Changes in PM2.5 Peat Combustion Source Profiles with Atmospheric Aging in an Oxidation Flow Reactor

Abstract. Smoke from laboratory chamber burning of peat fuels from Russia, Siberia, U.S.A. (Alaska and Florida), and Malaysia representing boreal, temperate, subtropical, and tropical regions was sampled before and after passing through a potential aerosol mass-oxidation flow reactor (PAM-OFR) to simulate ∼2- and 7-day atmospheric aging. Species abundances in PM2.5 between aged and fresh profiles varied by >5 orders of magnitude with two distinguishable clusters: around 0.1 % for reactive and ionic species and mostly >10 % for carbon. Organic carbon (OC) accounted for 58–85 % of PM2.5 mass in fresh profiles with low EC abundance (0.67–4.4 %). After a 7-day aging time, degradation was 20–33 % for OC, with apparent reductions (4–12 %) in low temperature OC1 and OC2 (thermally evolved at 140 and 280 °C), implying evaporation of higher vapor pressure semi-volatile organic compounds (SVOCs). Additional losses of OC from 2- to 7-days aging is somewhat offset by the formation of oxygenated organic compounds, as evidenced by the 12–19 % increase in organic mass (OM) to OC ratios. However, the reduction of OM abundances in PM2.5 by 3–18 % after 7 days, reconfirms that volatilization is the main loss mechanism of SVOCs. Although the ammonia (NH3) to PM2.5 ratio rapidly diminished with a 2-day aging time, it represents an intermediate profile – not sufficient for completed OC evaporation, levoglucosan degradation, organic acid oxidation, or secondary inorganic aerosol formation. Week-long aging resulted in an increase to ∼7–8 % of NH4+ and NO3− abundances, but with enhanced degradation of NH3, low temperature OC, and levoglucosan for Siberia, Alaska, and Everglasdes (FL) peats. Elevated levoglucosan was found for Russian peats, accounting for 35–39 % and 20–25 % of PM2.5 mass for fresh and aged profiles, respectively. Abundances of water-soluble organic carbon (WSOC) in PM2.5 was >2-fold higher in fresh Russian (37.0 ± 2.7 %) than Malaysian (14.6 ± 0.9 %) peats. While Russian peat OC emissions are largely water-soluble, Malaysian peat emissions are mostly water-insoluble, with WSOC/OC ratios of 0.59–0.71 and 0.18–0.40, respectively. Source profiles can change with aging during transport from source to receptor. This study shows significant differences between fresh and aged peat combustion profiles among the four biomes that can be used to establish speciated emission inventories for atmospheric modeling and receptor model source apportionment. A sufficient aging time (∼one week) is needed to allow gas-to-particle partitioning of semi-volatilized species, gas-phase oxidation, and particle volatilization to achieve representative source profiles for regional-scale source apportionment.


