MTT is publishing its research findings in two series of publications: MTT Science and MTT Growth. The MTT Science series includes scientific presentations and abstracts from conferences arranged by MTT Agrifood Research Finland. Doctoral dissertations by MTT research scientists will also be published in this series. The topics range from agricultural and food research to environmental research in the field of agriculture. MTT, FI-31600 Jokioinen, Finland. email julkaisut@mtt.fi 28 Dairy cow behaviour in relation to health, welfare and milking Doctoral Dissertation Jutta Johanna Kauppi MTT CREATES VITALITY THROUGH SCIENCE www.mtt.fi/julkaisut Academic Dissertation: To be presented, with the permission of the Faculty of Veterinary Medicine of the University of Helsinki, for public examination at Viikki Kampus, Biokeskus 3, lecture room 2402, Viikinkaari 1, on 10th October 2014, at 12 noon. Helsinki 2014 28 Dairy cow behaviour in relation to health, welfare and milking Doctoral Dissertation Jutta Johanna Kauppi Supervised by Satu Raussi, PhD, Director of the Finnish Centre for Animal Welfare Jukka Ahokas, Professor Department of Agrotechnology, University of Helsinki, Finland Joop Lensink, PhD, Director, ISA Institut Superiéur d’Agriculture Lille, France Supervising Professor Anna Valros Department of Production Animal Medicine, Faculty of Veterinary Science, University of Helsinki, Finland Reviewed by Lena Lidfors, Professor Department of Animal Environment and Health, Swedish University of Agricultural Sciences, Sweden David Arney, Associate Professor, Institute of Veterinary Medicine and Animal Sciences, Department of Nutrition, Estonian University of Life Sciences, Estonia Opponent Knut Bǿe, Professor, Norwegian University of Life Sciences, Department of Animal and Aquacultural Sciences, Norway ISBN 978-952-487-556-1 (Print) ISBN 978-952-487-557-8 (Electronic) ISSN 1798-1824 (Printed version) ISSN 1798-1840 (Electronic version) URN http://urn.fi/URN:ISBN:978-952-487-557-8 www.mtt.fi/mtttiede/pdf/mtttiede28.pdf Copyright MTT Agrifood Research Finland Jutta Johanna Kauppi Distribution and sale MTT Agrifood Research Finland, Media and Information Services, FI-31600 Jokioinen, e-mail julkaisut@mtt.fi Printing year 2014 Cover Reeta-Maija Siipola Printing house Tampereen Yliopistopaino Juvenes Print Oy MTT SCIENCE 28 3 Dairy cow behaviour in relation to health, welfare and milking Jutta Johanna Kauppi MTT Agrifood Research Finland, Animal Production Research, FI-31600 Jokioinen jutta.kauppi@mtt.fi Abstract Animal welfare includes both physio-logical and mental health and is af-fected by several external and inter- nal factors. In dairy cows, human care and the ability of the cows to cope with dai- ly challenges are the most significant fac- tors. The research presented in this dis- sertation focused on cow behaviour and aspects impacting on behavioural chang- es during mastitis and milking. To deep- en knowledge of the relationship between the behaviour and health of cows and asso- ciated detection methods, we experimen- tally manipulated the health status of dairy cows through mastitis induction. Moreover, we tested and validated a thermal infrared camera for recording the udder skin tem- perature, which could be helpful in the case of early mastitis detection. Furthermore, we examined the relationship between cow behaviour and milking in herringbone and automatic milking systems. We established that cow behaviour changed during masti- tis. The most apparent changes were in ly- ing, eating and stepping behaviour. It was shown that inflammation affected the cow health status, which changed the behav- ioural priorities. In contrast to our expecta- tions, visual signs in the udder and changes in milk composition occurred only 2 hours post-challenge, while clinical and behav- ioural changes were first recorded 4 hours post-challenge. However, changes in lying and restlessness behaviours were promis- ing indicators for detecting signs in cows exposed to mastitis. The transient increase in body temperature of cows with experi- mentally-induced clinical mastitis was suc- cessfully detected by udder skin tempera- ture detection with the help of a thermal camera. The udder skin temperature rose simultaneously with the rectal temperature. However, local inflammatory changes in the udder, appearing earlier than the rec- tal temperature increase, were not detected with udder skin temperature measurement by using the thermal infrared camera. Re- garding cow behaviour as an indicator of the success and quality of the milking pro- cess, we found that half of the deviations occurring during automatic milking orig- inated from cow behaviour, such as cows kicking, lifting their legs and moving dur- ing milking, and had their origins in ma- chine failures. To conclude, cow behaviour can be used as an indicator for detection of mastitis and (un)successful milking. How- ever, effectively functioning human-animal- technology interactions should be studied more in the future in order to enhance hus- bandry practices that can improve animal welfare. 4 MTT SCIENCE 28 Lehmän käyttäytymisen yhteys eläimen terveydentilaan, hyvinvointiin ja lypsyyn Eläimen hyvinvointi muodostuu sekä fyysisestä että psyykkisestä tilasta ja siihen vaikuttavat lukuisat eläi- men sisäiset mutta myös ulkoiset tekijät. Lypsylehmän hyvinvointiin vaikuttavista tekijöistä tärkeimpiä ovat karjanhoitajan hoitotoimenpiteet sekä lehmän kyky so- peutua päivittäisiin haasteisiin. Tässä väi- töskirjatyössä keskitytään lehmän käyttäy- tymiseen sekä käyttäytymisen muutoksiin, jotka liittyvät lypsyyn tai lehmän tervey- dentilan heikkenemiseen. Keskityin selvittämään lehmän käyttäy- tymisen, hyvinvoinnin sekä terveydenti- lan tunnistamiseen liittyvien menetelmi- en välistä vuorovaikutusta. Vaikutimme keinotekoisesti lehmän terveydentilaan ja saimme näin uutta tietoa siitä, kuinka utaretulehdus muuttaa lehmän käyttäy- tymistä yksilön terveydentilan heiketessä. Tämän lisäksi testasimme ja validoimme samanaikaisesti uuden teknologian (infra- punakameran) käyttöä utareen pintaläm- pötilan tunnistamisessa. Tavoitteenamme oli tehokkaampi alkavan utaretulehduksen tunnistaminen karjoissa. Tutkimme myös lehmän käyttäytymisen ja lypsyn onnistu- misen välisiä yhteyksiä sekä kalanruoto- että automaattilypsyssä. Lehmä muuttaa käyttäytymistään tervey- dentilan heiketessä, sairastuessaan uta- retulehdukseen. Selkeimmät muutokset havaittiin makuu- ja syömiskäyttäytymi- sessä - sekä levottomuus (jalkojen nostelu) käyttäytymisessä. Lehmän käyttäytymi- sen prioriteetit muuttuvat sen sairastuesa utaretulehdukseen. Toisin kuin odotim- me, maidon koostumuksen silmin näh- tävät muutokset voitiin tunnistaa jo kak- si tuntia utaretulehdukseen sairastumisen jälkeen, kun taas kliiniset- sekä käyttäy- tymisen muutokset voitiin tunnistaa vasta neljä tuntia lehmän sairastumisen jälkeen. Lehmän makuukäyttäytyminen ja levot- tomuus (jalkojen nostelu)käyttäytymisen muutokset osoittautuivat tutkimukses- samme lupaaviksi indikaattoreiksi tun- nistettaessa utaretulehdukseen sairastuvia lehmiä. Lehmän ruumiinlämpötilan ly- hytaikainen nousu kokeellisesti aiheute- tun utaretulehduksen aikana tunnistettiin infrapunakameralla utareen pintalämpöti- laa mittaamalla. Utareen lämpötila kohosi samanaikaisesti lehmän rektaalilämpötilan kanssa. Kuitenkin paikalliset tulehdus- muutokset utareessa ilmenivät ennen rek- taalilämpötilan nousua, eikä niitä kyetty tunnistamaan infrapunakameran avulla. Puolet automaattisen lypsyn aikaisista häiriötilanteista oli lehmän käyttäytymi- sen aiheuttamia (potkut, jalkojen noste- lu, liikehtiminen lypsyn aikana) ja leh- män käyttäytymisellä oli selvä vaikutus lypsytapahtuman onnistumiseen sekä lyp- syn kunnolliseen läpivientiin. Lehmän käyttäytymistä sekä sen muutos- ta voidaankin käyttää indikaattorina niin utaretulehduksen tunnistamisessa kuin myös lypsyn onnistumisessa. Jatkotutkimuksia tarvitaan, jotta ihmis- eläin-teknologia-vuorovaikutus saadaan mahdollisimman hyvin toimivaksi niin, että käytännön tilalla lehmän hyvinvoin- nin tarkkailu sekä terveydentilan heikke- nemistä ehkäisevät karjanhoidon ratkaisut saadaan tehokkaiksi. Näin voidaan ennal- taehkäistä lehmän terveydentilan heikke- nemistä ja edistää lehmän hyvinvointia. MTT SCIENCE 28 5 Lehm ja koiv Mää tahro olla lehm koivu al. Mää en tahro olla luav. Mää en tahro oppi uut taitto-ohjelma. Mää en tahro selvittä äit-tytär suhret. Mää en tahro viärä sitä kirjet posti. Mää en tahro soitta Kelan tätil. Mää en tahro muista yhtäkän pin-koori. Antakka mu olla lehm koivu al. Viäkkä mu väsyne nahk kamarim permanol, kakluni ette. The Cow and the Birch Tree I want to be a cow under the birch tree I don’t want to be creative I don’t want to learn a new software program I don’t want to sort out the relationships of the mothers and daughters I don’t want to take that letter to the post office I don’t want to phone that woman at the benefits office I don’t want to memorise any more pin numbers Take my tired hide and put it in front of the hearth Heli Laaksonen (Pulu uis 2000) (englanniksi kääntäneet Mark Phillips ja Christa Prusskij) 6 MTT SCIENCE 28 Acknowledgements Two persons made it possible for me to complete this thesis: the thesis super-visor, Satu Raussi, and Laura Hän- ninen. Without their timeless and invalua- ble effort, help and support, this would not have been possible. I am grateful to Professor Anna Valros for her guidance and advice during my PhD research. I am also grateful to Joop Lensink, ISA Lille Group, my remote supervisor, for his professional help and guidance and for taking such good care of me during the pleasant visits to Lille, France. I also thank him for inspiring discussions during my PhD research. I am grateful to Professor Jukka Ahokas at the Department of Agrotechnology. His help, advice and support during this long process has been considerable and highly important to me. I also want to express my deep gratitude to Professor Outi Vainio, who strongly sup- ported and guided me at the beginning of my PhD project. I sincerely thank my pre-examiners, Pro- fessor Lena Lidfors and associate profes- sor David Arney. Their invaluable com- ments significantly raised the standard of this work. I am warmly grateful to my superiors and colleagues at MTT Agrifood Research for positively supporting and encouraging me during the project. Special thanks to my research fellows, Mari Hovinen, Matti Pastell and Anna- Maija Aisla. It was such a pleasure to share research topics with you. I am also grateful to the staff at the Ani- mal Hygiene and Viikki Research Barn, as well as the staff of Suitia barn. I am ever so grateful for the enthusias- tic group spirit and the support I received from you, my working fellows around the Research Centre for Animal Welfare. I am also grateful to my tremendous- ly warm-hearted and supportive friends, Saynur, Mirva, Emilia and Piia. I highly appreciate the shared moments with you. I am warmly grateful to my parents, Riit- ta and Juhani: I thank you for supporting and helping me and my family patiently in so many ways during the last years. With- out you this would not have been possible. I thank my lovely daughter Alisa for her patience and reminding me that this is a project to be finalized in order to be able concentrate on more important issues in life: playing and having fun. This research was supported by the Facul- ty of Veterinary Medicine at the Univer- sity of Helsinki, MTT Agrifood Research Finland, the Finnish Ministry of Agricul- ture and Forestry. The work was complet- ed within the Research School for Animal Welfare 2006–2008. Thank you. MTT SCIENCE 28 7 List of original publications I. Siivonen, J., Taponen, S., Hovinen, M., Pastell, M., Lensink, B.J.,Pyörälä, S., Hänninen, L. Impact of acute clinical mastitis on cow behavior, Applied Animal Behaviour Science 132 (2011) 101–106. doi:10.1016/j.applanim. 2011.04.005 II. Hovinen M., Siivonen J., Taponen S., Hänninen L., Pastell M., Aisla A-M., and Pyörälä S. Detection of clinical mastitis with the help of a thermal camera, Journal of Dairy Science, 2008, 91, 12: 4592-4598. doi:10.3168/jds. 2008-1218. III. Kaihilahti J., Suokannas A. and Raussi S. Observation of cow behaviour in an automatic milking system using web-based video recording technology. Biosystems Engineering 2007, 96, 1: 91-97. doi:10.1016/j.biosystemseng. 2006.10.001. IV. Siivonen J., Pastell M., Aisla A-M., Jauhiainen L., Ahokas J., Vainio O. and Raussi S. Ef- fect of milking and management on cow behaviour. Submitted. Original articles have been printed with the kind permission of original scientific publishers. 8 MTT SCIENCE 28 Contents Acknowledgements ................................................................................................6 List of original publications ..................................................................................7 I Introduction .................................................................................................10 II Literature review ..........................................................................................12 Welfare definitions ........................................................................................... 12 Cow welfare and coping ................................................................................... 12 Cow behaviour .................................................................................................. 12 Cows and mastitis ............................................................................................. 13 Sickness behaviour ............................................................................................ 13 Cow behaviour and automatic milking ............................................................ 15 Technologies for health monitoring ................................................................... 15 III Aims of the study .........................................................................................16 The main research questions ............................................................................. 16 Specific aims of the study .................................................................................. 