JournalofForestEconomics17 (2011) 285–297
ContentslistsavailableatScienceDirect
Journal
of
Forest
Economics
j o u r n a l h o m e p a g e :w w w . e l s e v i e r . d e / j f e
The
evaluation
of
forest
crop
damages
due
to
climate
change.
An
application
of
Dempster–Shafer
method
Iacopo
Bernetti, Christian
Ciampi,
Claudio
Fagarazzi,
Sandro
Sacchelli
∗Dept.ofAgriculturalandForestEconomics,Engineering,SciencesandTechnology,18,PiazzaledelleCascine,I-50144Florence,Italy
a
r
t
i
c
l
e
i
n
f
o
Articlehistory:
Received19January2010
Accepted28April2011
Keywords:
Dempster–Shafertheoryofevidence
Climatechange Forestry Fuzzysets Spatialanalysis Uncertainty JELclassification: Q19 Q54
a
b
s
t
r
a
c
t
Inordertoassessdamageriskcausedbyclimatechangein for-estareas,Dempster–Shafertheoryofevidenceandfuzzymeasures wereappliedtodevelopaframeworkfortheestimationof eco-nomicforestdamage.Accordingtothedefinitionofrisksupported bytheIntergovernmentalPanelonClimateChange,afunctionof hazardandresiliencelinesofevidencewasdefined.Theresults ofthehazardandresilienceassessmentwereusedtodevelopan economicframeworkbasedonFaustmannstudies.Theevaluation model,implementedthroughaspatialanalysisprocedure,was car-riedoutlinkingFaustmann formulawithhazardandresilience rastermaps.Themodelpermittedtoestimateinmonetaryterms twopossiblecoststobesupported:thefirstoneisexpressedas theexpecteddamagetotheforestcroponthebasisofthecurrent obtainablewoodyassortmentsandthesecondonereferredtothe potentialexpensestopayinordertomitigatetherisk.Finally,the frameworkwastestedonanareaofcentralItaly(Tuscanyregion).
© 2011 Department of Forest Economics, SLU Umeå, Sweden. Published by Elsevier GmbH. All rights reserved.
Introduction
ResearchonthepossibleimpactofclimatechangeonforestsinEuropeandthedevelopment
ofadaptationandmitigationstrategiesstartedintheearly1990s;sincethen,assessmentonclimate
change,itsimpactandconsequencesonnaturalresourcesmanagementhavebeenthefocusof
contin-uousresearchefforts(EFIetal.,2008).Amongtheknowneffectsofclimatechangetherearealteration
∗ Correspondingauthor.
E-mailaddress:sandro.sacchelli@unifi.it(S.Sacchelli).
1104-6899/$–seefrontmatter © 2011 Department of Forest Economics, SLU Umeå, Sweden. Published by Elsevier GmbH. All rights reserved.
intreesgrowthandproductivity(KauppiandPosch,1988;Loustauetal.,2005),variationsin
for-estareaandcompetitionbetweenspecies(Solomon,1986;Lexeretal.,2002)anddamagevariation
causedbynaturaldisturbances(Flanniganetal.,2000).
Severalmethodologieswereimplementedtoassesshazardandresilienceinforestsector.One
promisingevaluation ofthesetwovariables hasbeenprovided byNitschkeand Innes(2008).In
theirresearchameta-modellinganalysiswasdevelopedtoincorporatethestrengthofmanysmaller
modelsintoaframeworkthatconsidersnaturaldisturbances,floralandfaunalspeciesandspecies
habitats.
Awidereviewonmodellingforestecosystemdisturbancesandonnaturalhazardassessingwere
developedinrecentworksofSeidletal.(2011)andHanewinkeletal.(2010a),respectively.Fromthese
studiesitispossibletodepictseveralexamplesofmethodologiesthatcanbeappliedtodescribeharm
andadaption.Seidletal.(2011)classifythemodelsaccordingtodisturbancestypology(event-based
orregime-based)in:(i)statistical,(ii)staticordynamicprocess-based,(iii)vegetationorlandscape
dynamicsand(iv)plantphysiologymodels.Hanewinkeletal.(2010a)dividemethodstomodelrisk
inexperimental(e.g.mechanisticmodel)andempirical(bothanalyticalandoutputorientedmodels).
Despitetheintensiveresearcheffortsintheabovetopics,inclimatechangescenariotheprecise
evaluationofeconomiclosseswithhighterritorialdetailisadifficulttask.Specifically,themainaspects
tobetakenintoaccountare:(1)theimpactassessmentvarieswidelydependingonthesimulation
modelapplied,theclimatescenarioinvestigatedandthegeographicandecologicalcharacteristicsof
thecrop;(2)thereisconsiderableuncertaintyconcerningtheinteractionsbetweenvulnerabilityand
adaptationabilityofforestecosystems.
Themainobjectiveofthisstudyistogiveapreliminaryassessmentoninnovativekindofeconomic
analysisinclimatechangeforestsector,inparticular: (i)toevaluatetheriskofvulnerabilityand
theprobabilityofresilienceoftheforeststhroughprobabilisticandsubjectiveestimationand(ii)to
estimatethefinancialconsequencesofclimatechangebywayofthecalculationofboththehazard
damageandthevalueofriskmitigationinforeststands.Theappliedprocedureimplementedland
expectationvalue(LEV)calculatedbyFaustmannapproachandDempster–Shafertheoryofevidence
(DS).
