Evaluation of cool-season perennial forage varieties as monocultures and legume–grass binary mixtures under intensive grazing

Abstract In Canada, new forage varieties need not undergo grazing trials before registration and sale. To evaluate the effect of grazing, six cool-season species including meadow bromegrass (Bromus riparius Rehmann), orchardgrass (Dactylis glomerata L.), sainfoin (Onobrychis viciifoila Scop.), and three alfalfa varieties (Medicago sativa L.) were established in monocultures and grass–legume binary mixtures in Saskatchewan, Canada. Forage treatments were randomly established within each of three 3.5 ha paddock replicates. In Year 1, 69 Bos taurus crossbred steers, and in Year 2, 149 steers were homogenously allocated to the three paddocks for the grazing trials. Alfalfa monocultures had the greatest (p < 0.05) pre-graze forage yields and crude protein content. Leaf area index was greater in alfalfa monocultures than in all other treatments (p < 0.01). Etiolated growth differed (p < 0.05) among binary mixtures but not monocultures. Alfalfa monoculture stands were preferred over grass monocultures and most grass–legume mixtures, and sainfoin was avoided relative to alfalfa. Killarney orchardgrass was the least productive and least preferred among forages evaluated. In summary, alfalfa monocultures had greater (p < 0.05) forage yields and quality than sainfoin monocultures, grass–legume binary mixtures, or grass monocultures.


Introduction
Perennial forages are an integral part of cattle production as they provide low-input feeds for cattle whilst optimizing the use of marginal land not suitable for annual cropping (Beacom 1991).Legume-grass mixtures are common in western Canada because of the ability to produce higher forage yields and withstand greater environmental variation than monocultures (Bélanger et al. 2017).As profit margins are narrow for cow-calf producers, having long-lasting, productive forage stands to support the production of several calf crops (i.e., 6-11 years) is crucial to maintain a return on investment (Entz et al. 1995;Rothwell 2005), especially under an intensive grazing system.Therefore, it is important to provide producers with accurate grazing data so that they can make informed forage variety selections for productive pasture establishment.
Under current regulations, new forage varieties are not required to be tested in grazing trials as a part of variety testing (Government of Canada 2019).Instead, most variety testing involves small plot trials and mechanical harvesting which does not accurately assess forage performance under grazing pressure (McCartney and Bittman 1994).Livestock intro-duce unique stressors to the plant community that mechanical harvesting cannot simulate, including trampling, nutrient deposition, irregular defoliation frequency, and plant selection.
Cattle exhibit increased intake of preferred forages relative to other forages (Rutter et al. 2004;Darambazar et al. 2014;Lardner et al. 2019).Thus, providing pasture composed of preferred species may increase animal gain and improve the uniformity of grazing, which is valuable from an economic standpoint (Ganskopp et al. 1997;Rutter et al. 2004).However, little is known about the forage preferences of beef cattle grazing multiple cool-season grasses and legumes used in western Canada (Rutter et al. 2004;Lardner et al. 2019).Therefore, the objective was to test the hypothesis that forage performance and nutritive value would be similar among alfalfa monoculture treatments, with sainfoin monocultures producing slightly less forage yield than the alfalfas.The performance of the grasses was hypothesized to be similar between treatments, with lower forage yield and nutritive value than legume monocultures.This study was designed to evaluate new cool-season perennial forage varieties as monocultures and legume-grass binary mixtures under intensive grazing Fig. 1.Experimental field layout of the trial.The research paddocks with forage treatments seeded in randomized adjacent 0.26 ha plots (21 m × 125 m) within three replicate blocks.Note: 3006ALF, alfalfa var.3006; CronusALF, alfalfa var.Cronus; FootholdALF, alfalfa var.Foothold; SF, sainfoin var.AC Mountainview; MBG, meadow bromegrass var.AC Armada; OG, orchardgrass var.Killarney.There was no fencing between treatment plots within blocks to allow free animal movement within a paddock.conditions over multiple grazing seasons to evaluate forage yield, grazing preference, and botanical composition.

Weather
The average monthly air temperature and rainfall values were sourced from the University of Saskatchewan Livestock and Forage Centre of Excellence (LFCE) weather stations onsite from 2018 to 2021.Long-term 30-year averages were retrieved from the Saskatoon Saskatchewan Research Council weather station (Environment Canada 2023) for comparison.

