors measured in this work is included.
Emission factors of the total speciated NMHCs analyzed by the GC-MSD/FID system were 5.04 ± 2.04, 2.47 ± 2.11 and 3.08 ± 2.43 g kg −1 for rice, corn and wheat straw, respectively (Table 1).Our results were higher than those reported by previous studies (Li et al., 2009;Wang et al., 2014), partly due to the fact that more NMHCs were analyzed in our study (67 species in total).As shown in Fig. 1a-c, olefins and acetylene accounted for 56-58 % of the total speciated NMHCs, followed by alkanes (22-28 %) and aromatic hydrocarbons (16-21 %).Table S1 and Fig. 2 show the emission factors of each NMHC for open burning of different straws.Emission factors of unsaturated hydrocarbons ranged from 1.37 (corn) to 2.91 g kg −1 (rice), with the majority being ethene, acetylene and propene.Emission factors of alkanes ranged from 0.69 (corn) to 1.09 g kg −1 (rice), with ethane and propane being the two most abundant compounds.The emission factors of aromatic hydrocarbons were in the range of 0.42 (corn) to 1.04 (rice), and benzene and toluene are dominant species.It is worth noting that major compounds in the three groups (alkanes, alkenes and aromatic hydrocarbons) were all negatively correlated with the modified combustion efficiency (Fig. S2), suggesting that more efficient combustion would reduce their emissions.
Based on their emission factors, we calculated the ozone formation potential for each NMHC.The summed ozone formation potentials were 22.5 ± 10.1, 13.7 ± 12.4 and 16.3 ± 13.5 g kg −1 for open burning of rice, corn and wheat straw, respectively.As shown in Fig. 1d-e, the relative contri-butions of olefins to the total ozone formation potential could reach > 80 %.Ethene was the largest ozone precursor (35-42 %), followed by propene (16-28 %), and these two compounds contributed 58-64 % of the total ozone formation potential.Although the emission factors of aromatic hydrocarbons were lower than those of alkanes, their ozone formation potential was dominant over those of alkanes, with toluene being the largest contributor among all the aromatic hydrocarbons.The contribution of alkanes to the total ozone formation potential was minor (2-3 %).It is noted that oxygencontaining organic vapors in agricultural residue burning plumes could also have large ozone formation potentials.For example, the OFPs of formaldehyde and acetaldehyde for all experiments were 0.57-2.46times those of the 67 speciated NMHCs.

Emission of particulate matter
The emission factors of particulate matter were 3.73 ± 3.28, 5.44 ± 3.43 and 6.36 ± 2.98 g kg −1 for rice, corn and wheat straw, lower than those reported in the previous studies (Table 1).As suggested by Robinson et al. (2007), the POA emission factors would decrease with increasing dilution ratios, due to evaporation of semi-volatile organic compounds.In this study, the dilution ratios ranged from 1300 to 4000, which were within the typical range of ambient dilution ratios (1000-10 000) (Robinson et al., 2007).Therefore, it can be expected that emission factors of primary organic car-bon (POC) measured in our study (2.05-4.11gC kg −1 ) were lower than those measured by previous work with dilution ratios of 5-20 (Li et al., 2007;Ni et al., 2015).Moreover, it has been shown that the modified combustion efficiency could affect emission factors (Heringa et al., 2011;Stockwell et al., 2015).Figure S3 shows negative correlations of the modified combustion efficiency with emission factors of PM and POC (p < 0.05 for both cases), indicating that enhancement of combustion efficiency could reduce the emissions of PM and POC.In our study, all straws were pre-baked to reduce the moisture content to < 1 %, and this treatment could increase the modified combustion efficiency and thus reduce emission factors of particulate matters (Ni et al., 2015).In addition, the amount of straw burned each time in our experiments was much less than that in the fields, which is expected to avoid oxygen deficit during burning to some extent and thus increase the modified combustion efficiency as well.
While POA emission factors showed large variability for different types of straw, BC emission factors were relatively constant (0.22-0.27 gC kg −1 ).Since BC is a mixture of non-volatile compounds in particulate matters, as expected, its emission factors measured in our work were comparable to those reported under lower dilution conditions (Li et al., 2007;Ni et al., 2015).The [POA] / [CO] ratios ranged from 0.022 to 0.133 in our study, larger than those (0.001-0.067) measured in chamber studies for hard-and soft-wood fires (Grieshop et al., 2009b) and vegetation commonly burned in North American wildfires (Heringa et al., 2011), but lower than those (0.051-0.329) obtained in field campaigns (Jolleys et al., 2012).

