Increased interest in precision agriculture and use of monitoring equipment on farms necessitates an in-depth understanding between the available data, animal biology and environmental factors. Monitoring of energy balance is an important aspect of dairy cow production. High producing cows are usually unable to consume adequate feed to meet the requirements for maintenance and milk production during early lactation (Patton et al., Reference Patton, Kenny, McNamara, Mee, O'Mara, Diskin and Murphy2007; Weber et al., Reference Weber, Hametner, Tuchscherer, Losand, Kanitz, Otten, Singh, Bruckmaier, Becker, Kanitz and Hammon2013). Hence, during this period cows mobilise energy from body fat which leads to negative energy balance (NEB) and loss of body condition. Early lactation energy deficit is normal in many mammals, however, in cattle rapid losses or prolonged periods of body energy deficit have negative effects on milk yield, cow health and reproduction (Chagas et al., Reference Chagas, Bass, Blache, Burke, Kay, Lindsay, Lucy, Martin, Meier, Rhodes, Roche, Thatcher and Webb2007; Leroy, et al., Reference Leroy, Vanholder, van Knegsel, Garcia-Ispierto and Bols2008; Aguilar-Perez et al., 2009). It is, therefore, important to monitor energy balance in order to ensure appropriate management of dairy cows.
Energy balance may not be directly measurable on farm but is usually monitored through change in body condition score (BCS: Roche et al., Reference Roche, Friggens, Kay, Fisher, Stafford and Berry2009; Friggens et al., Reference Friggens, Disenhaus and Petit2010). The NEB that occurs during early lactation is reflected through BCS loss (Pryce et al., Reference Pryce, Coffey and Simm2001). Besides BCS change, energy balance can be estimated based on changes in body weight and BCS (NRC, 2001; Banos et al., Reference Banos, Coffey, Wall and Brotherstone2006) or based on the differences between energy intake and energy requirements for maintenance and milk production (Patton et al., Reference Patton, Kenny, McNamara, Mee, O'Mara, Diskin and Murphy2007; Aguilar-Perez et al., 2009). Proxy measures such as BCS are useful for quick on farm assessment, however, estimates of actual energy balance are necessary for a better understanding of physiological changes.
Traditionally, accelerometers have been used to detect oestrus using within-animal variation in activity (Firk et al., Reference Firk, Stamer, Junge and Krieter2002; Lovendahl & Chagunda, Reference Lovendahl and Chagunda2010; Palmer et al., Reference Palmer, Olmos, Boyle and Mee2010). Monitoring cow activity has also been used to study animal behaviour in relation to health and welfare status (O'Callaghan et al., Reference O'Callaghan, Cripps, Downham and Murray2003; Barrientos et al., Reference Barrientos, Weary, Galo and von Keyserlingk2011; Thorup et al., Reference Thorup, Munksgaard, Robert, Erhard, Thomsen and Friggens2015). O'Callaghan et al. (Reference O'Callaghan, Cripps, Downham and Murray2003) and Thorup et al. (Reference Thorup, Munksgaard, Robert, Erhard, Thomsen and Friggens2015) used accelerometers in early detection of lameness while Barrientos et al. (Reference Barrientos, Weary, Galo and von Keyserlingk2011) linked lying behaviour of dairy cows to presence and absence of deep-bedded stalls. However, if activity data were used across farms, it would be crucial to account for the variation that arises from environmental and other bio-physical factors. An example of this would be the comparison of cow feeding time and feed barrier design in different buildings.
The current study was undertaken to determine the association between energy status and cow activity monitored using accelerometers. This paper explores the use of accelerometer data to improve monitoring of energy balance in dairy cows and examines the interaction of three factors; cow energy status, energy content of ration and cow genetic merit.
Materials and methods
Data were obtained from a database available at Scotland's Rural College (SRUC) Dairy Research and Innovation Centre, Crichton Royal Farm, Dumfries, Scotland which was compiled as part of a long-term Langhill herd genotype by environment study. The herd consisted of Holstein Friesian cows from two genetic lines (Select (S) and Control (C) selected based on genetic merit for kilograms milk fat plus protein. Experimental design of the long-term study has previously been described in detail by Pryce et al. (Reference Pryce, Nielsen, Veerkamp and Simm1999), Chagunda et al. (Reference Chagunda, Ross and Roberts2009) and Ross et al. (Reference Ross, Chagunda, Topp and Ennos2014). The experiments were carried out in accordance with the U.K. Animals (Scientific Procedures) Act, 1986 and associated guidelines and was approved by SRUC Animal Experiments Committee.
