Productivity Model for Cut-to-Length Harvester Operation ... · for. eng. 39(2018)1 1 ... There has been a concerted shift from traditional motor-manual and semi-mechanised timber

Croat. j. for. eng. 39(2018)1 1

Originalscientificpaper

Productivity Model for Cut-to-Length Harvester Operation in South African

Eucalyptus Pulpwood PlantationsJennifer Norihiro, Pierre Ackerman, Ben D. Spong, Dirk Längin

Abstract

There has been a concerted shift from traditional motor-manual and semi-mechanised timber harvesting systems to mechanised cut-to length (CTL) operations in South Africa. This is particularly true in Eucalyptus pulpwood felling and processing, South Africa’s largest com-mercial wood resources used in the pulp and paper industry. Mechanisation improvements are typically driven by increasing safety regulations, product quality and productivity concerns related to traditional harvesting systems. The objective of this study is to develop productivity models for mechanised Eucalyptus pulpwood CTL felling and processing operations by combin-ing the results of a number of individual studies done over a period of 24 months in the summer rainfall areas of South Africa. The study takes into account species, machine type (purpose built vs. excavator based), silvicultural practices (planted vs. coppiced) and slope. The pooled data revealed general productivity ranges from 5.16 m3 PMH-1 to 27.49 m3 PMH-1.

Keywords: cut-to-length, eucalyptus, pulpwood, full-mechanized system, productivity study

planted (18millionha in90countries)andvaluedhardwood,thereremainsaglobaldeficiencyofpub-lisheddataonmechanisedEucalyptusharvesterop-erations(FAO2006).AstheSouthAfricanindustryhasrapidlytransitionedtofullymechanisedCTLopera-tions,therehasbeenaneedtodeterminetheinfluenc-ingfactorsthataffectharvesterproductivitywithinaSouthAfricansetting.Inareviewofscientificandpeerreviewedpublications,domesticandinternational,atotalof13articleswerefoundtoberelatedtofullymechanisedharvester-basedEucalyptusoperations,buttheywereinconsistentinrecordingdatainonewayoranother.Althoughinconsistent,thesestudiesidentifiedand

analysedinfluencingfactorsthatarevitaltounder-standingharvestingproductivity.Factorsincludetreevolume (Spinelli et al. 2010), species composition(Nurminenetal.2010),equipment type(SirenandAaltio2003,Spinellietal.2010),sitecharacteristics(Puttocketal.2005,Andersson2011),silvicultureprac-tices(KelloggandBettinger1994,Ramantswanaetal.2013), operator training (Ovaskainen et al. 2004,

1. IntroductionCommercialforestryhasexperiencedaglobalshift

towardmechanisedharvestingoperations(FAO1997,Nurminenetal.2006,Jiroušeketal.2007).ThischangehasalsooccurredintheSouthAfricanForestIndustry,withthekeydriversbeingforestworkerhealthandproductquality.Withthistransition,therehasbeenanincreaseinstudiesdealingwithtimberharvestingandtransportproductivityaimedatdeterminingandmod-ellingequipmentproductivity.Theseinvestigationscanprovidethemeanstooptimiseeconomicgainsandvolumeyieldstomanagersandcontractors(WilliamsandAckerman2016).Althoughamultitudeofresearchrelatedtomechanisedharvestingsystemshavebeenconductedinternationally,littleresearchhasbeenpub-lishedinrelatedoperationsinSouthAfrica.InSouthAfrica,Eucalyptusisthepredominantge-

nususedforpulpwoodanditaccountsfor83%ofthecommercialwoodresourcesforthepulpandpaperindustryinSouthAfrican(FES2011,FSA2013).Al-thoughEucalyptusisconsideredthemostcommonly

J. Norihiro et al. Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13)

2 Croat. j. for. eng. 39(2018)1

PurfürstandErler2012),delimbinganddebarking(HartsoughandCooper1999).AccordingtoSpinellietal.(2010),treevolumehas

beenidentifiedasthemostsignificantvariabletode-termineharvesterproductivityandisareliablepredic-torofproductivity.Additionalstudiesnotonlyveri-fied this, but suggested that production rate ispositivelycorrelatedtoincreasingtreevolume(Akayetal.2004,ErikssonandLindroos2014).Otherprojectsuseddiameteratbreastheight(DBH)asthecontinualpredictorofproductivity,whichmadeitdifficulttocomparewithstudiesthatusedtreevolume(McEwanetal.2016,AcunaandKellogg2009,HartsoughandCopper1999).Literaturealsofoundoperatorperfor-manceasaninfluencingfactortoharvesterproductiv-ity,butithasbeenchallengingtoquantifybecausetrainingisnotstandardisedglobally(Ovaskainenetal.2004,PurfürstandErler2012).ThehumanfactorandworkshiftwereconsideredbyPassicotandMurphy(2013),butoperationsobservedconsistedoftreevol-umeexceedingthecommonSouthAfricanrangetobeapplicable.Inaddition,productivitywasoftenrecord-edasm3PMH-1,butinHartsoughandNakamura(1990)andAcunaandKellogg(2009),productivitywasre-cordedasbonedrytonneperscheduledhour(BDT/SH)ortonnesPMH-1withnoinformationonthemachineused.Terrain,morespecificallyslope,wasidentifiedinsomeofthestudiesandproventohaveaconsider-ableeffectonproductivity(DavisandReisinger1990,Spinellietal.2002,AcunaandKellogg2009).InAcunaandKellogg(2009),slope,rangingfromgentletomod-erateslope,wasidentifiedasasignificantfactor,butproductivitywasrecordedinconsistentlywhencom-paredtootherliterature.Despiteafewfactorswithineachpublishedpaper

applicabletoaSouthAfricancontext,mostwerein-consistentlyrecordedandcouldnotbeusedasapre-dictorofproductivitytrends.Asameanstoaddressthelimitedliterature,theindividualstudiesperformedinSouthAfricawerecombinedinanattempttode-velopgeneralproductivitymodels.Theobjectiveofthisstudyistodevelopgeneral

productivitymodelsformechanisedEucalyptuspulp-woodCTLharvesting(fellingandprocessing)opera-tionsbycombiningtheresultsoffiveindividualandindependentproductivitystudiescompletedoveraperiodof24monthsinEucalyptusclearfellingpulp-woodstandsinthesummerrainfallareaofSouthAf-rica.Thisstudywilltakeintoaccountspecies,silvicul-turalpractices(plantedvs.coppiced),machinetype(purposebuiltvs.excavatorbased)andslopeinherentinthefivestudies.

