Forage production in rotational systems generates similar yields compared to maize monocultures but improves soil carbon stocks

European Journal of Agronomy 97 (2018) 11–19

Contents lists available at ScienceDirect

European Journal of Agronomy journal homepage: www.elsevier.com/locate/eja

Forage production in rotational systems generates similar yields compared to maize monocultures but improves soil carbon stocks

T

⁎

Ralf Loges, Inga Bunne, Thorsten Reinsch , Carsten Malisch, Christof Kluß, Antje Herrmann, Friedhelm Taube Grass and Forage Science/Organic Agriculture, CAU Kiel, Hermann-Rodewald-Straße 9, 24118 Kiel, Germany

A R T I C LE I N FO

A B S T R A C T

Keywords: Forage production Carbon sequestration Grassland Continuous maize Net primary productivity (NPP) Roots

Ruminant livestock in agriculture is one of the largest contributors to anthropogenic greenhouse gas (GHG) emissions. One GHG mitigation strategy is to maintain or increase soil carbon stocks. However, the estimation of the impact of agricultural production systems on soil carbon stocks is often difficult due to lack of data regarding the above- and belowground allocation of the net primary production of plants. Hence, in a 7-year field experiment in northern Germany, the aboveground net primary productivity and carbon budget of three different forage production systems (a crop rotation (grass-clover, maize and winter wheat); continuous maize; and continuous grassland) were quantified, with belowground net primary productivity being determined in two production years. While the net primary production was similar across all systems and ranged between 12.2 and 13.3 t organic matter ha−1, the belowground fraction of the NPP was higher in grasslands with up to 35%, compared to 18 and 23% in continuous maize and the crop rotation. Accordingly after deduction of harvest removal also the carbon inputs as predicted by the soil carbon model were much higher in grassland and carbon stocks are projected to increase by +413 kg C ha−1 a−1 in fertilized grasslands, yet are projected to decrease by −183 kg C ha−1 a−1 in unfertilized continuous maize. However, the best option with respect to both carbon inputs and harvestable yields was the crop rotation, obtaining almost identical yields with the continuous maize with nearly balanced carbon stocks independent of the fertilization.

1. Introduction Carbon sequestration describes the capacity of an ecosystem to absorb CO2 from the atmosphere and store it as soil organic matter (SOM) in the soil (Lal 2004). Thus, carbon sequestration has received a lot of attention as a mechanism to achieve a negative greenhouse gas (GHG) balance (Rees et al., 2005; Lal 2003). Root growth and the turnover rate of the plant roots are the driving variables for the carbon amount that is sequestered. However, large differences exist in the performance of these two factors among various agricultural systems. Grasslands, for example, are characterised by a dense, fibrous root system and a substantially larger belowground biomass production than annual crops. Rees et al. (2005) compared different studies to find that the residual plant C from cereal roots into the soil was on average 1.3 t C ha−1, and thus less than half of the C input of perennial ryegrass, which was 2.8 t C ha−1. Moreover annual crops are linked to soil tillage, which reduces the ability of soils to sequester high rates of carbon in the mid- and long-term perspective (Johnston et al., 2009a,b; Six et al., 2000; Reinsch et al., 2018). Accordingly, continuous grasslands

⁎

compose one of the largest terrestrial C pools (Gobin et al., 2011). However, grasslands in intensive agricultural production in northwest Europe produce lower herbage yields compared to annual crops such as maize grown for silage (Muylle et al., 2015; Propheter et al., 2010) and the same can be seen for metabolic energy yields. This resulted in a large decrease of permanent grasslands, mainly in dairy production areas, which were predominantly substituted by maize cultivation (Souchère et al., 2003; Taube et al., 2014). Nevertheless, despite these developments being driven by an optimization of the systems for aboveground net primary productivity (ANPP), both the ANPP and belowground net primary productivity (BNPP) are important, with the ANPP being a major concern for economic viability of the production, while the BNPP is relevant for regulating ecosystem services, such as climate change mitigation. The ratio between the two can be expressed in the belowground fraction of the overall net primary production (NPP), also called the fBNPP. The fBNPP is dependent on plant species, but also on environmental changes and management operations, e.g. nutrient or water availability (Skinner and Comas, 2010), as under favourable growing conditions, growth is primarily allocated to shoots

Corresponding author. E-mail address: [email protected] (T. Reinsch).

https://doi.org/10.1016/j.eja.2018.04.010 Received 16 November 2017; Received in revised form 17 April 2018; Accepted 20 April 2018 1161-0301/ © 2018 Published by Elsevier B.V.

European Journal of Agronomy 97 (2018) 11–19

R. Loges et al.

• Continuous grassland (henceforth called ‘grassland’)

(Dodd and Mackey, 2011). Despite the importance of both, ANPP and BNPP formation, and the large differences among cultivation systems, studies that compare those regarding to different crops and cropping systems are rare, despite these data being the prerequisite for modelling the development of carbon stocks in the future. Hence, to enable a better understanding of the differences in fBNPP and C input of different production systems, a field experiment was conducted with three different forage production systems, which represent a perennial system (i.e. grassland), a complex annual system (i.e. crop rotation) and a simple annual system (i.e. continuous maize), as well as two levels of fertilization. With this general setup, this work aimed at answering the following questions: what is i) the above- and belowground performance of each system, ii) the amount of plant residual C of each crop, and iii) the resulting longterm impact on soil carbon stocks?

2 N-Fertilization

• 0N • 240N (240 kg N ha

−1

a−1, applied as cattle slurry)

The grassland was sown in autumn 2010 with a standard mixture (consisting by weight of 67 % perennial ryegrass (Lolium perenne), 17 % timothy grass (Phleum pratense), 10 % smooth meadow grass (Poa pratensis) and 6 % white clover (Trifolium repens)) at a sowing density of 30 kg ha−1. Maize (Zea mays cv. Ronaldinio) was sown after rotary cultivation and ploughing of grass clover swards in the crop rotation or ploughing of preceding maize stubble in the continuous maize system each year in the first half of May at a seed rate of 12 plants m−2 and a row distance of 0.75 m. Winter wheat (Triticum aestivum cv. Mulan) was sown between late October and mid-November with 350 grains m−2. In the crop rotation, a grass-clover mixture with predominantly red clover was used, that was undersown in winter wheat with 20 kg ha−1 perennial ryegrass, 8 kg ha−1 red clover (Trifolium pratense) and 2 kg ha−1 white clover and continued to grow during the subsequent year (Table S2). This treatment is subsequently called grass-clover to distinguish it from the grassland. Weed control occurred by spring time harrowing in winter wheat and inter-row hoeing in maize. Maize was harvested as silage maize with a target dry matter content of 300–350 g kg−1. Winter wheat was harvested as whole crop silage in the soft dough stage. Grassclover was cut once in autumn two months after the wheat harvest, while in the subsequent year it was cut four times in unison with the grassland. Slurry application in the N-fertilized treatment was split in four dressings for grassland (80/60/60/40 kg N ha−1; Table S3) and three dressings for winter wheat and maize (80/80/80 kg N ha−1). Slurry was applied with trailing hoses and had an average C/N-ratio of 6.9. The grass-clover was not fertilized because it was in itself considered as significant N-source in the crop ration. All plots were fertilized with phosphorous (45 kg P ha−1), potassium (100 kg K ha−1), magnesium (24 kg Mg ha−1) and sulphur (68 kg S ha−1) in the form of rock phosphate and potassium-magnesium sulphate. Liming occurred every two years with 1 t ha−1 calcium carbonate (23 % Ca and 1.4 % Mg). Each crop in the rotation was present in three replicates in each experimental year. The plot size was 12 m × 6 m.

