Uncertainties related to climate change and forest management with implications on climate regulation in Finland

Unce ain ies ela ed o clima e change and o es managemen wi h implica ions on clima e egula ion in Finland Ja i Vauhkonen ⇑ , Tuula Packalen Na u al Resou ces Ins i u e Finland (Luke), Bioeconomy and En i onmen uni , Yliopis oka u 6, FI-80100 Joensuu, Finland a icle in o A icle his o y: Recei ed 6 Oc obe 2017 Recei ed in e ised o m 8 Decembe 2017 Accep ed 19 Feb ua y 2018 A ailable online 3 Ma ch 2018 abs ac Fo es s play an impo an ole in one o he mos impo an ecosys em se ices, clima e egula ion. In o de o mi iga e clima e change, a ious in e na ional ag eemen s aim a dec easing emissions h ough Land-Use, Land-Use Change and Fo es y (LULUCF) ac i i ies. In a legisla i e p oposal by he Eu opean Union, emissions om o es s a e accoun ed o in ela ion o an es ima e o a e age emissions o a ange o yea s in he pas . Howe e , di e en o es s uc u es, managemen ac i i ies, g ow h a ia ions and impac s o changing clima e may esul in conside ably di e en u u e emissions. We assessed he magni ude o po en ial unce ain ies due o changing clima e and o es managemen o he p ojec ions o ca bon s ocked in abo e- and belowg ound o es biomass in Finland un il 2050. We used an a ea- based ma ix model, which was de eloped o inco po a e clima e-induced ee g ow h as a ime- inhomogeneous Ma ko chain. The po en ial amoun s o bo h he ca bon s o ed and ex ac ed a ied conside ably depending on he le el and alloca ion o u u e ha es s. I ealized, clima e- o managemen -induced g ow h imp o emen s could inc ease he ca bon s ocks by up o one hi d in he end o he simula ed pe iod. P ojec ions based solely on business-as-usual ansi ions and ha es s could he e o e lead o ine icien decisions ega ding u u e ca bon s ocks and ha es ing possibili ies. Ó2018 The Au ho s. Published by Else ie B.V. This is an open access a icle unde he CC BY license (h p:// c ea i ecommons.o g/licenses/by/4.0/). 1. In oduc ion Clima e egula ion is one o he mos impo an ecosys em se - ices (Cos anza e al., 2017), o which o es s and o es y play an impo an ole. On one hand, ca bon accumula es h ough g ow h o ees in o o es g owing s ock. On he o he hand, Land-Use, Land-Use Change and Fo es y (LULUCF) ac i i ies impac ca bon s ocks. Fi s , sus ainable land-use and o es managemen can con- se e o inc ease o es ca bon s ocks. Second, ha es ed wood- based p oduc s and hei bi-p oduc s can eplace ossil-based p oduc s, ma e ials and ene gy. The complexi y o ade-o s be ween ees le g owing o clima e egula ion se ice o hose ha es ed o p o isioning se ices – and consequen ly o clima e egula ion as di e se p oduc s – place challenges o decision mak- ing ega ding LULUCF ac i i ies and hei egula ion. The issues ela ed o LULUCF a e e lec ed by egula ion mea- su es in a ious in e na ional ag eemen s unde he Uni ed Na ions F amewo k Con en ion on Clima e Change (UNFCCC). In line wi h he Pa is Ag eemen (UNFCCC, 2017), he Eu opean Com- mission (EC) has p esen ed a legisla i e p oposal (EC, 2016) o se a binding commi men o each membe s a e o ensu e ha accoun ed emissions om land use a e en i ely compensa ed by an equi alen emo al o CO 2 om he a mosphe e h ough ac ion in he sec o , known as ‘‘no debi ule”. In he p oposed ules, emis- sions om o es s a e accoun ed o in ela ion o a so called na ional o es e e ence le el. The o es e e ence le el is an es i- ma e o he a e age annual ne emissions o emo als esul ing om managed o es land wi hin he e i o y o a membe s a e. The a e age alues a e calcula ed o a ange o yea s in he pas , e e ed o as a e e ence pe iod, which is a poli ical decision ha applies o all coun ies. In p ac ice, he pas o es g ow h, ha es s and, consequen ly, ne emissions may a y a lo be ween yea s wi hin a coun y and he a ia ion pa e n o e ime is no he same in all coun ies. Fi s , he e is a ia ion o sinks due o he g ow h a ia ion o ees (Mäkinen e al., 2002). Second, he e is in e annual a ia ion o emissions (Ande sson e al., 2007), due o he ma ke luc ua ion, o example. In addi ion, clima ic a ia ion a ec s he ee g ow h and he impac s o changing clima e a e assumed o a y be ween egions depending on hei ecological condi ions (Cha u e al., 2017) and managemen (Hen onen e al., 2017). Fu he mo e, glo- bal ma ke u bulence accele a es luc ua ion in oundwood ma - ke in Eu ope (Packalen e al., 2017) and he impac s o global h ps://doi.o g/10.1016/j.ecose .2018.02.011 2212-0416/Ó2018 The Au ho s. Published by Else ie B.V. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/). ⇑ Co esponding au ho . E-mail add ess: [email p o ec ed] (J. Vauhkonen). Ecosys em Se ices 33 (2018) 213–224 Con en s lis s a ailable a ScienceDi ec Ecosys em Se ices jou nal homepage: www.else ie .com/loca e/ecose changes a e expec ed o ealize a di e en phase in di e en coun- ies depending on hei socio-economic si ua ion. Consequen ly, he use o he same e e ence pe iod o all coun ies may penalize o bene i a coun y depending on how well he u u e o es esou ce and ma ke up u n and down u n coincide wi h each o he compa ed o he na ional e e ence alue. Fo es esou ce p ojec ions ha e been used o model u u e bio- mass supplies a he Eu opean and na ional le els. The e is a la ge a ia ion in da a and models used o he na ional p ojec ions ai- lo ed o local condi ions and in o ma ion needs (Ba ei o e al., 2016). Fo example, sus ainabili y and he p o isioning o o he ecosys em se ices han biomass-based p oduc s a e aken in o accoun as a se ies o ecological, economic and echnical con- s ain s, which limi he a ailabili y o accessibili y o o es s o wood supply (Albe di e