Introduction
In recent decades, peatland fires in Southeast Asia, especially the
Indonesian provinces of Sumatra, Kalimantan, and Papua as well as Malaysian
Borneo, have become more frequent in occurrence (Page et al., 2009;
Langner and Siegert, 2009; Van der Werf et al., 2010). Major peat-burning
events have coincided with the El Niño Southern Oscillation (e.g.,
1997–1998, 2006, 2015), during which warmer conditions decrease dry season
precipitation, which lowers the water table of peatlands, increases their
flammability, and promotes longer-range transport of the smoke (Reid et al.,
2013). Within a season, meteorological factors contribute to major
peat-burning pollution events and transport, including typhoons and wind
patterns (Wang et al., 2013). Notably, even in
non-El Niño years, peat burning remains an important source of biomass
burning emissions in Southeast Asia (Reid et al.,
2013). The 2015 peatland fire episode that occurred September–November
2015 occurred during an El Niño year and was reported as the strongest
peatland fire episode since 1997–1998 (Parker et al., 2016; Koplitz et al.,
2016; Huijnen et al., 2016). The 2015 fires burned ∼ 1 million ha of tropical forests and peatlands in Indonesia, releasing
∼ 0.2 Pg C of carbon to the atmosphere (Huijnen et al.,
2016). However, these values are well below the 1997–1998 estimates of
∼ 2 million ha of burned peatland area that released
∼ 1.7 Pg C of carbon to the atmosphere in 2015 (Page et al.,
2002; Chisholm et al., 2016; Huijnen et al., 2016; Tacconi, 2003). The 2015
peatland fire smoke also impacted neighboring Singapore, Malaysia, Thailand and
Philippines with an estimated economic loss greater than USD 16 billion to
their GDPs due to declines in productions and services during the event, in
addition to
long-term impacts to human health and the environment (Glover and Jessup,
2006; Chisholm et al., 2016; WorldBank, 2016). Negative health effects due
to inhalation of peat smoke were widely reported during this catastrophe
(Koplitz et al., 2016). In Palangka Raya, the capital of Central
Kalimantan, PM10 levels reached up to 3741 µg m-3, nearly
2 orders of magnitude higher than the World Health Organization (WHO)
guideline for 24 h PM10 exposure (Stockwell et al., 2016;
WHO, 2005). It was estimated that more than 40 million people suffered from
continuous exposure to peat smoke during this episode and significant
increase of premature deaths were reported due to respiratory and
cardiovascular diseases (Koplitz et al., 2016). Despite the substantial
environmental, socioeconomic, and health impacts, the peatland fire emissions
are still under-studied with respect to their chemical and physical
properties.
Peatlands are globally distributed over ∼ 400 Mha land area,
hold ∼ 550 MgC ha-1 of carbon per 1 m depth, and can
reach depths of 20 m. It has been estimated that ∼ 5.4×1014 kg
of carbon are stored underground in peat
deposits, accounting for a significant fraction (44–71 %) of the
terrestrial carbon pool (Maltby and Immirzi, 1993; Yu et al., 2010).
Tropical peatlands particularly in Malaysian and Indonesian lowlands are
frequently converted to agricultural croplands, commercial forests, or
pasture by draining the peatlands (Maltby and Immirzi, 1993). During
1996–1999 the Indonesian government excavated more than 4000 km of drainage
channels throughout 1 Mha of peatland to cultivate rice under the former
Mega Rice Project (Page et al., 2009). After the project was
abandoned in 1999, deforested and degraded peatlands were covered with
secondary vegetation (Page et al., 2009). In recent decades, Indonesian
peatland fires have occurred more frequently, intensively, and extensively.
Degraded peatlands are at high risk of uncontrolled fire, because dry peat
is highly combustible and secondary vegetation is more fire-prone than the
original forest (Langner and Siegert, 2009; Page et al., 2002, 2009).
Fires first occur in aboveground vegetation, then enter into the
carbon-rich soils where they smolder and can spread slowly beneath the
surface until the peatland is flooded during the next monsoon (Page et
al., 2009). The burned areas do not easily regenerate to primary
vegetation; instead, they are converted into ferns with patchy secondary
vegetation that are prone to repeat fires (Chisholm et al., 2016).
Peat contains more than 85 % organic matter (OM) by dry mass that is made of
plant tissues at varying stages of decomposition, with major organic
compound classes being cellulose, hemicellulose, lignin, cutin, humic acid,
and fulvic acid (Dehmer, 1995; Zulkifley et al., 2015; Dizman et al.,
2015). Peat is categorized as fibric, hemic, or sapric based on the degree
of decomposition. Fibric peat is the least degraded type with higher fiber
content, while sapric peat is the most degraded peat type with an amorphous
structure, and hemic peat has intermediate properties (Huat et al.,
2011). Thus, peat soils carry biomarkers indicative of floral origin and
these could be potentially used to identify peatland fire emissions.
Levoglucosan, mannosan, syringaldehyde (S), vanillin (V), syringic acid (SA), vanillic
acid (VA), and n-alkanes are common biomass burning tracers and specific ratios of
these compounds were suggested as indicators of peatland fire emissions in
previous studies that analyzed the ambient air impacted by peat smoke
(Fujii et al., 2014, 2015a, b). Some
organic compounds (e.g., polycyclic aromatic hydrocarbons, PAHs) are highly enriched in peat smoke compared to
raw peat biomass, showing over 100 times greater concentration in smoke than
soil and indicating their formation during combustion (Black et al., 2016).
Prior studies of peat-burning emissions involved either laboratory
experiments or collecting ambient aerosols at receptor sites impacted by
peat smoke. Many of these studies primarily focused on chemically
characterizing gaseous emissions (Benner, 1977; Chen et al., 2007;
Christian et al., 2003; Geron and Hays, 2013; May et al., 2014; McMahon et
al., 1980; Ward, 1990; Hatch et al., 2015; Stockwell et al., 2015, 2014;
George et al., 2016; Black et al., 2016; Iinuma et al., 2007;
Yokelson et al., 1997) while fewer focused on the PM fraction (Black et
al., 2016; Fujii et al., 2014, 2015a; Iinuma et al., 2007).
Peatland fire emissions were not considered in the biomass burning emission
inventory published by Andreae and Merlet (2001). Akagi et al. (2011)
updated this inventory to include peatland fires as a source of biomass
burning emissions but did not report an PM2.5 emission factor
(EFPM2.5).
Peat fire EFPM2.5 reported in the literature have varied by a large scale,
ranging from 5.9 to 66 g kg-1 with uncertainties
associated with measurements of emissions of black carbon (BC) and organic
carbon (OC) greater than 50 % of the associated value (Black et al.,
2016; Geron and Hays, 2013; Akagi et al., 2011). Thus, the global estimates
of peat fire PM2.5, OC, and BC emissions are associated with large
uncertainties. The variation across lab-measured emission factors (EFs) likely results from
different burning conditions. In addition, the dissection of peat soil
during sampling, handling, transport, and storage of peat can significantly
alter its physical properties and subsequent combustion. Thus, in situ sampling of
peat fire emissions under natural burning conditions is needed to accurately
represent peat fire emissions in global peat fire emission estimates and
parameterize human exposure studies and climate and air quality models
(Van der Werf et al., 2010; Page et al., 2002; Akagi et al., 2011).
