Paststudieshave extensivelycharacterizedchamise and otherchaparral species.Severalstudiesexaminedfuel characteristics of one or more chaparral species (Countryman and Philpot1970,Nord and Countryman ChamiseChaparralDead Fuel FractionIs Not Reliably Predictedby Age 1972, Wakimoto 1977, Rundel and Parsons 1979), but none of these studiesquantitativelydescribeda rela- tion of age to fraction dead. Other chaparralstudiesprovideage-related information concerning the dynamics of the stand speciescomposition Timothy E. PaysenandJackD. Cohen,• PacificSouthwest ForestandRangeExperiment Station, USDA ForestService, Riverside,CA 92507. ABSTRACT.Fire managers of southern California chaparral oftenassume thatthe amountof deadmaterialin chaparral shrubsis closelyrelatedto canopyage. Analysisof chamise (Adenostomafasciculatum),sampled from southern Californiashrublands, indicates thattheratio of deadto live components is not related reliablyto ageof shrubcanopy.Further statistical description of thedataindicates thatfractions ofdeadmaterial greater than 0.40arerare.Theresults provide grounds for seriously questioning currentassumptionsaboutthestrongrelationship between ageandfractiondeadin southern Californiachaparral fuels. (Hanes 1971, Patric and Hanes 1964, Stohlgrenet al. 1984),but the fraction of thedeadvegetative fuelwasnotde- of other species.This age-dynamic chaparralfuel modelhasbeendistributed to field personnel through classeson prescribedburningand fire behavior in California. scribed. Several authors believe that the fraction of dead vegetationincreaseswith age, but did not present quantitative studies (Philpot 1974, Green1981,Hanes1974). Also, the southern California chaparralfuel model option (SCAL) in the FIRECAST (Cohen 1986) computerprogram is based on the dynamic fuel model. FIRECAST The literature did not contain the information needed to resolve our questionaboutageand fractiondead, so we conducteda field study that sampledthe dead vegetativefraction of chamisechaparralwith regardto canopyage.Thisarticlereportsthe resuitsof thatfieldstudy. is a fire behavior predictor generallyavailableto fire managersfrom variousfire agencies throughthe computersystemof the CaliforniaOffice of EmergencySer- METHODS vices. Althoughthe conceptof a relationship between age of chaparraland fractionof deadfuel hasbeenwidely acceptedby fire managers,contradictory observationsduring prescribed Naturalresource managers of burns and wildfires compelledus to southernCaliforniashrublands (chap- questionit. Our initial field observaarral)addressthe planningaspects of tions of the fraction dead in various wildlandfire by usingthe conceptof age classesof chaparralindicateda age classmanagement.Wildland fire high degreeof variabilityand genermanagersassumethat the dead vege- allylesstotaldeadfuelthanpredicted. tativefuel is the mostimportantchar- We initiallyturnedour attentionto the acteristic thatinfluencesfire intensity. literaturefor guidancein determining And, they assumethat the older the whethera relationshipbetweenage Chamisewas chosenfor this study on the basis of three considerations: WestJ. Appl. For. 5(4):00-00, October1990. vegetativecanopy, the greater is the fraction of dead fuel. These two as- (1) chamisechaparraldatastronglyinfluencedthe age/fractiondead relationshipbelievedto exist;(2) Hanes (1971) indicated that chamisewas the most common chaparral speciesin southern California, and (3) Coun- trymanand Philpot(1970)quantified fuel information,includingthe fraction dead, for a 33-year-oldstandof chamisethat couldserveas onepoint for comparison. Two typesof samplingprocedures and fraction dead exists. were followed. The first was a multi- sumptions formthe basisfor management of chaparralfuel by age class (HunterandPhilpot1982).The goalof managementby age classis to keep S o u t h e r n chaparralstandsyoung enough to prevent accumulationsof dead fuel that resultin unmanageable, highly destructive wildland fires (Rogers El_ ' 1982). Chaparralfire managers havebeen usinga mathematical modelthat provides various fuel descriptors, including the fraction of fuel that is dead, as a function of canopyage (Rothermeland Philpot1973,Philpot 1974).Althoughthis modelis largely basedon chamisechaparral(Adenostomafasciculatum) data, it is generally appliedto chaparralstandscomposed • r• _ .... CaI i fo;;:r:i!i.n i aMojave•Nev•lda ..?' Desert ...;./ r a n s.v.::•"'" •" • ..'