In 2017 and 2018 a field experiment comparing different between-bed management strategies in organic plasticulture vegetable production was conducted at the Michigan State University (MSU) Southwest Michigan Research and Education Center (SWMREC) in Benton Harbor, Michigan (42.09°N, 86.36°W). The experimental treatments were repeated in the same field and plot boundaries across years of the study to investigate the combined in-season and cumulative impacts of between-bed management practices on soil quality parameters and weed management. The area used for the experimental site was a transitioning organic field that had been maintained as a rye-hairy vetch cover crop for five years prior to trial establishment. The soil at the experimental site was a Spinks (Sandy, mixed, mesic Lamellic Hapludalfs) and Selfridge (Loamy, mixed, active, mesic Aquic Arenic Hapludalfs) loamy sand (Soil Survey Staff, USDA Natural Resources Conservation Service, 2024). Initial soil chemical characteristics were as follows: organic matter 1.6% (loss on ignition); pH 5.9 (1,1 soil/water); CEC 4.0 cmol kg⁻¹; and P (Bray-P1 extractant), K, Mg, and Ca (1 N ammonium acetate extractant) levels of 48, 133, 54, and 397 mg kg⁻¹, respectively.

The experiment was arranged in a split-plot randomized complete block design with four replications. Between-bed management was the main plot factor and cash crop was the split-plot factor. Between-bed management strategies evaluated included cultivated bare ground, cereal rye residue dead mulch, mowed ambient weeds, and three cover crop living mulch treatments: rye (Secale cereale L.) monoculture, Italian ryegrass (Lolium multiflorum) monoculture, and a Dutch white clover (Trifolium repens)/rye mixture. Treatment names and description are listed in Table 2. Bell pepper (Capsicum annuum cv. Paladin) and summer squash (Cucurbita pepo cv. Lioness) were chosen as cash crops to represent a relatively long- and short-season crop commonly grown in plasticulture systems. While main plot (between-bed management) treatments were maintained in the same locations in both years of the study, the location of split-plots (cash crop factor) within main plots were rotated between years. Main plots measured 6.1 m wide × 5.0 m row length, contained four raised plastic mulch beds (each 0.61 m wide and 0.15 m tall), and three between-row soil areas that measured approximately 1 m wide.

Dates of key field operations are outlined in Table 3. Winter cover crops—a rye-vetch mixture in the first year and rye monoculture in the second—were mowed and incorporated with a rototiller in early May of both years. In mid-May, the field received 112 kg total N ha⁻¹ as organic fertilizer (Nature Safe “All-Season 10-2-8,” Irving, TX) derived from feather meal, meat and bone meal, blood meal, and sulfate of potash. Plastic mulch beds (0.15 m bed height and 0.61 m bed top width) were laid on 1.67 m centers within 1 week of fertilizer application. Plastic mulch was black, 1 mil thick and 1.22 m wide (Trickl-Eez Irrigation Inc., St. Joseph, MI). Single drip irrigation lines with 0.30 m emitter spacing were laid under the plastic mulch using a combined plastic mulch layer and bed shaper (Reddick Equipment Company LLC, Williamston, NC).

Bell peppers were sown in 98 cell flats in a heated greenhouse at the end of March and transplanted into the field at the end of May. Summer squash were direct seeded in the field at pepper transplanting. Peppers were grown staked in staggered double rows with 0.30 m between-row spacing and 0.46 m in-row spacing. Squash were grown in single rows with 0.61 m in-row spacing. All beds were drip irrigated simultaneously twice daily with a single drip line offset from the row (squash) or between rows (peppers), based on crop demand and standard grower practice in the area. No fertilizer was applied through the drip.

Between-bed management strategies were established simultaneously with cash crop planting in both years. In 2017, plastic mulch beds were laid 1 week before implementing between-bed management treatments and a stale seedbed cultivation using a wheel-hoe was completed in all between-bed areas immediately before living mulch seeding to reduce potential weed competition. However, wetter spring conditions in 2018 did not allow for the same stale seedbed window between bed formation and treatment establishment.

