Wind Tunnel Modeling of Atmospheric Emissions from Agricultural Burning: Influence of Operating Configuration on Flame Structure and Particle Emission Factor for a Spreading-Type Fire Bryan M. Jenklns,*lt Ian M. Kennedy,* Scott Q. Turn,t Robert B. Wllllams,t Steven G. Hall,? Stephen V. Teague,s Daniel P. Y. Chang,il and Otto G. Raabes
Department of Biological and Agricultural Engineering, Department of Mechanical, Aeronautical, and Materials Engineering, Institute for Toxicology and Environmental Health, and Department of Civil and Environmental Engineering, University of California, Davis, California 95616 Agricultural burning is a significant source of atmospheric pollutants throughout the world and has lately been subject to increased regulation in North America and Europe. A combustion wind tunnel having multiple operating configurations was developed to simulate field burning conditions and to provide direct measurement of emission factors. Sensitivity to operating configuration was determined for total suspended particulate matter (TSP) emissionsand flame structure. TSP emission factor varied from 0.505 to 0.727% of fuel mass as a result of controlled changes to the inlet flow velocity and turbulence characteristics and to ventilation through the fuel bed. Flame structure was examined through measurements of local temperatures, gas concentrations, and soot volume, the latter obtained by laser-light extinction. Although the range in TSP emission factor was not large, wind speed appeared to be the dominant parameter affecting the emission of particles. Results of this study will be used to determine operating protocols for future tests investigating criteria and toxic emissions from agricultural burning.
Introduction In the last decade, legislation has been enacted in California allowing emission offset credits to facilities that burn agricultural and forest biomass (crop and forest wastes) for the generation of steam and electricity. This legislation recognizes the potential for reducing air pollution when agricultural wastes are used as fuel for power plants rather than being burned in the field. Farmers burn approximately 1.7 X lo6dry t of crop wastes annually in the state (1, 2). Emissions of particulate matter and carbon monoxide are 2-3 times higher than those from the state’s electric utilities, and hydrocarbons and sulfur
* Author to whom correspondence should be addressed. t
Department of Biological and Agricultural Engineering.
* Department of Mechanical, Aeronautical, and Materials Engi-
neering. 5 Institute for Toxicology and Environmental Health. 11 Department of Civil and Environmental Engineering. 0013-936X/93/0927-1763$04.00/0
0 1993 Amerlcan Chemical Society
dioxide are of the same order. Numerous crop wastes are burned, but the most important contributions to air pollution come from rice straw, almond and walnut tree prunings, and wheat straw (2). More recent legislation restricts the amount of rice straw which can be burned, but retains the emission offset allowance for facilities that burn it. The procedure for determining the emission offset credits uses quantities expressing the mass of pollutant emitted per unit mass of fuel burned (the emission factors). Emission factors were developed using a burning tower by Darley (3-5) for criteria pollutants from a number of crop types important to California. Emission factors have also been reported in the compilation by the U.S.EPA (6). Some uncertainty exists in these laboratory results because of the high temperatures maintained in the sampling duct above the fire and because of the small quantity of fuel burned. Field studies have also been conducted (7-9). The primary disadvantage of field studies is the lack of direct knowledge of the mass of fuel burned. This parameter is usually computed by carbon balance from COZmeasurements in the plume of the fire. Wide ranges in reported emission factors are typical. Because of the inherent difficulties in field measurements, a wind tunnel technique was developed to quantify emission factors under more controlled conditions and to extend the range of compounds analyzed ( 1 0 , l I ) . The wind tunnel is an open circuit, forced draft type consisting of a fuel feed table, inlet section, combustion test section, and stack sampling section. A schematic of the wind tunnel appears in Figure 1. To extend the sampling time when burning spreading-type fires, a system of conveyors is used to translate the fuel bed downstream at a rate equal to the upstream fire spreading rate. In the inlet flow development section, this is accomplished by means of a flexible belt, which is exposed for loading on the fuel feed table and ends at the junction with the combustion test section. The fuel is transferred to a stainless steel rod conveyor at the front of the combustion test section, and the fuel is burned on this conveyor. The rod conveyor is located above a refractory brick floor and unloads into an ash bin at the end of the combustion test Environ.
Scl. Technoi., Voi. 27, No. 9, 1993 1763
-1 -re
m -4.8
9.8m+
7
.
4
m
4
1. Exterior schematic of combustion wind tunnel.
above the fire or by installing a sheet metal floor directly beneaththe fuelbedto block ventilation from below. With these two adjustments, four configurations were possible as follows: (1) ceiling extended downstream to the front or leading edge of the fire, with no floor installed between the fuel bed and the refractory brick (designated CENF for ceiling extended, no floor): (2) ceiling retracted to the front wall of the combustion test section, with no floor installed (designated CRNF for ceilingretracted,no floor); (3) ceiling extended to the front edge of the fire, with the floor installed beneath the fuel bed (designated CEWF for ceiling extended, with floor); and (4) ceiling retracted UnP.rm.lAW
Fhun 2. Schematic of wind tunnel conveyor system.
section. Ash is removed from the bin by a motor-driven auger emptying into a collection bag. A movable ceiling extending horizontally into the combustion test section from the front wall can be positioned anywhere above or ahead of the fire to aid in shaping the inlet velocityprofile. The inlet section is 1.2-m square, and the ceiling is located at a height just above the top of the inlet section. A schematicof the conveyor and ceiling arrangement appears in Figure 2. Stack samples are collected with a trackmounted traversing platform located near the top of the stack (Figure 1). The stack section tapers upward in two stagesto a final 1.2-m square cross section. The sampling location is approximately 9.8 m above the rod conveyor and is located two duct diameters downstream and onehalf duct diameter upstream from the nearest flow disturbance. This report summarizes results of 64 experiments conducted to investigate the influence of different wind tunnel operating configurations on the flame structure and particulate emissions. Results are presented on total suspended particulate (TSP) emissions, particle size distributions, flame temperatures, average soot volume fractions, local gas concentrations, inlet velocity profiles and turbulence intensities, and general operating conditions. Allexperimentswere condudedwithfiresspreadmg in opposition to the approach airflow as is currently required under existing regulations for field burns. Wind Tunnel Conjiguratiom The operating configuration of the wind tunnel can be altered by adjusting the position of the movable ceiling TIS4
E n W m Scl. Tschnd.. Vol. 27. No. 9, 1993
tothefrontwall,withthefloorinstalled(designatedCRWF for ceiling retracted, with floor). Two wind speeds were also tested-a low wind speed of about 2 m s-l and a higher wind speed of 3-3.5 m s-’-wbere the wind speed is the superficial velocity in the inlet duct, i.e., the volumetric flow rate divided by the duct cross sectional area.