.,',?,l:',"•':.:.::?:.':,+",..,.....q n Pa c i f i c o oe a ß '........ ':.':':.?Z:,: "e '•"'"'"":";;"" .... •"Z • 'l, • .-.+ n. • 'R. •'.."•.In '.•"."a "."i.'•"vC'.:i'."-'i:':•. ..:',.. e •o•oraao ;,:::..:,.,.. :::','a ' Coastalandinlandintensive samples xNow at the Southeastern ForestExperi- Thirty-one shrub sample Moxico ment Station, USDA Forest Service, De- partment of Agriculture, Macon, GA 31020. Figure1. Symbols indicate thelocations of samplesitesof chamise fractiondead. WJAFS(4)1990 127 Canopyages,for the sitessampled, rangedfrom9 to 100years.The samplinglocations werechosenarbitrarily with respectto severalfactors:(1) the availabilityof chamiseas the dominant shrubspeciesin the area, (2) our abilityto identifycanopyagefromage classmaps, (3) the representation of 1 A 0.8 therangeofknowncanopy ages(cur- 0.6 ++ 0.4 rent fire history maps limit this to 1-100 years),(4) the representation of the geographic and ecological (based mostlyon climateat this stageof the procedure)range of southern Californiachamise sites,and(5)ourability + 0.2 to access the site for shrub collection. o 22 44 66 88 With the use of vegetationmaps,fire history maps (for age class),dimate zonemaps,androadmapsfor accessibility, a generallocationwas chosen throughan iterativeprocess until all of 110 1 B 0.6 the above criteria were met. General locations were never chosen from on- 0.6 0.4 siteinspections. Using the intensive, site-specific, + samplingprocedure,a sampleof size ten was taken from each of two dif- ferent southernCaliforniachaparral sites.The generallocationswere arbitrarilyselected as in the multisiteprocedure,but with two major exceptions:(1) the chamise-dominated sites representedcanopy ages of 30-39 years sincethe last fire, and (2) the generallocationsrepresenteda gen- 0.2 o o 22 44 66 66 i c 0.8 eral climatic contrast between the coastal zones and the interior moun- tains. Within the selectedgenerallocations,specificsiteswere randomly 0.6 selected. 0.4 0.2 0 22 Individual shrubs, the sampling units for both samplingprocedures, were randomlyselected.The random shrub was selectedby choosinga random directionand distance(within 44 66 68 110 CANOPY AGE (years) Figure 2. Scatterplots of the 31-shrubsample(A) and intensivesampleplots at the coastaland interior mountains(B) and North Mountain (C) displaysampleresultsfor chamisefractiondeadby canopyage. site procedure that sampled individual chamiseshrubsextensively across southern California. The secondproceduresampledthreespecific southern California sites inten- The multisiteproceduresampled31 chamise southern shrubs from 15 different California locations within areasvegetatedby chaparralcommunity series(Paysenet al. 1980,1982). 100m) froman access point. If the resuitingvectordid notlocatean appropriatechamisesiteand shrub,the processwasrepeatedfromthelocationof the unusable sample location. The sampledshrubwas cut, bagged,and removedfrom the sitefor analysisat the USDA Forest Service Fire Labora- tory, in Riverside,CA. The shrub analysis followed the procedures usedby Countrymanand Philpot(1970).Shrubmaterial,foliage, and stemslessthan 76 mm, wasseparatedinto livingand deadcategories. The materialwas then put into containersfor oven-dryingovera 24-hour sively, with a greater number of shrubs taken from each site than was the casefor the first procedure.The intensivesampleswere used to provide insightinto the kind of withinsitevariabilitythat mightbe foundin southern California standsof chamise, and to providea comparison for evaluating the multi-site data. Figure 1 identifies locations samplesweretaken. 128 WJAF 5(4)1990 where shrub Table1. Summary statistics for all chamisesampledatasetsof fractiondead. Data set Statistic Median Mean S.D. Samplesize 3'1-shrub Coastal mtns. North Mtn. North Mtn. sample age 3'1yr age 36 yr age 33 yr age 55 yr 0.228 0.223 0. '120 0.249 0.237 0.092 0.259 0.259 0.053 0.259 0.244 0.088 0.303 0.286 0. '107 31 10 Interior 10 mtns. 16 16 Table2. Sampleestimates andbootstrap summary statistics for the mean,standard deviation, and median,for the 31-shrubdataset.The summary statistics were run on 500 bootstrap the actual data, and Table 2 presents estimates(0') of each statistic. timates. The small values for the stan- Summarystatistics Statistic estimates Mean dard errors of the mean, median, and S.D. Median 0• 0.233 0.118 0.228 Mean (0*) 2 0.230 0.117 0.220 S.E. (0*) 0.021 0.020 Bias (mean (0*) - 0), -0.003 indicatesthe sampleestimateof the statistic of interest. 