Asamatteroffactseveralauthors(e.g.Buongiorno,2001)demonstratedthattheFaustmann
for-mulaisabletoincorporatebiologicalandeconomicrisks.Themethodologychosentocarriedoutthe
research(DStheoryofevidence)permitstoovercomethelimitsofBayesianapproachinanuncertain
frameworkasfortheclimatechangeone.
DStheorywasalreadyusedinenvironmentalsectorindifferenttopics.Severalstudiesdefineland
coverestimationandlandcovermonitoringtechniquescarriedoutthroughtheuseofDStheory(see
e.g.Cayuelaetal.,2006;Comberetal.,2004).Thistheorywasalsointroducedinclimatechange
uncertaintyevaluation(RajeandMujumdar,2010;LouandCaselton,1997).Indifferentresearches
DStechniquewasappliedinforestsector;someexamplesconcern:(i)waystomeasuresustainability
offorestmanagementandtooperationaliseitintermsofutilitymaximisationinlandusestrategies
(Varmaetal.,2000);(ii)definitionofuncertaintyinforestmanagementandsilviculturaldecision (Ducey,2001)and(iii)provisionofpracticaltoolinordertoestimatestandregenerationmaps(Mora
etal.,2010).Actually,accordingtoauthorsknowledge,DStheoryofevidencehasneverbeenusedfor
economicevaluationofforestunderclimatechangeconditions.
Methodologicalframework
TheadoptedmethodologicalframeworkissummarizedinFig.1.Accordingtointernational
litera-tureriskhazardtoclimatechangeandresilienceofforestcropswasdefined.Then,basedonexperts’
knowledge,thefactors(criteria)defininghazardandresiliencewereanalysed,evaluatedthrougha
fuzzymethodologyandlinkedtoaGeographicInformationSystem(GIS).TheuseofDempster–Shafer
(DS)approachpermittedtoaggregateallcriteriainthreeindicatorsofsubjectiveprobabilitiesdefined,
accordingtoDStheory,as:beliefofhazard,beliefofresilienceanduncertainty.
Finallyexpectedlosscausedbyclimatechangewasestimatedbylinkingsubjectiveprobabilityand
I.Bernettietal./JournalofForestEconomics17 (2011) 285–297 287
Fig.1.Flowchartonmethodologicalframework.
Climatechange,vulnerabilityandadaptationcapability
Resilienceisdefinedastheabilityofaforesttoreacttoexternalenvironmentalchanges,
maintain-ingacertaindegreeofecologicalstability,bothintermofspeciescompositionandmaintenanceofthe
structure.Hazardrepresentsthedegreeofriskofdestructionand/ordamageofanarea.Inthisresearch
thepresentvalueofvariablesthatpotentiallyhaveagreaterinfluenceonthehazarddegreewasrelated
withthepossibilityofaclimatechange,accordingto:(i)averagespeedofwind,(ii)yearly
cumula-tiveprecipitationsand(iii)yearlyaveragetemperatures.Futurevalueoftheseclimaticvariablesare
basedonA2scenariooftheHadleyCentreMetOfficeClimateModelling3(Hadcm3)developedby
IntergovernmentalPanelonClimateChangeandre-elaboratedforTuscanyregion(Moriondoetal.,
2010).
Accordingtothenationalandinternationalliterature(Carraroetal.,2007;EFIetal.,2008)the
followingindicatorsarechosenfortheevaluation:
• Resilience
◦Managementinterventionsontheforestcrops(expressedasapercentageofforestsurfacewith
managementinterventions);
◦ Presenceofprotectedareas;
◦Ecologicaldiversity;
◦Landscapediversity.
• Hazard
◦ Wind,asnumberofyearlyextremeevents;
◦ Aridity(basedonDeMartonneindex);
◦ Slope; ◦ Aspect;
◦ Changeinphytoclimaticzone(basedonPavariclassification).
Theoryofevidence
Researchfornewmethodologiesdealingwithsubjectiveprobabilitiesandwiththeresulting
uncer-tainty(mainlyasanalternativeorarevisiontoBayesianapproach)wasintensifiedoverthelast50
years.Inforesteconomicandpolicysector,a varietyofnewapproachesforthis topichave been
proposed(e.g.HildebrandtandKnoke,2011;GadaudandRambonilaza,2010).Animportantissue
intheclimatechangeassessmentcanberepresentedbythedifficultyuncertaintyevaluation;ina
decisionproblem-solvingframework,uncertaintycouldbeaccessibleandusefultodecision-makers.
AsexpressedbyPattandDessai(2005)“theliteratureinbehavioureconomicsprovidesmanyexamples
ofhowpeoplemakedecisionsunderconditionsofuncertaintyrelyingoninappropriateheuristics,
lead-ingtoinconsistentandcounterproductivechoices”.Thereforeanin-depthanalysisabletorelaterisk
managementevaluationanduncertaintyisanecessarypolicytool.
Themethodologicalframeworkappliedinthispaperfortheclimatechangeconsequences
esti-mationderivesfromthesefollowingmethodologies:(i)fuzzysettheory,inwhichtheuncertaintyis
associatedwithvague(typicallylinguistic)variables(appliedtoclimaticresearchinScherm,2000);
(ii)Dempster–Shafertheoryofevidence(DS),thataimsatgoingbeyondthelimitsofaprobabilistic
formulation(Bayesianapproach)(LouandCaselton,1997).
