JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. D24, PAGES 32,013-32,030,DECEMBER 27, 1998
Physical and optical properties of young smoke from
individual
biomass
fires
in Brazil
JeffreyS. Reid• andPeterV. Hobbs Departmentof AtmosphericScience,Universityof Washington,Seatfie
Abstract. Physicalandopticalcharacteristics of particlesin smokefrom 19 fires weremeasured in Brazilduringthe 1995burningseasonaspartof theSmoke,Clouds,andRadiation-Brazil
(SCAR-B)project.TheUniversity of Washington C-131Ameasured particlesizesandabsorption andscattering properties in veryyoungsmoke(6CO 2+ 6H20+ 1.26x 106kJ
WhenCE >_90%, a fire is generallyin the flamingphase,and whenCE _710 K. If temperaturesare increasedto 910-980 K, CO can paper we wish simply to separatequalitativelythe properties
REID AND HOBBS:ABSORPTIONBY SMOKE AND BLACK CARBON CONTENT
32,015
Over the 12 s that it takes to fill the grab bag, the aircraft of smoke particles generated by low-intensity and high-intensityfires. To this end, we employtwo methods. travels about 960 m. All of the measurementspresentedin First,we can relateparticleproperties to the NOx emission this paper were derived from a samplecollectedacrossthe
can be factorof thefire. NOx is produced in firesfromfuelnitrogen width of a smokeplume. Thereforethe measurements
and by the oxidation of molecular nitrogen at high temperatures[Lobert and Warnatz, 1993]. Emissionfactors for fuel nitrogen shouldreach peak values in early flaming combustion,where fuel temperaturesare high. Similarly,
interpretedas averagesacrossthe width of a plume. On some occasions,two passesthrougha plume (separatedby less than 2 min) wererequiredto fill a bag. Someambientair (90%, with an averageof 95%. Smokesampleswere collected on averageat 480 m agl (950 hPa), with the highestat 880 m.
:
1 0 '1
1078 m. Excessparticle massconcentrations in the plumes
1 0 '2
variedfrom60 to 450 I.tg m-3.
1 0 '3
10 '4
4.2.
1 0 '5
To illustratethe basicnatureand variabilityof the size distributions of particlesproducedby biomassburningin Brazil, averageparticlenumberandvolumesizedistributions, basedon all the data,areshownin Figure1. The numberand volumedistributions for eachplumewerenormalized to unity.
1 0 '6
1 0 '7 0.01
0.1
(b)
1
10
Particle
Structures
and
Size
Distributions
By comparingthe areaunderthe curvesto the measuredmass concentrations, the averagedensityof the smokeparticleswas
i
foundto be 1.35+ 0.15g cm'3.
,
From Figure lb we seethat the submicronaerosolmode was
1.5
dominant,accountingfor over 90% of the total particle volume. This mode was dominatedby particlesproducedby the combustionprocess. Over 85% of the massin this mode wascomposed of organiccompounds, with theremaining15% beingblackcarbon(7%), potassium(4%), and otherelements (4%) [Ferek et al., this issue]. A transmissionelectron microscopy(TEM) micrographof particlescollectedin a _
1
0.5
0
0.01
0.1
Diameter •m) Figure 1. Average particle number (a) and volume (b) size distributionsfor all plume samplesin Brazil. Vertical lines show standard errors in the mean value.
The solid and dashed
lines are measurements from the differentialmobility particle sizer (DMPS) and forwardscatteringspectrometer probe300 (FSSP-300), respectively.
_
typicalforestfire showsthe internalstructure of theparticles (Figure2a). From this imagewe seethat the constituents of the aerosolsare both internallyand externallymixed, and both single(roughlyspherical)particlesand aggregates are present. Most of the sphericalparticleshad volatile shells coveringone or more involatilecoreparticles. The volatile shellswere mostlikely organics.
