Are polycultures for silage pragmatic medleys or gallimaufries?

Abstract

Polycultures, mixtures of different crop species in the same field, may provide both production and ecological benefits. Silage production in annual cropping systems may incorporate polycultures and take advantage of species’ niche partitioning, potentially stabilizing yield variation due to abiotic stress. Using maize (Zea mays L.) silage as the basis of our 3-yr study, we tested the impact on crop and soil attributes of replacing a fraction of maize with soy [Glycine max (L.) Merr.], sorghum (Sorghum bicolor var. bicolor × bicolor and var. sudanense [unnamed hybrid]), or a medley of soy, sorghum, and sunflower (Helianthus annuus L.). Compared to maize monocultures and on average for the 3 yr, a replacement mixture of maize + soy lowered yields (1.57 vs. 1.87 kg m^–2), but increased the N, P, and K concentration in the silage by 1.2x, 1.09x, and 1.03x, respectively. Maize + sorghum polycultures matched the biomass yields of maize monocultures (1.77 vs. 1.87 kg m^–2) and increased K concentration (10.2 vs. 8.2 g kg^–1). While random forest analysis revealed no change in post-harvest soil mineral N with depth among treatments, there was a tendency for higher total mineral N left in the soil for soy-containing vs. sorghum-containing treatments (12.4 vs. 10.9 g m^–2). Silage polycultures are a feasible alternative to maize silage monocultures and can improve silage nutrient concentration with no yield penalty if maize or sorghum dominate plant stands.

Abbreviations

B

biomass in kg m^–2

BMR

brown midrib sorghum

ET_o

reference evapotranspiration

HI

harvest index

K_C

potassium concentration in mg kg^–1

K_M

potassium by mass in g m^–2

N_C

nitrogen concentration in mg kg^–1

N_M

nitrogen by mass in g m^–2

NH₄⁺

ammonium

NO₃^-

nitrate

P_C

phosphorus concentration in mg kg^–1

P_M

phosphorus by mass in g m^–2

YMR

yellow midrib sorghum

Θ_g

gravimetric water content

ρ_b

bulk density

Ψ_L

water potential of the leaf in J kg^–1

Ψ_S

water potential of the soil in J kg^–1

Ψ_x

water potential of the xylem in J kg^–1

1 INTRODUCTION

Dairy farms often rely on maize (Zea mays L.) silage to provide a high-energy and nutritive feed for animals year-round. Among the U.S. dairy states, Pennsylvania ranks sixth for dairy production and has an estimated 7,059 farms that produce silage on 143,000 ha at an average yield of 15 Mg ha^–1 (dry basis) (USDA National Agricultural Statistics Service, 2017). While maize silage is important to dairies across the state and generally produces high yields compared to other ensiled crops, its production is vulnerable to abiotic stress stemming from climate change and its associated variability, as well as variable soils. Moreover, maize silage production, as part of the agricultural land-use portfolio in Pennsylvania, contributes N and P loading to the Chesapeake Bay watershed with detrimental effects downstream (Ator, Garcia, Schwarz, Blomquist, & Sekellick, 2019). The challenges of dairy production in Pennsylvania are experienced in other dairy states and watersheds. In this work, we investigated polycultures for silage production by combining species with complementary traits. By using these mixtures, we aimed to mitigate reductions in silage yield due to abiotic stresses while providing environmental benefits like increased nutrient uptake, reduced residual inorganic N, and increased biodiversity.

Adjustments to the current maize-dominated production system might be necessary because climate change can alter production conditions in temperate climates like that of Pennsylvania. Models predict that by the mid-21st century the northeastern United States will experience an increase of up to 21 d above 35 °C, with larger increases in the southern portion of the region (Kunkel et al., 2013). Additionally, early in the growing season more frequent intense precipitation events are expected (Vogel, Hauser, & Seneviratne, 2020; Wolfe et al., 2018). Despite this increase in precipitation, short-term yield-limiting droughts during critical stages of plant development will become more common in the Northeast due to increased evapotranspiration without an increase in precipitation during the summer months (Prasad et al., 2018; Wolfe et al., 2018). The frequency of extreme hot events is also expected to increase (Vogel et al., 2020). Moreover, shallow and stony soils of many dairy farms in central Pennsylvania have low total water holding capacity (Braker, 1981), exacerbating the risk and magnitude of droughts.

Crops can minimize production losses due to water stress through two main pathways. First, plants can regulate their timing of water use. Plants that have lower maximum transpiration rates can better tolerate droughts (Messina et al., 2015; Sinclair, Hammer, & Van Oosterom, 2005). Species like sorghum (Sorghum bicolor var. bicolor × bicolor var. sudanense [unnamed hybrid]) generally have lower stomatal conductance than maize and exhibit this behavior ( Khan, Link, Hocking, & Stoddard, 2007; Laffray & Louguet, 1990; Sinclair, Tanner, & Bennett, 1984). This prevents high water losses during high temperatures and dry air conditions and can delay the onset of severe water stress. These and other stress tolerance mechanisms, like improved silking and altered root architecture, continue to improve maize drought tolerance (Campos et al., 2006; Cooper et al., 2014; Hammer et al., 2009). However, these drought-tolerant traits are often priced at a premium in elite hybrids, or are grouped with other traits (herbicide tolerance, insecticide-coated seed) that make seed more costly than alternative options.

Second, plants can minimize production losses by more completely extracting water from the subsoil. Sunflower (Helianthus annuus L.), for example, tends to explore the soil profile faster and deeper than other species like sorghum (Stone et al., 2001), thus accessing deeper soil water reserves. Using deep water faster rather than decreasing production at the first sign of stress can also buffer a potential drop in productivity, provided that the dry period is short (Bremner et al., 1986). In addition, and at the right planting density, sunflower leaves cover the ground rapidly. As a result, shade cast by the leaves diminishes soil–water evaporation, improving the water capture efficiency by the plant stand (Aguera et al., 1997; Blum,  2009).

Core Ideas

• Biomass yield of maize + sorghum polyculture matched that of maize monoculture.
• Soy increased N and P concentration of silage, but reduced yield.
• Adding soy or sorghum to maize increased the K concentration of silage.
• Residual soil mineral N appeared to be lower in polycultures with sorghum.

For decades, production systems in the United States have exploited the efficiencies of monocultures (Giller et al., 2015), but one of the tradeoffs is that our current system may be more prone to regional yield losses due to the lack of local diversification. Polycultures have the potential to counteract biotic stresses like insect pests and provide weed suppression and functional diversity in cropping systems (Andow, 1991a, 1991b; Finney and Kaye, 2017; Iverson et al., 2014; Liebman, 1995; Martin et al., 1989; Picasso, Brummer, Liebman, Dixon, & Wilsey, 2008). However, it is unclear whether the adaptive attributes of polyculture can counteract abiotic stresses associated with climate change and other negative effects in annual silage production. Because maize silage production entails chopping and ensiling whole plants (i.e., the biomass gets mixed anyway), we propose using polycultures for silage production to exploit functionally diverse traits to mitigate abiotic stress.