Growth of particle size
Figure 3 shows the evolution of particle size distribution after photochemical aging of 0, 0.5, 2.5 and 5 h.Aerosol particles emitted from open straw burning peaked at 50-90 nm under ambient dilution conditions.The geometric mean diameters for primarily emitted particles in this study were smaller than those (100-150 nm) reported for crop residual burning under low dilution conditions (H.Zhang et al., 2011;Li et al., 2015), probably due to evaporation of organic vapors under the high dilution conditions (Lipsky et al., 2006) and coagulation of fine particles under the low dilution conditions (Hossain et al., 2012).After switching on black lamps, apparent growth of particle size was observed.In all the aging experiments, growth rates of particle diameters in the first 0.5 h were 10 times larger than those afterwards, and after 5 h aging the geometric mean diameters peaked at 60-120 nm.For instance, in the photochemical aging experiment for wheat straw burning (Fig. 3c), the growth rate of particles was 18 nm h −1 in the first 0.5 h and decreased to ∼ 1 nm h −1 during the following 4.5 h.The size distribution of aged aerosol particles in our study is similar to those of ambient particles under the severe biomass burning impact during haze events (Betha et al., 2014;Niu et al., 2016).

Particle mass enhancement
Figure 4 shows the chemical evolution of aerosol particles during the 5 h photochemical aging of wheat straw burning.During the whole process, OA kept increasing and was dominant over inorganic species.After 3 h of photochemical aging, the levels of all the inorganic species were constant, and nitrate was the second most abundant component, with a mass fraction of 7 %, followed by chloride (2 %), ammonium (1 %) and sulfate (< 1 %). Figure 4b depicts [OA] evolution as a function of OH exposure.OA increased slowly in the first ∼ 0.2 h, and then increased rapidly with OH exposure.
The OA enhancement ratio, defined as the mass ratio of aged OA at the end of each aging experiment to POA, was calculated.In the six aging experiments, the OH exposure and OA enhancement ratios ranged from (1.87-4.97)× 10 10 molecule cm −3 s to 2.4-7.6,respectively.Assuming an average OH concentration of 1.5 × 10 6 molecule cm −3 in the ambient air (Hayes et al., 2013), this means that rapid SOA formation would occur in 3.5-9.2h during the daytime after straw burning.The OA enhancement ratios determined in our study were higher than those (0.7-2.9) for the combustion of vegetation commonly burned in North American wildfires (Hennigan et al., 2011), and comparable to those (0.7-6.9) for wood burning (Grieshop et al., 2009b;Heringa et al., 2011).
Recently, Bruns et al. (2016) found that 22 NMOGs emitted from residential wood burning could explain the majority of the formed SOA.In our study, 20 of the 22 NMOGs were detected and quantified with the PTR-TOF-MS.Concentration differences of each compound before and after photo-oxidation were calculated to estimate the SOA formed from these precursors.Since SOA formation highly depends on oxidation conditions, SOA yields for a certain precursor vary with VOC / NO x ratios.In our work, we chose a set of SOA yields for these NMOGs based on the observed VOC / NO x ratio in the chamber experiments.More specifically, if the observed VOC / NO x ratio for a certain precursor in the chamber was within the VOC / NO x range reported in the literature, the mean value of the highest and lowest yields within the VOC / NO x range in the literature was used to estimate the SOA formed from the precursor in the chamber; if the observed VOC / NO x ratio for a certain precursor was higher than the maximum VOC / NO x ratio reported in the literature, we chose the yield reported at the maximum VOC / NO x ratio; if the observed VOC / NO x ratio was lower than the minimum VOC / NO x ratio reported in the literature, we chose the yield reported at the minimum VOC / NO x ratio.
Figure 5a shows the time series of POA, SOA predicted and unexplained SOA in a typical aging experiment.The contribution of SOA predicted by the 20 NMOGs was minor, and large fractions of observed SOA could not be explained.In all the experiments, only 5.0-27.3% of the observed SOA mass could be explained by the 20 NMOGs (Fig. 5b).Even if the highest SOA yield for each precursor reported in the literature were used, 60-90 % of the observed SOA mass could still not be explained.It has been suggested that aqueousphase oxidation of alkenes could produce substantial SOA (Ervens et al., 2011).Considering large emissions of olefins from straw burning (Fig. 1a-c), we also estimated the SOA formed from the three most abundant alkenes (ethene, acetylene, and propene) with their newly developed SOA yields (Ge et al., 2016(Ge et al., , 2017;;Jia and Xu, 2016), and their total contribution to the observed SOA was found to be negligible (< 0.5 %).It is noted that although over 80 VOCs species were quantified by the GC-MSD/FID and the PTR-TOF-MS in this study, only 20 species among them were taken into the SOA prediction because of the lack of published data for SOA yields.The unaccounted VOC species might be a reason for the discrepancy.On the other hand, as indicated by Deng et al. (2017), SOA yields obtained from chamber studies in purified air matrix might be lower than that in a real ambient air matrix.Consequently, using SOA yields from studies in a purified air matrix might also underpredict SOA yields in the complex biomass burning plume matrix.Moreover, oxidation of particulate organic matter (POM), like semi-volatile organic compounds (SVOCs) and intermediate volatility organic compounds (IVOCs), would also contribute substantially to SOA formation (Presto et al., 2009;Zhao et al., 2014), yet this is not accounted for in our prediction.Above all, there are still unknown precursors and/or physicochemical processes contributing the majority of SOA formed from open straw burning.