Living mulch seed was broadcast sown by hand into between-bed areas and lightly incorporated using a rake. Italian ryegrass was seeded at 35.9 kg ha⁻¹. Dutch white clover was seeded at 22.4 kg ha⁻¹. In 2017, cereal rye was seeded at 168.1 kg ha⁻¹, but was doubled to 336.3 kg ha⁻¹ in 2018 in an attempt to address poor establishment in 2017. In both years, Dutch white clover was seeded with rye at half the monoculture rate (84.0 and 168.1 kg ha⁻¹ in 2017 and 2018, respectively). Overhead irrigation was applied to benefit living mulch establishment in 2017, but not in 2018 due to higher precipitation around the time of seeding. Between-bed living mulches and weeds were mowed using a walk behind push mower with the mower deck set to 10.16 cm when the average height of weeds and living mulches were approximately 30 cm. Cultivated plots were maintained through hand cultivation using a wheel-hoe with a 20 cm blade (Glaser wheel hoe, Johnny’s selected seeds, Fairfield, ME). Dead mulch consisted primarily of cereal rye obtained from an adjacent field that had been established the previous fall and flail mowed two days before mulching. Dead mulch was applied at an approximate rate of 18 Mg ha⁻¹.

At the end of September, plastic mulch was removed, and the entire experimental area rototilled to a depth of 20 cm. A rye cover crop was then planted at 180 kg ha⁻¹.

All data were analyzed separately by year using a mixed model ANOVA with the MIXED procedure in SAS (Version 9.4, SAS Institute, Cary, NC). When necessary, data were square root or log transformed to meet assumptions of normality. Untransformed data are presented. Heterogeneous variance was corrected by using unequal variance models or with the appropriate variance–covariance structure in repeated measures analyses. Mean separations were conducted using Fisher’s protected least significant difference test at p < 0.05.

For data collected at a single time point, treatment was included as a fixed factor and block (replication) as a random factor. Since weed biomass in the cultivated treatment was negligible in 2017 and only collected in two of four treatment replicates in 2018, this treatment was excluded from total weed biomass statistical analysis. Weed seedbank data by species was summed across all sampling dates to get cumulative emergence as an estimate of the germinable seedbank by species. At the species level, high variability in estimates of germinable seedbank limited the statistical power to detect differences, so data was aggregated across all weed species.

A repeated measures mixed model ANOVA was used to analyze data collected at multiple time points or to multiple depths (i.e., soil cores). Treatment and sampling date or depth were included as fixed factors and block (replication) as a random factor. For inorganic N and soil moisture data, a first-order autoregressive variance–covariance structure was used given the roughly equally spaced timings of measurements. Soil microbial biomass, total extracellular enzyme activity, and the Shannon diversity index were determined in each treatment separately, but similarities in plant composition and the lack of significant differences in microbial response between weedy and living mulch treatments encouraged combining these treatments into a single category referred to as vegetative cover. Treatment group and sampling date were treated as fixed factors and block (replication) as a random factor. Variance–covariance structures were selected based on fit statistic results. For soil profile inorganic N, treatment and depth were treated as fixed factors and block (replication) as a random factor. An equal variance first-order autoregressive variance–covariance structure was used.

The relationships between marketable cash crop yield, cash crop tissue N, total between bed living mulch and weed biomass, soil inorganic N averaged over the growing season, and soil moisture averaged over the growing season were evaluated via Spearman’s rank correlation coefficient (rho) using the “cor.test” function in base R (Version 2023.03.1 + 446, R Core Team, 2023, Vienna, Austria).

Average temperatures were similar in both years of the study and comparable to the 30-yr average for the location, except for higher temperatures in May 2018 (Table 4). Compared to seasonal precipitation averages, 2017 was a relatively dry season and 2018 a relatively wet season. In 2017, overhead irrigation was used for two weeks following living mulch sowing to encourage establishment. This was not necessary in 2018; heavy spring rain caused flooding in some areas of the field in that year and may have hindered living mulch establishment.

Dominant summer annual weed species in the field included crabgrass species (Digitaria spp.), common lambsquarters (Chenopodium album L.), ladysthumb (Persicaria maculosa), oak-leaf goosefoot (Chenopodium glaucum), purslane (Portulaca oleracea L.), ragweed species (Ambrosia spp.), and wood sorrel species (Oxalis spp.). Perennial weeds including quackgrass (Elymus repens) and dock species (Rumex spp.) were less common than summer annual weeds, but present in localized areas of the field.

