Introduction

Increasing concern about depletion of non-renewable fossil fuels, and climate change driven by gas emissions has resulted in a high level of interest in renewable and sustainable energy resources such as bioethanol [1,2,3]. Bioethanol production in the USA has been primarily starch-based using corn (Zea mays) as a primary raw material [4,5,6].

Lignocellulosic biomass materials, such as agricultural residues remaining after grain harvest and dedicated energy crops (e.g., switchgrass (Panicum virgartum L.) and prairie cordgrass (Spartina pectinata Link.)), are abundant organic resources for biofuel production. Perennial energy crops have unique advantages in that they have greater environmental sustainability and are renewable [7, 8]. Prairie cordgrass has been shown to have high potential as a biomass crop in the northern Great Plains [9,10,11].

Prairie cordgrass is a perennial warm-season grass that spreads both vegetatively through rhizomes and non-vegetatively through seed. It is native to the North American Prairie and is found in most of the USA except for California, Nevada, and Arizona in the southwest and Louisiana to South Carolina in the southeast [12]. Prairie cordgrass is valued for stabilizing soil, preventing erosion on spillways and drainage channels, and revegetating wetlands. It also provides good cover for wildlife because of its thick stands, and can be used for hay for livestock if harvested when relatively immature [12]. Prairie cordgrass is a facultative wetland species, predominantly found in low floodplains and wetlands [13] where it occupies soils too wet and inadequately aerated for other species such as big bluestem (Andropogon gerardii Vitman) and switchgrass [12]. Studies have documented its potential for biomass production in short-season areas in Europe [14], southwestern Quebec [15], eastern South Dakota [9], and Kansas [11]. On agricultural soils in Minnesota, prairie cordgrass had consistently greater long-term yields than miscanthus (Miscanthus × giganteus), switchgrass, or big bluestem [16]. Prairie cordgrass has also been shown to have high tolerance for soil salinity in Canada [17] and the USA [18].

Proper N management is crucial in achieving economically viable biomass yields, reducing environmental contamination from N leaching, and maintaining stands of perennial biofuel crops [19]. For example, optimum biomass yield of switchgrass was reached with 140 kg N ha-1 in Iowa [20], but biomass yield response to N application varied with initial soil N concentration among several locations in the USA [19], and N application was not always economically viable at these same sites [21]. In Illinois, Guo et al. [22] found a yield response of prairie cordgrass to a N rate up to 84 kg N ha-1, but no additional response occurred above this rate. Otherwise, there is little information regarding biomass yield response of prairie cordgrass to N fertilization, especially across multiple diverse environments.

For successful biomass production, one of the critical phases is initial plant density. Prairie cordgrass can be propagated by seed or vegetatively by rhizomatous growth [10]. Due to significant rhizome production, it is possible that prairie cordgrass would not need high initial plant density for biomass production. This is economically advantageous considering the generally high cost of prairie cordgrass seed. For Miscanthus × giganteus, Lewanowski et al. [23] noted a slightly greater yield in the first 2–5 years when rhizomes were planted at higher densities (> 2 m-2), but this yield increase did not compensate for the increased establishment costs associated with the higher initial plant densities. Also, high initial plant density has been shown to increase tiller mortality because of competition for nutrients [24, 25]. Therefore, information on optimal planting density for any species, including prairie cordgrass, has both economic and agronomic ramifications [26].

Kura clover (Trifolium ambiguum M. Bieb) is a rhizomatous perennial legume that is a potential candidate for inclusion in mixtures with perennial grasses grown as bioenergy crops. It can withstand seasonal flooding; thus, it may grow well in areas where prairie cordgrass is also adapted. It has persisted for over 20 years in long-term grazing trials in Wisconsin and Minnesota [27]. It tolerates frequent defoliation in monoculture or in a mixture with perennial grass under continuous grazing [28, 29].