2. Material and methods 2.1. Experimental site and design The field experiment was established at the organically managed research farm „Lindhof“ of Kiel University (N 54°27′55 E 9°57′55; 15 m a.s.l.) in autumn 2010. Soil samples were taken until 2017 to determine changes in soil carbon content. Measurements regarding the ANPP and BNPP of the crops were conducted in two years only, namely between April 2012–March 2013 (PI) and April 2013–March 2014 (PII). Soils at the research farm are classified as either Eutric Luvisol or Cambisol. The soil texture comprised 61% sand, 26% silt and 13% clay, with 1.2% organic carbon in the 0–30 cm soil depth (Table S1). The historical management of the site prior to the establishment of the experiment in 2010 was a long-term arable cropping system with a 4-year crop rotation. The climate at the research station is oceanic with a mean longterm (1981–2010) annual temperature of 8.9 C and a mean annual precipitation of 778 mm (Table 1). Weather data (temperature and precipitation) were obtained from a weather-station in Kiel-Holtenau, located 8 km in a direct line to the experimental field. The first experimental period (April 2012 to March 2013; PI) was characterised by a long cold period in winter, with an average temperature between December 2012 and March 2013 of 0.8 °C, compared to the long term mean of 2.2 °C (Table 1). Contrary to that, the second experimental period (April 2013 to March 2014; PII) exhibited high temperatures in winter, with a mean temperature of 4.9 °C between December 2013 to March 2014. Precipitation was low during the observation period, with 664 mm and 646 mm, for PI and PII, respectively, in comparison to 778 mm for the long-term mean. The experiment was designed as a two-factorial split-plot design with three replicates, using the following factors:

2.2. Above- and belowground biomass Aboveground biomass samples were taken at each harvest during the vegetation period. Herbage yield on grass plots was measured using a forage plot harvester (Haldrup, Løgstør, Denmark), to a residual sward height of 5 cm. Maize plots were harvested by a Haldrup harvester at a stubble height of 20 cm. Weight of total fresh harvested biomass was directly determined. DM weight was measured after oven drying of fresh samples taken during the Haldrup harvest at 58 °C until constant weight. To determine the residual biomass, after each harvest the stubbles were cut at soil level in subsampling plots of 0.25, 0.5, and

1 Forage production system

• A crop rotation, consisting of grass-clover – maize – winter wheat (with grass-clover undersown) • Continuous silage maize

Table 1 Mean monthly air temperature (°C) and precipitation (mm) for the two experimental periods (PI and PII), as well as the long-term average (1981–2010). Temperature (°C)

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Jan

Feb

Mar

Year

PI (2012/2013) PII (2013/2014) 1981–2010

6.9 6.7 7.6

12.2 11.8 11.9

13.6 14.7 14.8

16.3 18.2 17.3

17.3 17.8 17.0

13.5 13.7 13.6

9.6 11.5 9.7

6.3 6.4 5.2

1.1 5.4 2.2

1.7 2.0 1.5

0.4 5.2 1.5

−0.1 6.8 4.0

8.2 10.0 8.9

Aug

Sep

Oct

Nov

Dec

Jan

Feb

Mar

Year

55 51 74

52 60 67

55 99 77

41 50 70

49 60 67

70 53 70

21 42 47

4 22 57

664 646 778

Precipitation(mm)

Apr

May

Jun

Jul

PI (2012/2013) PII (2013/2014) 1981–2010

55 15 40

40 93 54

103 65 71

120 37 84

12

European Journal of Agronomy 97 (2018) 11–19

R. Loges et al.

1m2 for grass, winter wheat and maize, respectively with subsequent drying at 58 °C until constant weight. For the grassland, stubbles were determined as described above. However, as no ploughing occurred, stubbles did not contribute directly to the carbon input of the soil and thus were not included as input in the soil carbon model. In the crop rotation, however, stubbles of the winter-wheat and the undersown grass-clover were sampled and separated manually into the respective groups. All aboveground biomass samples were subsequently milled to a particle size of 1 mm using a centrifugal mill (Cyclotec mill, Tecator, Foss, Hillerød, Denmark) for further analysis. The ANPP of each crop is reported as the total biomass above the soil surface. Due to the grassclover being undersown in winter wheat in the crop rotation, the ANPP of winter wheat consists of a sum of the harvested biomass of the undersown grass-clover and winter wheat. Belowground biomass growth was determined with the ingrowth core method (Steingrobe et al., 2000; Steingrobe et al., 2001), using a sampling interval of four weeks during Apr.-Oct. and four cores per plot. During winter the ingrowth cores were installed in November, and remained in the soil until the end of March. Cores were installed by using a spiral hand auger (diameter 4 cm) (Eijkelkamp, The Netherlands), which was drilled into the soil at an angle of 45° relative to the soil surface, to a vertical depth of 30 cm. In cereal plots, cores were installed between and parallel to the seeded rows. For maize, with the large distance of 75 cm between two rows, the distances between the four cores were spaced evenly across a gradient ranging from the plant base to the centre between two rows (i.e. 12.5 cm intervals). Mesh bags (synthetic fibre net, mesh size 1 mm, diameter 4 cm, length 60 cm) were pulled over PVC tubes and inserted into the holes. Afterwards the mesh bags were filled in several steps with pre-sieved (1 mm) soil from the same treatment, while simultaneously the PVC tubes were pulled out of the mesh bags. Care was taken to achieve a soil bulk density identically to the ambient soil bulk density. Finally, bags were sealed with cable ties. After the cores were removed, the roots were washed in a hydro pneumatic elutriation system (Ehmsomat GTI, Kiel, Germany) over a 0.63 mm sieve and manually separated from other soil constituents (Smucker et al., 1982). Accumulation of root growth in the bags over all dates resulted in total belowground net primary production (BNPP). The sum of ANPP and BNPP accounts for the whole net primary production (NPP), whereas fBNPP represents the ratio of BNPP to NPP, i.e.: fBNPP = BNPP/NPP. Fresh weight of roots were determined, before oven drying samples at 58 °C to constant weight, for determination of dry matter content. Afterwards, all root samples were milled in a ball-mill at a frequency of 25 Hz for 2 min (MM200, Retsch GmbH, Haan, Germany) for further analysis. C, N and ash content of ANPP and BNPP were estimated using near infrared reflectance spectroscopy (NIRS; NIRS-System 5000 monochromator; Foss Nirssystems Silver Spring, USA) with two replicates per analysis. The used material specific calibration represented the entire spectral variability and was done using the WinISI II software (Infrasoft Internationals, South Atherton St., PA, USA). For validation, the C- and N-concentration of a sub-set of samples was analysed using a C/N-analyzer (Vario Max CN, Elementar Analysensysteme GmbH, Hanau, Germany). The standard error of validation (SEV) for this subset ranged between 0.05 and 0.1% of the DM for nitrogen analyses in the aboveground biomass and between 0.3 and 0.7% for carbon analyses. In roots, the SEV were on average 0.1% and 0.8% for nitrogen and carbon analyses, respectively and in stubbles, the SEV were 0.1% and 1.7%. Ash free dry matter of calibration samples were determined by burning in a muffle furnace (24 h at 550 °C). As ash content varied substantially particularly in root systems (mean: 26%, range: 14–49%), but also in the aboveground biomass (mean: 10%, range 4–25%), all values are reported as organic matter (OM) weight.