16 IV Materials and methods .................................................................................17 Animals and housing ......................................................................................... 17 Measurements ................................................................................................... 17 Clinical measurements, sampling ............................................................... 17 Follow-up of clinical signs .......................................................................... 17 Behavioural parameters .............................................................................. 18 Measuring technologies .................................................................................... 19 Video recordings ...........................................................................................19 Detecting udder skin temperature with a thermal camera ........................... 19 Statistical analysis .............................................................................................. 20 V Results ..........................................................................................................21 1. Can we identify a change between typical and sickness behaviour in cows exposed to mastitis? (Article I-II) .................................................................21 Clinical signs ..................................................................................................... 21 Changes in behaviour ........................................................................................ 22 Daily behavioural rhythm ................................................................................. 22 2. Can a thermal infrared camera be used in detecting temperature changes in the cows’ udder during mastitis (Article II)? ..........................................24 Infrared thermal camera and body temperature ................................................ 24 3. Changes and main characteristics of cow behaviour during milking (Articles III–IV) ...........................................................................................25 Changes in milk ................................................................................................ 25 Behavioural changes following a change between herringbone (HRB) and automatic milking systems (AMS) ..................................................................... 25 Cow behaviour and deviations in automatic milking ........................................ 25 Origin of failures in automatic milking ............................................................. 27 Problem cows during automatic milking ........................................................... 27 MTT SCIENCE 28 9 VI Discussion ....................................................................................................29 1. Can we identify a change between typical and sickness behaviour in cows exposed to mastitis? (Articles I–II) ......................................................29 Change in behavioural priorities and patterns ................................................... 29 Increased vigilance ............................................................................................ 30 Daily rhythm .................................................................................................... 30 Eating and rumination ..................................................................................... 31 2. Can a thermal infrared camera be used to detect temperature changes in the cows’ udder during mastitis (Article II)? ...........................................31 3. Changes in and main characteristics of cow behaviour during milking (Articles III–IV) ...........................................................................................32 Factors impacting on the success of automatic milking ..................................... 32 Cow behaviour in changing milking systems .................................................... 33 Management of automatic milking ................................................................... 33 VII Conclusions ................................................................................................. 34 1. Can we identify the change between typical and sickness behaviour in cows exposed to mastitis (Articles I-II)? ...................................................... 34 2. Can a thermal infrared camera be used to detect temperature changes in the cows’ udder during mastitis (Article II)? .......................................... 34 3. Changes and main characteristics causing deviations during automatic milking (Articles III–IV) .............................................................................35 4. Future implications for welfare studies ........................................................36 Questions requiring further studies ................................................................... 36 VIII Reference list ................................................................................................37 10 MTT SCIENCE 28 I Introduction The current trend in Europe is a constantly increasing farm and av-erage herd size. At the same time, the use of new technology in the barn and automation of certain activities is growing accordingly. For example, increasing num- bers of dairy farms are switching from conventional twice per day milking sys- tems to milking robots (automatic milk- ing system, AMS). Furthermore, addi- tional tools and devices can be added in order to automatically detect, for exam- ple, lameness (e.g. Pastell et al., 2008) or mastitis through the conductivity of the milk. Some of these tools do actually pro- vide information to stockpersons, allow- ing them to improve cow welfare. How- ever, the information provided by these tools is not always easy for stockpersons to interpret and is based on physiological (e.g. milk conductivity) or physical chang- es (e.g. weight pressure of feet on a mat). Measures might be taken too late in cer- tain cases, and more modern observation technology is needed to provide further online information and to observe chang- es in the behaviour of individual cows and welfare in the barn. The importance of monitoring cow welfare will increase as di- rect human–cattle contact during milking disappears with the change from conven- tional milking to automatic milking sys- tems (AMS). As daily human–cattle inter- action decreases, other ways of observing changes in the welfare of cows will become more prominent. Mastitis is one of the most common diseas- es of high-producing dairy cows, and acute mastitis is the most important welfare and economic problem in dairy herds (Halasa et al. 2007). A recent study by Heikkilä et al. (2012) demonstrated that it is most profitable to treat mastitic cows and keep them in the herd as long as healthy ones. This supports the fact that more inten- sive observations on both behavioural and clinical changes within the herd are need- ed to detect and react to ongoing changes in cow welfare. Cows that develop mastitis display both behavioural and clinical changes (Kemp et al. 2008, Rousing et al. 2004). Acute mastitis can cause motivational conflict in the behavioural priorities of a cow, and thus change the classical patterns of sick- ness behaviour. However, there is still a lack of knowledge concerning which pre- cise behaviours change first at the begin- ning of acute mastitis, and whether behav- ioural changes might serve as a tool for its early detection. In general, numerous factors affect the be- haviour and welfare of dairy cows, with human care and the production environ- ment being the most significant. This the- sis focuses on cow behaviour and the ca- pacity of cows to adapt to challenges faced in daily life, such as mastitis and milking. Previously, several studies have report- ed the effects of different dairy manage- ment regimes on cow behaviour. Total ly- ing time and the number of lying bouts, or standing in the lying area, have been used as indicators of cow welfare (Cook et al. 2005, Fregonesi and Leaver 2001, Ha- ley et al. 2000). In addition, changes in lying down and rising movements as well as cow comfort in resting have been used to detect changes in cow welfare (Lidfors 1989, Plesch et al. 2010). MTT SCIENCE 28 11 Motivational priorities play a key role in affecting behavioural responses. For ex- ample, fear competes with sickness, and fear-motivated behaviours take prece- dence over sickness-motivated behaviours (Aubert 1999). Sickness behaviour is con- sidered to be more an expression of the motivational state than a consequence of weakness (Miller 1964). Sick animals ex- hibit distinct behavioural patterns; a sick individual shows decreased activity, explo- ration, body care and sexual behaviour, as well as having a poor appetite (Inter- national Dairy Federation 1999). In se- vere cases, rumen contractions of sick cows slow down (Radostits et al. 2007). If sick- ness behaviour is prevented, recovery from disease is poorer (Johnson 2002). The re- search presented in this thesis investigat- ed the effect of mastitis on cow behaviour, which altered the normal behavioural pat- terns, thus resulting in changes in behav- ioural priorities (I). The results have improved our under- standing of how a cow modifies its behav- iour when infected with mastitis. It was found that the detection of changes in be- havioural patterns can serve as an addi- tional tool when detecting clinical mas- titis. A study by Fogsgaard et al. (2012) confirmed that cows do react to pain by altering their behaviour, including chang- es in milking. As milking in automatic systems is carried out by machines, the stockperson does not have information on changes in a cow’s behaviour during the milking process. I suggest that there exists an intensive interaction between the de- velopment of mastitis and changes in cow behaviour. Moreover, success in milking is essential to maintaining the health and welfare of cows, but also has an impact on cow behaviour. New technologies will continuously be in- troduced in barns in order to secure on- line information on cow welfare parame- ters on a long-term basis. However, there is still a lack of information on the precise behavioural changes that best describe the changing health status in cows and that can be used to detect illnesses as early as possible. The interactions between humans, ani- mals and technologies will become increas- ingly important, as individual cow welfare must be ensured, despite the reduced time available to stockpersons to devote to in- dividual animals. In the following literature review, I dis- cuss some of the factors that affect cow behaviour and welfare. This dissertation focuses on cow behaviour and aspects im- pacting on behavioural changes during mastitis and milking. To deepen knowl- edge of the relationship between behaviour and health, and the associated detection methods, we experimentally manipulated the health status of dairy cows by induc- ing mastitis and observed the behaviour- al evolution of cows during mastitis devel- opment (I). Moreover, we tested a thermal infrared camera regarding its potential in automatic mastitis detection (II). In addi- tion, we observed and compared cows dur- ing milking in herringbone and automatic milking systems in order to assess whether cows can easily transfer from one system to another as expressed through their be- haviour (III–IV). 12 MTT SCIENCE 28 II Literature review Welfare definitions Animal welfare comprises physiological and mental health and is affected by sev- eral external and internal factors (Dawk- ins 2004, Webster et al. 2004). The wel- fare of animals is typically addressed by posing three questions: is the animal func- tioning well, is the animal feeling well and is the animal able to live according to its natural behaviour (Fraser et al. 1997). On a Europe-wide scale, scientists have devel- oped new models based on a multi-criteria- based approach founded on the four main principles of animal welfare: good feeding, good housing, good health and appropri- ate behaviour (Botreau et al. 2009). This “Welfare Quality (WQ) concept” concen- trates on assessing welfare using animal- based criteria, interpreting how the ani- mal performs in its environment. Cow welfare and coping To maintain an optimal level of welfare, a cow always attempts to cope with its en- vironment by adjusting its behaviour ac- cording to the prevailing circumstances (Broom 1996). When coping fails, signs of poor welfare can occur (Broom 1996). The coping abilities of cows have been assessed using various methods describing the changes in normal, expected behaviour. Norring et al. (2012) and Drissler et al. (2005) reported the effect of the amount and type of bedding on lying behaviour in cows. Furthermore, the time budget and behavioural preferences of dairy cows were investigated by Munksgaard et al. (2005). Adequate rest (Haley et al. 2000, Munksgaard et al. 2005, Norring et al. 2012) and sleep (Ruckebush 1974) have proven to be essential for cow welfare, in addition to frequent feeding and an ade- quate water supply. Identified welfare crite- ria, such as “comfort around resting”, have been used by Plesch et al. (2010). Moreo- ver, a decreased total lying time and num- ber of lying bouts or prolonged standing in the lying area have been used as indi- cators of poor cow welfare (Cook et al. 2005, Fregonesi and Leaver, 2001, Haley et al., 2000). In addition, abnormalities in lying down and rising movements have been used to indicate changes in cow wel- fare (Lidfors 1989). Cow behaviour Cows can alter their behaviour based on their ability to cope with changes in their environment. For example, Albright et al. (1993) found that cows practice so- cial facilitation; they eat more when fed in groups compared to cows fed as individ- uals. Furthermore, they adapt their feed- ing speed according to the feeding sys- tem (Wierenga and Hopster 1991). Calves and cows (Camiloti et al. 2012, Norring et al. 2012, Tucker et al. 2009) prefer dry, soft and organic lying surfaces that they are familiar with, such as straw compared to sand material. Moreover, cows change their weight distribution by decreasing their rear leg movements while avoiding abrasions of the rear legs and the swollen udder (Chapinal