Figure2b showsan enlargement of a portionof Figure2a. To enhancethe fine structure of the particles,the negative imageis shownin Figure2b. Severalaggregates (probably
32,018
REID AND HOBBS: ABSORFFION BY SMOKE AND BLACK CARBON CONTENT
Figure 2. Transmissionelectron microscopy(TEM) micrographsof fine particles collected in the smoke plume from a mixed-phaseforest fire in Brazil. (a) Low magnificationview. This is a positive image, so darker areasindicatehigher densitymaterial. Note the gray areasaroundthe core particleswhere material has evaporated. (b, c) Digitally enhancedand enlargementsof individual particles. Here the image negative is displayed. Thus lighter areasare densermaterial.
black carbon) with volatized shells are seen. An enlargement of anotherheterogeneous particleis shownin Figure 2c (also a negativeimage). Here a large "residue"ring is seen (this is likely to be somewhatlarger than the original size due to
3c), and soil (Figure 3d). These particlesare similar to those foundby Radke et al. [1991] in the smokeplumesfrom boreal fires. Small noncombustiblematter on (or in) the foliage, suchas soils on leavesor silicatesin stems,are likely to be flatteningof the particle on impaction). The majority of the suspendedinto a fire plume. Also, soil particles can be original particle has volatized, leaving the involatile core. suspended by saltation of surface dust driven by winds The core particle itself is made up of several particles, generatedby the fire [Radke et al., 1991]. Particles such as these can have equivalent mass and aerodynamicdiameters probablyBC and low volatility organics. Figure lb showsa secondprominentvolume mode between over an order of magnitude smaller than their geometric 1-4 [xmdiameteranda thirdmodenear9 [xmthatextendedto 14 diameters,and the relationshipbetweenthe geometricand the [xm (not shownbecauseof poor countingstatistics). Scanning equivalent optical diameters is uncertain [Reid et al., 1994]. electron microscopy (SEM) micrographsof these particles Thus even thoughsuchparticlesmay accountfor 10% of the (Figure 3) show them to be quite differentfrom the submicron measured particle volume, their contributions to the mass, particles. They consist of carbon aggregates(Figure 3a), absorption, and scattering properties of the aerosol are partially cornbustedfoliage (Figure 3b), ash particles(Figure uncertain. However, since coarse particles have low mass
REID AND HOBBS:ABSORPTIONBY SMOKE AND BLACK CARBON CONTENT
32,019
Figure 3. Scanningelectronmicroscopy(SEM) micrographsof coarseparticles collected from mixedphaseforestfires in Brazil. (a) Ash aggregate,(b) cornbusted plant fiber, (c) elongatedashparticle,(d) soil particle.
4.2.1. Forest fires. Average DMPS size spectra for flaming and smolderingforest fires are shownin Figure 4. There were considerable variations in the structure of the From these distributions, two statistically significant trends particle size distributionswith regard to fuel, combustion are evident: CMD increaseswith increasingMCE, and the efficiency, and other fire properties. A summary of the standard deviations of the number and volume distributions particle size characteristics by fuel type is given in Table 1. ((sg cand(sg v,respectively) decrease withincreasing MCE.The We will now discusstheseresultsfor each fuel type. first result is consistentwith that reported by Hobbs et al.
scatteringefficiencies,their impact on light scatteringin the visible wavelengthsis probably low.
Table 1. Particle Size Parameters(Mean + StandardError) for Smoke Less Than 4 min Old in Brazil
Forest
(Flaming)
(Smoldering)
Grass
Cerrado
Particle Number Distribution
Median diameterQl.m)
Geometric standard deviation (ogc) Core median diameter•m)
Coregeometric standard seviation (ogc)
0.13 1.68 0.08 1.80
+0.005 + 0.02 + 0.02 + 0.05
0.10+ 1.77 + 0.05 + 1.99 +
0.005 0.02 0.005 0.04
0.10 1.79 0.06 1.95
+ + + +
0.01 0.05 0.01 0.09
0.10 1.91 0.06 1.81
+ + + +
0.01 0.15 0.01 0.08
Particle
Median diameter(gm)
Geometric standard deviation (O•c) Coremediandiameter(gm)
Coregeometric standard deviation (•c)
Volume Distribution 0.24 _+0.01 0.29 _+0.01 1.62 _+0.07 1.84 _+0.05
0.30 _+0.04 1.87 _+0.07
0.23 _+0.02 1.80 _+0.14
0.19_+0.02 1.65_+0.11
0.16_+0.02 1.59_+0.05
0.13_+0.01 2.00_+0.2
0.18_+0.01 1.76_+0.17
32,020
REIDANDHOBBS: ABSORFFION BY SMOKEANDBLACKCARBON CONTENT 0.32
•
Flaming Phase
---"
Smoldering Phase
:i
':• VMD;= 0.65-0.44 {MC,'E){',r=0.49)
,
........................