Materials and Procedures For each operating configuration, measurements were made to deserihe the approach flow, flame structure, and particulate emissions. The approach flow velocityprofiles and turbulence characteristics were determined using a twewire x-probe-type hot-wire anemometer. Total suspended particulate matter (TSP) was determined using an in-stack sampling protocol. Particle size distribution was determined with a cascade impactor, also used instack. Flame structure was analyzed by measuring flame temperatures, flame radiation, and local gas concentrations. Average line-of-sight soot volume concentrations were also determined by laser-light extinction through the flame and postflame regions. Fuel loading rate waa kept constant, and fuel moisture content and residual ash weight were determined for each experiment. Inlet air temperature and relative humidity, conveyor speed, stack gas velocity,and stack gas temperature were continuously recorded. Due to instrument limitations, not all meaaurementa could be made simultaneously, and repeated tests were conducted with all controlled variables held constant. Measurements of TSP and operating parameters were always taken, the TSP results serving as a means of evaluatingrepeatability. Each test was replicated. Atotal of 64 experiments were conducted: 19 in the CRNF
configuration, 17 in the CENF configuration, 15 in the CEWF configuration, and 13 in the CRWF configuration. Inlet Velocity Profiles and Turbulence Characteristics. A two-axis, x-probe-type hot-film anemometer coupled to two anemometer signal conditioners was used to measure the inlet flow field along the vertical center line at a position approximately 120 mm upstream of the front or leading edge of the fire. Here, the leading edge of the fire at the fuel surface is used as a primary reference position. A second vertical traverse was made approximately 600 mm upstream of the leading edge of the fire. This position was designated the upstream position, while the first position closer to the flame was designated the downstream position. The absolute positions of the probe were in all cases similar because the fire was held at the same position in the tunnel. At the downstream position with a fire present in the tunnel, the flow could not be measured near the fuel surface because of probe heating and possible burnup. Velocity measurements were made in cold flow without the fire present, with the probe at the same absolute positions, and in this case descending to the fuel surface at the downstream position. The probe was oriented to measure both the streamwise and vertical velocity components of the flow. The airflow rate was adjusted by changing the vane angle and effective aperture on the inlet vanes of the blower. The anemometer was routinely calibrated over a range of 0.15-5.5 m s-l. Analog outputs from the signal conditioners were converted to digital form by a 12-bit A/D board and stored on a computer. The anemometer probe was mounted on a three-axis positioning system capable of traversing the height and width of the combustion test section and 470 mm in the streamwise direction. At each location, 8500samples from each channel were acquired at 2.2 kHz for approximately 3.9 s. The sampling rate was selected to be in excess of the Nyquist criterion based on the frequency spectrum of the flow. For the velocities tested, the highest observed frequency was 800 Hz. The sampling time was adequate to yield average velocities deviating less than 1%from those obtained with sampling times greater than 5 s. Air temperature at the probe tip was measured simultaneously and used to correct the velocity for differences in test and calibration temperatures. Each vertical traverse included 15-18 sampling locations and required approximately 30 min to complete with the transfer of data to hard-disk storage between samples. Replicated traverses were made at each position for each of the configurations, each of the two wind speeds, and with and without a fire present in the wind tunnel. Mean velocities and locl turbulence intensities were computed at each sampling location for both the streamwise and vertical directions. Frequency spectra were derived from fast Fourier transforms (FFT) of the velocity data. Measures of the integral length scales of wind tunnel turbulence were derived from autocorrelations of the same data. Total Suspended Particulate Matter and Particle Size Distribution. Emission factors for total suspended particulate matter were used to measure gross changes in emission characteristics as a result of changes in the wind tunnel operating configuration. TSP was determined using a California Air Resources Board Method 17 instack sampling protocol (12). Particle size distribution was determined using a cascade impactor in the stack.
The stack cross section was divided into 12 equal rectangular areas each 305 mm X 406 mm. Samples were drawn at the center of each area. The TSP and cascade impactor probes, along with a type-K thermocouple for monitoring stack gas temperature and a hot-film anemometer for monitoring stack gas velocity, were carried on a 100mm X 75 mm rectangular steel beam, which could be extended into the tunnel from the side wall. The beam itself was carried on a rail at the wall and could be moved along the wall. By moving the beam longitudinally and extending it transversely, the probes could be positioned at any point in the stack. A sheet steel curtain moving with the beam kept the stack wall closed against gas leakage at all probe positions. The TSP probe consisted of a 12.7mm stainless steel tube extending through the beam to which a 47-mm stainless steel filter holder and stainless steel buttonhook nozzle were attached. A Teflon tube was attached between the outer end of the stainless tube and the inlet of an impinger train. The impingers were connected through a check valve to an air-tight sample pump, dry test meter, and mass flowmeter. A thermocouple was used to continuously record impinger outlet temperature. The system was leak-checked before each test. A vacuum gage was connected ahead of the pump to monitor filter pressure drop. The standard S-type pitot tube was replaced with a calibrated hot-film anemometer because the flow dynamic pressure was too low to accurately gage gas velocity. The stack velocity was monitored by the operator at the sampling platform and was also continuously recorded. This anemometer was calibrated when clean and checked after operating in the stack flow. The probe retained calibration within 0.3% full scale over 64 tests. The cascade impactor was mounted in a manner similar to the TSP probe and also employed a stainless steel buttonhook nozzle. The outlet of the cascade impactor was connected to a stainless steel tube passing through the carrier beam, and by Teflon tubing to a desiccant column, an air-tight sample pump, a dry test meter, and an orifice meter. This system was also leak-checked before each test. Samples were collected isokinetically through the probes. Typical flow rates were 10 L min-l for the TSP probe and 20 L min-' for the impactor. TSP samples were collected on weighed, desiccated, glass fiber filters. Coated stainless steel membranes were used as collection surfaces in the cascade impactor. Eight stages were employed. Total sample time was 48 min, giving 4 min at each traverse location. Time-Temperature Histories of Flame. Timetemperature histories of the fires for each of the four configurations and each wind speed were made by burning past a stationary vertical array of thermocouples. In the case of the ceiling-extended configurations, the ceilingwas manually adjusted to follow the fire. Exposed bead type-K (chromel-alumel) thermocouples made from 125-pm diameter wire were used. The time constant of this wire was too large (approximately 150ms) to measure true flame temperatures, but gave representative values of mean temperature. Thermocouples were placed at the fuel surface and vertically above at 127-mm intervals. Thermocouples were sampled at 30 Hz using the same data acquisition system described for the x-probe anemometer. Temperature data were later compiled into contour and surface plots displaying approximate flame shape. Environ. Sci. Technol., Vol. 27, No. 9, 1993
