Oct 30, 1996 - Georgia. 3Biogeochemie, Max-Planck-Institut ffir Chemie, Mainz, Germany. 4Commissariat ... Fearnside [ 1991 ] and Comer), et al. [ 1981 ] for a ...
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 101, NO. D19, PAGES 23,651-23,665, OCTOBER 30, 1996
Black carbon formation by savanna fires: Measurements and implications for the global carbon cycle T. A. J. Kuhlbusch, 1,2M. O. Andreae, 3 H. Cachier, 4 J. G. Goldammer, 3 J.-P.Lacaux, 5 R. Shea, 6 andP. J. Crutzen I Abstract. Duringa fieldstudyin southern Africa(Southern AfricanFire-Atmosphere Research Initiative(SAFARI-92)),blackcarbonformationwasquantifiedin theresidues of savanna fires. The volatilizationratiosof C, H, N, and S weredeterminedby measuringtheir contentsin the fuel andresidueloadson six experimentalsites.The volatilizationof sulfur(864-8%)was significantly higherthanpreviously reported. Volatilization of H, N, andS wassignificantly correlated with thatof carbon,enablingusto estimatetheirvolatilizationduringsavannafiresby extrapolation fromthoseof carbon. By partitioning theresidues in variousfractions (unburned, partiallyburned,andash),a strongcorrelation betweentheH/C ratioin theresidueandthe formation of black carbon was obtained. The ratio of carbon contained in ash to carbon
contained in theunburned andpartiallyburnedfractionis introduced asan indicatorof the degreeof charring.As nitrogenwasenrichedin theresidue,especiallyin theashfractionof > 0.63 mm, thisindicatormaybeusefulfor an assessment of nutrientcycling.We showthatthe formationof blackcarbonis dependent on the volatilizationof carbonaswell asthe degreeof charring. Theratioof blackcarbonproduced to thecarbonexposed to thefire in thisfieldstudy (0.6-1.5%)wassomewhat lowerthm•in experi•nental firesunderlaboratory conditions (1.01.8%)whichmaybe dueto lesscompletecombustion. Theaverageratioof blackcarbonin the
residue tocarbon emitted asCO2ranged from0.7to2.0%.Usingtheseratiostogether with variousestimatesof carbonexposedor emittedby savam•afires,the worldwideblackcarbon
formation wasestimated to be 10-26Tg C yr'1 withmorethan90%of theblackcarbon remainingon thegrom•d.The formationof thisblackcarbonis a netsinkof biospheric carbon
andthusof at•nospheric CO2 aswellasa source of 0 2. 1. Introduction
supplyand fuel composition.Oxygensupplyand hencethe amountof energy liberatedare very much dependenton the The formationof black carbon(BC, definedhere as a highly stateof aggregation of the solidfuelswhich"mayrangefrom polyaromaticto elementalor graphiticcarbon fraction in the finely divided particlesto objectsas large as a building" particulatematerial) in the smokeand residuesof vegetation [Edwards,1974, p. 149]. Usingwoodburningas an example, fires and its importancefor the global carboncycle were first oxygenmustfirst diffuseintothe surfaceto furtheroxidizethe discussedby Seller and Crutzen [1980] and subsequentlyin carbonenriched,charred,nonvolatile,and dehydrated material. various publications[Houghton, 1991; Crutzen and Andreae, Further oxidation of nonvolatile matter in, e.g., savanna grass 1990; Mack, 1994]. BC in combustionresiduesand primary may be easiercomparedto wood becausethe fuel is not so carbonaceous aerosolsis formedby charringof the organicfuel condensed. Thus oxygenand heat may reachthe nonvolatiles duringcombustion.The processes involvedin the pyrolysisare from any direction. Thermal conduction,which in most solid well describedby Edwards[1974, pp. 61-68]. The main factors fuels is relatively slow, is less importantfor savannagrass influencingthe pyrolysisand thusBC formationare the oxygen burningthan for, e.g., woodburning. The formationof primaryaerosolsis describedby Edwards [1974, p. 152] as follows: "High surfacetemperaturesare achievedduringcombustion.Cracksand fissuresof the surface l Lufichemie,Max-Planck-Institut•r Chemie,Mainz, Germany. 2Now National ResearchCouncilAssociateat the National Exposure material result. Particulatematter, in fact particlesof rather ResearchLaboratory,U.S. EnvironmentalProtectionAgency, Athens large dimensionsmay be dislodgedin the process."This proGeorgia. cessis taking place during the combustion,but some charred 3Biogeochemie, Max-Planck-Institut ffir Chemie,Mainz,Germany. material may becomeairborneby wind erosionlong aRer the 4Commissariat • !'EnergieAtomique,CentreNationalde la Recherche fire extinguishedand thus may be significantlycontributingto Scientifique,Gif-sur-Yvette,France. atmosphericbackgroundconcentrationsof black and organic 5Laboratoire d'Aerologie,Universit6PaulSabatier,CentreNationalde la Recherche Scientifique, Toulouse,France. 6Departmentof RangelandResources,Oregon State University, Corvallis.
Copyright1996by theAmericanGeophysical Union. Papernumber95JD02199. 0148-0227/96/95JD-02199509.00
carbon rich aerosols. Visual
observations
of a burned savanna
site by one of the authors(T.K.) showedthat a large proportion of the black-coloredresiduewas transportedoff site in the absenceof rain within three to four weeks. Sumanet al. [1996] estimateda flux of 10 Tg BC yr-• to marinesediments,mainly to thoseof the continentalshelves,representing4-20% of the total black carbonproducedby vegetationfiresannually.
23,651
23,652
KUHLBUSCH ETAL.:BLACKCARBON FORMATION BYSAVANNA FIRES
plotson eachof the six experimental Onceformedby anycombustion processes, blackcarbon,in threeor foursampling plotconsisted of approximately 100m2 theresidues onthegroundandin atmospheric emissions, repre- sites.Eachsampling with above average vegetation density (by vegetation height and sentsa sinkof atmospheric CO2(anda sourceof 02) asit is no
to ensure thattheareawasburned during thefire, longeravailable for uptakeby plantsor microbial breakdown.patchiness) waspatchyowingto 2 yearsof drought Thelongresidence timeof blackcarboncanbe deduced from sincethevegetation its presence in soils,ice,andsediments [Herring,1977;Clark conditions.Thusthe determinedvegetationdensitiesdo not retheaverage vegetation density oftheexperimental sites. and Robinson,1993; Sternberg,1968; Chylecket al., 1987]. present The oldestcharcoalrecordin a carbonaceous matrix reportedin
Eachsampling plotconsisted of threesquares (each0.25m2)
density.Vegetation was a reviewby Copeand Chaloner[ 1980]wasfoundin Pennsyl- with low,medium,andhighvegetation anddividedinto grass,litter, andherbs. vanJancoalseamsthat formedapproximately 350 million years manuallysampled whichwasneithergrass ago.Thuscarbonis sequestered by the formation of black Herbsincludedall standingvegetation litterincluded all vegetation layingon carbon from the short-term biosphericcycle to the long norwoodyvegetation, standing dead -termgeological cycle.Thiseffectontheglobalcarbon cycleis the soil,mainlydeadgrass,andgrassincluded Onlyabout5-10%of thestanding grass could a significant factorin theunderstanding of pastandprediction andlivinggrass. be considered as living, "green"grass.The residueswere of futureatmospheric CO2 levels.
fromsquares justbeside theoneswherewe sampled CrutzenandAndreae[1990] estimatedthe annualformation sampled
within10min.to 2 hours afterthe rateof blackcarbonto be in therangeof 0.2-0.6Pg (1015g)C, priorto thefire,normally plots.Thisensured thatthepostfire resibutpointedto a highuncertainty dueto a deficiency of mea- firepastthesampling surements. Emissions of BC as a constituent of carbonaceous dueswe sampled werederivedfromvegetation typesandfuel aerosols havebeenthoroughly investigated [Changet al., 1982; load similarto thosesampledpriorto the fire. The residues intounburned (UB), partiallyburned(PB) Ogrenand Charlson, 1982;Penneret al., 1993]because of werepartitioned collected), andash(collected witha handheld their radiativeand catalyticproperties.On the other hand, (bothmanually cleaner)(Figure1). Thevacuumcleanertechnique was measurements of charcoalformationin postfireresidueshave vacuum preferred to trays, as by Sharrow and Wright [ 1977], Shea et al. only beenconducted for Amazoniandeforestation fires by et al. [1995],for several reasons. Fearnside[ 1991] andComer), et al. [1981] for a prescribed fire [thisissue],andKauffman would have hadtobeplaced below theli•ter, amajor fuel ona longleafpinesitein Florida.No fielddataontheformation Trays to ensure full recovery of theash,whichwould of blackcarbonin residuesfor agricultural waste,firewood,and component, savanna fireshaveyet beenpublished. Residues fromthecom- havemeanta disturbanceof the fuel, arbitrarilyinfluencingthe andcombustion process. Nevertheless, wetested metal bustionof differentvegetationtypes in a burningapparatus density traysin severalareasto compare thetwo techwere analyzedfor the first time for BC by Kuhlbusch and ashcollection andfounda poorrecovery. Theturbulence duringthe Crutzen[1995].Theywereableto showa negativecorrelation niques between black carbon as a fraction of the total residual carbon
fire had blown almostall of the ash out of the metal trays.To
(BCFFRC) andthe molarCO/CO 2 ratioin thegasesemitted, ensurethat vacuumcleaningdid not leadto a significantconindicating that flamingratherthansmoldering combustion is
tamination for the black carbon determination, we looked care-
the main sourceof BC. For nonwoodyfuel, about 1.5% (1.0-
fully at the soilpriorto the fire for anyblackcarbonparticles from previousfires,but couldfind only very few. Second,we
1.8%) of the fire-exposed carbon(CE) wasconverted to BC while90% of theexposed carbonwasvolatilized. Duringthe Southern AfricanFire-Atlnosphere Research Initiative(SAFARI-92),six experimental siteswereburnedfor
vacuumcleanedfive plots, after samplingthe biomassto determinethe vegetationdensityprior to the fire, and analyzed thosesamplesfor blackcarbon.Fromtheseanalyses we derived
joint multidisciplinaryresearchactivitiesin the KrugerNational Park, South Africa. At these study sites, severalinternational researchgroupsconductedmeasurements on the impactof savannafires on the savannaecosystemand the role of fire in the regionaland global atmosphereand in the globalcarboncycle. We quantifiedthe exposed,residualand black carbonon three plotsof eachexperimentalsite. Elementalcontentsandvolatilization ratios of hydrogen, sulfur, and nitrogen were also determined.