et al. 2013). The lying time of cows in their time budg- et is a reliable and relatively consistent welfare indicator that can be followed. A study by Ito et al. (2010) demonstrated that cows spend on average approximately 11 h per day lying down. Lying time is the most prioritised cow behaviour when com- pared with other behaviours (Munksgaard et al. 2005). Cows do have a strong moti- MTT SCIENCE 28 13 vation to lie down (Jensen et al. 2005), but they avoid doing so if the lying surface is uncomfortable (Haley et al. 2001). Sever- al studies have reported changes in lying time to be a relevant indicator of cow wel- fare (Fregonesi and Leaver 2001, Medra- no-Galarza et al. 2012). Fregonesi et al. (2007) found that overstocking reduces the lying time in cows and increases the competition in stalls through cows dis- placing each other. Cows and mastitis Mastitis has a detrimental effect on cow health and welfare because it causes sys- temic physiological changes, such as an increased body temperature, as well as impacting on milk quality by elevating the somatic cell count (Bannerman et al. 2005). Moreover, it affects cow behav- iour in various ways: increasing the hock- to-hock distance (Kemp et al. 2008), in- creasing restlessness behaviour (Rousing et al. 2004) and decreasing lying behav- iour (Haley et al. 2000). Medrano-Galarza and colleagues (2012) observed that mas- titis cows display significant differences in behaviour during mastitis days, including a higher frequency of kicks, lifts and steps per minute in milking during the most se- vere phase of mastitis. Reactivity and restless behaviour during milking are connected with the discomfort caused by mastitis (Medrano-Galarza et al. 2012). In addition, weight shifting de- creased in the rear legs when symptoms of inflammation were at their worst (Chap- inal et al. 2013). The weight distribution between the legs and hock-to-hock dis- tance appear to change during the onset of mastitis (Kemp et al., 2008), representing promising indicators of changes in the cow health status (Kemp et al. 2008). When exposed to mastitis, cows are af- fected through three levels: 1. the host (i.e. breed, parity and production level), 2. the cause (i.e. pathogen, trauma and chem- ical) and 3. the environment (i.e. season and management) (Polat et al. 2010). The risk of contracting mastitis is highest dur- ing the first month post-partum due to changes in udder adaptation and defence mechanisms and the negative energy bal- ance (Cai et al. 1994). The effect of automatic milking systems (AMS) on cow health and welfare and their use in detecting changes in the milk- ing process and management has recent- ly attracted wide research interest (Hov- inen et al. 2009, Hovinen and Pyörälä 2011, Jacobs and Siegford 2012). Kete- laar-de Lauwere and colleagues (1999) re- ported the impact of automatic milking on the milking, feeding and resting be- haviour of cows. In automatic milking sys- tems, the inspection of foremilk is fully automated and is carried out by a milking robot. Abnormal milk is detected by col- lecting and modelling the milking data from automatic milking systems (Kam- phuis et al. 2008 a and b, 2010, De Mol and Ouweltjes 2001). However, no perfect solution or technological combination for mastitis detection has been established to date. Since mastitis causes marked losses in farm income due to premature culling (Heikkilä et al. 2012), further preventive actions need to be taken to diminish losses at the farm level and to increase longevity in dairy herds. Therefore, effective masti- tis detection systems are needed, and it is essential that both clinical and behaviour- al changes in cows are effectively detected. Sickness behaviour Sickness, according to the Encyclopae- dia Britannica (2014), is considered as: “A harmful deviation from the normal structural or functional state of an organ- ism. A diseased organism commonly ex- hibits signs or symptoms indicative of its abnormal state. Thus, the normal condi- tion of an organism must be understood in order to recognize the hallmarks of dis- ease. Nevertheless, a sharp demarcation 14 MTT SCIENCE 28 between disease and health is not always apparent.” Changes in the behaviour of sick individ- uals are used by veterinarians and stock- persons in the diagnosis of disease (Broom 2006). Sick animals exhibit distinct behav- ioural patterns, including decreased ac- tivity, exploration, body care and sexu- al behaviour, as well as a reduced appetite (Gonzales et al. 2008, Urton et al. 2005, Huzzey et al. 2007). Acute clinical mastitis causing sickness in cows initiates a motivational conflict with respect to their behavioural priorities. Due to sickness, a cow will be motivated to lie down in order to rest and enhance the healing process by minimizing the con- sumption of body energy reserves (John- son 2002, Weary et al. 2009). However, lying down may cause uncomfortable or painful feelings because of a painful udder, which can limit the lying time (Fogsgaard et al. 2012, Medrano-Galarza et al. 2012). Hart (1988) considered sickness behav- iour more as an expression of the animal’s motivational state than a consequence of illness. Furthermore, Aubert (1999) con- cluded that the expression of certain be- haviours in animals requires a hierarchi- cal structure of motivational states that is influenced by different stimuli, which are continuously updated. Motivation is as- sumed to make an essential contribution to behaviours when an animal copes with different challenges, which in turn direct- ly affects the welfare status of the animal. An animal’s changed welfare status reflects a changed pathological status, affecting the behaviours linked to the cow’s cop- ing strategies and motivational hierarchy (Broom 2006). This has been demonstrat- ed in previous studies by Aubert (1999), Konsman (2002) and Dantzer (2001), in which sickness behaviours were considered as motivational states, originating from re- organized priorities and affected by differ- ent stimuli. If sickness behaviour is prevented, recov- ery from disease is poorer (Johnson 2002). Furthermore, fear-motivated behaviours are reported to override sickness-moti- vated behaviours (Aubert 1999). Sickness responses in animals exposed to painful challenges have been reported to affect be- havioural patterns, for example, in rats, lambs and cows (Barrientos et al. 2009, Molony et al. 2002, Tom et al. 2002, Chapinal et al. 2010a and b), demonstrat- ing that there is variation in an animal’s sickness behaviour. Some elements of sickness behaviour have been used to automatically detect disease outbreaks in cattle. Changes in feeding be- haviour and feed intake in cattle have been used to predict ketosis (Gonzales et al. 2008), bovine respiratory disease in beef cattle (Sowell et al. 1998, Buhman et al. 2000), calf morbidity in feedlots (Quim- by et al. 2001) and metritis in dairy cows (Urton et al. 2005, Huzzey et al. 2007). In addition, changes in sucking behaviour are useful for identifying sick unweaned calves (Svensson and Jensen 2007). Bor- deras et al. (2008) noted changes in the behavioural patterns of calves during li- popolysaccharide (LPS) challenge. Follow- ing the challenge, rumination, hay intake and self-grooming decreased. In addition, inactivity during lying and standing bouts increased (Fogsgaard et al. 2012). Experi- mentally induced mastitis has been report- ed to trigger sickness behaviour in cows (Fogsgaard et al. 2012) and affect cow be- haviour and well-being. Moreover, not only are lying and feeding behaviour af- fected (Cyples et al. 2012), but changes in dry matter intake and milk yield have also been reported (Yeiser et al. 2012). Knowledge of the consequences of masti- tis for cow behaviour is still rather limit- ed. Therefore, more precise information is needed on the type of behaviour and how it changes during mastitis, as well as whether it can be detected by observing a cow’s behavioural patterns. MTT SCIENCE 28 15 Cow behaviour and automatic milking Automatic milking systems offer good pos- sibilities to measure health-related param- eters in cows, such as the milk yield, milk flow, milk quality parameters, cow activ- ity and deviations in the milking process. A study by Miquel-Pacheco et al. (2014) confirmed that cows change their behav- iour in an AMS in response to a decrease in a particular health status, which in their case was lameness. Pastell et al. (2006) found that changes in milking behaviour provide valuable signals of leg or other health-related problems. They also noted that lame cows lift their legs more frequently and place less weight on a sore hoof during milking (Pastell and Kujala 2007). Borderas et al. (2004) report- ed that cows that regularly visit the milking robot walk better (less limping) than those that visit the milking robot less frequently. Furthermore, a higher kicking frequency during milking in cows may be a result of pain or discomfort caused, for instance, by teat lesions (Rousing et al. 2004). Along with the rapid change towards pre- cision livestock farming, the effectiveness of technology has become more prominent and its support for cow health and wel- fare. Data patterns obtained from AMS have been modelled to predict and detect clinical mastitis (Kamphuis et al. 2008 a and b). Moreover, Rasmussen et al. (2007) found that the frequency of insufficient milking increased from 5% to 30% one week before a mastitis outbreak, which indicates that mastitis can impact on the success of the milking process. In addition, the effectiveness and availability of auto- matic milking is affected by cow behav- iour (Jacobs et al. 2012). However, more information is needed on the link between cow behaviour and milking success, and the frequency of incomplete milkings in automatic milking systems. Technologies for health monitoring Numerous technologies have been devel- oped and tested for assessing the health status of animals, especially dairy cows. These have involved various novel sensors and technologies in data collection, con- centrating on animal observations. Pre- cision livestock farming, as described by Berckmans (2008), “consists of measur- ing variables on the animals, modelling these data to select information, and then using these models in real time for mon- itoring and control purposes.” A recent review on the use of novel sensor tech- nologies to support health management on dairy farms revealed that considerable effort has been put into finding effective technologies for detecting the most cost- ly welfare issues, such as mastitis, fertil- ity and locomotion problems (Rutten et al. 2013). For example, Polat et al. (2010) and Colak et al. (2008) reported that ud- der temperature correlates closely with a cow’s rectal temperature, and that an in- frared thermal camera can be used to de- tect changes in udder temperature caused by mastitis. Pastell et al. (2006, 2008) focused on leg health by measuring the weight distribution between legs during milking. Changes in the distribution of weight between the legs or changes in milking behaviour were found to be val- uable signals of leg or other health-relat- ed problems due to the more frequent shifting of weight off and lifting of a sore hoof (Pastell & Kujala 2007). Moreover, vision-based systems have been applied in lameness detection (Song et al. 2008). Image analysis has been used to monitor locomotion and posture in cows (Cangar et al. 2008). Various devices exist to re- cord activity in barns, locomotory activ- ity during milking (Pastell et al. 2009, Chapinal et al. 2011), rumination and changes in milk quality (Hovinen and Pyörälä 2011). 16 MTT SCIENCE 28 III Aims of the study The overall aim of the research re-ported in this thesis was to find new health- and behaviour-based indicators to help in establishing more ef- ficient technological tools for dairy man- agement in large herds. The more specific aims were to identify changes in cow be- haviour and udder skin temperature in re- sponse to mastitis, as well to assess the ap- plicability of a thermal infrared camera in detecting udder skin temperature chang- es in cows during mastitis. The objective was to determine how strongly mastitis af- fects the daily behavioural rhythm of cows and assess the sensitivity of novel tools, such as a thermal infrared camera, in de- tecting mastitis. Moreover, we examined cow–technology interactions and problems during milking by using video recording technology. The main research questions − Can a change from normal to sickness behaviour be identified in cows exposed to mastitis? • The hypothesis was that cows will change their normal behaviour pat- tern when sick. One such expected change was an increase in restlessness behaviour. − Can a thermal infrared camera be used to detect changes in the udder skin tem- perature of cows during mastitis? • The hypothesis was that udder skin temperature of cows increases during mastitis, and that this increase can be detected with a thermal infrared camera. − Can cow behaviour during milking be detected and described, and what are the main characteristics of cow–technolo- gy interaction causing deviations dur- ing automatic milking? • The hypothesis was that restless be- haviour in cows during milking caus- es incomplete milkings and impairs the milking process. Specific aims of the study − To investigate how cow behaviour changes when exposed to mastitis (ar- ticle I). − To investigate the viability and applica- tion of novel technological tools to de- tect changes in cows (articles II and III). − To study the effect of milking on cow behaviour and the effect of cow behav- iour on the success of milking and prob- lems occurring during the milking pro- cess (articles III and IV). MTT SCIENCE 28 17 IV Materials and methods The experimental protocols were ap-proved by the Ethics Committee for experimental studies on ani- mals of the University of Helsinki, Fin- land (experiment I, original articles I–II). For experiment II (original article III), an- imals were only video recorded. The Eth- ics Committee for the use of experimental animals at MTT Agrifood Research, Jok- ioinen, approved experiment III (original article IV). Animals and housing For the mastitis experiment (articles I–II), six cows were