Environ. Sci. Technol. 2008, 42, 8829–8834
Chemical, Microphysical and Optical Properties of Primary Particles from the Combustion of Biomass Fuels G A Z A L A H A B I B , †,| C H A N D R A V E N K A T A R A M A N , * ,† TAMI C. BOND,‡ AND JAMES J. SCHAUER§ Department of Chemical Engineering, Indian Institute of Technology Bombay, Powai, Mumbai - 400076, India, Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Newmark Civil Engineering Laboratory, MC-250, Urbana, Illinois 61801, and Environmental Chemistry and Technology Program, University of Wisconsin-Madison, Madison, Wisconsin 53706
Received April 8, 2008. Revised manuscript received August 28, 2008. Accepted September 15, 2008.
Biomass fuel combustion for residential energy significantly influences both emissions and the atmospheric burden of aerosols in world regions, i.e., east and south Asia. This study reports measurements of climate-relevant properties of particles emitted from biomass fuels widely used for cooking in south Asia, in laboratory experiments simulating actual cooking in the region. Fuel burn rates of 1-2 kg h-1 for wood species, and 1.5-2 kg h-1 for crop residues and dried cattle dung, influenced PM2.5 emission factors which were 1.7-2 g kg-1 at low burn rates but 5-9 gkg-1 at higher burn rates. Total carbon accounted for 45-55% and ions and trace elements for 2-12% of PM2.5 mass. The elemental carbon (EC) content was variable and highest (22-35%) in particles emitted from low burn rate combustion (wood and jute stalks) but significantly lower (2-4%) from high burn rate combustion (dried cattle dung and rice straw). The mass absorption cross-section (MAC, m2 g-1) correlated with EC content for strongly absorbing particles. Weakly absorbing particles, from straw and dung combustion, showed absorption that could not be explained by EC content alone. On average, the MAC of biofuel emission particles was significantly higher than reported measurements from forest fires but somewhat lower than those from diesel engines, indicating potential to significantly influence atmospheric absorption. Both for a given fuel and across different fuels, increased burn rates result in higher emission rates of PM2.5, larger organic carbon (OC) content, larger average particle sizes, and lower MAC. Larger mean particle size (0.42-1.31 µm MMAD) and organic carbon content, than in emissions from combustion sources like diesels, have potential implications for hygroscopic growth and cloud nucleation behavior of these aerosols. These measurements can be used to refine regional
* Corresponding author e-mail:
[email protected]. † Indian Institute of Technology Bombay. ‡ University of Illinois at Urbana-Champaign. § University of Wisconsin-Madison. | Current address: The Indian Institute of Technology Delhi, New Delhi, India. 10.1021/es800943f CCC: $40.75
Published on Web 10/28/2008
2008 American Chemical Society
emission inventories and derive optical parametrizations, for climate modeling, representative of regions dominated by primary particles from biomass fuel combustion.
1. Introduction Globally, fossil fuel combustion for energy in the industrialized world, open biomass burning of forests and grasslands and biofuel combustion for energy are the largest sources of anthropogenic combustion particles. Particle emissions from several sources like diesel and gasoline vehicles (e.g., refs 1, 2,), power plants and furnaces (e.g., ref 3,), tropical forest burning (e.g., refs 4, 5,) and savanna or grassland and crop residue burning (e.g., refs 5-7,) have been studied in regard to their microphysical, chemical, and optical properties needed for air quality and climate studies. The relative predominance of particle emitting combustion sources varies significantly in different world regions. Biomass fuel combustion for cooking and residential energy is an important combustion source in developing world regions (8, 9) and has largely been studied through measurements of health damaging pollutants like carbon monoxide and particles (10, 11) and their mutagenic constituents (14, 15) in several laboratory studies and fewer field studies. Significant burden of disease, in the form of respiratory and heart ailments, tuberculosis, and blindness, is attributed to the exposure to these emissions suffered by biomass fuel user populations both globally and particularly in developing world regions (16). Biomass fuel combustion for energy accounts for significant emissions of total and carbonaceous aerosols in east Asia, sub-Saharan Africa, and south Asia (8, 9). The large influence of biomass fuel emissions on the atmospheric burden of carbonaceous aerosols in east and south Asia has recently been demonstrated through model simulations (17, 18). A limited amount of information on particle chemical composition is available for biomass fuel emissions in early studies from Africa (19) and more recently south Asia (20); however, these measurements have not included the suite of climate relevant physicochemical and optical properties. A recent field study of biomass fuel emissions in Honduras (21) measured much larger emissions than was previously measured in laboratory studies. Real-time particle composition and absorption varied in different combustion phases, with emissions of very dark particles containing large amounts of elemental carbon during strong flaming events. It was suggested that representing biomass fuel combustion in global or regional emission inventories would need an understanding of the variables that control particle emission factors and properties such as EC/PM ratio and light absorption. In an attempt to address this knowledge gap, and in the absence of the opportunity for widespread field studies, the experiments in this study were designed to measure climaterelevant properties of aerosol emissions from biomass fuels widely used in traditional cooking stoves in south Asia. The laboratory experiments were conducted by simulating fuel burn rates and fuel charging practice likely to occur in actual cooking in the region (22). Measured properties include PM2.5 emission factors, particle size-distribution, elemental and organic carbon content (EC/PM2.5 and OC/PM2.5 ratio), and particle absorption. The dependence of climate relevant aerosol properties on fuel, stove and combustion parameters is investigated. VOL. 42, NO. 23, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8829
