• Full Text
• Scholarly Journal

# Dormant Season Grazing: Effect of Supplementation Strategies on Heifer Resource Utilization and Vegetation Use

Rangeland Ecology and Management; Lawrence Vol. 72, Iss. 6,  (Nov 2019): 878-887. DOI:10.1016/j.rama.2019.06.006

### Full text

Introduction

Heifer development is essential to the productivity and longevity of a cow/calf operation and typically results in additional feed costs to ensure heifers reach a target prebreeding weight that supports ideal reproductive performance (Patterson et al., 1992; Roberts et al., 2009; Mulliniks et al., 2013a). However, high feed costs have caused producers to explore low-input management strategies that minimize costs associated with providing harvested feeds (Adams et al., 1996). Late-season forage corresponding to fall weaned heifers is typically low in quality, potentially putting animals in a state of negative energy balance (Mulliniks et al., 2013b). Protein supplementation while grazing dormant range can enhance heifer growth and improve pubertal status at breeding and reproductive performance (Stalker et al., 2006; Martin et al., 2007). Thus, developing heifers on poor-quality forages may require considerable supplementation of nutrients in order to achieve target bodyweight and reproductive efficiency (Galyean and Goetsch, 1993; Bowman et al., 2004; Mulliniks et al., 2013a).

One of the primary goals of a forage-based livestock production system is to obtain optimal animal productivity while effectively using the forage resource base. Thus, the spatial component of herbivory is a central aspect of domestic livestock ecosystems (Coughenour, 1991). Topography, thermal environments, and forage resources (e.g., standing crop and nutritional quality) interact to determine space use (Jamieson and Hodgson, 1979; Adams et al., 1986; Beaver and Olson, 1997). Providing a supplement alters the nutrient status of animals, which can have strong influences on grazing behavior (Allison, 1985; Adams et al., 1986). Supplementation changes grazing distribution on rangelands (Ares, 1953; Bailey et al., 2001) and daily grazing activities (Adams, 1985), altering the distribution of vegetation use based on supplement form, delivery method, and location (Bailey and Welling, 2007; Bailey and Jensen, 2008). Therefore, it is likely that grazing behavior may vary with protein supplementation strategies and efficiency in dormant-season grazing systems (Murden and Risenhoover, 1993; Bailey and Welling, 2007; Odadi et al., 2013).

Dormant forage is often perceived to be more tolerant of grazing pressure (Petersen et al., 2014); however, improperly managed dormant season grazing can have detrimental effects on vegetation production and residual cover (Willms et al., 1986; Bullock et al., 1994; Petersen et al., 2014). Vegetation composition and structural heterogeneity are key habitat characteristics that influence wildlife species diversity and ecosystem function (Wiens, 1997; Bailey et al., 1998; Fuhlendorf et al., 2009). Habitat structural heterogeneity is recognized as a precursor to biological diversity at most levels of ecological organization and is proposed to be the foundation of conservation and ecosystem management (Christensen, 1997; Wiens, 1997; Fuhlendorf and Engle, 2001). Western rangelands are inherently heterogeneous as vegetation composition and structure vary with topographic and edaphic features (Patten and Ellis, 1995; Fuhlendorf and Smeins, 1998; Fuhlendorf et al., 2006). However, traditionally supplementation has been used to alter grazing distribution and promote uniform vegetation utilization across the landscape (Bailey and Welling, 1999). Management strategies that promote uniform utilization of vegetation results in the homogenization of rangeland landscapes and an overall decline of ecosystem structure, function, and biodiversity (Derner et al., 2009; Fuhlendorf et al., 2009; Hovick et al., 2015). Maintaining heterogeneity of vegetation composition and structure in rangelands is beneficial to ecological biodiversity (Fuhlendorf et al., 2006). However, little is known about the effects of supplementation strategies on winter grazing behavior of heifers and the potential short-term impacts on residual vegetation composition, structure, and heterogeneity.

