A Biomass Combustion Chamber : Design , Evaluation , and a Case Study of Wheat Straw Combustion Emission Tests

Open biomass burning is a significant source of trace gases and particulate pollutants on a global scale and plays an important role in both atmospheric chemistry and climate change. To study the emission characteristics of biomass burning, with a focus on crop residue combustion in Northwest China, a combustion chamber was established. This paper describes the design, structure, and operating principles of the chamber. A series of evaluation tests were conducted, demonstrating its applicability in emission studies. The combustion chamber was equipped with a thermoanemometer and a dilution sampler as well as multiple sampling ports for interfacing with different monitors. A case study of wheat straw combustion was performed to demonstrate reproducibility and comparability of the derived emission factors with prior studies. The combustion chamber may be applied to develop emission factors to update emission inventories and source profiles for improving source apportionment.


INTRODUCTION
Biomass burning includes the open burning of crop residues, forest fires, and grassland fires.It is a large source of many trace gases and fine particulate matter (PM, usually measured as PM 2.5 , particles with aerodynamic diameters < 2.5 µm) that cause adverse health hazards, visibility impairment, and other environmental impacts on regional and global scale (Crutzen and Andreae, 1990;Andreae, 1991;Penner et al., 1992;Duan et al., 2004;Li et al., 2010;Huang et al., 2012b;Malamakal et al., 2013).In China, ~180 terragram [Tg, 10 12 grams] of biomass was openly burned in the mid-1990s, emitting ~280 Tg of carbon dioxide (CO 2 ), ~16 Tg of carbon monoxide (CO), ~0.82 Tg of reactive nitrogen oxides (NO x ), ~0.08 Tg of sulfur dioxide (SO 2 ), ~0.23 Tg of ammonia (NH 3 ), ~0.54 Tg of methane acquire EFs and source profiles (e.g., Yokelson et al., 2008;Watson et al., 2012), but it limits the ability to evaluate a large number of fuels and burning conditions (Lobert et al., 1991;Lobert and Warnatz, 1993).Laboratory tests complement field measurements that allow for: 1) use of more complex instrumentation to measure a larger range of pollutants; 2) better control of dilution and cooling to account for condensation and near-source chemistry; 3) separation of emissions from specific burning phases, especially flaming and smoldering; 4) comparison of different fuels and fuel conditions, such as moisture content and soil content, that can guide more efficient burning (McMeeking et al., 2009;Chen et al., 2010); and 5) evaluation of emission uncertainties through replicate tests (Yokelson et al., 2008;Burling et al., 2010).
An open burning simulation chamber was established to acquire multi-pollutant emission rates and source profiles of crop residue combustion at the Institute of Earth Environment, Chinese Academy of Sciences (IEECAS).Combustion of wheat straw, a major form of crop residues in North/Northwest China, and its reproducibility is demonstrated.