0.022 -0.001 -0.008 O*represents a bootstrap estimate of the statistic of interest. period at 95øC.The dried vegetation was then weighed to determine its summarystatistics of the bootstrapes- RESULTS the standard deviation for the boot- strapsample,as well as the low bias values,imply that the actual31-shrub sampleis not significantly influenced by outliersor unevennessin the data. Similarly,the bootstrapresultssuggestthat the Kendall'stau measureof correlation providesa reasonable estimate of the relationshipbetweenthe fraction dead and the age. Figure 3 summarizes the actual and the boot- The plot of the 31-shrub,multisite strapestimateof Kendall'stau. A posAs a complement to thetwo sample data revealslittle, if any, relationship itive tau between 0.0 and 1.0 indicates sets described above, data were inbetween the fraction dead and age a positive relationshipbetween the cludedfromthe originalCountryman (Figure2). Inspectionof the site-spe- fractiondead and age. The resulting cific data sets shows that the varitau of 0.225, while statisticallydifand Philpot (1970) chamise fuels ferent from zero (with a two-tailed abilityfound on a specificsiteis simstudy, and from our resamplingof ilar to that found in the multisite data, probabilityof 0.078),indicatesa very their site. Countryman and Philpot and the rangeof the specificsitedata weak association. The values for the sampled 16 chamiseshrubsfrom a single site whose vegetativecanopy is comparableto that of the multisite bias and the standard error of tau indata. dicate that the Kendall's tau estimate age was 33 years old. Twenty-two The results of the bootstrap proceis reasonable for the 31-shrubsample. years later, we sampled16 chamise dure suggestthat unevennessand Furtherdescriptionof the multisite shrubs from their original site (the outlyingvaluesin the 31-shrub,multishrubsample,withoutregardto age, North Mountainsamplesin Figure1) influindicatesa strong clusteringof the whosecanopyage was then 55 years site sampledo not significantly old. ence the estimates of the mean, medata.Figure4, a histogram of the frequencyof occurrence of the fraction To establisha quantitativemeasure dian, and the standard deviation. dead, showsthe 31-shrubsampledata of of the relationshipbetweenfraction Table1 presentssummarystatistics deadand agein our multisitesample, we used a nonparametric coefficient, mass. Kendall's tau, as a measure of correlation. We chose not to use the usual correlationand regressiontechniques From 31 data points: because there was no reason to believe tau=.225 From bootstrapof 500 replications(n=31) mean(tau*)=.214 that the restrictions imposedby these techniques(for example,linearityin std. dev.(tau*)=.153 bias(tau)=mean(tau*)-t the case of the linear correlation coeffi- cient, r) were appropriate.Kendall's tau respondsonly to monotonicity, and thereforedoes not demand specificfunctionalrelationships (suchas linearity)of the data. Becauseof the small samplesize, we performeda bootstrapresampling procedure (Efron 1982, Efron and Gong1983)on the 31-shrubsampleto checkthe internalconsistency of the data as it might affectthe measureof correlation. The resampling comprised500 iterationsof resampling with replacement from the 31 pairsof fractiondead/agevalues.The size of each samplewas 31. For each iteration, an estimate of Kendall's tau was calculated; to summarize these estimates, mean and standard errors of the tau estimates lated. were then au =.214-.225=-.016 *bootstrap estimates 8O • 64 LU 32 LU U. 16 calcu- For the samemultisitesample,we calculatedsummarystatisticson the fraction dead data to generally de- -0.5 -0.2 0.1 0.4 0.7 TAU scribe the occurrence of the dead fuel componentfor southernCalifornia chamisechaparral.The fractiondead was describedusing the bootstrap procedureon the samplemean, median, standarddeviation, and quantiles(10ththrough90thpercenttie). Figure3. A