al., 2016). The Eu opean s udies a e o en summed up om na ional p ojec ions (Schelhaas e al., 2017) ca - ied ou using a pan-Eu opean da a se and a gene ic model ha canno ully accoun o di e ences be ween coun ies. Conse- quen ly, he esul s om he s udies summed up o he Eu opean le el o en show a la ge a ia ion, usually claimed o be a esul o inhe en unce ain ies ela ed o da a (e.g., Re enmaie e al., 2010; Ben sen and Felby, 2012) and biomass es ima ion models (Neumann e al., 2016). E en i con ex -dependen changes in u u e land-use, o es managemen , and clima e ob iously ha e implica ions ha p opaga e he p ojec ions as unce ain ies, s ud- ies add essing hese aspec s a e missing (Ba ei o e al., 2016; Schelhaas e al., 2017). Acco ding o he legisla i e p oposal o EC (2016), he u u e ca - bon pools o es ima ing he emissions should be p ojec ed assum- ing a ‘‘con inua ion o cu en o es managemen p ac ice and in ensi y” o make he emission accoun ing compa able be ween o he sec o s and membe s a es, bu also enable accoun ing o coun y-speci ic o es y dynamics. Based on his p inciple, G assi and Pilli (2017) desc ibed a simula ion amewo k, which is (a) pa ame e ized by he p e ailing o es age-class s uc u e, inc e- men s, and business-as-usual ha es ing p ac ices and in ensi y; and (b) used o p ojec he ca bon pools a e he e e ence pe iod, assuming ha ha es s a e con inued in a simila magni ude as in he e e ence pe iod, bu ela i e o he de elopmen o biomass a ailable o wood supply ( o de ails, see especially Box 1 in G assi and Pilli, 2017). Howe e , because o he mul iple ac o s causing a ia ion o he u u e scena ios, as e iewed abo e, he membe s a es would mos likely bene i om he assessmen o u u e o es ca bon sink unce ain ies when nego ia ing on he na ional o es e e ence le el. Bayesian in e ence echniques such as Ma ko chain models (e.g., Nabuu s e al., 2000; Thü ig and Schelhaas, 2006; E iksson e al., 2007; Ve ke k e al., 2011) may be applicable o quan i ying unce ain ies (c ., Smi h and Ma shall, 2008), due o he po en ial o lexibly a y he assump ions ela ed o u u e scena ios (see also Vauhkonen and Packalen, 2017). The aim o his s udy is o es a Ma ko chain model o assess- ing he deg ee o unce ain ies in he p ojec ions o ca bon s ocked in abo e- and belowg ound o es biomass in changing clima e and in he con ex o LULUCF egula ion in Eu ope. The main objec i e is o de elop me hodology o inco po a e ee g ow h a ia ion and impac s o changing clima e in o an a ea-based, Ma ko chain model de eloped o p ojec ing di e en managemen scena ios (Vauhkonen and Packalen, 2017). The seconda y objec i e is o apply he me hodology o quan i y he unce ain ies ela ed o o - es ca bon, and consequen ly, o he selec ion o na ional o es e - e ence le el in he LULUCF egula ion o Finland. Ou analyses undamen ally co e wo ecosys em se ices: oundwood ha es s as a p o isioning se ice and he ela ed e ec s on he ca bon ex ac ed and s o ed in he emaining g owing s ock as a egula- ion se ice. 2. Ma e ial and me hods 2.1. O e iew As he Ma ko chain model, we used . 2.0. o he Eu opean Fo es y Dynamics Model (EFDM), which is implemen ed in he R s a is ical modeling en i onmen (R Co e Team, 2016) and can be downloaded om he EFDM p ojec eposi o y (FISE, 2017) as open sou ce unde he Eu opean Union Public License (EUPL). The EFDM is an a ea-based ma ix model, in which he ma ices ep esen o es a eas classi ied acco ding o ecological and socio-economic ac o s. The EFDM simula es he de elop- men o he o es a ea dis ibu ion as a p oduc o i s ini ial s a e, p opo ions o a eas expec ed o be managed acco ding o di e en sil icul u al p ac ices, and he co esponding ansi- ion p obabili ies. The ansi ion p obabili ies a e condi ioned on he ac i i ies, which can bo h di e be ween ac o s such as si e ype, species, owne , and o he ac o s ei he a ec ing he o es dynamics o needed o epo ing. As elabo a ed by Si kiä (2012) and Packalen e al. (2014), he e is a ansi ion ma ix pe ac o combina ion and pe ac i i y. The ini ial s a e and ac i i y and ansi ion p obabili y ma ices can be de i ed h ough a simple classi ica ion and agg ega ion ou ine om Na ional Fo es In en o y (NFI) plo da a, i he u u e de elopmen is assumed o ollow ha ealized in he pas . G ow h models o simula o s modi ying he pai wise obse a- ions can be used o de i e ansi ion p obabili ies unde chang- ing clima e. The EFDM was pa ame e ized o he cu en clima e using ansi ion and ac i i y p obabili ies de i ed om pe manen NFI plo s as desc ibed in de ail in he open-access a icle by Vauhkonen and Packalen (2017). The amewo k was ex ended o include e ec s o clima e change and con e he ou pu s o ca - bon. The analyses ca ied ou he e aim a quan i ying he deg ee o unce ain y occu ing, when decisions a e made acco ding o he ansi ion p obabili ies obse ed in he pas , bu changes o hese ansi ions occu due o he clima e o managemen imp o - ing he g ow h. The gene al amewo k and especially hese changes a e desc ibed below, bu ega ding de ails o he pa ame- e s and hei e ec s o he ou pu , he eade is e e ed o he pape by Vauhkonen and Packalen (2017). The simula ions we e ca ied ou in 5-yea pe iods, which co - espond o he measu emen in e al in he pe manen plo da a. Al oge he eigh pe iods we e simula ed, i.e., he las yea o sim- ula ions is a ound 2050, depending on he ini ial measu emen yea . I was assumed ha he land-use es ic ions de e mined he sil icul u al sys em applied. The de elopmen o o es s wi h- ou es ic ions was simula ed acco ding o e en-aged manage- men and age and olume as he axes o he ma ices. Fo es s wi h es ic ions on wood supply we e simula ed acco ding o an une en-aged managemen , whe e inal ellings we e eplaced wi h hinnings om abo e and he simula ions we e based on s em numbe and olume ma ices. Only he na u al p ocesses we e simula ed o o es s no a ailable o wood supply. The de i a ion o he ini ial da a and ansi ion p obabili ies co e- sponding o obse a ions made om he pe manen NFI plo s a e desc ibed in Sec ions 2.2 and 2.3.1. Fo he unce ain y assess- men , expec ed e ec s o clima e and adap ing he ansi ion p obabili ies due o hese changes we e modeled using ‘‘a ansi- ion p obabili y d i e ”, as desc ibed in Sec ion 2.3.2. Th ee di - e en ha es ing a ge s and wo al e na i e alloca ions o he ha es s we e applied, as desc ibed in Sec ion 2.4. The ha es decisions we e based on oundwood olume (measu ed in m 3 ), bu ansla ed o ca bon ( onnes) using ou pu coe icien s as desc ibed in Sec ion 2.5. 