Information relating to supplementation strategies to individual grazing behavior by heifers and vegetation use on dormant forage is lacking. Thus, the specific objectives of this study were to evaluate how two different protein supplementation strategies during the dormant grazing season influence 1) grazing activity and resource utilization by heifers and 2) residual vegetation cover, structure, and heterogeneity. We expected supplementation strategy to affect grazing behavior and resource use of cattle, thereby altering structure and cover of residual vegetation. System-level impacts are likely mediated by the provision of supplement and uncontrolled environmental conditions.

Methods

The use of animals in this study was approved by the Institutional Animal Care and Use Committee of Montana State University (2015-AA04). This study was conducted at the Ft. Keogh Livestock and Range Research Laboratory near Miles City, Montana (46°22'N 105°5'W). Elevation of the study area ranged from 750 m to 837 m. Climate is characterized as continental and semiarid with an average annual pecipitation of 338 mm, 60% of which occurs during the 150-d growing season (mid-April to mid-September). Vegetation is dominated by western wheatgrass (Pascopyrum smithii [Rydb.] Love), threadleaf sedge (Carex filifolia Nutt.), needle and thread (Stipa comata Trin. and Rupr.), and blue grama (Bouteloua gracilis [H.B.K.]). Our study occurred during the winter grazing seasons of 2015–2016 and 2016–2017. Precipitation during the growing season was higher in 2016 than 2015 (258, 172 mm), likely reflected as higher vegetation production during the 2016–2017 winter grazing season (Table S1; available online at https://doi.org/10.1016/j.rama.2019.06.006). Preciptiation for both winters was similar (2.77, 2.92 cm); however, temperatures during the 2016–2017 winter were substantially cooler than during the winter of 2015–2016 (−4.85, 0.26°C), resulting in higher amounts and prolonged periods of snow in the second year of the study (Table S2).

The use of animals in this study was approved by the Institutional Animal Care and Use Committee of Montana State University (2015-AA04). This study was conducted at the Ft. Keogh Livestock and Range Research Laboratory near Miles City, Montana (46°22'N 105°5'W). Elevation of the study area ranged from 750 m to 837 m. Climate is characterized as continental and semiarid with an average annual pecipitation of 338 mm, 60% of which occurs during the 150-d growing season (mid-April to mid-September). Vegetation is dominated by western wheatgrass (Pascopyrum smithii [Rydb.] Love), threadleaf sedge (Carex filifolia Nutt.), needle and thread (Stipa comata Trin. and Rupr.), and blue grama (Bouteloua gracilis [H.B.K.]). Our study occurred during the winter grazing seasons of 2015–2016 and 2016–2017. Precipitation during the growing season was higher in 2016 than 2015 (258, 172 mm), likely reflected as higher vegetation production during the 2016–2017 winter grazing season (Table S1; available online at https://doi.org/10.1016/j.rama.2019.06.006). Preciptiation for both winters was similar (2.77, 2.92 cm); however, temperatures during the 2016–2017 winter were substantially cooler than during the winter of 2015–2016 (−4.85, 0.26°C), resulting in higher amounts and prolonged periods of snow in the second year of the study (Table S2).

