Yield and nutritive value of grazed complex legume–grass mixtures under increasing nitrogen application rates

Abstract: Complex pasture mixtures are advantageous, but little information exists on the best forage species and nitrogen (N) management in eastern Canada. We compared under mob stocking four complex mixtures of one of two legume species [alfalfa (Medicago sativa L.) and birdsfoot trefoil (Lotus corniculatus L.)] plus one of two grass mixes [No. 1 — timothy (Phleum pratense L.), meadow fescue (Schedonorus pratensis (Huds.) P. Beauv.), reed canarygrass (Phalaris arundinacea L.), and Kentucky bluegrass (Poa pratensis L.); No. 2 — tall fescue (Schedonorus arundinaceus (Schreb.) Dumort., nom. cons.), meadow bromegrass (Bromus biebersteinii Roem. and Schult.), reed canarygrass, and Kentucky bluegrass] under three N application rates at three sites over the first three post-seeding years. Legume species had little effect on most forage attributes mainly because of their low contribution to forage dry matter (DM) yield (<30%) in second and third years. The grass mix with tall fescue and meadow bromegrass (No. 2) yielded similar or slightly better than the one with timothy and meadow fescue (No. 1) but tended to have a greater neutral detergent fibre concentration, and lower N and total digestible nutrient concentrations. Nitrogen fertilization increased forage DM yield only in second and third years when the legume contribution to forage DM yield was <30%, and it increased N concentration and decreased nonstructural carbohydrate concentration. These results highlight the challenge of maintaining legume species in rotationally grazed pastures of eastern Canada and confirm recommendations of applying no or little N fertilizer on grazed legume–grass mixtures when the legume contribution to forage DM yield is >30%.


Introduction
Ecological research suggests that plant biodiversity benefits ecosystem functions including primary productivity, stability of production, and resistance of weed invasions in natural grasslands (Sanderson 2005). Similar benefits have been observed in managed grasslands (Frankow-Linberg et al. 2009;Picasso et al. 2011;Finn et al. 2013;Sturludóttir et al. 2013). Consequently, pastures with increased plant species diversity could improve forage yield and yield stability. In studies conducted in Atlantic Canada, more complex mixtures were shown to result in more resilient pastures that were ultimately more productive (McKenzie et al. 2005;McElroy et al. 2012;Papadopoulos et al. 2012).
Recommended forage mixtures in eastern Canada usually contain a maximum of three species, including both grasses and legumes. Designing forage mixtures, particularly complex mixtures, is challenging because several grass and legume species are recommended in eastern Canada. New grass species are now recommended in eastern Canada, but their performance when grown with other grass species and a legume species under grazing has not yet been determined. Meadow fescue [Schedonorus pratensis (Huds.) P. Beauv.] is not widely used in eastern Canada, but it is often seeded with timothy (Phleum pratense L.) in Scandinavian countries (Virkajärvi 2004). Both timothy and meadow fescue are well adapted to growing conditions in eastern Canada, but they are not tolerant to frequent defoliation (Kunelius et al. 2003;Drapeau and Bélanger 2009), and timothy has poor regrowth potential. Tall fescue [Schedonorus arundinaceus (Schreb.) Dumort., nom. cons.] and meadow bromegrass (Bromus biebersteinii Roem. and Schult.) have excellent regrowth potential and good tolerance to frequent cutting (Bélanger et al. 2018).
Alfalfa (Medicago sativa L.) is a fast-growing legume species with a high yield potential, but it is sensitive to poor drainage, soil acidity, and winter conditions (Bélanger et al. 2006). As compared with alfalfa, birdsfoot trefoil (Lotus corniculatus L.) is a slow-growing species with a lower yield potential but with a greater tolerance to poor growing conditions (e.g., soil acidity and poor drainage). Both birdsfoot trefoil and a grazing-type alfalfa 'AAC Trueman' (previously referred to as CRS1001) performed well in binary mixtures with a grass species under grazing or frequent cutting (Bélanger et al. 2018).
Very few studies have looked at the fertilizer nitrogen (N) requirements of complex mixtures including at least one legume species. In a study conducted in Switzerland, complex legume-grass mixtures fertilized with 50 kg N ha −1 yr −1 produced yields comparable to grass monocultures fertilized with 450 kg N ha −1 yr −1 when the legume proportion was maintained between 50% and 70% (Nyfeler et al. 2009). The benefits of having several species were reduced at the highest N fertilization rate of 450 kg N ha −1 yr −1 with a general trend towards a grass-dominated sward. Current recommendations for N fertilization of legume-grass mixtures in eastern Canada consider the proportion of legumes in the sward, and no or little N fertilizer is recommended on grazed legume-grass mixtures as long as the legume contribution to forage dry matter (DM) yield is greater than between 30% and 50%. More specifically, in Ontario, 60 kg N ha −1 yr −1 are recommended when the sward has between 33% and 50% legumes, but no N fertilizer is recommended for swards containing more than 50% legumes (OMAFRA 2017). In Québec, rates of 75-160 kg N ha −1 yr −1 are recommended when the sward has less than 40% legumes, whereas 0-75 kg N ha −1 yr −1 are recommended when legumes make up more than 40% of the sward (CRAAQ 2010). In Nova Scotia, only 20 kg N ha −1 yr −1 are recommended in the spring when the sward has more than 30% legumes (PERENNIA 2018).
Legume-grass mixtures and N fertilization can also affect the forage nutritive value. The high concentration of crude protein in forage legumes and the fast degradation of proteins compared with the amount of fermentable organic matter available in the rumen may lead to inefficient utilization of N by ruminants, resulting in high N losses to the environment mainly under grazing (Kleen et al. 2011). Enhanced levels of readily available energy from forages with a high concentration of water-soluble carbohydrates (WSC) and (or) nonstructural carbohydrates (NSC) improved microbial N synthesis in the rumen (Berthiaume et al. 2010) and the efficiency of N utilization by dairy cows (Miller et al. 2001;Brito et al. 2009). As compared with legume monocultures, mixtures of legume species with grasses have been shown to increase the WSC concentration of forages .
This study focused on four complex mixtures that had previously performed well in a study with grazing or frequent cutting in eastern Canada (unpublished data). The four complex mixtures included grass species with either slow regrowth and (or) poor tolerance to frequent defoliation or fast regrowth and (or) good tolerance to frequent defoliation, along with legume species with either low yield potential and good tolerance to poor growing conditions or high yield potential and poor tolerance to poor growing conditions. The four complex mixtures combining two grass mixes and two legume species were studied under different rates of N fertilization with the assumption that the presence of the legume species will reduce N fertilizer requirements. The objective was to compare the forage yield and nutritive value of four complex pasture mixtures of one legume and four grasses under varying N fertilization during the first three post-seeding years.

