Global Change Biology (2014) 20, 313–326, doi: 10.1111/gcb.

12292

A new baseline of organic carbon stock in European

agricultural soils using a modelling approach
E M A N U E L E L U G A T O , P A N O S P A N A G O S , F R A N C E S C A B A M P A , A R W Y N J O N E S and
LUCA MONTANARELLA
European Commission, Joint Research Centre, Institute for Environment and Sustainability, Via E. Fermi, Ispra, VA 2749
I-21027, Italy

Abstract
Proposed European policy in the agricultural sector will place higher emphasis on soil organic carbon (SOC), both as
an indicator of soil quality and as a means to offset CO2 emissions through soil carbon (C) sequestration. Despite
detailed national SOC data sets in several European Union (EU) Member States, a consistent C stock estimation at EU
scale remains problematic. Data are often not directly comparable, different methods have been used to obtain values
(e.g. sampling, laboratory analysis) and access may be restricted. Therefore, any evolution of EU policies on C
accounting and sequestration may be constrained by a lack of an accurate SOC estimation and the availability of tools
to carry out scenario analysis, especially for agricultural soils. In this context, a comprehensive model platform was
established at a pan-European scale (EU + Serbia, Bosnia and Herzegovina, Croatia, Montenegro, Albania, Former
Yugoslav Republic of Macedonia and Norway) using the agro-ecosystem SOC model CENTURY. Almost 164 000
combinations of soil-climate-land use were computed, including the main arable crops, orchards and pasture. The
model was implemented with the main management practices (e.g. irrigation, mineral and organic fertilization, till-
age) derived from official statistics. The model results were tested against inventories from the European Environ-
ment and Observation Network (EIONET) and approximately 20 000 soil samples from the 2009 LUCAS survey, a
monitoring project aiming at producing the first coherent, comprehensive and harmonized top-soil data set of the EU
based on harmonized sampling and analytical methods. The CENTURY model estimation of the current 0–30 cm
SOC stock of agricultural soils was 17.63 Gt; the model uncertainty estimation was below 36% in half of the NUTS2
regions considered. The model predicted an overall increase of this pool according to different climate-emission
scenarios up to 2100, with C loss in the south and east of the area (involving 30% of the whole simulated agricultural
land) compensated by a gain in central and northern regions. Generally, higher soil respiration was offset by higher
C input as a consequence of increased CO2 atmospheric concentration and favourable crop growing conditions, espe-
cially in northern Europe. Considering the importance of SOC in future EU policies, this platform of simulation
appears to be a very promising tool to orient future policymaking decisions.
Keywords: agriculture, CENTURY model, climate change scenario, sequestration, soil organic carbon stock, Tier 3 approach

Received 20 February 2013 and accepted 4 June 2013

categories have been adopted in the ‘Land Use, Land-

Introduction
Soil organic carbon (SOC) is the largest carbon (C)
stock in most terrestrial ecosystems (Lal, 2008a, b), con-
taining approximately 2344 Gt of organic C globally
(Stockmann et al., 2013). Moreover, soil is recognized
as the second largest C pool after the oceans and one
of the most important components of the biosphere,
delivering major ecosystem services and functions
(Ogle & Paustian, 2005). Policy frameworks influencing
land use and land use changes could trigger dramatic
changes in SOC levels. At a global scale, guidelines to
quantify changes in greenhouse gas emissions from
agricultural soils, land-use changes and forestry
Correspondence: Emanuele Lugato, tel. +39 0332 789672,
fax +39 0332 789034, e-mail: emanuele.lugato@jrc.ec.europa.eu
Use Change and Forestry’ (LULUCF) activities, listed
in articles 3.3 and 3.4 of the Kyoto Protocol (UN, 1997).
However, accounting on agricultural land (including
pasture) related to land management is not mandatory,
but rather made on a voluntary basis. These additional
activities identified under article 3.4 were not put into
a common EU framework, thus missing the opportu-
nity to offset the national balances through the imple-
mentation of C sequestration measures (Lantz et al.,
2001; Eaton et al., 2008; Smith et al., 2008; Wang et al.,
2011).
The decline in SOC is recognized as one of the eight
soil threats identified in the EU Thematic Strategy for
Soil Protection (EC, 2006, 2012a). One of the key goals
of the Soil Thematic Strategy is to maintain and
enhance SOC levels across the EU, while a recent policy

© 2013 John Wiley & Sons Ltd 313

314 E . L U G A T O et al.
document, the Roadmap to a Resource Efficient Europe
(EC, 2011a), sets the objective of increasing current lev-
els of organic matter in the EU by 2020.
Moreover, the quantification of SOC in the EU is an
important factor in the preparation of climate change
and agricultural policies (EC, 2008). The current eco-
nomic, environmental and territorial challenges in
agriculture are addressed in the post-2013 Common
Agricultural Policy (CAP), which will require the
introduction of specific ‘greening’ measures related to
agriculture practices that are beneficial for both cli-
mate and the environment (EC, 2011b, 2012b). Farm-
ers, under CAP policy, are encouraged to maintain
the agro-ecosystem by rural development measures
promoting environmentally sustainable farming prac-
tices (so called agri-environment measures) and by
sanctioning through cross-compliance (so-called Good
Agricultural and Environmental Condition – GAEC).
Through the GAEC scheme, soil erosion protection, soil
structure maintenance and soil organic matter levels
are recognized as minimum requirements to achieve a
good condition of agricultural land (EC, 2009).
Despite raising awareness on the importance of SOC,
the stock in agricultural soils at pan-European level is
still a very uncertain pool. To date, the most compre-
hensive pan-European estimation of SOC concentration
was made by Jones et al. (2005) using pedo-transfer
rules derived from the European Soil Database, an
extended Corine Land Cover data set, a digital eleva-
tion model and mean annual temperature data.
Although a robust attempt to create a recognized base-
line for 1990, uncertainties are connected to the diffi-
culty of disentangling SOC values from different land
uses and the time period covered, since the estimations
were based on input layers, such as the European Soil
Database, where some data reflect conditions from the
1960s. In 2009 the EIONET-SOIL network (European
Environment and Observation Network for Soil – a
partnership network of the European Environment
Agency (EEA, 2003) and its member and cooperating
countries) (Panagos et al., 2013a) was formally
requested to provide SOC data on a pan-European
1 km grid. However, many countries did not provide
data and, in several cases, the information provided
was neither complete nor homogeneous (e.g. varying
spatial extents, different methods were used for data
collection, laboratory analysis and spatial extrapola-
tion). As for many other inventories based on sampling
and spatial interpolation, the picture depicted is static
and any scenario analysis very difficult. In that context,
Tier 3 approaches, based on modelling using disaggre-
gated data at sub-national level and measured data set
for model validation, may be the most promising and
powerful solution (IPCC, 2003, 2006). In particular,
process-based SOC models allow for the simulation of
the mechanistic effect of the anthropogenic (land man-
agement) and natural (climate and soil) drivers, with
the possibility to assess the effect of alternative scenar-
ios.
In Europe, SOC model estimations exist at different
scales, ranging from site (Wattenbach et al., 2010),
national and sub-national studies that accurately
describe the land use and the management practices