Sampling

We established thirty, 30-m transects randomly within each paddock. We measured vegetation production, canopy cover, and visual obstruction readings (VOR) at six 0.1-m2 plots located every 5 m along each transect. Canopy cover of plant functional groups (grass, forb, shrub), cover of bare ground, and litter were estimated at each plot using the six-cover class Daubenmire method (Daubenmire, 1959). Ground cover of plant functional groups, bare ground, and litter have all previously been found to influence the abundance and demography of grassland birds, who serve as indicator species for grassland ecosystems (Bock and Webb, 1984; Bradford et al., 1998; Derner et al., 2009). Visual obstruction readings were measured in four cardinal directions using a 1-m Robel pole (Robel et al., 1970). Visual obstruction is typically correlated with aboveground biomass and represents a measure of the vertical structure and density of vegetation (Robel et al., 1970; Damiran et al., 2007). All measurements were taken before and after grazing prior to spring green up to evaluate how supplementation strategy affected residual vegetation structure across the paddocks. Pregrazing vegetation functional group production was estimated using the dry weight rank method and clipping each plot (Mannetje, 1963; Dowhower et al., 2001). Clipped samples were placed in a forced-air oven at 60°C for 48 h and then weighed. Vegetation composition was calculated by taking pregrazing vegetation functional group production divided by the total vegetation production for each transect (Coulloudon et al., 1999). Pregrazing samples from each paddock were composited by transect and ground to pass a 1-mm screen in a Wiley mill. Samples were then analyzed in duplicate for nitrogen (Leco CN-2000; Leco Corporation, St. Joseph, MI) and fiber (neutral detergent fiber [NDF] and acid detergent fiber [ADF]; Ankom 200 Fiber Analyzer, Ankom Co., Fairport, NY) as indicators of vegetation quality (see Table S1).

We established thirty, 30-m transects randomly within each paddock. We measured vegetation production, canopy cover, and visual obstruction readings (VOR) at six 0.1-m2 plots located every 5 m along each transect. Canopy cover of plant functional groups (grass, forb, shrub), cover of bare ground, and litter were estimated at each plot using the six-cover class Daubenmire method (Daubenmire, 1959). Ground cover of plant functional groups, bare ground, and litter have all previously been found to influence the abundance and demography of grassland birds, who serve as indicator species for grassland ecosystems (Bock and Webb, 1984; Bradford et al., 1998; Derner et al., 2009). Visual obstruction readings were measured in four cardinal directions using a 1-m Robel pole (Robel et al., 1970). Visual obstruction is typically correlated with aboveground biomass and represents a measure of the vertical structure and density of vegetation (Robel et al., 1970; Damiran et al., 2007). All measurements were taken before and after grazing prior to spring green up to evaluate how supplementation strategy affected residual vegetation structure across the paddocks. Pregrazing vegetation functional group production was estimated using the dry weight rank method and clipping each plot (Mannetje, 1963; Dowhower et al., 2001). Clipped samples were placed in a forced-air oven at 60°C for 48 h and then weighed. Vegetation composition was calculated by taking pregrazing vegetation functional group production divided by the total vegetation production for each transect (Coulloudon et al., 1999). Pregrazing samples from each paddock were composited by transect and ground to pass a 1-mm screen in a Wiley mill. Samples were then analyzed in duplicate for nitrogen (Leco CN-2000; Leco Corporation, St. Joseph, MI) and fiber (neutral detergent fiber [NDF] and acid detergent fiber [ADF]; Ankom 200 Fiber Analyzer, Ankom Co., Fairport, NY) as indicators of vegetation quality (see Table S1).

We monitored cattle grazing activity using Lotek GPS collars (3300LR; Lotek Engineering, Newmarket, Ontario, Canada) containing head position sensors that recorded daily timing and locations of grazing activities (Turner et al., 2000; Ungar et al., 2005; Brosh et al., 2010). Heifers within each treatment were stratified by bodyweight and then randomly selected for GPS collars (seven collars/treatment) with new animals selected every 28 d to minimize autocorrelation in responses among grazing periods. GPS data were used for each supplement treatment to evaluate the effects of supplementation strategy on animal grazing activity (Ungar et al., 2005; Brosh et al., 2010; Valente et al., 2013). Each collar was configured to record GPS positions, head position, and vertical/horizontal movements at 5-min intervals. Each collar stored the percentage of time the head position sensor registers in the down position (grazing activity) for each sampling period. We then used the binary classification methods developed by Augustine and Derner (2013) to separate grazing from nongrazing activities to examine time spent grazing and cattle foraging distribution. Limiting observations to grazing locations allowed us to determine important foraging areas rather than general pasture occupancy (Walburger et al., 2009).