histogram of the500bootstrap estimates of Kendall'stau. Thevalueof tau calculated from the actual field data (n = 31) is 0.225. From bootstrap estimatesof 500 replications, mean(tau*)= 0.214;S.D. (tau*)= 0.153;bias(tau)= mean(tau*)- tau = 0.214 - 0.225 = -.011. (Tau* indicatesthe bootstrapestimateof tau.) WJAF5(4)1990 129 rehableage esbmateexistsfor other chaparralspecies. Our resultsshowa symmetricdistribution 12 of chamise fraction dead valuesarounda relativelylow valuein the fraction dead scale. This distribu- tion suggeststhat 80% of the time we will find stands of chamise with frac- tion deadvalueslessthan or equalto 0.31--regardlessof canopyage. RECOMMENDATIONS Southern California vegetation managersrequiringan estimateof the fraction o 0.5 FRACTION DEAD Figure 4. Histogramshowingthe frequencydistributionof the actual31 chamisefrac- this distribution, the mean is 0.233 and the median is 0.228 (from Table 1). The closeness of the meanand median values indicates that the data dis- suitsof this studybring into serious questionthe efficacyof ageclassmanagementin chamisechaparralif that management is basedon an expected age-fractiondead relationshipas a predictorof potentialfire behavioror CONCLUSIONS Our data indicatethat a general fractiondead/agerelationship is very other effects. weak for southern California chamise chaparral(Adenostoma fasciculatum)-tribution is somewhatsymmetrical. at leastwithin the agerange(9 to 100 This concurs with the low measure of years)sampled.Thus,ageof chamise fractiondead/agecorrelation provided canopydoesnot providea reliableesby Kendall'stau. timateof the fractiondeadof vegetaMean percentlies,developedfrom tive fuel. Since the current method of the bootstrapprocedure,were usedto usingageto estimatethe fractiondead establisha cumulativefrequencydisis basedon chamisecharacteristics, no tribution and other not recommended. Further, the re- tion dead data. distributed around a central value. For dead of chamise chaparral speciesshould obtain an averagefraction dead estimatefrom on-sitesampling.The useof a canopy agemap (e.g., a fire historymap)as a tool for predicting fraction dead in chamiseand otherchaparralstandsis o for fraction dead, which The lackof an age/fraction deadrelationshipfor chamisechaparraldoes not negatethe factthata fractiondead exists.The lack of an agerelationship indicatesthat otherinteractingfactors suchas yearlyclimatevariability,insects, and diseasecould be responsible for the fraction of dead fuel. Be- causewe sampledover a relatively allowsestimationof the frequencies of fractionsdead relative to particular values.For example,asseenin Figure 5, 50% of the fractionsdead fall below approximately0.22, and 80% of the fractions dead fall below approximately 0.31. Fractions dead approaching0.50and greaterare rare. An inspectionof all datasets,summarized in Table 1, using box and whisker plots (Figure 6) indicatesa 10o 80, prevailing tendencyof the fraction dead data to cluster around a central value.With boxand whiskerplotswe cancomparethe samplesetswith regard to dispersion,centraltendency, and skewness(Emersonand Strenio 1983).Inspectionrevealsthat the me- m 60 ,r, "" ,r, 40 dian value for each data set is central to the rangevalues,andcentralto the values that delimit the central 50% of the data.Theoverlapof valueswithin the central 50% of the data for each data set, and containment of all data set medians within the 95% confidence limits for the North Mountain (age55 years)median,givesthe impressionthat all sampleswere from a singlepopulation.This overlapalso suggeststhat site, at the stand level (asopposedto the shrublevel),is not a dominantfactorin the development of fraction dead. 130 WJAF5(4)1990 o o.o 0.1 0.2 FRACTION 0.3 0.4 DEAD Figure5. Bootstrapestimateof the cumulativefrequencydistributionof chamisefraction deadfor the 31-shrubsample. COrn,FrRYMAN,C. M., AND C. W. PHmPOr.1970. Physicalcharacteristics of chamiseas a wildland fuel. USDA For. Serv.Res.Pap.PSW-66. 16 p. EFRON, B. 1982.Thejackknife,thebootstrap and other resamplingplans. Soc.Indus. Appl. Math. Philadelphia.92 p. EFRON,B., ANDG. GONe;.1983.A leisurelylook at the bootstrap,the jackknife,and cross-validation. Am. Stat. 37(1):36-48. 