Changes in carbon stocks of Danish agricultural

European Journal of Soil Science, September 2014, 65, 730–740
doi: 10.1111/ejss.12169
Changes in carbon stocks of Danish agricultural mineral
soils between 1986 and 2009
A . T a g h i z a d e h- T o o s i a, J . E . O l e s e n a, K . K r i s t e n s e n a, L . E l s g a a r d a,
H . S . Øs t e r g a a r d b, M . Læ g d s m a n d a,‡, M . H . G r e v e a & B . T . C h r i s t e n s e n a
a
Department of Agroecology, Aarhus University, Blichers Allé 20, DK-8830 Tjele, Denmark, and b Knowledge Centre for Agriculture, Crop
Production, Agro Food Park 15, DK-8200 Aarhus N, Denmark
Summary
To establish a national inventory of soil organic carbon (SOC) stocks and their change over time, soil was sampled
in 1986, 1997 and 2009 in a Danish nation-wide 7-km grid and analysed for SOC content. The average SOC
stock in 0–100-cm depth soil was 142 t C ha−1 , with 63, 41 and 38 t C ha−1 in the 0–25, 25–50 and 50–100 cm
depths, respectively. Changes at 0–25 cm were small. During 1986–97, SOC in the 25–50-cm layer increased
in sandy soils while SOC decreased in loam soils. In the subsequent period (1997–2009), most soils showed
significant losses of SOC. From 1986 to 2009, SOC at 0–100 cm decreased in loam soils and tended to increase
in sandy soils. This trend is ascribed to dairy farms with grass leys being abundant on sandy soils while cereal
cropping dominates on loamy soils. A statistical model including soil type, land use and management was applied
separately to 0–25, 25–50 and 50–100 cm depths to pinpoint drivers for SOC change. In the 0–25 cm layer, grass
leys added 0.95 t C ha−1 year−1 and autumn-sown crops with straw incorporation added 0.40 t C ha−1 year−1 . Cattle
manure added 0.21 t C ha−1 year−1 . Most interestingly, grass leys contributed 0.58 t C ha−1 year−1 at 25–50 cm,
confirming that inventories based only on top-soils are incomplete. We found no significant effects in 50–100 cm.
Our study indicates a small annual loss of 0.2 t C ha−1 from the 0–100 cm soil layer between 1986 and 2009.
Introduction
World soils contain the largest pool of carbon (C) in the
terrestrial environment with an estimated 2500 Pg C (0–100 cm)
including 1550 Pg of soil organic C (SOC) and 950 Pg inorganic
C (Lal, 2004). Soil OC accounts for two times more C than the
atmosphere (760 Pg C) and changes in SOC may significantly
impact atmospheric carbon dioxide (CO2 ) concentrations. Atmospheric CO2 has increased from about 285 in 1850 to 393 mg l−1
in 2012 and is currently increasing at an annual rate of about
2 mg l−1 mainly because of CO2 emissions from using fossil energy
(Blasing, 2013). However, losses of SOC derived from previous
and current land use and management remain an important source
of atmospheric CO2 , with SOC stocks continuing to decline for
decades to centuries after land has been brought into agriculture
(Johnston et al., 2009; Kirk & Bellamy, 2010). When long-term
arable soils are abandoned and left to revert to woodland, SOC may
‡ Present address: Unisense Fertilitech A/S, Tueager 1, DK-8200 Aarhus N,
Denmark.
Correspondence: A. Taghizadeh-Toosi. E-mail: Arezoo.Taghizadeh-Toosi
@agrsci.dk
Received 29 January 2014; revised version accepted 6 June 2014
730
continue to accumulate for similarly long periods (Poulton et al.,
2003).
Reliable estimates of changes in the distribution of C within
terrestrial ecosystems are crucial for projections of likely effects
of future climate changes. Changes in SOC can be estimated by
monitoring programmes, inventories based on historical records
and simulation models. Programmes for SOC monitoring involve
extensive soil sampling and C analysis but allow, when properly
designed, for rigid statistical analyses of SOC data from sites
with repeated sampling (Arrouays et al., 2012). Since changes in
SOC occur slowly, repeated samplings over decades are needed
to establish trends in SOC storage (Saby et al., 2008). Further,
the spatial variability of SOC can be considerable: therefore soil
samplings need to be sufficiently well replicated to provide reliable
estimates of SOC stocks at a given site. Costs associated with long
running national monitoring networks and with soil sampling and C
measuring campaigns can be substantial. Although modelling is less
expensive and essential for projections of future changes in SOC,
model performance and outputs need to be verified against results
from monitoring activities (Capriel, 2013).
Major uncertainties in the global C budget are associated with
changes in SOC in agricultural soils, and data from national SOC
monitoring networks are essential in reducing this uncertainty (van
© 2014 British Society of Soil Science
Soil carbon storage and management
Wesemael et al., 2011). Monitoring of SOC on agricultural land is
also important in order to isolate effects of management on SOC
stocks, recognizing that individual effects of management elements
are most often confounded. Previous analyses have shown effects
on SOC contents of grassland cultivation (Goidts et al., 2009; van
Wesemael et al., 2010), production intensity and changes in climate
(Bellamy et al., 2005), and SOC inventories have been used to
validate simulations of changes in SOC at regional and national
scales using information on land use and management (Smith
et al., 2012). However, it is essential to consider limitations arising
from confounded management effects, differences in sampling
procedures such as sampling depths, sampling intensity and time
span covered, methods used for C analysis and the statistical design
of sampling campaigns.
In Denmark, a nation-wide square grid monitoring net (termed
‘Kvadratnettet’) with sampling areas (830 areas; each 50 × 50 m2 )
fixed at a mutual distance of 7 km was established in 1986
(Østergaard, 1989). The sampling areas under agricultural use (590
areas in 1986) are hosted by farmers and subjected to their choice of
management. A given year’s management is reported by the farmer
in annual questionnaires. The present study draws on soil sampling campaigns in 1986/87, 1997/98, and 2009/10 (referred to as
1986, 1997 and 2009, respectively). All three campaigns depended
on comparable soil sampling and C analyses procedures but were
subject to advances in site location techniques and C analytical
equipment. Methodological issues associated with soil sampling
and laboratory routines were assessed in separate studies in 2009. A
preliminary report of SOC and soil nitrogen data extracted from the
1986 and 1997 sampling campaigns was issued by Heidmann et al.