Temperature, wind speed, and direction data were collected using a permanent weather station located at the experimental paddocks. Wind speeds at 30-m2 spatial resolution were predicted across all paddocks using average daily wind measurements collected on site, ArcGIS spatial analyst tool, a digital elevation model at 30-m2 resolution, and WindNinja wind prediction software (Brooks, 2012). In addition, HOBO Pendant Temperature/Light Data Logger (Onset Computer Corporation, Bourne, MA) were deployed at each randomly selected transect location within each paddock and programmed to collect ambient temperature every 30 min. Fine-scale ambient air temperature was modeled using the collected temperature data set and generalized linear models to evaluate the effects of physical properties (e.g., aspect, elevation and slope) on fine-scale temperature of the paddocks. Model results were used to create spatially explicit predictions of temperature and wind conditions across each paddock, which were used as covariates in subsequent resource utilization modeling.

All supplementation and water locations were located via handheld GPS (spatial error < 10 m). Using the spatial analysis tool in ArcGIS (Environmental Systems Research Institute, Redlands, CA) and a digital elevation model at 30-m2 resolution (USGS, 2017), we created additional spatial covariate layers representing slope, aspect, and horizontal and vertical distance from supplement locations at 30-m2 resolution.

Statistical Analysis

To evaluate the effects of supplementation strategy on vegetation functional group cover and visual obstruction, we calculated the differences in average vegetation measurements collected before and after grazing at each transect for each treatment paddock in each year. The effects of supplementation treatment on patch-level heterogeneity of functional group vegetation cover and VOR were obtained by subtracting the pregrazing transect-level standard deviation from the postgrazing transect-level standard deviation for each treatment paddock within each time period for both years. Due to only one treatment replication per grazing time period and paddocks alternating treatments the second year of the study, year was used as replication of treatment within grazing time period. Therefore, the effects of supplementation strategy on vegetation conditions and time spent grazing were analyzed using mixed-effects analysis of variance with a generalized mixed model including treatment, time period, and a treatment by time period interaction as fixed effects with a random intercept of treatment and random slopes effect of year. Data were plotted and log-transformed if needed to satisfy assumptions of normality and homogeneity of variance. An alpha ≤ 0.05 was considered significant.

To model the influence of supplementation treatment on space use, individual GPS-collared heifers in each treatment were defined as the biological unit of interest in modeling grazing resource utilization within paddocks. We quantified space use related to habitat covariates using multiple regression in a resource utilization function analysis (RUF; Marzluff et al., 2004; Winder et al., 2014). Resource utilization functions quantify interanimal variation in resource use and examine use as a probabilistic and continuous metric, allowing for an increase in sensitivity for detecting resource selection for individuals within each supplementation treatment (Winder et al., 2014). In addition, RUFs treat individual animals as the experimental unit, incorporating an individual’s entire distribution of use, independently, while accounting for spatial autocorrelation of multiple locations per individual and reducing errors associated with location estimation and focus on specific locations (Marzluff et al., 2004; Kertson and Marzluff, 2010).

Due to cattle home range being confined to paddock management units, GPS collar data were used to build resource utilization models on a third-order scale, defining animal movements and selection of environmental and vegetation conditions within each paddock (Johnson, 1980). Using the Geospatial Modeling Environment (Beyer, 2010), we created a raster representing the period-specific utilization density distribution for the grazing locations of individuals in each paddock. Relative use values were bounded between 1 and 99, for each 30 m2 cell, based on the relative volume of utilization distribution (Marzluff et al., 2004). Environmental covariates expected to influence heifer resource utilization included temperature, wind, distance to supplement (vertical and horizontal), supplementation treatment, slope and aspect, annual forage production, and quality expressed as CP, NDF, and ADF. Environmental covariates and individual relative use rasters were stacked and converted to spatially explicit data files using the “raster” function in R as an input for the ruf.fit package (Kertson and Marzluff, 2010). Individual relative use values were log-transformed before modeling to meet the assumptions of multiple regression models. Using the ruf.fit package in R, resource utilization functions with standardized β coefficients were generated and evaluated for each individual in both treatment groups for each time period to represent the influence of the environmental covariates on heifer resource utilization (Marzluff et al., 2004; Kertson et al., 2011).

