The experiment was carried out in the “Centro Regional Sur”, Experimental Station of the Faculty of Agronomy (34°36'S, 56°13'W), of the Universidad de la República (Canelones, Uruguay) during the spring, from 17 October to 27 November 2010.

Treatments consisted of two pastures with contrasting composition: one with 60% of herbage mass of the legumes Medicago sativa L. (lucerne) and Trifoliumrepens L. (white clover) and 40% of the grass Bromus auleticus Trin. ex Nees (“cebadilla”), referred to as Legume sward hereafter, and the other pasture with 24% herbage mass of the legume Lotus corniculatus L. (birds-foot trefoil) and 76% of grass Lolium multiflorum Lam. (ryegrass), referred to as Grass sward hereafter.

A 2 × 2 Latin square design was used to evaluate the two treatments. Eight lactating Holstein cows were used over two periods of 21 days (with seven days of dietary adaptation and 14 days of faeces collection and methane measurements). The animals were allotted to two groups of four cows with similar pre-experimental milk production (24.9 ± 4.15 kg/d milk), live weight (536 ± 18 kg) and lactation stage (195 ± 7 days). Each lot was randomly assigned to the treatments in the first period, and then assigned to the other treatment in the second period. Swards were strip-grazed at a daily minimum amount of 30 kg DM/cow/day (above 5 cm). This herbage allowance was not limiting for herbage intake, according to previous work from Peyraud et al. [10]. A new area of pasture was offered to the cows once a day after the morning milking. Back-grazing was prevented by electric fencing. Daily areas to be offered were determined by estimating forage availability as described below.

Pre-grazing herbage mass and mean sward height were measured on days 9, 12, 16 and 19 in each plot. Herbage mass was estimated by harvesting three diagonal strips (10 m × 0.5 m) with a motor scythe at a cutting height of 5 cm. Herbage mass below the motor scythe cutting height was determined by two samples cut within the frame of a 0.3 m × 0.3 m quadrat on each strip. Herbage samples were weighed fresh, sampled, and approximately 500 g was dried at 60 °C for 48 h for DM determination and for subsequent chemical analysis of forage offered. Each time, 30 tillers were taken at random and the extended height was measured to estimate the mean sward height before grazing. The same procedure was followed for determination of post-grazing herbage mass and sward height after grazing, on days 11, 14, 18 and 21. Total herbage mass to ground level was calculated as the sum of herbage mass measured above and under the motor scythe cutting height. Herbage utilization was calculated as the difference between herbage mass before grazing and after grazing and expressed as a percentage of herbage mass before grazing. The mean depth of defoliation was estimated by the difference between the values of sward height before and after grazing.

On the same days as for the determination of pre-grazing herbage mass, three handfuls of herbage were cut at ground level with scissors on the edge of each strip, to determine the proportion of legume and grass of the herbage offered (three samples per strip and three strips per treatment). The three handfuls of herbage were bulked (one sample per strip) and arranged in a bag to keep the sward structure unaltered and immediately frozen. The proportion of legume and grass in the total DM herbage mass was determined by manual separation from a first sub-sample prior to drying. These fractions were then dried at 60 °C over 48 h and then weighed dry. The second sub-sample, with its original structure still preserved, was cut at a height corresponding to the mean post-grazing sward height. The upper portion, considered as representative of the defoliated herbage, was dried before chemical analysis.

Simultaneously to the experiment with dairy cows (days 7 to 21), the digestibility of the offered herbage (cut above motor scythe height) was measured at each experimental period with twelve Corriedale wethers (six per treatment) to characterize herbage quality. The animals were kept in metabolism crates with free access to water. Each of the two experimental periods for the digestibility trial comprised an adjustment period of 10-days, followed by 5-days dedicated to collection of faeces, orts, and feeds. For the adjustment period, feed was available ad libitum until DM intake stabilized (at a level of an average refusal of about 10% of the feed offered). During the five faecal collection periods, feed offered and refused, and faeces of each animal were weighed daily, and samples taken for chemical analysis.