2. Material and methods

2.1 Case studiesFiveindividualproductivitystudysiteslocatedin

thenorth-eastofSouthAfricawereincludedinthisstudy.Thesiteshavebeensequentiallynumberedandreferredtobythisnumberingthroughoutthispaper(Fig.1).ThesestudiescoveredfourdifferentspeciesofEucalyptusandwereallclear-fellingpulpwoodcom-partmentsthatwereharvestedduringthedrywintermonths.Onlytwocomponentsoftheharvesteropera-tionwereconsidered:fellingandprocessing.Thefourspeciesharvestedincluded:Eucalyptus grandisxcamal-dulensis(GxC),Eucalyptus grandisxurophylla(GxU),Eucalyptus smitthii (ES)andEucalyptus dunnii (ED).Furtheroninthisstudy,specieswillbereferredtobytheiracronym.Harvestingsitescoveredadiverserangeofterrain

(slope),treecharacteristics(species,form,individualtreevolume)andharvestermachinetype(excavatorbasedandpurposebuilt)inordertoincorporatesiteconditionsandfactorsthatcontributetoproductivitytrends(Table1).Eventhoughthefiveindividualstudieshadvaryingoriginalobjectives,thedatawascollectedusingastandardisedtime-studyprotocol(Ackermanetal. 2014) that enables comparisonsbetween thestudies.TheobjectiveofStudy1wastodetermineproduc-

tivitydifferencesbetweenoneandthreepassdebark-inganddebranchingoperationinaGxCclonesoneventerrain.TheobjectiveofStudy2wastodeter-mine productivity differences between excavatorbasedandpurposebuiltmachinesonvaryingslopeterrain in a GxCclone.TheobjectiveofStudy3wastodetermineproductivitydifferencesbetweenthreeandfivepassdebranchinganddebarkinginaGxUcloneoneventerrain.TheobjectiveofStudy4wasapureproductivitystudyofanexcavatorbasedharvesting

Fig. 1 Locations of study areas

Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13) J. Norihiro et al.

Croat. j. for. eng. 39(2018)1 3

machine,fellingandprocessingpoorformESoneventerrain.TheobjectiveofStudy5wastodeterminepro-ductivitydifferencesbetweenthreeandfivepassde-barkinganddebranchingpassesoperationinEDoneventerrain.Debarkinganddebranchingpassesaredefinedasthenumberoftimestheharvesterheadtravelsalongthetreestemdebarkinganddebranch-ing.Thelastpasswillentailcross-cuttinginlogas-sortments.

2.2 Time studyDifferentresearcherscollectedtimestudydataat

eachofthestudyareasaccordingtotheSouthAfricanForestIndustryTime-studyStandard(Ackermanetal.2014).FieldtimestudyobservationswererecordedusingaTrimbleGeoXThandheldcomputer.Timere-cordedwascategorisedintooneoffourelementsiden-

tifiedinthestandard:fell,process,moveanddelay(Table 2).Allmachine operators, althoughnot thesameinallstudies,wereconsideredtrainedandca-pableofoperatingtheharvesterinEucalyptuspulp-woodoperationsconsistingoffelling,debarking,deb-ranching and crosscutting into assortments.Delaytimeswererecordedregardlessofduration.Producti-vityresultswereexpressedinproductivemachinehours(PMH).Individualtreevolume(m3)wascalculatedus-ing the Schumacher andHallmodel (Bredenkamp2012). Individual tree and compartment attributesrecordedarereflectedinTable1.Inthisstudy,slopeisconsideredasacontinuous

variable.ContinuousslopedatawereobtainedfromDigitalTerrainModels(DTMs).Thesemodelswerederived from large-footprintLiDARdatawith ap-proximate1mresolution.

Table 1 Individual site and stand characteristics of the five studies

Site characteristics Study 1Study 2

Study 3 Study 4 Study 5Study 21 Study 22

SpeciesEucalyptus grandis x

camaldulensis(G x C)

Eucalyptus grandis xcamaldulensis

(G x C)

Eucalyptus grandis xcamaldulensis

(G x C)

Eucalyptus grandis xurophylla(G x U)

Eucalyptus smithii(ES)

Eucalyptus dunii(ED)

DBH, cm

Average 15.5 15.3 16.3 21.6 15.9 16.4

Min. 7.0 9.0 7.3 8.6 5.2 8.0

Max. 21.2 27.2 25.3 29.1 35.7 30.5

SD 2.3 2.7 2.9 3.8 4.6 5.0

Age, y 12 8 8 9 7 12

SPH, n ha-1 987 1001 926 1087 1106 826

Average height, m 16.3 19.88 20.03 25.4 17.4 18.5

Average tree volumem3 tree–1 0.12 0.15 0.15 0.38 0.14 0.15

Slopea, %(continuous variable)

Level(0–10)

Level – very steep(0–61)

Level – very steep(0–61)

Level (0–10) Level (0–10) Level (0–10)

Silvilculture Planted Planted-Coppice Planted-Coppice Planted Planted Coppice

Carrier type Excavator Purpose Built Excavator Excavator Excavator Excavator

Machine manufacturer Hitachi Zaxis 200 Timberpro TL-725B Volvo EC-210bf Hitachi Zaxis 200 Hitachi Zaxis 200 Komatsu PC 200

Head Waratah H616 Maskiner SP 591-LX Maskiner SP 591-LX Waratah H616 Maskiner SP 591-LX Maskiner SP 591-LX

Location Zululand Melmoth Melmoth Kwambo KZN Midlands Piet Retief

Sample size 297 1156 1099 181 1478 177

a Slopes are classified using the National Terrain Classification for Forestry (Erasmus 1994)