1765
Local Flame Temperatures. A fast-response type R (Pt/Pt+l3% Rh) thermocouple was used to record local flame temperatures. This thermocouple was of the exposed bead type, made of 25-rm diameter wire with an estimated time constant of 5 ms. The wire was strung through a ceramic and stainless steel support, which was inserted through the wall of the tunnel into the flame. Each set of data was collected over 6 s a t 1kHz using the same data acquisiton system described above for the x-probe anemometer. Relative frequency densities of temperature were computed from each data set. No radiation correction was applied to these temperature measurements, although total flame radiation was measured from a position above and in front of the fire using a Schmidt-Boelter-type heat flux transducer. Assuming a Nusselt number, Nu = 2, and thermocouple emittance varying from 0.2 (clean) to 0.95 (soot covered), the temperature error due to radiation would be in the range of 2C-100 K. Local Gas Concentrations. Local gas samples were collected from the flame region using a 9.5-mm-0.d. stainless steel water-cooledprobe. The probe was inserted throughaslotinthetunnelwall,andthefmewastranslated on the conveyor to keep the probe in position during the sampling period. The gas sample was drawn through a 1.5-mm-diameter stainless steel tube running inside the water cooled probe. Initially, a static pressure tap located just inside the probe nozzle was to he used to adjust the flow for isokinetic conditions. This was later abandoned in favor of a mam flow transducer at the outlet of the sample line, which was used to set sample flow rate from velocity measurements near the sampling location. The outlet of the gas sampling line was connected through a Teflon line and glass fiber filter to a Teflon-lined sample pump, impinger, and 2-mL glass sample collector. The sample line was purged for 3 min at each location before taking the sample. Typical flow rates ranged from 0.1 to 0.25 L min-' depending on height. Sampleswere analyzed by gas chromatography for permanent gases and light hydrocarbons. Soot Volume. Line-of-sight average soot volume fractions were computed from measurements of laser-light extinction through the flame. An argon ion laser (488 nm wavelength, 50 mW maximum output power) was used in conjunction with a photodiode to monitor laser-light intensity on the opposite side of the tunnel. An interference filter and neutral density filter were placed ahead of the photodiode to restrict wavelength and attenuate intensity. All optical elements were protected from ambient light. The laser and photodiode were located on anexternalU-shapedmounting platformextendingaround the walls and under the floor of the tunnel. The laser and photodiode were moved vertically together by lifting the open end of the platform on a roller carriage. The photodiodeoutput was filtered to providearelativelylong time constant (on the order of 1-2 s) to obtain better averages of light extinction. The output signal was acquired on the same data acquisitionsystem used for the x-probe anemometer. The light beam was passed through a vertical series of 16-mm-diameter holes in the wind tunnel wall. AU holes not in use were kept plugged. Sampling heights were 76, 127, 178, 229,330,432,533, and 686 mm above the rod conveyor. Scans were conducted by holding the conveyor stationary and allowing the fire to burn past the laser. vea
-.
w. ~ m vol..27. . No. 0. m
3
:,.
._
I
I
ngv.3. TraMmlsslonslecbanmlaoqaph0fsootpemdescolected a poM-
from Ib R a m regkm. 63 mm above thn fuel suface. on tc-plane electmsta* preclpkator. The h g l h of thn refrepresents 250 nm.
bar
This technique proved more consistent than attempting to hold the fire stationary a t each sampling location. Samples were acquired a t 1 Hz,with a total scan duration of approximately 240 8 at each height. Average soot volume fraction across the tunnel width was determined using a Rayleigh formulation in which light extinction depends on path length, wavelength of the light, refractive index of the particles, and particle diameter and number density, the latter two parameters yielding the soot volume. In this ease, the soot volume, 0, was found as
+ = 8.18 X lo4 In(IJI)
(1) where I is attenuated light intensity at photodiode, Io in unattenuated light intensity at photodiode, and the refractive index used was that of Dalzell and Sarofii (13). To examine the assumption that light attenuation was principally due to soot absorption, a series of particulate samples were collected from the flame and postflame regions using a point-to-plane electrostatic precipitator. These samples were examined by transmission electron microscopy (TEM). Large numbers of soot particles in the size range of 10-30 nm were found in the flame regions (Figure 3). Ash particles were present in the postflame region, but soot still appeared to be the dominant form. Operating Conditions. Inlet air temperature was monitored by a thermistor located between the flow straightening screens in the inlet section and later by a type-T thermocouple placed near the inlet of the combustiontestsection. Inlet relative humiditywasmonitored near the screens by a sulfonated polystyrene humidity transducer. These sensors were connected to a separate electronic data logger with digital values stored on computer. Also connected to the data logger were the stack and impinger thermocouples,the stack anemometer, and a sensor for measuring conveyor speed. Conveyor travel (total displacement) was obtained by integrating conveyor speed over time. Fuel consumption rate was obtained from conveyor travel. All sensors were scanned at 1-min intervals, with the exception of the stack anemometer (scanned at 1Hz and averaged over 1min)
1400 7
200 I
I
I
I
.e
I
I
I
0 ,
I
1400
1000 -.
E
g Ea
-P
800-.
.
: 1
@
e 600-
I
I
I
I
I
I
l
* b
400F=T-200 0
I
~
-*
L i
4 0
20
-4
40
80
60
Local Turbulence Intensity ( O h ) (b)
Flgure 5. (a) Mean streamwise velocity profile. (b) Mean turbulence intenslty for the CRNF low wind speed configuration.
intensities up to around 30 % were fcund at a height of 200 mm above the conveyor or 136 mm above the mean fuel surface. Lower elevations were not probed because of the rough nature of the fuel surface and the potential for probe damage. Turbulence measurements with x-probes are well-known to be unreliable very close to rough surfaces at any rate (14). Freestream turbulence intensities were about 1%.The CRNF low wind speed configuration was characterized by a region of low and nearly constant velocity just above the fuel surface, which could be the result of an adverse pressure gradient leading to a separated flow in this configuration. Local turbulence intensities