Study site and management
The study was conducted at the LFCE, Cow-Calf Teaching Unit (Clavet, SK, Canada), between 2019 and 2021.Soil type in the area is Dark Brown (Saskatchewan Soil Survey 1992).Prior to this study, the research site was used for conventional annual crop production.In spring 2018, weeds were controlled with two pre-seed applications of glyphosate herbicide (1.9 and 2.5 L ha -1 ; Roundup; Monsanto, Creve Coeur, Greater St. Louis, Missouri, USA), and on June 2018, the site was seeded with barley (Hordeum vulgare L. var.AC Rosser) at a rate of 54 kg ha -1 and seeding depth of 3.8 cm, along with 46-0-0 nitrogen (N) fertilizer applied at a rate of 56 kg ha -1 .
The barley cover crop was harvested and baled as greenfeed in August 2018.Treatment forages were then seeded the week following the barley harvest using a 1.3 cm seeding depth and recommended seeding rates for each forage in the Dark Brown soil zone (Saskatchewan Forage Council 2007).Seeding was done using a pull type Wintersteiger (Wintersteiger AG, Ried, Austria) at row spacing of 30 cm.Forage treatments were seeded into three 3.5 ha replicate blocks (Fig. 1).The perimeter of each replicate block was fenced to create an individual paddock, leaving 18 m alleyways between paddocks.Each replicate block contained 14 treatments seeded in randomized, adjacent 0.26 ha (21 m × 125 m) plots.There was no fencing between treatment plots to allow free animal movement within a paddock.Six forages were seeded: three alfalfa (ALF, Medicago sativa L.) varieties (3006,3006ALF;Cronus,CronusALF,and Foothold,FootholdALF)

Grazing management
Animal use was approved by the University of Saskatchewan Animal Care Committee (protocol# 20010048) and performed in accordance with the guidelines put forth by the Canadian Council on Animal Care (CCAC 2009).
BW was recorded at a consistent time (∼0900 h) for each steer over 2 consecutive days at the beginning and end of each period using the inspected, permanent weighing and handling systems.Steers were stratified by weight and randomly allocated to each of the three trial paddocks for each grazing event, resulting in a similar number of steers and average BW allocated to each paddock.In Year 1, each of the three replicate-blocks held 23 steers (4 animal unit [AU] ha -1 ; estimated according to Lardner et al. 2013).Dry conditions in 2019 allowed for only one grazing period lasting 19 days (from 27 July to 15 August).Forage preference measurement was performed only the first 7 days of grazing to minimize the effect of limited forage on preference, after which stocking density was increased to 12 AU ha -1 with all 69 steers grazing each paddock sequentially.This reduced plant selection, and cattle uniformly grazed the forages to a 5 cm stubble height (Lardner et al. 2015).In Year 2, two grazing periods occurred, where grazing period 1 lasted 19 days (from 26 June to 14 July 2020) and grazing period 2 lasted 9 days (from 27 August to 4 September 2020).There were 45 days of rest period for the forages between the two grazing events.The same group of 149 steers was used for grazing periods 1 and 2. Each block contained 49 or 50 steers (8 AU ha -1 in grazing period 1; 9 AU ha -1 in grazing period 2).Like Year 1, forage preference was measured for the first 7 days of grazing, and then the forages were grazed until a uniform 5 cm stubble height was achieved.In Year 3, one grazing period occurred, where grazing period 1 lasted 20 days (from 6 to 26 July 2021).

Forage yield and botanical composition
Each year, available forage was determined immediately before grazing events, as described by Sanderson et al. (2001).Four 0.25 m 2 quadrats from each 0.26 ha treatment strip were clipped to a 5 cm stubble height.Quadrat samples were handseparated into their botanical components on-site at the time of clipping, with each component identified by plant species, and stored in a paper bag for further analysis.The components were categorized as target legume (if applicable), target grass (if applicable), and other.The samples were dried at 55 • C for 72 h in a forced-air oven (Heratherm OMH750; Thermo Scientific Ltd., Langenselbold, Germany).Botanical composition was determined on a dry matter (DM) basis as the average proportion of each component (by weight) per 0.25 m 2 in each treatment.Total available forage of each sampling point was determined by summing the aboveground biomass of all forage components removed from each plot and expressed in DM yield (DMY) kg ha -1 .After determining the botanical composition, all quadrat clippings taken from each treatment strip were combined, subsampled, and ground to pass through a 1 mm screen using a hammer mill (Christy and Norris Ltd., Chelmsford, UK) for laboratory analysis.

Chemical composition
Forage samples were analyzed using standard wet chemistry by Cumberland Valley Analytical Services (Waynesboro, PA, USA).Crude protein (CP) was quantified using a Leco FP-528 Nitrogen Combustion Analyzer (Leco Corporation, MI, USA) according to method 990.03 of the AOAC (2000).Method 973.18 of the AOAC (2000) was used to determine acid detergent fiber (ADF) content of the forages, and neutral detergent fiber (NDF) was quantified using the method described by Van Soest et al. (1991).Calcium and phosphorus contents were determined (method 985.01;AOAC 2000) using a Perkin Elmer 5300 DV ICP spectrometer (Perkin Elmer, CT, USA).Total digestible nutrients (TDN) were calculated according to Weiss et al. (1992).
Leaf area index, maturity staging, and etiolated growth Due to labor shortage caused by the COVID-19 pandemic, leaf area index (Years 1 and 2), maturity staging (Years 1 and 2), and etiolated growth (Years 2 and 3) data were not collected and therefore were presented as 2-year results of the study.Leaf area index (LAI) measurements were taken immediately prior to grazing using an AccuPAR LP-80 LAI ceptometer (Decagon Devices, Pullman, WA, USA) as an indicator of canopy density and forage yield.Three measurements of the light transmittance were taken above and below the canopy for an averaged reading at five sites per treatment strip according to operational manual (Decagon Devices 2013).Pregraze maturity staging was conducted in Years 2 and 3. Four 0.25 m 2 quadrats were thrown to randomly select plants in each treatment strip; all legume stems in each quadrat were staged according to the method described by Kalu and Fick (1981), and all grass tufts in each quadrat were staged using the method described by Moore and Moser (1995).
Non-structural carbohydrate reserves were estimated for each mixture in Year 2 (2020) and Year 3 (2021) using the etiolated growth technique as described by Lardner et al. (2003) and Ward et al. (2012).Individual plant tufts were defoliated to a near-ground level and covered using lightproof terracotta pots, ∼20 cm in diameter and 20 cm in height, and secured using polyethylene straps in the ground.Etiolated growth was sampled to a 2.5 cm stubble height every 14 days from 11 May to 22 June 2020, and then every 7 days until growth was exhausted.For Year 3, etiolated growth was sampled every 14 days from 20 April to 13 July 2021, and then every 7 days until growth was exhausted.