OA mass spectrum evolution
In the high-resolution W mode of AMS, ions generated from particles could be identified by their exact mass-charge ratio (m/z) and then grouped into the CHON, CHO, CHN and CH families.Figure 6 presents the evolution of OA mass spec- tra.For POA (Fig. 6a), the CH-family was the major component, with a mass fraction of 68 %, followed by CHO (23 %), CHN (6 %), and CHON (2 %).The ions at m/z 43, 41 and 55 were the dominant peaks in the POA mass spectrum.The major ions at m/z 27, 39, 41, 55, 57, 67 and 69 belonged to the CH-family and could be the fragments of hydrocarbons (Weimer et al., 2008).The peaks at m/z 28, 29, 43, 44 and 55 contained considerable CHO ions, and the corresponding ions (CO + , CHO + , C 2 H 3 O + , CO + 2 and C 3 H 3 O + ) could be the fragments of aldehydes, ketones and carboxylic acid (Ng et al., 2011a).The peak at m/z 91 was mainly attributed to C 7 H + 7 , possibly originating from aromatic compounds.The mass spectra of aged OA were quite different from those of POA (Fig. 6b-c).The mass fraction of the CHfamily decreased to 46 % and was comparable to that of the CHO-family, while the contribution of N-containing OA (CHN and CHON) increased to ∼ 11 %.The ions at m/z 44 and 43, mainly coming from the CHO-family, became the dominant peaks for the aged OA.The fractions of two major masses at m/z 44 (f 44 ) and m/z 43 (f 43 ) in OA can be used to generate an f 44 vs. f 43 triangular space, in which oxygenated organic aerosol (OOA) moves towards the apex  (Ng et al., 2011b) and the ambient OOA range (Ng et al., 2010).The typical f 44 ranges of ambient SV-OOA and LV-OOA are indicated with the vertical arrows.(b) Comparison of f 44 vs. f 60 for straw burning OA with those for other types of biomass burning OA (Alfarra et al., 2007;Hennigan et al., 2011;Cubison et al., 2011;Brito et al., 2014;May et al., 2015).
during the aging process (Ng et al., 2010).In addition, f 44 in the ambient air was suggested to be 0.07 ± 0.04 for semivolatile OOA (SV-OOA) and 0.17 ± 0.04 for low-volatility OOA (LV-OOA), respectively (Ng et al., 2010).Figure 7a plots f 44 and f 43 of the POA and the aged OA in all six experiments.Most of the data are within the f 44 vs. f 43 triangular space and close to the left margin.Photochemical aging led to an increase in f 44 for all the experiments, suggesting transformation of OA from SV-OOA to LV-OOA.For comparison, the f 43 did not change significantly in all the experiments.The main ions at m/z 43 were C 2 H 3 O + and C 3 H + 7 .It can be observed in Fig. 6c that the increased contribution of C 2 H 3 O + and the decreased contribution of C 3 H + 7 were comparable during photoreaction.
The ion at m/z 60, mainly consisting of C 2 H 4 O + 2 , is regarded as a BBOA marker, and the mass fraction of this ion in OA, f 60 , is widely used to probe the evolution of BBOA (Brito et al., 2014;May et al., 2015).Figure 7b plots the evolution of f 44 and f 60 in all the experiments conducted in this study, in order to compare them with measurements in aging biomass burning plumes (Cubison et al., 2011) and those in the POA from different types of biomass burning (Alfarra et al., 2007;Brito et al., 2014;May et al., 2015).Photooxidation caused an increase in f 44 and a decrease in f 60 , and this is consistent with the general evolution of OA in ambient biomass burning plumes (Cubison et al., 2011).However, our measured f 60 , 0.003-0.006 in the POA from open straw burning and 0.002-0004 in aged OA, were all lower than those from other field campaigns and quite near the background f 60 level of 0.003 for ambient OA (Cubison et al., 2011;Fig. 7b).Low values of f 60 (0.005-0.02) were also reported by Hennigan et al. (2011) in a chamber study for fuels commonly burned in wildfires.In their study, biomass burning took place in a 3000 m 3 combustion chamber, and the smokes were then injected into another chamber for aging experiments with a dilution ratio of ∼ 25.Previous studies have demonstrated that levoglucosan is a semi-volatile compound with a saturation concentration of ∼ 8 µg m −3 at 293 K (Grieshop et al., 2009b;Huffman et al., 2009;Hennigan et al., 2011).As a result, the high dilution conditions used in our study would cause levoglucosan to evaporate, and this may at least partly explain the low f 60 observed in the POA from straw burning.From previous studies, the levoglucosan / OC ratios of straw burning ranging from 4.92 to 16.8 % (4 types of vegetation summarized; Dhammapala et al., 2007;Kim Oanh et al., 2011;Hall et al., 2012) were not significantly (two-sample t-test, p > 0.05) lower than those of prescribed fuel burning, wildfire and wood burning ranging from 1.46 to 13.5 % (20 types of vegetation summarized; Hosseini et al., 2013;Shahid et al., 2015).So the difference in fuel type cannot explain the lower f 60 observed in our study.