were higher in this region, largely as a result of the low air speed. The velocity profiles were little affected by the presence of absence of a floor. At the lower speeds, retracting the ceiling produced a more erect flame as the flow decelerated into the combustion test section. Vertical turbulence intensities were lower than streamwise, ranging up to about 10% near the fuel surface for the CEWF configuration and 20% for the CRNF configuration. The decrease relative to the streamwise intensity was consistent with other observations of wind tunnel and atmospheric turbulence (14, 15). Particle Size Distributions. Particle size distributions for replicated experiments of the CEWF high wind speed configuration are given in Table IIIa, and those for the CRNF low wind speed configuration are given in Table IIIb. Computed mass median aerodynamic diameters (MMAD)and geometricstandard deviations (ag)are shown for each experiment. All distributions showed. a large fraction of fine particulate matter, with 70% or more collected on the filter stage. The distributions generally reached a minimum at around 2-5 pm, with increasing mass collected on the larger stages. No substantial differences were seen between the CEWF high speed and CRNF low speed configurations. The CEWF experiments bracket the CRNF experiments in MMAD. Temperature Profiles. Flame temperature profiles are shown in Figures 6 and 7 for the CEWF high wind speed and CRNF low wind speed configurations, respectively. The central graphic in each figure is the isotherm as derived from the type-K thermocouple array. Surrounding it are the relative frequency densities of temperature for the flame regions indicated, derived from measurements with the fast-response type-R thermocouple. The time scale of the original measurements was converted to distance (travel) by use of the measured spreading rate for each experiment. The relative frequency density was computed by taking the relative frequency of temperature for intervals of 50 K and dividing it by the interval width. The integral over the full temperature
Table I11
stage 1 2 3 4 5 6 7 filter
stage 1 2 3 4 5 6 7 filter
Section a: Particle Size Distributions for Replicates of CEWF High Wind Speed Configuration MMAD = 0.062 pm; ug = 3.38 MMAD = 0.101 pm; ug = 3.92 effective mass mass cumulative effective mass mass cumulative diameter (pm) (mg) fraction (%) mass fraction (%) stage diameter (pm) (mg) fraction (%) mass fraction (%) 13.69 7.67 4.09 2.16 1.24 0.73 0.38 0.00
0.3 0.1 0.1 0.1 0.2 0.4 0.5 5.4
4.2 1.4 1.4 1.4 2.8 5.6 7.0 76.1
100.0 95.8 94.4 93.0 91.5 88.7 83.1 76.1
1 2 3 4 5 6 7 filter
13.57 7.60 4.05 2.14 1.23 0.73 0.38 0.00
0.5 0.1 0.0 0.0 0.2 0.3 0.5 5.2
7.4 1.5 0.0 0.0 2.9 4.4 7.4 76.5
100.0 92.6 91.2 91.2 91.2 88.2 83.8 76.5
Section b: Particle Size Distributions for Replicates of CRNF Low Wind speed Configuration MMAD = 0.091 pm; a, = 4.32 MMAD = 0.087 pm; ug = 3.57 effective mass mass cumulative effective mass mass cumulative diameter (pm) (mg) fraction (76) mass fraction (9%) stage diameter (pm) (mg) fraction (%) mass fraction (%) 15.61 8.74 4.66 2.47 1.42 0.83 0.44 0.00
0.5 0.2 0.3 0.2 0.1 0.4 0.5 8.1
4.9 1.9 2.9 1.9 1.0 3.9 4.9 78.6
1768 Envlron. Sci. Technol., Vol. 27, No. 9, 1993
100.0 95.1 93.2 90.3 88.3 87.4 83.5 78.6
1 2 3 4 5 6 7 filter
15.56 8.72 4.65 2.46 1.41 0.83 0.43 0.00
0.7 0.4 0.4 0.3 0.3 0.5 0.7 9.1
5.6 3.2 3.2 2.4 2.4 4.0 5.6 73.4
100.0 94.4 91.1 87.9 85.5 83.1 79.0 73.4
0020
-
0010
1
0.004 0.002 0.000
0 005 0 004
0.015 0010
0.005 0 000
0 003
0
,
~
1000
2000
0
1000
0002 0 001 0 000
2000
Temperature ("C)
ky-; 'T ,
500
600
1
'
"
\
ih 0
Temperature ("C)
I000
,
2000
/ '
2ooo
O
600
0010 T
I
500
I
400 E 300 200
c
100
I
O
Em
g
200 100
L
0
0
0.005T 0 004
/
\
:::i':, ;q& 0.005 r
0.005 T
0.005 [\
0 002
0 001 0 000
0 000 0
1000
2000
0
1000
0 001
0 000
2000
0
1000
2000
Temperature ("C)
Flgure 6. Temperature contours and relative frequency densities of temperature from fast-response thermocouple at indicated positions in flame, CEWF high wind speed configuration. Contour interval = 100 K, height given above fuel surface.
range has a value of 1.0. In this, the relative frequency densities resemble the expected probability densities of temperature at each location. The shape of the relative frequency distributions yields pertinent information about the prevailing flame conditions at each point and is important in understanding the contribution of the various flame regions to the emission of pollutants. Mean temperature profiles for two positions-upward through the flame center and upward over a region of glowing combustion 150-200 mm behind the flame-are given in Figures 8 and 9. Mean values are shown by the central solid line in each graph, with the dashed lines indicating 1 standard deviation. The assumption of a normal distribution is, of course, inaccurate for temperature, but serves to illustrate the shorter flame length and greater unsteadiness of the CEWF high wind speed configuration. Distinct differences existed in the flame shape for the two configurations. The CRNF low speed configuration gave tall, erect flames with flame angles in excess of 60" from the horizontal. The burning zone in the fuel bed was 600-650 mm in length corresponding to a duration of about 50 s. The flaming zone extended about 200 mm behind the front, with a region of glowing combustion extending beyond. Coolingwas evident behind the flame, apparently due to entrainment of air from below the fuel bed. Increasing the wind speed for this configuration caused the flame to become noticeably tilted, with an angle of 40-50" and a total burning length between 500 and 1000 mm. The CEWF high wind speed configuration exhibited high flame tilt and short flame length. The burning length
0000
,
1000
0 000
2000
0
1000
~
2000
ZLj; 0 000
0
1000
2000
Temperature ("C)
Flgure 7. Temperature contours and relative frequency densities of temperature from fast-response thermocouple at indicated posklons in flame, CRNF low wind speed configuration. Contour interval = 100 K, height given above fuel surface.
was 550-600 mm, with a duration of approximately 60 s. The flame angle was about 30") substantially shallower than the CRNF configuration. The flaming zone was longer, about 300 mm. Cooling behind the flame was not observed to the same extent as when the floor was removed (cf. Figures 6 and 7). The flame length was approximately 400 mm, roughly two-thirds that of the CRNF low speed configuration. The relative frequency densities of temperature are similar in many respecta, differences being due principally to the greater unsteadiness of the flame at the higher wind speed in the CEWF configuration. Peak temperatures at the fuel surface were somewhat lower in the CEWF configuration (1200 vs 1350 "C for the CRNF configuration). The lower temperatures were likely the result of inadequate response time for the thermocouple and possibly to greater heat transfer to the floor. The temperature profile for the CENF high wind speed configuration (Figure loa) differs markedly from the CEWF case at the same speed. The flame angle is also about 30°, but the flame length is comparable to that of the CRNF configuration (about 600 mm). The CENF configuration also closely resembles temperature profiles collected for comparison during an actual field burn in rice straw (Figure lob). The field data were collected in the same manner as the low-resolution temperature data in the wind tunnel, that is, by allowing the fire to burn past a vertical array of thermocouples. In the field, however, the region between the fuel bed and the soil surface could also be instrumented. As in the wind tunnel, the field burn was conducted with the fire spreading against Environ. Scl. Technol., Vol. 27, No. 9, 1993
1769
'
600 -
0
100
200
300
400
'
,
1 ow
500
0
Height above conveyor (mm)
I
0
200
100
I
300
,
-
1
Wind Tunnel
1500
2000
(a)
400
Height above conveyor (mm)
I
(b)
Figure8. Temperatureprofiles showing mean (solidline) and standard deviation (dashed lines) for the CEWF high wind speed configuration. (a) Upward through the flame center. (b) Upward over the zone of glowing combustion. Height of fuel surface approximately 64 mm.
