2.1. Site characteristics

The experiment was conducted over three growing seasons (2018–2020) at Agriculture and Agri-Food Canada's experimental farm in Frelighsburg, QC, Canada (lat. 45°03′16.9″N, long. 72°51′39.2″W). This area has a continental climate characterized by cold winters and humid and warm summers (Fig. S1). The soil at the study site is part of the Blandford loam series within the Brunisolic order (41% sand, 38% loam, and 22% clay), which is typically well drained and associated with forested lands (Cann et al. 1948; Soil Classification Working Group 2002; Institut de recherche et de développement en agroenvironnement 2008).

The selected field had a slope of 2.75%, was not irrigated, and was restored from fallow. The soil was harrowed and cleared of stones before vine plantation. The study site was fertilized based on provincial recommendations (Barriault 2012): potassium sulphate (0–0–50) at 456 kg·ha⁻¹, triple superphosphate (0–46–0) at 315 kg·ha⁻¹, and sulphate of potash magnesia (0–0–22 + 21S + 11Mg) at 100 kg·ha⁻¹. Fertilization was adjusted in the following years based on the vine's response (Barriault 2012). Semihardy grapevines (cv. “Vidal”) were planted on 17 May 2018 at 0.9 m spacing. Interrows were 3 m wide and oriented north×–south in the direction of the slope (Fig. S2). Vines were either cane or spur trained after planting, depending on the winter protection method used, which was specific to each treatment. The guidelines for the Gobelet system (spur-trained vines) were followed for hilled vines [cultivation, annual intercrop, and unmanaged (weedy)], whereas vines covered with a geotextile fabric during winter were double cordon trained (cane-trained vines; perennial intercrop). All vines were supported upwards by a trellis with a screw anchor end-post design. Every fall, vines were hilled or protected with a geotextile fabric in early November in preparation for vine dormancy (Table 1 and Fig. 1). Soil mounds in the vine row and the geotextile fabric were removed every spring in April. Starting in 2019, vines were pruned twice a year (spring and fall) (Table 1). Clusters were thinned in-season at bloom in 2019, leaving just the first clusters starting from the bottom to promote vegetative growth during the establishment period (Barriault 2012).

Table 1.

Main field operations in various weed management method treatments throughout the experiment.

Fig. 1.

Winter protection: hilling (top) and geotextile fabric (bottom) (Agriculture and Agri-Food Canada 2018).

2.2. Experimental design

The experimental layout was a randomized complete block design with four treatments and four blocks (Fig. S2). A single experimental unit measured 12 m × 12 m and consisted of four vine rows with 12 vines per row. Units were secluded by a 5 m buffer alley. Treatments consisted of four interrow weed control management methods paired with a winter protection method:  (1) cultivation, (2) annual intercrop, (3) perennial intercrop, and (4) unmanaged (weedy). The vines were hilled in treatments (1), (2), and (4); during hilling, the vines were buried using soil taken between the rows, killing all vegetation in the interrow. The vines were protected by a geotextile fabric in treatment (3) (Fig. 1).

A 1-m-wide strip centred under the vine row was kept weed-free throughout the experiment in all the treatments. This weeding was done using tractor-mounted finger weeders (Kult Kress, Vaihingen an der Enz, Germany) and hand weeding. Dimethenamid-P (Frontier Max^®) was applied once in-row at a rate of 691 g a.i.·ha^–1 on 21 May 2018. The product must be applied before vine bud break to avoid crop damage; however, weather conditions did not permit applications before bud break in 2019 and 2020, and thus herbicides were not used again. The cultivation treatment consisted of keeping the interrow weed-free by using a rotary cultivator three to four times during the growing season (Frontier RT2283 model, John Deere US, Moline, IL, USA). All intercrops were planted in a 12 m × 2 m strip at the centre of the interrow. The annual intercrop was a mixture of cultivated oats (Avena sativa L. cv. “Fiona”) from Pedigrain (Saint-Hyacinthe, QC, Canada; 38.4 g per 1000 seeds) and Italian ryegrass (Lolium multiflorum Lam. cv. “Aubade”) from Semican Inc. (Plessisville, QC, Canada; 4.1 g per 1000 seeds). The perennial intercrop consisted of a sward commercial mix (Common No. 1 forage seed mix) from Centre Agricole Petit Bernier Inc. (Saint-Jean-sur-Richelieu, QC, Canada) with the following grass species ratio: 20% perennial ryegrass (Lolium perenne L.), 15% Italian ryegrass (L. multiflorum L.), 30%, red fescue (Festuca rubra L.), and 35% Kentucky bluegrass (Poa pratensis L.). In 2018, both intercrop treatments underwent seed bed preparation using a rotary cultivator pass, followed by vibrating tine cultivator passes (Track Curry, MK Martin Enterprise Inc., Elmira, ON, Canada). The annual and perennial intercrop treatments were first planted on 22 May 2018 and the annual intercrop was reseeded under similar conditions in mid-to-late May in the next 2 years (Table 1).

