Pasture-based dairy systems are receiving special attention around the world because they may offer benefits to several environmental aspects (Aguirre-Villegas et al., 2017), may present profitability advantages (O'Brien et al., 2015), and are well-regarded by consumers in terms of animal welfare (Cardoso et al., 2019). Uruguayan pasture-based dairy farming systems have the potential for production growth and must achieve it to maintain international competitiveness (Fariña and Chilibroste, 2019). The optimized utilization of grazed pastures to reduce the costs associated with feeding dairy cows is a common goal for these dairy production systems (Finneran et al., 2010).

Grazing is a complex process that involves a set of decisions at different scales with bite mass as a basic unit. Hence, daily pasture dry matter intake by grazing animals is largely dependent on the mass of each bite and the frequency of them, which is affected by the sward height (Gibb et al., 1997) and herbage mass (Laca et al., 1992) available to animals. The daily grazing pattern adopted by the animals will depend mainly on their intrinsic behavior, photoperiod (Linnane et al., 2001), sward characteristics (Chapman et al., 2007), internal state (e.g., ruminal fill, energy demand, hormone release, physiological state, and adiposity, etc.; Chilibroste et al., 2005), spatial memory (ability of herbivores to use their previous experience to choose where to graze; Bailey et al., 1996), and anthropogenic management (Chilibroste et al., 2015).

The actions of a grazing animal can be divided at different spatio-temporal scales (Carvalho and Moraes, 2005). Considering the level of anthropogenic management, such as the use of fences to manage the forage allowance, in dairy farming could be identified at small spatial scales like patch, feeding stations, and bite scale. When the grazing animal has its front forefeet immobile and take bites in front of it, this area is defined as a feeding station (Ruyle and Dwyer, 1985). The selection of a feeding station is highly related to sward mass, structure, and quality, so it is a good indicator of grazing conditions (Carvalho and Moraes, 2005). In this context the animal behavior could be summarized as searching time and displacement rate between feeding stations and bite mass (Griffiths et al., 2003) and residence time in each feeding station (Gonçalves et al., 2009), as a general relationship has been postulated that large herbivores tend to spend more time exploring each feeding station as forage biomass increases (Searle et al., 2005).

Grazing management strategies will also have direct effects on pasture growth, where low post-grazing sward heights have an unfavorable impact on pasture regrowth and on the total forage production (Chapman, 2016). In Uruguay, previous research has indicated that increasing post-grazing sward height presents benefits for both pasture and animal performances (Mattiauda et al., 2009; Zibil et al., 2016). However, studies to understand the relationships between daily dry matter intake in eating behavior and milk production of dairy cows at different residual sward height management are scarce. The potential to increase milk production by increasing post-grazing sward height, and seeking an optimal point in the plant-animal interface, is still unknown in fescue. Therefore, an experiment was conducted to study the influence of post-grazing sward height on dry matter intake, milk production, daily grazing and ruminating time as well as frequency and length of grazing meals and ruminating bouts as well as exploration of feeding stations and patches. The hypothesis was that higher post-grazing sward height will increase intake rate, total dry matter intake, and milk production, through a more selective ingestive behavior, leading to a diet with a higher nutritive characteristic. These cows will do that by dividing the daily grazing time into shorter meals followed by a ruminating session and will devote more time to exploring each feeding station with fewer steps of searching between them.

The experiment was carried out at the Experimental Research Station “Dr. Mario A. Cassinoni” (EEMAC) of the Faculty of Agronomy–University of Uruguay (Paysandú, Uruguay, 32°S, 58°W), from late October to early November 2017. A pasture of tall fescue (Lolium arundinaceum cv. INIA Fortuna) was sown in May 2016 at 10 kg/ha. At the time of sowing, it was fertilized with 18 kg of N and 46 kg of P/ha and in the 2nd year, 180 kg of N/ha was applied in four equal applications from June to October. In February 2017, the sward was homogenized by mechanically mowing to 6 cm and the grazing treatments were implemented in April. Animal procedures were approved by the Animal Experimentation Committee of the University of Uruguay, application number 02174500019718 (Montevideo, Uruguay).

