Milk yield and feeding behaviour responses to two flat-rate levels of concentrate supplementation fed over a period of 8 months to cohorts of grazing dairy cows, differing in genotype, bodyweight, or milk yield

Site details

All data were collected from the Tasmanian Institute of Agriculture Dairy Research Facility (TDRF) at Elliott (41′08″S, 145′77″E), Tasmania, during the lactation season of 2017–2018. The climate at this site is characterised as cool-temperate with high annual rainfall (long-term mean 1180 mm), with the majority of rain falling between April and October (winter and early spring). The site has well established predominantly perennial ryegrass (Lolium perenne) and white clover (Trifolium repens) pasture. Approximately 40% of the paddocks available for grazing by the milking herd were irrigated. Pasture management was based on rotational grazing, with a new pasture allocation after every milking, a target pasture biomass at the start of grazing of ~3000 kg DM/ha and a residual biomass of 1500 kg DM/ha. The site was undulating, with a clay loam red Ferrosol (Humic Etrodox) soil (Soil Survey Staff 1990; Isbell 1996).

Animals

The milking herd at TDRF consisted of 320 spring-calving cows rotationally grazing approximately 120 ha of pasture (48 ha under irrigation). The milking herd consisted of approximately 30% Holstein-Friesian animals, with the remainder primarily being cross-bred Friesian × Jersey, with fewer numbers of Friesian × Aussie Red and Friesian × Montbeliarde × Aussie Red. The cows were milked twice daily for the duration of the lactation. One hundred and eighty early lactation multiparous (parity range 2–7) dairy cows were selected from the TDRF milking herd as they started their lactation, and allocated to different cohorts on the basis of three cow parameters tested (60 cows per parameter). Cows were between 9 and 15 days in milk (DIM) when allocated to a cohort.

Experimental design and treatments

In early lactation, in a spring-calving herd, six cohorts of 30 cows each were selected on contrasting differences for either their genotype (Friesian (GF) or cross-breed (GC)), production at start of lactation (either high (PH) or low (PL) milk yield), or BW (high (WH) or low (WL) weight) as proxy for cow size (Table 1). Each cohort was divided into two balanced groups of 15 cows each, receiving concentrate at either 2 or 6 kg DM per cow per day (Table 1). Treatments for each cow started at recruitment into a cohort, and trial data collection after an adaptation period of one fortnight. The trial period was divided into three seasonal periods of 10, 12 and 10 weeks (P1, spring; P2, late spring–summer; P3, late summer–autumn), with the trial cows a mean of 28, 98 and 182 days in milk at the start of P1, P2 and P3 respectively. All trial cows remained part of the milking herd at TDRF, with similar opportunity for all cows to graze pasture or feed on supplemented grass silage during periods of pasture shortage (typically ~20% of the diet over summer and autumn). Milk yield was recorded at each milking, and feeding behaviour (as proxy for pasture intake) was continuously measured with MooMonitor+ collars.

Data collection

Cows were milked twice per day in a 20-unit aside swing-over dairy with automatic cup removers. All concentrate supplementation was provided individually in the dairy by an automated feeding system, checked weekly to ensure that concentrate delivery amounts from each individual feeder were accurate. Each cow had their own individual numbered ear tag as well as a National Livestock Identification System (NLIS) ear tag, which contained an electronic radio-frequency identification device (RFID). The tag had a unique ID number that could be read by sensors in the dairy. This system is used to ID individual cows as they enter the dairy for milking, or when they pass over the walk-over scales to measure BW. The dairy at TDRF uses DeLaval equipment (DeLaval International AB, Tumba, Sweden), including DeLaval’s DELPRO® system, which uses the cow’s RFID device to control and collect data for each individual cow and during each milking, using walk-over scales (DeLaval AWS100 automatic weighing system) for BW, the concentrate feeding system in the dairy (DeLaval individual bail concentrate feeding system), and the DeLaval milking-point controller MPCII with DeLaval milk meter MM27BC, which is an ICAR (International Committee for Animal Recording)-accredited milk-monitoring system. Each cow was fitted with a neck collar-mounted activity and behaviour monitoring system, MooMonitor+ (Dairymaster Inc., Kerney, Ireland). The commercially available MooMonitor+ uses a three-dimensional accelerometer to determine cow movement and head direction, and, on the basis of proprietary algorithms, distinguishes among feeding, ruminating and resting behaviours. Using the MooMonitor+, cows were continuously monitored, and the time each cow spent feeding, ruminating and resting reported as minutes per 15-min time block.