J. Norihiro et al. Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13)

4 Croat. j. for. eng. 39(2018)1

Table 2 Time study elements breakdown (Ackerman et al. 2014)

Time element Description

FellStarts when the operator begins moving the head to a tree, ends when the butt end begins to move through the head

ProcessStarts when the butt end begins to move through the head, ends when the head has released the last piece of the tree

MoveStarts when the tracks begin moving, ends when the tracks come to a stop

DelayStarts when the machine unexpectedly stops working, ends when work begins again

2.3 Experimental designAteachofthefivestudysites,diameteratbreast

height,measuredoverbark(DBH),wasrecordedforeverytreeusingadiametertapewithanaccuracyof0.1cm.WhilemeasuringDBH,eachtreewasallocatedauniquenumberperstudyareainordertoidentifyeachtreewhenrecordingcycletimesduringtheac-tualharvestingofthesamples.Heightsofatleast50representativetreespersite,chosenfromvariousloca-tions in the allocated compartment and spanningacrosstherangeofDBHavailable,weremeasuredus-ingaHaglofVertexlaserhypsometerwithanaccu-racyof0.1m.TheheightsandDBHoftheserepresen-tativetreeswereusedtoderivearegression,whichallowedtheheightsoftheremaining,notmeasuredtrees,tobeestimatedbasedontheDBHmeasuredforeachtree.Everytreewasnumberedtofacilitatethepairing

oftreedimensionswithfellingandprocessingtimestocalculateproductivity(m3PMH-1).Numberswerepaintedontreestemsatanangletoensurevisibilityduringtiming.Priortoharvesting,arandomisedblockexperimentaldesign(RBD)(ClewerandScarisbrick2001)wasappliedtoeachstudyareatoreducebias.

2.4 Statistical analysisBasicstatistics,correlationanalysisandlinearre-

gressionmodellingwereperformedtodetermineandclarifyvariablesaffectingharvesterproductivity.Treevolumewasusedasthecontinuouspredictorforre-gressionmodelswithadditionalcorrelationanalysesappliedtoidentifythesignificanceofvariables,suchasspecies,carrier type,silviculture,slope,andde-barkingpassonproductivity.Wheresignificantfac-torswereidentified,additionalmodelsweredevel-oped.Asasecondaryanalysis,multipleregressionanal-

ysiswasconductedtobetterfitthedataset.Thepooleddatasetwascategorisedaccordingtopotentialinflu-

encingfactors,notablyspeciesandcarrier type, todetermineifthesefactorsweresignificanttoharvest-erproductivity,whileusingtreevolumeasthepre-dictivevariable.Tocompensatecategoricalinfluenc-ingfactorswithmorethantwocategories,suchasspeciesandcarriertype,datawasgroupedandanal-ysedregardingtheirrespectivecategories.Multipleregressionanalysiswasconductedasameanstocap-tureresiduals,andmoreaccuratelyrepresentproduc-tivity.Aftereachmultiplelinearregressionproductiv-

itymodelwasdeveloped,ananalysisofcovariance(ANCOVA)wasconductedinordertoverifypoten-tial significant differences between the individuallinearregressionmodels thatmakeupeachof thefullmultiple linearregression. If theresultsof theANCOVAshowthattheindividuallinearregressionsarenon-parallel,thentheANCOVAisrejectedandthemultiplelinearregressionmodelissignificant.However,ifthetestcannotrejectthattheindivid-

uallinearregressionsareparallel,thensignificanceofthefullmultiplelinearregressionisnotestablished.Furthertestingofinterceptequalityisconductedinordertoestablishthatthemodelsarenotthesame.Ifequalinterceptcannotberejected,themultiplelinearregressionmodeldevelopedisnotsignificantlydif-ferentandasinglelinearregressionmodelcanade-quatelyfitthedataset.However,ifinterceptequalityisrejected,themultiplelinearregressionproductivitymodelisabetterfitforthedataset.Allanalysisandmodelswereconductedandde-

velopedthroughExcelandSTATISTICA13(StatSoft,Tulsa,OK,USA).

3. ResultsAllfiveindividualdatasetswerepooledtoproduce

ameanproductivityfigureof14.5m3PMH-1(Table3).Literatureandcorrelationanalysisidentifiedtreevol-umeasthemostsignificantcontributortoharvesterproductivity(p<0.001).Thepooledharvesterproduc-tivitywasplottedagainsttreevolumeandanalysedtodevelopasinglelinearregressionmodel.Theresultofthesingleregressionequationwaspositivelycorre-latedwiththedataset(r2=0.64,p<0.001),wherethere-gressionequationisy=4.536+63.801x(wherex = tree volume)(Fig.2andTable3).Theaverageproductivityforeachoftheindividu-

alstudiesvariedbetween13.80and27.49m3PMH-1.Regressionmodelswerealsodevelopedforeachofthedifferentstudies(Table3).Theproductivitymodelsweredevelopedwith»x«equaltotreevolume.

Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13) J. Norihiro et al.

Croat. j. for. eng. 39(2018)1 5

terexplainvariationofthepooleddataset.Allmultiplelinearregressionequations,whensignificant,weredevelopedconsideringspecies,carriertype(excavatorbasedversepurpose-built),silviculture (plantedorcoppice),slope,debarkinganddebranchingharvesterheadpassesandtreevolume.Inthisanalysis,slopeandtreevolumearecontinuousvariables,whilesilvi-culture,harvestedheadpasses,carriertypeandspe-ciesarecategorical.

3.2 SpeciesProductivityequationsweredevelopedbycate-

gorisingdatabyspecies.Alongwithspecies,equa-tionsofcarriertype,silviculture,slope,debarkingpassandtreevolumewereconsidered.Multiplelinearregressionmodelsweredeveloped

foreachspecies.ModelsforEucalyptus smitthii(ES)andEucalyptus dunnii(ED)werenotsignificantfromeachotherafteranANCOVAtest(p=0.48).Asthein-dividualmodelsforESandEDwerenotsignificant,bothspeciesdatawerepooledtodevelopanewcom-binedmodel(ES+ED).Theoverallandthreespeciesbasedmodels,ES+ED; G xC; G xU(Table4),showapositiverelationshipwithincreasingtreevolume.Eachproductivitymodelwasdevelopedwithre-

specttoinfluencingfactors.Forinstance,theinfluenc-ingfactorstoES+EDproductivityweresilviculture,passandtreevolume,while,GxCproductivitywasinfluencedbycarriertype,silviculture,slope,passandtreevolume.GxUproductivitywasonlyinfluencedbypassandtreevolume.Asmultiplevariableswereusedtodevelopthese

models,predictedvaluesversusobservedvalueswereplotted(Fig.3).Eachoftheproductivitymodelsrep-resentthepooleddatasetwithr2greaterthan0.60.

