Geosci. Model De ., 9, 4169–4183, 2016
www.geosci-model-de .ne /9/4169/2016/
doi:10.5194/gmd-9-4169-2016
© Au ho (s) 2016. CC A ibu ion 3.0 License.
Fo es 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
in Finland
Aleksi Leh onen1, Tapio Linkosalo1, Mikko Pel oniemi1, Ris o Sie änen1, Raisa Mäkipää1, Pekka Tamminen1,
Maija Salemaa1, Tiina Nieminen1, Bo is ˇ
Tupek1, Juha Heikkinen1, and Alexande Koma o 2,†
1Na u al Resou ces Ins i u e Finland, Na u al esou ces and biop oduc ion, (LUKE), P.O. Box 2, 00791 Helsinki, Finland
2Ins i u e o Physicochemical and Biological P oblems in Soil Science o he Russian Academy o Sciences,
142290 Ins i u skaya ul., 2, Pushchino, Moscow, Russian Fede a ion
†deceased
Co espondence o: Aleksi Leh onen ([email p o ec ed])
Recei ed: 6 June 2016 – Published in Geosci. Model De . Discuss.: 10 Augus 2016
Re ised: 2 No embe 2016 – Accep ed: 4 No embe 2016 – Published: 22 No embe 2016
Abs ac . Dynamic soil models a e needed o es ima ing
impac o wea he and clima e change on soil ca bon s ocks
and luxes. He e, we e alua e pe o mance o Yasso07 and
ROMUL models agains o es soil ca bon s ock measu e-
men s. Mo e speci ically, we ask i li e quan i y, li e qual-
i y and wea he da a a e su icien d i e s o soil ca bon
s ock es ima ion. We also es whe he inclusion o soil wa-
e holding capaci y imp o es eliabili y o modelled soil ca -
bon s ock es ima es. Li e inpu o ees was es ima ed om
s em olume maps p o ided by he Na ional Fo es In en-
o y, while unde s o ey ege a ion was es ima ed using new
biomass models. The li e p oduc ion a es o ees we e
based on ea lie esea ch, while o unde s o ey biomass hey
we e es ima ed om measu ed da a. We applied Yasso07
and ROMUL models ac oss Finland and an hose mod-
els in o s eady s a e; he ea e , measu ed soil ca bon s ocks
we e compa ed wi h model es ima es. We ound ha he
ole o unde s o ey li e inpu was unde es ima ed when he
Yasso07 model was pa ame e ised, especially in no he n
Finland. We also ound ha he inclusion o soil wa e hold-
ing capaci y in he ROMUL model imp o ed p edic ions,
especially in sou he n Finland. Ou simula ions and mea-
su emen s show ha models using only li e quali y, li e
quan i y and wea he da a unde es ima e soil ca bon s ock in
sou he n Finland, and his unde es ima ion is due o omission
o he impac o d ough s o he decomposi ion o o ganic
laye s. Ou esul s also imply ha he ecosys em modelling
communi y and g eenhouse gas in en o ies should imp o e
unde s o ey li e es ima ion in he no he n la i udes.
1 In oduc ion
Soil ca bon is a signi ican componen o e es ial ca bon
s ocks and unde s anding i s dynamics unde changing cli-
ma e is c ucial. Howe e , he signi icance and in e ac ions
o di e en mechanisms o long- e m ca bon accumula ion
a e s ill unknown and a e he e o e o en lacking in models.
I we wan o unde s and he ela ionship o di e en abi-
o ic and en i onmen al ac o s o soil ca bon s ocks and hei
dynamics, we ha e o combine expe imen al esea ch wi h
p ocess-based models. One way o wa d is o es ablish soil
ca bon in en o ies in o de o quan i y soil ca bon s ocks and
hei change. Con en ional soil in en o ies measu ing a i-
ous nu ien s, ca bon con en s, bulk densi ies and s oniness
(e.g. Gam eld e al., 2013) allow us o s udy he dis ibu ion
o soil ca bon ac oss landscapes and co ela ions be ween
di e en soil p ope ies. Gene ally, i is shown ha soil ca -
bon in en o ies a e able o p oduce soil maps and co a ia es
be ween soil ca bon quan i ies and o he a iables, such as
a ious nu ien s, al hough he sample size o hese in en-
o ies is usually no adequa e o na ional-le el soil ca bon
s ock change assessmen , wi h ew excep ions (e.g. in Swe-
Published by Cope nicus Publica ions on behal o he Eu opean Geosciences Union.
4170 A. Leh onen e al.: Fo es soil ca bon s ock es ima es in a na ionwide in en o y
den and Ge many) (Gam eld e al., 2013; G ünebe g e al.,
2014).
Acco ding o he Uni ed Na ions clima e con en ion,
coun ies a e eques ed o epo annual ca bon s ock changes
o soils unde di e en land uses and unde land-use change.
The majo i y o coun ies apply soil ca bon models, like
Yasso07 (Tuomi e al., 2011) and CENTURY (Pa on e al.,
1987) o es ima e soil ca bon s ock changes. These annual
submission a e e iewed annually and me hods should be
anspa en and e i iable, a ou ing simple soil models in-
s ead o complex ones.
The scien i ic communi y is also aiming o p edic u u e
soil–clima e change eedbacks on a global scale using Ea h
sys em models (ESMs). The ESMs a e es ed agains soil ca -
bon measu emen s in o de o e alua e model pe o mance
bu un o una ely esul s ha e been poo . Guene e al. (2013)
p esen a es whe e soil ca bon s ocks p edic ed by he OR-
CHIDEE model we e plo ed a a plo le el agains measu e-
men s and ailed o display any co ela ion. Simila ly, Todd-
B own e al. (2013) concluded ha mos ESMs a e no able
o p oduce measu ed soil ca bon s ocks a a g id le el. This
inding is somewha ala ming due o ac ha i is essen ial o
ha e co ec ini ial ca bon s ocks wi h soil models, because
ca bon s ock change es ima es depend on hem. Ini ial soil
ca bon s ock es ima es a e pa icula ly impo an when ca -
bon s ock changes o de o es a ion e en s a e modelled.
Indi idual soil ca bon models a e es ed agains epea ed
soil in en o ies and i has been ound ha models a e able o
es ima e soil ca bon s ock change o he same magni ude as
was measu ed (O iz e al., 2009; Ran aka i e al., 2012). The
limi a ion o his conclusion was ha unce ain ies o bo h
measu emen s and model es ima es a e o en highe han ac-
ual es ima es (O iz e al., 2013; Ran aka i e al., 2012).