Feed analysis

Concentrate samples were collected each time the silo was filled. Samples were analysed by wet chemistry at Cumberland Valley Analytical Services, Waynesboro, PA, USA. Pre-grazing pasture samples (taken above residual height) were collected approximately once a week for a consecutive AM and PM pasture allocation throughout the 2017–2018 milking season. Samples were dried at 60°C and ground through a 0.1 mm sieve. Ground dried samples were analysed (Table 2) using near-infrared (NIR) analysis at Cumberland Valley Analytical Services (Waynesboro, PA, USA).

Milk yields

At each of the two daily milking events, each cow’s milk yield (L) was recorded using the DeLaval milk metering system. Milk-composition samples were collected during alternate morning or afternoon milkings on 1 day in each month of the trial. Milk samples were assayed for fat and protein concentrations (% as w/v), and for somatic cell count (SCC) by using a Bentley B2000 Infrared Milk Analyser (Bentley Instruments Inc., Chaska, MN, USA) by TasHerd Pty Ltd (Hadspen, Tas., Australia).

Milk yields from once monthly alternating AM or PM herd tests were converted to fat and protein-corrected milk (FPCM) production for each month and cohort. FPCM was calculated using the following equation derived from Tyrrell and Reid (1965):

Bodyweight

Cow BW was measured twice daily with automatic in-race walk-over scales (DeLaval AWS100 automatic weighing system) as animals exited the dairy. The AWS100 utilises an algorithm that discounts recorded weights that differ greatly from the 7-day mean weights for individual animals (e.g. instances when more than one cow is on the scales at any one time). We used the 7-day average weight data as reported by the AWS100.

Data analyses

Results were analysed separately for each of the three parameters (genotype, production and BW) and periods. Data for milk yield, time feeding, rumination and resting, and BW were averaged across fortnights, with P1 consisting of five fortnights from early September until mid-November, P2 of six fortnights from mid-November until early February, and P3 of five fortnights from early February until mid-April. Outcomes were analysed statistically within a linear mixed-modelling framework in which the experimental units are given by Cow ID, for which a random-effects model was applied, with cow ID as the random effect and treatment (parameter × feeding level) as the fixed effect. Calculations were performed using Proc mixed and Proc PLM in SAS ver. 9.4. For all terms in the resulting model, the predicted means with standard errors and 95% confidence intervals were reported. Contrasts of pairs of specific treatments were calculated. These were shown as the mean differences with standard errors of the differences. Contrasts were tested for equality; significance is reported under adjusted P-values. Significances of contrasts were adjusted for multiplicity using Scheffe’s method. Confidence intervals for the contrasting parameter difference are shown, including after adjustment for multiplicity. Diagnostic plots were examined to assess conformance to modelling assumptions. The statistical analysis results report per cohort the fortnightly predicted means with standard errors, and the levels of adjusted P-values between the contrasting 6 kg and 2 kg DM treatments for each of the correlated cohorts as an indicator of the size of responses (Tables 3, 4 and 5 for the genotype, production and weight cohorts respectively).

Ethics

Animal ethics approval was granted prior to all animal procedures (University of Tasmania Animal Ethics Committee, A0016635) according to guidelines set out in the Tasmanian Animal Welfare Act 1993, and the ‘Australian code of practice for the care and use of animals for scientific purposes, 8th edition’ (National Health & Medical Research Council Australia 2013).