(2002). Recently, Rubæk et al. (2013) published a detailed account
of changes in soil phosphorus accumulation and spatial distribution
using soils from the same sampling campaigns.
The objectives of the present study were to establish SOC stocks
and their change in agricultural mineral soils in Denmark between
1986 and 2009 and to identify impacts on SOC of individual
management elements.
Materials and methods
Denmark covers a total area of 43 095 km2 of which nearly
two-thirds is under agricultural management. Most of the agricultural land is intensively cultivated with autumn and spring sown
cereals, oilseed rape, silage maize and grass-clover leys. The climate is Atlantic coastal temperate with an annual mean temperature of 7.7∘ C, ranging from 0∘ C in January-February to 15.6∘ C
in July-August (means for 1961–90). The average annual precipitation is 712 mm. Annual potential evapotranspiration is 550 mm
with surplus precipitation in late autumn, winter and early spring.
Figure 1 shows monthly mean temperatures and precipitation
between 1986 and 2010.
Sampling campaigns in 1986 and 1997
In 1986, soil was sampled from the 590 (50 × 50 m2 ) grid areas
under agricultural use (Figure 2). At each grid area, located by
731
Figure 1 Monthly mean temperature (a) and precipitation (b) recorded
between 1986 and 2010. Dashed lines show the average values for period
1 (1986–97) and period 2 (1997–2009).
1:25 000 maps, 16 soil cores were taken along three parallel
transects with an Ultuna soil corer (diameter, 22.8 mm) at four soil
depths; 0–25, 25–50, 50–75 and 75–100 cm. These depths were
taken to represent the plough layer (0–25 cm), the main rooting
zone (0–50 cm), and drainage depth (75–100 cm). For each grid
area and soil depth, the 16 samples were mixed into one bulk
sample. The coring procedure ensured that carry-over of soil across
different profile depths was avoided. The soil was left to air-dry
and stored in the dark. Soil texture was analysed by standard sieving
and sedimentation methods (Sparks et al., 1996). The SOC analysis
is described later and for the present study, soil from 50–75 and
75–100 cm depths were combined to represent the 50–100 cm
layer.
In 1997, soil samples were collected from 336 agricultural
grid areas located < 40 m from those of the 1986 campaign. The
sampling protocol was as in 1986, but in 1997 samples were not
subject to texture analyses. Samples were air-dried and stored. Only
a few of the samples taken in 1997 from the 50–75 and 75–100 cm
soil depth were available for C analysis and these soil depths were
ignored in this study.
Soil samples retrieved in 1986 and 1997 were analysed in 1997
using ball-milled sub-samples. Total carbon (TC) content was
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
732 A. Taghizadeh-Toosi et al.
outside the acceptance range, the soils in the analytical batch were
re-analysed.
The reproducibility of the soil sampling and laboratory procedures was tested in 2009 using TC analysis data. In one study, 40
of the grid areas were re-sampled, using 16 grid cells (selected randomly a priori) not included in the regular 2009 campaign. Samples
from the three depth intervals (n = 120) were analysed in order to
evaluate variation within a grid area. In another study, 151 individual soil samples from the regular 2009 campaign (randomly selected
among grid areas and soil depths) were re-analysed for TC content repeating all steps of the laboratory procedure. This was done
to isolate the variation associated with laboratory sub-sampling,
ball-milling and analysis.
Information on land use and management
Figure 2 The location of sampling areas under agricultural use in 1986,
when the nationwide monitoring grid was established.
determined by IR analysis of the amount of CO2 produced after dry
combustion in pure O2 at 1250∘ C using a LECO-CNS 1000 analyser
(LECO Corporation, St. Joseph, MI, USA). Total carbon was taken
as soil organic carbon (SOC) unless a precedent test indicated
the presence of carbonates. When carbonate was present, this was
determined separately by a volumetric method (Sparks et al., 1996)
and SOC was taken as the difference between TC and carbonate C.
Every year, farmers hosting grid areas submitted reports with
information on the previous year’s land use and management.
This information was subsequently categorised into land use (crop)
groups: (i) grass leys, (ii) autumn sown cereals and oilseed rape with
straw removal, (iii) autumn sown cereals and oilseed rape with straw
incorporation, (iv) silage maize, spring sown cereals and oilseed
rape with straw removal, (v) spring sown cereals, oilseed rape and
maize with straw incorporation and (vi) spring sown row crops.
For management the following groups were established: (i) main
crop followed by a cover crop or under-sown with grass, (ii) soil
ploughed, (iii) cattle manure applied, (iv) pig manure applied and
(v) application of other types of organic material.
Calculations and statistics
The concentration of SOC (Cdi ) determined for individual
soil samples from a given depth interval (di ) was converted
to t C ha−1 as:
Sampling campaign in 2009
In 2009, 504 agricultural grid areas were again located by maps
and then positioned with GPS technology. This will provide a
more exact location (< 5 m) of the grid areas in future sampling
campaigns. Each 50 × 50 m2 grid area was sub-divided into 100 grid
cells (each 5 × 5 m2 ) and 16 of these (drawn randomly a priori) were
sampled to 100 cm depth using three depth intervals: 0–25, 25–50
and 50–100 cm. Samples were air-dried, ball-milled and analysed
for C as in 1997, but this time using a Thermo Flash 2000 NC
Analyser (Thermo Fisher Scientific, Delft, The Netherlands).
Validity of soil sampling and C analysis procedures
During C analysis in 1997 (including soils sampled in 1986) and
2009, four reference soils (laboratory standards) and aspartic acid
were routinely included to verify the quality of the C analysis.
One randomly chosen reference soil was included for every ten
grid samples. The quality of the C analysis was accepted when the
measured C content of the reference soils was within the ranges
indicated in Figure 3. If the C analysis of the reference soil fell
(
)
t C ha−1 = Cdi (%) × 𝜌d t m−3 × d (cm) ,
(1)
in which 𝜌d is the soil density of the depth interval d.
Guided by soil texture determined in 1986 for the 0–25 cm layer,
representative soil densities were established for each grid area, soil
depth, and sampling campaign by retrieving average soil densities
from matching soil profiles in the national soil profile database
(Krogh et al., 2003) that includes a total of 1001 profiles. The
grid areas were grouped according to soil types in the Danish soil
classification system (Table 1). Table 2 shows the soil bulk densities
used throughout.
