Explaining planted-tree survival and growth in urban neighborhoods: A social–ecological approach to studying recently-planted trees in Indianapolis

14 

Full text

(1)

ContentslistsavailableatScienceDirect

Landscape

and

Urban

Planning

jou rn al h om ep a ge :w w w . e l s e v i e r . c o m / l o c a t e / l a n d u r b p l a n

Research

Paper

Explaining

planted-tree

survival

and

growth

in

urban

neighborhoods:

A

social–ecological

approach

to

studying

recently-planted

trees

in

Indianapolis

Jessica

M.

Vogt

a,b,f,∗

,

Shannon

Lea

Watkins

b,c

,

Sarah

K.

Mincey

c,d,e,f

,

Matthew

S.

Patterson

g

,

Burnell

C.

Fischer

c,f,b

aFurmanUniversity,DepartmentofEarthandEnvironmentalSciences,3300PoinsettHighway,Greenville,SC29613,UnitedStates

bIndianaUniversityBloomington,CenterfortheStudyofInstitutions,Population,andEnvironmentalChange,408N.IndianaAve.,Bloomington,IN48407,

UnitedStates

cIndianaUniversityBloomington,SchoolofPublicandEnvironmentalAffairs1315E.10thSt.,Bloomington,IN47405,UnitedStates dIndianaUniversityBloomington,IntegratedProgramintheEnvironment,702N.WalnutGroveAve,Bloomington,IN47405,UnitedStates eIUResearchandTeachingPreserve,702N.WalnutGroveAve,Bloomington,IN47405,UnitedStates

fTheVincentandElinorOstromWorkshopinPoliticalTheoryandPolicyAnalysis,513N.ParkAve.,Bloomington,IN57508,UnitedStates gUniversityofWashington,UrbanEcologyResearchLaboratory,432GouldHall,394915thAve.NE,Seattle,WA,UnitedStates

h

i

g

h

l

i

g

h

t

s

•Weexaminedtherelationshipbetweensocial–ecologicalsystem(SES)factors&streettreesuccess.

•VariablesfromallSESfactorsinfluencerecently-plantedtreesurvival&growth.

•Theimpactofneighborhoodwateringstrategyontreesuccessdependsonplantingseason.

•Futureresearchshouldconsidersocial–ecologicalcontextofplantedurbantrees.

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received28February2014

Receivedinrevisedform

24November2014

Accepted27November2014

Availableonline30December2014

Keywords: Treegrowth Treesurvival Streettrees Social–ecologicalsystems Communitycharacteristics Institutions

a

b

s

t

r

a

c

t

Thisresearchseekstoanswerthequestion,whatfactorsoftheurbansocial–ecologicalsystem pre-dictsurvivalandgrowthoftreesinnonprofitandneighborhoodtree-plantingprojects?TheOstrom social–ecologicalsystemframeworkandClarkandcolleagues’modelofurbanforestsustainabilityinform ourselectionofvariablesinfourcategoriesinthesocial–ecologicalsystem;thesecategoriesarethe trees,thebiophysicalenvironment,thecommunity,andmanagementinstitutions.Weusetree inven-torymethodstocollectdataonthesurvival,growth,andthesocial–ecologicalgrowingenvironmentof recently-plantedstreettreesinIndianapolis,INtoanswerourresearchquestion.Weuseaprobitmodel topredicttreesurvival,andalinearregressionmodeltopredicttreegrowthrate.Thefollowingvariables arepositivelyrelatedtotreesuccess(survivaland/orgrowth):ball-and-burlaporcontainerpackaging,a visiblerootflare,goodoverallconditionrating,thesizeofthetree-plantingproject,plantingareawidth, medianhouseholdincome,percentofrenteroccupiedhomes,residenttenure,priortreeplanting experi-ence,correctmulching,andacollectivewateringstrategy.Thefollowingvariablesarenegativelyrelated totreesuccess:caliperatplanting,crowndieback,andlowertrunkdamage.Additionalvariables mea-suredhavelessclearconnectionstotreesuccessandshouldbeexaminedfurther.Giventhatmodels includingvariablesfromallfourcategoriesofthesocial–ecologicalsystemgenerallyoutperformmodels thatexcludesomecomponents,werecommendthatfutureresearchonurbantreesurvivalandgrowth shouldconsidertheholisticsocial–ecologicalsystemscontextoftheurbanecosystem.

©2014TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/3.0/).

∗ Correspondingauthor.Tel.:+0019208502016.

E-mail addresses: jessica.m.vogt@gmail.com, jessica.vogt@furman.edu (J.M. Vogt), shawatki@indiana.edu (S.L. Watkins), skmincey@indiana.edu (S.K. Mincey),

tertiarymatt@gmail.com(M.S.Patterson),bufische@indiana.edu(B.C.Fischer).

http://dx.doi.org/10.1016/j.landurbplan.2014.11.021

(2)

1. Introduction

Inthelasttwodecades,manycitiesintheUnitedStateshave

increasedtreeplantingactivitiesandsettreeplantingorcanopy

covergoals(McPherson&Young,2010).However,relativelylittle

isknownaboutthefactorsthatinfluencethesuccessoftheseyoung

urbantrees.Treesinurbanenvironmentsfacechallengestotheir

survivalandgrowththataredifferentfromthosefacedbytreesin

forestsornurseries(Whitlow&Bassuk,1987).Treesinurban

sett-ingsareaffectednotonlybyenvironmentalconditions,butbythe

peoplewhoplant,own,maintain,passby,andbenefitfromthese

trees.However,muchresearchontreeoutcomeshastakenplace

asexperimentsingreenhousesornurseries,whichcannotsimulate

theactualgrowingconditionsofurbantreesthatgrowinsuchclose

proximitytopeople.Thispaperstudiesthesurvivalandgrowth

ofyoungtreesplantedalongcitystreets.Itusesaholistic

frame-worktoexplainrecentlyplantedurbantreesuccessthataccounts

forcharacteristicsofthetrees,thebiophysicalenvironment,the

surroundingcommunity,andmaintenanceinstitutions.Webuild

uponpreviousresearchinurbanforestryandonsocial–ecological

systemsbyconductinginsituresearchonurbantreesurvivaland

growthandbyexplicitlyconsideringthatplantedtreesarepartof

alargerurbansocial–ecologicalsystem.

1.1. Studyingurbantreesurvivalandgrowthinsitu

Ourreviewoftheliteraturefindsthatthemajorityofresearch

abouturbantreesuccesscomesfromexperimentsconductedin

relativelycontrollednurserysettingsratherthanintheurban

envi-ronmentwherestreettreesgrow.Fewstudiesattempttocontrol

fortheadditionalstressesthatcomefromtheurbanenvironment.

Fewcomprehensivelymeasurethecombinedeffectsof

biophysi-calconditionsandmanagementfactorsontreesuccess,muchless

combinesocialorcommunityinfluencewiththesebiophysical

fac-tors.OneexceptionistherecentstudybyLuetal.(2011),which

examinedtheinfluenceoflocalbiophysicalfactors(urbandesign,

biologicalcondition,etc.)andsocialfactors(e.g.,aweededtreeplot

asevidenceoftreestewardship)onthemortalityratesofyoung

streettreesinNewYorkCity.Jack-Scott,Piana,Troxel,

Murphy-Dunning,andAshton(2013)alsomakeuseofinformationabout

treesurroundingstoinformtheirstudyoftreesuccess.

1.2. Urbanforestsassocial–ecologicalsystems

Theurbanforestcanbeunderstoodasasocial–ecologicalsystem

oflinkedhumanandnaturalcomponents(Mincey,2012;Mincey

etal.,2013;Vogt&Fischer2014).Thisperspective(seeTable1)

buildsontwotheoriesofsustainableresourcemanagement:the

modelofurbanforestsustainability(Clark,Matheny,Cross,&Wake,

1997)andthesocial–ecologicalsystemframework(Ostrom,2009)

andhighlightspotentialfactorsthatmightinfluencetreesurvival

andgrowth.Themodelofurbanforestsustainabilitywas

devel-opedinthefieldofurbanforestmanagementinthemid-1990s.

Themodelidentitiesthreeelementsthatarenecessaryforanurban

foresttobesustainable(i.e.,abletocontinueproducingbenefits

atthe samelevel over time): (1)a healthy vegetative resource

(thetreesandtheirgrowingenvironment),(2)asupportive

com-munity,and (3) an adequatemanagement regime(Clarket al.,

1997).Thesocial–ecologicalsystem(SES)frameworksuggests

sim-ilarcategoriesoffactorsthatappearmostrelevanttosocialand

ecologicaloutcomesin ruralnatural resource systems.The late

NobelLaureateElinorOstromandcolleaguesdeveloped theSES

frameworkthroughdecadesof casestudyresearchoncommon

poolresourcemanagementinruralforests,fisheries,andirrigation

systems(Ostrom,2009).TheSESframeworkusesfourcoresetsof

variablestocategorizeinfluencesonoutcomesoflinkedhuman

andnaturalsystems:(1)theresourceunits(e.g.,fish,trees),(2)

theattributesofthebiophysical resourcesystem(e.g.,size ofa

lakeorforest),(3)thecharacteristicsofthecommunityofactors,

orresourceusers(e.g.,numberofusers),and(4)theinstitutional

factorsofthegovernancesystem(e.g.,rulesforfishingortimber

harvesting;Ostrom,2009).Specificvariables inthesefour

cate-goriesinteractwithone anotherandwiththelargerecological

andsocio-politicalcontexttoproducesocialandecological

out-comes(Ostrom,2009;Epstein,Vogt,Mincey,Cox,&Fischer,2013,

Vogt,2014).Ascoupledhuman-naturalsystems(Liuetal.,2007)of

treesandpeople,urbanforestsaresocial–ecologicalsystems,and

theSESframeworkcanhelpexplainobservedoutcomes.However,

theoriginalSESframeworkwasdevelopedlargelyusingresearch

conductedinextractiveresourcesystemsinruralsettings;thus,

weadaptthisframeworkforourapplicationtourbanforeststhat

providenon-extractivebenefits.

