Underestimation of boreal soil carbon stocks by mathematical soil carbon models linked to soil nutrient status

Biogeosciences, 13, 4439–4459, 2016 www.biogeosciences.ne /13/4439/2016/ doi:10.5194/bg-13-4439-2016 © Au ho (s) 2016. CC A ibu ion 3.0 License. Unde es ima ion o bo eal soil ca bon s ocks by ma hema ical soil ca bon models linked o soil nu ien s a us Bo is ˇ Tupek1, Ca ina A. O iz2, Shoji Hashimo o3, Johan S endahl2, Jonas Dahlg en4, E ik Ka l un2, and Aleksi Leh onen1 1Na u al Resou ces Ins i u e Finland, P.O. Box 18, 01301 Van aa, Finland 2Swedish Uni e si y o Ag icul u al Sciences, P.O. Box 7014, 75007 Uppsala, Sweden 3Fo es y and Fo es P oduc s Resea ch Ins i u e, Tsukuba, Iba aki 305-8687, Japan 4Swedish Uni e si y o Ag icul u al Sciences, Skogsma ksg änd, 90183 Umeå, Sweden Co espondence o: Bo is ˇ Tupek ([email p o ec ed]), Aleksi Leh onen (aleksi.leh onen@luke. i) Recei ed: 21 Decembe 2015 – Published in Biogeosciences Discuss.: 18 Janua y 2016 Re ised: 6 July 2016 – Accep ed: 7 July 2016 – Published: 10 Augus 2016 Abs ac . Inaccu a e es ima e o he la ges e es ial ca bon pool, soil o ganic ca bon (SOC) s ock, is he majo sou ce o unce ain y in simula ing eedback o clima e wa ming on ecosys em–a mosphe e ca bon dioxide exchange by p ocess- based ecosys em and soil ca bon models. Al hough he mod- els need o simpli y complex en i onmen al p ocesses o soil ca bon seques a ion, in a la ge mosaic o en i onmen s a missing key d i e could lead o a modeling bias in p edic- ions o SOC s ock change. We aimed o e alua e SOC s ock es ima es o p ocess- based models (Yasso07, Q, and CENTURY soil sub-model 4) agains a massi e Swedish o es soil in en o y da a se (3230 samples) o ganized by a ecu si e pa i ioning me hod in o dis inc soil g oups wi h unde lying SOC s ock de elop- men linked o physicochemical condi ions. Fo wo- hi ds o measu emen s all models p edic ed ac- cu a e SOC s ock le els ega dless o he de ail o inpu da a, e.g., whe he hey igno ed o included soil p ope ies. How- e e , in e ile si es wi h high N deposi ion, high ca ion ex- change capaci y, o mode a ely inc eased soil wa e con en , Yasso07 and Q models unde es ima ed SOC s ocks. In com- pa ison o Yasso07 and Q, accoun ing o he si e-speci ic soil cha ac e is ics (e. g. clay con en and opsoil mine al N) by CENTURY imp o ed SOC s ock es ima es o si es wi h high clay con en , bu no o si es wi h high N deposi ion. Ou analysis sugges ed ha he soils wi h poo ly p edic ed SOC s ocks, as cha ac e ized by he high nu ien s a us and well-so ed pa en ma e ial, indeed ha e had o he p edomi- nan d i e s o SOC s abiliza ion lacking in he models, p e- sumably he myco hizal o ganic up ake and o gano-mine al s abiliza ion p ocesses. Ou esul s imply ha he ole o soil nu ien s a us as egula o o o ganic ma e mine aliza ion has o be e-e alua ed, since co ec SOC s ocks a e decisi e o p edic ing u u e SOC change and soil CO2e lux. 1 In oduc ion In spi e o he his o ical ne ca bon sink o bo eal soils, 500Pg o ca bon since he las ice age (Rapalee e al., 1998; DeLuca and Bois enue, 2012; Scha lemann e al., 2014), bo eal soils could become a ne sou ce o ca bon dioxide o he a mosphe e as a esul o long- e m clima e wa m- ing (Ki schbaum, 2000; Amundson, 2001). They ha e he po en ial o elease la ge quan i ies o ca bon han all an- h opogenic ca bon emissions combined (337Pg; Boden e al., 2010). In o de o p ese e he soil ca bon pool and o u ilize he soil ca bon seques a ion po en ial o mi iga e an h opogenic CO2emissions, mi iga ion s a egies o cli- ma e o cing aim o imp o e soil o ganic ma e managemen (Schlesinge , 1999; Smi h, 2005; Wiesmeie e al., 2014). Suppo ing soil managemen decisions equi es an ac- cu a e quan i ica ion o spa ially a iable soil o ganic ca - bon (SOC) s ock and SOC s ock changes (Scha lemann e al., 2014). The ini ial le el o SOC s ock is essen ial in o de o es ima e SOC s ock changes (Palosuo e al., 2012; Todd-B own e al., 2014), especially when es ima ing ca bon emissions due o land-use change, e.g., a o es a ion o g ass- Published by Cope nicus Publica ions on behal o he Eu opean Geosciences Union. 4440 B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y lands (Be h ong e al., 2009). P ocess-o ien ed soil ca bon models like CENTURY, Ro h-C, Biome-BCG, ORCHIDEE, JSBACH, ROMUL, Yasso07, and Q a e impo an ools o p edic ing SOC s ock change, bu he e a e also isks o poo p edic ions (Todd-B own e al., 2013; DeLuca and Bois - enue, 2012). The models need u he alida ion and im- p o emen as hey show poo spa ial ag eemen on ine scale and mode a e ag eemen on egional scale agains SOC s ock da a (Todd-B own e al., 2013; O iz e al., 2013). Despi e he po en ially quan i a i e impo ance o CO2emissions he ex- pec ed change will be small in ela ion o he SOC s ock. The e o e, he unce ain y o measu emen s and/o model es ima es could p e en conclusions on SOC s ock changes (Palosuo e al., 2012; O iz e al., 2013; Leh onen and Heikki- nen, 2015) especially o he soils wi h he la ges SOC s ocks, which a e he mos sensi i e o ca bon loss. Beside la ge unce ain ies, he poo ag eemen be ween he modeled and measu ed SOC s ocks (Todd-B own e al., 2013) could also indica e missing bio ic o abio ic d i e s o long- e m ca bon s o age (Schmid e al., 2011; A e ill e al., 2014). Fo example, igno ing he essen ial ole o soil nu ien a ailabili y in ecosys em ca bon use e iciency (Fe nández- Ma ínez e al., 2014) could lead o missing impo an con- ols o plan li e p oduc ion and soil o ganic ma e s abi- liza ion mechanisms. Soil nu ien s a us is linked o he mo- bili y o nu ien s in he wa e solu ion (Husson e al., 2013), p oduc ion, quali y and mic obial decomposi ion o plan li - e (O win e al., 2011), and o ma ion o he soil o ganic ma e (SOM). The SOM a ec s soil nu ien s a us by e- cycling o mac onu ien s (Husson e al., 2013), and wa e e en ion and wa e a ailabili y (Rawls e al., 2003). In spi e o s a e o he a soil ca bon modeling based on he amoun and quali y o plan li e “ ecalci ance”, a ec ed by clima e and/o soil p ope ies as in he Yasso07, Q, and CENTURY models, hese ypes o p ocess-based models do no include mechanisms o SOM s abiliza ion by (a) he o - ganic nu ien up ake by myco hizal ungi; (b) humic o - ganic ca bon in e ac ions wi h sil -clay mine als; and (c) he inaccessibili y o deep soil ca bon and ca bon in soil agg e- ga es o soil bio a (O win e al., 211; Sollins e al., 1996; To n e al., 1997; Six e al., 2002; Fan e al., 2008; Dun- gai e al., 2012; Clemen e e al., 2011). Al hough he models do no con ain a o emen ioned mechanisms and con ols o changes in SOM s abiliza ion p ocesses, hey ha e been pa- ame e ized using a wide a ie y o da a se s and can ea soil bio ic, physicochemical, and en i onmen al changes im- plici ly. The Yasso07 model (Tuomi e al., 2009, 2011) is an ad anced o es soil ca bon model and i is used o Kyo o p o ocol epo ing o changes in soil ca bon amoun s o he Uni ed Na ions F amewo k Con en ion on Clima e Change (UNFCCC) by Eu opean coun ies, e.g., Aus ia, Finland, No way, and Swi ze land. The Q model (Åg en e al., 2007) is a mechanis ic li e decomposi ion model de eloped in Sweden and used, e.g., o compa e esul s p oduced wi h Swedish na ional in en o y da a (S endahl e al., 2010; O iz e al., 2011) and also wi h o he models a na ional o global scales (O iz e al., 2013; Yu o a e al., 2010). The CEN- TURY model (Pa on e al., 1987, 1994; Adai e al., 2008) is one o he mos widely applied models and i is used o soil ca bon epo ing o he UNFCCC by Canada, Japan, and USA. Al hough indi idual pa ame e s and unc ions a y, ma hema ical models such as Yasso07, Q, and CENTURY ha e simila s uc u es. Fo example, hese models a e d i en by he decomposi ion a es o li e inpu and SOM. Decom- posing li e and SOM is di ided in o pools based on li e quali y, and i s ans e om one pool o ano he is, apa om model unc ions and pa ame e s, a ec ed by empe - a u e (Q), and/o wa e (Yasso07), and/o soil ex u e and s uc u e (CENTURY). The Q model does no include ex- plici mois u e unc ions, whe eas p ecipi a ion a ec s de- composi ion o he Yasso07 and CENTURY models (Tuomi e