TheBayesianformulationrepresentsthestartingpointoftheplausibilitynotiontreatmentinDS
theory:thetwoapproachessharetheideathattheplausiblereasoningisatypeofuncertainreasoning
becauseitiscarriedoutbyusingsourcesthatgiveinformationcharacterizedbyadegreeofreliability
butnotofcertainty.UnliketheBayesianprobabilities,DStheorydoesnotneedcompleteinformation
inthespaceoftheevents;thus,thistheoryallowstheuseoftwodifferentvaluesinordertoexpress
thecredenceinaspecificproposition,orthecredenceinitsnegation.
Inthisstudytwopropositions(calledlinesofevidence)weredefined:hypothesish=vulnerabilityof
thegeographiclocalizationandhypothesiss=resilienceofthelocalization.The“non-singular”
hypoth-esis[h,s]representsthevalueoflocalizationsthatareatthesametimevulnerableandresilient.The
conceptofDSprobabilitypdiffersfromtheconceptofBayesianprobabilitybecausefortwo
hypothe-sesA1 andA2 intheDStheorywewillhavep(h)+p(s)+p(h,s)=1and thusp(h)+p(s)<1,whilein
Bayesp(h)+p(s)=1.Theremainingp(h,s)representsthecontributiontouncertainty.Theevaluation
ofthehypothesisisbasedontheconceptofBasicProbabilityAssignment(BPA).TheBPArepresents
thecontributionthatacertainfactor(ai)givesasasupportforaspecifichypothesis(forinstancethe
vulnerabilityofastand).TheassessmentofBPAisbasedonthecombinationoffuzzyfunctionson
environmentalandsocio-economicvariablesandlinguisticevaluatorswhichwereusedinthemodel
(expressedasadegreeofbelief).Theevaluatorsweregivenbyexpertsasoralterms(Bentabetetal.,
2000);modelformulationwasexplainedinthefollowingformula:
BPA(ai,x)=linguistic(ai)·ai(xai) (1)
withlinguistic(ai)assessmentthroughafuzzylinguisticevaluatorofthecontributeinforestcrops
damage–inclimatechangescenario–ofvariableaiandai(xai)assessmentthroughamembership
functionoftheenvironmentaleffectofthevariableaionthelocalizationx.
Theaggregationforthehypotheses“hazardtoclimatechange”and“resiliencetoclimatechange”,
I.Bernettietal./JournalofForestEconomics17 (2011) 285–297 289
(orthogonalsum–Shafer,1976).FortwobasicprobabilityassessmentBPA(ai,x)andBPA(aj,x)(for
examplei=ecologicaldiversityandj=landscapediversity)theorthogonalsumis:
BPA(ai,aj)=
BPA(ai,x)·(1−BPA(aj,x))
1−BPA(ai,x)·BPA(aj,x)
(2)
Allfactorsareprogressivelyaggregatedinpairstocalculatethemassprobabilityofhazardm(h)
andthemassprobabilityofresiliencem(s).Theaggregationofthetwohypothesisfallsinthecaseof
linesofevidencemarkedbyconflictandthusitiscarriedoutbynormalizingthesumofthejointly
probabilitiesnotinconflict:
Bel(h)= m(h)·(1−m(s))
1−m(h)·m(s)
Bel(s)= m(s)·(1−m(h))
1−m(h)·m(s)
U(h,s)=Bel(s)−(1−Bel(h))= (1−m(h))·(1−m(s))
1−m(h)·m(s)
(3)
Bel(h)andBel(s)arebeliefofhazardandbeliefofresiliencemeasures;U(h,s)beliefintervalrepresents
theuncertaintydegreeoflocation,derivingbothfromoccurrenceofhighhazardandhighresilience
inthesamearea.
Evaluationoftheeconomicdamage
OnceDempster–Shafermethodologywasappliedtothetwolinesofevidenceandtheresultswere
aggregated,thequantificationoftheeconomicdamagethatclimatechangecancauseuptotheyear
2036wascalculated.
ThebasicelementforthatdefinitionistheapplicationoftheFaustmannformula,whichinits
foundationisdefinedasfollows:
F= Sf−R+
St·(1+r)f−t (1+r)f−1 −R+ a−e r (4)withSfnetstumpagevalueoffinalcut,Rregenerationcost,Ststumpagevalueofthinninginterventions,
rdiscountrate,frotation,tageofthethinningoftheforestcrop,aannualrevenuesandeannual
expenses.TheFaustmannformulaisastand-leveleconomicmodelthatwasoriginallyconceivedfor
pureeven-agedstands.Forthecurrentcase,landexpectationvalue(LEV)isusedasaproxyforthe
assessmentoftheperceivedutilitybytheownerfortheuseoftheforestrycrops.Thisformulawas
implementedthroughaGISprocedureofspatialanalysis(Bernettietal.,2009a).
Theproposedmethodusestheconceptofsubjectiveexpectedutilityvariationinordertotryto
quantifyinmonetarytermstheexpecteddamageduetoclimatechange.Thetheoryofsubjective
expectedutility(Savage,1954)combinesanutilityfunctionandasubjectiveprobabilitydistribution.
BasedonthemapsofthehazardandofthevulnerabilitydegreeoftheTuscanyforestitispossible
toapplytheFaustmannformularesultsinordertoquantifyinmonetarytermstwopossibleexpected
utilityvariations:thefirstoneexpressedastheexpecteddamagetotheforestcrop(onthebasisofthe
currenttimberassortmentsobtained–Eq.(5))andthesecondonereferredtotheexpectedexpenses,
tobesupportedinordertotrytoprotectforestfromdamages(Eq.(6)).