systems on long-term changes in soil carbon stocks, the CN-SIM (Petersen et al., 2005a,b) model was used, which is calibrated for northwest European conditions. It accounts for seven conceptual compartments, which comprise added organic matter (two compartments, AOM1 and AOM2); soil microbial biomass (two compartments, SMB1 and SMB2); one compartment for microbial residues (SMR); native organic matter (NOM); inert organic matter (IOM). Heterotrophic soil respiration (CO2) in the model results from decay of organic matter of soil decomposers (SMB) and represents the pathway for soil carbon losses. The allocation of initially measured carbon pools to these seven conceptual compartments was set according to recommendations of the farm simulation model FASSET (Berntsen et al., 2006). Decay rates and partitioning of incorporated matter between AOM1 and AOM2 pools are based on Petersen et al. (2005a,b).With regards to environmental data, hourly soil temperature was predicted using a simple model approach by Plauborg (2002). The model is considering the measured air temperature as input in order to calculate the corresponding soil temperature (Plauborg, 2002). The model was validated using regular soil temperature measurements in 2011 and 2012. For the long term projections with the CN-Sim model, no soil temperature increments were considered, but the long-term mean values were set as constant. Soil water content was modelled according to Mohrlok (2009). This model requires daily measurements of rainfall amounts and solar radiation combined with the water retention capacity of the soil at the experimental site as input to calculate the evaporation rate and the subsequent soil moisture. This approach was validated using several soil moisture measurements and the calculated soil moisture was again assumed to remain constant in long-term projections of the CN-Sim model. Carbon input of plant residues in annual systems, such as the crop rotation and continuous maize was defined as the cumulated roots measured with the ingrowth core method during plant growth and stubbles that were incorporated by ploughing, while potential preharvest losses of senescence leaves were not accounted for. Since no ploughing was conducted in the grassland system in addition to carbon derived by roots, a shoot turnover (leaf litter and standing dead) of 28% was assumed, as reported in Schuman et al. (2002). This high number seems reasonable, given that perennial ryegrass has a particularly high shoot turnover due to its inability to maintain more than three to four leaves per pseudostem, after which senescence of the oldest leaf is initiated (Schneider et al., 2006). Given that plants were about 20 cm high at cutting, plants will have continued to grow for some time after the three leave stage. The fine roots and root exudates of different crops which were not captured by the ingrowth core method were calculated as a proportion of measured net primary production, using the mean value for each culture as calculated in the review of Kuzyakov and Domanski (2000). These values were 9% for winter wheat and 5% for maize, grass-clover and grassland. The model was calibrated using soil carbon data from 2010 to 2017 to assess the model accuracy. Soil carbon measurements were conducted annually since the experiment was established in 2010. Soil samples were taken to a soil depth of 0–30 cm. Each sample was bulked from three replicates per plot. Soil samples were dried in an oven at a temperature of 30 °C until constant weight and afterwards stored in paper bags until further processing. In pre-treatment for laboratory analysis, dried soil was sieved to pass a 2 mm sieve. To provide homogenous conditions for each sample, soil samples were ball milled. Total concentration of soil carbon was measured by dry combustion using a C/N-Analyser (Vario Max CN, Elementar Analysensysteme, Hanau, Germany). Subsamples were treated with HCl (4 mol l−1) to check for presence of carbonates. On this site no carbonates were present, therefore measured soil carbon was assumed to be equivalent to total organic carbon. All procedural steps were taken according to the international standard ISO 10694. To estimate the total amount of carbon stocks, soil density (Db) of undisturbed soil cores (100 cm3) was

2.3. Carbon modelling For the prediction of the effect of the different forage production 13

European Journal of Agronomy 97 (2018) 11–19

R. Loges et al.

measured in all plots. Total carbon (t C ha−1) for the upper soil layer (0–30 cm) was calculated considering Corg (g C 100 g−1 soil−1), Db (g cm−3) and soil depth (30 cm) (VandenBygaart et al., 2007). Longterm simulation runs of 100 years were performed, whereby C-inputs from AGB and BGB were assumed constant over time.

the crop rotation (Fig. 1). However, the harvestable aboveground biomass was lower in the grassland with 7.1 t OM ha−1 compared to 9.7 t OM ha−1 and 9.3 t OM ha−1 for continuous maize and the crop rotation, respectively. Similarly, the harvested energy yield was lower in the grassland system, with 84 GJ ME ha−1 compared to 114 GJ ME , ha−1 and 121 GJ ME ha−1, for continuous maize and the crop rotation, respectively (results not shown). The comparable NPP of the grassland is mainly due to the high belowground net primary productivity (BNPP), as expressed in the fBNPP of 32%, which was higher (P < 0.01) than the 18% and 21% for continuous maize and the crop rotation, respectively (Fig. 1). The lower dry matter and carbon input of roots in continuous maize systems is exacerbated when looking at the stubble biomass, where grassland contributed 2.0 t OM ha−1, while continuous maize and the crop rotation contributed 0.5 t OM ha−1 and 0.8 t OM ha−1, respectively. Differences existed also between the systems in the BNPP formation during winter (P < 0.001; Fig. 2). With maize being harvested in autumn, roots growth is restricted to the growing season. Contrary to that, both grassland and crop rotation were able to accumulate root dry matter with on average 0.6 t OM ha−1 and 0.2 t OM ha−1, respectively, with a share of 0.3 t C ha−1 in grassland and 0.1 t C ha−1 in winter wheat during winter. Fertilization showed no significant effect on plant growth during winter. Consequently, the average annual C input into the soil from roots and stubbles accounted for 3.0 t C ha−1 in grassland, and thus was higher (P < 0.001) than the continuous maize (1.3 t C ha−1) or the crop rotation (1.6 t C ha−1). However, stubble biomass from grassland was not directly considered in the model evaluation (see section 2.3) as the turnover rate of stubbles in grassland was unknown. To assess the resources that plants allocate in either above- and belowground biomass formation, we tested the correlation between the cumulative root growth (BNPP) to the cumulative aboveground biomass (ANPP, i.e. harvested biomass and stubble). Generally, over all harvests and production systems ANPP and BNPP were positively correlated (r = 0.75, P < 0.001, result not shown). However, large differences existed among the systems. Because there was no significant interaction between the production system and the fertilization (P = 0.24), all production systems were subsequently pooled over the

2.4. Statistical analysis The statistical software R (R Core Team, 2016) was used to analyse the data. The data evaluation started with the definition of an appropriate statistical mixed model (Laird and Ware, 1982; Verbeke and Molenberghs, 2000) and implemented using the R package “nlme” (Pinheiro et al., 2016). The data were assumed to be normally distributed and to be heteroscedastic due to the different levels of crop and fertilization. These assumptions are based on a graphical residual analysis. The statistical model included the crop sequence (i.e. all three simultaneously grown crops of the crop rotation averaged to a “mean crop rotation” value) and fertilization as well as all their interaction terms as fixed factors. The year and the interaction between repetition and year were regarded as random factors. Based on this model, an analysis of variances (ANOVA) was conducted to answer the questions of the trial. After this, multiple contrast tests were conducted using the R package “multcomp” (Hothorn et al., 2008) in order to compare the several levels of the influence factors, respectively. Statements about correlations were determined using the Pearson correlation coefficient. The overall model performance was evaluated using the coefficient of determination (R2), the Nash-Sutcliffe model efficiency coefficient (NSE) and the root mean squared error (RMSE) (Loague and Green, 1991). 3. Results 3.1. Net primary production Net primary production (NPP) was not significantly different among the three production systems across the 0N and 240N treatment, showing values of 13.3 t organic matter (OM) ha−1 in grassland compared to 12.2 t OM ha−1 for continuous maize in and 12.7 t OM ha−1 in

Fig. 1. Biomass production of different production systems and the individual cultures of the crop rotation system, divided in belowground biomass (roots) and aboveground biomass (stubble and harvestable biomass). Different lower case letters indicate significant differences between unfertilized (N0) and fertilized (N240) treatments, while capital letters indicate differences among production systems (p ≤ 0.05). Error bars are the standard error of the mean of the entire NPP (sum of all three fractions).