Elemental ratio and oxidation state of OA
In this study, the O / C and H / C ratios in the POA from different straws burning were in the ranges of 0.20-0.38 and 1.58-1.74,respectively.After 5 h aging, O / C increased and H/C decreased (Table 2).Kroll et al. (2011) proposed a metric, the average carbon oxidation state (OS c ), to describe the degree of oxidation of atmospheric organic species.OS c could be calculated from the elemental composition of OA measured by AMS, given by Eq. ( 7): In this study, the OS c values for the fresh POA from open straw burning ranged from −1.25 to −0.89, consistent with those suggested for BBOA (−1 to −0.7) (Kroll et al., 2011).During photochemical aging, the OS c values increased linearly (p < 0.001) with OH exposure (Fig. 8), and the slopes were quite near each other even for different types of straws,   implying AMS measured OS c might be a good indicator OH exposure and thereby of photochemical aging.
Figure 9 shows the Van Krevelen diagram of OA.In this study, the slopes of linear correlations between H / C and O / C range from −0.49 to −0.24 for the five experiments.Slopes of −1, 0.5 and 0 in the Van Krevelen diagrams indicate addition of carboxylic acids without fragmentation, addition of carboxylic acids with fragmentation, and addition of alcohols/peroxides, respectively (Heald et al., 2010;Ng et al., 2011a).Therefore, the slopes determined in our study suggest that open straw burning OA aging resulted in net changes in chemical composition equivalent to addition of carboxylic acid groups with C-C bond breakage and addition of alcohol/peroxide functional groups.