applied (Van Wesemael et al., 2010; Alvaro-Fuentes
et al., 2011), to large scale but at lower spatial resolution
(Smith et al., 2005).
In this context, a comprehensive pan-European model
application was established, using the agro-ecosystem
SOC model CENTURY (Parton et al., 1988). SOC stocks
were calculated for approximately 164 000 combina-
tions of soil–climate–land use, including the main ara-
ble crops, orchards and pastures. The model was
implemented with the main management practices (irri-
gation, mineral and organic fertilization, tillage, etc.)
derived from official statistics maintained by EURO-
STAT (the statistical office of the European Union) and
tested with the EIONET-SOIL inventory (Panagos et al.,
2013a) and the soil sample from 2009 LUCAS survey
(Land Use/Cover statistical Area frame Survey)
(EUROSTAT, 2011), providing SOC measured data.
The aim of the exercise was to create a new baseline
of SOC in EU agricultural soils and to provide a power-
ful dynamic tool to orient future European policies for
C sequestration.
Material and methods
Model
CENTURY is a process-based model designed to simulate C,
Nitrogen (N), Phosphorous (P) and Sulphur (S) dynamics in
natural or cultivated systems, using a monthly time step
(Parton et al., 1988; Shaffer et al., 2001). The soil organic matter
submodel includes two fresh residue pools (litter) and three
SOC pools. Litter is subdivided in two pools (metabolic and
structural) and SOC into three pools (active, slow and pas-
sive). The metabolic litter pool represents easily decomposable
constituents of plant residues, while the structural litter pool
is composed of more recalcitrant, lignocellulose plant materi-
als. The three SOC pools represent a gradient in decompos-
ability. Active SOC is microbial biomass and associated
metabolites having a rapid turnover (months to years), slow
SOC has intermediate stability and turnover times (decades),
and passive SOC represents the highly processed and stabi-
lized products with longer turnover times (centuries). The
decomposition of both plant residues and SOC are assumed to
be microbially mediated with an associated loss of CO2 (as a
result of microbial respiration), which is influenced by soil
texture. Decomposition products flow into a surface microbe
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
 S O I L O R G A N I C S T O C K A T P A N - E U R O P E A N L E V E L 315
pool or one of three SOC pools. The potential decomposition
rate is reduced by multiplicative functions of soil moisture
and soil temperature and may be increased as an effect of cul-
tivation. The model is also able to simulate the water balance,
using a weekly time step, and a suite of simple plant growth
models are included to simulate C, N, P and S dynamics of
crops, grasses and trees. Plant growth and net primary pro-
duction is simulated according to the genetic potential of the
plant, temperature, and availability of moisture and nutrients.
The plant growth model determines the type and timing of
the net primary production that is allocated to the different
SOC pools.
CENTURY was selected as the most suitable model for a
pan-European SOC assessment since: (i) crop growth routines
are integrated for both herbaceous and trees crops (e.g. orch-
ard, vineyard), including the possibility to simulate mixed
systems; (ii) the model has been tested successfully in several

European countries (Lugato et al., 2006, 2007; Alvaro-Fuentes
et al., 2012a); (iii) the effect of the main management practices
in the agricultural fields are simulated (tillage, grazing, irriga-
tion, fertilization, etc.); (iv) the monthly time step reduces the
computational time when dealing with a large number of
combinations.
For this study, the model was run with the coupled C-N
sub-models.
Data sets. Unique homogenous territorial units (Soil and Cli-
mate Unit – SCU) were identified by the overlay of soil poly-
gons with a climatic data grid, according to the following
data sets. All EU Member States plus Serbia, Bosnia and
Herzegovina, Croatia, Montenegro, Albania, Former Yugoslav
Republic of Macedonia and Norway states were included in
the simulation. The total number of resulting SCU combina-
tions was 102901.
Soil data
The soil data used by the model were derived from the Euro-
pean Soil Database-ESDB (King et al., 1994) available at the
European Soil Data Centre (ESDAC). The properties consid-
ered for the top-soil layer (0–30 cm) included soil texture, bulk
density, pH, drainage class and rock content at 1 km 9 1 km
grid resolution. Those layers have been classified as integers to
an ordinal scale (Hiederer, 2013) and are available internally
for modelling activities at the ESDAC. However, to reduce the
number of soil combinations, the polygon cover named as Soil
Mapping Unit (SMU) of ESDB was used. The average value of
the overlaying raster cells was assigned to each SMU.
Although CENTURY has a simple water bucket model, the
hydraulic properties (field capacity and wilting point) were
estimated using a pedotransfer rule (Rawls et al., 1982). These
two parameters were corrected for the presence of rock
according to the factor: [1–(Rv/100)], where Rv is the rock
fragment content by volume. Data on soil depth or the pres-
ence of an impediment layer, derived from the ESDB, were
used to define the bottom boundary layer.
The CENTURY model, originally parameterized to simulate
SOC dynamics in the 0–20 cm depth, was modified to
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326