EUV (damage)=F·Bel(h) (5)
EUV (adaptation)=
ea r + Pt·(1+r)f−t (1+r)f−1 ·(1−Bel(s)) (6)withEUV(damage)expectedutilityvariationcausedbytheriskofcropdamage,Fcapitalizationvalue
offorestcropsfromFaustmannformula,EUV(adaptation)utilityvariationcausedbytheadaptation
cost,eacostfortheannualadaptationactions(viabilitymaintenance,firesurveillance,monitoringof
toavoidoverstockedstandssusceptibletoincreasedmortalityfromdrought,insects,diseaseand
wildfire)(AndersonandChmura,2009).
Therationalstrategyoftheforestowneristhemaximizationofhisutilityandthustheminimization
ofitsvariation(inourcasealwaysnegative).Thenthestrategicbehaviour(orstrategiccost)derives
fromtheminimizationofthetwopreviousexpectedutilityvariations:
strategic behaviour=min{EUV(damage), EUV (adaptation)} (7)
Studyareaanddata
Themodeldevelopedforthequantificationoftheeconomicdamageduetoclimatechangerefers
toforestcropsoftheTuscanyregion,localizedincentralItaly.Inthisregion,forestsurfacereaches
about1,150,000ha,withastrongvariabilitybothintermsofspeciescompositionand
geomorpho-logicalconditions.Themainimportantforestformationsaretheonescharacterizedbyprevalenceof
deciduousbroadleaved(79%),followedbyevergreenbroadleaved(13%)andconifers(8%).Themain
silviculturalsystemiscoppice(RegioneToscana,2008).
Theevaluationmodelwasimplementedthroughaspatialanalysisprocedureappliedtoraster
mapswitharesolutionof75m×75m.Thegeoreferreddatausedintheworkwerelinkedtothe
characteristicsoftheregionalforestsasforboththemanagerialaspectsdirectlyinfluencedbythe
humanactivityandthevariablesnotdependingondirectinterventions,suchasclimaticchangesand
morphologicalpeculiarities.
Thefirstpartofthestudyconsideredthenormalizationandtheweightsevaluationofthevariables
set(criteria)foreachlineofevidence.
ThesecriteriawerechosenthroughafuzzyDelphiprocedure(Cheng,1999),whichwascarried
outwiththecontributionofeightexpertsoftheforestsector,inordertodefineseveralparameters
oftheanalysismodel,bymeansofaniterativequestionnaire.Thesurveyconsideredmanagersofthe
ConsortiumofMountainousAreas,oftheStateForestryCorpsandofforestacademicstaff.
Thequestionnairemadepossibletoobtainneededparametersforthecreationofthevaluesof
environmentaleffects(fuzzymaps)ai(ai)andoftheimportanceofvariableaiinclimatechange
sce-nario(degreeofcredence)linguistic(xai);thesevalueswerenecessarytocalculatethebasicprobability
assignments(Eq.(1)).
Theexperts’opinionsweredescribedbyparametersandlinguistictermswhichcanbeexpressed
intrapezoidal(ortriangular)fuzzynumbers.Thepossibilityofcomparingdifferentindicatorsamong
eachother(suchasthepresenceofmanagedareasandthelandscapediversity)necessarilyrequires
anormalization(re-classificationofthefactorsinarangebetween0and1,thatisminimumand
maximumprobabilitythatthefactorcontributestotheevent,respectively).Table1showsthefuzzy
functionparametersusedinnormalizationofhazardandresilienceindicators.
Thefuzzylinguisticoperatorsarestrictlyrelatedtothefuzzylogicfunctions(ChenandHwang,
1992).Theserepresentamethodologywhichallowstogettoanumericquantificationofqualitative
opinionsgivenbytheexpertsofthesectorforparticulardecisionalprocesses,throughanevaluation
doneonspecificfunctionalshapes.Thus,withafuzzylinguisticoperatorwecanturntheverbal
evaluation(i.e.highinfluenceofavariableonthelineofevidence“hazard”)intoanumber,maintaining
theintrinsicuncertaintyoftheexpertestimation.ChenandHwangidentified8scalesoflinguistic
terms.Theconversionscaleadoptedforeachevaluation(Fig.2)waschosenbyexpertsintheprocess.
Theevaluationoftheothervariableswasthenimplementedthroughalinguistictermsetsderived
fromScale2oftheChenandHwangasitisshowedinTables2and3.Thelinguistictermswere
convertedintocrispvalues(high:0.808;medium:0.5;low:0.192)usingafuzzyrankingmethod.
TheBasicProbabilityAssignment(BPA)mapswereobtainedthroughtheproductofeachfuzzy
mapandtherelativedegreeofcredence.Thislastdegreederivesfromthenumericquantification
ascribedtoeachvariablethroughafuzzylinguisticevaluatorfollowingthescale4ofChenandHwang
(linguisticvalues:low:0.115,medium-low:0.3,medium:0.5,medium-high:0.7,high:0.885).The
degreeofcredenceofeachvariablewasdefinedinTable4.
Tomaketheconsensusoftheexpertsconsistent,itwasutilizedfuzzyDelphimethodtoadjust
I. Bernetti et al. / Journal of Forest Economics 17 (2011) 285– 297 291 Table1
Definitionofthemembershipfunctionsoffuzzymaps.