al., 2009; Adai e al., 2008). On he o he hand, he mod- els do no explici ly o by de aul include mechanisms ha educe decomposi ion by excessi e p ecipi a ion/mois u e (Falloon e al., 2011). We hypo hesized ha (1) soil ca bon es ima es o he Yasso07, Q, and CENTURY models would de ia e o soils whe e SOC s abiliza ion p ocesses no implici ly accoun ed by he models a e p edominan , (2) he Yasso07 and Q mod- els igno ing soil p ope ies would ail on he nu ien - ich si es o he sou hwes e n coas o Sweden and on occasion- ally paludi ied clay and sil soils, and (3) he CENTURY model ou pe o ms he Yasso07 and Q models due o ac ha i includes soil p ope ies as inpu a iables. We g ouped Swedish o es soil in en o y da a in o ho- mogenous g oups wi h speci ic soil physicochemical con- di ions using a eg ession ee and ecu si e pa i ioning modeling me hods. A e ha we an he models un il hey eached an equilib ium wi h a li e inpu ha was de i ed om he Swedish o es in en o y. The ea e , we compa ed he model es ima es agains da a by g oups ha we e ob- ained om he eg ession ee model. In discussion we ad- d ess he easons why he models de ia e and indica e di ec- ions o u he imp o emen s. 2 Ma e ial and me hods 2.1 Measu emen s We analyzed da a om he Swedish o es soil in en o y (SFSI), which is a s a i ied na ional g id su ey o ege- a ion and physicochemical p ope ies o soils (SLU, 2011; Olsson e al., 2009). The soil da a dis inguished be ween he o ganic, B (0–5cm o B ho izon), BC (45–55cm below g ound su ace), and C (55–65cm om he op o he mine al soil) ho izons (Olsson e al., 2009). All analysis was done using R so wa e o s a is ical compu ing and g aphics (R Co e Team, 2014). The soil da a we e iden ical o a da a se used in S endahl e al. (2010). We es ic ed ou sample plo s Biogeosciences, 13, 4439–4459, 2016 www.biogeosciences.ne /13/4439/2016/ B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y 4441 Table 1. Desc ip ion o he Swedish Fo es Soil In en o y (SFSI) da a educ ion o soil so ing o pa en ma e ial and humus ypes; SFSI con e sion es ima e o soil classes o soil mois u e o nume ical ep esen a ion o soil wa e con en ; and SFSI con e sion es ima e o classes o nume ical ep esen a ion o soil ex u e (sand, sil , and clay con en o sedimen s by Lindén (2002) and o ills by Albe A e be g’s dis ibu ion o he di e en g ain size ac ions). So ing pa en ma e ial Humus ype Mois u e SFSI Reduced SFSI Reduced SFSI SFSI Nume ic Bed ock Bed ock Mode No-pea Wa e Long- e m Poo ly so ed sedimen s Unso ed Mo 1 No-pea le el (m) mois u e % Tills Unso ed Mo 2 No-pea D y <2 10 Well-so ed sedimen s So ed Mull No-pea F esh 1–2 20 Mull-Mode Pea F esh-mois <1 30 Pea Pea Mois <0.5 50 Pea -Mo Pea Tex u e SFSI Nume ic Sedimen s Tills Sand % Sil % Clay % Sand % Sil % Clay % Bed ock 0 0 0 0 0 0 Boulde 0 0 0 0 0 0 G a el 10 0 0 10 0 0 Coa se sand 40 5 0 40 5 0 Sand 80 10 0 45 10 0 Fine sand 70 25 5 55 15 0 Coa se sil 50 40 10 65 20 5 Fine sil 10 75 15 55 35 10 Clay 0 65 35 0 85 15 Pea 0 0 0 0 0 0 o mine ogenic soils since he Q, Yasso07, and CENTURY models we e no de eloped o use on pea soils, and only o plo s o o es land use wi h Swedish o es in en o y da a (SFI). We also excluded samples wi h o al SOC s ock be- low 2.8 and abo e 470.5 ( Cha−1), i.e., samples wi h SOC s ock below 0.01 and abo e 99.9 pe cen ile. Measu emen da a o igina ed om 1993 o 2002, which cons i u e a ull in en o y, and om 2020 sample plo s loca ed a ound Swe- den, and in o al i included 3230 samples. Fo each sample plo he wea he (yea s 1961–2011) and N deposi ion (yea s 1999–2001) da a we e e ie ed om he nea es s a ions o Swedish Me eo ological and Hyd ological Ins i u e (SMHI) ne wo k (Fig. 1). The plo s, which we e linked by he closes dis ance o he gi en wea he s a ion had he same wea he and N deposi ion da a, and he numbe o soil samples pe s a ion anged be ween 10 and 70. The mean o al SOC s ock o samples co esponding o wea he s a ions anged om 40 o 200 ( Cha−1), and he SOC s ock le el dec eased om sou he n o no he n Sweden (Fig. 1). Each sample plo con ained ca ego ical da a om he ield su ey on he so ing o soil pa en ma e ial, humus ype, soil ex u e, and soil mois u e. In ou analysis we educed ca ego ical classes by basing hem on he so ing o soil pa - en ma e ial and humus ype (Table 1). We de e mined nu- me ic alues o sil , clay, and sand con en om soil ex u e ca ego ies by Albe A e be g’s dis ibu ion o he di e en g ain size ac ions in ills and dis ibu ions o sedimen s by Lindén (2002) (Table 1). We also de e mined nume ic al- ues o olume ic soil wa e con en (SWC) om ca ego ical ield da a classi ied acco ding o he dep h o he g ound wa- e le el (WL; Table 1). As is ypical o soil ca bon in en o ies, he a ia ion o da a was la ge (Table 2). Fo example, he mean o al SOC s ock o all samples was 93 ( Cha−1) while 1s and 99 h pe - cen iles we e 17 and 309 (Table 2). The mean SOC s ock was 33.3 and 66.8 ( Cha−1) o he humus ho izon and he mine al soil. The mean alues o ca ion exchange capaci y (CEC) (23.9mmolckg−1), he base sa u a ion (36.4%), and he C/N a io (16.5) indica ed condi ions o medium e il- i y, al hough he soils we e mos ly acidic (mean pH was 5.2). The mean p e ailing soil wa e con en (22.3) was ypical o he well-d ained o es soils. The mean annual empe a u es anged om below 0 o abo e 8◦C, and annual p ecipi a- ion a ied be ween 392 and 1154mm (Table 2). To al SOC s ock o all he samples gene ally inc eased o pea and pea like humus o ms, o well-so ed sedimen s, o soils wi h high ac ion o sil and clay and wi h inc easing soil mois- u e (Fig. S1 in he Supplemen ). www.biogeosciences.ne /13/4439/2016/ Biogeosciences, 13, 4439–4459, 2016 4442 B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y Table 2. Desc ip i e cha ac e is ics (mean, con idence in e al, 1s , 50 h, and 99 h pe cen ile) o selec ed a iables (n=3230 samples). The alues o he bulk densi y, ca ion exchange capaci y, base sa u a ion, C /N a io, and pH a e shown only o BC soil ho izon ( ixed 45–50cm dep h below he g ound su ace) due o he s ong co ela ion o he o al soil ca bon s ock. The soil was cu o a 1 m. The si e p oduc i i y index (H100, m) is an app oxima ion o he si e e ili y exp essed as he heigh o ees a 100 yea s o age. S and and unde s o y biomass, and li e inpu a e modeled alues o app oxima ed equilib ium condi ions based on obse a ions. Mean CI 1s pe cen ile 50 h pe cen ile 99 h pe cen ile To al soil ca bon s ock ( Cha−1) 93.24 1.95 17.02 79.68 308.68 Humus ca bon s ock ( Cha−1) 33.29 1.17 3.89 22.82 176.66 Mine al soil ca bon s ock ( Cha−1) 66.82 1.7 6.92 54.81 273.91 Dep h o humus (cm) 10.52 0.27 1 8 36 Dep h o soil (cm) 93.37 0.6 18 99 99 S oniness (%) 39.91 0.54 3.96 42.37 65.05 Bulk densi y o BC (gdm−3) 1267.1 5.5 790.55 1294.9 1522.13 Ca ion exchange capaci y o BC (mmolckg−1) 23.94 1.28 1.53 12.33 203.25 Base sa u a ion o BC (%) 36.44 1.02 4.33 25.73 100 C/N a io o BC 16.5 0.35 3.33 14.98 62.45 pH o BC 5.17 0.02 4.36 5.08 7.26 Sil con en (%) 19.98 0.57 0 15 85 Clay con en (%) 3.16 0.25 0 0 35 Sand con en (%) 51.25 0.63 0 55 80 Long- e m soil mois u e (%) 22.36 0.2 10 20 30 Mean ai empe a u e (◦C) 4.63 0.09 −0.44 5.34 8.47 To al p ecipi a ion (mm) 697.87 7.13 392.54 637.11 1154.55 Ni ogen deposi ion (kgNha−1y−1) 7.17 0.14 2.35 6.56 17.67 P oduc i i y class (H100, m) 23.61 0.21 12 23 36 To al s and biomass ( Cha−1) 56.02 1.39 1.34 51.14 156.52 To al unde s o y biomass ( Cha−1) 2.69 0.05 0.96 2.37 6.02 To al li e all inpu ( Cha−1) 3.17 0.03 1.65 3.07 5.28 2.1.1 Biomass and li e all es ima es Fo he biomass and li e all es ima ion we adop ed a s an- da d me hod o na ional g eenhouse gas in en o ies o es i- ma ing soil ca bon s ock changes (S a is ics Finland, 2013). In o de o model SOC s ocks o o es in equilib ium (no SOC s ocks changes), we modi ied he me hod by es ima ing he long- e m li e all o o es in equilib ium. Fo es s and biomass was es ima ed by allome ic biomass unc ions o s em wi h ba k, b anch, oliage, s ump, coa se oo s and ine oo s applied o basic ee dimensions (b eas heigh diame- e , o al heigh o ee, numbe o ees) o SFI s ands (Ma k- lund, 1988; Pe e sson and S åhl, 2006; Repola, 2008; Leh