FEEDING BEHAVIOUR

Changes in dry matter intake (DMI) and feeding time on the day of calving have been well documented in dairy cows housed within intensive (indoor) group-housed sys tems (e.g., freestalls barns, compost barns). It has been reported that cows decreased their DMI by approximately 30% on the day of calving (^{Huzzey et al 2007}, ^{Proudfoot et al 2009}, ^{Schirmann et al 2013}) compared to the previ-ous day. This decrease occurred from 24 h to 6 h before parturition (Table 1). Time spent at the feed bunk on the day of calving follows a similar pattern as DMI (Huzzey et al 2007, Schirmann et al 2013, ^{Büchel and Sundrum 2014}), however, with a larger variation between individual cows (Miedema et al 2011^a, ^{Jensen 2012}).

Reduction in DMI and feeding time a few hours before calving may be due to a shift in the motivational priorities of the cow, as hypothesized by ^{Proudfoot et al (2009}), or due to pain and discomfort associated with the onset of parturition (^{Huzzey et al 2007}). More recently, ^{Neave et al (2017}) reported that primiparous dairy cows consumed less feed than multiparous cows during the days around calving (d -1 to d +1). Additionally, primiparous cows spent more time at the feed bin and were more often replaced (i.e., cases in which a cow replaced another cow by physically making contact and taking her place at the feed bin) than multiparous cows in competitive interactions, reducing their ability to access the feed. This finding suggests that primiparous cows may shift their feeding times, visiting the feed bin when occupancy is lower, and thus avoiding competition with multiparous animals. Further research is necessary to understand how competitive interactions differentially affect the feeding behaviour of primiparous and multiparous animals as calving approaches.

Automated systems for measuring feed intake and feeding behaviour are more applicable to confinement housing systems where cows a receive total mixed ration, and thus it is not surprising that the majority of research has been conducted on cows housed in these types of systems. Feeding behaviour monitors are primarily used in research settings (^{DeVries et al 2003}, ^{Chapinal et al 2007}), but commercially available feeding behaviour quantification methods have recently been developed and evaluated (^{Bikker et al 2014}, ^{Borchers et al 2016}). Feeding behaviour monitoring is more challenging in grazing systems since visually observing grazing behaviour is time-consuming and open to the interpretation of the observer. Although there is some equipment available to identify grazing activity such as leg switch movement, jaw movement (^{Umemura et al 2009}) and biting and chewing sounds (^{Laca and Wallis DeVries 2000}), with additional information regarding bite mass and intake rate (^{Rutter et al 1997}), most of these techniques are expensive and labour intensive. ^{Delagarde and Lamberton (2015}) using affordable sensor equipment (a portable accelerometer that records physical activity level) in their study and conclud-ed that this type of technology is indeed a useful tool for the continuous automatic recording of grazing behaviour in dairy cows at pasture. To our knowledge, there are no electronic systems commercially available that will allow for detailed recording of the feeding behaviour of grazing cows, especially at calving time.

RUMINATION BEHAVIOUR

Changes in rumination behaviour have been evaluat-ed as a signal of impending calving for cows housed in confinement (^{Soriani et al 2012}, ^{Schirmann et al 2013}, ^{Calamari et al 2014}, ^{Pahl et al 2014}, ^{Büchel and Sundrum 2014}, ^{Ouellet et al 2016}, ^{Borchers et al 2017}). Soriani et al (2012) observed that rumination time declined progressively in the week before parturition, moreover, on the day of calving it decreased 3 h compared with the dry period. In concordance, Calamari et al (2014) found that the decrease in rumination time on the day of calving amounted to 70% on average of the value observed during the dry period. Table 1 shows that rumination time declines in the hours before parturition, supporting the hypothesis that rumination time measured on an hourly scale is a useful short-term predictor of calving.

Data on rumination behaviour of grazing dairy cows affected by the onset of calving is limited and the few available studies show results similar to those reported in dairy cattle in group-housed systems. In both confinement (^{Ouellet et al 2016}) and pasture-based systems (^{Clark et al 2015}), pregnant dairy cows ruminate less during morning and afternoon hours than in the evening and night in the days before parturition, which is in line with the diurnal normal rumination pattern of cows (^{Gregorini et al 2012}). These results suggest that the reduction in rumination time during the period before calving might be independent of the type of production system (indoor versus pasture). Further investigations are needed to quantify the behavioural differences within a few hours before calving since they would be useful to identify predictors of the onset of calving in grazing cows.