3.3 Species and harvester typeAssuggestedbySirénandAaltio(2003)andSpinelli

etal.(2010),machinedifferencesmayhaveaneffectonproductivity.Therefore,thepooleddatasetwasreana-lysedandnewproductivityequationsweredeveloped

Fig. 2 Single linear regression model of pooled productivity

Table 3 Mean productivity per study

StudyMean productivity

m3 PMH–1 Equation R2 Significance

Overall 14.47 (0.35–69.22) y=4.536+63.801x 0.64 ***

Study 1 17.93 (2.92–43.78) y=5.800+102.784x 0.45 ***

Study 2 14.45 (1.90–44.32) y=4.754+63.611x 0.61 ***

Study 3 23.61 (2.46–58.57) y=3.283+53.041x 0.79 ***

Study 4 27.49 (0.35–59.24) y=1.073+82.817x 0.76 ***

Study 5 13.80 (1.56–69.22) y=1.085+84.778x 0.75 ***

x = tree volume, m3; *** refers to significance at p<0.001

Table 4 Regression equation by species

Species Equation R2 Significance

Overall y=23.684+(0.497)*x1+ (–0.734)*x2+(0.027)*x3+ (–3.963)*x4+(64.430)*x5 0.68 ***

ES+ED y=0.847+(1.189)*x2+(83.087)*x5 0.76 ***

G x C y=21.246+(0.174)*x1+ (–1.906)*x2+(–0.052)*x3+ (–2.633)*x4+(65.652)*x5 0.60 ***

G x U y=3.283+(53.041)*x5 0.78 ***

x1 = model type (purpose-built = 1 or excavator = 2); x2 = silviculture (planted = 1 or coppice = 2); x3 = slope (percent); x4 = number of processing passes;x5 = tree volume (m3); *** refers to significance at p<0.001

3.1 Multiple linear regressionAlongwithsinglelinearregressionmodels,mul-

tiplelinearregressionmodelsweredevelopedtobet-

J. Norihiro et al. Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13)

6 Croat. j. for. eng. 39(2018)1

Fig. 3 Productivity regression models per species, including predictive values versus observed values

Table 5 Regression equation based on harvester machine make per species

Machine make Species Equation R2 Significance

Hitachi ES+ED y=4.368+(63.286)*x5 0.65 ***

Komatsu ES+ED y=1.052+(83.114)*x5 0.76 ***

TimberPro G x C y=10.559+(–2.300)*x2+(–0.094)*x3+(62.286)*x5 0.56 ***

Volvo G x C y=4.979+(–1.455)*x2+(0.003)*x3+(73.665)*x5 0.64 ***

Hitachi G x C y=22.427+(–3.196)*x4+(52.717)*x5 0.62 ***

Hitachi G x U y=20.197+(–2.064)*x4+(40.857)*x5 0.56 ***

x1 = Silviculture (planted = 1 or coppice = 2); x3 = Slope (percent); x4 = Number of Processing Passes; x5 = Tree volume (m3);*** refers to significance at p<0.001

Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13) J. Norihiro et al.

Croat. j. for. eng. 39(2018)1 7

Fig. 4 Productivity regression models per species and harvester manufacturer and model, including predictive values versus observed values

J. Norihiro et al. Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13)

8 Croat. j. for. eng. 39(2018)1

withbothspeciesandharvestermanufacturerascat-egoricalvariables.TheTimberProharvester,usedatonesite,wastheonlypurposebuiltmachine.Alltheothersiteswereharvestedusingexcavatorbasedhar-vesters.Silviculture,slope,debarkingpassesandtreevolumewereeachtestedforsignificanceandincludedintheappropriateproductivitymodels.Again,eachofthemultiplelinearregressionmodelswaspositivelycorrelatedwithincreasingtreevolume(Table5).TheHitachi andKomatsuES+ED productivity

wasonlyinfluencedbytreevolume.TheTimberProGxCandVolvoGxCproductivitywasalsoinflu-encedbytreevolume,butalsobysilvicultureandslope.IntheGxCandGxUstandswiththeHitachimachine, theproductivitywasonly influencedbypassandtreevolume.As previously completed for the species based

models, predicted values verses observed valuesgraphswereplottedtodemonstratetheaccuracyofdevelopedmodelsbyplottingthemodeloverthere-cordedproductivityofeachcarriermakeandspecies(Fig.4).

4. DiscussionWhencomparingfiveoriginalstudiesusingmul-

tiplelinearregressions,thehighestproductivitywasobservedinStudy1,whilethelowestproductivitywasrecordedinStudy3.DatacollectedinStudy4andStudy5hadthesecondhighestproductivitywhen

stemsizeexceeded0.19m3,regardlessofpoortreeform. However, as tree volume decreased below0.19m3,productivityrecordedinStudy2andStudy3exceededthevaluesofStudy4andStudy5.InStudy2,steepandvaryingslopemayberespon-

siblefor thehighrecordedprocessingtime(Fig.5)and,hence,lowerproductivitysimilartoAcunaandKellogg(2009).Study3hadthesecondhighestmeanproductivityasaresultoflargerandhighervolumetrees.Whileproductivitywouldbeexpected tobeevenhigheronthissitebasedonmostpublishedlit-erature,considerableadditionaltimewasrequiredforprocessingeachtree, loweringoverallproductivitysimilartotheresultsfoundinNakagawaetal.(2007,2010).

