Voluntary heat stress abatement system for dairy cows: Does it mitigate the effects of heat stress on physiology and behavior?

Many cooling strategies are used to keep cows in thermal homeostasis; however, most of them are applied to the group level, commonly at the feed bunk or milking parlor. The variance of heat stress effects on animals are well known, but with more individualized management in dairy farms, group cooling opportunities are becoming restricted. It is known that dairy cattle are variable in their responses to an increase in heat load. Thus, the first objective of this study was to investigate the effect of 2 mandatory soakings at the exit of the milking parlor and free access to a voluntary soaking system compared with cows with access to a voluntary soaking system only, with no mandatory soakings. The second objective of this study was to assess the heat abatement capability of voluntary soaking of cows by assessing cow physiology, behavior, and milk production. Last, this study aimed to determine the individual use of the voluntary heat abatement system and its relationship with temperature-humidity index (THI). Fifteen mid-lactation Holstein cows were enrolled in this study and had free access to a motion-activated soaker (Cool Sense, Edstrom) located adjacent to the research pen for an 8-wk data collection period. Cows were paired according to parity, milk production, and body weight, and assigned a treatment with or without mandatory soakings twice per day. In the mandatory soaking treatment (MS), cows were soaked using a motion-activated soaker at the exit of the milking par-lor and had free access to the voluntary soaker in the pen. Cows in the treatment without mandatory soak-ings (NMS) were not soaked at the exit of the milking parlor and had free access to the voluntary soaker in the pen. The effects of soaker treatment were analyzed using mixed linear models. The model included treat-ment, soaker uses per day, pair, mean daily THI, days in milk, daily milk yield, and interaction of treatment with mean daily THI. Study day was specified as a repeated measure, and cow as the subject, using an autoregressive structure. Also, we assessed the relationship of mean soaker use and THI against all variables. There was great individual variation in voluntary soaker use, ranging from 0 to 227 soakings/d (mean ± standard deviation, 13 ± 30 voluntary soakings/d). Treatment did not affect voluntary soaker use (MS, 12.4 ± 1.4 soakings/d; NMS, 14.8 ± 1.4 soakings/d), respiration rate (MS, 57.3 ± 0.4 breaths/min; NMS, 56.4 ± 0.4 breaths/min), or milk yield (MS, 36.5 ± 0.6 kg/d; NMS, 36.2 ± 0.6 kg/d). However, MS cows spent more time ruminating (MS, 558.6 ± 5.2 min/d; NMS, 543.4 ± 5.4 min/d). Temperature-humidity index had a positive relationship with voluntary soaker use and mean respiration rate. In conclusion, voluntary soaker use related positively to the THI, but no major productive, physiological, or behavioral differences were observed between soaking treatments. Furthermore, we found that voluntary soaker use is highly variable among cows and it was related positively to milk yield, where higher producing cows used the soaker more frequently.


INTRODUCTION
Cows in environmental conditions that exceed their thermoneutral zone (5 to 25°C) (McDowell, 1972) have increased metabolic requirements, increased respiration rate (RR), sweat, and pant to regulate body temperature (Collier et al., 1982). Temperature-humidity index (THI) is a common method of assessing heat stress affecting dairy cows, because it has been shown to be highly associated with production losses, health, and behavioral changes [see review by ]. Production losses have been found to occur at around THI ≥ 72 (Armstrong, 1994;Ravagnolo et al., 2000); however, behavioral and motivational changes were found to occur at lower THI, such as ≥68 (De Rensis et al., 2015).