An effective polyculture for silage would ideally produce large and stable yields and reduce inputs. Polycultures are common in perennial pastures (Dobson, Fisher, & Beaty, 1976; Mallarino, Wedin, Perdomo, Goyenola, & West, 1990; Wedin, Donker, & Martin, 1965) and can produce greater and more stable yields compared to monocultures (Bonin & Tracy, 2012; Serajchi, Schellenberg, Mischkolz, & Lamb, 2018). Additionally, annual high-biomass grasses mixed with legumes can produce yields comparable to those of grass monocultures and are more stable across years and environments (Bybee-Finley, Mirsky, & Ryan, 2016). However, this is not always the case (Cardinale et al., 2011; Lauriault & Kirksey, 2004). Silage production with annual crops is comparable to that of perennial forage production in that the entire aboveground portion of the plant is used by the animal. Unlike forage, silage prevents the animal from selecting one plant over another, but silage quality and palatability must remain high.

Total production and nutrient composition of silage polycultures depends on the species composition of the mixture. Maize + soy (Glycine max (L.) Merr.] silage polycultures can produce similarly to maize monocultures, but with more yearly variation (Carruthers et al., 2000; Herbert, Putnam, Poos-Floyd, Vargas, & Creighton, 1984). Additionally, aboveground biomass is highly dependent on total maize density because increasing maize density increases total aboveground biomass (Erdal, Pamukcu, Curek, Kocaturk, & Dogu, 2016; Martin, Voldeng, & Smith, 1990; Putnam, Herbert, & Vargas, 1985). Despite the potential for reduced production of maize + soy polycultures, these mixtures have been shown to provide other benefits over maize monocultures including residual soil inorganic N to the following crop that translates into better nutritional quality indicators like increased protein content and decreased acid detergent fiber (Erdal et al., 2016; Putnam, Herbert, & Vargas, 1986). Conversely, maize + sorghum mixtures can maintain production under drought stress due to sorghum's ability to capture water efficiently in low-moisture soils and withstand lower soil water potentials (Muchow, 1989; Schittenhelm, 2010; Singh & Singh, 1995). The nutritional value of sorghum and maize silage is similar and dairy cows produce comparably on either one (Nichols, Froetschel, Amos, & Ely, 1998; Nordquist & Rumery, 1967). This is especially true with the introduction of the low-lignin brown midrib (BMR) hybrids whose digestibility is similar to that of maize (Oliver, Grant, Pedersen, & O'Rear, 2004). Choosing one species over the other depends largely on environmental conditions (Nordquist & Rumery, 1967). While these studies have provided insight into some of the potential benefits of polycultures, they have not addressed impacts on N, P, and K plant uptake or the potential for polycultures to mitigate nutrient pollution in agricultural systems.

Nutrient management is important for environmental quality. Excess N and P lost from agricultural fields contribute to air pollution and to water pollution in rivers, lakes, estuaries, and oceans. Under changing climate conditions, mitigating pollution to large bodies of water is expected to become more challenging due to greater precipitation (Bosch, Wagena, Ross, Collick, & Easton, 2018). Residual end-of-season inorganic N is prone to leaching and residual P may be moved into water bodies through soil erosion or tile drains, both of which are increased with greater precipitation. Efficient and complementary nutrient use in polycultures may result in fewer pollution-prone nutrients, like N and P, in the environment (Bracken & Stachowicz, 2006; Cardinale et al., 2006; Tilman, Wedin, & Knops, 1996), suggesting that polycultures could be used in agricultural systems to reduce nutrient losses.

Forage quality and environmental quality, both a function of N and P availability, can be at odds. While N and P in water leaving fields are considered pollutants, these elements, along with K, are important nutritional components and indicators of silage quality. When feeding maize silage, soy meal and sunflower meal are used as protein supplements in rations (Eastridge, 2006). Previous work has shown that silage comprising maize and legumes, either grown together or mixed post-harvest, can increase protein and overall quality (Maasdorp & Titterton, 1997; Martin et al., 1990; Titterton & Maasdorp, 1997). A silage polyculture would ideally balance N, P, and K in the feed, increase the protein content compared to maize silage, and provide environmental benefits through nutrient uptake from the soil and potential nutrient excretion reduction by livestock.

Our goal was to determine if silage polycultures can confer tolerance to abiotic stresses and improve forage quality without hampering yield. Additionally, we hoped that polycultures may mitigate negative environmental impacts by reducing residual mineral N. To address our goals, we tested mixtures that replaced a fraction of maize with sorghum, soy, and sunflower. For comparison, we also tested maize, sorghum, and soy monocultures. Our hypotheses were as follows: (a) polycultures with sorghum replacing a fraction of maize will produce similar yields as maize monocultures in normal years and buffer yield in dry years and (b) soy will increase the N and P content and N and P yield of the polycultures but may reduce biomass yield. We intended to test the performance of sunflower in the mixture, but the density of this species was below target and was a minor component of the stands. Overall, we expected that polycultures, when compared to maize monocultures, may not result in increased yields every year, but may improve yield stability across years due to their varying traits and niche partitioning while providing other environmental benefits.

2 MATERIALS AND METHODS

2.1 Experimental site and design

The experiment was conducted from 2016 to 2018 at the Penn State Russell E. Larson Agricultural Research Center at Rock Springs, PA (40°42′58.3″ N, 77°56′05.3″ W; elevation 330 m above sea level).

Four species—maize, soy, forage sorghum, and sunflower—were included in the experiment. Two hybrids or cultivars of each species were used, denoted as species A and species B (Supplemental Table  S1). For maize, this included a drought-tolerant hybrid and a hybrid commonly used in the area. Soy included one glyphosate [N-(phosphonomethyl) glycine]-tolerant variety specifically bred as a forage variety, and one high-protein, conventionally bred variety. Sorghum included a conventional yellow midrib forage sorghum hybrid (YMR, sorghum hegari) and, as an insurance against a loss of silage quality, a brown midrib (BMR) sorghum–sudangrass hybrid. The BMR sorghum is derived from a genetic mutation that lowers lignin content, which increases overall digestibility of the plant material (Bucholtz, Cantrell, Axtell, & Lechtenberg, 1980). Sunflowers consisted of two conventionally bred black oilseed varieties. Soy plots in 2017 had poor stands, first due to herbicide damage and second due to groundhog (Marmota monax) damage after replanting.