14

European Journal of Agronomy 97 (2018) 11–19

R. Loges et al.

Fig. 2. Mean biomass production of different production systems and the individual cultures of the crop rotation system, divided in belowground biomass (roots) and aboveground biomass (stubble and harvestable biomass) for both unfertilized (N0) and fertilized (N240) plots. Error bars are standard error of the mean.

the grassland, continuous maize and crop rotation, respectively. Accordingly, C/N ratios were stable across a large range of BNPP yields (i.e. similar slope), yet the fNBNPP was larger for any given fBNPP in maize monoculture compared to the crop rotation and grassland (i.e. parallel shift of slope among cultures).

fertilization levels. A correlation existed for both maize in continuous cultivation (r = 0.42, P < 0.001), and for maize in the crop rotation (r = 0.47, P < 0.001; Fig. S1). However, no such link existed for either the grassland system (r = −0.08, P = 0.52) or the grass-clover (r = 0.07, P = 0.58). Nevertheless, while in both systems the ANPP did not affect the BNPP, the clover share had an impact in both grassland (r = −0.43, P < 0.001) and grass-clover (r = −0.28, P < 0.05; result not shown). Winter wheat is not shown, because the undersown grassclover prevents an attribution of roots to either the winter wheat or grass-clover component. Across all production systems, plots fertilized with 240 kg N ha−1 in form of slurry yielded 9.4 t OM ha−1 harvestable biomass, which was 18% more (P < 0.05; Fig. 1) than unfertilized plots. Similarly, fertilized plots achieved a NPP of 14.0 t OM ha−1, which was 21% higher (P < 0.01) than in unfertilized plots. Additionally, the fraction of biomass being located belowground (fBNPP) increased (P < 0.05) overall from 22 to 25%. The C/N ratios for roots were comparable among the different systems and fertiliser levels (Table 2). However, for the harvestable shoot biomass these values were evidently different for the different tested crops with highest mean C/N values for wheat (28:1) and maize (39:1) when compared to grassland and grass-clover (16:1) respectively. Due to the large differences in C/N-ratios aboveground, yet their similar nature belowground, the overall relationship between the fBNPP and the belowground fraction of nitrogen of the total nitrogen (fNBNPP) also differs among the production systems and can be described using the formulas: fBNPPcont.grassland = 8.2 + 0.89 x fNBNPP (R2 = 0.82, P < 0.001), fBNPPmaize = 1.9 + 0.7 x fNBNPP (R2 = 0.92, P < 0.001) and fBNPPcrop-rotation = 1.4 + 0.77 x fNBNPP (R2 = 0.95, P < 0.001) for

3.2. Observed and modelled changes in soil carbon stocks On basis of the annual soil samples seven years after establishment of the experiment (in 2017), the unfertilized grassland had increased its soil carbon stocks on average by 85 kg C ha−1 a−1, and thus had more carbon input (P < 0.001) than the continuous maize, which decreased C stocks by 530 kg C ha−1 a−1, and the crop rotation with decrements of 120 kg C ha−1 a−1 in 0–30 cm soil depth. When these systems were fertilized with 240 kg N ha−1 of slurry, soil carbon stocks in grassland increased by 1300 kg C ha−1 a−1, and consequently produced higher (P < 0.001) carbon inputs than the continuous maize, which still decreased by 550 kg C ha−1 a−1, and the crop rotation, where soil carbon stocks increased by 250 kg C ha−1 a−1 (results not shown). When utilizing these data in the soil carbon model, in long-term the grassland C-sequestration rates accounted for 306 and 156 kg C ha−1 a−1 in the 20 (20y) and 100 (100y) year perspective for the upper soil layer 0–30 cm (Fig. 3). The model simulated long-term soil carbon losses in continuous maize of −510 (20y) and −183 (100y) kg C ha‐1 a−1. For the crop rotation, a slight decrease of −215 (20y) and −81 (100y) kg C ha−1 a−1 was simulated. After 100 years, the initial carbon stocks are expected to have increased by 16 t C ha−1 in the grassland, yet decreased by 18 t C ha−1 in continuous maize, while having been reduced by less than 8 t C ha−1 in the crop rotation.

Table 2 Mean values for the fraction of the net primary productivity (NPP) that is growing belowground (fBNPP) and the C/N ratio of the three production systems (grassland, maize and the crop rotation (crop rot.)), as well as of the three components of the crop rotation (grass-clover, maize and winter-wheat with grass-clover undersown). N-fert is the level of nitrogen fertiliser applied in kg N ha−1. C/N ratios are expressed as only the numerator of the ratio, with the denominator always being 1. Production systems grassland N-fert.

C/N ratio

fBNPP (%) Shoot Stubble Root

0

maize 240

29 17 20 21

Crop rotation

35 17 23 25

crop rot.

0

240

18 42 197 28

18 41 231 25

0 20 26 79 20

240 23 27 82 20

15

red-clover grass

maize

0

0

240

14 37 137 20

18 37 148 19

240 17 14 17 17

24 15 18 19

w.-wheat + red clover grass 0

240 29 26 77 23

27 30 78 21

European Journal of Agronomy 97 (2018) 11–19

R. Loges et al.

4. Discussions 4.1. Net primary production While the NPP of the overall crop rotation was in the same range as the continuous maize and the grassland, they were variable within the components. Accordingly, winter wheat only had NPP yields of 9.6 t OM ha−1, of which 6.7 t OM ha−1 were harvestable biomass, while maize in crop rotation yielded 15.1 t OM ha−1 NPP, of which 12.1 t OM ha−1 were harvestable biomass. Generally, pre-crop derived green manures are advantageous for the following maize, because it is well adapted to exploit the positive effect of organic nitrogen (Komainda et al., 2018; Herrmann et al., 2014). Accordingly, maize in crop rotation was reported to take up higher N amounts than continuous maize (Maloney et al., 1999) and reached higher DM, N and C amounts both above- and belowground. Although, this effect was previously observed to be dependant on nitrogen fertilization, with strongest effects occurring only in low fertilization regimes (Nevens and Reheul, 2001), the additional rotational N effect was observable in our results despite the comparably high N-fertilization rate used. Accordingly, maize in crop rotation yielded 26% more harvestable biomass than continuous maize. Moreover, the crop rotation showed a lower fNBPP value when compared to continuous maize. Consequently, more of the nitrogen uptake is used for the production of harvestable plant material and lower rates of reactive nitrogen are potentially released to the environment by root residues after harvest. Another advantage of the crop rotation compared to continuous maize is the nearly whole year round soil cover as was observed in the winter biomass accumulation. This can result in a reduced risk of soil erosion and surface runoff, and an improved nutrient use efficiency. Hence, particularly the crop rotation provides additional benefits other than yield, while having almost similar aboveground biomass yields compared to high yielding forage crops such silage maize. Independent of this fact, generally C inputs in winter are highly influenced by the weather and were dominated by root growth in winter of the second period (PII). During the cold winter period in PI (Table 1), both above- and belowground biomass formation were nearly zero. Contrary to that, the warmer winter period PII resulted in comparably high belowground biomass growth, in grassland and grassclover plots. Temperatures during February and March reached up to 13 °C. Garwood (1967) described that under these conditions, an establishment of a high rate of produced new roots early in the year is possible. Accordingly, with the climatic changes proposed for the climate of northern Germany, where climate change is expected to result in a substantial reduction of frost-periods (Trnka et al., 2011), these benefits of an additional carbon input from a crop rotation or grassland during winter are likely to gain more importance. Regarding the methodological uncertainties that exist for estimating BNPP, one common issue with the method of ingrowth cores is that the ingrowth duration can result in an underestimation of the BNPP when root mortality starts (Chen et al., 2016a), and thus the selection of an appropriate ingrowth period is essential. The onset of root mortality usually starts after 1.5 months of ingrowth duration in arable crops (Steingrobe et al., 2000). Hence, we established a pretest to identify the optimal duration of the ingrowth period for the ingrowth core. This test compared the annual BNPP from ingrowth cores that were either replaced in a four-weekly cycle or were always left in the ground for the duration between two cuts in grassland (i.e. roughly 1.5 months). Annual BNPP was not different between these two ingrowth durations (P = 0.24; results not shown) and neither were the interactions between the ingrowth duration and fertilization (P = 0.54) or the production system (P = 0.09) significant. Thus, we can be confident that under normal conditions no root turnover will have occurred during a four week period, thus making our results comparable across all production systems. Even in the six months ingrowth period in winter, root mortality is unlikely to have had a substantial impact, due to the

Fig. 3. Development in soil carbon stocks (Csoil) as simulated with the CN-Sim model until the year 2100. Production systems are the continuous grassland (Grass), continuous maize (Maize), and the crop rotation (CR).