The objectives of this paper are to characterize in situ peat PM emissions from
different peat-burning sites in Indonesia during the 2015 El Niño
period, compute PM emission factors and develop source profiles for peat-burning aerosols,
and compare the peat PM emission factors from the
literature with our in situ measurements. A moveable
lab was deployed during the 2015
fire episode in Palangka Raya, Central Kalimantan, to make in situ ground-based
measurements of trace gases and aerosols directly from authentic peatland
fire smoke. Samples discussed in this paper were collected from 18 peat fire
smoke plumes across 6 sites and were chemically speciated for
∼ 90 gas-phase species and ∼ 70 particulate-phase species.
This paper focuses on the particulate-phase chemistry, and a
comprehensive description of gas-phase emissions and optical properties
(brown carbon (BrC), BC, and the mass absorption coefficients
for the bulk OC due to BrC) is given in Stockwell et al. (2016). Combined
together, EFs for more than 150 gaseous and particulate species were
determined, providing a wealth of chemical detail on these emissions and
enabling the evaluation of the magnitude of PM2.5 emissions and the
ratio of particulate to gaseous carbon emitted from the 2015 El Niño
peat fires.
Experimental details
Site description
A comprehensive description of sampling sites is given in Stockwell et
al. (2016) and a brief overview is given here. PM2.5 samples were
collected from 18 separate plumes from 6 different peatland areas in Central
Kalimantan, Indonesia, from 1 to 7 November during the 2015 El Niño. The
sites were carefully selected to represent different peat types (fibric,
hemic, or sapric) and cover a range of burning depths ranging from 18 to 62 cm,
averaging (± standard deviation) 34±12 cm. The sampled
sites were located where the maximum fire activity is typically reported, in
moderately to heavily disturbed areas by roads, canals, and/or previous
fires. The aboveground vegetation was nonexistent (most often due to surface fires that ignited the peat hours to days prior to sampling) or
limited to ferns or patchy secondary vegetation that was not burning. The samples were collected directly from visible plumes in smoldering
peat. Sampling was immediately stopped during any occasional flaming
combustion events within aboveground vegetation in the vicinity to ensure
sampling of pure smoldering peat emissions.
Each plume was identified by an English letter (E–Z to AA) and the complete
description of the plumes including peat type, burning depth, and surface
fuel is given in Table S1 of Stockwell et al. (2016). Two PM samples were
collected from plumes E, F, and W, bringing the total number of PM samples to
21. Because of the variability in PM emissions within a single plume, both
values were used in calculating study averages. Plume Y showed a different
emission profile from the others likely due to co-burning of leaf litter at
this shallow peat-burning site. Thus, plume Y was excluded from average
calculations but individual values are reported in Table S2 and
corresponding figures.
Sample collection
A comprehensive description of sample collection is given in Stockwell et
al. (2016). In brief, PM2.5 was collected using a custom-built,
two-channel PM sampler. The sampling inlet was mounted on a ∼ 2.5 m
pole to allow sampling of smoke from a safe distance. The inlet was
positioned approximately 2–3 m downwind of the smoldering peat, at a point
where the plume of smoke had cooled to near-ambient temperature. The sample
inlet was not fixed to a point and always followed the plume path when the
plume direction changed due to variations in ambient air flow. PM samples were collected over a period
of 9–30 min each, at PM2.5 concentrations that averaged 15 mg m-3
and ranged from 1 to 40 mg m-3. The duration of filter sample
collection and PM2.5 concentrations sampled are summarized in Table S1
for each plume. For plumes with two samples collected, the time over which
samples were collected were comparable and the sampled PM2.5
concentrations were within a factor of 3. The PM was collected on
pre-cleaned 47 mm quartz fiber filters (QFFs) and pre-weighed Teflon filters
(PALL, Life Sciences, Port Washington, NY) preceded by two 2.5 µm sharp-cut cyclones (URG). The filtered air was then passed to the land-based
Fourier transform infrared (LA-FTIR) spectrometer multipass cell for the
measurement of gas-phase species as described by Stockwell et al. (2016).
Sampled filters were stored in the dark and frozen (-20 ∘C) and
were shipped frozen to the University of Iowa for chemical analysis.
Field blanks were collected for every fifth sample. For some samples a
second (backup) QFF was placed in series behind the first (front) QFF
in order to assess the positive sampling artifacts from carbonaceous
gas adsorption. Filter samples were collected upwind of the plumes for
∼ 20 min (similar to smoke sampling duration) in order to
account for background PM2.5.
PM2.5 mass, elemental carbon (EC), and organic carbon measurement
A complete description of PM mass, EC, and OC measurements is given in Stockwell et al. (2016). In brief, PM mass was
calculated as the difference of pre- and post-sampling filter weights of
Teflon filters after conditioning for 48 h in a desiccator. The relative
error in the PM mass measurements was propagated from the standard deviation
of the triplicate measurements of pre- and post-sampling filter weights, the
standard deviation of background PM masses, and 10 % of the PM mass
concentration, which is a conservative estimate of the analytical
uncertainty associated with the mass measurement. Ambient background
PM2.5 concentrations were very similar across all the sites and on
average the ambient PM2.5 contributed only 0.60 % of the sampled
PM2.5 mass, indicating that the ambient PM contribution was very
small compared to PM concentration in the peat smoke. Nevertheless, the
average background concentration was subtracted from the sample
concentrations in order to calculate pure peat fire emissions.
EC and OC were measured by thermal optical analysis following the NIOSH 5040
method using 1.00 cm2 punches of QFFs (Sunset
Laboratories, Forest Grove, OR) (NIOSH, 2003). The uncertainty in OC
measurements was propagated from the standard deviation of the background
filters, the standard deviation of the backup filters, and 10 % of the OC
concentration, a conservative estimate of the method precision in replicate
measurements (NIOSH, 2003). The uncertainty of EC measurements was
propagated from the instrumental uncertainty (0.05 µg cm-2),
5 % of the measured EC, and 5 % of pyrolyzed carbon, which refers to
organic carbon that charred during analysis.
Water-soluble organic carbon (WSOC)
A 1.053 cm2 subsample of QFF was analyzed for WSOC using a total organic carbon analyzer (Sievers
5310 C, General Electric). WSOC was extracted into 15.0 mL of > 18.2 MΩ
resistivity ultra-pure water (Thermo, Barnstead Easypure II) using acid
washed (10 % nitric acid) and pre-baked (550 ∘C for 5.5 h)
glassware. Inorganic carbon was removed with an inorganic carbon remover
(Sievers ICR, General Electric). WSOC was measured in triplicate and quantified using
standard calibration curves prepared from potassium hydrogen phthalate
(Ultra Scientific). The WSOC concentration in the sampled plumes was
calculated using the extraction volume, total filter area, and sampled air
volume. The uncertainty of the WSOC measurement was propagated using the
standard deviation of the triplicate measurements, standard deviation of the
background filters and 10 % of the WSOC concentration. The fraction of
water-insoluble organic carbon (WIOC) was calculated by subtracting the WSOC
concentration from total OC concentration. The error of WIOC concentration
was propagated from individual uncertainties of OC and WSOC.