As an alternative to synthetic N fertilizer, biologically fixed N has economic and environmental advantages [30, 31]. When grown in mixture with grasses, N fixed by the legume can be utilized by grasses through direct transfer via mycorrhizal hyphae from legume to grass, degradation and decay of dead legume tissues, or nitrogen exudation by legume roots [32, 33]. Several studies have documented improvements in both forage yield and quality in grass-legume mixtures compared with monoculture grass swards [34,35,36]. Zemenchik et al. [36] estimated the fertilizer nitrogen replacement value (FNRV) of kura clover grown with orchardgrass (Dactylis glomerata L.) or smooth bromegrass (Bromus inermis L.) to range from 74 to 325 kg N ha-1. Barnett and Posler [34] also found yield advantages when various legumes were included in binary mixtures with diverse perennial grasses. However, there is no information regarding the use of kura clover, particularly as a source of N, with prairie cordgrass for sustainable biomass production.

Our hypotheses were that N fertilizer would increase biomass production in monoculture prairie cordgrass stands and that kura clover would provide a N benefit when grown in mixture with prairie cordgrass. Based on the rhizomatous growth of prairie cordgrass, we also hypothesized that initial plant density would not have long-term effects on biomass production. Therefore, the objectives of this study were to determine (1) N rate and plant spacing effects on yield and yield components of monoculture prairie cordgrass swards and (2) to determine the fertilizer N replacement value of kura clover in prairie cordgrass/kura clover binary mixtures.

Materials and Methods

Location Description

This study was conducted from 2010 to 2013 at four locations in the USA: Arlington, WI; Rosemount, MN; Brookings, SD; and Urbana, IL. Specific information (GPS coordinates and soil type) for each location is described in Table 1. Plots were established on land at each location that was either poorly drained or commonly flooded for some period during spring.

Table 1 Location and soil type for each site in the USA where prairie cordgrass was grown alone or in a mixture with kura clover
Full size table

Field Management

Prairie cordgrass was grown in monoculture or in a binary mixture with kura clover. Prairie cordgrass germplasm was from a native population collected in South Dakota and the kura clover population was developed by AgResearch New Zealand. Individual seedlings of both species were grown in conetainers (Stuewe, Inc., Corvallis, OR) during late winter-early spring 2010 in the greenhouse and then transplanted to the field in late spring 2010. Kura clover was inoculated with R. leguminosarum biovar trifolii strains 162C11, 162C13, and 162C14 mixture from Liphatech (Liphatech, INC. Milwaukee, WI) in the greenhouse. Kura clover was transplanted on 30-cm centers (30 cm between and within rows for a total density of 111,111 kura clover plants ha-1) in the field in the prairie cordgrass-kura clover mixture treatments. Prairie cordgrass seedlings were transplanted within the kura clover on 60- or 90-cm centers (populations of 26,896 and 11,881 plants ha-1, respectively) to obtain initial prairie cordgrass stand percentages of approximately 25 and 10%, respectively. Monoculture prairie cordgrass stands were also established at the same population densities as mixed stands. Four levels of N fertilizer (0, 75, 150 and 225 kg N ha-1) were applied as urea to the prairie cordgrass monocultures once annually in May from 2011 to 2013 at all locations. Individual plots were 3.0 m wide and 5.7 m long. This size allowed us to overcome potential border effects since a strip (approximately 1 m wide) was harvested through the center of the length of each plot for prairie cordgrass yield determination. A broadleaf herbicide, 2, 4-D (2, 4-dichlorophenoxy acetic acid), was applied at a rate of 0.6 kg a.i. ha-1 to suppress the kura clover in WI and MN in 2011, 2012, and 2013 and in SD and IL in 2012 and 2013. Herbicide was sprayed across the entire experimental area, including monoculture prairie cordgrass plots, to maintain consistency of treatments.

Biomass Yield

Prairie cordgrass yield (Mg ha-1) was determined at all locations from 2011 to 2013. A strip (120 cm wide for 60-cm row spacing treatment or 180 cm wide for 90-cm row spacing treatment) from the middle of each plot was harvested with a sickle-bar mower at a stubble height of about 10 cm once annually in late autumn after a killing frost and significant senescence had occurred. Little to no kura clover was present in the harvested biomass because most kura clover had fallen below the cutting height after a killing frost. For this reason, reported yields include only prairie cordgrass. Fresh samples were harvested and weighed in the field and then dried in a forced air oven at 60 °C for 72 h and reweighed for dry matter determination.