Strict management protocols were operated both within and between systems. Cows were housed in the same building and managed by the same staff. Within a system, one complete diet was offered to all cows irrespective of milk yield and stage of lactation. The complete diet was offered at 1.05% of daily requirement and refusals removed daily. The diets had target crude protein (CP) and metabolisable energy (ME) content of 185 g/kg DM and 12.3 MJ/kg DM for the BP diet while it was 180 g/kg DM, and 11.5 MJ/kg DM for the HG diet, respectively. Table S1 in the online supplementary file provides details on the feed chemical composition of rations and target milk production.
Animals within S and C genetic lines were allocated to two different feeding systems, namely, by-product (BP) and a home-grown feeding systems (HG), making four experimental groups comprising two production systems (BP and HG) and within each feeding system two genetic lines (S and C). The four experimental groups constituted the four production systems herein referred to as BPS, BPC, HGS and HGC. Each production system was comprised of approximately 50 cows in their first 3 lactations. At the end of their third lactation the cows were replaced by heifers due to calve within 2 months. If there were no suitable replacements then cows remained on the system for an additional lactation (Roberts and March, Reference Roberts and March2013).
Data recording and management
Data collected between December 2012 and February 2013 were retrieved from the SRUC Langhill database using the Microsoft SQL Server Management Studio 2008. The total period used was three months during which time all the cows were housed for winter. Cows included had at least 21 d from which daily activity and eating time were recorded. A total of 8602 records from 124 cows between their first and fourth lactations were available for analysis. Cows in their first, second and third or more lactation accounted for 34, 31 and 35% of the study population, respectively. There were 33, 37, 31 and 23 cows from BPC, BPS, HGC, and HGS, respectively.
Herd management followed strict disease control management where vaccination and routine treatments such as deworming and hoof trimming were followed. There were also on-going checks for mastitis and infected animals were treated accordingly. All vaccinations and treatments were carried out by qualified veterinary surgeons or experienced farm staff. A veterinary surgeon visited the farm on a weekly basis for routine veterinary work, mainly related to fertility. If required, the veterinary surgeon would also visit the farm within an hour to attend to difficult calvings or other urgent veterinary issues. Routine foot-trimming was once every 6 months and cows walked through a footbath containing copper sulphate twice a week. Severely lame animals were examined as soon as possible by the head dairyman otherwise the veterinary surgeon visited fortnightly and cows locomotion scored 4 and above were walked around an enclosure and those considered lame were examined and any findings recorded and loaded to the database. The locomotion score was on a scale of 1–5. Sick animals were isolated and kept in a sick bay where they received treatment and were returned to the feeding groups upon recovery. Hence, the data used in this study are from cows that were considered healthy.
The data collected included animal identification, genetic group, feeding system and lactation number as well as calving date, milk yield, weights and body condition score (BCS), activity and feeding duration. Milk yields of individual cows were recorded at each milking and individual cow milk samples taken weekly for analysis of fat, protein and somatic cell contents. Live weights were measured after each milking. Body condition score was estimated weekly using the tail head system with a score of between 0 and 5 (with 0 being thinnest and 5 the fattest) (Mulvaney, Reference Mulvaney1977). The individual feed and water feed intake was recorded on 3 d out of six using Hoko gates (Insentec BV, Marknesse, The Netherlands). Individual time budgets (weekly averages of duration of lying and standing, motion index, number of steps and lying bouts) were monitored in cows using accelerometers (IceQube Sensors®, Icerobotics Ltd, UK).
Weekly body energy content (BEC) was calculated using weights and body condition score according to formulae described by Banos et al. (Reference Banos, Coffey, Wall and Brotherstone2006) as summarised in the online Supplementary file. Energy corrected milk yield was calculated using the formula reported by Sjaunja et al. (Reference Sjaunja, Baevre, Junkkarinen, Pedersen and Setala1990) given as:
Kg ECM = kg milk × (0:25 + 0.122 × Fat % + 0.077 × Protein %)
Data analysis
Data were analysed using descriptive statistics, crosstabs, frequencies and mixed models using Statistical Analysis System (SAS 9.3). All data on milk yield, fertility, body energy content (BEC), motion index, number of steps, standing, lying and feeding durations were subjected to the generalised mixed linear model (GLIMMIX) procedure of SAS 9.3 where differences in the response variables were determined between the feeding systems, genotypes and other management practices. The normally distributed data on milk yield were analysed using generalised mixed linear models (GLMM) with a normal error distribution and a log link function, while the data on activity and BEC were analysed using GLMM with negative binomial error distribution and a logit link function. A negative binomial error distribution was opted for in the count data analysis as a Poisson error distribution resulted in over-dispersion. Spearman correlation analysis was used to determine the relationship between milk yield, BEC, cow activity and feeding duration.