Individual herbage OM intake was determined using chromic oxide (Cr₂O₃) to estimate faecal organic matter output, and Nitrogen (Nf) and acid detergent fiber (ADFf) contents in the faeces (g/kg OM) to estimate herbage OM digestibility (OMd) of herbage, according to the equation established by Comeron and Peyraud [11] using herbage-based diets without supplements:

                        OMd = 0.791 + 0.0334 Nf – 0.0038 ADFf, (R² = 0.89; RSD = 0.013)

Concentrate pellets containing chromic oxide (ca. 20 g Cr₂O₃ kg of supplement) were supplied in two equal portions of 150 g each, at milking, starting on day one of the experimental period to achieve a ruminal balance (steady state). Faeces were rectal-sampled after morning milking and after evening milking from days 8 to 21, and oven dried at 60 °C over 96 h in order to measure the dry matter, the chrome concentration, and the chemical composition.

The CH₄ emission was measured using the SF₆ tracer technique reported by Johnson et al. [8]. Seven days before the beginning of the experiment, a SF₆ permeation tube (PT, provided by the NIWA, National Institute of Water and Atmospheric Research, New Zealand) was introduced per os into the rumen of each animal. The permeation rates were 6.422 ± 0.416 mg/d (mean and standard deviation, respectively), as given by a 6-weeks calibration assay. The breath gas sampling system consisted of a 0.5 L stainless steel collecting vessel (canister), a ball-baring inflow restrictor adjusted to accumulate 0.5 bar of air sample during a 5-day period and a short tube used to connect both. The inflow restrictor was located just above the animal’s nostrils and protected against water and dust by means of a double filter reported by Gere and Gratton [9]. Two collecting canisters were fitted to each animal’s head by means of especially designed halters. Immediately prior to the sampling period, each collecting canister was evacuated (<0.5 mb) after cleaning with high purity nitrogen gas (N₂). The breath gas samples were measured over two sub-periods of five days during each period of measurements (on days 10 to 14 and 16 to 20). At the end of the first 5-day sub-period, the set of canisters (two per animal) were replaced by another similar set. So at the end of each period, two sets of canisters per animal were collected across two sub-periods of five days. Additionally, an identical set as used with the cows was used to collect background air samples during each 5-day sub-period. The breath gas samples collected were analyzed immediately after the end of the experimental period. Daily CH₄ emissions were calculated from SF₆ release rate and CH₄/SF₆ ratio of concentrations in breath samples, after correction for background gas concentrations [8].

The cows were milked twice, from 6:00 to 6:30 h in the morning and from 17:00 to 17:30 h in the afternoon. Individual milk production was measured each day. Milk fat and protein contents were determined on four consecutive days, twice in each experimental period (on days 9 to 12 and 16 to 19). Cows were weighed on the last day of each experimental period.

All the dried samples were ground through a 1 mm screen before chemical analysis. The dry matter (DM) concentration was determined by drying at 105 °C in an oven for 24 h and ash content was determined by incineration at 600 °C for 4 h for organic matter (OM) calculation. Crude protein (CP) content was determined using the Kjeldahl method (7.021 procedure) in A.O.A.C [12]. Content of a neutral detergent fiber (NDFom) was determined without sodium sulfite and with heat stable amylase. Acid detergent fiber (ADFom) and acid detergent lignin (lignin) were determined using sequential analysis and are expressed exclusive of residual ash according to Van Soest et al. [13]. An Ankom apparatus (Ankom 220, Fairport, NY, USA) was used for extraction and filtering. Condensed tannins were analyzed according to Schofield et al. [14]. Gross Energy (GE) was determined using an adiabatic bomb calorimeter (Gallenkamp Autobomb; Loughborough, Leics, UK).

Milk samples at every milking of each collection period were analyzed for fat, protein, and lactose content with infrared spectroscopy (Bentley 2000 Infrared Milk Analyzer, Bentley Instruments Inc., Chaska, MN, USA). Chromium (Cr) concentration in faecal samples was determined by atomic absorption spectrophometry (Perkin-Elmer 2380, Norwald, CT, USA), using air and an acetylene flame according to William et al. [15]. Cr standards were prepared using pre-trial faecal collections that contained no Cr.

