The trial was conducted on the experimental farm of the University of Kassel, the state-owned Frankenhausen farm (51.412 N, 9.439 E; 231 m above sea level), on the “Untere Kiebitzbreite” field. The soil type is Haplic Luvisol. The soil texture in the plough horizon ranges from strong clayey silt (Ut4) to very silty clay (Tu4) [10]. The pH value in the plough horizon is 6.6, and the C o ⁢ r ⁢ g content is 1.16%.

The temperature profile during the trial period from 2011 to 2013 was similar to the 30-year average (Figure 1). The greatest deviation was a significantly cooler March 2013.

Precipitation was lower than average in many months. May 2013 stood out due to its above-average total rainfall (data from 2011 to 2013 from the weather stations at Frankenhausen, Kassel Calden airport and Kassel Harleshausen; long-term average data from the German Meteorological Service (DWD), Kassel weather station).

A multi-year trial was carried out in 2007 as a randomized block design with four field replications. Originally the experiment was set up to test if the stubble cleaner is as effective for the control of Cirsium arvense as pre crop alfalfa. Therefore, one crop rotation with alfalfa grass was established in combination with conventional tillage and a cereal based crop rotation in combination with conventional tillage and in combination with reduced tillage. The conventional tillage in autumn consisted of a double stubble breaking with the cultivator followed by ploughing to a depth of 25 cm, and reduced tillage was carried out exclusively with the stubble cleaner to a maximum depth of 10 cm [11].

Alfalfa grass was sown in 2008 as a second crop rotation. The mixture consisted of 80% alfalfa and 20% grass (based on the seed weight). The cultivation of alfalfa grass lasted three years. The alfalfa grass was mowed two to three times per year, and the harvested biomass was removed. Tillage in this crop rotation took place after 2010 in the same way as the conventional tillage in the cereal-based crop rotation (Figure 2). The details on tilling and sowing are given in Table 1.

In autumn 2011, the single factor trial was extended by the factor of green manure in a split plot design. The fourfold replication was maintained. The following legume and non-legume green manure species were cultivated in pure stands and in a mixture of two species (Table 2):

In the trial year 2011/12, all the treatments were sampled or assessed (= 84 plots). In the trial year 2012/13, only selected plots were sampled or assessed in order to check whether the various green manure species still had an effect in the second year of cultivation after growing green manure. The following treatments were selected for this purpose: Bare fallow, S. alba and V. sativa (= 36 plots).

In the trial year 2011/12, growing the green manure crops was followed by sowing the main crop oats (Avena sativa, Cv. Scorpion, 400 germinable grains per m 2 ). In the trial year 2012/13, field beans were sown (Vicia faba, Cv. Bioro, 40 germinable grains per m 2 ).

No fertilization was carried out during the trial period. No weed control was carried out in 2011/12. The area was hoed once in 2012/13.

The above-ground biomass yield of green manure was determined on November 17, 2011. For this purpose, a square area with a side length of 1.5 m per plot was cut off directly above the ground by hand and then weighed, and the dry matter was determined by drying a subsample of the harvested material at 60 ∘ C for 48 hours (until constant weight). In addition to the determination of yield, the samples were used to analyse the total nitrogen content (N t ) and carbon content (C t ) of the above-ground biomass.

Oat was also harvested by hand on August 13, 2012. For this purpose, half a square metre of land was harvested twice per plot. The beans were also harvested by hand on August 26, 2013. For this purpose, five metres of two rows of crops per plot were cut off just above the ground.

To analyse the nitrogen available to plants in the soil, samples up to a depth of 90 cm in 30 cm intervals were taken on three (2011/12) or four (2012/13) dates spread over the trial year. If sampling to a depth of 90 cm was not possible due to the soil conditions, the soil was sampled to a depth of 60 cm. A mixed sample was taken from three cores per plot. The samples were immediately cooled and carried to a freezer as soon as possible. The samples were examined according to DIN ISO 14255 and DIN EN ISO 11732 in the Landesbetrieb Hessisches Landeslabor, Kassel. The entire sample material was tested for NO 3 -N, and the samples from the top soil were also tested for NH 4 -N. For the top soil N m ⁢ i ⁢ n content was calculated by summating the NO 3 -N and NH 4 -N values, whereas N m ⁢ i ⁢ n for the subsoil consists only of the NO 3 -N values, as NH 4 -N is negligible below the top soil on this site.