214 J. Vauhkonen, T. Packalen / Ecosys em Se ices 33 (2018) 213–224 2.2. Ini ial s a e space The sample plo da a om he 11 h Finnish Na ional Fo es In en o y (NFI11; al oge he 51,827 o es plo s measu ed in 2009–2013) we e used o he es ima e he ini ial dis ibu ion o o es in an a ea o al oge he 21.28 mill ha on p oduc i e and poo ly p oduc i e o es land in Finland. The g owing s ock was 2,234 mill m 3 , which is app oxima ely 95% o he en i e g owing s ock in Finland, and only a eas wi h low impo ance o o es y we e excluded. Fo he analyses, he o es s we e classi ied o ma ices wi h axes co esponding o ei he age and olume o s em numbe and olume unde e en-aged o une en-aged manage- men , espec i ely. To p oduce an adequa e amoun o obse a- ions o he es ima ion o he ansi ion p obabili ies, he class limi s o bo h he olume and s em numbe we e de i ed as he alues o he 10 h, 20 h, ..., 90 h and 95 h quan iles o he pai - wise obse a ions made om he pe manen NFI plo s. The age classes we e de ined as 0, 5, 10, .. ., 120, 120+ yea s, he class in e - al o i e yea s co esponding o bo h he measu emen s and he simula ion s ep used in he analyses. The class limi s o he con- inuous a iables a e p esen ed as Appendix A. These ma ices we e de i ed sepa a ely applying he ollowing, s a ic land-use classes: (i) known land-use es ic ions: o es s a ailable, o es s wi h es ic ions on a ailabili y, and o es s no a ailable o wood supply; (ii) o es owne ship: p i a e, public + o he ; (iii) si e e il- i y: al oge he , i e ca ego ies co esponding o ou axa ion classes adi ionally used in Finland + i h class including all poo ly p oduc i e o es land; (i ) dominan species: pine, sp uce, deciduous ees. 2.3. T ansi ion p obabili ies 2.3.1. Cu en clima e The ansi ion p obabili ies co esponding o he cu en cli- ma e we e de i ed using pai wise obse a ions om pe manen plo s o NFI11 (al oge he , 11,987 plo s), which we e measu ed app oxima ely i e yea s ea lie in he p e ious in en o y (NFI10). Posi i e di e ences in he o al olumes on plo s wi h no ea men s based on da a ha could be ma ched wi h ce ain y be ween he wo subsequen in en o ies we e eco ded as he pai wise da a. The es ima ed ansi ions he e o e included only g ow h and no po en ial educ ions due o calami ies o na u al dis u bances, o example. The ansi ions due o managemen ac i i ies we e based on simula ions o hei expec ed de elopmen . The o es s a ec ed by inal ellings we e o ced o ansi o he beginning o he e en-aged o a ion. A hinning simula o was implemen ed o de i e pai wise obse a ions be o e and a e he ea men s. The hinnings ook place in he beginning o each simula ion pe iod and he g ow h o he hinned o es s in ha pe iod was simula ed applying he ansi ion p obabili ies o o es s no managed. 2.3.2. Adap ing he ansi ion p obabili ies acco ding o he expec ed clima e change We used he ollowing wo k low, called ‘‘ ansi ion p obabil- i y d i e ”, o include he e ec s o clima e-induced o es g ow h in he Ma ko chain model. In ou d i e , a g ow h end wi h s ochas ic a ia ion was modeled in ou s eps, he i s wo o which a e ela ed o p edic ing inc eased CO 2 and empe a u e unde clima e change scena ios and he la e wo s eps o using hese alues o p edic ing he esul ing g ow h inc emen . 1. CO 2 . G eenhouse gas (GHG) emissions esul ing om clima e change we e expec ed o de elop acco ding o he Rep esen a- i e Concen a ion Pa hway (RCP) scena ios. We used h ee ou o he ou scena ios, igno ing RCP6.0, as i s e ec s wi h espec o o es g ow h we e p ac ically simila o hose o RCP4.5 based on ou modeling app oach. In he scena ios RCP2.6, RCP4.5, and RCP8.5, he le el o ambien CO 2 is expec ed o ise om he cu en le el o 350 ppm o 443, 487, and 541 ppm, espec i ely, by 2050 (Meinshausen e al., 2011). 2. Tempe a u e. The Finnish Me eo ological Ins i u e has p o- jec ed he annual mean su ace ai empe a u es in Finland o inc ease by 2040–2069 depending on he le el o GHG emis- sions ealized in he di e en RCPs. The expec ed changes in he annual mean empe a u es, ela i e o 1981–2010, a e exp essed as no mal dis ibu ions wi h pa ame e s o he di - e en RCPs (Ruos eenoja e al., 2016). The alue ep esen ing he change in empe a u e by 2050 was ob ained as a andom alue om he RCP-speci ic dis ibu ions. The empe a u e co - esponding o 10%, 50%, and 90% alues o he dis ibu ions we e compu ed o he g ow h end modeling. 