Water-soluble inorganic ions
Water-soluble inorganic ions were quantified in aqueous extracts of Teflon
filters by ion exchange chromatography coupled with conductivity detection
as described in detail elsewhere (Jayarathne et al., 2014). In brief,
half of the Teflon filter was uniformly wet with 50 µL of isopropyl
alcohol and subsequently extracted into 15.0 mL ultra-pure water
(> 18.2 MΩ resistivity) by shaking 12 h at 125 rpm.
For cation analysis, a Dionex IonPac CS12A column was used with the mobile
phase of 20 mM methane sulfonic acid at 0.5 mL min-1 flow rate. A
Dionex IonPac AS22 anion column with the mobile phase of 4.5 mM sodium
carbonate (Na2CO3) and 1.4 mM sodium bicarbonate (NaHCO3) at
a flow rate of 1.2 mL min-1 was used for anion separation. A
conductivity detector (Thermo) was used for detection and was preceded by a
self-regenerating suppressor: CERS-500 and AERS-500 for cations and anions,
respectively.
Total metals
Teflon filters were cut in half using ceramic blades and then digested in
mixtures of 2:1 concentrated nitric and hydrochloric acid (TraceMetal Grade,
Fisher Chemical) using a MARS 6 microwave-assisted digestion system (CEM
Corporation, Matthews, NC) at 200 ∘C for 13 min following US
EPA method 3052 (USEPA, 1995). Extracts were filtered
(0.45 µm PTFE) and analyzed for metals using a Thermo X-Series II quadrupole ICP-MS
instrument (Thermo Fisher Scientific Inc., Waltham, MA, USA) (Peate et
al., 2010). The instrument was calibrated against IV-ICPMS-71A ICP-MS
standard (Inorganic Ventures) at concentrations ranging from 0.1 to 50 ppb.
The metal concentration in the extract was converted to metal oxide
concentration in the sampled plumes (µg m-3) using extraction
volume, total filter area, sampled air volume, metal to metal oxide mass
ratio, and the natural metal isotope abundance (Rosman and Taylor,
1999). The uncertainty of the measurement was propagated using the method
detection limits, standard deviation of the field blank filters, and 10 %
of the metal concentration.
Organic species
Organic species were quantified in organic extracts of QFF by gas
chromatography coupled to mass spectrometry (GC-MS) as described in detail elsewhere
(Al-Naiema et al., 2015). In brief, QFFs were
subsampled to obtain ∼ 200 µg C prior to organic
species characterization. These subsamples were spiked with deuterated
internal standards which were used in quantification: pyrene-D10,
benz(a)anthracene-D12, cholestane-D4, pentadecane-D32,
eicosane-D42, tetracosane-D50, triacontane-D62,
dotriacontane-D66, hexatriacontane-D74,
levoglucosan-13C6, and cholesterol-D6. Each subsample
was then stepwise extracted in 2×20 mL aliquots of hexane followed
by 2×20 mL aliquots of acetone by ultra-sonication (60 sonics
min-1, 5510-Branson) for 15 min. The solvent extracts were
subsequently concentrated to a final volume of ∼ 100 µL
using Turbovap (Caliper Life Sciences, Turbo Vap LV Evaporator) and
micro-scale nitrogen evaporation system (Thermo Scientific,
Reacti-Vap™ Evaporator) upon high-purity nitrogen (PRAXAIR Inc.).
These extracted samples were stored at -20 ∘C until chemical
analysis.
Organic species in filter extracts were quantified using GC-MS (Agilent Technologies GC-MS 7890A) equipped
with an Agilent DB-5 column
(30 m × 0.25 mm × 0.25 µm) with electron ionization
source using a temperature program ranging from 60 to
300 ∘C. Helium was utilized as the carrier gas, and 3 µL
aliquots of the extracts were injected in splitless mode. Oxygenated compounds were analyzed following trimethylsilyl derivatization
(Stone et al., 2012). Briefly, 10 µL of the extract was blown
down to complete dryness and reconstituted in 10 µL of pyridine
(Burdick & Jackson, Anhydrous). A 20 µL of the silylation agent
N,O-bis-(trimethylsilyl)trifluoroacetamide (Fluka Analytical, 99 %) was
added to the mixture and was heated for 3 h at 70 ∘C to
complete the silylation reaction. The silylated samples were immediately
analyzed.
Responses of analytes were normalized to the corresponding
isotopically labeled internal standard and five-point linear calibration
curves (with correlation coefficients, R2≥0.995) were utilized
for the quantification of organic species. Compounds that were not in the
standards were measured by assessing the response curve from the compound
that was most analogous in structure and retention time. The analyte
concentration in the extract was converted to ambient concentrations
(µg m-3) using extraction volume, the total filter area, and
sampled air volume. The analytical uncertainties for the measured species
were propagated from the method detection limits, standard deviation of the
field blank filters, and 20% of the measured concentration, which is based
upon the spike recoveries of individual species being allowed to vary within
100±20 %.
Emission factor calculation
The mixing ratios of CO2, CO, CH4, and ∼ 90 other
gases were quantified by a field-deployed FTIR
spectrometer combined with whole air sampling (Stockwell et
al., 2016). The carbon mass balance approach was used to determine
fuel-based EFs for gases, in units of mass of analyte per
kilogram of fuel burned (g kg-1) (Stockwell et al., 2016).
Carbon monoxide was used as the reference species to calculate the EFs of
particulate species. For this purpose, carbon monoxide mass drawn through
the filter (MCO) that was measured in series by FTIR, the mass of the
analyte (MX; i.e., PM mass, EC, OC), and emission factor of carbon
monoxide (EFCO) were used to calculate the emission factors of the
desired analyte (EFX) using Eq.(1).
EFX=MXMCO×EFCO
Uncertainty in EFX was propagated from the relative uncertainty of
EFCO, conservatively estimated as 5 % of the value and the analytical
uncertainty of the considered analyte.
Modified combustion efficiency (MCE)
MCE was calculated as
MCE = ΔCO2/ΔCO+ΔCO2
and was used as an indicator of predominantly flaming
combustion (MCE > 0.9) and smoldering combustion (∼ 0.72–0.84) (Yokelson et al., 1996). Notably, the
filter-integrated MCE values reported herein correspond to the duration of
filter sample collection and could differ slightly from those reported by
Stockwell et al. (2016) that were interjected, unfiltered samples of the
same plumes.