simulate the 0–30 cm layer, according to Alvaro-Fuentes et al.
(2011).
The initial SOC values (see model spin-up section for fur-
ther details) were derived from the current pan-European
SOC estimates based on Jones et al. (2005), hereafter referred
as the OCTOP model.
Climatic data
Monthly temperature and precipitation were taken from a
10′ 9 10′ cell climate data set provided by the Climate
Research Unit, University of East Anglia, UK (Mitchell et al.,
2004). Monthly values were provided for the interval 1900–
2000, based on interpolated observed data. For the period
2000–2100, values were obtained from four different Global
Climate Models (GCMs) forced by four Intergovernmental
Panel on Climate Change (IPCC) CO2 emissions scenarios, as
reported in the Special Report on Emissions Scenarios (SRES)
(Nakicenovic et al., 2000). For this study, the two contrasting
scenarios selected were HadCM3-A1FI (‘world markets-fossil
fuel intensive’) and PCM-B1 (‘global sustainability’) as they
encompass a wide range of climatic variations, the former
more extreme and the latter more conservative.
The CENTURY model can simulate the effects of increasing
atmospheric CO2 concentration (Parton et al., 1994) consider-
ing (i) the increase of Net Primary Productivity (NPP) with a
different response for C3 and C4 plant species; (ii) the transpi-
ration reduction which is supposed to happen in relation to a
decrease in stomatal conductance; and (iii) the C/N and
shoot/root ratio change of grasses and crops. A linear growth
rates in CO2 concentration to reach 954 ppmv for the A1FI sce-
nario and 540 ppmv for the B1 scenario was assumed for 2100
(Nakicenovic et al., 2000).
Land use and management
Within each SCU, the land use was defined by overlaying
either the Corine Land Cover (CLC) database for 2006 (http://
www.eea.europa.eu/publications/COR0-landcover) or for
2000 in areas where the 2006 data were not available. CLC is
one of the most comprehensive and spatially detailed data-
base of land use (100 m resolution). Since this study was ori-
ented to agricultural soils, only the thirteen classes defined as
‘agricultural areas’ were considered. For each SCU, the origi-
nal CLC classes were aggregated into seven classes and the
area (ha) for the following new categories was calculated:
arable, rice, vineyard, olive, orchard, pasture and complex
system.
This information was complemented by statistics on crop
production area for NUTS2 regions, from the EU Statistical
Office (EUROSTAT) (http://epp.eurostat.ec.europa.eu/portal/
page/portal/agriculture/data/database) to implement the
model with appropriate crop rotations for the categories ara-
ble system. Specifically, space was substituted by time,
hypothesizing a four year rotation in which each crop occu-
pies 25% of the time (equivalent to 25% of the space within
the SCU). The ‘relative’ crops distribution was calculated
and fitted into the four year rotation (the maximum level of
316 E . L U G A T O et al.
complexity considered), adopting some approximation rules
based on the proximity to the class limit. For example, if a
crop had a relative distribution of 40% (from EUROSTAT
statistics), half of the four year rotation simulated was cov-
ered by that crop. The schedules files for 18 arable/fodder
crops (barley, wheat, maize grain, silage maize, soybean,
sugarbeet, sunflower, tobacco, ryegrass, alfalfa, rice, pulses,
oilseed, rape, cotton, potato, tobacco, rice) were created to be
simulated according the partition rules created. According to
the CLC definition, complex systems were created adding
fodder crops to the arable rotation. The final number of Soil-
Climate-Land use (SCL) combinations generated was
163 924.
EUROSTAT’s Agricultural Statistic Database (http://epp.
eurostat.ec.europa.eu/portal/page/portal/agri_environmen-
tal_indicators/data/database) was used to derive spatial
information on irrigated areas, mineral N fertilization and
livestock units. Irrigated area statistics were provided at
NUTS3 level, while mineral N consumption at nation level.
In case of missing countries in EU official statistic, N con-
sumption was taken from FAO databases (http://faostat.fao.
org/). To scale the N amount from national level to each
SCU, a linear function based on the agricultural area within
the SCU was used (e.g. the more is the cultivated area
the more is the N allocated). However, since some land
uses received higher fertilization, a weight (arable = 0.8,
orchard = 0.16, pasture = 0.04) was assigned to take into
account regional distribution (supplementary material Fig.
S1). Statistic about livestock units were provided both at
NUTS3 and NUTS2 level. The more disaggregated data were
generally used when available; in case of data availability
only at NUTS2 level, they were partitioned at the lower level
(NUTS3) according to the area covered by pasture and arable
(assigning a higher weight to pasture). Livestock unit values
were converted in organic N load using species dependent
excretion coefficient (Grizzetti et al., 2007) (supplementary
material Fig. S1). In addition the C input was calculated
according to the average C : N ratio of the different type of
manures.
Due to the difficulty in defining specific tillage techniques
for every SCL, a standard management practice was imple-
mented based on a main tillage (moldboard plough) during
the autumn and a secondary tillage (more superficial) close
to the planting date. Minimum tillage techniques were not
Table 1 Spin-up sequences simulated in each SCL unit
Equil. 1 Equil. 2
Time 1700 yrs 300 yrs
Land use 3 yr (W-O-F) + pasture 4 yr (B-C-W-M) + pasture
Fertilization Org Org
Tillage intensity Low Low
Irrigation No No
W-O-F = low yield wheat–oat –fallow (‘maggese’) rotation typical of roman and middle-age agriculture.
B-C-W-M = low yield barley–clover–wheat–meadow rotation introduced in XVII–XVIII century.
Org = organic fertilization; Min = mineral fertilization.
simulated due to the difficulty of retrieving quantitative and
spatial information about this management. Indeed, the
no-till adoption at EU level (Derpsch et al., 2010) is esti-
mated to be only 1.1 Mha, which is <1% of the agricultural
land simulated.
Model spin-up
The CENTURY model, initialized with SOC values derived
from the OCTOP, was run for a long (2000 years) and complex
spin-up (Table 1). This process was necessary as (i) SOCOCTOP
values were the weighted average of several land use classes,
including forests which are richer in SOC than agricultural
fields; (ii) new SOC estimations completely independent from
initial values were aimed; and (iii) the initial SOC pool parti-
tion needed to be defined.
Some basic assumptions were necessary due to the impossi-
bility to reconstruct past land use for such a broad territory. In
particular, the main assumption was that the areas cultivated
at present, being the most fertile, were likely to have been con-
tinuously cultivated. Recent studies have highlighted exten-
sive European deforestation since 1000 BC (Kaplan et al.,
2009). The first equilibrium sequence (Equil.1), spanning
1700 years, was characterized by a typical 3 year rotation with
wheat–oats and a fallow period called ‘maggese’, that was
undertaken to recover soil fertility [‘maggese’ derives the
Latin names ‘maius’ (May), since it was the month the fallow
field was tilled]. This agricultural system was practiced by
European farmers until the introduction of the four-year rota-
tion, developed in Holland and introduced into Great Britain
the mid-1700s. This second equilibrium (Equil. 2) was charac-
terized by the presence of a N-fixing crop (clover) and by
equal amount of the rotation surface dedicated to livestock
feeding (fodder crops) and food crops (mostly cereals). The R1
sequence consisted of the actual land use in each SCU, with
some modifications related to input intensity (fertilization, till-
age, etc.), including lower crop yields and higher presence of
fodder crop in the rotation (Table 1). The R2 sequence was
based on current management and land use/crop distribution.
The evolution of crop productivity and harvest index in the
last century was based on previous experience of model appli-
cation in long-term experiments (Lugato et al., 2007). Besides
site-specific parameters, all the run time coefficients were left
unchanged.
R1 R2
1901–1960 1961–2010
Actual with more fodder crops in arable Actual
Org + low Min Org + Min
Moderate Intensive
Yes Yes
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
 S O I L O R G A N I C S T O C K A T P A N - E U R O P E A N L E V E L 317

Model validation and uncertainty analysis eurostat.ec.europa.eu/portal/page/portal/nuts_nomencla

The modelling results were validated against two indepen-
dent data sets:
1. LUCAS (Land Use/Cover statistical Area frame Survey)
direct field observations gathering fully harmonized data
on land use/cover and their changes over time in the EU,
that included a soil survey in 2009. Top-soil samples were
collected from 10% of the general survey points, thus
providing approximately 20 000 soil samples. LUCAS soil
samples were taken from all land use/land cover types, but
mainly on agricultural areas (EUROSTAT, 2011). The sam-
ples were analysed in a single ISO-certified laboratory, pro-
viding the top-soil SOC expressed in g kg
1. To convert
this concentration to a stock, an empirically derived pedo-
transfer function, developed by Hollis et al. (2012), was
used to predict bulk density in European soils. A compari-
son was made using the LUCAS points and the simulated
value of the intersected SCL unit, for the matching land use
category (arable, pasture and permanent crops). However,
to avoid the comparison between one point vs. a polygon,
data were aggregated at higher hierarchical level corre-
sponding to administrative regions (NUTS2). The same
level of aggregation (NUTS2) was adopted as the most suit-
able for the comparison of LUCAS data with OCTOP map
(Panagos et al., 2013b).
2. The EIONET-SOIL database containing SOC concentration
(g kg
1) and SOC stocks (t ha
1) for 1 km cells for the
depth range of 0–30 cm (Panagos et al., 2013a). Six coun-
tries (Table 2) provided measurements or a best ‘estimate’
(e.g. based on models) which represents an official stand-
point of the country.
The model uncertainty was quantified at NUTS2 scale, since
these territorial units are considered basic regions for the
application of regional policies by EU bodies (http://epp.
ture/introduction).
Precisely, the absolute errors (AE) were calculated in each i
NUTS2 region as
AEni ¼ Xni
Yni ; ð1Þ
where Xni and Yni are the n SOC values simulated in the n SCLi
units and measured in the overlaying LUCAS points, respec-
tively. Within each i NUTS2 region, the AEni were then used
to generate an error distribution using a bootstrap procedure
(R core team, library ‘simpleboot’), based on 5000 replicates
with resampling. The confidence interval at 95% level (2r)
was derived from the error distribution and added to the
mean error of the distribution, to obtain the final error E95i.
Finally the uncertainty was expressed as follows:
uncertainty (%) ¼ E95i =
xi  100; ð2Þ
where x i is the average of modelled SOC in each NUTS2
region.
Results
SOC stock distribution
The top-soil SOC values across Europe in 2010,
derived from the model run through all the manage-
ment sequences and agricultural land uses, are pre-
sented in Figure 1. In the Mediterranean area, almost
all values were located in the lowest two classes and
often below 40 t C ha
1, while high SOC values were
estimated for the north-eastern part of Europe (on
average between 80 and 250 t C ha
1). The interaction
between pedo-climatic and agronomic conditions
resulted in a complex SOC distribution in eastern