3. G ow h end. Ma ala e al. (2005, 2006) used p edic ions ob ained om a physiological g ow h model, FinnFo , o desc ibe he impac s o ele a ed empe a u e and CO 2 on ee g ow h. The models p esen ed gi e a a io (o Rela i e Scena io E ec , RSE ) o he olume g ow h unde he changing clima e o ha unde he cu en clima e. Sepa a e models, con olled by s and densi y, compe i ion, si e e ili y and cu en empe - a u e sum, a e p esen ed o Sco s pine, No way sp uce and sil- e bi ch. To implemen he models in con en ional g ow h simula ions, Ma ala e al. (2005, 2006) also exp ess RSE as a shape e ec SE, which is a a io o he heigh and olume g ow h, and p esen eg ession models o SE. The models o Ma ala e al. (2005, 2006) a e a ailable only o mine al soils; howe e , Nuu inen e al. (2006) modeled SE also o pea lands. We used he CO 2 and empe a u e alues ob ained acco ding o he desc ip ion abo e in he models o de i e RSE V and used his a io as he g ow h end o ou analyses. We compu ed RSE o bo h mine al and pea lands using he models o Ma ala e al. (2006), bu adjus ed he RSE alues o pea lands acco ding o he a io o SE o mine al soil (Ma ala e al., 2006) o SE o pea lands (Nuu inen e al., 2006). 4. S ochas ic annual a ia ion in ee g ow h. In addi ion o he g ow h end, we wan ed o include he a ia ions in ee g ow h as obse ed in he pas g ow h se ies (c ., Hen onen e al., 2017). We used he au o eg essi e mo ing-a e age mod- els i o a end-clea ed g ow h-index se ies obse ed om Fin- land om 1890 o 1988 (Pasanen, 1998), which ake in o accoun he posi i e au oco ela ion be ween successi e yea s and he posi i e c oss-co ela ion be ween he g ow hs o di - e en ee species. To ob ain se ies o g ow h ends wi h s ochas ic a ia ion, he p e ious ou s eps we e epea ed 1000 imes o each RCP and plo . The ini ial and clima e-induced olume alues we e classi ied o olume classes using class limi s desc ibed abo e (see also Appendix A). Conside ing each o he simula ion pe iods sepa a ely, i was compu ed how many imes o he 1000 d aws he plo s wi h he enhanced g ow h inc eased o a highe olume class han wi h he ini ial ansi ion p obabili ies. Mul iplying his p opo ion wi h he a ea ep esen ed by a plo ga e he o al a ea es ima e ha was expec ed o ansi as e unde he gi en RCP. These a ea p opo ions we e ansla ed o ansi ion p obabili ies simila o hose based on he ini ial pai wise NFI obse a ions (Sec ion 2.3.1). In he EFDM simula ions, he ini ial J. Vauhkonen, T. Packalen / Ecosys em Se ices 33 (2018) 213–224 215 ansi ion p obabili ies ( he single p obabili y ma ix de i ed om he NFI obse a ions) we e used o compu a ions ep esen ing he cu en clima e, and RCP and simula ion pe iod speci ic ansi ions ob ained using he d i e o hose ep esen - ing clima e change. 2.4. Ha es scena ios and managemen ac i i ies We p ojec ed he de elopmen o he ca bon s ock assuming h ee di e en ha es ing le els o oundwood: Business-As- Usual (BAU), Na ional Fo es P og amme (NFP) and Non- Declining Volume (NDV). The BAU and NFP scena ios we e based on ixed ha es ing le els o 61.75 mill m 3 /a and 76 mill m 3 /a, espec i ely, which we e ob ained om he Na ional Fo es S a - egy o Finland (Minis y o Ag icul u e and Fo es y, 2015)by adjus ing he p oposed le els o he 95% amoun o g owing s ock conside ed he e. The NDV scena io had an adap i e le el o ha - es s ha was de e mined as he le el, which did no dec ease he olume o he g owing s ock. The alloca ion o ha es s o di e en ypes o o es s was de e mined acco ding o ac i i y p obabili ies, which gi e he p o- po ions o managemen ac i i ies in he da a classi ied acco ding o he ac o combina ions. Two al e na i e app oaches we e es ed as he alloca ion o he u u e ha es s. Fi s , a business- as-usual alloca ion (A BAU ) was ob ained as he p opo ion o a eas wi h no managemen , hinning o inal elling ealized du ing he mos ecen i e-yea pe iod, i.e. based on obse a ions om he pe manen plo s be ween he wo subsequen in en o ies. Second, a schoolbook-alloca ion (A SB ) was ob ained as he p opo ion o a eas, which we e ma ked in he ield wi h a need o be ha es ed wi hin he nex i e yea s s ic ly acco ding o he ins uc ions o o es managemen (Y jölä, 2002). Fig. 1 depic s he di e ences in he wo al e na i e ha es alloca ions. To ul ill he ha es ing objec i es, he ac i i y p obabili ies based on bo h al e na i e ha es alloca ions we e i e a ed o yield a ha es ing d ain o oundwood co esponding o BAU, NFP, o NDV scena ios (Vauhkonen and Packalen, 2017). The ac i i y p obabili ies used in all analyses we e compu ed assuming g ow h a es o he cu en clima e, i.e., using ansi- ion p obabili ies de i ed om he pai wise NFI obse a ions. ‘‘Unce ain y” in he impac analyses he e o e o igina es om po en ial clima e o managemen induced addi ional g ow h ha occu s on op o he g ow h ha is expec ed based on he ansi ions obse ed a he ime when making he ha es - ing decisions. 2.5. Ou pu coe icien s Coe icien s o he mean alues o a ea classes we e de e - mined o quan i y he imbe asso men d ain and ca bon s ocked in o ex ac ed om he o es . Simila o Vauhkonen and Packalen (2017), we de i ed he imbe asso men d ain by compu ing he ela i e p opo ion o log- and pulpwood p opo ions o he en i e olume sepa a ely o inal ellings and hinnings, using all NFI plo s o hose plo s o which he hinnings we e simula ed, espec i ely. The biomass in componen s (s em, b anches, oliage, s ump, oo s) was calcula ed o each plo (Repola, 2008, 2009). To ob ain he ca bon con en , he biomasses we e mul iplied by species-speci ic expansion ac o s (a ound 0.5; see Table 1 o Pukkala, 2014). 