Our urban forests as social–ecological systems perspective

(Table1)containsfourbroadcategories ofvariables thatmight

influencethesuccessoftheurbanforest:(1)thetrees,(2)their

biophysicalenvironment,(3)thesurroundingcommunity,and(4)

themaintenanceinstitutionsthataffectthetree.Weusethis

theo-reticalframeworktomodeltreesuccess.Intherestofthissection,

wedescribewhatpreviousresearchtellsusabouthoweachofthese

categoriesmightinfluencetreesuccessintheurbanforest.

1.2.1. Trees

Thesurvivalandgrowthofplantedtreesisinfluencedbythe

characteristicsofthosetrees.Previoushorticulturaland

arboricul-turalresearchprovidessomeinsight here.Forinstance,thesize

of the tree when it is planted (Neal &Whitlow, 1997; Struve,

Burchfield, &Maupin,2000; Watson, 2005;Lambert,Harper, & Robinson,2010),thetypeofplantpackaging(Gilman&Beeson, 1996;Lambertetal.,2010),andthetreespecies(e.g.Iakovoglou, Thompson,Burras,&Kipper,2001;Grabosky&Gilman,2004)may

influenceitssurvivalandgrowth.Plantingdepthcanimpacttree

survival: trees thatare plantedtoo deeply, withtoomuch soil

coveringtherootball,areatgreaterrisk ofmortality(Gilman&

Grabosky,2004).Additionally, tree health and conditionreflect

overalltreevigorandshouldalsoberelatedtothesurvival(e.g.

Roman,2013)andgrowth(Berrang,Karnosky,&Stanton,1985; Achinelli,Marquina,&Marlats,1997)ofthetree.

1.2.2. Biophysicalenvironment

Thebiophysicalenvironmentalsoinfluencestreesuccess.

Evi-dencesuggeststhattreesurvivalisinfluencedbysurroundingland

usetype(MillerandMiller,1991;Rhoades&Stipes,1999;Luetal.,

2011), as wellasavailable growingspace (Lu etal., 2011)and

rootingvolume,whichconstrainsrootgrowthandthereforealso

abovegroundgrowth(Krizek&Dubik,1987;Grabosky&Gilman,

2004;Day,Wiseman,Dickinson,&Harris,2010).Treegrowthis

also impacted by water stress (Kramer, 1987; Krizek &Dubik,

1987;Graves,Joly,&Dana,1991),poorsoilconditions(Smith,May, &Moore,2001;Scharenbroch,Lloyd,&Johnson-Maynard,2005; Scharenbroch,2009)andcompetitionforspacewithotherurban

infrastructure both above and belowground (Green &Watson,

1989;Gilman,1990a;Kjelgren&Clark,1992;Grabosky&Gilman,

2004).Competitionwithothertreesforrootingspace,nutrientsand

waterbelowgroundandforspaceandlightabovegroundinfluences

growthrates(Nowak,McBride,&Beatty,1990;Rhoades&Stipes,

1999;Iakovoglouetal.,2001),ascantheseasoninwhichatreeis

planted(Solfjeld&Hansen,2004).

1.2.3. Community

Wedefinecommunitytobethepeoplewithinandsurroundinga

resourcesystemwhoprovide,use,andbenefitfromthatresource

(3)

Table1

Urbanforestsaresocial–ecologicalsystems.Thisperspectivecombinesthesocial–ecologicalsystem(SES)frameworkdevelopedinruralcommonpoolresourcemanagement

settings(Ostrom,2009)andthemodelofurbanforestsustainabilitydevelopedinthefieldofurbanforestsustainability(Clarketal.,1997).

Social–ecologicalsystem(SES)framework(Ostrom,2009) Modelofurbanforestsustainability(Clarketal.,1997) Urbanforestsassocial–ecologicalsystems

Resourcesystem Vegetativeresource Biophysicalenvironment

Resourceunits Trees

Users Supportivecommunity Community

Governancesystem Adequatemanagement Institutionsaandmanagement

a

Institutions”aredefinedasrules,normsandstrategiesthatgoverninteractionsbetweenthebiophysicalresourceandthecommunityofusersorbeneficiaries(Ostrom, 2005).

influencetreesuccess(Clarketal.,1997).ResearchbehindtheSES

frameworkhasdemonstratedthatcertaincommunity

character-isticscan belinkedtosuccessof theresource system(Ostrom,

1990).For instance, in ruralcommunity-managed forests, Yang

etal.(2013)findthatthelargestcontributionstoward resource

monitoringeffortscamefromintermediatecommunitysizeand

socialcapitalacrossgroupsandamonggroupmembers,leadingto

thebiggestgainsinforestcover(Yangetal.,2013).

Residentsofcommunities(i.e.,neighborhoods)inwhichtrees

areplantedvaryintheircapacityandresourcestomaintaintrees

andin theirlevelof commitmenttoalong-termneighborhood

improvementprojectliketreeplanting.Highertreeplanting

suc-cessshouldinpartbedrivenbyneighborhoodcharacteristicslinked

tohighercapacityfor–ornormsof–bettertreeplantingandcare.

Nowaketal.(1990)identifyapositivecorrelationbetweenpercent

unemploymentandtreemortalityrates.Theseauthorsspeculate

that,“increasedunemploymentsignifiesmoretimespentinthe

neighborhoodandincreasedactivityinthestreetenvironment,”

whichmaylead togreaterstreet treemortality(: p. 128).This

samestudyalsofinds anegativecorrelationbetweentheratios

ofowner-to-renteroccupiedhousesandtreemortalityrates,and

hypothesizesthat lower ratesof homeownership transferto a

lackofstreettree‘ownership’andthereforehighermortalityrates

(Nowaketal.,1990).RecentresearchinConnecticut,U.S.,by Jack-Scottetal.(2013)demonstratedtherelevanceofothercommunity

groupcharacteristicstotreesurvivalandgrowth(Jack-Scottetal.,

2013).Theyobservedthatthelongevityofthegroupengagingin

treeplantingand prior treeplanting experienceofa

neighbor-hoodarepositivelyrelatedtotreesurvivalandgrowth,andthat

morepeopleparticipatinginplantingduringaplantingyearalso

yieldedgreatersurvivalandgrowth(Jack-Scottetal.,2013).Other

studieshaveexaminedmotivationsforindividualinvolvementin

treeplanting.For example,Austin(2002)reportedthat

individ-ualsinvolvedintreeplantingandmaintenanceactivitiesinDetroit,

Michigan,U.S.weremotivatedbyadesiretoworkinnatureandan

opportunitytobettertheneighborhood.Treesmightfarebetterin

neighborhoodswhereresidentsaremoreinterestedinworkingin

natureandinbetteringtheneighborhood.

1.2.4. Institutionsandmanagement

Institutionsaretherules,norms,andsharedstrategies–formal

andinformal–thatstructuretheinteractionsbetweenthe

commu-nity,resources,andbiophysicalenvironment(Ostrom,2005).Any

rules,normsormanagement strategiesrelatedtoneighborhood

maintenanceofplantedtreesmightinfluencethesuccessofplanted

streettrees.Evidenceexiststhatinstitutionscaninfluenceurban

forests:Larsen etal.(2008)reportedthat rulesofhomeowners’

associationsactivelyregulatedthecompositionandmaintenance

ofresidentiallandscapes.Inadditiontoevidenceaboutstrictrules,

Groveet al.(2006)foundthat homeownersare likely to

main-tainlandscapessimilartotheirneighbors’becauseofsocialstatus

or norms. In multiple cities in Michigan, Nassauer, Wang, and

Darrell(2009)foundthatcommunitynormsregardinglandscape

appearanceinfluenced exurbanhouseholdpreferences forfront

yarddesign.

Maintenance strategies – thetype, frequency, duration, and

intensityofmaintenanceperformedontrees–canalsoinfluence

outcomesintheurbanforest(Vogtetal.inreview).Importantly,

maintenance throughout a tree’s life mayactually mitigatethe

impacts of poor biophysical growing conditions. For example,

irrigationandpropersoildrainagecanreducewaterstress(Gilman,

1990b,2001,2004).Researchinnurseriesandgreenhousesreveals

thatconsistentwateringofyoungtreesislinkedtogreatertree

sur-vivalandgrowth(Gilman,2001,2004).Thus,weexpectthatany

strategythatproducesmoreconsistentwateringwillleadtohigher

neighborhoodtreesurvivalrates.

Pruning andmulching strategies also influencetree success.

Pruningofrootsduringplanting(Gilman,1990b;Solfjeld&Hansen,

2004)orofbranches(Carvell,1978)canhaveasignificantinfluence

ontreegrowthandsurvival,dependingontheproportionofatree’s

rootsystemorcanopyremoved.Inparticular,pruningbranches

fromatreeremovesphotosyntheticareaandthusweexpecttrees

togrowmoreslowlyiftheyarepruned(Whitcomb,1979;Nowak

etal.,1990).Incorrectpruningcanleavealargeandslow-to-heal

wound,exposingatreetodiseasesthatmaydecreasetreevigor

(Clark&Matheny,2010).However,minimal,correctlyperformed

pruningatthetimeofplantingcanactuallyenhancetreegrowth

rates(EvansandKlett,1985).