Prairie Cordgrass Yield Components

Prairie cordgrass tillers were collected from each plot before harvest for yield component analysis (i.e., tiller density and tiller mass). All prairie cordgrass tillers were hand-harvested to a stubble height of about 10 cm using a rice knife from two 0.36 m2 quadrats in the 60-cm row spacing treatment and two 0.81 m2 quadrats in 90-cm row spacing treatment. The different quadrat sizes allowed sampling from the entire width of the middle row of each plot. Collected tillers were dried at 60 °C for 72 h in a forced air oven and then the number of tillers was counted and weighed to determine tiller density and tiller mass.

Nitrogen Use Efficiency

Nitrogen use efficiency (NUE) represents yield per unit of applied N and was calculated following Zemenchik and Albrecht [37]:

$$ \mathrm{NUE}=\frac{\mathrm{N}-\mathrm{fertilized}\ \mathrm{plot}\ \mathrm{yield}\hbox{--} \mathrm{unfertilized}\ \mathrm{plot}\ \mathrm{yield}}{\mathrm{N}-\mathrm{fertilization}\ \mathrm{rate}} $$

Fertilizer N Replacement Value of Kura Clover

Second-order linear regressions were fitted for prairie cordgrass yield on fertilizer N rate in 2013 at each location before treatment means were calculated [36]. The regression procedure of Statistix 9 (Analytical Software, Tallahassee, FL) was used for regression analysis. Mean yield of prairie cordgrass in the binary mixture treatments for 2013 were substituted for Y in the regression model and then solved for X to obtain the FRNV of kura clover.

Statistical Analysis

Analyses of variance were performed to evaluate the effect of N rate, plant spacing, and all interactions on biomass production (balanced data) and yield components (unbalanced data for tiller mass and tiller density) of the prairie cordgrass monocultures using the linear models procedure in Statistix 9 (Analytical Software, Tallahassee, FL). Fisher’s least significant difference (LSD) was used to separate treatment means at P < 0.05. Nitrogen rate, plant spacing, and year were considered to be fixed effects. Locations were analyzed separately due to significant location × treatment interactions.

Results

Weather Conditions

Weather conditions during the growing season varied by year and location (Tables 2 and 3). Generally, precipitation was below long-term averages from 2011 to 2013 at all locations except for WI in 2013. In 2012, a widespread drought affected much of the central US, and precipitation from May to July of 2012 was often significantly below the 30-year average at the four locations. Particularly dry months in 2012 included June (6.0 mm) in WI, July (31.0 mm) in SD, June and July (33.0 and 17.8 mm, respectively) in MN, and July (14.2 mm) in IL. Average monthly temperatures approached the 30-year mean in most years with the notable exception of higher temperatures in July 2012 in SD and May and July 2012 in IL.

Table 2 Growing season (April through October) precipitation and the 30-year mean in 2011, 2012, and 2013 at Arlington, WI (WI), Brookings, SD (SD), Rosemont, MN (MN), and Champaign, IL (IL) where prairie cordgrass was grown alone or in a binary mixture with kura clover
Full size table
Table 3 Average monthly temperature and the 30-year mean in 2011, 2012, and 2013 at Arlington, WI (WI), Brookings, SD (SD), Rosemont, MN (MN), and Champaign, IL (IL) where prairie cordgrass was grown alone or in a binary mixture with kura clover
Full size table

Biomass Yield

Averaged across years, N rate, and plant spacing, prairie cordgrass yields varied from 6.4 Mg ha-1 in IL to 16.6 Mg ha-1 in MN (Table 4). Differences in prairie cordgrass yield occurred among years and N rates at all locations (Table 5, Fig. 1), and there was a N × year interaction in SD and IL and a plant spacing × year interaction at all locations. Biomass yields were 17.8, 13.2, 11.3, and 6.0 Mg ha-1 in WI, MN, SD, and IL, respectively in 2011. Yields of prairie cordgrass were lowest at all locations in 2012 due to abnormally low precipitation, but then increased in 2013 after the drought of 2012 (Fig. 1).