Table 2 Comparison between modelled SOC stock and those provided by EIONET–SOIL plus other data sources

EIONET-SOIL* Others sources CENTURY†

SOC 0–30 cm St. dev. SOC stock SOC 0–30 cm St. dev. SOC stock
Country (t C ha
1) (t C ha
1) (Mt) SOC stock (Mt) (t C ha
1) (t C ha
1) (Mt)

Bulgaria 28.0 7.7 315.2 (100%) 65.0 (62.6) 21.4 363.6

Denmark 86.4 64.9 370.6 (100%) 563–598§ 58.5 (53.8) 29.4 180.8
Italy 56.3 29.1 993.9 (57.6%) 490  122¶ 58.7 (55.1) 28.9 807.8
Netherlands 100.1 34.0 298.8 (77.3%) 286‡ 150.6 (130.5) 85.5 321.5
Poland 79.6 63.1 1752.7 (70.1%) 69.6 (57.4) 43.5 1400.9
Slovakia 45.3 30.6 122.3 (54.0%) 72.9 (65.6) 29.9 127.7
France 3260  0.87k 78.3 (70.3) 41.8 2563

*Percentage coverage of the total country area in parenthesis. Poland, the Netherlands, and Slovakia refer mainly to agricultural
soils; Italy refers to data from 10 regions; Bulgaria refers to the 0–25 cm depth.
†SOC values are arithmetic mean of all SCL units; median values are in brackets; SOC stock is the sum of the stock for all agricul-
tural land uses simulated.
‡(Kuikman et al., 2003).
§(Krogh et al., 2003).
¶(Chiti et al., 2012): the estimate refers to arable and orchard crops.
k(Martin et al., 2011): all the land uses included.

© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
318 E . L U G A T O et al.
Fig. 1 Simulated soil organic carbon (SOC) stock in the top-soil layer (0–30 cm) of European agricultural soils.
countries. Even at latitudes > 50°N, the model simu-
lated values <40 t C ha
1 in parts of Denmark, north-
ern Germany, Poland and Lithuania, which are
characterized by coarse parent material deposited
during the last glacial period (Supplementary material
Fig. S2). On the contrary, the model simulated SOC
stock ranging between 80 and 120 t C ha
1 across
Hungary and Romania, where the soils were very rich
in clay (Supplementary material Fig. S3). Some hot-
spot situations were predicted in Ireland, UK, Nether-
lands and Finland with values >250 t C ha
1, corre-
sponding to peatland areas.
The country with the highest simulated SOC stock
was France (2588 Mt), followed by the UK with 2367
and Germany with 1866 Mt (Fig. 2). Land used as pas-
ture was the dominant SOC reservoir in the UK, Ireland
and the Netherlands and also significant in France and
Germany. Permanent crops (vineyard, olive and orch-
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
ard) contributed for only 3% of the total simulated SOC
stock, with highest values of these classes being found
in Spain, Italy and France.
The arable land, which was the prevailing land use
simulated in terms of area (112.75 Mha representing
53% of the total Utilized Agricultural Area) (Fig. 2),
was predicted to store 7.65 Gt of C, corresponding to
43% of the total SOC stock. The pasture land use
contained 5.5 Gt of C.
The average top-soil SOC for the whole area consid-
ered was simulated as 82.4 t C ha
1, corresponding to
a C stock of 17.63 Gt.
Model performance evaluation
The simulated values were generally in agreement
with measurements for all three aggregated land uses
(Fig. 3). In the arable class, the median values were
 S O I L O R G A N I C S T O C K A T P A N - E U R O P E A N L E V E L 319
Fig. 2 Modelled soil organic carbon (SOC) stock (0–30 cm)
aggregated at national level for different land uses (permanent
crops = vineyard + olive + orchard; other = complex system +
rice). The embedded figure represents the distribution area
(Mha) of the agricultural systems simulated.
close to 70 t C ha
1 for both simulated (cnt) and mea-
sured (lcs) data. Instead, the initial SOC values imple-
mented into the model (OC_tp), derived from the
OCTOP data set were almost more than double.
Hence, the long equilibrium phase (2000 years) and
the following management sequences (Table 1) appear
to have driven the CENTURY model towards an equi-
librium state completely independent form the initial
values. The absolute error distribution was normal and
centred around 0, while the RMSE was 36.5 t C ha
1.
In the other land uses (pasture and permanent crops),
the model showed a similar behaviour with a higher
dispersion in the permanent crops. Modelled SOC
in pasture lands showed a median value of 126.5
t C ha
1 with a RMSE of 52.2 t C ha
1, with 59.3
t C ha
1 and a RMSE of 30.0 t C ha
1 in the perma-
nent crops.
The uncertainty calculated was <40% in half of the
NUTS2 regions (Fig. 4). Uncertainty between 60 and
80% was estimated in Netherlands, north UK and
Finland, while the highest values were shown in
Denmark and north and south eastern Italy.
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
SOC prediction
For both climatic scenarios, the CENTURY model pre-
dicted an increase in the total SOC stock from 17.63 Gt
to about 18 Gt of C until 2080 (Fig. 5). However, the
effect of the projected climate on C balance was particu-
larly marked until 2050, when SOC trends began to
diverge among the two scenarios. In particular, when
CENTURY was driven by the HAD_A1F1 scenario, the
simulated SOC began to decrease up to 2080 while,
under the PCM_B1 scenario, agricultural soils still con-
tinued to accumulate C. The average trend of both sce-
narios reached a new steady state value in the last
20 years of the simulation. The C balance clearly
showed that crop productivity NPP in the last decade
of 2100 would be higher than the in the period 1990–
2000, with median values of 609.1 and 497.7 g C m
2
respectively (supplementary material Fig. S4 and S5
showing the average NPP in the period 1990–2000 and
2090–2100 across Europe). Consequently, the C inputs
(Ci) were higher, but counterbalanced by an accelerated
soil respiration (RH).
Despite this average trend, the SOC changes varied
on a regional basis (Fig. 6). In the short to medium
term (2020), the model predicted a widespread small
increase of SOC up to 2.5 t C ha
1, but also a
decrease in some parts of southern Spain, central and
southern Italy, Bulgaria and the Czech Republic.
However, close to the end of the century, the simula-
tions depicted a dichotomist situation in Europe.
Mediterranean countries and those beyond the 20°E
meridian were characterized by a net loss of SOC,
although not of high magnitude (<2.5 t C ha
1), while
in other countries such as the UK, Ireland, France and
Germany, the model predicted a general SOC accu-
mulation.
Discussion
Model spin up
The partition of carbon pools is one of the main factors
affecting the accuracy of the model, especially when the
aim is the estimation of SOC stock. Pools with a fast-
intermediate turnover of C can reach the equilibrium
quite quickly, hence the simulation of current climatic
and management conditions for some decades may be
sufficient to reach a new steady state. Instead the
passive pool, often relevant in SOC partition under
agricultural simulations, has a turnover time of hun-
dreds to thousands of years, consequently requiring a
very long and frequently complex spin up phase. A
more robust approach would be the definition of man-
agement sequences starting from natural vegetation
320 E . L U G A T O et al.