Changes in rumination time at calving might be affected by parity, however, the literature on this topic is scarce. It was reported that primiparous Italian-Friesian dairy cows housed in a freestall barn spent less time ruminating than multiparous cows in the days before calving and at calving time, being this difference particularly pronounced in the week after parturition (^{Soriani et al 2012}). These findings suggest that first lactation cows have more difficulties to cope with calving and the onset of lactation. Daily rumination time differs substantially between housed (^{Schirmann et al 2012}) and pasture-based grazing animals (^{Gregorini et al 2012}). Thus, future work should examine how rumination behaviour is affected by parity under pasture-based conditions.

In recent years, technology that allows rumination time to be easily quantified has become available (^{Clark et al 2015}, ^{Ouellet et al 2016}, ^{Borchers et al 2017}), and their performance for automated calving prediction has been recently reviewed (^{Saint-Dizier and Chastant-Maillard 2018}). Ouellet et al (2016) evaluated the performance of a 3-dimensional accelerometer attached to the ear tag to predict the onset of calving within 24 h, 12 h, and 6 h in housed dairy cows. The study showed that the reduction in rumination time was more accurate to predict calving within 6 h before parturition than 12 h or 24 h, however, with lower performance (63% sensitivity and specific-ity). Borchers et al (2017), using a collar consisting of an accelerometer and a microphone and applying three machine-learning techniques, found a greater level of sensitivity (72% to 79%) and specificity (80% to 89%) to predict calving within 8 h in housed dairy cows. In pasture-based dairy cows, Clark et al (2015) used a similar collar and achieved 70% sensitivity and 70% specificity in predicting the day of calving.

LYING BEHAVIOUR AND ACTIVITY

Reduction in lying time, more frequent posture chang-es and increase in walking activity are among others the most frequently reported behaviours in prepartum cows (R0rvang et al 2018^a). During the last 24 h before parturi tion, cows managed in indoor housed conditions spent less time lying down compared with previous days, showing a greater number of shorter duration lying bouts - transi-tions between standing and lying - per day (Miedema et al 2011^a, ^{Jensen 2012}, ^{Ouellet et al 2016}). Similarly, cows on pasture also showed shorter lying times around calving (^{Rice et al 2017}, ^{Sepúlveda-Varas et al 2018}, ^{Hendriks et al 2019}) (table 1).

An increased frequency of lying bouts has been ob-served in the 6 h prior to the onset of parturition (Miedema et al 2011^a, ^{Jensen 2012}, ^{Ouellet et al 2016}), reflecting the increased degree of restlessness (i.e., characterised by increased frequency of postural changes) (Miedema et al 2011^a, Jensen 2012) and the growing discomfort of the cow associated with the first stage of parturition (Ouellet et al 2016).

An increase in cow activity as calving approaches has been reported in animals housed in both confinement (Miedema et al 2011^a, ^{Jensen 2012}, ^{Titler et al 2015}, ^{Black and Krawczel 2016}, Borches et al 2017) and pasture settings (Black and Krawczel 2016, ^{Hendriks et al 2019}). However, comparison between studies is difficult because of the different methodological approaches to defining active behaviour. Miedema et al (2011) used continuous focal observations from video to quantify daily durations of walking and observed that cows walked for longer during the 24 h prior to calving compared to a 24-hour control period during late pregnancy. Jensen (2012) measured an activity index using an accelerometer attached to the cows' legs and reported that cows were more active the day before calving compared to 2-4 d before calving and, on an hourly scale, the level of activity increased throughout the 6 h prior to calving compared with the same time of day in the 3 preceding days. Titler et al (2015) monitored cow activity by counting steps using electronic data loggers and reported an increase in the number of steps within the 12 h immediately prior to calving. ^{Borchers et al (2017}) measured the prepartum activity using two types of de vices. Firstly, they used data loggers and showed that the number of steps taken by the animals increased 8 h before calving, compared with a baseline period. Secondly, neck movements were measured using 3-axis accelerometers and it was observed that the overall neck activity decreased to its lowest value 18 h before calving, and then increased to its highest value 2 h before calving. More recently, the work of Hendriks et al (2019) showed an increased activity, measured by the number of steps taken using electronic data loggers, on the day of calving in grazing dairy cows.