To develop treatment-level inferences, we calculated mean standardized $β¯ˆ$ coefficients and a population-level variance that incorporated individual animal variation for each environmental covariate by supplementation treatment and time period (Marzluff et al., 2004). Standardized coefficients with 95% confidence intervals that do not overlap zero are considered significant predictors of space use (Marzluff et al., 2004; Winder et al., 2014). If a resource utilization coefficient is significantly different from zero, we inferred that resource use was greater or less than expected on the basis of the availability of the resource within the treatment paddock (Marzluff et al., 2004; Winder et al., 2014).

A primary step in RUF analyses is to overlay spatially explicit utilization distributions of individuals (e.g., heifers) onto georeferenced rasters representing spatially explicit estimates of covariate values across the entire study area (Marzluff et al., 2004; Kertson and Marzluff, 2010; Winder et al., 2014). Unfortunately, spatially continuous estimates of vegetation composition, production, and quality were not available at appropriate spatial resolution for our entire study area. Therefore, to model the effects of vegetation conditions on resource use, we stacked the transect location shapefile with the predicted relative resource use rasters for all individuals to extract and pair individual relative use to the corresponding transect vegetation measurements. To avoid overfitting our resource utilization models, we conducted a preliminary multicollinearity analysis to select uncorrelated (|r| > 0.6) variables that are ecologically relevant and feasible to measure (Dormann et al., 2013). If two covariates were correlated, we fitted preliminary resource utilization models and evaluated relative support of each variable using Akaike Information Criterion adjusted for small sample sizes (AICc, Burnham and Anderson, 2002); we retained the variable with more relative support for further modeling and discarded the correlated variable (Fieberg and Johnson, 2015).

We evaluated the effects of vegetation (e.g., production, composition, quality) and supplement treatment on relative use using generalized linear mixed models with a Gaussian (normal) error structure and individual animal as a random effect. We used AICc to evaluate support for competing models reflecting hypotheses about the effects of various vegetation attributes and supplement treatment on relative use by heifers (Table S3; available online at https://doi.org/10.1016/j.rama.2019.06.006; Burnham and Anderson, 2002). Models with ΔAICc ≤ 2 that differed from the top model by a single parameter were excluded if confidence intervals of parameter estimates overlapped 0 (ie., were noninformative; Arnold, 2010). Model fit was then evaluated by calculating marginal and conditional r2 values for generalized linear mixed models (Nakagawa and Schielzeth, 2013). All statistical analyses were performed in R (R Core Team, 2017).

We evaluated the effects of vegetation (e.g., production, composition, quality) and supplement treatment on relative use using generalized linear mixed models with a Gaussian (normal) error structure and individual animal as a random effect. We used AICc to evaluate support for competing models reflecting hypotheses about the effects of various vegetation attributes and supplement treatment on relative use by heifers (Table S3; available online at https://doi.org/10.1016/j.rama.2019.06.006; Burnham and Anderson, 2002). Models with ΔAICc ≤ 2 that differed from the top model by a single parameter were excluded if confidence intervals of parameter estimates overlapped 0 (ie., were noninformative; Arnold, 2010). Model fit was then evaluated by calculating marginal and conditional r2 values for generalized linear mixed models (Nakagawa and Schielzeth, 2013). All statistical analyses were performed in R (R Core Team, 2017).