(a) (b) (c) (d)

Fig. 3 Soil organic carbon (SOC) model comparison against LUCAS inventory. Data were averaged at NUTS2 level, for three different
land uses: arable, pasture and permanent crops. (a) distribution frequency of modelled (black) and simulated data (grey); (b) 1 : 1 com-
parisons of modelled (cnt) and measured (lsc) values; (c) boxplot values (OC_tp refers to values of OCTOP, used as initial values); (d)
absolute error distribution between modelled (cnt) and measured (lsc) values.

(Easter et al., 2007) and encompassing all the main
agro-technology or climatic changes. The limit of this
latter approach is the requirement of a large volume of
information that is necessary to reconstruct the past
land use, that makes such an approach more feasible to
limited case studies (Vaccari et al., 2012).
To encompass the long-term spin up, initialization
procedures based on Bayesian calibration have been
evaluated (Yeluripati et al., 2009). This procedure has
the advantage of associating the model uncertainty
with the initial SOC distribution, but is very time con-
suming as it requires the model to be run interactively
for thousands of iterations with a set of measured val-
ues to maximize the likelihood function. Other
approaches are based on relaxation schemes, scaling
SOC pools to match measured data. In the fast relaxa-
tion scheme (Carvalhais et al., 2008), the scaling is done
at the end of initialization, which produces an unstable
model behaviour thereafter. Conversely, the slow relax-
ation scheme, as proposed by Hashimoto et al. (2011),
enables a smoother transition from the spin up phase to
the forward model run. This approach appears promis-
ing but still requires measured data and higher compu-
tational time.
Due to the large number of SCL combinations in this
study (163 924), a complete simulation with the spin up
and the forward run, with two climatic scenarios,
currently requires more than ten days. All the initializa-

© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
 S O I L O R G A N I C S T O C K A T P A N - E U R O P E A N L E V E L 321

Fig. 4 Percentage uncertainty in soil organic carbon (SOC) stock estimates across NUTS2 regions.

(a) (b)

Fig. 5 (a) Predicted soil organic carbon (SOC) stock (Gt) trend, according to HAD3_A1FI (thin dotted line) and PCM_B1 (thick dotted
line) scenario. Blue line is the average, while grey region delimited the 2r confidence interval. (b) Distribution of the net primary
productivity (NPP), C input (Ci) and soil heterotopic respiration (RH), averaged in 1990–2000 (90) and 2090–2100 (00) decade.

tion procedure based on iteratively model runs were,
consequently, not feasible.
The common model spin up, consisting of running
the model in reverse with the current conditions, would
have been a feasible solution, leading the CENTURY
model to a steady state. However, this approach is
quite unrealistic since agricultural practices (e.g. rota-
tion, crop productivity, fertilization) have varied
strongly over time; thereby, pools with high mean resi-
dence time may not be in equilibrium, as postulated
when running backward the model with the actual con-
ditions. For this reason, a spin up scenario (Table 1)
based on the main agricultural changes that have
occurred in the past 2000 years was created, with the

© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
322 E . L U G A T O et al.
(a)
(c)
Fig. 6 Predicted soil organic carbon (SOC) stock change (t C ha
1) with respect to the actual value, in 2020 (a) and 2100 (b). Values are
the average of the two climatic scenarios (HAD3_A1Fi and PCM_B1). Figures (c) and (d) represent the corresponding uncertainty calcu-
lated as 2r.
awareness that this procedure is inevitably a compro-
mise due to the difficulty in reconstructing all the exact
land uses and managements for such a broad terri-
tory. A similar attempt to reconstruct the agricultural
management for the period 1901–2000 was made also
by Gervois et al. (2008) in their modelling exercise.
Starting from that work, Ciais et al. (2010) suggested
that modelling the effect on land use change has a neg-
ligible impact on C fluxes over the area that is currently
cropland, which are instead driven by historical agro-
technology changes.
(b)
(d)
SOC estimation in European agricultural soils
The accurate estimation and prediction of SOC in agri-
cultural soils is strongly related to the specific charac-
teristics of agro-ecosystems. The SOC balance is often
mainly driven by anthropogenic actions (Smith et al.,
2007; Lugato & Berti, 2008; Van Wesemael et al., 2010;

Alvaro-Fuentes & Paustian, 2011) rather than by
climate change. Farmers can undertake adaptation
measures through land management practices (e.g. irri-
gation, crop variety choice, land use), but also take
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
 S O I L O R G A N I C S T O C K A T P A N - E U R O P E A N L E V E L 323
advantage of climate change in some areas. For this
reason, any modelling application dealing with agricul-
tural soils should include the mechanistic effect of the
main production factor such as tillage, fertilization, irri-
gation and crop rotation since they strongly affect the C
balance. Another important modelling aspect that is
often under estimated is the relatively consistent
fraction of the passive pool in agricultural soils (Basso
et al., 2011), often more than 50% of the total simulated
SOC. This passive pool contributes very little to short-
medium term simulated ecosystem fluxes. Thus, model
validation against C budget components (e.g. NEE,
respiration) may not guarantee an accurate estimation
of SOC stock. At a European level, modelling estimates
of ecosystem fluxes, such as done for grassland by
Vuichard et al. (2007) or for natural and cultivate lands
by Zaehle et al. (2007) and Ciais et al. (2011), are easier
to find without truly assessing the absolute SOC stock
distribution, which remains a poorly understood com-
ponent of the C budget. Accurate predictions of SOC
stock still rely on a good estimation of the passive pool
that, since is not a measurable entity (Zimmermann
et al., 2007), is often dependent to a long-term spin up
or to an initial partition function (Basso et al., 2011).
The methodology presented in this article attempts to
take into account the specific characteristics of agro-eco-
systems both (i) by making a long-term spin up phase
that considers the main agricultural technological
stages, and (ii) by using a well known and calibrated
SOC model in agro-ecosystems, with spatially explicitly
management inputs, such as organic and mineral fertil-
ization, irrigation, rotation and crop distribution, resi-
dues management and tillage according to the crop
type.
Indeed, SOC modelling at EU scale is still limited,
especially for agricultural soils. Vleeshouwers & Verg-
hagen (2002) estimated average fluxes in cropland and
grassland with a simple model (CESAR), for a business
as usual and alternative scenarios considering the time-
frame 2008–2012.Their study, on a grid of 0.5°, did not
indicate absolute SOC values but only differential
fluxes.
Subsequently, Smith et al. (2005) ran the Roth-C
model in combination with C input form the Lund–
Potsdam–Jena model, on a finer grid (10′, EU-25 plus
Norway and Switzerland) and in combination with
climatic scenarios. While that paper was oriented to
show regional SOC changes, a SOC stock baseline
(1990) of about 96 t C ha
1 (equal to 6 Gt) and 87 t
C ha
1 (equal to 11 Gt) can be calculated for grass-
lands and croplands respectively. These values are in
agreement with the estimates from this study for the
same year, with 5.5 Gt for grassland and 12.1 Gt for
cropland (including orchard and complex system);
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
precisely, the geographical extent in this study was
larger than that of Smith’s (213.9 vs. 189.2 Mha
respectively) and more spatially detailed, since SCL
units simulated were an order of magnitude higher
than Smith’s work.
Gervois et al. (2008) ran the coupled vegetation-crop
model ORCHIDEE-STICS with prescribed historical
agricultural practice changes over the western part of
the EU. Despite this important work, the overall SOC
balance in European croplands may be biased by the
simulation of only maize and wheat crops.
In a comprehensive study on European cropland C
balance (EU-25 using inventory data and modelling),
Ciais et al. (2010) calculated the SOC variation by
summing the different components of the C balance
(e.g. NPP, soil respiration). Even in this case, despite
there was not a specific representation of the absolute
SOC stock but its change over time, some C compo-
nents could be compared. For instance, NPP values
simulated with CENTURY for the arable category
(485 g C m
2) were in agreement with the values
reported by Ciais, ranging from 483 to 646 g C m
2
depending on the methodology used. Although some
detailed modelling studies on SOC stock and changes