THE BIOMASS COMBUSTION CHAMBER
Previous simulation facilities and their applications are evaluated to optimize the design, as summarized in Table S-1 (Supporting Information).Laboratory tests use small quantities (approximately 10 g to 13.6 kg) of fuel burned under controlled combustion conditions with portions of the exhaust directed to various continuous and integrated (e.g., filters, canisters, and adsorbents) sampling systems.
The following combustion chamber features were identified and incorporated: 1) a tilting stage to evaluate heading and backing fires [i.e., flames moving with and against the wind, respectively]; 2) provision for diluting and aging of samples prior to collection to allow for vapor condensation and fast chemical reactions (Hildemann et al., 1989;Lipsky and Robinson, 2005;Watson et al., 2012;ISO, 2013); 3) control of combustion air and outlet flows to simulate flaming and smoldering phase of combustion; and 4) availability of numerous sampling ports and sufficient throughput to accommodate multi-pollutant measurements.
The schematic of the combustion chamber is illustrated in Fig. 1.The 3 mm thick aluminum is used to withstand high combustion temperatures with dimensions of 1.8 m (L) × 1.8 m (W) × 2.2 m (H) and a volume of ~8 m 3 .No rubber, plastics, greases, or oils are used to minimize organic contamination.Valves on the top and bottom of the chamber allow for chamber sealing and control of combustion air and updraft speeds.The combustion air inlet is preceded by High Efficiency Particulate Air (HEPA) filters (3M Company, St Paul, MN, USA) with nominal filtration efficiencies of 95% for > 0.3 µm PM, and an air blower.Fuels of 50-300 g (depending on study objectives) are placed on the fuel tray during tests, and ignited by a butane pilot light.A ~0.15 m 2 fuel tray is affixed to 0.6 m (L) × 0.6 m (W) × 0.55 m (H) stage at the center of the chamber that can be adjusted to different angles to simulate different wind directions (i.e., heading and backing fires).An electric fan mixes the smoke within the chamber.This is accompanied by a lamp inside the chamber (with switch outside) allowing the fire to be observed and recorded through a window.The initial fuel and final residual masses are weighted using a balance with a resolution of ± 0.1 g.
Exhaust is drawn from the chamber through a 0.15 m diameter exhaust duct by a venting fan with adjustable speed.The thermoanemometer (960 probe, TSI Inc., Shoreview, MN, USA) measures gas velocity, temperature, and pressure in the duct.Sampling ports in the duct allow for extraction of the exhaust into various devices, including the dilution sampler (Model 18, Baldwin Environmental Inc., Reno, NV, USA).Three non-dispersive infrared CO 2 sensors (PP Systems, Amesbury, MA, USA) are used to measure CO 2 concentrations in the undiluted effluent, dilution (ambient) air, and diluted plume at 1-sec resolution, respectively.Dilution ratios range from 0 to 40.The smoke residence time in the dilution tunnel is 4-11 s to allow for condensation, coagulation, and rapid reactions before sampling.Tests were made to evaluate: 1) the enclosure seal; 2) uniformity of exhaust flow and pollutant concentrations across the exhaust duct; 3) stability of the air flow; and 4) PM levels in the combustion and dilution air.
To perform a leak test, the top and bottom valves were closed and the electric fans inside the chamber were switched on.Approximately 30 g of biomass, loosely packed on the stage, was burned to produce CO 2 .CO 2 concentration was continuously monitored by a Q Trak (TSI Inc., Shoreview, MN, USA) located inside the chamber.Fig. 2 shows a rapid increase of CO 2 , ~1200 s after ignition, from the background level of ~510 ppm to a maximum of ~3200 ppm (1500 s; fired died before 1500 s).After the burn was completed, the CO 2 concentration declined by only ~2% after more than 30 min.The valves, venting fan, and blower were opened at ~3300 s, the CO 2 concentration reduced to below ~50% of the peak concentration within 1 minutes achieved back ground levels after another 9 min.
Sampling probes may draw the effluent from different parts of the exhaust duct, which is valid if there is a homogeneous distribution across the duct's cross-section.An inertial droplet separator (IDS) nozzle, sampling the smoke isokinetically (Baldwin et al., 2012), was inserted into one of the sampling ports at various distances from the duct's centerline (Fig. 3(a)), with CO 2 concentrations varied by < 5% from the average concentration of 1240 ppm (Fig. 3(b)).Flow velocities, measured with a thermoanemometer at five different positions, also reported < 5% variation from the average velocity of 3.6 m s -1 (Fig. 4).
Fig. 5 shows the stability of CO 2 concentrations with the venting fan and blower turned on.Background CO 2 was about 550 ppm, while a 100,000 ppm CO 2 stream was introduced into the combustion chamber.The CO 2 concentration increased gradually to a steady-state level of ~919 ppm with a standard deviation (SD) of 24 ppm, ~240 s after injection.
Since average ambient PM 2.5 concentrations in Xi'an achieve 375 µg m -3 in winter and 131 µg m -3 in summer owing to contributions from biomass burning, coal combustion, and vehicle exhaust (Cao et al., 2005(Cao et al., , 2007)),  both combustion and dilution air needs to be filtered.The combustion air introduced into the chamber is preceded by High Efficiency Particulate Air (HEPA) filters and an air blower.Fig. 6 shows the variation of PM 2.5 concentration measured with a DustTrak (Wang et al., 2009) before and after HEPA filtration with a dilution ratio of ~5.The dilution air is generated by a compressor, and filtered by actived carbon and a HEPA filter (Wang et al., 2012a).Filtration reduced PM 2.5 levels from 55 µg m -3 to 2 µg m -3 , corresponding to a removal efficiency of > 95%.For other pollutants, such as CO, NO x , and SO 2 , the background concentration can be measured to correct for emission calculation.