Additional C-inputs due to higher NPP and carbon derived from applied slurry increased the simulated rates of sequestration to +832 (20y) and +413 (100) kg C ha−1 in the grassland; −215 (20y) and −76 (100y) kg C ha−1 in continuous maize; and almost equilibrated soil carbon stocks of +21 (20y) and +15 (100y) kg C ha−1 in the crop rotation. Accordingly, with fertilization the C stocks were simulated to increase by 41 t C ha−1 in the grassland and 2 t C ha−1 in the crop rotation, while still decreasing by 8 t C ha−1 in the fertilized continuous maize in long-term (100y). In the model validation, the simulated soil carbon stocks for the years 2011–2017 showed a good fit to the measured soil carbon data over all treatments (R2adj = 0.54, P < 0.001, NSE = 0.49, RMSE = 2.79; Fig. 4). However, differences existed among production systems, with continuous maize showing a close match in the trend between modelled and observed soil carbon stocks (y = −3.3 + 1.08x, R2adj = 0.41, P < 0.01, NSE = 0.43, RMSE = 1.22), compared to the crop rotation (y = −24.7 + 1.55x, R2adj = 0.28, P < 0.05, NSE = −1.67, RMSE = 2.70) and grassland (y = −54.0 + 2.06x, R2adj = 0.53; P < 0.01, NSE = 0.39, RMSE = 3.82). However, the crop rotation and the grassland appear to be slightly underestimated by the model based on the slope of the regression lines.

Fig. 4. Measured vs. modelled values soil carbon stocks (Csoil) with measurements from 2011 to 2017 in layer of 0–30 cm. The regression line across all production systems identifies the fit of the model to the observed changes in the carbon pool in the soil. 16

European Journal of Agronomy 97 (2018) 11–19

R. Loges et al.

et al., 2017), thus rendering the relative importance of subsoil carbon negligible. Hence, we assumed that the observed soil depth (0–30 cm) is suited best to estimate the impact of different forage production systems on soil organic matter. The predicted absolute C inputs derived from plant residues of winter wheat with undersown grass (3.1 t C ha−1), silage maize (1.5 t C ha−1) and grassland (5.3 t C ha−1) are comparable to modelled values from the literature (Meersmans et al., 2013). Modelled changes in carbon storages showed a good fit with the measured soil carbon data (Fig. 4). Even though the NSE was poor within the crop rotation, the other treatments indicated a good performance of the modelled results and the RMSE values were small across all treatments. This indicates the suitability of the CN-sim model and that the accuracy is comparable to other soil carbon models reported in the literature (Smith et al., 1997). The model used estimated carbon sequestration rates of +832 kg C ha−1 a−1 for fertilized grassland under a cutting regime in a 20 years perspective. This is in accordance with values reported for grasslands in existing studies with comparable management techniques. Hence, Jones and Donnelly (2004) reported a range of carbon sequestration rates from +123 to +2372 kg C ha−1 a−1, based on a survey of measured data for temperate grasslands. Soussana (2008) reported values of +1040 kg C ha−1 a−1, yet Skinner (2008) reported annual losses of −810 kg C ha−1 a−1. However, the annual rate of sequestration is dependent of the initial carbon stocks and management. When compared to a 20 years old permanent grassland swards close to the experimental site (Reinsch et al., 2018) soil carbon data showed low figures due to the historical arable management, thus after grassland establishment high initial sequestration rates were achieved. Similarly, the fertilized forage crop rotation maintained the soil carbon levels because historical pre-management of the site also contained annual red-clover grass and manure amendments in a crop rotation (Wheatgrass-clover − Oats − potatoes). On the contrary, continuous maize led to a decrease of soil carbon stocks irrespective of slurry application. Other studies also obtained positive sequestration rates using continuous maize silage (Kristiansen et al., 2005) but with dominated cereal production and without grass-clover in the crop rotation in premanagement. The poor carbon sequestration potential of maize is even reported elsewhere (Gregorich et al., 2001), whereas the main reasons for the carbon losses in silage maize are the high amount of herbage removal and the bare soil after harvest until the subsequent maize is established and the mechanical soil disturbance intensity such as ploughing and hoeing, resulting in less annual carbon inputs. Consequently, the rate of carbon sequestration of silage maize is mainly related to the crop residues (BNPP and stubble). Yet while the C/N ratio of roots was generally favourable for decomposition across all systems, with a range of 17:1–28:1 (Table 2), the differences were much larger in stubbles. Here, the grassland and grass-clover had by far the lowest ratio, being always around 20:1, while the stubbles of maize had C/N ratios between 137:1 and 231:1. An increase in the C/N ratio from roughly 20–100 can decrease the decomposition constant k by approximately a factor 100 (Enríquez et al., 1993). However, this did not result in substantial increments of the carbon inputs due to the small contribution (∼11%) that stubbles accounted for in the total C input into the soil. A larger impact has the fact that maize lacks inputs from plant litter, whereas grassland and grass-clover swards add around 35% of their carbon into the soil via plant litter. This is another factor for the higher C inputs of grasslands compared to maize. Despite the generally discovered negative relationship between particularly the C/N ratio of belowground biomass and carbon stocks, the predictive capacity of C/N ratios for the soil carbon stocks in grassland is limited, with management techniques such as ploughing having a much larger effect on soil carbon stocks than changes in C/N ratios (Heyburn et al., 2017) Even though slurry application induced an additional input of plant residues due to an increase in NPP and the slurry itself of 2.9 t C for grassland, 1.9 t C for maize and 1.6 t C for the crop rotation, continuous