Forage preference
Forage preference was evaluated using the instantaneous scan sampling technique, both visually from a vehicle driving along paddock fence lines and via aerial drone footage (Mufford et al. 2019).Two-hour observation periods occurred twice per day beginning at 0600 and 1900 h.Five scans were performed at 30 min intervals in each 2-h observation period (10 total scans per day).At each scan, the number of animals present per plot was recorded.Forage preference was calculated as the proportion of animals actively grazing a given treatment at each scan.Preference index for treatment (PI, %) was then calculated as the number of steers grazing a treatment plot divided by the total number of steers grazing on the paddock.

Statistical analysis
Data were analyzed as a randomized complete block design replicated in four grazing periods (n = 4) nested within 3 years.Each paddock was a replicate block (n = 3), and the experimental unit was the forage treatment plot (n = 14).Replicate block and grazing period nested within year were considered random effects, with forage treatment as a fixed effect.All data were analyzed using the PROC GLM procedure of SAS (SAS Institute, Inc. 2020).The UNIVARIATE procedure was used to determine that data were normally, identically, and independently distributed.Differences between the grazing periods 1 and 2 were also determined using PROC GLM.Five covariance structures were tested for analysis, including simple, compound symmetry, ante-dependence, unstructured, and heterogenous compound symmetry.The Kenward Roger option was used to estimate denominator degrees of freedom.The simple covariance structure was selected, having the lowest Akaike and Bayesian information criterion values.Differences in treatment means were considered significant when p < 0.05, and trends were considered when 0.05 < p < 0.10.To assess the relationship between forage preference and nutrient profiles, Pearson correlation coefficients were calculated between the forage PI of the grazing steers and DMY, CP, NDF, ADF, TDN, Ca, and P using the CORR procedure of SAS (SAS Institute, Inc. 2020).

Weather
The monthly temperatures at the site (Table 1) were below average in January and February of Year 1, preceding the first grazing season of this study.Otherwise, temperatures were similar to the long-term average (LTA).Drought conditions existed in 2018 (year before study), the year that all forage treatments were seeded, with less than 210 mm of annual precipitation.The long-term average annual precipitation for Saskatoon area was 340 mm.Below-average snowfall and rainfall from January to May of Year 1 (26 vs. 97 mm) exacerbated dry conditions leading up to the first grazing event in Year 1.After this initial drought, rainfall rose to above the LTA in June of Year 1 (152 vs. 117 mm).The Year 2 saw above-average rainfall during May and June (149 vs. 100 mm).The Year 3 saw below-average total precipitation compared with LTA (181 vs. 340 mm).Overall, these weather data suggested that the current experiment was conducted in an environment with similar temperatures and 30% lower precipitation compared to the 30-year average weather conditions.
During the grazing period 2 in Year 2, a higher CFY (p < 0.05) was obtained for 3006ALF, but it did not differ from the other ALF and SF treatment combinations, with 3006ALF-OG, 3006ALF-MBG, CronusALF-OG, CronusALF-MBG, FootholdALF-OG, FootholdALF-MBG, SF-OG, and SF-MBG being not different (p > 0.05) from OG.At the grazing period 1 in Year 3, 3006ALF and FootholdALF monocultures produced greater CFY (p < 0.01) compared to all other treatments except for CronusALF or the binary mixtures of ALF with OG or MBG, or SF-MBG.The SF-MBG was not different (p > 0.10) from SF-OG and OG.
For period 1 between Year 1 and Year 3, 3006ALF was greater in CFY (p < 0.01) but not different (p > 0.05) from the other ALF, with CronusALF-MBG being similar to 3006ALF-OG or MBG, CronusALF or FootholdALF-OG, FootholdALF-MBG, SF-OG or SF-MBG, and MBG monoculture.A grazing period effect was observed in Year 2 (p < 0.01), where period 1 accumulated greater CFY than period 2. No interaction (p = 0.82) was observed between treatment and grazing period.Overall, when pooled all 3 years and periods by year, 3006ALF (3127 kg ha -1 ) was significantly greater (p < 0.05) in CFY relative to FootholdALF-OG (1641 kg ha -1 ) and SF-OG (1218 kg ha -1 ), and OG (990.5 kg ha -1 ; data not shown) was numerically greater relative to other treatments in CFY.

