Getting Legume Cover Crops to Work in Mid-Atlantic Field Crop Rotations

Study sites. Field experiments were conducted at the USDA- ARS Beltsville Agricultural Research Center near Beltsville, MD (BARC; 39.03 ˚N, 76.93 ˚W; North Farm, South Farm, East Farm, and Powder Mill Farm); the University of Maryland’s Central Maryland Research and Education Center near Glenn Dale, MD (39.01˚ N, 76.83 ˚W; Hayden Farm); and the University of Delaware’s Carvel Research and Education Center near Georgetown, DE (38.64 ˚N, 75.46 ˚W; Carvel Farm; Table 1). We had a total of nine site-years, two sites in 2017 (North and South Farms), three sites in 2018 (Hayden, East and Carvel Farms), and four sites in 2019 (Powder Mill Farm). Information on soil composition and bulk density for each field site is available in Table 1.

Table 1. Soil properties and study years of double-crop soybean-cover crop trials in the mid-Atlantic US.

All nine field experiments were initiated in the double-crop soybean phase of a two-year rotation of winter wheat/double-crop soybean - corn. Double-crop soybean at five of the sites were interseeded with cover crop treatments (North, Hayden, South, East, and Carvel Farms). Cover crop establishment was successful at three of the interseeded sites, which were maintained through the winter (North, Hayden, and East Farms). Two sites were discontinued after soybean harvest due to insufficient measurable cover crop biomass growth (South and Carvel Farms). Of the three sites where cover crop establishment was successful, North and Hayden Farms for continuation of the experiment past cover crop termination into a second year. These two sites were planted with corn and the experiment was concluded the following fall at corn harvest. The four fields at Powder Mill Farm in 2019 were included to expand the number of site years examining the effects of soybean row spacing on yield. Therefore, these fields were not interseeded with cover crops, and the experiment concluded at double-crop soybean harvest.

Experimental design. The soybean row spacing and cover crop treatments were established in a randomized complete block design with at least four replicates in each field. Each replicate block consisted of two soybean row spacing treatments (38 and 76 cm) without cover crops (no cover crop control), along with five cover crop treatments interseeded into double-crop soybean on a 76 cm wide row spacing for a total of seven treatments. The five cover crop treatments included: (1) cereal rye alone; (2) cereal rye + hairy vetch (Vicia villosa Roth); (3) cereal rye + crimson clover (Trifolium incarnatum L.); (4) cereal rye + red clover (Trifolium pratense L.); and (5) cereal rye + winter pea (Pisum sativum var. arvense (L.) Poir). Each cover crop plot was at least 6 m wide and 61 m long.

After cover crop termination in the spring, we continued the experiment at North and Hayden Farms and planted corn into the cover crop treatment plots. In June, we divided the corn growing in the cover crop plots into six subplots of 3 m wide and 9 m long (North Farm, 2017–2018) or 18 m long (Hayden Farm, 2018–2019). These subplots received the same amount of N at planting (34 kg N ha^-1), but each received a different N sidedress fertilizer rate at corn growth stage V6 (0, 34, 67, 100, 135, and 190 kg N ha ^-1). Placement of fertilizer subplots was randomized within each larger cover crop plot.

Field operations. Double-crop soybean was planted at 72,720 seeds ha^-1 in 38 and 76 cm wide row spacing following winter wheat harvest in June or early July. Soybean seed variety and planting dates are listed in Table 2. Postemergence herbicides with no residual control were applied to eliminate weed competition without interfering with cover crop establishment. Cover crop interseeding occurred at soybean growth stage R5 (late August - early September) in most of the fields, except for at East Farm, which took place at R7 (October) due to wet field conditions (Table 3). The four fields at Powder Mill Farm that were designated for row-spacing treatment trials only, were not interseeded with cover crops. The cover crop mixtures were planted into the double-crop soybean plots at the rates described in Table 4 using an inter-seeding no-till cover crop drill (Interseeder Technologies, LLC, Woodward, PA). This planting equipment sows three rows of cover crop seeds between established 76 cm wide cash crop rows using the same technology as cash crop no-till grain drills.

Table 2. Soybean planting dates, variety information, and harvest dates at field sites where soybean yield was compared between 38 and 76 cm row spacing (Objective 1).

Table 3. Soybean and cover crop planting, harvesting and termination dates at field sites where soybean yield and cover crop biomass were analyzed by cover crop treatment  were collected (Objectives 2 and 3).

Table 4. Cover crop planting species and seeding rates for a double-crop soybean interseeding experiment in the mid-Atlantic US.

The following spring, the entire experimental site was treated with herbicide to terminate cover crops and control weeds.