We tested seven treatments, four monocultures and three polycultures where polyculture seeding rates were based on fractions of monoculture seeding rates (Table 1). Maize was often seeded at a higher fraction in the mixture compared to other species to both ensure high biomass and because maize is the backbone of the mixtures. Maize monocultures were used as the baseline for the experiment. The monocultures were maize A, maize B, soy A + B (soy), and sorghum A + B (sorghum), and the polycultures were maize + soy, maize + sorghum, and maize + soy + sorghum + sunflower (medley). All treatments except for maize A and maize B included two genotypes or cultivars in equal proportion for each species present in the treatment. Sunflower was only considered as a small proportion of the medley based on prior work (Fisher, Bittman, Mir, Mir, & Shelford, 1993; He, Mir, Beauchemin, Ivan, & Mir, 2005) and we did not test pure stands of sunflower. Proportions of each species in mixtures were based on ideal monoculture planting densities for that species (e.g., the maize in medley was planted at density of 8 × 0.33 = 5.3 plants m^–2 where 8 plants m^–2 is the usual density in this location). The experiment was set up in a randomized block design with five replicates. This design was used in all 3 yr in different fields.

TABLE 1. Monocultures seeding rate and fraction (0–1) of monoculture seeding rate for each species when in mixtures. When maize was in a mixture, it included equal parts of maize A and B

Species	Maize	Soy	Sorghum	Sunflower	Sum
Monoculture seeding rate, plants m^–2	8	25	10	8
Fractional planting rate
Maize A	1	0	0	0	1
Maize B	1	0	0	0	1
Sorghum	0	0	1	0	1
Soy	0	1	0	0	1
Maize + sorghum	0.66	0	0.34	0	1
Maize + soy	0.66	0.34	0	0	1
Medley	0.33	0.24	0.33	0.1	1

Details of site management and field characteristics are given in Supplemental Table S2. Soils were primarily silt loams with 0–3% slope (NRCS-USDA, 2019). All crops were no-till planted and glyphosate was used as a burndown. Herbicide programs specific to each treatment varied year to year based on the recommendation of the Penn State Weed Extension Specialist and were applied at labeled rates. Plots were fertilized according to soil test results by the farm manager. The planting dates were 26 May 2016, 2 and 3 June 2017, and 15 June 2018. Planting was delayed due to rains in 2018. All treatments were planted at 76-cm row spacing. Maize was planted with a maize planter; all other seeds were planted in one pass with the InterSeeder (Interseeder Technologies, LLC.).

2.2 Biomass

Aboveground biomass harvests were made when moisture content was roughly 65%, the ideal moisture content for harvesting silage, usually in mid-September. However, in 2016 the harvest was late and closer to maize physiological maturity. We sampled biomass by hand from the center of the plots (1 m in 2016 and 2017, and 2 m in 2018). Within the sampled area and to measure total aboveground biomass, we harvested plants at a 1-cm cutting height and kept the plant species separated for all subsequent processing. The two genotypes of each maize, soy, and sunflower were not separated when in a mix; soy was also not separated when in a monoculture. Sorghum was separated as YMR or BMR in both mono- and polycultures since they were distinguishable. Freshly harvested material was weighed and dried at 50 °C until the drying mass was stable. After drying, maize grain was separated from other aboveground biomass (leaves, sheaths, cob, and stalk) and weighed. As an indicator of silage quality, harvest index (HI) was calculated for maize as the ratio of the grain biomass to total aboveground biomass. Harvested tissue samples of each individual species in each treatment were ground to 2 mm and sent to the Penn State Agricultural Analytical Services Lab for composition analysis of N, P, and K using acid digestion.

2.3 Plant nutrient analysis

For each nutrient, the concentration (N_C, P_C, K_C, g kg^–1) and mass (N_M, P_M, K_M, g m^–2) were calculated for each species in every treatment. The concentration of N, P, and K functions as an important aspect of silage quality while the mass of these elements taken up by plants may have environmental implications. In maize where grain and stalk were separated, and in sorghum where YMR and BMR were separated, a weighted analysis was used to reconstruct the unseparated sample based on the proportion by dry weight of each component.

2.4 Water relations

Water potential (Ψ) readings were used as a measure of water stress and as potential indication of differential root distribution of the species (Camargo & Kemanian, 2016). Measurements were taken twice per year in 2016 and 2017 using a Scholander pressure chamber (Model 615, PMS Instrument Company). The first reading was taken after a rain that brought soils to field capacity where plants should be minimally stressed and the second reading after a minimum of a 7-d drying period, at which point plants should experience greater water stress as the soil dries. Sunlit leaves at pre-dawn and noon were harvested and measured to determine the soil (Ψ_S) and leaf (Ψ_L) water potential, respectively. In 2017, a foiled leaf was harvested at noon and was used to measure the xylem water potential (Ψ_X). These potentials allow assessing the direction of water flow in the plant as well as the comparative magnitude of the resistances to liquid water flow between root (mostly endodermis) and leaf (mostly the mestome sheath in monocots and similar structure in dicots). Pre-dawn potentials reflect an average soil water potential as seen by the root system, integrating root density or activity per layer and soil moisture per layer (Camargo & Kemanian, 2016; Campbell et al., 1976). Four cultivars (maize, BMR, YMR, and soy) were measured in 2016 in maize, sorghum, maize + sorghum, maize + soy, and medley treatments. In 2017, water potential was measured in maize, BMR, and YMR species in the monoculture plots only. Maize A and B were not measured separately as they were indistinguishable in mixes and maize in monocultures were treated as a single entity. Sorghum A and B were easily distinguishable and therefore measured separately. Additionally, the fraction of resistance to water flow in the root was calculated in 2017 as (Ψ_X – Ψ_S)/(Ψ_L – Ψ_S). In both years, measurements were made across all five blocks.

2.5 Soil nitrogen

Soil cores were taken after harvest using a Giddings soil probe (Giddings Machine Company). Samples were taken as deep as possible up to 1.1 m. Rocky soil prevented most cores from reaching this depth and core length ranged from approximately 0.4–1.1 m. Cores were frozen horizontally at –20 °C until processed. During processing, cores were split by genetic horizon and the gravimetric water content (Θ_g) as well as nitrate (NO₃^–) and ammonium (NH₄⁺) concentrations determined. These cores could not be used to reliably determine soil bulk density (ρ_b, Mg m^–3); this variable was estimated using a pedotransfer function that included organic matter (Saxton & Rawls, 2006). The rock-free Θ_g was measured by taking a 20-g sample of wet soil, drying it at 105 °C until the weight was stable, sieving rocks >2 mm, and drying and weighing the rocks. Inorganic N was extracted using 2 M KCl. Nitrate was detected using vanadium (III) chloride and ammonium was determined using salicylate/nitroprusside (Doane & Horwath, 2003; Sims, Ellsworth, & Mulvaney, 1995). Concentrations of NO₃^– and NH₄⁺ were converted to mg kg^–1 (mg of N species per kg of soil), and to g m^–2 using ρ_b and thickness of each soil layer.