Cultivation and dead mulch reduced in-season weed biomass by an average of 94 and 78% respectively, more than any of the living mulch treatments (Figure 1). The cultivation schedule in 2017 resulted in these plots remaining weed-free during the season. However, in 2018, wet field conditions and delayed cultivation led to some weed growth in the cultivated treatment at the end of July. Although we are unable to statistically compare weed biomass in the cultivated treatment to the others, cumulative weed biomass production in the cultivated treatment was substantially less than weedy and living mulch treatments (Figure 1). In 2017, Italian ryegrass established quickly, producing 191 g m⁻² biomass and reducing weed biomass by >50% compared to the weedy treatment (Table 5). However, heavy spring rain in 2018 contributed to poor Italian ryegrass establishment and modest biomass accumulation in that year (50.7 g m⁻²). In comparison, rye establishment was poor across years, and the development of leaf rust by mid-July nearly eliminated stands. The clover/rye treatment was slow to establish in 2017 and failed to establish in some plots in 2018. Neither rye nor clover/rye significantly reduced in-season weed biomass compared to the weedy treatment in either year (Table 5).

Dominant weed species found in the germinable seedbank were similar to those in the field, with the exception of perennials, which were excluded from seedbank evaluation. Overall, the most abundant species were crabgrass species, ladysthumb, and oak-leaf goosefoot.

After two years, reductions in the germinable seedbank relative to weedy treatment were detected only for the cultivation, dead mulch and Italian ryegrass treatments. The cultivated and dead mulch treatment had the smallest germinable weed seedbank, and the weedy treatment the largest after both seasons (Table 5). The 2017 pattern of in-season weed suppression was closely reflected in estimates of the germinable weed seedbank from between-bed areas in spring 2018. Compared to weedy, the reduction in the germinable seedbank increased from 18 to 72% as weed biomass suppression increased from 9 to 100% (Table 5). Among cover crops grown in 2017, the Italian ryegrass treatment had the greatest impact on weeds, reducing both weed biomass and subsequent weed seedbank density by more than half of that observed in the weedy treatment. Treatments applied in 2018 generally had less impact on the seedbank than those from 2017, and seedbank effects were not as closely associated with weed biomass suppression. Despite minimal in-season weed suppression by Italian ryegrass in 2018 (12%), the weed seedbank in that treatment was reduced by 36% relative to weedy and was comparable to that in both the dead mulch and cultivated treatments at experiment termination (Table 5). Although in-season weed control by Italian ryegrass living mulch was variable and small relative to cultivation and dead mulch, it still reduced the germinable weed seedbank across both years of our study compared to the weedy treatment. In contrast, rye and the clover/rye mixture living mulches did not reduce the weed seedbank compared to weedy in either year.

Squash performance metrics were unaffected by between-bed treatment in both years, and in 2018 pepper performed similarly across all treatments (Table 6). However, in 2017 living and dead mulch treatments resulted in pepper yield losses of between 41 and 54%, and we detected a negative correlation between total living mulch and weed biomass between plastic mulch beds and marketable pepper yield (Table 7). Total N concentration in pepper leaves sampled at the last harvest in 2017 was greater in the cultivated treatment relative to all other treatments (Table 6). Soil inorganic N levels in the crop row ranged from 28 to 57 mg N kg soil⁻¹ in the weedy and rye treatments respectively, and no significant differences were resolved among treatments (Figure 2B). Correlation analysis revealed no relationship between marketable yield and in-row soil inorganic N, but a positive correlation between marketable pepper yield and in-row soil moisture averaged across sampling dates in 2017 (Table 7). Soil moisture under plastic mulch was numerically higher in the cultivated control compared to all other treatments during pepper fruit development in 2017 (Figure 2A). In 2018, pepper marketable yield was comparable across treatments, and no significant differences in in-bed soil moisture (Figure 2C) or soil inorganic N levels (Figure 2D) were detected in pepper beds.