Materials and Methods
This study was conducted at three sites in eastern Canada (Nappan, NS; Normandin, QC; New Liskeard, ON). The initial soil characteristics for the three sites are presented in Table 1. The soils at the three sites differed by their texture, pH, and available soil phosphorus and potassium.
Four complex mixtures were evaluated under three N rates (0, 60, and 120 kg N ha −1 yr −1 ) with half of the N applied in the spring and the second half applied after the second grazing. The N application rates were chosen to represent a wide range of N nutrition while ensuring that, in some treatments, the benefits of the presence of legumes were maximized. Ammonium nitrate was used at Nappan and New Liskeard, while calcium ammonium nitrate was used at Normandin; the N availability of the two forms of fertilizer is similar.
The four complex forage mixtures were a combination of a grazing-type alfalfa ('AAC Trueman'; 6 kg seeds ha −1 ) or birdsfoot trefoil ('AC Langille'; 6 kg seeds ha −1 ) with two four-grass species mixes based on either timothy (Tm, 'Express'; 4 kg seeds ha −1 ) and meadow fescue (Mf, 'Pradel'; 7 kg seeds ha −1 ) or tall fescue (Tf, 'Courtenay'; 6 kg seeds ha −1 ) and meadow bromegrass (Mb, 'Fleet'; 5 kg seeds ha −1 ). 'AAC Trueman' is characterized by its unique rhizomatous growth habit and is lateflowering, winter-hardy, and tolerant to midsummer drought, spring and fall water-logging, and frequent grazing. Reed canarygrass (Phalaris arundinacea L.) (Rc, 'Venture'; 2 kg seeds ha −1 ) and Kentucky bluegrass (Poa pratensis L.) (Kb, 'Troy'; 3 kg seeds ha −1 ) were included in both grass mixes. Their roles to enhance sward productivity were clearly demonstrated in a previous study . The two grass mixes were referred to in the tables as TmMfRcKb and TfMbRcKb. The four species of each grass mix were seeded in equal rates with an overall seeding rate of 1000 seeds m −2 .
The 12 treatments, within a factorial arrangement of 3 × 4, were replicated three times at each site within a split-plot design with three N application rates as main plots and four complex mixtures combining two legume species with two grass mixes as subplots. The N application rates were latinised so that not all N Values at the start of the experiment in the spring of the first post-seeding year prior to nitrogen (N) fertilization (0-15 cm). The soil pH was determined following the method of McKeague (1978) in a 1:1 soil:water ratio solution. Potassium (K) and phosphorus (P) were extracted following the Mehlich III method (Tran and Simard 1993) and determined by inductively coupled plasma optical emission spectroscopy (ICP-OES, optical emission spectrometer, model Optima 4300 DV, Perkin Elmer, Woodbridge, ON, Canada). application rates appear in one end of all blocks. Phosphorus and potassium were applied at the beginning of each growing season based on soil test and local recommendations. Plots were seeded in the spring of 2013. Dates of N fertilizer application, soil sampling, grazing events, and botanical separation are presented in Table 2.
Mob stocking, that is, grazing of plots with animals in 1 or 2 d, was used at the three sites. Mob stocking is a useful technique for evaluating new species and germplasms intended for grazing in small experimental plots (McCartney and Bittman 1994;Papadopoulos et al. 2012). The number of animals varied as a function of the forage DM yield at the time of grazing with the objective of grazing the plots in 1 or 2 d. Growing beef steers, beef cows with their calves, and heifers were used at Nappan, Normandin, and New Liskeard, respectively. Due to the plot size, fencing of the subplots was not practical, and all mixtures were grazed at the same time at each site when timothy reached about 33 cm in height.
Grazing events lasted a maximum of 2 d. The exit grazing height was 4-5 cm in the initial rotation grazing and 10 cm in subsequent grazing cycles. Mob stocking at a high stocking rate was used to help achieve quickly a uniform exit grazing height.
At each grazing event, DM yield was measured prior to grazing by sampling every plot at a 5 cm height with two quadrats of 0.25 m 2 in area. Forage samples were dried at 55°C for at least 72 h prior to recording the sample dry weight and then ground for the analysis of nutritive value. Seeded grasses, seeded legumes, and weeds were manually separated twice during the season generally at the first and third grazing, except at Nappan in 2016 when it was done once at the second grazing and at New Liskeard in 2015 when it was done at three grazing events (Table 2). One forage sample was taken at a 5 cm height in each plot from a 0.25 m 2 quadrat; each component was dried at 55°C in a force-draft oven for 72 h and weighed to determine their proportion assessed as their contribution to DM yield. Note: NA, information not available.
All forage samples collected from the 3 yr at the three sites (n = 991) were scanned by visible near-infrared reflectance spectroscopy (VNIRS) using a Foss NIR System 6500 monochromator instrument (Foss NIRSystems Inc., Silver Spring, MD, USA). From VNIRS sample scans, a set of 356 forage samples (285 for calibration and 71 for validation) was selected by the WinISI software version 4.5.0.1407 (Infrasoft International, LLC, Silver Spring, MD, USA). This set of forage samples were chemically analyzed for the concentrations of DM, ash, acid detergent fibre (ADF), neutral detergent fibre assayed with a heat-stable α-amylase (aNDF), N, neutral detergent insoluble crude protein (NDICP), ether extract (EE), WSC, and starch, as well as for the in vitro neutral detergent fibre digestibility (NDFd).