Pre-grazing individual species yield
Grass component yield did not differ (p = 0.51; Table 3) among treatments during Year 1.In Year 2, MBG yield was greater (2063 vs. 761 kg ha -1 , p < 0.01) compared to OG in monocultures and binary mixtures.In addition, there was a tendency (p = 0.07) for a greater MBG yield at grazing period 2 in Year 2. In Year 3, there was a greater grass component yield (p < 0.01) of MBG compared to OG.During the first grazing period of each year, MBG had the greatest yield as grass component (p < 0.01), with CronusALF-MBG being compara- ble to all other mixtures and monocultures.No interaction (p = 0.83) was observed between treatment and grazing period for grass.During all grazing periods in Years 1 and 2, legume monocultures had higher yields (p < 0.01) than the legumes in binary mixtures (ranging from 919 to 2578 vs. 86 to 645 kg ha -1 , respectively).During the grazing period in Year 3, however, legume component yields differed within the binary mixtures, with the highest (1968 kg ha -1 , p < 0.01) being 3006ALF in the mixture with OG, and the lowest-yielding legume components were FootholdALF and SF (178 and 189 kg ha -1 , respectively) in the mixtures with MBG.

Forage chemical composition
The 3006ALF and FootholdALF had greater CP content (p < 0.01) when compared to the other treatments (Table 4).CronusALF-OG was not different from CronusALF-MBG, 3006ALF-OG, FootholdALF-OG, SF, and SF-OG, and SF-MBG had lower CP but did not vary from 3006ALF-MBG, FootholdALF-MBG, SF-MBG, and OG.MBG and its binary mixtures contained the highest ADF content (p < 0.01), compared to FootholdALF-OG with intermediate ADF level, but did not different from either 3006ALF, CronusALF, and SF or their binary mixtures with OG, while FootholdALF contained the lowest ADF content.In addition, MBG had greater NDF content (p < 0.001) than the other treatments, and FootholdALF had lower NDF but not different from 3006ALF and SF.The SF had greater TDN (p < 0.01) as opposed to lower TDN in MBG as was in CronusALF-MBG, FootholdALF-MBG, SF-MBG, and OG.Higher Ca concentration (p < 0.01) was detected in all ALF monocultures and CronusALF-OG, with MBG having lower Ca concentration, but not different from FootholdALF-MBG, SF-MBG, and OG.Phosphorus concentration was greater (p < 0.01) in OG compared with the other treatments and was lower in MBG and its mixtures with all legumes.
The SF was in the late bud stage (p < 0.01), while all ALF were in the early bud stage during the grazing period 1 in Year 2 (Table 6).However, at the grazing period 2 in Year 2, all legumes were in the late flowering stage.In 2021, 3006ALF and SF were in the late flower stage, whereas CronusALF and FootholdALF were in the early flower stage.Among grasses, MBG was in the reproductive stage (p < 0.01), while OG was in the stem elongation stage during the grazing period 1 in Year 2, but both grasses were at the vegetative stage during grazing period 2. However, in Year 3, MBG was at the reproductive stage (p < 0.01), but OG was at the vegetative stage.Overall, this was indicative of OG struggling during the grazing by steers over the study years.
Cumulative etiolated growth was similar among forages (p = 0.12) in Year 2 (Table 7).However, ALF varieties had greater (p = 0.02) cumulative etiolated growth as compared with OG, with SF and MBG being intermediate but not dif-  ferent from others.In addition, cumulative etiolated growth was higher in Year 2 than in Year 3 for all forage treatments.