The number of grid area samples for different soil types and
depths is shown in Table 3. The number of grid areas allowing
comparison between SOC in 1986, 1997 and 2009 was 258 for
0–25 cm and 252 for 25–50 cm; for 50–100-cm depth data were
available only for 1986 and 2009 and comparisons were based on
457 grid areas. Only data available for all depths in all sampling
years were used for 0–100 cm, thus for this integration the number
of grid areas was restricted to 252. Analysis of variance was
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
Soil carbon storage and management
733
Figure 3 Box plots of total-C (TC) contents in reference samples during the 2009 analysis campaign.
Box plots show the mean (grey line), median (black
line), interquartile range (box) and 10–90th percentile (bars). Lines below the box plots show means
and acceptance ranges for C contents in previously
analysed reference samples (including data from the
1986 and 1997 soil sampling campaigns). Note: different scales on y axes.
Table 1 Classification of Danish mineral soils with less than 10% organic
matter and less than 10% lime based on top-soil texture. Data are based on
weight percentage; soil JB number refers to the Danish soil classification
system (Madsen et al., 1992)
Soil type
Soil
JB no.
Clay
/%
Silt
/%
Fine
sand / %
Total
sand / %
Coarse sand (CS)
Fine sand (FS)
Loamy sand (LS)
Loamy sand (LS)
Sandy loam (SL)
Sandy loam (SL)
Loam (LO)
1
2
3
4
5
6
7
0–5
0–5
5–10
5–10
10–15
10–15
15–25
0–20
0–20
0–25
0–25
0–30
0–30
0–35
0–50
50–100
0–40
40–95
0–40
40–90
0–85
75–100
75–100
65–95
65–95
55–90
55–90
40–85
Table 2 Average soil bulk density (g cm−3 ) according to soil type of the
top-soil and depth in the soil profile. Soil JB number refers to the Danish
soil classification system (Madsen et al., 1992)
0–25 cm
25–50 cm
50–75 cm
Soil
JB no. g cm−3 SD n
g cm−3 SD n
g cm−3 SD n g cm−3 SD n
1
2
3
4
5
6
7
1.50
1.47
1.43
1.45
1.55
1.52
1.60
1.53
1.49
1.56
1.54
1.60
1.60
1.60
1.44
1.40
1.43
1.39
1.51
1.46
1.49
0.11
0.08
0.14
0.14
0.16
0.18
0.18
185
38
96
143
23
87
75
0.12
0.10
0.14
0.15
0.20
0.24
0.12
118
25
58
83
14
37
71
0.12
0.15
0.18
0.18
0.10
0.17
0.21
75–100 cm
79
23
19
36
3
18
51
1.53
1.53
1.58
1.57
1.86
1.70
1.67
0.17
0.08
0.11
0.19
–
0.11
0.20
56
17
8
14
1
13
47
SD, standard deviation; n, number of samples for bulk density measurement;
– , only one sample.
performed to test for changes in SOC over time for each soil
type and for all soils. The data were also analysed for effects
of land use and management factors on changes in SOC stocks.
For this purpose, we applied a statistical model that assumes that
under constant management SOC will gradually approach a steady
state equilibrium defined by soil type, land use and management.
The rate at which SOC approaches the steady state depends on
Table 3 Number of grid areas included in this study (see Calculations and
statistics section) for different soil types and profile depths
Soil type
0–25 cm
25–50 cm
50–100 cm
Coarse sand (CS)
Fine sand (FS)
Loamy sand (LS)
Sandy loam (SL)
Loam (LO)
Total
30
24
74
86
44
258
28
24
73
84
43
252
70
34
146
132
75
457
soil type. The following statistical model was therefore applied to
relate the changes in SOC over time to determining variables. The
model was applied separately to each of the depths 0–25, 25–50
and 50–100 cm.
Yis = 𝜆is + 𝜏is + 𝜂s + Eis ,
(2)
where: i = grid area number (1, 2, . . . ., p); s = sampling number
(s = 0, 1 and 2 for the 1986, 1997 and 2009 campaigns, respectively); Y is = the measured SOC in grid area i at sampling s;
Y is − 1 = the measured SOC in grid area i at sampling s − 1; 𝜆is = the
expected SOC in grid area i at sampling s when no C is added to or
removed from the soil (when 𝜏 is = 0);
) 𝛽n
⎧𝛼 + (Y
1
is−1 − 𝛼 e
⎪
⎪
⎪
⎪
𝜆is = ⎨
(
) 𝛽n
⎪𝛼 − 𝛼 − Yis−1 e 2
⎪
⎪
⎪
⎩
if the actual value is greather
than the long − term value,
i.e. if Yis−1 > 𝛼
if the actual value is smaller
than the long − term value,
i.e. if Yis−1 ≤ 𝛼,
𝜏 is = the net addition of C added by management in area i between
m
∑
campaigns; 𝜏is =
𝛾j xisj ; 𝜂 s = a constant that accounts for the
j
systematic differences between campaigns (only applied when
analysing 25–50 cm soil); 𝛼 = the soil type dependent steady
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
734 A. Taghizadeh-Toosi et al.
state equilibrium of SOC (termed long-term value); 𝛽 1 = a soil
type-dependent constant describing the exponential decrease in
SOC over time; 𝛽 1 = k𝛽 2 ; 𝛽 2 = a soil type dependent constant
describing the exponential increase in SOC over time; 𝛾 j = a
constant describing the expected change in SOC induced by a
crop or management operation j (relative to crops and management
operations not included in the model, that is non-significant effects
and crops and management operations not explicitly defined in the
model); n = the number of years between campaigns s − 1 and s;
p = the number of grid areas included (see Table 3); m = the number
of crops and management operations taken into account (crops and
management, operations were included in the model when effects
were significant at the 10% level); xisj = the number of management
operation of type j for grid area i between campaign s − 1 and
s; Eis = the random error which is expected to follow a normal
distribution (mean = 0; variance = 𝜎 2 ); and k = a quotient between
rate constants for SOC increase and SOC decrease.