Conclusion
In this study, primary emissions of open burning of rice, corn and wheat straw and their photochemistry were investigated using a large indoor chamber.Emission factors of NO x , NH 3 , SO 2 , 67 NMHCs, PM and particle number were measured under dilution ratios ranging from 1300 to 4000.Emission factors of PM (3.73-6.36g kg −1 ) and POC (2.05-4.11gC kg −1 ) were lower than those reported in previous studies conducted at lower dilution ratios, probably due to the evaporation of semi-volatile organic compounds.Emission factors of POC, PM and major NMHC compounds were all negatively correlated with the modified combustion efficiency, suggesting that incomplete burning of agricultural residues could lead to larger primary emission.
Both agricultural residue burning and domestic coal combustion have been recognized as contributing substantially to the deteriorating regional air quality, especially in rural areas of China (Pan et al., 2015;J. Liu et al., 2016;Zhu et al., 2016).The emission factors of the speciated NMHCs, PM, NO x , CO and SO 2 from combustion of raw bituminous coal, which is currently prevalent for cooking and heating in rural areas, have been reported to be 0.56-5.40,25.49 ± 2.30, 0.97 ± 0.03, 208 ± 5 and 2.43-5.36g kg −1 , respectively (Du et al., 2016;Li et al., 2016;Liu et al., 2017).Annually burned crop residues and domestic coals were estimated to be 160 Tg (Q.Li et al., 2016) and99.6 Tg (NBSPRC, 2014) in China.Therefore, with the emission factors of the speciated NMHCs (2.47-5.04g kg −1 ), PM (3.73-6.36g kg −1 ), NO x (1.47-5.00g kg −1 ), CO (46.1-63.5 g kg −1 ) and SO 2 (0.07-0.99 g kg −1 ) measured for agricultural residue burning in this study, agricultural residue burning might emit more NMHCs and NO x but less primary PM, CO and SO 2 than domestic coal burning on a national scale.
Photochemical aging of primary emissions was investigated with OH exposure equal to 3.2-9.2h under typical ambient conditions, and at the end of experiments the OA mass concentrations increased by a factor of 2.4-7.6,suggesting that SOA could be rapidly produced within several hours.Our estimation suggests that phenols are the most important identified SOA precursors, and more than 70 % of the formed OA still cannot be explained by the oxidation of known precursors.Measurements using HR-TOF-AMS reveal that after photochemical aging, signals for oxygen-and nitrogencontaining compounds were largely increased, with OS c increased in a highly significant linear way with OH exposure.

Figure 1 .
Figure 1.(a-c) Non-methane hydrocarbon (NMHC) compositions and (d-f) their relative contribution to ozone formation potential (OFP) for open burning of rice, corn and wheat straw.

Figure 2 .
Figure 2. Emission factors (EFs) of NMHCs for straw burning of rice, corn and wheat.Only species with emission factors > 0.01 g kg −1 are shown.The order of NMHC species is the same as TableS1, in which a comprehensive dataset of emission factors measured in this work is included.

Figure 4 .
Figure 4. (a) The evolution of particulate matter components (Burn 2).(b) OA mass growth as a function of OH exposure (Burn 5).

Figure 5 .
Figure 5. (a) Time series plots of concentrations of POA, secondary organic aerosol that can be explained by the reacted precursors (SOA predicted ), the difference between the formed SOA and the predicted SOA (unexplained SOA) in Burn 6.(b) Contribution of 20 NMOGs to the formed SOA at the end of photoreactions.Error bars correspond to the range of contributions when the lowest/highest SOA yields in references were used for all precursors.

Figure 6 .
Figure 6.(a) Mass spectrum of POA; (b) mass spectrum of aged OA; (c) difference in mass spectra between aged OA and POA.The data were all taken from Burn 5.

Figure 7 .
Figure 7. (a) Comparison of f 44 vs. f 43 determined in our work with those for the ambient BBOA data sets(Ng et al., 2011b) and the ambient OOA range(Ng et al., 2010).The typical f 44 ranges of ambient SV-OOA and LV-OOA are indicated with the vertical arrows.(b) Comparison of f 44 vs. f 60 for straw burning OA with those for other types of biomass burning OA(Alfarra et al., 2007;Hennigan et al., 2011;Cubison et al., 2011;Brito et al., 2014;May et al., 2015).

Figure 8 .
Figure8.The growth of the OA carbon oxidation state with OH exposure for burning corn (Burns 3 and 4) and wheat (Burns 5 and 6) straws.Data for burning rice straws were not included since in Burn 1 AMS was then not run in W-mode.

Figure 9 .
Figure 9. Van Krevelen diagram for the OA.Each slope corresponds to the addition of a specific functional group to an aliphatic carbon.

Table 1 .
Primary emission factors measured for agricultural residue burning.All the units are g kg −1 , except that the unit for particle number (PN) is 10 15 particle kg −1 .MCE: modified combustion efficiency; NMHCs: non-methane hydrocarbons; POA: primary organic aerosol; POC primary organic carbon; BC: black carbon.

Table 2 .
Overview of important experimental conditions and key results in the photochemical oxidation experiments.The unit for OH exposure is 10 10 molecule cm −3 s.NA: data were not available because no data were recorded in the W-mode.