Lying behaviour and activity at calving seem to be influenced by the housing system. ^{Black and Krawczel (2016}) observed that cows lay down more frequently on pasture during the day of calving when compared to freestall conditions, suggesting that cows may find pasture a more comfortable surface for changing lying position compared to the mattresses commonly used in freestall calving pens. On pasture, the time required for grazing requires an increase in walking and, therefore, an increase in daily steps. In prepartum beef cows, it has been proposed that the increase in steps may also be due to cows walking in search for a safe place to calve, seeking isolation from the herd, or pacing due to discomfort (^{Duncan and Meyer 2019}). Thus, step count differences may be detected in dairy cows kept in open grazing areas more readily than cows kept in free stalls barns because of differences in the environmental characteristics. On the contrary, another study found no change in lying bouts and numbers of steps during the days before calving in cows housed in the pasture which, according to the authors, may indicate that the pasture provides an adequate environment for cows to lie down (^{Rice et al 2017}).

Differences in lying time (Miedema et al 2011^b, ^{Titler et al 2015}), frequency of transitions from lying to standing (^{Neave et al 2017}) and activity (Titler et al 2015, ^{Borchers et al 2017}) have been reported between primiparous and multiparous cows in freestall housing conditions. Primiparous cows spent less time lying compared with multiparous cows during the final 24 h preceding calving (453 ± 44 min/d vs. 598 ± 40 min/d, respectively; Titler et al 2015) and 2 h prior to calving (Miedema et al 2011^b). Authors speculated that this difference might be related to longer labour time and a higher number of contractions during parturition in primiparous cows (Miedema et al 2011^b, ^{Schuenemann et al 2011}). Primiparous cows also lay down more frequently, but for shorter periods, during the week before calving and during the period between the day before and after calving (Neave et al 2017) and became more active compared to multiparous cows in the last 6 h before the onset of parturition (Borchers et al 2017). It has been hypothesised that these behavioural differences related to parity may reflect an increased discomfort in primiparous cows as parturition approaches.

Relatively few studies looked at differences in lying behaviour of primiparous and multiparous dairy cows at calving in pasture-based systems. It was reported that first parity cows have lower daily lying times and higher frequency of lying bouts than multiparous cows during the first week after calving (^{Sepúlveda-Varas et al 2014}). Moreover, increasing parity was associated with lower step counts during the period immediately before and after calving (^{Hendriks et al 2019}). These findings indicate that parity number affects the frequency of lying behaviour and activity around parturition; thus, parity needs to be con-sidered in the analyses of cow behaviour at calving time.

An advantage of measuring lying behaviour and activity is that the available technology is practical and affordable. Although increasing the number of validating sensors are available at the market which can automatically measure lying behaviour (^{McGowan et al 2007}, ^{Ledgerwood et al 2010}, ^{Mattachini et al 2013}, ^{Borchers et al 2016}) and activity (^{Champion et al 1997}, ^{Robert et al 2009}, ^{Bikker et al 2014}), few studies have attempted to evaluate the performance of these automated sensors to predict calving events using these measures. ^{Titler et al (2015}) used an activity index combining the number of steps, lying bouts, and standing time measured by an electronic data logger that was able to predict parturition with a time interval greater than 4 hours in 76% of the primiparous and multiparous Holstein cows evaluated in freestall barns. Using the same device, ^{Ouellet et al (2016}) evaluated lying time and lying bouts in housed cows (freestalls) to check the accuracy in calving prediction and found that a combination of these variables had a greater level of prediction accuracy within the next 6 h than considering the next 12 h and 24 h, but with high rates of false negatives and false positives (58% sensitivity, 61% specificity). Similarly, Borchers et al (2017) described calving prediction methods applying three machine-learning techniques to a combination of collected lying behavioural variables (numbers of steps, lying bouts and lying time) during the 14 prepartum days and achieved a sensitivity of 38% to 75% and a specificity of 88% to 91% in predicting the day of calving in housed dairy cows.

Since there are favourable results presented in the scientific literature on the automated detection of lying behaviours and activity to predict calving in dairy cows under confinement (^{Saint-Dizier and Chastant-Maillard 2018}), future research should investigate whether the changes in daily lying time, lying bouts, or activity around the calving event could be used to predict the timing of calving in grazing cows.

OTHER BEHAVIOURS

In cattle, tail raising is mentioned as a sign of im-minent calving (^{Owens et al 1985}, ^{Lidfors et al 1994}, ^{Wehrend et al 2006}, Miedema et al 2011^a, ^{Lange et al 2017}). The frequency and duration of tail raising increase significantly the day before calving and peaks during the 6 h prior to parturition (Miedema et al 2011^a). There is a growing consensus that raising the tail is one of the most consistent behavioural changes observed in the hours prior to parturition (Wehrend et al 2006, Miedema et al 2011^a, Lange et al 2017), making it a potentially useful indicator of calving. To date, several systems for calving prediction including inclinometers and accelerometers that detect tail raising are available (^{Mee et al 2019}), but the performance of these marketed devices under field conditions still lack scientific support (^{SaintDizier and Chastant-Maillard 2015}).