Feed quality

Cows grazed ryegrass (Lolium perenne)- and white clover (Trifolium repens)-dominated pastures for the duration of the trial. Pasture quality ranged for metabolisable energy (ME) from 9.3 to 12.1 MJ (mean 10.9 MJ), crude protein (CP) from 12.6% to 25.5% (mean 18.5%) and neutral detergent fibre (NDF) from 39% to 63% (mean 48.1%). In total, 50 pasture samples were analysed, with the mean per period showing the lowest ME and CP and highest NDF in P2 (late spring–summer) (Table 2). In P2 and P3 (late summer–autumn), grass silage was supplemented to fill the feed gap when pasture growth was insufficient. During these occasions, the amount of silage fed was usually ~20% of the diet, with the long-term average silage quality at TDRF typically ~9.7 MJ ME, 13.6% CP and 53.5% NDF. The calculated energy content in the pelleted concentrate ranged from 10.9 to 11.9 MJ ME, and CP from 11.2% to 17.6%. The means per period for the nutrients analysed are shown in Table 2. The higher CP in the concentrate in P2 and P3 is due to an increased protein concentration from the start of P2 onward, to account for the decline in %CP in the pasture that time of the year, and the silage fed in P2 and P3 with a lower %CP than for pasture.

Cohort responses

The fortnightly predicted means with standard errors and adjusted P-values between concentrate supplementation at the 6 kg DM (6 kg) and their contrasting treatments at 2 kg DM (2 kg) for genotype, production and BW are presented in Tables 3, 4, 5. The 6 kg cohorts that achieved a significant (P < 0.05) difference in milk yield compared with the 2 kg cohort were for GF6 in all periods and GC6 in P3; PH6 in P2 and P3 and PL6 in all periods; and both WH6 and WL6 in all three periods. There was a significant difference for feeding time for GC6 in all three periods; PH6 in P1 and PL6 in P1 and P2; and WL6 in P1 and P2. A significant difference was found for resting time only for GC6 in P3, and for BW only for PL6 in P3. No significant difference was found for rumination time.

The mean for all cohorts and all periods for milk yield, and feeding, rumination and resting time (minutes per cow) per day was 18.85 L, and 472, 492 and 360 min respectively. The mean results for cohorts in each of the three parameters tested were very similar and the difference to the mean for each of the three parameters ranged from −0.05 to 0.08 L, −2.95 to 4.16 min, −3.85 to 5.42 min, and −6.46 to 6.59 min for milk yield and feeding, rumination and resting respectively. The differences between the 6 kg and their contrasting 2 kg cohorts are reported in Tables 6 and 7 and are calculated from the statistically analysed results presented in Tables 3, 4, 5. The calculated results include marginal milk responses, and are reporting the difference between cohorts, and provide comparison for the results that differ significantly. Mean marginal milk response (L milk per 1 kg DM extra of concentrate) for all 6 kg cohorts over the trial period was 0.88 L, increasing from 0.71 L in P1, to 0.92 L in P2 and 1.03 L in P3 (Table 6). GF, PL and WL achieved the highest marginal response, whereas GC achieved the lowest.

The higher level of concentrate supplementation resulted in a mean reduction in feeding time (a proxy for pasture substitution) of 37 min per cow per day between all 6 kg and 2 kg cohorts, and the difference reduced from 46 to 29 min from P1 to P3, with GC6 showing the greatest overall reduction (58 min, Table 6). The higher amount of concentrate supplementation resulted in a mean reduction in rumination time of 5 min per cow per day (1% of total rumination time) between all 6 kg and 2 kg cohorts, with P1 reporting an increase of 5 min rumination time, reducing to −13 min in P3 (Table 6). Resting time increased for all 6 kg cohorts by 29 min, with not much difference among periods (Table 6), and the overall average was 360 min (6 h) per day resting.

The ratio between the difference in feeding minutes per cow per day and litres milk yield between 6 kg and 2 kg cohorts was used, to infer a magnitude of substitution of pasture by concentrate (i.e. more substitution, as indicated by a greater reduction in feeding time, will result in a lower ratio value). The 6 kg supplementation resulted in a mean ratio of difference in feeding minutes to difference in litres of milk of −10.5 between all 6 kg and 2 kg cohorts, with that ratio increasing (less pasture substitution) from −16.3 in P1 to −7.0 in P3. The lowest overall mean ratio for GC at −26.5 (Table 6), with a nadir for GC in P1 of −89.8 (Table 7), indicated the largest pasture substitution.