We have found that the establishment and maintenance of treatments for 8 months before the period of measurement generated a different spatial arrangement of the height (proportion of HH, MH, and LH) and a distinctive canopy vertical structure in each treatment that affected the behavioral pattern of the animals. The grazing started with equal sward height and herbage mass between treatments, but the different depletion rate during grazing down resulted in a different post-grazing sward height, although no differences in herbage mass, probably because the different vertical structures of the canopy between treatments (sheath height) changed the density of the residual strata. Therefore, as suggested by Gibb et al. (1997), the productive and behavioral performance of the animals should be analyzed in a context of a complex environment, because mean values of biomass or sward surface height would be poor predictors of cattle foraging behavioral responses (Laca et al., 1992). For a better correlation of sward characteristics and animal behavior, an analysis of sward height distribution during the beginning and the end of the occupation period was performed. The heterogeneity increased from TC to TL, given that it had an increase in the range, variance, and standard deviation of sward height distribution. Although heterogeneity is a complex, rich, and multidimensional concept (Laca, 2008), we refer to it as the variation on sward surface height across the paddock and days (space and time scale). When a large variance is found, it means that both good and poor bites are relatively abundant; in the opposite, when a small variance is found, most of the bites are similar (Laca, 2008). The variance was almost double on TL than on TC (29.3 vs. 16.1) and, despite no effect of treatment by moment interaction, the variance decreased 71% from the beginning to the end of the occupation period on TC, whereas the same reduction was about 7% on TM and 29% on TL. Heterogeneity can affect intake through effects on bite formation and dimensions, as well as selectivity, which will be beneficial when searching costs do not increase because of it (Utsumi et al., 2009).

The daily grazing time was not affected by treatment, but was due to the interaction with the time of day and moment of the occupation period. It is argued that cows on each treatment adopted a different behavioral strategy during the grazing process as a result of the changing sward structure. The daily grazing time is defined by Gibb et al. (1998) as a sum of all grazing meals. At a post-grazing sward height of 12 cm (TM), the fact that cows increased grazing time by ≈20 min, although this was not statistically significant, is highlighted by the lower ruminating time found in this treatment when compared with TC and TL (69 and 41 min lower, respectively). The time spent on ruminating and grazing could have had a compensatory relationship, that is, when the grazing time is longer, the ruminating time tends to decrease (Dado and Allen, 1994), given that the total chewing time is less variable (Beauchemin, 2018). The number of grazing meals found in this study is in accordance with that expected for dairy cattle (Gregorini, 2012), presenting a maximum value on TL at the beginning moment (6.3), where forage was abundant. The length of each grazing meal is highly dependent on rumen fill (Chilibroste et al., 1997), although it has been related to a multi-factorial response (Gregorini, 2012; Chilibroste et al., 2015). Normally, between two consecutive grazing meals, the animal ruminates (Gregorini, 2012) to release soluble nutrients from feeds (Chilibroste et al., 2007). On TC, the ruminating bout length was greater than other treatments, and, according to Amaral et al. (2013), by decreasing post-grazing sward height the animal is forced to explore strata with a predominance of stems and senescent material, increasing the intervals between meals.

In the literature there is evidence that the dusk grazing event is the longest and most intense (Gibb et al., 1998; Gregorini, 2012; Chilibroste et al., 2015); however, only cows managed on TL in our experiment followed this grazing pattern. This is possibly due to the less favorable grazing conditions on both TC and TM, because the cows increased both morning and evening grazing meal length to compensate for low intake rates. On both TC and TM the compensation was limited by requirements of rumination and idling time (Chilibroste et al., 2015), but on TC the low intake rate and lower estimated digestibility of the diet resulted in a reduced dry matter intake and consequently milk production. Furthermore, sward height depletion has been associated with short-term behavior adaptations, such as bite mass and intake rate reductions (Gregorini et al., 2009a). On TM, the sward structure allowed a more efficient compensation mechanism, that resulted in intermediary milk production which was achieved through the same dry matter intake as TL from a diet with lower digestibility.

The chemical composition of sward changes through the day through photosynthates accumulation, with the maximum levels of dry matter and non-structural carbohydrate concentrations in the evening (Griggs et al., 2005) and in the upper layers of the sward (Delagarde et al., 2000). Therefore, analyzing the two most important grazing events (first morning and evening meal; Gibb et al., 1998), the TL cows grazed double in the evening compared to the morning, while on TC and TM grazing time was equal across the day. In addition, the higher intake rate achieved on TL occurred at the time when the nutritional value of the sward was highest (Abrahamse et al., 2008). The theoretically higher non-structural carbohydrate concentrations in the evening, associated with the highest crude protein level in herbage samples on TL led to higher dry matter intake and milk production. Controversially, the neutral detergent fiber and the diet digestibility was higher on TL. It is possible that a better match of nutrients along the day in the rumen of TL cows allowed a better diet digestibility despite the higher level of NDF. In other words, a higher crude protein and probably soluble carbohydrate intake added to shorter grazing meals followed by a ruminating session determined a better ruminal environment and thus a higher extent of fermentation. According to Filho et al. (2012), the pattern of ruminal fermentation clearly depends on changes in nutritional values that occur in different sward layers during grazing.