Intercrops were seeded using a minimum-till compact drill seeder (model 3P600-1106, Great Plains Manufacturing Inc., Salina, KS, USA), with 12 units and 15 cm spacing. The annual intercrop was seeded by offsetting two subsequent seeder passes to generate alternating rows of oats and Italian ryegrass. Seeding rates were 169 kg·ha⁻¹ (440 seeds per m²) for oats and 17 kg·ha^–1 (424 seeds per m²) for Italian ryegrass. The perennial intercrop was seeded at a rate of 100 kg·ha⁻¹. All the rates were adjusted based on seed germination tests done prior to seeding. In the unmanaged (weedy) treatment, weeds were left to grow only to mowing height (14 cm) to limit the invasion of perennial woody plants and to allow tractors and workers to move between the rows. All the treatments except cultivation were mowed five to seven times a year using a lawn tractor (ZTrak^™ Z915E, John Deere US, Moline, IL, USA; Table 1), before the dispersal of seeds from weeds or intercrops. Residues were left on the ground in the interrow after mowing. Finally, pest management and disease management were identical for all units and based on standard recommendations.

2.3. Weed and intercrop sampling (2018–2020)

Weed and intercrop density, percentage cover, and shoot and root biomass data were taken using two 0.25 m², 50 cm × 50 cm, quadrats placed at approximately the same location (using flags located on the rows to mark the location) in the centre interrow of each experimental unit. Weed and intercrop density sorted by species was evaluated yearly in all the treatments in early June once the annual intercrop had emerged and before the first rotary cultivator pass in the cultivation treatment (Table 1 and   Table S1 (cjps-2021-0213_suppla.pptx)). Intercrop density was evaluated every year in the annual intercrop and once in 2018 in the perennial intercrop (the four species in the latter crop were not distinguished) due to early tillering of the perennial grasses. Weed and intercrop cover percentages were estimated visually at intercrop emergence, before each mowing (five to seven times per year), and at the end of the growing season. Weed and intercrop species were not distinguished for cover percentage evaluation. Weed shoot biomass sorted by species and the shoot biomass of intercrop plants (all species combined) were evaluated at the end of October. Specifically, both perennial white clover (Trifolium repens L.) and red clover (Trifolium pratense L.) were found in plots but were pooled for analysis, owing to the difficulty of differentiating between the two at emergence and because of the time required to separate them when harvesting biomass. Shoot biomass samples were then dried at 70 °C until the weight stabilized and were weighed. Weed shoot biomass samples where I = (1, 2, …, i) weed species present in plots also served to evaluate weed species richness (S; eq. 1), relative abundance (p_i; eq. 2), and diversity (Shannon's diversity index, H′; eq. 3) based on the corresponding equations adapted from Hayek and Buzas (2010). When a plot was barren, a value of 0.0001 g·m⁻² was recorded for the most common species [brown knapweed (Centaurea jacea L.), present in all blocks every year] to retain this plot in the analysis (Baumgartner et al. 2008).