used as experimental animals (5 Finnish Ayrshires and 1 Holstein-Frie- sian); of these, five cows were in their first lactation and one cow in the second lacta- tion. The cows were housed in a stanchion barn of 68 milking cows in tie stalls bed- ded with wood shavings. The cows had free access to good quality silage and wa- ter and were fed with concentrate six times daily according to their state of lactation. All cows were placed in the same row, be- side each other. The cows were milked with a pipeline milking machine (DeLaval Har- mony, DeLaval International AB, Tumba, Sweden) twice a day at 05:30 and 17:30. The lights were on between 05:00 and 20:00 and a dim night-light was provid- ed in addition to the natural light coming from the windows. The second experiment (article III) was carried out at the Suitia experimental farm of the University of Helsinki and the third experiment (article IV) was conducted at Haapajärvi Agricultural School Farm, Fin- land. Housing in experiments II (article III) and III (article IV) were mostly similar to each other. Cows were housed in a warm loose housing system with a single unit au- tomatic milking system (VMS, DeLaval International Ab, Tumba, Sweden). The experimental group at Suitia experimental farm (article III) consisted of 38 Holstein- Friesian cows. Of these, 58% were primi- parous, 24% in the second, 10% in the third and 8% in the fourth lactation. The cows had free access to good quality silage and water, as well as to concentrate feeders adjusted to feed them according to their state of lactation. The Haapajärvi Agricul- tural School Farm (article IV) had similar arrangements as Suitia. The experimental group consisted of 10 multiparous Ayrshire cows. In addition to an AMS, the Haapa- järvi barn also included a herringbone par- lour with three milking places (DeLaval international, Ab, Tumba, Sweden). Both milking systems in Haapajärvi were in- stalled in the same loose-housing barn and under the same herd management. Forced cow traffic with separation gates in the par- lour were used on both farms. Measurements Clinical measurements, sampling For the mastitis experiment (articles I–II), the clinical examination of the cows and milk sampling of the experimental and control quarters are presented in Table 1. Follow-up of clinical signs Local udder signs were considered slight- ly changed if the udder was slightly swol- len and sore and severely changed if it was painful, firm and severely swollen. Milk ap- pearance was considered slightly changed if milk was slightly discoloured and/or con- tained small flakes. Severely changed milk was strongly discoloured or watery, and/or contained large clots. 18 MTT SCIENCE 28 Behavioural parameters The definitions of the most relevant behav- iours recorded during the research of this the- sis appear in Table 2 (articles I–II), where the mean bout length, total daily duration and frequencies were determined for each of the behaviours. In addition, to examine the ef- fects of time from the induction of mastitis, the data from the recordings were divided into two-hour periods. Cow body postures were scored as either standing or lying. To establish whether cows avoided lying on the affected udder quarter, the side on which the cows were lying was registered. In addition, we registered eating, drinking and ruminat- ing of the cows. Specific observed behaviours during milking (articles III–IV) are present- ed in Table 3. Table 2. Ethogram for observations in the mastitis experiment Behaviour/Posture Description Standing Cow is standing on four legs Lying down Abdomen of cow touches the floor Lying on the right/left udder Right/left udder side is on the ground Stepping Cow lifts a leg (up, front, side, hind directions) Eating concentrate Cow has muzzle in the concentrate bucket or above it and shows chewing movements or licks concentrate from the floor under the bucket Eating silage/hay Cow has muzzle on the silage or shows chewing movements above the feed Drinking water Cow has muzzle in the water bowl Rumination The jaws move rhythmically from side to side, not related to eating. Body care Cow licks its body, or head moves rhythmically while muzzle touches any body part Table 1. Protocol for experimental mastitis in six cows induced with E. coli lipopolysaccharide. Time of the day Days 05.00 07.00 09.00 11.00 13.00 15.00 17.00 19.00 -1 A1, B2, Y3 NaCl4 A A A A A, Y A 0 A, B, Y LPS5 A A A A A, Y A 1 A, B, Y A A A A A, Y A 2 A, B, Y A A A A A, Y A 3 A, B, Y A A A A A, Y A 1 A = clinical examination of the cow, milk sampling and thermal imaging of experimental and control quarters 2 B = bacterial sampling of milk 3 Y = milk yield measurement 4 NaCl = NaCl infusion of the experimental quarter 5 LPS = Escherichia coli lipopolysaccharide infusion of the experimental quarter Table 3. Ethogram for observations in the milking experiments Behaviour Description Kick-off Cow kicks the milking cluster off from the udder Weight transfen Cow lifts a leg and shifts her weight to the other leg Stepping Cow lifts a leg from the ground and puts it back down for a while Kicking Cow lifts a leg to make kicking movements MTT SCIENCE 28 19 Measuring technologies Video recordings Cow behaviour was filmed analogically for 12-hour periods (articles I–III) or a digital camera was used to record milking during the two-week experimental period (article IV). The cameras were installed (articles I–II) above the cows on the barn ceiling. Two cameras recorded the behaviour of two cows from the rear and two cameras from the front. Twelve cameras in total were con- nected to a multiplexer and one VHS video recorder (Panasonic 6070). Three cameras (for article III) and one camera (Panasonic for article IV) were attached in front of the milking parlour on the wall and floor struc- ture. The data were recorded with a dep- 800 IDVR digital recorder (Dynamic Elec- tronics and Protection, Espoo, Finland). The recording system consisted of a PC with a 4-channel video capture card, Sky- view RX (Dynamic Electronics and Protec- tion, Espoo, Finland) surveillance software and the Windows XP Professional operat- ing system. Recordings were saved automat- ically on removable 120 gigabyte hard discs. Recording was set at five frames per second. Behaviours in each experiment were scored continuously with The Observer© (Noldus, the Netherlands) video analysis software. We assessed the functioning of the AMS based on its success in performing the milk- ing process without deviations. The milking process in the AMS was analysed from vid- eo recordings of 300 milkings observed to assess cow behaviour linked to AMS func- tioning and deviations (article III). The re- sults were then compared with reports pro- vided by the AMS computer programs. The reports comprised basic cow information and data on AMS deviations. For the fourth article, a total of 200 milk- ings were observed. Cow behaviour and deviations during milking were recorded and subsequently scored continuously with the same method by using The Observer© (Noldus, the Netherlands) software. The first group of cows was brought to a sepa- rate waiting area in front of the herringbone milking (HRB) and automatic milking sys- tem (AMS) units. Heart rate measuring de- vices (Black box, Tampere University of Technology, 2006) were attached around the cows with flexible belts to record HR during milking. Detecting udder skin temperature with a thermal camera Thermal images were taken to monitor the change in the quarterly udder temperature of cows (articles I–II). Three consecutive images were taken at one-second intervals before clinical sampling at a distance of ap- proximately 50 cm from each angle. The orientation for images was from the lateral and the medial angles from the experimental and control quarters, and from the cranio- lateral side of the experimental quarter. Im- ages were taken at each sampling through- out the five-day experimental period with a handheld thermal camera (IR FlexCam Pro, Infrared solutions, USA). The thermal camera used operated in the 8 to 14 µm spectral band. The thermal res- olution of the camera was 0.09 oC and it was calibrated to the temperature range of 0–100 oC. The camera used microbolome- ter detectors and had a 160 x 120 pixel fo- cal plane array. The camera had an internal recalibration feature, which automatically calibrated the detector to provide readings corrected for the ambient temperature. The emissivity value was set to 0.98, which is the value measured from human skin (Jones and Plassmann, 2002). The maximum temperatures of the three images were averaged and recorded as the udder temperature for the sampling. The values from the lateral side were selected for use in the analysis, since they were found to most precisely show the temperature rise of the udder during the LPS challenge (Fig- ure 1.). 20 MTT SCIENCE 28 Statistical analysis To analyse the impact of acute endotox- in mastitis on cow behaviour (article I), a paired t-test was applied to study the ef- fects of day (placebo vs. induction) on the daily durations, bout frequencies and bout durations of registered behaviours. To ex- amine the effects of the time of day, two- hour means for the behaviours were cal- culated. The means were analysed with repeated mixed models. All the statistical analyses were conducted with SPSS 13.0 for Windows (SPSS Inc. Chicago, IL). Inflammation indicators in milk and ud- der temperature were analysed with linear mixed models, taking repeated measures into account (article II). The fixed fac- tors were hour (the time from the induc- tion of mastitis), quarter (experimental or control), and the interaction between hour and quarter. The cow was introduced as a random effect. No covariates were used. Homogeneity of variances was evaluated with a scatter plot of residuals and pre- dicted values. Pearson’s correlation coef- ficient was calculated between the rectal temperature and udder temperature. All statistical analyses were performed with SPSS 13.0 for Windows (SPSS Inc., Chi- cago, IL). Linear mixed models were used (arti- cles III–IV) for the analysis of the ef- fect of milking on cow behaviour. All the statistical analyses were conducted with SAS/MIXED software (version 9.1 SAS, 2004). This revealed that square- root transformation was needed for the analysis of kicks, weight transfers, steps and the somatic cell count. Respiration and heart rate data were not transformed. Statistical analysis was based on a crosso- ver experimental design. The experiment had two periods, and two measurements were taken within each period: one dur- ing the morning milking and another during the evening milking. The milk- ing time was used as a repeated factor in statistical analyses. Figure 1. Thermal images of the lateral angle of the experimental udder quarter (left fore) a) before and b) during the induction of acute endotoxin mastitis. A1 = 40× 40 pixel area above the teat; Hot+ = position of the maximum udder skin temperature of the image. MTT SCIENCE 28 21 V Results The most important results of the experiments appear in this sec-tion, which summarizes the find- ings from articles I–IV. For more detailed results, the reader may refer to the origi- nal papers included at the end of the thesis. 1. Can we identify a change between typical and sickness behaviour in cows exposed to mastitis? (Article I-II) Clinical signs All cows developed clinical mastitis, show- ing both systemic and local signs after the Escherichia coli lipopolysaccharide (LPS) challenge (Figure 2.). Rectal and udder temperatures increased from 4 to 8 h af- ter the induction (post challenge, PC; P < 0.001) and the udder quarters became swollen. The udders of cows were visibly swollen from 2 to 4 h after the induction, and did not return to normal during the experimental period (5 days, 100 hours). Rectal temperatures of the cows started to rise from 4 to 6 h PC and remained above 39.2 °C from 6 to 10 h PC, reach- ing peak values at 6 h. Body temperatures returned to normal within 12 h PC (Fig- ure 3.). Milk electrical conductivity and the somatic cell count (SCC) started to in- crease at 4 h PC and milk NAGase activ- ity at 8 h PC, and remained higher than in the control quarters during the follow- ing 24 h. All cows had mild to moderate system- ic and local signs and changes in milk ap- Figure 2. Local udder signs also started to appear in cows 4 hrs after challenge, and the time spent lying down decreased 0 1 0 2 0 3 0 4 0 5 0 6 0 - 2 2 4 6 8 10 12 14 16 18 20 22 No local signs Swollening of the trial quarter Lying down during the control day Lying down during the trial day * * * * * Some local signs Severe local signs * 22 MTT SCIENCE 28 pearance 2 h PC, with a few exceptions: changes in the milk appearance of one cow were seen 4 h PC, and local signs and changes in the milk appearance of anoth- er cow were seen 4 h PC. Both the rectal and udder skin temperatures and system- ic signs remained above normal until from 8 to 10 h PC, as illustrated in Figure 3. All quarters except 1 were swollen until the end of the experimental period, and milk appearance did not normalize dur- ing this period. Local signs in the udder were severe in five cows on the day of the challenge. Severe changes in milk appear- ance were visible for three cows in sporad- ic samplings. No local signs or changes in the milk appearance were detected in the control quarters. The average milk yield of the cows on d −1 and from the morn- ing milking of d 0 was 13.0 ± 0.4 kg per milking, and from the evening milking of d 0 and at both milkings of d 1 was 11.5 ± 0.1 kg per milking (P = 0.50). Changes in behaviour On the day of acute endotoxin mastitis in- duction, compared with the control day before induction, cows exhibited more frequent stepping behaviour (P = 0.02), tended to spend less time lying and less time lying on the side of the affected ud- der quarter (P < 0.07, for both). The cows also tended to stand longer on the induc- tion day than on the control day and the standing bouts were longer (P < 0.07). They spent a longer time eating silage dur- ing the induction day than the control day (P < 0.05) and they also elevated their step- ping frequency. Daily behavioural rhythm Mastitis had an effect on the daily rhythm of the cows. Statistically significant (P < 0.05) interactions between the time of the day and day of the experiment were found for the mean hourly durations of lying, ru- minating and drinking water, as illustrated in Figures 4., 5., 6 and Table 4. Figure 3. Rectal temperature (solid line), maximum udder skin temperature of the image (dotted line) and udder skin temperature of the 40 × 40 pixel area above the teat (dashed line) for the lateral angle of the experimental quarters of 6 cows throughout the experimental period, excluding d −1, x ± SE. Escherichia coli LPS was infused at time point 0. a–f Different letters indicate statistically significant differences between mean temperatures (P < 0.05) at different sampling times. MTT SCIENCE 28 23 Figure 4. The impact of LPS mastitis challenge on six dairy cows; time spent resting and body temperature. 