Heat stress has been associated with reductions in DMI (Spiers et al., 2004;Bernabucci et al., 2010;Soriani et al., 2013) and feeding bouts (Bernabucci et al., 2010), consequently affecting rumination (Kadzere et al., 2002;Bernabucci et al., 2010;Soriani et al., 2013). This can have consequences for rumen pH because more feeding bouts and ruminating produce saliva, which acts as a pH buffer and is important for healthy rumen pH (Bernabucci et al., 2010). Heat stress also affects cow behavior, including preference for standing rather than lying (Tucker et al., 2008;Allen et al., 2015), even after lying deprivation , which is a welfare issue. The increase of maintenance metabolism (Collier et al., 1982) and reduction in DMI (Spiers et al., 2004;Bernabucci et al., 2010;Soriani et al., 2013) during heat stress result in cows often being in a state of negative energy balance (Drackley, 1999). Negative energy balance subsequently diminishes milk production (Spiers et al., 2004;Soriani et al., 2013)-the primary source of income to dairy producers-and is therefore a serious economic issue for the dairy industry.
It is important to investigate options for cooling heat-stressed dairy cows to improve milk production to support dairy farm economics and to improve cow welfare during periods of heat stress. Options for heat abatement have previously included the use of shade, fans, and soaking the cows, which typically reduce the negative effects of heat stress on physiology and behavior Chen et al., 2016;Tresoldi et al., 2018). For decades, one of the most efficient methods of cooling dairy cattle was based on repeated soaking to attain maximal water trapping in the coat, followed by its rapid evaporation (Flamenbaum et al., 1986;Chen et al., 2016). Water as a method of cooling is often delivered via automated cycling sprinklers installed above feed alleys; however, cows may perceive them as a deterrent or obstacle, because wetting their head is uncomfortable (Chen et al., 2016). Conversely, cows have not been observed to display head-wetting avoidance behavior in a voluntary use soaking system (Legrand et al., 2011).
Many cooling strategies are used to keep cows in thermal homeostasis; however, most of them are applied to the group level, at the feed bunk or milking parlor . Providing cows with longer soaking sessions in a holding pen (20 to 140 min) has been shown to be effective in aiding cows in reducing and maintaining lower body temperatures (Araki et al., 1985;Flamenbaum et al., 1986;Valtorta and Gallardo, 2004;Kendall et al., 2007).
Although it may not be the most common heat abatement practice, producers will opt to provide cows with a short soaking at the exit of the milking parlor (Van Os, 2019). Limited research is available on the benefits of soaking cows at the exit of the milking parlor, yet Collier et al. (2006) stated that providing cows soakings at the exit of the milking parlor can provide animals with an extra 15 to 25 min of additional cooling, depending on the weather conditions. Hence, there is a need to investigate further the validity and effects of the adoption of short soaking cycles at the exit of the milking parlor, and whether this practice is beneficial when paired with other heat abatement practices.
Dairy cattle differ in how they respond to an increase in heat load. Voluntary soaking stations for heat stress abatement provide freedom of choice and cow self-management, which focuses on individual cow needs instead of the group. Other voluntary use equipment (e.g., automatic milking systems) is perceived as advantageous because it removes the necessity of daily laborious tasks (e.g., daily milking) and adds freedom of choice while avoiding herding and interaction with humans (Webster, 2001;Holloway et al., 2014). Because individual cows have different tolerances to heat stress, it seems logical to offer heat abatement at an individually self-managed level, such as by a voluntary use soaker station. Cow heat tolerance differs among individuals because of genetics (Aguilar et al., 2009;Liang et al., 2013;Alfonzo et al., 2016), parity (Aguilar et al., 2009;Stone et al., 2017), milk production (Liang et al., 2013;Stone et al., 2017), body size, hair structure (Alfonzo et al., 2016), and a myriad of other factors.
The variance of heat stress effects on animals are well known; however, with more individualized management in dairy farms, group cooling opportunities are becoming restricted. It is known that dairy cattle are variable in their responses to an increase in heat load. Thus, there is an opportunity to provide dairy cows with a voluntary heat stress abatement method that allows animals to self-manage their heat stress. Because animals may have different requirements and therefore motivations to use heat stress abatement tools, it would be beneficial to investigate whether there is a need to provide extra heat abatement strategies in addition to the use of voluntary heat abatement systems. Thus, the first objective of our study was to investigate the effect of 2 mandatory soakings at the exit of the milking parlor and free access to a voluntary soaker compared with cows with access to a voluntary soaker only, with no mandatory soakings. The second objective of this study was to assess the heat abatement capability of voluntary soaking of cows by assessing cow physiology (RR and body temperature), behavior (rumination, feeding, lying time, and steps per day), and milk production. Last, this study aimed to determine the individual use of the voluntary soaking system and its relationship with THI.