Table 4 Mean monoculture prairie cordgrass yields at Arlington, WI (WI), Brookings, SD (SD), Rosemont, MN (MN), and Champaign, IL (IL). Data are averaged across years, plant spacing, and N rate. Standard error of the mean for each location is in parentheses
Full size table
Table 5 Level of significance of F tests from model I analysis of variance for biomass yield of prairie cordgrass at four locations (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL) in the USA
Full size table
Fig. 1
figure 1

The effect of year on biomass yield of prairie cordgrass at four locations in the USA (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL). Data are averaged across plant spacing and N rate. Values within a location followed by different letters are significantly different at 0.05 level of probability

Full size image

A significant year × plant spacing interaction was observed for biomass yield of prairie cordgrass at each location (Table 5, Fig. 2). Biomass yield was higher for the 60-cm plant spacing than the 90-cm plant spacing at all locations in 2011. Prairie cordgrass with 60-cm plant spacing yielded 18.0, 39.3, 13.0, and 39.3% more compared with the 90-cm plant spacing in WI, MN, SD, and IL, respectively in that first year (Fig. 2). In 2012, plant spacing affected yield in SD only, and by 2013 plant spacing had no effect on biomass production.

Fig. 2
figure 2

Prairie cordgrass biomass production as influenced by the plant spacing × year interaction at four locations in the USA (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL). Data are averaged across N rate. Values within a location followed by different letters are significantly different at 0.05 level of probability

Full size image

Nitrogen rate consistently affected biomass yield of prairie cordgrass at all locations (Table 5, Fig. 3). Averaged across years and plant spacing, biomass yield was highest when 225 kg N ha-1 was applied in WI and SD, but yields peaked at the 150 kg N ha-1 rate in MN and the 75 kg N ha-1 rate in IL. Compared with the 0 N control, biomass yield increased by 8.5, 11.0, and 14.1% in WI, 3.0, 16.3, and 16.4% in MN, 35.6, 36.6, and 45.8% in SD and 30.1, 36.5, and 33.8% in IL when N was applied at rates of 75, 150, and 225 kg ha-1, respectively.

Fig. 3
figure 3

The effect of N rate on prairie cordgrass yield at four locations in the USA (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL). Data are averaged across year and plant spacing. Values within a location followed by different letters are significantly different at 0.05 level of probability

Full size image

Nitrogen Use Efficiency

Averaged across years and plant spacing, NUE decreased in response to increasing N rates (Table 6). NUE was highest at 75 kg N ha-1 and significantly decreased at the 150 and 225 kg N ha-1 rates in SD and IL. Numerically, the highest NUE for WI and MN was also at the 75 kg N ha-1; however, there was no statistical difference among the various N rates at either of these locations (Table 6).

Table 6 Nitrogen use efficiency (NUE) of prairie cordgrass to which four N rates had been applied over 3 years at four locations (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL) in the USA
Full size table

Averaged across N rate and year, there was no significant difference in NUE between 60- and 90-cm prairie cordgrass plant spacings, except for WI (Table 6) where the NUE in 90-cm plant spacing was higher than that in 60-cm plant spacing.

Averaged across N rates and plant spacing, NUE tended to increase over years at all locations (Table 6). The first year after establishment had the lowest NUE and the third year the highest NUE at all locations except for MN and IL which had comparable NUE in 2012 and 2013.

Yield Components

Tiller density of prairie cordgrass was negatively correlated with tiller mass at all locations (Fig. 4). Averaged across years and plant spacing, there were significant N rate effects on tiller density at SD and IL, but not at WI and MN, but there were also various 2-way interactions for these yield components (Table 7). Tiller density ranged from 110 to 134, 158 to 184, 169 to 216, and 131 to 160 tillers m-2 at WI, MN, SD, and IL, respectively (Table 8).

Fig. 4
figure 4

Relationship between tiller density (tiller m-2) and tiller mass (g tiller-1) in prairie cordgrass monocultures at four locations in the USA (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL). Data are averaged across years, plant spacing, and N rate

Full size image
Table 7 Level of significance of F tests from model I analysis of variance for tiller density and tiller mass of prairie cordgrass at four locations (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL) in the USA
Full size table
Table 8 Tiller density and tiller mass of prairie cordgrass to which four N rates had been applied over 3 years at four locations (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL) in the USA
Full size table

N rate affected tiller mass at all locations (Table 7). Tiller mass was greater in N-fertilized prairie cordgrass compared with the unfertilized control (Table 8). Across the 75, 150, and 225 kg N ha-1 rate, average tiller mass was 18 g at WI, 8 g at MN, 7 g at SD and 4 g at IL, while the average tiller mass of unfertilized prairie cordgrass was 14 g at WI, 6 g at MN, 5 g at SD, and 3 g at IL. There was no significant difference in tiller mass among the 75, 150, and 225 kg N ha-1 rates at any location.