At North and Hayden Farms, corn (‘Pioneer 0506AM’) was planted at 74,000 seeds ha^-1 into the terminated cover crop plots on 26 May 2018 and 9 May 2019, respectively. At planting, 34 kg N ha^-1 of starter fertilizer was applied as urea ammonium nitrate solution (30% N; UAN). The six different rates of N fertilizer (0, 34, 67, 100, 135, and 190 kg N ha ^-1) were applied to the corn subplots at corn stage V6 by dribbling UAN solution between corn rows on 2 July 2018 and 26 June 2019.

Data collection. Within one day of interseeding the cover crop treatments, light transmission through the soybean canopy was measured using a point sensor and light bar sensor (LI-COR BioSciences, Lincoln, NE). Soybean was harvested in November for all three years, except at East Farm in 2018, which was harvested in December due to wet field conditions. Soybean yields were collected from the whole plot with a combine and weigh wagon, except at East Farm in 2018. At East Farm, the middle two rows of the plots were collected for a length of at least 6 m with an Almaco SPC-40 Specialized Plot Combine (Almaco, Nevada, IA) with integrated weight scale to avoid driving the harvest equipment through the flooded and waterlogged sections of the field. Soybean moisture levels were measured on sub-samples of each plot harvest using a harvest moisture meter (Churchill Industries, Minneapolis, MN) and then adjusted to 13% moisture.

Just prior to cover crop termination in the spring, aboveground biomass was collected at North, Hayden and East Farm in each plot within two 0.5 m² quadrats by cutting plants at ground level; mixtures were separated into component species. This biomass was dried at 60 °C for a minimum of seven days, weighed, and then analyzed for total C and N using dry combustion [2018 samples - LECO CN Analyzer (LECO Corporation, St. Joseph, MI); 2019 samples - Perkin Elmer 2400 Series II Elemental CHN Analyzer (PerkinElmer Corporation, Waltham, MA)].

To evaluate corn yield, the center two rows of corn from each sidedress treatment subplot were harvested with the Almaco SPC-40 specialized plot combine. Corn grain moisture was measured by the integrated combine moisture meter for each N rate plot in 2018 and as a whole plot area mean in 2019, and then grain weight was adjusted to 15.5% moisture content.

Statistical approach. In order to test the effect of row spacing on double-crop soybean yield at the individual sites, a linear mixed effects (LME) model was used with row spacing as the main effect and block as the random effect. To test the effect of cover crop interseeding and fall cover crop growth on soybean yield at North, Hayden and East Farms, an LME model was used with cover crop treatments (e.g., no cover crop, cereal rye, cereal rye + crimson clover, etc.) as the main effect and block as the random effect. Additional LME models were used to test the effect of cover crop treatment on spring cover crop biomass production and cover crop N concentration with cover crop treatment as the main effect and block as the random effect. All LME models were run using the lmer function in the nlme package in R (Bates et al., 2015; Pinheiro et al., 2020).

To analyze subsequent corn yields, we tested an ensemble of models with cover crop treatment, N rate, and their interaction as fixed effects and block as the random effect. All LME models for corn yield were run using the lme function in the nlme package in R with the maximum likelihood method (Pinheiro et al., 2020). We first assessed curves describing plateau, linear, quadratic, linear plateau, and quadratic plateau mixed effects models of corn yield response to N rate by cover crop species. Model selection was conducted using likelihood ratio tests on hierarchically nested pairs of models (e.g., adding or removing a single coefficient at a time). The following two models were fit; each coefficient was estimated for each cover crop treatment:

Linear:                        Y = 𝜷₁·X + 𝜷₀ + B_Block + 𝜺                                                   (1)

Quadratic:                   Y = 𝜷₂·X² + 𝜷₁·X + 𝜷₀ + B_Block + 𝜺                                       (2)

where Y is the dependent variable [corn grain yield (Mg ha^-1)] and X is the N rate (kg ha^-1). P is the plateau Y value and X_P is the value for X at which the linear or quadratic response meets the plateau. After selecting the model according to the likelihood ratio tests, coefficient contrasts were conducted with the emmeans (Lenth, 2019) and multcomp (Hothorn et al., 2008) packages. Goodness-of-fit for each model was reporting using an R² analog that is derived from the likelihood ratio test (Magee, 1990; Kramer, 2005):

R² = 1-exp(- (2/n)(logL_m-logL_o))                                                                              (3)

where logL_M is the log-likelihood of the model of interest, logL_O is log-likelihood of the intercept only model, and n is the number of observations. No transformations were necessary to improve the normality of the soybean yield, cover crop biomass, or corn yield data sets. Analyses were conducted in the R environment for Macintosh (R Core Team, 2020).

At this stage, I have finished collecting and analyzing all of the data pertaining to Project Objectives 1, 2, 3, and 4.