Animal Housing and Diet
The study was conducted from July 10 to October 3, 2018, at the University of Kentucky Coldstream Research Dairy Farm (Lexington, KY;latitude, 38.1103759;longitude, −84.5164302), and was approved by the Institutional Animal Care and Use Committee of the University of Kentucky (Protocol No. 2018-2914. Cows were housed in a compost-bedded pack barn that was tilled twice per day (at approximately 0520 and 1415 h). The experimental pen was equipped with a total of 8 automatic intake recording feeders (Insentec, Hokofarm Group), one 4.9-m low-speed highvolume fan (Powerfoil X3.0, Big Ass Fans) over the bedded pack area, and three 91-cm fans (Yellow Jacket, Big Ass Fans) over the feed alley. A motion-activated voluntary soaker (Cool Sense, Edstrom) was installed adjacent to the experimental pen as an extension of the feed alley walkway. The 20-m 2 voluntary soaking area had a grooved concrete floor and was covered with a shade cloth that blocked 75% of the solar radiation. The experimental pen layout, approximate measurements of the compost-bedded pack barn, and fan placement are illustrated in Figure 1. Cows were housed in the experimental pen starting at enrollment and then returned to the main herd at the end of the data collection period. Cows were fed a TMR formulated according to NRC guidelines (NRC, 2001) to meet or exceed the requirements of lactating dairy cows producing at least 39 kg of milk daily. Composition of the TMR as fed was 40.7% corn silage, 27.8% lactating cow grain mix, 23.6% alfalfa silage, 5.1% cotton seed, 1.8% alfalfa hay, and 1.0% mineral mix. Cows were fed ad libitum twice per day at approximately 0800 and 1400 h. Orts were removed daily before the morning feeding. Animals had ad libitum access to fresh water provided from a selffilling water trough located in the feeding alley. Milking occurred twice daily at 0730 and 1800 h.
Fifteen confirmed-pregnant, mid-lactation Holstein cows were chosen randomly from the herd for enrollment in this study and were moved to the experimental pen for a 2-d adaptation period. After the adaptation, a 4-wk training period with the voluntary soaker started. On the first day of the training period, cows were 233 ± 38 (mean ± SD) DIM, had a parity of 2 ± 1, weighed 673 ± 69 kg, and had a milk yield of 38.0 ± 5.4 kg/d.

Experimental Design
Cows were split into 7 pairs, balanced for parity, milk production, and body weight; 1 cow was unpaired. Cows were assigned within pairs to 1 of the 2 treat-ments by random block design for eight 1-wk treatments, totaling 8 wk of data collection per pair. The possible treatments were (1) the mandatory soaker treatment, which comprised 2 mandatory soakings/d with free access to the voluntary soaker; and (2) the no mandatory soaker treatment, which comprised no daily mandatory soakings with free access to the voluntary soaker. Mandatory soakings were given by sorting cows via a sort gate (AfiSort, Afimilk) after milking, and guiding them under a motion-activated cattle soaker (Cool Sense, Edstrom) located adjacent to the milking parlor. Cows had free access to the voluntary soaker adjacent to the experimental pen at any time except during milking or when the feeding alley was scraped (mean access ± SD, 18.3 ± 1.6 h/d).
Cows were trained individually to use the voluntary soaker during the 4-wk training period. Training for the voluntary soaker involved a single experimenter encouraging cows to pass under the motion-activated soaker 3 times/d (at 1000, 1230, and 1500 h). Training was considered complete after a cow used the soaker voluntarily at 2 separate instances within 3 d (monitored via video footage). All cows enrolled were trained successfully to use the soaker. In addition to voluntary soaker training, cows were acclimated to the mandatory soaker during the 4-wk training period by using a sort gate to guide them under a motion-activated soaker at the exit alley from the milking parlor. After the 4-wk training period, the 8 wk of treatments were started.