Results Grazing Behavior and Space Use

Average daily time spent grazing varied across treatments and time periods (P < 0.01). Cattle supplemented with concentrated protein spent more time grazing than cattle supplemented with cake (6.92 ± 0.18, 6.24 ± 0.17 h; see Fig. S1A). In addition, both treatments spent more time grazing during time period 3 (7.64 ± 0.21 h) than time periods 1 and 2 (6.31 ± 0.20, 6.12 ± 0.18 h; see Fig. S1B).

Average daily time spent grazing varied across treatments and time periods (P < 0.01). Cattle supplemented with concentrated protein spent more time grazing than cattle supplemented with cake (6.92 ± 0.18, 6.24 ± 0.17 h; see Fig. S1A). In addition, both treatments spent more time grazing during time period 3 (7.64 ± 0.21 h) than time periods 1 and 2 (6.31 ± 0.20, 6.12 ± 0.18 h; see Fig. S1B).

We estimated herd-level grazing RUFs from 42 heifers under 2 supplementation treatments during 3 winter time periods (7 heifers/treatment/time period), with an average of 1 852 ± 42 grazing locations per individual. Relative resource utilization by grazing heifers in the cake treatment were negatively related to horizontal distance from the supplement feeding location in time periods 1 and 2 ($β¯ˆ$ = −0.41 ± 0.16SE, −0.53 ± 0.17; Fig. 1). Relative resource selection by heifers in the protein treatment tended to decrease with horizontal distance from supplement during time periods 1 and 2 ($β¯ˆ$ = −0.22 ± 0.25, −0.39 ± 0.26); however, individual variability in habitat selection resulted in confidence intervals overlapping 0 for the herd-level responses (see Fig. 1). Low values of relative use coefficients indicate average temperature, wind speed, aspect, slope, and vertical distance from supplement had little influence on grazing space use for all time periods for both supplementation groups. However, vertical and horizontal distance from supplement and temperature were highly variable among individuals as drivers of space use for both supplementation treatments (Fig. 2A−C).

The relationships among vegetation composition, production, and quality; supplementation strategy; time period of grazing; and relative resource utilization by heifers were evaluated for 42 heifers (7 heifers/treatment/time period) using pregrazing vegetation conditions measured at 180 transect locations (30 transects/paddock) in the study area. A single model containing crude protein (%), standing biomass of perennial grass (kg/ha), and a supplementation strategy by time period interaction had 88% of the relative support of the data (Table 1). Predicted relative resource utilization of grazing heifers for both treatments increased by 5% ± 0.4% with every 100 kg/ha standing biomass of perennial grasses ($β¯ˆ$ = 0.0005 ± 0.00004; Fig. 3B). In addition, for every 1% increase in CP, relative use increased by an average of 12% ($β¯ˆ$ = 0.12 ± 0.007; see Fig. 3A). The protein concentrate self-fed treatment tended to have a 43% ± 28% higher overall relative use than the cake supplementation group in time period 1; however, confidence intervals overlap 0, indicating no significant supplement treatment-level effects ($β¯ˆ$ = 0.43 ± 0.28; see Fig. 3A and 3B). The effect sizes of relative grazing use for the self-fed protein supplement treatment decreased in time periods 2 and 3 ($β¯ˆ$ = −0.09 ± 0.44, −0.56 ± 0.07). Conversely, relative use for the hand-fed cake treatment increased in time periods 2 and 3 ($β¯ˆ$ = 0.07 ± 0.28, 0.39 ± 0.05), resulting in similar relative use between treatments in time periods 2 and 3 (see Fig. 3A and 3B). The top model had a conditional r2 of 0.71; however, the marginal r2 was only 0.07, suggesting the majority of the variation in space use by grazing heifers was explained by individual-level variation.