Results and discussion
Emission of PM2.5
EFPM2.5 for in situ Indonesian peat burning ranged from
6.04 to 29.6 g kg-1 for 18 plumes, averaging (± standard deviation) 17.3±6.0 g kg-1
(Fig. 1). The percent difference across samples collected
sequentially from the same plume was 57, 37, and 8 % for plumes E,
F, and W, respectively, indicating some temporal variability in emissions
within the fire as it progresses. This in-plume variability in the field
could result from the spread and progression of the fire, consistent with
peat samples burned batch-wise in laboratory settings that show EFPM2.5
decreases on the timescale of hours during combustion (Black et al.,
2016). The overall relative standard deviation of EFPM2.5 in
this study was 35 %, indicating that variability across plumes is on the
same order as in-plume variability.
Emission factors of PM2.5, EC, OC, and water-soluble ions for
the average and individual peat smoke plumes. Error bars represent 1
standard deviation of the average or the propagated analytical uncertainty.
EFPM2.5 was dominated by OC (72 %) with minor contributions from EC
(< 1 %), ions (< 1 %), and metal oxides (< 0.15 %, not shown).
Comparison of the averaged in situ Indonesian peat emission data to prior
laboratory studies of peat combustion.
Peat location
PM
No. of
EFPM
OC
EC
WSOC %
of origin (and type)
Size
samples
(g kg-1)
(%) Mass
(% Mass)
of OC
OC:EC
MCE
Reference
Indonesia
PM2.5
21
17
72
1.1
16
60
0.78
This study
Indonesia
PM2.5
1
6.06a
99a
1a
–
151
0.838
Christian et al. (2003)
Indonesia
PM1
1
34.9
99b
0.03c
–
–
0.891
May et al. (2014)
Indonesia
PM10
1
33
24
1.7
39
14
–
Iinuma et al. (2007)
German
PM10
1
44
29
2.2
52
13
–
Iinuma et al. (2007)
North Carolina (ARNWR)d
PM2.5
4
7.1
89
0.73
–
122
0.89
Black et al. (2016)
North Carolina (PLNWR)e
PM2.5
4
5.9
73
1.4
–
52
0.88
Black et al. (2016)
North Carolina (ARNWR)d
PM2.5
4
48–66
–
–
–
–
0.79–0.86
Geron and Hays (2013)
North Carolina (PLNWR)e
PM2.5
4
35–55
–
–
–
–
0.77–0.83
Geron and Hays (2013)
North Carolina (Green Swamp)
PM2.5
4
44–53
–
–
–
–
0.80–0.81
Geron and Hays (2013)
Florida (sawgrass)f
PM2.5
6
30
–
–
–
–
–
McMahon et al. (1980)
Alaska (tundra core)
TSP
–
41.3
93.5
2.6
–
36
0.87
Chen et al. (2007)
a PM mass was not directly measured and was estimated as the sum of EC and
OC. b Measured as organic aerosol. c Measured as refractory black
carbon.
d Alligator River National Wildlife Refuge. e Pocosin Lakes National
Wildlife Refuge. f Corresponds to dry peat within the first 24 h of
combustion.
The average EFPM2.5 for Indonesian peat burning is within the range of
values reported in prior laboratory studies (6–66 g kg-1; Table 1).
Generally, the variability is attributed to natural variation within the
fuel (including its chemical composition (e.g., C content), moisture
content, and fuel density) and burn conditions
(e.g., extent of flaming
versus smoldering combustion) (Iinuma et al., 2007). The average
EFPM2.5 reported by Black et al. (2016) for two samples of North
Carolina peat (7.1±5.6 and 5.9±6.7 g kg-1)
are in the lower range of EFPM2.5 observed in this study. The peat
fires studied by Black et al. (2016) exhibited higher MCE values (0.80–0.88) compared
to this study (0.73–0.83), in which the former corresponds
to lower PM emissions (McMeeking et al., 2009) and may have resulted
from oven-drying their peat samples prior to combustion. Meanwhile, the
observed EFPM2.5 value for in situ Indonesian peat burning is lower than the
EFPM2.5 values reported by other laboratory studies: 46±21 g kg-1
by Geron and Hays (2013), 33–44 g kg-1 (for PM10)
by Iinuma et al. (2007), 42 g kg-1 by Chen et al. (2007),
35 g kg-1 by May et al. (2014), and 30±20 g kg-1 by
McMahon et al. (1980). These higher EFPM2.5 could be due to natural
variability in the peat composition and/or experimental variables such as
sampling from early stage of fires or extent of dilution. We also cannot
rule out that the smoke in some previous laboratory studies was concentrated
enough to increase gas–particle partitioning beyond the level in our
samples. Further, alterations to peat between the field and laboratory may
have an effect through the transporting and handling of peat soils,
differences associated with igniting the peat sample (e.g., heated coil
vs. propane torch), the edge effects due to igniting small chunks of peat, and
sustainability of the fire during the time of sample collection could also
affect the EFPM2.5. Because the EFPM2.5 computed during this
study corresponds to natural conditions of peat burning that was not handled,
transported, or processed disturbing the peat soil micro-properties, the
reported measurements are not subject to potential fuel alteration.
Emission of OC, EC, and WSOC
Across the studied plumes, EFOC ranged from 1.76 to 26.9 g kg-1,
averaging 12.4±5.4 g kg-1 (Fig. 1). The high OC mass fraction
of PM (72±11 %) is in a good agreement with literature reported
values of 73–89 % by Black et al. (2016) and 94 % by Chen et al. (2007) for PM2.5 from peat combustion in laboratory studies.
The EFEC ranged from 0.09 to 0.44 g kg-1, averaging 0.24±0.10 g kg-1 (Table 2). The high EFOC and low EFEC values are
consistent with purely smoldering combustion with MCE values of 0.725–0.833 as discussed by Stockwell et al. (2016). The optically measured
EFBC in PM1 by photoacoustic extinctiometry (PAX) (0.006±0.002 g kg-1) was noticeably lower than that of filter-based EFEC
likely due to sampling of char particles by filters, different PM size cuts,
and differences in measurement methods (Stockwell et al., 2016).
Overall, both optical and chemical measurement methods employed in Indonesia
and prior studies of EC in peat-burning emissions (Table 1) agree that
EFEC and EFBC are very small compared to EFOC.
Average emission factors for PM2.5, EC, OC, water-soluble
ions, metals (as mass fraction of PM2.5), and organic species
normalized to organic carbon mass. Individual EF data are given in Table S2.