o- nen e al., 2016a). In o de o simula e “equilib ium” soil ca - bon s ock, we es ima ed long- e m mean o es biomass, e- e ed o as “equilib ium o es ” below. We adop ed an obse ed ac ion o pho osyn he ically ac- i e abso bed adia ion ( APAR; Fig. A1 in Appendix A) as a ela i e indica o o a si e’s capaci y o p oduce biomass (minimum is 0, maximum is 1) by accoun ing o he o es s and s uc u e, anging om he absen s and APAR =0 o he closed canopy s and APAR =1, h ough i s majo ole on limi ing o he po en ial g oss p ima y p oduc ion (Pel- oniemi e al., 2015). The APAR was calcula ed based on SFI measu emen s o basic ee dimensions as in Hä könen e al. (2010) and o he main ee species (pine, sp uce, de- ciduous) i was well co ela ed wi h he s and basal a ea (Ap- pendix A). The equilib ium o es APAR alues we e assumed o be in a ange be ween he median and he maximum ac ion o he obse ed s a e o es APAR o a gi en species, la i u- dinal deg ee, and si e p oduc i i y index (Appendix A). We selec ed equilib ium APAR as he 70 h pe cen ile ( APAR70) ou o a ange om he 50 h o 95 h, because he modeled soil ca bon dis ibu ions wi h a li e inpu om he APAR70 biomass ag eed bes wi h he measu ed soil ca bon dis i- bu ions (Fig. S2). The APAR70 was he es ima ed 70 h pe - cen ile o he obse ed ac ion o abso bed adia ion speci ic o a gi en species, la i udinal deg ee, and si e p oduc i i y index H100 (heigh o ees a 100 yea s o age; m; Fig. B1 in Appendix B). The si e index H100, ha can be ansla ed o a speci ic p oduc i i y (m3ha−1y −1), was o Swedish o - es in en o y plo s de e mined based on heigh de elopmen cu es and obse ed si e p ope ies by using he me hodol- ogy o Hagglund and Lundma k (1977) (Swedish S a is ical Yea book o Fo es y, 2014). Ins ead o modeling o equi- lib ium biomasses o e e y ee s and componen sepa a ely o he species, la i ude, and si e p oduc i i y index, we sim- pli ied he biomass modeling i s by es ima ing only equilib- Biogeosciences, 13, 4439–4459, 2016 www.biogeosciences.ne /13/4439/2016/ B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y 4443 10 15 20 25 56 58 60 62 64 66 68 68 ● 34 and numbe o Me eo.s a ion nea es soil samples ( ha−1) Soil ca bon La i ude (°) Longi ude (°) 50 100 150 Figu e 1. Geog aphical loca ions o me eo ological s a ions wi h co esponding numbe o nea es soil samples (n, size o he ci cle) and hei mean measu ed soil o ganic ca bon s ock ( Cha−1, colo o he ci cle) ac oss Sweden. ium o es s and s uc u e o he species, la i ude, and p o- duc i i y ( APAR70, Table A1 in Appendix A) and secondly by using APAR70 wi h APAR biomass models (Table B1 in Appendix B) o es ima e he biomass componen s. We modeled he equilib ium biomass by applying he i - ed exponen ial unc ions be ween he obse ed s a e o es biomass componen s (s em, b anch, oliage, s ump, coa se oo s, ine oo s, es ima ed by ee s and measu emen s and he allome ic biomass unc ions) and he obse ed ac ion o abso bed adia ion ( APAR; Appendix B) o he es ima ed APAR70 o he equilib ium o es . The unde s o y ege a- ion o he equilib ium o es was es ima ed by applying ou g ound ege a ion models (Appendix C) o he modeled equi- lib ium o es cha ac e is ics, and plo -speci ic en i onmen- al condi ions. In o de o de i e he li e inpu s, he annual u no e a e (TR), he ac ion o li ing biomass ha is shed on o he g ound pe yea o biomass componen s, was applied o he modeled biomass componen s o he equilib ium o - es . The needle li e TR was a linea unc ion o la i ude o pine and sp uce and a cons an o deciduous species (Åg en e al., 2007). The TR o b anches and oo s we e om Mukkonen and Leh onen (2004) and Leh onen e al. (2004) and he TR o s ump and s em we e om Vi o (1955), Mälkö- nen (1974, 1977) and Liski e al. (2006). Fo ee ine oo s, we assumed he e was a di e ence be ween ee species and be ween sou he n and no he n Sweden. Fo pine, sp uce, and bi ch he TR ine oo s we e 0.811, 0.868, and 1.0, e- spec i ely, as epo ed by Maidi (2001), Ku z e al. (1996), and Liski e al. (2006). Kleja e al. (2008) and Leppälampi- Kujansuu e al. (2014) epo ed di e en ine- oo TR o sou he n (1 and 0.83) and no he n Finland (0.5). We in e po- la ed TR acco ding o he mean annual empe a u e g adien be ween TR o ine oo s in he sou h and he no h. The ine- oo s TR o 0.811, 0.868, and 1.0 in he wa mes sou he n- mos soil plo s we e hus educed down o 0.5 in he coldes no he nmos soil plo s. The unde s o y TR was applied as in Leh onen e al. (2016b). The majo pa o he li e inpu o igina ed om he ee s and biomass componen s, which we e modeled by he non- linea unc ions wi h R2 alues close o 0.9 (Fig. B1, Ta- bles A1 in Appendix A, and B1 in Appendix B). The linea unde s o y ege a ion models had low R2 alues (Table C1 in Appendix C). Howe e , when he unde s o y models (Ap- pendix C) we e applied only o plo s close o equilib ium o - es , as in ou applica ion, he R2 alues o p edic ed and ob- se ed unde s o y componen s we e la ge (Fig. S9). In com- pa ison o majo unde s o y li e all o igina ing om ea- sonably well-p edic ed dwa sh ubs and mosses (Figs. S9 and S10), he in luence o poo e unde s o y models ( o he bs, g ass, and lichens) was small on p edic ions o he un- de s o y li e and ma ginal on p edic ions o he o al o es li e all (Fig. S10). The main imp o emen on he accu acy o o al li e inpu was achie ed by a oiding he con ounding e ec o managemen on obse ed o es s a e by modeling he biomass/li e all es ima es ep esen ing he mean long- e m condi ions (de ined by es ima ed equilib ium APAR70) o small egions (de ined by deg ee o la i ude and p oduc- i i y index o dominan species; Fig. A1 in Appendix A). Thus he es ima es accu a ely e lec ed he long- e m spa ial a iabili y in dominan species, nu ien s a us and clima e (Fig. S11) and lacked highe spa ial and empo al p ecision; as a emp s o high p ecision o he es ima es applied o he pe iod o he las ew housand yea s would be unce ain due o high a ia ion o ac o s a ec ing plo his o y. 2.1.2 Co ela ion analysis O e all ou da a consis o 3230 soil samples and hei ca - bon s ocks linked o soil physicochemical a iables, s and and g ound ege a ion biomass and li e all componen s, and nea es wea he s a ion en i onmen al a iables. We pe - o med he Spea man’s ank co ela ion analysis be ween he o al soil ca bon s ock and he o he soil a iables, si e, cli- ma e, and ege a ion cha ac e is ics. As expec ed he o al soil ca bon s ock mos s ongly co ela ed wi h he measu ed www.biogeosciences.ne /13/4439/2016/ Biogeosciences, 13, 4439–4459, 2016 4444 B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y CEC.BC < 17 CEC.BC < 7.9 pa en .ma e ial: bed ock,unso d CN.BC < 16 pa en .ma e ial: unso d CEC.BC < 33 N.deposi ion < 9.9 Bound.H2O.C < 2.4 Humus: no.pea >= 17 >= 7.9 So ed >= 16 So ed >= 33 >= 9.9 >= 2.4 Pea 65 n=959 30 % 82 n=909 28 % 130 n=136 4 % 86 n=335 10 % 126 n=182 6 % 104 n=296 9 % 137 n=180 6 % 269 n=8 0 % 144 n=142 4 % 203 n=83 3 % Measu ed soil ca bon s ock ( C ha−1) (a) Ca bon low medium high medium high medium high ex a high ex a Mois u e d y− esh esh mois − esh esh mois − esh esh esh esh esh mois − esh Fe ili y low medium medium medium high low high high medium medium Soil g oup 1 2 3 4 5 6 7 8 9 10 (b) n=959 n=909 n=136 n=335 n=182 n=296 n=180 n=8 n=142 n=83 Figu e 2. (a) Classi ica ion/ eg ession ee o he measu ed soil ca bon s ock ( Cha−1), soil physicochemical p ope ies, and si e en i- onmen al cha ac e is ics; he ca ion exchange capaci y o BC ho izon (CEC.BC, (mmolckg−1)), he C/N a io (CN.BC), he ni ogen deposi ion (N.deposi ion kgNha−1y−1), he highly bound soil wa e o C ho izon (bound.H2O.C, %), and soil class a iables as ype o so ed o unso ed soil pa en ma e ial and humus ype. No e ha a iables used o calcula e he soil ca bon s ock (bulk densi y, ca bon con en , dep h, and s oniness) we e excluded om he eg ession ee analysis. The alues in he lea es o he ee show o he dis inc en- i onmen al condi ions mean soil ca bon s ock ( Cha−1), numbe and pe cen age o samples. (b) The in e p e a ion o 10 physicochemical soil g oups o he eg ession ee model in o he le els o ca bon, soil mois u e, and e ili y oughly inc easing om le o igh . a iables used o i s calcula ion, e.g., bulk densi y, dep h o humus and mine al soil, ca bon con en , and s oniness. These a iables we e excluded om u he eg ession ee analysis, which aimed o g oup da a acco ding o he p ocesses o soil ca bon s ock de elopmen . 