1.5c•
,
0.28 0.26
0.24
._N_ , , , ,
0.50
0.5
,
, ,
, ,
...........................:..............[............................ !--e.......................
0.22 o z
,
, , ,
Z 0.2
0.86
0.84
0.88
0.9
0.92
0.94
0.96
0.98
o
0.1
o.ol
(b)
1
ogv = 4.5 - 0.32 (MCE) ß . ...........
• .............
(r=0.74)
Diameter(gm)
1.9
............................
.'...........................................
Figure 4. Normalized accumulationmode number and volume particle size distributionfor aerosolsproducedby flaming(solid)andsmoldering (dashed) forestfiresin Brazil.
1.8
............. '•......... •-............ i........................................................
1.7
1.6
[1997a] for prescribedfires in the PacificNorthwest,but it differsfromEinfeldet al. [1991] resultsfor a prescribed fire in Montana. However, our relationshipbetweenthe CMD and theMCE is strong(Figure5a), with a correlation coefficient of
1.5 , ,
, ,
,
,
1.4
0.83.
0.86
0.84
0.88
The geometricstandarddeviationof the particlenumber
0.9
0.92
0.94
0.96
0.98
ModifiedCombustionEfficiency
distributiondecreaseswith increasingcombustionefficiency
Figure 6. (a) Volume median diameter and (b) geometric standard deviation of accumulation-mode aerosol in smoke from forest fires in Brazil versus the modified combustion
0.15
(a)
CMD=-0.10 , 0.24 (MCE)
(r=0.83) ,
efficiency.
0.14 :
!
:
............. ',, .............. •.............. :•.............. •............. •...........i........... ß
0.13
(Figure 5b). This, in turn, gives rise to a slight decreasein particle VMD with increasingMCE (Figure 6a). This agrees
withtheresults ofEinfeld etal. [1991].Likeoigc., Ogvis
negatively correlated with the MCE. It is surpnszngthat unlike the CMD, the VMD doesnot show a strongcorrelation with the MCE. However, becauseof the relatively few data points, the regression is not statistically significant at the
0.12
0.11
95% confidence
level.
,
0.1 0.86
0.84
0.88
0.9
0.92
0.94
0.96
0.98
1.9
._o ß•, ß•>
"• •
(b)
ogc = 2.1 - 0.39(MCE)
(r=0.27)
1ß85 .....................................................................................................
1.8
ß
1.75 .................................................................... •............................
The relationships between the particle size distribution parametersand the MCE are somewhat puzzling. Current theories, supportedby some data, suggestthat the CMD,
VMD,ogc,andogv should allbestrongly anticorrelated with the MCE [Einfeldet al., 1991;Radkeet al., 1991]. To explore this, we correlatedthe particlenumberandvolumedistribution moments with the CE, MCE, the logarithm of the flux-intensityscale, and the emissionfactorsfor particlesand gases.The correlationsandp valuesare listedin Table 2. As noted earlier, the CMD was highly correlatedwith the
• 1.7 ..... -..,-.------------,--------------r--------• withtheemission factors for CO2 andCO,respectively. Also, MCE.