Additionally,mulchingcanhavepositiveornegativeimpacts

ontreesuccessdependingonthedepth,placement,type,and

tim-ingofmulchapplications.Mulchthatislessthanabout5cm(2in.)

deepandispulledawayfromthebaseofthetrunkhelpsretain

soilmoistureandpreventsweedgrowthandcanimprovetree

suc-cess(Gilman&Grabosky,2004).However,impropermulching(i.e.,

mulchthatistoodeeparoundthebaseofa tree)mayincrease

atree’sirrigationneeds(Gilman&Grabosky,2004)andcanalso

encourageadventitiousrootgrowthorgirdlingroots.

2. Methods

Toexamine treesuccessin theurbangrowingenvironment,

this studyuses data collectedduring a re-inventory ofplanted

urbantreestocreatestatisticalmodelsoftreesurvivalandgrowth.

We partnered with Keep Indianapolis Beautiful, Inc. (KIB), an

urbangreening nonprofitinMarion County,Indiana.KIB works

withneighborhoodgroupsinIndianapolistoplanttreesthrough

itsNeighborWoods program.In theirproposalfor a tree

plant-ingproject,groupsthatplanttreesthroughtheNeighborWoods

programmustidentifyastrategythat theywillfollowtowater

theplantedtrees.Wesortedtree-plantingprojectsthatoccurred

between2006and2009bythetypeofwateringstrategyadoptedby

theneighborhoodandchoseastratifiedsampleof23projects(in16

neighborhoods)wheretreeswerewateredbyindividualresidents

(individualwateringstrategy)and12projects(in9neighborhoods)

whereatleastsometrees werewatered bygroupsof residents

together(collectivewateringstrategy).Overall,oursampleincluded

35projectsin25neighborhoods.

Inthesummermonthsof2011and2012,were-inventoried

street trees that were planted in our sample neighborhoods.

(4)

neighborhoods(1462trees)andsystematicallysampledand

re-inventoried20–30livingtreesperproject(673livingtreesintotal).

Someoftheseobservationshadmissinginformationabout

partic-ularvariablesandwerenotusedforanalysis.Overall,ouruseable

samplecontains1345treesforsurvivalmodelsand616treesfor

growthmodels.

Duringthesummerof2011wegatheredinformationaboutthe

tree,biophysicalinformationabouttheplantingsiteandevidence

oftreemaintenanceactivitiesaccordingtoadatagatheringguide

(Authoretal.,2014maskedforblindreview).Highschoolstudent

employeesofKIB’sYouthTreeTeamweretrainedbytwoofthe

authorsandcollectedadditionaltreedataduringthesummerof

2012.Additionalinformationaboutthebiophysicalenvironment

camefromtheCityofIndianapolisandtheNationalLandCover

Dataset(NLCD)2006.KIBprovidedinformationfromthetimeof

plantingaboutthetreesandthetree-plantingproject.

Socio-demographiccommunitycharacteristicscamefromthe

U.S.CensusBureau(U.S.DepartmentofCommerceCensusBureau,

2012).Eachtreewasassigned thevaluesforsocio-demographic

variablesofthecensustractinwhichthetreewasplanted.Sources

forallvariablescanbefoundinTable2.

Selectinformation aboutinstitutionsrelatedtotree

manage-ment (in addition to our knowledge of the watering strategy)

wasgatheredfromsemi-structuredinterviewswithneighborhood

leadersandnonprofitemployees(seeMincey&Vogt[2014]for

completeinterviewmethods).Interviewresponseswereusedto

generatehypothesesandtoinforminterpretationofthestatistical

models.

2.1. Dependentvariables

Wemeasuretreesuccessintwoways:treesurvivalandgrowth

rate.Survivalanditsconverse,mortality,arecommonmeasures

oftreesuccessinurbanforestrystudies,thoughspecificestimates

ofsurvivalrateswithinthefirstfewyearsafterplantingarerare

(Nowaketal.,1990;Roman&Scatena,2011),asaregeneral

esti-matesofurbantreemortality(L.Roman,personalcommunication,

November13,2013).Ourmeasureofsurvivalisabinaryindicatorof

whethertheplantedtreewasstillaliveatthetimeofre-inventory.

Growthrateisalsoausefulmeasureoftreesuccess;as

larger-sized(i.e.,mature)treesprovidegreaterecosystemservices,atree

withafastergrowthratethatreachesmaturitysooneryieldsfaster

rateofreturn(Nowaketal.,1990).Also,treesthatgrowwellearly

inlifeareconsideredmorelikelytobeestablishedandfreetogrow

withreducedneedforwateringorothermaintenanceduring

nor-malweatherconditions.Ourmeasure ofgrowthis the“relative

growthrate,”calculatedas

relativegrowthrate= lnC2−lnC1

(t2−t1)/365

whereC1andC2 aremeasurementsoftreecaliperatthetimeof

planting(t1)andtimeofre-inventory(t2),respectively(afterBrand,

1991asadaptedbySamynanddeVos(2002)).Thedifferenceinthe

timeofplantingandtimeofre-inventoryismeasuredindays,so

wedividethenumberofdaysinthedenominatorby365daysper

yeartoobtainrelativegrowthrateinyears.

2.2. Analysis

Wemodelsurvivalandgrowthasafunctionofthefour

cate-goriesofsocial–ecologicalvariablesdescribedabove.

Weusemaximumlikelihoodestimation(probit)topredictthe

likelihood that a plantedtree wasstill alive at thetime of

re-inventory.Weusedataonalltrees thatwere plantedbetween

2006and2009inourselectedneighborhoods(atotalsampleof

1345).Ourbasicsurvivalmodel(ModelS.1)includesindicatorsof

thesizeandtheconditionofeachtreeatplanting,indicatorsof

thesurroundingenvironmentthatarenottree-specific(the

per-centof impervious surfaceand the speed limit),and all of the

socialandinstitutionalindicatorsdescribedbelow.InModelS.2,

wealsoincludetheinteractionbetweencollectivewateringand

ifatreewasplantedinthefall.Wesuspectthattheeffectof

col-lectivewateringmightvaryacrosscircumstance.Weexpectthe

positiveeffectofcollectivewateringtobegreaterinspring

plant-ingsbecausethetreesfaceapotentiallylong,hot,drysummerand

collectivewateringmayprovidemoreconsistentwatering.

Weuselinearregressiontopredictannualcalipergrowthfor

thetrees thatwere livingat thetimeof re-inventory. Ourfirst

model (Model G.1) includes thetree, biophysical environment,

community,andinstitutionalindicatorsdescribedbelowaswell

asnurseryandtaxonomicfamilydummyvariables.Inoursecond

model(ModelG.2),weaddtheinteractionbetweenfallplanting

andcollectivewatering.We alsosuspecttheeffectofcollective

watering mightdiffer acrossoverall treecondition.Tomeasure

theseeffects,wealsointeractcollectivewateringwithourbinary

indicatorsoftreecondition.

The dataare clusteredin two ways—bytree speciesand by

neighborhood.Toaccountforobservedandunobserved

species-levelvariation,weincludedummyvariablesfortreefamilyinall

models,with an “other” category for families in ourdata with

fewerthan20trees,andthebeech/oakfamily(themost

numer-ous family in our dataset) as the baseline. Tree survival and

growthmay alsovary acrossneighborhood inways wecannot

observe.Wecontrolforsomeneighborhoodvariationby

includ-ingasuiteofcommunitycovariates(socio-demographicvariables).

Usingneighborhood-levelrandomeffectswouldcontrolfor

unob-served neighborhood-level characteristics. We tested whether

randomeffectswerenecessaryusinganintraclasscorrelation

coef-ficient (ICC). Given the low ICCs for both survival (ICC=0.087)

and growth(ICC=0.187)datasets,neighborhood randomeffects

models are not shown. Robust standard errors are used in all

models.

2.3. Independentvariables

2.3.1. Treecharacteristics

We includeindicatorsoftreesizeand conditionatplanting:

caliper-at-plantingandwhetherthetreewaspackagedin

ball-and-burlappackaging,acontainer,orothertypeofpackaging(bare-root

ornylonbags;theexcludedpackagingtype).Weincludedummy

variables for treefamily (withthe mostnumerous family,Oak,

beingtheexcluded,orbaselinefamily),andalsofornursery(i.e.,

originofplantingstock;themostnumerousnursery–nursery5

–isexcludedasthebaseline).Wecontrolforageofthetreeat

re-inventory(yearssinceplanting).Transplantedtreesmaytake

a fewgrowingseasonstobecomefullyestablishedinthe

land-scape(Gilman,Black,&Dehgan,1998; Struveetal., 2000), and

yearssinceplantinghelpscontrolforthis.ThesedatacamefromKIB

records.

We also include indicators oftree conditionat re-inventory

in ourgrowth models.Condition variables include binary

indi-catorsof leafchlorosis, a visible rootflare,and damage onthe

lowertrunkofthetree.Acanopydiebackrating(modifiedfrom

InternationalUnionofForestResearchOrganizations,International Society of Arboriculture, United States Forest Service, &Urban NaturalResourcesInstitute,2010)isalsousedtoindicatecondition

ofthetree.Weincludearatingoftheoverallconditionofthetreeas

twodummyvariablesforgoodconditionandpoorcondition(the

excludedconditionwasfair).Wealsoaccountforlight

availabil-itybyincludingacrownexposurerating(afterInternationalUnion

ofForestResearchOrganizationsetal.,2010).Post-plantingtree

(5)

Table2

Categoriesofvariablesandsourcesofdatausedinanalyses.