2.1.3 Reg ession ees In o de o o ganize SOC da a in o g oups acco ding o he physicochemical soil a iables and o be e unde s and he na u e o measu ed da a, we gene a ed eg ession ees o SOC s ocks by using ecu si e pa i ioning (RPART; Th- e neau and A kinson, 1997). RPART is based on de eloping decision ules o p edic ing and c oss- alida ion o con inu- ous ou pu o soil ca bon s ocks ( eg ession ee). The clas- si ica ion ee was buil by inding a single a iable, which bes spli s he da a in o wo g oups. Each sub-g oup was e- cu si ely sepa a ed un il no imp o emen could be made o he soil ca bon s ock es ima ed by using he spli -based e- g ession model. The complex esul an eg ession ee model was c oss- alida ed o a nes ed se o sub- ees by compu - ing he es ima e o soil ca bon s ock o im back he ull ee. When building he eg ession ee models, we excluded a iables such as bulk densi y, ca bon con en s o soil lay- e s, soil dep h, and s oniness, since hese measu ed a iables we e used o de e mining he o al soil ca bon s ock. The selec ed a iables o he RPART da a mining we e based on he co ela ions analysis (see Sec . 2.1.2), he p ocesses o soil o ganic ma e o ma ion (e.g., Husson e al., 2013) and decomposi ion, and ep esen ed he soil ca ego ical a i- ables (so ing o pa en ma e ial, soil ex u e, long- e m soil mois u e, and humus o m), soil physicochemical a iables Biogeosciences, 13, 4439–4459, 2016 www.biogeosciences.ne /13/4439/2016/ B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y 4445 (sand, clay, and sil con en , long- e m soil mois u e, highly bound wa e , C /N a io, pH, CEC o o ganic, B, BC, and C ho izons), clima ic a iables (annual mean ai empe a u e, annual p ecipi a ion sum), and s and and si e cha ac e is ics ( ee species co e age o pine, sp uce and deciduous, o al olia li e inpu , p oduc i i y class and N deposi ion). Al e - na i ely, we also an eg ession and classi ica ion analysis by excluding all measu ed soil a iables because soil a iables a e o en una ailable o landscape le el modeling. The eg ession ee model sepa a ed he measu ed o- al SOC s ocks ( Cha−1) in o 10 g oups. The ca ion ex- change capaci y o he BC ho izon (CEC, mmolckg−1) di- ided all he samples in o wo- hi ds o lowe SOC s ock g oups (means be ween 65 and 130 Cha−1) and one- hi d o la ge g oups (means be ween 86 and 269 Cha−1; Fig. 2a). The g oup o he smalles SOC s ock consis ed o 959 sam- ples compa ed o eigh samples o he g oup wi h he la ges SOC s ocks. We acknowledge ha his is a small dis inc g oup based only on eigh obse a ions. Howe e , we did no ha e any easons o exclude hese da a poin s as ou lie s. These obse a ions indica ed highly e ile condi ions (high N deposi ion, he la ges H100 among g oups (31m), second la ges li e inpu , he highes empe a u e and p ecipi a ion on well-d ained soil) (Fig. 2, Table S1 in he Supplemen ). Two- hi ds o samples wi h smalle SOC s ocks we e sub- di ided by CEC and he ype so ing o soil pa en ma e ial (so ed o unso ed). One- hi d o samples wi h la ge SOC s ocks was subdi ided by he C /N a io, CEC, N deposi ion among o he s. Roughly gene alized, g oups om le o igh o om 1 o 10 o med a g adien in le els o SOC s ock, mois u e, nu ien s a us, and p oduc ion (Fig. 2, Table S1). The al e na i e eg ession ee model was buil wi h a i- ables o he han soil p ope ies. The eg ession ee wi h he annual mean ai empe a u e, he annual p ecipi a ion sum and he pe cen age o pine ees in he s and, and he ni- ogen deposi ion sepa a ed measu ed SOC s ocks ( Cha−1) in o i e g oups (Fig. S3). Colde g oups wi h smalle SOC s ocks (wi h means 67 and 85) had less li e inpu (below 3 Cha−1) and a low si e p oduc i i y index (H100 <20 m; Table S2). 2.2 Soil ca bon s ock modeling The Q model (Rol and Åg en, 1999) is a con inuous mech- anis ic li e decomposi ion model desc ibing change o soil o ganic ma e o e ime. The decomposi ion a e o he b anch, s em, needle, ine oo , and woody li e ac ions is con olled by he empe a u e, li e quali y, mic obial g ow h, and li e in asion a e. The model has been cali- b a ed o se en clima ic egions o Sweden in o de o ac- coun o Swedish empe a u e and p ecipi a ion g adien s (O iz e al., 2011; Table 3). The Q model was applied in se e al s udies o SOC s ock and change es ima ion in Swe- den (e.g., S endahl e al., 2010; O iz e al., 2013; Åg en e al., 2007). The Q model was un o se en Swedish cli- ma ic egions (O iz e al., 2011). The mean egional pa ame- e iza ion om he calib a ion o he 2011 Q model was used o he plo simula ions. Thus, he simula ions in each e- gion ep esen a ia ions in clima e and li e inpu and no pa ame e a ia ions. The equilib ium soil ca bon s ocks a e es ima ed in he model using he equa ion o equilib ium soil ca bon s ock, which is de i ed om he decomposi ion unc ions wi h cons an amoun s and quali y o li e inpu . The Yasso07 model (Tuomi e al., 2009, 2011) is one o he mos widely applied SOC models. The model was calib a ed based on almos 10 000 measu emen s o li e decomposi- ion om Eu ope, No h and Sou h Ame ica (Table 3). The equi ed annual inpu s o li e all, i s size and chemical com- posi ion, empe a u e, and p ecipi a ion de e mine he de- composi ion and seques a ion a es o soil o ganic ma e . Yasso07 es ima es SOC s ock o a dep h o 1 m (o ganic and mine al laye s), change o SOC s ock, and he e o ophic soil espi a ion. Species-speci ic chemical composi ion o di e - en li e compa men s o Yasso07 we e used acco ding o Liski e al. (2009). The ini ial soil o ganic ma e o Yasso07 was ze o. The simula ed soil ca bon s ock co esponding o equilib ium be ween he li e inpu and decomposi ion was achie ed by a Yasso07 spin-up un o 10 000 yea s. Yasso07 uns used li e inpu s o he equilib ium o es biomasses (see Sec . 2.1.1) and clima e a iables (annual ai empe a- u e, mon hly empe a u e ampli ude, and annual p ecipi a- ion). The global pa ame e alues o decomposi ion a es, low a es, and o he dependencies o he Yasso07 soil ca - bon model we e adop ed om Tuomi e al. (2011) and he es ima es o Yasso07 SOC s ocks we e used in compa ison wi h measu emen s and o he models. We did no use he SOC s ocks simula ed wi h he mo e ecen Yasso07 pa am- e e s based on he li e decomposi ion da a om he No dic coun ies (Ran aka i e al., 2012), because he SOC s ocks simula ed wi h he global pa ame e alues p oduced a be e i wi h SFSI measu emen s. The CENTURY ma hema ical model o iginally de el- oped o g assland sys ems (Pa on e al., 1987, 1992) has been since modi ied o a ious ecosys ems including bo eal o es s (Nalde and Wein, 2006). The CENTURY is also one o he mos widely applied models. The soil o ganic ma - e in he model consis s o ac i e, slow, and passi e pools, which ha e di e en TR (Table 3). The decomposi ion a es a e modi ied by empe a u e and mois u e, and in addi ion he decomposi ion a es o he slow and passi e pools ely on lignin o N and C o N a ios, while he ac i e pool de- composi ion a e elies on soil ex u e. The model simula es soil o ganic ma e o a dep h o 20cm. The model simu- la es plan p oduc ion and pools o li ing biomass, while TR o biomass pools de e mine he li e all inpu s o soil. To compa e he pe o mance o he soil sub-model wi h o he soil ca bon dynamics models, Q and Yasso07, we only used he CENTURY soil sub-model. We used he same li e all inpu s as used by he Q and Yasso07 simula ions, which we e es ima ed by ou li e all modeling (see Sec . 