Dry matter and ash concentrations were determined by thermogravimetry (Leco Corporation 2009) using an auto-analyser (model TGA701, Leco Corporation, St. Joseph, MI, USA). Nitrogen concentration was measured using a method adapted from Isaac and Johnson (1976). Ground samples (100 mg) were digested for 60 min at 380°C in a 1.5 mL mixture of selenious and sulfuric acid plus 2 mL of 30% H 2 O 2 . After cooling, the mixture was diluted to 75 mL with deionized water. An auto-analyzer (QuikChem 8000 Lachat Zellweger Analytics Inc., Lachat Instruments, Milwaukee, WI, USA) was used to measure N with the method 13-107-06-2-D and P with the method 13-115-01-2-A (Lachat Instruments 2019; Zellweger Analytics Inc.). The ADF concentration was determined according to AOAC (1990). The aNDF was analyzed following Mertens (2002) with addition of a sodium sulfite. These fibre extractions were done using the Ankom filter bag technique (ANKOM Technology, Macedon, NY, USA). The concentration of NDICP was determined according to Licitra et al. (1996). The NDFd was measured using the method of Goering and Van Soest (1970) based on a 48 h incubation with buffered rumen fluid followed by an aNDF determination of the post-digestion residues. The rumen fluid incubation was performed with Ankom F57 filter bags and an Ankom Daisy II incubator, using the bath incubation procedures outlined by Ankom Technology Corp. Rumen fluid was obtained from a ruminally fistulated dairy cow that was offered a diet of 37% grass silage, 15% corn silage, 8% hay, 30% corn grain, and 10% concentrate mix formulated to meet the nutritional requirements of a lactating dairy cow expected to produce 10 200 kg milk yr −1 . The NDFd (g kg −1 aNDF) was calculated as below: NDFd = 1 − postdigestion dry weight following aNDF wash predigestion dry weight of aNDF × 1000 Crude fat (EE) was determined using Ankom xt15 Extractor Technology Method (American Oil Chemists' Society 2003). Concentrations of WSC and starch were measured according to dos Passos Bernardes et al. (2015), and the concentration of NSC was calculated as the sum of WSC and starch concentrations. Using concentrations of ash, N, ADF, aNDF, NDFd, NDICP, and EE, concentration of total digestible nutrients (TDN) was calculated based on NRC (2001) using the Excel spreadsheet Milk2013 of the University of Wisconsin Alfalfa/Grass Evaluation System (Undersander et al. 2013) for all chemically analyzed forage samples.
Results from chemical analyses and calculations for the calibration set of samples were then used to develop calibration equations using the WinISI version 4 software (Foss NIRSystems Inc., Silver Spring, MD, USA) for the following nutritive attributes: N, aNDF, NDFd, NSC, and TDN. Calibration equations were validated using the same software by comparing predicted values against reference laboratory values obtained for the 71 validation samples. The VNIRS predictions were considered excellent with the ratio of prediction to deviation (ratio of standard deviation of the reference data used in the validation set to standard error of prediction corrected for bias) values greater than 3, ranging from 3.0 to 6.6 among all nutritive attributes (Table 3). Using the modified PLSR method of the WinISI version 4 software, predicted values of each nutritive attribute were then generated for all forage samples.
Soil samples (0-15 cm) were taken in each plot prior to the N application in the first, second, and third post-seeding years except at Nappan in the first postseeding year when soil sampling was done 13 d after the N application (Table 2). A second set of soil samples were taken in late summer or early fall in the 3 yr. The soil N-NH 4 and N-NO 3 concentrations were determined by extraction with 1 mol L −1 KCl using a 1:25 material: extractant by weight ratio. Soil samples taken prior to the N application in the spring of the first post-seeding year were used to determine the initial values of soil mineral N (N-NH 4 and N-NO 3 ) at each site. In the second and third post-seeding years, however, soil sampling prior to the N application in spring reflected the residual soil mineral N (N-NH 4 + N-NO 3 ) in each of the plots following the fertilization treatments in the previous year.
Data were assessed by analyses of variance (ANOVA) using the GENSTAT version 17 statistical software (VSN International 2013). The data were analyzed for each site and post-seeding year with blocks, main plots, and subplots considered random effects, while N application rates, legume species, and grass mixes were considered fixed effects. To provide more general information, the data were also analyzed with combining the three sites in each post-seeding year with sites as a random factor, and with combining both sites and post-seeding years with sites as a random factor and post-seeding years as a repeated measure. Differences were considered significant when P < 0.05. Seasonal values of DM yield were calculated as the sum of the DM yield at each grazing. Seasonal values of the nutritive attributes were calculated as the average of their values at each grazing. Seasonal values were reported with the objective of presenting and discussing the overall response over the 3 yr of the study.