UFasSEScomponent Variables Datasource

Outcomevariables Alive Protocol,V13

Annualcalipergrowth(cm) KIB/Protocol,V5 Tree Treespecies(aggregatedtofamilylevel) KIB/Protocol,V3

Nursery KIB

Caliperatplanting(cm) KIB

Ball-and-burlappackaging(0,1) KIB

Containerizedpackaging KIB

Age(#yearssinceplanting) KIB

Crowndiebackrating Protocol,V9

Lowertrunkdamage(0,1) Protocol,V12

Leafchlorosis(0,1) Protocol,V10

Rootflarevisible(0,1) Protocol,V11 Goodoverallconditionrating(0,1) Protocol,V13 Pooroverallconditionrating(0,1) Protocol,V13 Biophysical

environment

%Impervioussurface NLCD2006

Speedlimit Indianapolis

#Treesplantedinproject KIB

Fallplantingseason KIB

Plantingareawidth(naturallog) Protocol,V26 Treelawnplantingarea(0,1) Protocol,V24

Crownexposurerating Protocol,V9

#Treeswithin10m(naturallog) Protocol,V29 #Treeswithin10-20m(naturallog) Protocol,V29/V30 Community Neighborhoodname(usedforrandomeffects) KIB

%Unemployment ACS,2011

Medianhouseholdincome($) ACS,2011 %Lessthanhighschooleducation ACS,2011 %Singleparenthouseholds Census,2010

%Nonwhitepopulation Census,2010

%Renteroccupiedhomes Census,2010

%Movedinlast5years ACS,2011

%Vacanthouses Census,2010

#Totaltree-plantingprojects KIB Institutionsand

management

Plantingyear KIB

Correctpruning(0,1) Protocol,V35

Incorrectpruning(0,1) Protocol,V34

Correctmulching(0,1) Protocol,V34

Collectivewateringstrategy(0,1) Interviews

Protocol,Vogtetal.,2014maskedforblindreview,includingthevariablenumberfromtheProtocol(e.g.,V4).KIB,KeepIndianapolisBeautiful,Inc.datacollectedattime

ofplanting.NLCD2006,NationalLandCoverDataset2006(http://www.mrlc.gov/nlcd2006.php).ACS,2011,UnitedStatesDepartmentofCommerce,CensusBureau,2011

AmericanCommunitySurveyfromAmericanFactFinder(http://facfinder2.census.gov).Census,2010,UnitedStatesDepartmentofCommerce,CensusBureau,2010complete

U.S.CensusfromAmericanFactFinder(http://facfinder2.census.gov).Indianapolis,IndianapolisCityGovernmentgeographicinformationsystemshapefileofcitystreets,

obtainedfromKIB.Interviews,interviewswithKIBemployeesaboutthetreeplantingproject,includingreviewoftreeplantingprojectapplications.

2.3.2. Biophysicalenvironment

Weaccountforthesurroundingbiophysicalenvironmentusing

severalindicators.Forallmodels,weusethepercentofimpervious

surfaceinthe30×30mcellsurroundingthetreeasaproxyfor

additionalstressontherootzone(e.g.,limitedrainfallinfiltration).

ThesedatacomefromtheNationalLandCoverDataset2006.The

speedlimitonthestreetadjacenttothetreeservesasaproxyfor

howbusyastreetisandwhetheritislikelytobehighlysaltedinthe

winter.ThisdatacamefromCityofIndianapolisroadlayerfiles.We

alsoaccountforwhetherthetreewasplantedinthefallorspring

andincludethenumberoftreesplantedinaprojecttocontrolfor

projectsize.ThesevariablescamefromKIBrecords.

Forgrowthmodels,wealsoincludetwomeasuresofgrowing

space—abinaryindicatorofwhetherthetreewasplantedinatree

lawnanda measureofthewidth(narrowestdimension)ofthe

plantingarea.We usethenaturallogofthewidthofthe

plant-ingareaaroundthetreeasaproxyforavailablerootingvolume.

Wealsoincludemeasuresofcompetition:crownexposurerating

(reflectingshading),andthenumberofothertreeswithina10-m

radiusandbetween10and20moftheplantedtree,bothinnatural

logform.Thesevariablesweregatheredduringre-inventory.

2.3.3. Community

We use socio-demographic indicators to proxy community

capacity and commitmentto tree care. To capture community

capacityweincludethefollowingvariablesfromtheU.S.2010

Cen-sus:ameasureofthepercentageofindividualsinthelaborforce

thatareunemployed,themedianhouseholdincome,the

percent-ageofindividualsthat haveless thanahighschooldegreeand

thepercentageofhouseholdsthatareheadedbyasingleparent.

WeincludeameasurefromKIB’srecordsofwhetherthetreewas

plantedaspartofthecommunity’sfirstprojectwithKIB,the

sec-ond,thethird,etc.,tocapturecapacity specifictotree-careand

learningwithexperience.Tocapturecommitmentweinclude a

measureofthepercentageofindividualsthathavelivedinthesame

residenceforatleastfiveyears(i.e.,residenttenure)andthe

per-centageofhousingunitsthatareoccupiedbyrenters.Weinclude

thepercentofunitsthatarevacant.Vacanciesmightreflecttwo

things:anindicationofdeeperneighborhoodproblemsandlack

ofcapacityand/orconditionsinwhich neighborsareanxiousto

restoretheneighborhood’sappearanceandthusmaybemore

ded-icatedtotreecare.Wealsocontrolforthepercentofindividualsin

theneighborhoodthatarenon-white.Commitmentvariablescome

fromtheU.S.Census.

2.3.4. Institutionsandmanagement

Tomeasuretreemaintenanceinstitutions,weincludea

mea-sureofwhethertheneighborhood’sapplicationfortreesproposed

acollectivewateringstrategyinwhichatleasttwoneighborswould

(6)

Table3

Descriptivestatisticsforsurvivalmodels.Forbinaryvariables,themeanrepresentstheproportionofobservationswiththatindicator.

UFasSEScomponent Variables N Mean Median Std.dev.

Outcomevariable Alive(0,1) 1345 0.894 1 0.307

Tree Caliperatplanting(cm) 1345 3.536 3.81 0.506

Ball-and-burlappackaging(0,1) 1345 0.125 0 0.331

Containerizedpackaging(0,1) 1345 0.573 1 0.495

Age(#yearssinceplanting) 1345 4.545 4.72 0.970

Biophysicalenvironment %Impervioussurface 1345 43.903 46 19.115

Speedlimit(mph) 1345 32.093 30 6.605

#Treesplantedinproject 1345 67.297 57 36.070

Fallplantingseason(0,1) 1345 0.460 0 0.499

Community %Unemployment 1345 7.488 6.5 4.082

Medianhouseholdincome($1000s) 1345 $45.552 $43.221 $17.213 %Lessthanhighschooleducation 1345 15.716 13.8 9.883

%Singleparenthouseholds 1345 10.987 11.1 4.642

%Nonwhitepopulation 1345 29.362 28.8 14.417

%Renteroccupiedhomes 1345 43.021 49.5 18.995

%Movedinlast5years 1345 54.043 56.1 9.941

%Vacanthouses 1345 16.105 14.6 11.004

#Totaltree-plantingprojects 1345 1.317 1 0.561

Institutionsandmanagement Plantingyear 1345 2007 2007 1.149

Collectivewateringstrategy(0,1) 1345 0.410 0 0.492 Tree—nurserydummyvariables

(nursery5excluded)

Nursery3(0,1) 1345 0.093 0 0.290

Nursery6(0,1) 1345 0.371 0 0.483

Nursery7(0,1) 1345 0.082 0 0.274

Othernursery(0,1) 1345 0.062 0 0.241

Tree—familydummyvariables (beech/oak[Fagaceae]family excluded)

Maple(Aceraceae)family(0,1) 1345 0.150 0 0.357

Birch(Betulaceae)family(0,1) 1345 0.087 0 0.282 Dogwood(Cornaceae)family(0,1) 1345 0.057 0 0.231

Legume(Fabaceae)family(0,1) 1345 0.083 0 0.276

Pine(Pinaceae)family(0,1) 1345 0.029 0 0.168

Planetree(Platanaceae)family(0,1) 1345 0.029 0 0.168

Rose(Rosaceae)family(0,1) 1345 0.139 0 0.346

Other+family(0,1) 1345 0.076 0 0.265

+The“Otherfamily”categoryincludestreesofthefollowingfamilies,eachrepresentedbyfewerthan20individualsinourdataset:Altingiaceae,Apocynaceae,Celastraceae,

Cupressaceae,Ebenaceae,Ginkgoaceae,Juglandaceae,Lauraceae,Magnoliaceae,Malvaceae,Oleaceae,Styracaceae,andUlmaceae.

willyieldmoreconsistentwateringbecauseitmightsupportde

factoorformalmonitoring(Wade,1994),andbecauseAuthorand

Author(2014amaskedforblindreview)foundasignificantinfluence

ofwateringstrategyontreesurvival.Forourgrowthmodels,we

includewhethertherewasevidenceduringre-inventorythatthe

treewascorrectlypruned,incorrectlyprunedorlackedevidence

ofpruning. Weinclude abinarymeasureofwhethertherewas

evidenceofcorrectmulchingattimeofre-inventory.Weexpect

correctpruningandcorrectmulchingtoincreasegrowthrate.