2.1.1). www.biogeosciences.ne /13/4439/2016/ Biogeosciences, 13, 4439–4459, 2016 4446 B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y Table 3. Desc ip ion o models and da a inpu s ele an o his s udy. Model Yasso07 Q CENTURY 4.0 soil submodel Time s ep Yea Yea Mon h Pa ame e iza ion Global Scandina ian Combined global wi h si e speci ic Ca bon pools Labile (acid -, wa e -, and e hanol- sol- uble and non-soluble), ecalci an (hu- mus) Coho s ( oliage, s ems, b anches, coa se oo s, ine oo s, “g ass”), soil o ganic Li e (su ace s uc u al and me abolic, belowg ound s . and me .), su ace mic obial, soil o ganic ma e (ac i e, slow and passi e) Biomass Biomass componen s es ima ed by allome ic biomass unc ions and p o ided s and da a o li e inpu es ima ion Li e amoun Annual o mon hly ac ions o biomass componen s (species speci ic, same o al li e inpu s o all models) Li e quali y Li e a u e-based solubili ies Es ima ed coho s quali ies C/N a ios and lignin /N a ios Tempe a u e ai Annual mean, mon hly ampli ude Annual mean Max and min mon hly mean P ecipi a ion Annual o al – Mon hly o al Soil p ope ies – – Bulk densi y, sand, sil , and clay con en Soil dep h (m) 1 – 0.2 The li e inpu s e lec ed N deposi ion and si e p oduc i i y (Fig. S11). Fo CENTURY we adop ed gene al pa ame e s om he pa ame e ile “ ee.100”, pa ame e s o si e “AND H_J_ANDREWS” o coni e s, and si e “CWT Cowee a” o deciduous ees. The N dynamics in CENTURY sub-model included uning si e-speci ic pa ame e s o opsoil mine al N ela i e o N deposi ion (Th oop e al., 2004) and educ ion o C/N a io o he li e all up o 15% o mos p oduc i e si es (Me ilä e al., 2014). We also accoun ed o si e-speci ic soil d ainage by a ying i s pa ame e be ween 1 and 0.6 el- a i e o long- e m soil wa e con en anging be ween 10 and 50% (Raich e al., 2000). The CENTURY SOC s ocks sim- ula ion we e un wi h equilib ium o es li e inpu s, si e- speci ic C/N a ios o li e all, si e-speci ic soil pa ame e s (speci ic bulk densi y, sand, sil , and clay con en , mine al N in opsoil, and d ainage) and clima e a iables (mon hly ai empe a u e, and mon hly p ecipi a ion). In o de o accoun o he deep soil ca bon (Jobbágy and Jackson, 2000), we scaled CENTURY es ima es ep esen ing he opsoil ho izon by adding 40% o es ima ed si e-speci ic SOC s ock. The simula ed equilib ium SOC s ocks we e es ima ed by a spin- up un o 5000 yea s. The numbe o yea s o each equi- lib ium (equilib ium be ween he li e inpu and decomposi- ion) was sough empi ically on 100 andom si es, and di e s om Yasso07 and Q models. 3 Resul s The dis ibu ions o Yasso07, Q, and CENTURY model es- ima es o o al SOC s ocks ( Cha−1) we e in ag eemen o wo- hi ds o he measu ed da a wi h lowe SOC s ock (Fig. 3, dis ibu ions o g oups 1, 2, and 4). The emain- ing one- hi d o SOC da a we e unde es ima ed by models. This one- hi d o da a we e sepa a ed in o se en physico- chemical soil g oups (means o g oups anging om 104 o excep ionally la ge 269 Cha−1, see Fig. 3, dis ibu ions o g oups 3, and 5–10). The linea eg ession o mean le els o all 10 physicochemical soil g oups (weigh ed by he numbe o samples in each g oup) be ween he modeled and mea- su ed SOC s ocks showed smalle unde es ima ion o CEN- TURY compa ed o Yasso07 and Q models (Fig. 4). The weigh ed oo mean squa e e o (RMSE) was 27.5 ( Cha−1) o CENTURY and 31.6 and 38.8 o Yasso07 and Q, espec- i ely. The p opo ion o explained a iance was la ge o Q ( 2=0.58) han o Yasso07 and CENTURY ( 2=0.42 and 0.32; Fig. 4). The de ia ion o he dis ibu ions o CENTURY SOC s ocks, simula ed using soil bulk densi y, sand, sil , and clay con en , we e lowe han hose o Yasso07 and Q es i- ma es o 10 physicochemical soil g oups (Fig. 3). Accoun - ing o si e-speci ic soil ex u e (clay, sil , and sand con en ) and s uc u e (bulk densi y) by he CENTURY model im- p o ed SOC s ock es ima es o e ile si es wi h high clay con en , bu no o si es wi h high N deposi ion. Va ying CENTURY pa ame e s o si e-speci ic opsoil mine al ni o- gen and C/N a io o he li e all showed ha his impac on SOC s ocks es ima es was small in compa ison o sensi i i y o SOC s ock es ima es o li e all (Fig. S12). The applica- ion o si e-speci ic d ainage on ou mos ly well-d ained soils showed mino impac on es ima ed CENTURY SOC s ocks. As expec ed, he models clea ly showed less a ia ion han he measu emen s. The shi o he mean alues om he cen- e o dis ibu ion, he wid h o con idence in e als o means, and he wid h o he ails o dis ibu ions we e clea ly la ge o he measu emen s han o he modeled es ima es (Fig. 3). The modeled dis ibu ions ag eed o he poo –medium e - ili y soils wi h low and medium measu ed SOC s ocks, low and medium CEC, unso ed pa en ma e ial, low empe a- u es, and low p oduc ion (g oups 1, 2, and 4; Figs. 2, 3, Table S1). Disag eemen be ween modeled and measu ed SOC s ock dis ibu ions we e o med on e ile soils wi h so ed pa en ma e ial (g oups 3 and 5), soils wi h highe wa e con en (g oups 3, 5, and 10), whe e ni ogen depo- si ion was la ge (g oups 7 and 8), and whe e CEC was me- dian o la ge (Figs. 2, 3). The la ges de ia ion be ween he Biogeosciences, 13, 4439–4459, 2016 www.biogeosciences.ne /13/4439/2016/ B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y 4447 0 100 200 300 Measu emen s Yasso07 CENTURY Q 0 100 200 300 n=959 n=909 n=136 n=335 n=182 n=296 n=180 n=8 n=142 n=83 12345678910 Soil ca bon s ock ha ( −1) Soil g oups Figu e 3. Bean plo o dis ibu ions o he soil ca bon ( Cha−1) measu emen s (g ay ill) and es ima es o 10 physicochemical g oups. The ull and dashed ho izon al lines ep esen he g oup means and hei con idence in e als. The nis he numbe o samples. Fo desc ip ion o g oup le els o SOC s ocks, mois u e, and e ili y see Fig. 2 and Table S1. measu ed and modeled dis ibu ions was ound o he ela- i ely small physicochemical g oups o soils (3%) ypical o highly bound wa e and pea humus ypes (g oups 8 and 10; Figs. 2, 3). The dis ibu ions o measu ed o al SOC s ocks ( Cha−1) gene ally inc eased o he g oups wi h highe nu- ien s a us (Figs. 3, S4). The dis ibu ions o SOC s ocks in mine al soil we e la ge han hose in humus ho izon, and dis ibu ions o mine al SOC s ocks inc eased wi h e ili y sligh ly mo e han dis ibu ions o SOC s ocks in humus ho i- zon (Fig. S4). A e excluding all he soil physicochemical cha ac e is- ics om he ecu si e pa i ioning, he SOC s ock dis ibu- ions o i e g oup eg ession ee models (Fig. S3, Table S2) we e in ag eemen be ween he measu emen s and model es- ima es o h ee g oups (77% o samples) and de ia ed o wo g oups (23%; Fig. S5). The modeled SOC s ock dis- ibu ions ag eed wi h measu emen s o all models on si es wi h low annual empe a u es <3 ◦C in no he n si es (low- C.cold.pine, low-C.cold.o he ) and o wa me condi ions in middle Sweden on si es wi h low ni ogen deposi ion and me- dian SOC s ocks (Fig. S5). Howe e , he models unde es i- ma ed SOC s ocks on si es wi h high (>10 kgNha−1y−1) N deposi ion (21% o samples) and on si es wi h wa m and d y clima e (2% o samples; Fig. S5). The a ia ion o densi y unc ions o modeled SOC s ocks o 10 physicochemical g oups (Fig. 3) was simila o he a ia ion o he o al annual plan li e inpu ( Cha−1; Fig. S6) indica ing ha li e all was he main d i e o SOC accumula ion in he models . The mean le els o annual plan li e inpu and mean SOC s ocks o 10 soil g oups we e mo e s ongly co ela ed o Yasso07 and Q models (wi h 2 alues 0.86 and 0.96, espec i ely) han o CENTURY ( 2=0.52). Al hough, models pe o med easonably well o he la ges soil g oups o nu ien and p oduc ion le els (Figs. 3 and 4), none o he models was able o p edic a i- a ion o indi idual samples (Fig. S7). The model es ima es we e well co ela ed be ween Yasso07 and CENTURY wi h 2 anging om 45 o 73% o indi idual samples o 10 soil g oups, whe eas he co ela ions o es ima es be ween Q and he o he wo models we e lowe (Fig. S8). 