(1)

(2)

(3)

To evaluate root biomass (intercrops and weeds not differentiated), six core samples were taken in October ( Table S1 (cjps-2021-0213_suppla.pptx)) from the middle interrow of each plot in 2019 and 2020 at two soil depths (0–15 and 15–30 cm) using a hydraulic soil corer (Giddings Machine Company, Windsor, CT, USA) with sampling tubes (diameter of 7.1 cm) mounted on a post (Multi-Pro^™, Rhino Tool Company, Kewanee, IL, USA). Core samples from 2019 were kept frozen at −20 °C until evaluation in November 2020, whereas the core samples from 2020 were kept refrigerated at 4 °C. The root biomass and ash content of the core samples were determined based on protocols adapted from Beyaert and Fox (2007) and Thivierge et al. (2016). All the cores were soaked individually in a 10% w/v (g·L⁻¹) sodium hexametaphosphate solution for 16 h to disperse soil aggregates trapped within the root system before washing.

Cores were then prewashed by rinsing them through a sieve nest with 1 mm and 250 µm openings to eliminate large rock fragments and part of the soil. Subsequent root washing was conducted using a hydropneumatic elutriation washing system (Gillison's Variety Fabrication Inc., Benzonia, MI, USA) and a 760 µm sieve, with the air and water pressure set at 10 kPa and 69 PSI, respectively. After washing, the remaining root biomass was processed through a 250 µm sieve by flotation to separate roots from most other mineral and (or) organic particles that had passed through as well. The recovered root biomass was dried at 60 °C until weight stabilization. The ash content was measured by heating ground samples (1600 MiniG, SPEX SamplePrep, Metuchen, NJ, USA) to 420 °C in a muffle furnace until weight stabilization. The root biomass was corrected afterwards by subtracting the ash biomass in proportion to the sample weight (Beyaert and Fox 2007).

2.4. Soil sampling (2019–2020)

Aggregate stability to water (mean weight diameter: 0–7 cm) and soil gravimetric water content at 15–30 and 45–60 cm depths were measured (1) in spring before intercrop seeding, (2) midseason, and (3) at the end of the season in 2019 and 2020 ( Table S1 (cjps-2021-0213_suppla.pptx)). Four samples were randomly collected in the interrow of each plot using a square shovel. These samples were bulked and sieved through a 6 mm mesh and then stored in rigid-wall plastic containers at 4 °C within hours of sampling. Subsamples were then processed using a custom-built wet-sieving apparatus based on the model used by Angers et al. (2007), which uses the oscillations applied to soil samples placed atop a sieve nest immersed in water to disrupt soil aggregates. The selected sieve sizes were 1, 0.50, and 0.25 mm to distinguish between classes of stability based on Le Bissonnais (2016). Samples were covered in deionized water only on the downward movement and the sieve nest was moved up and down 30 times per minute at an amplitude of 4.5 cm for 10 min. Soil-stable aggregates of each fraction (1) >2 mm, (2) 1–2 mm, (3) 0.50–1 mm, and (4) 0.25–0.50 mm were recovered into Erlenmeyer flasks and dried at 105 °C for 48 h in the first step (part one). Primary particles of each fraction were recovered in the same Erlenmeyer flasks and dried at 105 °C for 48 h in the second part of the analysis. A solution of 5% w/v (g·L⁻¹) sodium hexametaphosphate was used to facilitate the dispersion of stable aggregates and the samples were placed on a lab shaker for 10 min. The mean weight diameter and particle size of water-stable aggregates (fraction weight) were calculated according to Angers et al. (2007) and aggregate stability was determined using the classification developed by Le Bissonnais (2016). Soil gravimetric water content was evaluated by collecting one soil core at 15–30cm and another at 45–60 cm depth in the middle interrow of each plot and on adjacent vine rows with an Edelman auger from Hoskin Scientific (Saint-Laurent, QC, Canada). These soil samples were bagged and refrigerated at 4 °C until processing, and then weighed before and after drying at 105 °C for approximately 48 h.

2.5. Vine growth, yield, and fruit quality (2018–2020)

Measurements were taken in October from three randomly selected vines per preidentified harvest row ( Table S1 (cjps-2021-0213_suppla.pptx)). Every year, all leaves from the same three randomly selected grapevines were sampled a few days before senescence and dried at 70 °C until weight stabilization to determine vine leaf biomass.