^{Jensen et al (2012}), using continuous observation during the final 12 h prior to calving, reported that multiparous cows frequently turned their head toward their abdomens, behaviour that increased during the final 2 h prior to calv-ing and typically occurred during contractions. Pain and discomfort associated with abdominal contractions during the second stage of parturition (^{Mainau and Manteca 2011}) likely explain that cows might be turning their head due to pain experienced during contractions as described by Jensen et al (2012), especially during the last hours before calving. On the contrary, ^{Lange et al (2017}) used repeated observations lasting 15 s every hour during the 24 h period before calving and observed that turning the head toward the abdomen did not occur often and did not increase before parturition in primiparous cows. These contradictory results may be explained by the different observational methods used since this behavioural sign lasts a few seconds and may be better detected during continuous observation rather than during repeated observations.

The literature also reports that cows display other be-haviours before parturition such as stepping (^{Lange et al 2017}), sniffing or licking at the ground (Miedema et al 2011^a), and an increased amount of time spent in lateral recumbency (^{Schuenemann et al 2011}). More recently, Lange et al (2017) built mathematical equations including different behaviours recorded via direct observations (i.e. tail raising, stepping, turning the head toward the abdomen, and lying lateral with abdominal contractions) to predict the second stage of parturition in dairy cows. These re-searchers found that the sensitivity and specificity were 69 and 88%, respectively, for predicting calving within 12 h for heifers (n = 29) examined from d 269 until calving and the positive predictive values were between 35% and 73% depending on the day of gestation, and concluded that predicting the second stage of calving by direct ob-servation of plausible signs is imprecise and therefore not recommendable.

MANAGEMENT PRACTICES AND ENVIRONMENTAL FACTORS AFFECTING BEHAVIOUR AT CALVING

It is well established that the provision of a comfort-able environment around parturition minimises the risk of dystocia and enhances the subsequent health of the cow and calf (^{Mee 2004}). In recent years, there has been an increased interest in housing and management practices to ease the calving process. Next, we discuss behavioural responses to common farm management practices and housing conditions including the time of moving a cow to the calving area, the social environment, and the use of a secluded area to calve. We also review the effects of cow exposure to environmental stressors such as hot and cold weather near to calving time on dairy cow's behaviour, with particular emphasis on pasture-based systems.

TIME OF MOVING A COW TO THE CALVING SITE

Intensively managed dairy farms, cattle is generally moved to a calving area when calving is imminent, and this decision is often based on their due date and/or calving signs. ^{Mee (2004}) recommended moving cows to a calving area within 1 or 2 d before calving to allow animals to adapt to their new environment, social group, and diet. This management practice may be particularly important in heifers, where such stresses can negatively impact the calving process. ^{Heuwiser et al (1987}) reported that cows that were moved into a maternity pen 3 d before calving had lower blood concentrations of cortisol at calving compared with cows moved close to calving, suggesting a lower stress level in cows that had time to adapt to the new environment of the maternity pen. Additionally, ^{Mee et al (2013}) in a case-control study, using 30 dairy Irish herds, found that the herds with cows being transferred to the calving unit within 12 h before calving had a higher risk of perinatal mortality than herds moving cows 12 to 24 h, or more than 2 days before calving.

To date, only ^{Proudfoot et al (2013}) have studied the effect of moving cows from a group pen (bedded pack) to a calving area on the expression of normal calving behaviour, reporting that multiparous cows that were moved to an individual maternity pen closer to the time of calving (late first stage of parturition; moved about 2 h before calving and showing imminent signs of parturition, such as bloody mucous outside of the vulva and the start of abdominal contractions), experienced lower lying time and shorter lying bouts than cows moved with early signs of calving (early first stage of parturition; moved about 12 h before calving, showing signs of early first stage of parturition, such as raised tail, engorged, tense and leaky udder, and/or relaxed pelvic ligaments), or before any signs (moved on average 74 h before calving). These be-havioural changes may be related to cows spending more time exploring their new environment to ensure that it is a safe place to give birth.