Studies of the formation
D•grN of
Chmrin8 - not •xpo•l to thereal degradation
PartiallyBurned
- only partially themlanydegraded
of black carbon and its
relation to gaseousemissionsand fuel and residuecharacteristics were conducted.The presentstudyfocuseson black carbon in savannafires to obtain a better estimateof the sequestering of carbonfrom the short-termto the long-termcarboncycle.
2. Experimental Techniques
Residue
- totally "blackened' but still some unaffected
organicmatterinside
:...::...':'• ß Sieve (0.63 *0.63rr•n) •
- totallythetonally degraded
sampled mandy
2.1. Sampling
Theexperimental sitesin KrugerNationalParkcanbe cha- Figure1. Thedifferent residue fractions andthedegree of charring concerning theirstructure. The unburned andpartly
racterizedas a deciduous, open-treesavannaparkland.More detailedinformationaboutthe vegetationcomposition is given
burnedfractionswere manuallysampled,whereasash was
witha handheld vacuum cleaner. Ashwasthendivided by Sheaet al. [thisissue]andTrollope et al. [thisissue].To sampled quantifythe carbonloadbeforeandafterthe fire,we chose intoash2 andash1 by sieving(0.63x 0.63mm).
KUHLBUSCH ET AL.: BLACK CARBON FORMATION BY SAVANNA FIRES
23,653
a maximum contaminationof black carbonfrom previousfires 2.2. Black Carbon Measurements of 9% of the black carbondeterminedin ash 1 (ranging from Black carbon was measured in the different residue fractions 0.3% on KP/3 to 8.9% on KPE/1), and hence a maximum contaminationof 6% of the total quantifiedblack carbon.Trace by a two-stepmethod developedby Kuhlbusch[1995] here elementalanalysesby W. Maenhaut(personalcommunication, briefly explained.Duringthe first stepthe sampleswere treated 1994) showedthat soil contaminationwas not significantin the with a seriesof aqueoussolvents(70% HNO3, 1% HC1, 1 M ash 1 and ash 2 fractions. NaOH, H20) to removeall inorganiccarbon(IC) and as much Prior to the determinationof the elementalcompositionwe organiccarbon(OC1) as possible.For severalresiduesamples separated the ashfractionby sieving(0.63x 0.63ram) into ash1 (12 are shownin Table 2) an elementalanalysisafter this first (A1) and ash 2 (A2) (Figure 1). The fractionleft in the sieve step(TC1 and TH1, Table 2) was performed.Thusthe IC and (A2) was found to have a differentcompositionthan A1. A1 OC1 removedby the solventextractionwere obtainedby sub-
consisted of particlessmallerthan0.63 mm diameterand was traction of TC1 from TC. Similar calculations were conducted composed of burnedandthermallytotallydegradedmatter.The for hydrogen.In the secondstep all remainingorganiccarbon A2 fractionwastotally blackenedfrom outsidebut still hadthe (OC2) was removed by a thermal treatment at 340øC for 2 original cell structureof the vegetationand did containsome hoursin pure oxygen.Afterwards,black carbonas well as the "blackhydrogen"(BH) was measuredby an eleless pyrolyzedorganicmatter inside. In comparison,partially corresponding burnedmaterial showedthe originalstructureand color as well mental analysis.OC2 and OH2 were calculatedby subtraction as pyrolyzedparts.Then all sampleswere oven-driedat 80øC of BC andBH from TC1 andTH1, respectively.To quantifythe for 24 hoursto determinethe dry weight. They were milled to inorganiccarboncontentin the residues,a separatefraction an averageparticlesize of 40 Inn andanalyzedfor carbon(TC), (approximately1 g) of the samplewas treatedwith 25% HCI the hydrogen(TH), sulfur,andnitrogen.To avoidany influenceby for 24 hoursto evolve the carbonatesas CO•_.Subsequently moisture depositionon the dried sampleson the H/C ratio, sampleswere oven-driedat 80øC for 24 hours,reweighedafter sampleswere redriedat 100øCfor 3 min. directly prior to the 12 hours acclimatizationin a desiccator,and analyzed for elementalanalysis.Usingthe fuel loadsand the corresponding carbon(TC2). By subtractingthis time TC2 from TC we calcumeasuredcarboncontents,we calculatedthe carbonexposedto latedthe IC contentof the samplesand thusthe OC1 fractionas the fire (CE), total carbonin the residue(TRC), and thus the well. Resultsfor four different plots are listed in Table 2 and volatilizedcarbon(VC). The samecalculationswere conducted will be discussedin section3.1.2. A generalcarbonbalanceof a for hydrogen,sulfur,andnitrogen.Otherabbreviations usedare fire visualizingthe carbon partitioningby the fire and in the smnmarized in Table 1. residueby the analyticalmethodis shownin Figure2.
Table 1. Explanationsof the Abbreviation, Abbreviation
Explanation Elemental
TC+TH TCI
+ THI
BC + BH OCI
+ OHI
OC2 + OH2
TC2 1C
VC, VH, VN, VS CE TRC
removed %.d.m.
mass
Fractions
Totalcarbonandhydrogencontentof a sample Carbonandhydrogencontentof a sampleafter solventextraction Blackcarbonandhydrogen: carbon andhydrogencontentof a sampleaftersolventextractionand O ß thermaltreatmentat 340 C in pureoxygenfor 2 hours Organiccarbonandhydrogen1: carbonandhydrogenremovedby the solventextraction;calculated by subtraction of TC 1 fromTC (TH-TH 1) Organiccarbonandhydrogen2: carbonandhydrogenremovedby thethermaltreatment;calculated by subtraction of BC from TCI (TH-BH) Carboncontentof a sampleafteracidtreatment Inorganiccarbon,carbonates: removedby acidtreatment(TC-TC2) Volatilizedcarbon,hydrogen,nitrogen,andsulfurby thefire;expressed in percentof theelement beingexposedto the fire Carbonexposedto thefire: sumof thecarbonbeingquantifiedin thedifferentvegetationfractions Totalcarbonin the residue:sumof thecarbonbeingquantifiedin the differentresiduefractions Massbeingremovedby the solventextraction Massof theelementor specificelementalfractionexpressed in percentof thedry massof theoriginal sample;e.g., BC 3.5 [%-d.m.] meansthat 3.5 % of the totaldry massof the residuewasblack carbon Residue Fractions
UB
A2
Unburned:residuefractionwhichdid not showany visualalterationdueto the fire Partiallyburned:residuefractionwhichshowedtheoriginalstructure andcoloraswell assome pyrolysedparts Ash;all residuematerialsmallerthan2 cmcollectedwitha handheldvacuumcleaner:totally blackenedfromoutsidebut somematerialstillshowedthestructure of thegrass Ash 2: ashfractionwith a particlesizebiggerthan0.63 mm diameter:blackenedfrom outsidebut still
AI
Ash 1: fineash,ashfractionwitha particlesizesmallerthan0.63 mmdiameter:totallythermally
PB
some unaffected matter inside
degraded
23,654
KUHLBUSCH ETAL.:BLACKCARBON FORMATION BYSAVANNA FIRES
Table2. Results of theElemental Analyses toDetermine theDifferentCarbon Fractions andTheirH/CRatios TC
TH
TCI
THI
IC
%-d.m. %-d.m. %-d.m. %-d.m. %-d.m.
Plot
KPFJ2PB KPE/2A2 KPFJ2AI 56/1PB 56/1A2
27.1 44.6 18.2 27.5 42.9
2.70 3.28 1.13 2.61 3.09
7.62 18.71 10.35 5.81 17.13
0.61 0.84 0.39 0.48 0.69
12.6
KPFJlPB KPFJlA2 KPFJlAI
KP3/2PB KP3/2A2 KP3/2AI
34.0 44.8 13.5
35.4 45.2 14.4
0.86
3.54 3.46 1.03
3.90 3.44 0.96
8.41
6.05 15.11 7.33
8.89 17.35 9.42
0.24
0.58 0.71 0.21
0.94 0.63 0.33
OC2
BC
BH*
%-d.m.