2.6 Data analysis

Total dry biomass was determined by summing all the dry masses for each species per m² per plot. To detect outliers, we calculated the median absolute deviation (MAD) within a treatment and considered observations more than 3 MADs from the median as outliers (Leys, Ley, Klein, Bernard, & Licata, 2013). Only one outlier was detected (2017, maize + sorghum treatment) and discarded. Additionally, one plot of soy monoculture failed in 2017 and was considered missing.

Least squares means for biomass yields and silage nutrient content were modeled in SAS (v9.4, SAS Institute) using the Mixed procedure to account for both fixed and random effects with year (random with block nested within year), treatment (fixed), and the year × treatment interaction as regressors and missing plots. A slice test (an analysis of simple effects accounting for interactions) by year and a P value of .05 was used to determine statistical significance between treatments within each year. Biomass yields and silage N concentration for individual block observations each year were plotted in RStudio using the boxplot() function.

Water potential was analyzed similarly to biomass, using the Mixed procedure in SAS with a slice test, with the time of day (pre-dawn or noon), leaf (sunlit or foiled), and treatment (silage type, when applicable) as regressors. Random effects were kept as block nested within year.

Laboratory processing of soil cores were by genetic horizon, thus making it difficult to compare inorganic N concentrations across plots since the depth of each genetic horizon varies by sample. To amend this, after laboratory processing ρ_b, NO₃^– (mg NO₃^– kg^–1 soil), and NH₄⁺ (mg NH₄⁺ kg^–1 soil) concentrations were estimated by fixed depth (0.1, 0.3, 0.5, 0.7, and 0.9 m) using the package random forest, Breiman and Cutler's Random Forests for Classification and Regression (version 4.6-14, CRAN; Liaw & Wiener, 2018) in RStudio (RStudio Inc., 2019). For each of the three variables, 300 regression trees were created. The random forest models considered the following predictors: year, treatment, depth, horizon, row, and column. Depth is the midpoint depth of the soil layer, horizon is the soil genetic horizon, and row (block) and column represent the plot location in the field (the row and column were unique for each year). While depth and horizon have some redundancy, they do not affect the models’ accuracy and allow using them for prediction. Plant N uptake was also tested as an additional variable in the random forest model but resulted in a minimal effect on total NO₃^– and NH₄⁺ left in the soil (Supplemental Figure S1) and was subsequently left out of the model reported here. The g m^–2 of NO₃^– and NH₄⁺ in each layer were calculated using ρ_b and depth from the random forest output. The total mass of NO₃^–, NH₄⁺, and total N per plot to fixed 0.9-m depth was calculated by summing the soil layers in each profile for each N species. The random forest predictions are used only to present the modeled estimates of soil inorganic N in the soil profile.

Reference evapotranspiration (ET_o) was calculated using the method described in the FAO Factsheet no.  56 for daily estimations (Allen, Pereira, Raes, & Smith, 1998); weather data were obtained from NOAA Earth Systems Research Laboratory SurfRad, USDA NRCS National Water and Climate Center, and NASA POWER (NASA Langley Research Center, 2019; National Water & Climate Center et al., 2019; NOAA, 2020).

3 RESULTS

3.1 Biomass

Climatically, the 3 yr encompassed a dry year (2016, 279 mm precipitation in the growing season), a year with timely precipitation (2017, 445 mm), and a year with above average precipitation (2018, 685 mm; Supplemental Table S3; NASA Langley Research Center, 2019; National Water & Climate Center et al., 2019)NASA Langley Research Center, 2019; National Water & Climate Center et al., 2019). Only 2016 had greater reference evapotranspiration than precipitation in the growing season (Supplemental Table S3, cumulative from January). Accordingly, the average biomass yield of maize was 1.9, 2.3, and 1.4 kg m^–2 for 2016, 2017, and 2018, respectively.

Maize biomass yield tended to be greater or equal to yields of polycultures in all years (Figure 1, Table 2). For this reason, and while there were differences between the two maize hybrids, we treat maize A and maize B as a single entity. Overall, maize and sorghum monocultures and maize + sorghum polycultures produced the same biomass (P < .05). Soy monocultures yielded significantly less biomass than all other treatments. In the driest year (2016), we found greater differences in yield between treatments than in the two wetter years (2017 and 2018). Replacing maize with soy reduced yields, while replacing maize with sorghum rendered yields comparable to those of maize monocultures (Table 2). Proportions of each species in the harvested biomass are reported in Supplemental Table S4.

FIGURE 1

Silage biomass harvested from seven experimental treatments in 2016, 2017, and 2018 in central Pennsylvania. Black dots indicate individual observations for each block. Interquartile range is represented by the orange boxes. The median of each treatment is represented by the red horizontal line within the interquartile range. Whiskers extend through the lower and upper quartiles. The horizontal line in each year represents the average biomass (kg m^–2) of all maize monoculture plots

TABLE 2. ANOVA and least squares means for biomass yield and biomass concentration of nitrogen (N_c), phosphorus (P_c), and potassium (K_c) of silage harvested in central Pennsylvania in 2016, 2017, and 2018. Within year and variable, different letters indicate significant differences at P < .05 according to Tukey groupings

ANOVA	df	Biomass	NC	PC	KC
			P value
Year	2	<0.0001	.5000	<.0001	<.0001
Treat	6	<0.0001	<.0001	<.0001	<.0001
Year × treatment	12	0.0027	<.0001	.4188	.3130
Error terms		kg m^–2	g kg^–1
Block(year)	12	0.37	3.7	0.3	1.6
Residual	70	0.35	2.6	0.3	2.0
Least square means
2016
Maize A		1.92a	10.7c	1.7bc	6.3bc
Maize B		1.93a	10.3c	1.8bc	5.9c
Sorghum		1.70ab	11.0c	1.6c	8.7ab
Soy		0.75c	30.0a	2.5a	11.0a
Maize + sorghum		1.87a	10.4c	1.5c	8.7ab
Maize + soy		1.31b	14.8b	2.1b	6.9bc
Medley		1.42b	12.6bc	1.8bc	9.7a
2017
Maize A		2.12a	10.4b	2.3bc	8.3d
Maize B		2.43a	11.7b	2.2bc	9.8cd
Sorghum		2.01a	13.0b	1.9c	13.9ab
Soy		0.22b	24.7a	3.0a	14.7a
Maize + sorghum		2.03a	12.1b	2.5ab	11.8bc
Maize + soy		2.11a	11.9b	2.3b	8.9d
Medley		2.13a	11.1b	2.4b	10.6cd
2018
Maize A		1.45a	10.5cd	2.2b	9.0d
Maize B		1.38a	11.2bcd	2.3b	9.8cd
Sorghum		1.20a	10.2cd	2.2b	15.0a
Soy		0.62b	35.2a	3.0a	14.2ab
Maize + sorghum		1.41a	8.9d	2.2b	10.1cd
Maize + soy		1.28a	12.9bc	2.4b	9.5d
Medley		1.24a	14.0b	2.6ab	12.1bc