Lineofevidence Criteria Mapdata Fuzzymembership Controlpoint
a b c d
Resilience
Management PercentageofMunicipalityforest surfacewithmanagement interventions
Linearincreasing 30 70
Accessibility Accessibilitytotheforestareasforthe managingintervention,expressedin termsof“cumulative”differencein heightsfromtheforestviability(m)
Lineardecreasing 0 100
Ecologicaldiversity Varietydensityindex Linearincreasing 0 Maximumvalueinarea
Landscapediversity Edgedensityindex Linearincreasing 0 Maximumvalueinarea
Hazard
Wind No.ofyearlyextremeevents Linearincreasing 0 Maximumvalueinarea
Aridity DeMartonneindex Lineardecreasing 15 30
Slope % Linearincreasing 0 100
Fig.2.Scale2and4ofChenandHwang.
Table2
Evaluationofthevariable“forestsinnaturalprotectedareas”.
Typologyofprotectedareaincludingforestcrops Definitionoflinguisticevaluator
NationalPark High
RegionalPark,NationalReserve Medium
ProvincialPark,RegionalorProvincialReserve Low
experts’opinions,weconsideredthemeanoffuzzyratingsai(ai)andthemodalvalueofdegreeof
credencelinguistic(xai)(ChenandLin,1995;ChengandLin,2002).Theadoptedprocedurediffersfor
trapezoidalfuzzynumbers(ChengandLin,2002)andlinguisticsoperator,inparticularthisphasewas
carriedoutasfollows:
(a)Aggregationoffuzzynumbers.Step1:expertsprovideaparametersforeachfuzzymembership
function.Step2:foreachexpert,firsttheaverageofresultsofstep1andthen,thedifferences,
arefoundandsentbacktotheexpertsforre-examination.Step3:eachexpertpresentsarevised
Table3
Evaluationofthevariable“changeofphytoclimaticzone”.
CorineLandCover–CLCIV◦levelcodeanddescription
(forestswithprevalenceof...)
Definitionoflinguisticevaluator Fromcastanetum tolauretum Fromfagetum tocastanetum Frompicetum tofagetum 3111:evergreenoaksandotherevergreenbroadleaves Low Low Notpresent
3112:deciduousoaks Medium Low Notpresent
3113:autochthonousbroadleaves High High Notpresent
3114:chestnut High Medium Low
3115:Europeanbeech High High Medium
3116:hydrophilicspecies High High Notpresent
3117:exoticbroadleaves(locusttree) High Medium Notpresent
3121:Mediterraneanpinesandcypresses Low Low Notpresent
3122:mountainpines Medium Low Notpresent
3123:Norwayspruceand/orsilverfir High High Medium
3125:exoticconifers Medium Medium Notpresent
313:mixedforest Low Low Low
I.Bernettietal./JournalofForestEconomics17 (2011) 285–297 293
Table4
Degreeofcredenceofhazardandresiliencecriteria.
Lineofevidence Variable Degreeofcredence
Resilience
Management Medium-low
Accessibility Medium
Protectedareas Medium-low
Ecologicaldiversity Medium
Landscapediversity Medium
Hazard
Wind Medium-high
Aridity Medium-high
Changeofphytoclimaticzone High
Slope Medium
Aspect Medium-low
trapezoidalfuzzynumber.Thisprocess,startingwithstep2,isrepeateduntilthesuccessivemean
becomesreasonablyclose.
(b)Aggregationoflinguisticsterms.Step1:sameasprevious.Instep2themethodpresentedinChen
(1998)isusedtodealwithfuzzyopinionaggregationforhomo/heterogeneousgroupsofexpert,
inparticular:(i)calculationofthedegreeofagreement(ordegreeofsimilarity)oftheopinions
betweeneachpairofexperts;amethodintroducedbyChenandLin(1995)isusedformeasuring
thedegreeofsimilaritybetweentrapezoidalfuzzynumbers.(ii)Constructionoftheagreement
matrix.(iii)Calculationoftheexpertaveragevalueofagreementbytheuseofagreementmatrix.
(iv)Calculationoftheconsensusdegreecoefficient.(v)Allparametersaresentbacktoexpertfor
re-examination.Step3:eachexpertprovidesrevisedevaluationuntilconsensusdegreebecomes
acceptable.
Resultsanddiscussion
Fig.3showsthemapsandthebox-whiskerdiagramsforthebeliefofhazard,beliefofresilience
anduncertaintyinterval.Thebox-whiskerplotsinFig.3showthemean(blacklineinsidethe
box-whiskers),thefirstandthirdquartileandfinally,theupperandlowerlimitsreached.
Ingeneraltheriskofhazardisrelativelylimited,withanaverageprobabilityequaltoabout0.4.Half
oftheTuscanyforestspresentanhazardriskbetween0.35and0.49.Therearealsosomelocalizations
witharelativehigherrisk,withanupperwhiskerequalto0.71.Thevaluesregardingresilientforests
toclimatechangearehoweverrelativelylow,withameanvalueof0.38andfirstandthirdquartile
equalto0.29and0.44,respectively.Inthiscasealsotheupperlimitdoesnotseemhigh(0.66).The
uncertaintyintervalshowsanaveragevalueof0.21(firstandthirdquartileof0.18and0.24,upper
limit0.32),thusafewTuscanyforestsseemtohavebothhighhazardandhighresilienceevaluation
Fig.4.Expecteddamage,adaptationcostandstrategiccost(D /ha).
inthesamearea.Onthegeographicpointofview,theforestlocalizationsshowingthehigherdegree
ofhazardandthelowerresiliencearesituatedinthemountainousareaoftheApennines
(northwest-southeastsideoftheregion),whiletheMediterraneanforestsseemtoshow,asawhole,alowerrisk.
TheseresultsmatchwiththeItalianandinternationalliteratureontheissue(EFIetal.,2008;Carraro
etal.,2007).
