Introduction

The potential contribution of agriculture to carbon (C) sequestration is considerable (Lal 2004) even though the carbon sink function of agricultural soils is finite and temporally limited (Smith 2004). Within agricultural systems, it has been suggested that conversion to organic farming would be one appropriate and effective measure to promote C sequestration due to enhanced soil organic matter (SOM) levels compared with conventional farming (Freibauer et al. 2004; Lal 2006).

In field surveys, a number of studies (e.g., Foissner 1992; Munro et al. 2002; Piorr and Werner 1998; Pulleman et al. 2003; Reganold 1988) have identified higher SOM levels as a characteristic property of organically managed compared with conventionally managed soils. These observations are also supported by results from long-term field experiments (e.g., Fließbach et al. 2007; Hepperly et al. 2006; Mazzoncini et al. 2010). Using the CENTURY model (Parton et al. 1987), Foereid and Høgh-Jensen (2004) calculated an increase in soil organic carbon (SOC) of 10–40 g C m−2 year−1 over a period of 50 years after conversion of arable land from conventional to organic farming practices. However, higher SOM levels are not always found in organic compared with conventional farming systems (van Diepeningen et al. 2006; Gosling and Shepherd 2005; Marinari et al. 2006). Recently, Kirchmann and Bergstroem (2008) concluded that organic farming will lead to a loss of soil organic matter compared with intensive crop production, at least where comparable crop yield levels are assumed as basic condition.

Consequently, it is necessary to carry out appraisals at farming system scale and use a method for the quantitative analysis of actual SOM balances, as suggested by Leifeld and Fuhrer (2010) and Shepherd et al. (2002). Both papers conclude that management impact on soil organic matter is likely to be dependent on actual organic matter input and not on farming system labels. We therefore hypothesize that the impact on SOM levels is not an intrinsic characteristic of any farming system but is the result of the actual structure of the farming system, in particular, the composition and management of crop rotations, and the availability and utilization of organic manures. Our aim is therefore to contribute to this discussion by (1) analyzing the impact of specific organic versus conventional farming systems on SOM levels in four long-term field experiments in Germany and Switzerland with a simple SOM balance model, and (2) scaling-up the results with scenario calculations based on data on farm system structures on organic and conventional farming in Germany obtained from agricultural censuses.

Materials and methods

Evaluation of the HU-MOD model in the comparison of organic and conventional farming systems

Four long-term field experiments, three in Germany and one in Switzerland, were included in the model evaluation. Relevant site and management data are given in Tables 1 and 2, and more detailed descriptions of the experiments are provided by Debruck (2003), Deike et al. (2008), Mäder et al. (2006), and Reinicke et al. (2003). Two of the four long-term field experiments are located in eastern Germany on silt soils with a high potential yield level (BL, BN). One other experiment is located in eastern Germany, too, but on sand soils (DDF). The fourth experiment (DOK) is located on loam soils near Bale (Switzerland). All of these field experiments compare organic versus conventional farming. Due to limitations in capacity, we evaluated only some treatments from each experiment, where these treatments were considered by the site managers to best reflect typical farming practice under organic and conventional management.

Table 1 Overview of long-term field experiments (LTFE) included
Full size table
Table 2 Description of LTFE treatments included
Full size table

For each experimental treatment included, SOC change in the topsoil was calculated for the survey period using data provided for each site, including additional data from a comparative survey carried out in all experiments in Spring 2006. SOC was usually analyzed according to ISO 10694 (1995) in all experiments.

Furthermore, we calculated SOM balances for all treatments using the SOM balance model HU-MOD (Brock et al. 2012). Briefly, the balance is calculated as the sum of management-induced SOM loss (negative) and SOM supply (positive). Soil organic matter loss in HU-MOD is calculated by estimating the contribution of soil organic nitrogen (N) to crop nutrition, considering the supply by other pools (e.g., recently added organic fertilizers) and rates of mineralization stimulated by cultivation operations. Constitutively, SOC loss is calculated from SON loss assuming a default soil C/N ratio of 11.6. SOM supply is estimated based on organic inputs in plant biomass and organic fertilizers, taking account of organic matter losses as a result of rapid turnover of fresh OM additions. To run the model, the minimum data required are information on crop type together with the type and amount of fertilizer used in each year in the assessment period. For better calibration, the yields of main and side products (e.g., grain and straw) can be used as inputs. Side product yields and yield amounts of fodder crops were not available for all experiments. Data availability for all experiments is shown in Table 1.