Harvest index and fraction of maize biomass significantly varied by both year and treatment (P < .0001; Table 3 and Supplemental Figure S2) and the year × treatment interaction indicates that treatment responses varied across years. When plots received precipitation below that of the ET_o (2016), maize in treatments that were majority maize had a greater HI than in treatments with large proportions of other species (approximately 0.55 vs. 0.44 kg kg^–1). However, when timely precipitation occurred (2017), maize HI was not affected by treatment (a mean HI of 0.31 kg kg^–1, P = .05). The larger HI in 2016 compared to 2017 reflects a later silage harvest in 2016 where maize was closer to physiological maturity. In the year with precipitation that well exceeded ET_o (2018), the drought-tolerant maize A had a greater HI than the non-drought-tolerant maize B and medley (0.48 vs. 0.43 kg kg^–1). With some variation, the fraction of maize in the biomass in polycultures was preserved across years—medley with the smallest, maize + soy with the largest, and maize +sorghum in between.

TABLE 3. Harvest indices (HI) of maize and fraction of biomass that is maize (F_maize) on a mass basis of silage harvested in central Pennsylvania in 2016, 2017, and 2018. Within year and variable, different letters mean significant differences at P < .05 according to Tukey groupings

ANOVA	df	Fmaize	HI
		P value
Year	2	<.0001	<.0001
Treat	6	<.0001	<.0001
Year × treatment	12	<.0001	<.0001
Error terms		kg kg^–1
Block(year)	12	0.08	0.05
Residual	72	0.09	0.04
Least square means
2016
Maize A		1.00a	0.55a
Maize B		1.00a	0.54a
Sorghum		—	—
Soy		—	—
Maize + sorghum		0.39c	0.45b
Maize + soy		0.77b	0.55a
Medley		0.25d	0.42b
2017
Maize A		1.00a	0.32
Maize B		1.00a	0.30
Sorghum		—	—
Soy		—	—
Maize + sorghum		0.73b	0.30
Maize + soy		1.00a	0.33
Medley		0.72b	0.32
2018
Maize A		1.00a	0.48a
Maize B		1.00a	0.43b
Sorghum		—	—
Soy		—	—
Maize + sorghum		0.66b	0.45ab
Maize + soy		0.89a	0.46ab
Medley		0.48c	0.43b

3.2 Plant nutrient analysis

On average, across years and compared with maize monocultures, polycultures that included soy increased N_c (10.8 vs. 13.2 and 12.6 g kg^–1 for the maize + soy and medley) (Figure 2, Table 2), and polycultures that included sorghum increased K_c (K_c, 8.2 vs. 10.2 and 10.8 g kg^–1 for maize + sorghum and medley) (Table 2). Additionally, the same N_M was harvested from all stands (approximately 20 g m^–2) (Table 4), despite treatments having different biomass yield. Soy monocultures had lower biomass in 2016 and 2018 compared to other treatments, but still had a N_M of approximately 20 g m^–2. On average, adding sorghum increased the harvested K_M from 15.3 g m^–2 in maize to >17 g m^–2 when including sorghum.

FIGURE 2

Silage nitrogen concentration (N_C, g kg^–1) in polycultures and monocultures with no soy, polycultures with soy, and soy monocultures harvested in central Pennsylvania in 2016, 2017, and 2018. If soy was not present in the harvested biomass, the data point belonged to the polycultures or monocultures with no soy category, regardless of whether soy was planted in the polyculture. Black dots indicate individual observations for each block within year. Interquartile range is represented by the orange boxes. The median of each treatment is represented by the red horizontal line within the interquartile range. Whiskers extend through the lower and upper quartiles

TABLE 4. Least squares means estimates of the mass of N, P, and K (N_M, P_M, and K_M) in harvested biomass of silage harvested in central Pennsylvania in 2016, 2017, and 2018. Within year and variable, different letters indicate significant differences at P < .05 according to Tukey groupings

ANOVA	df	NM	PM	KM
		P value
Year	2	.0086	<.0001	<.0001
Treat	6	.2712	<.0001	<.0001
Year × treatment	12	<.0001	<.0001	.0020
Error terms		g m^–2
Block(year)	12	7.01	0.76	4.20
Residual	70	4.82	0.73	3.89
Least squares means
2016
Maize A		20.6	3.4a	12.2abc
Maize B		19.9	3.5a	11.5abc
Sorghum		18.7	2.7ab	14.8a
Soy		22.5	1.9b	8.2c
Maize + sorghum		19.5	2.9a	15.9a
Maize + soy		19.1	2.7ab	9.1bc
Medley		17.7	2.6ab	13.7ab
2017
Maize A		22.7b	4.7a	17.7c
Maize B		28.8a	5.3a	23.8b
Sorghum		25.1ab	3.6b	26.7b
Soy		6.2 c	0.6c	37.7a
Maize + sorghum		24.1ab	5.2a	24.0b
Maize + soy		25.3ab	4.8a	18.8c
Medley		23.9ab	4.9a	22.1cb
2018
Maize A		15.4b	3.2a	13.0abc
Maize B		15.3b	3.2a	13.4abc
Sorghum		12.2b	2.7ab	17.8a
Soy		21.6a	1.8b	8.8c
Maize + sorghum		12.7b	3.1a	14.2ab
Maize + soy		16.7ab	3.0a	12.1bc
Medley		17.2ab	3.1a	15.2ab

In all treatments, the concentration of N is between 8.9 and 35.2 g kg^–1 (Table 2), with lower concentrations corresponding to maize and sorghum monocultures. Both are outside the recommended range of 17.3–31.2 g kg^–1 (Sutton & Lander, 2003). The concentration of P and K across all years and treatments (Table 2) were within or slightly above the nutrient recommendations for dairy cattle, which are 2.3 to 4.2 and 4.8 to 12.4 g kg^–1 for P and K, respectively (Sutton & Lander, 2003). Soy and sorghum monocultures exceeded K recommendations in years with adequate moisture (2017 and 2018). The interaction year by treatment for nutrient concentration and mass are displayed in Figure S4.