Grazing preference
The grazing period had an effect on grazing preference (p < 0.05), where steers demonstrated preferences among forage treatments in Year 1 and period 1 of Year 2, but not in period 2 of Year 2 (Table 8).Steers tended (p = 0.06) to se-lect legume monocultures during Year 1.During the grazing period 1 in Year 2, steers preferred (p = 0.02) CronusALF, with CronusALF-MBG, SF-OG, and SF-MBG being intermediate but not different from the others.Pooled averages from all grazing periods in Year 1 and 2 showed that the three ALF varieties were the most preferred forages (p < 0.01, Table 8).The MBG was less preferred (p < 0.05) than Foothold and 3006ALFs, but it was preferred similarly (p > 0.05) with CronusALF.Killarney OG was the least preferred compared to the four legumes and MBG (p < 0.05).The remaining forage treatments (all binary mixtures, SF, and MBG) were similar in preference values.In Year 2, a grazing period effect was observed (p = 0.05), where animals displayed a preference for 3006ALF, CronusALF, CronusALF-OG, CronusALF-MBG, FootholdALF-MBG, SF, and OG over 3006ALF-OG, 3006ALF-MBG, FootholdALF, FootholdALF-OG, SF-OG, SF-MBG, and MBG during period 1, but steers preferred 3006ALF-MBG, FootholdALF, FootholdALF-OG, SF-OG, and SF-MBG at period 2.
The results of Duncan grouping of forage treatments by preference (Table 9) suggested that steers mostly preferred ALF monocultures and their mixtures with grass (p ≤ 0.05) compared with OG or OG mixtures with legumes, and the latter may cause cattle metabolic disorder (i.e., bloat).In western Canada, pasture bloat is more frequent in the moist areas of parkland regions, especially in Gray Wooded soil zone, and is becoming progressively less frequent in Dark Brown to Brown soil zones.ALF bloat has been reported in 40% of the livestock farms in the Peace River region of northern Alberta (Majak et al. 2003).Many studies have reported that grazing cattle on grass-legume mixed pasture reduces the incidence of bloat and that ALF should constitute no greater than 50% of the mixture to minimize bloat risk (Popp et al. 2000; Majak  Kalu and Fick (1981) and Moore and Moser (1995), respectively.a,b, Means within a column and within forage type with different letters differ at p < 0.05.
x,y, Means within a row with different letters differ at p < 0.05.et al. 2003).This can dilute the potential for bloat because cattle are less likely to consume large quantities of legumes in a mixed pasture.Feed additives, such as poloxalene (Dougherty et al. 2006), can be mixed with feed or provided in mineral supplements to help reduce the likelihood of bloat.No incidence of bloat occurred in steers in the current study, probably because the strip (with 14 different strips) grazing system was employed.When all data were pooled, forage components and forage preferences of the grazing steers (Table 10) were weak but positively correlated (p ≤ 0.01) with DMY (r = 0.33), CP (r = 33), TDN (r = 0.21), and Ca (r = 0.38), and weak and negatively correlated (p < 0.01) with ADF (r = -0.31)and NDF (r = -0.37).

Discussion
Forage yield and leaf area index Climatic conditions (precipitation) allowed for one grazing period in Year 1 and two periods in Year 2 in this study.With most of Saskatchewan's forage production occurring within a 60-day window in early to mid-summer (Agriculture and Agri-Food Canada 2008) and the 42-day post-graze rest period recommended for western Canada by the Beef Cattle Research Council (2021), one to two grazing periods per season are typical for this region.In the current study, similar forage yield was accumulated in the first grazing period of each year and lower in the grazing period 2 of Year 2. A high first cut yield was expected.As the first growth of forages reaches the reproduc- tive growth stage, the second cut mostly remains as the vegetative growth (Foster et al. 2013;Damiran et al. 2022).As no interaction between grazing period and treatment was observed in the study, it was indicated that grazing period affected yield equally among treatments.The LAI of grasses observed in the current study were considerably lower than those reported by others (Peri et al. 2007;Biligetu and Coulman 2010).Peri et al. (2007) and Biligetu and Coulman (2010) reported an LAI of 4.1 and 4.2 for OG and MBG, respectively, whereas grass species in the current study reached only 1.4 LAI units each.However, Papadopoulos et al. (1995) reported a similar LAI of 1.59 to our study for OG under rotational grazing.
The maximum light interception for these grasses was observed at LAI 5.0 (Pearce et al. 1965).Low tiller development in OG has been noted when the grass is rested for long periods between defoliations (Papadopoulos et al. 1995;Oates et al. 2011); frequent grazing of OG encourages prostrate leaf growth (Papadopoulos et al. 1995), which increases LAI available for recovery.In the current study, both grasses experienced limited regrowth after the grazing period 1 of Year 2, likely due to low water stress, but OG suffered to a greater extent, suggesting that the 45-day rest period preceding the grazing period 2 in Year 2 might not be adequate for OG.A minimum LAI of 2.4 is required for first-cut ALF to intercept 95% of incident light and optimize growth rate (Wilfong et al. 1967).All ALFs exceeded this threshold in the current study.
The reported LAI for SF ranges from 2.8 to 5.0 (Sheehy and Popple 1981), but SF had an LAI of 1.6 in this experiment.The leaf water potential of SF was higher than that of ALF, which theoretically makes SF more drought resistant (Sheehy and Popple 1981).Yet, all three ALF varieties, in the present study, exceeded the optimum LAI for growth, while AC Mountainview SF failed to meet previously reported LAI values under the same conditions.Meyer (1975) reported that this is caused by SF's inability to effectively fix N. Without substantial N fertilization, SF yields are poor despite adequate rainfall (Meyer 1975).Yield and tiller production rely on precipitation, particularly in the early months of the growing season (White 1985), which limited in the year of establishment ( 2018) and the spring of both production years studied.This suggests that the dry conditions combined with heavy grazing pressure may have limited production of the forages in the current study.
Legume, grass monoculture, and binary mixture pre-grazing yields Legume monocultures out-yielded grass-legume mixtures, with ALF monocultures producing twice the herbage biomass of ALF-grass mixtures.This was in contrast with past studies, which demonstrated similar yields between ALF and ALFgrass mixtures in the first two production years (Aponte et al. 2019).This may be explained by the proportion of grass and legume observed in the stand.In the current study, grasslegume mixtures were composed of 12%-40% legume only, whereas Aponte et al. (2019) acknowledged that their ALFgrass mixtures were primarily ALF.
ALF-grass mixtures produced 100% of the yield of grass monocultures, consistent with results from Sleugh et al. (2000), but lower than numbers reported by Aponte et al. (2019).Mixing ALF with a grass often increases forage yield (Sleugh et al. 2000;Bélanger et al. 2017;Damiran et al. 2022); this is likely because grass production is bolstered by ALF's N fixation and deep-rooted water access (Høgh-Jensen et al. 2004).Higher forage yield in grass-legume mixtures was not observed in the current study, although this may be because the legume composition of mixtures was lower than in other studies.
It is estimated that ALF stand in Saskatchewan yields 3700-4900 kg ha -1 without any irrigation (McArton et al. 2020).ALF monocultures in our study yielded approximately half that amount per harvest.FootholdALF produced 43% of the yield of Beaver (a common control variety) reported in Saska- ).An earlier study in Swift Current, SK, reported yields closer to those observed in our study with Fleet MBG producing 1500 kg ha -1 (Biligetu et al. 2014).
Killarney OG also failed to meet expected yields based on values reported by previous research.Acharya et al. (2005) demonstrated an average OG yield of 4022 kg ha -1 across four dryland locations in Saskatchewan and 6234 kg ha -1 in 28 locations throughout western Canada. Moreover, McArton et al. (2020) found that Killarney OG yielded much less in Saskatoon, SK, at 1632 kg ha -1 .Killarney OG yield in the present study was lower still, producing only 45% of the result McArton et al. (2020) reported.
Over 3 years, the ALF varieties had greater (p < 0.01) yields (ranging from 631 to 758 kg ha -1 ) when they were mixed with OG compared to MBG (ranging from 193 to 308 kg ha -1 ), indicating ALF were more competitive with OG.In contrast, SF did not vary (p > 0.05) in yield in the binary mixtures with the grasses.For legume component yield, no grazing period effect (p = 0.60) in Year 2 or interaction between treatment and grazing period (p = 0.44) was observed.