4.1 General productivity modelsInpreviousstudies,treevolumewasidentifiedas

asignificantpredictorofharvesterproductivityand,asaresult,regressionequationsweredevelopedbasedontreevolume(SirénandAaltio2003,Nurminenetal.2010,AcunaandKellogg2009,Strandgardetal.2013,Standgardetal.2016).Inordertocomparethepooleddatasettothelit-

eraturedata,asinglelinearregressionmodelwasde-velopedbasedon21previouslypublishedpapers.Inordertodothis,themeanproductivityvaluesandthemeantreevolumeineachpublicationwereplottedandanewsinglelinearregressionmodelwasdevel-oped.Theliteraturebasedmodelwasthenoverlappedwith the single linear regressionmodeldevelopedfromthepooleddataset(Table6).Unfortunately,duetothesmallsamplesizefromliteraturedata,thecom-parisonwaslimited.Specifically, inthiscomparison,allproductivity

datainthecombineddatasetandtheliteraturemodelsassociatedwithtreevolumesgreaterthan0.5m3wereremovedfromtheanalysis.Thisprocessallowedthedatasettostaywithinanappropriateharvestedtreevolumerange.Atypical10yearoldharvestedGxC grownonahighsiteindexSouthAfricanplantation,wouldhaveavolumeof0.23m3(Kotzeetal.2012),withfeweverexceedingthis0.5m3limit.

Fig. 5 Individual time consumption per work element per study in centi-minutes

Table 6 Regression model equation of literature based data against dataset

Regression model R2 Significance N

Current study y=4.0582+67.3274x 0.624 *** 4388

Literature y=2.4658+52.6189x 0.623 *** 21

x = tree volume in m3

Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13) J. Norihiro et al.

Croat. j. for. eng. 39(2018)1 9

Fig.6plots the literatureandcombineddatasetmodelsinrespecttotreevolumeandproductivity.Additionally,allindividualdatapointsareplottedtoillustratethespreadofdataaroundthemodels.Over-all, bothmodels showclearproductivity increaseswithincreasingtreevolumes.Themeanproductivityrecordedis14.47m3PMH-1,

whereastheproductivityrecordedfortheliteraturemodelis9.91m3PMH-1.Whencomparedtothelitera-turethroughtheleastsquaredmethod,themeanpro-ductivitycapturedbythecombinedstudydatawassignificantlymoreproductive(p<0.001).AlthoughtheANCOVAwasultimatelyrejectedaftertestinginter-ceptequality,itcouldnotrejectthatthemodelsmaybeparallel.(p=0.28).Thismayimplythatthemodelshavesimilarities,eventhoughproductivityissignifi-cantlydifferent,oritcouldbepotentiallyattributedtosystematicerrorrelatedtotheremovalofdatatolimittheeffectofthelargetreesizesinliteraturemodels.

4.2 Other influencing factors

4.2.1 SpeciesSimilartoNurminenetal.(2010),thisstudyidenti-

fiedspecieshavingasignificanteffectonproductivity(p<0.001).TheGxC,EDandESproductivitymodels(Table4)havearelativelyhigherspreadofproductiv-ityvaluesoflessthan30m3PMH-1,whiletheGxU productivitymodelhasamoreconsistentandregularspreadofdatawithvaluesoflessthan45m3PMH-1.

Aseachspecies-specificmodelhasitsowninfluencingfactors,itisdifficulttocomparethemodels.Forin-stance,carriertypeandslopeappearonlyintheOver-allandGxCmodels,whereassilviculture,numberofpassesandtreevolumeappearinallofthemodels.Overallproductivityestimatescanbecalculatedwiththebasicdataoninfluencingfactors.Theseestimatesareanimportantcomponentinthemanagementofloggingcrewsandtheextendedforestproductssup-plychain.

4.2.2 Carrier typeIntheliterature,machineandequipmentselection

hasbeenconsideredtomakeasignificantdifferenceonharvesterproductivity (SirénandAaltio2003,Spinellietal.2010).Oneofthereportedpotentialdif-ferences is the influenceofharvesterheadmodels(LaitilaandVäätäinen2013).Thisrelationshipwasnotconfirmedbythecurrentstudy;itwasonlyabletoestablishsignificanceforthespecificharvestermanu-facturerandmodelwhen testedwithacorrelationanalysis.Furthermore,nopublishedliteraturewasfoundon

productivitybasedonmachineselectionbetweenex-cavator basedmachines verses purpose-built ma-chines,especiallyinrelationtoEucalyptusCTLhar-vesting operations. This study compared the twocarriertypesandconfirmedpurpose-builtmachinesasbeingmoreproductiveformosttreevolumes,butastreevolumedecreasedsodidthemarginofsignifi-cance.AlthoughlesscommoninSouthAfricabecauseofthehighinitialinvestmentcost,purpose-builtma-chinesspecialiseintreefellingandprocessing,whichkeepstheirproductionratestableandlessaffectedthanexcavatorbasedmachinesbyfactorssuchaster-rainchanges(Martin2016).

4.2.3 SlopeGroundslopeofthesitesinthisstudyrangedfrom

flattoover60%.InallstudiesexceptStudy2,slopewasclassifiedasperErasmus(1994)aslevel(0–10%)and,afteranalysis,itwasfoundtobeinsignificanttoproductionrate(p=0.07).Incontrast,Study2hadvary-ingslopesrangingfromleveltoverysteep.Thelitera-turesuggeststhatregardlessoftreevolume,asteeperslope leads toadecrease inharvesterproductivity(Spinelli2002,AcunaandKellogg2009,Magagnottietal.2011,McEwanetal.2016).Theinfluenceofslope,asstatedintheliterature,wasonlysignificantinStudy2,wherethereweremoredataonsteeperterrainusedintheanalysis.Atthesametime,thelesssteepterrainhadverylittleinfluenceonproductivityinthefulltreevolumerange.