Averaged across plant spacing and N rate, there was a significant effect of year on tiller density and tiller mass at all locations except for tiller mass at IL (Table 7). Tiller density and mass were not measured in 2013 in MN and in 2011 in IL. In MN, tiller density and tiller mass decreased significantly from 2011 to 2012. In IL, tiller density increased from 2012 to 2013 while tiller mass did not change (Table 8).

There was a significant plant spacing × year interaction on tiller mass and tiller density (Table 7). Overall, tiller density in 60-cm plant spacing was higher than that in 90-cm plant spacing in 2011 (Fig. 5). However, differences in tiller density between 60- and 90-cm plant spacing decreased over time at all locations except for SD where tiller density differences were noted again in 2013 between the 60- and 90-cm plant spacings. Although not consistent, tiller mass in 60-cm plant spacings tended to be lower than that in 90-cm plant spacing at all locations across the 3 years (Fig. 5).

Fig. 5
figure 5

Tiller density (tiller m-2) and tiller mass (g tiller-1) in prairie cordgrass monocultures as affected by the plant spacing x year interaction at four locations in the USA (Wisconsin, WI; Minnesota, MN; South Dakota, SD; and Illinois, IL). Data are averaged across N rates. Values within a location followed by different letters are significantly different at 0.05 level of probability

Full size image

Fertilizer Nitrogen Replacement Value of Kura Clover

Averaged across plant spacing, biomass yield of prairie cordgrass monocultures to which 0 kg N ha-1 was applied annually remained relatively similar over the 3 years of this study. In contrast, biomass yield of prairie cordgrass-kura clover mixtures generally increased over time at all locations (Fig. 6). In 2011, biomass yield of prairie cordgrass monocultures with 0 kg N ha-1 was greater than that of prairie cordgrass-kura clover mixtures in WI and IL, while yields for these comparable treatments were similar in SD and MN. In 2012, yield of the prairie cordgrass monoculture control (0 kg N ha-1) was similar to that of the prairie cordgrass-kura clover mixture at MN, SD, and IL, but not WI. In 2013, prairie cordgrass yielded more when mixed with kura clover than the 0 N control at all locations except WI (Fig. 6). The FNRV varied among locations in 2013, but was highest in MN (82 kg N ha-1) and lowest in WI (25 kg N ha-1) (Table 9).

Fig. 6
figure 6

Comparison of biomass yield of prairie cordgrass grown as monocultures with 0 kg N ha-1 (PCG + 0 kg N ha-1) and prairie cordgrass in binary mixture with kura clover (PCG + KC). Data are averaged across plant spacing. Values within a location and year followed by different letters are significantly different at 0.05 level of probability. ns*, not significantly different at 0.05 level of probability (p > 0.05)

Full size image
Table 9 Equations and coefficients of simple determination (R2) for prairie cordgrass yield response (Y) regressed on fertilizer N rate for monocultures of prairie cordgrass in 2013, and fertilizer N replacement values (FNRV) calculated from equations for kura clover grown with prairie cordgrass
Full size table

Discussion

Prairie cordgrass yield varied temporally at all locations, and differed among locations as well. Lowest overall annual yields occurred at the IL location (5–10 Mg ha-1) and the highest yields were observed at the WI location (14–20 Mg ha-1). Yields at MN and SD were intermediate to these other locations. Based on an evaluation of multiple switchgrass cultivars at several diverse environmental locations in the USA, Casler et al. [38] recommended that switchgrass not be moved more than one USDA hardiness zone north or south of their origin. However, they noted that moving cultivars across a longitudinal gradient may be possible if field tests demonstrate this type of adaptation. We suspect that the prairie cordgrass germplasm derived from SD was not well adapted in IL because of the latitudinal difference between the two locations. In addition, IL has since identified and collected IL-local prairie cordgrass germplasm that has yielded up to 20 Mg ha-1 (DoKyoung Lee, personal communication). Biomass yield in 2012 was significantly lower than in the other 2 years as a result of widespread drought throughout the USA, including the locations in this study. Despite dry conditions in 2012, prairie cordgrass yield increased in 2013 to levels above that seen in 2011, the year after establishment in the field. Perennial grass yield often increases during the first several years after establishment, and the extent to which increases occur is dependent on several factors including weather conditions, environment, and N and harvest management [26]. Thus, although our prairie cordgrass was transplanted, it was not unexpected to have prairie cordgrass yields be lower the first production year, and in the second production year because of dry conditions.