Regression analysis was used to relate the calculated SOM balances to the measured SOC change between treatments for each experiment with SOM balance as the driving variable (x) and SOC change as the dependent variable (y). The impact of site-specific factors are not fully recognized in HU-MOD, hence the regression coefficient (b) varies with site; the coefficient of determination (r 2) was considered as the key assessment parameter. The results are then interpreted on a three-step ordinal scale (positive–none–negative deviation of impact on SOC in scenario comparison; Brock et al. 2012). The advantages and limitations of the model will be addressed in the discussion.

SOM balances for cropland in Germany

SOM balances for conventional versus organic farming in Germany were then calculated using HU-MOD, applying three different scenarios that differ in complexity with regard to the data used as data inputs to the model. All the input values estimated and then applied in each scenario are shown in Table 4.

Scenario A is a pure cropland balance based on a new survey of the structure of agricultural production in Germany by BLE (2012). These data are comprised of absolute organic and conventional cropland area for all agricultural crops, respectively, but there are no data on livestock and fertilization. As the applied SOM balance model requires input data on mineral N supply (Brock et al. 2012), we applied an estimated mean application of 100 kg N ha−1 year−1 for conventional cropland based on a nitrogen balance survey carried out for Germany by Panten et al. (2009).

Scenario B additionally includes data on straw incorporation and the use of catch crops. As the BLE survey does not include data on catch crop area, straw yield, or incorporation, we estimated these parameters from Statistisches Bundesamt (2011) for catch crop area and IFEU (2008) on straw removal. Both sources provide average values for German cropland, without regional differentiation or differentiation according to the farming system.

Finally, scenario C includes additional input data on animal manure application that was calculated from farming-system-related data of BMELV (2007) on livestock numbers on organic and conventional farms, and in addition, average data for manure application for both grassland and cropland in Germany by Statistisches Bundesamt (2011).

Results

Model evaluation in field experiments

The SOC changes indicated both increases and decreases in SOC in the experimental treatments over the period of the survey (Table 3). In the BL experiment, the effect of different cropping system structure types was stronger than the impact of farming systems. SOC change followed the order CON–MF > ORG–MF > CON–SFL > ORG–SFL. With BN, the order was ORG–MF > ORG–SFC > CON–SFC, and in DDF, both systems had very negative SOC change with the ORG–MF system showing a greater loss of SOC than the conventional reference. In the DOK experiment, SOC change was slightly positive in BIODYN and CON–MF, slightly negative in ORG–MF, and strongly negative in CON–MIN.

Table 3 SOC change and SOM balances for organic versus conventional treatments in four long-term field experiments
Full size table

SOM balances calculated by HU-MOD do not closely match the estimated values for SOC change (Table 3), but the direction and relative extent of SOC change still is reflected in the variation in the SOM balance (Fig. 1). These relationships were significant, even though correlation coefficients often only were around r = 0.5; the regression coefficients strongly differed from 1, and the intercept was very variable as well, indicating that an absolute quantification of SOC change was not possible based on SOM balances calculated by HU-MOD.

Fig. 1
figure 1

Relation between SOM balances and SOC change in four long-term field experiments comparing organic and conventional farming systems. SOC change calculated as linear trend of SOC change in the survey period, SOM balances calculated with the HU-MOD model (Brock et al. 2012)

Full size image

Upscaling—cropland SOM balances for organic versus conventional farming in Germany

Cropping structure differs considerably between organic and conventional farming systems, with a higher proportion of perennial legumes in organic farming systems (Table 4). The pure cropland balance (scenario A) therefore shows a higher SOM balance for the organic cropland. Where catch crop area and straw fertilization are considered (scenario B), SOM balances increase in with both farming systems but remain higher in organic farming systems.

Table 4 SOM balance calculation for organic versus conventional cropland in Germany
Full size table

With scenario C (inclusion of animal manure availability), SOM balances for the two systems are at a comparable, slightly negative level. Conventional farming systems have a higher average livestock stocking rate of 1.28 LU ha−1 compared with 0.63 LU ha−1 with organic farming. In the scenario C calculations, it has been assumed that roughly 50 % of available animal manure is applied to arable land in conventional agriculture, as reported by Statistisches Bundesamt (2011). No specific numbers exist for organic farming, but, as extensive grassland management is very common with organic farming in many regions (Statistische Ämter 2011), we preliminarily assumed that 60 % of the available manure is applied on arable land in organic farming systems.