Living mulch establishment was a challenge and weed biomass on average accounted for 99, 74, and 94% of organic matter inputs in rye, ryegrass, and clover/rye treatments, respectively. Italian ryegrass produced significantly more biomass than rye or clover/rye in both years (Table 5). Total shoot biomass inputs in the weedy and living mulch treatments averaged 490 g m⁻² in 2017 and 647 g m⁻² in 2018 (Figure 1). In comparison, dead mulch was applied at an average rate of 18 Mg ha⁻¹ and had an additional input of 130 g m⁻² of weed shoot biomass in 2017 and 122 g m⁻² in 2018.

At trial initiation, we detected significantly greater SOM levels in weedy treatment plots relative to all others, a random spatial effect not attributable to treatments. SOM in the between-bed areas of the weedy treatment was again significantly greater than all other treatments at trial termination. However changes in SOM in between-bed areas over the course of the experiment were comparable across treatments, decreasing by an average of 0.2% (Table 8).

Given similarities in plant community composition between living mulch treatments and the weedy treatment (i.e., dominated by the ambient weed community), these treatments were collapsed into a single treatment category referred to as “vegetative cover” from here forward.

No significant differences were observed in microbial biomass C between cultivated, dead mulch, or vegetative cover plots in either year (Figures 3A,B).

By the end the of 2017 season, total extracellular enzyme activity was similar across treatments (Figure 3C). This was true at the beginning of the 2018 season too, but the treatments began separating in June (Figure 3D). As the season progressed, the dead mulch treatment consistently had the greatest total enzyme activity, with vegetative cover as an intermediate, and cultivation having the lowest (Figure 3D).

Significant differences in the Shannon Diversity Index were observed between treatments at the end of the 2017 season (Figure 3E). However, this pattern did not carry over into 2018. No significant differences between treatments were detected for the entirety of the 2018 season. However, in 2018 the diversity of enzyme activities was on average greater in dead mulch and the vegetative cover treatments than the cultivated treatment (Figure 3F).

Where plants were actively growing between plastic mulch beds during the 2017 season, potentially leachable N in the soil profile (0–60 cm) was reduced by approximately 61% (Figure 4A). Differences in soil profile N were most pronounced in the top 20 cm and decreased with increasing soil depth. In 2018 no significant differences in potentially leachable nitrogen were observed (Figure 4B).

Our results confirm the importance of living mulch establishment and biomass accumulation in controlling weeds (e.g., Bybee-Finley et al., 2017; Alonso-Ayuso et al., 2018), and highlight the inherent risks of managing weeds with cover crops (Teasdale, 1996; Sarrantonio and Gallandt, 2003; Fernando and Shrestha, 2023). The relatively slow growth rate of clover and stress experienced by rye planted outside of its climatic niche made these species unable to compete with aggressive weeds, results confirmed elsewhere (Reid and Klotzbach, 2013). Italian ryegrass was better suited for this niche, but performance was still variable between years. In previous work we found that teff (Eragrostis tef) was the only living mulch treatment among nine to significantly reduce in-season weed biomass between plastic mulched beds (Tarrant et al., 2020). In both cases, weed suppression by Italian ryegrass and teff was variable across years, but these results suggest that vigorous summer annual grasses may be better suited for this niche than slower growing clovers and winter grains.

The close relationship between weed biomass suppression and weed seedbank suppression observed in the first year of this study suggests that competition effects on seed production played an important role in explaining seedbank effects. Reductions in weed seed production and seed viability in the presence of competition have been previously demonstrated in agronomic cropping systems (Nurse and DiTommaso, 2005) and in response to living mulches within vegetable cropping systems (Brainard and Bellinder, 2004; Brainard et al., 2005). However, treatment differences in the germinable seedbank may also have been influenced by indirect effects on the rates of predation, decay, or fatal germination of weed seeds. Seed burial and predation studies demonstrate that tillage and cover crops can accelerate the rate of seed loss through impacts on soil micro-organisms, seed predators, or fatal germination (Gallandt et al., 2004; Mohler et al., 2018; Frost et al., 2019; Fernando and Shrestha, 2023).

Not surprisingly, mowing ambient weeds in our study led to increases in the weed seedbank compared to cultivation, dead mulching, and Italian ryegrass living mulch. While the weed seed bank can serve as a “free” living mulch in subsequent seasons when managed this way (Hartwig and Ammon, 2002), unchecked weeds may be more competitive and harder to control than a cover crop living mulch (Westbrook et al., 2021). Mowing may also lead to a shift in the ambient weed community composition toward difficult to manage and competitive species (Butler et al., 2013). Because of the impact on future management, these potential drawbacks of mowing ambient weeds are important for growers to consider.