exist (Van Wesemael et al., 2010; Alvaro-Fuentes et al.,
2011, 2012b), these are generally limited to national
and sub-national areas. In these cases, the model
parameterization is certainly more detailed and accu-
rate than continental scale assessments, due to a better
knowledge of the territory being studied and local
agricultural practices. Nevertheless, the SOC stock

pattern in northeast Spain as simulated by Alvaro-Fu-
entes et al. (2011), who also used CENTURY, was sim-
ilar to the one presented in this article, as was the
total calculated stock for cropland (244 vs. 211 Mt C
respectively).
Modelling performance
Large scale simulations often report model uncertainty
related to variations in the quality of input data, but
may lack an evaluation of the results against measured
data. For the simulation presented in this article, the
output from the CENTURY model showed a good
agreement with LUCAS measured values in all the
aggregated land uses considered (Fig. 3), with no par-
ticular bias. One particular interesting aspect was that
in the first simulation of orchards, the predicted SOC
was underestimated. This bias was corrected by grow-
ing a grass cover on the orchard sites, since the model
could simulate mixed systems and their interaction.
This issue highlights the importance of C inputs pro-
vided by the grass component of this agro-ecosystem,
which cannot be neglected.
324 E . L U G A T O et al.
The current estimate of SOC stock across European
agricultural soils is a very uncertain pool and mostly
derived from national inventories, based on sampling
and data inference (geostatistics-digital soil mapping)
(Panagos et al., 2013a). Despite the heterogeneity of
the information provided by the EIONET-SOIL
(Table 2), some national figures appear comparable.
The difference between the simulated SOC stocks and
those reported by EIONET-SOIL was less than 20% in
Bulgaria, the Netherlands, Poland and Slovakia. Mar-
tin et al. (2011) reported a higher SOC stock in France,
but the estimation included also the forest land which
is approximately 30% of the area considered. In Italy,
considering the same crop categories used by Chiti
et al. (2012), the model estimation was comparable
(490 vs. 526 Mt of C respectively). A marked differ-
ence was noticed in Denmark where the CENTURY
model clearly underestimated the SOC stock, in
particular in the western part of the country. Several
factors may explain this behaviour such as: model
tendency to underestimate stock on coarse-textured
soils (Ogle et al., 2007), being the implemented texture
very sandy (>80%; supplementary material Fig. S2);
an unrepresentative spin-up since the original soils
were poorly drained and accumulating organic mat-
ter; a general inaccuracy of other model inputs (e.g.
climate, pedological characteristics, agronomy). For
instance, the abrupt SOC variations between Lithuania
and Latvia are clearly correlated to differences in
national soil texture data (Supplementary materials
Fig. S2–S3). Model uncertainty (Fig. 4) confirmed the
model bias in Denmark; however, the high uncertainty
values present in north eastern Italy were unexpected,
since CENTURY model was extensively calibrated
against data of long-term experiments present in that
area (Lugato et al., 2006, 2007). Indeed Panagos et al.
(2013b), in a recent comparison of LUCAS data
against previous SOC estimations, found a systematic
bias of LUCAS values exactly in the same regions of
north eastern Italy.
SOC predictions
Future predictions in SOC levels were made using cur-
rent land use and management practices, with two very
contrasting GCM-SRES scenarios. In this case, the scope
of the exercise was to establish the possible effects of
climate change on a future SOC baseline, without con-
sidering any adaptation or mitigation strategies.
The results suggest that, under current land manage-
ment practices, the overall SOC stock for the EU is pro-
jected to increase (Fig. 5), but with strong regional
differences involving net losses in about 30% of the sim-
ulated area (Fig. 6). The interaction of climatic parame-
ters (precipitation and temperature) and the increasing
CO2 atmospheric concentration were the main drivers
of that behaviour, through NPP and hence C input
increase. Many modelling applications, not only agri-
cultural oriented, report a similar NPP trend (Morales

et al., 2005; Smith et al., 2005; Sitch et al., 2008; Alvaro-
Fuentes et al., 2012b; Gottschalk et al., 2012). Moreover,
despite the higher vulnerability of southern and eastern
European to SOC loss, the magnitude was limited
(almost <5 t C ha
1 by 2100), suggesting that SOC stock
should be enough resilient against climate change (Za-
ehle et al., 2007; Gottschalk et al., 2012; Smith, 2012).
What is important to point out is that large margins of
uncertainty still exist in the current model application,
that are not only related to input accuracy or state of
the art SOC models structure (Ogle et al., 2010). The
impact of extreme events, such as heat waves or flood-
ing, may be difficult to be predicted and simulated.
However, these should be taken into account since heat
waves are very likely to decrease their return period in
EU, within the end of the century (IPCC, 2012). While
flooding events will likely have a less incidence than
temperature extremes, there is medium confidence that
droughts will intensify in the 21st century, especially in
central Europe and Mediterranean regions (IPCC, 2012).
Furthermore, biotic factors such as parasites prolifera-
tion, community weed change and pollination function
may be affected by a changing climate, thereby impact-
ing the agro-ecosystem functions and the C balance.
Nevertheless, the results of this study will contribute
to an improvement of the SOC stock estimation in
European agricultural soils, which is still a poorly
understood component of C balance. The process-
based modelling approach adopted appears promising,
as it is economically feasible compared to other
approaches and is continuously improvable, with local
expert knowledge and data. While SOC models are
upgraded and tested, demonstrating the ability to pre-
dict the SOC evolution in agricultural systems, many
unknown aspects of the processes are still unknown or
poorly represented by current models (Stockmann
et al., 2013). A new model generation should be inevi-
tably oriented to (i) modelling SOC pools that are
directly measurable; (ii) better understand the contro-
versial hypothesis of temperature sensitivity on pool
decomposition; and (iii) integrate the peculiarity of
modelling aspect in agro-ecosystems (e.g. process-
based effect of management practices) on land- sur-
face-ocean global circulation scheme to represent more
realistic feedbacks.
As long as these targets are not reached, the actual
model generation could be usefully implemented espe-
cially if models are well constrained and calibrated
with measurements.
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
 S O I L O R G A N I C S T O C K A T P A N - E U R O P E A N L E V E L 325
Acknowledgements
This paper is based on work carried out as preparation to
support to the European Commission DG Agriculture’s CAPR-
ESE-SOILS project (CArbon PREservation and SEquestration in
agricultural soils: Options and implications for agricultural
production).
References

Alvaro-Fuentes J, Paustian K (2011) Potential soil carbon sequestration in a semiarid
Mediterranean agroecosystem under climate change: quantifying management
and climate effects. Plant and Soil, 338, 261–272.

Alvaro-Fuentes J, Easter M, Cantero-Martinez C et al. (2011) Modelling soil organic
carbon stocks and their changes in the northeast of Spain. European Journal of Soil
Science, 62, 685–695.

Alvaro-Fuentes J, Morell FJ, Plaza-Bonilla D et al. (2012a) Modelling tillage and nitro-
gen fertilization effects on soil organic carbon dynamics. Soil and Tillage Research,
120, 32–39.