In spring 2013, the soil was tested for the content of the microbial biomass in addition to the N m ⁢ i ⁢ n sampling. The samples were taken according to the layers of the tillage depths, i.e., 0–10 cm (corresponding to the working depth of the stubble cleaner), 10–25 cm (corresponding to the plough depth) and 25–50 cm (an uncultivated layer). A mixed sample was taken from seven cores per large plot according to the crop rotation—tillage system treatments. The analysis was performed using the chloroform fumigation extraction method [12]. Bulk density was also examined in these layers. For its determination calibrated soil rings of 100 cm 3 volume were inserted vertically in the middle of each layer. They were dried at 105 ∘ C until weight constancy. The dry weight was determined and after subtracting the weight of each ring and the lids the soil weight of the ring volume was obtained and bulk density could be calculated.

The mean values and standard errors were calculated to describe the distribution of N m ⁢ i ⁢ n , the C and N contents, and the green manure and main crop yields. Each data set was assessed for normally distributed residuals (Kolmogorov-Smirnov test). The large plot, small plot and block as fixed factors were examined for significant effects and interactions using a univariate analysis of variance. If the analysis of variance showed significant effects or interactions, a post hoc test (Tukey-B) was then carried out for the factor combination of green manure x tillage or for the individual factors (alpha ≤ 0.05). Significant differences between treatments are indicated by different letters in the figures except in the figures for N m ⁢ i ⁢ n -content. For these figures letters are shown in tables in the attachment for better legibility.

The statistical analyses were performed with SPSS-21 [13].

The analysis of variance of green manure crop and tillage/crop rotation system for green manure above ground biomass yield, green manure above ground biomass N content and green manure C/N ratio showed a significant influence of the green manure crops, the tillage/crop rotation systems and a significant interaction at the 0.001 probability level each (Table 3).

The stubble cleaner system (SC) led to the lowest green manure yield for each green manure species, with the exception of V. sativa, which had a similar yield in each system. L. perenne, P. tanacetifolia, S. alba and the mixture of S. alba and T. resupinatum each had the lowest yield, approximately 0.5 t dry matter per hectare, in the stubble cleaner system. T. resupinatum could not be sampled due to its poor field emergence. The highest yield, approximately 1.5 t DM per hectare, was achieved by S. alba in the plough-alfalfa grass (PLALF) system and by V. sativa in all three systems (Figure 3).

The green manure in the PLALF system had the highest N content, and green manure in the SC system had the lowest, with the exception of V. sativa, which had the highest N content of all green manure crops in all three tillage systems. The N content of V. sativa ranged from 47.3 kg N ha - 1 in the PLALF system to up to 57.5 kg N ha - 1 in the PL system. In the SC system, the N content of V. sativa was 52.8 kg N ha - 1 . In this system, the difference between V. sativa and the other green manure species was particularly great (Figure 4).

Of the green manure species studied, L. perenne had the highest C/N ratio, ranging from 16.1 in the PLALF system to 21.3 in the SC system. S. alba had a slightly lower C/N ratio (15.3 – 16.7) than L. perenne. The mixture of S. alba and T. resupinatum had a C/N ratio between 12.8 and 17.1. P. tanacetifolia had a C/N ratio between 12.8 and 15.9. V. sativa had the lowest C/N ratio (9.4-10.6). For each green manure treatment except for V. sativa, the C/N ratio was lowest in the PLALF system and highest in the SC system. For the C/N ratio of V. sativa, there were no detectable differences between the systems (Table 4).