Therearesometypesofrulesanddecision-makingprocesses

thataffecttreesuccessthatoccuratthenonprofitlevel,suchas

choiceoftreespecies,depthoftreeplanting(whichaffects

pres-enceofarootflare),andothermethodsofplanningandorganizing

treeplantingactivities.Manynonprofitdecisionsarerepresented

in tree,biophysical environment, and community categories of

variables. The single nonprofit-level institution we included in

our model is year of planting, which we believe may

repre-sentorganizationallearningandinstitutionalchange,andvaries

acrosstrees.InterviewsandinformalconversationswithKeep

Indi-anapolisBeautiful,Inc. employeesrevealedthat over time they

changedtree-plantingstrategies,adaptingandrespondingto

infor-malobservations that KIB employees weremaking about their

tree-plantingprojects.Forthisreason,weincludeyearof

plant-ingasaseparatevariable,distinctfrom–althoughstillcorrelated

with–theageofthetree.Nonprofitrulesanddecisionsthatdonot

varyacrossprojectshavenovariationinourdatasetandsoarenot

includedinthissingle-cityanalysis.

3. Results

Descriptive statistics for variables included in survival and

growthmodels are displayed in Tables 3 and 4. Overall, 89.4%

oftreesplantedbetween2006and2009werealiveatthetime

ofre-inventory. Averagecaliper growthrateof livingtrees was

1.12cm/year.

3.1. Treesurvivalmodels

Table5presentscompleteresultsfromtreesurvivalprobit

mod-els.Coefficientsinprobitmodelsaredifficulttointerpret,sowe

relyheavilyonsignificanceandexpecteddirectiontointerpretour

results.Positive,significantcoefficientsinTable5indicatethatan

increaseinthevalueofavariableincreasestheprobabilityofa

tree’ssurvival;negativecoefficients indicatethatanincreasein

thevalueofavariablereducestheprobabilityofatree’ssurvival.

Fig.1presentstheoddsratios(exponentialformofthecoefficients

inModelS.2inTable5)foreachindependentvariable.Theodds

ratioistheoddsofatreesurvivinggivenaone-unitchangeinthe

meanoftheindependentvariable.Forpresence/absence

indepen-dentvariables,theoddsratioistheoddsofatreesurvivinggiventhe

presenceofthatvariablerelativetotheoddsofatreesurvivingin

theabsenceofthatvariable.Oddsratiosgreaterthanoneindicate

increasedprobabilityoftreesurvival,whileoddsratioslessthan

oneindicatedecreasedprobabilityoftreesurvival.Oddsratiosnot

significantlydifferentfromoneindicateanindependentvariable

thatdoesnotaffecttheoddsoftreesurvival.

Treeandbiophysicalvariablesarenotasstronglyrelatedaswe

expectedtosurvival.Inourbestmodelsthatcontrolfortree

fam-ily,wefindnosignificantrelationshipsbetweensurvivalandthe

characteristicsofthetreeexceptfornurserywhereitwasgrown.

However,wefindsomerelationshipbetweensurvivalandthe

bio-physical environment:impervioussurface negatively influences

survival,andthenumberoftreesplantedinagivenproject

(7)

Table4

Descriptivestatisticsforvariablesingrowthmodels.Forbinaryvariables,themeanrepresentstheproportionofobservationswiththatindicator.

UFasSEScomponent Variables N Mean Median Std.dev.

Outcomevariable Annualcalipergrowth(cm) 675 1.126 1 0.605

Relativegrowthrate 675 0.192 0.189 0.070

Tree Caliperatplanting(cm) 675 3.477 3.81 0.530

Ball-and-burlappackaging(0,1) 675 0.079 0 0.269

Containerizedpackaging 675 0.607 1 0.489

Age(#yearssinceplanting) 675 4.472 4.69 1.067

Crowndiebackrating 673 0.201 0 0.816

Lowertrunkdamage(0,1) 656 0.474 0 0.500

Leafchlorosis(0,1) 630 0.144 0 0.352

Rootflarevisible(0,1) 658 0.274 0 0.446

Goodoverallconditionrating(0,1) 646 0.853 1 0.354 Pooroverallconditionrating(0,1) 646 0.023 0 0.151 Biophysical

environment

%Impervioussurface 675 43.947 47 18.361

Speedlimit 675 31.785 30 5.710

#Treesplantedinproject 675 59.939 49 35.339

Fallplantingseason 675 0.447 0 0.498

Plantingareawidth(naturallog) 650 1.806 1.5 1.241

Treelawnplantingarea(0,1) 675 0.535 1 0.499

Crownexposurerating 674 4.733 5 0.782

#Treeswithin10m(naturallog) 675 1.372 1.39 0.467 #Treeswithin10–20m(naturallog) 675 1.813 1.79 0.482

Community %Unemployment 675 7.557 6.8 3.923

Medianhouseholdincome($1000s) 675 $45.609 $39.375 $16.798 %Lessthanhighschooleducation 675 16.316 15 9.438

%Singleparenthouseholds 675 11.110 11.1 4.397

%Nonwhitepopulation 675 28.347 25.8 14.986

%Renteroccupiedhomes 675 41.620 46.8 19.546

%Movedinlast5years 675 55.641 58.1 10.005

%Vacanthouses 675 15.954 14.6 10.214

#Totaltree-plantingprojects 675 1.320 1 0.572

Institutionsandmanagement Plantingyear 675 2007.23 2007 1.209

Correctpruning(0,1) 658 0.169 0 0.375

Incorrectpruning(0,1) 658 0.207 0 0.405

Correctmulching(0,1) 658 0.105 0 0.307

Collectivewateringstrategy(0,1) 675 0.450 0 0.498 Tree—nurserydummyvariables

(nursery5excluded)

Nursery3(0,1) 675 0.062 0 0.242

Nursery6(0,1) 675 0.419 0 0.494

Nursery7(0,1) 675 0.044 0 0.206

Othernursery(0,1) 675 0.030 0 0.170

Tree—familydummyvariables (beech/oak[Fagaceae]family excluded)

Maple(Aceraceae)family(0,1) 675 0.099 0 0.299

Birch(Betulaceae)family(0,1) 675 0.096 0 0.295

Dogwood(Cornaceae)family(0,1) 675 0.052 0 0.222

Legume(Fabaceae)family(0,1) 675 0.080 0 0.271

Pine(Pinaceae)family(0,1) 675 0.031 0 0.174

Planetree(Platanaceae)family(0,1) 675 0.041 0 0.200

Rose(Rosaceae)family(0,1) 675 0.159 0 0.365

Other+family(0,1) 675 0.062 0 0.242

+The“Otherfamily”categoryincludestreesofthefollowingfamilies,eachrepresentedbyfewerthan20individualsinourdataset:Altingiaceae,Apocynaceae,Celastraceae,

Cupressaceae,Ebenaceae,Ginkgoaceae,Juglandaceae,Lauraceae,Magnoliaceae,Malvaceae,Oleaceae,Styracaceae,andUlmaceae.

Somecommunitycharacteristicsare relatedtotree survival.

Medianhouseholdincome,thepercentofrenter occupiedunits

andthepercentofpeoplewhohavemovedinwithinthelast5

yearsareallsignificantlyandpositivelyrelatedtotreesurvivalin

ModelS.2(ourtheoreticallypreferredmodel).

Weseeevidencethattreesthatwereplantedinlateryearshave

ahigherprobabilityofsurviving,evenaftercontrollingforageofthe

tree.Thiscouldbearesultoforganizationlearningand

improve-mentsinplantingmethodsbyKIB.However,thesummerof2007

wasaparticularlydryyearandrainfallduringthesummermonths

(May–August)wasmorethan7inchesbelowaverage(National

WeatherServiceIndianapolis,INWeatherForecastOffice,2013).

Wecannotdisentanglewhetherapositivecoefficientontheyear

ofplantingpicksuptheimpactofadrysummerontreesplantedin

earlieryearsoroforganizationallearningresultinginimproved

plantingtechniquesinlater years. Additionally,mortalitycould

simplybehigherforearlierplantingyearsbecauseearliercohorts

oftreeshavehadmoretimetoaccruehighermortality(Roman,

2013).Whateverthecausalmechanism,mortalityratesdodiffer

significantlybyyearofplanting,rangingfromahighof16.4%for

treesplantedin2007toalowof7.5%fortreesplantedin2008.

Thesignificanceoffallplantingseason,collectivewatering

strat-egy,andtheinteractiontermbetweenthesevariablesmeansthat

thecombinedimpactofthesevariablesisdifferentforeachofthe

foursub-populationsoftreesdividedbythesevariables(Table7).

Holdingallothervariablesconstant,wefindthattreesplantedin

thespringandwateredcollectivelyhavegreateroddsofsurvival

thantreesplantedinthespringandwateredindividuallyandthan

alltreesplantedinthefall.However,treesplantedinthefalland

wateredcollectivelyhaveslightlyloweroddsofsurvivalthantrees

plantedinthefallandwateredindividually.Wenotethatthe

inter-actionterm(collectivewatering×fallplanting)ishighlynegatively

correlated with ball-and-burlap planting packaging (polychoric

correlation coefficient of −0.96; see Supplementary material),

and highly positivelycorrelated withexperience (project

num-ber;polychoriccorrelationcoefficientof0.90;seeSupplementary

material);thesevariablescouldbeconfoundingtherelationship

(8)

Table5

Survivalmodelresults.Coefficientsshownwithstandarderrorsinparentheses.VariablesthataresignificantinModelS.2,themosttheoreticallysoundsurvivalmodel,are inbold.ModelS.1:probitmodel(nointeractionterm,noneighborhoodrandomeffects).ModelS.2:probitmodelwithinteractionterm(noneighborhoodrandomeffects).