Additional growth and yield variables to vine leaf biomass were measured from the same three randomly selected vines in 2019 and 2020: total vine leaf area and pruning weight. The leaf area of a subsample of 20 leaves per vine was evaluated using a planimeter to estimate the total vine leaf area. The fall pruning weight including vine leaves was measured upwards from the first trellis wire. The vine yield and cluster number per vine were evaluated at first commercial harvest in 2020 according to the normal production cycle in Quebec vineyards (Barriault 2012). Grape must (without skin) was then extracted from a subsample of 20 berries taken randomly from 20 clusters in the harvest row. Brix measurement was performed with an optical refractometer (Reicher 13853500 High-Precision Brix Refractometer, Buffalo, NY, USA) and pH and titratable acidity were measured using a pH meter (model AR15, Fisher Scientific, Ottawa, ON, Canada).

2.6. Statistical analysis

Data analyses were conducted using the MIXED procedure in SAS University Edition 2.8.1 9.4 M6 (SAS Institute Inc. 2021). The block and sampling year were set as random effect and repeated fixed effect, respectively, for the analysis of the following variables: weed and intercrop densities, weed abundance and diversity, shoot and root biomass, grapevine leaf area, canopy biomass, and pruning weight. Only the two intercrop treatments were taken into consideration when comparing intercrop densities, intercrop cover, and intercrop shoot biomass. When soil data were analyzed, the block was treated as a random effect and the sampling event as a repeated fixed effect. For the analysis of weed and intercrop cover, the model included the block as a random effect and the sampling month × year interaction as a repeated fixed effect. All the variables were analyzed using the variance component covariance structure according to the best model fit based on the Akaike information criterion, except vegetation, intercrop, and weed cover, for which the un@ar(1) (Direct product AR[1]) covariance structure was specified. The vine yield and cluster number per vine were analysed separately with the block as a random effect and sampling year as a repeated fixed effect. For all the analyses, homogeneity of variance was verified by plotting residuals using the SGPLOT procedure. The normal distribution of residuals was assessed graphically and by performing a Shapiro–Wilk test using the UNIVARIATE procedure. The standard error of means was also calculated using the UNIVARIATE procedure. Logarithmic (weed density, annual weed biomass, pruning weight, and particle size distribution of water-stable aggregates) or square root transformations (perennial weed biomass, intercrop, and weed cover) were applied to data that did not meet these criteria, and back-transformed data were presented. The statistical significance between multiple comparisons was determined using Tukey's honestly significant difference test based on a 95% confidence interval.

3.1. Weed and intercrop density

There was a significant effect of weed management method × year on weed density at intercrop emergence in early June (P < 0.0001; Fig. 2). Every year, weed density was not significantly different between the annual intercrop and unmanaged (weedy) treatments, while it was the lowest in the cultivation treatment in 2020. Weed density was also lower in the perennial intercrop treatment than in the unmanaged (weedy) treatment in 2020. In 2019, weed density was only higher in the perennial intercrop treatment relative to the cultivation treatment. Weed density remained at similar levels between years within the annual intercrop treatment, and reached its highest value in the perennial intercrop treatment in 2019 and in the unmanaged (weedy) treatment in 2020 (Fig. 2).

Fig. 2.

Mean weed density at intercrop emergence (early June) throughout the experiment. Bars with different lowercase letters are significantly different between weed management methods and within years. Asterisks indicate significant interyear differences within weed management method. All multiple comparisons are based on Tukey's honest significant differences (α = 0.05). Error bars indicate + standard error.

Intercrop density in the annual intercrop (Italian ryegrass and oats pooled together) treatment was lower in 2020 (354 plants per m²) than in 2018 (605 plants per m²) and 2019 (624 plants per m²) (P = 0.0024; data not shown). The perennial intercrop treatment had a density of 170 plants per m² in 2018 (data not shown).

3.2. Weed and intercrop cover

Vegetation cover (weeds and intercrops pooled together) varied as a result of a weed management method × month × year interaction (P < 0.0001; Fig. 3a). It was consistently lower in the cultivation treatment regardless of the sampling month or year. In June 2018, vegetation cover was higher in the annual intercrop treatment (greater than 40%) but was surpassed by the unmanaged (weedy) and perennial intercrop treatments (greater than 60%) by the end of the season. Vegetation cover was greater in the perennial treatment than in the annual intercrop or unmanaged (weedy) treatment in June 2019 (40% vs. 10%, and 3%) and 2020 (90% vs. 45%, and 37%; Fig. 3a).