The analysis of variance of N m ⁢ i ⁢ n from the 15.03.2012 showed a significant influence of the green manure crops, the tillage/crop rotation system and a significant interaction for all three soil layers at the 0.001 probability level each (Table 5). On the 15.05.2012 there was a significant influence of the tillage/crop rotation system for all three soil layers at the 0.001 probabililty level each and a significant influence of the green manure crops and a significant interaction for the top soil at the 0.01 probability level each (Table 5). On the 05.09.2012 there was a significant influence of the tillage/crop rotation system for the two sampled soil layers at the 0.001 probabililty level each and there was a significant interaction at the 0.05 probability level for the top soil and at the 0.01 probability level for the layer 30–60 cm (Table 5).

On March 15, 2012, V. sativa achieved the significantly highest N m ⁢ i ⁢ n values in all tillage systems. In the PLALF system, the difference between V. sativa and the other green manure treatments was smaller than that in the PL and SC systems. L. perenne led to the lowest N m ⁢ i ⁢ n values in the PLALF and PL systems (Figure 5a and Table NaN).

On May 15, 2012, the PLALF system had the significantly highest N m ⁢ i ⁢ n content in each layer. With regard to green manure, L. perenne and bare fallow led to the significantly lowest value, but V. sativa led to the significantly highest N m ⁢ i ⁢ n content in the 0–30 cm layer in each tillage system. T. resupinatum led to similarly high N m ⁢ i ⁢ n values in the PL system as V. sativa but not in the SC system (Figure 5b and Table NaN).

On September 05, 2012, the N m ⁢ i ⁢ n content in the PLALF system was approximately twice as high as that in the other two tillage systems. The N m ⁢ i ⁢ n content was significantly highest in the SATR treatment of the PLALF system. In the other two tillage systems, the N m ⁢ i ⁢ n content in the SATR treatment was as low as that in the other green manure treatments (Figure 5c and Table NaN).

The analysis of variance of oats yield showed a significant influence of the green manure crops, the tillage/crop rotation systems and a significant interaction each on the 0.001 probability level, both for grain yield and straw yield.

The SC system had the lowest grain yield, the PLALF system had the highest grain yield, and the PL system had a medium grain yield. Bare fallow, L. perenne, P. tanacetifolia, the mixture and T. resupinatum led to the lowest grain yields in the SC and PL systems, whereas in the PLALF system, all green manure treatments and the bare fallow treatment led to the highest yields. In the SC system, S. alba also led to the lowest oat yields. The only green manure that led to the highest grain yields in the SC system was V. sativa (Figure 6).

The SC and PL systems had significantly lower straw yields than did the PLALF system. S. alba resulted in a significantly lower straw yield in the SC system than in the PLALF system. V. sativa led to the highest straw yields in each system (Figure 7).

The analysis of variance of N m ⁢ i ⁢ n from all the sampling dates in 2013 showed a significant influence of the tillage/crop rotation system for all three layers, with the exception of the top layer on September 08, 2013. The green manure crops had no significant effect on N m ⁢ i ⁢ n in the second year after green manure crop cultivation, and there was no significant interaction (Table 6).

On April 08, 2013, and May 23, 2013, the N m ⁢ i ⁢ n content in the PLALF system was significantly higher in all three layers than in the other two systems (Figure NaN; Table NaN). On June 18, 2013, the N m ⁢ i ⁢ n content in the 0–30 cm layer was significantly higher in the PLALF system than in the PL system. The N m ⁢ i ⁢ n content in the 0–30 cm layer in the SC system was between that of the other two systems. In the 30–60 cm and 60–90 cm layers, the PLALF system had the significantly highest N m ⁢ i ⁢ n content (Figure NaN and Table NaN).