1.200 1
I
---__
-1 1
0
200
100
300
400
500
600
700
Helght above conveyor (mm)
i
,
I
I
1,2007-
O
\
C 0
100
200
300
400
500
600
700
Height above conveyor (mm) (b)
FlgureS. Temperatureprofilesshowing mean (solid line)and standard devlatlon (dashed lines) for the CRNF low wind speed configuration. (a) Upward through the flame center. (b) Upward over the zone of glowing combustion. Helght of fuel surface approximately 64 mm.
the wind. The field data, as the wind tunnel data suggest, show peak temperatures to lie in the fuel bed. In general, the flame shape in the wind tunnel was sensitive to the position of the ceiling and the floor 1770 Environ. Scl. Technol., Vol. 27, No. 9, 1993
-400
0
SoilSurface ' ' '
'
500
'
,
,
,
,
,
1 ow
,
,
,
I 1500
I
I
~,
i
1
2000
Travel (mm) (b)
Figure 10. Temperature contours for (a) CENF high wind speed configuration In wind tunnel, and (b) field burn In rice straw with flre spreadingagainst the wind (wind velocity approxlmateiy 3 m s-' at 1-m elevation). Height in panel a is above fuel surface.
configuration. Retracting the ceiling caused the flames to become erect, installing the floor shortened the flame length. Local Gas Concentrations. Concentration profiles for oxygen, carbon monoxide, and carbon dioxide through the flame center and over the glowing combustion zone behind the flame (at approximately the same locations as in Figures 8 and 9) are plotted in Figures 11and 12 for the CEWF high wind speed and CRNF low wind speed configurations,respectively. The shorter flame height with the CEWF configuration is immediately apparent in Figure 11. In the flame, oxygen concentrations decreased to about 5 % in the vicinity of the fuel surface for both configurations. The increase to 7% at the fuel surface for the CRNF low wind speed configuration (Figure 12) is not considered to be significant and represents the type of variability demonstrated in flames of this type. CO concentrations approach 10% in the flame, but remain low in the region behind the flame due to dilution with inlet air, especially in the CRNF configuration where air could pass through the fuel bed from below. C02 concentrations were found in the range of 12-14 % in the flame near the fuel surface. Behind the flame and over the zone of glowing combustion, the COSconcentration was 4 % in the CEWF configuration, but only 1% in the CRNF configuration, again most likely because of dilution air passing upward through the fuel bed (note the nearly constant oxygen concentration behind the flame in the CRNF configuration). The lowest 02 concentrations were observed in the CEWF configuration with low wind speed,
*
02
* co
u
CEWF High Wind Speed
coz
- . - - - -CRNF - - Low Wind Speed
30
p -
4
8.
20
.-
E
3
6
1
0
C
1
I
I
5
g
10
00
~~
100
200
300
400
600
500
700
v)
Height (mm)
(a)
*
02
* co
0
* co2
0
100
200
300
400
500
600
700
Height above conveyor (mm)
Flgure 13. Profiles of peak soot volume fraction through flame center for the CEWF high wind speed (solid line) and CRNF low wind speed (dashed line) conflguratlons.
0
100
200
400
300
500
600
700
Height (mm)
(b)
Flgure 11. Gas concentration profiles for the CEWF high wind speed configuration. (a) Upward through the flame center. (b) Upward over the zone of glowing combustion.
u
02
co
--A-
--a-
near the fuel surface was impeded by the plugging of the probe tip with tar condensing from the gas phase. Soot Volume Fractions. Peak soot volume fractions for the two configurations are plotted in Figure 13. Shown are profiles taken upward through the flame center. The peak volume fractions for the CRNF low wind speed configuration were at least double those for the CEWF high wind speed configuration. There appeared to be a higher concentration for the CRNF configuration in the plume behind the fire as well, which was consistent with the higher TSP emission for this configuration.
co2
Discussion
0
100
200
300
400
500
600
700
Height (mm)
(a)
* 02
--A-
co
u
co2
30
F
2
20
w
;, I
g
3
10
5 0
Turbulent flows are characterized by a range of length scales from the largest energy-containing eddies that are described by the integral length scale down to the smallest scales of the flow, the Kolmogorov scales, in which viscous dissipation terminates the cascade of energy through the spectrum. When a flow contains chemicallyreacting zones within it, such as a flame, it is useful to compare the length or time scales of the turbulence with characteristic length or time scales of the reaction zone. A comparison of these scales serves to identify regimes of combustion (16)which may extend from a distributed reaction zone regime that resembles a chemical reactor to a system that resembles laminar flamelets that are strained and contorted by the turbulent flow field. An analysis of these time or length scale ratios is instructive in terms of determining if model experiments behave in a similar manner to full-scale systems. Variations in these ratios may affect the structure of the flame and the emission of pollutants. A fundamental comparison of time scales in combustion is the Damkohler number, Da, which relates a characteristic fluid mechanical time scale, 7%to a time scale for chemical reaction, T C , i.e.
I 0
100
200
300
400
500
600
700
Height (mm)
(b)
Flgure 12. Gas concentratlon profiles for the CRNF low wind speed conflguration. (a) Upward through the flame center. (b) Upward over the zone of glowing combustlon.
where the concentration near the surface in the flame center was 1.2 % In all cases, sampling in the fuel bed or
.
--
Da = rF/rC
(2)
As 7c 0, the reactions tend toward equilibrium. As T F 0 and Da approaches zero, the reactions are far from equilibrium as the chemistry cannot keep up with the supply of reactants that is imposed by the flow field. In a turbulent flow, it is possible to associate the fluid time Environ. Sci. Technoi., Vol. 27, No. 9, 1993
1771
lo' N -
in'
in1
in' IO-'
in"
10'
in'
in*
in'
io'
io'
io?
io3
in4
inc
in
in'
io'
in4
ino
io1
io*
in3
K~
io4
(m-')
Flgure 14. Spectral energy distributionsfor the CEWF high wind speed configuration at four heights above the conveyor surface (upstream anemometer positlon,flre present in wind tunnel). Stretchedabscissa, lowest energy regions between lo3and lo4 wavenumber In all cases.
approximately 2.5 decades up to the maximum frequency allowed by the sampling (1100 Hz).For convenience, the spectra have been separated along the abscissa (each spectrum having then a separate abscissa) so that the lowest energy regions correspond to the wavenumber region between lo3 and lo4m-l, with the exception of the 180-mm position in the CRNF configuration, where the lowest energyregion lies in the range of 104-105m-l. Lower wavenumbers represent long wavelengths or large eddies, which contain most of the turbulent energy. The Kolmogorov length scale is found at higher wavenumbers in the viscous dissipation range. Intermediate is the inertial subrange, where the energy is proportional to K1-5/3,which can be seen in the figures for heights 510 rnm and lower within the boundary layer. This relationship would not be expected to hold in the freestream above the boundary layer (i.e., for the 800-mm height). The distinct difference in the spectral energy for the 180-mmheight in the CRNF configuration is due to the unusual nature of the flow just above the fuel surface (cf. Figure 5). Approximate turbulent kinetic energies were estimated from the anemometer data by assuming that the intensities of the vertical and transverse fluctuations were half that of the streamwise (consistent with measurements in the vertical direction)
k -
io-'
ioo
io1
in2
in!
io'
io:
in'
io'
in4
ioo
in1
io2
in3
io4
io'
io'
io'
in'
in4
io4
inJ
K (&I)
I
Flgure 15. Spectral energy distributions for the CRNF low wind speed configuration at four helghts above the conveyor surface (upstream anemometer position, fire present in wind tunnel). Stretchedabscissa, lowest energy regions between lo3 and lo4 wavenumber in all cases except 180 mm between lo4 and lo5.