reduced soil activity and hence, a much slower root turnover. Nevertheless, the possibility that root mortality occurred cannot be excluded entirely, but because only grassland and the crop rotation had a soil cover in winter, any root mortality would lead to an underestimation of the grassland and crop rotation root growth. Hence, this would only increase the benefits of an additional C input during winter. Nevertheless, the differences across systems with regards to BNPP were similar to previously reported carbon inputs from roots of grass-clover systems of 1.8 t C ha−1 (Dodd and Mackay, 2011; Chen et al., 2016b), and 1.1 t C ha−1 (van Noordwijk et al., 1994) for winter wheat roots. For maize a good correlation was found for ANPP and BNPP, particularly in absence of a nitrogen deficiency (i.e. the fertilized treatments). This is in accordance with a survey of studies summarised for maize (Amos and Walters, 2006). Regarding the effect of fertilization, this generally results in a reduced fBNPP as fertilization usually reduces the requirement for a large root body and favours shoot growth instead (Hui and Jackson, 2006), due to a lower requirement for a large root mass to obtain the required nutrients (Dodd and Mackay 2011). The fBNPP was similar across fertilization treatments in maize, yet increased in the fertilized compared to the unfertilized grassland. While appearing unusual, this can be explained as the fertiliser application changed the botanical sward composition. Due to the loss of a competitive advantage for the legumes, this resulted in a lower clover share of 33%, compared to 52% in the unfertilized plots, albeit with a large variability within the systems. With a denser root system and a higher root/shoot ratio of grasses compared to white clover, this resulted in a higher root biomass in fertilized plots (Reijs et al., 2007; Griffith et al., 2000; Davidson, 1969a,b). This effect was exacerbated for the grassland system in our experiment due to the lack of difference in ANPP of grassland swards between fertilized and unfertilized plots, due to the increase in legumes, which compensated for the lack of fertiliser with the biological nitrogen fixation (Schmeer et al., 2014). Consequently, with denser roots at similar aboveground biomass, the fBNPP slightly increased due to the fertilization. Hence, the clover content showed a strong negative correlation with BNPP but this was not necessarily correlated with lower yields. However, for the grass-clover in the crop rotation another reason might be that ingrowth bags favour the root ingrowth of lateral roots, whereas deep-rooting species, such as red clover produce mainly a taproot with only few lateral fine roots. 4.2. Observed and modelled changes in soil carbon stocks For maize, it has been shown that during a ten-year period, 44% of the C input from roots occurred into the subsoil below 30 cm depth (Rasse et al., 2006). Hence, it has to be taken into account that in our study, only the upper 30 cm soil layer was analysed, which could potentially result in an underestimation of the carbon input from maize. This is particularly true, as the proportion of root dry matter found in the topsoil (i.e. upper 30 cm) is around 95% for small grain cereals, such as winter wheat and perennial ryegrass, yet only around 85% for red clover (Bolinder et al., 2002), and some studies have only found 80% of the root biomass even in the 0–40 cm layer for maize (Qin et al., 2006). Nevertheless, the proportion of maize roots in deep soil layers, and hence the C input in deep soil layers vary substantially, based on the observed large differences in the proportion of root dry matter that has been formed in the topsoil (Amos and Walters, 2006). Another issue increasing the complexity of assessing the importance of subsoil carbon inputs is that carbon in the subsoil is usually highly heterogeneous, as the C input is largely dependent on macropores that stabilize particular carbon. Together with a higher stability of carbon in the subsoil, due to the physical separation between carbon and the decomposing microorganisms, resulting in a reduced decomposability and higher retention times, the impact of subsoil carbon is difficult to estimate (Kautz et al., 2013). Moreover, several studies reported substantial differences in the topsoil SOM after management changes in mid-term but no significant differences in deeper soil depths (Franzluebbers et al., 2000; Walia 17

European Journal of Agronomy 97 (2018) 11–19

R. Loges et al.

maize still exhibited a decline in soil carbon stocks, whereas for grassland and the crop rotation soil carbon stocks were increased substantially. Furthermore, it has to be considered that due to the high rate of slurry application to ensure the value of 240 kg N ha‐1 also high rates of slurry derived carbon were achieved. The proportion of slurry C on total C inputs in the N-fertilized treatments accounted for 20% in grassland, 34% in the crop rotation and 49% for continuous maize. According to the assumptions of the model, 20% of the slurry derived carbon are considered to contribute directly to the native organic matter pool (Petersen et al., 2005a, 2005b), whereas the rest is decomposed very rapidly (Glaser et al., 2001), only on average of 6 kg and 3 kg of carbon were sequestered per ha−1 a−1 of each unit (m−3) of slurry applied in the 20y and 100y perspective, respectively. These assumptions have proven reasonable for the continuous maize system, where both fertilized and unfertilized treatments were evenly distributed above and below the 1:1 line in the model validation (Fig. 4). However, this was slightly less true for the crop rotation and grassland, which modelled values were generally slightly underestimating the observed measurements (skewed above 1:1 line for the fertilized treatment). Despite these minor issues, the validation has proven the general accuracy of the model to be excellent.

Chen, S., Lin, S., Reinsch, T., Loges, R., Hasler, M., Taube, F., 2016a. Comparison of ingrowth core and sequential soil core methods for estimating belowground net primary production in grass-clover swards. Grass Forage Sci. 71, 515–528. http://dx. doi.org/10.1111/gfs.12214. Chen, S.-M., Lin, S., Loges, R., Reinsch, T., Hasler, M., Taube, F., 2016b. Independence of seasonal patterns of root functional traits and rooting strategy of a grass-clover sward from sward age and slurry application. Grass Forage Sci. 71, 607–621. http://dx.doi. org/10.1111/gfs.12222. Davidson, R.L., 1969a. Effects of edaphic factors on the soluble carbohydrate content of roots of Lolium perenne L and Trifolium repens L. Ann. Bot. 33, 579–598. http://dx.doi. org/10.1093/oxfordjournals.aob.a084310. Davidson, R.L., 1969b. Effects of soil nutrients and moisture on root/shoot ratios in Lolium perenne L and Trifolium repens L. Ann. Bot. 33, 571–577. http://www.jstor.org/ stable/42908747. Dodd, M.B., Mackay, A.D., 2011. Effects of contrasting soil fertility on root mass, root growth, root decomposition and soil carbon under a New Zealand perennial ryegrass/ white clover pasture. Plant Soil 349, 291–302. http://dx.doi.org/10.1007/s11104011-0873-0. Enríquez, S., Duarte, C.M., Sand-Jensen, K., 1993. Patterns in decomposition rates among photosynthetic organisms: the importance of detritus C:N:P content. Oecologia 94, 457–471. http://dx.doi.org/10.1007/bf00566960. Franzluebbers, A.J., Stuedemann, J.A., Schomberg, H.H., Wilkinson, S.R., 2000. Soil organic C and N pools under long-term pasture management in the Southern Piedmont USA. Soil Biol. Biochem. 469–478. Garwood, E.A., 1967. Seasonal variation in appearance and growth of grass roots. Grass Forage Sci. 22, 121–130. http://dx.doi.org/10.1111/j.1365-2494.1967.tb00514.x. Glaser, B., Bol, R., Preedy, N., McTiernan, K.B., Clark, M., Amelung, W., 2001. Short-term sequestration of slurry-derived carbon and nitrogen in temperate grassland soil as assessed by 13C and 15N natural abundance measurements. J. Plant Nutr. Soil Sci. 164, 467–474. http://dx.doi.org/10.1002/1522-2624(200110)164:5<467:AIDJPLN467>3.0.CO ;2-Q. Gobin, A., Campling, P., Janssen, L., Desmet, N., van Delden, H., Hurkens, J., Lavelle, P., Berman, S., 2011. Soil organic matter management across the EU – best practices, constraints and trade-offs. Final Report for the European Commission’s DG Environment, 2011. http://dx.doi.org/10.2779/17252. Gregorich, E.G., Drury, C.F., Baldock, J.A., 2001. Changes in soil carbon under long-term maize in monoculture and legume-based rotation. Can. J. Soil Sci. 81, 21–31. http:// dx.doi.org/10.4141/s00-041. Griffith, G.S., Cresswell, A., Jones, S., Allen, D.K., 2000. The nitrogen handling characteristics of white clover (Trifolium repens L.) cultivars and a perennial ryegrass (Lolium perenne L.) cultivar. J. Exp. Bot. 51, 1879–1892. http://dx.doi.org/10.1093/ jexbot/51.352.1879. Herrmann, A., Claus, S., Loges, R., Kluß, C., Taube, F., 2014. Can arable forage production be intensified sustainably? A case study from northern Germany. Crop Pasture Sci. 65, 538. http://dx.doi.org/10.1071/CP13362. Heyburn, J., MkKenzie, P., Crawley, M.J., Fornara, D.A., 2017. Effects of grassland management on plant C:N:P stoichiometry: implications for soil element cycling and storage. Ecosphere 8, e01963. http://dx.doi.org/10.1002/ecs2.1963. Hothorn, T., Bretz, F., Westfall, P., 2008. Simultaneous inference in general parametric models. Biometr. J. 50, 346–363. http://dx.doi.org/10.1002/bimj.200810425. Hui, D., Jackson, R.B., 2006. Geographical and interannual variability in biomass partitioning in grassland ecosystems: a synthesis of field data. New Phytol. 169, 85–93. http://dx.doi.org/10.1111/j.1469-8137.2005.01569.x. Johnston, A., Edward Poulton, P.R., Coleman, K., 2009a. Chapter 1 soil organic matter. Advances in Agronomy. Elsevier, pp. 1–57. http://dx.doi.org/10.1016/S00652113(08)00801-8. Johnston, A.E., Poulton, P.R., Coleman, K., 2009b. Soil organic matter: its importance in suitable agriculture and carbon dioxide fluxes. Adv. Agron. 101, 1–57. http://dx.doi. org/10.1016/s0065-2113(08)00801-8. Jones, M.B., Donnelly, A., 2004. Carbon sequestration in temperate grassland ecosystems and the influence of management, climate and elevated CO2: Tansley review. New Phytol. 164, 423–439. http://dx.doi.org/10.1111/j.1469-8137.2004.01201.x. Kautz, T., Amelung, W., Ewert, F., Gaiser, T., Horn, R., Jahn, R., Javaux, M., Kemna, A., Kuzyakov, Y., Munch, J.-C., Pätzold, S., Peth, S., Scherer, H.W., Schloter, M., Schneider, H., Vanderborght, J., Vetterlein, D., Walter, A., Wiesenberg, G.L.B., Köpke, U., 2013. Nutrient acquisition from arable subsoils in temperate climates: a review. Soil Biol. Biochem. 57, 1003–1022. http://dx.doi.org/10.1016/j.soilbio. 2012.09.014. Komainda, M., Taube, F., Kluß, C., Herrmann, A., 2018. Effects of catch crops on silage maize (Zea mays L.): yield, nitrogen uptake efficiency and losses. Nutr. Cycl. Agroecosyst. 110, 51. http://dx.doi.org/10.1007/s10705-017-9839-9. Kristiansen, S.M., Hansen, E.M., Jensen, L.S., Christensen, B.T., 2005. Natural abundance and carbon storage in Danish soils under continuous silage maize. Eur. J. Agron. 22, 107–117. http://dx.doi.org/10.1016/j.eja.2004.01.002. Kuzyakov, Y., Domanski, G., 2000. Carbon input by plants into the soil. Rev. J. Plant Nutr. Soil Sci. 163, 421–431. http://dx.doi.org/10.1002/1522-2624(200008) 163:4<421:aid-jpln421>3.0.co;2-r. Laird, N.M., Ware, J.H., 1982. Random-effects models for longitudinal data. Biometrics 38 (4), 963–974. Lal, R., 2003. Global potential of soil carbon sequestration to mitigate the greenhouse effect. Crit. Rev. Plant Sci. 22, 151–184. http://dx.doi.org/10.1080/713610854. Lal, R., 2004. Soil carbon sequestration impacts on global climate change and food security. Science 304, 1623–1627. http://dx.doi.org/10.1126/science.1097396. Loague, K., Green, R.E., 1991. Statistical and graphical methods for evaluating solute transport models. Overview and application. J. Contam. Hydrol. 7, 51–73. http://dx. doi.org/10.1016/0169-7722(91)90038-3.