Results and Discussion
There were very few significant interactions between factors (N application rate, legume species, and grass mix) for seasonal DM yield and nutritive attributes. When they occurred, the interactions were more related to the size of the differences among treatments than the direction of the differences. Consequently, the main effects of any of the three factors were considered stable across the other factors at all sites and post-seeding years.

Seasonal forage DM yield
Increasing N application rates did not significantly increase seasonal forage DM yield at any of the three sites in the first post-seeding year but significantly increased the forage DM yield at New Liskeard in the second post-seeding year and at Normandin and New Liskeard in the third post-seeding year (Table 4). This positive response to N fertilization was quite limited in the second post-seeding year but was more important in the third post-seeding year. In the second post-seeding year at New Liskeard, forage DM yield with no applied N was 89% of the maximum forage DM yield that was obtained with 60 kg N ha −1 . With 120 kg ha −1 , N was in excess of the crop N demand as there was no increase in yield compared with 60 kg N ha −1 . In the third postseeding year, forage DM yield with no applied N was 69% of the maximum forage DM yield obtained with 120 kg N ha −1 at Normandin and 79% at New Liskeard. Although the response was not statistically significant at Nappan, forage DM yield with no applied N was around 85% of the maximum forage DM yield obtained with 60 kg N ha −1 in the second and third post-seeding years. Furthermore, there was no interaction between N application rate and any of the other two factors (legume species and grass mix), except at New Liskeard in the third post-seeding year where the response to N application rate was greater for the tall fescue-meadow bromegrass-based mixtures (7.4-10.0 Mg DM ha −1 ) than for the timothy-meadow fescue-based mixtures (6.3-7.1 Mg DM ha −1 ). Averaged across the three sites, applying 60 kg N ha −1 compared with no applied N significantly increased forage DM yield from 6.9 to 7.7 Mg ha −1 in the second post-seeding year and from 5.3 to 5.8 Mg ha −1 in the third post-seeding year.
The lack of N response in the first post-seeding year could be attributed to the significant contribution of the legume species to forage DM yield. In the first postseeding year, alfalfa contributed 38%-49% to forage DM yield, while birdsfoot trefoil contributed 18%-45% to forage DM yield (Table 5). In the second and third postseeding years, the contribution of the legume species to forage DM yield was less than 30%. The absence of a response to N fertilization in the first post-seeding year suggests that sufficient N was available to the forage mixtures, either from legume N fixation or soil mineral N. Legume N fixation was not measured, and although soil mineral N was measured in the spring of each growing season, it is a poor indicator of N availability for the growing season of forage grasses under the humid conditions of eastern Canada (Ziadi et al. 2000). Our results, however, confirm our hypothesis and results from other studies (Nyfeler et al. 2009) that the presence of legume species reduces N fertilizer requirements of complex forage mixtures.
The legume species did not affect seasonal forage DM yield in any of the post-seeding years at the three sites (Table 4). The grass mix did not significantly affect forage Note: n, number of samples in the validation set; SD, standard deviation; SEP, standard error of prediction; SEP(C), standard error of prediction corrected for the bias; RSQ, coefficient of determination for the prediction; RPD, ratio of prediction to deviation [SD/ SEP(C)]; aNDF, neutral detergent fibre assayed with a heat-stable α-amylase; DM, dry matter; NSC, nonstructural carbohydrates (water soluble carbohydrates plus starch); NDFd, in vitro aNDF digestibility; TDN, total digestible nutrients. DM yield in the first post-seeding year at Normandin and Nappan, but at New Liskeard the tall fescue-meadow bromegrass-based mixtures with either alfalfa or birdsfoot trefoil had a greater forage DM yield than the two timothy-meadow fescue-based mixtures (Table 4). The grass mix did not significantly affect forage DM yield at any of the sites in the second post-seeding year. In the third post-seeding year, the two tall fescue-meadow brome-based mixtures had a greater seasonal forage DM yield than the two timothy-meadow fescue-based mixtures at New Liskeard, but there was no significant difference at Nappan and Normandin. Averaged across sites and years, forage DM yield of the two tall fescue-meadow brome-based mixtures tended (P = 0.07) to be greater than that of the two timothy-meadow fescue-based mixtures (6.8 vs. 6.5 Mg ha −1 ).
The choice of grass mixes had a very limited effect on seasonal forage DM yield even though grass species are known to differ in their tolerance to frequent defoliations (Cullen et al. 2006;Drapeau and Bélanger 2009). The grass mix containing timothy and meadow fescue, known to be sensitive to frequent defoliations, performed nearly as well in terms of seasonal forage DM yield over three post-seeding years as the grass mix with tall fescue and meadow bromegrass, known for their greater tolerance to frequent defoliations. In a previous study conducted in eastern Canada, timothy, meadow fescue, tall fescue, and meadow bromegrass grown in binary mixtures with a legume species and grazed or frequently cut persisted well over five postseeding years (Bélanger et al. 2018). Our results confirm that four recommended forage grass species in eastern Canada, including timothy and meadow fescue, perform well under the type of grazing management used in our study and the study of Bélanger et al. (2018), whether they are in binary or complex legume-grass mixtures. The legume species had little effect on forage DM yield.