While he unce ain ies be ween model ou pu and eal mea-
su emen s e eal whe he models ag ee wi h da a o no , hey
pu less emphasis on whe he all he mos impo an soil ca -
bon s ock d i e s we e included in hese models.
Simplis ic soil ca bon models like Yasso07 (Tuomi e al.,
2011) a e d i en only by wea he condi ions and by li e
inpu ; while mo e complex models, like ROMUL (Che o
e al., 2001) also include he ni ogen cycle and he impac
o soil wa e holding capaci y on decomposi ion. I is e i-
den ha soil p ope ies a ec soil ca bon s ocks (Schimel
e al., 1994; Six e al., 2002), and he e o e hey a e explic-
i ly included in complex models. Fo example, in he CEN-
TURY model, clay con en limi s decomposi ion (Pa on e
al., 1987) and due o he low speci ic su ace a ea o clay
mine als in he model, clay- ich soils ha e la ge passi e soil
ca bon s ocks and lowe C:N a ios (Pa i e al., 1997). In
simplis ic models, soil p ope ies a e omi ed. Fo example,
in Yasso07, he e ec o soil p ope ies on soil ca bon s ock
accumula ion is included only implici ly h ough he model
calib a ion wi h la ge da ase s (Tuomi e al., 2011). Al hough
he simple models lack some p edominan d i e s o soil
ca bon accumula ion, he s eng h o hese models lies in
hei easie calib a ion wi h da a; howe e , he impac o soil
p ope ies, especially nu ien s a us, on he accu acy o es-
ima ed soil ca bon s ock es ima es needs o be e-e alua ed
in bo h he CENTURY and Yasso07 models (ˇ
Tupek e al.,
2016).
I is well known ha decomposi ion and soil espi a ion
is con olled by wa e con en , whe eby in d y soils, lack
o wa e slows down decomposi ion, while excess wa e e-
duces i by limi ing oxygen di usion (Skopp e al., 1990).
Fo bo eal o es condi ions, Pumpanen e al. (2003) p o-
posed a model whe eby maximum soil espi a ion d ops a e
he ela i e wa e con en o soil eaches a le el o 60%. The
limi ing e ec o soil mois u e on decomposi ion in d y o
wa e -sa u a ed soils is widely included in models, al hough
he deg ee o his dependency a ies widely be ween models
(Sie a e al., 2015).
In addi ion o he model s uc u e, p ecise and accu a e es-
ima es o li e inpu quan i y and quali y a e also essen ial
o success ul model applica ions. S and-alone soil models
ely on o es in en o y da a o o he ex e nal es ima es o
ge ing co ec li e inpu s, while ecosys em models u ilise
plan submodels o desc ibe ege a ion p oduc i i y and li e
inpu s. A common ea u e o ecosys em models and s and-
alone soil models is ha o en unde s o ey ege a ion is ne-
glec ed du ing he calib a ion and applica ion o models, e-
sul ing in biased li e inpu s. This omission is mo e c i i-
cal in bo eal landscapes whe e he con ibu ion o unde -
s o ey ege a ion inc eases wi h no he ly la i ude. Fo ex-
ample, Yuan e al. (2014) epo ha he b yophy e biomass
con ibu es 20–60 % o he o al no malized di e ence ege-
a ion index (NDVI) in high no he n la i udes.
In o de o assess he necessa y d i e s o soil ca bon
models in Finland, we es ed he pe o mance o Yasso07
and ROMUL models agains measu ed na ion-wide soil C
da a. The ROMUL e e s o a modi ied ROMUL model e -
sion whe e decomposi ion a e unc ions a e de i ed om he
model p esen ed by Pumpanen e al. (2003), and a simple
olume ic soil wa e model (Linkosalo e al., 2013) is ap-
plied o d i e hose decomposi ion unc ions.
We we e speci ically in e es ed i he addi ional complex-
i y o p ocesses in oduced by ROMUL imp o es he soil C
s ock es ima e o e Yasso07 model ha is p esen ly used by
g eenhouse gas (GHG) in en o ies o se e al coun ies. The
Yasso07 and ROMUL models di e ed in hei ime s eps
(annually s. daily), de e mina ion o he li e quali y (li -
e solubili y s. ni ogen con en and li e sou ce) and com-
plexi y o d i e s (ROMUL including he impac o ni ogen
and soil wa e holding capaci y o decomposi ion). Speci i-
cally wi h hese models, he ollowing de ails we e es ed.
1. We in es iga ed whe he he li e quan i y, species-
speci ic li e quali y and wea he da a a e enough o
es ima e spa ial ends o ca bon s ocks in he upland
soils o Finland. We hypo hesise ha by accoun ing o
soil wa e holding capaci y, ROMUL would ou pe -
Geosci. Model De ., 9, 4169–4183, 2016 www.geosci-model-de .ne /9/4169/2016/
A. Leh onen e al.: Fo es soil ca bon s ock es ima es in a na ionwide in en o y 4171
o m (i) ROMUL wi h cons an soil wa e holding ca-
paci y and (ii) he Yasso07 model wi h di e en pa am-
e e isa ions (Yasso07 excludes soil p ope ies en i ely).
2. We hypo hesised whe he imp o ing he es ima ion o
unde s o ey li e inpu would posi i ely a ec accu acy
o p edic ed soil ca bon s ocks.
We use Yasso07 and ROMUL soil models o es ima e
s eady-s a e ca bon s ocks o upland soils in Finland on a
spa ial 10×10km2g id. We an he Yasso07 model wi h pa-
ame e s based on Scandina ian da a (Ran aka i e al., 2012)
and also wi h pa ame e s based on global da a (Tuomi e al.,
2011). The pa ame e isa ion o he ROMUL model was he
same as in he o iginal ROMUL model (Che o e al., 2001),
excep o decomposi ion a e unc ions depending on soil
wa e con en de i ed om Linkosalo e al. (2013). Yasso07
and ROMUL models un wi h iden ically es ima ed li e
quan i y, quali y and clima e da a. In addi ion, he ROMUL
model was es ed wi h bo h cons an soil wa e holding ca-
paci y as well as a iable wa e holding capaci y based on
digi al soil maps o Lilja and Ne alainen (2006). Fu he -
mo e, we de eloped new models o he unde s o ey li e
inpu and apply hem alongside soil ca bon models o im-
p o ed es ima es o spa ial a ia ion o li e inpu s. Simu-
la ed soil ca bon s ocks a e e alua ed agains measu ed soil
ca bon s ocks.