The mean differences between each 6 kg and 2 kg cohorts and each period are reported in Table 7. The smallest marginal milk response (0.18 L/kg DM) was for GC in P1, and the highest (1.22 L/kg DM) for WL in P3. The GC6 cohort showed the largest reduction in feeding time (i.e. largest pasture substitution) for all three periods, compared with all other cohorts. This was reflected in the feeding to milk ratio, especially in P1 (−89.8, Table 7).

Results from once monthly alternating AM or PM herd tests were investigated to calculate the mean litre FPCM production for each month and cohort. The results of these 8 monthly herd tests (not reported) indicated that FPCM response generally was higher than the litre response. The only exception was for WL, for which the mean litre FPCM tended to be just below the litre response.

Bodyweight

The mean 7-day average BW per cow at the end of each period, as a mean for all cohorts and all periods, was 492 kg. The mean BW results for the cohorts in each of the three parameters tested were very similar and within ±10 kg of the mean of 492 kg. Similar change trends between cohorts were observed over the trial period for the 7-day mean BW at the end of each period, i.e. a drop from recruitment (9–15 DIM) to the end of the adaptation period (mean 27 DIM), a start of recovery in BW by the end of P1 (mean 97 DIM), an increase of BW by the end of P2 (mean 181 DIM), and another small increase by the end of P3 (mean 251 DIM). Overall, the 7-day mean BW gain for all cohorts was 23 kg from recruitment. The mean weight gain between the end of P1 and the end of P3 for all cohorts was 29 kg. All 2 kg cohorts gained 20 kg BW, and the 6 kg cohorts achieved a 38 kg BW gain. Only cohort PL6 reached significance (adj. P < 0.05) for BW at the end of P3 compared with PL2 (Table 4), gaining an extra 27 kg BW between the end of P1 and P3 compared with PL2. Cohort PH6 only gained 14 kg more compared with PH2. The 6 kg cohorts WH6 and WL6 had similar additional gains (18 and 15 kg respectively) compared with WH2 and WL2, which is similar to the mean 18 kg higher gain for all 6 kg cohorts.

Milk

The treatment cows in this investigation received a flat rate of 6 kg DM of concentrate per day, 4 kg DM more per day than did the control cows. Although a positive response for milk yield is expected when feeding more supplementary concentrate, substitution effect diminishes its overall effect (Bargo et al. 2003; Wales and Kolver 2017). A kilogram DM of concentrate has an approximate ME content equal to the requirement of producing 2 L of milk. However, the mean marginal response during this trial was 0.88 L. Table 7 shows the highest marginal response at 1.22 L for the WL6 cohort in P3, and the lowest at 0.18 L for GC6 in P1. The lowest response for all 6 kg cohorts (Table 6) was in P1 (0.71), increasing towards the highest response in P3 (1.03). The pasture quality is lowest in late spring and summer (P2). And under dry conditions such as those that occur in P2 at the trial site, pasture growth can reduce, necessitating supplementing the diet with silage of quality below that of pasture. The highest response was in autumn (P3), when pasture growth reduces because of declining temperatures, with pasture growth dropping below herd requirement, necessitating supplementing the diet with silage. The 6 kg cohorts produced 2.9 L (13.2%) more milk in P1, and showed a more persistent milk production, declining less in milk in P2 with 4.6 L (18.6%) and 3.4 L (13.7%) in P3, totalling 8.0 L (32.3%), than the 2 kg cohorts which declined 5.4 L (24.7%, and 3.9 L (17.6%) respectively, totalling 9.3 L (42.3%). We postulate that if this trial had continued until the end of lactation, the marginal responses for the 6 kg cohorts would have increased, similar to findings by Kellaway and Porta (1993), Thompson and Holmes (1995) and Roche et al. (2013). In P1, the strongest response to the extra concentrate fed was for the GF6 cohort and the PL6 and WL6 cohorts (Table 7). Underlying causes could be that the Friesian genotype is genetically inclined to produce more milk, and the PL cows struggled in early lactation to reach a normal milk peak when they were recruited into the trial (at 9–15 DIM), with the extra concentrate from then onwards helping them reach a normal peak. Cabezas-Garcia et al. (2021) reported cows normally achieving 85–90% of their peak at ~2 weeks in milk (when we recruited cows into the trial), and their milk production peak approximately 5–7 weeks in milk (which was in P1 for cows in our trial). We postulate that the smaller cows in WL might suffer more competition from bigger cows in the herd, reducing opportunities to feed, and that impact might have been lessened with the additional concentrate fed. Cross-breed cows have a lower proportion of Friesian genotype than do GF, and the GC6 cohort had the lowest milk response, most pronounced in P1 (Table 7). The low response of the GC6 cows supports the general consensus that cross-breed cows are better suited for low supplemental concentrate-input grazing systems.