2. Materials and Methods 2.1. Combustion Experiments and Particle Sampling. Combustion experiments were carried out using four species of wood, six types of crop waste, and dried cattle dung (tabulated in Supporting Information (SI) Table S1), procured from the parts of India where they are widely used as cooking fuels (22). In addition, two commonly used cooking fossil fuels, kerosene and liquid petroleum gas (LPG), were included in this study for comparison. A U-shaped, single-pot, mud stove (SI Figure S1a), most frequently used in rural India (22), was used in this study. Emissions were diluted and entrained through a hood into a duct by an induced draft fan. A sample was isokinetically extracted into a dilution sampler developed and optimized in previous work (11) (details in SI Figure S1b). The gas stream from the dilution plenum was drawn through a cyclone-inlet particle sampler (SI Figure S2) collecting particles smaller than 2.5 µm aerodynamic diameter (PM2.5), on five different filter substrates for downstream analysis. Emission factors per kg of fuel burnt were calculated (see the SI). The dilution ratio for the direct combustion exhaust, estimated from temperature measurements of the combustion and diluted gas streams, ranged 10-20 for most and 25-40 for a few experiments. Higher fan speeds were avoided as they increased the excess air rate into the stove, placed directly below the hood, resulting in dilution and reduction of combustion temperature. A residence time of 200-300 s in the dilution plenum (11) allowed for near-complete condensation of semivolatile compounds. Estimated combustion gas flow rates through stoves are of 0.001-0.005 m3 h-1 in this work. Typical air exchange rates in rural households are 5-10 m3 h-1 (10, 16) suggesting that dilution ratios could be about 100 times larger in actual indoor environments, even in the absence of complete mixing. Recent literature suggests that larger dilution ratios close to 5000 are needed to dilute automotive exhaust to near ambient conditions (23) in dilution tunnels with residence time on the order of 1-2 s. The lower dilution ratios in this work could result in enhanced partitioning of semivolatile organic compounds to the particle phase, than would occur under natural dilution in typical indoor environments. 2.2. Measurement of Aerosol Chemical Constituents and Absorption. The particles were analyzed for carbonaceous aerosol (elemental carbon, EC, and organic carbon, OC) using a thermal-optical-transmittance (TOT) carbon aerosol analyzer (model 3, Sunset Laboratory, OR) at the Nano Chemistry Laboratory, at University of California Los Angeles (5) and repeated at the University of Wisconsin Madison, using the same protocol, to evaluate interlaboratory variability. The repeat analysis agreed within 28% for total carbon and within 45% for OC for all samples, but the variability was higher for EC ranging 35-100%, with larger uncertainty in the high EC values. A fraction of OC pyrolyzes to EC, and evolves at higher temperatures. This fraction, which is subtracted from the total EC, is sensitive to choice of split time. As EC is significantly less than OC, the amount that could be misclassified creates a larger fractional uncertainty in EC than in OC. Differences in split time (SI Figure S3) therefore influenced the EC amounts in the outlier samples. The results from the repeat analysis were used to calculate the geometric mean and standard deviation on EC-OC filter loading. Note that these mean values were used only for composition estimates reported here. The optical properties were measured separately as described in next section. Particles collected on Teflon filters (Teflo 47 mm diameter, Pall Corporation) were analyzed for ionic species by ion chromatography and trace elements by inductively coupled plasma mass spectroscopy (ICP-MS) at the University of Wisconsin Madison (details in SI, Section A2). Samples 8830
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 23, 2008
FIGURE 1. Scatter plot of (a) PM2.5 emission factors and burn rate (b) EC/PM2.5 ratio and burn rate. Errors bar are one standard deviation around mean indicate the variability among species in each biomass fuels category. Error bar on fibrous stalk, rice straw and dried cattle dung were derived from the variability estimated for average emission factor and burn rate of woods-LBR. collected on Nuclepore filters (0.2 µm pore size) were weighed to determine mass emissions, which ranged from 100-500 µg for various biomass fuels, and analyzed with an integrating plate (24) for absorption at four wavelengths (435, 525, 660, and 800 nm) by a method detailed in SI Section A3. As concurrent scattering measurements were not made, an estimate of scattering was used to correct the absorption artifact at each wavelength (detailed in SI Section A3). Doubling the estimated scattering would change the absorption by only about 5%, making this a minor source of uncertainty, unlike ambient aerosol where scattering is the dominant component of extinction.