%-d.m.
%-d.m.
0.03
1.30
1.02
0.11
0.09
1.16
0.68
0.15
0.07
1.22
0.50
0.22
0.02
1.18
1.07
0.02
0.09
1.16
0.62
0.15
0.06
3.48
0.35
0.18
0.02
1.32
1.21
0.02
0.08
1.12
0.73
0.14
0.05
1.62
0.34
0.22
0.03
1.46
1.34
0.06
0.12
1.23
0.69
0.14
0.07
1.65
0.49
0.19
0.38
19.14
6.74
0.88
1.4%e
70.5% e
24.9% e
3.2% e
0.88
24.98
13.17
2.0%e
56.0% e
29.5% e
5.54
0.64
7.23
7.55
2.80
39.7% ½
41.4% ½
15.4% e
0.14
21.50
5.07
0.74
0.5% ½
78.3% ½
18.4% ½
2.7% e
1.11
24.68
11.51
5.62
57.5% ½
26.8% e
13.1% e
2.11
2.12
6.14
2.27
16.7% ½
16.8% e
48.6% e
18.0% e
1.23
26.72
3.6% e
78.6%e
OCI
OC2
BCb
H/C mol H/C mol H/C mol
12.4% e
3.5% e
2.6%e
56/1AI
O(21 %-d.m.
5.37
0.68
15.8% e
1.8% e
0.58
29.15
10.35
4.76
1.3%e
65.0% e
23.1% e
10.6% e
0.15
6.02
5.64
1.69
1.1%c
44.6% c
41.8% c
12.5% c
2.44
24.08
8.13
0.76
6.9%c
68.0% c
23.0% c
2.1%c
0.64
27.24
8.74
8.61
1.4% c
60.2% c
19.3% c
19.0% c
0.48
4.54
6.27
3.15
3.3% ½
31.4% c
43.4% ½
21.8% c
Abbreviations areasfollows:%-d.m.,% of thedry massof thesampled fraction; sample abbreviations: exp./jx; exp.,theexperimental site;
j, thenumber of theplot;x, thesampled residue fraction; exp.-- SPS,Shambeni 5;KPE,Kambeni extra; FP4,Faai4; 56,block56;55,block55; other abbreviations see Table 1. anot corrected.
bcorrected forthehydrogen notbondedtocarbon, onaverage 0.022%-d.m.,seesection 3.1.3. cPercent of TC.
Carbon Balance of a Fire
IcO•anic arbonI /
Volatilized Carbon
(0½•).•
Carbon in the Panculate
Total Carbon before theFire
Total Carbon inthe Particulate in Smoke and Residue
Figure 2. Generalcarbonpartit•omng causedby savanna fire. (left) Partitioning of thecarbonexposed to fire intovolatilizedandresidualcarbon.(right)The differentcarbonfractionsin the residueas determined by the analyticalmethod.
KUHLBUSCHET AL.: BLACKCARBONFORMATIONBY SAVANNAFIRES
23,655
Table 3. Elemental Content of the Different Fuel and ResidueFractions Ca-b%-d.m. Fuel Grass
H a.b %-d.m.
N a-• %-d.m.
Sa-• %-d.m.
H/C molar
43.94-1.4
5.74.0.1
0.574-0.14
0.214.0.04
1.554.0.02
44.04-1.3
5.94.0.2
0.534-0.21
0.204.0.05
1.614.0.03
Litter
43.64-3.4
5.44.0.3
0.634.0.16
0.224.0.05
1.494.0.07
herbs
44.14-4.7
5.64.0.8
1.124-0.34
0.214.0.05
1.514.0.08
Residue Unburned
26.44-6.0
2.34.0.7
0.754.0.09
0.154.0.05
1.01M).10
44.14-2.6
5.34.0.7
0.834.0.18
0.184.0.04
1.424.0.12
Partiallyburned
35.94-8.3
3.64-0.9
0.674.0.13
0.154.0.06
1.224.0.09
Ash
21.94-5.3
1.54-0.5
0.804.0.16
0.174.0.08
0.834.0.08
Ash 2 (>0.63 mm)
44.34-2.7
3.24.0.6
1.134-0.19
0.204.0.09
0.864.0.14
AshI ( 1 mm. The main residuefractionwith a particlesize < 0.63 mm is especiallyaccessible for winderosionandis readily 3.1.1. Elemental content of C, H, S, and N and volatilizatransported off site.Detailsaboutthe vegetationtypesand fire behaviorfor the differentexperimental sitesare givenby Shea tion ratios. Table 3 liststhe averageelementalcomposition(C, H, N, S) for the different fractionsof biomassand the residue et al. [thisissue]andTrollopeet al. [thisissue]. The averageC and N contentsin the biomassand residues after the fire (Figure 1). The average mass ratio of grass/litter/herbs for all experimentalsiteswas 1/0.75/0.03,in- are comparableto those of prescribedfires in the tropical dicatingthatherbsdo not represent a significantfractionof the Australiansavanna[Hurst et al., 1994]. Like theseauthors,we averagetotal dry vegetationmass of 6133+2235 kg/ha. The observedan increaseof the N contentin ash comparedto the averagemassratio after the fire for ash/PB/UBfor all experi- fuel.Especially interesting is theincrease of nitrogenin theun3.1.
Biomass Fuel and Residue Characteristics
mental sites was 1/0.71/0.05
with unburned as a minor fraction.
This ratio was significantlydifferentfor site FP4 (1/1.14/0.32), showingthatfor thissitethe PB andUB fractionsrepresent the majorfractionsandhenceexplainingthelow BC/TRCratiosfor thesesitesin Table 6. The site FP4 had by far the lowestvegetation density (Table 6), the greatestpatchinessof the vegetation and thus the lowest
volatilization
ratio
for carbon.
The
burned and A2 fraction. In the case of the unburned matter the
increaseis quite surprisingand indicatesthat in spiteof the absenceof any visual evidencesomealterationdue to the fire
hadtakenplace.A2 as explainedabovewas totallyblackened on the surfacebut still containedsomelesspyrolyzedmatter inside.Thereforethe doublingof the nitrogencontentin A2 (with respectto fuel N) is possiblydueto dryingand elimination processes with the organiccompounds containingnitrogen being far less volatile than other organiccompounds.Cook
average total dry mass of the residue for the six sites was 1055+247kg/ha. In all cases,the ashwas dominatedby the fine fraction.The averagemassratio of A1/A2 was 1/0.2, meaning [1994] made a similar observationin residues of Australian that only about17% of the total massof ashhada particlesize savanna fires. He measured nitroeen contents of 0.4 %-d.m. for largerthan 0.63 ram. The last ratio is in goodagreementwith ash< 1 mm and1.1%-d.m.for ash> 1 min.Alongwith the
23,656
KUHLBUSCH
ET AL.: BLACK CARBON FORMATION
Table 4: Carbon,Hydrogen,Nitrogen, and SulfurVolatilization in Percentof the ElementExposedto the Fire Plot
VCa
VHa
VNa
VSa
SPS/1
96.8ñ0.9
97.9ñ0.6
94.5+1.8
95.9+1.3
SP5/2
91.9ñ2.3
94.9ñ 1.1
89.9+3.3
98.0ñ2.1
SP5/3
87.5ñ6.5
91.8ñ4.4
73.0ñ12.0
91.3ñ4.1
KPE/1
85.4ñ8.8
89.2ñ5.8
76.5ñ14.2
85.2ñ7.4
KPFJ2
84.2ñ2.2
89.7ñ2.5
54.8ñ2.8
82.5ñ1.4
KPE/3
84.0+2.2
89.8+ 1.6
60.8+3.7
81.8+ 1.0
KP3/1
88.1+6.7
93.6+3.8
57.3+25.5
93.5+4.0
BY SAVANNA FIRES
processof pyrolysismay even lead to an enrichmentof carbon can be seenby lookingat charcoalsor forestfires residueswith carboncontentsof up to 80 %-d.m.; this explainsthe carbon
contentof A2. Furtherpyrolyticreactionsare mainly temperature-caused crackingof the fuel moleculesand partial oxidization. That the latter processdid not influencethe generalcompositionis indicatedby the similarity of the H/C ratios for A1 and A2 and stressedby similar BC to TC ratios determinedin A1 andA2 of 17.0+5.2%and 14.6+8.0%respectively. Table4 givesthe amountof carbon,hydrogen,nitrogen,and sulfur released(in percentof elementexposedto the fire) for each plot. Comparingthe volatilizationratios of the different
KP3/2
91.0+1.4
94.3+ 1.4
70.2+3.1
89.0+0.9
elements, we see an increase of the release ratios in the
KP3/3
95.5+ 1.8
97.2+ 1.3
80.1+6.0
94.3+2.8
FP4/1
78.2+3.8
83.7+3.7
74.9+5.2
77.1+2.2
FP4/2
72.0+ 1.0
80.3+ 1.3
68.0+ 1.8
70.4+6.0
FP4/3
87.8+1.6
91.1+1.0
85.0+2.1
74.0+3.3
56/1
89.8+1.7
93.0+1.4
78.1+4.4
87.2+2.9
followingorder:N, S, C, H. The overallvolatilizationratiosfor the elementsare fairly similar to thosein the literature(Table 4). Jenkinset al. [1991] found a higher volatilizationrate for nitrogen relative to sulfur. The volatilization ratio for sulfur accordingto Dehnas [1982] is in the range of 30 to 60%,
56/2
88.3+0.7
92.5+0.8
65.7+7.8
82.6+4.0
whereas we found a volatilization
56/4
92.6+2.2
96.1+0.8
61.2+ 17.5
92.8+2.5
averagearound80%. This may be due to the form in which sulfur is storedin the biomass.Detailed investigationsconductedby Sanbornand Ballard [1991] show that increasing concentrations andproportions of sulfatein Douglasfir foliage significantlyreducethe percentageof total S lost duringcombustionaccordingto the thermal stability of the main salts of sulfate(e.g., K2SO4,Na2SO4, CaSO4).A more detaileddiscussion on the nitrogenand sulfur balanceas well as their emissionsduringSAFARI-92 is givenby Lacauxet al. [thisissue].