4 Discussion 4.1 SOC s ock dis ibu ions linked o mechanisms o SOM s abiliza ion I has been sugges ed ha p ocess-based soil ca bon mod- els wi h he cu en o mula ion lacking majo soil en i on- men al and biological con ols o decomposi ion would ail o condi ions whe e hese con ols p edomina e (Schmid e al., 2011; A e ill e al., 2014). E en so, he e ec o he soil p ope ies on SOC s ocks, e.g., soil nu ien s a us in he widely used models such as Yasso07, Q, and CENTURY, ha e no p e iously been quan i a i ely e alua ed. We ound ha in compa ison wi h Swedish o es soil in en o y da a, he models based on he amoun and quali y o inhe en s uc- u al p ope ies o plan li e (Q, Yasso07, and CENTURY) p oduced accu a e SOC s ock es ima es o wo- hi ds o no he n bo eal o es soils in Sweden. Two- hi ds o he dis ibu ions o SOC s ocks measu emen s o SFSI ag eed wi h dis ibu ions o SOC s ock es ima es o he Q, Yasso07, and CENTURY soil ca bon models (Fig. 3, dis ibu ions o g oups 1, 2, and 4). Howe e , he SOC s ocks unde es ima- ion by hese models o one- hi d o he da a (Fig. 3, dis i- www.biogeosciences.ne /13/4439/2016/ Biogeosciences, 13, 4439–4459, 2016 4454 B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y Appendix B: Models o o es d y weigh biomass wi h APAR We i ed species-speci ic exponen ial eg ession models be ween he biomass componen s (s em, b anch, oliage, s ump, coa se oo s, ine oo s, all in kgha−1) o obse ed s a e o es and he obse ed ac ion o abso bed adia ion ( APAR) (s a is ics o he eg ession models in Table B1 in Appendix B). The biomass componen s de i ed wi h allo- me ic models (measu ed) and hose de i ed wi h APAR models (modeled) showed s ong co ela ions (Fig. B1 in Appendix B). In o de o model he long- e m mean o es biomass “equilib ium o es biomass” we applied he APAR biomass models o he modeled APAR70 alues. Table B1. Pa ame e es ima es and hei s anda d e o s o he co- e icien s o he d y weigh biomass (kgha−1) models wi h he ac- ion o abso bed adia ion (y=ab APAR ) o Sco s pine, No way sp uce, and deciduous s ands. y=ab APAR Species a±SE b±SE adj.R2 B anch pine 610±21 122±6 0.92 sp uce 877±35 54±2 0.92 deciduous 290±26 156±16 0.89 Fine oo pine 422±13 21 ±1 0.84 sp uce 317±14 15±1 0.80 deciduous 453±28 14±1 0.82 Foliage pine 361±24 86±8 0.71 sp uce 766±40 33±2 0.83 deciduous 141±28 71±16 0.56 Roo pine 703 ±26 183 ±10 0.92 sp uce 629±32 113±7 0.90 deciduous 359±33 150±16 0.89 S em and ba k pine 1793±84 254 ±17 0.89 sp uce 974±72 229±19 0.86 deciduous 972±98 161±18 0.88 S ump pine 232±10 214±13 0.89 sp uce 171±10 129±9 0.88 deciduous 80±8 216±25 0.87 p < 0.001 o all pa ame e s. Biogeosciences, 13, 4439–4459, 2016 www.biogeosciences.ne /13/4439/2016/ B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y 4455 0 50 100 150 0 50 100 150 S em+ba k Modeled ( C ha−1) Measu ed ( C ha−1) 2 = 0.95 0 10 20 30 0 10 20 30 B anch Modeled ( C ha−1) Measu ed ( C ha−1) 2 = 0.97 0 5 10 15 0 5 10 15 Foliage Modeled ( C ha−1) Measu ed ( C ha−1) 2 = 0.94 0 5 10 15 0 5 10 15 S ump Modeled ( C ha−1) Measu ed C ha ( −1) 2 = 0.96 0 10 30 50 0 10 20 30 40 50 Roo Modeled ( C ha−1) Measu ed C ha ( −1) 2 = 0.97 01234 0 1 2 3 4 Fine oo Modeled ( C ha−1) Measu ed ( C ha−1) 2 = 0.96 Figu e B1. Sca e plo s (n=3698 in each panel) o he d y weigh ee biomass componen s ( Cha−1) be ween “modeled” (es ima ed based on ac ion o abso bed adia ion, APAR, and ou APAR models) and “measu ed” (es ima ed based on basic ee dimensions and allome ic biomass models). The 2 alues ep esen he coe icien o de e mina ion indica ing how close he modeled alues i he measu ed alues. www.biogeosciences.ne /13/4439/2016/ Biogeosciences, 13, 4439–4459, 2016 4456 B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y Appendix C: Models o unde s o y ege a ion We used Swedish o es in en o y g ound ege a ion co - e age (%) da a isually moni o ed be ween 1993 and 2002 on 2440 plo s a ound Sweden wi h al oge he 4472 obse a- ions sepa a ely o species o o es loo ege a ion o hei classes (Table S3). In o de o de i e he g ound ege a ion biomass and o apply he co e age/biomass con e sion unc- ions (Leh onen e al., 2016), we g ouped he species co - e age obse a ions in o i e unc ional ypes (dwa sh ubs, he bs, g asses, moss, and lichen; Table S3). The applied co e age/biomass con e sion unc ions es ima ed sepa a ely he abo e- and below-g ound biomass componen s o dwa sh ubs, he bs, and g asses, and o al biomass o moss, and lichen. Excep he unde s o y co e age, he o es in en o y da a also con ained basic ee dimensions (diame e and heigh o ees) and s and a iables (species dominance, age, basal a ea, si e p oduc i i y class indica ed by he heigh o la ges ess a 100 yea s o s ands age), and also we linked he plo s by hei closes p oximi y o SMHI wea he s a ions wi h wea he da a (ai empe a u e, p ecipi a ion) and loca ion a - ibu es o he wea he s a ions (la i ude, longi ude, al i ude). Table C1. Pa ame e es ima es and hei s anda d e o s o he coe icien s o he o es unde s o y ege a ion d y weigh biomass (kgha−1) models (Eq. C1) o unc ional ypes (1 – dwa sh ubs, 2 – he bs, 3 – g asses, 4 – mosses, and 5 – lichens) wi h in e cep (a) and n– numbe o p edic o s (b1 – age (yea s), b2 – basal a ea (m2ha−1), b3 – annual ai empe a u e (◦C), b4 – la i ude (◦), b5 – H100 (heigh o ees a 100 yea s o age, m), b6 – H100 o sp uce ees (m), b7 – H100 o pine ees (m), b8 – pine dominance (0/1), and b9 – sp uce dominance (0/1)). Fo he la in names o species included in o unde s o y unc ional ypes see Table S3. Wa±Eb1±SE b2±SE b3±SE b4±SE b5±SE b6±SE b7±SE b8±SE b9±SE adj.R2 Abo e 1 24.28 ±0.32 0.13 ±0.01 -0.43 ±0.02 7.13±0.33 0.29 g ound 2 −82.13 ±6.8 −0.1 ±0.1a1.23 ±0.1 0.77±0.03 0.12 3 4.07±0.30 −0.16 ±0.01 0.27 ±0.01 −1.36 ±0.15 0.21 4 32.9±0.62 −0.78 ±0.04 0.48 ±0.06 3.66 ±0.3 5.76 ±0.29 0.22 5 19.91±0.57 −0.13 ±0.01 −0.45 ±0.02 6.31 ±0.29 0.25 o al 43.68±0.29 0.12 ±0.01 −0.41 ±0.01 6.34 ±0.3 0.30 Below 1 −256.3±3.5 0.1±0.01 −0.35±0.02 5.05±0.06 8.56 ±0.35 0.75 g ound 2 −89.34 ±7.85 −0.03 ±0.1b1.4 ±0.12 0.78±0.04 −4.97 ±0.27 0.19 3 5.97±0.37 −0.19 ±0.01 0.32 ±0.01 −1.78 ±0.19 0.21 o al −251.9±3.3 −0.2 ±0.01 5.15 ±0.05 0.7 To al −222.7 ±4.0 0.12 ±0.01 −0.44±0.02 4.9 ±0.07 0.67 p < 0.001 o all pa ame e s excep o ap=0.44, and bp=0.84. We buil linea models o d y weigh biomass o unde s o y ege a ion (kgha−1) in a wo le el selec ion o he p edic o s om s and, wea he and loca ion a iables. Fi s , we selec ed he p edic o s in o linea models by using R package “Mass” and i s s epwise model selec ion by exac Akaike’s in o ma- ion c i e ion (AIC; Venables and Ripley, 2002). Second, we e ined he model by using “ elaimpo” R package es ima ing use ulness (G ömping, 2006), o ela i e impo ance o each o he p edic o s in he model, and by selec ing only p edic- o s wi h ela i e impo ance ≥0.1. The gene al o m o he models was yi=a+b1x1+...+bnxn+ε, (C1) whe e yiis he unde s o y d y weigh biomass (kgha−1), x1 ... xna e he p edic o s, a,b1... bna e pa ame e s o he i h unde s o y unc ional ype (Table C1 in Appendix C), and εis he esidual e o . S a is ics o he models a e shown in Table C1 in Appendix C. Sca e plo s be ween he measu ed co e age de i ed biomass and modeled d y weigh biomass (kgha−1) o he unc ional ypes o g ound ege a ion o he o es s in hei obse ed s a e close o he es ima ed equilib- ium a e shown on Fig. S9. Biogeosciences, 13, 4439–4459, 2016 www.biogeosciences.ne /13/4439/2016/ B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y 4457 The Supplemen ela ed o his a icle is a ailable online a doi:10.5194/bg-13-4439-2016-supplemen . Acknowledgemen s. We hank he Finnish Minis y o En i on- men and he Finnish Minis y o Ag icul u e and Fo es y o unding his wo k h ough he Me la p ojec 7509 “Imp o ing soil ca bon es ima ion o g eenhouse gas in en o y”, and Academy o Finland o unding he mobili y p ojec s 276300 and 276602. We would like o hank he edi o and he e iewe s o hei aluable commen s imp o ing he manusc ip . Edi ed by: A. V. Elisee Re iewed by: h ee anonymous e e ees Re e ences Adai , E. C., Pa on, W. J., Del G osso, S. J., Sil e , W. L., Ha mon, M. E., Hall, S. A., Bu ke, I. C., and Ha , S. C.: Simple h ee-pool model accu a ely desc ibes pa e ns o long- e m li e decompo- si ion in di e se clima es, Global Change Biol., 14, 2636–2660, 2008. Åg en, G. I., Bosa a, E., and Magill, A. H.: Combining heo y and expe imen o unde s and e ec s o ino ganic ni ogen on li e decomposi ion, Oecologia, 128, 94–98, 2001. Åg en, G. I. and Hy önen, R.: Changes in ca bon s o es in Swedish o es soils due o inc eased biomass ha es and inc eased em- pe a u es analysed wi h a semi-empi ical model, Fo es Ecol. Manage., 174, 25–37, 2003. Åg en, G., Hy önen, R., and Nilsson, T.: A e Swedish o es soils sinks o sou ces o CO2—model analyses based on o es in en- o y da a, Biogeochemis y, 82, 217–227, 2007. Amundson, R.: The ca bon budge in soils, Annu. Re . Ea h Plane . Sci., 29, 535–562, 