Contribution of the seeded species to forage DM yield Legume species
The contribution of the legume species to forage DM yield across sites and grass mixes ranged from 18% to 49% in the first post-seeding year, 1% to 17% in the second post-seeding year, and 2% to 29% in the third postseeding year ( Table 5). The poor contribution of both legume species to forage DM yield in the second and third post-seeding years could be attributed to poor grazing tolerance, poor winter survival (Bélanger et al. 2006), a strong competition from the forage grasses, and (or) selective gazing in the first post-seeding year observations (Bélanger et al. 2018). In a recent study conducted in Nova Scotia with the grazing of binary legume-grass mixtures, the low proportion of alfalfa and birdsfoot trefoil after the second post-seeding year was attributed to selective grazing by cattle based on visual observations (Bélanger et al. 2018).
Our measurements, however, did not allow for the determination of the exact cause of this reduction in the contribution of the legume species to forage DM yield.

Grass mixes
The contribution of timothy to forage DM yield varied with sites and grass mixes from 14% to 40% in the first post-seeding year, 35% to 70% in the second post-seeding year, and 13% to 59% in the third post-seeding year ( Table 5). The contribution of meadow fescue to forage DM yield ranged from 6% to 17%, 5% to 19%, and 10% to 39% in the first, second, and third post-seeding years, respectively. The contribution of meadow fescue to forage DM yield in the timothy-meadow fescue-based mixtures increased in the third post-seeding year at the expense of timothy. Changes in the botanical composition of the sward and the identity of the species with the largest contribution to forage DM yield have previously been reported (Picasso et al. 2011).
The contribution of tall fescue to forage DM yield varied from 14% to 42% in the first post-seeding year, 8% to 72% in the second post-seeding year, and 24% to 49% in the third post-seeding year ( Table 5). The contribution of meadow bromegrass to forage DM yield ranged from 5% to 30%, 0% to 70%, and 1% to 41% in the first, second, and third post-seeding years, respectively. The contribution of meadow bromegrass was low (<2%) in both second and third post-seeding years at Nappan. Meadow fescue and tall fescue seem to have performed better at New Liskeard than at Normandin, while timothy and meadow bromegrass seem to have performed better at Normandin.
Reed canarygrass and Kentucky bluegrass, the two other grasses seeded in the two grass mixes, did not contribute much to forage DM yield in all three post-seeding years at Normandin and New Liskeard; their contributions were greater at Nappan but still below 15% in all 3 yr (Table 5). The contribution of weeds to forage DM yield averaged across mixtures, sites, and years was 11% and was always less than 23%.

Forage nutritive value N concentration
Increasing N application rates significantly increased forage N concentration in all post-seeding years at the three sites (Table 6). Averaged across the three sites, forage N concentration with no N applied divided by that with 120 kg N ha −1 was 0.94 in the first post-seeding year, 0.79 in the second post-seeding year, and 0.81 in the third post-seeding year. Averaged across sites and years, forage N concentration increased from 23.9 g kg −1 DM with no applied N to 28.2 g kg −1 DM with 120 kg N ha −1 . This result confirms the greater response of forage N concentration to N application rates in the second and third post-seeding years than in the first post-seeding year, and the greater benefits of applying N when the contribution of the legume species to forage DM yield is less than 30%. Increases in N concentration of forage grasses with increasing N fertilization have often been reported in eastern Canada (Tremblay et al. 2005;Pelletier et al. 2009;Bélanger et al. 2016).
Forage N concentration was significantly greater with the alfalfa-based mixtures than with the birdsfoot trefoil-based mixtures in the first post-seeding year at New Liskeard and in the second post-seeding year at Nappan (Table 6). This result could possibly be explained by the greater contribution of alfalfa than birdsfoot trefoil to forage DM yield in the first post-seeding year at New Liskeard (41% vs. 26%; Table 5) and in the second post-seeding year at Nappan (16% vs. 3%; Table 5). Averaged across sites and years, however, forage N concentration of the alfalfa-based mixtures (26.2 g kg −1 DM) and the birdsfoot trefoil-based mixtures (25.9 g kg −1 DM) did not differ significantly.
Forage N concentration of the timothy-meadow fescue-based mixtures was significantly greater than that of the tall fescue-meadow bromegrass-based mixtures in the second and third post-seeding years at New Liskeard and in the second post-seeding year at Normandin, but was significantly less in the first and second post-seeding years at Nappan. Averaged across sites and years, forage N concentration of the tall fescuemeadow bromegrass-based mixtures (26.0 g kg −1 DM) and the timothy-meadow fescue-based mixtures (26.2 g kg −1 DM) did not differ significantly.