3. Resul s 3.1. E ec s o clima e-induced addi ional ee g ow h o he ansi ion p obabili ies The ansi ion p obabili ies ha we e adap ed o he changing clima e di e ed be ween he RCPs as expec ed. Fig. 2 depic s hese di e ences as he p opo ion o a ea ha is expec ed o ansi as- e han wi h he p obabili ies de i ed om he pai wise NFI obse a ions. Fig. 2 indica es ha he a ea expec ed o ansi as- e inc eased acco ding o he ime s eps o he simula ion, bu his p opo ion a ied acco ding o si e cha ac e is ics such as e ili y. When he sou ce and a ge classes o he as e ansi ions a e examined a he class-le el (Fig. 3), i can be seen ha a e age ansi ions we e usually equal o hose based on he ini ial pai - wise obse a ions. Howe e , he e we e mo e equen ly jumps o mo e han one class han was obse ed in he case o he an- si ion p obabili ies de i ed om he pai wise obse a ions. The esul s abo e a e based on using RCP4.5 as an example scena io. Wi h espec o o he RCPs, he esul s did no essen ially di e excep o he magni udes o clima e-induced changes. The con e sion o a eas o ca bon using he ou pu coe icien s yielded a de elopmen pa e n ha can be assumed o ealis ically mimic he ca bon dynamics in bo eal o es . Among indi idual ca - bon (o biomass) componen s, s em o oliage g ew mos o leas apidly, espec i ely. The di e ences be ween si e ypes we e in he o de o 5.7–8 pe cen age poin s, species 8.5–11 pp, and geo- g aphic a eas 8.4–10.7 pp o he componen g ow h, and clima e change ampli ied hese igu es. Howe e , he main di e ences Fig. 1. P opo ions o managemen ac i i ies in he business-as-usual (A BAU , le ) o schoolbook (A SB , igh ) ha es alloca ion. Re e o Appendix A o he de ini ion o olume classes. 216 J. Vauhkonen, T. Packalen / Ecosys em Se ices 33 (2018) 213–224 we e ela ed o age-size dis ibu ions: Fig. 4 shows ha clima e- induced p edic ions had simila ends han hose de i ed om he pai wise NFI obse a ions. Especially he g ow h o he leas ma u e o es s was ampli ied by he clima e-adap ed ansi ion p obabili ies. 3.2. De elopmen o he ca bon s ock and he ela ed unce ain ies due o clima e and managemen Figs. 5–6 p esen he esul s o he de elopmen o ca bon s ocks and d ains, when he ansi ion p obabili ies desc ibed in Sec ion 3.1 we e applied oge he wi h he ac i i y p obabili ies o ha es s. The di e ence be ween Figs. 5 and 6 is ha in he o - me , he ac i i y p obabili ies (Sec ion 2.4) we e i e a ed in e e y simula ion s ep o yield he desi ed ha es goal, i.e., an equal ol- ume o oundwood was ha es ed in e e y simula ion pe iod. In Fig. 6, he ac i i y p obabili ies we e ixed o he le el ha yielded he ha es goal in he beginning o he simula ions, i.e., an equal p opo ion o a ea was always ha es ed, whe eas he olume ha - es ed a e he i s pe iod depended on he (in-)g ow h o o es a ea o he speci ic class. In all cases, he managemen decisions we e made acco ding o he ansi ion and ac i i y p obabili ies de i ed om he o iginal pai wise NFI obse a ions, which is illus- a ed using g ey ba s in Figs. 5–6. The lines in he igu es depic al e na i e cou ses o de elopmen , which we e ob ained by eplacing ei he ansi ion p obabili ies wi h hose induced by cli- ma e (Sec ion 3.1) o business-as-usual ha es alloca ion wi h ha based on schoolbook (Sec ion 2.4) o applying bo h o hese changes. 3.2.1. Ca bon dynamics unde cu en clima e and business-as-usual managemen Based on he business-as-usual ansi ions and ac i i ies, wi h ac i i y p opo ions i e a ed o yield he desi ed ha es goal (Fig. 5), ha es ing 61.75 mill m 3 o oundwood pe yea (BAU sce- na io) inc eased he ca bon s ock om 810 o 1065 mill onnes (31%) by he end o he simula ion. Ha es ing 76 mill m 3 acco d- ing o he NFP scena io i s sligh ly inc eased he s ock, bu ended up o educe he s ock o 757 mill onnes (7%) in he end o he simula ion. The equi emen o no decline he olume esul ed in a linea ly educing amoun o ha es s and, subsequen ly, ca - bon ex ac ed. The esul s di e ed conside ably, i p opo ions o a ea yielding he desi ed ha es le el we e ixed in he beginning o he simu- la ions (Fig. 6). The di e ences we e also mos p onounced wi h he BAU ha es ing scena io, whe e he le el o ha es s inc eased by ime. This was because o highe p opo ion o o es ha ma u ed due o low ini ial ha es s and he ixed p opo ion o his a ea o be ha es ed acco ding o he ac i i y p obabili ies. As a esul o inc easing he ha es s, howe e , he g owing s ock dec eased. Wi h he NFP scena io, he de elopmen was ai ly sim- ila as desc ibed in he p e ious pa ag aph. Wi h he NDV scena io, he g owing s ock dec eased unlike when he ac i i y p obabili ies we e con inuously i e a ed du ing he simula ions, whe eas he ha es le els educed less han when i e a ed. 3.2.2. Le el o unce ain ies due o clima e and managemen Bo h he imp o ed g ow h and he change o ha es alloca ion om A BAU o A SB inc eased he ha es ing possibili ies in all scena - ios desc ibed abo e. Compa ed o business-as-usual, he clima e change alone esul ed o 5–33% highe ca bon s ock in abo e- and belowg ound o es biomass in addi ion o ob aining 8–20% highe ha es d ain. The a o emen ioned le els depended on he ha es ing goals and i he ac i i y p obabili ies we e i e a ed du - ing he simula ions (Figs. 5–6). A compa ison o he clima e- and managemen induced e ec s shown in Figs. 5–6 indica es ha he impac s o clima e change we e pa ly ela ed o hose p oduced by he di e en ha es allo- ca ions. Al hough he same amoun o oundwood was ha es ed in A BAU and A SB , he amoun o o al ca bon (including all compo- nen s in addi ion o ee s ems) a ec ed was smalle in A SB ;an obse a ion ha is u he examined in he nex sec ion. The com- Fig. 2. The p opo ion o a ea ha was expec ed o ansi as e based on he ansi ion p obabili ies adap ed o expec ed clima e (RCP4.5) compa ed o hose based on pai wise NFI obse a ions. The end lines ep esen he a ea p opo ions ob ained using he median alue o he clima e-induced empe a u e dis ibu ion, while he e ical lines show he a ia ion be ween he 10% and 90% alues o he dis ibu ion. The legend e e s o dominan species in di e en si e e ili y classes – he i h class including all poo ly p oduc i e o es land was omi ed om he igu e. Fig. 3. T ansi