ReducedModel:significantvariablesfromModelS.2only(includingallnurserydummyvariables).

UFasSEScomponent Variables ModelS.1 ModelS.2 ReducedModel

Tree Caliperatplanting(cm) −0.142(0.167) −0.141(0.170) Ball-and-burlappackaging(0,1) −0.081(0.394) −0.212(0.408) Containerizedpackaging(0,1) 0.063(0.225) −0.003(0.228) Age(#yearssinceplanting) 0.492*(0.273) 0.430(0.286) Biophysical

environment

%impervioussurface −0.007**(0.004) −0.007**(0.004) −0.009***(0.003)

Speedlimit(mph) 0.005(0.016) 0.006(0.016)

#treesplantedinproject 0.009***(0.003) 0.011***(0.003) 0.010***(0.002)

Fallplantingseason(0,1) 0.336(0.220) 0.681***(0.262) 0.307**(0.127)

Community %unemployment 0.003(0.022) −0.001(0.023)

Medianhouseholdincome($1000) 0.016(0.010) 0.017*(0.010) 0.018***(0.005)

%lessthanhighschooleducation −0.017(0.012) −0.020(0.012)

%singleparenthouseholds 0.066*(0.039) 0.049(0.040)

%nonwhitepopulation 0.005(0.008) 0.009(0.008)

%renteroccupiedhomes 0.036***(0.011) 0.033***(0.011) 0.024***(0.007)

%movedinlast5years 0.015(0.013) 0.024*(0.012) 0.017**(0.008)

%vacanthouses −0.033*(0.018) −0.009(0.019)

#totaltree-plantingprojects −0.128(0.165) 0.078(0.182)

Institutionsand management

Plantingyear 0.752***(0.262) 0.681**(0.273) 0.372***(0.059)

Collectivewateringstrategy(0,1) 0.466**(0.201) 1.030***(0.241) 0.853***(0.154)

Interactions Collectivewatering×Fallplanting −1.193***(0.313) −0.986***(0.262)

Tree—nurserydummy variables(nursery5 excluded) Nursery3(0,1) −0.430*(0.254) −0.486*(0.262) −0.309(0.197) Nursery6(0,1) 0.055(0.245) −0.025(0.250) 0.179(0.113) Nursery7(0,1) −0.124(0.313) −0.190(0.316) 0.146(0.240) Othernursery(0,1) −0.944***(0.270) −0.957***(0.279) −0.734***(0.214) Tree—familydummy variables(beech/oak familyexcluded) Maplefamily(0,1) 0.272(0.275) 0.334(0.266) Birchfamily(0,1) −0.264(0.244) −0.067(0.255) Dogwoodfamily(0,1) −0.082(0.233) −0.089(0.238) Legumefamily(0,1) 0.329(0.273) 0.361(0.275) Pinefamily(0,1) −0.287(0.368) −0.240(0.382) Planetreefamily(0,1) 0.088(0.324) 0.033(0.322) Rosefamily(0,1) −0.247(0.210) −0.324(0.213) Otherfamily(0,1) 0.131(0.219) 0.053(0.225) Constant −1512.816***(526.900) −1372.498**(549.493) −749.010***(118.372) No.ofobservations 1345 1345 1345 Log-likelihood −389.192 −384.031 −396.350

Modelsignif.(p-value) 0.000 . 0.000

Pseudo-R2 0.142 0.153 0.126 AIC 840.383 832.063 820.699 BIC 1001.712 998.595 893.557 *p<0.10. **p<0.05. ***p<0.01.

Ourmodelisfairlyrobusttochangesinmodelspecification;

thecoefficientsare quiteconsistentacrossmodels.While there

are a few changes in significance across models, there are no

signchangesinsignificantvariables.Inthereducedmodelwhere

insignificantvariables aredropped, coefficients change slightly,

althoughnotinsignorsignificance.

3.2. Relativegrowthratemodels

Table6presentscompleteresultsfromthetreegrowthmodels.

Asexpected,wefindthatnearlyalltreecharacteristicsincludedin

ourmodelssignificantlyinfluencegrowth.Treesthataresmaller

atplantingappeartogrowfaster.Wefindthattreespackagedas

ball-and-burlaporcontainersgrewfasterrelativetoother

packag-ingtypes.Treesfromothernurseriesalsoexhibitedslowergrowth

ratesrelativetothebaselinenursery(nursery5,themostcommon).

Manyofthefamilydummyvariablesarealsosignificantandinthe

expecteddirectiongivenpreviousworkontreegrowth:treesinthe

dogwood,pine,androsefamilygrowmoreslowlythanoaktrees

(thebaselineandmostcommonfamilyoftrees),whereastreesin

thelegumesand planetreefamiliesexhibithighergrowthrates,

holdingallothervariablesconstant.

Indicatorsoftreeconditionatthetimeofre-inventoryhelp

pre-dictrelativegrowthrate.Treesexhibitinghigherdiebackratings

andtreeswithdamagetothelowertrunkgrewmoreslowly,and

treeswithavisiblerootflareandthoseratedingoodcondition

grewmorequickly.

Generalcharacteristicsofthesurroundingenvironmentimpact

treegrowthlessthantreecharacteristics.Widerplantingareais

associatedwithfastergrowth.Wefindnoevidenceofarelationship

betweentreegrowthandimpervioussurface,speedlimit,project

size,treelawnplantingarea,crownexposure,ornumberofnearby

trees.

Thereissomeevidenceofarelationshipbetweencommunity

characteristicsandtreegrowth.Wefindthatthepercentofsingle

parenthouseholdsisnegativelyrelatedtotreegrowth,whichmay

beindicativeofthecapacityfortreecarebysingleparent

house-holds.Thenumberofpreviousprojectsthattheneighborhoodhas

undertakenisalsosignificant—treesthatareplantedaspartofa

laterprojectgrowfaster.Noneoftheotherdemographicorhousing

variablesappearedtoinfluencetreegrowth.

Wefindevidencethatsomeinstitutionandmanagement

vari-ablesmatter.Plantingyearisnegativelyandsignificantlyrelated

togrowthrate(seefollowingparagraph).Correctmulchingalsois

positivelyandsignificantlyrelatedtogrowthrate,asexpected.

Boththe ageof thetreeand the yearof plantingare

nega-tivelyandsignificantlyrelatedtotreegrowthrate;thesefindings

(9)

Fig.1. OddsratiosforvariablesincludedinsurvivalModelS.2,themosttheoreticallysoundmodel.Oddsratiosgreaterthan1(circles)indicateavariablewithapositive effectonsurvival;oddsratiosoflessthan1(squares)indicateavariablewithanegativeeffectonsurvival.+Confidenceintervalextendsoutsideoftherangeshownonthe graphto4.16.

betweenplantingandre-inventory)andtheyearofplanting(2006,

2007,etc.)arenegativelycorrelated.Anegativecoefficientforage

ofthetreeindicatesthat oldertrees(i.e.,plantedless recently)

havesloweraveragegrowthrates.Anegativecoefficientforyear

ofplantingmeansthat trees plantedinmore recent years(i.e.,

youngertrees)havegrownmoreslowly.Becauseofthis,wesuspect

yearofplantingdoesnotindicatelearningandinsteadmight

cap-turetheimpactofanestablishmentperiodcharacterizedbyslower

growth.

Thesignificanceofthefallplantingseasonindicatorandthe

interactionbetweencollectivewateringstrategyandfallplanting,

butnotofcollectivewateringstrategymeansthatthewatering

strategyonlyhasasignificantimpactongrowthfortreesplantedin

thefall(Table7).Holdingallothervariablesconstant,fallplanting

hasanegativeimpactongrowth,butitsimpactismagnifiedby

choiceofacollectivewateringstrategy.

Ourmodelfortreegrowthisfairlyrobust.Mostsignificant

envi-ronmentalandinstitutioncoefficientsaresignificantacrossboth

ModelG.1andG.2,andremainsignificantinthereducedmodel.

PercentvacanthousesaresignificantinModelG.1,butnotwhen

theinteractionterms areaddedinModel G.2.Collective

water-ingstrategyishighlysignificantinModelG.1,buttheadditionof

theinteractionbetweenwateringstrategyandplantingseasonin

ModelG.2completelycapturestheeffectofcollectivewatering.

CoefficientschangeslightlybetweenModelG.2andthereduced

model,whenallinsignificantvariablesaredropped,butno

vari-ableschangesign.