aNDF concentration
Forage aNDF concentration was not significantly affected by N application rates at any of the sites in the 3 yr of the study (Table 7). The legume species did not significantly affect forage aNDF concentration in the second and third post-seeding years. In the first post-seeding year, however, the alfalfa-based mixtures had a greater forage aNDF concentration than the birdsfoot trefoil-based mixtures at Normandin but a lower forage aNDF concentration at New Liskeard. These differences in forage aNDF concentration could be due to legume species differences in aNDF concentration or to their relative contribution to forage DM yield. Alfalfa is known to have a greater aNDF concentration than birdsfoot trefoil (Cassida et al. 2000), and this might explain the result at Normandin where the contribution to forage DM yield of the two legume species was similar. Because forage legume species are known to have a lower aNDF concentration than forage grasses, the greater contribution of alfalfa to forage DM yield might explain the lower forage aNDF concentration of the alfalfa-based mixtures at New Liskeard.
Note: Probability values for the effects of N application rate, legume species, and grass mix along with their interactions are also presented. Tm, timothy; Mf, meadow fescue; Kb, Kentucky bluegrass; Tf, tall fescue; Rc, reed canarygrass; Mb, meadow bromegrass; DM, dry matter; NS, not significant at P ≤ 0.05.
The effect of the grass mix on forage aNDF concentration varied with sites and years ( Table 7). The timothymeadow fescue-based mixtures had a greater forage aNDF concentration than the tall fescue-meadow bromegrass-based mixtures in the first two post-seeding years at Nappan and in the first post-seeding year at Normandin, but a lower forage aNDF concentration in the three post-seeding years at New Liskeard and in the second post-seeding year at Normandin. Averaged across the three sites, the timothy-meadow fescue-based mixtures had a lower forage aNDF concentration than the tall fescue-meadow bromegrass-based mixtures in the second (526 vs. 541 g kg −1 DM) and third (516 vs. 538 g kg −1 DM) post-seeding years.

The NDFd
Increasing N application rates significantly increased forage NDFd at Normandin and New Liskeard in the second post-seeding year and at Normandin in the third post-seeding year (Table 8), but this increase was less than 24 g kg −1 aNDF. At Nappan in the first post-seeding year, forage NDFd decreased when the N application rate increased from 0 to 60 kg N ha −1 but then increased when the N application rate went from 60 to 120 kg N ha −1 . Averaged across sites and years, forage NDFd increased from 712 g kg −1 NDF with no applied N to 724 g kg −1 NDF with an application of 120 kg N ha −1 .
The legume species did not significantly affect forage NDFd in any of the 3 yr at the three sites, except in the first post-seeding year at New Liskeard where the alfalfa-based mixtures had a lower forage NDFd than the birdsfoot trefoil-based mixtures ( Table 8). The timothy-meadow fescue-based mixtures had lower forage NDFd than the tall fescue-meadow bromegrass-based mixtures in the first post-seeding year at Normandin and in the first and second post-seeding years at New Liskeard, but a greater forage NDFd in the third post-seeding year at New Liskeard. Differences in forage NDFd, however, were less than 22 g kg −1 aNDF. Across sites and years, forage NDFd of the tall fescue-meadow bromegrass-based mixtures (721 g kg −1 aNDF) and the timothy-meadow fescue-based mixtures (719 g kg −1 aNDF) did not differ significantly.

NSC concentration
Increasing N application rates significantly decreased forage NSC concentration in all post-seeding years at the three sites (Table 9), except at Nappan in the first post-seeding year. Differences in forage NSC concentration between no applied N and 120 kg N ha −1 varied across sites from 5.6 to 8.3 g kg −1 DM in the first postseeding year, 15.4 to 23.5 g kg −1 DM in the second post-seeding year, and 17.7 to 26.3 g kg −1 DM in the third post-seeding year. Averaged across sites and years, the forage NSC concentration decreased from 84.0 g kg −1 DM Table 7. Forage neutral detergent fibre assayed with a heat-stable α-amylase (aNDF) concentration of complex legume-grass mixtures as influenced by nitrogen (N) application rate, legume species, and grass mix in the first three post-seeding years at three sites.