Year × plant spacing interactions affected prairie cordgrass yield at all locations (Table 5). Biomass yield differences in the first production year can likely be attributed primarily to initial plant density. Plots with narrower row spacing (60 cm) initially allowed for greater utilization of space and light based on the fact that we generally observed higher tiller density in these plots. Similarly, previous studies showed that biomass and essential oil yield of rose-scented geranium (Pelargonium species) was 132% and 99% higher for narrower spacing (60 × 30 cm) than that for wider spacing (120 × 30 cm) [39], and dry bean (Phaseolus vulgaris L.) cultivar yields were 52% higher on 250 mm compared to a 750-mm row width [40]. As we hypothesized, however, the effect of plant spacing on prairie cordgrass yield decreased over time, to the extent that no differences were noted in 2013. In addition, by the 2013 season, it was difficult to detect rows, or individual plants, in any of the plots, regardless of initial plant spacing. Prairie cordgrass is strongly rhizomatous [10], and these rhizomes give rise to new tillers that populated the inter-row and inter-plant areas over time. Initially, the wider row spacing had more inter-row and inter-plant area than the narrow plant spacing. Thus, the 90-cm plant spacing allowed for greater yield increases as tiller density increased while the high initial plant density in 60-cm plant spacing may have increased competition among plants for available nutrients. Mahmood and Honermeier [41] reported that row spacing of sorghum (Sorghum bicolor L. Meonch) had no clear effect on yield in the first experimental year, but in the second year narrower spacing yielded higher than wider row spacing. Also, other studies have demonstrated that higher initial plant density led to greater tiller mortality over time as a result of competition for nutrients and light [24, 25]. Because prairie cordgrass is a perennial, rhizomatous species, yield differences between the two plant spacings decreased over time until tiller density was similar for both the 60- and 90-cm plant spacings at three of the four locations (Fig. 5).

In previous studies in SD, the mean yield of different prairie cordgrass populations was 8.3 Mg ha-1 [9] while average yield of “Red River” and “Atkins” prairie cordgrass in Kansas was 9.5 Mg ha-1 [11]. We observed yields lower than, similar to, or higher than these at the four locations used in this study (Fig. 1).

Prairie cordgrass responded positively to N fertilizer at all locations (Fig. 3.). However, a previous 4-year study reported no yield response with N fertilization (N rates up to 168 kg ha-1) on “Red River” and “Atkins” prairie cordgrass in either South Dakota or Kansas [11]. Owens et al. [19] noted variable responses in switchgrass yield as related to N application at five locations across 2 years at diverse locations across the USA. They speculated that initial soil N concentration was at least partially responsible for the variable responses. In a follow-up to that study, that included an additional 5 years of yield data from the same locations, Lee et al. [42] found similar results with the exception that switchgrass at the Iowa site began to respond positively to N application.

Higher biomass production has often been achieved with higher N inputs. However, higher N does not always give higher NUE, and the maximum biomass produced per unit of N applied may be at lower N fertilizer rates [19]. Considering the environmental and economic impacts of N fertilizer application, determining the NUE is important when selecting biomass energy crops. Pedroso et al. [43] reported the maximum NUE was reached at the lowest N fertilizer level with switchgrass, miscanthus, and tall wheatgrass (Agropyron elongatum). Lemus et al. [44] also reported the NUE of switchgrass was highest at the lowest N rate (56 kg N ha-1) in their study. In our study, the lowest N fertilizer rate (75 kg N ha-1) had the highest NUE at all locations (Table 6).

Christian et al. [45] suggested that 1-year-old plants of switchgrass and coastal panic grass (Panicum amarum A.S. Hitchc. & Chase) for biomass production would not require a high level of N because it has relatively low biomass yield. However, more N may be necessary in succeeding years when above- and belowground mass of the crop is further developed allowing for improved nitrogen use efficiency [26]. Similar results were observed in our study when the NUE was higher for prairie cordgrass in the second or third growing season compared to the first year.