2 Ma e ial and me hods
Soil ca bon simula ions we e pe o med on a 10×10km2
g id ac oss Finland. This g id is used o me eo ological da a
p edic ion (Venäläinen e al., 2005). Li e inpu was es i-
ma ed o he same g id. F om he g id, only loca ions ha
we e on upland soils and o es (acco ding o he Food and
Ag icul u e O ganiza ion o he Uni ed Na ions (FAO) o es
de ini ions) we e chosen. This classi ica ion is based on Mul-
isou ce Na ional Fo es In en o y p oduc s, which combine
digi al maps (including, e.g. land use and pea lands), o es
in en o y da a and Landsa images (Tomppo e al., 2008).
2.1 T ee biomass
Fi s ly, s em olume maps by ee species om he Na ional
Fo es In en o y 9 (NFI9, 1996–2003) we e used, acco ding
o Tomppo e al. (2011), o accoun o la ge-scale a ia ion
o s em olume ac oss Finland. These a ia ions a e p ima -
ily d i en by soil p ope ies, clima e, si e p oduc i i y, o es
managemen echniques and ee species dis ibu ion.
Secondly, biomass expansion ac o s (BEFs, Mgm−3)
we e es ima ed o main ee species g oups, namely Sco s
pine (Pinus syl es is), No way sp uce (Picea abies) and o
b oadlea ed species (mainly Be ula sp.). These BEFs we e
es ima ed o each biomass componen ( oliage, b anches,
ba k, s emwood, s ump and woody oo s). Biomass was es-
ima ed o sample ees in NFI10-based biomass models by
Repola (2008, 2009); he ea e , mean BEFs we e es ima ed
a a clus e le el (a clus e is o med by 10 o 15 ield plo s)
by di iding he sum o gi en biomass componen s by he es-
ima ed sum o s em olumes. Biomass es ima es o ees
we e based on biomass models, whe e diame e a b eas
heigh , ee heigh , c own heigh , an inc emen o 5 yea s
and ba k hickness a e used as p edic o s. To upscale BEFs
ac oss Finland, we applied colloca ed co-k iging (Bi and e
al., 2008) by species g oup, o accoun o la ge-scale spa-
ial co ela ion and co- a ia ion o ee allome y. We used
he “gs a ” (Pebesma, 2004) package o R (R Co e Team,
2014) o es ima ion. Fo Sco s pine and No way sp uce,
we emo ed linea ends o la i ude and longi ude (using
uni o m coo dina e sys em o Finland, YKJ), while o de-
ciduous ees only ends o la i ude we e emo ed. Fo all
species g oups and componen s we assumed sphe ical a i-
og am unc ions. Fo he de ails o he used biomass mod-
els, see Appendix 7c by S a is ics Finland (2014). Biomass
componen s o each g id poin we e ob ained by mul iply-
ing s em olume maps by species wi h componen species-
speci ic BEF es ima es ha we e es ima ed ia co-k iging o
he same g id.
Biomass o ha es esidues and na u al mo ali y we e
es ima ed based on o es s a is ics and NFI da a (Yli alo,
2013). F om s a is ics, we a ained an es ima e o he s em
olume o annual logging and na u al mo ali y by egion
( o es y cen es) and hese we e subsequen ly con e ed o
biomass wi h BEFs. These BEFs we e based on a subse o
pe manen sample plo s o NFI9 (1996–2003) and NFI10
(2004–2008). BEFs o logging we e es ima ed sepa a ely
om he subse o logged plo s and hese logging-speci ic
BEFs we e used in he es ima ion o biomass o ha es
esidues. Fu he mo e, ene gy use o s umps and ha es
esidues we e deduc ed om egional soil inpu s, based on
egional wood ene gy use (Yli alo, 2013). Fo biomass o
na u al mo ali y, BEFs we e es ima ed based on da a om
hose ees ha died on pe manen sample plo s be ween
measu emen s. The ea e , he olume o na u al mo ali y
was mul iplied wi h co esponding BEFs. This p ocedu e ol-
lowed p inciples o Finnish GHG in en o y (S a is ics Fin-
land, 2014).
Fine oo biomass was es ima ed based on he wo k o
Leh onen e al. (2016). We selec ed a simple model o mu-
la ion, whe e a na u al loga i hm o ine oo mass was es i-
ma ed as a unc ion o he na u al loga i hm o s em olume.
See model 1 in Leh onen e al. (2016) o de ails ( he s udy
has 95 si es wi h ine oo da a, and model 1 has an adjus ed
R2o 0.217). This model was used o app oxima e gene al
ine oo mass le els as a unc ion o s em olume o each
10×10 km2g id poin .
2.2 Unde s o ey ege a ion biomass
In o de o es ima e he li e inpu o unde s o ey ege a-
ion o soils, we de eloped models o ege a ion biomass.
www.geosci-model-de .ne /9/4169/2016/ Geosci. Model De ., 9, 4169–4183, 2016
4172 A. Leh onen e al.: Fo es soil ca bon s ock es ima es in a na ionwide in en o y
Table 1. Si e desc ip ion o ICP Fo es s Le el II plo s s udied o unde s o ey ege a ion biomass.