J. C. Chow et al.: Peat combustion source profiles These patterns, termed "source profiles," have been measured in diluted exhaust emissions and resuspended mineral dusts for a variety of representative emitters. Many of these source profiles are compiled in country-specific source profile data bases (Cao, 2018;CARB, 2019;Liu et al., 2017;Mo et al., 2016;Pernigotti et al., 2016; and have been widely used for source apportionment and speciated emission inventories. Chemical profiles measured at the source have been sufficient to identify and quantify nearby, and reasonably fresh, source contributions. These source types include gasolineand diesel-engine exhaust, biomass burning, cooking, industrial processes, and fugitive dust. Ambient VOC and PM concentrations have been reduced as a result of control measures applied to these sources, and additional reductions have been implemented for toxic materials such as lead, nickel, vanadium, arsenic, diesel particulate matter, and several organic compounds. As these fresh emission contributions in neighborhood-and urban-scale environments (Chow et al., 2002) decrease, regional-scale contributions that may have aged for intermediate (∼ 2 d) or long (∼ 7 d) periods prior to arrival at a receptor gain in importance. These profiles experience augmentation and depletion of chemical abundances owing to photochemical reactions among their gases and particles, as well as interactions upon mixing with other source emissions.
Peatland fires produce long-lasting thick smoke that leads to adverse atmospheric, climate, ecological, and health impacts. Smoke from Indonesian and Malaysian peatlands is a major concern in the countries of southeastern Asia (Wiggins et al., 2018) and elsewhere; it is transported over long distances. Aged peat smoke profiles are likely to differ from fresh emissions, as well as among the different types of peat in other parts of the world.
Ground-based, aircraft, shipboard, and laboratory peat combustion experiments have been carried out to better represent global peat fire emissions and estimate their environmental impacts (e.g., Akagi et al., 2011;Iinuma et al., 2007;Nara et al., 2017;Stockwell et al., 2014Stockwell et al., , 2016. Most peat fire studies report emission factors (EFs) for pyrogenic gases (e.g., methane, carbon monoxide, and carbon dioxide) and fine particle (PM 2.5 , particles with aerodynamic diameter < 2.5 µm) mass, with a few studies reporting EFs for organic and elemental carbon (OC and EC) .
Despite this lack of peat-specific fresh and aged source profiles, results have been published for source apportionment in Indonesia (See et al., 2007), Malaysia (Fujii et al., 2017), Singapore (Budisulistiorini et al., 2018), and Ireland (Dall'Osto et al., 2013;Kourtchev et al., 2011;Lin et al., 2019). These have involved sampling under environments dominated by near-source and far-from-source emissions, such as the 2015 Indonesia burning episode, to determine changes in thermally derived carbon fractions with aging (Tham et al., 2019) and inference of aged peat burning profiles from positive matrix factorization (PMF) application to chemically speciated ambient PM samples (Fujii et al., 2017). Budisulistiorini et al. (2018) observe that ". . . atmospheric processing of aerosol particles in haze from Indonesian wildfires has scarcely been investigated. This lack of study inhibits a detailed treatment of atmospheric processes in the models, including aerosol aging and secondary aerosol formation." Changes in source profiles have been demonstrated in large smog chambers (Pratap et al., 2019), wherein gasparticle mixtures are illuminated with ultraviolet (UV) light for several hours and their end products are measured. Such chambers are specially constructed and limited to laboratory testing. A more recent method for simulating such aging is the oxidation flow reactor (OFR), based on the early studies of Kang et al. (2007), revised and improved by several researchers (e.g., Jimenez, 2018;Lambe et al., 2011), and commercially available from Aerodyne (2019a, b). Although the Aerodyne potential aerosol mass (PAM)-OFR has many limitations, as explained in the Supplement (Sect. S1), it is a practical method for understanding how profiles might change with different degrees of atmospheric aging. A growing users group (PAMWiki, 2019) provides increasing knowledge of its characteristics and operations.
Laboratory peat combustion EFs for gaseous carbon and nitrogen species corresponding with the profiles described here, as well as PM 2.5 mass and major chemical species (e.g., carbon and ions), are reported by Watson et al. (2019). The PM 2.5 speciated source profiles derive from six peat fuels collected from Odintsovo, Russia; Pskov, Siberia; northern Alaska and Florida, USA; and Borneo, Malaysia, representing boreal, temperate, subtropical, and tropical climate regions. Comparisons between fresh (diluted and unaged) and aged (representing intermediately aged (∼ 2 d) and wellaged (∼ 7 d) laboratory-simulated oxidation with an OFR) PM 2.5 speciated profiles are made to highlight chemical abundance changes with photochemical aging. The objectives of this study are to (1) evaluate similarities and differences among the peat source profiles from four biomes; (2) examine the extent of gas-to-particle oxidation and volatilization between 2 and 7 d of simulated atmospheric aging; and (3) characterize carbon and nitrogen properties in peat combustion emissions.

Experiment
The Supplement describes the sampling configuration shown in Fig. S1 and OFR operation. Briefly, peat smoke generated in a laboratory combustion chamber (Tian et al., 2015) was diluted with clean air (by factors of 3 to 5) to allow for nucleation and condensation at ambient temperatures (Watson et al., 2012). These diluted emissions were then passed through an unmodified Aerodyne PAM-OFR in the OFR185 mode without ozone (O 3 ) injection. Hydroxyl radical (OH) production as a function of UV lamp voltage was estimated by inference from sulfur dioxide (SO 2 ) decay using wellestablished rate constants. UV lamps were operated at 2 and 3.5 V with a flow rate of 10 L min −1 and a plug-flow residence time of ∼ 80 s in the 13.3 L anodyne-coated reactor, which translates to OH exposures (OH exp ) of ∼ 2.6 × 10 11 and ∼ 8.8 × 10 11 molecules s cm −3 at 2 and 3.5 V, respectively.
Transport times between source and receptor of 1 to 10 d are typical of peat burning plumes, and the two OH exp estimates were selected to examine intermediate (∼ 2 d) and long-term (∼ 7 d) atmospheric aging. Other emissions aging experiments (e.g., Lambe et al., 2011) cite Mao et al. (2009) for a 24 h average atmospheric OH concentration (OH atm ) of 1.5 × 10 6 molecules cm −3 . This number appears nowhere in the text of Mao et al. (2009), but it corresponds to the groundlevel median value in Mao's Fig. 8 plot of OH vs. altitude for Asian outflows over the Pacific Ocean. The individual measurements in the plot range from OH atm near zero to 5.3 × 10 6 molecules cm −3 . Altshuller (1989) concluded that "the literature contains reports of atmospheric OH radical concentrations measured during daylight hours ranging from 10 5 to over 10 8 molecule cm −3 , but almost all of the values reported are below 5×10 7 molecules cm −3 ." Stone et al. (2012) report atmospheric values ranging from 1.1 × 10 5 molecules cm −3 in polar environments to 1.5 × 10 7 molecules cm −3 in a vegetated forest. Uncertainties in OH exp within the OFR are, therefore, not the controlling uncertainty in estimating profile aging times. Added to this uncertainty are reactions among emission constituents that are not embodied in the OFR185 mode that tend to suppress OH exp with respect to that estimated by the SO 2 calibration Peng et al., 2015Peng et al., , 2016Peng et al., , 2018Peng and Jimenez, 2017). The "OFR Exposure Estimator" available from the PAMWiki (2019) intends to estimate this OH exp , but detailed VOCs from these experiments are insufficient to apply it. The nominal 2 and 7 d aging times determined by dividing OH exp by Mao's 1.5 × 10 6 molecules cm −3 are subject to these uncertainties, which may increase or decrease the aging time estimates. However, these uncertainties, along with other uncertainties related to peat sample selection, moisture content, and laboratory burning conditions, do not negate the value of the measurements reported here. There are distinct differences in the fresh, intermediately aged, and well-aged profiles that address the concerns expressed by Budisulistiorini et al. (2018).
A total of 40 smoldering-dominated peat combustion tests were conducted that included three to six tests for each type of peat fuel (Table S1). The following analysis uses timeintegrated (∼ 40-60 min) gaseous and PM 2.5 filter pack samples collected upstream and downstream of the OFR, representing fresh and aged peat combustion emissions, respectively.