ions om he ini ial (Class#, 0) o he subsequen olume class (Class#, 1) in he ini ial pai wise obse a ions (g ey symbols) and hose adap ed o clima e-induced g ow h (black symbols; ansi ions adap ed o RCP4.5 a e shown). The hick ho izon al lines depic he median, he bo om and op o he boxes he in e qua ile ange be ween he 25 h and 75 h pe cen iles, he whiske s he lowes da a wi hin 1.5 imes he in e qua ile ange, and ci cles he da a no included in ca ego ies abo e. J. Vauhkonen, T. Packalen / Ecosys em Se ices 33 (2018) 213–224 217 bined e ec s o clima e and managemen we e mo e p onounced in he case, whe e he ac i i y p obabili ies we e ixed in he begin- ning o simula ions, o which eason he ex below mainly e e s o Fig. 6. In i , a co esponding inc emen in he g owing s ock could be obse ed due o clima e change (abou 8% using he expec ed alues o RCP4.5) as i he ha es alloca ion was changed om A BAU o A SB (7–9% depending on he ha es a ge ). In addi- ion, i bo h he clima e change and he change in ha es alloca- ion occu ed, i inc eased he amoun o ca bon ha es ed o app oxima ely he same le el as wi h clima e change alone by he end o he simula ion, bu wi h a comple ely di e en de elop- men pa e n du ing he simula ion pe iod (see Fig. 6). 3.2.3. Unce ain ies ela ed o he dynamics o indi idual ca bon componen s A dec easing amoun o he o al ca bon ex ac ed could be obse ed based on Fig. 5 (e.g., lowe le panel), e en i he ac i i y p obabili ies we e i e a ed o yield he desi ed ha es goals. This obse a ion and also he conside able di e ences in o al ca bon s ocks and d ains depending on he ha es alloca ion (Fig. 6) migh no seem in ui i e, when conside ed as o al ca bon, bu can be be e easoned when he analyses a e b oken down o he le el o indi idual ca bon componen s (Fig. 7). Al hough he ha es decisions a e based on he amoun o oundwood, he o al ca bon a ec ed by he ha es s includes oliage, b anches, s ump, Fig. 4. The mean annual inc emen o he o al biomass in he NFI plo s as a unc ion o age – e e o Appendix A o he de ini ion o he age classes. The hick line depic s he inc emen in he ini ial pai wise obse a ions and he hin lines he inc emen s o he eigh simula ion pe iods, when he pai wise obse a ions we e adap ed o he RCP2.6 (le ), RCP4.5 (middle), and RCP8.5 ( igh ) clima e scena ios. Fig. 5. The de elopmen o o al ca bon du ing he eigh simula ion s eps, when he ac i i y p obabili ies we e i e a ed be ween he indi idual s eps o ob ain ha es ing goals BAU (le column), NFP (middle column), and NDV ( igh column). The g ey ba s depic he de elopmen based on business-as-usual ha es alloca ion (A BAU ) and ansi ion p obabili ies de i ed om he pai wise NFI obse a ions. The black ‘‘e o ba s” a e based on he same ansi ion p obabili ies, bu schoolbook-alloca ion (A SB ) o he ha es s. The lines abo e he ba s depic he de elopmen , when he ansi ion p obabili ies a e adap ed o he clima e: he ed, black and g een lines e e o RCP2.6, RCP4.5, and RCP8.5 scena ios. The solid o b oken lines e e o he use o A BAU o A SB , espec i ely, and he e a e h ee lines pe clima e scena io and ha es alloca ion, depic ing he 10%, 50% and 90% alues o he expec ed empe a u e dis ibu ion. 218 J. Vauhkonen, T. Packalen / Ecosys em Se ices 33 (2018) 213–224 and oo s o he ees ha es ed. The p opo ion o he di e en componen s a ec ed is ela ed o he di e en alloca ion o he ha es s (Fig. 1): in A SB , mos o he d ain is ob ained om e- gene a ion ha es s, which a e essen ially inal ellings o e en- aged o es s. The ca bon ha es ed in A SB he e o e o igina es om mo e ma u ed o es s, whe e a highe p opo ion o he o al ca - bon is s o ed in he s em wood. Thus, e en i bo h ha es alloca- ions yielded he same le el o oundwood, he le el o o al ca bon ex ac ed a ied due o he di e en p opo ion o ca bon componen s in he o es s subjec o he ha es s. 4. Discussion 4.1. Modeling clima e-induced o es g ow h a ia ions using a ea- based Ma ko chain models To accoun o he e ec s o clima e change, some ea lie simu- la ion s udies ca ied ou in Finland (Nuu inen e al., 2006; Kallio e al., 2013) ha e used he models o Ma ala e al. (2005, 2006) o p edic how inc easing annual mean empe a u e and ambien CO 2 a ec he o es g ow h. When inco po a ing hese e ec s o he o es de elopmen scena ios, he s udies men ioned abo e ha e conside ed ei he immedia e o g adual inc ease in he g ow h, nei he o which is ealis ic acco ding o he g ow h pa - e ns obse ed in he pas (Pasanen, 1998; Hen onen e al., 2017). Al e na i e app oaches conside ing he s ochas ici y o g ow h ha e also been p esen ed (Pukkala and Kellomäki, 2012), bu addi ional conside a ions we e needed o implemen hese e ec s wi h ma ix models ha assume he s a iona i y o he ansi ions. Ou app oach used he RCPs and subsequen clima e p ojec- ions calib a ed o Finland (Ruos eenoja e al., 2016) o de i e p obabili y dis ibu ions o inc ease in CO 2 and annual mean em- pe a u e. The dis ibu ions we e sampled o p o ide CO 2 and em- pe a u e alues o be used as p edic o s o he g ow h e ec (Ma ala e al., 2005, 2006). By means o sampling and adding he ob ained g ow h end wi h s ochas ic a ia ion, we we e able o accoun o he unce ain ies ela ed o he RCP p edic ions. Finally, o be applicable in ou simula ions, he g ow h se ies we e no used as such, bu as classi ied o espec i e olume classes. The classi ica ion s ep undamen ally ‘‘smoo hs” he g ow h se ies, as he addi ional g ow h modeled o e e y plo is no ans e ed o he p ojec ions as such. Ins ead, he a ea ac ion ep esen ed by he plo is di ided acco ding o he p obabili y o he plo o ansi om he o iginal olume class o a highe class. The p oba- bili y depends on bo h he