Fig. 6 Combined dataset (CS), published literature (LT) models and data points in respect to tree volume and productivity

J. Norihiro et al. Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13)

10 Croat. j. for. eng. 39(2018)1

4.2.4 Passes for debarking and delimbingAsEucalyptus treesare typicallydebarkedand

delimbedatthestumpinCTLoperations,theseac-tivitiesareconsideredinthedevelopmentofproduc-tivity models. Debarking effort is related to thestrengthofthebark/woodbond;thestrongerthebark/woodadhesion,thegreatertheimpactondebarkingproductivity (HartsoughandCooper 1999, vandeMerwe2014).The literaturehassuggestedthatcli-maticconditionscansignificantlyaffectthebarkwoodbondof logsduetovaryingmoisturecontentand,therefore,theproductivityrateofimmediatein-fielddebarking(Öman2000,Araki2002,Nuutineneetal.2010,vandeMerwe2014).Twostudiesdidnothavethenumberofpasses included in theirmodels. InStudy2,themainfocusoftheprojectwastoinvesti-gate carrier type interactionswithproductivityonvariableterrain,solittletonodatawascollectedonthenumberofpassesrequiredfordebarkinganddelimb-ing.Likewise,thefocusofStudy4hadlimitedinterestinthenumberofpassesandthesedatafelloutofthemodelasinsignificant(p>0.05).

4.2.5 Independent literature modelsAspreviouslystated,manystudieshaveshown

treevolumetobethemostconstantvariabletodeter-mineharvesterproductivity(Spinellietal.2002,Ovas-kainenetal.2004,Jiroušeketal.2007,Nakagawaetal.2007,Spinellietal.2010,McEwan2012,Picchioetal.2012,SeixasandBatista2012).Thestrongcorrelation

betweentreevolumeandproductivityisconfirmedbytheanalysisinthisstudy,wheretreevolumewasidentifiedasthemostsignificantpredictorofharvest-erproductivity(p<0.001).Inthegeneralproductivitymodelsdiscussedinthefirstpartofthissection,theliteraturebasedmodelwasgeneratedusingvolumeandproductivitydatapointsfrommultiplepaperstodevelopalinearregressionmodel.Threeadditionalpublishedstudiesfullydevelopedproductivitymod-elsthatallowafurthercomparisonwiththecombineddatasetmodel.AllfourofthesemodelsareplottedinFig.7.The Spinelli et al. (2002) and Strandgard et al.

(2016)modelsfocusedondevelopingharvestingpro-ductivitymodelsforEucalyptuswithregardtosouth-ernEuropeandAustralia,respectively.Ramantswanaetal.(2013)consideredharvesterproductivityeffectsondifferentlymanaged silviculture (coppiceverseplanted)Eucalyptusplantations.Despitedifferentpri-maryobjectives, themodelswereallbasedontreevolumeas the continuouspredictor and thus theywerecomparablewiththecombineddatasetmodel.Whenmodelswerecompared,theproductivitymod-eldevelopedwiththedatasetmodelfitsintotheexist-ingrangeandfollowsthecommontrendbasedonliteraturemodels(Spinellietal.2002,Ramantswanaetal.2013,Strandgardetal.2016).Theseregressionmodelsnotonlyreveal,butvali-

datetheincreaseinproductivityoftheharvesterastreevolumeincreases,regardlessoftheconsiderationofadditionalvariables(i.e.terrain,silviculture,carriertype).Theseequationsarethestartofapotentialpro-ductivityequationtohelplocalstakeholdersandcon-tractorstodetermineproductivityandcostmodelsforfutureSouthAfricanoperations.

4.3 LimitationsThemainlimitationsofthisstudyareasfollows:Þ asthisstudyconsistsofacombinationofdis-creetdatasetswithdiverseobjectivesandvari-ables,notnecessarily recorded inall studies,analysesandcomparisonswerecomplicated

Þalthoughconsideredtrained,differentoperatorswereusedoverthetwo-yeardatacollectionpe-riodofthisstudy.Operator’sefficiencywasex-cludedfromanalysis

Þweatherconditionsforeachofthestudieswerenotincludedinthiscombineddataset.Thepro-ductivityofdifferenttasks,likedebarking,canvarybetweenwetanddryweather,sowhilethesedatawereassumedtobecollectedduringnormal dry conditions, actual dailyweathercouldresultinproductivitydifferences.Weath-ereffectswerenotincludedinthisanalysis.

Fig. 7 Harvester productivity (m3 PMH-1) for three independent lit-erature models and the combined dataset model

Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13) J. Norihiro et al.

Croat. j. for. eng. 39(2018)1 11

5. ConclusionsThisstudydevelopedgeneralproductivitymodels,

specificforSouthAfrica,formechanisedEucalyptuspulpwoodCTLharvesting(fellingandprocessing)op-erationsthroughthecombinationoftheresultsoffiveindividualstudies.Themodelsconsideredspecies,sil-viculturalpractices(plantedvs.coppiced),carriertype(purposebuiltvs.excavatorbasedmachines),numberofpassesfordebarkinganddelimbingandslope.Whenstudieswerecombined, theoverallmean

productivityfromthedatasetwas14.47m3PMH-1witharangebetween0.35m3PMH-1and69.22m3PMH-1.Throughacorrelationanalysis,treevolumewasfoundtobethemostsignificantpredictorofoverallproduc-tivity,confirmingthepublishedresults.Basedonthisresult,asinglelinearregressionmodelwasdevelopedwithrespecttotheindividualtreevolume.Tofurtherstrengthenthemodels,theadditional

influence of species, silvicultural practices, carriertype,numberofpassesfordebarkinganddelimbingandslopewereincorporatedintoageneralproductiv-itymodelthroughmultiplelinearregressionanalysis.Thedatasetwasthencategorisedbyspecies,showingthattherewereproductivitydifferencesforeachspe-ciesgroups.Aseachspeciesgroupuseddifferentcon-tributingfactors,itwasimpossibletomakesignificantcomparisonsbetweenthegroups.Anewmodelbasedonexistingdatapointsfrom

publishedliteratureandthreeotherpublishedcom-pleteproductivitymodelswerealsocomparedwiththemodelsdevelopedinthisstudy.Similaritiesbe-tweenthemodelsconfirmedthatharvesterproductiv-ityincreasesastreevolumeincreases,regardlessoftheconsiderationofadditionalvariables(i.e.slopeorsil-viculture).Asthefirststepinrefiningalocallyrelevantpro-

ductivitymodelformechanisedCTLsystems,theseresultscanhelpstakeholdersandcontractorstodeter-mineproductivityandcostsforfutureoperations.ThisworkbynomeansaddressesallaspectsofEucalyptuspulpwoodclearfellingproductivity,butcontinuedef-fortsinthisfieldandbroadeningthedatabasewithmoreanddiversedata,willleadtoarobustSouthAf-ricanspecificproductivitymodel.