To ensure alignment with assumptions on variance homogeneity,
the method of transform both sides (Kettl, 1990) was applied by
taking the logarithms of the dependent variable as well as the model
prediction. The method assumes that residuals are independent,
which was tested by calculating semivariograms. These showed
no spatial correlation of residuals. The initial parameters were
estimated with a non-linear model that minimized the residual sum
of squares. The minimizing was done iteratively by regressing the
residuals on to partial derivatives of the model with respect to the
parameters until the difference between two subsequent steps were
sufficiently small (George & Seber, 2003). This was performed
using the procedure NLIN of SAS (SAS, 2010). Significant differences were extracted by Student’s t-test at a significance level
of 10%.
The changes in SOC for different soil types in different periods
were tested for significance using pair-wise Student’s t-test adjusted
for multiplicity (14 or 7 tests in Figures 7, 8, respectively) using
Bonferoni’s method.
A Monte Carlo simulation was run in SAS to address the
uncertainty introduced by using fixed mean bulk densities. Firstly,
as standard deviation (SD) of bulk density means for individual
soil types and depths (n = 28) were similar, a weighted common
SD was calculated (0.15 g cm−3 ). Then, for each combination of
sampling points, depth and time, t SOC ha−1 was calculated 10 000
times using the relevant mean bulk density and again after adding a
random bulk density uncertainty (BDu ) from the normal distribution
BDu ∼ N(0, 0.0225). The SD on the differences between these two
estimates was 3.6–7.5 t SOC ha−1 for each depth across time and
soil types. These additional uncertainties corresponded to relative
increases of 1.3–4.1% when compared with the standard deviation
on the means of measured data.
Results and discussion
Studies of sampling and analytical procedures
Concentrations of C in soils sampled in 1986 and 1997 were
determined in 1997 with a different instrument to those analysed
Figure 4 Correlation between (a) total-C (TC) concentrations in soil from
the 16 grid cells of the regular 2009 sampling campaign and in soil
from 16 alternative grid cells sampled in the same 40 grid areas, (b) TC
concentrations in soil from the regular 2009 sampling campaign and in 151
soil samples randomly selected for re-analysis. Data are plotted for each
of the three depths ( ) 0–25 cm, ( ) 25–50 cm, and ( ) 50–100 cm. Lines
show the 1:1 relationship.
in 2009. Results for the four laboratory reference soils, included
for laboratory quality control of the C analyses in 1997 and 2009,
showed that the analyses performed in 2009 were in accordance
with those previously established in 1997 (Figure 3), substantiating
the validity of comparing SOC results across the two analytical
runs. Comparing the means of analyses of the reference soils and
aspartic acid showed only minor systematic trends; mean C contents
determined in 2009 were 1.3% (range 0.3–2.7%) larger than those
determined in 1997.
The C content in soils sampled in the 16 grid cells selected
for the regular 2009 campaign was in accordance with that of
soils retrieved in 16 alternative cells within the same grid area
(Figure 4a; paired-sample t-test; mean difference −0.036% C;
P = 0.13; n = 120; r = 0.95). Also, the C content in re-analysed
samples aligned with that found in the regular analytical run
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
Soil carbon storage and management
(Figure 4b; mean difference 0.015% C; P = 0.31; n = 151; r = 0.98).
From this we surmised that the uncertainties associated with the
sampling and laboratory procedures were sufficiently small to allow
the detection of any changes in SOC stocks between 1986 and 2009.
Reports on SOC storage in agricultural soils most often lack determinations of stone content (> 2 mm) in the sampled profiles nor
apply specific corrections for stone content in their calculations of
SOC stocks. This is also true for the three sampling campaigns
included in our study. However, previous studies associated with
the Danish Soil Profile Database found that soil fractions > 2 mm
accounted for < 5% of the soil mass (M. H. Greve; unpublished
results). This is generally acknowledged as a small stone content.
We therefore considered that the uncertainty associated with variations in stone content within a grid area would not jeopardize the
calculations of SOC stocks and their change over time. To retain
transparency in our study, we refrained from applying a general 5%
reduction to the calculated SOC stocks as the impact of correcting for the volume occupied by stones differs for soil depths with
different concentrations of C in the < 2 mm soil.
We used one set of soil bulk densities for each soil type and soil
depth to calculate SOC stock at the three samplings. These bulk
densities were derived from a large data-set of soil profiles representative for Danish arable soils. Although individually measured
and stone-corrected bulk densities are preferred when calculating
SOC stocks, Schrumpf et al. (2011) concluded that C concentrations were of greater importance than bulk density for the variability
in SOC stocks. Sensitivity analyses by Monte Carlo simulation indicated that standard errors in the SOC inventories (Figures 5, 6) could
be under-estimated by a factor of up to 1.04, which was interpreted
as a modest additional uncertainty of using fixed bulk densities.
Overall storage and distribution of SOC
Between 1986 and 2009, the amount of SOC stored in the top-soil
(0–25 cm) averaged 63 t C ha−1 , ranging from 78 t C in the coarse
sandy (CS) soil to 54 t C in the sandy loam (SL) soil (Figure 5).
Similar SOC stocks have been reported for agricultural mineral
top-soils in Finland (Heikkinen et al., 2013), northern Belgium
(Meersmans et al., 2009) and southeast Germany (Wiesmeier et al.,
2012) despite the substantial climatic gradient and the different
geological origin of the soils. This suggests that mineral top-soils
under long-term agricultural use may approach comparable SOC
stocks, probably defined by similar management intensity, similar
choice of crop types and comparable rotation schemes.
The 25–50 cm soil layer contained 41 t C ha−1 (Figure 5), ranging
from 48 t C in the CS soil to 36 t C in the SL soil. The 50–100 cm
layer held an almost equal amount of SOC (Figure 6; 38 t C ha−1 )
with small differences among soil types. The overall stock of
SOC in the top 100 cm was 142 t C ha−1 , ranging from 155 t C
in CS soil to 125 t C in the SL soil. A previous inventory of
SOC in the 0–100-cm layer of Danish soils (Krogh et al., 2003)
found that agricultural soils held 140 t C ha−1 , while wetland soils,
forest soils and soils under natural vegetation stored 356, 169 and
144 t C ha−1 , respectively. Somewhat smaller estimates have been
735
Figure 5 (a–c) Soil organic carbon (SOC) stock in 0–25, 25–50 and
0–50-cm depths of different soil types sampled in 1986, 1997 and 2009
(CS, coarse sand; FS, fine sand; LS, loamy sand; SL, sandy loam; LO, loam).