dis ance o he ini ial olume alue o he class limi and he magni ude o he ela i e scena io e ec , which u he depends on ac o s such as si e ype and ee compe- i ion (Ma ala e al., 2005, 2006). In p inciple, his ype o p oba- bilis ic app oach could mode a e he model-based p edic ions o clima e-induced g ow h a es, which may o he wise seem o e ly op imis ic (c ., Pukkala, 2017b). Pukkala (2017b) came o his con- clusion using an al e na i e app oach, which p edic s ee su i al a es in addi ion o s aigh o wa d changes in he clima e- p oduc ion ela ionship due o he ele a ed empe a u e. The mean annual clima e-induced inc emen s o s em olume we e in he o de o 3.5–12.5%, bu as much as 30% du ing he simula- ion pe iod (Pukkala, 2017b). In ou s udy, he mean annual inc e- men o he o al biomass (compu ed as an a e age o e he age classes in Fig. 4) was 6.8% and was expec ed o inc ease o 8.4– 9.4% in he las simula ion pe iod depending on which RCP was assumed. Also he esul s o he de elopmen scena ios, when compa ed in he end o he simula ions, a e undamen ally in line wi h Pukkala (2017b). E en hough he igu es a e o e all di icul o compa e due o di e en biomass componen s and compu a ion Fig. 6. The de elopmen o o al ca bon du ing he eigh simula ion s eps, when he ac i i y p obabili ies we e ixed in he beginning o he simula ion. Re e o he cap ion o Fig. 5 o he in e p e a ion o he image. J. Vauhkonen, T. Packalen / Ecosys em Se ices 33 (2018) 213–224 219 me hods (in Pukkala, 2017b, he igu es we e based on maximizing he ne p esen alue wi h a ying in e es a es), he app oach es ed he e is conside ed p omising and o wa an es s wi h o he applica ions han hose aiming a Ma ko chain models. Adap ing he ansi ion p obabili ies o he changed clima e sligh ly changes he concep o he EFDM, which is o a high deg ee buil upon he Ma ko ian p ope y, whe e he u u e s a e depends only on he p esen s a e, no on e en s ha p eceded i . This p ope y is no a ec ed, bu he e is a equi emen o es i- ma e mul iple ansi ion ma ices and apply each ma ix sepa- a ely wi hin he gi en ime s eps. Concep ually, he changes esul o a ime-inhomogeneous Ma ko chain, which is much less applied o e en s udied han he heo y o homogeneous Ma ko chains. Ma ko chain models wi h andom o condi ional ansi- ions ha e been es ed in o he applica ions han o es y (Shamshad e al., 2005; Meidani and Ghanem, 2013). To da e, Liéna d and S igul (2016) a e appa en ly he only ones o desc ibe ime-inhomogeneous Ma ko chains applied o o es p ojec ions. Thei model was ope a ed a he o es o pa ch le el, which is no di ec ly compa able o ou analyses because o la ge a eas ep esen ed by he ma ix cells and he use o ansi ions condi ional o managemen ac i i ies in ou pape . The e o e, he wo k on modeling he ansi ion p obabili ies desc ibed he e may p o ide an in e es ing con ibu ion as an applica ion o ime- inhomogeneous Ma ko chain models wi h ac i i y-condi ional ansi ions. Fig. 7. The ca bon d ain p esen ed in he lowe ow o Fig. 6 b oken down o indi idual (biomass) componen s. The columns ep esen he ha es ing goals BAU (le ), NFP (middle), and NDV ( igh ). The ba s con ain he p opo ions o s em, oliage, b anches, s ump, and oo s, espec i ely, om bo om o op, and sepa a ely o inal- ellings (dashed ba s) and hinnings. The wo ba s o each simula ion pe iod show a compa ison o A BAU s. A SB wi h ansi ion p obabili ies de i ed om he pai wise NFI obse a ions ( i s ow); o ini ial ansi ion p obabili ies s. hose assuming he ealiza ion o he 90% alue o he expec ed empe a u e dis ibu ion o RCP8.5 unde A BAU and A SB (second and hi d ow, espec i ely). 220 J. Vauhkonen, T. Packalen / Ecosys em Se ices 33 (2018) 213–224 4.2. Limi a ions o ou esul s and u he unce ain ies ela ed o p ojec ing u u e ca bon balance The de elopmen o he o es esou ces was p ojec ed un il a ound 2050. Acco ding o sensi i i y analyses based on his o ical da a o compa isons o scena io p ojec ions (Nabuu s e al., 2000; Vauhkonen and Packalen, 2017), simila ma ix model p ojec ions as applied he e could be used o pe iods o up o 50–60 yea s. E en i he p ojec ions could hus ha e been made o a sligh ly longe pe iod and also he clima e scena ios ex end beyond 2050, he end in many scena ios changes app oxima ely a ound 2050. Ye , nume ical p edic ions o he GHG concen a ions a e p o- ided only o yea s 2050 and 2100 (Meinshausen e al., 2011). Because he p ojec ions we e composed o sequences o 5-yea pe iods, we eel ha excessi e assump ions would ha e been ela ed only o modeling he clima e end beyond 2050, o which eason he p ojec ions we e no ex ended u he . By ‘unce ain ies’ in ou analyses, we essen ially e e o u u e de elopmen ha canno be p edic ed using ansi ion p obabili- ies de i ed om he pai wise obse a ions based on measu e- men s. Ou ocus was pa icula ly in he e ec s o clima e- and managemen -induced addi ional ne g ow h o he ca bon s ocked in abo e- and belowg ound o es biomass. Howe e , bo h he emphases on ne g ow h and o es biomass also p oduce limi a- ions owa ds he in e p e abili y o ou esul s, which is discussed below wi h espec o ea lie li e a u e. 4.2.1. Nega i e impac s o clima e change Clima e wa ming likely a ec s no only o es g ow h, bu also heal h: isks o na u al dis u bances and e en calami ies inc ease. I models o occu ence and deg ee o damages we e a ailable, hose could be used as addi ional ac i i y and ansi ion p obabil- i ies, espec i ely, in ou model, and he e o e also he nega i e impac s could easily be assessed wi h espec o he u u e scena - ios. Howe e , e en hough con inen -speci ic indica ions on he inc ease o bo h bio ic and abio ic damages due o clima e wa m- ing ha e