2001. A e ill, C., Tu ne , B. L., and Finzi, A. C.: Myco hiza-media ed compe i ion be ween plan s and decompose s d i es soil ca bon s o age, Na u e, 505, 543–545, 2014. Be h ong, S. T., Jobbágy, E. G., and Jackson, R. B.: A global me a- analysis o soil exchangeable ca ions, pH, ca bon, and ni ogen wi h a o es a ion, Ecol. Appl., 19, 2228–2241, 2009. Boden, T. A., Ma land, G., and And es, R. J.: Global, e- gional, and na ional ossil- uel CO2emissions, Ca bon Diox- ide In o ma ion Analysis Cen e , Oak Ridge Na ional Labo- a o y, US Depa men o Ene gy, Oak Ridge, Tenn., USA, doi:10.3334/CDIAC/00001_V2010, 2010. Cas ellano, M. J., Muelle , K. E., Olk, D. C., Sawye , J. E., and Six, J.: In eg a ing Plan Li e Quali y, Soil O ganic Ma e S abiliza- ion and he Ca bon Sa u a ion Concep , Glob. Change Biol., 21, 3200–3209, doi:10.1111/gcb.12982, 2015. Cla holm, M. and Skyllbe g, U.: T ansloca ion o me als by ees and ungi egula es pH, soil o ganic ma e u no e and ni ogen a ailabili y in acidic o es soils, Soil Biol. Biochem., 63, 142– 153, 2013. Clemen e, J. S., Simpson, A. J., and Simpson, M. J.: Associa ion o speci ic o ganic ma e compounds in size ac ions o soils un- de di e en en i onmen al con ols, O g. Geochem., 42, 1169– 1180, 2011. Cools, N., Ves e dal, L., De Vos, B., Vanguelo a, E., and Hansen, K.: T ee species is he majo ac o explaining C: N a ios in Eu- opean o es soils, Fo es Ecol. Manage., 311, 3–16, 2014. Co u o, M. F., Wallens ein, M. D., Boo , C. M., Dene , K., and Paul, E.: The Mic obial E iciency-Ma ix S abiliza ion (MEMS) amewo k in eg a es plan li e decomposi ion wi h soil o ganic ma e s abiliza ion: do labile plan inpu s o m s able soil o - ganic ma e ?, Glob. Change Biol., 19, 988–995, 2013. Deluca, T. H. and Bois enue, C.: Bo eal o es soil ca bon: dis i- bu ion, unc ion and modelling, Fo es y, 85, 161–184, 2012. Dungai , J. A. J., Hopkins, D. W., G ego y, A. S., and Whi mo e, A. P.: Soil o ganic ma e u no e is go e ned by accessibili y no ecalci ance, Glob. Change Biol., 18, 1781–1796, 2012. Falloon, P., Jones, C. D., Ades, M., and Paul, K.: Di ec soil mois- u e con ols o u u e global soil ca bon changes: An impo - an sou ce o unce ain y, Global Biogeochem. Cy., 25, GB3010, doi:10.1029/2010GB003938, 2011. Fan, Z., Ne , J. C., Ha den, J. W., and Wickland, K. P.: Bo- eal soil ca bon dynamics unde a changing clima e: A model in e sion app oach, J. Geophys. Res.-Biogeo., 113, G04016, doi:10.1029/2008JG000723, 2008. Fe nández-Ma ínez, M., Vicca, S., Janssens, I. A., Sa dans, J., Luyssae , S., Campioli, M., Chapin III, F. S., Ciais, P., Malhi, Y., Obe s eine , M., Papale, D., Piao, S. L., Reichs ein, M., Roda, F., and Penuelas, J.: Nu ien a ailabili y as he key egula o o global o es ca bon balance, Na u e Clima e Change, 4, 471– 476, 2014. F anklin, O., Högbe g, P., Ekblad, A., and Åg en, G. I.: Pine o - es loo ca bon accumula ion in esponse o N and PK addi- ions: bomb 14C modelling and espi a ion s udies, Ecosys ems, 6, 644–658, 2003. G ömping, U.: Rela i e impo ance o linea eg ession in R: he package elaimpo, J. S a . So w., 17, 1–27, 2006. Hagglund, B. and Lundma k, J.: Si e index es ima ion by means o si e p ope ies, Sco s pine and No way sp uce in Sweden, S ud. Fo . Suec., 138, 38, 1977. Hä könen, S., Pulkkinen, M., Duu sma, R., and Mäkelä, A.: Es i- ma ing annual GPP, NPP and s em g ow h in Finland using sum- ma y models, Fo . Ecol. Manage., 259, 524–533, 2010. Husson, O.: Redox po en ial (Eh) and pH as d i e s o soil/plan /mic oo ganism sys ems: a ansdisciplina y o e iew poin ing o in eg a i e oppo uni ies o ag onomy, Plan Soil, 362, 389–417, 2013. Jobbágy, E. G. and Jackson, R. B.: The e ical dis ibu ion o soil o ganic ca bon and i s ela ion o clima e and ege a ion, Ecol. Appl., 10, 423–436, 2000. Kamme , A. and Hagedo n, F.: Mine alisa ion, leaching and s a- bilisa ion o 13C-labelled lea and wig li e in a beech o - es soil, Biogeosciences, 8, 2195–2208, doi:10.5194/bg-8-2195- 2011, 2011. Ki schbaum, M. U. F.: Will Changes in Soil O ganic Ca bon Ac as a Posi i e o Nega i e Feedback on Global Wa ming?, Biogeo- chemis y, 48, 21–51, 2000. Kleja, D. B., S ensson, M., Majdi, H., Jansson, P., Lang all, O., Be gk is , B., Johansson, M., Weslien, P., T uusb, L., and Lin- d o h, A.: Pools and luxes o ca bon in h ee No way sp uce ecosys ems along a clima ic g adien in Sweden, Biogeochem- is y, 89, 7–25, 2008. www.biogeosciences.ne /13/4439/2016/ Biogeosciences, 13, 4439–4459, 2016 4458 B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y Ku z, W. A., Beukema, S. J., and Apps, M. J.: Es ima ion o oo biomass and dynamics o he ca bon budge model o he C ana- dian o es sec o , Can. J. Fo es Res., 26, 1973–1979, 1996. Leh onen, A. and Heikkinen, J.: Unce ain y o upland soil ca bon sink es ima e o Finland, Can. J. Fo es Res., 45, 1–13, 2015. Leh onen, A., Sie änen, R., Mäkelä, A., Mäkipää, R., Ko honen, K. T. and Hokkanen, T.: Po en ial li e all o Sco s pine b anches in sou he n Finland, Ecol. Modell., 180, 305–315, 2004. Leh onen, A., Linkosalo, T., Pel oniemi, M., Sie änen, R., Mäkipää, R., Tamminen, P., Salemaa, M., Nieminen, T., Tupek, B., Heikki- nen, J., and Koma o , A.: Soil ca bon s ock es ima es in a na- ionwide in en o y: e alua ing pe o mance o he ROMUL and Yasso07 models, Geosci. Model De . Discuss., in e iew, 2016a. Leh onen, A., Pal iainen, M., Ojanen, P., Kalliokoski, T., Nöjd, P., Kukkola, M., Pen ilä, T., Mäkipää, R., Leppälammi-Kujansuu, J. and Helmisaa i, H.: Modelling ine oo biomass o bo eal ee s ands using si e and s and a iables, Fo es Ecol. Manag., 359, 361–369, doi:10.1016/j. o eco.2015.06.023, 2016b. Leppälammi-Kujansuu, J., A o, L., Salemaa, M., Hansson, K., Kleja, D. B., and Helmisaa i, H.: Fine oo longe i y and ca bon inpu in o soil om below-and abo eg ound li e in clima ically con as ing o es s, Fo es Ecol. Manage., 326, 79–90, 2014. Lindén, A., G.: Swedish Geological Su ey epo , pp. 10, a ail- able a : h p:// esou ce.sgu.se/p oduk e /ae/ae118-besk i ning. pd (las access: 3 Augus 2016), 2002. Liski, J., Tuomi, M., and Rasinmäki, J.: Yasso07 use -in e ace manual, Finnish En i onmen Ins i u e, Helsinki, 2009. Liski, J., Leh onen, A., Palosuo, T., Pel oniemi, M., Egge s, T., Muukkonen, P., and Mäkipää, R.: Ca bon accumula ion in Fin- land’s o es s 1922–2004–an es ima e ob ained by combina ion o o es in en o y da a wi h modelling o biomass, li e and soil, Ann. Fo es Sci., 63, 687–697, 2006. Majdi, H.: Changes in ine oo p oduc ion and longe i y in ela ion o wa e and nu ien a ailabili y in a No way sp uce s and in no he n Sweden, T ee Physiol., 21, 1057–1061, 2001. Malkonen, E.: Annual p ima y p oduc ion and nu ien cycle in a bi ch s and, Commun. Ins . Fo . Fenn., 91, 1–35, 1977. Manzoni, S., Taylo , P., Rich e , A., Po po a o, A., and Åg en, G. I.: En i onmen al and s oichiome ic con ols on mic obial ca bon use e iciency in soils, New Phy ol., 196, 79–91, 2012. Ma klund, L.: , Biomassa unk ione ö all, g an och bö k i S e ige, Biomass unc ions o pine, sp uce and bi ch in Sweden, Swedish Uni e si y o Ag icul u al Sciences, Depa men o Fo es Su - ey, Repo 45, ISSN 0348–0496, 1988. McCo mack, M. L., C is ield, E., Raczka, B., Schnekenbu ge , F., Eissens a , D. M., and Smi hwick, E. A.: Sensi i i y o ou eco- logical models o adjus men s in ine oo u no e a e, Ecol. Model., 297, 107–117, 2015. Me he ell, A. K.: Cen u y: Soil O ganic Ma e Model En i on- men : Technical Documen a ion: Ag oecosys em Ve sion 4.0, Colo ado S a e Uni e si y, 1993. Me ilä, P., Mus ajä i, K., Helmisaa i, H., Hilli, S., Lind oos, A., Nieminen, T. M., Nöjd, P., Rau io, P., Salemaa, M., and Ukon- maanaho, L.: Abo e-and below-g ound N s ocks in coni e ous bo eal o es s in Finland: Implica ions o sus ainabili y o mo e in ensi e biomass u iliza ion, Fo . Ecol. Manage., 311, 17–28, 2014. Muukkonen, P. and Leh onen, A.: Needle and b anch biomass u no e a es o No way sp uce (Picea abies), Can. J. Fo es Res., 34, 2517–2527, 2004. Nalde , I. A. and Wein, R. W.: A model o he in es iga ion o long- e m ca bon dynamics in bo eal o es s o wes e n Canada: I. Model de elopmen and alida ion, Ecol. Model., 192, 37–66, 