The information about feeding station exploration is widely variable between studies (Roguet et al., 1998b; Searle et al., 2005; Gregorini et al., 2007, 2009b, 2011; Gonçalves et al., 2009; Da Trindade et al., 2012; Amaral et al., 2013; Mezzalira et al., 2013, 2014), but there is a consensus that it is highly dependent of sward vertical structure and spatial distribution, herbage mass, and the internal state of the animals, resulting in a trade-off determining the behavior pattern adopted throughout the grazing down process within the day and between days.

The exploration of feeding stations enhanced the adaptation of the behavioral patterns of cows to adapt to different scenarios. According to Gregorini et al. (2007), feeding station behavior is partially controlled by ruminal fill and their interaction with external factors such as sward characteristics. The decision to move to a new feeding station will depend of the profitability of each bite (trade-off between the benefit and cost of moving on) which is highly connected to the animal perception of resource availability (Roguet et al., 1998a). We found that cows on TL explored a higher number of feeding stations but spent less time at each feeding station, probably because the benefit outweighed the cost of travel (Utsumi et al., 2009).

Utsumi et al. (2009) using artificial patches of fescue and lucerne found that the cows increased their patch residence time when the distance between patches was increased, leading to increased speed between patches. The authors suggest that the animals adopted a strategy linearly related to the maximization of intake rate. In this study, the cows on TC remained at each feeding station longer but took more steps between feeding stations than TL cows. According to Utsumi et al. (2009), foragers maximizing intake rate increase travel time between patches by increasing the residence time in each patch. We, however, did not find an increase in travel time between patches, but the residence time was lower on TM compared with TL, resulting in a greater number of patches explored by TM cows. On the other hand, the feeding stations explored per patch was greater on TL. These results suggest that TL achieved the selectivity at the feeding station and bite scale while TM cows did it at the patch scale. The patch can be defined as a spatial aggregation of feeding stations (and bites, consequently) over which intake or movement rate remains relatively constant (Bailey et al., 1996). Therefore, considering that the number of patches per hour was lower on TL than on TM, the TL cows had a more constant intake rate.

If departure decisions at patch scale are influenced by trade-offs between maximizing intake rate and food quality (Searle et al., 2005), we can think that TL patches were more attractive to the cows, because on this treatment the animals spent more time at each patch with greater diet digestibility. On the other hand, TM cows explored more patches achieving the same dry matter intake as TL cows, but the selected diet was inferior in terms of quality and, consequently, the milk production was lower as well.

A general relationship has been postulated that large herbivores tend to spend more time in feeding stations as forage biomass increases (Searle et al., 2005). However, in this case, the mean herbage mass did not differ between treatments, giving spatial arrangement and a distinctive canopy vertical structure. The cows on TL grazed more “high structures” (Giles et al., 2018) than on TM and on TC and increased the residence time per feeding station, conversely conforming to our hypothesis but still generating a greater nutrients intake. Similar results about structure selection were reported by Faber (2012) with the same pasture species. In a short-term study on natural grasslands, Gonçalves et al. (2009) found the opposite. The animals increased the residence time and decreased the number of feeding stations visited when sward height increased. In accordance with our results, Mezzalira et al. (2013) observed that under low sward heights, animals explored more feeding stations, and they argue that probably grazing patterns were affected by spatial memory, giving that it was a large study (more than 4 months). On the same line, according to Roguet et al. (1998a) and Searle et al. (2005), grazing herbivores tend to utilize feeding stations in such a way as to remove more digestible forage before moving on; however, the animal can prioritize intake rate rather than selection when the transmission of signals to the brain in response to low amounts of digesta in the rumen occurs (Gregorini et al., 2009a). The latter probably occurred on TC and TM, but with a different trade-off, in response to the different canopy structures.