3.3. JointsignificanceofSEScategoriesandmodelselection

Oneofourobjectivesistodeterminewhetherincluding

charac-teristicsofthecommunityandofinstitutionsimprovesour

under-standingof treesuccessinurbanenvironments.We jointlytest

thesignificanceofcategoriesofSESvariablesbycomparingAkaike

andBayesianinformationcriteria(AIC,BIC)valuesbetween

mod-elsexcludingonecategoryofvariables(restrictedmodels—results

shown in Supplementary material) to Model S.2 and G.2, the

theoreticallysuperiormodels.Forexample,wecomparetheAIC

of Model S.2 to a model that excludes all of the community

(10)

Table6

Relativegrowthratemodelresults.Standardizedcoefficientsshownwithstandarderrorsinparentheses;coefficientswithlargermagnitudehaveagreaterrelativeinfluence ongrowthrate.VariablesthataresignificantinModelG.2,themosttheoreticallysoundgrowthmodel,areinbold.Notethatselectvariablesareincludedinnaturallog

formtoachieveamorenormaldistribution.ModelG.1:ordinaryleastsquares(OLS)regressionmodel(nointeractionterms,noneighborhoodrandomeffects).ModelG.2:

OLSmodelwithinteractionterms(noneighborhoodrandomeffects).Reducedmodel:onlysignificantvariablesfromModelG.2(includingallnurseryandfamilydummy

variables).

UFasSEScomponent Variables ModelG.1 ModelG.2 ReducedModel

Tree Caliperatplanting(cm) −0.496***(0.00740) −0.485***(0.00731) −0.450***(0.00578)

Ball-and-burlappackaging(0,1) 0.302***(0.0252) 0.290***(0.0246) 0.293***(0.0220)

Containerizedpackaging(0,1) 0.205***(0.00870) 0.185***(0.00907) 0.136***(0.00701)

Age(#yearssinceplanting) −0.922***(0.0112) −0.934***(0.0113) −0.690***(0.00900)

Crowndiebackrating −0.107***(0.00598) −0.108***(0.00641) −0.096**(0.00585)

Lowertrunkdamage(0,1) −0.101***(0.00497) −0.098***(0.00497) −0.097***(0.00484)

Leafchlorosis(0,1) 0.067*(0.00722) 0.080**(0.00720) 0.087**(0.00691)

Rootflarevisible(0,1) 0.061*(0.00554) 0.066*(0.00547) 0.073**(0.00530)

Goodoverallconditionrating(0,1) 0.131***(0.00700) 0.173***(0.0110) 0.139***(0.00657)

Pooroverallconditionrating(0,1) −0.001(0.0223) 0.068(0.0349)

Biophysical environment

%impervioussurface 0.031(0.000203) 0.038(0.000200)

Speedlimit −0.054(0.000762) −0.069(0.000759)

#treesplantedinproject 0.059(0.000122) 0.092(0.000124)

Fallplantingseason −0.350***(0.00919) −0.261***(0.0101) −0.141***(0.00618)

Plantingareawidth(naturallog) 0.131*(0.00376) 0.141**(0.00379) 0.079**(0.00209)

Treelawnplantingarea(0,1) −0.001(0.00748) 0.012(0.00755)

Crownexposurerating 0.015(0.00309) 0.020(0.00303)

#treeswithin10m(naturallog) 0.024(0.00646) 0.023(0.00657)

#treeswithin10–20m(naturallog) 0.053(0.00510) 0.030(0.00523)

Community %unemployment 0.034(0.00106) 0.026(0.00106)

Medianhouseholdincome($1000) −0.028(0.000531) 0.045(0.000532)

%lessthanhighschooleducation 0.093(0.000639) 0.111(0.000640)

%singleparenthouseholds −0.203*(0.00181) −0.207*(0.00182) −0.152***(0.000546)

%nonwhitepopulation 0.077(0.000397) 0.116(0.000409)

%renteroccupiedhomes −0.085(0.000490) −0.070(0.000491)

%movedinlast5years −0.013(0.000567) 0.016(0.000594)

%vacanthouses −0.197*(0.000781) −0.126(0.000812)

#totaltree-plantingprojects 0.202***(0.00729) 0.259***(0.00781) 0.314***(0.00610)

Institutionsand management Plantingyear −1.095***(0.0107) −1.086***(0.0107) −0.820***(0.00750) Correctpruning(0,1) −0.040(0.00631) −0.043(0.00628) Incorrectpruning(0,1) −0.013(0.00599) −0.024(0.00611) Correctmulching(0,1) 0.065(0.00994) 0.079*(0.00989) 0.072*(0.00952)

Collectivewateringstrategy(0,1) −0.270***(0.00847) −0.025(0.0150)

Interactionterms Collectivewatering×Fallplanting −0.201***(0.0155) −0.281***(0.0111)

Collectivewatering×Goodcondition −0.107(0.0133)

Collectivewatering×Poorcondition −0.084(0.0421)

Tree—nurserydummy variables(nursery5 excluded) Nursery3(0,1) −0.086*(0.0133) −0.097**(0.0134) −0.100**(0.0117) Nursery6(0,1) 0.021(0.0109) 0.014(0.0113) −0.039(0.00942) Nursery7(0,1) 0.055(0.0146) 0.050(0.0148) 0.061(0.0130) Othernursery(0,1) −0.162***(0.0241) −0.168***(0.0237) −0.204***(0.0220) Tree—familydummy variables(beech/oak familyexcluded) Maplefamily(0,1) −0.077(0.0118) −0.073(0.0116) −0.043(0.0113) Birchfamily(0,1) −0.061(0.0111) −0.034(0.0113) −0.020(0.00932) Dogwoodfamily(0,1) −0.124***(0.0127) −0.117***(0.0131) −0.089**(0.0117) Legumefamily(0,1) 0.106**(0.0133) 0.103*(0.0132) 0.120**(0.0124) Pinefamily(0,1) −0.097**(0.0280) −0.096**(0.0273) −0.109**(0.0275) Planetreefamily(0,1) 0.194***(0.0152) 0.191***(0.0156) 0.177***(0.0147) Rosefamily(0,1) −0.314***(0.0102) −0.318***(0.0104) −0.331***(0.0100) Otherfamily(0,1) 0.105**(0.0136) 0.101**(0.0136) 0.109**(0.0129) No.ofobservations 605 605 605 F 14.07 14.21 .

Modelsignificance(p-value) 6.13E−66 4.50e−69 .

Adjusted-R2(overallR2forr.e.models) 0.471 0.436 0.435

AIC −1796.3 −1801.5 −1816.9

BIC −1598.1 −1590.0 −1689.2

*p<0.10. **p<0.05. ***p<0.01.

jointlysignificant.AICandBICvaluesweighthebenefitsof

addi-tionalinformationwhenmorevariablesareaddedtothemodel

againstthecostsoffewerdegreesoffreedomandtheadditionof

irrelevantvariables.SignificantlylowerAICandBICvaluesindicate

bettermodels;BICpenalizesmoreharshlythanAICforaddition

ofirrelevantvariables.However,AICandBICshouldonlybeused

toguidemodelselection,andnottodefinethe“best”modelwith

absolutecertainty.

Forsurvivalmodels,wefindcoefficientestimatesforsignificant

variablestobefairlyconsistentacrossmodels(seeSupplementary

material).AICandBICvaluesindicatethatModelS.2outperforms

mostmodelsexcludingentirecategoriesofSESvariables.Wefind

thatalthoughBICvaluesforrestrictedmodelsareslightlylower

thanforModelS.2,AICvaluesaretypicallymuchhigher.Thefour

modelsthatexcludealltreecharacteristics,allenvironmental

vari-ables, allcommunity variables, or allinstitutional variables are

outperformed by ourtheoretically preferablemodel (S.2); only

therestrictedmodelthatexcludesprojectcharacteristics(a

sub-setofcommunityvariables)performsbetterthanourtheoretically

(11)

Table7

Interactionbetweenfallplantingandcollectivewateringstrategy.Coefficients

forthe combinedeffects ofplanting seasonand wateringstrategy, basedon

coefficientsfromModelS.2(survival,Table5andModelG.2(growth,Table6).

Plantingseason Wateringstrategy Collective Individual Survival(probitmodel

oddsratios) Fall 1.679 1.976 Spring 2.801 1.000 Growth(standardized coefficients) Fall −0.462 −0.261 Spring 0 0

Forgrowthmodels,coefficients acrossmodelsareconsistent

insignandsignificance,exceptforageandyearvariables,eachof

whichflipsign(topositive)andlosetheirsignificancewhentheSES

componentcontainingtheotherisexcludedfromthemodel(see

Supplementarymaterial).AICandBICvaluesindicatethatour

the-oreticallybestmodel(ModelG.2)outperformsrestrictedgrowth

modelsthatexcludealltreecharacteristics,allcommunity

vari-ables,orsubsetsofcommunityvariables;amodelthatexcludes

allbiophysicalenvironmentvariablesperformsslightlybetterthan

ModelG.2(seeSupplementarymaterial).

4. Discussion

Ourresultssuggestthatbiophysical,socialandinstitutional

fac-torsareallimportant inexplaining thesuccessofyoungurban

plantedtrees,butthatthefactorsthataffectsurvivalandgrowth

aredifferent.Fig.2comparesthemosttheoreticallysound

mod-elsforsurvivalandgrowth(ModelsS.2andG.2).Sometimesthe

samefactorsignificantlyinfluencesbothoutcomes,butinopposite

directions.

Ingeneral,treevariablesappeartomattermorefortreegrowth

thanforsurvival.Theonlytreecharacteristicthataffectssurvivalis

thenursery,whiletreegrowthissignificantlyinfluencedbynearly

alltreevariables,includingseveraltreefamilyvariables(Fig.2).

Thisisnotsurprising,becausewedonothaveinformationon

per-hapsthemostvitaltree-levelfactorthatmightinfluencesurvival—a

tree’sconditionwhile itwasalive andpresumably decliningin

healthandvigor.Otherauthorshavelinkedtreeconditionto

prob-abilityofsurvival(e.g.,Koeser,Hauer,Norris,&Krouse,2013).