Results
The results showed that there were differences in productivity, energy status and activity of the cows associated with feeding system as well as genotype. Cows on by-product feeding system had significantly (P < 0.05) higher average daily milk yield and body energy content (BEC) than cows on home-grown feeding system (Table 1). Both within the by-product and home feeding systems, there was significant difference in milk yield between select (higher yield) and control cows. However, there was no significant difference in energy corrected milk yield (ECM) between control cows on by-product diet and select cows fed home grown feeds. This suggests that control cows were able to mobilise energy from by-product diet which had relatively higher energy similar to select cows on home-grown feed. The results can be related to the higher (P < 0.05) feed intake by select cows on home grown diet compared to control cows on by-product diet. This higher feed intake probably enabled mobilisation of energy similar to the control cows on by-product feed. Cows on by-product diet generally had lower feed intake and higher BEC than cows on homegrown diet (Table 1).
Average daily body energy content, milk production and activity of high and average genetic merit cows on either home grown or by-product feeds
a,b Means with different superscripts within the same row are significantly different (P < 0.05); *BPC, by-product control; BPS, by-product select; HGC, home-grown control; HGS, home-grown select; ECM, energy corrected milk; BEC, body energy content.
Although select cows on by-product diet and control cows on homegrown diet had similar BEC, their milk yield was significantly different suggesting that select cows on by-product diet had the ability to mobilise more nutrients towards milk production than control cows on homegrown diet. This is further confirmed by the fact that control cows on homegrown diet had similar BEC compared to select cows on the same diet, but there was a significant difference (P < 0.05) in their milk yield. This is most likely demonstrating the interaction between feeding system and genotype.
Cow activity varied with the feeding system used. Cows on home grown feeds were significantly (P < 0.05) more active than cows on by-product feeds as indicated by higher motion index and number of steps per day (Table 1). Within feeding systems, high genetic merit cows were more active than average genetic merit cows. All the cows had a similar number of lying bouts as well as a similar minimum lying bout duration. However, cows on by-product feeds had significantly (P < 0.05) longer duration of maximum lying bouts (2.47 h) than cows on home grown feeds (2.03 h). Cows on home grown feeds also stood almost 1 h longer than those on by-product feeds for both genetic groups. The activity data compared well with feeding duration data, where cows on home grown feeds also spent more time feeding than cows on by-product feeds. However, there was no significant difference in the feeding time between select and control cows on by-product and homegrown diets, respectively. This aspect was also reflected in that these cows had similar average daily BEC content.
Production system had an effect on energy balance, cow activity and milk production. The results showed that the same cows that had negative balance were also those that had high milk yield, while cows in positive energy balance generally had low milk yield. This was the case in both feeding systems (Table 2) However, there was an interaction between feeding system and energy balance with regard to activity and feeding duration. Cows in NEB on the by-product feeding system were significantly less active than cows in positive energy balance (PEB) while on home-grown feeding systems all cows had similar activity. Control cows on by-product diet that were in NEB had longer feeding duration (4.9 h) than the same cows in PEB (4.2 h). Select cows on homegrown diet that were in NEB had a shorter duration (5 h) of feeding than the same cows in PEB (6 h).
The cow activity and milk yield of cows in positive and negative energy balance in high and average genetic merit cows on either home grown or by-product feeds
a,b Means with different superscripts within the same row are significantly different (P < 0.05); *BPC = by-product control; BPS = by-product select; HGC = home-grown control; HGS = home-grown select
Generally, cow activity, energy status and feeding durations varied widely within production systems as evidenced by coefficients of variation ranging from 38 to 49%. This may suggest that there were other differences in cow activity, energy balance and feeding duration associated with individual cows within systems. Correlation analysis showed that both milk yield and BEC were significantly positively correlated to some aspects of cow activity and feeding duration (P < 0.05, Table 3). Milk yield was significantly correlated with motion index and number of steps per day but not lying time and feeding duration. There was a negative correlation between milk yield and motion index as well as number of steps per day. This implies that cows with higher milk yield were less active than cows with lower milk yield. There was no significant correlation between milk yield and feeding duration and this could reflect the interaction between genetic merit and milk yield. Regardless of feeding duration and BEC, select cows seemed to maintain higher milk yield than control cows. It is likely that select cows were able to mobilise additional nutrients from body reserves to sustain milk production. This is further reflected in that there was no significant correlation between BEC and milk yield.