The timing of movement to a calving area is also a management practice that affects the ease of calving and health outcomes. ^{Proudfoot et al (2013}) showed that moving multiparous cows in the advanced first stage of parturition may interrupt and prolong the second stage by approximately 30 min. Also, cows with a longer second stage of parturition had higher inflammation levels (mea-sured using haptoglobin) after calving compared with cows moved earlier. Cows experiencing prolonged labour may also be experiencing additional pain and restless behaviour (^{Mainau and Manteca 2011}), increased need for assistance, and are at increased risk of dystocia (^{Schuenemann et al 2011}, ^{Barrier et al 2012}). Similarly, ^{Carrier et al (2006}) reported a 2.5-fold increase in stillbirths in cows that were moved during the late part of the first stage of parturition. These findings suggest that moving a calving cow to an individual pen late in the birth process should be avoided as it represents a stressor capable of altering behaviour and affect their ability to calve normally.

Cows can also be moved too early to the calving area. In a recent study, farmers were asked about the moment when the cows and heifers were brought to the calving area, and an overall 42% of the producers said they brought their cows at the start of the prepartum period, 3 wk before their due date, while 25% said they moved their cows when the first signs of calving were detected (^{Villettaz-Robichaud et al 2016}). However, first signs of calving could be anything from first to the second stage of labour (i.e. restlessness and udder fill to bloody mucus, amniotic sac, or part of the calf emerging from the vagina). Field observations have also reported that cattle are at risk of developing postpartum health disorders such as ketosis and displaced abomasum and elevated blood non-esterified fatty acid concentrations when they remain in an individual calving pen for longer than 3 d (^{Cook and Nordlund 2004}). It has been proposed that prolonged isolation may be stressful to cattle, which are naturally gregarious, increasing stress and the risk of disease (^{Sepúlveda-Varas et al 2013}). To our knowledge, this approach has not been tested in dairy cows.

SOCIAL ENVIRONMENT

In intensive indoor dairy production systems, cows usually calve in individual calving pens or group calv-ing pens, depending on the practices related to calving management of the farm (^{Cook and Nordlund 2004}, ^{Villettaz-Robichaud et al 2016}). The use of individual calving pens has been recommended to protect the dam from potential social disturbances during calving and to reduce the risk of mismothering (^{Edwards 1983}, ^{Illmann and Spinka 1993}). Edwards (1983) observed that when calving in groups, cows found it more difficult to main-tain proximity to their calf and spent less time licking their calves compared with cows that calved alone. Also, approximately half of cows showed interest in an alien calf (a calf born from another cow which was not yet removed) before giving birth themselves. Interestingly, these behaviours did not produce the rejection of the cow's own calf, but on the contrary, could lead to the adoption of the alien calf (Illmann and Spinka 1993).

Larger commercial dairy farms are more likely to use group pens than individual pens (^{Villettaz-Robichaud et al 2016}), which may not be consistent with the choice of the cows. R0rvang et al (2018^b) examined whether periparturient multiparous dairy cows would voluntarily move into individual calving areas placed in a group calving location (referred to as motivation-based calving facilities). In this study, about half of the cows moved away from the group and calved in an individual calving area (34 of 66), with social factors having an important influence on the chance of a cow using the individual calving area; dominant cows were most likely to use the individual calving areas, whereas the presence of a calf born from another cow reduced the chance of cows calving inside the individual calving area.

In group calving areas, the choice of calving site may be also affected by the site of previous calving. In a study, a group of 10 cows was followed recording the location of the breaking of the amniotic sac as well as the place of birth and found that 9 out of the 10 cows calved within a distance of 1 cow length from where the previous calving took place, suggesting that the cows did not select calving site at random and potentially explained by the presence of amniotic fluid from other cows (R0rvang et al 2017^a).

USE OF A SECLUDED AREA TO CALVE

The use of secluded areas in confinement calving settings has been the subject of considerable interest and revision over the last 5 years (Proudfoot et al 2014^a,b, R0rvang et al 2017^b, 2018^a,b). Proudfoot et al (2014^a,b) provided the first evidence that indoor-housed cows preferred a secluded environment to calve, particularly when they were housed individually and there was a high level of activity in the barn during the daytime. R0rvang et al (2017^b) investigated dairy cows' preference regarding the degree of isolation at calving in an individual pen by offering different types (in terms of height, width or both) and levels of isolation (in terms of percentage coverage, 50% or 75%). Results showed that periparturient dairy cows had no preference for a specific calving pen design or level of isolation. However, cows experiencing a longer calving duration (longer second stage labour) were more likely to give birth in the most isolated individual calving pen (75% coverage) compared with those without prolonged labour, possibly explained by an increased level of restlessness and increased discomfort during calving that may have motivated them to seek further seclusion. In a follow-up study, researchers examined whether dairy cows would voluntarily isolate in an individual calving pen placed in a group calving setting were access to individual calving pens was free (with a gate permanently open) or for only one cow at a time (with a gate closed that a cow could push through) (R0rvang et al 2018^b). Results showed that 50% of the cows calved in an individual pen, regardless of the presence of a functional gate in the pen or not. Cows housed with permanently open gates tended to be more likely to calve in the individual calving pens compared to cows housed with functional gates, suggesting that cows in labour and experiencing discomfort or pain, may have perceived the gates as obstacles rather than an advantage for isolating before calving.