5. Conclusions While the aboveground net primary productivity was highest in maize and lowest in grassland systems, all systems showed similar net primary productivities (NPP), as the low aboveground yields in grassland were compensated by higher belowground biomass yields. Root functional traits obviously differ between multispecies swards and pure grass stands, but grassland or grass-clover systems can be highly productive belowground even though low aboveground biomass is produced. Consequently, the additional carbon input into the soil needs to be taken into account as an ecosystem service, when comparing the productivity of different forage production systems. Particularly after an implementation of carbon emission trading schemes for agriculture, this could shift the overall assessment of the viability of forage production in favour of grasslands. This is especially true in unfertilized systems, where the share of legumes can compensate the nitrogen input from mineral or organic fertiliser. In order to simulate the carbon budgets, it has become apparent that better knowledge about the net belowground productivity is required to increase the performance of soil carbon models. Acknowledgements These investigations were supported by the European Commission through the 7th Framework Programme (Project ID: 289328, Funded under: FP7-KBBE, CANTOGETHER (Crops and ANimals TOGETHER) project). We would like to thank Germany's National Meteorological Service (Deutscher Wetterdienst (DWD)) for their provision of the climatic data. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.eja.2018.04.010. References Amos, B., Walters, D.T., 2006. Maize root biomass and net rhizodeposited carbon. Soil Sci. Soc. Am. J. 70, 1489. http://dx.doi.org/10.2136/sssaj2005.0216. Berntsen, J., Petersen, B.M., Olesen, J.E., 2006. Simulating trends in crop yield and soil carbon in a long-term experiment − effects of rising CO2, N deposition and improved cultivation. Plant Soil 287, 235–245. http://dx.doi.org/10.1007/s11104-006-9070-y. Bolinder, M.A., Angers, D.A., Bélanger, G., Michaud, R., Laverdière, M.R., 2002. Root biomass and shoot to root ratios of perennial forage crops in eastern Canada. Can. J. Plant Sci. 82, 731–737.