Error bars are standard error.
reported for cropland (90 t C ha−1 ), grassland (118 t C ha−1 ) and
forest soils (98 t C ha−1 ) in southeast Germany (Wiesmeier et al.,
2012), mainly because there is less SOC below the top-soil (A
horizon). Meersmans et al. (2009) also found less SOC stored in
the 0–100-layer under cropland and grassland in northern Belgium
(86 and 108 t C ha−1 , respectively), while another study of North
Belgian soils (Mestdagh et al., 2009) reported SOC storage ranging
from 111 to 163 t C ha−1 in grasslands (excluding grasslands on
polders and dunes). For Scottish arable soils, Chapman et al. (2013)
reported SOC stocks of 111 t C ha−1 .
Most regional and national inventories of SOC storage in agricultural soils are restricted to SOC in top-soils (plough layer).
Although the effects of cultivation are expected to affect mainly this
soil layer, it is important to note that SOC in the 0–25 cm accounts
for less than half the quantity of SOC stored in the entire soil profile.
The SOC in the deeper layers may also show long-term C changes
following the cultivation of native soils as a substantial part of the
rooting system of plants extends below the plough layer and because
of vertical transport of SOC down the soil.
Changes in SOC storage between 1986 and 2009
Averaged across all soil types, there was no significant change
between 1986 and 2009 in the amount of SOC stored in 0–25 cm
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
736 A. Taghizadeh-Toosi et al.
(a)
(a)
(b)
(b)
(c)
(c)
Figure 6 (a–c) Soil organic carbon (SOC) stock in 0–50, 50–100 and
0–100-cm depths of different soil types sampled in 1986 and 2009 (CS,
coarse sand; FS, fine sand; LS, loamy sand; SL, sandy loam; LO, loam).
Error bars are standard error.
Figure 7 (a–c) Changes in SOC stocks in 0–25, 25–50 and 0–50-cm
depths from 1986 to 1997 (period 1) and from 1997 to 2009 (period 2) for
different soil types (CS, coarse sand; FS, fine sand; LS, loamy sand; SL,
sandy loam; LO, loam).
layer (Figure 7), and there was no systematic trend when comparing
changes observed for the periods 1986–97 and 1997–2009. In contrast, SOC in the 25–50 cm depth increased significantly between
1986 and 1997 in sandy soils (CS, fine sand (FS) and LS) while
it decreased significantly in the loam soil (LO). In the subsequent
period (1997–2009), all soils (except the FS) had a significant loss
of SOC. It may be speculated that the greater mean temperature
between 1997 and 2009 than between 1986 and 1997 (Figure 1)
may have promoted greater microbial activity and caused enhanced
losses of CO2 from the soil. Climate warming is likely to increase
SOC decay in the humid boreal zone. According to Smith et al.
(2007) and Kirk & Bellamy (2010), climate change accounted for at
most 10–20% of the observed SOC changes in England and Wales
between 1978–2003, while van Wesemael et al. (2010) found no
significant effect of climate trends between 1960–2006 on SOC
changes observed for Belgian agricultural soils.
Most SOC inventories are based solely on measurements in the
top-soil, and it is noteworthy that changes in SOC were substantially
larger in the 25–50-cm than in the 0–25-cm layer. Inventories based
only on changes in top-soil SOC may therefore be inconclusive
and even misleading (Chapman et al., 2013). Averaged across all
soil types, the 25–50-cm layer accumulated 0.27 t C ha−1 year−1
during 1986–97 and then lost 0.57 t C ha−1 year−1 between 1997
and 2009. Changes in the SOC present in the 0–50-cm layer were
dominated by the changes observed at 25–50 cm. The SOC at
0–50 cm decreased significantly between 1997–2009 when, for all
soils, the loss corresponded 0.59 t C ha−1 year−1 . Although we can
offer no explanation for the different trends observed during the
two periods, our results demonstrate that estimates of changes in
SOC stocks should be consolidated by being based on more than
two sampling campaigns decades apart. Seen over the entire period
1986–2009, the most dramatic change was a decrease in SOC
present in the LO soil corresponding to an average annual loss of
0.86 t C ha−1 from the 0–50 cm layer (Figure 8). The average annual
loss of SOC from the SL soil was 0.32 t C ha−1 whereas the CS and
FS soils gained similar quantities of SOC.
Averaged across all soils, the amount of SOC at 50–100 cm
remained almost constant between 1986 and 2009 although in the
LO soil it tended to decrease and in sandy soils (CS, FS and LS) it
tended to increase (Figure 8).
For the entire profile (0–100 cm), the LO soil lost SOC at a
mean annual rate of 1.24 t C ha−1 while in the sandy soils SOC
tended to increase by 0.16–0.59 t C ha−1 year−1 . However, when
integrating data for all soils, we found a small and non-significant
mean annual reduction of 0.20 t C ha−1 over the entire period
1986–2009 (Figure 8). Similar losses have been observed in other
European countries (Meersmans et al., 2009; Mestdagh et al., 2009;
Heikkinen et al., 2013) and in a compilation of SOC inventories
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
Soil carbon storage and management
737
Table 4 SOC in 0–25 cm depth: estimated model parameters with standard
errors (SE) for effects of land use and management on changes in soil organic
carbon (SOC) between 1986 and 2009. Results are shown for significant and
non-significant effects. Model parameters; see Calculations and statistics
section.
(a)
Parameter
Estimate
Significant effects (P ≤ 0.10)
𝛼
41.3
𝛽 sand
−0.02
𝛽 loam
−0.05
k
0.51
Grass
0.95
Autumn sown crops,
0.40
straw incorporation
Cattle manure
0.21
Non-significant effects (P > 0.10)
Autumn sown crops,
0.01
straw removed
Spring sown crops,
−0.22
straw incorporation
Spring sown crops,
−0.12
straw removed
Cover crop
0.12
Ploughing
−0.13
Pig manure
0.07
Other organic manures
0.27
(b)
(c)
Figure 8 (a–c) Changes in SOC stocks in 0–50, 50–100 and 0–100-cm
depths from 1986 to 2009 for different soil types (CS, coarse sand; FS, fine
sand; LS, loam sand; SL, sandy loam; LO, loam).
covering European (EU 25) croplands (Ciais et al., 2010). The
tendency for SOC to increase in sandy soils and decrease in loamy
soils may be linked to different land uses. Dairy farms with frequent
perennial grass-clover leys in their rotations are abundant on Danish
sandy soils whereas loamy soils are dominated by spring and
autumn sown cereals and other cash crops, often combined with
straw removal for energy purposes.