55/1
94.6+1.6
96.7+1.1
85.6+1.1
55/2
89.5+2.7
94.0+1.6
68.6+3.3
81.8+3.8
55/3
94.0+1.8
96.0+1.3
84.8+5.1
90.7+1.6
88+6
92+5
74+11
86+8
Average
84.9+3.7
Data for
comparison Austr. Savanna b
96+2
89+5
85.0
Aft. Savanna c
97-98
93-95
35-67
Aft. Savanna a
80.4
Rice Straw ß
94+ 1
79.6 98+0
92+ 1
Correlation 80+8
aVC: volatilized carbon; VH: hydrogen; VN: nitrogen; VS: sulfur: standarddeviationof the threesquaresper plot. bHurstet ai. [1994], Cook [1994]. cDeimas[ 1982]. dLacauxet.ai. [ 1994b]. eJenkinset ai. [ 1991].
ratio of 70 to 98% with an
studies of the volatilization
ratios of carbon to
the other determinedelements(VH, VN, VS) for the different
plotsgavegoodlinear regressions for VH and VN anda fairly goodcorrelationto VS (Table 5). Two plots (FP4/1 and FP4/2) were not used in this correlationstudy becauseof the high amountof unburnedmatter in the residue(see low VC in Table 6). The linear regressionparametersfor the differentelements are given in Table 5 which might be useful to estimatethe releasedS, N, and H.
In Table6, the carbonloadbeforeandafterthe fire, the type doublingof the nitrogencontentin A2 we notea carboncontent of fire, the black carboncontent,the ratio of grasscarbonto thatis similarto thatof the fuel andis on averagehigherthanin litter carbon,andthe ratio of ashcarbonto partiallyburnedplus PB. For sulfur,no significantchangebetweenthe fuel and resi- unburnedcarbon(DoC) for the differentplotsare summarized. due contentwas observed,which is in agreementwith results The average relative standarddeviationsfor the carbon load with rice straw by Jenkinset al. [1991]. The absolutesulfur before and after the fire were 40ø,4and 50%, respectively,decontentin the biomass(0.2%-d.m., dry matter)is at the low end scribing the range of the vegetationdensity in that area. In of the range given by De Kok [1990] for sulfur contentsof general,our grassloads are a factor 1.8+0.4 higher than the plants(0.1-1.5%-d.m.)anda factor2 higherthanthe average averagedatagiven by Sheaet al. [this issue]for the whole exvalue given by Menaut et al. [1991] of 0.1%-d.m. for savanna perimentalsites.This is dueto the fact that we choseour plots grass.
The molar H/C ratios for the different fuel and residue frac-
tionsare givenin the fight-handcolumnof Table 3. The molar H/C ratio was fairly constant(1.6ñ0.1) for the differentfuels (grass,litter, herbs,wood, rice straw). The decreaseof the H/C ratio of the unburnedfraction(1.4ñ0.1mol) is anotherindicator of somechemicalchangesin this fraction.In Figure1 we show the different residue fractionstogetherwith the degreeof charringwhich increasesfrom unburnedto A1. Comparingthe
Table 5. Correlationof VH, VN, and VS to VC Linear Rregression Parameters
Confidence Interval
rl
quite similar. This indicatesthat A2 alreadyunderwentthe initial drying/distillingprocessof pyrolysis,duringwhichmost of the eliminationreactionstake placeas well. That this initial
Slope
(t Test)
VH
0.95
0.32ñ0.01
0.69ñ0.01
VN
0.42
-1.05ñ0.10
1.99ñ0.62
99%
0.33
0.04ñ0.05
0.93ñ0.36
>98%
H/C ratios,a decrease of theratiowith an increase of thedegree VS of charringwas observed.Despite the fact that the carbon Datafor comparison contentsof A1 andA2 differ by a factorof 3, the H/C ratiosare
Constant
Deimaset al. [1994]a 0.60 -0.90+0.09 1.91+1.12 Lobert[1989]a 0.89 -0.80ñ0.03 1.78ñ0.10 Without FP4/1 and FP4/2.
aVN, fromexperimental firesinthelaboratory.
>99%
99%
KUHLBUSCHET AL.: BLACKCARBONFORMATIONBY SAVANNAFIRES
23,657
23,658
KUHLBUSCH ET AL.: BLACK CARBON FORMATION BY SAVANNA FIRES •o
0.85 •
4•7% Slo!•: 1.53:•0.0g
r•: 0.79, Intercept: 18.5%, Slope:-0.00218
8O 7O
•
5
r•: 0.52, Intercept:3.59%, Slope:-0.00044
3o 20 AI: mmov•! ma• 19-40% mr, ov•! cas'oon2345%
0 0
10
i
i
i
2o
3o
4o
0 5o
20
40
60
80
100
R•aov•l Mass[%]
Total CarbonContent[%]
Figure 5. Blackcarbonas a functionof the massremovedby the solventextraction,showingthat the higherthe BC content function of the carbon content of that residue fraction. Solid the less mass is removed. Solid squaresrepresentAI; solid squares represent AI; soliddiamonds, A2; solidstars,PB; and diamonds,A2; solid stars,PB; and asterisksash sampledby asterisksashsampledby Sheaet al. [thisissue]. Shea et al. [this issue].
Figure 3. The massremovedby the solventextractionas a
in areaswith an aboveaveragefuel load to ensurethat this area got burned. 3.1.2. Results of the residue analysis. Plotting the mass removed from the residue samplesby the solvent extraction versus their carbon contents (TC), a positive correlation 0a=0.85, Figure 3) was obtained.The A2 fraction, however, was not included. The slope of the linear regressionline indicatesthat the massremovedby solventextractiondepends
ferentiated.If BC was producedby the pretreatmentsteps,a correlation shouldbe seen,at least within a residue fraction. As no correlation was found and because of blank measurements
able to distinguishthe threegroupsAI, A2, andPB whichwere initially divided becauseof different color and structure.The high variability of the massremovedby the solventextraction for A2 represents the high variabilityof the degreeof charting of this fraction(the higherthe degreeof charting,the lessmass
[Kuhlbusch,1995], it was concludedthat charring(formationof BC by the analyticaltechnique)is negligibleandthatthe measuredblackcarbonwasproducedby the combustion andnot by the analyticalmethod. For comparablecarboncontentsmore masswas removed from partiallyburnedsamples(PB) than from A2 (Figure 3) becausePB sampleswere lessexposedto thermaldegradation. The highvariabilityof A2 (Figures3 and4) andits degreeof chartingindicatethe dependence of especiallythis fractionon fire characteristics like heat transfer,oxygenavailability,and fire type(smoldering versusflaming),asreflectedby themolar
is removed).
CO/CO2 ratio.
on the initial carbon content. Besides this correlation, we are
When black carbon is shown as a function of the total carbon
Figure 5 also showsthat A2 was different from the other content(both in %-d.m.), the samevariability in A2 was ob- residuefractions;it alsohad doublethe amountof nitrogen(see servedas in the massremovedby the solventextraction(Figure section3.1.1). In Figure5, BC (%-d.m.) is plottedagainstthe 4). Again, the threekinds of residues(A1, A2, PB) can be dif- massremovedby the solventextraction.The linear regressions obtained for the A2 fraction and the A I + PB fractions show
that the removedmassdisplayshow much the organicmatter was exposedto thermal degradation.The lessblack carbonwas produced,the more masscould be removedby the solventextraction.The slopeof the linearregressionfor A2 is steeperby a factorof 5. Thesedifferencesin A2 may be associated with the charredsurfaceprotectingthe inner area and with the fact that dehydrationand eliminationprocesses were predominantto oxidation.A2 representedon averageabout 11% of the total mass after the fire but contained about 35% of the total amount
0
10
20
30
40
50
60
Total CarbonContent[%-d.m.]