Plo A ea Region No h Eas T ee Si e Fo es S and Basal DF GR HE BR LI Yea o nabo e/
no. coo d. coo d. species ype ype age a ea m2sampling below
1 Se e ijä i 1 7723 3573 1 5 UVET 210 13.52 x – – x x 2009 28/12
2 Pallasjä i 1 7543 3377 1 4 EMT 100 17.85 x – – x x 2003 28/28
3 Pallasjä i 1 7549 3384 2 3 HMT 150 15.42 x x x x – 2003 28/28
4 Sodankylä 1 7472 3485 1 4 EMT 80 19.66 x – – x x 2003 28/28
5 Ki alo 1 7360 3484 2 3 HMT 80 17.95 x x – x – 2002 28/28
6 Ki alo 1 7364 3488 1 4 EMT 65 18.94 x – – x x 2002 28/28
32 Ki alo 1 7371 3486 3 3 HMT 58 14.39 a a a x – 2009 28/0
21 Oulanka 1 7359 3612 2 3 HMT 180 21.03 a a a x – 2009 28/0
20 Lieksa 2 7012 3687 1 4 EVT 140 22.27 a – a x – 2009 28/0
10 Juupajoki 2 6866 3353 1 4 VT 90 23.55 x x x x – 2002 28/28
11 Juupajoki 2 6863 3359 2 2 OMT 90 30.4 x x x x – 2002 28/28
12 Tammela 2 6730 3325 2 3 MT 70 33.08 x x x x – 2002 28/28
13 Tammela 2 6727 3330 1 4 VT 70 29.27 x x x x – 2002 28/28
16 Punkaha ju 2 6854 3627 1 4 VT 90 31.95 x – x x – 2002 28/28
17 Punkaha ju 2 6858 3622 2 2 OMT 80 30.76 – – x x – 2002 28/28
33 Punkaha ju 2 6862 3622 3 1 OMaT 27 16.13 – a a – – 2009 28/0
34 Luumäki 2 6763 3515 1 5 CT 60 13.55 x – – x x 2009 28/12
35 Luumäki 2 6756 3513 2 3 MT 60 28.21 x x x x – 2009 28/12
Region (1 indica es no he n Finland, 2 indica es sou he n Finland); ee species (1 indica es Sco s pine, 2 indica es No way sp uce and 3 indica es deciduous ees); si e ype (1–5, om ich o poo
e ili y le el); o es si e ypes (abb e ia ions explained in Table 1 o Salemaa e al. (2008)); plan species g oups in biomass samples: DF indica es dwa sh ubs, GR indica es g asses, HE indica es
he bs, BR indica es b yophy es and LI indica es lichens. The “x” indica es biomass sample includes bo h abo eg ound and belowg ound (in o ganic soil laye ) pa o ege a ion, he “a” indica es sample
includes only abo eg ound ege a ion, “–” indica es plan g oup does no g ow in he plo . nindica es numbe o samples (each sized 30×30 cm2) o abo eg ound and belowg ound biomass.
The ela ionship be ween he isually es ima ed pe cen age
co e o plan species including ascula plan s, b yophy es
and lichens (p ojec ed on o he o es loo using 30×30cm2
ames) and hei li ing biomass was s udied in 18 o es
s ands ac oss Finland (Table 1). The plo s a e pa o he
In e na ional Co-ope a i e P og amme on Assessmen and
Moni o ing o Ai Pollu ion E ec s on Fo es s (ICP) in en-
si e moni o ing plo ne wo k (e.g. Me ilä e al., 2014). Al-
oge he , 28 sys ema ically selec ed biomass samples we e
aken once om each plo du ing he pe iod 2002–2009. In
o al, he sample size was 504 o 0.3×0.3m2 ege a ion plo s
ha ing de ailed biomass measu emen s by species. Vascula
plan s we e di ided in o abo eg ound (shoo s) and below-
g ound ( hizomes and oo s in o ganic laye ) pa s. The s udy
was ca ied ou a he ime o maximum biomass and ege a-
ion g ow h be ween he end o July and end o Augus . Hal
o he plo s we e loca ed in he no h ( i e in Pinus syl es is,
h ee in Picea abies and one in Be ula pubescens domina ed
s ands) and he o he hal we e in he sou h ( ou in Pinus
syl es is, ou in Picea abies and one in Be ula pendula
domina ed s ands). The si e ypes o he plo s anged om
poo o ich e ili y le el (Table 1).
In o de o p edic he ege a ion biomass, we buil linea
mixed models based on he ela i e co e age o i e unc-
ional plan g oups (dwa sh ubs, g asses, he bs, b yophy es
and lichens). The biomass o unde s o ey ege a ion was
es ima ed using models ha co ela e ege a ion co e age
wi h measu ed biomass. These models we e es ima ed o
he i e a o emen ioned main species g oups and o hei
belowg ound pa s, i applicable. These models we e es i-
ma ed sepa a ely o sou he n and no he n Finland. We used
a linea mixed model wi h plo -le el andom e ec s using
he “lme” command in he “nlme” package (Pinhei o e al.,
2012) o he R en i onmen o he es ima ion (R Co e Team,
2014).
Each model was weigh ed acco ding o he land a ea o
di e en si e ypes in sou he n and no he n Finland using
he weigh s op ion in lme. This was done o ensu e ha he
sample o unde s o ey biomass plo s did no unde es ima e
he weigh o he mos common si e ypes ( hose o medium
e ili y). The weigh ing was pe o med on land a eas based
on NFI11 da a (Yli alo, 2013). Models we e linea ised by
aking na u al loga i hms om biomass and co e age. Fo
model pa ame e s, bias co ec ion and hei unce ain ies, see
he Supplemen (Table S1, Figs. S1 and S2).
To quan i y he biomass o unde s o ey ege a ion, we
used species-speci ic co e age measu emen s o 2501 pe -
manen sample plo s om 1995 o es in en o y da a
(Mäkipää and Heikkinen, 2003). We applied colloca ed co-
k iging me hods o accoun o he co ela ion be ween
species g oups and he ea e we gene alised measu ed un-
de s o ey co e age on a 10×10 km2g id on upland soils
ac oss Finland (Bi and e al., 2008). The co-k iging me hod
was simila o ha o BEF es ima ion desc ibed abo e. Fi s ,
we de- ended all species g oup co e age (sh ubs, he bs and
g asses, lichens and b yophy es) by linea ends o la i-
ude and longi ude (using uni o m coo dina e sys em o Fin-
land, YKJ). A e ha , we es ima ed a iog ams and c oss-
a iog ams and applied co-k iging me hods o p edic co e -
age o unde s o ey co e age o Finland by using he gs a
package o R (R Co e Team, 2014) ( o R code, a iog ams
and c oss- a iog ams, see he Supplemen ). Fo all species
Geosci. Model De ., 9, 4169–4183, 2016 www.geosci-model-de .ne /9/4169/2016/
A. Leh onen e al.: Fo es soil ca bon s ock es ima es in a na ionwide in en o y 4173
Table 2. Li e u no e a es o ee and unde s o ey biomass by componen , species and egion.