Under natural extensive pasture environment, cattle search for secure places to hide at calving to find protec-tion from predators and bond with the calf (Rorvang et al 2018^b). The early work of ^{Lidfors et al (1994}) reported that some dairy and beef cows isolated themselves before calving but this behaviour only occurred when suitable conditions were present, such as tall grass or tree cover with appropriate grazing sites nearby. Moreover, most cows made use of a shelter for calving when this was available. However, under an intensive pastoral grazing systems, prepartum group cows are commonly managed on small paddocks sizes (defined grazing areas) that are rotationally grazed in order to provide cows with access to a fresh allocation of pasture (^{Roche et al 2005}); in these systems, the majority of calvings occur within the paddock, where there are fewer opportunities to separate and hide from the herd compared to an extensive, pas-ture-based system. A recent study evaluated dairy cow's preference for calving environment in group-housed in a pasture setting (^{Edwards 2018}). Cows had free access to a covered bedded-pack and to 2.1 ha of pasture subdivided into two different sections: open pasture and pasture with natural forage cover. Results showed that the bedded-pack barn (39%) and natural forage cover (35%) were most frequently selected for calving, compared with the pasture with no forage cover (26%). Similarly, a previous study reported that, when given a choice, most dairy cows chose to calve in a straw-covered yard rather than in a pasture paddock without forage cover (Edwards 1983). These results suggest that each cow has its preferences for the environment at calving, and the characteristics of the calving area in pasture-based systems may be of importance for determining whether a cow will use it or not at calving.

Collectively, these findings provide evidence that behavioural studies can contribute to understanding the needs and motivation of the periparturient cow to find a secluded area to give birth under both indoor-housed or pasture conditions and, therefore, may also promote animal welfare. However, more research is needed to better understand calving behaviour associated with the design of calving facilities in terms of size, shape and location of the secluded area.

ENVIRONMENTAL STRESSORS

In grazing production systems in temperate regions (e.g. Chile, New Zealand, and Ireland), dairy cows are kept on pasture all year round or at least for some part of the year. In these regions, cattle are sometimes exposed to adverse weather conditions particularly during summer and winter (^{Mee and Boyle 2020}). In natural extensive environments, cattle will often have opportunities to find shade or shelter, but shade and shelter may not be avail-able under intensive grazing dairy farming conditions (^{Webster et al 2015}). Research has shown that exposure to inclement weather has negative effects on behaviour but most of the work to date has focused on lactating dairy cows, with little research during the prepartum period or when the process of parturition is imminent. Next, we will describe behavioural changes that may be associated with the exposure to hot and cold weather conditions in temperate regions and discuss the effect of providing shade or shelter during periods of adverse weather on the behaviour of dairy cows housed outdoors.

Hot conditions. Heat stress occurs when an animal's heat load is greater than its capacity to lose heat (^{Van laer et al 2014}). Heat stress in cattle not only depends on air temperature, but on another weather factors as well, such as relative humidity, solar radiation, and air movement. For instance, moderate temperatures may already induce heat stress in conjunction with high humidity and intense solar radiation (e.g., 21°C and 75% humidity) (Van laer et al 2014, ^{Webster et al 2015}).

Behavioural coping strategies used by lactating cows to avoid a rise in body temperature include modified feeding (^{Kendall et al 2006}, ^{Karimi et al 2015}) and grazing be haviour (Kendall et al 2006) (i.e. shifting feeding/grazing times to cooler periods during the day), increased water intake and time spent around a water source (^{Muller et al 1994}, Schütz et al 2010^a), reduced lying time (^{Tucker et al 2008}, Schütz et al 2010^a), decreased activity and movement (^{West 2003}, Schütz et al 2008) and reduced rumination time (^{Soriani et al 2013}). A study that evaluated the effects of heat stress (defined as temperature-humid-ity index, THI > 72, equivalent to 25°C at 50% relative humidity) on behavioural changes in prepartum dairy cows under confinement systems, showed that during the last 10 d before calving the cows adapt to heat stress through increasing meal size and reducing meal duration, as well as increasing standing times (Karimi et al 2015). Similarly, ^{Paudyal et al (2016}) reported that the average rumination time during the days prior to calving (d -14 to d -1) in healthy cows was lower in the hot season (monthly average THI >76) compared to the cool season (monthly average THI <76; 428 vs. 447 min/d), suggesting that variations in rumination behaviour can be explained by weather conditions.