36.5 37 37.5 38 38.5 39 39.5 40 40.5 41 0 10 20 30 40 50 60 6 8 10 12 14 16 18 20 22 24 2 4 Plasebo day Induction day T * * * * * Figure 5. The impact of LPS mastitis challenge on six dairy cows; time spent ruminating and body temperature. 37 37.5 38 38.5 39 39.5 40 40.5 41 0 5 10 15 20 25 30 6 8 10 12 14 16 18 20 22 24 2 4 Plasebo day Induction day T + * * The mean time spent ruminating de- creased between 4 and 8 h PC (at 12:00 and 16:00 PM) compared with the con- trol day (P < 0.05). Cows also tended to reduce their drinking between 4 and 6 h PC (from 10:00 to 12:00 AM) compared with the control day (P < 0.05). Further- more, cows increased their drinking at 10 h PC (at 16:00 PM) (P < 0.05). 24 MTT SCIENCE 28 Figure 6. The impact of LPS mastitis challenge on six dairy cows; time spent drinking and body temperature. 37 37.5 38 38.5 39 39.5 40 40.5 41 0 20 40 60 80 100 120 140 6 8 10 12 14 16 18 20 22 24 2 4 Plasebo day Induction day T + * * + 2. Can a thermal infrared camera be used in detecting temperature changes in the cows’ udder during mastitis (Article II)? Infrared thermal camera and body temperature All cows developed mastitis after the Es- cherichia coli lipopolysaccharide (LPS) challenge. The mean temperature of the udder and control quarters was elevated 4 h post-challenge (P < 0.01). However, the udder temperature of the cows did not in- crease before an increase was recorded in the rectal temperature. The correlation between the rectal temperature and mean udder temperature for the lateral angle of the drawn circle was r = 0.92 (P < 0.001), and the correlation between the rectal temperature and maximum udder tem- perature for the lateral angle of the image was r = 0.98 (P < 0.001). Table 4. Resting behaviour of six dairy cows before (placebo day) and after induction of acute en- dotoxin mastitis (induction day). Results are presented as the daily mean and the standard error of paired differences (SE). Behavioural variables Frequency (No.) Total daily duration (min) Mean bout length (min) Day Placebo Day Induction SE Placebo Day Induction SE Placebo Day Induction SE Standing 23.33 21.0 2.19 638.61 759.59* 30.03 27.73 36.63† 3.71 Lying down 23.33 21.0 2.19 800.40 670.30* 30.08 34.88 32.59 3.46 Lying down on the control quarter (right) 10.33 8.5 1,64 381.30 326.04 29.33 38.28 40.94 5.56 Lying down on the induced quarter (left) 11.17 10.67 0.96 418.49 353.15† 28.43 37.61 33.76 3.62 †; P ≤ 0.07; *; P ≤ 0.05, se = standard error of mean difference MTT SCIENCE 28 25 The maximum udder temperature (TU) of the image and the mean TU of the meas- ured area increased in parallel for both an- gles of the experimental and control quar- ters. At 6 h PC, 5 of the 6 cows had a clearly increased mean TU in the measured area. The udder temperature decreased to normal levels within 8 to 10 hours after the challenge, in parallel with or a slightly later than the rectal temperature. 3. Changes and main characteristics of cow behaviour during milking (Articles III–IV) Changes in milk We found no statistically significant dif- ference between milking systems (auto- matic milking system, AMS, and herring- bone system, HRB) in the milk somatic cell count (article IV). When milked in the AMS and the HRB, respectively, the aver- age somatic milk cell count was 45.9 cells/ µl and 51.1 cells/µl. The cell count varied between 16.1 and 114.2 cells/µl. Howev- er, a clear difference in the milk cell count was detected between morning and even- ing milkings, with higher milk cell counts in the morning (F = 17.2, P < 0.001). The change in milking management caused no marked peaks in the cell count. Behavioural changes fol- lowing a change between herringbone (HRB) and auto- matic milking systems (AMS) A change between the AMS and HRB did not cause marked changes in cow be- haviour (article IV). Independent of the milking system, cows expressed greater be- havioural activity, as measured by weight transfer, in morning than evening milk- ings (F = 8.21, P < 0.05). On average, cows exhibited 2.2 weight transfers per milking in the AMS and 2.6 in the HRB. An interaction effect between the type and time of milking was detected in bovine stepping behaviour. In the AMS, cows stepped more often during the morning than the evening milking (21.2 vs. 16.1 steps, F = 6.02, P < 0.05). Kicking behav- iour did not differ according to the time of milking or the type of milking system. Kicking varied on average between 1.3 and 1.8 kicks during the morning milking and between 1.6 and 2.0 kicks during even- ing milking. Kick-offs, i.e. kicking the milking device completely off the udder, were also record- ed, but the total number remained too low for statistical analysis. In the HRB, 16 kick-off cases occurred during the experi- mental period; the corresponding number in the AMS was 9. The health records of cows were checked during and after the experiment. No changes in health due to the experiment were detected. Cow behaviour and devia- tions in automatic milking Six of the 300 milkings (2%) were not as- sessed due to poor visibility of the milk- ing process or the stockperson standing in front of the camera while treating the cow. The number of milkings in the automatic milking system (AMS) per cow during the 3-day observation period varied from 5 to 10. The milking interval of the group aver- aged 8 h 55 min (SD 0.07), ranging from 6 h 46 min to 13 h 2 min. Of the 300 milk- ings, 247 (82%) were completed success- fully. The types of deviations are present- ed in Figures 7 and 8. Although failures causing incomplete milkings occurred, in most cases the ma- chine was able to compensate the proce- dure such that all phases of the milking were properly performed. If the AMS was 26 MTT SCIENCE 28 Figure 7. Types of deviations during automatic milking (in 300 observed milkings). 0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5 (1) Failure in teat cleaning (2) Failure in milking (3) Failure in both cleaning and milking (4) Kick-offs during cleaning (5) Kick-offs during milking Type of deviation D ev ia tio ns % Figure 8. Types of deviations during automatic milking (in 300 observed milkings). 82 % 16 % 2 % Successfully completed milkings Failed milkings Unclear milkings able to compensate for failures and the cow was milked, the milking was recorded as successful in the results. However, a total of 47 (16%) incomplete milkings were ob- served. Failures occurred in either the teat cleaning or milking processes, or in both. Two unclear cases occurred because a cow moved continuously such that the observer could not be sure of what was happening. In the teat cleaning process, 11 (4%) de- viations occurred. In the milking process, milking was incomplete in only 3 (1%) cases. However, 11 (4%) cases were defi- cient in both teat cleaning and milking. Kick-offs were detected in 31 of 294 milk- ings, with 14 of them occurring during teat cleaning; 1 (0.3%) of them was com- pensated by the machine, such that the procedure was recorded as successful, and 13 (4%) were uncompensated. Dur- ing milking, 17 kick-offs were detected; 8 (2.7%) of them were compensated and 9 MTT SCIENCE 28 27 Table 5. The origin of failures during milking were caused both by cow behaviour and AMS functioning. Failures caused by cow Number Proportion, % 1) Kick-offs during washing (uncompensated) 13 28% 2) Kick-offs during milking (uncompensated) 9 19% 3) Cow caused failures in teat washing 1 2% 4) Cow caused failures in milking 1 2% 5) Cow caused failures in both washing and milking 2 4% Sub-total 26 55% Failures caused by AMS     1) Failures in teat washing 10 21% 2) Failures in milking 2 4% 3) Failures in both washing and milking 9 19% Sub-total 21 45% Total 47   (3%) were uncompensated. Only 29% of all deviations during the teat cleaning and milking processes were successfully com- pensated by the robot. Observed kick-offs were compared with kick-offs in the AMS computer report. The latter consisted of normal kick-offs, i.e. the milking unit being kicked of by the cow, a teat cup dropping on the floor when the teat was empty, and a cow stand- ing on the milking pipe, resulting in de- tachment of the teat cup. Some teat cups also fell when the robot gripper tried to attach the teat cup to another teat. All of these situations were registered as kick-offs in the AMS software. Origin of failures in automatic milking Deviations during milking were of two types: cows disturbed the milking process in 55% of the failures, and failures orig- inating from the AMS were detected in 45% of the failed cases, which are shown more precisely in Table 5. Problem cows during automatic milking Six problem cows out of 38 were clearly identified. Three of these problem cows caused 52% of all the failures in teat clean- ing, milking, or both. These six problem cows caused the most important deviations in the milking process. After a failed milk- ing of these six cows, the stockperson had to complete the milking procedure man- ually on each observation day. The most severe problems were detected with two cows that had no successful un- aided milkings during the 3-day observa- tion period. The first problem cow had three unsuccessful milkings that were not manually assisted and two successful man- ually assisted milkings during the obser- vation period. The second problem cow made seven milking attempts, but none of them were successful. The stockperson had to put the milking robot into man- ual mode to assist the robot in milking the cow. Four of the six problem cows experienced milder problems in the milking process. 28 MTT SCIENCE 28 Individual failures in the milking process varied from 40% to 83% for these four cows. Their failures comprised incomplete teat cleaning, milking or uncompensated kick-offs. Restless behaviour, i.e. kicking the teat-washing device and detaching teat cups, had a marked impact on their milk- ing procedure. Problems during milking became progressively more serious if im- mediate action by the stockperson was not taken to correct the situation. The status of these six problem cows was rechecked six months after the experi- ment. The situation had not improved, and two of the cows were culled. The rea- son for culling the first problem cow was her character and the problem behaviour exhibited in the AMS. The cow could not become accustomed to the milking pro- cedure, continuing to kick the machine, resulting in failed milkings. The second problem cow was culled because of a con- tinuously high somatic cell count. Of the four remaining problem cows, three had been treated several times for fertility and reproduction problems. One cow had no health problems. MTT SCIENCE 28 29 VI Discussion Cows in the studies presented in this thesis changed their behav-iour after the induction of mas- titis. However, a change of milking tech- nologies did not impact on cow behaviour. When the health status of the cows deteri- orated, changes were detected in both their behavioural pattern and clinical status. At present, sick cows are often detected in the late stages of infection when their milking interval is already prolonged (Miquel-Pa- checo et al. 2014). Our results support the view that automated observation tools pro- viding combined information on cow be- haviour, clinical status and milk appear- ance would best serve in detecting early warning signs of a change in health status. 