scale with the turbulence characteristics, in particular, the turbulence intensity and a length scale. Most mixing in turbulent flows is due to the large eddies so that an appropriate fluid time scale would be the large eddy life time, found in terms of the integral length scale, 1, and the turbulence kinetic energy, k, as 7F =
1/(2k)112
(3)
The integral length scale can be found in the wind tunnel by examination of the velocity spectra and velocity autocorrelations. The latter quantities were used to determine the integral length scale through integration under the correlation function, followed by application of Taylor's hypothesis (17). Results of spectral analyses on both the CEWF high speed and CRNF low speed configurations are given in Figures 14 and 15, respectively. Plotted is the spectral energy, E , at four different heights, against longitudinal wavenumber, ~1 = 27rffU, where U is the streamwise velocity (m s-l) andfis the frequency (s-l)of the turbulent fluctuations given from the FFT. The spectra cover 1772 Environ. Sci. Technol., Vol. 27, No. 9, 1993
= -ut2 3-
4
(4)
where u" is the variance of the velocity fluctuations in the streamwise direction. The time scale for chemical reaction will be approximately the same for rice straw combustion in both the wind tunnel and in the field burn. Therefore, the question of comparative Damkohler numbers becomes one of relative hrbulence kinetic energies arid length scales. For the CEWF high wind speed configuration, at a typical height of 200 mm (approximate height of the flame tip), the rms velocity fluctuation is about 0.5 m s-l. Tbe value of k is then about 0.19 m2 s-~. The integral length scale at this height is 0.16 m; this yields an approximate fluid time scale of 260 ms. In the CRNF low speed configuration at 200 mm height, the rms velocity fluctuation is about 0.06 m s-1, the integral length scale is 0.12 m, and the corresponding fluid time scale is about 1700 ms. The time scales converge for the two configurations above 300 mm, increasing for the CEWF high wind speed and decreasing for the CRNF low wind speed configuration. Both are approximately 500 ms at 350-mm height. The low intensity of turbulence gives very large values for 7F outside the boundary layer. Above the canopy of a field in an atmospheric boundary layer, the integral length scale is much greater than in the tunnel, of the order of 100 m (18). However, this length scale is clearly inappropriate in terms of mixing within a flame burning within the first 200-300 mm above the surface. In this case, a more appropriate measure of largescale turbulence would be of the order of the flame height; larger eddies would appear simply as a relatively slow unsteadiness of the flow on the scale of the fire. Consequently, the relevant large length scales would be of the same order in both field and tunnel flows, Le., a flame height of several hundred millimeters for the opposed flow condition over a straw layer. The estimated turbulence intensity in the atmospheric boundary layer is approximately 155% of the local mean velocity at heights from 2
Table IV. Computed Wind Speed, Boundary Layer and Friction Velocity Thickness (a), Roughness Height (a), (U*)
integrated config- wind speed uration (m s-l) CEWF CEWF CEWF CEWF CEWF CEWF CEWF CEWF CRNF CRNF CRNF CRNF CRNF CRNF CRNF CRNF
2.98 3.19 2.93 3.16 1.53 1.71 1.53 a 2.75 2.77 2.52 a 0.96 1.29 0.65
*
position upstream upstream downstream downstream upstream upstream downstream downstream upstream upstream downstream downstream upstream upstream downstream downstream
flow 6 condition (m) cold hot cold hot cold hot cold hot cold hot cold hot cold hot cold hot
0.45 0.65 0.60 0.6P 0.70 0.55 0.60 a 0.60 0.80 0.70 a b 0.60b b a
20
u*
(m)
(m s-l)
0.023 0.050 0.055 a 0.096 0.131 0.119 a
0.42 0.61 1.26 U
0.44 0.60 0.45 a 0.047 0.49 0.090 0.63 0.069 0.51 a a 0.352b 0.6gb 0.298b 0.94b 0.540b 0.81b a U
*
Partial traverse. Fullvdeveloped boundary layer not observed.
to 30 m above the ground (19), and just above plant canopies measurements range from approximately 30 % to 120% (20-23). The lower values are typical of the turbulence intensities in the tunnel at the height of the fire. Consequently, the relevant large eddy time scale in the field will be of the order of 100-500 ms at similar wind speeds. Another important comparison of scales relates to the local structure of the flame itself. This comparison necessitates an examination of the small-length scales, the Kolmogorov scales, in relation to a characteristic flame thickness. A typical laminar diffusion flame reaction zone thickness of 1 mm will be an adequate measure of the latter parameter. The estimation of the Kolmogorovscale is somewhat more difficult. The Kolmogorov scale in the atmosphere is about 1mm (18). In the CEWF high wind speed configuration, the Kolmogorov scale was estimated to lie in the range of 0.10.5 mm. The uncertainty was greater in the case of the CRNF low wind speed configuration, where the Kolmogorov scale was estimated to lie in the range of 0.2-3 mm. The flame thickness is of the same order or slightly larger than the estimated Kolmogorov scales, at least where a reasonably well-developed boundary layer can be observed. Although the conditions for a flamelet description of the flame structure may not be rigorously satisfied, we may expect nevertheless that the structure of the flames will be similar in the field and in the wind tunnel, given the similarity of the flame kinetics and the Kolmogorov scales. A similar conclusion applies with regard to the large-scale structures of the flames in the field and tunnel. In the case of the CEWF configuration, at both high and low wind speeds, and in the case of the CRNF high wind speed configuration, it has been possible to estimate surface roughness heights, zo (m), and friction velocities, u* (m s-l), from models of the flow based on the logarithmic law-of-the-wall: (5)
where z is the height (m) and K is von Karman's constant (K= 0.41). The results of the analysis are given in Table
IV, along with estimated boundary layer thickneas (derived from considerations of Reynolds stresses) and total integrated inlet wind speed. Results for the upstream and downstream anemometer positions, as well as hot (fire present) and cold (fire absent) flows, are included for the two configurations at both wind speeds. Where only partial traverses were possible (downstream position with a fire present), z, and u* could not be determined. For the CEWF configuration at the higher wind speed, 20 Izo I50 mm, and 0.4 Iu* I1.2 m s-l. At lower wind speeds, zo z 100 mm, and 0.4 Iu* 5 0.6 m s-l. Similar conditions apply to the CRNF high wind speed configuration, in which 40 Iz, I90 mm, and 0.5 Iu* 50.6 m s-1. However, for the CRNF low wind speed configuration, the logarithmic law-of-the-wall model is not appropriate to determine zo and u* because of the poorly developed boundary layer. Roughness heights for the atmosphere were reported by Counihan (19)to range from 1.4 mm over open ground to 150mm over rural terrain. A grassy surface with height of 150 mm was reported to have a roughness height of 6.6 mm. Raupach et al. (14) gave zo from 15 to 50 mm for wheat fields depending on canopy height and 4-10 mm for model crops and regularly arrayed elements in wind tunnels. Estimates from Counihan (19) show the friction