The mandatory and voluntary soakers were identical in model and setting. The soaking system is described in Mazon et al. (2021). Briefly, both soakers were set to a 5-s cycle; the water flow rate was approximately 4.1 L/5-s soaker cycle (or 49.2 L/min) via 2 shower heads (accumulatively). The soaker system was activated after both motion sensors were triggered. The soaker at the exit alley was installed per manufacturer instructions. Two motion sensors were installed above the exit alley approximately 1.8 m apart to allow 1 cow to activate the soaker while walking through the alley (soaking only 1 cow at a time). However, the voluntary soaker was modified; both motion sensors were next to each other, immediately adjacent to the shower heads. This modification was made so a cow could activate both motion sensors (and therefore initiate a soaking cycle) while standing under the water flow.

Data Collection
Physiological measurements of the cows measured during the study were reticulorumen temperature (RT), RR, and panting score. Reticulorumen temperature measurements were collected with an automated data logging bolus (Herdstrong TruCore, DVM Sys-tems), which was validated previously for recording of RT (Bewley et al., 2008). The bolus manufacturer supplied the researchers with RT data after removing temperature changes caused by water intake (Cantor et al., 2018) using a proprietary algorithm. Boluses were assigned to cows 6 ± 1 d before the start of the training period.
Two observers recorded RR and panting scores (observer 1, 88%; observer 2, 12% of the observations). High interobserver agreement was achieved for RR and panting score as defined by Hinkle (1988) (r = 0.98 and 0.87, P < 0.001; R 2 = 0.96 and 0.75, P < 0.001; respectively), and no bias was observed using Bland-Altman plots (Bland and Altman, 1995a,b). Respiration rate was recorded thrice per day at 0645, 1230, and 1720 h. Respiration rate was recorded by counting flank movements for 1 min according to the methodology of Rhoads et al. (2009) andMin et al. (2015). The observer recording RR was stationed approximately 5 m from the focal cow and ensured flank movements were visible for the duration of the observation. At the same time, panting score was recorded using a predefined ethogram (Table 1).
Rumination (min/d) and feeding (min/d) were recorded for each cow by a behavior-monitoring collar (MooMonitor+, DairyMaster). The collar was validated previously for all the used behaviors (Grinter et al., 2019). Collars were assigned to cows per farm protocol (306 ± 197 d before the beginning of the training period). A behavior-monitoring leg tag (AfiTagII, Afimilk) was used to monitor lying time (min/d), lying bouts (bouts/d), and steps (steps/d). The leg tag was  validated previously for lying time, lying bouts, and steps (Higginson et al., 2010). Leg tags were assigned to cows per farm protocol (>30 d before the start of the training period).
Daily milk yield was recorded during each milking and was summarized by day using an automatic meter (AfiMilk, AfiMilk). Milk fat and protein were measured at each milking using an in-line milk analyzer (AfiLab, AfiMilk) that was validated previously (Kaniyamattam and De Vries, 2014). A 3-d rolling average was calculated each day for milk yield, milk fat, and milk protein by taking the mean value of a day with the previous 2 d.
Use of the voluntary soaker was monitored 24 h/d for the duration of the study via video footage (Hikvision model DS-2CD2342Wd-I, Hangzhou Hikvision Digital Technology Co. Ltd.). Video was recorded and later played back to record soaker use by each animal and the part of the body targeted for soaking. Four observers recorded voluntary soaker use. When soaker use was detected, the observers recorded the cow identification number, time of cycle activation (hh: mm: ss), and area of cow under the soaker according to an ethogram (Table 2). In addition, at the end of each soaking cycle, observers recorded whether the cow continued using the soaker, was displaced (by another cow or farm personnel), stopped using the soaker voluntarily (no other cows present), or stopped using the soaker for an unknown reason (not obvious whether displacement occurred or the cow chose to exit). In any event of displacement, the cow displacing and the cow that was displaced were both recorded. Voluntary soaker use was summed by experimental day (0000 to 2359 h) and averaged by hour for each and all animals during the experimental period. To determine interobserver reliability, fifteen 1-h blocks of video were watched simultaneously by all observers to determine the use of the soaker by each cow during the period. Very high correlations (r > 0.95) and very high linear regressions (R 2 > 0.95) were used to deem observer reliability for discerning soaker use per day.