1. Can we identify a change between typical and sickness behaviour in cows exposed to mastitis? (Articles I–II) Cows suffering from endotoxin-induced acute mastitis showed some typical fea- tures of sickness behaviour, e.g. a poorer appetite. In contrast to typical sickness be- haviour, cows did not spend more time ly- ing, but stood more, and avoided lying on the side of the sick quarter. Moreover, in contrast to earlier findings describing in- creased lying and lethargy, cows exhibited increased stepping. Cows increased their standing time simultaneously with the lo- cal swelling of the udder quarter and an el- evated temperature. The most promising health indicators were identified in lying, ruminating and drinking behaviour pat- terns. Moreover, changes in cow behaviour occurred rather concurrently with clini- cal changes, with changes in udder swell- ing becoming visible between 2 to 4 hours post-challenge. Change in behavioural priorities and patterns Mastitis was found to have consequenc- es for both cow behaviour and physiolo- gy that made the cows change their be- havioural pattern. Due to sickness, cows reorganized their behavioural priorities and partly shifted their normal behav- iour towards sickness behaviours. Mas- titis caused changes in behavioural pat- terns, such as decreased lying time and changes in ruminating, as well as chang- es in the clinical health status. These find- ings are consistent with those of by Cy- ples et al. (2012), who noted that cows reduced their lying time during the first 20 h following a mastitis outbreak, when the most severe signs of experimentally induced mastitis were present. They re- ported similar indications of the poten- tial of lying time as a valuable cow health indicator. Fogsgaard et al. (2012) also de- tected similar changes in cows, reporting increased standing and decreased lying during mastitis. Changes in clinical sta- tus were additionally visible, such as udder swelling and modified milk composition. However, the findings in this study con- trasted with the previous results of John- son (2002) and Weary et al. (2009), who described an increase in lying time. Fur- thermore, Weary et al. (2009) suggested that due to sickness, cows were motivat- ed to lie down in order to rest and en- hance the healing process by minimizing the consumption of body energy reserves. In the present study, cows appeared to compromise their need for lying in or- 30 MTT SCIENCE 28 der to avoid pain in the affected quarter of the udder, as also found by Bolles and Fanselow (1980). Since some of the be- havioural changes in our study were op- posite to those expected in sick animals, we suggest that pain experienced in the udder overrode the motivational state of the cows to express sickness behaviour, as also suggested by Aubert (1999). The changes in the behaviour of the cows with acute mastitis can be interpreted to reflect changes in behavioural priorities (Johnson 2002). We suggest that the cows started to reduce their lying time after the induction of mastitis, accompanied by a rise in the body temperature, but also along with the increasing local signs of udder swelling. Increased vigilance Cows in the present study showed more restlessness behaviours, a decreased lying time and increased preference for lying on one side, in line with the findings of Me- drano-Galarza et al. (2012). It is also pos- sible that discomfort caused by mastitis prevented the cows from lying and shift- ed their behaviour towards increased step- ping to temporarily reduce their physical discomfort, as suggested by Cooper et al. (2007). Increased standing and stepping may, however, be a sign of increased vigi- lance, which helps the cows to detect a giv- en stimulus at a given time (Dimond and Lazarus 1974). Increased vigilance has also been observed in lactating mice during an LPS challenge (Aubert et al. 1997). More- over, this finding may originate from a dif- ferent motivational hierarchy of cows in favour of the preservation of survival in- stincts, keeping the cows in a more vigi- lant state (Aubert 1999). We found that cows increased their step- ping behaviour when suffering from acute mastitis. The increase in stepping activ- ity may have originated from the innate need of the cows to avoid or escape the uncomfortable experience, and may repre- sent pain behaviour, as suggested by Mol- ony and Kent (1997) and Chapinal et al. (2011). Daily rhythm Mastitis affected the daily rhythm of the cows. Significant interactions between the time of the day and day of the experiment were found for the mean hourly durations of lying, ruminating and drinking wa- ter. The mean time spent ruminating de- creased between 4 and 8 h PC compared with the control day. Cows also tended to reduce their drinking between 4 and 6 h PC compared with the control day. Moreover, cows compensated their drink- ing at 10 h PC. Numerous findings con- firm that lying down and rising may cause uncomfortable or painful feelings, which can change the locomotory activities of the cows, i.e. prohibit or prolong lying down movements (Niss et al. 2009, Galindo & Broom 2002, Juarez et al. 2003, Walker et al. 2008). The present results are consistent with findings on sickness behaviours in cows during oligofructose overload challenge reported by Niss et al. (2009), and dur- ing ruminal acidosis challenge report- ed by DeVries et al. (2009). Lying time during sickness has been found to be a changing variable in cows (Fogsgaard et al. 2012). Previous studies have demon- strated that cows have a strong motivation to lie down (Jensen et al. 2005, Munks- gaard et al. 2005), but they avoid doing so if the lying surface is uncomfortable (Ha- ley et al. 2001). In this study, standing be- haviour changed, as cows tended to stand longer on the induction day than on the control day and the standing bouts were longer. The extended standing after LPS challenge contrasts with the behaviour re- ported in calves, which tended to lie down after LPS challenge (Borderas et al. 2008). The motivational priorities of calves and their ability to cope with acute illness such as LPS challenge can differ from those of dairy cows. Moreover, different experi- MTT SCIENCE 28 31 mental settings may also explain this dif- ference, as the calves were challenged in- traperitoneally, while cows in the present study were infused via the teat. In addi- tion, changes in the somatic cell count and NAGAse remained changed until the end of the induction day. This may be because unpleasant feelings preventing the cows from lying down, but changes in milk are not the best indicators of pain in cows. Eating and rumination We found that cows spent a longer time eating silage during the induction day compared with the control day. Ruminat- ing behaviours are related to feeding and lying behaviour patterns (Schirmann et al. 2012) such that diurnal patterns exist for both ruminating and feeding and ly- ing when observing cows for a longer peri- od. However, no correlation was observed in our experiment between daily rumi- nation times and daily lying time. In our study, cows adjusted their eating behav- iours during mastitis by spending more time eating, especially eating silage. The cows were probably not eating more, but rather more slowly, due to fever-induced lethargy. Fitzpatrick et al. (2013) noted that cows ruminated less in the hours fol- lowing LPS infusion and compensated for rumination later during the induction day, as was also the case in our study. Cows spent proportionally more time ru- minating while lying down. We suggest that cows adapted to a shortened lying time by increasing the proportional time spent ruminating. The increased rumina- tion time has been explained by Schir- mann et al. (2012), who reported a posi- tive correlation between rumination time and lying time, indicating that periods of rumination are more frequent when cows are lying down. To conclude, we suggest that a change in the lying pattern can serve as a valu- able health indicator when observing be- havioural changes related to mastitis, but is not sensitive enough. Thus, additional indicators are needed for effective health observation. 2. Can a thermal infrared camera be used to detect temperature changes in the cows’ udder during mastitis (Article II)? Thermal imaging was found to be a vi- able tool to detect changes in the udder skin temperature of cows during masti- tis. As expected, the udder skin temper- ature rose simultaneously with the rectal temperature. Moreover, a slightly unex- pected observation was that the elevated udder skin temperature was detected 4 h PC, while mild systemic signs, local signs in the udder and changes in milk appear- ance were already detected 2 h PC. Fur- thermore, automated tools for detecting changes in both udder skin temperature and modelling visual changes in udder ap- pearance would enable earlier warning of a changing health status in cows. Infrared thermal imaging was a viable and non-invasive tool in detecting the rise in the udder skin temperature during mas- titis. Thermal images revealed that the mean udder temperature of the experimen- tal and control quarters was increased 4 h PC. The udder temperature rose simulta- neously with a rise in the rectal tempera- ture. The maximum udder temperature measured by a thermal infrared camera increased in parallel for both angles of the experimental and control quarters. At 6 h PC, 5 of the 6 cows had a clearly increased mean udder temperature in the measured area. The thermal infrared camera was able to produce reliable information that was consistent with experiments reported by Colak et al. (2008), confirming that in- frared thermography is sensitive enough to detect differences in udder skin tempera- ture tested against the California Masti- 32 MTT SCIENCE 28 tis Test. Moreover, studies by Polat et al. (2010) confirmed our finding that an in- frared thermal camera is sensitive enough to detect subclinical mastitis measured by observing changes in the udder surface temperature. Thermal infrared cameras have also been used in mastitis detection in sheep by Saraiva Martins et al. (2013). Their studies confirmed that the accuracy of thermal infrared cameras was sufficient to detect subclinical mastitis. To conclude, our study confirmed that in- frared thermal cameras can be used to de- tect changes in udder temperature caused by mastitis. However, when designing nov- el technological skin temperature-based tools for health detection, more research and development is needed to design farm- level applications to be integrated into ex- isting technologies, for example in auto- matic milking systems. 3. Changes in and main characteristics of cow behaviour during milking (Articles III–IV) We found that cows easily adapt to trans- fer between herringbone and automatic milking systems. However, some cows ap- peared to have difficulties in being proper- ly milked in an automatic milking system. The underlying reasons for these difficul- ties originated from both the cows and the milking system. We found that the milk- ing process should be regularly observed to ensure an optimal and effectively working human–animal–technology interaction, which is essential in dairy management. Behavioural activities and deviations dur- ing milking can be detected by using web- based video recording technology. Devia- tions in the milkings and causes of failures at milking are easy to analyse from record- ings in order to ensure that cows are prop- erly and regularly milked. Factors impacting on the success of automatic milking The overall success of the automatic milk- ing process was on a reasonably high level (82%) in our experiment. Failures in milk- ing originated from cow- and machine- based failures. This was in line with Ha- mann et al. (2004), who concluded that the occurrence of incomplete milkings was conditioned by both technical and cow-re- lated circumstances. In some cases in our study, the milking robot apparently did not find a teat to be milked because the cows moved or kicked excessively. Fur- thermore, the milking interval of a sin- gle udder quarter or even the whole udder may have been prolonged if cows needed to queue in front of an AMS and the wait- ing time for milking exceeded their nor- mal milking rhythm. Long milking in- tervals impair tight junctions in the cow udder, causing an influx of somatic cells into the milk. This influx of neutrophils has been found to continue even after cows were returned from once-a-day to twice-a- day milking (Stelwagen and Lacy-Hulbert 1996). In our study, this may have caused more pressure on the udder and changed the udder size, resulting in subsequent dif- ficulties of the milking unit in attaching a teat cup to a teat. Hovinen et al. (2005) also found that an abnormal teat or udder structure in general caused, for example, unsuccessful teat cleaning. Moreover, the milking frequency in automatic milking systems should be modified according to the cow’s state of lactation (Hovinen and Pyörälä 2011). The milking of problem cows should reg- ularly be checked to ensure that all ud- der quarters are milked completely. In ad- dition, the deviation report that an AMS automatically produces is a valuable tool to identify potential problems in milking. MTT SCIENCE 28 33 Cow behaviour in changing milking systems We found that cows were able to cope with a change between milking systems and did not alter their behaviour when the milking system was changed within the same barn. However, cows displayed more behaviour- al activity, as measured by weight transfer between the legs, in morning than evening milkings. The difference in weight transfer was similar for the two milking systems. An interaction effect between the type and time of milking was detected in stepping behaviour. When cows were milked in the AMS, they exhibited more stepping dur- ing the morning than in the AMS during evening milking or in the HRB during both the morning and evening milkings. Kicking behaviour did not differ in time or according to the type of milking sys- tem. We hypothesised that changes in the milking procedure might disturb the dai- ly life of cows, which might be reflected in a change in their milking behaviour. This was not the case in our study. It has to be remembered that in our study, both milking systems were situated inside the same barn, and thus the social and physi- cal environment of the cows remained the same. Cows in loose housing, and especial- ly with automatic milking systems, devel- op certain behavioural patterns such as the milking interval, synchrony in the feeding time and resting behaviours (Rousing et al. 2006, Borderas et al. 2008). DeVries et al. (2011) found that standing and lying pat- terns in cows are especially affected by the milking frequency and milk yield. Management of automatic milking We found that cows having difficulties in being properly milked were dependent on the activity of the stockpersons to en- sure complete milking. These findings are supported by Hovinen and Pyörälä (2011), who described the actions of stockpersons as being the most significant external fac- tor in management affecting the udder health of a cow. Moreover, they found that proper milking management in automat- ic milking affects the udder health, clean- liness and milk yield of cows. In addition, cow traffic in AMS barns has an impact on cow behaviours, as sep- aration gates may affect milking, resting and feeding frequencies. DeVries et al. (2011) and Munksgaard et al. (2011) also observed that cow behaviour in an AMS is affected by daily care routines, such as fetching cows to be milked or placing feed in front of the cows. We suggest that cows were able to cope with the change in the milking process because the change was properly managed. Cows are adaptive in- dividuals and exposed to various chang- es during their life on a farm. The im- portance of management routines and maintenance decisions, especially with cows having more important problems in adjusting to AMS, should be taken seri- ously. Rasmussen et al. (2007) found the frequency of unsuccessful milkings to in- crease from 5% to 30% one week before clinical mastitis occurred. This confirms that it is essential to observe the success rate of milkings and be able to take ac- tions immediately after failures occur dur- ing milking. If the stockperson does not react in time to problems occurring dur- ing automatic milking, more problems are likely to occur and lead to deterioration in the welfare of the cows. Our findings em- phasise the importance of an effectively working human–animal–technology re- lationship to secure proper and complete milking for cows in AMS (III), and are in line with those reported by Deming et al. (2013), who suggested that the success of an AMS system is largely affected by the combination of housing, management and cow characteristics. 