PM 2.5 mass and chemical analyses
Measured chemical abundances included PM 2.5 precursor gases (i.e., nitric acid (HNO 3 ) and ammonia (NH 3 )) as well as PM 2.5 mass and major components (e.g., elements, ions, and carbon). Water-soluble organic carbon (WSOC), carbohydrates, and organic acids that are commonly used as markers in source apportionment for biomass burning were also quantified (Chow and Watson, 2013;Watson et al., 2016).
The filter pack sampling configurations for the four upstream and two downstream channels along with filter types and analytical instrument specifications are shown in Fig. 1. Multiple sampling channels accommodate different filter substrates that allow for comprehensive chemical speciation. Additional upstream Teflon-membrane and quartz-fiber filters were taken for more specific nitrogen and organic compound analyses that are not reported here. The limited flow through the OFR precludes additional downstream sampling.
Detailed chemical analyses along with quality assurancequality control (QA-QC) measures are documented in Chow and Watson (2013). For each analysis, a minimum of 10 % of the samples were submitted for replicate analysis to estimate precisions. Precisions associated with each concentration were calculated based on error propagation (Bevington, 1969) of the analytical and sampling volume precisions .

PM 2.5 source profiles
Concentrations of two gases (i.e., NH 3 and HNO 3 ) and 125 chemical species acquired from each sample pair (fresh vs. aged) were normalized by the PM 2.5 gravimetric mass to obtain source profiles with species-specific fractional abundances. The following analyses are based on the average of 24 paired profiles (shown in Table 1), grouped by upstream (fresh) and downstream (aged) samples for 2 and 7 d aging (i.e., denoted as Fresh 2 vs. Aged 2 and Fresh 7 vs. Aged 7) for each of the six peats with 25 % fuel moisture. Composite profiles are calculated based on the average of individual abundances and the standard deviation of the average within each group (Chow et al., 2002). Although the standard deviation is termed the source profile abundance uncertainty, it is really an estimate of the profile variability for the same fuels and burning conditions, which exceeds the propagated measurement precision.
To assess changes with fuel moisture content, tests of three sets of Putnam (FL1) peats at 60 % fuel moisture were conducted with the resulting profiles shown in Table S2. A few samples were voided due to filter damage or sampling abnormality, which produced five unpaired (either fresh or aged) individual profiles (Table S3). These profiles are reported as they might be useful for future source apportionment studies.

Equivalence measures
The Student's t test is commonly used to estimate the statistical significance of differences between chemical abundances. Two additional measures are used to determine the similarities and differences between profiles: (1) the correlation coefficient (r) between the source profile abundances (F ij , the fraction of species i in peat j ) divided by the source profile variabilities (σ ij ) that quantifies the strength of association between profiles, and (2) the distribution of weighted dif- i2 ] 0.5 ) for < 1σ, 1σ -2σ, 2σ -3σ , and > 3σ . The percent distribution of R/U ratios is used to understand how many of the chemical species differ by multiples of the uncertainty of the difference. These measures are also used in the effective variance chemical mass balance (EV-CMB) receptor model solution that uses the variance (r 2 ) and the R/U ratio to quantify agreement between measured receptor concen-trations and those produced by the source profiles and source contribution estimates (Watson, 2004).