Fig.4highlightsthemapsthatrepresenttheexpecteddamage,themitigationcostandthestrategic
cost.
Resultsanalysis,showsthatgenerallytheadaptationcostishigherthantheexpecteddamages.The
averagevalueoftheexpecteddamageisabout332D/ha,withthefirstandthirdquartileequalto158
and524D/ha.Theadaptationcost,instead,showsameanof950D/ha(firstandthirdquartile719and
1,215D/ha,respectively).Forthisreason,thestrategiccostcomestobe,inthemajorityoftheforest
localizations,equaltotheexpecteddamage(mean326,firstandthirdquartile155and510D/ha).
Theseresultspointoutvaluesclosetotheonespresentedinanotherrecentstudycarriedoutby
Hanewinkeletal.(2010b)withasimilarmethodology(LEVreductioncalculatedwiththeFaustmann
formula)forsouthwestGermanyNorwayspruce;theauthorscalculateanaveragereductionsofLEV
equalto521D/hainA2IPCCscenario.
Thestrategiccostrepresentedinformula(7)andinFig.4,isexpressedasacapitalvalue.However,
consideringseveralofthebasicconditionsnecessaryfortheapplicationoftheFaustmannformula(i.e.
alongtimeconstitutedforestryfundandthuscharacterizedbyconstantrevenuesandcosts,aswell
asbyapermanenceinthedestinationandintheproductiveconditions),itispossibletohypothesize
thepresenceofregulatedanduneven-agedforestsinwhichthestrategiccostisreportedaspresent
valueofannuity.ThefrequencydistributionofthisvalueisrepresentedinFig.5.
Fortheentireregionalterritorythestrategiccostamountstoabout12,500,000D/year,
correspond-ingtothe33%ofthegrossproductofthesilviculturalsectorinTuscany(IRPET,2008).Thisreduction
canbecomparedwiththeonecalculatedforothersectorssensitivetoclimaticchanges.Inthe
agricul-turalsectorBernettietal.(2009b)evaluatedreductionsbetween1930and101,718D/haforthewine
sectorand134D/haforthearablelands.Carraroetal.(2007)estimatedthat,innorthernItaly,
reduc-tionofthetouristincomecouldvaryfromaminimumof17.3%toamaximumof21.2%(asafunction
ofthescenarioandoftheconsideredregion).Internationalstudies(e.g.Stern,2007)evaluatedthat
thecostderivedfromtheclimaticchangeisvaryingbetween−4%ofGDP(thatis,netgains)to+15%
ofGDP;mostestimatesarestillcenteredaround1%ofGDP.Thus,theforestsector(togetherwiththe
agriculturalandthetouristones)seemstobemoresensitivewhencomparedtotheeconomicsystem
initscomplexity.
Theseresultsmayallownotonlytoestimatea monetaryquantificationbutalsotodefinethe
localizationofthedamage.Thegeographiclocalizationandquantificationofhazard,resilienceand
uncertainty value, permitsto downscale resultsaccording to optimal administrativeand policy
I.Bernettietal./JournalofForestEconomics17 (2011) 285–297 295
Fig.5.Annualstrategiccost.
Fig.6. Adaptationcost(leftbars)andexpecteddamage(rightbars)forTuscanyProvinces.
asamatteroffacttheoperativeactionsregardingtheItalianagriculturalpolicyaredelegatedtothe
Provinces(NUTS-3subdivision,accordingtoEurostatclassification)andtheyarethemainlocal
stake-holders,abletoaddressfundsinparticularagro-forestrysectors.Quantificationofadaptationcostand
expecteddamageinTuscanyProvincesisexplainedinFig.6.
Conclusions
TheexaminedDempster–Shaferapproachappearstobeapromisingalternativeforconducting
economicanalysisinnear-ignoranceconditionsandinanenvironmentoflimiteddataandrelevant
uncertaintysuchastheclimatechangeinforestrysector.ThefuzzymeasureandDempster–Shafer
theoriesusedinthisstudyseemtobeefficienttoolsfordataintegrationandenvironmentalimpact
assessment.
InparticulartheapplicationofDStheoryprovedtobeparticularlyusefulfromtheexpertsystems
viewpointasapowerfultoolforuncertaintyanalysis;inadditionitaggregatesallevidentialvariables
inanuniqueframeofdiscernmentthatisanefficientkindofdamageanalysisinclimatechange
riskconditions.Finally,inenvironmentaldecisionmakingsector,itcouldprovideusefulguidancefor
naturalresourceplansbasedonthelevelofprobabledamageindifferentlocations.
Ontheotherhandtheworkcanbeimprovedinsomeaspectsasfollows.Monetarycomputations
inexpectedutilityvariationcausedbytheriskandbytheadaptationcostworkundertheassumption
thatallpricesandcosts(timber,labour,interest,regeneration,etc.)areconstantovertimeaccording
totheclassicalFaustmannmodel.Howeverinrealforestry,timberandinputpricesaresubjectto
(extremeevent,increaseintemperature,etc.).Similarly,inFaustmannformulaanassumptionisthat
theinterestrateisknownandconstantovertheperiod.
In ordertoovercometheselimitstheintegrationofriskinclimatechangedamageestimation
canbeachievedalsobytheadditionofapremiumtothediscountrate(Hanewinkeletal.,2010a).