Alvaro-Fuentes J, Easter M, Paustian K (2012b) Climate change effects on organic car-
bon storage in agricultural soils of northeastern Spain. Agriculture, Ecosystems and
Environment, 155, 87–94.
Basso B, Gargiulo O, Paustian K et al. (2011) Procedures for initializing soil organic
carbon pools in the DSSAT-CENTURY model for agricultural systems. Soil Science
Society of America Journal, 75, 69–78.
Carvalhais N, Reichstein M, Seixas J et al. (2008) Implications of the carbon cycle
steady state assumption for biogeochemical modelling performance and inverse
parameter retrieval. Global Biogeochemical Cycles, 22, GB2007.
Chiti T, Gardin L, Perugini L et al. (2012) Soil organic carbon stock assessment for the
different cropland land uses in Italy. Biology and Fertility of Soils, 48, 9–17.
Ciais P, Wattenbach M, Vuichard N et al. (2010) The European carbon balance. Part 2:
croplands. Global Change Biology, 16, 1409–1428.
Ciais P, Gervois S, Vuichard N et al. (2011) Effects of land use change and manage-
ment on the European cropland carbon balance. Global Change Biology, 17, 320–338.
Derpsch R, Friedrich T, Kassam A, Li H (2010) Current status of adoption of no-till
farming in the world and some of its main benefits. International Journal of Agricul-
tural and Biological Engineering, 3, 1–25.
Easter M, Paustian K, Killian K et al. (2007) The GEFSOC soil carbon modelling sys-
tem: a tool for conducting regional-scale soil carbon inventories and assessing the
impacts of land use change on soil carbon. Agriculture, Ecosystems and Environment,
122, 13–25.
Eaton JM, Mcgoff NM, Byrne KA et al. (2008) Land cover change and soil organic car-
bon stocks in the Republic of Ireland 1851-2000. Climatic Change, 91, 317–334.
EC (2006) (COM(2006) 231 final) Communication from the Commission to the Council, the
European Parliament, the European Economic and Social Committee of the Regions. The-
matic Strategy for Soil Protection, Brussels.
EC (2008) Review of Existing Information on the Interrelations between Soil and Climate
Change (CLIMSOIL) – Final Report. Contract number 70307/2007/486157/SER/
B1:208. (ed. Schils RE), European Commission, Brussels.
EC (2009) (EC No 73\2009) Council Regulation Establishing Common Rules for Direct
Support Schemes for Farmers under the Common Agricultural Policy and Establishing
certain support Schemes for Farmers, Amending Regulations (EC) No 1290/2005, (EC)
No 247/2006, (EC) No 378/2007 and Repealing Regulation (EC) No 1782/2003. Official
Journal of the European Union, Brussels.
EC (2011a) (COM (2011) 571 final) Communication from the Commission to the European
Parliament, the Council, the European Economic and Social Committee and the Committee
of the Regions, Roadmap to a resource Efficient Europe. Official Journal of the European
Union, Brussels.
EC (2011b) (SEC (2011) 1153 final/2) Commission staff working paper. Impact assessment
Common Agricultural Policy towards 2020 accompanying the document proposals for a
Regulation of the European Parliament and of the Council. Official Journal of the Euro-
pean Union, Brussels.
EC (2012a) (COM(2012) 46 final) Report from the Commission to the European Parliament,
the Council, the European Economic and Social Committee and the Committee of the
Regions, The implementation of the Soil Thematic Strategy and on going activities. Offi-
cial Journal of the European Union, Brussels.
EC (2012b) (COM (2010) 672 final) Communication from the Commission to the Euro-
pean Parliament, the Council, the European Economic and Social Committee and the
Committee of the Regions, The CAP towards 2020: Meeting the Food, Natural
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326
Resources and Territorial Challenges of the Future. Official Journal of the European
Union, Brussels.
EEA (2003) EIONET, European Environment Information and Observation Network. Euro-
pean Environment Agency. Available at: www.eionet.europa.eu (accessed 22
October 2012).
EUROSTAT (2011) Land Use and Land Cover Area Frame Survey. Available at: http://
epp.eurostat.ec.europa.eu/portal/page/portal/lucas/methodology (accessed 10
August 2012).
Gervois S, Ciais P, De Noblet-Ducoudre N et al. (2008) Carbon and water balance of
European croplands throughout the 20th century. Global Biogeochemical Cycles, 22,
GB2022.
Gottschalk P, Smith JU, Wattenbach M et al. (2012) How will organic carbon stocks in
mineral soils evolve under future climate? Global projections using RothC for a
range of climate change scenarios. Biogeosciences, 9, 3151–3171.
Grizzetti B, Bouraoui F, Aloe A (2007) Spatialised European Nutrient Balance. Office for
Official Publications of the European Communities, Luxembourg.
Hashimoto S, Wattenbach M, Smith P (2011) A new scheme for initializing process-
based ecosystem models by scaling soil carbon pools. Ecological Modelling, 222,
3598–3602.
Hiederer R (2013) Mapping Soil Typologies - Spatial Decision Support Applied to European
Soil Database. Publications Office of the European Union, Luxembourg. 127
pp. EUR 25932 EN. Scientific and Technical Research series, ISSN 1831-9424,
doi:10.2788/87286.
Hollis JM, Hannam J, Bellamy PH (2012) Empirically-derived pedotransfer functions
for predicting bulk density in European soils. European Journal of Soil Science, 63,
96–109.
IPCC (2003) Tier level. In: Good practice guidance for land use, landuse change and for-
estry (GPG-LULUCF). (eds Penman J, Gytarsky M, Krug T, Kruger D, Pipatti R,
Buendia L, Miwa K, , Ngara T, Tanabe K, Wagner F) pp. 3.16–3.20. Ipcc-Iges,
Kanagawa.
IPCC (2006) Guidelines for National Greenhouse Gas Inventories. Institute for Global
Environmental Strategies. Cambridge University Press, Cambridge.
IPCC (2012) Managing the risks of extreme events and disasters to advance climate
change adaptation. In: A Special Report of Working Groups I and II of the Intergovern-
mental Panel on Climate Change, (eds Field CB, Barros V, Stocker TF, Qin D, Dokken
DJ, Ebi KL, Mastrandrea MD, Mach KJ, Plattner G-K, Allen SK, Tignor M, Midgley
PM), pp. 582, Cambridge University Press, Cambridge, UK, and New York, NY,
USA.
Jones RJA, Hiederer R, Rusco E et al. (2005) Estimating organic carbon in the soils of
Europe for policy support. European Journal of Soil Science, 56, 655–671.
Kaplan JO, Krumhardt KM, Zimmermann N (2009) The prehistoric and preindustrial
deforestation of Europe. Quaternary Science Reviews, 28, 3016–3034.
King D, Daroussin J, Tavernier R (1994) Development of a soil geographic database
from the Soil Map of the European Communities. Catena, 21, 37–56.
Krogh L, Noergaard A, Hermansen M et al. (2003) Preliminary estimates of contem-
porary soil organic carbon stocks in Denmark using multiple datasets and four
scaling-up methods. Agriculture, Ecosystems and Environment, 96, 19–28.
Kuikman PJ, De Groot WJM, Hendriks RFA et al. (2003) Soil carbon stocks in the
Netherlands. In: Stocks of C in Soils and Emissions of CO2 from Agricultural Soils in
The Netherlands. (ed. Alterra-Report 41) Alterra, Wageningen.
Lal R (2008a) Carbon sequestration. Philosophical Transactions of the Royal Society B:
Biological Sciences, 363, 815–830.
Lal R (2008b) Sequestration of atmospheric CO2 in global carbon pools. Energy and
Environmental Science, 1, 86–100.