3.3 Water relations

In 2016, neither treatment nor species influenced Ψ (P > .05, Table 5), indicating that a species (e.g., maize) had a similar Ψ in all treatments sampled (e.g., maize monoculture, maize + sorghum, and medley) and that all species had Ψ similar to each other. The only minor exceptions were the Ψ_S of YMR and the Ψ_L of maize in the 27 July reading. These readings were 280 and 100 J kg^–1 greater (less negative, or wetter) than the average of the other three species, respectively. After a period of drying, the 10 August readings were similar among species, all within 160 and 50 J kg^–1 of each other for Ψ_S and Ψ_L, respectively. Somewhat similar results were obtained in 2017. On 8 Aug. 2017, maize had a Ψ_S lower than both sorghum cultivars; however, this was not the case in the 15 August reading and all species had a similar Ψ_L. In BMR on 15 August, the Ψ_X at noon was slightly greater than the Ψ_S, which means that there was either an error reading the pressure gauge, fast root advance through the soil profile with roots reaching wetter portions of the soil profile between morning and noon, or roots with very little resistance to water flow. The fraction of resistance to water movement in the root (PR_root, Table 5) appeared to be higher when plants were less stressed (less negative Ψ_L).

TABLE 5. Means of the water potential, analyzed by date within species, from silage grown in central Pennsylvania in 2016 and 2017. P value < .05 indicates significant differences between readings within each row

Date	Species	Treatment	ΨS	ΨX	ΨL	P value	PRroot
			J kg^–1
27 July 2016	BMR	ns	–1,180	na	–1,700	<.0001	na
ETo = 4.50 mm	YMR	na	–850	na	–1,720	.0189	na
	Maize	ns	–1,050	na	–1,510	<.0001	na
	Soy	ns	–1,170	na	–1,700	.0001	na
10 Aug. 2016	BMR	ns	–1,220	na	–1,560	<.0001	na
ETo = 4.15 mm	YMR	ns	–1,160	na	–1,540	.0338	na
	Maize	ns	–1,060	na	–1,570	<.0001	na
	Soy	ns	–1,140	na	–1,520	<.0001	na
8 Aug. 2017	BMR	na	–280	–550	–790	.0124	0.53
ETo = 4.19 mm	YMR	na	–220	–400	–870	.0002	0.28
	Maize	ns	–570	–700	–760	.0680	0.45
15 Aug. 2017	BMR	na	–630	–600	–1,030	.0274	—
ETo = 4.01 mm	YMR	na	–570	–650	–980	.0679	0.20
	Maize	ns	–590	–760	–1,170	<.0001	0.29

• Note. PR_root, proportion of resistance in the root; BMR, brown midrib; ns, not significant (P = .05); na, not available; ET_o, reference evapotranspiration; YMR, yellow midrib.

3.4 Soil nitrogen

Soil mineral N (NO₃^– and NH₄⁺) concentration was influenced the most by depth and horizon, followed by row and column (the position of each plot in a grid), and year, respectively (Supplemental Table S5). In general, mineral N decreased with depth, though treatment had a moderate effect on the mineral N left in the soil. Averaged across 3 yr, residual mineral N in the soil tended to be greater in treatments that contained soy (soy, maize + soy, and medley, 12.4 g m^–2) and soy monocultures (12.7 g m^–2) compared to treatments with maize or sorghum (maize, sorghum, maize + sorghum, medley, 10.9 g m^–2; Table 6). Averages for mass of soil NO₃^– and NH₄⁺ by year and treatment were calculated and reported in Table 6; soil NO₃^– and NH₄⁺ by depth are shown in Supplemental Table S6 and Supplemental Figure S3.

TABLE 6. Averages of nitrate and ammonium by treatment for 0.9 m of soil. Results obtained with random forest regression based on end-of-season soil samples taken in silage plots grown in central Pennsylvania in 2016, 2017, and 2018

Year and treatment	NO3^–	NH4^+	Total
	g m^–2
2016
Maize A	6.9	5.3	12.1
Maize B	10.2	6.1	16.3
Sorghum	6.2	5.9	12.1
Soy	9.8	6.1	15.9
Maize + sorghum	7.0	6.8	13.8
Maize + soy	9.5	8.0	17.5
Medley	6.7	5.7	12.4
2017
Maize A	5.2	5.5	10.7
Maize B	5.6	4.8	10.4
Sorghum	4.2	5.7	10.0
Soy	4.9	5.5	10.3
Maize + sorghum	6.0	5.1	11.1
Maize + soy	6.9	5.4	12.3
Medley	7.8	5.7	13.5
2018
Maize A	4.7	4.0	8.7
Maize B	4.1	3.9	8.0
Sorghum	4.0	4.2	8.3
Soy	8.2	3.8	11.9
Maize + sorghum	3.9	4.3	8.2
Maize + soy	6.1	3.8	9.8
Medley	4.3	4.1	8.4

4 DISCUSSION

While the results of this study are not unequivocal, they are encouraging and suggest that the polycultures tested in this experiment can be beneficial. The important considerations in selecting the proper mixture of species are high total biomass, adequate nutritional content, the ability to withstand abiotic stress, and the ability to mitigate negative environmental impacts like nutrient pollution. Although in our research none of the tested mixtures met all these criteria, biomass yield, nutritional content, and the amount of N left in the ground post-harvest could be manipulated towards these goals using sorghum and soy.

4.1 Biomass yield

Biomass yield for maize monocultures was greater or equal to that of the polycultures. Notably, when sorghum replaced maize, across all 3 yr the polyculture yield was similar to maize in monoculture (Table 2). We expected that the benefit of sorghum would emerge with water stress. If it did, it did so in modest terms, for drought was not intense enough in any year—not even in 2016, the driest year in our experiment—to test our first hypothesis. When soy replaced maize or any other species, it was unable to account for lost maize biomass and led to significantly less yield in the dry year. Results for all years are displayed graphically in Figure S5.

In terms of biomass production, maize + sorghum appeared to be the most promising of the tested polycultures. A similar result was found in a study where, compared to maize monoculture, a maize + sorghum mixture produced a similar forage yield and quality, but also had a lower environmental impact (Samarappuli & Berti, 2018). Because sorghum is more tolerant to drought than maize (Turner, 1974; Assefa, Staggenborg, & Prasad, 2010), maize + sorghum polycultures may withstand drought stress better than a pure maize monoculture (Klocke et al., 2011, 2012)—a proposition that we were unable to test under severe stress. Experiments with maize + sorghum polycultures should be conducted in environments that experience more intense drought than central Pennsylvania or in shallower soils. However, a record drought in western New York state in 2016 resulted in large yield losses in silage production (Sweet, Wolfe, DeGaetano, & Benner, 2017) and should serve as a warning to the region for the future. In the face of climate change, it would be wise to have a defensive strategy for silage production.

