**4. Discussion**

#### *4.1. Cover Crop Biomass Production*

On average across all sites, despite the difference in cover crop height prior to rolling (i.e., rye taller than triticale), the cover crop biomass did not differ between the two species. However, ANOVA per site determined a significant effect of species on biomass production under certain pedoclimatic conditions, with the most productive species varying between years and locations (Figure 3). Among the six trials, cereal rye biomass before rolling varied from 2936 kg ha−<sup>1</sup> (18-Frce C) to 12,588 kg ha−<sup>1</sup> (17-Arl. A1) and triticale biomass ranged from 3977 kg ha−<sup>1</sup> (18-Frce E) to 16,994 kg ha−<sup>1</sup> (17-Arl. A1). Environment (soil and climate) was thus identified as a factor explaining part of the variability observed in the cover crop biomass. This was consistent with other findings in the scientific literature [38,45,48,49]. Smith et al. [37] also found contrasting results between years and locations in North Carolina on sandy and loamy sand soil characterized by a warm humid subtropical climate. For example, a decrease in rainfall accumulation in 2009 was correlated with lower rye biomass (4450 kg ha−1) compared to the previous year (10,854 kg ha−1).

As discussed by Smith et al. [37], improved cover crop managemen<sup>t</sup> including fertilization, planting date, seeding rate, species, and cultivar choice is fundamental to successful cover crop establishment and biomass production. This "management x environment" effect was observed at site A, with more than 10,000 kg ha−<sup>1</sup> of biomass for both species in 2017 and less than 7000 kg ha−<sup>1</sup> in 2018. The first year, the cover crop was planted after alfalfa and manure was applied before planting. The second year, planting occurred after corn harvested for silage and did not receive manure, with temperatures reaching above 0 ◦C a month later than the previous year. The shorter period of cover crop biomass production at 18-Arl. A2, combined with both lower precipitation during cover crop establishment in the fall of 2017 and lower nitrogen availability, resulted in lower biomass.

A similar impact of both environment and managemen<sup>t</sup> was observed in France, where in 2018, cover crop biomass was lower than 4000 kg ha−<sup>1</sup> for every species at every site except cereal rye on site D. The 2017–2018 growing season was characterized by below-average rainfall during cover crop establishment which affected cover crop emergence followed by above average rainfall in the winter which led to reduced tillering. With a fine loam clay soil type, the 18-Frce C was the most affected by the wet conditions. The water did not readily infiltrate through the soil, leading to cover crop stand losses (e.g., 2963 kg ha−<sup>1</sup> of cereal rye biomass). A week of frost in February after mild January temperatures which had brought the cover crops out of dormancy also negatively impacted cover crop development in France in 2018. At site D, the earlier planting date (25 August), nitrogen credit from the preceding alfalfa crop, and mild fall temperatures (above 10 ◦C until November) led to rapid cover crop development before winter. The cover crop was thus at more sensitive stage than at other locations during the period of frost in February, which affected its biomass production potential. The significant difference in cover crop biomass between rye and triticale at site D in 2018 (6668 and 4314 kg ha−1, respectively) was likely explained by the superior winter hardiness of rye compared to triticale (Figure 3).

Cover crop planting and termination timing have often been observed to play a key role in cover crop biomass production, explaining part of the variability between sites [50,51]. Mirsky et al. [33] discussed the increase of cover crop biomass production in May in mid-Atlantic region of US following earlier cover crop planting by comparing six planting dates across 10 day intervals under high annual precipitation condition evenly distributed (760–1012 mm) and silt loam soil. Delayed cover crop termination is critical to both improve cover crop termination and increase biomass production. Depending on specific annual conditions, cereal biomass can increase by 200 kg ha−<sup>1</sup> per day after the stem elongation stage (i.e., after the 39 Zadok stage) [17]. In our study, planting dates varied from mid-August to early October and termination dates from mid-May to mid-June (Table 2).

One strategy to increase the resilience of the CCBRT may include the use of cover crop species mixtures. As suggested by Liebert et al. [39] in New York, mixing tall species such as cereal rye with species that are shorter with wider leaves such as triticale or barley can optimize early soil shading and hasten canopy closure. This strategy could improve cover crop establishment and early spring weed control as well as increase the probability of achieving adequate cover crop biomass at rolling under challenging conditions (e.g., soil type heterogeneity, drought, excess of water, etc.). The main drawback of using species mixtures is the lack of synchronization of anthesis of the different cultivars, which would need to be assessed for successful implementation of a roll-crimp system. Cover crop termination of cereal using a roller-crimper has been shown to be most effective when done between anthesis and early dough stage (Zadoks growth stage 61 to 85), with termination increasingly effective as the cereal matures to the soft dough stage [16,17,40,48,51].

The different cover crop cultivars used in the study particularly between the Upper Midwest and Southern France trials have to be considered as cultivar might impact the potential of cover crop biomass production, the cover crop sensitivity to cold temperatures and change climate as well as the cover crop flowering period [40,52]. In North Carolina, Wells et al. [40] observed higher cereal rye biomass (>9000 kg ha−1) and greater cover crop control (100%) using earlier-flowering cultivar compared with late-flowering cultivar where cover crop biomass was inferior to 9000 kg ha−<sup>1</sup> and

the cover crop control e ffectiveness was inferior to 65%. Thus, despite the cover crop biomass e ffect, ability to provide an adequate cover crop termination also might influence the weed pressure as well as soybean yield. To address the cultivar e ffect, interest in breed early-flowering fall rye is growing in North America to a specific adaptation for organic CCBRT as observed in Canada with the "CETAB + H ÂTIF" cultivar [53,54].