Species
Study
Standard
average
deviation
EF PM2.5 (g kg-1)
17.3
6.0
EC (as mass fraction of PM2.5; g gPM2.5-1)
0.011
0.005
OC (as mass fraction of PM2.5; g gPM2.5-1)
0.72
0.11
Water-soluble OC fraction
0.16
0.11
Water-insoluble OC fraction
0.84
0.11
Water-soluble ions (as mass fraction of PM2.5; mg gPM2.5-1)
Sodium
0.054
0.065
Ammonium
5.1
3.0
Potassium
0.26
0.43
Fluoride
0.66
0.63
Chloride
4.2
2.4
Nitrate
0.16
0.13
Sulfate
1.41
1.42
Metals (as mass fraction of PM2.5; mg gPM2.5-1)
Fe
0.27
0.10
Cu
0.74
NA
Zn
0.40
NA
As
0.007
0.002
Cd
0.0002
0.0001
Ba
0.014
0.010
Pb
0.04
NA
Organic species (as mass fraction of organic carbon; mg gOC-1)
PAHs
Anthracene
0.0062
0.0036
Fluoranthene
0.036
0.017
Pyrene
0.056
0.031
Methylfluoranthene
0.043
0.021
Benzo(ghi)fluoranthene
0.0056
0.0029
Cyclopenta(cd)pyrene
0.0045
0.0022
Benz(a)anthracene
0.023
0.013
Chrysene
0.054
0.021
1-Methylchrysene
0.019
0.010
Retene
0.031
0.028
Benzo(b)fluoranthene
0.023
0.013
Benzo(k)fluoranthene
0.0036
0.0028
Benzo(j)fluoranthene
0.0031
0.0023
Benzo(e)pyrene
0.029
0.016
Benzo(a)pyrene
0.0081
0.0066
Perylene
0.0041
0.0034
Benzo(ghi)perylene
0.016
0.011
Dibenz(ah)anthracene
0.0098
0.0085
Picene
0.0139
0.0051
Hopanes
17α(H)-22,29,30-Trisnorhopane
0.344
0.058
17β(H)-21α (H)-30-Norhopane
0.85
0.13
17α(H)-21β(H)-Hopane
0.218
0.066
NA indicates not applicable, as the analyte was detected in two or fewer samples.
Continued.
Species
Study
Standard
average
deviation
n-Alkanes
Octadecane
0.39
0.46
Nonadecane
1.1
1.3
Eicosane
2.2
2.2
Heneicosane
3.8
2.8
Docosane
4.3
3.2
Tricosane
4.8
2.1
Tetracosane
4.1
2.2
Pentacosane
5.4
2.4
Hexacosane
4.1
2.1
Heptacosane
5.5
2.2
Octacosane
4.8
2.0
Nonacosane
6.5
1.9
Triacontane
4.7
1.4
Hentriacontane
6.7
1.4
Dotriacontane
3.03
0.52
Tritriacontane
2.83
0.54
Tetratriacontane
1.25
0.23
Pentatriacontane
0.66
0.15
Heptatriacontane
0.82
0.26
Octriacontane
2.5
1.3
Nonatriacontane
0.98
0.47
Branched alkanes
Norpristane
0.35
0.47
Pristane
1.0
1.2
Squalane
1.31
0.74
Anhydrosugars
Levoglucosan
46
40
Mannosan
0.93
0.76
Galactosan
0.14
0.13
Lignin decomposition products
Vanillin
0.030
0.044
Syringaldehyde
0.93
0.46
Vanillic acid
3.7
2.2
Syringic acid
1.69
0.91
Sterols
Stigmasterol
0.22
0.11
β-Sitosterol
0.53
0.34
Campesterol
0.29
0.20
The OC:EC ratio for in situ Indonesian peat burning ranged from 27 to 129, averaging
67±26. This is in the middle of the range of OC:EC values reported
previously for peat combustion (Table 1). The PAX results showed that the
ratio of light absorption at 405 nm relative to 870 nm wavelength was
approximately 50 (Stockwell et al., 2016), whereas a ratio close to 2.2 is
indicative of absorption by pure BC (Bond and Bergstrom, 2006). Thus,
the light absorption by peat smoke is largely due to BrC and the measured
high BrC:BC absorption ratio (52) is similar to the measured OC:EC ratio
(Stockwell et al., 2016). The bright yellow color of the PM
collected filters (Fig. S1) is also an indication of the light-absorbing
nature of the OC and a very small relative emission of EC.
The prior lack of information on light absorption by peat-burning emissions
could potentially limit the accuracy of direct radiative forcing estimates
in Southeast Asia (Ge et al., 2014). Previously, Ge et al. (2014) modeled
radiative forcing using OC:EC values up to 17. Our much larger OC:EC values
could imply that a more strongly scattering aerosol is relevant, depending on the
extent to which regional emissions are dominated by peat burning. In
addition, with new measurements of BrC presented in our companion paper
(Stockwell et al., 2016), the role of BrC in
direct radiative forcing should be evaluated in future assessments of this
kind.
Linear regression of the measured organic carbon (OC)
concentration versus the estimated organic matter (OM) concentration in
sampled plumes that was calculated as the difference between PM2.5 mass
and the sum of EC, water-soluble ions, and metal oxides.
On average, only a minor fraction of OC was water soluble (16±11 %)
and the majority (84±11 %) was water insoluble (Table 2).
Hence, the majority of OC is composed of hydrophobic organic compounds.
These results are consistent with prior observations of high relative
concentrations of aliphatic organic species in peat and peat-burning aerosol
reported previously (Iinuma et al., 2007; McMahon et al., 1980).
The low water solubility and presence of hydrophobic organic species likely
contribute to the hydrophobicity and low cloud condensation nuclei activity of fresh peat-burning
emissions (Dusek et al., 2005).
Chemical composition of PM2.5
OC accounted for the major fraction of PM2.5 (72±11 %) while
EC was detected in only 15 plumes and on average comprised 1.2 % of
PM2.5 (Table 1). Minor contributions to PM2.5 were observed for
water-soluble ions (1.2 %), and metal oxides (less than 0.1 %)
(Table 2). The sum of OC, EC, water-soluble ions, and metal oxide masses comprised,
on average, 74±11 % of gravimetrically measured PM2.5 mass.
The remaining PM2.5 mass is expected to be primarily from elements
associated with carbon in forming OM (e.g., O, H, N). Assuming
that no major chemical species were unmeasured, we estimate OM as the difference between PM2.5 mass and the sum of EC,
water-soluble ions, and metal oxides (OM = PM2.5–[EC + ions + metals oxides]).
The linear regression analysis of this
estimate of OM and measured OC correlated strongly (R2=0.93),
indicating their dependent co-variance (Fig. 2). The slope of the
regression line is 1.26±0.04 OM OC-1 and provides the conversion
factor of OC to OM for fresh peat-burning aerosols. This OC to OM factor is
in the range of values typically observed for gasoline combustion (1.1–1.3)
(Schauer et al., 2002, 1999) and below those used for other types of
biomass burning (1.4–1.8) (Reid et al., 2005), which is expected to
result from the semi-fossilized nature of the peat fuel and the
water-insoluble (Sect. 3.2) and aliphatic-rich (Sect. 3.5) nature of OC.
MCE
The calculated MCEs were indicative of smoldering combustions with values
ranging 0.725–0.833 (average = 0.78±0.04) (Yokelson et
al., 1996). Burn depth and MCE were negatively correlated (r=-0.738; p=0.001;
Fig. S2),
consistent with higher emission of CO(g) relative to CO2(g) for deep peat combustion, potentially due to
less oxygen supply. Over the small range of observed MCEs and for the purely
smoldering combustion, neither MCE nor burn depth were correlated with PM
mass, EC, or OC emission factors (p>0.23) and thus did not
noticeably affect PM emissions.