EMISSION FACTORS FOR WHEAT STRAW COMBUSTION
Wheat straw was collected from Hebei Province, China, and stored at ambient temperature (~20°C) and humidity (35 to 45%) for at least one month before the experiment.Its characteristics are summarized in Table 1.The wheat straw (~100 g) and ashes were weighed before and after each burn.For each of the five replicate burns, wheat straw was stacked horizontally on the fuel tray (Fig. 1), and ignited with a butane pilot lighter.Each burn lasted ~30 min from ignition until concentrations of measured pollutants (i.e., CO 2 , CO, and PM 2.5 ) returned to the background levels.Dilution ratio was ~5 for each test.PM 2.5 samples were collected on three parallel 47-mm filters (1 Teflonmembrane filter (2 µm pore size, R2PJ047, Pall Life Sciences, Ann Arbor, MI, USA), and 2 quartz microfiber filters (QM/A, Whatman, Midstone, Kent, England) of the dilution sampler at a flow rate of 5 L min -1 per channel.13.64 * Proximate and Ultimate analysis were conducted to determine the carbon (C) and nitrogen (N) content in dry mass.In addition, moisture, volatile matter, ash, and fixed carbon content were measured.See details in Liao et al. (2004).
Gravimetric analysis was conducted on the Teflon membrane filters after 24-hr equilibrium at ~25°C and 35% relative humidity using a microbalance with a ± 1 µg sensitivity (Sartorius, Göttingen, Germany).Total carbon (TC), organic carbon (OC) elemental carbon (EC), and their thermal carbon fraction were determined on quartz fiber filters with a Desert Research Institute (DRI) Model 2001 Thermal/Optical Carbon Analyzer (Atmoslytic Inc., Calabasas, CA, USA) following the IMPROVE_A thermal/ optical protocol (Chow et al., 2007).Real-time measurements of CO 2 , CO and PM 2.5 were made at 1-s resolution.
Emission factors were calculated by dividing the mass of pollutant released by the mass of the consumed fuel, and expressed as grams of emission per kilogram of dry fuel consumed (g kg -1 ) (Andreae and Merlet, 2001).For CO 2 and CO, the time-integrated EF i is: and for particulate pollutants (i.e., PM 2.5 , OC and EC), the EF p is: where C Dil,i is the diluted concentration of pollutant i in mg m -3 under standard conditions (temperature [T Std ] = 293 K and pressure [P Std ] = 1 atm); t sample is the sampling duration in seconds; V Stk is the average stack flow velocity in m s -1 at standard conditions; D is the stack cross section in m 2 ; m fuel is the fuel consumption in g determined from the difference of initial and final fuel weights; m filter is the mass of pollutants collected on the filter in mg; Q filter is the sampling volume through the filter in m 3 at standard temperature and pressure (Wang et al., 2012a).DR is the dilution ratio, controlled by the flow balance (i.e., the ratio of sample flow (from stack) and total inflow (sample flow + dilution flow)) of the dilution sampler, where: Total Inflow or Outflow Flow-Based DR Sample Flow  (3) Total inflow equals total outflow, which is the sum of flows through all the filter packs, online monitors, and make-up flow.DR can also be verified by CO 2 concentrations (CO 2 -Based DR) measured at the exhaust duct (undiluted CO 2,Stk ), sampler (diluted CO 2,Dil ), and background (ambient CO 2,Bkg ).Thus: The agreement of the two DR estimates (Flow-Based and CO 2 -Based DR) indicates that either method is sufficient to obtain an accurate EF.
Emissions vary between different burning phases (e.g., flaming versus smoldering), which can be differentiated by combustion efficiency (CE) -the ratio of carbon (C) emitted as CO 2 to the total amount of C emitted (Ward and Hardy, 1991).When only CO 2 and CO are monitored, CE can be simplified as the modified combustion efficiency (MCE), defined as: When ∆CO 2 and ∆CO exceed the molar mixing ratio, MCE indicates the relative importance of flaming and smoldering combustion (Koppmann et al., 2005;Chen et al., 2007).MCE is typically close to 1 during the flaming phase, and ranges between 0.7 and 0.9 for the smoldering phase (Hao and Ward, 1993;Yokelson et al., 1997;Reid et al., 2005).
Table 2 shows the PM 2.5 mass collected on the three parallel filters from each of the five replicate wheat straw combustion tests.The consistent results between both types of filters and among replicates, as indicated by the low relative standard deviations (RSD) of PM 2.5 mass (< 10%), demonstrate the capability of the system and reproducibility of the results.The mass on Teflon filter was always more than that on quartz fiber filters.PM 2.5 samples were collected on three parallel 47-mm filters (1 Teflon-membrane filter and 2 quartz microfiber filters) of the dilution sampler at a flow rate of 5 L min -1 per channel.So theoretically, the mass on these three filters from each test should be the same.However, it becomes complex when considering the positive and negative sampling artifacts.Adsorption of organic vapors onto the quartz filters leads to overestimation of the PM 2.5 mass load (positive artifact), while volatilization of the collected PM from the filter results in the underestimate of the PM 2.5 mass load (negative artifact) (Turpin et al., 2000).Positive and negative artifacts occur simultaneously and hence are difficult to isolate and quantify.Previous studies report that the positive artifact appears to dominate on samples taken with bare quartz filters (Turpin et al., 1994;Kirchstetter et al., 2001), so it is reasonable to expect more mass on quartz filters than on Teflon.However, quartz filters are known to be fragile and may be prone to fiber loss during weighing, handling, and sampling, which may cause weight loss on quartz filter.The mass difference between quartz and Teflon filters ranged from 8.6% to 14.5%, which falls into a reasonable range considering the sampling and handling uncertainty.
The homogeneity of PM 2.5 filter deposits was also tested by dividing the quartz filter from one channel into four quadrants for separate carbon analyses.Average carbon concentration are presented in Table 3, good reproducibility was found with RSD ranging from 3-9% for TC, 3-9% for OC, and 1-10% for EC.
Response to combustion emissions by the real-time instruments is usually delayed due to the time it takes to travel from the sample probe to the sensor and the residence time in each instrument's sensing volume.The delays vary with instrument and need to be accounted for in order to align instrumental responses for post-sampling data processing (Wang et al., 2012b).Instruments were first sampled with ambient air for a few minutes to obtain background readings.Then a match was lit near the inlet of the sampling probe to measure time delays.Table 4 shows the delays determined for each instrument to detect a 10% change from background concentration.Delay times of 11-16 s were determined relative to the DustTrak, since the DustTrak responds first (i.e., the delay time of DustTrak was assumed to be 0).

