Cover crops and weed suppression in the U.S. Midwest: A meta‐analysis and modeling study

In addition to soil health and conservation benefits, cover crops (CCs) may offer weed control in the midwestern United States, but individual studies report varying effects. We conducted a meta‐analysis of studies measuring weed biomass (WBIO) or density (WDEN) in paired CC and no‐cover treatments in corn (Zea mays L.)–soybean [Glycine max (L.) Merr] rotations in the U.S. Midwest. Fifteen studies provided 123 paired comparisons of WBIO and 119 of WDEN. Only grass CCs significantly reduced WBIO, while no CC reduced WDEN. We found no evidence CC management factors (e.g., termination method) directly affected outcomes. Our dataset showed that a 75% reduction in WBIO requires at least 5 Mg ha−1 of CC. Simulations from a process‐based model (SALUS) indicated achieving 5 Mg ha−1 requires substantially earlier fall planting and later spring termination in most years, conflicting with typical cash‐crop planting and harvesting. We conclude CCs significantly reduce WBIO, but current CC management constraints render these reductions variable and uncertain.

soil health and water quality. Despite clear environmental benefits (Daryanto, Fu, Wang, Jacinthe, & Zhao, 2018;Kaspar & Singer, 2011), less than 10% of midwestern cropland is currently managed with CCs (Seifert, Azzari, & Lobell, 2018). The lack of short-term economic returns from growing CCs overwhelms long-term environmental benefits, creating a major barrier to wide adoption (Plastina, Liu, Miguez, & Carlson, 2018;Roesch-Mcnally et al., 2018). If CCs can reduce weed management costs, this could provide immediate monetary incentives for adoption. Previous literature syntheses have found CCs reduce weed pressure across various cropping systems, but the direction and magnitude of effects are context-specific (Osipitan, Dille, Assefa, & Knezevic, 2018). Given its ubiquity and significance in the U.S. Midwest, the corn (Zea mays L.)-soybean [Glycine max (L.) Merr] production system merits explicit examination. Unfavorable fall-winter climatic conditions in the Midwest are known to limit CC establishment and growth (Baker & Griffis, 2009;Strock, Porter, & Russelle, 2004), which in turn may affect factors governing CC performance relative to weed management. Region-specific analyses can also provide more precise CC biomass (CCBIO) production targets for weed suppression (Baraibar et al., 2018;Mirsky et al., 2013) and explore how planting or termination timing affects the feasibility of achieving those targets.
To address these gaps, we synthesized data from published field studies measuring weed responses to CCs in corn-soybean systems in the Midwest. Our objectives were (a) to quantify how environmental conditions and management practices affect weed responses to CCs, (b) to identify Midwest-specific CCBIO targets for providing significant weed suppression, and (c) to evaluate the feasibility of achieving these targets under different CC planting and termination scenarios.

Meta-analysis of weed-responses to cover crops
We conducted a systematic search of the literature using Web of Science Core Collection (Clarivate Analytics) and CAB Direct (CAB International) databases. Search details, including a PRISMA diagram and list of included publications, are in the supplementary material (Supplemental Material S1). In our database, we included weed biomass (WBIO), weed density (WDEN), and cash-crop yield as response variables. We recorded values in a paired format, requiring each pair of response variables to be measured in the same crop at the same time with all aspects of management held constant except for a treatment of a

Core Ideas
• Cover crops reduce weed biomass but not weed density. • Grass monoculture cover crops offer the most consistent weed suppression. • At least 5 Mg ha −1 of cover crop is required to reduce weed biomass 75%. • Producing 5 Mg ha −1 of cover crop requires early planting and late spring termination. • Managing cover crops for weed suppression will require changes in policy and agronomy.
fall-planted CC. Ancillary data included geographical location, climate, and soil characteristics of the study site; cashcrop and CC management including species, tillage system, planting and termination methods and dates; and experimental information such as timing of weed measurements and type of weed (Supplemental Material S1). The complete database is published and available on Iowa State University's DataShare platform (Nichols, Basche, & Weisberger, 2020). All data manipulation and statistical modelling were done in R version 3.6.1 (R Core Team, 2019) using the tidyverse meta-package (Wickham, Averick, Bryan, Chang, & McGowan, 2019) and others (Firke, 2019;Grolemund & Wickham, 2011). A detailed account of statistical methods is presented in Supplemental Material S2, and all R code is available on github (https://github.com/vanichols/ ccweedmeta-analysis). In brief, all statistical models used the log-transformed response ratio (measurement in the CC treatment over measurement in the no-cover treatment) as the response variable (Gurevitch, Koricheva, Nakagawa, & Stewart, 2018). Mixed-effect models were used with the modifier of interest as a fixed effect and a random intercept for each study using nonparametric weighting based on the number of replicates (Adams, Gurevitch, & Rosenberg, 1997). All linear models were fit using the lme4 package (Bates, Mächler, Bolker, & Walker, 2015), and results were analyzed using lmerTest (Kuznetsova, Brockhoff, & Christensen, 2017) and emmeans (Lenth, Singmann, & Love, 2018). Means and 95% confidence intervals were back-transformed for reporting purposes. To identify suites of practices predictive of achieving both a reduction in weeds and an increase in cash-crop yield with CCs, we fit random forest models (Kuhn & Johnson, 2013) using several R packages (Hothorn, Hornik, & Zeileis, 2006). All statistical results are in Supplemental Material S3.