Organic species
A subset of samples (n=10), representing at least one sample per sample
collection site, was analyzed for anhydrosugars, lignin decomposition
compounds, alkanes, hopanes, PAHs, and sterols. On average, the quantified
organic compounds accounted for ∼ 9 % of the total OC mass
on a carbon basis with major contribution from alkanes (6.2 %), followed by
anhydrosugars (2.1 %), lignin decomposition products (0.36 %), hopanes
(0.12 %), sterols (0.06 %), and PAHs (0.03 %) (Fig. 3). Up to
approximately 5% more of the OC is expected to come from n-alkenes, some
oxy-PAH, additional lignin decomposition products, and nitrophenols that
were measured in peat emissions by Iinuma et al. (2007). The remaining OC
remains unresolved and is likely to include isomers of the abovementioned
compounds (e.g., branched alkanes) and high-molecular weight organic
compounds. Plume Y that was obtained from shallow peat-burning sites with
plant roots observed in the burn pit had a different emission profile with a
larger contribution from anhydrosugars (16 %) compared to lignin
decomposition products (2.8 %) and alkanes (1.6 %). Plume Y thus
represents the co-burning of peat with surface vegetation and was excluded
from average calculations that represent subsurface burning of peat. The
full emission profile for each individual plume is reported in Table S2.
Organic carbon mass fraction of the speciated compound classes in
selected peat-burning emission samples. Plume Y was excluded from the
average calculation as discussed in Sect. 2.1.
Alkanes
The homologous series of n-alkanes and select branched alkanes were
quantified in emissions from Indonesian peat burning. The n-alkanes with
carbon numbers ranging C18–C34 were detected in all samples
analyzed by GC-MS,
with higher-carbon number homologues observed in many
samples (Table S2). The n-alkane emission factor (EFalk) for the
quantified species ranged 456–3834 mg kg-1 (Table S2).
On average, n-alkanes contributed 6.2 % of OC mass. This OC mass fraction is
consistent with results from Iinuma et al. (2007) for Indonesian and German
peat burning and is remarkably higher than other types of biomass burning OC
for which this OC fraction is typically less than 1 % (Hays et
al., 2002; Iinuma et al., 2007). The high n-alkane contribution to OC
results from the high lipid content of peat that accumulates from plant
waxes (e.g., cutin, suberin) during decomposition (Ficken et al.,
1998). The in situ source emissions and prior measurements of peat combustion in the
field (Fujii et al., 2015a) and in the laboratory (Iinuma et
al., 2007) agree that n-alkanes can be used to distinguish peat emissions
from other types of biomass burning and other combustion sources by their
high contribution to particle-phase OC.
The most abundant n-alkane (Cmax) was consistently observed for the
C31 carbon homolog (Table S2). This is the same Cmax value
observed by Iinuma et al. (2007) for Indonesian peat, while in ambient air
impacted by Indonesian peat burning, Fujii et al. (2015a) and bin Abas et
al. (2004) reported Cmax at C27. This variability in Cmax
likely derives from in the peat material but may be influenced by
atmospheric aging as the differences in Cmax are aligned with fresh and
aged peat-burning aerosol.
Molecular distribution of n-alkanes for selected plumes (n=10).
Y axis indicates individual n-alkane mass fraction of OC. The horizontal
lines (black) in the box represent the 25th, 50th (median), and
75th percentiles and mean values are indicated by the blue lines.
As shown in Fig. 4, n-alkanes demonstrated a slight odd carbon preference
that is indicative of biogenic material, particularly plant waxes
(Fine et al., 2002; Oros and Simoneit, 2001a, b; Baker, 1982). The carbon
preference index (CPI) was calculated using concentrations of C24-32 n-alkanes
following Fujii et al. (2015a) and ranged 1.22–1.60,
averaging 1.42±0.10. Comparable CPI values have been reported
previously for laboratory emissions from peat collected in Indonesia (1.5),
Germany (1.8) (Iinuma et al., 2007), and North Carolina (1.4–1.5)
(George et al., 2016). These CPI values are low in comparison to
emissions from foliage, softwood, and hardwood combustion emissions that
range 1.6–6.2 (Hays et al., 2002; Yamamoto et al., 2013).
Together, the high n-alkane mass fraction and CPI values of 1.4±0.2 are
characteristic features of Indonesian peat fire emissions.
Organic carbon mass fractions of select anhydrosugars for study
average and selected individual plumes. Plume Y was not included in average
calculation as discussed in Sect. 2.1. On average, the galactosan mass
fraction was 0.14 mg gOC-1 (maximum = 0.77 mg gOC-1); due to its
low concentrations, it was not included in the plot.
Anhydrosugars
Pyrolysis of cellulose and hemicellulose generates anhydrosugars, of which
levoglucosan, mannosan, and galactosan were quantified. Anhydrosugar EF
(EFanh) ranged 157–2041 mg kg-1 and averaged 543±598 mg kg-1.
The dominant anhydrosugar was levoglucosan (averaging 46±40 mg gOC-1), followed by mannosan (0.93±0.76 mg gOC-1) and
galactosan (0.14±1.13 mg gOC-1) (Fig. 5, Table 2).
Levoglucosan was the most abundant individual species quantified and
contributed 0.3–6.0 % of OC mass (Table S2). A significant correlation was
not observed between EFOC and EFlevoglucosan (p=0.4) in
contrast to Sullivan et al. (2008), who observed the correlation of these
values for biomass burning emissions from grass, duff, chaparral, softwood,
and hardwood fuels (R2=0.68) . The variable cellulose content across
peat soils likely contributes to this lack of correlation.
While relative ratios of levoglucosan, mannosan, and galactosan have been
used to distinguish between various types of biomass combustion emissions
(Engling et al., 2014), peat-burning emissions did not exhibit
consistent ratios of these species. The levoglucosan to mannosan ratio
ranged widely from 27 to 160 with an average (± standard deviation) of
55±41. Meanwhile, Iinuma et al. (2007) reported this ratio to be 11
and Fujii et al. (2015a) reported it to average 15. Because of the
variability across studies and the expected dependence of this ratio on
biomass cellulose content and composition (Sullivan et al., 2008),
this ratio is insufficient to distinguish peat combustion from other biomass
types.
Lignin decomposition compounds
S, V, SA, and VA
derived from lignin pyrolysis were quantified (Fig. 6), with a combined EF ranging
15–154 mg kg-1 and averaging 80±50 mg kg-1 (Table S2).
Correlations among aldehydes (V and S) were not significant, possibly due to
V partitioning to the gas phase, as indicated by its detection on backup
filters, whereas other species (S, VA, and SA) were detected only on front
filters indicative of particle-phase species. We examined the potential of
the VA:SA ratios to be useful in distinguishing this source from other types
of biomass burning, since VA:SA depends on the lignin composition of the
biomass (Simoneit et al., 1999). A significant moderate correlation
was observed between EFVA and EFSA (R2=0.65; p=0.004).