velocity for the lower atmosphere to be of the order of 0.2-0.4 m s-l. For the experiments described here, the higher wind speeds and the configurations with the ceiling extended give values of zo and u* which are at least comparable with atmospheric values. The values of z, computed from the CRNF low wind speed velocity profiles are not. That the flame temperature profiles are affected strongly by the change in wind tunnel conditions is immediately apparent from Figures 6 and 7. As one would expect, the flame burning under the high velocity conditions is bent over by the approaching flow and appears to be considerably shorter than the low speed flame. The peak temperatures that were measured under both conditions were similar, about 1200-1300 "C,although it must be kept in mind that these temperatures represent lower bounds due to the averaging effect of the thermocouple, which had a finite response time to flame fluctuations. The accompanying frequency distributions of temperature at similar positions in both flames are quite similar in both cases. The most interesting distributions are at the foot of the flames. The distributions appear to be bimodal, with the lower mode indicating freestream cold air at these locations. The form of the distributions may be important eventually in understanding the sources of NO in these flames; however, measurements of NO were not obtained in these tests. The reduction in the flame height with the increase in wind speed suggests that mixing may be enhanced. The estimates of the time scale for large eddy mixing showed that in the high speed condition this time scale was about 300 ms near the surface, and about 6 times this value in the low speed case. Therefore, it is plausible, and not unexpected, that increased wind speed would improve the mixing rates of fuel and air in these diffusion flames. This is supported by an examination of the probe sampling results through the flame that were presented in Figures l l a and 12a. In particular, the 0 2 concentration profiles are of interest. At low speeds, the 0 2 mole fraction reaches the freestream value at about 400-500 mm above the fuel Envlron. Sci. Technol., Vol. 27, No. 9, 1993
1773
bed. On the other hand, at high speeds the freestream O2 mole fraction is attained at a height of 200 mm above the fuel bed. CO and CO2 concentrations confirm the trend. It is worth noting that there are no significant differences (within the experimental uncertainty) in the 02,COO,and CO concentrations in the flame near the fuel bed with and without the floor. However, the 0 2 concentration is lower, and the product concentrations are higher downstream of the flame front when the floor is present. This observation suggests that transport of air through the floor may be important in the downstream smoldering zone. Laser extinction was used to measure line-of-sight soot volume fractions. Typical peak values measured in the flames were of the order of lo4. Much higher values have been obtained in measurements of the local soot volume fractions in hydrocarbon flames. For example, Santoro et al. (24) reported peak values of about in a laminar ethylene air diffusion flame. Neil1 and Kennedy (25) measured values of about 6 X in a turbulent ethylene air flame and about 2 X 10-7 in a turbulent propane air flame. In the wind tunnel, portions of the laser path were outside the flame region. However, it was necessary to define the path length in terms of the tunnel width due to the unsteadiness of the flame. For this reason, the reported soot volume fractions are much lower than one would obtain by measuring over a much shorter distance through a laminar or confined turbulent diffusion flame. Upon comparing the two sets of profiles, the wind speed appears to have an important impact on the soot formation within the fire itself. The soot loading under the low speed conditions is at least a factor of 2 higher than for the highspeed conditions. The trend i s consistent with the TSP results obtained from sampling in the stack of the wind tunnel, although the effect is greatly diminished through sampling in the stack gases. At least part of the difference may be due togreater unsteadiness in the flow at the higher wind speed (with lower flame volume in the line of sight) and possibly to errors resulting from the fire crossing the beam at an angle, although this latter effect would apply equally to both cases and is reduced by averaging over several runs. In general, soot volume fraction was reduced by increasing the inlet velocity. Soot formation in flames is a relatively slow process and its kinetics are far from equilibrium. Therefore, the amount of soot that forms in a diffusion flame depends critically on the time that is available. A shorter residence time will lead to a reduction in the amount of soot that forms. Both laminar and turbulent diffusion flames exhibit this effect. Increasing the flow rate of fuel in a laminar diffusion flame will increase the amount of soot that forms as a result of the greater length of the flame. Eventually, the flame will start to emit soot when it reaches its so-called sooting height (26). On the other hand, a turbulent flame at a fairly high Reynolds number does not change its height as the flow rate of fuel is increased. Instead, the intensity of the turbulence and the mean velocity increase. As a result, the rate of mixing of fuel and air increases and less soot forms (27). These examples serve to demonstrate the impact of residence times or mixingrates on soot production. Increases in mixing rates within the rice straw flames appear to be manifested as a reduction in the soot loading. The magnitudes of the TSP emission factors determined here can be compared with those of Darley (4,5),who was also able to perform direct mass balances for computing 1774
Environ. Sci. Technol., Voi. 27, No. 9, 1993
the emission factor. At similar moisture content (8-10% wet basis), Darley reports three values for TSP with backing fires in rice straw: 0.11 % (5),0.35% (4, fuel bed on a 25' slope), and 0.46 % (4, fuel bed on a 15' slope). The TSP emission factor in the 64 tests over all four configurations of the wind tunnel using the same source of straw ranged from a low of 0.45% (CEWF, 3.4 m 9-9 to a high of 0.79% (CRNF, 2.2 m s-9. The lowest values here are about the same as the highest values obtained by Darley. Rice straw from other sources tested at the same moisture and loading rate in the wind tunnel have subsequently yielded values as low as 0.2% in the CEWF high wind speed configuration. MMAD found here of about 0.1 pm are similar to those reported for field burns with rice straw (8). Reasons for the differences between the results reported here and those of Darley are therefore uncertain, but may be due to differences in the burning conditions, fuel composition, or sampling technique. Some effects due to differences in burning conditions have been noted above. Differences due to sampling technique probably exist; the wind tunnel samples were withdrawn from the stack flow at lower temperatures than Darley encountered ( 4 , 5 ) ,and this could affect aerosol particle mass. Influences of physical and chemical differences, other than moisture content, within fuels of the same type have not been explored, but may be important contributors to the variation in emission factor observed.