On September 08, 2013, there were no significant differences in the N m ⁢ i ⁢ n content among the different systems in the 0–30 cm layer. The high N m ⁢ i ⁢ n value for the V. sativa treatment in the PL system was due to the unusually high value of one subsample, from which the large standard error also originated. The PLALF system had the significantly highest N m ⁢ i ⁢ n content in the 30–60 cm layer, and the SC system had the significantly lowest N m ⁢ i ⁢ n content. The PL system had a medium N m ⁢ i ⁢ n content. In the 60–90 cm layer, the PLALF system again had the significantly highest N m ⁢ i ⁢ n content compared to that of the other two systems. However, the differences among the treatments amounted to only a few kilograms per hectare (Figure NaN and Table NaN).

The analysis of variance of N m ⁢ i ⁢ c and C m ⁢ i ⁢ c showed a significant influence of the tillage/crop rotation system for the top layer on the 0.01 probability level. After six years of differentiated tillage, the SC system had significantly more N m ⁢ i ⁢ c and C m ⁢ i ⁢ c in the top layer than the PLALF and PL systems. In the two lower layers, the three systems did not differ significantly (Figure NaN).

In terms of the bean and straw yield of the field beans, the analysis of variance showed no significant influence neither of the tillage/crop rotation system nor of the green manure treatments in the second year after their cultivation. The bean yield at 86% DM varied between 3.1 t ha - 1 with the V. sativa treatment in the PL system and 3.7 t ha - 1 with the bare fallow treatment in the PLALF system (data not shown). The straw yield at 86% DM was between 3.5 t ha - 1 with the V. sativa treatment in the SC system and 4.3 t ha - 1 with the bare fallow treatment in the PLALF system (data not shown).

V. sativa proved to be very effective in achieving a high N m ⁢ i ⁢ n content in the SC and PL systems, which both had a reduced N m ⁢ i ⁢ n content compared to the PLALF system. V. sativa also led to high N m ⁢ i ⁢ n values in the PLALF system, but the choice of the green manure crop was of minor importance in this system. The greatest differences between the N m ⁢ i ⁢ n values of the V. sativa-treatment and the other treatments were found in the SC system. The high N m ⁢ i ⁢ n values of the V. sativa-treatment can be attributed to the high N content of V. sativa, especially in the PL and SC systems, as well as to its low C/N ratio (Figure 4 and Table 4). Two regressions of N m ⁢ i ⁢ n on March 15, 2012 in the 0–30 cm layer on the N content of green manure crops and the C/N ratio of green manure crops show a slightly greater influence of the N content (R 2 = 0.7555) than the C/N ratio (R 2 = 0.6278) on the N m ⁢ i ⁢ n content of the soil.

T. resupinatum as a second legume species in this trial was not as effective as V. sativa due to its low growth. Heavy rainfall two days after the drilling led to silting which probably hampared the germination of this small-seeded legume. T. resupinatum seems to be more sensitive to growing conditions than other green manure crops. For T. resupinatum strongly fluctuating N contents between 140 kg ha - 1 and 15 kg ha - 1 with yields between 3.5 t ha - 1 and 0.4 t ha - 1 above ground dry matter biomass are reported in a study by Grosse et al. [14].

The C/N ratio of T. resupinatum could not be determined due to its low growth. The C/N ratio of T. resupinatum is given as 15.8 by Fageria et al. [15], i.e., higher than the value of V. sativa (9.4 to 10.6) measured in this experiment. Due to the low growth of T. resupinatum, very little N was likely bound.

The mixture of T. resupinatum and S. alba suffered from the low growth of T. resupinatum. However, despite half the amount of seeding density for S. alba in the mixture compared to S. alba in pure stand, the yield and the N content of the mixture was comparable to that in pure stand in the PLALF and the SC system.

L. perenne, as an overwintering species, was the only treatment in the PLALF and PL systems that produced lower N m ⁢ i ⁢ n values before being tilled into the soil than the bare fallow treatment (Figure 5a). Even after being tilled in, L. perenne still had the lowest N m ⁢ i ⁢ n values in the PLALF and PL systems (Figure 5b). L. perenne had both the lowest N content and the highest C/N ratio of the green manure crops tested (Figure 4 and Table 4) in each system. A high C/N ratio can lead to a temporary immobilization of N, as reported in a study by Baggs et al. [16].