E o
(D
•
Therefore
it is also well correlated
and anticorrelated
',
',
'
1.65 .......................................................... :: ................................ ß ........ ß :
a reasonable
anticorrelation
exists between the CMD
and the
NMHC, sincetheseare highly correlatedwith CO. Thusthose ,, aspectsof a fire that determinethe MCE (such as fuel type, 1.6 , , , • , , , • , , , i , , , i , , , • , , , i , , , 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98 oxidationrate, fire intensity,and fuel-air ratio) are also likely ModifiedCombustionEfficiency to determine the CMD. As noted in section 2, particles are nucleated in flamesby PAHs, and they grow by condensation Figure 5. (a) Count median diameter and (b) geometric and coagulation. In such high temperatureand particle-rich standard deviation of accumulation-mode aerosol in smoke environments,coagulation and condensationare essentially from forest fires in Brazil versus the modified combustion the same process because of the very small sizes of the efficiency.
REID AND HOBBS:ABSORPTIONBY SMOKEAND BLACKCARBONCONTENT
Table 2. Correlations andp Values(in Parentheses) of
32,021
correlation between theNOx emission factorandthec•g c,as
Particle SizeDistribution Parameters (CMD,Ogc, VMD, andc•gv) WithThreeFireParameters (CE,MCE,and
well as a slight (though not statisticallysignificant) positive
correlationbetweenthe NOx emission factorandthe CMD. Recall that, NOx can be interpretedas a surrogate for flame
Flux-IntensityScale) and EmissionFactorsfor Nine Samplesof SmokeFrom Flamingand SmolderingForestFires in Brazil
temperature. Thustheanticorrelation between the{Jgc andthe NOx emission factormayindicate a correlation between {Jgc and temperature. The correlationbetween the flux-intensity
CMD
o
gc
Fire Parameters 0.82 -0.24 MCE
Log of flux-intensityscale
VMD
gv
-0.46
-0.51
(0.002)
(0.478)
(0.208)
(0.128)
0.83
-0.27
-0.48
-0.54
(0.00!)
(0.430)
(0.183)
(0.103)
0.41
-0.75
-0.83
-0.91
(0.300)
(0.033)
(0.006)
(0.003)
0.67
0.54
andseveralemission factorssuchasNOx (Table2). Similarly
Emission Factors
Aerosol mass(PM4)
-0.29
(0.413)
(0.112)
(0.048)
(0.105)
CO2
0.81
-0.66
-0.56
0.61
0.61
0.15
CO
(0.003) -0.83
NOx NMHC
(0.053)
(0.117)
(0.059) -0.53
(0.002)
(0.080)
(0.477)
(0.109)
0.46
-0.85
-0.74
-0.82
0.27
0.60
(0.182) -0.58
(0.130) Ethane
0.56
-0.72
(0.007) (0.518) 0.33
(0.033) (0.114) 0.43
(0.007) 0.52
(0.148) 0.36
(0.018)
(0.419)
(0.292)
(0.347)
Propane
-0.66 (0.070)
0.24 (0.571)
0.33 (0.417)
0.32 (0.195)
Ethene
-0.28
-0.61
(0.505)
(0.110)
Propene
-0.63 (0.105)
Benzene
-0.65
(0.081) Toluene
-0.59
(0.125)
0.68
(0.067)
0.78
(0.013)
0.52
0.85
0.89
(0.195)
(0.007)
(0.004)
0.45
(0.259) 0.55
(0.154)
scale andtheNOx emission factor with{Jgc (Table 2) istobe expected for particles growing by coagulation and condensationat high temperatures. If condensationand coagulationwere the only mechanisms responsiblefor particle growth, the VMD shouldincreasewith the MCE. However, the VMD is negativelycorrelatedwith the MCE (albeit not significantly). Also, very strong anticorrelationsexist among VMD, the intensity-flux scale,
0.81
(0.014) 0.86
(0.005)
0.59
(0.088) 0.51
(0.153)
Valuesin italic typeindicatecorrelations at the 95% confidence level or better.
strong correlations existbetween {Jgvandthese parameters. Thus it appearsthat while the CMD is governed by flame processes, the VMD may be governedby processes outsidethe combustion
zone.