Significant N translocation to belowground storage organs occurs in perennial grasses during senescence [46,47,48], and this may contribute to higher N use efficiency. We suggest that prairie cordgrass has high N use efficiency because it is able to recycle nutrients via well-developed rhizomes. Nitrogen translocation from shoot to root biomass during senescence contributes to plant survival through the winter [49, 50], and this stored N can be re-translocated to aboveground tissues during the early growing season to help reduce the use of N fertilizer [43]. This may be particularly relevant when perennial species are harvested after a killing frost, thus maximizing N translocation to roots and rhizomes and minimizing N removal during biomass harvest, as in this study.

In our study, tiller density and tiller mass varied by N application rate, plant spacing, and year (Table 7). These variations may result from different environmental conditions among locations. Smart et al. [51] reported that switchgrass stands with low tiller number and high tiller weight yielded 25% more biomass than stands with high tiller number and low tiller weight. Similarly, in our study, WI had low tiller number, high tiller mass, and thus the highest yield. Tiller number of Miscanthus × giganteus in central Greece [52] and of two different prairie cordgrass varieties in South Dakota and Kansas [11] was not affected by N rate. Similar results in our study were found in WI and MN. However, N fertilization affected tiller density and tiller mass in SD (Table 8). Similarly, mean tiller density of tall fescue increased with N application [53], and there was high variation in tiller density of perennial ryegrass from year to year [54, 55]. Mahmood and Honermeier [41] reported that narrow row spacing caused a significant increase in tiller density compared with wide row spacing. Also, Snider et al. [56] reported that wider row spacing of sorghum produced the highest stem densities, and stem diameter declined as stem density increased at one of their experimental sites, similar to what we observed in this study (Fig. 4). Fraser and Kindscher [57] reported that stem density of transplanted prairie cordgrass plugs significantly changed over 3 years. Stem density in the small plug treatment decreased sharply while stem density in the large plug treatment increased, and at the end of 3 years the stem density between the different treatments appeared to converge.

Yield of prairie cordgrass mixed with kura clover increased annually, and by the third year had out-yielded the non-fertilized prairie cordgrass in the monoculture plots. Similarly, previous studies showed greater yields in pastures and grasslands composed of grass-legume mixtures [58,59,60,61]. As kura clover development increased over time, the transferable N from legume to grass may have also gradually increased or accumulated in the soil through degradation and decomposition of dead legume tissue [32, 62,63,64] or N exudation from the roots [33]. Nesheim and øyen [65] and Boller and Nösberger [62] reported that legumes grown in a mixture with grasses provide a large majority of their N through symbiotic N2 fixation, up to 300 kg N ha-1. Thus, N transfer from prairie cordgrass to kura clover may have boosted the prairie cordgrass yield in mixtures. Zemenchik et al. [36] estimated the FNRV of kura clover to range from 71 to 274 kg N ha-1 when grown with various cool season grasses for forage, and their results varied by location and year. In our study, the FNRV of kura clover ranged from 25 to 82 kg N ha-1 in 2013 (Table 9). This range of values is not surprising considering the variable environments utilized in this study. Regardless of the range of values, however, it demonstrates the potential benefit of the binary mixture of prairie cordgrass-kura clover.

Conclusions

Biomass of prairie cordgrass varied by location, ranging from 6.4 to 16.6 Mg ha−1, and yield was affected by N rate, year, and plant spacing. We also observed a significant influence from the interaction between year and plant spacing. Monoculture prairie cordgrass yield increased with N application; however, it had the greatest nitrogen use efficiency at the 75 kg N ha-1 rate and in later years when prairie cordgrass was well established. Prairie cordgrass tiller density tended to increase in response to increased N. Yield of prairie cordgrass with 0 N was equal to or less than the yield of prairie cordgrass grown with kura clover at all locations in 2011 and 2012; however, there was a positive N value associated with prairie cordgrass grown with kura clover in 2013. Kura clover has potential to provide N to prairie cordgrass in binary mixtures of these two species and on land that may not be easily farmed due to poor drainage or excessive wet periods.