Effect of land use and management on SOC
Autumn sown crops (cereals and oilseed rape) with straw incorporation increased SOC in the 0–25 cm layer by 0.40 t C ha−1 year−1
when compared to crops not included in the model (Table 4). Annual
incorporation of straw may provide a dry matter input of 5 t ha−1 ,
corresponding to 2.3 t C ha−1 for a typical straw with a C concentration of 45%. Field experiments operated over decades have
demonstrated that 14% of the straw-derived C is retained in the
plough layer (Thomsen & Christensen, 2004). Thus, straw incorporation may lead to a SOC increase of 0.32 t C ha−1 year−1 , which
aligns with our estimates. However, spring sown crops with straw
incorporation did not provide a significant effect on SOC in the
0–25 cm layer, indicating that the effect of straw incorporation
probably interacts with effects related to the crop planting time. This
may be because of greater inputs of non-harvestable crop residues
(including roots) in autumn-sown crops than in spring-sown crops
(Chirinda et al., 2012).
SE
Unit
2.7
0.01
0.02
0.23
0.36
0.17
t C ha−1
Year−1
Year−1
None
t C ha−1 year−1
t C ha−1 year−1
0.12
t C ha−1 year−1
0.15
t C ha−1 year−1
0.33
t C ha−1 year−1
0.20
t C ha−1 year−1
0.36
0.11
0.09
0.30
t C ha−1 year−1
t C ha−1 year−1
t C ha−1 year−1
t C ha−1 year−1
Cattle manure (mainly in the form of cattle slurry) increased SOC
by 0.21 t C ha−1 year−1 in the 0–25 cm layer (Table 4). Assuming
a fresh-weight application rate of 30 t ha−1 year−1 , a dry matter
content of 10% and a C concentration in the dry matter of 30%,
this would be equivalent to an annual application of 0.9 t C ha−1 .
Provided that 30% of the slurry C is retained over decades (Witter,
1996; Kätterer et al., 2008), application of cattle manure would
represent a SOC increase of 0.3 t C ha−1 year−1 .
The most conspicuous effect extracted from our dataset was
related to the presence of grass leys. These significantly increased
SOC by 0.95 t C ha−1 year−1 in the 0–25 cm layer and by
0.58 t C ha−1 year−1 in the 25–50 cm layer (Tables 4, 5). The
increase in SOC in 0–25 cm is in accordance with an annual
increase of 1.1 t C ha−1 year−1 in the 0–20 cm soil layer of a field
experiment with 1–6 years old grass leys (Christensen et al.,
2009). Several studies have shown that grasslands accumulate
SOC: possible reasons include greater above- and below-ground
residue inputs, lack of soil tillage operations and larger livestock
densities with intense inputs of animal manure (Goidts & van
Wesemael, 2007). For the deepest soil layer (50–100 cm) effects
of land use and management on SOC were non-significant. However, grass leys tended to increase SOC in the 50–100 cm layer
(Table 6). Including this effect, the presence of grass leys promoted
an average annual increase of 1.65 t C ha−1 in the entire soil profile.
The model estimates show some differences for sandy soils (CS,
FS and LS) and loamy soils (SL and LO). For the 0–25-cm layer
the rate of change in SOC was a factor two larger for the loamy soils
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
738 A. Taghizadeh-Toosi et al.
Table 5 Soil organic carbon at 25–50 cm depth: estimated model parameters with standard errors (SE) for effects of land use and management on
changes in SOC between 1986 and 2009. Results are shown for significant
and non-significant effects. Model parameters; see Calculations and statistics section
Table 6 Soil organic carbon at 50–100 cm depth: estimated model parameters with standard errors (SE) for effects of land use and management
on changes in SOC between 1986 and 2009. All management effects are
non-significant. Model parameters; see Calculations and statistics section
Parameter
Parameter
Estimate
Significant effects (P ≤ 0.10)
47.3
𝛼 sand
𝛼 loam
34.6
𝛽
−0.05
k
1
𝜂
−7.28
Grass
0.58
Non-significant effects (p > 0.10)
Autumn sown crops,
0.02
straw incorporated
Autumn sown crops,
−0.18
straw removed
Spring sown crops,
0.21
straw incorporated
Spring sown crops,
0.09
straw removed
Cover crop
−0.08
Ploughing
0.10
Cattle manure
−0.09
Pig manure
0.03
Other organic manures
0.15
SE
Unit
2.4
2.0
0.01
0.00
0.88
0.31
t C ha−1
t C ha−1
Year−1
None
None
t C ha−1 year−1
0.19
t C ha−1 year−1
0.19
t C ha−1 year−1
0.39
t C ha−1 year−1
0.23
t C ha−1 year−1
0.42
0.21
0.12
0.10
0.34
t C ha−1 year−1
t C ha−1 year−1
t C ha−1 year−1
t C ha−1 year−1
t C ha−1 year−1
than the sandy soils (Table 4). For the 25–50 cm layer the reference
steady state equilibrium content of SOC was greater for sandy
than for loamy soils (Table 5). This probably reflects that most of
the coarse sand soils in the western part of Denmark are located
on outwash plains generated during the Weichsel glaciations and
subject to prolonged water-stagnation. As with other sandy soils,
the coarse sand soils were subsequently exposed to podsolization.
The spodic horizons Bh and Bhs , enriched in stable C (Schmidt
et al., 2000; Vejre et al., 2003) co-precipitated with aluminium, iron
and manganese, became part of the top-soil when the soils were
converted to agricultural land. In the model estimates, relict C in the
sandy soils will result in either a slower SOC turnover rate or greater
steady state equilibrium of SOC. The faster turnover rate estimated
for SOC in 50–100 cm of sandy soils than in loamy soils (Table 6)
remains unexplained. However, one contributing factor could be
a very small effect of management in this soil depth, providing
uncertain estimates of changes in SOC.