Fig. 3.

Mean vegetation (a), weed (b), and intercrop cover (c) before mowing through the experiment. Weed management method data were grouped by month–year for analysis. 06 = June; 07 = July; 08 = August; 09 = September; 10 = October. Bars with different lowercase letters are significantly different between weed management methods within month and year. All multiple comparisons are based on Tukey's honest significant differences (α = 0.05). Error bars indicate + standard error.

Weed cover was affected by a weed management method × month × year interaction (P < 0.0001; Figs. 3b and 4a). Weed cover was generally lowest in the cultivation treatment (less than 5%) than in other treatments, except in June 2018 and 2019, when the weed cover in the annual intercrop and perennial intercrop treatments was also low. In 2018, the highest weed cover values were consistently observed in the unmanaged (weedy) treatment, whereas the annual and perennial intercrop treatments had similar values (less than 20%) throughout the season. During all years, the weed cover in the unmanaged (weedy) treatment reached values above 85% in August (Fig. 3b). In 2019, the weed cover in the annual intercrop treatment peaked at 50% in September, while the weed cover in the perennial intercrop treatment was higher (greater than 55%) starting in July (Fig. 4a). In 2020, the weed cover in both intercrop treatments was high and equivalent to that of the unmanaged (weedy) treatment in August and at the end of the season (Fig. 3b). Moreover, the weed cover increased significantly within season in the unmanaged (weedy) and annual intercrop treatments (except for October 2019; Fig. 4a), reaching significantly higher values after 2018 (  Fig. S3 (cjps-2021-0213_suppld.pptx)a). In contrast, the weed cover was significantly higher after 2018 in the perennial intercrop treatment (  Fig. S3 (cjps-2021-0213_suppld.pptx)a) and remained approximately at the same level within season except in June 2019, with values ranging from 59% to more than 85% (Fig. 4a).

Fig. 4.

Mean weed (a) and intercrop (b) cover before mowing through the experiment. Weed management method data were grouped by month–year for analysis. 06 = June; 07 = July; 08 = August; 09 = September; 10 = October. Bars with different lowercase letters are significantly different between months within year for each management method. All multiple comparisons are based on Tukey's honest significant differences (α = 0.05). Error bars indicate + standard error.

Finally, intercrop cover was also influenced by a weed management method × month × year interaction (P = 0.0002; Figs. 3c and 4b). Intercrop cover was not different between the annual and perennial Intercrop treatments except during three specific sampling months: July and August 2018 and July 2019 (P = 0.0002; Fig. 3c). Intercrop cover was constant throughout the season in the annual intercrop treatment in 2018 only (Fig. 4b). Perennial intercrop cover was also relatively stable throughout the season (except in June 2018 and October 2019; Fig. 4c) and tended to decrease in 2020 after July compared with prior years (  Fig. S3 (cjps-2021-0213_suppld.pptx)b).

3.3. Weed and intercrop shoot and root biomass

Both total vegetation and weed shoot biomass were affected by a weed management method × year interaction (Table 2). Vegetation shoot biomass was consistently lower in the cultivation treatment but the values for the annual intercrop, perennial intercrop, and unmanaged (weedy) treatments were similar, except in 2020 when the perennial intercrop shoot biomass value was 1.49 higher than the unmanaged (weedy). Similarly, total weed shoot biomass was lowest in the cultivation treatment and the two intercrop treatments had similar values in each year. Total weed shoot biomass remained lower in the annual and perennial intercrop treatments than in the unmanaged (weedy) treatment until 2020, when both intercrop treatments had high values that were similar to that of the unmanaged (weedy) treatment. Intercrop shoot biomass did not vary between years and was 1.42 times higher in the perennial treatment than in the annual intercrop treatment (Table 2). In 2019 and 2020, annual weed shoot biomass was higher in the annual intercrop treatment compared with the perennial intercrop treatment. The shoot biomass of perennial weeds in the perennial intercrop treatment was greater than that of the annual intercrop treatment in 2019 and 2020 (Table 2).