Results and discussion
3.1 Similarities and differences among peat profiles The equivalence measures are used to provide guidance in compositing and comparing the 40 sets of fresh vs. aged profiles. The first comparison is made between two Florida samples from locations separated by ∼ 485 km (i.e., Putnam County lake bed, FL1; and Everglades National Park, FL2), representing different geological areas and land uses. Panel A of Table S4 shows that the two profiles yield high correlations (r > 0.994), but are statistically different (P < 0.002), with over 93 % of the chemical abundance differences within ±3σ . However, when combining both fresh Florida profiles (i.e., all Fresh 2 vs. all Fresh 7 in Panel B), statistical differences are not found, with over 98 % of abundance differences within ±1σ and P > 0.5. Notice that statistical differences are found between the two fresh Florida profiles (i.e., FL1 Fresh 2 vs. FL2 Fresh 2 and FL1 Fresh 7 vs. FL2 Fresh 7 in Panel A) with few (< 0.81 % and 5.6 %) R/U ratios exceeding 3σ ; combining the two Florida profiles may cancel out some of the differences. However, paired comparisons of other combined profiles show statistical differences with low P values (P < 0.002). To further demonstrate the differences, these two Florida profiles are classified as Subtropical 1 and Subtropical 2 to compare with other biomes.
Group comparisons between fresh and aged samples (Panel A of Table 2) show statistical differences for all but Putnam (FL1) peat (P > 0.94). This is consistent with Watson et al (2019) where atmospheric aging (7 d) reduced organic carbon EFs (i.e., EF OC ) by ∼ 20 %-33 % for all but Putnam (FL1) peats (EF OC remained within ±0.5 %). As OC is a major component of PM 2.5 , no apparent changes in OC and carbon fraction abundances may dictate the lack of statistical differences between the fresh and aged profiles.
Paired comparisons for 2 d aging (Panel B of Table 2) show no statistical differences between the Fresh 2 vs. Aged 2 Putnam (FL1) and Malaysian profiles (P > 0.30 and 0.95), which may be due to the low number of samples (n = 2) in the comparison; this results in no statistical differences for combined Putnam (FL1) and Malaysian peat comparison (P > 0.62). Similar to the findings of combining both fresh     0.00 ± 0.0000041 0.00 ± 0.0000036 0.00 ± 0.0000038 0.00 ± 0.0000022 0.00 ± 0.0000025 0.00 ± 0.0000028 0.00 ± 0.0000030 0.00 ± 0.0000024  Table 1. Only one sample was analyzed for elements by X-ray fluorescence with abundance and measurement uncertainty. c Peat ID code, detailed operation parameters are reported in Watson et al. (2019). d Data not available; water-soluble K + data were contaminated for aged samples due to the use of potassium iodide denuder downstream of the oxidation flow reactor. e WSOC measures from peat sample ID PEAT028 were invalidated due to a crack in the test tube. Therefore, only two measurements are used to calculate the average and standard deviation. f Data not available due to the invalidated citric-acid-impregnated filter sample. g The carbon analysis follows the IMPROVE_A thermal-optical reflectance protocol  that is applied in long-term US non-urban IMPROVE and urban Chemical Speciation Network. Organic carbon (OC) is the sum of OC1 + OC2 + OC3 + OC4 plus pyrolyzed carbon (OP).
Elemental carbon (EC) is the sum of EC1 + EC2 + EC3 minus OP. Total carbon is the sum of OC and EC. Since a large fraction of OP (7 %-13 %) is found in smoldering peat combustion emissions -indicative of higher-molecular-weight compounds that are likely to char -the resulting EC fractions are lower than the individual EC fraction after OP correction. a For the t test, a cutoff probability level of 5 % is selected; if P < 0.05, there is a 95 % probability that the two profiles are different. For correlations, r > 0.8 suggests similar profiles, 0.5 < r < 0.8 indicates a moderate similarity, and r < 0.5 denotes little or no similarity. The R/U ratio indicates the percentage of the > 93 reported chemical abundances differ by more than an expected number of uncertainty intervals. The normal probability density function of 68 %, 95.5 %, and 99.7 % for ±1σ , ±2σ , and ±3σ , respectively, is used to evaluate the R/U ratios. The two profiles are considered to be similar, within the uncertainties of the chemical abundances when 80 % of the R/U ratios are within ±3σ , with r > 0.8 and P > 0.05. Species with R/U ratios > 3σ are further examined as these may be markers that further allow source contributions to be distinguished by receptor measurements. They may also reflect the sampling and analysis artifacts that are not representative of the larger population of source profiles. b Unless otherwise noted, boreal represents Russia and Siberia regions, temperate represents the northern Alaska region, subtropical is divided into Subtropical 1 for Putnam (FL1) and Subtropical 2 for Everglades (FL2) peats, and tropical represents the island of Borneo, Malaysia, region. c n1 and n2 denote number of samples in comparison. d Student's t test P values.
Florida profiles (i.e., all Fresh2 vs. all Fresh 7 in Table S4), the two fresh Alaskan profiles (Fresh 2 vs. Fresh 7 in Panel D of Table 2) do not show statistical differences (P > 0.12).
Compositing profiles by averaging each of the measured abundances may disguise some useful information. For receptor model source apportionment, region-specific profiles are most accurate for estimating source contributions.
Student's t tests for the gravimetric PM 2.5 mass concentrations (µg m −3 ) measured upstream and downstream of the OFR (Table S5) show statistically significant differences (P < 0.05) between fresh vs. aged PM 2.5 (i.e., Fresh 2 vs. Aged 2 and Fresh 7 vs. Aged 7). Fresh 2 and Fresh 7 PM 2.5 mass concentrations are similar, as expected from replicate tests for the same conditions. Increases in some species abundances offset decreases on other abundances, resulting in similar PM 2.5 levels for the "all fresh vs. all aged" comparison.