18

European Journal of Agronomy 97 (2018) 11–19

R. Loges et al.

elevated pCO2 and nitrogen supply. New Phytol. 172, 117–126. http://dx.doi.org/10. 1111/j.1469-8137.2006.01796.x. Schuman, G.E., Janzen, H.H., Herrick, J.E., 2002. Soil carbon dynamics and potential carbon sequestration by rangelands. Environ. Pollut. 116, 391–396. http://dx.doi. org/10.1016/S0269-7491(01)00215-9. Six, J., Elliott, E.T., Paustian, K., 2000. Soil macroaggregate turnover and microaggregate formation: a mechanism for C sequestration under no-tillage agriculture. Soil Biol. Biochem. 32, 2099–2103. http://dx.doi.org/10.1016/S0038-0717(00)00179-6. Skinner, R.H., Comas, L.H., 2010. Root distribution of temperate forage species subjected to water and nitrogen stress. Crop Sci. 50, 2178–2185. http://dx.doi.org/10.2135/ cropsci2009.08.0461. Skinner, R.H., 2008. High biomass removal limits carbon sequestration potential of mature temperate pastures. J. Environ. Qual. 37, 1319–1326. http://dx.doi.org/10. 2134/jeq2007.0263. Smith, P., Smith, J.U., Powlson, D.S., McGill, W.B., Arah, J.R.M., Chertov, O.G., Coleman, K., Franko, U., Frolking, S., Jenkinson, D.S., et al., 1997. A comparison of the performance of nine soil organic matter models using datasets from seven long-term experiments. Geoderma 81, 153–225. Smucker, A.J.M., McBurney, S.L., Srivastava, A.K., 1982. Quantitative separation of roots from compacted soil profiles by the hydropneumatic elutriation system. Agron. J. 74, 500. http://dx.doi.org/10.2134/agronj1982.00021962007400030023x. Souchère, V., King, C., Dubreuil, N., Lecomte-Morel, V., Le Bissonnais, Y., Chalat, M., 2003. Grassland and crop trends: role of the European Union Common Agricultural Policy and consequences for runoff and soil erosion. Environ. Sci. Policy 6, 7–16. http://dx.doi.org/10.1016/S1462-9011(02)00121-1. Soussana, J.-F., 2008. Towards a full accounting of the greenhouse gas balance of european grasslands. In: Dolman, A.J., Valentini, R., Freibauer, A. (Eds.), The Continental-Scale Greenhouse Gas Balance of Europe, Ecological Studies. Springer, New York, pp. 263–283. http://dx.doi.org/10.1007/978-0-387-76570-9_13. Steingrobe, B., Schmid, H., Claassen, N., 2000. The use of the ingrowth core method for measuring root production of arable crops — influence of soil conditions inside the ingrowth core on root growth. J. Plant Nutr. Soil Sci. 163, 617–622. http://dx.doi. org/10.1002/1522-2624(200012)163:6<617:AID-JPLN617>3.0.CO;2-0. Steingrobe, B., Schmid, H., Gutser, R., Claassen, N., 2001. Root production and root mortality of winter wheat grown on sandy and loamy soils in different farming systems. Biol. Fert. Soils 33, 331–339. http://dx.doi.org/10.1007/s003740000334. Taube, F., Gierus, M., Hermann, A., Loges, R., Schönbach, P., 2014. Grassland and globalization − challenges for north-west European grass and forage research. Grass Forage Sci. 69, 2–16. http://dx.doi.org/10.1111/gfs.12043. Trnka, M., Olesen, J.E., Kersebaum, K.C., Skjelvag, A.O., Eitzinger, J., Seguin, B., Peltonen-Sainio, P., Rotter, R., Iglesias, A., Orlandini, S., Dubrovsky, M., Hlavinka, P., Balek, J., Eckersten, H., Cloppet, E., Calanca, P., Gobin, A., Vucetic, V., Nejedlik, P., Kumar, S., Lalic, B., Mestre, A., Rossi, F., Kozyra, J., Alexandrov, V., Semeradova, D., Zalud, Z., 2011. Agroclimatic conditions in Europe under climate change. Glob. Change Biol. 17, 2298–2318. http://dx.doi.org/10.1111/j.1365-2486.2011.02396.x. VandenBygaart, A.J., Janzen, H.H., Bremer, E., Ellert, B.H., 2007. Measuring change in soil organic carbon storage. Soil Sampling and Methods of Analysis, second edition. CRC Presshttp://dx.doi.org/10.1201/9781420005271. (ch3). Verbeke, G., Molenberghs, G., 2000. Linear Mixed Models for Longitudinal Data. Springer. van Noordwijk, M., Brouwer, G., Koning, H., Miejboom, F.W., Grzebisz, W., 1994. Production and decay of structural root material of winter wheat and sugar beet in conventional and integrated cropping systems. Agr. Ecosyst. Environ. 51, 99–113. http://dx.doi.org/10.1016/0167-8809(94)90037-X. Walia, M.K., Baer, S.G., Krausz, R., Cook, R.L., 2017. Deep soil carbon after 44 years of tillage and fertilizer management in southern Illinois compared to forest and restored prairie soils. J. Soil Water Conserv. 72, 405–415. http://dx.doi.org/10.2489/jswc.72. 4.405.

Maloney, T.S., Silveira, K.G., Oplinger, E.S., 1999. Rotational vs. nitrogen-fixing influence of soybean on corn grain and silage yield and nitrogen use. J. Prod. Agric. 12, 175–187. http://dx.doi.org/10.2134/jpa1999.0175. Meersmans, J., Martin, M.P., Lacarce, E., Orton, T.G., De Baets, S., Gourrat, M., Saby, N.P.A., Wetterlind, J., Bispo, A., Quine, T.A., Arrouays, D., 2013. Estimation of soil carbon input in France: an inverse modelling approach. Pedosphere 23, 422–436. http://dx.doi.org/10.1016/S1002-0160(13)60035-1. Mohrlok, U., 2009. Bilanzmodelle in der Grundwasserhydraulik: Quantitative Beschreibung von Strömung und Transport im Untergrund. Habilitation thesis. KIT Scientific Publishing, University Karlsruhe. http://dx.doi.org/10.5445/KSP/ 1000012015. Muylle, H., Van Hulle, S., De Vliegher, A., Baert, J., Van Bockstaele, E., Roldán-Ruiz, I., 2015. Yield and energy balance of annual and perennial lignocellulosic crops for biorefinery use: a 4-year field experiment in Belgium. Eur. J. Agron. 63, 62–70. http:// dx.doi.org/10.1016/j.eja.2014.11.001. Nevens, F., Reheul, D., 2001. Crop rotation versus monoculture; yield, N yield and ear fraction of silage maize at different levels of mineral N fertilization. NJAS Wageningen J. Life Sci. 49, 405–425. http://dx.doi.org/10.1016/S1573-5214(01) 80026-9. Petersen, B.M., Berntsen, J., Hansen, S., Jensen, L.S., 2005a. CN-SIM: a model for the turnover of soil organic matter. I. Long-term carbon and radiocarbon development. Soil Biol. Biochem. 37, 359–374. http://dx.doi.org/10.1016/j.soilbio.2004.08.006. Petersen, B.M., Jensen, L.S., Hansen, S., Pedersen, A., Henriksen, T.M., Sørensen, P., Trinsoutrot-Gattin, I., Berntsen, J., 2005b. CN-SIM: a model for the turnover of soil organic matter. II. Short-term carbon and nitrogen development. Soil Biol. Biochem. 37, 375–393. http://dx.doi.org/10.1016/j.soilbio.2004.08.007. Pinheiro, J.C., Bates, D.M., DebRoy, S., Sarkar, D., R Core Team, 2016. Nlme: Linear and Nonlinear Mixed Effects Models. Plauborg, F., 2002. Simple model for 10 cm soil temperature in different soils with short grass. Eur. J. Agron. 17, 173–179. http://dx.doi.org/10.1016/S1161-0301(02) 00006-0. Propheter, J.L., Staggenborg, S.A., Wu, X., Wang, D., 2010. Performance of annual and perennial biofuel crops: yield during the first two years. Agron. J. 102, 806–814. http://dx.doi.org/10.2134/agronj2009.0301. Qin, R., Stamp, P., Richner, W., 2006. Impact of tillage on maize rooting in a Cambisol and Luvisol in Switzerland. Soil Tillage Res. 85, 50–61. http://dx.doi.org/10.1016/j. still.2004.12.003. R Core Team, 2016. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Rasse, D.P., Mulder, J., Moni, C., Chenu, C., 2006. Carbon turnover kinetics with depth in a French loamy soil. Soil Sci. Soc. Am. J. 70, 2097–2105. http://dx.doi.org/10.2136/ sssaj2006.0056. Rees, R.M., Bingham, I.J., Baddeley, J.A., Watson, C.A., 2005. The role of plants and land management in sequestering soil carbon in temperate arable and grassland ecosystems. Geoderma 128, 130–154. http://dx.doi.org/10.1016/j.geoderma.2004.12.020. Reijs, J.W., Sonneveld, M.P.W., Sørensen, P., Schils, R.L.M., Groot, J.C.J., Lantinga, E.A., 2007. Effects of different diets on utilization of nitrogen from cattle slurry applied to grassland on a sandy soil in the Netherlands. Agric. Ecosyst. Environ. 118, 65–79. http://dx.doi.org/10.1016/j.agee.2006.04.013. Reinsch, T., Loges, R., Kluß, C., Taube, F., 2018. Renovation and conversion of permanent grass-clover swards to pasture or crops: effects on annual N2O emissions in the year after ploughing. Soil Tillage Res. 175, 119–129. http://dx.doi.org/10.1016/j.still. 2017.08.009. Schmeer, M., Loges, R., Dittert, K., Senbayram, M., Horn, R., Taube, F., 2014. Legumebased forage production systems reduce nitrous oxide emissions. Soil Tillage Res. 143, 17–25. http://dx.doi.org/10.1016/j.still.2014.05.001. Schneider, M.K., Luscher, A., Frossard, E., Nosberger, J., 2006. An overlooked carbon source for grassland soils: loss of structural carbon from stubble in response to

19