Contrary to expectations, trees exhibiting leaf chlorosis

experienced faster growth. We recognize this could be from

misidentificationofchlorosisbythedatacollectors.While

chloro-sis–lackofchlorophyll–istheoretically linkedtostuntedtree

growth(Graves,1994),othercausesofleafdiscolorationappearing

aschlorosistotheuntrainedeyeandthatmighthavebeenrecorded

assuchinthere-inventorymaynotbelinkedtostuntedgrowth.We

alsorecognizethatthepresenceofchlorosisisametricofpresent

treeconditionandmaybemorereflectiveofrecenttreestresses

ratherthanthestressesoverthe(albeitshort)lifetimeofthetree.

Wefindevidencethatthebiophysicalenvironmentaffects

sur-vivalandgrowth.Plantingwidth(aproxyforarea)limitsgrowth

butnotsurvivalinourmodels.Plantingareamayrepresent

avail-ablerootingvolume,whichlimitsgrowthasotherauthorshave

found(Kopinga,1991; Kjelgren&Clark,1992;although Nowak

etal.,1990findnosignificantimpactofplantingareaongrowth).

Alternatively,itmightbethatplantingwidthisaproxyforwater

availability(i.e.,asmallerplantingareameansasmallerareainto

whichrainfallmayinfiltrate),whichmaybemoretightlytiedtothe

successofsmall,recentlyplantedtreesthatmaynotyetbelimited

byavailablerootingvolume

We find evidence that people (i.e., characteristics of the

communityandmaintenanceinstitutionsusedbypeopleto

man-age trees) influence the survival and growth of trees. Some

(12)

socio-demographiccharacteristicsofneighborsarounda

planted-treearesignificantlyrelatedtotreesurvivalandgrowth.However,

apositivecoefficientonpercentrentersinsurvivalmodelsis

con-trarytothefindingofNowaketal.(1990).Wealsofindevidence

thatneighborhood experiencewithtreeplantingcontributesto

treesuccess—treesplantedinlaterprojectsinthesame

neighbor-hoodhadfastergrowth.Thesefindingssuggestthatmoredetailed

researchabout maintenance motivationsmight illuminate why

particularsocio-demographiccharacteristicsmightberelatedto

treesuccess.

Thereisevidencethatinstitutionsmatter,particularlyfortree

survival,butalsothathowinstitutionsimpacttreeoutcomesdepends

onplantingseason. We expectedtrees plantedin the falltobe

morelikelytosurvivebecausetheydonotexperienceasummerof

hot,dryweatherimmediatelyaftertransplanting(although,adry

fallwithoutsufficientwateringcanbepotentiallydeadlyfortrees,

sincethetreesarebeingplantedintoapotentiallyverydrysoil).

However,whenplantingseasonisconsideredincombinationwith

wateringstrategy,wefindthatafallplantingseasononlyimproves

survivalratesfortreesthatareindividuallywatered(Table7).A

col-lectivewateringstrategywaspositivelyrelatedtotreesurvivalfor

springplantings.Forgrowthmodels,collectivewateringstrategy

compoundsthenegativeimpactoffallplantingontreegrowth,

butwateringstrategyhasnoimpactontreeoutcomesforspring

plantings.Wesuspectthesefindingsmayreflectdifferencesinhow

acollectivewateringstrategymaybeimplementedforplantings

indifferentseasons.Recallthatourcollectivewatering strategy

variableisanindicatorofthewateringstrategychosenbythe

neigh-borhood,andthatwecanonlyinfertowhatextentthisstrategywas

actuallyimplementedbytheneighborhood(i.e.,howconsistently

treeswerewatered).Treesplantedinthespringmustbewatered

immediatelyduringthesummerfollowingplanting.However,for

treesplantedinthefall,wateringactivitiesmightnotcommence

untilthefollowingspringandsummer.Itmaybethatcollective

wateringiseasiertoimplementwhenwateringactivities

imme-diatelyfollowplantingactivities,andthereforewateringhappens

moreconsistentlyduringthefirstsummerafterplanting.

Weincludedplantingyearasavariablewethoughtmight

indi-cateinstitutionallearningandchangeintreeplantingmethodsby

thenonprofit.However,ourresultsindicatethatthisvariablemay

actuallybecapturingmoreinformationabouttreeageand

estab-lishmentthanaboutinstitutions.Treesplantedinlateryearsexhibit

ahigherlikelihoodofsurvivalandlowergrowthrates.Higher

sur-vivalratesareexpectedifwethinkthenonprofithasimproved

itsplanting techniquesover time. Buthighersurvival rates are

alsoexpectedduetothefactthatcumulativemortalityislower

forcohortsoftrees(i.e.,treesplantedinthesameyear)thathave

beeninthegroundforlesstime(Roman,2013).Lowergrowthrates

makessensein thecontext oftreebiology:trees plantedmore

recentlyarealsoyoungerandless likelytobeoutofthe

estab-lishmentperiodduringwhichtrunkdiametergrowthisslowed

(Gilmanetal.,1998).

5. Conclusion

Wefindthatattributesofthetree,biophysicalenvironment,

andsurroundingcommunity,aswellasmanagementinstitutions

appeartoimpacturbantreeplantingoutcomes.Severalfindings

fromthisworkcaninformdecision-making.Nonprofitshavea

rea-sonabledegreeofchoiceoverthesizeofplantedtrees:planting

smallertrees willyieldtrees thatbecomeestablishedand grow

more quickly in thelandscape. They also have somechoice in

plantinglocation,includingthesizeoftheplantingarea:

choos-inglarger,widerplantingareaswherepossiblemayyieldhigher

treegrowthrates.Andlocatingtreesinareaswithloweramountsof

impervioussurfacecovermayimprovesurvivalrates.However,we

alsorecognizethatareaswithnarrowtreelawnsandhigh

impervi-ouscoverarealsosomeoftheareasinhighestneedofthebenefitsof

trees(Wilson&Lindsey,2009).Wealsofindthatmany

characteris-ticsofthesurroundingenvironmentdonotsignificantlyaffecttree

growthforthetreesinoursample,whichsuggestsnonprofitsmight

notneedtobeconcernedaboutthesecharacteristicswhen

choos-ingplantinglocations:forinstance,thenumberofothernearby

trees,thespeedlimitoftheadjacentroad,andwhetherthelocation

isatreelawnorothertypeofplantingspotappearnottoinfluence

treesuccess.Maintenancepractices,however,domatter:Correct

mulchingpracticespositivelyimpacttreegrowthrates,suggesting

thatinvestmentinmulchingortrainingthecommunitytomulch

willyieldimprovementstotreesuccess.

Thereare otherpotential decision pointsfor which the

evi-dence is less clear.Although we observe a negative coefficient

on therelationship between fall planting and tree growth,we

hesitatetosaythat fallplantingsshouldbeavoideddue toour

inability to disentangle the relationship between water

avail-ability – a combination of neighborhood-determined watering

strategy(known),precisewateringfrequency/amount(unknown),

anduncontrollableweather/rainfallconditions–andplanting

sea-son.Theunderlyingrelationshipbetweencollectivewateringand

treesuccessisunclearanddependentontheseasonofplanting.

Morefine-grainedinformationonthefrequencyofwateringand

variationinseasonalrainfallwillhelptodetailtherelationships

between watering strategy, environmentalconditions, and tree

outcomes.Nurseryalsoappearstoinfluencesurvivalandgrowth,

butthisvariablemaybeconfoundedbyrelationshipswithplant

packagingtreeandspecies;itisalsonotsomethingthatcanbe

controlledeasilybynonprofitspurchasing treeswhere theyare

available.

Thispaperoffersamoreholisticmodelofthesurvivalandearly

growthoftreesplantedincities.Futureresearchthatexamines

treesuccessshouldexaminetreesacrossmultiplecitiestotestthe

generalizabilityoftheseresults.Longitudinalmonitoringoftree

populationswouldhelpbuildastrongercausalcaseforthefactors

thatcontributetotreesurvivalandgrowthovertime.

Acknowledgements

Thismanuscripthasbenefittedgreatlyfromdiscussionswith

manycolleagues,includingDr.DavidGood,Dr.RichardJ.Hauer,

and LukeShimek.Additionally, theauthorswould liketothank

currentandpastemployeesofKeepIndianapolisBeautiful,Inc.for

theirtimeandenthusiasticandpatientsupportofthisresearch,

inparticularDaveForsell,JeromeDelbridge,NateFaris,Andrew

Hart,BobNeary,andMollyWilson.Inaddition,wethanktheKIB

YouthTreeTeamsummer2012DataTeamandteamleaderJennifer

Swilikfortreedatacollection.Additionalfinancialsupportwas

pro-videdbytheEfroymsonFamilyFund,aswellastheGardenClubof

America’sUrbanForestryFellowshipawardedtoJMVin2012and

2013.Administrativeandfinancialsupportwasprovidedbythe

CenterfortheStudyofInstitutions,Populationand

Environmen-talChange;TheOstromWorkshopinPoliticalTheoryandPolicy

Analysis;and,theSchoolofPublicandEnvironmentalAffairsat

IndianaUniversity,Bloomington.Thisresearchfulfilledpartofthe

dissertationrequirementsoftheSchoolofPublicand

Environmen-talAffairsatIndianaUniversityforauthorJMV.Finally,wethank

twoanonymousreviewersforhelpfulcomments.

AppendixA. Supplementarydata

Supplementary data associated with this article can be

found, in the online version, at http://dx.doi.org/10.1016/

Figure

Updating...

Related subjects :