Pre-grazing individual species yield
There was no effect (p > 0.05) of year on the botanical composition in this study, suggesting no effect of competition or winterkill on treatment forages between years.However, both grass and legume species yield interacted (p < 0.01) with the treatment in period 1 over 3 years.Killarney OG made up ∼60% of grass-legume mixtures, which was consistent with the results of Bélanger et al. (2017) in eastern Canada.However, OG significantly decreased in all treatments from grazing period 1 to grazing period 2, while the legume fraction of Killarney OG mixtures increased.This may be attributed to slow tillering of OG under dry conditions.
OG is known to regrow rapidly in comparison to other grasses but exhausts its root reserves in the process, making itself susceptible to winterkill (Davidson and Milthorpe 1966;Lardner et al. 2003).If sufficient stubble is not left behind after grazing, inhibited photosynthesis combined with minimized root reserves will hinder regrowth (Davidson and Milthorpe 1966).As such, OG persistence is poor under intense grazing pressure, particularly in drought conditions (Beacom 1991).A combination of these factors was present in the current study, leading to poor regrowth and persistence of Killarney OG.
Unlike OG, the proportion of MBG in binary mixtures increased from Year 1 to Year 2. Armada MBG is a short rhizomatous grass, making it more competitive than other bunch grasses (Knowles et al. 1993).MBG regrowth is superior to many other grasses because its growth originates from existing tiller bases, further increasing persistence under grazing because it is not limited by the removal of elevated apical meristems (Knowles et al. 1993;Biligetu and Coulman 2010).However, MBG' regrowth has been shown to reduce ALF persistence in rotationally grazed stands (Katepa-Mupondwa et al. 2002).Conversely, Billman et al. (2020) demonstrated that mob-grazed paddocks favor ALF growth over grasses.
ALF with a higher lateral root density (creeping-or branching-rooted varieties) were positively correlated with greater grass proportions in ALF-grass mixtures (Hakl et al. 2018).The present study did not observe differences in the botanical composition of ALF-grass binary mixtures among ALF varieties or root types.SF monoculture plots consisted of 24% other species or weeds, as inferior weed competitiveness is a well-known trait of SF (Moyer 1985).The weed component of SF stands in Saskatchewan is widely variable, ranging from 5% to 96% when left uncontrolled (Waddington et al. 1985).Weed control must be a consideration for producers interested in grazing pure SF stands, or a grass should be seeded with the legume.As a companion study showed, SFgrass mixtures contained fewer (p < 0.05) weeds than SF as a monoculture stand (Sim 2021).