0.8 EMERSON, J. D., ANDJ. S•,•Nuo. 1983.Boxplots 0.8 and batchcomparison.P. 58-96 in Understanding robustandexploratory dataanalysis, D. C. Houglin,F. Mosteller,and J. W. Tukey 0.6 0.6 (eds.).Wiley, New York. 447p. GREEN,L. R. 1981.Burningby prescriptionin 0.4 0.4 0.2 0.2 0.0 0.0 chaparral.USDAFor. Serv.Gen.Tech.Rep. PSW-51.36p. PlANms,T. L. 1971. Successionafter fire in the chaparral of southern California. Ecol. Monogr.41(1):27-52. PlANms, T. L. 1974.The vegetationcalledchap- arral.p. 1-5 in Proc.syrup.onlivingwiththe chaparral.SierraClub, SanFrancisco. HUNTS. R, S.C., ANDC. W. PHILPtrr.1982.Fire be- havior and managementin Mediterranean- typeecosystems: a summary andsynthesis. P. 520-522in Proc.syrup.on dynamics andmanagementof Mediterranean-type ecosystems. USDAFor.Serv.Gen.Tech.Rep.PSW-58. NORD,E. C., ANDC. M. COUNTRYMAN. 1972.Fire Sample relations.P. 88-97 in Proc.syrup.on wildland shrubs--theirbiologyand utilization.USDA For. Serv. Gen. Tech. Rep. INT-1. PATiUC, J. H., ANDTSDL. •. 1964.Chaparral suCcessionin a San Gabriel mountain area of Fiõure6. Boxandwhiskersplotsfor all dataon chamisefractiondeaddiscussed in this paper.Thelenõthof theboxapproximates theinterquartileranõeof thedata.The cross bar showsthe median,with notchesapproximatinõ95% confidencelimits for the median. Ninety-ninepercentof the data pointsfall within the ranõeindicatedby the "whiskers." Outliers are indicatedby asterisks. California.Ecology 45(2):353-360. PAYSEN, T. E., rr At. 1980.A vegetation dassificationsystemappliedto southernCalifornia. USDA For. Serv. Gen. Tech. Rep. PSW-45. 33 p. PA¾sWN, T. E., J. A. DœP,S¾, AND C. E. CONIReD. 1982.Vegetation classification system forusein California:its conceptualbasis.USDA For. short time period (1984-85), we cannotbe certainasto how persistent the value of centralclusteringand the for example, we assumea level of fractiondead (basedupon a canopy age map) and that assumptionis not distribution correct, then observed fire behavior of fraction dead are. Our descriptions of the fraction dead shouldbe viewedasthe sampledcondition and subjectto change. Important new understanding about southern California chaparral would include how the fraction dead is producedand how longit takesfor significantfractionsdead to accumulate. This information would indicate for how longsampleestimates of fraction dead would be viable. Giventhe lackof any consistent relationshipbetweenage and fraction dead in chamise,we as fire professionalsneedto reexamineourperceptionsof the way fractiondeadaffects fire behaviorin these stands--especiallyif thatunderstanding isbasedon field observation of fire behavior. If, should not fit our expectations--if there is a necessary relationshipbetween fire behavior and fraction dead. If, however, the fire behavior does fit our expectations under thesecircumstances,then perhapsfire behavioris not as heavilydependenton the fraction dead as we have been led to believe. We are forced to conclude that fire behavior variations in chamise chaparralare more complexthan can be describedby a singlefuel characteristiclikethefractionof deadvegetation. LiTERATURE [] user's manual. Francisco. ROGERS,M.J. 1982. Fire management in southernCalifornia.P. 496-501in Proc.syrup. on dynamicsand management of Mediterranean-typeecosystems. USDA For. Serv.Gen. Tech.Rep.PSW-58. RcrrrmRM•r,R. C., ANDC. W. •. USDA Gen.Tech.Rep.PSW-90.11p. For. Serv. 1973.Fire in wildland management,predictingchanges in chaparralflammability.J. For.71:164-169. RuNo•I,, R. W., AND D. J. PAI•ONS.1979. Struc- tural changesin chamise(Adenostoma fasciculatum)alonga fire-inducedage gradient.J. RangeManage.32(6):462-466. STOrtLGi•œN, T. J., D. J. PARSONS, ANDP. •/. RUNDEr. 1984.Population structureof Adenostomafasciculatum in maturestandsof chamise chaparral in the southernSierraNevada,California.Oecologia64:87-91. WAKIMO?O, R. R. 1977.Chaparralgrowthand fuel assessment in southern CITED Corms,J. D. 1986.Estimatingfirebehaviorwith FIRECAST: Serv.Gen.Tech.Rep. PSW-63.14 p. PHna,o?, C. W. 1974.The changingroleof fireon chaparrallands.P. 131-150in Proc.symp.on living with the chaparral.SierraClub, San California. P. 412-418in Proc.syrup.on the environmental consequences of fire and fuel management in Mediterraneanecosystems. USDA For. Serv. Gen.Tech.Rep.WO-3. WJAFS(4)1990 131