Even though sorghum proved able to retain high biomass levels, competition between plants and reduction of maize in the mixture may have negative effects on maize HI, and thus silage quality, for two reasons. First, kernels act as the main sink for N in mature maize. Though N is relocated from other parts of the plant (leaves, stems, and roots) during grain filling, up to 40% of N is derived from the soil and transferred to the kernel where N becomes more highly concentrated (Ta & Weiland, 1992). Reducing maize HI may lower overall N in the silage if the fewer kernels result in a smaller N sink. Second, the lesser amount of grain may decrease the concentration of easily decomposable sugars needed for fermentation. In the driest year, competition from sorghum in the polyculture significantly decreased maize HI (0.55 maize monoculture vs. 0.45 maize + sorghum and 0.42 medley) (Table 3). In addition, treatments dominated by maize biomass (maize monocultures and maize + soy) had a significantly higher HI compared to treatments in which biomass competition was greater (maize + sorghum and medley). Drought-tolerant maize in the extremely wet year (2018) had a higher HI compared to the traditional maize variety, a curious result indicating that drought-tolerant maize may provide benefits in years with excessive precipitation. Previous research on the HI is unclear about its response to plant population density, and research has not addressed maize HI with respect to community makeup that includes species with similar stature and growth habit to that of maize, like sorghum (Begna, Hamilton, Dwyer, Stewart, & Smith, 1997; Li et al., 2015). Our results indicate that the proportion of sorghum in maize + sorghum polycultures needs to be managed carefully, because in dry years overly competitive sorghum growth may negatively affect silage quality by reducing maize HI.

4.2 Plant nutrient analysis

Plant nutrient analysis was analyzed both as concentration and by mass. Concentration is indicative of nutritive quality of the silage and mass gives an estimate of nutrient removal from the soil. Like biomass, available N (a surrogate for protein), P, and K depended on species composition of the silage. Averaged across years and without considering soy monocultures, treatments had an N_M of ≈20 g m^–2 of N, with a range across years of 15–25 g m^–2 corresponding to 2018 and 2017, respectively (Table 4); that is, the lowest N_M occurred in the wettest year. Supporting our second hypothesis, the addition of soy led to a significant increase in N_C (P < .05), and therefore protein concentration, of the silage compared to treatments without soy (Table 2). However, when sunflower was in our most complex polyculture, it was not possible to disentangle the individual contributions of soy and sunflower to N_C.

These results corroborate previous research that found replacing rows of maize with soybean in silage increased the protein concentration by 11–51% compared to maize-only silage where greater row-replacement increased protein content more dramatically (Herbert, Putnam, Poos-Floyd, Vargas, & Creighton, 1984; Putnam, Herbert, & Vargas, 1986; Martin, Voldeng, & Smith, 1990). While our target was to replace approximately one-third of the maize plants with an equivalent relative number of soy plants (based on ideal soy planting populations) (Table 1), it is not clear from our research what the exact proportion of soybean in the mixture is that would retain yield while meaningfully changing N_C, though other research has suggested 10% of the silage as a minimum proportion (Toniolo, Sattin, & Mosca, 1987). It is clear, however, that increases in the proportion of soy can change the silage composition, increasing protein content and therefore milk production (Daniel, Friggens, Chapoutot, Van Laar, & Sauvant, 2016; Oldham, Broster, Napper, & Siviter, 1979).

Analogous to N_C, and across all 3 yr, treatments that contained soy also contained greater P_C (Table 2). Total P_M, however, was less consistent across years (Table 4), and no treatment in any year outperformed the P_M content of maize monocultures. Phosphorus uptake may have been slightly decreased by most of the polycultures we tested, but the chemical form of P in the biomass was likely different when soy and sunflower were part of the medley. Our analysis did not discriminate between P stored as phytate vs. other forms of P. In most seed, P is contained in phytate (myo-inositol hexakisphosphate) but the percentage of phytate varies (from smallest to largest) in cereals (e.g., maize), legumes (e.g., soy), and oil crops (e.g., sunflower; Cheryan & Rackis, 1980; Raboy et al., 2000; Reddy, Sathe, & Salunkhe, 1982). Phosphorus as phytate is less available for animals, but microbes in the rumen break the bonds in phytate, thus making P available for uptake (Yank, Bae, Selinger, & Cheng, 1998). Feed is often supplemented with either additional P to ensure adequate amounts of P, or with phytase to facilitate the breakdown of phytate already in the feed, thus increasing costs and making the total amount of phytate in silage an important quality factor (Harland,  1989; Nelson, Ferrara, & Storer, 1968; Yank, Bae, Selinger, & Cheng, 1998).

As for K_C and K_M, the most distinguishable pattern was that treatments containing sorghum and/or soy tended to have higher K_C (approximately 11.7 g kg^–1; Table 2) and K_M (approximately 18.3 g m^–2; Table 4) than maize monocultures or treatments in which a large amount of biomass was maize, like in maize + soy (approximately 8.3 g kg^–1 and 14.6 g m^–2, respectively). Dairy farms commonly return manure to fields, including a large return of K. Sorghum tends to take up more K than maize and preventing soil K deficits by monitoring K uptake and K return from manure becomes important when sorghum is part of the silage mix. However, excess K and N fertilization—especially K from manure where K levels may be higher—has been linked to high K levels in fodder and hypomagnesemia in cattle (Cherney, Mikhailova, & Cherney, 2002; Mayland & Grunes, 1979; Schonewille, 2013). Taken together, our results indicate that replacing maize with or adding sorghum and/or soy to maize silage may increase K_C while remaining below harmful thresholds.

4.3 Water relations

In addition to producing adequate yield with proper nutritional content, we aimed to develop polycultures that would produce well in environments that experience periodic droughts. In the driest year of our experiment (2016), maize showed clear signs of moisture stress (e.g., rolling leaves) throughout the season while sorghum did not. Moreover, the HI revealed that maize grown with sorghum was stressed during the reproductive phase. Nevertheless, total biomass showed that the water stress response of maize did not alter biomass productivity since maize was the most productive or tied with the most productive treatments each year.

Our results show that treatment did not control variation in plant water potential in 2016 (the dry year) or 2017. Furthermore, the similar Ψ_L for nearly all the readings suggests that all species experienced a similar level of water stress as measured by this variable (Table 5). Osmotic adjustment and perhaps changes in plant hydraulic conductance can make Ψ_L alone an unreliable measure of water stress, or at least one that would not always permit distinguishing responses among species at the levels of stress in this experiment.

Although sorghum has been reported to reach lower Ψ_L compared to maize (Turner, 1974), that difference was not evident in our measurements, despite the visible symptoms of water stress in maize (like leaf rolling) that were absent in sorghum. Sorghum is a drought-tolerant crop that can withstand arid conditions and produce greater biomass than maize under dry conditions (Beadle, Stevenson, Neumann, Thurtell, & King, 1973; Schittenhelm & Schroetter, 2014). Its drought tolerance is due to its ability to withstand low water potentials and ability to transpire under drier conditions (Wright, Smith, & Morgan, 1983), but this was not observed here even under the driest conditions. It should be considered that within species—including sorghum—some varieties may be more drought tolerant than others, or express different hydraulic behaviors (Salih et al., 1999). Thus, broad generalization about species behavior might be misleading and combining species might not always result in the desired mixture of physiological diversity.