Discussion

Method performance in the long-term field experiments

The assessment of the SOC change from a series of measured SOM/SOC parameters is problematic because of the high spatial heterogeneity of SOM distribution (Hülsbergen et al. 2000; Papritz and Webster 1995). Moreover, the reliability of SOC and SON trends was limited for BL, BN, and DDF, as trends had to be calculated based on few, non-annual values (BN, DDF) or short time periods (BL). The significant correlation between SOC change modelled SOM balance at the farm-type aggregation level suggests that the simple model can be used in practice for the analysis of the likely trends in the impact of farming systems and/or farming structures on SOM level. The relevance of data quality also became evident in a separate validation of HU-MOD, where the model performed best in a field experiment with highly detailed data on SOM levels and SOM level development (Brock et al. 2012). This HU-MOD validation also showed a better performance of the model when considering treatment means rather than plot data, which is in line with the results in this paper.

The main advantage of the HU-MOD tool is that the model does not require any information on soil parameters, and texture or current SOM contents in particular. This allows the easy application of the model where such data are not available at the required scale to allow true quantitative modelling, as was the case in our survey. Of course, the advantage of this approach also leads to the limitation that quantification of absolute changes is not possible. This situation is recognized in the model scope (Brock et al. 2012) and is the reason why we did not apply common validation criteria that refer to a quantitative divergence between observed and predicted values, as root mean square error (RMSE), or modelling efficiency (EF) (see overview by Smith et al. 1997). However, the model allows for the assessment of the likely trends of impact on SOC change (better, indifferent, worse) in assessments of farming systems or management change where more sophisticated modelling is prevented by lack of (soil) data.

Our results show that organic farming does not in general show higher SOM balances and a more favorable impact on SOC levels than conventional farming. The BL and DDF experiments comprise a comparison of integrated versus organic farming systems; under these conditions, conventional (integrated) farming had higher SOM balances and a more favorable impact on SOC levels than the organic farming systems with which they were compared. This observation is in line with results from long-term fertilization experiments, where additional mineral N fertilization usually produced higher yields (and returns of crop residues) which lead to SOM levels, if crop rotations and organic fertilization were held constant (Carlgren and Mattsson 2001; Shevtsova 2003). Leithold (1991) also concluded that mineral nitrogen fertilization of arable crops decreases the demand for nitrogen from SOM mineralization and thus contributes to a lower demand for organic matter supply to maintain SOM levels. Leithold’s assumption is supported by the analytical results of Nissen and Wander (2003), and this mechanism has been implemented in the SOM balance model HU-MOD. A possible priming effect of mineral fertilization (Kuzyakov et al. 2000) must also be taken into account. However, the priming effect may be overcompensated by the lower N demand from SOM mineralization and the higher supply of crop residues as an effect of higher yield levels.

In the well-known DOK experiment, the two conventional systems M and K feature the same crop rotation as the organic systems D and O but differ in fertilization (Table 1). Crop rotation effects on SOM dynamics therefore do not occur, even though they may be more crucial than fertilization effects (Gregorich et al. 2001; Liu et al. 2006). Within identical crop rotations, SOM inputs in crop residues will be higher within conventional farming systems, if yield levels are higher and nitrogen demand is supplied by fertilizers, and not by a higher absorption of SON (Brock et al. 2011). As a consequence, the different fertilization regimes in the DOK treatments did not lead to a clear differentiation between organic and conventional treatments (Leifeld et al. 2009). In this trial, significantly higher SOM levels were only observed comparing the D system (biodynamic management with compost-based fertilization) to the M system (conventional management with mineral fertilizer application only) (Fließbach et al. 2007; Mäder et al. 2002).