Ambient temperature and relative humidity (RH) were recorded by a portable weather station (HOBO External Temperature/Relative Humidity Data Logger model U23-002, Onset Computer Corp.) located in the study pen. Wind speed was recorded by an anemometer (Digi-Sense model 20250-22, Cole-Parmer) located in the study pen. The calculations for THI include air temperature (T) and RH, were performed using the following equation: THI = (1.8 × T + 32) -(0.55 -0.0055 × RH) × (1.8 × T -26), where T is the ambient temperature in degrees Celsius and RH is the percentage of relative humidity (National Oceanic and Atmospheric Administration, 1976).
Hourly THI weather values were averaged to calculate mean daily THI, and the maximum daily value was taken for the daily maximum THI. In addition, mean daily THI was averaged by hour for the duration of the   Table 3.

Statistical Analysis
All statistical analyses were performed using SAS (version 9.4, SAS Institute Inc.). Before analysis, all data were checked for normality using the UNIVARI-ATE procedure probability distribution plots. For daily lying time, values equal to 0 min/d or greater than 1,440 min/d were removed because they were considered recording errors by the device. The first (<366 min) and 99th (>1,014 min) percentiles of data were also removed (Stone et al., 2017). Seven percent of lying bouts data was removed because the device reported 0 lying bout/d, which was considered a recording error.
Originally there were 28,498 recordings for RT. Any hour when the mean RT was <35.6°C or >42.2°C was removed for biological relevance, as outlined by Bewley et al. (2008). This step removed 8,654 data points. After this, an additional 190 were removed because they were deemed as outliers for exceeding 3 SD of the mean, according to the same data preparation steps as Bewley et al. (2008), resulting in 19,654 points.   Data from cows detected having mastitis (2 cows) or in estrus (1 cow) by standard farm procedures during the study were removed the day before detection, day of detection, and day after detection.
The effect of soaker treatment was determined by ANOVA using mixed linear models (MIXED procedure). The model included treatment (mandatory soakings or no mandatory soakings), soaker use per day, pair, mean daily THI, DIM, daily milk yield (kg/d), and interaction of treatment with mean daily THI. Study day was specified as a repeated measure, and cow as the subject, using an autoregressive structure. Effects with a P-value >0.30 were removed from the model using a stepwise backward elimination process starting with the least contributing effect. Treatment, soaker use per day, pair, and mean daily THI remained in the model regardless of significance. Significance was declared at P ≤ 0.05, and trends were defined as P ≤ 0.10.
When the fixed effects of voluntary soaker use per day or THI were significant for any outcome variable (RR, panting score, RT, rumination, feeding, lying, steps, milk yield, milk protein percentage, and milk fat percentage), an additional extraction of estimates was performed in the mixed models (ESTIMATE state-ment) for that variable. The relationship between the significant variables and voluntary soaker use per day and THI were classified as positive or negative.

Daily Soaker Use
Treatment did not influence soaker use. The frequency of voluntary soaker use during the no mandatory soakings treatment was approximately 15 voluntary soakings/d, similar to the 2 mandatory soakings treatment of about 12 voluntary soakings/d (F 1,14 = 1.74, P = 0.21; Table 4).