A growing body of work has shown that the seeking of shade is also an important behaviour coping strategy for pasture lactating cows when exposed to hot summer conditions in temperate climates (^{Kendall et al 2006},^{Tucker et al 2008}, ^{Schütz et al 2008}, 2009, 2010^a). The work of Kendall et al (2006) showed that during mild summer conditions (average temperature 25°C) dairy cows preferred to stand, as opposed to lying, in shaded areas probably to maximize their surface area exposed to the environment to regulate body temperature. A study by Schütz et al (2008) found that dairy cows in warm condi-tions (air temperature >30°C) chose to stand beneath the shade rather than lying after experiencing 12 h of lying deprivation. Since dairy cattle are highly motivated to lie down (^{Jensen et al 2005}), the authors suggested the motivation to find shade appeared strong relative to the motivation to undertake other valued behaviours such as resting. A follow-up study showed that cows preferred shade that provided more protection from solar radiation (50 and 99% blockage versus 25%) (Schütz et al 2009). In the latter study, the average ambient air temperature was 20°C (range: 13-24°C). Similar results were reported by Tucker et al (2008) who found that cows with access to a shade that provided more protection from solar radiation, spent more time using it than cows with access to a shade that provided less protection. In both studies, cows spent more time under the shade on days with higher levels of solar radiation and air temperature. Shade use peaked when solar radiation levels were highest (Tucker et al 2008, Schütz et al 2009). In a complementary study, the effect of the amount of shade was tested using 1 of 3 treatments for 5 d: access to 2.4 m² or 9.6 m² shade/cow, or no shade during mild summer (Schütz et al 2010^a). Cows that had more access to shade spent more than twice as much time in the shade and engaged in fewer aggressive interactions (defined as the contact between any part of one cow and another that resulted in immediate hoof movement), compared to the moment when less shade was available.

^{Edwards (2018}) evaluated the effect of heat stress on calving location preference in dairy cows housed in pasture, where cows could choose to calve either in a covered bed ded-pack barn or in a pasture paddock with natural forage. Results showed that cows were more likely to calve in the pasture when they were not experiencing heat-stressed conditions (defined as THI > 68, equivalent to 21°C at 75% relative humidity). On the contrary, cows preferred to calve inside the barn during heat-stress conditions due to better protection from solar radiation.

Current research concludes that heat-stressed dairy cows, under pasture-based systems, exhibit behavioural changes that indicate that animal welfare is compromised. Cows showed a high motivation to seek and use shade to reduce the impact of heat, but if the shade is insufficient in terms of space, they will increase aggressive behaviour. Furthermore, shade availability seems to be particularly important at calving, playing a role in determining where cattle chose to calve. However, little is known about the effect of heat stress on calving progress. A better under-standing of the effects of shade availability, and other mitigation alternatives, on calving behaviour of dairy cows on pasture is needed.

Cold and wet conditions. In general, cattle are much better adapted to cope with low ambient temperatures than hot ambient environments. However, in temperate regions, exposure to winter weather that combines cold, wind and rain, causes cattle to increase heat losses to the surrounding environment, increasing energy requirements to maintain body temperature (^{Van laer et al 2014}, ^{Webster et al 2015}).

Dairy cows usually change their behaviour to adapt to increased energy demand. For example, studies using non-lactating Holstein cows have reported that animals decrease food intake and time spent lying, stand with lowered heads, and adopt postures that might reduce heat loss when exposed to a combination of simulated and natural rain and wind in winter (^{Tucker et al 2007}, mean air temperature 5°C, wind chill factor -10°C; ^{Webster et al 2008}, mean air temperature 3°C, wind chill factor -0.3°C; Schütz et al 2010^b, mean air temperature 10°C, wind chill factor -11°C). Since a high body condition helps insulates cows and to maintain more stable body temperatures, thinner cows changed their behaviour to reduce heat loss to a greater extent than well-conditioned cows under cold and wet conditions (Tucker et al 2007).