Ratios of sum of species to PM 2.5 mass
The sum of the major PM chemical abundances should be less than unity since oxygen, hydrogen, and liquid water content are not measured (Chow et al., 1994(Chow et al., , 1996. As shown in Table S6, the sums of elements, ions, and carbon explain averages of ∼ 70 %-90 % of PM 2.5 mass for fresh profiles except for Russian peat (62 %-64 %). The "sum of species" decreased by an average of 6 % and 11 % after 2 and 7 d, respectively. These differences are consistent with loss of semivolatile organic compounds (SVOCs) in the low-temperature carbon fractions, although they are offset by formation of oxygenated compounds during aging. This is true for all but Putnam (FL1) peat, for which the sum of species explains nearly the same fraction of PM 2.5 for the fresh and aged profiles.

Comparison between fresh and aged profiles
Fresh and aged chemical abundances are compared in Fig. 2. Species abundances vary by several orders of magnitude but exhibit two distinguishable clusters: centered around 0.1 % for reactive and secondary ionic species (e.g., NH + 4 , NO − 3 , and SO = 4 ) and centered around 10 % for carbon compounds (e.g., OC fractions and WSOC). While most gaseous NH 3 /PM 2.5 ratios exceed 10 %, HNO 3 /PM 2.5 ratios are well below 1 % in fresh emissions. Reactive-ionic species and carbon components are mostly above and below the 1 : 1 line, respectively, implying particle formation and evaporation after atmospheric aging. Large variabilities are found for individual species as noted by the standard deviations associated with each average. Figure 3 shows the ratio of averages between aged and fresh profiles. Atmospheric aging increased oxalic acid, NO − 3 , NH + 4 , and SO = 4 abundances (likely due to conversion of nitrogen and sulfur gases (e.g., NH 3 , NO, NO 2 , and SO 2 ) to particles), but decreased NH 3 , levoglucosan, and low-temperature OC1 and OC2 abundances in most cases. Large variations are found among measured species (left panels in Fig. 3) as ratios range several orders of magnitude for mineral and ionic species. Consistent with Fig. 2 where most carbon compounds are close to but below the 1 : 1 line, the right panels in Fig. 3 show the reduction of carbonaceous abundances with aged / fresh ratios between 0.1 and 1. Higher aged / fresh ratios in low-temperature OC1 and OC2 after 7 d aging are consistent with additional volatilization with longer aging time.
Atmospheric aging should not change the abundances of mineral species (e.g., Al, Si, Ca, Ti, and Fe), except to the extent that the PM 2.5 mass (to which all species are normalized) increases or decreases with aging. Large standard deviations associated with the ratio of averages for mineral species in the left panels of Fig. 3 illustrate variabilities among different combustion tests for the less abundant species.

Organic carbon and thermally evolved carbon fractions
Total carbon (TC, sum of OC and EC) constitutes the largest fraction of PM 2.5 (Table 1), accounting for 59 %-87 % and 43 %-77 % of the PM 2.5 mass for the fresh and aged profiles, respectively. OC dominates TC with low EC abundances (0.67 %-4.4 %), as commonly found in smolderingdominated biomass combustion (Chakrabarty et al., 2006;Chen et al., 2007). The largest OC fractions are hightemperature OC3 (15 %-30 % of PM 2.5 ), consistent with past studies for biomass burning emissions Chow et al., 2004). OC abundances decreased with aging time. As shown in Fig. S2, upstream (Fresh 2 and Fresh 7) OC abundances ranged from 58 % to 85 % and decreased by 4 %-12 % and 20 %-33 % after 2 and 7 d aging, respectively. The exception is for Putnam (FL1) peat, where the OC abundances were similar (changed by ∼ 0.5 to 1.5 %) between fresh and aged profiles. Part, but not all of this reduction is due to increasing abundances of non-carbon components, particularly nitrogen-containing species that add to PM 2.5 mass. OC abundance decreases after aging for other profiles may have contributed to the statistical differences found between fresh and aged PM 2.5 mass (Table S5). With the exception of Putnam (FL1) peat, the additional 7 %-22 % OC degradation from 2 to 7 d aging implies that much of the OC changes require about a week of aging time.
The Student's t test for fresh and aged profiles shows statistical differences (P < 0.05) for TC, OC, and lowtemperature OC1 and OC2, but similarities for OC3 and OC4. High-temperature OC3 and OC4 contain more polar and/or high-molecular-weight organic components  that are less likely to photochemically degrade. Large fractions of pyrolyzed carbon (OP of 7 %-13 %) Figure 2. Comparison between fresh and aged profile chemical abundances for each of the six types of peat with 2 and 7 d aging times. Standard deviations associated with averages for x and y variables are also shown. Vertical dashed lines (red) at 1 % on the x axis delineate the two distinguishable clusters: centered around 0.1 % for reactive and ionic species and centered around 10 % for carbon compounds. are also found, indicative of higher-molecular-weight compounds that are likely to char (Chow et al., , 2004. Reduction in OC abundances after atmospheric aging is attributed mostly to decreases in low-temperature OC1 and OC2 abundances in the OFR as shown in the fresh vs. aged ratios of average abundances (Fig. 3). Figure S3a shows reductions in OC1 abundances after 2 and 7 d of atmospheric aging are apparent but at a similar level: ranging from 2 % to 10 % and 3 % to 14 %, respectively. Additional OC1 reductions from 2 to 7 d are most apparent for Russia and Everglades (FL2) peats at the 6 %-10 % level. Similar reductions are found for OC2 (Fig. S3b): ranging from 3 % to 11 % and 3 % to 12 % after the 2 and 7 d of aging, respectively. Prolonged aging times resulted in additional 4 %-8 % OC2 reduction for all but Russian and Putnam (FL1) peats. As oxidation of organic compounds with OH radicals is an efficient chemical aging process (Chim et al., 2018), some of the VOCs and SVOCs may have been liberated (Smith et al., 2009).