Biomass componen Species Region Value (%) Re e ence
Lea es Sco s pine 27% (18–34%) Tupek e al. (2015)
Lea es No way sp uce 13% (9–15%) Tupek e al. (2015)
Lea es B oadlea ed 79% Tupek e al. (2015)
B anches Sco s pine 2% Leh onen e al. (2004)
B anches No way sp uce 1.25% Muukkonen and Leh onen (2004)
B anches B oadlea ed 1.35% Leh onen e al. (2004)
Ba k Sco s pine 0.3% Mälkönen (1977), Vi o (1955)
Ba k No way sp uce – – Mälkönen (1977), Vi o (1955)
Ba k B oadlea ed 0.01% Mälkönen (1977), Vi o (1955)
Coa se oo s Sco s pine 2% Leh onen e al. (2004)
Coa se oo s No way sp uce 1.25% Muukkonen and Leh onen (2004)
Coa se oo s B oadlea ed 1.35% Leh onen e al. (2004)
Fine oo s Sco s pine Sou h 85% Kleja e al. (2008)
Fine oo s No way sp uce Sou h 85 % Kleja e al. (2008)
Fine oo s B oadlea ed Sou h 85% Kleja e al. (2008)
Fine oo s Sco s pine No h 50% Leppälammi-Kujansuu e al. (2014)
Fine oo s No way sp uce No h 50 % Leppälammi-Kujansuu e al. (2014)
Fine oo s B oadlea ed No h 50% Leppälammi-Kujansuu e al. (2014)
Abo eg ound Dwa sh ubs – 37% This s udy
Belowg ound Dwa sh ubs – 8% This s udy, Helmisaa i e al. (2015)
Abo eg ound G asses – 33% This s udy
Belowg ound G asses – 59% Leppälammi-Kujansuu e al. (2014)
Abo eg ound He bs – 100%
Belowg ound He bs – 59% This s udy, Leppälammi-Kujansuu e al. (2014)
Abo eg ound B yophy es – 42% This s udy
Abo eg ound Lichen – 10% Kumpula e al. (2000)
g oups, we assumed sphe ical a iog ams wi h a common
ange o 100km.
2.3 Li e inpu es ima ion
To es ima e li e inpu om li ing biomass componen s o
he soil, li e u no e a es we e used (Table 2). Fo nee-
dles, we used he de ailed 10×10 km2g id li e u no e
a es epo ed by Tupek e al. (2015). Fo ine ee oo s, we
assumed ha he li e span was 1.18 yea s o sou he n Fin-
land and 2 yea s o no he n Finland (Kleja e al., 2008;
Leppälammi-Kujansuu e al., 2014). In o de o in e pola e
ine oo u no e a es o Finland, we assumed a linea de-
pendency o u no e a e and mean empe a u e sum (mean
o 1981–2010). Fo a eas wi h a empe a u e sum (5◦C
h eshold) highe han 1200◦C, a u no e a e o 85% was
used, and o a eas whe e empe a u e sums we e less han
700◦C, a a e o 50 % was used. Tu no e a es we e in e -
pola ed be ween empe a u e sums o 700 and 1200◦C (see
he Supplemen ).
Fo b anch and coa se oo li e , he cons an u no e
a es o he b anches we e used ac oss Finland (Leh onen e
al., 2004; Muukkonen and Leh onen, 2004). Fo ba k li e ,
we assumed a cons an a io o s em biomass acco ding o
Vi o (1955) and Mälkönen (1977). Fo ellings and na u al
mo ali y, we assumed ha all biomass was le o he si e,
excluding s ems o wood used and ha es ing esidues o
ene gy use. The quan i y o wood use was based on egional
o es y s a is ics (Yli alo, 2013).
The u no e a es o di e en unc ional plan g oups
we e pa ly based on li e a u e (Table 2) bu we also es i-
ma ed some a es om ou own biomass samples o unde -
s o ey ege a ion desc ibed abo e. We calcula ed he p o-
po ion o he cu en -yea g ow h ou o he o al li ing
biomass o dwa sh ub shoo s o es ima e he u no e a es
o abo eg ound plan pa s. We used he a e age alues o
deciduous and e e g een dwa sh ubs. Mos o he g asses
g owing in bo eal o es s (like Deschampsia lexuosa) a e
pe ennial, and we used he a io be ween he li ing and
dead biomass o es ima e he u no e a e o hose species.
The abo eg ound pa s o he mos he b species (like Ma-
ian hemum bi olium) a e annual, so we used 100% o hei
u no e a e. Fu he mo e, we used he numbe o annual
g ow h segmen s in he g een uppe pa s o he b yophy es
o es ima e hei a e. We assumed ha o e a long ime pe-
iod he es ima e o annual biomass g ow h co esponds o
he amoun o li e inpu o he soil om unde s o ey ege-
a ion.
www.geosci-model-de .ne /9/4169/2016/ Geosci. Model De ., 9, 4169–4183, 2016

4174 A. Leh onen e al.: Fo es soil ca bon s ock es ima es in a na ionwide in en o y
Ni ogen con en o all li e componen s was de i ed om
Koma o e al. (2007); o de ails, see he Supplemen .
2.4 Applica ion o soil ca bon models
Ca bon s ocks we e es ima ed by unning Yasso07 and RO-
MUL models in o a s eady s a e (i.e. a s a e whe e ca bon
inpu o he model equals ca bon lux due o decomposi-
ion). Fo a dynamic model, a s eady s a e is a s a e o which
he model aims wi h gi en inpu s. He e, model inpu s we e
a e age clima ic condi ions (1961–2012) and li e inpu s o
each g id poin , depending on dominan ee species and un-
de s o ey ege a ion co e age plus egional es ima es o li e
om ha es ing and na u al mo ali y (sc. o es y cen es).
I we assume ha his a e age le el o inpu s and clima e
has emained s eady o e cen u ies, hen ou soils should ap-
p oach s eady-s a e condi ions. Li e inpu s we e gi en o
hese models as 10×10 km2g id poin s and wea he da a
o ha gi en g id poin we e hen used o model inpu . Fo
Yasso07, s eady s a e was simula ed by unning he model
10000 yea s, a e which ela i e change o ca bon s ock was
less han 1: 10 000. Simila ly, o ROMUL he model was
un o housands o yea s, and hen we compa ed consecu-
i e ca bon s ocks and s eady-s a e s ock was ob ained when
ela i e change was less han 1: 10 000.