The large proportion of the area in Denmark under agricultural
management means that the agricultural sector remains the primary
manager of the national SOC stock. It has been estimated that 69%
of the SOC at 0–100 cm resides under agricultural land (Krogh
et al., 2003). This is very different from the global picture where
agricultural soils are estimated to account for only 6% of the total
SOC pool (Krogh et al., 2003). Although we were able to isolate
some management elements of quantitative importance to SOC storage, our study also shows the difficulties in using the information
Estimate
Significant effects (P ≤ 0.10)
𝛼
29.0
𝛽 sand
−0.22
𝛽 loam
−0.05
k
0.74
Non-significant effects (P > 0.10)
Grass
0.12
Autumn sown crops,
−0.02
straw incorporated
Autumn sown crops,
−0.04
straw removed
Spring sown crops,
0.01
straw incorporated
Spring sown crops,
0.04
straw removed
Cover crop
0.07
Ploughing
−0.08
Cattle manure
0.01
Pig manure
0.00
Other organic manures
−0.07
SE
Unit
2.0
0.01
0.02
0.32
t C ha−1
Year−1
Year−1
None
0.17
0.17
t C ha−1 year−1
t C ha−1 year−1
0.14
t C ha−1 year−1
0.30
t C ha−1 year−1
0.16
t C ha−1 year−1
0.36
0.11
0.06
0.07
0.21
t C ha−1 year−1
t C ha−1 year−1
t C ha−1 year−1
t C ha−1 year−1
t C ha−1 year−1
provided by the farmers on land use and management. In practical
farming situations, management operations with positive and negative impact on SOC storage are often implemented simultaneously,
and the information provided by the farmer may be not sufficiently
detailed to isolate effects of individual managements by statistical
analyses.
It is well established that changes in SOC contents of temperate
agricultural soils occur slowly and are experimentally verifiable
only over extended periods. Therefore, field experiments with long,
continued treatments represent a most valuable additional source
of information on the impact of individual management elements
on SOC development (Johnston et al., 2009), and such experiments
provide unique data-sets for testing models that simulate SOC
dynamics over decades to centuries (Bruun et al., 2003). It is
reassuring that our estimates of the impact of selected management
on SOC based on monitoring farmers’ fields align well with
estimates extracted from long-term field experiments. This provides
confidence in model estimates that have been calibrated against
results from long-term field experiments.
Conclusions
The overall land use and management intensity did not change
substantially during our study period stretching over only two
decades. Accordingly, we found only a small overall loss of SOC
(0.2 t C ha−1 year−1 from 0 to 100 cm). Some distinct changes in
SOC storage emerged when we considered individual soil types,
soil depths and soil sampling campaigns, but we were able to
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
Soil carbon storage and management
isolate significant effects of only a few management elements.
The most conspicuous trends were gains of SOC in sandy soils
and loss of SOC from loamy soils. These trends are ascribed
to differences in land use; dairy farms with frequent grass leys
being abundant on sandy soils and cereal cropping with straw
removal dominating on loamy soils. This is supported by our
model estimates on effects of management. Grass leys added
0.95 t C ha−1 year−1 to SOC at 0–25 cm and, most interestingly, also
0.58 t C at 25–50 cm. In addition to this, cattle manure was found
to contribute 0.21 t C ha−1 year−1 in the 0–25 cm layer.
The importance of reliable SOC inventories has long been
recognized. However, a large number of previous studies have relied
only on SOC present in the top-soil, even though the sub-soil usually
accommodates more than 50% of the SOC in the entire profile.
In this respect, our study has clearly shown that changes in the
25–50 cm soil layer can be significant within decades. The setup
of the national square grid net and the SOC database accumulated
so far represent a unique basis for studies on future changes on SOC
in agricultural mineral soils in Denmark.
Acknowledgements
This study was financially supported by the Danish SINKS project
and by the FP-7 EU project SmartSOIL (Grant no. 289694).
References
Arrouays, D., Marchant, B.P., Saby, N.P.A., Meersmans, J., Orton, T.G.,
Martin, M.P. et al. 2012. Generic issues on broad-scale soil monitoring
schemes: a review. Pedosphere, 22, 456–469.
Bellamy, P.H., Loveland, P.J., Bradley, R.I., Lark, R.M. & Kirk, G.H.D.
2005. Carbon losses from all soils across England and Wales 1978–2003.
Nature, 437, 245–248.
Blasing, T.J. 2013. Recent Greenhouse Gas Concentrations. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S.
Department of Energy. doi: 10.3334/CDIAC/atg.032. [WWW document].
URL http://cdiac.ornl.gov/pns/current_ghg.html [accessed on 24 January
2014].
Bruun, S., Christensen, B.T., Hansen, E.M., Magid, J. & Jensen, L.S. 2003.
Calibration and validation of the soil organic matter dynamics of the
Daisy model with data from the Askov long-term experiments. Soil
Biology & Biochemistry, 35, 67–76.
Capriel, P. 2013. Trends in organic carbon and nitrogen contents in
agricultural soils in Bavaria (south Germany) between 1986 and 2007.
European Journal of Soil Science, 64, 445–454.
Chapman, S.J., Bell, J.S., Campbell, C.D., Hudson, G., Lilly, A., Nolan, A.J.
et al. 2013. Comparison of soil carbon stocks in Scottish soils between
1978 and 2009. European Journal of Soil Science, 64, 455–465.
Chirinda, N., Olesen, J.E. & Porter, J.R. 2012. Root carbon input in organic
and inorganic fertilizer-based systems. Plant & Soil, 359, 321–333.
Christensen, B.T., Rasmussen, J., Eriksen, J. & Hansen, E.M. 2009. Soil
carbon storage and yields of spring barley following grass leys of different
age. European Journal of Agronomy, 31, 29–35.
Ciais, P., Wattenbach, M., Vuichard, N., Smith, P., Piao, S.L., Don, A. et al.
2010. The European carbon balance. Part 2: croplands. Global Change
Biology, 16, 1409–1428.
739
George, A.F. & Seber, C.J.W. 2003. Nonlinear Regression. John Wiley &
Sons Inc., Hoboken, NJ.
Goidts, E. & van Wesemael, B. 2007. Regional assessment of soil organic
carbon changes under agriculture in southern Belgium (1955–2005).
Geoderma, 141, 341–354.
Goidts, E., van Wesemael, B. & van Oost, K. 2009. Driving forces of soil
organic carbon evolution at the landscape and regional scale using data
from a stratified soil monitoring. Global Change Biology, 15, 2981–3000.