The challenge of competition between cash crops and living mulches in integrated plasticulture living mulch systems has been documented in several previous studies (Table 1). In our study, summer squash was particularly robust to competition, with yields unaffected by between-bed management in either year (Table 5). We observed that the squash canopy shaded plants most likely to compete for in-bed resources along bed edges (Figure 5). These results are aligned with previous work in plasticulture acorn squash (Cucurbita pepo cv. Table Ace) production (Nelson and Gleason, 2018), but contrast with results from Bruce et al. (2022) where researchers found that clover/ryegrass and ryegrass living mulch between plastic mulched beds reduced zucchini (Cucurbita pepo cv. Dunja F1) yields compared to a cultivated and straw mulch control. These diverging results highlight the complex interactions between cash crop and living mulch species, management, and environmental factors that mediate competitive dynamics (Dzvene et al., 2023).

Pepper performance in our study was variable across study years, suggesting it may be more susceptible to competitive interference than squash. Law et al. (2006) also documented bell pepper yield reductions when living mulches or weeds were grown between plastic mulched beds. The researchers speculated yield reductions were driven by competition, but competitive mechanisms were not investigated. A negative relationship between marketable yield and between-bed living mulch and weed biomass suggest competition hindered pepper performance in 2017, driven by water limitations (Table 7 and Figure 2A). Reduced pepper leaf N content in all treatments compared to the cultivated control also suggest competition for N (Table 6). However, no significant differences in soil inorganic N levels in the crop row were resolved among treatments, and considerable plant available N remained under pepper plastic mulch beds at the end of the season. Pepper N uptake may therefore have been limited by soil moisture rather than N availability. Interestingly, the dead mulch treatment also reduced pepper yields in 2017. Weed escapes in dead mulch treatments were concentrated along bed edges where they may have been most likely to compete for in-bed resources, likely driving this result.

We would expect supplemental irrigation and/or fertility to help mitigate competition. However, Warren et al. (2015) found that even after increasing fertility to eliminate potential N competition and ensuring adequate irrigation, yields of broccoli (Brassica oleracea L. var. italica) grown on plastic mulched beds were reduced where living mulch was grown in the between-bed areas compared to cultivation. Alternative explanations for cash crop interference by living mulches and weeds in this case and in our study include a cooler microclimate caused by between-bed plant transpiration and the induction of a shade avoidance response.

Additional strategies could be employed to mitigate potential competition, for instance keeping the between-bed area weed-free for some critical period unique to plasticulture systems before planting living mulch, applying a dead mulch, or mowing ambient weeds. Research in inter-seeding living mulches in vegetable cropping systems has revealed great promise in delayed living mulch planting to mitigate yield penalties (Brainard and Bellinder, 2004; Gibson et al., 2011; Ciaccia et al., 2015). Though previous research has been done to evaluate the influence of in-bed weeds on plasticulture vegetable production (Buckelew et al., 2006; Norsworthy et al., 2008; Garvey et al., 2013; Chaudhari et al., 2016; Bertucci et al., 2019), research toward identifying the critical weed-free period for plants growing between plastic mulch beds is limited.

By contributing organic matter inputs, building microbial biomass, and physically protecting SOM from erosion, planting cover crops in fallow windows can increase SOM stocks over time (Kallenbach and Grandy, 2011; Peng et al., 2023). The ambient weed pressure at our study site was high enough to ensure complete soil coverage and resulted in biomass contributions comparable to our living mulch treatments, showing that in some cases the ambient weed community can be functionally like a cover crop in terms of biomass contributions. The maintenance of vegetative covers between plastic mulched beds in our trial led to average shoot residue contributions of 4,900 kg ha⁻¹ and 6,460 kg ha⁻¹ to between-bed areas (approximately 55% of the field) in 2017 and 2018 respectively, in addition to unmeasured root contributions. Considering typical biomass production of a productive summer cover crop like sorghum-sudangrass is 4,000–6,000 kg ha⁻¹ in the north central United States (Clark, 2007), this represents a significant addition of organic material. However, we did not find a detectable difference in SOM changes between trial initiation and termination among treatments. The relatively short duration of our study (2 years) combined with the difficulty in detecting changes in bulk SOM on sandy soils (Tiemann and Grandy, 2015) may explain similar changes in SOM across our treatments. It should also be noted that SOM was evaluated before living and dead mulch incorporation at experiment termination in the second year of the study, which may have underestimated the impact of residues on SOM.