When e alua ing model esul s o Yasso07, esh woody
li e and ecen ly dead wood li e we e excluded om
model s a e a iables, while all non-woody ma e ial and hu-
mus boxes (including mo e hea ily decomposed ma e ial
om ine-wood and coa se-wood li e ) we e accoun ed as
soil ca bon s ock. This sepa a ion was done as Yasso07 slows
down decomposi ion o logs o high diame e , which in-
c eases soil s eady-s a e ca bon s ock es ima es. Da a om
Biosoil do no include dead wood masses in soil ca bon
s ocks. Fo ROMUL , we included all model s a e a iables
o he compa ison, no ing ha log size does no a ec decom-
posi ion.
2.4.1 Clima e da a
Daily wea he da a o s eady-s a e simula ions a e a ailable
as k iging es ima es since 1961 a a 10×10 km2 esolu ion
ac oss Finland (Venäläinen e al., 2005). Wea he da a o he
g id poin s on upland soils and on o es ed land, acco ding o
mul isou ce NFI, we e included (Tomppo e al., 2008).
Daily wea he p edic ions o a 10×10km2g id we e ag-
g ega ed o he Yasso07 model om 1961 o 2012. The a -
e age empe a u e, empe a u e ampli ude and p ecipi a ion
o each g id cell we e p o ided o Yasso07 o es ima e he
impac o clima e on soil ca bon s ock s eady s a es. The RO-
MUL model was d i en by mean daily empe a u e and p e-
cipi a ion o each g id cell.
2.4.2 Yasso07
The Yasso07 soil ca bon model (Tuomi e al., 2011) is
d i en by li e quan i y, li e quali y and wea he ( empe -
a u e and p ecipi a ion). This model is widely used in he
g eenhouse gas in en o ies o Eu opean coun ies (e.g. Fin-
land, No way and Swi ze land). This model has been also
success ully coupled wi h he clima e–ca bon cycle model
ECHAM5/JSBACH (Thum e al., 2011).
The Yasso07 is a simple dynamic model wi h luxes and
s a e a iables. The model has i e compa men s: acid, wa-
e , e hanol, non-soluble and humus boxes. Di ision o he
li e inpu acco ding o he li e solubili y was species spe-
ci ic and ollowed ha o Finnish GHG in en o y, based on
he appendix o he Yasso07 manual by Liski e al. (2009).
O ganic ma e lows be ween hese boxes and o he a mo-
sphe e a iably acco ding o wea he condi ions (Fig. 1). The
model builds on he assump ion ha o ganic ma e solubil-
i y de ines li e quali y, while decomposi ion a es du ing
luxes a e es ima ed nume ically, wi hou a p io i assump-
ions. The model was calib a ed using a la ge da abase o
li e and wood decomposi ion measu emen s and measu e-
men s o age ch onosequences o soil ca bon s ocks (Liski e
al., 2005, 1998; Ran aka i e al., 2012; Tuomi e al., 2011).
The Yasso07 soil ca bon model pa ame e alues we e es-
ima ed wi h Ma ko chain Mon e Ca lo (MCMC) me h-
ods o which he sou ce code is publicly a ailable h ough
he model websi e (h p://www.syke. i) (Tuomi e al., 2011).
He e, we used he Yasso07 model wi h Scandina ian pa am-
e e s (Ran aka i e al., 2012) and wi h global pa ame e s, his
being a p elimina y e sion om Tuomi e al. (2011) pa am-
e e s (p ac ically iden ical). Yasso07 was un wi h o al li e
inpu ( om ees and unde s o ey ege a ion) and also wi h
li e excluding ha o unde s o ey ege a ion. The model
was applied wi h annual ime s eps. Model has been cali-
b a ed wi h li e inpu es ima es and soil ca bon measu e-
men s down o 1m soil dep h. See he Supplemen o mo e
de ails on Yasso07.
2.4.3 ROMUL
ROMUL is a soil o ganic ma e decomposi ion model de el-
oped by Che o e al. (2001). I desc ibes he lux o o ganic
ma e h ough he decomposi ion p ocess, sepa a ed in co-
ho s o di e en li e o igins: lea es, shoo s, unks, coa se
oo s, ine oo s and g ound ege a ion. Li e om abo e he
g ound (lea es, shoo s and s ems) alls on o he o es loo ,
whe eas he oo li e decomposes in he mine al soil laye .
In he inal s ages, all decomposing ma e ends up in a com-
mon s o age o semi-s able humus esiding in he mine al soil
laye (Fig. 1). Decomposi ion a es o each coho depend on
he ni ogen and ash con en o he speci ic li e ype, and o
all coho s he soil empe a u e and wa e con en modi y he
decomposi ion a e. He e, we used species-speci ic ni ogen
Geosci. Model De ., 9, 4169–4183, 2016 www.geosci-model-de .ne /9/4169/2016/
A. Leh onen e al.: Fo es soil ca bon s ock es ima es in a na ionwide in en o y 4175
Figu e 1. Schema ic illus a ion o Yasso07 (le ) and ROMUL ( igh ) soil ca bon models. Solid a ows indica e luxes o o ganic ma e ,
while dashed a ows indica e CO2 luxes o a mosphe e.
a ios o di e en li e ac ions; hese a ios we e he same
ac oss he coun y in a gi en ac ion o he gi en species.
In his pape , we u ilised a e sion o he ROMUL model
whe e we adop decomposi ion a e unc ions depending on
soil wa e con en and he model o soil wa e dynamics as
desc ibed in Linkosalo e al. (2013). We call his e sion o
he model ROMUL (whe e he “ ” e e s o decomposi-
ion a es based on olume ic soil wa e measu es). The soil
wa e holding capaci y da a we e ob ained om digi al soil
maps o Finland (Lilja and Ne alainen, 2006). To al soil wa-
e holding capaci y da a we e ex ac ed o a 10×10km2g id,
when a ailable. To e alua e he impac o soil wa e holding
capaci y a iabili y, we also epea ed he ROMUL simula-
ion assuming a cons an soil wa e holding capaci y o he
whole simula ion a ea. As he ROMUL model seg ega es
he o ganic and mine al soil laye s, we assumed ha 20 %
o he o al soil wa e holding capaci y was in he o ganic
laye and 80% in mine al soil laye . See he Supplemen o
mo e de ails on ROMUL , he sou ce code and pa ame e
es ima es.