AcknowledgementsTheauthors acknowledgeMondi forproviding

studyareas.WerecogniseJohnRabieandChadMar-tin,MScstudentsatStellenboschUniversity,forpro-vidinguswithdata.Inaddition,thankstoJohnEggersandNonkululekoNtingafromMondiforprovidingsupport,aswellas,ProfessorDaanNelatStellenboschUniversityforassistancewithdataanalysis.

6. ReferencesAckerman,P.,Gleasure,E.,Ackerman,S.,Shuttleworth,B.,2014:StandardsforTimeStudiesfortheSouthAfricanFor-estIndustry(Accessed17March2015).Availableat:http://www.forestproductivity.co.za/?page_id=678.

Acuna,M.A.,Kellogg,L.D.,2009:Evaluationofalternativecut-to-lengthharvestingtechnologyfornativeforestthin-ninginAustralia.InternationalJournalofForestEngineering20(2):17–25.

Akay,A.,Erdas,O.,Session,J.,2004:Determiningproductiv-ityofmechanizedharvestingmachines.JournalofAppliedSciences4(1):100–105.

Andersson,R.,2011:Productivityofintegratedharvestingofpulpwoodandenergywoodinfirstcommercial thin-nings.Matersthesis,SwedishUniversityofAgricultureSci-encesDepartmentofForestResourcesManagement.Umeå,Sweden.

Araki,D.,2002:Fibrerecoveryandchipqualityfromde-barkingandchippingfire-damagedstems.Report,ForestEngineeringResearchInstituteofCanada.

Bredenkamp,B.V.,2012:Thevolumeandmassoflogsandstandingtrees.In:Bredenkamp,B.,Upfold,S.(eds.),SouthAfricanForestryHandbook(5thedn.).Pretoria:SouthernAfricanInstituteforForestry:239–267.

Clewer,A.G.,Scarisbrick,D.H.,2001:Practicalstatisticsandexperimentaldesignforplantandcropscience.WestSussex,England:JohnWileyandSons,Ltd.

Davis,C.J.,Reisinger,T.W.,1990:Evaluatingterrainforharvestingequipmentselection.JournalofForestEngineer-ing2(1):9–16.

Erasmus,D.,1994:NationalTerrainClassificationSystemforForestry:Version1.0.InstituteforCommercialForestryRe-search.ICFRBulletin11/94.Pietermaritzburg,SouthAfrica.

Eriksson,M.,Lindroos,O.,2014:ProductivityofharvestersandforwardersinCTLoperationsinnorthernSwedenbasedonlargefollow-updatasets.InternationalJournalofForestEngineering25(3):179–200.

FAO,1997:StateoftheWorld’sForests,Rome,Italy:FoodandAgricultureOrganizationoftheUnitedNations.

FAO,2006:GlobalForestResourcesAssessment2005:Prog-resstowardsustainableforestmanagement.ForestryPaperNo.147.Rome,Italy:FoodandAgricultureOrganizationoftheUnitedNations.

ForestryEconomicsServicesCC,2011:ReportofcommercialtimberresourcesandprimaryroundwoodprocessingSouthAfrica.DepartmentofAgriculture,ForestryandFisheries,Pretoria.

ForestrySouthAfrica,2013:AbstractofSouthAfricanfor-estryfactsfortheyear2010/2011.DepartmentofAgriculture,ForestryandFisheries.Johannesburg.

J. Norihiro et al. Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13)

12 Croat. j. for. eng. 39(2018)1

Hartsough,B.R.,Cooper,D.J.,1999:Cut-to-lengthharvest-ingofshort-rotationEucalyptus.ForestProductsJournal49(10):69–75.

Hartsough,B.R.,Nakamura,G.,1990:HarvestingEucalyp-tusforfuelchips.CaliforniaAgriculture44(1):7–8.

Jiroušek,R.,Klvač,R.,Skoupý,A.,2007:Productivityandcostsofthemechanisedcut-to-lengthwoodharvestingsys-tem inclear-fellingoperations. JournalofForestScience53(10):476–482.

Kellogg,L.,Bettinger,P.,1994:Thinningproductivityandcostforamechanizedcut-to-lengthsysteminthenorthwestPacificcoastregionoftheUSA.JournalofForestEngineer-ing5(2):43–54.

Kotze,H.,Kassier,H.W., Fletcher,Y.,Morley, T., 2012:Growthmodellingandyieldtables.In:Bredenkamp,B.V.andUpfold,S.J.(eds.),SouthAfricanForestryHandbook.TheSouthAfricanInstituteofForestry.5thEdition.ColourPlanet.Pinetown.

Laitila,J.,Vaatainen,K.,2013:Thecuttingproductivityoftheexcavatorbasedharvesterinintegratedharvestingofpulp-woodandenergywood.BalticForestry19(2):289–299.

Magagnotti,N.,Nati,C.,Pari,L.,Spinelli,R.,Visser,R.,2011:Assessingthecostofstump-sitedebarkingineucalyptplan-tations.Biosystemsengineering110(4):443–449.

Martin,C.,2016:Assessingtheeffectofslopeoncostsandproductivity of single-grip purpose-built and excavatorbased harvesters. MSc. Thesis. Stellenbosch University,SouthAfrica.

McEwan,A.,2011:TheeffectoftreeandbundlesizeontheproductivityandcostsofCut-To-Lengthandmulti-stemharvestingsystemsinEucalyptuspulpwood.MSc.Thesis.UniversityofPretoria,SouthAfrica.

McEwan,A.,Magagnotti,N.,Spinelli,R.,2016:Theeffectsofnumberofstemsperstooloncuttingproductivityincop-piceEucalyptusplantations.SilvaFennica50(2):1–14.