Correlation between milk yield, body energy content, cow activity and feeding duration in dairy cows under home grown and by-product feeding systems
Figures within parentheses show the P value.
BEC was positively and negatively correlated to lying time and feeding duration, respectively. Cows that had longer lying durations had higher BEC than those with shorter lying durations. Cows with higher BEC also had shorter feeding durations. Standing durations were positively correlated with feeding duration indicating that part of the time that cows stood was spent feeding. However, a correlation analysis by feeding system showed some differences in some of the relationships in Table 3. For instance, online Supplementary Tables S2 and S3 show that BEC was negatively correlated with milk yield in cows on by-product diet while it was positively correlated to the same trait in cows on homegrown. This result further shows the interaction between genotype and feeding systems.
Discussion
Both high and average genetic merit cows had distinct and consistent patterns of milk production and activity which were related to the feeding systems. The results showed an interaction between genetic merit and feeding system where cows of average genetic merit fed BP diets with higher crude protein (CP) and ME produced higher milk yields than high genetic merit cows fed HG diets. These results are similar to earlier findings on the same herd by Pollott and Coffey (Reference Pollott and Coffey2008) and other herds (Nielsen et al., Reference Nielsen, Pedersen and Toivonen2003; Horan et al., Reference Horan, Dillon, Faverdin, Delaby, Buckley and Rath2005; Windig et al., Reference Windig, Beerda and Veerkamp2008) and demonstrate that productivity is a function of both genetic merit and feeding systems. Such responses in productivity are related to feeding efficiency which is a ratio of output to intake (Brody, Reference Brody1945). Apart from feed quality, feed efficiency is also dependent on genotype and physiological state (Blake and Custodio, Reference Blake and Custodio1984). The genotype effect is through the genetically determined potential for milk yield as shown by higher milk yields in cows of high genetic merit.
The results also showed an association between production system, energy status and cow activity. HGS were the most active while BPC cows were least active. This could be associated with the quality of the feeds in the systems and the genetic merit of the cows. HG feeds were less dense in terms of metabolisable energy and crude protein than BP feeds. Hence high genetic merit cows on HG spent more time standing and feeding to increase feed intake to match with their high milk production while average genetic merit cows on BP feeds were less active as their feed quality was better and milk production relatively lower. However, there was no significant difference in time spent standing and lying. Heublein et al. (2017) found similar results in dairy cattle in pasture-based systems where cows that are supplemented with concentrates are less active than those that are solely on pasture.
The results showed that there were differences in cow activity that could be attributed to body energy status. Cows fed diets with relatively lower ME and CP were more active and spent more time standing and eating than those on diets with higher ME and CP. Although ME and CP levels were different, these differences may not have been large in relation to much wider differences that exist in different production systems. However, the diets represent two different acceptable feeding levels that demonstrate that accelerometers can potentially detect and be used to appropriately address subtle management differences. This suggests, therefore, that when such a technology is used as a tool for management it may be able to quickly detect even more adverse problems that may otherwise take too long to detect using conventional management techniques.
The high activity may also imply that the cows were restless. Their greater activity was probably to increase feed intake to counter negative energy balance. Hence, there is an association between energy status and cow activity that could be determined using accelerometers. There is need for further analysis to investigate more details on cow activity in terms of time budgets such as feeding bouts and their durations. This may be useful in precision farming where feeds may need to be provided in accordance with specific animal behaviour and requirements.
In conclusion, the implication of the findings is that activity monitoring can be used beyond oestrus detection and animal welfare management to include diet quality and body energy status. When setting up the baseline activity for each cow, there may be a need to account for the cow's feeding system. Secondly, if cow activity is being used to evaluate building design then feed system, milk yield and the general well-being of the animals all need to be considered. This could be achieved through either having integrated or complementary technologies. For example, a combination of data from cow activity meters and BCS imaging cameras may further improve management approaches in dairy cows.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0022029922000267
Acknowledgements
The authors acknowledge Flanders International Cooperation Agency (FICA), Scotland's Rural College (SRUC), The Roslin Institute and Biotechnology and Biological Sciences Research Council (BBSRC {BB/J004316/1}) for financial and technical support.