Since organic matter inputs act as substrate for soil microbes, we expected microbial abundance would be elevated in all treatments compared to the cultivated control. We further expected that the greatest increases may be observed in vegetative cover treatments based on past research showing that cover crop rhizo-deposition alone can stimulate microbial growth (Finney et al., 2017), in some cases more than surface applied dead mulches (Buyer et al., 2010). However, microbial biomass C was unaffected by our treatments at any sampling date (Figure 3). Discrepancies between our results and those from previous work may be explained by methodological differences including how MBC was measured and when it was measured in relation to cover crop growth. The chloroform-fumigation method is a course resolution assessment of microbial biomass and often does not capture small changes through time (Tiemann and Billings, 2011). Distinct soil types and the study site climate are also likely to mediate results, with the effects of cover crops on soil microbial communities more likely to accumulate on medium-textured compared to fine- or coarse-textured soils (Buyer et al., 2010; Muhammad et al., 2021).

Despite no differences in microbial abundance between our treatments, we did observe greater overall extracellular enzyme activity in dead mulch and vegetative cover plots at the last cash crop harvest. This result suggests improved nutrient recycling within these management strategies compared to cultivation (Xiao et al., 2018), and aligns with a meta-analysis that found removal of plant litter and living roots reduces extracellular enzyme activity in agroecosystems (Ai et al., 2023). This also provides some evidence for improved soil health, although additional time may be needed to realize and detect benefits. Based on the Shannon diversity index calculated from extracellular enzyme activity data, microbial communities appeared to be functionally similar across our treatments. Previous research has shown that organic matter input quality can influence enzyme activity diversity (Bending et al., 2002). However, Bending et al. (2002) evaluated extracellular enzyme activity following residue incorporation, whereas the smaller organic inputs derived from surface mulch and living roots during our sampling period may reduce the likelihood of detecting microbial community effects.

The presence of living cover between plastic mulch beds may also benefit system N cycling by scavenging mineral N unlikely to be accessed by the cash crop (Xie et al., 2017). In fact, end of season soil profile (0–60 cm) N sampling in 2017 demonstrated a 61% reduction in potentially leachable N in living mulch and weedy treatments compared to cultivated and dead mulch treatments (Figure 4A). While these results were not replicated in 2018, significant in-season N losses prior to sampling may have reduced the potential to resolve differences by season’s end. In particular, heavy spring rain in May and June of that year may have promoted early season N losses via leaching and denitrification. Nevertheless, greater N uptake by living mulches and weeds in between-bed areas are likely to reduce N losses relative to bare cultivation. However, the benefit of potential reductions in N losses from between-bed areas must be balanced with the potential for competition with cash crops for in-bed N resources (Xie et al., 2017).

This research contributes to a limited body of literature quantifying systems tradeoffs between common organic grower practices for managing weeds and soil between plastic mulch beds versus living mulch alternatives. Within the context of our study, living mulches integrated into plasticulture production had limited potential to provide adequate weed control, introduced some risk to cash crop yields, and had minimal impacts on short-term soil health metrics despite notable organic matter inputs. However, the potential for soil health benefits to accrue over time, along with other possible benefits of living mulches not measured in this study, including biological pest control, increased farm biodiversity, reduced soil erosion, and improved harvesting conditions, continue to make this practice an attractive alternative to other between-bed management practices. Growers interested in planting living mulches between plastic mulch beds should proceed with caution, acknowledging variable weed control and potential risks to cash crop productivity. Future research identifying management levers on living mulch establishment, including species selection and appropriate establishment conditions could improve the weed suppressive ability of living mulches. In terms of living mulch species selection, summer annual grass species appear to be better suited than winter grasses or clovers to this niche. Additional research into irrigation and fertility management modifications could help mitigate competition between cash crops and living mulches. Identifying the critical weed-free period for between plastic mulch beds could help further mitigate competition and inform living mulch management.