Site
Nappan Normandin New Liskeard Year Note: Probability values for the effects of N application rate, legume species, and grass mix along with their interactions are also presented. Tm, timothy; Mf, meadow fescue; Kb, Kentucky bluegrass; Tf, tall fescue; Rc, reed canarygrass; Mb, meadow bromegrass; NS, not significant at P ≤ 0.05. with no applied N to 68.5 g kg −1 DM with 120 kg N ha −1 . This result confirms the greater negative response of forage NSC concentration to N application rates in the second and third post-seeding years than in the first post-seeding year. Both negative effect (Tremblay et al. 2005;Bélanger et al. 2016) or no effect (Pelletier et al. 2009) of increasing N fertilization on NSC or WSC concentration of forage grasses have been reported.
The legume species only significantly affected forage NSC concentration in the first post-seeding year at Normandin where forage NSC concentration was greater in the birdsfoot trefoil-based mixtures than in the alfalfa-based mixtures (Table 9). Averaged across sites and years, however, forage NSC concentration of the alfalfa-based mixtures (75.1 g kg −1 DM) and the birdsfoot trefoil-based mixtures (76.4 g kg −1 DM) did not differ significantly.
The grass mix did not significantly affect forage NSC concentration in the first post-seeding year at any of the sites (Table 9). The timothy-meadow fescuebased mixtures had a significantly greater NSC concentration than the tall fescue-meadow bromegrass-based mixtures in the second post-seeding year at Normandin and in the third post-seeding year at New Liskeard, but a lower NSC concentration the third post-seeding year at Nappan. Across sites and years, forage NSC concentration of the timothy-meadow fescue-based mixtures (76.8 g kg −1 DM) tended (P = 0.068) to be greater than that of the tall fescue-meadow bromegrass-based mixtures (74.7 g kg −1 DM).

TDN concentration
Increasing N application rates significantly decreased forage TDN concentration at Nappan in the first postseeding year but increased it at New Liskeard in the second post-seeding year (Table 10). Across sites and years, N application rates did not significantly affect forage TDN concentration. Nitrogen fertilization generally did not affect the nutritive attributes associated with forage digestibility such as TDN concentration and NDFd. The effect of N fertilization on forage DM digestibility is the subject of conflicting reports (Bélanger et al. 2001).
The legume species did not significantly affect forage TDN concentration at most sites and post-seeding years (Table 10). The significant effect of the legume species at Normandin in the first post-seeding year was quite small. The timothy-meadow fescue-based mixtures had a significantly greater TDN concentration than that of the tall fescue-meadow bromegrass-based mixtures in Table 8. Forage in vitro neutral detergent fibre digestibility (NDFd) of complex legume-grass mixtures as influenced by nitrogen (N) application rate, legume species, and grass mix in the first three post-seeding years at three sites.
Note: Probability values for the effects of N application rate, legume species, and grass mix along with their interactions are also presented. Tm, timothy; Mf, meadow fescue; Kb, Kentucky bluegrass; Tf, tall fescue; Rc, reed canarygrass; Mb, meadow bromegrass; NS, not significant at P ≤ 0.05. Table 9. Forage nonstructural carbohydrate (NSC) concentration of complex legume-grass mixtures as influenced by nitrogen (N) application rate, legume species, and grass mix in the first three post-seeding years at three sites.
Note: Probability values for the effects of N application rate, legume species, and grass mix along with their interactions are also presented. Tm, timothy; Mf, meadow fescue; Kb, Kentucky bluegrass; Tf, tall fescue; Rc, reed canarygrass; Mb, meadow bromegrass; DM, dry matter; NS, not significant at P ≤ 0.05. the three post-seeding years at New Liskeard and in the second and third post-seeding years at Normandin, but a lower TDN concentration in the first post-seeding year at Nappan (Table 10). Across sites, the TDN concentration of the timothy-meadow fescue-based mixtures was significantly greater than that of the tall fescue-meadow bromegrass-based mixtures in the second (606 vs. 590 g kg −1 DM) and third (613 vs. 586 g kg −1 DM) postseeding years.
The timothy-meadow fescue-based mixtures generally had a lower forage aNDF concentration, and greater forage N and TDN concentrations than the tall fescuemeadow bromegrass-based mixtures in the second and third post-seeding years. This result can be explained by differences in DM yield and the negative relationship between forage nutritive value and DM yield (Bélanger et al. 2001), by differences in the proportion of grasses and legumes, or by differences in the aNDF concentration of timothy and meadow fescue compared with tall fescue and meadow bromegrass. Because the aNDF, N, and TDN concentrations of the individual species in the mixtures were not measured, we cannot conclude with certainty on the reasons for the observed effects.