Organic mass (OM) and OM/OC ratios
Reduction of the sum of species and OC abundances from fresh to aged profiles can be offset by the formation of oxygenated organic compounds as the profiles age. Different assumptions have been used to transform OC to organic mass (OM) to account for unmeasured H, O, N, and S in organic compounds (Cao, 2018;Chow et al., 2015a;Riggio et al., 2018). As single multipliers for OC cannot capture changes by oxidation in the OFR, OM is calculated by subtracting mineral components (using the IMPROVE soil formula by Malm et al., 1994), major ions (i.e., NH + 4 , NO − 3 , and SO = 4 ), and EC from PM 2.5 mass to account for unmeasured mass in organic compounds (Chow et al., 2015a;Frank, 2006). This approach assumes that no major chemical species are unmea- Figure 3. Ratios of average aged (A) to fresh (F) chemical species for 2 d (A2/F2) and 7 d (A7/F7) of atmospheric aging for six types of peats. Error bars represent the standard deviations associated with each ratio. Note that different scales were used for the two Y axes, with 0.001 to 10 000 on the left axis and 0.1 to 100 on the right axis (species abbreviations are shown in Table 1; OM is organic mass). sured and that the remaining mass consists of H, O, N, and S associated with OC in forming OM. Table 3 shows that OM / OC ratios ranged from 1.1 to 1.7 and 1.3 to 2.2 for fresh and aged profiles, respectively. The lower OM / OC ratios in fresh emissions are consistent with those reported for other types of biomass burning Reid et al., 2005). Figure S4 shows a general upward trend in OM / OC ratios after atmospheric aging with additional 14 %-21 % increases from 2 to 7 d for all but Putnam (FL1) peat. The increase in OM / OC ratios with aging is likely due to an increase in oxygenated organics. The OM / OC ratio of 1.20 ± 0.05 for fresh Borneo, Malaysia, peat is consistent with the 1.26 ± 0.04 ratio for fresh peat burning emissions in Central Kalimantan, Indone-sia (Jayarathne et al., 2018), both located on the island of Borneo.
The highest OM / OC ratios are found for Russian peat, ranging from 1.6 to 1.7 for fresh profiles and increasing to 2.1-2.2 for aged profiles, consistent with formation of lowvapor-pressure oxygenated compounds in the OFR. Watson et al. (2019) report that the Russian peat fuel contains the lowest carbon (44.20 ± 1.01 %) and highest oxygen (38.64 ± 0.78 %) contents among the six peats. The low carbon contents in peat fuel and source profiles are consistent with the lowest sum of species found in Russian peat, with 62 %-64 % and 50 %-52 % of PM 2.5 mass for the fresh and aged profiles, respectively. After 7 d aging for Siberian peat, the increasing OM / OC ratios from 1.2±0.14 to 1.5±0.18 are similar to the increase from 1.22 to 1.42 reported by Bhattarai et al. (2018).  0.039 ± 0.0035 a Uncertainty associated with each ratio is calculated based on the square root of the individual uncertainties multiplied by the ratio (Bevington, 1969

Water-soluble organic carbon (WSOC)
WSOC abundances in PM 2.5 were over 2-fold higher in fresh Russian peat (36 %-37 %) than Malaysian (15 %-17 %) peat. The 15 %-17 % WSOC in PM 2.5 for fresh Borneo, Malaysia, peat (Table 1) is consistent with the 16 ± 11 % from Central Kalimantan, Indonesia, peat (Jayarathne et al., 2018). However, the WSOC / PM 2.5 ratio is not a good indicator of changes in WSOC abundances during atmospheric aging as PM 2.5 also contains non-water-soluble and noncarbonaceous aerosol. Table S7 shows large variabilities associated with the differences (i.e., aged minus fresh), suggesting that no differences exist within ±3 standard deviations. The only exceptions are for the 7 d Putnam (FL1) peat and 2 d Malaysian peat, where aging resulted in 7 %-8 % increases in WSOC abundances in PM 2.5 .
As WSOC is part of the OC, the WSOC / OC ratio is a better indicator of atmospheric aging. WSOC / OC ratios (Table 3) vary between fresh (0.18-0.64) and aged (0.31-0.71) profiles. Figure S5 shows a general increase in WSOC / OC ratios from fresh to aged profiles. Longer aging time from 2 to 7 d results in 5 %-10 % higher WSOC / OC ratios for all but the two Florida peats. OC water solubility also varies by peat type. Russian peat OC emissions are largely watersoluble, whereas Malaysian peat emissions are mostly waterinsoluble, with WSOC / OC ratios of 0.59-0.71 and 0.18-0.40, respectively.