There was large individual variation in voluntary soaker use, ranging from 0 to 227 soakings/d (mean ± SD, 13 ± 30 voluntary soakings/d) (Figure 2). Four cows had a maximum use of less than 10 soakings/d, whereas 4 other cows had a maximum of more than 100 soakings/d (Figure 2). Cows preferred to use the soaker predominantly on their back, followed by their side, rump, and neck and head, and licking was the least common soaker use area (Figure 3). At the end of each 5-s soaker cycle, cows typically continued to use the soaker (Figure 4). Cows were least likely to leave the soaker because of displacement (by cow or farm personnel), followed by leaving by choice (Figure 4). The main reason cows left the soaker was "unknown" (when the observer was unable to differentiate between leaving by choice versus a displacement; Figure 4). Cows with a greater frequency of soaker use were more likely to displace another cow from the soaker (F 1,14 = 9.28, P < 0.01; Figure 5) or be displaced from the soaker (F 1,14 = 30.61, P < 0.001; Figure 5). Soaker use peaked between 1900 and 2000 h, and the nadir (excluding hours manipulated by limited access) was between 1100 and 1200 h (Figure 6).

Temperature-Humidity Index
Temperature-humidity index affected physiological and behavioral variables, and almost all milk-related variables in the mixed model (Table 5). The relationship and its direction between the response variables with daily soakings and THI are given in Table 5. In summary, rumination, feeding and lying times, and milk fat percentage had a negative relationship with Grinter et al.: VOLUNTARY HEAT STRESS ABATEMENT Figure 4. Behaviors recorded at the end of each 5-s soaking cycle. Each point represents the percentage of time each cow (n = 15) performed the corresponding behavior. If cows "continued use," the soaker was activated again for another cycle. Leaving the soaker area because of being displaced is indicated as displaced by farm personnel [Displaced (p)] or displaced by another cow [Displaced (c)]. A cow that left by "own choice" left with no other cows in the vicinity of the soaker to influence choice for leaving. An "unknown" reason was recorded when the observer was unable to attribute displacement or cow's own choice. In each boxplot, the center line represents the median for each cow, the whiskers represent the interquartile range of each cow, and the dots represent individual observations for each cow.
THI. Voluntary soaker use (Figure 7) and all other significant variables had a positive relationship with THI.

DISCUSSION
Voluntary soaking opportunities were provided to dairy cows during a time of elevated THI. This study is the first to compare the use of a voluntary soaker method with or without a mandatory cooling opportunity for dairy cows, in combination with provision of fans and shade. We found limited differences in heat stress alleviation between the 2 treatments of 2 mandatory soakings at the exit of the milking parlor with voluntary soaker use and no mandatory soakings with voluntary soaker use. We did not find an additional heat abatement between treatments nor within any physiological variables measured, including RR associated with heat stress (Rhoads et al., 2009;Schütz et al., 2010;Min et al., 2015).
The only behavioral variable affected by treatment was rumination time. Shorter daily rumination time has been suggested as being an indication of heat stress (Kadzere et al., 2002;Bernabucci et al., 2010;Soriani et al., 2013;Müschner-Siemens et al., 2020) and was observed in the nonmandatory soaking treatment in comparison to longer daily rumination time in the compulsory soaking treatment. In addition to the difference in rumination time between treatments, there was a positive relationship between soaker use and rumination time. Because rumination had a negative relationship with THI (reducing daily rumination time with increasing THI), the positive relationship between the mandatory soaking treatment and rumination time may suggest that soaker use mitigated, to some extent, the negative effects of heat stress on rumination. However, because there were no differences between treatments for RR, panting score, RT, daily feeding time, daily lying time or bouts, steps per day, or milk production and components, the results suggest cows assigned either treatment had a comparable level of heat alleviation. Cows in our study were highly productive, at more than 38 kg/d of yield, and, at this production level and with the high THI during the study, we found no differences between soaking treatments. As a limitation of this study, we did not have a true negative control with no heat abatement to quantify the heat stress effects in milk yield. Future research could investigate the magnitude of heat alleviation of heat Grinter et al.: VOLUNTARY HEAT STRESS ABATEMENT Figure 5. Mean displacement actions of cows (n = 15) for the voluntary soaker, by cow (left y-axis). Depicted is the number of times a cow displaced another cow (diagonal striped bars) or a cow was displaced by another cow (horizontal striped bars). In addition, mean soaker use is displayed by the scatter graph (right y-axis). Displacements and soaker use were recorded by 24-h surveillance of the soaker. soakers for cows in dairies with different heat abatement strategies (e.g., without fans, with sprinklers over the feed bunk, without shade) or include a treatment of no voluntary soaker.