Table 2.

Vegetation, intercrop, and weed shoot biomass at the end of growing seasons.

Brown knapweed, smooth crabgrass (Digitaria ischaemum [Schreb.] Muhl.), witchgrass (Panicum capillare L.), narrow-leaved plantain (Plantago lanceolata L.), prostrate knotweed (Polygonum aviculare L.), and clovers (Trifolium spp. L.) represented the majority of the weed species present in the unmanaged (weedy) and annual and perennial intercrop treatments throughout the experiment (Table 3).

Table 3.

The richness, diversity, and relative abundance of the most frequent weed species.

Weed species richness (S) and diversity (H′) were both influenced by a weed management method × year interaction (Table 3). Weed species richness was lowest in the cultivation treatment except in 2020 when it was also low in the annual and perennial intercrop treatments. Weed species richness was consistently similar between the annual intercrop and unmanaged (weedy) treatments, the latter showing the highest values. Weed species diversity (Shannon's diversity H′) was consistently lower in the cultivation treatment than in the other treatments except in 2020 when it was equivalent to that of the annual intercrop treatment. The highest values were recorded in the unmanaged (weedy) treatment (Table 3).

Root biomass varied between weed management methods only at depths of 0–15cm (P < 0.0001) and 15–30 cm (P = 0.0005; Fig. 5). At a depth of 0–15cm, root biomass was the highest in the perennial intercrop treatment and the lowest in the cultivation treatment, with intermediate values in the unmanaged (weedy) and annual intercrop treatments. At a depth of 15–30 cm, root biomass was significantly higher in the perennial intercrop treatment than in the cultivation treatment (Fig. 5).

Fig. 5.

Mean root biomass (ash-free) at 0–15cm (a) and 15–30 cm (b) soil depth in weed management method before hilling. Root biomass data from 2019 and 2020 were pooled by weed management method for presentation. Bars topped with different letters are significantly different between weed management methods within soil depth. All multiple comparisons are based on Tukey's honest significant differences (α = 0.05). Error bars indicate + standard error.

3.4. Soil properties

There was a significant effect of a weed management method × sampling event interaction on the average size (mean weight diameter) of water-stable aggregates and on the proportion of medium and stable (1.3–2 mm) aggregates and macroaggregates (>0.25 mm; fraction weight) (Table 4). Over the 2 year study, water-stable aggregation remained relatively constant over time within each of the living-mulch treatments, although differences between weed management methods occurred in October of both years. In October 2019, the lowest values for mean weight diameter and proportions of macroaggregates, medium aggregates, and stable aggregates were found in the cultivation treatment. In October 2020, the mean weight diameter as well as the proportions of macroaggregates and medium and stable aggregates were higher in the perennial intercrop treatment than in the cultivation treatment (Table 4).

Table 4.

Mean water-stable aggregate characteristics (2019–2020).

Soil gravimetric water content did not vary among the weed management methods (P > 0.05), regardless of soil depth or field sampling spot (in-row or interrow). It varied only between sampling events (P < 0.01) and was generally the lowest in August (  Fig. S4 (cjps-2021-0213_supple.pptx)).

3.5. Vine growth, yield, and fruit quality

Vine leaf biomass just like vine leaf area was affected by single weed management (P < 0.0001) and year (P < 0.0001) effects (Table 5). Vine leaf biomass and leaf area were both lower in the intercrop treatments than in the cultivation treatment, and higher in 2019 than in 2020, regardless of treatment. There was a single-year effect on pruning weight, which significantly decreased year after year in all the treatments (P < 0.0001; Table 5). Vine yield (P = 0 2563), cluster number (P = 0.3363), Brix (P = 0.3850), pH (P = 0.7592), and titratable acidity (P = 0.3250) were not affected by the weed management method in 2020 (Table 5).

Table 5.

Mean vine yield, cluster number, fruit quality, and vine growth at the end of the season.