Figure 4. Blackcarbonas a functionof the carboncontentof that residue fraction. The three different residue fractions can
be differentiatedby the BC content.Solid squaresrepresentAI; solid diamonds, A2; and solid stars, PB.
of blackcarbonproduced. 3.1.3. The different carbon fractions and the corresponding H/C ratios. In Table 2 the resultsof the differentcarbon fractions(IC, OC1, OC2, and BC) for the different kinds of residueson four exemplaryplotsare shown.For all plotsa clear
dependence of OC1, OC2, and BC on the degreeof charting was evident.The structureandthe H/C ratio of the remaining materialwereconsidered asindicators of the degreeof charting (Figure1 andTable 3). BC as a fractionof TC (Table2) clearly increaseswith the degree of charting (A1 > A2 > partially burned)and decreases with increasingH/C ratios(Table 3). As
KUHLBUSCH ET AL.' BLACK CARBON FORMATION BY SAVANNA FIRES
OC1 with an averageratio of 1.2 is close to that of unburned residues(1.4), whereas OC2 (0.77+0.34) and BC (0.13+0.07 correctedfor hydrogen not bondedto carbon, Table 2) have much lower ratios. It is interestingto note the dependenceof the molar I-I/C ratio of OC2 material on the degreeof charting (PB > A2 > A1). This may indicatethat OC2 consistsof two
0.94 Intercept: 0.022%Slope: 0.013
0.30 0.25
o•.• 0.20
carbon fractions with different
0.10 0.05
o
I-I/C ratios and that the carbon
fractionwith the high ratio was consumedin A2 and A1 by the fire. As discussed by Kuhlbuschand Crutzen[1995] this molar H/C ratio is an indicatorfor the degreeof aromaticityand may be useful as a first approximationto predict the resistanceof this carbonto microbial and chemicaldegradation.A decrease in the I-I/C ratios expressesa higher degree of aromaticity leadingup to elementalor graphiticcarbon.An overviewof the
0.15
o
23,659
5
lO
15
H/C
ratios
for the different
residues and carbon
fractions
is
given in Figure 7. Black Carbon[%-d.m.] 3.1.4. Backing/headingfires carbon volatl!lzatlon and BC Figure6. Blackhydrogen asa function of blackcarbon to production. Table 6 containssomeindicatorsof the type of determinethe I-I/C ratio of black carbon. fire and the residues produced. In general, backing fires (spreadingagainstthe wind) havehighervolatilizationratiosfor carboncomparedto heading fires (spreadingwith the wind). The latter spreadmuch fasterand bum biggerareasin the same for BC in Table 2 the samerelationshipfor OC2 but a negative time as backingfires, albeit they are less efficient in burning one for OC1 was observed,indicatingthat BC and at leastpart vegetation.If we comparethe type of fire with our measured of OC2 were producedin the field fires. We emphasizeagain, VC values(90.8+3.7% headingfires; 86.9+7.1% backingfires) astreatedin detail by Kuhlbuschand Crutzen[ 1995], that some no significantdifferencesare observed.This may be due to the of the OC2 carbonmay, like black carbon, decomposeonly inhomogeneityof the vegetationon the experimentalsites and very slowly by microbial action.The actual carbonpartitioning the high variability in headingfires. The ratio betweencarbon differs widely dependingon the burning conditions;BC can in the ash residue to carbon in the partially burned and unrangefrom 3-18% of the average270+110 kg TRC/ha as can be burnedresidue(DoC) did not showany relationshipto VC or to seenin Table 6. Factorsdeterminingthe burningconditionsare, the type of fire. e.g., wind speed,humidity, fuel moisture, fuel density, fuel Laboratory experiments [Kuhlbusch and Crutzen, 1995] composition,and heat release.Since OC1, OC2, and IC were showedfor similarvegetationhigherratiosof blackcarbonas a not quantifiedfor all plots, we give the rangesobtainedfrom fractionof the total carbonin the residue(BC/TRC) for backing the plots listed in Table 2 which are 62+10%, 26+6%, and comparedto heading fire. Cachier et al. [1994] stated that 3+1% of TRC, respectively.BC in Table 2 comprises8.7+4.4% headingandbackingfirescannotgenerallybe distinguished just of the total residualcarbon.The amountof inorganiccarbonis by the BC/TC ratio in the aerosols but "samples collected under not dependenton the degreeof charring.It was in the rangeof burningregimes 0.5-6.9% of total carbon(the exception56/1 AI, 16.7%, may be backingfire conditionsduringwell established from larger fires, seem to produce aerosols significantly due to woody characteristicsof the material). This inorganic carbondoes not representan importantmass fractionbut is enriched in black carbon" thus indicating similar formation partly responsiblefor the alkalinecharacterof ashand aerosols. laws of BC in aerosolsand combustionderivedresidues.During CO2 may be producedfollowing the depositionof acid rain or SAFARI-92, residuesof heading as well as backing fires on four experimentalsiteswere collected.Two sitesshowedhigher its mixing with acid material. Before discussingthe molar H/C ratios for the different BC/TRC ratios for backing and two for heading fire. We carbonfractions,correctionsare requiredfor the H/C ratio for measuredaverageBC/TRC ratios of 12.0+5.0% and 8.3+3.7% for seven heading and 11 backing plots, respectively,thus BC, sincesomemeasuredhydrogenis bondedto elementsother than carbonas shownby Kuhlbusch[1995]. Therefore"black" hydrogenwas plottedversusblack carbonin Figure 6 (both in 2.0 %-d.m., dry matter). BC measurementsof savanna grass sampledprior to the fire are givenas "blanks"in Figure6. The massratio of 0.013 derivedfrom the regressionslope(Figure 6) 1.5 was used to calculate the correct H/C ratio for black carbon. An
averagemolar H/C ratio for BC of 0.16 was obtained,which is slightlyhigherthan that derivedfrom experimentalfires (0.11 [Kuhlbuschand Crutzen,1995]). This H/C ratio is a goodindicator of higherpolyaromticor elementalcarbonderivedfrom combustion processes. The interceptof 0.022+0.014%(Figure 6) represents the amountof hydrogenbondedto elementsother thancarbon,mostprobablysilicate.Additionally,we showwith a correlationcoefficient(r2) of 0.94 that the BC fractioncan be definedby the I-I/C ratio. It is possibleto differentiate thethreecarbonfractions(OCI, OC2, andBC) in Table 2 whencomparingthe molar I-I/C ratios.
i l.0
:::::....:.. g ........... ...... ..........
o
.... !.....:5i'"-5•i-" i...... '...... •................... ::."...:•• ..... ."'.'-:.. ........ Orsms
LIB
PB
A2
AI
O(21
OC2
BC
Residue and Carbon Fractions
Figure 7. The molar H/C ratio for the different residueand carbon fractions. For abbreviations
see Table 1.
23,660
KUHLBUSCH ET AL.' BLACK CARBON FORMATION BY SAVANNA FIRES
g•ving no significantindicationof the dependenceon the type of fire. Similarly, no dependence of the conversionratio BC/CE on the type of fire (heading1.00ñ0.38%;backing0.94ñ0.48%) was observed.Thus no dependenceof the type of fire and the productionof BC was observedunder field conditions.We believe that this is due to the high variability of the fire regimes, the few measurements,and especiallythe patchiness of the vegetationon the plots. 3.2. RelationshipBetweenBC Formation and the Molar CO/CO2 Emission Ratio
by the amountof thermallyunaffectedmatter.Table 7 summarizes the BC/TC ratios of the ash fractions and the cortes,
ponding CO/CO2ratiosmeasured by Lacauxet al. [thisissue] and Ward et al. [this issue].Lacauxet al. [this issue] determined the CO/CO2ratioswith power-autonomous mobile pumpingunits, a NDIR-CO-CO2spectrometer, and emissions
sampled via a poleheldin theplmnes.Wardet al. [thisissue] usedfixed-upright polesstanding in the field, sampling the smokein canisters whenthefireburned through. TheCO/CO2 emission ratiosdetermined by Lacauxet al. [thisissue]were measurednear the samplingplots and were used for ashes
sampled byus.Theratiosdetermined by Wardet al. [thisissue] In residuesderivedfrom variousexperimentalfuresunderlaboratoryconditions[Lobeftet al., 1991] a negativecorrelation betweenthe formationof BC andthe emissionratio of CO/CO2 as an indicator of the burning conditions was obtained [Kuhlbuschand Crutzen, 1995]. This indicatesthat flaming ratherthansmolderingcombustion is the mainsourceof BC. To
wereusedfor ashessampledby Sheaet al. [thisissue]around the fixed poles.
Forthefollowingcorrelation studyit is important to notethe variabilityof the CO/CO2ratioevenwithinoneexperimental site andcombustion phase.The lack of oxygenandthermal quenching especially duringheadingfiresmay causedramatic
comparethisfindingwith the field data,the BC/TC ratioof the changes in thecombustion conditions expressed by CO/CO 2
ash fractionwas used,which eliminatedany differencescaused ratiosof up to 20%asobserved by oneof theauthors (H.C.).
Table7: BlackCarbon andCO/CO 2 Emission Ratio Plot
BC/TC for ash,
CO/CO2,
Typeof Fire
Thusmeasuring theCO/CO2rationotexactlyfor thesampled residue mayleadto higherrorsandexplainthelackof any relationship between theBC/TCratioandtheCO/CO2ratiofor mixedtypeof fires.Thecombustion process is notsubject to suchdramatic changes duringbacking fireconditions, resulting in a negative regression whenonlybackingfire samples were considered (ra: 0.62; intercept:24.2%; slope:-1.89; eight degrees of freedom). To improvethisrelationship, additional
%
mol-%
SP5/I
12.99
4.80
B
SP5/2
10.33
5.83
H
measurements andintegrativesamplingof residuesandthe cor-
SP5/3
12.92
5.83
H
responding emissions in thefieldwouldberequired.