2006. Olsson, M. T., E landsson, M., Lundin, L., Nilsson, T., Nilsson, Å., and S endahl, J.: O ganic ca bon s ocks in Swedish Podzol soils in ela ion o soil hyd ology and o he si e cha ac e is ics, Sil a Fennica, 43, 209–222, 2009. O iz, C., Ka l un, E., S endahl, J., Gä denäs, A. I., and Åg en, G. I.: Modelling soil ca bon de elopmen in Swedish coni e ous o - es soils—An unce ain y analysis o pa ame e s and model es i- ma es using he GLUE me hod, Ecol. Model., 222, 3020–3032, 2011. O iz, C. A., Liski, J., Gä denäs, A. I., Leh onen, A., Lundblad, M., S endahl, J., Åg en, G. I., and Ka l un, E.: Soil o ganic ca bon s ock changes in Swedish o es soils – A compa ison o unce - ain ies and hei sou ces h ough a na ional in en o y and wo simula ion models, Ecol. Model., 251, 221–231, 2013. O win, K. H., Ki schbaum, M. U., S John, M. G., and Dickie, I. A.: O ganic nu ien up ake by myco hizal ungi enhances ecosys- em ca bon s o age: a model-based assessmen , Ecol. Le ., 14, 493–502, 2011. Palosuo, T., Foe eid, B., S ensson, M., Shu pali, N., Leh onen, A., He bs , M., Linkosalo, T., O iz, C., Rampazzo Todo o ic, G., Ma cinkonis, S., Li, C., and Jandl, R.: A mul i-model compa ison o soil ca bon assessmen o a coni e ous o es s and, En i on. Modell. So w., 35, 38–49, 2012. Pa on, W. J., Schimel, D. S., Cole, C. V., and Ojima, D. S.: Analysis o Fac o s Con olling Soil O ganic Ma e Le els in G ea Plains G asslands, 51, 1173–1179, 1987. Pa on, W., Ojima, D., and Schimel, D.: En i onmen al change in g asslands: Assessmen using models, Clima e Change, 28, 111– 141, 1994. Pa on, W.J., McKeown, R., Ki chne , V., and Ojima,D.: CEN- TURY Use s’ Manual, Na u al Resou ce Ecology Labo a o y, Colo ado S a e Uni e si y, F . Collins., 1992. Pel oniemi, M., Pulkkinen, M., Au ela, M., Pumpanen, J., Kola i, P. and Mäkelä, A.: A semi-empi ical model o bo eal- o es g oss p ima y p oduc ion, e apo anspi a ion, and soil wa e – calib a ion and sensi i i y analysis, Bo eal En i on. Res., 20, 2015. Pe e sson, H. and S åhl, G.: Func ions o below-g ound biomass o Pinus syl es is, Picea abies, Be ula pendula and Be ula pubescens in Sweden, Scand., J. Fo . Res., 21, 84–93, 2006. Raich, J. W., Pa on, W. J., Russell, A. E., San o d J , R. L., and Vi- ousek, P. M.: Analysis o ac o s egula ing ecosys emde elop- men on Mauna Loa using he Cen u y model, Biogeochemis y, 51, 161–191, 2000. Ran aka i, M., Leh onen, A., Linkosalo, T., Tuomi, M., Tammi- nen, P., Heikkinen, J., Liski, J., Mäkipää, R., Il esniemi, H., and Sie änen, R.: The Yasso07 soil ca bon model–Tes ing agains e- pea ed soil ca bon in en o y, Fo es Ecol. Manage., 286, 137– 147, 2012. Rapalee, G., T umbo e, S. E., Da idson, E. A., Ha den, J. W., and Veldhuis, H.: Soil Ca bon s ocks and hei a es o accumula ion Biogeosciences, 13, 4439–4459, 2016 www.biogeosciences.ne /13/4439/2016/ B. ˇ Tupek e al.: Unde es ima ion o modeled SOC s ocks linked o soil e ili y 4459 and loss in a bo eal o es landscape, Global Biogeochem. Cy., 12, 687–701, 1998. Repola, J.: Biomass equa ions o bi ch in Finland, Sil a Fenn., 42, 605–624, 2008. Rawls, W. J., Pachepsky, Y. A., Ri chie, J. C., Sobecki, T. M., and Bloodwo h, H.: E ec o soil o ganic ca bon on soil wa e e- en ion, Geode ma, 116, 61–76, 2003. Rol , C. and Åg en, G. I.: P edic ing e ec s o di e en ha es ing in ensi ies wi h a model o ni ogen limi ed o es g ow h, Ecol. Model., 118, 193–211, 1999. R Co e Team R: A language and en i onmen o s a is ical com- pu ing. R Founda ion o S a is ical Compu ing, Vienna, Aus- ia, a ailable a : h p://www.R-p ojec .o g/ (las access: 3 Au- gus 2016), 2014. Scha lemann, J. P., Tanne , E. V., Hiede e , R., and Kapos, V.: Global soil ca bon: unde s anding and managing he la ges e - es ial ca bon pool, Ca bon Managemen , 5, 81–91, 2014. Schlesinge , W. H.: Ca bon Seques a ion in Soils, Science, 284, 2095–2095, 1999. Schmid , M. W. I., To n, M. S., Abi en, S., Di ma , T., Guggen- be ge , G., Janssens, I. A., Klebe , M., Kogel-Knabne , I., Lehmann, J., Manning, D. A. C., Nannipie i, P., Rasse, D. P., Weine , S., and T umbo e, S. E.: Pe sis ence o soil o ganic ma - e as an ecosys em p ope y, Na u e, 478, 49–56, 2011. Six, J., Conan , R. T., Paul, E. A., and Paus ian, K.: S abiliza ion mechanisms o soil o ganic ma e : Implica ions o C-sa u a ion o soils, Plan Soil, 241, 155–176, 2002. SLU: Ma kin o, a ailable a : h p://www-ma kin o.slu.se/eng/ index.h ml (las access: 3 Augus 2016), 2011. Smi h, P.: An o e iew o he pe manence o soil o ganic ca bon s ocks: in luence o di ec human-induced, indi ec and na u al e ec s, Eu . J. Soil Sci., 56, 673–680, 2005. Sollins, P., Homann, P., and Caldwell, B. A.: S abiliza ion and des abiliza ion o soil o ganic ma e : mechanisms and con ols, Geode ma, 74, 65–105, 1996. S a is ics Finland: G eenhouse gas emissions in Finland 1990– 2011, in: Na ional In en o y Repo o he UNFCCC Sec e a ia , Minis y o he En i onmen , Helsinki, Finland, 285–286, 2013. S endahl, J., Johansson, M., E iksson, E., Nilsson, Å., and Lang- all, O.: Soil o ganic ca bon in Swedish sp uce and pine o es s—di e ences in s ock le els and egional pa e ns, Sil a Fenn., 44, 5–21, 2010. Swedish S a is ical Yea book o Fo es y: O icial S a is ics o Swe- den, 370 pp., Skogss y elsen, 2014. S ensson, M., Jansson, P., and Kleja, D. B.: Modelling soil C se- ques a ion in sp uce o es ecosys ems along a Swedish ansec based on cu en condi ions, Biogeochemis y, 89, 95–119, 2008. The neau, T. M. and A kinson, E. J.: An in oduc ion o ecu si e pa i ioning using he RPART ou ines, 1997. Th oop, H. L., Holland, E. A., Pa on, W. J., Ojima, D. S., and Keough, C. A.: E ec s o ni ogen deposi ion and insec he - bi o y on pa e ns o ecosys em le el ca bon and ni ogen dy- namics: esul s om he CENTURY model, Glob. Change Biol., 10, 1092–1105, 2004. Todd-B own, K. E. O., Rande son, J. T., Hopkins, F., A o a, V., Ha- jima, T., Jones, C., She liako a, E., Tjipu a, J., Volodin, E., Wu, T., Zhang, Q., and Allison, S. D.: Changes in soil o ganic ca bon s o age p edic ed by Ea h sys em models du ing he 21s cen- u y, Biogeosciences, 11, 2341–2356, doi:10.5194/bg-11-2341- 2014, 2014. Todd-B own, K. E. O., Rande son, J. T., Pos , W. M., Ho man, F. M., Ta nocai, C., Schuu , E. A. G., and Allison, S. D.: Causes o a ia ion in soil ca bon simula ions om CMIP5 Ea h sys em models and compa ison wi h obse a ions, Biogeosciences, 10, 1717–1736, doi:10.5194/bg-10-1717-2013, 2013. To n, M. S., T umbo e, S. E., Chadwick, O. A., Vi ousek, P. M., and Hend icks, D. M.: Mine al con ol o soil o ganic ca bon s o age and u no e , Na u e, 389, 170–173, 1997. Tuomi, M., Rasinmäki, J., Repo, A., Vanhala, P., and Liski, J.: Soil ca bon model Yasso07 g aphical use in e ace, En i on. Modell. So w., 26, 1358–1362, 2011. Tuomi, M., Thum, T., Jä inen, H., F onzek, S., Be g, B., Ha mon, M., T o ymow, J. A., Se an o, S., and Liski, J.: Lea li e de- composi ion – Es ima es o global a iabili y based on Yasso07 model, Ecol. Model., 220, 3362–3371, 2009. Vi o, P.: In es iga ions on o es li e , Val ioneu os on Ki jap, 1955. Wiesmeie , M., Hübne , R., Spö lein, P., Geuß, U., Hangen, E., Reischl, A., Schilling, B., on Lü zow, M., and Kögel-Knabne , I.: Ca bon seques a ion po en ial o soils in sou heas Ge many de i ed om s able soil o ganic ca bon sa u a ion, Glob. Change Biol., 20, 653–665, 2014. Xia, M., Talhelm, A. F. and P egi ze , K. S.: Fine oo s a e he dominan sou ce o ecalci an plan li e in suga maple- domina ed no he n ha dwood o es s, New Phy ol., 208, 715– 726, doi:10.1111/nph.13494, 2015. Yu o a, A. Y., Volodin, E. M., Ag en, G. I., Che o , O. G., and Koma o , A. S.: E ec s o a ia ions in simula ed changes in soil ca bon con en s and dynamics on u u e clima e p ojec ions, Glob. Change Biol., 16, 823–835, 2010. Venables, W. N. and Ripley, B. D.: Mode n applied s a is ics wi h S-PLUS, Sp inge Science & Business Media, 2013. Zandonadi, D. B., San os, M. P., Busa o, J. G., Pe es, L. E. P. ,and Façanha, A. R.: Plan physiology as a ec ed by humi ied o ganic ma e , Theo e ical and Expe imen al Plan Physiology, 25, 13– 25, 2013. www.biogeosciences.ne /13/4439/2016/ Biogeosciences, 13, 4439–4459, 2016