4.1. Weed control and weed community composition

As predicted, intercrops did not outperform cultivation as a weed management method but still provided significant weed control as measured by weed biomass in at least 2 years out of 3 relative to unmanaged interrows, thus confirming our first hypothesis. Results show that both annual and perennial intercrops did not prevent weed establishment during the study, but prevented weed growth during the first 2 years. The weed density generally increased, except in the cultivated treatment, where most weeds did not grow to maturity, whereas the seedbank was replenished in the other treatments despite regular mowing. Weed emergence may have been curbed only by the dense cover of perennial weeds in 2020 as perennial plant species can effectively limit the weed density in pastures (Wardle et al. 1992; Demjanova et al. 2009; Meiss et al. 2010; Luna et al. 2020).

Although deep tillage such as mouldboard ploughing can normally bury weed seeds deeper in the soil and thus reduce weed germination and emergence during the subsequent growing season (Rahman et al. 2000; Cordeau et al. 2020), ploughing to a depth of 30 cm (hilling and dehilling every year) did not reduce the weed density in the hilled treatments where weeds were only mowed. This is probably because the first 30 cm top soil layer could still hold a lot of weed seeds despite repeated ploughing as reported by Vasileiadis et al. (2007) and that seeds do not become deeply buried during the growing season. Weed seeds were buried in the soil mound on the vine row during winter, only to be redistributed in the interrows and rows in the spring.

The perennial intercrop treatment did prevent weed establishment during the third year, as mentioned above, but allowed for significantly higher weed cover in June (mostly brown knapweed, clovers, and prostrate knotweed). This is most likely attributable to management techniques, such as seeding date, mowing height, and fertilization, which failed to promote the competitiveness of the species in the perennial grass mixture. Mowing alone will not prevent weed establishment, even if the cut is done higher (14 cm vs. <5 cm; Busey 2003). Typical grass management techniques, such as reapplication of fertilizers, interseeding, or chemical weed control are often required to replenish grass cover, thus limiting weed establishment (Christians et al. 2017). This intensive management is however contrary to the idea that intercrops have to be low maintenance to remain cost effective for winegrowers. Finally, seeding the perennial intercrop prior to the experiment in August or September 2017 would probably have allowed a more rapid establishment in 2018 and increased intercrop competitivity (Reicher et al. 2000).

Our results also demonstrated that living-mulch management affects weed community assembly within a short period of time (first 3 years) (Ryan et al. 2010; Sanguankeo and León 2011; Steenwerth et al. 2016). Smith (2006) and Cordeau et al. (2017), among others, showed that soil tillage is an important driver of weed community assembly. Lower plant species richness and diversity are generally associated with soil tillage used for weed control, as observed in our cultivated treatment (Bàrberi and Mazzoncini 2001; Hyvönen and Salonen 2002; Ngouajio and McGiffen 2002; Mas and Verdú 2003; Moonen and Bàrberi 2004; Baumgartner et al. 2008). Furthermore, annual weeds are usually associated with tilled annual intercrops (Tuesca et al. 2001; Chauhan et al. 2006) and they became dominant in our annual intercrop in 2020. Drier weather conditions during the spring of 2020 may also have promoted the proliferation of smooth crabgrass, which is known for its tolerance to dryness owing to its C₄-type photosynthesis (Dore and McNeill 1980; Turner et al. 2012). Smooth crabgrass proliferation increased weed cover in August and decreased weed species diversity in 2020.

The decrease in weed species richness and diversity in the perennial intercrop relative to the control and the annual intercrop is in accordance with the conclusions of Jiang et al. (2008), who reported lower plant species richness and density in no-till vineyards compared with hilled vineyards. This lower diversity coincides with the predominance of perennial weeds such as brown knapweed and clovers in the perennial intercrop. Perennial weeds are typically more persistent under perennial intercrop management because they benefit from the absence of soil disturbance, which leaves their root systems and underground plant propagation asexual organs intact. Mowing is thus mostly ineffective for weed species that grow near the ground, which explains how brown knapweed, prostrate knotweed, narrow-leaved plantain, and clovers came to be the most common weed species in these vegetated treatments (Gago et al. 2007; Jiang et al. 2008). Although the cultivation treatment was hilled, the frequent rotary cultivator passes kept weed infestation levels low and thus weed species richness and diversity always remained inferior compared with the perennial intercrop treatment.