The results from the 8-monthly herd test FPCM response investigation, which indicated that the L FPCM response (adjusted to a standard concentration of 4.00% fat and 3.40% protein) by the 6 kg cohorts generally was higher than the litre response, was expected because this is a reflection of the mean fat and protein concentration in the milk from the whole herd, which was 4.67% fat and 3.45% protein for the 2017–2018 season. However, higher milk-yield responses to feeding extra concentrate can cause some reduction of fat and protein concentration owing to higher rumen fermentation rates compared with forages (Auldist et al. 2013, 2014). We indeed found that with higher milk-yield responses, the difference in FPCM responses tended to diminish somewhat compared with litre responses. For all 6 kg cohorts, overall response in FPCM was greater than was the response in litre. The only exception was for WL, for which the mean FPCM tended to be just below the litre response, mostly owing to low fat and protein concentrations in P1.

Feeding time

Soca et al. (2014) found in a trail similar to our investigation, lower grazing time and lower DM intake from pasture (measured using the chromic oxide method) for cows grazing unrestricted pasture and receiving 6 kg than for cows receiving 3 kg of concentrate. The overall reduction in feeding time in this investigation for the 6 kg compared with the 2 kg cohorts was 37 min per cow per day, with the largest reductions in feeding time for 6 kg cohorts being in P1 with 47 min (Table 6). This is when the herd had access to high-quality spring grass similar in quality to concentrate, and hence with a small difference in rumen fill between these two feeds. When the average diet quality reduced in P2 and P3 (the forage became more rumen filling than was concentrate), the difference in feeding time dropped to 35 min in P2 and 29 min in P3 (Table 6). Using change in feeding time (minutes per cow per day) as proxy for substitution of pasture intake, the feeding time dropped, on average, overall for the 6 kg cohorts with 9.3 min extra per kilogram DM concentrate. As reported in Bargo et al. 2003, this is in the range found by Arriaga-Jordan and Holmes (1986), Kibon and Holmes (1987) and Bargo et al. (2002), who reported a mean grazing-time reduction per extra kilogram DM concentrate supplemented of 9.9 min, ranging between 8.3 and 12.3 min. The largest reduction in feeding time overall was for the GC6 cohort (58 min, Table 6), and in P1 with a low milk response, the feeding time for GC6 declined 65 min (Table 7), resulting in the lowest feeding to milk ratio of −89.8 (Table 7). Generally, the reduction in feeding time and the ratio for feeding to milk diminish as lactation progresses (Table 6, All 6–2, P1, P2 and P3), partially also due to the mid- to late lactation cow directing some of the energy intake towards recovery of body reserves as evidenced in this study by the 6 kg cohorts gaining more weight than the 2 kg cohort.