Conclusions Average emission factors for total suspended particulate matter were found to differ significantly among the four wind tunnel configurations, ranging from 0.505% for the ceiling extended configuration with the floor installed (CEWF) and the wind speed over 3 m s-1 to 0.727% for the ceiling retracted configuration without the floor (CRNF) and the wind speed at 2 m s-l. Adding the floor tended to reduce the TSP emission factor, as did increasing the wind velocity. The ceiling position by itself was not found to significantly alter the emission factor. Instead, the effect of the ceiling position appeared to be related more to the change in wind speed, whereby the flow decelerated ahead of the fire after entering the combustion test section. Velocity profiles from tests with low wind speeds and the ceiling retracted were distinctively different from those of other tests. A region of very low velocity extended upward from the fuel surface for 100-200 mm, suggesting an extremely adverse pressure gradient with separated flow. The gas sampling results did not suggest significant effects of the tunnel floor on the gas composition near the fuel bed at the front of the flame. Further downstream it is possible that airflow through the smoldering bed is important. Particle entrainment by such airflow may also be important. Although uncertain, the laser measurements lend credence to the hypothesis that the wind speed is a dominant parameter in the emission of particles, especially soot which is an important component of the stack particle loading. This suggestion is supported by the estimates of mixing times in the fires; these time scales were based on the velocity measurements. Some correlation between wind speed and particulate loading can be expected on the basis of these results. Although there were significant differences found in TSP emission factor, the total variation was not large,
being on the order of 30-40%. For the purposes of estimating emission offset credits, an average value of emission factor determined under the extreme cases in the wind tunnel should be adequate, given that properly conducted field burns are likely to occur over a similar range of conditions, even in the same field. At least some of the variability in field data can be attributed to differences in wind speed, turbulence characteristics, and fuel bed position. Additional effects include those of high winds and gusts, which enhance particle entrainment, and variable fuel properties, especially moisture, which also influence pollutant emissions. Gust effects are not evaluated in the existing facility; variable fuel properties are the subject of additonal research. The difficulty in applying these results for the regulatory purposes of offset credits and protection of ambient air quality lies in the extent to which field burning is, in fact, conducted properly. Violations of the prescribed agricultural burning methods are known to occur (e.g., burning with the wind rather than in opposition, burning with excess fuel moisture) and for the most part increase TSP emissions (other emissions, such as CO, will also increase, although some, such as NO, may not). Common violation of prescribed burning methods would cause, for most pollutants, the offset determination procedures to underestimate the benefits of crop waste utilization by the power stations. Since there is actually very little enforcement of burning technique (for reasons due principally to limited staff and, hence, limited surveillance), the extent to which the prescribed methods are followed is unknown. The wind tunnel results cannot, therefore, represent actual emissions with certainty, but instead must represent the expected emissions under reasonable compliance. In light of the possible conditions under which field burning can occur, the range in TSP emission factors observed in the wind tunnel under these disparate conditions seems rather narrow. Understanding the reasons for the significant differences in TSP among the configurations is important for a better understanding of fire spread and emission formation mechanisms; from a regulatory perspective, the narrow range is preferred. Whether the emission factors for other pollutants are confined to such relatively narrow ranges remains to be seen. Acknowledgments This work was supported under Contract A932-161 of the California Air Resources Board (CARB). We are grateful to Jack Paskind and Manjit Ahuja of CARB for their support. Literature Cited Jenkins, B. M.; Turn, S. Q.; Williams, R. B.; Hickman, S.; Stanghellini,E.; Bounds,L. An assessment of burn fractions and seasonal burn profiles for agricultural crop residues in the SunJoaquin Valley of California;Final Report,ARB Interagency Agreement No. A847-110;California Air Resources Board Sacramento, CA, 1990. Jenkins, B. M.; Turn, S.Q.; Williams, R. B. Calif. Agric. 1991,45 (4),12-16.
(3) Darley, E. F. Air pollution from forest and agricultural burning; CARB Project 2-017-1;Statewide Air Pollution
Research Center, University of California: Riverside, CA, 1972. (4) Darley, E.F. Emission factors from burning agricultural wastes collected in California; Final Report, CA/ARB Project 4-011;Statewide Air Pollution Research Center, University of California: Riverside, CA, 1977. (5) Darley,E. F. Hydrocarbon characterization of agricultural waste burning; Final Report, CARB Project A7-068-30; Statewide Air Pollution Research Center: University of California, Riverside, CA, 1979. (6) EPA. Compilation of air pollutant emissionfactors AP-42; U.S.Environmental Protection Agency: Research Triangle Park, NC, 1985. (7) Boubel, R. W.; Darley, E. F.; Schuck, E. A. APCA J. 1969, 19 (7),497-500. (8) Goss, J. R.;Miller, G. E. Study of abatement methods and meteorological conditions for optimum dispersion of particulates from field burning of rice straw; ARB Project 1-101-1; University of California: Davis, CA, 1973. (9) Carroll, J. J.; Miller, G. E.; Thompson, J. I?.; Darley, E. F. Atmos. Environ. 1977,11, 1037-1050. (10) Jenkins, B. M.; Chang, D. P. Y.; Raabe, 0. G.; Jones, A. D.; Miller, G. E.; Turn, S. Q.; Williams, R.; Teague, S.; Mehlschau, J.; Raubach, N.; Uyeminami, D. Development of test procedures to determine emissions from open burning of agricultural and forestry wastes; Phase I Final Report, Contract No. A5-126-32;California Air Resources Board Sacramento, CA, 1990. (11)Jenkins, B. M.; Turn, S. Q.; Williams, R. B.; Chang, D. P. Y.; Raabe, 0. G.; Paskind, J.; Teague, S. In Global Biomass Burning: Atmospheric, Climatic, and Biospheric Implications: Levine, J. S., Ed.; MIT Press: Cambridge, MA, 1991;Chapter 37. (12)Stationary Source Test Methods; CaliforniaAir Resources Board Sacramento, CA, 1987;Vol. 1. (13) Dalzell, W. H.;Sarofim, A. F. Trans. ASME: J . Heat Transfer, 1969,91,100-104. (14) Raupach, M. R.; Antonia, R. A.; Rajagopalan,S.Appl. Mech. Rev. 1991,44 (l),1-25. (15)Stein, W.; White, B. R.; Kavanagh, J.; Brucker, D.; Castro, E.; Bagheri, N.; Strataridakis, C. J. Wind-tunnel study of atmospheric dispersion of exhausts from the stack of building 222; Lawrence Livermore National Laboratory: Livermore, CA, 1989;UCRL-21200. (16) Williams, F. A. Combustion Theory, 2nd ed, Benjamin/ Cummings: Menlo Park, CA, 1985;p 393. (17) Hinze, J. 0.Turbulence; McGraw-Hill: New York, 1975; p 46. (18)Wyngaard,J. C. Probing the Atmospheric Boundary Layer; Lenschow, Donald H., Ed.; AmericanMetrological Society: Boston, MA, 1986. (19) Counihan, J. Atmos. Environ. 1975,9,871-905. (20) Cionco,R. M. Boundary-Layer Meteorol. 1972,2,453-465. (21) Maitani, T. Boundary-Layer Meteorol. 1979,17,213-222. (22) Ohtaki, E.Boundary-Layer Meteorol. 1980,19,315-336. (23) Ohtou, A.; Maitani, T.; Seo, T. Boundary-Layer Meteorol. 1983,27,197-207. (24) Santoro, R. J.; Yeh, T. T.; Horvath, J. J.; Semerjian, H. G. Combust. Sci. Technol. 1987,53,89-115. (25) Neill, T.; Kennedy, I. M. AIAA J. 1991,29,932-935. (26) Haynes, B. S.;Wagner, H. Gg. Prog. Energy Combust. Sci. 1981, 7,229-273. (27)Kent, J. H.;Bastin, S. Combust. Flame 1984,56, 29-42.
Received for review June 8,1992.Revised manuscript received March 31, 1993.Accepted April 6,1993.
Envlron. Sci. Technol., Vol. 27, No. 9, 1993
1775