34 MTT SCIENCE 28 VII Conclusions The research reported in this the-sis focused on the use of behav-ioural observations as an indicator of management changes (e.g. the milking system) and health problems (i.e. mastitis). Furthermore, technology (i.e. an infrared camera) could help to improve the accu- racy of behavioural observations and im- prove the early detection of specific health problems. Our results improved knowl- edge of how a cow modifies its behaviour when developing mastitis. We found that detecting changes in behavioural patterns and assisting clinical detection with a ther- mal infrared camera measuring udder skin temperature (II) can serve as an additional tool for clinical mastitis detection (I). Fur- thermore, our findings on cow behaviour confirm that cows are able to cope with changes in management if they are prop- erly managed and do not disturb their rou- tines to a large extent. Moreover, we found evidence that half of the deviations occur- ring during the automatic milking process originated from cow behaviour, such as a cow kicking, lifting the legs and moving during milking (III). When exposed to mastitis, cows also alter their behaviour during milking. As milk- ing in automatic systems is carried out au- tomatically by machines, the stockperson does not have information on the changing behaviour of cows during the milking pro- cess. Thus, there might exist an interplay between mastitis, the success of the milk- ing process and cow behaviour. 1. Can we identify the change between typical and sickness behaviour in cows exposed to mastitis (Articles I-II)? We found that behavioural changes in cows due to mastitis could be identified. The most interesting changes were detect- ed in lying, eating and stepping behaviour during the bacterial LPS challenge. It was shown that inflammation simultaneous- ly affected the health status and changed the behavioural priorities of cows. In con- trast to our expectations, visual signs in the udder and changes in milk composi- tion occurred only 2 h PC, while clinical and behavioural changes appeared at 4 h PC. However, the changes in lying and restless behaviours appear to be promis- ing indicators for detecting signs in cows exposed to mastitis. We found that cows were more restless during the morning compared with the evening milking, sug- gesting that both individual and time-de- pendant behavioural variations need to be taken into account when considering rest- less behaviour in cows as a welfare indi- cator. The present study also yielded new information on cow behaviours during au- tomatic milking that could be used when further developing cow observation meth- ods in this system. 2. Can a thermal infrared camera be used to detect temperature changes in the cows’ udder during mastitis (Article II)? The transient increase in body temper- ature following experimentally induced clinical mastitis was successfully detect- ed with the help of a thermal camera. The udder skin temperature rose simultaneous- ly with the rectal temperature. However, local inflammatory changes of the udder, MTT SCIENCE 28 35 appearing earlier than the rectal temper- ature increase, were not detected by the thermal infrared camera. The capacity of thermal imaging to detect naturally occur- ring clinical mastitis needs to be examined under field conditions, where the system- ic inflammatory reaction may be less pro- nounced in a longer process. This study demonstrated, however, that the udder is a sensitive site for the detection of diseas- es of dairy cattle causing a rise in tempera- ture by infrared thermal camera. Thermal infrared imagining is a promising meth- od for obtaining information on chang- es in the health status of cows. However, the changes in udder temperature did not serve as an early warning indicator, since visible changes in the udder and milk ap- peared first. Therefore, technologies based on visual modelling might further improve the timing of mastitis detection. To con- clude, there remains a need to further de- velop more effective technologies to be in- tegrated into barn conditions. 3. Changes and main characteristics causing deviations during automatic milking (Articles III–IV) Half of the problems during milking were caused by the cows, meaning that the milking robot itself works at a satisfac- tory level. This demonstrates the impor- tance of monitoring cow behaviour as well as incomplete milkings in the automatic milking system. The web-based video re- cording technology we used proved to be highly useful in observing cow behaviour, especially problem behaviour during the milking process. We suggest that the sys- tem can be used as a supplementary tool for observing the behaviour of animals in the barn as well as during procedures such as the automatic milking process. These tools provide additional information on the effectiveness of automatic milking, the success of teat cleaning and milking as well as deviations occurring during milk- ing. However, further development in au- tomating the observation system is needed to register and combine both the changes in cow behaviour and failures produced by the milking machine to provide a valuable support tool for herd management. Our findings concerning cow behaviour when transferring between milking sys- tems confirmed that cows, as adaptive in- dividuals, do not show changes in behav- iour during milking, even though their social synchronisation changed in con- nection with a change in milking system. Cows appear to be flexible individuals and cope well with properly managed changes in daily routines. The cost of changes in herd management in terms of cow health and welfare are often difficult to deter- mine, since many factors cause variability, such as the impact of a different produc- tion environment, the management sys- tems and also humans. In our study, these other factors were controlled, and the only parameter that changed for the cows was the milking procedure. Cows react to their changing health status by altering their behaviour. This can be detected both by using technological tools and by observations of the stockperson, es- pecially during milking. Moreover, the im- portance of the stockperson still remains in automatic milking to observe cows and the success of milking. I suggest that the daily working hours of the stockperson should be focused on observing and en- suring the functioning of human–animal– technology interaction in dairy manage- ment, as illustrated in Figure 9. 36 MTT SCIENCE 28 4. Future implications for welfare studies Questions requiring further studies • Cows do change their behaviour in milking when exposed to mastitis. Therefore, the success of the milking procedure itself should be regularly ob- served. The role of the stockperson is even more important in large herds, and especially when automatic milking is used to observe cows and the milking process (Hovinen and Pyörälä 2011). This can only be done by ensuring a well-functioning human-animal-tech- nology interaction in herd management. Therefore, the important interplay be- tween behaviour, mastitis, milking and the human-animal-technology interac- tion requires further study. • What is the cost of a change in motiva- tional priorities for cows during the re- covery process? • What combined parameters are opti- mal for the early detection of acute dis- ease outbreaks in terms of the health sta- tus of cows? • What are the optimal combinations of detecting and modelling behavioural, physiological and clinical changes in cow health and welfare on a long-term basis? • How can a well-functioning human- animal-technology relationship be en- sured, especially when introducing new technologies into a herd? Stockperson Animal -cow Technology - AMS Attitudes Behaviour Fear Productivity Failures/deviations Management Welfare and behaviour Workability, capacity Stress Figure 9. Complexity of the interaction between humans and animals (Kaihilahti et al. 2007, modified from Hemsworth & Coleman 1998) and technology, and how the workability of an automatic milking sys- tem impacts on the capacity of the system, the welfare status of the cows and management decisions. MTT SCIENCE 28 37 VIII Reference list Albright, J.L. 1993. Nutrition, Feeding and calves; Feeding behavior of dairy cat- tle. Journal of Dairy Science. 76, 2, 485–498. Aubert, A., Goodall, G., Danzer R., Gheusi, G. 1997. Different effects of lipopoly- saccahride on pup retrieving and nest building in lactating mice. Brain, Behav- ior, and Immunology. 11, 107–118. Aubert, A., 1999. Sickness and behaviour in animals: a motivational perspective. Neuroscience and Biobehavioral Re- views. 23, 1029–1036. Bannerman, D.D., Chockalingam, A., Paape, M.J., Hope, J.C. 2005. The bo- vine innate immune response during experimentally-induced Pseudomonas aeruginosa mastitis. Veterinary Im- munology and Immunopathology. 107, 201–215. Barrientos, R.M., Watkins, L.R, Rudy, J.W., Maier, S.F. 2009. Characterization of the sickness response in young and aging rats following E. coli infection. Brain, Behavior, and Immunity. 23, 450–454. Berckmans, D. 2008. Precision livestock farming (PLF). Computers and Electron- ics in Agriculture. 62: 1. Bolles, R. C., Fanselow, M.S. 1980. A per- ceptual-defensive-recuperative model of fear and pain. Behavioral and Brain Sciences. 3,2, 291–301. Borderas T. F., Ter Maat S., De Passillé, A.M., Rushen, J., Fournier, A. 2004. Effect of lameness on dairy cows’ fre- quency of visits. to an automatic milking system (AMS). Proceedings of the 38th International Congress of the ISA. 171. Borderas, T.F., De Passillé, A.M., Rushen, J., 2008. Behavior of dairy calves after a low dose of bacterial endotoxin. Journal of Animal Science. 86, 2920–2927. Botreau, R., Veissier I., Perny P. 2009. Overall assessment of cow welfare: strategy adopted in Welfare Quality®. Animal Welfare. 18, 4, 363–370. Broom, D.M. 1996. Animal welfare defined in terms of attempts to cope with the environment. Acta Agriculturae Scandi- navica –Section A: Animal Science Agri- culturae. 27: 22–28. Broom, D.M. 2006. Behaviour and welfare in relation to pathology. Applied Animal Behaviour Science. 97, 73–83. Buhman, M.J., Perino, L.J., Galyean, M.L., Wittum, T.E., Montgomery, T.H., Swingle, R.S. 2000. Association between changes in eating and drinking behav- iors and respiratory tract disease in newly arrived calves at a feedlot. Ameri- can Journal of Veterinary Research. 61, 1163–1168. Cai, T.-Q., Weston, L.A., Lund, L.A., Brodie, B., McKenna, D.J.,Wager, W.C. 1994. Association between PMN function and periparturient disorders in dairy cows . American Journal of Veterinary Re- search. 55,93 4–943 Camiloti , T.V., Fregonesi, J.A., Von Key- serlingk, M.A.G., Weary, D.M. 2012. Short communication: Effects of bed- ding quality on the lying behavior of dairy calves. Journal of Dairy Science. 95, 3380–3383. Cangar, Ö., Leroy, T., Guarino, M., Vranken, E., Fallon, R., Lenehan, J., Mee, J., Ber- ckmans, D. 2008. Automatic real-time monitoring of locomotion and posture behaviour of pregnant cows prior to calving using online image analysis. Computers and electronics in agricul- ture. 6, 4, 53–60. Chapinal, N., De Passillé, A.M., Rushen, J., Wagner, S.2010. Automated methods for detecting lameness and measuring analgesia in dairy cattle. Journal of Dairy Science. 93, 5. Chapinal, N., Tucker, C.B. 2010. Validation of an automated method to count steps while cows stand on a weighing plat- form and its application as a measure to detect lameness. Journal of Dairy Sci- ence. 95, 11. 38 MTT SCIENCE 28 Chapinal, N., De Passillé, A.M., Rushen, J., Tucker, C.B. 2011. Short communica- tion: Measures of weight distribution and frequency of steps as indicators of restless behavior. Journal of Dairy Sci- ence. 94, 2. Chapinal, N., Fitzpatrick, C.E., Leslie, K.E., Wagner, S.A. 2013. Short communica- tion: Experimentally induced mastitis reduces weight shifting between the rear legs while standing in dairy cows. Journal of Dairy Science. 96, 5. Cook, N.B., Bennett, T.B., Nordlund, K.V., 2005. Monitoring indices of cow com- fort in free-stall-housed dairy herds. Journal of Dairy Science. 88, 11. Colak, A., Polat, B., Okumus, Z., Kaya, M., Yanmaz, L.E., Hayirli, A. 2008. Short Communication: Early Detection of Mastitis Using Infrared Thermography in Dairy Cows. Journal of Dairy Science. 91, 11. Cooper, M.D., Arney, D.R., Phillips, C.J.C. 2007. Two- or Four-Hour Lying Deprivation on the Behavior of Lactating Dairy Cows. Jour- nal of Dairy Science. 90, 3. Cyples, J.A., Fitzpatrick, C.E., Leslie, K.E., DeVries, T.J., Haley, D.B., Chapinal, N. 2012. Short communication: The effects of experimentally induced Escherichia coli clinical mastitis on lying behavior of dairy cows. Journal of Dairy Science. 95, 5. Dantzer, R. 2001. Cytokine-Induced Sick- ness Behavior: Where Do We Stand? Brain, Behavior, and Immunity. 15, 7–24. Dawkins, M.S., 2004. Using behaviour to assess animal welfare. Animal Welfare. 13, 3–7. Deming, J.A., Bergeron, R., Leslie, K.E., DeVries, T.J. 2013. Associations of housing, management, milking activity, and standing and lying behavior of dairy cows milked in automatic systems. Journal of Dairy Science. 96, 1. De Mol, R.M., Ouweltjes, W. 2001. Detec- tion model for mastitis in cows milked in an automatic milking system. Preven- tive Veterinary Medicine. 49, 71–82. Dimond S. Lazarus J. 1974. The Problem of Vigilance in Animal Life. Brain Behavior and Evolution. 9,60–79. DeVries, T.J., Beauchemin, K.A., Dohme, F., Schwartzkopf-Genswein, K.S. 2009. Repeated ruminal acidosis challenges in lactating dairy cows at high and low risk for developing acidosis: Feeding, ruminating, and lying behavior. Journal of Dairy Science. 