We did not observe any differences in the number of voluntary soakings between treatments. We believe that, regardless of treatment, cows were relatively cool. The minimum, mean, and maximum RT in our study were not at a level of concern for cattle in heat stress, as seen by Lees et al. (2018) in beef cattle and Ammer et al. (2016) in lactating dairy cows. The research on the use of mandatory soakings at the exit of the parlor is still limited, yet some producers will opt to provide cows a short soaking at the exit of the milking parlor (Van Os, 2019). Future research should investigate the possible benefits of soaking cows at the exit of the milking parlor as well as investigate the effects of heat stress on cows with access to a voluntary heat abatement system only.
Although there was no difference in voluntary soaker use between treatments, it related positively to THI, as seen in previous studies that offered a voluntary soaker (Legrand et al., 2011), voluntary sprinklers (Parola et al., 2012), or voluntary use of sprinklers over the feed bunk (Parola et al., 2012;Chen et al., 2013Chen et al., , 2016. This suggests that cows may be more motivated to use a voluntary soaker during periods of elevated heat stress potential. However, Parola et al. (2012) and Chen et al. (2013Chen et al. ( , 2016 used constantly running sprinklers (as opposed to the voluntary soaker in our study). Constantly running water without an escape route may be aversive to cows, because the animals would need to get their head wet, which has been associated with discomfort (Chen et al., 2016). For instance, cows show headwetting avoidance behaviors, such as a lowered head or keeping their head outside the sprinkler (Kendall et al., 2007;Schütz et al., 2011;Chen et al., 2016). Cows have also been observed standing with their head through headlocks when sprinklers over the feed bunk were activated, despite not feeding (Chen et al., 2013), and moving out of the sprinkler radius when sprinklers were activated (Marcillac-Embertson et al., 2009). In support of this work, we also found that cows wet their head and neck much less than other body areas when using the voluntary soaker. Conversely, Legrand et al. (2011) found that cows had their head near the voluntary soaker heads for more than half the time when in the soaker. This difference in behavior may be a result of flow rate or water droplet size between the work of Legrand et al. (2011) and our study. Legrand et al. Grinter et al.: VOLUNTARY HEAT STRESS ABATEMENT Figure 6. Mean hourly soaker use by cows (n = 15) as recorded from 24-h video recording (solid bars; left axis) with temperature-humidity index (THI; line graph; right axis). Mean daily THI is depicted by the dotted line with circles, and mean daily THI + 5-h delay is depicted by the dotted line with squares. Mean daily THI + 5-h delay was found to be the best predictor of mean hourly soaker use by the regression analysis (P < 0.01). Treatments were 2 mandatory soakings per day exiting the milking parlor with access to a voluntary soaker, and no mandatory soakings per day with access to a voluntary soaker located immediately adjacent to the feed alley of the pen. No effects of treatment were found; therefore, all data are presented. Cows were unable to access the soaker for approximately 5.7 ± 1.6 h/d. Soaker use was affected by being locked in the pen for other cows to milk (approximately 0600 to 0730 h and 1645 to 1750 h), leaving the pen for milking (approximately 0720 to 0740 h and 1750 to 1820 h), and leaving the pen while the alley was being scraped (approximately 0430 to 0510 h and 1500 to 1550 h).
(2011) speculated this may be a result of cows having control over the water source (as opposed to sprinklers in Schütz et al. (2011) and Chen et al. (2013Chen et al. ( , 2016, although this was not evident in our study. Future research should explore the motivation of cows to use a voluntary use soaker, or should experiment with different soaker structures (e.g., water flow, droplet size, design to avoid cow ears and head).