CONCLUSION
A combustion chamber was designed and tested, and a series of tests were conducted to verify its performance in terms of the enclosure seal against gas/particle loss, flow stability and uniformity in the stack, and purification of the dilution air.In this study, the combustion chamber was equipped with a dilution sampler and real-time instruments for measuring multi-pollutant emissions (e.g., CO 2 , CO, PM 2.5 , OC, and EC).Data obtained from wheat straw burning, with a focus on its emission factors and combustion efficiencies, were reported to demonstrate the reproducibility of the results and comparability with prior studies.Biomass of interest in Northwest China can be burned in the combustion chamber to develop emission factors for refining emission inventories and source profiles for more accurate source apportionment.

Fig. 1 .Fig. 2 .
Fig. 1.Schematic of the combustion chamber.The smoke are first diluted by a dilution sampler (Model 18, Baldwin Environmental Inc., Reno, NV, USA), then collected and sampled by three parallel filter packs, three CO 2 sensors (PP Systems, Amesbury, MA, USA), a CO analyzer (Thermo 48i, Thermo Scientific Inc., Franklin, MA, USA), and a DustTrak (Model 8543, TSI Inc., Shoreview, MW, USA).The listed flow rates are for operation with a flow-based dilution ratio of 5 (see Eq. (3) in the text) which can be adjusted by changing the dilution air and make-up flows.The undiluted, diluted, and ambient CO 2 concentrations measured by corresponding CO 2 sensors

Fig. 3 .Fig. 4 .Fig. 5 .
Fig. 3. (a) Gas and effluent velocity measurement locations across the sampling duct; and (b) CO 2 concentrations corresponding to locations in (a), demonstrating concentration homogeneity and that sampling at a single location provides a reasonable estimate of emission rates and composition.

Fig. 6 .
Fig. 6.PM 2.5 mass concentrations before and after purification by High Efficiency Particulate Air (HEPA) filters.

Table 2 .
PM 2.5 mass on each filter (mg) for the five replicate wheat straw combustion tests.Average, standard deviation (SD), and relative standard deviations (RSD) for each test; *** Average, SD, and RSD for the five replicate tests.With the stable system, experiments have a good reproducibility.

Table 3 .
Average concentrations of total carbon, organic carbon, element carbon, and their thermal fractions measured from each quadrant of sample filter (µg cm -2 ) for the five replicate wheat straw combustion tests.

Table 4 .
Delay time for each instrument.

Table 5 .
Comparison of emission factors in wheat straw burning among studies.