92, 10. DeVries, T.J. ,Deming, J.A, Rodenburg, J., Seguin, G., Leslie, K.E. , Barkema, H.W.2011. Association of standing and lying behavior patterns and incidence of intramammary infection in dairy cows milked with an automatic milking sys- tem. Journal of Dairy Science. 94, 8. Drissler, M., Gaworski, M., Tucker, C.B. and Weary, D.M. 2005. Freestall main- tenance: effects on lying behaviour of dairy cattle. Journal of Dairy Science. 88, 7. © Encyclopedia Britannica, Inc. Sickness. Retrieved April 05, 2014, from Diction- ary.com website:http://dictionary.refer- ence.com/browse/Sickness. Fitzpatrick, C.E., Chapinal, N., Petersson- Wolfe, C.S., DeVries, T.J., Kelton, D.F., Duffield, T.F., Leslie, K.E. 2013. The ef- fect of meloxicam on pain sensitivity, ru- mination time, and clinical signs in dairy cows with endotoxin-induced clinical mastitis. Journal of Dairy Science. 96, 5. Fogsgaard, K.K., Røntved, C.M., Sørensen, P., Herskin, M.S. 2012. Sickness behav- ior in dairy cows during Escherichia coli mastitis. Journal of Dairy Science. 95, 2. Fraser, D., Weary, D.M., Pajor, E.A., Mil- ligan, B.N. 1997. A scientific conception of animal welfare that reflects ethical concerns. Animal Welfare. 6, 187–205. Fregonesi, J.A., Leaver, J.D. 2001. Behav- iour, performance and health indica- tors of welfare for dairy cows housed in strawyard or cubicle systems. Livestock Production Science. 68, 2–3. Fregonesi, J.A., Tucker, C.B., Weary, D.M. 2007. Overstocking Reduces Lying Time in Dairy Cows. Journal of Dairy Science. 90, 7. Galindo, F., Broom, D.M. 2002. The ef- fects of lameness on social and indi- vidual behaviour of dairy cows. Journal of Applied Animal Welfare Science. 5, 193–201. Gonzales, L.A., Tolkamp, B.J., Coffey, M.P., Ferret, A., Kyriazakis, I. 2008. Changes MTT SCIENCE 28 39 in feeding behavior as possible indica- tors for the automatic monitoring of health disorders in dairy cows. Journal of Dairy Science. 91,3. Haley, D.B., Rushen, J., De Passillé, A.M. 2000. Behavioural indicators of cow comfort: activity and resting behaviour of dairy cows in two types of housing. Canadian Journal of Animal Science. 80, 257–263. Halasa, T., K. Huijps, O. Østerås, and H. Hogeveen. 2007. Economic effects of bovine mastitis and mastitis manage- ment: A review. The Veterinary Quarterly. 29,18–3. Haley, D.B., De Passillé, A.M., Rushen, J., 2001. Assessing cow comfort: ef- fects of two floor types and two tie stall designs on the behaviour of lactating dairy cows. Applied Animal Behaviour Science. 71, 105–117. Hamann,  J., Halm,  H., Redetzky,  R., Grabowski, N.T.H. 2004. Influence of incomplete milkings on milk yield under automatic milking conditions (VMS®). In:  Meijering, A., Hogeveen, H., de Koning,  C.J.A.M. Editors. Automatic Milking: A Better Understanding. Wage- ningen, the Netherlands: Wageningen Academic Publishers. 490–491. Hart, B.L. 1988. Biological basis of the be- havior of sick animals. Neuroscience & Biobehavioral Reviews. 12, 2. Heikkilä, A.-M., Nousiainen, J.I., Pyörälä, S. 2012. Costs of clinical mastitis with special reference to premature culling. Journal of Dairy Science. 95, 1. Hemsworth, P.H., Coleman, G.J. 1998. Hu- man–Livestock Interactions: the Stock- person and the Productivity and Welfare of Intensively-Farmed Animals. CAB In- ternational, Oxon, UK. Hovinen, M., Aisla, A.-M., Pyörälä, S. 2005. Visual Detection of Technical Success and Effectiveness of Teat Cleaning in Two Automatic Milking Systems. Journal of Dairy Science. 88, 9. Hovinen, M., Rasmussen, M.D., Pyörälä, S. 2009. Udder health of cows changing from tie stalls or free stalls with con- ventional milking to free stalls with ei- ther conventional or automatic milking. Journal of Dairy Science. 92, 8. Hovinen, M., Pyörälä, S. 2011. Review Article: Invited review: Udder health of dairy cows in automatic milk- ing. Journal of Dairy Science. 94, 2. Huzzey, J.M., Veira, D.M., Weary, D.M., Von Keyserlink, M.A.G., 2007. Prepar- tum behavior and dry matter intake identify dairy cows at risk for metritis. Journal of Dairy Science. 90, 7. International Dairy Federation. 1999. Sug- gested interpretation of mastitis termi- nology. In IDF Bulletin 338, 3–26, Bel- gium. Ito, K., Von Keyserlingk, M.A.G., LeBlanc, S.J., Weary, D.M.2010. Lying behavior as an indicator of lameness in dairy cows. 2010. Journal of Dairy Science. 93, 8. Jacobs, J.A., Siegford, J.M. 2012. Invited review: The impact of automatic milk- ing systems on dairy cow management, behavior, health, and welfare. Journal of Dairy Science. 95,5. Jensen, M.B., Pedersen, L.J., Munks- gaard, L., 2005. The effect of reward duration on demand functions for rest in dairy heifers and lying requirements as measured by demand functions. Applied Animal Behaviour Science. 90, 207–217. Johnson, R.W., 2002. The concept of sick- ness behavior: a brief chronological ac- count of four key discoveries. Veterinary Immunology and Immunopathology. 87, 443–450. Jones, B.F., Plassmann, P. 2002. Digital in- frared thermal imaging of human skin. Engineering in Medicine and Biology Magazine, IEEE. 21, 6, 41–48. Juarez, S.T., Robinson, P.H., DePeters, E.J., Price, E.O. 2003. Impact of lameness on behaviour and productivity of lactat- ing Holstein cows. Applied Animal Be- haviour Science. 83, 1–14. Kamphuis, C., Pietersma, D., Van der Tol, R., Wiedemann, M., Hogeveen, H. 2008a. Using sensor data patterns from an automatic milking system to develop predictive variables for classifying clini- cal mastitis and abnormal milk. Com- puters and Electronics in Agriculture. 62,169–181. Kamphuis, C., Sherlock, R., Jago, J., Mein, G., Hogeveen, H. 2008b. Automatic de- tection of clinical mastitis is improved by 40 MTT SCIENCE 28 in-line monitoring of somatic cell count. Journal of Dairy Science. 91, 12. Kemp, M.H., Nolan, A.M., Cripps, P.J., Fitzpatrick, J.L. 2008. Animal-based measurements of the severity of masti- tis in dairy cows. Veterinary Record. 163, 175–179. Ketelaar-de Lauwere, C.C., Ipema, A.H., Van Ouwerkerk, E.N.J., Hendriks, M.M.W.B., Metz, J.H.M., Noordhuizen, J.P.T.M., Schouten, W.G.P. 1999. Volun- tary automatic milking in combination with grazing of dairy cows Milking fre- quency and effects on behavior. Applied Animal Behaviour Science. 64, 91–109 Konsman, J.P., Parnet, P. and Danzer, R. 2002. Cytokine-induced sickness be- haviour: mechanisms and implications. Trends in Neurosciences. 25, 3. Lidfors, L., 1989. The use of getting up and lying down movements in the evaluation of cattle environments. Veterinary Re- search Communications. 13, 307–324. Medrano-Galarza, C., Gibbons, J., Wagner, S., De Passillé, A.M., Rushen, J. 2012. Behavioral changes in dairy cows with mastitis. Journal of Dairy Science. 95, 12. Miller, N.E. 1964. Some psycho-physio- logical studies of motivation and of the behavioral effects of illness. Bulletin of British Psychological Society. 17, 1–20. Miguel-Pacheco, G.G., Kaler, J., Remnant, J., Cheyne, L., Abbott, C., French, A.P., Pridmore, T.P., Huxley, J.N., 2014. Be- havioural changes in dairy cows with lameness in an automatic milking sys- tem. Applied Animal Behaviour Science. 150, 1– 8. Molony, V., Kent, E.J. 1997. Assessment of acute pain in farm animals using behav- ioral and physiological measurements. Journal of Animal Science. 75, 266–272. Molony, V., Kent, J.E. and McKendrick, I.J. 2002. Validation of a method for assess- ment of an acute pain in lambs. Applied Animal Behaviour Science. 76, 212–238. Munksgaard, L., Jensen, M.B., Pedersen, L.J., Hansen, S.W., Matthews, L. 2005. Quantifying behavioural priorities- ef- fects of time constraints on behaviour of dairy cows, Bos taurus. Applied Animal Behaviour Science. 92, 3–14. Munksgaard, L., Rushen, J., De Passillé, A.M., Krohn, C.C. 2011. Forced versus free traffic in an automated milking sys- tem. Livestock Science. 138, 244–250 Niss, D.B., Herskin, M.S., Danscher, A.M., Thoefner, M.B. 2009. Short communica- tion: Rising and lying behavior of heifers before and after alimentary oligofruc- tose overload. Journal of Dairy Science. 92, 617–620. Norring, M., Valros, A., Munksgaard, L. 2012. Milk yield affects time budget in dairy cows kept in tie-stalls. Journal of Dairy Science. 95, 1. Pastell, M., Takko, H., Gröhn, H., Hautala, M., Poikalainen, V., Praks, J., Veermae, I., Kujala, M., Ahokas, J. 2006. Assess- ing Cows’ Welfare: weighing the Cow in a Milking Robot. Biosystems Engineer- ing. 462, 93, 81–87. Pastell, M., Kujala, M. 2007. A probabilistic neural network model for lameness de- tection. Journal of Dairy Science. 90, 5. Pastell, M., Hautala, M., Poikalainen, V., Praks, J., Veermae, I., Kujala, M., Aho- kas, J. 2008. Automatic observation of cow leg health using load sensors. Computers and Electronics in Agricul- ture. 62, 48–53 Pastell, M., Tiusanen, J., Hakojärvi, M., Hän- ninen, L. 2009. A wireless accelerom- eter system with wavelet analysis for as- sessing lameness in cattle. Biosystems Engineering. 104, 545–551. Plesch,G., Broerkens, N., Laister, S., Winckler,C., Knierim, U. 2010. Reliabil- ity and feasibility of selected measures concerning resting behaviour for the on- farm welfare assessment in dairy cows. Applied Animal Behaviour Science. 126, 19–26. Polat, B., Colak, A., Cengiz, M., Yanmaz, L.E., Oral, H., Bastan, A., Kaya, S., Hay- irli, A. 2010. Sensitivity and specificity of infrared thermography in detection of subclinical mastitis in dairy cows. Jour- nal of Dairy Science. 93, 8, 3525–3532. Quimby, W.F., Sowell, B.F. Bowman, J.G.P., Branine, M.E., Hubbert, M. E., Sher- wood. H.W. 2001. Application of feeding behaviour to predict morbidity of newly received calves in a commercial feedlot. Canadian Journal of Animal Science. 81, 3, 315–320. Radostits, O.M., Gay, C.C., Hinchcliff, K.W. MTT SCIENCE 28 41 et al. (Eds). 2007. Veterinary Medicine – A Textbook of the Diseases of Cattle, Horses, Sheep, Pigs, and Goats. 10. ed. Philadelphia: Saunders. 673–762. Rasmussen, M. D., Bennedsgaard, T.W., Pedersen, L.H. 2007. Changes in quar- ter yield and milking frequency during clinical mastitis. NMC Annual Meeting Proc., San Antonio, TX. National Mastitis Council, Madison, WI, USA. 214–215. Rousing, T., Bonde, M., Badsberg, J.H., Sø- rensen, J.T. 2004. Stepping and kicking behaviour during milking in relation to response in human-animal interaction test and clinical health in loose housed dairy cows. Livestock Production Sci- ence. 88, 1–8. Rousing T, Badsberg, J.H., Klaas, I.C., Hind- hede, J., Sørensen, J.T. 2006. The as- sociation between fetching for milking and dairy cows’ behaviour at milking, and avoidance of human approach- An on-farm study in herds with automatic milking systems. Livestock Science. 101, 219–227. Ruckebush, Y., 1974. Sleep deprivation in cattle. Brain Research. 78, 495–499. Rutten, C.J., Velthuis, A.G.J., Steeneveld, W., Hogeveen, H. 2013. Invited review: Sensors to support health management on dairy farms. Journal of Dairy Science. 96, 4. Saraiva Martins, R.F., Do Prado Paim, T., De Abreu Cardoso, C., Lima Dallago, B.S., Barros de Melo, C., Louvandini, H., McManus, C. 2013. Mastitis detec- tion in sheep by infrared thermography. Research in Veterinary Science. 94, 722–724. Schirmann, K., Chapinal, N., Weary, D.M., Heuwieser, W., Von Keyserlingk, M.A.G. 2012. Rumination and its relationship to feeding and lying behavior in Holstein dairy cows. Journal of Dairy Science. 95, 6. Song, X.Y., T. Leroy, E. Vranken, W. Mae- rtens, B. Sonck, and D. Berckmans. 2008. Automatic detection of lameness in dairy cattle—Vision-based trackway analysis in cow’s locomotion. Comput- ers and Electronics in Agriculture. 64, 39–44. Sowell, B.F., Bowman, J.G.P., Branine, M.E., Hubbert, M.E. 1998. Radio frequency technology to measure feeding behav- ior and health of feedlot steers. Applied Animal Behaviour Science. 59, 277–284. Stelwagen, K., Lacy-Hulbert, S.J. 1996. Ef- fect of milking frequency on milk somatic cell count characteristics and mammary secretory cell damage in cows. Ameri- can Journal of Veterinary Research. 57, 902–905. Svensson, C., Jensen, M.B. 2007. Short Communication: Identification of Dis- eased Calves by Use of Data from Au- tomatic Milk Feeders. Journal of Dairy Science. 90, 2. Tom, E.M., Duncan, I.J.H., Widowski, T.M., Bateman, K.G., Leslie, K.E. 2002. Ef- fects of tail docking using a rubber ring with or without anesthetic on behaviour and production of lactating cows. Jour- nal of Dairy Science. 85, 9. Tucker, C.B., Weary, D.M., Von Keyserlingk, M.A.G., Beauchemin, K.A. 2009. Cow comfort in tie-stalls: Increased depth of shavings or straw bedding increases ly- ing time. Journal of Dairy Science. 92, 6. Urton, G., Von Keyserlink, M.A.G. and Wea- ry, D.M. 2005. Feeding behavior identi- fies dairy cows at risk for metritis. Jour- nal of Dairy Science. 88, 8. Yeiser, E.E., Leslie, K.E., McGilliard, M.L., Petersson-Wolfe, C.S. 2012. The effects of experimentally induced Escherichia coli mastitis and flunixin meglumine ad- ministration on activity measures, feed intake, and milk parameters. Journal of Dairy Science. 95, 9. Walker, S.L., Smith, R.F., Routly, J.E., Jones, D.N., Morris, M.J., Dobson, H. 2008. Lameness, Activity Time-Budgets, and Estrus Expression in Dairy Cattle. Jour- nal of Dairy Science. 91, 4552–4559. Weary, D.M., Huzzey, J.M., Von Keyser- lingk, M.A.G. 2009. BOARD-INVITED REVIEW: Using behavior to predict and identify ill health in animals. Journal of Animal Science. 87, 770–777. Webster, A.J.F., Main, D.C.J., Whay, H.R., 2004. Welfare assessment: indices from clinical observation. Animal Welfare. 13, 98. Wierenga, H.K., Hopster, H. 1991. Behaviour of dairy cows when fed concentrates with an automated feeding system. Applied Animal Behaviour Science. 30, 223–246. MTT is publishing its research findings in two series of publications: MTT Science and MTT Growth. The MTT Science series includes scientific presentations and abstracts from conferences arranged by MTT Agrifood Research Finland. Doctoral dissertations by MTT research scientists will also be published in this series. The topics range from agricultural and food research to environmental research in the field of agriculture. MTT, FI-31600 Jokioinen, Finland. email julkaisut@mtt.fi 28 Dairy cow behaviour in relation to health, welfare and milking Doctoral Dissertation Jutta Johanna Kauppi MTT CREATES VITALITY THROUGH SCIENCE www.mtt.fi/julkaisut