Lantz A, Lal R, Kimble J (2001) Sustaining the global farm: land use effects on soil car-
bon pools in two major land resource areas of Ohio, USA. In: 10th International Soil
Conservation Organization Meeting. (eds Stott DE, Mohtar RH, Steinhardt GC),
pp. 499–502. Purdue University and the USDA-ARS National Soil Erosion
Research Laboratory, West Lafayette, IN.
Lugato E, Berti A (2008) Potential carbon sequestration in a cultivated soil under dif-
ferent climate change scenarios: a modelling approach for evaluating promising
management practices in north-east Italy. Agriculture, Ecosystems and Environment,
128, 97–103.
Lugato E, Berti A, Giardini L (2006) Soil organic carbon (SOC) dynamics with and
without residue incorporation in relation to different nitrogen fertilisation rates.
Geoderma, 135, 315–321.
Lugato E, Paustian K, Giardini L (2007) Modelling soil organic carbon dynamics in
two long-term experiments of north-eastern Italy. Agriculture, Ecosystems and Envi-
ronment, 120, 423–432.
Martin MP, Wattenbach M, Smith P et al. (2011) Spatial distribution of soil organic
carbon stocks in France. Biogeosciences, 8, 1053–1065.
326 E . L U G A T O et al.
Mitchell TD, Carter TR, Jones PD et al. (2004) A Comprehensive Set of High-Resolution
Grids of Monthly Climate for Europe and the Globe: the Observed Record (1901–2000)
and 16 scenarios (2001–2100). Working Paper 55. Tyndall Centre for Climate
Change Research, University of East Anglia, Norwich, UK.
Morales P, Sykes MT, Prentice IC et al. (2005) Comparing and evaluating process-
based ecosystem model predictions of carbon and water fluxes in major European
forest biomes. Global Change Biology, 11, 2211–2233.
Nakicenovic N, Alcamo J, Davis G et al. (2000) Special Report on Emissions Scenarios: A
Special Report of Working Group III of the Intergovernmental Panel on Climate Change.
Cambridge University Press, Cambridge, UK.
Ogle SM, Paustian K (2005) Soil organic carbon as an indicator of environmental qual-
ity at the national scale: inventory monitoring methods and policy relevance. Cana-
dian Journal of Soil Science, 85, 531–540.
Ogle SM, Breidt FJ, Easter M et al. (2007) An empirically based approach for estimat-
ing uncertainty associated with modelling carbon sequestration in soils. Ecological
Modelling, 205, 453–463.
Ogle SM, Jay Breidt F, Easter M et al. (2010) Scale and uncertainty in modeled soil
organic carbon stock changes for us croplands using a process-based model. Global
Change Biology, 16, 810–822.
Panagos P, Hiederer R, Van Liedekerke M et al. (2013a) Estimating soil organic
carbon in Europe based on data collected through an European network. Ecological
Indicators, 24, 439–450.
Panagos P, Ballabio C, Yigini Y et al. (2013b) Estimating the soil organic carbon
content for European NUTS2 regions based on LUCAS data collection. Science of
The Total Environment, 442, 235–246.
Parton WJ, Stewart JWB, Cole CV (1988) Dynamics of C, N, P and S in grassland soils:
a model. Biogeochemistry, 5, 109–131.
Parton WJ, Ojima DS, Cole CV et al. (1994) A general model for soil organic matter
dynamics: sensitivity to litter chemistry, texture and management. Quantitative
modelling of soil forming processes. Proc. symposium, 1992, 147–167.
Rawls WJ, Brakensiek CL, Saxton KE (1982) Estimation of soil water properties.
Transactions - American Society of Agricultural Engineers, 25, 1316–1320, 1328.
Shaffer MJ, Liwang M, Hansen S (2001) Modelling Carbon and Nitrogen Dynamics for
Soil Management, lewis publishers, Boca Raton, FL.
Sitch S, Huntingford C, Gedney N et al. (2008) Evaluation of the terrestrial carbon
cycle, future plant geography and climate-carbon cycle feedbacks using five
Dynamic Global Vegetation Models (DGVMs). Global Change Biology, 14,
2015–2039.
Smith P (2012) Soils and climate change. Current Opinion in Environmental Sustainability.
4, 539–544.
Smith JO, Smith P, Wattenbach M et al. (2005) Projected changes in mineral soil car-
bon of European croplands and grasslands, 1990–2080. Global Change Biology, 11,
2141–2152.
Smith P, Chapman SJ, Scott WA et al. (2007) Climate change cannot be entirely
responsible for soil carbon loss observed in England and Wales, 1978-2003. Global
Change Biology, 13, 2605–2609.
Smith P, Martino D, Cai Z et al. (2008) Greenhouse gas mitigation in agriculture.
Philosophical Transactions of the Royal Society B: Biological Sciences, 363, 789–813.
Stockmann U, Adams MA, Crawford JW et al. (2013) The knowns, known unknowns
and unknowns of sequestration of soil organic carbon. Agriculture, Ecosystems and
Environment, 164, 80–99.
UN (1997) Kyoto Protocol (KP) to the United Nations Framework Convention on Climate
Change. Conference of the Parties on Its Third Session, FCCC/CP/1997/L.7/
Add.1. 10 December, Kyoto.
Vaccari FP, Lugato E, Gioli B et al. (2012) Land use change and soil organic carbon
dynamics in Mediterranean agro-ecosystems: the case study of Pianosa Island.
Geoderma, 175–176, 29–36.
Van Wesemael B, Paustian K, Meersmans J et al. (2010) Agricultural management
explains historic changes in regional soil carbon stocks. Proceedings of the National
Academy of Sciences of the United States of America, 107, 14926–14930.
Vleeshouwers LM, Verghagen A (2002) Carbon emission and sequestration by agri-
cultural land use: a model study for Europe. Global Change Biology, 8, 519–530.
Vuichard N, Soussana JF, Ciais P et al. (2007) Estimating the greenhouse gas fluxes of
European grasslands with a process-based model: 1. Model evaluation from in situ
measurements. Global Biogeochemical Cycles, 21, GB1004.
Wang D, Hejazi M, Cai X et al. (2011) Climate change impact on meteorological, agri-
cultural, and hydrological drought in central Illinois. Water Resources Research, 47,
W09527.
Wattenbach M, Sus O, Vuichard N et al. (2010) The carbon balance of european crop-
lands: a cross-site comparison of simulation models. Agriculture, Ecosystems and
Environment, 139, 419–453.
Yeluripati JB, Van Oijen M, Wattenbach M et al. (2009) Bayesian calibration as a tool
for initialising the carbon pools of dynamic soil models. Soil Biology and Biochemis-
try, 41, 2579–2583.
Zaehle S, Bondeau A, Carter TR et al. (2007) Projected changes in terrestrial carbon
storage in Europe under climate and land-use change, 1990–2100. Ecosystems, 10,
380–401.
Zimmermann M, Leifeld J, Schmidt MWI et al. (2007) Measured soil organic matter
fractions can be related to pools in the RothC model. European Journal of Soil Sci-
ence, 58, 658–667.
Supporting Information
Additional Supporting Information may be found in the
online version of this article:
Fig. S1. Input distribution: (a) irrigation, (b) mineral nitro-
gen and (c) organic nitrogen load. High (H) and moderate
(M) irrigation intensity refers to automatic irrigation applied
by the CENTURY model, when the remaining water holding
falls below 50 and 25%.
Fig. S2. Sand content (%) of the top-soil layer as derived
from ESDAC.
Fig. S3. Clay content (%) of the top-soil layer as derived
from ESDAC.
Fig. S4. Modelled total agricultural NPP (Net Primary
Productivity), averaged for the period 1990–2000.
Fig. S5. Modelled total agricultural NPP (Net Primary
Productivity), averaged for the period 2090–2100.
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 313–326