Simulation of cover crop biomass
To investigate the feasibility of growing CCs for effective weed control in the Midwest, we used the System Approach to Land Use Sustainability (SALUS) model (Basso & Ritchie, 2015) to simulate winter rye (Secale cereal L.) biomass across a range of soils and weather conditions of the Midwest. Rye is the most prevalent CC species used in the Midwest (Singer, 2008) and represents the best choice for maximizing CCBIO production in this region (Appelgate, Lenssen, Wiedenhoeft, & Kaspar, 2017;Ruis et al., 2019). Specific simulation details are provided in Supplemental Material S4. Three CC planting dates were explored: 15 September (optimistic), 7 October (realistic), and 1 November (late).

Meta-analysis results
Fifteen articles fit our criteria, producing 123 response ratios for WBIO and 119 response ratios for WDEN . The studies include a range of site characteristics and management representative of midwestern cornsoybean production systems (Supplemental Material S1).
Weed suppression was significantly affected by CCBIO for both WBIO (p = .03) and WDEN (p < .01). We found an estimated 5 Mg ha −1 of CCBIO at termination reduced WBIO by 75% for grass CCs but only 40% for other CCs (Figure 1).
The response of WBIO or WDEN to CC did not depend on any other modifiers tested. A full list of nonsignificant modifiers can be found in Supplemental Material S3 and included production system tillage regime; CC planting and termination method; termination-planting gap; and study-site latitude, aridity, and soil type.
In our database only 23% of the comparisons exhibited a "win-win" situation, with a concomitant increase in cashcrop yield and decrease in weed pressure (Figure 1). Using a random forest model, we found no scenarios that were strong predictors of whether an observation would fall in the win-win category, suggesting maximizing cash-crop yields and weed suppression may not have overlapping CC management strategies.

Simulation model results
For the "realistic" planting date (7 Oct.), 2% of counties achieved 5 Mg ha −1 by 1 May in ≥80% of the weather-years, increasing to only 30% under an "optimistic" CC-planting scenario (15 Sept.; Figure 2). With "late" planting (1 Nov.), none of the counties reached the threshold by 1 May, and only half did so by 1 July. Aggregated on a state level, Illinois, Missouri, and Kansas were the only states that could consistently achieve 5 Mg ha −1 of biomass before typical cash-crop planting dates of early May with optimistic CC planting dates (Figure 2).