KPE/I
11.91
KPE/2
15.02
H 4.58
B
KPE/3
20.92
4.58
B
KPF• a
12.21
7.10 b
H
KPF•I a
10.56
5.30 b
H
KPF•II
11.09
a
Byusingthemeasured CO/CO2 ratiosandassuming thatCO + CO2accounted for 95% of the quantifiedvolatilizedcarbon
(CE(kilograms perhectar)-TRC (kilograms perhectar)), it was possibleto computetheemittedCO2.Thisenabledus to calcu-
latetheBC/CO2ratios(in gramsC-BC/grams C-CO2)foreach
H
plot as shownin Table 6. Thesedatahave to be lookedat as a
first approximation. On averagethe ratio of BC to CO2is 1.3ñ0.6%.Thisratiois at the low endof the rangeof those obtained fromlaboratory studies (BC/CO2 1.2-2.6%[Kuhlbusch
KP3/I
16.44
4.80
B
KP3/2
21.45
5.73
H
KP3/3
25.57
5.73
H
KP3/I a
24.51
6.20 b
H
KP3/II a
18.94
8.60 b
H
KP3/III a
and Crutzen,1995]).
20.11
7,20 b
H
3.3. Relations Between Black Carbon Formation, Carbon
FP4/I
11.94
6.91
B
Volatilization, and Degree of Charring
FP4/2
10.14
6.91
B
FP4/3
9.79
6.91
B
56/1
17.11
3.41
B
56/2
16.65
3.41
B
56/4
16.47
3.41
B
56/I a
14.80
7.00 b
H
56/II a
17.26
8.00 b
H
In this sectionwe will discussthe mechanisms andquantitativeaspectsof BC formationin the residues.Figure8a shows BC as a fractionof the total carbonin the residue(BC/TRC) plotted againstVC. The positive correlationin Figure 8a indicatesthatthe BC fractionof TRC becomesmoreimportant with an increasing volatilizationratioof carbon.Thisis in good
56/III a
14.41
7.50 •
H
agreementwith results found for the different residuefractions
(A1, A2, partiallyburned,seeabove)and to a laboratorystudy [Kuhlbuschand Crutzen, 1995]. In this laboratorystudy an equationwasderivedwith the assumptions that the residuesare homogeneous andthat low VC valuesare accompanied by less thermal degradationof the fuel, thus forming less BC. The followingequationwasused:
55/1
25.12
H
55/2
27.12
H
55/3
19.01
55/Ia
12.30
6.50 •
B
55/II a
19.30
8.10 •
55/III a
19.50
H H H
CO/CO2:measured by Lacauxet al. [ 1996]. Abbreviationsare asfollows:H, headingfire, fire spreading with thewind;B, backingfire, fire spreading againstthe wind. aSampled by Sheaet al. [thisissue]. t•Measured by Wardet al. [thisissue].
BC/TRCm•
BC/TRC=
(1) A
(vcm - vc)
+1
A is a constantterm, BC/TRCmax represents the maximum BC/TRCratio,VC is thetotalvolatilizedcarbon,andVCm is
KUHLBUSCH ET AL.: BLACK CARBON FORMATION
BY SAVANNA FIRES
23,661
2.5 the VC value at which BC/TRC is half of BC/TRCmax.With BC/TRCmax=I6.3%, VC]a-85.3%, andA-I. 15, we calculateda (0.ss) 2.0 correlationcoefficient(xa) of 0.38. This correlationlies within a confidence interval of 99% (t test, 16 degrees of freedom)[Miller and Miller, 1986]. The BC/TRCmaxobtained 1.5 FP4/2 (0.7]) from the laboratorydatais about10% lower, whereasVC!a is .... • .... FP4/] of similar magnitude(laboratoryvalue 88.4%). The compara1.0 tively low value of BC/TRCma x duringthis field studycan be explainedby the fact that laboratoryfire residueshad hardly 0.5 any unburnedor partiallyburnedfraction. (0.49) Figure8b showsthe conversionratio BC/CE plottedagainst 0 70 75 80 85 90 95 100 volatilized carbon (VC), showinggood agreementwith the Volat'fiized Carbon[%] laboratorydata [Kuhlbuschand Crutzen,1995]. By multiplying equation(1) (see above) with TRC/CE (CE=100%) we derive Figure 9. The conversionfactor as a functionof VC. The two equation(2) (Figure8b): lines indicatethe dependence of the conversionfactor of VC (site FP4) and on DoC (site KPE), with DoC values in BC/TRCmax 100 - VC BC/CE = x (2) parentheses. (VC]t2- VC) 100
.............
....I 'a4)
i
A
i
+1
With a correlationcoefficient(ra) of 0.24 the dataare within a confidence interval of 95% (t test, 16 degrees of freedom)[Millerand Miller, 1986]. The fit is not a good one, reflectingthe uncertaintyin the valuesof CE and TRC (section 3.•.•). The wide scatterof the datain Figure8 comparedto thoseof the laboratorystudyis due to severalfactors.In part it can be explainedby the large standarddeviationof TRC and CE (50% and 40%, respectively)from which VC was calculated.The error in BC/TRC was alsohigherin the field studythan in the laboratorystudy [Kuhlbuschand Crutzen, 1995]. During the
0•'3'vc)+1 la)1.15]6.3
laboratorystudiesthe massburnedwasmeasuredcontinuously duringthe whole experimentand only one integrafiveresidue samplehadto be analyzed.In comparison, threedifferentfractionshadto be sampled,weighed,andanalyzedduringthis field studyto calculatethe BC/TRC ratio. Another reason for the above mentioned wide scatter in the
field data comparedto the laboratory study could be the inhomogeneityof the residuescollectedin this field campaign. Comparingthe residuesof the experimentalfires [Kuhlbusch and Crutzen,1995] to thosesampledduringSAFARI-92 there were hardlyany partiallyburnedandunburnedfractionsin the laboratoryexperiments. Thereforean indicatorof the degreeof charting (DoC) was defined by calculatingthe ratio of ash carbonto partiallyburnedandunburnedcarbon(Table6):
C(•h)
= DoC (of the wholeresidue) (3)
C(unbumed +partially burned) This ratio shouldbe calculatedexcludingcarbonates(IC); otherwise,potash(mainlyK2CO3,residuesfrom woodburning)
E
would have an infinite DoC, VC < 100%, and a BC/TRC ratio
of zero. But for this investigationDoC was calculatedwithout excluding carbonates,which producedno significant error becauseof generallylow IC contents. On averagewe find a DoC ratio of 1.4, rangingfrom 0.3 to 3.0 (Table 6). As alreadymentionedfor the laboratorystudywe I i I hardly had any partially burnedor unburnedmatter, giving an infinite but constantDoC ratio. Thus it is not surprisingto find 1.1501•.3-vc)+1 100 2.0 relatively goodcorrelationswith equations(1) and (2) for the ß 55/2 laboratoryresidueswith uniform charting.This differencein KPE/3KP3/I © the residuesof the laboratoryand field studymay be significant KPE/2•56/2 KP3/2 E 1.5 for the use of experimentallaboratoryfires, e.g., concerning SP5/••-.•,• 56/4 nutrientcycling. One reasonfor the comparablehigh amountof 1.0 FP4/I KPE/l - 56/1••5•1ß • UB and PB, besidesthe high variability of the combustion ß FP4/3 © SP•5/2 •'• process in the field, is that especiallythe aboveground biomass 0.5 nearthe soil surfaceremainsunburnedor partiallyburned. To investigatethe dependence of the conversionratio BC/CE i i i 85 90 95 100 0 70 •80 75 on VC and DoC in more detail, data from two experimental VolatilizedCarbon[%] sites,FP4 and KPE were plotted in Figure 9. Figure 9 shows that plots on FP4 have nearly constantDoC values while the Figure8. Blackcarbonin percentof thetotalresidual andof ratio decreases with increasing VC. In .contrast, VC the fire-exposed carbonas a functionof the total volatilized conversion carbon.The curveswere obtainedby fitting equations(1) and valuesfor KPE are nearly constant,andthe conversionratio increaseswith the degree of charting. However, it is not as (2) to thedetermined data. i
ß
55/3 x•
23,662
KUItLBUSCH
ET AL.' BLACK CARBON FORMATION BY SAVANNA FIRES
carbonpeak(~500øC)in the PB sampleis straightforward to showthecombined effectof VC andDoCon the low temperature the formationof BC. Furtherinvestigation to getmoredetailed of similarheightto the high temperaturecarbonpeak(~600øC), it nearlycompletelyvanishesin the A1 sample(Figure10). The informationon the formationof BC in residuesis necessary. Anotherparameter whichmightinfluencethe formationof same results can be seen for hydrogen.The thermogram BC is whetherthe burnedmatterwasgrassor litter. For all plots obtainedfrom the A1 sampleshowsthe same peaks and has the carbonratiosof grassto litter are shownin Table6, andit nearlythe sameshapeas one obtainedfrom ashof pine needle seemsthat no significantdependence of the formationof BC litter [Kuhlbusch,1995]. In section 3.3 we defined an indicator for the degree of betweengrassandlitter exists.Still suchan interpretation must be considered cautiouslysincemostof the litter wasdeadgrass chartingby the mount of carboncontainedin the different which was not very differentfrom the yellow standinggrass. residuefractions.The thermogramsand the observationspreOnly about5-10% of the standinggrasswas living, "green" sentedin Figure 1 indicatethat each residue fraction has a differentdegreeof charring.In section3.1.3 we saw that BC grass. and to some extend OC2 were combustion derived. Thus we
defined the following ratio as an indicator for the degreeof charringwithin a residuefraction.