There are strong correlations between VMD and the emission factors of propene, benzene, and toluene (Table 2). At high temperaturesthesedouble-bondedcarbonspecieshave acceleratedreaction rates. If these unsaturatedhydrocarbons survive the combustionprocess,it is likely that longer chain hydrocarbons also survive. After smoke leaves the combustion zone, it cools rapidly and the longer chain hydrocarbons condense onto the larger particles, thereby increasingthe VMD. This interpretationis supportedby the positive correlationbetween the VMD and the aerosolmass emissionfactor. Thus much of the aerosolmassis likely to be a result of condensationof thesehydrocarbons. Condensationalgrowth of aerosolparticlesshoulddecrease
C•g,[McMurry andWilson, 1982].Thusif theVMD of the particle size distribution is controlled by condensation,one
wouldexpectOgvto decrease withincreasing VMD (and decreasingMCE). However, contraryto this expectation,we
foundthat{Jgvincreased withdecreasing MCE. A similar tendency is seenfor C•gc;thusCMD and VMD are
particles. SEM and TEM analyses show an association anticorrelated. This behavior could be due to several between particle aggregatesand flaming phase combustion processes.As can be seenfrom the TEM imagesin Figure 2, (see section4.2) [Martins et al., this issue (a)]. In and above the particles were both internally and externally mixed. the flame zone, temperaturesand particle concentrationsare Condensationmight be favored on specific particle species, high. The coagulation rate is linearly dependent on temperature,on the squareof the particle concentration,and 0.08 roughly inversely proportional to the square of particle diameter. Thus particleswith diameters. IntergovernmentalPanel on Climate Change (IPCC), Climate Change 1995: The Scienceof Climate Change,editedby J. T. Houghtonet al., CambridgeUniv. Press,New York, 1996. Kaufman Y. J., C. J. Tucker, and I. Fung, Remote sensingof biomass burningin thetropics,J. Geophys.Res.,95, 9927-9939,1990. Kent, J. H., A quantitativerelationshipbetweensootyield and smoke
regionalhazesdominatedby smokefrom biomassburningin Brazil: Closuretestsand directradiativeforcing,J. Geophys.Res.,thisissue. Scorer,R. S., Environmental Aerodynamics,JohnWiley, New York, 1978.
Tangten, C.D., Scattering coefficient and particulate matter concentrationin forest fire smoke,J. Air Pollut. Control Assoc.,32, 729-732, 1982.
Tums,S. R., An Introduction to Combustion, Concepts andApplications, pp. 291-297,McGraw-Hill,New York, 1996. Kiel, J. T., and B. P. Briegleb,The relative rolesof surfateaerosolsand Vines,R. G., L. Gibson,A. B. Hatch,N. K. King, D. A. MacArthur,D. greenhouse gasesin climateforcing,Science,260, 311-314, 1993. R. Packham,andR. J. Taylor,On the nature,propertiesandbehavior Kotchenmther, R. A., and P. V. Hobbs, Humidification factors of of brashfire smoke,Div. Appl. Chem.,Tech.Pap. 1, Commonw.Sci. aerosolsfrom biomassburningin Brazil, J. Geophys.Res.,this issue. andInd. Res. Organ.,Melbourne,Australia,1971. Lin, C., M. Baker, and R. Chaffson, Absorption coefficient of Ward, D. E., Factorsinfluencingthe emissions of gasesandparticulate atmospheric aerosols:A methodfor measurement, Appl. Opt., 12, matter from biomass burning, in Fire in the Tropical Biota 1356-1363, 1973. (EcologicalStudies84), editedby J. G. Goldammer,pp. 418-436, Liousse,C., C. Devaux, F. Dulac, and H. Cachier,Aging of savannah Springer-Verlag, New York, 1990. biomassburningaerosols:Consequences on their opticalproperties, Ward, D. E., and W. M. Hao, Air toxic emissionsfrom burning of J. Atmos.Chem., 22, 1-17, 1995. biomassglobally--Preliminaryestimates,in Proceedingsof the 85th Lobert,J. M., andJ. Wamatz, Emissions from the combustion process in Annual Meeting and Exhibition, Air and Waste Management vegetation,in Fire in theEnvironment:The Ecological,Atmospheric Association,1992. pointmeasurements, Combustion andFlame,63, 349-358,1986.