Chemical composition and plant maturity
The energy and fiber contents of ALF determined in the current study were comparable to those reported by Bélanger et al. (2017) in eastern Canada and by McArton et al. (2020) in Saskatchewan.The CP content of FootholdALF was ∼3% lower in the current study than in previous Saskatoon trials (McArton et al. 2020), but the available soil N at the Clavet study site was also nearly 60% lower (data not shown) than it was reported for the Saskatoon site.CronusALF contained 2% less CP than 3006 or Foothold despite having equal or greater soil N levels at the study site and being of the same maturity, suggesting that genetic differences make Cronus less efficient at assimilating N into its plant protein (Wery et al. 1986).The protein content of CronusALF was similar to that of an older bloat-reduced ALF variety, AC Grazeland ALF (Lardner et al. 2019).
The energy and protein contents of AC Mountainview SF were equivalent to that of common SF studied by Khalilvandi-Behroozyar et al. (2010), and fiber values (ADF and NDF) were comparable to those reported by McArton et al. (2020) for AC Mountainview SF.Acid detergent fiber and NDF components were similar between the ALFs and SF, as others have noted (Chung et al. 2013;McArton et al. 2020).
Killarney OG was generally more nutrient dense compared to AC Armada MBG in the current study.The energy and protein content of both OG and MBG were lower than values previously observed for these species (Bélanger et al. 2017;McArton et al. 2020) and their fiber content was higher (Ferdinandez and Coulman 2001;Billman et al. 2017;Bélanger et al. 2017;McArton et al. 2020).The ADF portion of both grasses was high, indicating a high degree of lignification in plant tissues.High lignin is associated with environmental stressors such as water and N deficiency (Moura et al. 2010), both of which existed in the present study.
The chemical composition of binary mixtures was dependent on their proportion of grass and legume, with higher legume content conducive to better quality forage.Killarney OG mixtures were also higher quality than mixtures of the same legume paired with AC Armada MBG because legume-OG mixtures contained more legume, and OG itself lent to greater nutrient density to the mixture than MBG.The difference in CP content was likely more affected by the grasses in the binary mixtures than the legumes due to the treatments' botanical compositions.
Vegetative grasses are generally lower in fiber content than mature grasses (Ferdinandez and Coulman 2001;Rawnsley et al. 2002;Darambazar et al. 2013).MBG and OG in the present study contained more ADF in their vegetative regrowth than they had when initially harvested in the reproductive stages of maturity, while NDF remained unchanged.Water stress may have contributed to this effect.Baron et al. (2000) reported that ADF was lower in MBG and OG after regrowth; however, the authors noted that the grasses experienced above-average rainfall during regrowth.Conversely, the present study experienced dry conditions throughout.Water stress can increase lignin and/or cellulose proportions in plants (Lee et al. 2007;Le Gall et al. 2015), perhaps due to changes in cell wall thickness that enhance water retention and withstand rehydration turgor (Lee et al. 2007;Le Gall et al. 2015).As such it is possible that the grasses in the present study contained greater ADF due to the dry conditions experienced by central Saskatchewan in 2020.

Etiolated growth
Etiolated growth lasted 84 days before growth ceased, consistent with the timeframe reported by previous studies (Lardner et al. 2003;Ward et al. 2012).Thirty-eight percent of cumulative etiolated growth occurred within the first 14 days after plants were covered (data not shown), making this the period with the greatest rate of total non-structural carbohydrates' (TNC) mobilization.Richards and Caldwell (1985) also demonstrated this effect in several Agropyron species, and as reported in their study, the rate of root carbon utilization was greatest in the first 7 days.The quadratic effect of time × treatment in the current study (data not shown) indicated that the growth rate was different among forages.CronusALF was the only tap-rooted variety evaluated in this study.
Although differences among the ALF performances were small, CronusALF produced lower etiolated growth in the monoculture plots compared to 3006 and FootholdALF varieties.Fibrous root mass is correlated with fall dormancy (Johnson et al. 1998) which may explain why CronusALF produced less etiolated growth than either 3006ALF or FootholdALF which have more branching root mass.The presence of grass also promotes branching of ALF roots (Hakl et al. 2018) with the absence of grass, as competition in the ALF monoculture treatments likely further limited Cronus's number of lateral roots, making its poorer regrowth more evident in monoculture than binary mixture.
The etiolated growth of AC Mountainview SF was comparable to that of the ALF varieties, which was unexpected in light of SF's poor persistence.Cooper and Watson (1968) suggest that SF's larger crown allows it to store more TNC reserves than ALF.However, SF only begins storing non-fiber carbohydrates in the crown in the fall, shortly before winter (Mowrey and Volesky 1993).The late start in combination with its poor competitiveness likely caused the poor persistence of SF despite having the capacity for adequate regrowth.As evidenced, the cumulative etiolated growth of AC Mountainview SF in monoculture was 55% greater than etiolated growth in binary mixtures, supporting the theory of Vasileva and Vasilev (2012) that SF root matter is constricted by competition with grasses.AC Armada MBG produced ∼10 mg/cm 2 more etiolated growth than Paddock MBG in a similar study by Ward et al. (2012).However, there is no indication of whether MBG rhizomes were severed prior to covering the plant tufts in that study.Rhizomes were not severed in the present experiment.
Killarney OG generated slightly greater than 50% of the total etiolated growth produced from MBG in Year 2. These results agree with those of Lardner et al. (2003), who also observed that MBG produced more etiolated growth than OG and other grass species.OG' habit of expending all its root TNC rapidly after defoliation is most likely responsible for its poor etiolated growth following intensive grazing in the previous year (Davidson and Milthorpe 1966).Finally, forage varieties can maintain energy reserves, overwinter, and then produce adequate spring growth may be advantageous in the subsequent grazing season.