3.4. Thermograms
In Figure10,thermograms takenof PB,A2, andA1 residue samplescollectedon plot KPE/1 are shown.The different carbonfractionsfor eachresiduefractionof KPE/1 are given in
Table 2. The heating rate for these thermogramswas 108øC/min.with a gasflow of 20 vol.% 02 in N2 (for details see Kuhlbusch[1995]). The data for hydrogenand carbonin Figure 10 are plottedon the samescaleso that equivalent heightsexpressa molarH/C ratioof 1. In Figure1 we expressed the degreeof chartingby the color andstructureof the sampledfractionandsawan increasefrom PB to A2 and then to A1. The same order of the degree of
OC2+BC
= DoC
(within a residuefraction)
(4)
OC1
With the data given in Table 2 we calculatedan average degreeof chartingfor PB of 0.32+0.08,for A2 of 0.65+0.1,and for A1 of 2.2+1.3. The two indicatorsfor the degreeof charring
(equation(3) + (4)) arenot directlycomparable but increasethe more the organicmatter was thermally affected. Since the degreeof chartingis constantwithin a residuefraction,the degreeof chartingfor the wholeresidueis determinedby the mass share of each residue fraction.
chartingcanbe seenin the thermograms (Figure10). Whereas
4. Estimation of Black Carbon production by Savanna !.6
- a: PB
Fires
0.8
:f•
...... Hydrogen
:
1.2
0.6
0.8
0.4
0.4
0.2
The global BC formationin the residuesof savannafires is estimatedin three different ways: (1) by the BC/CO2 ratio (1.3%, seesection3.2); (2) by the BC/CE-conversion ratio; and (3) by the BC/TRC ratio. The rangefor the BC/CE-conversion ratio derivedfrom Figure8b for an averagevolatilizationratio of carbonof approximately 90% (89% for the six experimental fires duringSAFARI-92) was between0.5 and 2%. To estimate
0.6
!.2
the formationof BC by savannafires in Table 8 by usingthe exposedcarbonweightingan averageconversionratio of 1% was used.The averageBC/TRC ratio derived from Table 6 is 9.8+4.6%.Thusa ratioof 10%wasusedin Table8 whenusing the amountof residueleft on the ground. A comparisonof the BC/CO2 ratio of 1.3% for the black
0.2
-
c:A!
carbon formation in the residue to the one for BC in the carbo-
naceousaerosolfraction (0.11% [Cachier et al., 1994]) indicatesthatmorethan90% of the BC producedby a fire remains on site in the residues.A BC/CO2 ratio of 0.11% was usedin
-
Table 8 to calculate the amount of BC emitted in the smoke. !.2
0.6
0.8
0.4
0.4
0.2
o 100-:.......... :........ 300
500
700
Temperature [øC]
-............ o 900
In Table 8 we summarizeregionaland worldwideestimates of vegetationaffectedby savannafires. With theseestimates and the abovementionedfactorswe calculatedthe production rate of BC. In the secondcolumnfrom the right we averaged the different estimates of BC in the residues and added the BC
emittedin the smoke.About70% of all BC formedby savanna firesis producedin Africa. On average,18 Tg BC (10-26 Tg BC) is annuallyproducedby savannafires. This formationrate is 4-52ø,4of the total amountof BC producedannuallyby vegetationfires worldwide (50-270 Tg BC [Kuhlbuschand
Figure10. Thermograms of thedifferentresidue fractions on plotKPE/1.Equivalent heights forcarbon andhydrogen express Crutzen,1995]). The BC formedby savannafiresrepresents a net 0 2 source a molarH/C ratioof 1. The peakat temperatures above700øC of 27-69Tg yr-I anda CO2sinkof 10-26Tg C yr-l. Thecarbon is mainlydueto inorganic carbon, e.g.,calciumcarbonate.
KUHLBUSCH ET AL.: BLACK CARBON FORMATION BY SAVANNA FIRES
23,663
23,664
KUHLBUSCH ET AL.: BLACK CARBON FORMATION BY SAVANNA FIRES
freedprior to the fire in the vegetation was part of the biospheric carbon cycle.Byconversion to BCit is sequestered fromthe short-term to the long-term(geological) carboncycle.
Blackcarbon produced in savanna firesrepresents a netcarbon sinkconsidering thatperiodicsavannafiresdo not represent a net sourceof atmospheric CO2because of theregrowthduring the growingseason.
ClimaticImportanceof Vegetation Fires,editedby P. J.CrutzenandJ. G. Goldammer,pp. 193-213,DahlemWorkshop,JohnWiley, New
York, 1993. Comeor,J. A., G. R- Fahnestock,and S. G. Pickford,Elementalcarbon
deposition andflux fromprescribed burningon a longleafpinesitein Florida,in final report to the National Centerfor Atmospheric Research,Collegeof ForestResources, Universityof Washington, Seattle,WA, 1981.
Cook,G. D. The fateof nutrientsduringfiresin a tropicalsavanna, Aust. J. Ecol., 19, 359-365, 1994.
Cope, M. J., and W. G. Chaloner,Fossilcharcoalevidenceof past 5. Summary and Conclusions atmospheric composition, Nature,283, 647-649, 1980. Crutzen,P. J.,andM. O. Andreae,Biomass burningin thetropics:Impact In this field studythe C, H, S, andN loadingsin the biomass on atmospheric chemistorandbiogeochemical cycles,Science,250,
beforeand after the fires were determined,and averagevolatili-
1669-1678, 1990.
zation ratios for these elements of 88+6%, 92+5%, 86+8%, and
De Kok, L. J., Sulfurmetabolism in plantsexposedto atmospheric sulfur, in SulfurNutritionandSulfurAssimilation in HigherPlants,editedby 74+11%, respectively,on the six experimentalsitesobtained. H. Rennenberg, C. Brunold,L. J. De Kok, and I. Stulen,pp. 111-130, By partitioningthe residuein unburned,partiallyburned,A2 SPBAcad.,The Hague,Netherlands, 1990. (> 0.63 ram), and AI (< 0.63 ram) fractions,a correlation Delmas,R. A., On the emissionof carbon,nitrogen,and sulfurin the atbetweenthe degreeof charringof a residuefractionand the mosphere duringbushfiresin intertropical savannahzones,Geophys. Res.Lett., 9, 761-764, 1982. corresponding molarH/C ratio was established. A2, comprising approximately11% of the residualmass, was chemically Delmas,R. A., P. Loudjani,A. Podaire,andJ.-C.Menaut,Biomassburning in Africa: Assessment of annuallyburnedbiomass,in Global different from the otherresiduefractions.A2 was significantly BiomassBurning,editedby J. S. Levine,pp. 126-131,MIT Press, enrichedin nitrogenand carbon,and had a molar H/C ratio Cambridge,Mass.,1991. (0.86+0.14) of that of A1. Black carbonwas quantifiedin the Delmas,R-, J.-P.Lacaux,J. C. Menaut,L. Abbadie,X. Le Roux,G. Helas, different
residue
fractions.
Various
correlation
conducted to derive detailed information
studies were
on formation mecha-
andJ. M. Lobeft,Nitrogencompound emission frombiomass burning in tropicalAfrican SavannaFOS/DECAFE91 Experiment(Lamto, Ivory Coast),J. Atmos.Chem.,in press,1994.
nismsand quantitiesof black carbon.Comparingthe results Edwards, J.B.,Combustion: Formation andEmission of TraceSpecies, obtainedfrom this field studyto thosefrom experimentalfires Technomic,Lancaster,Pa., 1974. underlaboratoryconditions,it was foundthat the field residues Fearnside,P.M., Greenhousegas contributionsfrom deforestationin BrazilianAmazonia,in Global BiomassBurning,editedby J. S. were much more inhomogeneous and of different structure. Levine,pp.92-105,MIT Press,Cambridge, Mass.,1991. Thus,besidesthe dependence of the formationof black carbon on the volatilized carbon a second factor was identified, the
Hao, W. M., M.-H. Liu, and P. J. Crutzen, Estimates of annual and
regionalreleases of CO2 andothertracegasesto theatmosphere from
degreeof charring.This latter factormight affect nutrientsin firesin the tropics,basedon the FAO statistics for the period1975the vegetationfire residuesas well. Basedon theseresultsand 1980,in Fire in the TropicalBiota,editedby J. G. Goldammer, pp 440-462,Springer,New York, 1990. variousestimatesof carbonexposedor releasedby savannafires we estimatetheworldwideblackcarbondeposition in theresidueof Herring, J. R-, Charcoalfluxes into Cenozoicsedimentsof the North Pacific,Ph.D. thesis,ScrippsInst.of Oceanogr.,Univ. of Calif., San savannas to be 9-24Tg C yr-l. Addingthe 1-2Tg BC yrl in the Diego,La Jolla,1977. smoke of savannafires, 10-26 Tg black carbon are produced Houghton,R- A., Tropicaldeforestation andatmosphere carbondioxide, Clim. Change,19, 99-118, 1991. annually.Sincethis blackcarbonwill not be brokendownover from geologicaltimescalesit representsa net sink of atmospheric Hurst,D. F., D. W. T. Griffith,andG. C. Cook,Tracegasemissions biomass burningin tropicalAustralian savannas, J. Geophys. Res.,99, CO•.anda sourceof 27-69 Tg O2 yr•. 16441-16456, 1994. Acknowledgments.The authorswould like to thankthe staff of the Kruger National Park, A. Potgieter,and L. Trollope for enablingand
organizingSAFARI-92in the KrugerPark.We alsowish to thankK. WeiB,R- Wirz, H. Abberger,andU. Petersen for theircollectionof vegetationandresiduesamples,and W. Dindorffor theelementalanalysesfor S andN. Thanksare extendedto M. Berges,R. Cox, and the reviewersN. Cheney and P. Khanna for reviewing and certainly improvingthe manuscript.It was a greatpleasureto work with all the peopleinvolvedin SAFARI-92.