DISCUSSION
Cover crops affect weeds through interference mechanisms of resource competition and allelopathy (Teasdale & Mohler, 1993), delaying weed germination and development that manifests as lower WBIO. Management that disrupts rather than interferes with weed trajectories, such as crop rotation, may be more effective at reducing WDEN (Weisberger, Nichols, & Liebman, 2019). However, given that reductions in WBIO can increase susceptibility to herbicides (Wallace, Curran, & Mortensen, 2019) and weed size is directly related to seed output (Thompson, Weiner, & Warwick, 1991), reductions in WDEN may be possible with long-term CC use. More long term (>5 yr) work is needed to answer this question.
While CCs had a stronger effect on weeds before cashcrop planting (Figure 1), weeds measured after planting are likely of more interest to producers, as they directly represent resource competition with the cash crop. The stronger reduction in winter annual weeds is not surprising, given the winter growth period of the CC.
The environmental context of the studies had no significant effect on the weed responses or on CCBIO. This could simply reflect the lack of plotspecific information (Eagle et al., 2017;Gerstner et al., 2017), but it does suggest environmental context has only an indirect effect on CC-mediated weed suppression.
To prevent an increase in weed seedbanks, reductions in WDEN of 90% (comparable to herbicide effectiveness) are needed (Liebman & Nichols, 2020); our study shows that even with 5 Mg ha −1 of CCBIO, producers are unlikely to achieve this level of weed control, consistent with studies from other areas (Baraibar et al., 2018;Mirsky et al., 2013). Moreover, our SALUS simulations indicate achieving 5 Mg ha −1 of rye CCBIO regularly under typical Midwest production scenarios and climates would be challenging (Figure 2). Even with optimistic CC planting dates (15 Sept.), achieving 5 Mg ha −1 of CCBIO would require a mid-May or later termination date most years (≥80%) F I G U R E 2 Earliest termination date with rye biomass in excess of 5 Mg ha −1 as predicted by the SALUS crop model using 30 yr of historical weather for three rye planting date scenarios (15 Sept., 7 Oct., 1 Nov.). (Left) Results summarized by state at 80% probability levels. In Iowa, for example, rye biomass was >5 Mg ha −1 in 80% of the years if planted on 7 Oct. and terminated on or after 17 June (highlighted in red). (Right) Results corresponding to the 7 Oct. planting scenario, summarized by county at the 80% probability level in the majority of counties, well after typical cash-crop planting dates. It should be noted our simulations assumed direct CC seeding with uniform germination (Supplemental Material S4) and are therefore not to be extrapolated to other planting methods. While aerial-or interseeding can be used to establish CCs into standing crops, these methods are often unreliable (Wilson, Allan, & Baker, 2014), and standing crops prevent full sunlight penetration for CC growth well into October. Delayed corn and soybean planting consistently reduces yields (Baum, Archontoulis, & Licht, 2019;De Bruin & Pedersen, 2008), and delayed CC termination could be hindered by concerns over crop insurance eligibility (USDA-NRCS, 2019). High CCBIO production could increase other ecosystem services (Blanco-Canqui et al., 2015;Thapa, Mirsky, & Tully, 2018) but may also introduce issues with nitrogen immobilization and CC termination (Whalen et al., 2020). Other studies examined the effects of CCs on subsequent cash-crop yields (Marcillo & Miguez, 2017), showing no yield benefit from grass CCs. Choosing a CC species to maximize cash-crop yields may be at odds with choosing one for maximizing weed suppression, and while no-till may amplify yield responses (Marcillo & Miguez, 2017), it may not enhance weed control from CCs. The existence of these trade-offs is supported by the low percentage of observations with a "win-win" scenario ( Figure 1) in our database.

CONCLUSIONS
Our study, which synthesized work from the Corn Belt region of the U.S. Midwest, shows that grass CCs effectively reduce WBIO. We estimated 5 Mg ha −1 of grass CCBIO decreases WBIO by 75%, a threshold at which reduction of herbicide use is possible, but not always advisable. Furthermore, consistently achieving that level of CCBIO in the Midwest may not be feasible within the traditional corn-soybean fallow season. In our dataset, concomitant increases in yields and decreases in weeds with the use of CCs were minimal, highlighting the need to evaluate CC practices using multiple metrics. Therefore, we conclude that although CCs significantly reduce WBIO, which may render other weed management strategies more effective and reduce WDEN in the long-term, current CC management does not consistently suppress weeds. Optimizing CCs for weed suppression will entail both agronomic (e.g., use of different cash-crop maturity groups) and policy (change in insurance structure around CC termination requirements) changes at a broad scale.

A C K N O W L E D G M E N T S
We would like to acknowledge Alisha Bower who assisted with literature searches, Stefan Gailans who provided helpful feedback, Megan O'Donnell who assisted with dataset publication, Katherine Goode who provided statistical advice, and Matt Liebman who provided moral support. We also thank two anonymous reviewers whose insightful comments improved this manuscript. This material is based upon work supported by the National Science Foundation (Grant No. DGE-1828942), USDA-NIFA (award: 2019-67012-29595), and the North Central Region Sustainable Research and Education Program (Grant No. 2017-38640-26916).

C O N F L I C T O F I N T E R E S T
The authors declare no conflict of interest.

D ATA AVA I L A B I L I T Y
All data associated with this analysis have been published  and are publicly available at https: //iastate.figshare.com/. Additionally, the data are available as an R package on github (https://github.com/vanichols/ ccweedmetapkg), and all R code used to analyze the data is available in a github repository (https://github.com/ vanichols/ccweedmeta-analysis).