High variability in voluntary soaker use among cows was observed in our study (ranging from 0 to 227 daily soaking cycles). Similarly, Schütz et al. (2011) and Legrand et al. (2011) found some cows did not seek further heat alleviation from a voluntary choice soaker. Legrand et al. (2011) speculated that individual variation in soaker use may be related to a lack of learning of the cooling properties of water. However, in each of these studies, including our own, cows were previously accustomed to water. In addition, steers naive to sprinklers have been shown to use sprinklers for their cooling properties (Parola et al., 2012). The high variation in soaker use could also be because cows experience heat stress differently as a result of differences in genetics (Aguilar et al., 2009;Liang et al., 2013;Alfonzo et al., 2016), milk variables (Liang et al., 2013;Macciotta et al., 2017;Stone et al., 2017), body size, or hair structure (Alfonzo et al., 2016). In our study, we attempted to account for such individual variation in heat stress tolerance (parity, milk yield), but we were unable to consider differences in genetics that may have contributed to voluntary soak use variability. Differences in genetics and therefore heat tolerance could be a reason for the high variation in daily soaker use. Voluntary cooling options provide cows with the choice of when to use the soaker and, importantly, whether they want to use the soaker at all. Yet, it is important to notice that assessing individual soaker use is labor intensive. Both our study and that of Legrand et al. (2011) relied on manual review of video recordings to assess individual use of the voluntary soaking systems, which would not be feasible outside of a research setting. However, there are voluntary radiofrequency-controlled soaking systems that perform automatic identification of animals that activate the soaker (Mazon et al., 2021). This automatic identification would allow producers to identify easily those animals that are more prone to heat stress, and to provide early interventions that could benefit Grinter et al.: VOLUNTARY HEAT STRESS ABATEMENT If the result was significant from this model, an estimate was extracted from the mixed model for the response variables with mean daily temperature-humidity index and soaker use per day; associations were classified as positive or negative. Treatments were (1) 2 mandatory soakings per day exiting the milking parlor with access to a voluntary soaker, and (2) no mandatory soakings per day with access to a voluntary soaker located immediately adjacent to the feed alley of the pen. No effects of treatment found; therefore, all data are presented.
2 Panting score was recorded 3 times/d utilizing a scale that ranged from 0 (no panting, normal) to 4.5 (severe panting, with animal displaying open mouth with tongue fully extended for prolonged periods, head held down, and intense flank movements).
the entire herd. Therefore, a voluntary soaker may offer a good opportunity for heat-intolerant cows to alleviate heat stress. We suggest further research should investigate different voluntary cow soaker designs, individual motivation of cows for voluntary soakings, and the combination with other cooling strategies. Future research should also consider comparing cows with genetic testing for heat tolerance genes to investigate whether some cows that are less heat tolerant use the voluntary soaker or whether soaker use is related to individual preference.

CONCLUSIONS
This study found limited physiological and behavioral differences between the use of a voluntary soaker with or without 2 mandatory soakings. Rumination time only was longer in the treatment with 2 mandatory soakings. We conclude the results indicate similar heat alleviation between the treatments. However, voluntary soaker use related positively to THI. We encourage future research to investigate potential advantages of a voluntary soaker by comparing it to cows with no voluntary soaker, and different levels of heat abatement (such as with and without shade and fans) to explore heat abatement at the individual level further. In addition, future studies are necessary to understand cow preference to soaker or sprinkler designs to investigate why some cows prefer the soaker more than others.

ACKNOWLEDGMENTS
We gratefully acknowledge and thank the staff and students of the Coldstream Dairy Research Farm (Lexington, KY) who helped in this experiment, especially Adrien Lebreton, Charlotte Pertuisel, Brittany Core, Amelia Fendley, Joey Clark, and Matt Collins. We also thank Olga Vsevolozhskaya, Michelle Arnold, and Eric Vanzant from the University of Kentucky (Lexington, KY) for their contributions to this project. This project was funded by DairyMaster Co. Values for each cow's (n = 15) daily soaker use are represented by different symbols. Daily soaker use was recorded by 24-h surveillance of the soaker. Treatments were 2 mandatory soakings per day exiting the milking parlor with access to a voluntary soaker, and no mandatory soakings per day with access to a voluntary soaker located immediately adjacent to the feed alley of the pen.