We applied he ROMUL model using daily ime s eps
o he en i onmen al a iables impac ing he decomposi ion.
The es ima ed annual li e inpu was e enly dis ibu ed o
each day o he yea . The ROMUL model explici ly sim-
ula es he lux o ni ogen h ough he decomposi ion p o-
cess and he e o e p oduces es ima es o mine alised N in
addi ion o he soil o ganic ma e and N s o ages and CO2
elease om he soil. All model pa ame e s, excep he de-
composi ion a e, which is dependen on soil wa e con en ,
we e aken om he Che o e al. (2001) pape . Al hough he
ROMUL model does no explici ly de ine he dep h o soil
laye s, he soil wa e laye o 1m has been assumed implic-
i ly when g a ime ic wa e con en is es ima ed. The e o e,
we assume ha decomposi ion occu s in ha same laye o
1m.
2.4.4 Biosoil–soil ca bon measu emen s
The Biosoil da ase esul ed om EU- unded Fo es Focus
moni o ing p og am and includes measu emen s o soil ca -
bon and a ious o he ecosys em a iables om 2006. Da a
consis o 521 sample plo s loca ed ac oss Finland. These
plo s we e es ablished as a subse o 3009 pe manen sam-
ple plo s om 1985 o 1986 (see Mäkipää and Heikkinen,
2003).
Biosoil sample plo s ha e adii o 11.28m (400 m2). T ees
and unde s o ey ege a ion co e age we e measu ed. Soil
ca bon samples we e aken sepa a ely om he li e and hu-
mus laye s and mine al soil laye s a 0–10, 11–20, 21–40 and
41–80cm dep h. In o al, 10 o 20 subsamples we e aken
om he o ganic laye s wi h a cylinde (diame e =60mm)
and 10 ubes (diame e =23 mm). Al e na i ely, i e spade
subsamples we e aken om he 0–10cm mine al soil laye ,
ano he i e spade subsamples we e aken om he 11–20
and 21–40cm laye s and one subsample om he 41–80cm
laye . The Vi o (1952) od pene a ion me hod was used o
quan i y he olume o s ones and boulde s in he su ace
soil laye . Se e al p ope ies we e analysed om soil sam-
ples in he labo a o y: soil ex u e, ca ion exchange capaci y
(CEC), base sa u a ion (BS), pH, o ganic ca bon and o al ni-
ogen concen a ions. Aqua egia ex ac able P, Ca, K, Mg,
Mn, Pb, Na, Ni, Fe and S we e also measu ed (De ome e
al., 2007). Soil ca bon s ocks we e es ima ed down o 1 m
dep h by ex apola ion o 40–80cm measu emen s o he
80–100cm laye .
Measu ed soil ca bon s ocks we e g ouped in o di e en
la i udinal bands o ming a g adien in Finland o which we
www.geosci-model-de .ne /9/4169/2016/ Geosci. Model De ., 9, 4169–4183, 2016
4176 A. Leh onen e al.: Fo es soil ca bon s ock es ima es in a na ionwide in en o y
compa ed soil ca bon es ima es om Yasso07 and ROMUL
models agains measu ed da a. Compa isons we e made wi h
wo di e en wid hs o la i ude bands: i s ly we used bands
o 100km and he ea e bands o 20 km. Bands we e in-
cluded in he analysis when he e we e ou o mo e Biosoil
obse a ions in a band. We also iden i ied hose Biosoil plo s
ha had been unde he Li o ina Sea 7000–8000BP (Mie -
inen, 2004; Sohlenius e al., 1996) using map p oduc s p o-
duced by he Geological Su ey o Finland (E onen, 1974).
We used oo mean squa e e o (RMSE) o ank hese model
applica ions due o hei abili y o ake in o accoun bo h ac-
cu acy and p ecision when compa ing model es ima es and
da a.
We also s udied di e ences be ween mean s and and soil
p ope ies o wo egions in sou he n Finland (Region 1 wi h
a la i ude below 60◦460N, and Region 2 wi h a la i ude be-
ween 60◦460and 62◦340N).
3 Resul s
3.1 Pe o mance o he soil ca bon models
The soil ca bon s ock es ima es by Yasso07 model us-
ing global pa ame e isa ion (Tuomi e al., 2011) we e
sys ema ically la ge han measu ed da a ac oss Finland,
whe eas hose ob ained wi h Scandina ian pa ame e isa ion
(Ran aka i e al., 2012) we e in he same magni ude as he
Biosoil da a (Fig. 2a and b). In addi ion o he ealis ic le el
o soil ca bon s ock, he model ha was based on pa am-
e e isa ion wi h he Scandina ian da a also ep oduced de-
c easing soil ca bon s ock end om sou h o no h, as dis-
played in he Biosoil da a. Excluding unde s o ey li e om
he model inpu imp o ed he ma ch be ween Yasso07 soil
ca bon s ocks simula ed using global pa ame e s and Biosoil
da a, whe eas o Yasso07 using Scandina ian pa ame e s,
he same exclusion esul ed in unde es ima ion o soil ca bon
s ocks, especially in no he n Finland (Fig. 2c and d). The
ROMUL model p edic ions gene ally ag eed wi h Biosoil
da a when soil wa e holding capaci y was aken in o ac-
coun . The inclusion o soil wa e holding capaci y in RO-
MUL in oduced high a ia ion in soil ca bon s ocks be-
ween d y and mois g id poin s in sou he n Finland (Fig. 2e).
When he ROMUL model was d i en wi h cons an soil wa-
e holding capaci y, i was unable o ep oduce dec easing
soil ca bon s ocks ac oss Finland and he model unde es i-
ma ed ca bon s ocks, especially in he sou h (Fig. 2e and ).
La ge de ia ions be ween he da a and he model es ima es
we e also seen o he la ges Biosoil soil ca bon s ocks o
he sou he nmos plo s (Fig. 2).
