conventional variety; ¶ organic variety; † WAV 512 was renamed in 2018 to WAV 612; ‡ Anellino verde not available in 2019; Only tested in 2019.

At TH and FAK, a randomized complete block design with four replicates, and nine (2018) and 12 (2019) IFP treatments were used, respectively. In TH and FAK, no nitrogen was applied (0% N), and the IFP were only sown in the maize rows (IR).

Maize and IFP were sown on the same day (ET 3 May 2018, TH 28 May 2018 and 15 May 2019, FAK 9 May 2018 and 14 May 2019). Maize (*Zea mays* L., cv. "Figaro") was sown at 0.03 m sowing depth, with a row distance of 0.75 m and a planting density of 8 plants m<sup>−</sup><sup>2</sup> by a four-row pneumatic precision planter. Each plot consisted of four rows of maize. Four rows of IFP were sown after maize sowing by a plot seeder (type "Hege 80 PNI," Zürn Harvesting GmbH & Co. KG. Schöntal-Westernhausen, Germany, sowing width = 3 m, 4 rows with 0.75 m inter-row spacing). The sowing rate of 8 maize plants m<sup>−</sup><sup>2</sup> was chosen to enable the comparison to the maize–common bean intercropping treatments. In this system, maize is typically established with 8 plants m<sup>−</sup>2.

At the conventional managed location ET in 2018 weed control was done in treatment M, MS1, MS2, MB1 and MB2 with a mixture of 2.8 L ha−<sup>1</sup> *Pendimethalin* (455 g active ingredient <sup>L</sup>−1) and 1.0 L ha−<sup>1</sup> *Dimethenamid-P* (720 g a.i. <sup>L</sup>−1) as pre-emergence treatment. Treatment MN was treated in pre-emergence with 1.75 L ha−<sup>1</sup> *Pendimethalin* (455 g a.i. <sup>L</sup>−1). Treatment MA and MC were treated in post-emergence with 2.2 L ha−<sup>1</sup> *Pendimethalin* (455 g a.i. <sup>L</sup>−1). At TH in 2018 only treatment M, MB1 and MB2 were treated with 2.8 L ha−<sup>1</sup> *Pendimethalin* (455 g a.i <sup>L</sup>−1) and 1.0 L ha−<sup>1</sup> *Dimethenamid-P* (720 g a.i. <sup>L</sup>−1) as pre-emergence treatment and treatment MA was treated with 2.2 L ha−<sup>1</sup> *Pendimethalin* (455 g a.i. <sup>L</sup>−1) in post-emergence. In 2019, herbicide application for M, MB1, MB2, and MA were the same as in 2018, additionally MV was treated with 2.2 L ha−<sup>1</sup> *Pendimethalin* (455 g a.i <sup>L</sup>−1) in post-emergence and treatment MN with 1.75 L ha−<sup>1</sup> *Pendimethalin* (455 g a.i. <sup>L</sup>−1) in pre-emergence. No chemical weed control was done in MS1, MS2, MS3, MM1, and MM2. Under organic managemen<sup>t</sup> hoeing was carried out in 2018 three times (26 May 2018, 28 May 2018, and 5 June 2018). In 2019, the experimental site was two times cultivated by tine harrow (24 May 2019 and 27 May 2019) and three times hoed (1 June 2019, 13 June 2019, and 18 June 2019). In TH and FAK, hand weeding was done on demand. Applications of *Trichogramma brassicae* (Latreille) against *Ostrinia nubilis* (Hübner) were done according to the recommendations [36].

Plots were 12 m (ET), 11 m (TH), 10 m (FAK) long, and 3 m broad, consisting of a total of four rows maize and four rows of IFP per plot. Core plots for chopping were 10 m (ET), 9 m (TH), 8 m (FAK) long, and 1.5 m broad including two maize and IFP rows and leaving two rows of maize and IFP on the left and the right of the plot as a border. The complete experiments were enclosed to the right and the left by border plots (four maize rows, 0.75 cm row spacing) in order to protect the experiments from external influences.

Harvest was done by a plot harvester (type "BAURAL SF 2000", Zürn Harvesting GmbH & Co. KG, Schöntal-Westernhausen, Germany, cutting width=1.50 m) at dough stages of maize (ET 13 August 2018, TH 31 August 2018 and 17 September 2019, FAK 24 August 2018 and 5 September 2019).

#### *2.3. Data Collection and Statistical Analysis*

The dates of beginning and end of flowering of maize and the IFP were determined. For maize flowering was the time of anthesis (♂: from tassel visible until flowering completed; ♀: tip of ear emerging from leaf sheath until stigmata completely dry; [37]), for IFP when the first flowers were present until no more flowers were present. Prior to harvest plant height from soil surface to tassel tip of maize was measured. To determine the share of maize and IFP in the harvested biomass in ET a section harvest of 1.5 m × 0.67 m, in TH 1.5 m × 1.0 m and in FAK 1.5 m × 0.33 m was cut two weeks before harvest. Cutting height was the height at which the plot harvester later cuts the crops (0.15 m). All plant biomass in this area (maize, IFP, and weeds) was cut. The total weights of the fractions were determined, and the dry matter yield (DMY) and dry matter content (DMC) were determined gravimetrically after oven drying the material at 105 ◦C for 48 h.