4.2. Soil conservation

As expected, maintaining living mulches in the interrow preserved the short-term physical integrity of soil aggregates to a greater extent than did mechanical weeding throughout the season, thus confirming our second hypothesis. This difference at the end of the season seems to be the net result of the positive influence of groundcover root systems and the negative effect of repeated cultivation on aggregate stability. Although this improvement is transient because groundcover dies off and aggregate stability is susceptible to soil wetting–drying cycles occurring between seasons, the groundcover is effective during the fall when erosion from rainfall is important (Le Bissonnais 2016 mulch shoot biomass and mowing residues were incorporated into the soil only once a year, during fall hilling, in the annual intercrop and unmanaged treatments, compared with the three to four passes of a rotatory cultivator made in the cultivation treatment each year. The fact that the root biomass and mean weight diameter were constantly higher in the perennial intercrop treatment in October suggests that root-derived carbon from rhizodeposition had more of an effect than the decomposition of fresh material from shoot biomass by soil fauna and microorganisms in terms of mitigating aggregate breakdown (Bronick and Lal 2005). The higher proportions of stable aggregates and macroaggregates relative to sample weights in October also support this idea. We observed no short-term benefit from the use of the perennial intercrop compared with our annual intercrop on soil water-stable aggregation. This is likely because the selected annual and perennial intercrop species had similar functional traits, such as root architecture, whereas higher plant species diversity can increase the benefits of intercropping on soil properties, including aggregation, through the positive influence of complementary root traits (Gould et al. 2016; Saleem et al. 2020).

Soil tillage was detrimental to soil structure, as observed by Jiang (2010) in vineyards. Tillage for mechanical weed control is associated with the loss of particulate organic matter (POM), a primary source of organic carbon metabolized into binding agents contributing to aggregate cohesion (Six et al. 1998, 1999). The idea that tillage limits the benefits of intercropping has been documented in vineyards growing in a Mediterranean climate. For example, a study conducted by Belmonte et al. (2018) in the Napa Valley of California showed that tillage limited the long-term capacity of an annual intercrop to accumulate soil organic carbon, thus preventing aggregate stability improvement in comparison to a no-till system (lower POM). Garcia et al. (2019) showed how aggregate stability rapidly degraded with the introduction of tillage even though intercropping had been done for years.

4.3. Vine growth, yield, and fruit quality

Vine leaf biomass was higher in the cultivation treatment than in the intercrops but berry yield and quality were equivalent. This indicates that intercrops competed with grapevines for the available resources at the beginning of the season (Jackson 2008), but not enough to cause yield loss or differences in fruit quality. Similar results were observed in semipermanent and permanent intercrops established in North American vineyards (Ingels et al. 2005; Smith et al. 2008; Steinmaus et al. 2008; Sweet and Schreiner 2010).

However, we acknowledge that the yield and fruit quality comparisons between the perennial intercrop and other treatments are imperfect because the vines did not have the same training and winter protection. This may have caused variations in bud fruitfulness and in the microclimate within the vine canopy during grape maturation (Greven et al. 2014; Provost and Barriault 2019).

Yield and soil gravimetric water content results indicated that there was limited competition for water between intercrops and vines. Although no vine water status analyses were conducted, monthly cumulative precipitation (yearly totals averaged 113.3 cm) likely provided a sufficient supply of water. Hartwing and Ammon (2002) found that competition for water in vineyards would be low if the average annual rainfall exceeded 110 cm. Vine response to intercrops is tempered in a cooler and humid climate compared with areas where water is scarcer (Tesic et al. 2007). In addition, intercrop management by mowing or destroying annual intercrops before vine budbreak can help mitigate competition between intercrops and vines (Garcia et al. 2018). Studies in North American vineyards also demonstrated that weeding on the row limits competition by intercrops by preventing the establishment of interrow crops on the trellis (Baumgartner et al. 2007; Olmstead et al. 2012). Whether this conclusion also applies to weeds left to grow in the interrow in the long term remains to be tested.