The ROMUL model using he soil wa e holding capac-
i y was he only model able o ep oduce he inc ease in mea-
su ed soil ca bon s ocks a he sou he nmos plo s. All o he
models p edic ed a subs an ial dec ease in he soil ca bon
s ocks o he sou he n egion, which was no obse ed in
Table 3. Mean soil and o es s and p ope ies o R1 and R2 and 1.96 imes hei s anda d e o o mean (SEM). R1 is sou h o 60◦460N, while R2 lies be ween 60◦460and 62◦340N
no he n la i ude. P alues based on a wo-sided es , whe e g oup a iances a e assumed o be di e en a iables whe e p alue< 0.05 a e in bold. Sample sizes we e 31 and 127 plo s,
o sou he n (R1) and no he n (R2) a eas, espec i ely. Fo mean empe a u e and p ecipi a ion, sample sizes we e mo e han 150 based on he Finnish Me eo ological Ins i u e (FMI)
wea he g id (Venäläinen e al., 2005).
Sand Sil Clay BS O BS M CEC O CEC M pH C :N O C :N M Sphagnum Lichen He b GDecid Sp uce Li o
(%) (%) (%) (%) (%) (cmol(+) (cmol(+) (H2O) (%) (%) (%) (m2) (%) (%) (%)
kg−1) kg−1)
R1 mean 64.92 30.08 4.98 67.2 23.26 32.23 5.27 4.23 26.41 23.11 0.3 0.86 4.43 23.45 15.55 39.81 29
R1 sem 7.31 5.39 2.64 6.08 5.5 3.22 0.54 0.12 1.25 1.57 0.53 1.05 1.83 3.29 6.87 12.01 16
R2 mean 57.76 37.09 5.15 68.8 27.72 29.06 4.91 4.17 25.13 21.6 1.51 0.45 8.56 23.9 19.28 37.1 6
R2 sem 3.34 2.55 1.2 2.29 3.28 0.87 0.43 0.05 1.56 0.79 0.98 0.26 2.18 1.43 4.58 6.09 4
p alue ( es ) 0.09 0.03 0.91 0.63 0.18 0.07 0.31 0.4 0.21 0.1 0.04 0.48 0.01 0.81 0.38 0.69 0.01
Tmean P ec ex Cd ex Pb ex K ex P ex Na ex Ni ex Ca ex Mg ex Fe ex S ex Mn TWI Al i ude SOC O SOC M
(C) (mm) (mg kg−1) (mg kg−1) (mgkg−1) (mgkg−1) (mg kg−1) (mgkg−1) (mgkg−1) (mgkg−1) (mg kg−1) (mg kg−1) (mg kg−1) (dm) (Mgha−1) (Mg ha−1)
R1 mean 4.91 580.71 0.48 47.88 1149.16 943.61 139.71 10.8 4638.39 992.52 5157.74 1435.39 459.47 6.97 588 25.82 31.65
R1 sem 0.09 6.25 0.05 3.62 138.59 80.16 14.87 1.03 1044.39 329.89 984.63 85.95 167.12 1.83 124 5 7.15
R2 mean 3.62 545.66 0.38 36.02 1225.03 940.32 131.19 11.07 4058.02 1063.51 4675.6 1380.96 537.84 8.79 1119 17.46 35.23
R2 sem 0.05 2.28 0.03 5.77 85.23 38.99 11.11 1.05 256.21 159.48 643.98 55.04 148.54 1.14 55 1.57 3.31
p alue ( es ) 0 0 0 0 0.37 0.94 0.38 0.73 0.3 0.71 0.43 0.31 0.5 0.1 0 0 0.38
He e, sand, sil and clay con en a e gi en as a pe cen age. Base sa u a ion (BS), ca ion exchange capaci y (CEC) and C: N a ios a e gi en sepa a ely o o ganic laye (O) and o mine al soil laye (M). pH is measu ed in wa e .
Abundance o g ound ege a ion gi en as pe cen age co e age. Basal a ea o ees (G) and ela i e sha es o deciduous species and No way sp uce om basal a ea. Li o indica es he p opo ions o si es ha we e unde he Li o ina
Sea 8000–7000BP. Aqua egia ex ac able K, P, Na, Ni, Ca, Mg, Fe, S and Mn uni mgkg−1. Topog aphical we ness index (TWI). Al i ude o plo s based on 10m esolu ion map laye . Amoun o ca bon in o ganic and mine al soil
ho izon.
Geosci. Model De ., 9, 4169–4183, 2016 www.geosci-model-de .ne /9/4169/2016/
A. Leh onen e al.: Fo es soil ca bon s ock es ima es in a na ionwide in en o y 4177
Figu e 2. La i udinal ends o measu ed and modelled soil ca bon s ocks ac oss Finland. The xaxis is he no h coo dina e acco ding o
Finnish YKJ sys em and yaxis is he soil ca bon s ock Mgha−1. G ey do s a e indi idual model es ima es and he ed line is a second-o de
LOESS i . Panel (a) is Yasso07 wi h Ran aka i e al. (2012) pa ame e s, while panel (b) is wi h Tuomi e al. (2011) pa ame e s. Panels (c)
and (d) a e same as panels (a) and (b), bu wi hou unde s o ey li e inpu . Panels (e) and ( ) show he ROMUL model whe e (e) includes
soil wa e holding capaci y da a and ( ) includes cons an soil wa e holding capaci y. Black do s a e means om Biosoil da a o each la i ude
band and whiske s a e 1.96 imes he s anda d e o o he mean.
measu emen s (Fig. 4). Soil p ope ies o he sou he nmos
plo s (Region 1, R1, wi h a la i ude below 60◦460N) we e
di e en compa ed o soil p ope ies o o es s u he no h
(Region 2, R2, wi h a la i ude be ween 60◦460and 62◦340N).
The sou he n coas had lowe sil con en and highe sand
con en , bu simul aneously highe ca bon s ocks in he o -
ganic laye han he o he egion (Table 3). Also unde s o ey
ege a ion di e ed be ween hese egions and he sou he n
coas had lowe Sphagnum species and he bs species co e -
age, indica ing ha soils he e we e d ie . We also es ed si e
ype dis ibu ions be ween R1 and R2 bu he p alue om a
chi-squa e es was g ea e han 0.4, indica ing ha si e e -
ili ies did no di e . This is also suppo ed by he measu ed
C: N a ios, which did no di e be ween hese egions (Ta-
ble 3). These sou he n si es we e also younge and 29% o
plo s we e unde he Li o ina Sea ∼7000–8000BP, while
he sha e o younge soils was only 6% o he sligh ly mo e
no he n egion.
www.geosci-model-de .ne /9/4169/2016/ Geosci. Model De ., 9, 4169–4183, 2016