InastudycarriedoutbyKnokeandHahn(2007)simulationapproachdevelopedthroughMonte
CarloanalysiswasusedtointroduceriskevaluationinFaustmannformula.Uncertaintyrelatedto
timberpricescanbetakenintoaccountbymeansofstochasticdynamicprogramming(Lohmander,
2000).OnLEV reductionviewpointotherstudyestimatevalueclosetotheonesdepictedinthis
research,e.g.forsouthwestGermanyNorwayspruce,theapplicationofFaustmannprocedurelinked
bygeneralizedlinearmodelandsingle-treegrowthsimulatordefinedanaverageLEVreductionof
521D /havs.326D /haofthepresentresearch(Hanewinkeletal.,2010b).
Furthermorethemodelassumesthatalsodamageandmitigationsworkperpetuallyandwith
con-stantquantification.Inordertoconsiderpossiblefluctuationofclimatechange,ecosysteminteractions
andtheirconsequencesonforestcrops,process-basedapproachescanbeasolution.Dynamicsprocess
inparticularlandscapeandvegetationdynamicsorplantphysiologymodelsaresomesuitable
exam-plesofthese.Finallyrecentmethodologicaladvanceshavebeenusedtoconsiderhighlyvariableand
incompleteandnoisycharacteristicsofmostdisturbancedatasets(e.g.,machinelearningalgorithms
suchasrandomforestsandartificialneuralnetwork)(Seidletal.,2011).
Furtherfuturedevelopmentofthemodelcouldinvolvetheextensionofanalysistonon-market
datainordertoachieveamoreholisticframeworktosupportdecisionsinclimatechangeplanning.
Acknowledgements
AuthorsthanktwoanonymousreviewersaswellasDr.M.BracaliniandDr.E.Locandrofortheir
helpfulandstimulatingcommentsandsuggestions.
References
Anderson,P.D.,Chmura,D.J.,2009.Silviculturalapproachestomaintainforesthealthandproductivityundercurrentandfuture
climates.WesternForester54,6–8.
Bentabet,L.,Zhu,Y.M.,Kaftdandjian,V.,Babot,D.,MassandRombaut,D.M.,2000.Useoffuzzyclusteringfordetermining
functionsinDempster–Shafertheory.In:5thInternationalSignalProcessingProceedings,Beijing,China,pp.1462–1470.
Bernetti,I.,Fagarazzi,C.,Sacchelli,S.,Ciampi,C.,Ragaglini,G.,Villani,R.,Triana,F.,Tozzini,C.,Bonari,E.,2009a.Bacini
agroen-ergetici:stimadellapotenzialitàproduttivadelleagrienergieinToscana.In:ManualeARSIA,Firenze.
Bernetti,I.,Ciampi,C.,Sacchelli,S.,2009b.Stimadeidannieconomiciindottidaicambiamenticlimaticisulsettorezootecnico
eagricolotoscano:previsioniperilperiodo2007–2099.Aestimum54,23–50.
Buongiorno,J.,2001.GeneralizationofFaustmann’sFormulaforstochasticforestgrowthandpriceswithMarkovDecision
ProcessModels.ForestScience47,466–474.
Cayuela,L.,ReyBenayas,J.M.,Echeverria,C.,2006.ClearanceandfragmentationoftropicalmontaneforestsintheHighlandsof
Chiapas,Mexico(1975–2000).ForestEcologyandManagement226,208–218.
Carraro,C.,Crimi,J.,Sgobbi,A.,2007.LavalutazioneeconomicadegliimpattideicambiamenticlimaticiinItaliaedellerelative
misurediadattamento.In:Carraro,C.(Ed.),CambiamenticlimaticiestrategiediadattamentoinItalia:unavalutazione
economica.IlMulino,Bologna.
Chen,S.M.,1998.Anewapproachtohandlingfuzzydecisionmakingproblems.IEEETransactionsonSystems,Man,and
Cybernetics18,1012–1016.
Chen,S.M.,Lin,S.Y.,1995.Anewmethodforfuzzyriskanalysis.In:Proceedingsofthe1995ArtificialIntelligenceWorkshop,
Taipei,Taiwan,RepublicofChina,pp.245–250.
Chen,S.J.,Hwang,C.L.,1992.FuzzyMultipleAttributeDecisionMaking:MethodsandApplications.Springer-Verlag,Berlin.
Cheng,C.H.,1999.Asimplefuzzygroupdecisionmakingmethod.In:FuzzySystemsConferenceProceedings,vol.2,Seoul,
Korea,pp.910–915.
Cheng,C.H.,Lin,Y.,2002.Evaluatingthebestmainbattletankusingfuzzydecisiontheorywithlinguisticcriteriaevaluation.
EuropeanJournalofOperationalResearch142,174–186.
Comber,A.J.,Law,A.N.R.,Lishman,J.R.,2004.AcomparisonofBayes’,Dempster–ShaferandEndorsementtheoriesformanaging
knowledgeuncertaintyinthecontextoflandcovermonitoring.Computers,EnvironmentandUrbanSystems28,311–327.
Ducey,M.J.,2001.Representinguncertaintyinsilviculturaldecision:anapplicationoftheDempster–Shafertheoryofevidence.
ForestEcologyandManagement150,199–211.
EFI,BOKU,INRA,AISF,2008.ImpactsofclimatechangeonEuropeanforestsandoptionsforadaptation.ReporttotheEuropean
CommissionDirectorate-GeneralforAgricultureandRuralDevelopment,AGRI-2007-G4-06.
Flannigan,M.D.,Stocks,B.J.,Wotton,B.M.,2000.Climatechangeandforestfires.TheScienceoftheTotalEnvironment262,
221–229.