The concentrations of CH₄ and SF₆ were determined by chromatography in a specialized Laboratory (INFQC, Universidad de Córdoba, CONICET, Córdoba, Argentina). The samples were injected at once in two different setups. For CH₄ a 3 mL loop was used, a HP-PLOT Q column and a FID detector. For SF₆ a 10 mL loop, a HP-MOLSIV column and an ECD detector. Each sample was analyzed at least twice and the average values were used to obtain methane concentration and methane emission. Maximum delay between the collection and the determination of CH₄ and SF₆ concentrations was 15 days.

All the animal data were analyzed according to a 2 × 2 Latin square design, using PROC MIXED function of SAS (version 9.1; SAS Inst. Inc., Cary, NC, USA) using the model: Y = μ+ Pi + Tj + Ak + εijk, where μ was the overall mean, Pi was the fixed effect of period (i = 1 to 2), Tj was the fixed effect of sward (j = 1 to 2), Ak was the random effect of animal (k = 1 to 8) and εijk was the associated error. The production and composition of the milk were analyzed as repeated measures over time, according to an autoregressive model of order one reported by Littel et al. [16]. The pasture data were analyzed according to a 2 × 2 Latin square design by ANOVA using the GLM procedure of SAS (SAS Institute, Inc., 2001) with the following model: Y = μ+ Pi + Tj + εij. In both MIXED and GLM models initially the interaction treatment x period was included, and it was not significant, therefore, following the recommendations of Lorenzen and Anderson [17] it was not included in the final model. In order to compare the mean treatment values, the test of the minimum significant difference was used.

Total rainfall (186 mm) and mean temperature (15.5 °C) in the spring 2010 were below seasonally climatic conditions (309 mm, and 20.0 °C, mean over 10 years).

Total herbage mass was lower on the Grass sward (−869 kg DM/ha) (P = 0.0002). However, herbage mass above 5 cm (2,165 kg DM/ha on average) and sward height (29.5 cm on average) were similar between pastures (Table 1). Botanical composition of both swards was significantly different as intended (P < 0.0001): Legume sward presented a higher proportion of legumes (60% of herbage mass above 5 cm as Medicago sativa and Trifoliun repens) while Grass sward presented a higher proportion of grasses (76% of herbage mass above 5 cm as Lolium multiflorum). Both swards were at reproductive stage during the experiment.

Chemical composition (above 5 cm) was also different between swards. Grass sward had higher DM content (P = 0.0008), aNDFom content (P = 0.0290), and ADFom content tended to be higher too (P = 0.0618), while CP content was two times lower (−102 g/kg DM) (P < 0.0001) compared to Legume sward. Gross energy was also lower in Grass sward (P = 0.0017). Condensed tannins were higher on Legume sward than on Grass sward (+1.8 g /kg DM, P = 0.0205), and for both pastures, these values were slightly higher than those reported by Barry and McNabb [18] for lucerne and ryegrass forage.

The in vivo digestibility trial resulted in higher DM (+73 g/kg DM, P = 0.0006), OM (+39 g/kg DM, P = 0.0103), and aNDFom (+67 g/kg DM, P = 0.00016) digestibility for the Legume sward than the Grass sward. However, the digestibility of ADFom was similar for both swards (P = 0.1516) (Table 1). In particular, the DM digestibility of the Grass sward is similar to values reported by Molano and Clark [19] for reproductive ryegrass in an experiment to determine the effect of quantity and quality of forage intake on CH₄ emissions on wether lambs.

Cows on the Grass sward exhibited a higher herbage allowance than in the Legume sward (+10 kg DM/cow/day, P = 0.0226), as the area allocated per cow was higher on Grass sward (Table 2). However, in both treatments herbage allowance was above the stipulated minimum amount of 30 kg DM/cow/day (above 5 cm) which allows an ad libitum intake according to Peyraud et al. [10]. The depth of defoliation (9.5 cm on average) and the herbage utilization (42% on average) were similar for both swards.