The bare fallow was overgrown with weeds in every system despite flame weeding. This result means that N was also retained by the vegetation in these treatments. N mineralisation in all bare fallow treatments was comparable or higher in spring than N mineralisation in the non-legume green manure treatments. This result is in line with results from Baggs et al. [16], which showed that N retention through weed cover was as effective as that retained through sown green manure crops. However, it can be assumed that uncontrolled growth will increase the weed pressure in subsequent field crops.

The biomass yield of S. alba was significantly higher in the PLALF and PL system than in the SC system. This promotes the hypothesis that high N using crops will suffer more from reduced tillage. However, the N m ⁢ i ⁢ n level of the PL and the SC system was similar at the beginning of 2012, albeit it was somewhat lower in the SC system than in the PL system. This difference, despite beeing small, may result in a difference in yields on the given low N m ⁢ i ⁢ n level. However, other factors, such as reduced soil loosening at greater depths in the SC system, may also have played a role in the difference in yields.

In the second year after the green manure crops were grown, no effects of the green manure crops on the N content of the soil were detected. In contrast, the influence of crop rotation was measurable throughout the entire trial year. The cereal based crop rotation either with conventional tillage or with reduced tillage had similar N m ⁢ i ⁢ n levels, while the crop rotation which had included alfalfa grass showed a significantly higher N m ⁢ i ⁢ n content on each assessment date.

The SC system had significantly more microbial N and C in the top soil (0–10 cm) than did the other two systems, while there were no significant differences in the underlying layers. This result is consistent with that from studies by Berner et al. [17], Emmerling [18], Fließbach et al. [19] and Kuntz et al. [20]. The higher content of microbial biomass in the SC system is a positive result in terms of soil fertility, as the microbial biomass is an important indicator of this [12]. Other studies reported a lower C m ⁢ i ⁢ c content with reduced tillage than with conventional ploughing, which was attributed to soil compaction and the related adverse conditions for microbial activity [21, 22]. It is therefore essential to avoid soil compaction at all costs or to loosen the soil before switching to reduced tillage, since the soil can be less easily loosened with reduced tillage than with deep tillage [2]. Farmers who successfully employ reduced tillage methods are very much aware of this point [23].

The yield of oats in the PLALF system was statistically the same for all green manure treatments. In the PL and SC systems, V. sativa led to an oat yield comparable to that of the PLALF system. In the SC system, the difference between the V. sativa treatment and the other green manure treatments was more pronounced than that in the PL system. The yield of oats in the PLALF system was statistically the same for all green manure treatments. In the reduced tillage system, it is therefore particularly important to cultivate a green manure crop that grows well within the system, whereas the choice of green manure crop is less important for ploughing systems. When ploughing in a system with a varied crop rotation, including a rotation with a perennial forage, the choice of green manure crop is almost irrelevant. No differences among tillage or crop rotation systems were discernible for the yield of field beans in 2013. The leguminous main crop was thus able to compensate for the existing differences in the N m ⁢ i ⁢ n baseline levels.

The late sowing date (April 22, 2013) could be an explanation for the relatively low field bean yield of less than 4 t ha - 1 . However, late sowing does not necessarily lead to yield losses, as regional variety trials between 1988 and 2012 have shown [24]. It is possible that spring ploughing would have had a positive effect on the field bean yield rather than tilling with the spring tine cultivator. In a study by Gruber und Claupein [25], the use of cultivators led to large losses in the grain yield of the main crop of field beans. The weather may have had an adverse effect on the grain yield of the field bean. An adequate water supply during the flowering period and the development of the pods is of particular importance for field beans [26]. Drought stress at this time can increase the shedding of flowers and young pods. In this respect, the levels of precipitation in the months of June and July 2013, which were well below average, probably had a negative impact on the field beans.