3.0. Results and Discussion 3.1. PM2.5 Emissions. The measurements in this study are composite particle properties emitted from different biomass fuels, burned in a manner that would represent regional cooking practice in south Asia. Measured emissions factors of PM2.5 (g (kg of fuel burnt)-1) from burning four species of wood (at two different burn rates), dried cattle dung, and six types of crop residues increased with increasing fuel burn rate. Emission factors were lowest (Figure 1a) from woods burned at low burn rates (abbreviated hereafter as woodsLBR) (1.9 ( 0.8 gkg-1) and from jute stalks (1.7 ( 0.7 gkg-1) and higher from woods burnt at high burn rates (abbreviated hereafter as woods-HBR) (5.1 ( 1.4 gkg-1), dried cattle dung (5.4 ( 2.4 gkg-1), woody crop residue stalks (7.5 ( 3.3 gkg-1), and rice straw (9.3 ( 4.1 gkg-1). The error bars are one standard deviation of the mean of experiments among wood species and woody crop residue stalks. For fibrous hollow stalks, dried cattle dung and rice straw, the error bars show the maximum uncertainty in group of woods or woody stalk which is 44% for PM2.5, 48% for EC, and 24% for OC. Comparing these PM2.5 emissions to previously reported total particle emissions is reasonable as biomass combustion
emissions are expected to be largely in the small particle size range (11). Different fuels, burn practices and measurement techniques were used in the various studies (10-13), which would affect the comparison. Previously reported emission factors from wood stoves lie between the LBR and HBR measurements in this work, whereas those from crop residues and cattle dung lie in the upper end of the range measured here. There is a fair match with emission factors measured in a recent Honduran field study (25), though lower average values in this study reflect variability from differences in burn rate, fire-tending, and measurement methods. For similar burn rates, cattle dung had lower mean emissions than woody crop residues and woods-HBR had lower mean emissions than straw. Proximate and ultimate analysis of the fuels, made in a commercial laboratory using standard methods for the composition of solid fuels (SI Figure S4 and Section A4) do not show significant differences in composition except for dried cattle dung, which had higher mineral matter than other fuels. Thus, fuel chemical composition differences do not appear to explain differences in particle emission factors. The moisture content of biomass fuels was measured as percent weight loss on a dry basis at 110 °C until fuels attained constant weight. The moisture content in this and previous works (10, 11) was relatively constant (5-6%), and therefore could not explain differences in emissions. However, physical differences exist among the biomass fuels. Rice straw has the lowest density and highest surface to volume ratio, followed by woody crop residues, dried cattle dung, with highest values for wood species and hollow fibrous jute-stalks. A high surface to volume ratio would result in a larger rate of fuel devolatilization. Further, low density fuels, like straws and crop residue stalks, tended to fill the stove combustion chamber volume more than high density fuels, like woods and cattle dung, providing a higher resistance to air flow by natural convection through the fuel bed. Both factors would lead to emissions of unburned organics that can condense into particulate organic matter. Such reasoning is consistent with the observation of relatively white-smoky emissions from straw and woody crop residues, especially at the start of the fire. Thus the higher emissions from straw and woody crop residues, than woods-HBR and cattle dung, respectively, likely result from their shape, density and tendency to fill the stove combustion chamber, all leading to more incomplete combustion. Fuel physical parameters and stove combustion chamber size influence particle emissions, whereas no link to fuel chemical composition could be established. 3.2. PM2.5 Bulk Chemical Composition. While EC and OC emissions factor from these biomass fuels have been reported earlier (9), this work reports new measurements of PM2.5 bulk chemical composition, including carbonaceous, ionic, and elemental constituents and their dependence on combustion parameters. Total carbon (elemental carbon, organic carbon, and carbonate carbon together) was 47-55% of PM2.5, being lowest in dried cattle dung and highest in wood emissions. However, the relative amounts of EC and OC as individual species varied significantly as discussed in a following section. Inorganic ion content was highest in emissions from dried cattle dung (10%), and 2-5% in emissions from the other fuels. Trace element content was low (