4.4 Mineral soil nitrogen leftover

A complementary side of nutrient uptake is the amount that is left behind in the field and exposed to losses to the environment. End-of-season soil mineral N can provide insight into practices that help mitigate N pollution. In our experiment, we expected that soy may increase residual N, but that the use of polycultures may result in complementary uptake and ultimately reduce overall residual N. Depth and genetic horizon played the largest role in controlling residual mineral N in the soil (Supplemental Table S5) as determined by random forest, with higher concentrations of mineral N in the topsoil, consistent with previous research (Cassman & Munns,  1980).

Though the output presented here from random forest represent this model's prediction and not a statistical analysis of main effects, we note that across all 3 yr, polycultures that contained soy tended to leave more mineral N in the soil, while those that contained sorghum left less of it (Table 6). Moreover, more mineral N seem to have been left in the soil on average in the driest year (14.6 vs. 11.2 g m^–2 in 2016 and 2017, respectively) and compared to wetter years, NO₃^– may have made up a larger percentage of the mineral N leftover. In contrast, 2018 was wet and soils seemed depleted of mineral N (9.0 g m^–2), presumably due to mineral N lost to the environment. In maize, sorghum, and maize + sorghum treatments, average N extraction was approximately 20, 25, and 15 g m^–2 in 2016, 2017, and 2018 (Table 4). These numbers indicate that more N was removed with the harvest than what was placed as fertilizer each year, and yet soil analysis also indicates that a large amount of mineral N was left in the soil (Table 6). This mineral N is prone to losses over the winter (Bundy & Malone, 1988; Roth & Fox, 1990). Residual N from prior crops or the mineralization of N from organic matter during the growing season might have been substantial and led to the large amount of mineral N left in the soil.

To boost N content of silage or provide residual N for other crops, producers may add N₂-fixing crops like soy to silage, but this addition comes at an environmental cost. Our data indicate that polycultures with maize and sorghum may have prevented N losses more so than polycultures that included soy, and losses may have increased as the proportion of soy plants increased. One of the reasons is that the addition of soy may not be accompanied by a reduction in N fertilization, simply exacerbating N excess when it occurs. Further, research indicates N₂ fixed by legumes in a mixture may be transferred to grasses in perennial mixtures (Mallarino et al., 1990) but not in annual mixtures (Schipanski & Drinkwater, 2012), likely due to temporal differences in growth. Therefore, the addition of legumes in an annual silage mixture may increase nutritional content but lead to greater amounts of N left in the soil post-harvest—an environmental risk in wet years. Using cereal cover crops to reduce N leaching (Finney, White, & Kaye, 2016) may be a critical component of a management package for silage that includes soybean to help farm operators balance nutritional requirements of feed and environmental concerns.

Beyond studying specific crop species, further research should be done to determine if polycultures that utilize various functional traits (C₃, C₄, perennial, annual, grass, legume) may be calibrated to reduce on-farm N losses to the environment. Given our results, silage polycultures that included legumes may need to be coupled with other agricultural practices to minimize negative externalities. For example, manured systems that utilize heavy N applications may increase N concentrations, and thus protein content, of polyculture silage by adding sunflower rather than soy. This type of tailored choice of N supplementation may help increase N content but reduce residual N in the soil compared to using soy.

4.5 Other considerations

Producing polyculture silage required additional attention to management when compared to monocultures of maize silage. Planting mixtures of maize and other species entailed two passes over the field using specialized equipment like the InterSeeder, adding cost and potentially contributing to soil compaction. Ideally, planting would be reduced to a single pass over the field. Weed management and herbicide programs also became more complex as additional species were added, and especially when diversifying beyond monocots like in the medley, and when using sorghum which needed a safener. Producer adoption of silage polycultures will require training to ensure that biomass yield and nutrient requirements are met, and to ensure adequate field management and identifying and harvesting at optimal moisture content for ensiling. The increased complexity in management can only be justified by improved economic and environmental outcomes.

5 CONCLUSION

In this 3-yr experiment, we tested four monocultures and three polycultures of silage containing maize, soy, sorghum, and sunflower. We evaluated the ability of sorghum to help stabilize silage yield when the year-to-year water regime varies, and the ability of soy to increase N and P content in mixtures. Replacing maize with relatively low amounts of sorghum or other species (less than one-third replacement) produced yields comparable or slightly lower than those of maize monocultures. In the relatively dry year, there was no clear benefit to biomass yield with the addition of sorghum, although the composition of the yield varied—more grain with maize and more leafiness with sorghum. Replacing maize with soy reduced yield. However, adding even small quantities of soy increased N_C and P_C and adding sorghum increased K_C and K_M. With respect to soil N and on average, treatments with soy appeared to leave more mineral N in the ground while treatments with sorghum left less, indicating the potential for polycultures with sorghum to reduce pollution.

Depending on the producer's goal of either increasing biomass or altering nutritional composition, some silage polyculture mixtures are indeed beneficial medleys while others are gallimaufries that may not provide the needed production benefits. Maize + sorghum polycultures may be a viable choice for farm operators who want to produce biomass similar to that of conventional maize monocultures, whereas maize + soy polycultures may be viable for farmers who aim to increase silage N_C. Future research may test adding soy to maize rather than replacing maize, and replacement polycultures of maize + sorghum in more water-limiting environments.

Adding small amounts of sorghum, soy, and sunflower increases biodiversity without harming quantity and quality of the biomass. However, compared to monocultures weed management and planting is more complex due to a reduced number of herbicide choices and the need for specialized planting equipment like the InterSeeder. If the services provided by these polycultures can be monetized through green payments, then the expansion of polycultures for silage production in Pennsylvania and beyond can provide a steppingstone toward environmentally friendly and productive systems.

AUTHOR CONTRIBUTIONS

ARK, GWR, and JFT conceived the study; ABB, ARK, GWR, JFT designed the experiment; ABB conducted the experiments; ABB and ARK executed the data analysis; ABB performed the literature search; ABB and ARK defined the organizational structure after the literature search; ABB drafted the manuscript and prepared the figures; ARK, GWR, JB, and JFT contributed to the discussion and edited the manuscript. ARK, GWR, and JFT secured funding to support this research.

ACKNOWLEDGMENTS

This work was supported by USDA-NIFA Award no. 2016-67019-25209, the College of Agricultural Sciences at Penn State via the National Institute of Food and Agriculture and Hatch Appropriations under Projects no. PEN04571 and no. PEN04606 and Accessions no. 1003346 and no. 1009362, and the Annie's Sustainable Agriculture Scholarship Program. The authors thank the expert and invaluable support from Rodrigo Masip, Felipe Montes, and the personnel at the Russell E. Larson Agricultural Research Farm at Rock Springs.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Open Research

DATA AVAILABILITY STATEMENT

Data is available at https://doi.org/10.26208/5bhe-8c07.