Heidmann, T., Christensen, B.T. & Olesen, S.E. 2002. Changes in Soil C and
N Content in Different Cropping Systems and Soil Types. DIAS Report,
Plant Production no. 81, Danish Institute of Agricultural Sciences, Tjele.
77-86.
Heikkinen, J., Ketoja, E., Nuutinen, V. & Regina, K. 2013. Declining
trend of carbon in Finnish cropland soils in 1974–2009. Global Change
Biology, 19, 1456–1469.
Johnston, A.E., Poulton, P.R. & Coleman, K. 2009. Soil organic matter:
its importance in sustainable agriculture and carbon dioxide fluxes.
Advances in Agronomy, 101, 1–57.
Kätterer, T., Andersson, L., Andrén, O. & Persson, J. 2008. Long-term
impact of chronosequential land use change on soil carbon stocks on a
Swedish farm. Nutrient Cycling in Agroecosystems, 81, 145–155.
Kettl, S. 1990. Accounting for heteroscedasticity in the transform both sides
regression model. Applied Statistics, 40, 261–268.
Kirk, G.J.D. & Bellamy, P.H. 2010. Analysis of changes in organic carbon in
mineral soils across England and Wales using a simple single-pool model.
European Journal of Soil Science, 61, 406–411.
Krogh, L., Noergaard, A., Hermansen, M., Greve, M.H., Balstroem, T.
& Breuning-Madsen, H. 2003. Preliminary estimates of contemporary
soil organic carbon stocks in Denmark using multiple datasets and four
scaling-up methods. Agriculture, Ecosystems & Environment, 96, 19–28.
Lal, R. 2004. Soil carbon sequestration impacts on global climate change
and food security. Science, 304, 1623–1627.
Madsen, H.B., Nørr, A.H. & Holst, K.A. 1992. Atlas of Denmark, Series I.
Volume 3: The Danish Soil Classification. The Royal Danish Geographical Society C. A. Reitzel Publishers, Copenhagen.
Meersmans, J., van Wesemael, B., de Ridder, F., Dotti, M.F., de Baets, S. &
van Molle, M. 2009. Changes in organic carbon distribution with depth
in agricultural soils in northern Belgium, 1960–2006. Global Change
Biology, 15, 2739–2750.
Mestdagh, I., Sleutel, S., Lootens, P., Van Cleemput, O., Beheydt, D.,
Boeckx, P. et al. 2009. Soil organic carbon-stock changes in Flemish
grassland soils from 1990–2000. Journal of Plant Nutrition & Soil
Science, 172, 24–31.
Østergaard, H.S. 1989. Analytical methods for optimization of nitrogen
fertilization in agriculture. In: Management Methods to Reduce Impact
of Nitrates (ed J.C. Germon), pp. 224–234. Elsevier, London.
Poulton, P.R., Pye, E., Hargreaves, P.R. & Jenkinson, D.S. 2003. Accumulation of carbon and nitrogen by old arable land reverting to woodland.
Global Change Biology, 9, 942–955.
Rubæk, G.H., Kristensen, K., Olesen, S.E., Østergaard, H.S. & Heckrath,
G. 2013. Phosphorus accumulation and spatial distribution in agricultural
soils in Denmark. Geoderma, 209-210, 241–250.
Saby, N.P.A., Bellamy, P.H., Morvan, X., Arrouays, D., Jones, R.J.A.,
Verheijen, F.G.A. et al. 2008. Will European soil-monitoring networks be
able to detect changes in topsoil organic carbon content? Global Change
Biology, 14, 2432–2442.
SAS Institute Inc., 2010. SAS/STAT 9.22 User’s Guide. SAS Institute Inc.
Electronic Version, Cary, NC.
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740
740 A. Taghizadeh-Toosi et al.
Schmidt, W.I., Knicker, H. & Kögel-Knabner, I. 2000. Organic matter accumulating in Aeh and Bh horizons of a Podzol – chemical characterization in primary organo-mineral associations. Organic Geochemistry, 31,
727–734.
Schrumpf, M., Schulze, E.D., Kaiser, K. & Schumacher, J. 2011. How
accurately can soil organic carbon stocks and stock changes be quantified
by soil inventories? Biogeosciences, 8, 1193–1212.
Smith, P., Chapman, S.J., Scott, W.A., Black, H.I.J., Wattenbach, M., Milne,
R. 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., Davies, C.A., Ogle, S., Zanchi, G., Bellarby, J., Bird, N. et al.
2012. Towards an integrated global framework to assess the impacts of
land use and management change on soil carbon: current capability and
future vision. Global Change Biology, 18, 2089–2101.
Sparks, D.L., Page, A.L., Helmke, P.A., Loeppert, R.H., Soltanpour, P.N.,
Tabatabai, M.A. et al. 1996. Methods of Soil Analysis, Part 2. American
Society of Agronomy Inc., Madison, WI.
Thomsen, I.K. & Christensen, B.T. 2004. Yields of wheat and soil
carbon and nitrogen contents following long-term incorporation of
barley straw and ryegrass catch crops. Soil Use & Management, 20,
432–438.
Vejre, H., Callesen, I., Vesterdal, L. & Raulund-Rasmussen, K. 2003.
Carbon and nitrogen in Danish forest soils – contents and distribution
determined by soil order. Soil Science Society of America Journal, 67,
335–343.
van Wesemael, B., Paustian, K., Meersmans, J., Goidts, E., Barancikova, G.
& Easter, M. 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.
van Wesemael, B., Paustian, K., Andrén, O., Cerri, C.E.P. & Dodd, M. 2011.
How can soil monitoring networks be used to improve predictions of
organic carbon pool dynamics and CO2 fluxes in agricultural soils? Plant
& Soil, 338, 247–259.
Wiesmeier, M., Spörlein, P., Geuss, U., Hangen, E., Haug, S., Reischl, A.
et al. 2012. Soil organic carbon stocks in southeast Germany (Bavaria)
as affected by land use, soil type and sampling depth. Global Change
Biology, 18, 2233–2245.
Witter, E. 1996. Soil C balance in a long-term field experiment in relation to
the size of the microbial biomass. Biology & Fertility of Soils, 23, 33–37.
© 2014 British Society of Soil Science, European Journal of Soil Science, 65, 730–740