Due to the high post grazing sward height, the chemical composition of defoliated herbage on Grass sward showed a different composition from the offered herbage (Table 2). On the Grass sward the grazed forage contained more CP (+31 g/kg DM), more GE (+2.3 kJ/kg DM), less aNDFom (−58 g/kg DM), and ADFom (−55 g/kg DM) than on the offered sward above 5 cm (Table 2). These results are in agreement with the large vertical gradients observed in the different chemical constituents from the upper to the lower layer of a ryegrass sward reported by Delagarde et al. [20], The content of condensed tannins of the herbage defoliated was similar for both pastures (P = 0.5119).

As observed in the offered grass, the defoliated Grass sward had a lower CP content than defoliated Legume sward (−47 g/kg DM, P = 0.0225). Nevertheless, the CP content of the Grass sward is above the minimum critical content for ruminal digestion, according to Poppi and McLennan [21] (12.5% DM for pastures ranging on 62% DM digestibility). The lower CP content of defoliated herbage in Legume sward compared to the offered herbage may be explained by a lower proportion of white clover, because the post-grazing height (20 cm) was above the white clover height.

Faecal OM output was similar for Legume and Grass treatments (4.5 kg OM/cow on average). The OM digestibility of defoliated herbage, estimated by faecal index according to Comeron and Peyraud [11], did not differ between treatments (707 g/kg OM on average) and was higher than values measured on the digestibility trial with wethers on the offered herbage (above 5 cm). Due to the high herbage allowance per cow and the relatively high levels of herbage height in both swards, the post grazing height remained substantially higher (20 cm on average) than the cutting height of the motor scythe (5 cm on average), allowing the cows to consume a higher quality of herbage, in particular in Grass sward as discussed above (Table 2). Daily herbage OM intake was similar among treatments and averaged 15.7 kg OM/cow (Table 3), which agrees with the calculated amount of defoliated herbage presented in Table 2 (herbage allowance x herbage utilization). Therefore, the early stage of heading in ryegrass did not reduce herbage OM intake as reported by Astigarraga and Peyraud [22]. Finally, digestible OM intake was also similar (11.2 kg DOM/cow on average) among swards. The digestible OM intake is in accordance with the requirements for maintenance, milk production, and live weight gain calculated from the standards given by INRA [23] and adjusted for the energy expenditure used for muscular activity while grazing according to Ribeiro et al. [24], indicating that any bias in the estimation of OM intake was not significant.

Milk yield (20.3 kg/cow on average) and milk composition (37.1 and 33.3 g/kg milk for fat and protein contents respectively, on average) were not affected by the experimental pastures (Table 4). Live weight variation (+16 kg LW on average) was also not affected by treatments. These results are clearly associated with a similar total DOM intake and thus in digestible energy intake reported above.

The efficiency of sample collection was 94% (30 successful samples were collected from 32 expected). The methane emission (g/d and L/d) was similar among treatments, and averaged 368 g/d (Table 5). The CV for absolute CH₄ emission was 13.8%.

Methane yield per unit of DMI (22.2 g/kg on average) and as a percentage of gross energy intake (GEI) (Ym = 6.6% on average) did not differ between treatments, likely because DM intake and ingested energy were similar for Legume and Grass swards in this study (Table 3). These values are in the range reported in New Zealand by Lassey [25] and Boadi et al. [26] for dairy cattle grazing on temperate forages. Ramirez-Restropo and Barry [27], suggest that feeding forage legumes like lucerne or red clover tends to decrease CH₄ losses (g/kg DMI) compared to grass. Nevertheless, the results of our study do not seem to confirm what. Hammond et al. [28,29] recently suggested that methane emissions could be more related to DM intake, which allows variations in the composition of the diet selected at grazing, in particular for Grass swards. In fact, the methane emission and the methane yield (as g/kg DMI or Ym) reported by Lee et al. [30] on dairy cows fed increasing proportions of legume (white clover), are closer at similar DMI than for similar proportions of legume in the forage mixture. Archimède et al. [31] found no differences in methane per kg OM intake between C3 grasses and cold legumes in a meta-analysis compared for the same fiber content, digestibility and intake. Finally, methane emission expressed by unit of milk production was similar between treatments, as no differences in milk performance were observed.