Based on linear regression analysis, VA:SA was found to be 1.9±0.2
for freshly emitted peat smoke in this study (Fig. 7). This value agrees
well with observations of VA:SA in PM2.5 in Malaysia affected by
Sumatran peat fires, which had a VA:SA ratio of 1.7±0.4 (Fujii et
al., 2015b) and the ratio of vanillyl phenols to syringyl phenols of
2.0 reported for Kalimantan peat (Orem et al., 1996).
Meanwhile, other studies indicate lower VA:SA ratios for near-source
emissions of Sumatran peat burning (1.1±0.4) (Fujii et al., 2015a)
and laboratory burning of South Sumatran peat (0.11) (Iinuma et al.,
2007). Because other biomasses in South Asia have VA:SA that fall in this
range, such as bamboo (1.17) and sugar cane (1.78) (Simoneit et al.,
1999), this ratio is unlikely to be useful in distinguishing peat burning
from other types of biomass burning in the absence of other distinguishing
chemical or physical properties. Further, syringyl compounds degrade more
quickly in peat compared to vanillyl compounds (Orem et al.,
1996) and post-emission SA degrades more quickly than VA by photolysis in
the atmosphere, such that VA:SA is likely to increase with smoke transport
(Fujii et al., 2015b). Consequently, this ratio has limited utility in
source identification and apportionment.
Organic carbon mass fraction of measured lignin decomposition
products for study average and selected individual plumes. Plume Y was not
included in average calculation as discussed in Sect. 2.1.
PAHs, hopanes, and sterols
PAHs were observed in emissions from
Indonesian peat burning and the 18 PAHs that were quantified are listed in
Table 2. For the measured species, EFPAH ranged 1.7–17 mg kg-1 and
were consistent with previously reported EFPAH values, 6–25 mg kg-1
for laboratory peat-burning studies (Black et al., 2016; Iinuma
et al., 2007). PAH composition was dominated by pyrene, chrysene,
methylfluoranthene, fluoranthene, and retene, which accounted for
∼ 56 % of the measured PAH emissions (Table 2). Several
biomass burning studies have reported retene, a biomarker of softwood
combustion, as the most abundant PAH in wood smoke (Fine et al., 2002;
Hays et al., 2002; Schauer and Cass, 2000), whereas it contributed only
8 % of the measured PAH in this study.
Emission ratios of vanillic acid to syringic acid.
Benz(a)anthracene, benzo(a)pyrene (B[a]P), benzo(b)-fluoranthene,
benzo(k)fluoranthene, chrysene, and dibenz(a,h)anthracene, which are
categorized as probable human carcinogens by the US Environmental Protection
Agency (USEPA, 2008), were detected in peat-burning aerosols and
together these PAHs accounted for 39 % of total quantified PAH species.
The toxic equivalency factor was estimated for quantified PAHs to estimate
the overall human health hazard level (Nisbet and LaGoy, 1992).
The estimated B[a]P
equivalent toxicity value ranged 0.05–0.39 mg kg-1,
averaging 0.13±0.10 mg kg-1, and was comparable
to previously reported toxicity values for peat smoke, such as 0.12–0.16 by Black et
al. (2016). The total PAH concentration in undiluted peat smoke ranged
0.3–18 µg m-3 and was similar to PAH concentrations reported for
exhaust smoke of a coke oven (25 µg m-3), aluminum smelting
(15 µg m-3), diesel engines (5 µg m-3), and gasoline
engines (3 µg m-3) (Khalili et al., 1995; Armstrong et
al., 2004).
To the best of our knowledge, hopanes have not been previously quantified in
peat fire emissions. 17α(H)-22,29,30-Trisnorhopane, 17β(H)-21α (H)-30-norhopane, and 17α(H)-21β(H)-hopane
were identified using authentic standards and quantified in pure peat smoke
for the first time. EFhopanes ranged 11–37 mg kg-1, averaging
17±8 mg kg-1 (Table S2). Terpenoid and hopanoid hydrocarbon
compounds that have the hopane skeleton are ubiquitous in peat soils
(Ries-Kautt and Albrecht, 1989; Venkatesan et al., 1986; Quirk et al.,
1984; López-Días et al., 2010; Del Rio et al., 1992; Dehmer, 1995).
Thus, the presence of hopanes in peat smoke is not unexpected. Norhopane had the
highest OC mass fraction, followed by trisnorhopane and hopane (Table 2). A
fairly consistent ratio of 0.25:0.60:0.15 was observed among trisnorhopane,
norhopane, and hopane irrespective of the sampling site and burning depth,
indicating the formation of hopanes is independent of burning conditions
(Fig. S3). The observed hopane ratio is clearly distinct from that of
diesel (0.04:48:48) (Schauer et al., 1999) and non-catalytic gasoline
(0.10:0.42:0.48) (Schauer et al., 2002) engine emissions. However,
it is comparable to the hopane ratio of lignite (0.23:0.66:0.11) and
subbituminous (0.29:0.49:0.22) coal smoke (Oros and Simoneit, 2000).
This indicates similarities of terpenoid and hopanoid hydrocarbons in peat
soils and coal deposits and these are younger on the geological timescale
than crude oil.
Stigmasterol, β-sitosterol, and campesterol were detected in peat
smoke and accounted for 0.14–1.7 mg gOC-1 of OC mass fraction (Table S2).
Sterols have been identified in peat soils with a major contribution
from β-sitosterol (Del Rio et al., 1992;
López-Días et al., 2010). Similarly, β-sitosterol is the
predominant sterol in PM (Table 2), indicating the emission of peat
constituents to the atmosphere as PM during smoldering.
Water-soluble inorganic ions
Water-soluble ions accounted for only 1.1 % of the PM mass and total
quantified EFions ranged 45–490 mg kg-1, averaging 201±144 mg kg-1. Ammonium and chloride were detected in all the samples
with average EFs of 92±61 mg kg-1 and 75±52 mg kg-1, respectively. Frequency of detection for sulfate, nitrate,
and fluoride was 83, 61, and 56 % and EFs ranged 2–133, 0.2–6.8, and 0.4–45.9 mg kg-1, respectively. PM
mass fractions of ammonium vs sulfate (r=0.95, p<0.001) and
ammonium vs. chloride (r=0.89, p<0.001) were strongly correlated
indicating that a major fraction of inorganics in PM is in the form of
(NH4)2SO4 and NH4Cl. The molar concentrations of gaseous
NH3 and NO + HONO were 33 times and 312 times higher than that of
NH4+ and NO3-, respectively, consistent with a dominance
of gas-phase precursors in fresh peat-burning emissions (Stockwell et
al., 2016). The atmospheric oxidation of NO and HONO could increase the
concentration of NO3- (Gankanda and Grassian, 2013;
Gankanda et al., 2016), while acid–base reactions convert NH3 to
NH4+, thus leading to increased concentrations of these secondary
inorganic products in aged peat smoke.
Potassium has been used as an indicator of biomass burning, both on its own
and in concert with levoglucosan (Simoneit et al., 1999; Sullivan et al.,
2008; Chuang et al., 2013; Gao et al., 2003). From peat smoldering fires,
extremely low potassium emissions (0.03 % of PM mass) were observed, at
concentrations too low to be a useful indicator species, as described by
Sullivan et al. (2014) and Fujii et al. (2015a).