Nakagawa,M.,Hamatsu,J.,Saitou,T.,Ishida,H.,2007:Ef-fectoftreesizeonproductivityandtimerequiredforworkelementsinselectivethinningbyaharvester.InternationalJournalForestEngineering(18):24–28.

Nakagawa,M.,Hayashi,N.,Narushima,T.,2010:Effectoftreesizeontimeofeachworkelementandprocessingpro-ductivityusinganexcavator-basedsingle-gripharvesterorprocessoratalanding.JournalforForestResearch15(4):226–233.

Nurminen,T.,Korpunen,H.,Uusitalo,J.,2006:Timecon-sumptionanalysisofthemechanizedcut-to-lengthharvest-ingsystem.SilvaFennica40(2):335–363.

Nuutinen,Y.,Väätäinen,K.,Asikainen,A.,Prinz,R.,Hei-nonen,J.,2010:Operationalefficiencyanddamagetosaw-logsbyfeedrollersoftheharvesterhead.SilvaFennica44(1):121–139.

ÖmanM.,2000:Influenceoflogcharacteristicsondrumde-barkingofpulpwood.ScandinavianJournalofForestRe-search15(4):455–463.

Ovaskainen,H.,Uusitalo,J.,Väätäinen,K.,2004:Character-isticsandsignificanceofaharvesteroperators’workingtechniqueinthinning.InternationalJournalofForestEngi-neering15(2):67–77.

Passicot,P.,Murphy,G.E.2013:EffectofworkscheduledesignonproductivityofmechanizedharvestingoperationsinChile.NewZealandJournalofForestryScience43(2):10p.

Picchio,R.,Sirna,A.,Sperandio,G.,Spina,R.,Verani,S.,2012:MechanizedharvestingofEucalyptCoppiceforBio-massProductionUsingHighMechanizationLevel.CroatianJournalofForestEngineering33(1):15–24.

Puttock,D.,Spinelli,R.,Hartsough,B.R.,2005:Operationaltrialsofcut-to-lengthharvestingofpoplarinamixedwoodstand.InternationalJournalofForestryEngineering16(1):39–49.

Purfürst,F.T.,Erler,J.,2012:TheHumanInfluenceonPro-ductivityinHarvesterOperations.InternationalJournalofForestofForestEngineering22(2):15–22.

Rabie,J.,2014:Analysisofamechanisedcut-to-lengthhar-vesting operationworking in a poor growthEucalyptus smithiistandthroughuseofdiscrete-eventsimulationinR.MSc.thesis,StellenboschUniversity,SouthAfrica.

Ramantswana,M.,McEwan,A.,Steenkamp,J.,2013:Acom-parisonbetweenexcavator-basedharvesterproductivityincoppicedandplantedEucalyptus grandiscompartmentsinKwaZulu-Natal,SouthAfrica.SouthernForests:aJournalofForestScience75(4):239–246.

Seixas,F.,Batista,J.L.F.,2012:UseofWheeledharvesterandexcavatorsinEucalyptusharvestinginBrazil:In:Proceed-ingsofthe35thCouncilonForestEngineeringAnnualMeet-ing:EngineeringNewSolutionsforEnergySupplyDemand.NewBern,NorthCarolina,7p.

Sirén,M.,Aaltio,H.,2003:ProductivityandCostsofThin-ningHarvestersandHarvester-Forwarders.InternationalJournalofForestEngineering14(1):39–48.

Spinelli,R.,Owende,P.M.O.,Ward,S.,2002:ProductivityandcostofCTLharvestingofEucalyptus globulusstandsus-ingexcavator-basedharvesters.ForestProductJournal52(1):67–77.

Spinelli,R.,Hartsough,B.,Magagnotti,N.,2010:Productiv-ityStandardsforHarvestersandProcessorsinItaly.ForestProductsJournal60(3):226–235.

StatSoft, 2012: Statistica 13. Tulsa,OK,United States ofAmerica.

Strandgard,M.,Walsh,D.,Acuna,M.,2013:Estimatinghar-vesterproductivityinradiatapine(Pinus radiata)plantationsusingStanForDstemfiles.ScandinavianJournalofForestResearch28(1):73–80.

Strandgard,M.,Mitchell,R.,Walsh,D.,2013:ProductivityandcostoftwoEucalyptus nitensharvestingsystemswhenbarkisretainedonlogs.AustralianForestsOperationsRe-searchAlliance(AFORA):IndustryBulletin5.

Productivity Model for Cut-to-Length Harvester Operation in South African Eucalyptus ... (1–13) J. Norihiro et al.

Croat. j. for. eng. 39(2018)1 13

Strandgard,M.,Walsh,D.,Mitchell,R.,2015:Productivityandcostofwhole-treeharvestingwithoutdebarkinginEu-calyptus nitensplantationinTasmania,Australia.SouthernForests77(3):173–178.

Strandgard,M.,Mitchell,R.,Acuna,M.,2016:Generalpro-ductivitymodelforsinglegripharvestersinAustralianeu-calyptusplantations.AustralianForestry79(2):108–113.

VanderMerwe,J.,2014:Theimpactofmechanicallogsur-facedamageonfibrelossandchipqualitywhenprocessingEucalyptuspulpwoodusingasingle-gripharvester.MSc.Thesis,StellenboschUniversity,SouthAfrica.

Williams,C.,Ackerman,P.,2016:Cost-productivityanalysisofSouthAfricanpinesawtimbermechanisedcut-to-lengthharvesting.SouthernForests78(4):267–274.

Received:December12,2016Accepted:May1,2017

Authors’addresses:

JenniferNorihiroe-mail:[email protected],PhD.*e-mail:packer@sun.ac.zaUniversityofStellenboschDepartmentofForestandWoodSciencePrivateBagX17602MatielandSOUTHAFRICA

DirkLaengine-mail:Dirk.Laengin@mondigroup.co.zaMondi380OldHowickRoad3245HiltonSOUTHAFRICA

BenD.Sponge-mail:ben.spong@mail.wvu.eduWestVirginiaUniversityDivisionofForestryandNaturalResourcesPOBox6125WV26506MorgantownUSA

*Correspondingauthor