Rumination and resting time

Differences in rumination time exist, but they are generally small. This is not unexpected, because even though concentrate triggers less rumination per kilogram DM than does forage, and some of that forage is substituted, a total increase in DM intake when feeding more concentrate moderates that impact (Mertens 1997; Bargo et al. 2002). None of the results for rumination time in the 6 kg cohorts was significantly different from the others (Tables 3, 4, 5). The longest rumination times were recorded for P2, the period with the highest NDF concentrations (i.e. the lowest diet quality), and as a reflection P2 also recorded the shortest resting time. The MooMonitor+ sensors report only feeding, rumination and resting time, which leaves unreported for the daily time budget of a cow (1440 min) the not classified time, which cows spend on behaviours such as walking (mostly between dairy and paddocks), socialising, and drinking water. The unclassified time was 117 min/cow.day over the trial period (not reported), with hardly any variation between cohorts. Hence, the increase in resting time for the 6 kg cohorts was due to an overall decrease in the combined time for feeding and ruminating. Resting time (i.e. no activity) as reported by MooMonitor+ is different from the resting time, i.e. referring to lying time, often reported for cows in housed systems. Resting time for the 6 kg cohorts was higher with a mean of 29 min per cow per day overall (Table 6). The mean resting time within the 6 kg as well as within the 2 kg cohorts varied little within each period. The overall resting time was 345 min (5.75 h) for 2 kg cohorts and 374 min (6.25 h) for 6 kg.

BW

Bodyweight changes were similar for both treatments, but all 6 kg cohorts gained more weight than did 2 kg cohorts. In an invited review, Roche et al. (2009) reported correlations between BW and body condition score (BCS). Enevoldsen and Kristensen (1997) investigated the correlation between BCS and BW and stated that a 50 kg increase in BW with each unit of increase on the 5-point BCS scale is a generally accepted rule of thumb in Denmark (p. 1995). Using BW as proxy for BCS in this trial with only multiparous cows, the difference in BW gain corroborated that feeding a higher amount of concentrate can also be used strategically to gain BCS in mid- to late lactation when need be. However, if cows have reached their BCS target, a reduction in the amount of supplemented concentrate could be warranted to avoid over-conditioning these cows; this could be investigated in future research. BCS recovery when not lactating is less efficient, with the same BCS gain requiring about 26% extra energy (Freer et al. 2007). And intending for a cow to gain a mean of 23 kg BW during the dry period would require an additional 3 kg DM intake per day over a period of 45 days, which is difficult to achieve under current dry period conditions and diets for spring-calving pasture systems in Oceania. However, body condition gain in late lactation should be monitored, and the concentrate supplementation reduced if BCS moves above target levels.

General

All 6 kg cohorts spent less combined time feeding and ruminating, resulting in an increase of resting time. Whereas understanding impacts of changes in resting time on production and feeding behaviour of dairy cows in pasture-based systems warrants further research, we postulate that with the average 360 min (6 h) resting time per cow per day, that resting time was not limiting production in this trial. We postulate that for cows with production levels and diets as in this trial, there is still some scope to adjust transition management of those cows in a direction that increases the peak milk production, and, consequently, the appetite of the cows, converting some of the resting time to grazing and rumination time.

Feeding a higher amount of concentrate to lactating dairy cows inherently comes with increased inefficiencies such as those associated with wastage of feed, partitioning of nutrients, and NDF digestibility (Roche et al. 2017). In this trial, we postulate that the following energy flows from DM intake may have occurred. The mean marginal response for all 6 kg cows was 0.88 L of milk (ME requirement 5.85 MJ/L) per 1 kg DM of extra concentrate fed, which, on the basis of its energy density, requires ~0.50 kg DM of that concentrate. Feeding-time reduction was 37 min, which infers a reduced pasture intake (substitution) of 0.28 kg DM assuming 1.8 kg DM pasture intake per hour feeding time (on the basis of the daily feeding time in this trial and findings by Bargo et al. (2002)). The additional weight gain for the 6 kg cohorts was 23 kg, which is approximately a BCS score of 0.5 in the 8-point Australian BCS scoring system, which requires over the duration of this trial about 0.10 kg DM per day to achieve. Combining the marginal response in milk, reduced pasture intake, and extra BW gain, this could account for a total of 0.88 kg DM for each kg DM concentrate fed, leaving 0.12 kg unaccounted for, which could be labelled ‘loss owing to inefficiencies’.