References
Jenkins,B. M., S. Q. Turn, R. B. Williams,D. P. Y. Chang,O. G. Raabe, J. Paskind,and S. Teague,Quantitativeassessment of gaseousand condensedphase emissionsfrom open burning of biomassin a combustion wind tunnel,in GlobalBiomassBurning,editedby J. S. Levine,pp. 305-317, MIT Press,Cambridge,Mass.,1991. Kauffman,J. B., D. L. Cunnings,D. E. Ward, and R- Babbitt,Fire in the Brazilian Amazon: Biomass,nutrientpools, and lossesin slashed primaorforests,Oecologia,in press,1995. Kuhlbusch,T. A. J., Method determiningblack carbonin residuesof vegetationfires,Environ.ScL& Techn.,29, 2695-2702, 1995. Kuhlbusch,T. A. J., and P. J. Crutzen,Towarda globalestimateof black carbon in residuesof vegetationfires representinga sink of
atmospheric CO2 and sourceof O2, GlobalBiogeochem. Cycles,9, 491-501, 1995. Cachier,H., C. Liousse,P. Buat-M6nard,and A. Gaudichet,Particulate Lacaux,J.-P.,H. Cachier,andR- Delmas,Biomassburningin Africa:An contentof savannafire emissions, J. Atmos.Chem.,in press,1994. overviewof its impacton atmospherechemistry,in Dahlem WorkChang,S. G., R. Brodzinsky, L. A. Gundel,andT. Novakov,Chemical shopReports:Fire in the Environment.' TheEcological,Atmospheric, andcatalyticproperties of elementalcarbon,in ParticulateCarbon: and ClimaticImportanceof VegetationFires, editedby P. J. Crutzen Atmospheric Lifecycle, editedby G. T. Wolff andR. L. Klimisch,pp. andJ.G. Goldammer,pp. 159-188,JohnWiley, New York, 1993. 159-181, Plenum,New York, 1982.
Cheney,N. P., R. J. Raison,andP. K. Khanna,Release of carbonto the atmosphere in Australianvegetation fires,in CarbonDioxideand
Lacaux, J.-P., D. Brocard,C. Lacaux, R- Delmas, A. Brou, V. Yobou6, and M. Koffi, Traditionalcharcoalmaking:An importantsourceof
Climate:AustralianResearch,editedby G. I. Pearman,pp. 153-158,
atmospheric pollutionin the Africantropics,Atmos.Res.,35, 71-76,
Aust. Acad. of Sci., Canberra, 1980.
1994a.
Lacaux, J.-P., J. M. Brustet, R. Delmas, J. C. Menaut, L. Abbadie, B. Novakov,T. L. Cook,andB. D. Hinds,Aerosolandgraphiticcarbon Bonsang,H. Cachier, J. Baudet, M. O. Andreae, and G. Helas,
Chylek,P., V. Srivastava, L. Cahenzli,R. G. Pinnick,R. L. Dod, T.
contentof snow,J. Geophys. Res.,92, 9801-9809,1987. Clark,J. S., andJ. Robinson, Paleoecology of fire,in DabletoWorkshop Reports:Fire in theEnvironment: TheEcological, Atmospheric, and
Biomass burningin thetropicalsavannas of Ivory Coast:An overview of the field experimentFireof Savannas (FOS/DECAFE91), J. Atmos. Chem.,in press,1994b.
KUffLBUSCH
ET AL.: BLACK CARBON FORMATION
Lacaux,J.P., IC Delmas,C. Jambertand T. A. J. Kuhlbusch,NOx emissions fromAfricansavannafires,d. Geophys. Res.,thisissue. Lobert, J. M., Verbrennung pflanzlicher Biomasse als Quelle atmosph•rischer Spurengase: Cyanoverbindungen, CO, CO2, und NOx,Ph.D.thesis,Johannes Gutenberg-Univ. Mainz,Germany,1989. Lobert, J. M., D. H. Scharffe, W. Hao, T. A. J. Kuhlbusch,IC Seuwen,P. Warneck, and P. J. Cmtzen, Experimentalevaluation of biomass
BY SAVANNA FIRES
23,665
in savannaecosystems of SouthAfrica andZambia,d. Geophys. Res., this issue.
Sternberg, H. O., Man and environmental changein Soutl•America,in Biogeography and Ecologyin SouthAmerica,editedby E. J. Fittkau, T. S. Elias,H. Klinge,C. H. Schwabe, andH. Sioli,pp.413-445,SPB Acad.,The Hague,Netherlands,1968.
Suman,D. O., T. A. J. Kuhlbusch,and B. Lim, Marine sediments:A reservoirfor black carbon and their use as spatial and temporal burningemissions:Nitrogenand carboncontainingcompounds,in recordsof combustion,in Sedimentrecordsof biomassburningand Global BiomassBurning, editedby J. S. Levine, pp. 289-304, MIT globalchange,editedby J. S. Clark,H. Cachier,J.G. Goldammer, and Press,Cambridge,Mass.,1991. B. J. Stocks,Springer,in press,1996. Mack, F., The influenceof vegetationfires on the global carboncycle, Ph.D. thesis, Inst. for Plant Ecol., Justus-Liebig-Univ.,Giegen, Trollope,W. S. W., A. L. F. Potgieter,N. Zambatis,and L. A. Trollope, SAFARI-92: Characterization of biomass and fire behavior in Germany,1994. controlledburns in the Kruger National Park, d. Geophys.Res., this Menaut, J.-C., L. Abbadie, F. Lavenu, P. Loudjani, and A. Podaire, issue. Biomassburning in west African savannas,in Global Biomass Burning,editedby J. S. Levine, pp. 133-142, MIT Press,Cambridge, Ward, D. E., W. M. Hao, IC A. Susott,IC A. Babbitt, IC W. Shea, J. B. Kauffman, and C. O. Justice, Effect of fuel compositionon Mass., 1991. combustionefficiency and emissionfactors for African savanna Miller, J. C., and J. N. Miller, Statistics Jbr analyticalchemistry,pp. 89ecosystems, ./. Geophys. Res.,thisissue. 91, EllisHorwood,Chichester, England,1986. Ogren,J. A. and IL J. Charlson,Elementalcarbonin the atmosphere: Cycleandlifetime,Tellus,35, 241-254, 1982. M. O. Andreaeand J. G. Goldammer,Biogeochemie,Max-PlanckPatterson, E. M., C. K. McMahon,andD. E. Ward, Absorptionproperties andgraphiticcarbonemissionfactorsof forestfire aerosols,Geophys. Institut •r C'hemie,Posffach3060, 55020 Mainz, Germany. (e-mail: moa•cco.caltech.edu; jggold•ru f.uni-freiburg.de) Res. Let., 13, 129-132, 1986. H. Cachier, Centre des Faibles Radiactivites, Laboratoire mixte Penner,J. E., H. EddiemanandT. Novakov,Towardsthe developmentof a globalinventoryfor black carbonemissions, Atmos.Environ.,27,4, CNRS-CEA, avenue de la Terrasse,91198 - GIF sur Yvette CEDEX, France.(e-mail:Helene.Cachier•cfr. cnrs-gif. fr) 1277-1295, 1993. P. J. Crutzen, Luffchemic, Max-Planck-Institutfiir Chemic, Postfach Sanborn,P. T., and T. M. Ballard,Combustion lossesof sulphurfrom coniferfoliage:Implications of chemicalformandsoilnitrogenstatus, 3060, 55020 Mainz, Germany. T. A. J. Kuhlbusch,NERL, U.S. EnvironmentalProtectionAgency, Biogeochemistry, 12, 129-134,1991. (e-mail: Seiler,W., and P. J. Crutzen,Estimatesof grossand net fluxesof carbon 960 College Station Rd., Athens, GA 30605. Thomas•epamail.epa.gov) betweenthe biosphereand the atmospherefrom biomassburning, Kuhlbusch. J.-P.Lacaux,Laboratoired'A•rologie,CNRS-UPS,14 avenueEduard Clim. Change,2, 207-247, 1980. Sharrow,S. andH. A. Wright,Effectsof fire,ash,andlitteron soilnitrate, Belin,31400Toulouse,France.(e-mail:lacjp•aero.obs-mip.fr) R. Shea,2855 NW JohnsonAvenue,Corvallis,OR 97330. temperature,moistureand to Tobosagrass productionin the rolling ,,
planes,J. RangeManage.,30, 260-270, 1977. Shea,IC W., B. W. Shea,J. B. Kauffman, D. E. Ward, C. I. Haskins, M. C. Scholes,Fuel biomassand combustionfactorsassociatedwith fires
(ReceivedJuly20, 1994;revisedJuly8, 1995; acceptedJuly8, 1995.)