Grazing preference
In the current study, the lack of an interaction between treatment and grazing period indicated that forage preference was consistent among treatments over time.Previous studies (Gesshe and Walton 1981;Billman et al. 2017) have shown that forage preferences were absent when the plants were vegetative, and in the current study, 10 of the 14 treatments were primarily composed of vegetative grasses during this period.
The results of our study further confirmed the hypothesis that legume monocultures are most preferred by grazing beef steers and that binary mixtures are moderately preferred (Gesshe and Walton 1981;Rutter et al. 2004;Boland et al. 2011).The ALFs and SF were equally preferred by grazing steers, contrary to results of Smoliak and Hanna (1975) and Parker and Moss (1981) who found that animals strongly preferred SF over ALF, and unlike the findings of Gesshe and Walton (1981) who reported that cattle demonstrated a strong aversion to SF.
Interestingly, SF was selected as much as ALF despite being more mature than the ALFs.The legume monoculture treatments were the highest yielding, most nutrient dense, and lowest in fiber, presumably making them the most attractive and selective to the grazing steers.However, the correlation between forage nutrient composition and forage preference was lower (weak or no relationship) than we expected.The hypothesis that grasses are least preferred when compared with legumes was partly true, as OG was the least preferred forage on offer, whereas MBG did not differ in preference compared to other forages, except for 3006ALF and FootholdALF.The current finding that OG was the least preferred forage contradicts the results elsewhere that grazers were strongly attracted to grazing OG over other forages (Billman et al. 2017;Catalano et al. 2020).In the current study, forage preference was weakly (−0.4 < r < 0.4) correlated with DMY and CP, with OG producing the least forage yield and ranking lower than the other treatments in terms of protein content.Despite similar fiber content between MBG and OG, the MBG was grazed more often by steers.This was in agreement with the findings of Billman et al. (2017) that high fiber is not necessarily a deterrent in all cases.

Conclusion
The three ALF cultivars, 3006, Cronus, and Foothold exhibited similar performance characteristics.AC Mountainview SF was poor yielding under dry conditions.SF was a high-quality forage, but its persistence under intensive grazing challenges was poor.AC Armada MBG produced higher yields and was persistent under intensive grazing relative to Killarney OG.The legumes paired with Killarney OG produced higher quality forage compared to the legume-MBG mixtures.The steers preferred pure ALF stands over grass monocultures and most grass-legume mixtures, were more reluctant to graze OG than other species, and did not prefer or avoid SF relative to ALF.Being the most preferred, ALF monoculture pasture could be better managed and utilized in intensive grazing programs.
toon, SK, by McArton et al. (2020), 3006ALF 52%, and Cronus 41%.The same trial by McArton et al. (2020) also observed FootholdALF yielding 3675 kg ha -1 , which was over 1600 kg ha -1 more than the same cultivar in the present study.It should be noted that the soil texture at the site used by McArton et al. (2020) was clay loam, whereas the present study site was a sandy soil, but both trials took place in the Dark Brown soil zone.Neither 3006 nor CronusALF has previously been tested in Saskatchewan under dryland conditions.SF yield in the current study was moderate in comparison to values reported by others (McArton et al. 2020; Biligetu et al. 2021).McArton et al. (2020) recorded AC Mountainview yield over 4000 kg ha -1 near Saskatoon, whereas Biligetu et al. (2021) reported that SF grown in Lanigan, SK, yielded 597 kg ha -1 , only 47% of that produced in the present study.Likewise, SF yield was 690 kg ha -1 lower compared to the ALFs (average 1147 vs. average 1836 kg ha -1 , SF and ALFs, respectively) in the present study.McArton et al. (2020) showed similar results, with unirrigated AC Mountainview SF producing 333 kg ha -1 less than the check ALF.Although AC Mountainview SF produced substantially more in the trial by McArton et al. (2020) than in the present study, the relationship between ALF yield and SF yield was consistent.Saskatchewan Forage Test trials in Saskatoon demonstrated an average yield of 2200 kg ha -1 among the four MBG cultivars tested, including AC Armada MBG (McArton et al. 2020).Similarly, McCartney et al. (2004) observed yields of nearly 2000 kg ha -1 in unfertilized MBG.In contrast, MBG in the present study produced only 60% and 67% of those amounts (1337 kg ha -1

Table 1 .
Monthly mean air temperature and precipitation at the research site in the year prior to data collection (2018) and trial years.

Table 4 .
Effect of treatments on pre-graze forage chemical composition (%, DM basis) over 3 years.

Table 5 .
Average leaf area index of forages over 2 years (Years 1 and 2).

Table 6 .
Pre-graze maturity of forage species over 2 years.

Table 7 .
Cumulative etiolated growth of plants in monoculture and binary mixture treatments over 2 years.
a,b, Means within a column with different letters differ at p < 0.05.

Table 9 .
Duncan grouping of forage treatments by preference from most to least preferred.
Groupings with different letters differ at p ≤ 0.05.

Table 10 .
Correlations between forage components and forage preference index of the grazing steers.