SOM balance for organic versus conventional cropland in Germany

For cropland in Germany, organic farming systems have a less negative SOM balance compared with conventional systems where no animal manure applications are considered. Where inputs of livestock manure are estimated, there is no difference between the SOM balances of the farming systems. However, these balances are a mean for the whole cropland area and therefore do not reflect the situation on single farms. From the BMELV (2007) report, it becomes obvious that livestock and thus animal manure application shows a strong regional concentration in Germany that does not coincide with the main cash crop production areas. According to that source, livestock density on arable cash crop farms on average was only about 0.2 LU ha−1 within both conventional and organic farming systems. On the other hand, livestock density was roughly 1.6 LU ha−1 on average on conventional, but only 0.9 LU ha−1 on organic animal production farms. Furthermore, Zerger and Haas (2003) reported that livestock on organic farms was between 0.06 LU and 1.26 LU ha−1 in the German state of North Rhine-Westfalia when assessment was made at a communal level. Best (2007) found that at least 10 % of organic farms in a comprehensive survey had no livestock at all, while, on the other hand, roughly 30 % of farms had no cropland but only had grassland as they specialized in livestock production. With conventional farming, the specialization is even stronger: According to the 2010 official census, 43 % of farms in Germany were specialized in livestock; 25 % were specialized in arable cash crops, and only 15 % were mixed farms (Statistische Ämter 2011).

Specialization of farms and regional concentration of particular enterprises will have a considerable impact on the true SOM balances at farm and regional scales. However, the scenarios A and B without animal manure application point towards negative SOM balances in arable cash crop farming in both organic and conventional management systems. On the other hand, the concentration of livestock may lead to an excessive supply of nitrogen with regard to crop N demand balances in these regions, as indicated by the nitrogen balances (Osterburg and Schmitt 2008). An excessive supply of nitrogen thus decreases the N use efficiency and N utilization in the soil plant system. This situation is recognized in the HU-MOD model with the calculation of the N supply from SOM mineralization (Brock et al. 2012). While moderate N supply with fertilizers (i.e., N amounts that may be absorbed by the crops) leads to a decreased demand of N from SOM mineralization in the model, excessive supply will not lower the SOM loss parameter below a crop and yield specific threshold value. With regard to the calculated cropland scenarios, this situation means that a (considerable) amount of available N as indicated by average livestock size may actually not be applicable in the SOM balance due to specialization and concentration processes. Furthermore, the possibility of priming effects of N supply has been reported (Kuzyakov et al. 2000) but has not yet been included in the model. Even though balances calculated with HU-MOD did not have a strong link to absolute SOC change in the field experiments, there is some evidence from surveys on SOC change in European croplands that the true SOM balance on arable land in fact may be negative on average (Bellamy et al. 2005; Kutsch et al. 2010; Schrumpf et al. 2011).

It would be desirable to conduct a spatially more differentiated assessment supported by GIS, as, e.g., done by Sleutel et al. (2007) in a case study on agricultural management impact on SOM change in arable soils of Flanders (Belgium) at the community scale. These authors thus yielded spatially more explicit results. However, a database that would allow for a comparable spatially disaggregated assessment could not be compiled for our calculations and is not publicly available for Germany at present. Available GIS databases for European countries with a comparable high resolution, as LUCAS data (European Commission 2009), allow for a spatial analysis of cropland with different crops and could, with repeated surveys, be utilized to calculate a cropland SOM balance (e.g., Britz et al. 2011; Leip et al. 2008), comparable to Scenario A in our study. However, LUCAS data do not provide differentiated information on actual crop rotations, fertilization, and farming systems and would therefore have to be supplemented by other databases for SOM balance calculation.

The high spatial aggregation substantially limits the power of our results but, still, the results both in general approve the applicability of a simple SOM balance approach for the intended assessment and provide valuable information on structural aspects of organic versus conventional farming that seem to be drivers of SOM dynamics. A regional, GIS-based modelling approach would, however, be highly desirable, and this would be particularly valuable, if explicit data on SOM change were included for validation of the results.

Conclusions

It was possible to explain the differentiation of SOM levels and SOM level development between treatments in long-term field experiments applying the SOM balance model HU-MOD. Higher SOM balances under comparable basic conditions (i.e., within a single field experiment) coincided higher SOM levels and a relatively more positive SOM level development. We therefore conclude that a shift from conventional to organic agriculture will have a positive impact on SOM levels, if the SOM balance level of the new management is higher than with the former management. Referring to the actual cropland structure in Germany, SOM balances of organic farming were more positive than with conventional farming in the scenarios without consideration of animal manure application, but balances for the two systems were not different in the scenario with consideration of animal manure application as an average for the whole cropland. However, the considerable specialization patterns in the agricultural production sector challenge the validity of this calculation, as especially animal manure availability and application shows a strong regional concentration that would affect the mean cropland balance if calculated based on disaggregated data.

Altogether, our results support the applicability of simple SOM balance models for the comparison of farming systems and cropland structures with regard to their impact on SOC levels but at the same time imply a strong demand for studies with a sufficient spatial and structural resolution.