Another natural behavioural response of dairy cows to wet and windy conditions is seeking shelter. For example, Schütz et al (2010^b), exposed non-lactating pregnant dairy cattle to one of four treatments (control, wind, rain, wind and rain) created with fans and sprinklers for 22 h, and assessed their motivation to use the shelter by creating a tradeoff between time spent feeding while exposed to the weather treatments and time spent in the shelter. Results showed that cows used the shelter approximately 50% of the time but with no difference between weather treatments. Those authors have argued that cows were motivated to use the shelters for other reasons than protection from the weather, possibly as a response to social isolation. This behaviour might be particularly important in the hours prior to calving. To our knowledge, this hypothesis has not been tested in grazing dairy cows under inclement weather.

Research has shown that temporarily managing dairy cows on wet and muddy surfaces has negative effects on lying behaviour (^{Chen et al 2017}, ^{Schütz et al 2018}). Chen et al (2017), in a simulated stand-off situation exposed pregnant, nonlactating Holstein cows to three levels of soil moisture: 90 (dry), 74 (muddy), or 67% (very muddy) dry matter for 5 d each. Results showed that cows spent less time lying down in muddier conditions (dry, muddy, and very muddy treatments: 13, 12, and 9 h/24 h, respectively), and this response was more marked for heifers than for cows. When cattle chose to lie down on wetter soil, they limited the surface area exposed to their surroundings by tucking their legs beneath their bodies (Chen et al 2017). More recently, ^{Schütz et al (2018}) assessed the use and preference of pregnant, nonlactating Holstein cows for different wood chip surface types: clean and dry, dirty (contaminated with manure) and clean and wet. During 5 d of observation, cows were kept 18 h on woodchip surface and 6 h on pasture to allow for daily feed intake. Cows on the wet surface spent the least amount of time lying when restricted to one surface for 18 h (wet: 21%, dirty: 57%, clean: 64%), and spent more time lying when on pasture for 6 h (wet: 13%, dirty: 4%, clean: 3%), a time when ideally they should be grazing. Also, when given a choice, they clearly showed that they will avoid wet and dirty surfaces. This later research also suggests that wet surfaces not only influence the duration of rest but also the quality of rest. Cows on wet woodchips spent less time lying in a lateral position and with their heads supported, indicating reduced cow comfort and quality of rest on this surface (^{Schütz et al 2018}).

Spring calving season predominates in grazing sys tems and the cows are sometimes moved from pasture to a separate area in periods of wet weather (Shütz et al 2019). For instance, some farms in southern Chile use "sacrifice" paddocks to keep prepartum cows (i.e. 3 wk prior to calving) during late winter and early spring. A "sacrifice" paddock can take the pressure off the rest of the farm by allowing grass cover to build up while vulner able soils are wet. In these winter paddocks, depending on stocking density and usage, surfaces can become wet and muddy. In a recent study of our own, we followed Holstein dairy cows during the prepartum period (e.g., from 3 wk to 1 wk prior to calving) in the winter Chilean season (^{Cartes et al 2019}, mean 24/h air temperature 7°C, range = -2 to 16 °C). Cows were housed in pairs in 6 paddocks with (n=12) and without (n=12) access to a shelter with clean and dry wood chip bedding. Cows spent 60% of their daily time in the shelters and 75% of this time they were lying down. Furthermore, cows that had access to shelter during the prepartum period spent more time lying down during the wk 3 (706 min/d vs. 559 min/d) and wk 2 (742 min/d vs. 566 min/d) before calving compared to cows without shelter access. Shelter use in the wk prior to calving was similar between treatments. This study suggests that the provision of a protected area to rest during the weeks before calving might be beneficial for cow welfare.

In summary, cows can generally tolerate low air tem-peratures but they seek shelter in wet, windy and muddy conditions, suggesting that protection from inclement winter weather is important even in temperate regions. Also, newborn calves exposed to cold temperatures (<10°C), wet, and windy conditions show longer times to stand up than calves born during warm and dry weather (^{Diesch et al 2004}). Therefore, the use of a shelter with dry and clean bedding can mitigate the negative effects of inclement weather and wet and muddy underfoot condi-tions in pasture-based systems. However, there is a lack of information about the effect of continued exposure to wet and muddy conditions throughout the prepartum period. During that time, lying down and getting up safely and comfortably may be of increasing importance due to the growing fetus, as well as on the day of calving when the cow becomes more restless (^{Lidfors et al 1994}, ^{Campler et al 2014}). The potential effect of shelter availability in winter paddocks during the prepartum period on the ease of calving has not yet been investigated.