and Climatic Importanceof VegetationFires, editedby P. J. Cmtzen andJ. G. Goldammer,pp. 15-37,JohnWiley, New York, 1993. Martins, J. V., P. Artaxo, P. V. Hobbs, C. Liousse, H. Cachier, Y. Kaufman, and A. Plana-Fattori, Particle size distributions,elemental
Ward, D. E., and C. C. Hardy, Smoke emissionsfrom wildland fires, Environ. Intern., 17, 117-134, 1991. Ward, D. E., A. W. Setzer, Y. J. Kaufman, and R. A. Rasmussen, Characteristics
of
smoke
emissions
from
biomass
fires
of the
compositions, carbonmeasurements, and opticalpropertiesof smoke Amazonregion-BASE-Aexperiment, in BiomassBurningand Global from biomassbuming in the Pacific Northwestof the United States, Change,editedby J. S. Levine,pp. 394-402,MIT Press,Cambridge, in BiomassBurningand Global Change,editedby J. S. Levine,pp. Mass., 1991. 716-732,MIT Press,Cambridge,Mass., 1997. Ward,D. E., R. A. Susott,J. B., Kaufman,R. E. Babbit,D. L. Cummings, Martins,J. V., P. V. Hobbs,R. E. Weiss,andP. Artaxo,Sphericityand B. Dias, B. N. Holben, Y. J. Kaufman, R. A. Rasmussen,and A. W. morphologyof smokeparticlesfrom biomassburningin Brazil, J. Setzer, Smoke and fire characteristicsfor cerrado and deforestation Geophys.Res., this issue(a). bums in Brazil: Base B Experiment, J. Geophys. Res., 97,
Martins, J. V., P. Artaxo, C. Liousse, J. S. Reid, P. V. Hobbs, and Y.
14,601-14,619, 1992.
Kaufman,Effectsof black carboncontent,particlesize,andmixing Weiss,R. E., and P. ¾. Hobbs,Optical extinctionpropertiesof smoke on light absorptionby aerosolfrom biomassburningin Brazil, J. from the Kuwait oil fires, J. Geophys.Res., 97, 14,537-14,540,1992. Geophys.Res.,this issue(b). Westphal,D. L., and O. B. Toon, Simulationsof microscale, radiative, McMurry, P. H., and J. C. Wilson, Growth laws for the formationof and dynamicalprocessesin a continental-scale forest fire smoke secondaryambientaerosols:Implicationsfor chemicalconversion plume,J. Geophys. Res.,96, 22,379-22,400,1991. mechanisms,Atmos. Environ., 16, 121-132, 1982. Winklmayer, W., G. P. Reischl, A. O. Lindner, and A. Bemer, A new Patterson, E. M., andC. K. McMahon,Absorption characteristics of electromobility spectrometer for the measurement of the aerosolsize forestfire particulate matter,Atmos.Environ.,18, 2541-2551,1984. distributions in the size rangesof 1 to 1,000 nm, J. Aerosol Sci.,22,
Penner, J. E., R. Dickinson, and C. O'Neil, Effects of aerosol from biomass burning on the global radiation budget, Science, 256, 1432-1434, 1992.
Pyne, S., Introductionto Wildland Fires: Fire Managementin the UnitedStates,pp. 1-34,JohnWiley, New York, 1984. Quinn, P. K., S. F. Marshall, T. S. Bates, D. S. Covert, and V. N.
289-296, 1991.
P. V. Hobbs(corresponding author)andJ. S. Reid,Department of Atmospheric Sciences,Universityof Washington, Box 351640,Seatfie, WA 98195-1640. (e-mail:
[email protected]; jreid@ spawar.navy.mil.)
Kapustin,Comparison of measuredandcalculatedaerosolproperties relevantto the radiativeforcing of troposphericsurfateaerosolon (ReceivedAugust25, 1997; revisedDecember8, 1997; climate,J. Geophys. Res.,100, 8977-8991,1995. acceptedJanuary9, 1998.)