Gadaud,J.,Rambonilaza,M.,2010.Amenityvaluesandpaymentschemesforfreerecreationservicesfromnon-industrialprivate
I.Bernettietal./JournalofForestEconomics17 (2011) 285–297 297
Hanewinkel,M.,Hummel,S.,Albrecht,A.,2010a.Assessingnaturalhazardinforestryforriskmanagement:areview.European
JournalofForestResearch,doi:10.1007/s10342-010-0392-1.
Hanewinkel,M.,Hummel,S.,Cullmann,D.A.,2010b.Modellingandeconomicevaluationofforestbiomeshiftsunderclimate
changeinSouthWestGermany.ForestEcologyandManagement259,710–719.
Hildebrandt,P.,Knoke,T.,2011.Investmentdecisionsunderuncertainty—amethodologicalreviewonforestsciencestudies.
ForestPolicyandEconomics13,1–15.
IRPET,2008.10◦RapportoeconomiaepoliticheruraliinToscana.SupplementofAgrisole.
Kauppi,P.,Posch,M.,1988.AcasestudyoftheeffectsofCO2-inducedclimaticwarmingonforestgrowthandtheforestsector:
aproductivityreactionsofnorthernborealforests.In:Parry,M.L.,Carter,T.R.(Eds.),TheImpactofClimaticVariationson
Agriculture,vol.1.AssessmentinCoolTemperateandColdRegions.Kluwer,Dordrecht,TheNetherlands,pp.183–195.
Knoke,T.,Hahn,A.,2007.BaumartenvielfaltundProduktionsrisiken:EinForschungseinblickund-ausblick.Schweizerischen
ZeitschriftfürForstwesen158,312–322.
Lexer,M.J.,Hönninger,K.,Scheifinger,H.,Matulla,C.,Groll,N.,Kromp-Kolb,H.,Schadauer,K.,Starlinger,F.,Englisch,M.,2002.
ThesensitivityofAustrianforeststoscenariosofclimaticchange:alarge-scaleriskassessmentbasedonamodifiedgap
modelandforestinventorydata.ForestEcologyandManagement162,53–72.
Lohmander,P.,2000.Optimalsequentialforestrydecisionsunderrisk.AnnalsofOperationsResearch95,217–228.
Lou,W.B.,Caselton,B.,1997.UsingDempster–Shafertheorytorepresentclimatechangeuncertainties.JournalofEnvironmental
Management49,73–93.
Loustau,D.,Bosc,A.,Colin,A.,Ogée,J.,Davi,H.,Franc¸ois,C.,Dufrêne,E.,Déqué,M.,Cloppet,E.,Arrouays,D.,LeBas,C.,Saby,N.,
Pignard,G.,Hamza,N.,Granier,A.,Bréda,N.,Ciais,P.,Viovy,N.,Delage,F.,2005.Modelingclimatechangeeffectsonthe
potentialproductionofFrenchplainsforestsatthesub-regionallevel.TreePhysiology25,813–823.
Nitschke,C.R.,Innes,J.L.,2008.IntegratingclimatechangeintoforestmanagementinSouth-CentralBritishColumbia:an
assess-mentoflandscapevulnerabilityanddevelopmentofaclimatesmartframework.ForestEcologyandManagement256,
313–327.
Mora,B.,Fournier,R.A.,Focher,S.,2010.Applicationofevidentialreasoningtoimprovethemappingofregeneratingforest
stands.InternationalJournalofAppliedEarthObservationandGeoinformation,doi:10.1016/j.jag.2010.10.001.
Moriondo,M.,Pacini,C.,Trombi,G.,Vazzana,C.,Bindi,M.,2010.SustainabilityofdairyfarmingsysteminTuscanyinachanging
climate.EuropeanJournalofAgronomy32,80–90.
Patt,A.,Dessai,S.,2005.Communicatinguncertainty:lessonslearnedandsuggestionsforclimatechangeassessment.External
Geophysics,ClimateandEnvironment337,425–441.
Raje, D., Mujumdar, P.P., 2010. Hydrologic drought prediction under climate change: uncertainty modeling with
Dempster–ShaferandBayesianapproach.AdvancesinWaterResources33,1176–1186.
RegioneToscana,2008.RapportosullostatodelleforesteinToscana2007.Litografeditor,Perugia.
Savage,L.J.,1954.TheFoundationsofStatistics.NewYork,NY,Wiley.
Scherm,H.,2000.Simulatinguncertaintyinclimate-pestmodelswithfuzzynumbers.EnvironmentalPollution108(3),373–379.
Seidl,R.,Fernandes,P.M.,Fonseca,T.F.,Gillet,F.,Jönsson,A.M.,Merganicová,K.,Netherer,S.,Arpaci,A.,Bontemps,J.D.,
Bug-mann,H.,González-Olabarria,J.R.,Lasch,P.,Meredieu,C.,Moreira,F.,Schelhaas,M.J.,Mohren,F.,2011.Modellingnatural
disturbancesinforestecosystem:areview.EcologicalModelling222,903–924.
Shafer,G.,1976.AMathematicalTheoryofEvidence.PrincetonUniversity,Princeton.
Solomon,A.M.,1986.TransientresponseofforeststoCO2-inducedclimatechange:simulationmodelingexperimentsineastern
NorthAmerica.Oecologia68,567–579.
Stern,N.,2007.TheEconomicsofClimateChange:TheSternReview.CambridgeUniversityPress,Cambridge.
Varma,V.K.,Ferguson,I.,Wild,I.,2000.Decisionsupportsystemforthesustainableforestmanagement.ForestEcologyand