Carbohydrates
Bates et al. (1991) found that peat from Sumatra, Indonesia, consisted of 18 %-46 % carbohydrate (mainly levoglucosan) relative to total carbon based on nuclear magnetic resonance spectroscopy. Levoglucosan and its isomers (mannosan and galactosan) are saccharide derivatives formed from incomplete combustion of cellulose and hemicellulose (Kuo et al., 2008;Louchouarn et al., 2009) and have been used as markers for biomass burning in receptor model source apportionment (Bates et al., 1991;Watson et al., 2016). These carbohydrate-derived pyrolysis products undergo heterogeneous oxidation when exposed to OH radicals in the OFR (Hennigan et al., 2010;Kessler et al., 2010).
Only five of the 17 carbohydrates (Table 1) were detected, with noticeable variations (e.g., > 2 orders of magnitude) in levoglucosan for boreal and temperate peats. Levoglucosan abundances account for 35 %-39 % and 20 %-25 % of PM 2.5 mass for fresh and aged Russian profiles, respectively. On a carbon basis, Table 3 shows that levoglucosan carbon (with an OM / OC ratio of 2.25) accounts for 42 %-48 % and 30 %-35 % of WSOC and 27 %-28 % and 21 %-24 % of OC for fresh and aged Russian profiles, respectively. These levels are less than the 96 ± 3.8 % levoglucosan or ∼ 42.7 % of levoglucosan carbon in OC reported for German and Indonesian peats (Iinuma et al., 2007). Elevated levoglucosan is also found for Siberian and Alaskan peats, ranging from 4 % to 18 % in PM 2.5 . However, the levoglucosan abundances are low (1 %-4 %) for the subtropical and tropical peats. An aging time of 7 d resulted in an additional 1 %-4 % levoglucosan degradation relative to 2 d aging with the exception of an additional 9 % reduction for Russian peat.
The extent of levoglucosan degradation depends on organic aerosol composition, OH exposure in the OFR, and vapor wall losses (Bertrand et al., 2018a, b;Pratap et al., 2019). Figure 4 shows the presence of levoglucosan carbon for the Russian and Alaskan peats after 2 and 7 d aging, at the levels of 8 %-11 % and 2 %-9 %, respectively, in line with a chemical lifetime longer than 2 d. This is consistent with the estimated 1.2-3.9 d of levoglucosan lifetimes under different environments reported by Lai et al. (2014). However, other studies (Hennigan et al., 2010;May et al., 2012;Pratap et al., 2019) found that levoglucosan experiences rapid gas-phase oxidation, resulting in ∼ 1-2 d lifetimes at ambient temperatures.
While the presence of levoglucosan in peat smoke is apparent, its isomer galactosan was not detectable. Mannosan is detectable in cold climate peats with 1 %-5 % of PM 2.5 for the Russian and Alaskan peats and up to 1.3 % for Siberian peat. Apparent degradations from 3.9 % to 2.5 % and from 5.0 % to 2.1 % in mannosan abundances are found for Russian peat (Table 1) after 2 and 7 d, respectively. A 2-to 3fold reduction in mannosan is also evident after 7 d aging for the Siberian and Alaskan peats. Similar observations apply to glycerol in Russian peat, ranging from 1.9 % to 3.5 % and 1.3 % to 1.7 % of PM 2.5 for fresh and aged profiles, respectively. Other detectable carbohydrates are galactose and mannitol, typically present at < 5 % of the levoglucosan abundance.

Organic acids
Organic acids have been associated with many anthropogenic sources, including engine exhaust, biomass burning, meat cooking, bioaerosol, and biogenic emissions. Past studies show the presence of low-molecular-weight dicarboxylic acids in biomass burning emissions (e.g., Falkovich et al., 2005;Veres et al., 2010).
Only four of the 10 measured organic acids (Table 1) (i.e., formic acid, acetic acid, oxalic acid, and propionic acid) were detected with variable abundances (< 0.02 %-3.9 %). The largest changes between fresh and aged profiles are found for oxalic acid, ranging from < 0.02 % to 0.43 % of PM 2.5 for fresh profiles, with an ∼ 10-to 20-fold increase after 2 d (0.6 %-1.3 %) and with 1 to 2 orders of magnitude increases after 7 d (1.1 %-3.9 %). With the exception of Figure 4. Abundances of fresh and aged carbon-containing components in PM 2.5 . Levoglucosan (C 6 H 10 O 5 ) is divided by 2.25 and oxalic acid (C 2 H 2 O 4 ) is divided by 3.75 to obtain the carbon content. These levels are subtracted from the water-soluble organic carbon (WSOC) to obtain the remainder, and WSOC is subtracted from organic carbon (OC) to obtain non-soluble carbon. Elemental carbon (EC) is unaltered.
Acetic acid abundances are stable between fresh and aged profiles, mostly in the range of 0.2 %-0.5 % except for a 6fold increase from 0.23 ± 0.15 % (Fresh 7) to 1.5 ± 2.0 % (Aged 7) for Siberian peat with large variability among the tests. Formic acid and propionic acid abundances are low (< 0.5 % and < 0.02 %, respectively), but increase with aging. Extending the aging time from 2 to 7 d resulted in a notable increase in organic acid abundances, consistent with the increases in WSOC / OC ratios (Table 3). By biome, the highest abundances for organic acids in PM 2.5 are found for aged (Aged 7) Siberian peat, with 3.9 ± 1.4 % oxalic acid, 1.5 ± 2.0 % acetic acid, and 0.44 ± 0.28 % formic acid (Table 1).