Metals
Metal species accounted for a maximum of 0.15 % of the PM mass and their
EF ranged from below the detection limit to 28 mg kg-1 in plume P
(which had the highest PM mass loading on the filter; Table S1). The metal
fraction was dominated by Fe, Cu, Zn, and Ba, many of which have been
previously observed in peat soil (Dizman et al., 2015). The lower
EFmetal values relative to other quantified species (i.e., OC) indicate
the minimum influence of re-suspended soil dust to PM2.5. Further,
combustion at temperatures lower than 400 ∘C, indicative of
smoldering conditions, precludes metal transfer to the aerosol phase
(Raison et al., 1985; Usup et al., 2004).
Emission estimates from 2015 Indonesian peat fires
The emissions from Indonesian peat fires during the 2015 El Niño were
estimated using mean EFs calculated in this study for an estimated burned
area of 8.5× 105 ha (Whitburn et al., 2016), an
average burning depth of 34±12 cm (calculated during this study; Stockwell et al., 2016), and a peat bulk density of 0.120±0.005 g cm-3 (Konecny et al., 2016). The uncertainty of the
estimated value is propagated using standard deviation of the mean EFs, burn
depth, and peat bulk density. However, the uncertainty of burned area is not
defined.
Estimated emissions from Indonesian peat fires during 2015 El
Niño, based on a burned area of 8.5×105 ha (Whitburn et
al., 2016), an average burning depth of 34±12 cm (Stockwell et al.,
2016), and peat bulk density 0.120±0.005 g cm-3 (Konecny et al.,
2016). The uncertainty of the estimated value is propagated using standard
deviations of the mean EFs, burn depth, and peat bulk density.
Species
Total estimated emission
C-mass based
Mass based
(TgC)
(Tg)
PM2.5
–
6.0±5.5
C-containing compounds
OC(PM2.5)
4.3±4.3
–
EC(PM2.5)
0.083±0.081
–
CO2(g)a
149±71
547±259
CO(g)a
44±30
102±69
CH4(g)a
2.5±2.6
3.3±3.5
Other C-containing trace gasesa
5.5±1.3
9.3±2.6
Total C
205±77
–
Water-soluble ions in PM2.5
NH4+
–
0.032±0.039
Cl-
–
0.026±0.032
NO3-
–
0.0010±0.0013
SO42-
–
0.0096±0.0151
Other atmospheric gases
NH3(g)a
–
1.00±0.91
HCl(g)a
–
0.012±0.014
NO(g)a
–
0.11±0.17
HONO(g)a
–
0.073±0.061
a EFs are based on Stockwell et
al. (2016).
In this way, the total PM2.5 released to the atmosphere from this
fire event was estimated to be 3.2–11 Tg, averaging 6.0±5.5 Tg with
major contribution from OC (4.3 Tg) followed by EC (0.08 Tg) and
water-soluble ions (0.07 Tg) (Table 3). Combining our OC and EC emission
factors with gas-phase EFs of CO2, CO, CH4, and other carbon-containing gases from Stockwell et al. (2016), we estimate a total carbon
emission of 205±77 TgC to the atmosphere, of which 73 % was as
CO2 (149±71 TgC), 21 % as CO (44±30 TgC), 1.2 % as
CH4 (2.5±2.6 TgC), 2.7 % as other carbon-containing gases
(5.5±1.3 TgC), 2.1 % as OC (4.3±4.3 TgC), and 0.04 % as EC
(0.083±0.081 TgC). Our carbon emission estimates are in good
agreement with Huijnen et al. (2016) who estimated total C emissions of
227±67 TgC for this fire event. However, this is ∼ 8
times lower than the carbon emissions estimated for the 1997–1998 Indonesian
peat fires (810–2570 TgC) (Page et al., 2002).
Conclusions
PM2.5 was collected from authentic in situ peat smoke during the 2015 El
Niño peat fire episode in Central Kalimantan, Indonesia, and was
chemically characterized for PM mass, EC, OC, water-soluble ions, metals,
and organic species. Fuel-based EFPM2.5 ranged from 6.0 to
29.6 g kg-1,
averaging 17.3±6.0 g kg-1, and we estimate 3.2–11 Tg
of PM2.5 was released to the atmosphere during the 2015 El Niño
peat fire episode. OC accounted for the major fraction of PM mass while EC,
water-soluble ions, and metal oxides comprised only a minor fraction of PM
mass. Combining our EFOC and EFEC with gas-phase EFs of CO2,
CO, CH4, and other carbon-containing gases from Stockwell et al. (2016),
we estimate a total carbon emission of 205±77 TgC to the
atmosphere. OC and EC comprised 2.1 and 0.04 % of total carbon
emissions, respectively.
Overall, chemical speciation of OC revealed the following characteristics of
peat-burning emissions: high OC mass fractions (72 %), primarily
water-insoluble OC (84±11 %, low EC mass fractions (1 %), and
relatively high n-alkane contributions to OC (6.2 %C) with odd carbon
preference CPI (1.2–1.6). This chemical profile is in good agreement with
prior studies of Indonesian peat burning using laboratory measurements
(Christian et al., 2003; Iinuma et al., 2007) and ambient aerosol
studies in Indonesia (Fujii et al., 2015a, b) as well
as laboratory studies of peat emissions from other locations (Black et
al., 2016; Geron and Hays, 2013; Chen et al., 2007). The similarities of the
peat-burning chemical profiles determined in this in situ emissions characterization
and prior and laboratory studies reveal that laboratory studies can
accurately capture the fractional composition of PM and OC. However, greater
discrepancies arise in the absolute EFPM2.5 emissions (Table 1) across
field and laboratory studies, with the former typically yielding lower
EFPM2.5 values. Differences in EFPM across studies are expected to
result from several factors, such as fuel composition and moisture content,
combustion conditions, and timing and concentration of PM sampling.
Knowledge of chemical characteristics of peat emissions can be used in
source identification and apportionment modeling at a receptor site that is
impacted by peatland fire emissions. Further, they can allow for assessment
of acute and chronic hazards associated with exposures to high
concentrations of PM and PAH from peat smoke during the fire season
(Armstrong et al., 2004; Kim et al., 2013).
The quantitative emission factors developed in this study for Indonesian
peat burning are the most representative of natural peat-burning conditions
and may be used to update regional and global emission inventories which are
currently based on EFs computed from laboratory studies. The most recent
emission inventory compiled by Akagi et al. (2011) does not include an EF
value for PM2.5 for peat fire emissions. Further, the EFOC
reported in Akagi et al. (2011) is 50 % lower than the average EFOC
observed in this study, which would underestimate the PM2.5 OC
emissions observed in the field. Thus, the use of these in situ EFs in updates to
emission inventories can provide more accurate emission estimates. Moreover,
more studies should be carried out downwind to evaluate the effects of
atmospheric dilution and atmospheric photochemical reactions on the chemical
composition of peat fire PM.