Due to the high number of plots in ET in 2018 only 0% and 100% nitrogen fertilizer levels and only IR sowing were rated in combination with the IFP. Common vetch (MV) was not included due to a high weed infestation. In plots with the summer squashes I and II (MS1, MS2) no squash grew up, and were excluded from statistical analysis. In FAK treatment MA in 2018, and in 2019, the treatments MA, MC, MS1, MS2, and MM1 were not tested. The alfalfa in treatment MA was damaged due to mechanical weed management, also the yellow sweet clover (MC). Both squashes (MS1, MS2) had no biomass contribution to the yield due to the flat growth habits. The field emergence of the squashes in mixture I (MM1) was uneven with large gaps. Such strong weed infestation as seen in ET did not occur in TH and FAK. In addition, weeds that had survived the chemical and mechanical weed treatments in TH and FAK could be controlled by hand weeding due to the smaller scope of the test compared to ET. The sites in TH and FAK also did not had as much weed potential as ET.

At harvest all plants of all core plots were chopped by plot harvester (0.15 m above surface level). The chopper gravimetrically determined the fresh weight of the biomass. Two samples of 2 kg of each plot were extracted; one sample was dried for 48 h at 105 ◦C to determine the DMY and DMC, and the other sample was dried for 48 h at 60 ◦C for chemical analysis.

Chemical analyses of the plant material were done for the parameter's crude protein (CP), crude fat (CL), crude fiber (CX), and crude ash (CA). The analyses were done as described in Bassler [38]. The nitrogen-free extracts (NfE) was calculated as the difference between 100% and the sum of the percentage amount of CP, CF, CX, and CA. Biogas and methane yield were calculated after the formula of Schattauer and Weiland [39]. The calculation of the feed parameters gross energy (GE), metabolizable energy (ME) and net energy for lactation (NEL) for dairy cattle feeding were done after Steinhöfel et al. [40]. Soil mineral N (NO3-N) analyses "after harvest" (direct after chopping) and

at "end of vegetation period" (= end of growing season, when the weather causes a plant growth stop) were done separately for the three soil layers, 0–30 cm, 30–60 cm, and 60–90 cm. After drying (105 ◦C), milling (<2 mm), and homogenization, 100 mL of CaCl2 were added to 25.0 g of soil material. After overhead shaking (30 min, 30 rpm), the suspension was filtered and the NO3-N was determined by continuous-flow analysis [41,42].

Statistical analyses were done with the free software R (version 3.6.2) as mixed-model. After finding significant differences via F-Test, differences between treatments were compared at α = 5% using Tukey's Honestly Significant Difference (HSD) test. For creating the letter, the display package "multcomp" was used [43].

The experimental design was a randomized complete block design with four replicates for TH and FAK, where only IFP treatments were tested. For analyzing the growth, yield and quality parameters the used, fitted model was

$$y\_{i\rangle} = \mu + r\_i + \varrho\_{\rangle} + e\_{i\rangle} \tag{1}$$

where *yij* is the response, μ the general effect, *ri* the fixed effect of the *i*-th replicate, *gj* the fixed effect of the *j*-th IFP, and *eij* the residual error of *yij*.

In ET, the experimental design was a three-times replicated split-split-plot. The main plots were the level of nitrogen fertilization (0%, 50%, and 100%). The subplots 1 were the position of the IFP (BR vs. IR). Subplots 2 were the different IFP treatments. For analyzing the growth, yield, and quality the used, fitted model was

$$y\_{ijkl} = \mu + r\_i + d\_j + o\_k + g\_l + (rd)\_{ij} + (rdo)\_{ijk} + (d\emptyset)\_{jl} + (\log)\_{kl} + (d\alpha\emptyset)\_{ikl} + e\_{ijkl} \tag{2}$$

where *yijkl* is the response, μ the general effect, *ri* the fixed effect of the *i*-th replicate, *dj* the fixed effect of the *j*-th nitrogen level, *ok* the fixed effect of the *k*-th seed placement, *gl* the fixed effect of the *l*-th IFP, *rdij* the random interaction between the *i*-th replicate and the *j*-th nitrogen level, *rdoijk* the random interaction between the *i*-th replicate, the *j*-th nitrogen level and the *k*-th seed placement, *dgjl* the random interaction between the *j*-th nitrogen level and the *l*-th IFP, *ogkl* the random interaction between the *k*-th seed placement and the *l*-th IFP, *dogikl* the random interaction between the *i*-th nitrogen level, the *k*-th seed placement, and the *l*-th IFP and *eijkl* is the residual error of *yijkl*.

Normal distribution and homogeneity of variance were checked graphically.

Due to the weather extreme the statistical analyses were done for each year separate; the significant effect of the year masked the variance within the IFP. The analyses were done separately for each experimental location due to the differences in managemen<sup>t</sup> and the differences in experimental set-up (ET: conventional, three-factorial, TH: conventional, one-factorial and FAK: organic, one-factorial). Pearson's coefficient of correlation for ET in 2018 were calculated using the R package "corrplot" [44].