Soil mineral N
Increasing N fertilization significantly increased the soil mineral N concentration measured in late summer or early fall at all sites in the first and second post-seeding years and at Normandin in the third post-seeding year (Table 11). This increase was particularly important (>39 mg kg −1 ) at New Liskeard in the first two post-seeding years and at Nappan in the second post-seeding year. Averaged across the three sites, the soil mineral N concentration increased from 10.9 to 21.0 mg kg −1 in the first post-seeding year, from 8.8 to 41.2 mg kg −1 in the second post-seeding year, and from 13.0 to 23.4 mg kg −1 in the third post-seeding year when the N application rate was increased from 0 to 120 kg N ha −1 . The soil mineral N concentration in the spring of the second and third post-seeding years was not affected significantly by the N fertilization, and it was less than 20 mg kg −1 at Normandin and Nappan and less than 28 mg kg −1 at New Liskeard (data not shown). Differences in soil mineral N concentration observed in the late summer or early fall disappeared over the winter either due to crop uptake in late fall or early spring or to N losses to the environment. Table 11. Residual soil mineral nitrogen (N) a in late summer or early fall as affected by N application rate, legume species, and grass mixes in the first three post-seeding years at three sites.
Note: Probability values for the effects of N application rate, legume species, and grass mix along with their interactions are also presented. Tm, timothy; Mf, meadow fescue; Kb, Kentucky bluegrass; Tf, tall fescue; Rg, reed canarygrass; Mb, meadow bromegrass; NS, not significant at P ≤ 0.05. a Soil sampling was done to a depth of 20 cm at Nappan and Normandin and to a depth of 15 cm at New Liskeard.
The legume species and the grass mix did not significantly affect the soil mineral N concentration in late summer or early fall of the three post-seeding years. Although the choice of forage species, particularly the presence of legumes with grasses, has been reported to affect soil mineralizable N (Dhakal andAnowarul Islam 2018) andsoil mineral N (Enriquez-Hidalgo et al. 2016;Crème et al. 2016), the differences in the proportion of legumes and grasses in the different mixtures were probably too small to affect the soil mineral N concentration.

Agronomic applications
No or little N fertilization was required to reach maximum forage DM yield when 30% or more legumes were present in the legume-grass mixtures in the second and third post-seeding years. The need for N fertilization in legume-based complex mixtures, therefore, increases with time as the contribution of the legumes to forage DM yield decreases. The current recommendations in eastern Canada of not applying or applying very little fertilizer N when the contribution of the legumes to forage DM yield is greater than between 30% and 50% was therefore confirmed. Nitrogen fertilization of legume-grass mixtures when a significant proportion of legumes is present could increase soil mineral N in late summer or early fall, with a corresponding greater risk of N losses to the environment.
Our results also indicate that increasing N fertilization of complex legume-grass mixtures decreases the readily available energy in the form of NSC and increases N concentration of grazed forages with possible negative consequences for N-use efficiency by ruminants. Recent studies conducted in eastern Canada have shown that a difference of 10 g NSC kg −1 DM is sufficient to affect milk production and N-use efficiency by dairy cows (Brito et al. 2008(Brito et al. , 2009). This decrease in forage NSC concentration with increasing N fertilization combined with an increase in forage N concentration has, therefore, the potential to reduce animal production and N-use efficiency.
The legume species (alfalfa and birdsfoot trefoil) affected seasonal forage DM yield and some nutritive attributes only in the first post-seeding year. This effect of the legume species in the first post-seeding year, however, varied with sites, and no general conclusion can be drawn. The absence of an effect of the legume species in the second and third post-seeding years on forage DM yield, most nutritive attributes, and the soil mineral N concentration is not surprising considering the low contribution of the legume species to forage DM yield (<30%). The low contribution of the two legume species to forage DM yield in the second and third postseeding years (<30%) at all sites highlights the challenge of maintaining legume species in the rotationally grazed pastures of eastern Canada.
The two grass mixes in the complex legume-grass mixtures affected the forage DM yield and some nutritive attributes but mostly at one site. Averaged across the three sites, the tall fescue-meadow bromegrass-based mixtures had a greater seasonal DM yield than the timothy-meadow fescue-based mixtures. This superior performance of the tall fescue-meadow bromegrass-based mixtures, however, was mostly seen at New Liskeard where the contribution of tall fescue to DM yield was greater than at the other sites (32%-72% vs. 8%-36%). The performance of grass mixes in complex legume-grass mixtures might vary with the soil and climate conditions of different locations.
Our results suggest that forage DM yield and nutritive value are not greatly influenced by the composition of complex grass-legumes mixtures compared with other factors such as N fertilization, sites, and post-seeding years. Consequently, the species composition of complex forage mixtures might not be a critical factor in ensuring adequate forage productivity under grazing. Our results, however, are based on four forage mixtures made of species that are relatively well adapted to conditions of eastern Canada. Further research is needed to consider a broader range of forage mixtures.

Conclusions
The low contribution of alfalfa and birdsfoot trefoil to forage DM yield of the legume-grass mixtures in the second and third post-seeding years (<30%) at all sites resulted in very little effect of the legume species on most forage attributes and it highlights the challenge of maintaining legume species in the rotationally grazed pastures of eastern Canada. Regardless of legume species, the legume-grass mixtures based on tall fescue and meadow bromegrass performed similarly or slightly better than the legume-grass mixtures based on timothy and meadow fescue in terms of seasonal forage DM yield, but they tended to have a greater forage aNDF concentration, and lower forage N and TDN concentrations. Our results confirm the current recommendations in eastern Canada of not applying or applying very little N fertilizer on grazed legume-grass complex mixtures as long as the contribution of the legumes to forage DM yield is greater than 30%. Increasing the N application rates increased forage N concentration and decreased forage NSC concentration, but had no effect of forage TDN concentration and aNDF digestibility. crew of the Nappan Research Farm and Leo Gieson, Hal