Nitrogen species, sulfate, and chloride abundances
Ammonia normalized to PM 2.5 mass is high for fresh profiles, ranging from 17 % to 64 %, except for the low NH 3 content in Russian peat (6 %-8 %). These abundances are reduced to 3 %-14 % and 1 %-7 % after 2 and 7 d aging, respectively. As shown in Fig. 5, most of the NH 3 rapidly diminished after 2 d, with increasing particle-phase NH + 4 and NO − 3 after 7 d. The highest NH 3 -to-PM 2.5 ratios are found for fresh Everglades (FL2) peat profiles (51 %-64 %), ∼ 2-8fold higher than other peats. These high and low NH 3 /PM 2.5 ratios are consistent with the nitrogen contents in peat fuel: 3.93 ± 0.08 % for Everglades and 1.50 ± 0.52 % for Russian peats (Watson et al., 2019).
files. As shown in Fig. 6, the largest component of PM 2.5 is OM, accounting for 94 %-99 % and 80 %-95 % of PM 2.5 mass for fresh and aged profiles, respectively. Although the 7 d aging time increased the OM / OC ratios (by 12 %-19 %), the abundances of OM in PM 2.5 are reduced (3 %-18 %). This can be attributed to the combined effects of increased oxygenated organics, SVOC volatilization (Smith et al., 2009), and an increase in ionic species as shown in the average aged / fresh ratios in Fig. 3. Figure 6 shows increases in ionic species (i.e., sum of NH + 4 , NO − 3 , and SO = 4 ), with low abundances (0.3 %-1.7 %) in fresh profiles, increasing 3 %-16 % after aging. The sum of ionic species accounts for 11 %-16 % of PM 2.5 mass for the Siberian, Alaskan, Everglades (FL2), and Malaysian peats after 7 d, mainly due to the increase in NH + 4 and NO − 3 as shown in Fig. 5. Elemental abundances are low (< 0.0001 %), mostly below the lower quantifiable limits. Table 1 only lists 34 of the 51 elements (Na to U) detected by XRF. Using the IM-PROVE soil formula (assuming metal oxides of major mineral species; Malm et al., 1994) yielded 0.07 %-2.9 % of mineral components. The IMPROVE soil formula has been applied in many other studies (e.g., Chan et al., 1997;Pant et al., 2015;Rogula-Kozlowska et al., 2012), which provides an adequate estimate of geological mineral in reconstructed mass. Since geological minerals are not a major component of PM 2.5 , variations in the assumption regarding metal oxides or multipliers do not contribute to large variations in reconstructed mass (Chow et al., 2015a).
This study indicates that an aging time of ∼ 2 d represents the intermediately aged source profile, whereas ∼ 7 d repre-sents the profile with adequate residence time to complete the atmospheric process.

Changes in source profiles by fuel moisture content
The effect of fuel moisture content on source profiles is mostly unknown. The 25 % fuel moisture content selected for this study intends to better simulate the conditions of moderate to severe droughts where most peat fires occur. Increasing fuel moisture content from ∼ 25 % to 60 % for the three Putnam (FL1) peat fuels yielded 12 % higher EFs for CO 2 (EF CO 2 ), but 12 %-20 % lower EFs for CO, NO, NO 2 , and PM 2.5 mass (Watson et al., 2019). Tests of fuel moisture content on profile changes are available for only 2 d aging. Equivalence measures (Table S8) show statistical differences (P < 0.001) between 25 % and 60 % moisture profiles for either fresh or aged profiles with high correlations (r > 0.997), and over 93 % of species abundance falls within ±3σ . While OC abundances in PM 2.5 are comparable for the fresh and aged profiles (70 %-72 %) for 25 % fuel moisture, a reduction of 18 % OC in PM 2.5 is found for 60 % fuel moisture (from 82 % to 64 %) after aging (Table S2). The higher fuel moisture content also reduced WSOC by 6 % and levoglucosan by 1.3 % with < 1 % increases for NH + 4 and organic acids. After aging, the NH 3 -to-PM 2.5 ratios decreased from 28 % to 5 % and from 20 % to 8 % for the 25 % and 60 % fuel moisture, respectively. These results are not conclusive as most measurements are associated with high variabilities.