Studying the effect of manipulated ruminal fill on beef heifers, Gregorini et al. (2007) observed that the “hungrier” (fewer ruminal fill) the animals were, the greater the time invested in each feeding station was, and the faster they searched. This can help explain why TC cows explored for a longer time at each feeding station, since dry matter intake on this treatment remained below potential. This is in agreement with the assumption that cattle can reduce forage manipulation time during feeding to increase the intake rate (Parsons et al., 1994). On the opposite side, TL cows with a large bite mass (short-term measurement; Oborsky et al., 2019) and finally greater dry matter intake, probably spent more time manipulating each bite while searching for a new feeding station to explore. This leads us to believe that the difference in time spent ruminating per kg of NDF (15 min per kg of NDF less on TL than on TC) can be influenced by differences in processing during eating.

The results from the current study confirm our hypothesis that dry matter intake and milk production increases with a reduction in defoliation intensity through more selective ingestive behavior leading to a higher digestibility diet. Given that the first limitation on grazing animal performance is the feed intake (Holmes et al., 2002), the main reason the milk production increased from TC to TM and TL was the difference in the dry matter intake (3 kg of DM per cow per day). From TC to TM the dry matter intake increased 1 kg per extra cm in post-grazing sward height and milk yield response was 0.9 kg of energy corrected milk per extra cm in post-grazing sward height. The same milk yield response was found from TC to TL and from TM to TL. However, an increase on dry matter intake between TM and TL was not found, but it was on digestible dry matter intake, which was 1 kg lesser on TM than on TL. Ganche et al. (2013) contrasting 2.7 vs. 3.5 cm of post-grazing sward height, with supplemented cows, found a 30% increase on grass digestible organic matter intake, while in this study we found an increase of 23, 35, and 10% for TC vs. TM, TC vs. TL, and TM vs. TL, respectively on digestible dry matter intake. McEvoy et al. (2008), comparing 3.8 vs. 4.6 cm of post-grazing sward height reported a higher milk yield response (2.1 kg of milk per extra cm in post-grazing sward height) than we found. A similar milk yield response was reported by Delaby et al. (2001; 0.8 kg of milk per extra cm in compressed plate meter post-grazing sward height; 5.7 vs. 6.8 cm). The increase in milk production was related to a higher energy supply from an increasing digestible dry matter intake.

The management of the defoliation intensity determined different levels of heterogeneity in the pasture, resulting in the adoption of different behavior patterns of the animals at different spatio-temporal scales. This indicates that animal behavior is the basis of knowledge required for the decision-making process based on animal and pasture monitoring. The integration of the different behavioral scales evaluated in this experiment (feeding station, patch, grazing and ruminating session, and daily grazing and ruminating time) allowed us to understand the how and why of cow's grazing behavior.

The behavior data must be hierarchically integrated to gain a better understanding of the processes and allow the development of decision-making support models. In this sense, technological innovations that allow real-time monitoring of the animals in their pastoral environment should be used to capitalize the animal and pasture heterogeneity more than trying to homogenize it.

The intensity of defoliation modulates the animal-plant interaction and constitutes a valuable management tool that can be used to boost forage intake and milk production. The new developments on information and communication technology would allow for linking behavioral data with response variables (e.g., milk production, health, welfare, etc.) at different spatio-temporal scales and support either short (day-to-day) or long (months to years) term management decisions.

The height and structure of the pasture on control treatment resulted in lower intake rate with less selectivity as cows' strategy to compensate for the lower sward height. In the medium treatment, cows adopted a behavioral strategy different from the others, achieving the same dry matter intake than TL cows but without increases in milk production. The lax management of the grass resulted in a more digestible dry matter intake and therefore higher milk production, achieved through behavioral adaptations throughout the grazing process expressing higher selectivity. The cows' response to changes in mean residual sward height and their associated heterogeneity is not the adoption of a uniformed pattern since we observed changes throughout the occupation period and throughout the day. The dairy grazing cow has great capacity to adapt its behavior to different scenarios and does it at different scales of time and space.

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.

The animal study was reviewed and approved by Animal Experimentation Committee of the University of Uruguay, application number 02174500019718 (Montevideo, Uruguay).

GM participated in experiment design, performed the field experiment, analyzed the results, promoted data analysis discussions, wrote the original draft, and revised versions of the manuscript. PG conceptualized and designed the field experiment, performed the field experiment, participated in data analysis discussions, and reviewed and edited the manuscript. MO participated in experiment design, performed the field experiment, and participated in data analysis discussions. OF participated in data analysis discussions. DM conceived, designed, supervised field experiments, and participated in data analysis discussions. TG designed and supervised field experiments, alkane lab analysis, and participated in data analysis discussions. PC conceived, designed, obtained funding, administrated the project, supervised field experiments, results analyses, and participated in data analysis discussions. All authors reviewed the manuscript and approved the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.