been p esen ed (e.g., Seidl e al., 2017), he e a e no nume ical es ima es a ailable o be used as p obabili ies. Using a simila Ma ko chain app oach han in his s udy, he ela ed e ec s need o modeled indi ec ly unless he damages a e speci i- cally ela ed o he main axes o he ma ices (e.g., olume and age). Al hough his can be done in he EFDM ia ou pu coe i- cien s, hei use may add u he unce ain ies o he p ojec ions. The use o models de eloped by Ma ala e al. (2005, 2006) in he way desc ibed in he p e ious sec ions allows di ec modeling o olume inc ease as a unc ion o CO 2 , empe a u e, and o es - speci ic cha ac e is ics. Excep o hose models, we a e no awa e o any o he clima e-adap i e models o Finland ha could be in eg a ed o p ac ical simula ion sys ems making use o o es da a collec ed o ex ensi e a eas. Liéna d and S igul (2016), who used a Ma ko chain based app oach, ound a di e gence o ±5% be ween he de elopmen scena ios o mean biomass o ha dwood o es s o Quebec, Canada, by he beginning o 2090. Howe e , hey assumed he inc easing CO 2 and empe a u e o a ec mo e on i e a es han g ow h enhancemen s. In Finland, s udies in eg a ing u u e isks in o es p ojec ions ha e conside ed especially s o m- ela ed damages (e.g., Reye e al., 2017). Also pes a acks o pa hogen in ec ions may be expec ed o inc ease, bu hese a e mo e speci ic in e ms o occu ence a eas and species, and may he e o e equi e e y delica e species-speci ic modeling (e.g., Ne alainen e al., 2015). Al ahahleh e al. (2016), using a o es ecosys em model ha p edic s e-gene a ion, g ow h, and mo ali y acco ding o empe a u e sum, ee compe i ion, and soil, ni ogen, and ambien ligh and CO 2 a ailabili y, concluded ha clima e wa m- ing a ec ed he ees in no he n and sou he n Finland indi e - en ly. They elabo a ed hese indings wi h discussion on join clima e wa ming, si e-speci ic wa e holding capaci y, and species-speci ic esponses o hese phenomena. As men ioned abo e, ou analyses did no accoun o ei he bio ic o abio ic damages o dis u bances excep o added g ow h. We acknowl- edge ha including only posi i e e ec s o clima e change may be simplis ic, bu he discussion abo e also sugges s he complex- i y o conside ing all possible clima e- ela ed impac s. 4.2.2. O he componen s o ca bon balance han biomass The analysis p esen ed in Sec ion 3.2.3 explains he a ia ions in ca bon dynamics, when all ca bon (o biomass) componen s a e included as he o al ca bon. Al hough we acknowledge ha he possibili y o u ilize all hese componen s especially om hin- nings can be ques ioned, we ound his analysis bene icial om wo aspec s. Fi s , an idea on he compu a ional unce ain ies in ol ed is p o ided: he accu acy o biomass models and con e - sion ac o s ega ding he di e en componen s may a y (c ., Neumann e al., 2016). Second, i he componen s a e no ex ac ed and used o biomass-based p oduc s, hose p o ide he li e and deb is ha accumula es as dead o ganic ma e and a ec s he soil ca bon. Ou analyses a e no comple e wi h espec o he o al ca bon balance o o es s, as we did no explici ly include he ca bon seques e ed in he o es soil and p oduc s. Rega ding soil ca bon, howe e , mainly he ini ializa ion o he ca bon pools is p oblem- a ic, whe eas simula ing he decomposi ion can be based on exis - ing soil ca bon models (c ., Pukkala, 2014, 2017a; Akujä i e al., 2016). Howe e , all a o emen ioned s udies assume ha he cu - en clima e p e ails and no models simila o hose applied o li - ing biomass in Sec ion 2.3.2 can be ound om he li e a u e o p edic ing i he decomposi ion should be assumed o accele a e o slow down in he wa ming clima e. The dynamics o hese pools could be es ima ed by means o coe icien s (Pukkala, 2014; Heinonen e al., 2017), bu when no clima e-adap i e models o he decomposi ion exis , he changes in he soil ca bon would only be ela ed o he a ying amoun o ha es s. Howe e , he e ec s o ha es ing o hese s ocks can, o a ce ain deg ee, be deduced om he ea lie s udies (Pukkala, 2014, 2017a; Zubiza e a- Ge endiain e al., 2016). In he simula ions o he de elopmen o biomass, soil, and p oduc pools unde ou managemen scena ios (Pukkala, 2017a), he soil ca bon a ied much less han ca bon s ocked in biomass and p oduc s. In hose simula ions, a as decomposi ion o all pools was s a ed a e a ha es , esul ing o a nega i e o al ca bon budge in he sho e m ( h ee o i e decades), bu a posi i e budge in he longe e m due o he seques a ion in he g owing ees and ha es ed p oduc s. Thus, acco ding o Pukkala (2017a), he conclusions depend on he ime ho izon and also on how much weigh is se o he subs i u ion e ec s, i.e., he educ ion o consump ion o ossil uels due o wood-based p oduc s (see also Pukkala, 2014). The examples abo e illus a e he complexi y and assump ions equi ed o model he ca bon balance beyond he li ing biomass s ocks, in changing clima e and in he na ional scale. Due o ocusing on he ca bon s ocked in abo e- and below- g ound biomass, ou s udy canno be di ec ly compa ed o hose epo ed ea lie . O e all, he mul i ude o s udies and di e en app oaches indica es he challenges in he ela ed modeling ask. Ea lie ca bon balance s udies o p o ide ins uc ions o o es managemen (Pukkala, 2014, 2017a; Zubiza e a-Ge endiain e al., 2016) we e ocused on single o es s ands o small o es p ope ies (up o a ound 1000 ha). Heinonen e al. (2017) used NFI11 da a and conside ed he ca bon balance o en i e Finland, bu assumed di e en ha es alloca ion and no clima e change. The s udy by Al ahahleh e al. (2016) elied on se e al assump ions behind he o es ecosys em model used. Compa ed o ha , he J. Vauhkonen, T. Packalen / Ecosys em Se ices 33 (2018) 213–224 221