## *4.2. Weed Biomass*

As observed by Liebert and Ryan [55] and Ryan et al. [56] in humid continental climate and silt loam soil, results showed that when adequate biomass is produced prior to termination, the cover crop can significantly limit weed development. A su fficient amount of cover crop biomass remaining on the soil surface can reduce weed development by acting as a physical barrier, competing with weeds for nutrients, light, and water, and releasing allelopathic compounds [20,57,58]. Previous research has concluded that cover crop biomass should reach from 6000 to 10,000 kg ha−<sup>1</sup> before termination to ensure adequate weed control until cash crop harvest, with more reliable control at biomass rates closer to 8000 kg ha−<sup>1</sup> [17,18,37]. In our study, high levels of cover crop biomass (>8000 kg ha−1) were reached at two sites: 17-Arl. A1 and 17-Frce B (Figure 3). At these sites, weed biomass increased by 127 to 204 kg ha−<sup>1</sup> between Date 1 and Date 2, while at the other southern French sites, the weed biomass increased by more than 1000 kg ha−<sup>1</sup> within the same timeframe. In Wisconsin, within the 18-Arl. A2 conditions, cover crop biomasses averaged 6615 and 6548 kg ha−1, which although on the lower end of the anticipated acceptable range suppressed weed establishment throughout soybean season.

While species did not significantly di ffer in their biomass produced in our multi-site comparative study, cover crop species did di ffer in their weed suppression. Indeed, results showed that compared to triticale, cereal rye more e ffectively suppressed weeds through the entire soybean growing season. These results were consistent with previous organic CCBRT studies conducted in soybean or corn production systems. These studies also found that cereal rye used as a cover crop in CCBRT systems provided better weed control than other winter cereals or mixes of winter cereals and legume cover crops (e.g., winter wheat and winter pea, winter wheat, hairy vetch) [18,59,60]. In Iowa, located within the same cold temperate climate as Wisconsin, Delate et al. [18] observed lower weed pressure (broadleaf species) on silty clay loam soil with a cover crop mixture including cereal rye and hairy vetch compared to a mix of wheat and winter pea, with weed populations at the beginning of June of 2.2 plant m-<sup>2</sup> and 6.5 plant m-2, respectively. According to numerous researchers, the greater allelopathic e ffect of cereal rye may explain the greater weed control observed [61–63]. While few published studies directly address this phenomenon, within these systems where the cover crop remains on the soil surface, the release of allelopathic compounds could be delayed providing greater season-long e ffects [42].

Several studies have compared cereal rye and triticale as cover crops in organic CCBRT soybean production system. In conventional systems in Ontario, Canada, Moore et al. [60] indicated that cereal rye provided better control of redroot pigweed (*Amaranthus retroflexus* L.) than triticale and wheat. In organic systems, Silva [45] did not find any di fference in weed biomass using cereal rye, triticale or barley as cover crop neither before cover crop termination nor 12 weeks after cover crop termination in Wisconsin in 2010 and 2011. These results contrast with our study, and the di fference could be explained by the lower variability in cover crop biomass observed by Silva [45] in 2010 and 2011. In our study, the cover crop biomass was particularly low at three out of six sites (e.g., biomass less than 5000 kg ha−1) while Silva [45] obtained more than 10,000 kg ha−<sup>1</sup> of cereal rye, triticale and barley cover crop in 2010 and 2011 (with the exception of triticale in 2011 with 6380 kg ha−1). When cover crop biomass is lower than 8000 kg ha−1, according to Teasdale and Mohler [64] and Smith et al. [37] a di fference of 1000 to 2000 kg ha−<sup>1</sup> of rye biomass between cover crops may explain the success or failure of a CCBRT system. The broad range of pedoclimatic conditions encountered in our study did not allow for the confirmation of this hypothesis, but greater weed growth was observed between summer and fall when the cover crop biomass was less than 6000 kg ha−1. At 18-Frce C, D and E sites, where the cover crop biomass was less than 5000 kg ha−1, weed biomass increased between Date 1 and Date 2 was high (918 to 3162 kg ha−1). Conversely, at the 17-Arl. A1, 18-Arl. A2 and 17-Frce B sites where cover crop biomass was greater than 6000 kg ha−<sup>1</sup> before rolling, the weed biomass remained stable or increased only slightly.

Despite of cover crop biomass and allelopathic effect, others factors related to the species characteristics also might influence weed managemen<sup>t</sup> such as potential of tiller number production ensuring soil covering, leaf area, vegetative/reproductive ratio, decay dynamic of cover crop on soil surface (i.e., C/N ratio) and root growth [65]. These remain poorly documented in the literature, but recent promising paper promote the interest in species mixtures which can provide benefits for weed managemen<sup>t</sup> and hasten canopy cover before cover crop rolling [39]. For instance, cereal rye combined with other species characterized by shorter height and wider leaves such as triticale or barley could increase light interception and shading.