Vegetation Structure and Heterogeneity

We found no treatment effects on before and after grazing differences in mean residual cover of litter, grass, forbs, or shrubs (P > 0.24; Table 2). However, pre-post differences in mean litter, residual grass, and shrub cover differed among time periods (P < 0.01). Litter cover increased after grazing in time period 1 (11.45% ± 2.05% SE), was unchanged after time period 2 (−2.78% ± 1.53%), and was decreased after grazing in time period 3 (−15.28% ± 1.80%; Fig. 4B). Grass residual cover increased with grazing in time period 1 (7.57% ± 1.98%), decreased in time period 2 (−13.85% ± 1.67%), and further decreased in time period 3 (−35.95% ± 3.06%, see Fig. 4A). Shrubs showed no grazing effect in time period 1 (1.31% ± 0.83%) but slightly decreased in time periods 2 and 3 (−1.84% ± 0.83%, −1.53% ± 0.68%; see Fig. 4C). Bare ground cover displayed a treatment-level effect (P < 0.01), where bare ground was decreased 7.29% ± 1.12% with livestock grazing in paddocks treated with protein supplementation but not in paddocks treated with cake (0.17% ± 1.31%; see Fig. S2A). Bare ground cover was also associated with the time period of grazing (P = 0.02), decreasing in time periods 1 and 2 (−6.88% ± 1.70%, −3.42% ± 1.49%) with no effect in time period 3 (−0.89% ± 1.32%; see Fig. S2B). A treatment by time period interaction occurred when evaluating the effects of grazing on visual obstruction (P < 0.04). Visual obstruction was significantly reduced by heifers grazing during time period 1 under both supplementation treatments but only significantly reduced in the cake group during periods 2 and 3 (Fig. 5A).

We found no treatment effects on before and after grazing differences in mean residual cover of litter, grass, forbs, or shrubs (P > 0.24; Table 2). However, pre-post differences in mean litter, residual grass, and shrub cover differed among time periods (P < 0.01). Litter cover increased after grazing in time period 1 (11.45% ± 2.05% SE), was unchanged after time period 2 (−2.78% ± 1.53%), and was decreased after grazing in time period 3 (−15.28% ± 1.80%; Fig. 4B). Grass residual cover increased with grazing in time period 1 (7.57% ± 1.98%), decreased in time period 2 (−13.85% ± 1.67%), and further decreased in time period 3 (−35.95% ± 3.06%, see Fig. 4A). Shrubs showed no grazing effect in time period 1 (1.31% ± 0.83%) but slightly decreased in time periods 2 and 3 (−1.84% ± 0.83%, −1.53% ± 0.68%; see Fig. 4C). Bare ground cover displayed a treatment-level effect (P < 0.01), where bare ground was decreased 7.29% ± 1.12% with livestock grazing in paddocks treated with protein supplementation but not in paddocks treated with cake (0.17% ± 1.31%; see Fig. S2A). Bare ground cover was also associated with the time period of grazing (P = 0.02), decreasing in time periods 1 and 2 (−6.88% ± 1.70%, −3.42% ± 1.49%) with no effect in time period 3 (−0.89% ± 1.32%; see Fig. S2B). A treatment by time period interaction occurred when evaluating the effects of grazing on visual obstruction (P < 0.04). Visual obstruction was significantly reduced by heifers grazing during time period 1 under both supplementation treatments but only significantly reduced in the cake group during periods 2 and 3 (Fig. 5A).

Supplementation treatment did not significantly alter the differences in transect-level standard deviation of vegetation conditions before and after grazing (P > 0.09; Table 3). However, the time period during which grazing occurred had significant effects on the pre-post transect-level standard deviation of litter, residual grass, and shrub cover (P < 0.01). Litter transect-level standard deviation was increased 3.99% ± 1.19% in time period 1 and was unchanged in time periods 2 and 3 (−0.17% ± 0.97%, 0.15% ± 0.96%; see Fig. 4E). Residual grass pre-post transect-level standard deviation increased 6.67% ± 1.21% in time period 1, was unchanged in time period 2 (−2.19% ± 1.30%), and decreased 9.23% ± 1.14% in time period 3 (see Fig. 4D). Transect-level standard deviation of shrub cover increased 3.22% ± 1.11% in time period 1, decreased 2.87% ± 1.12% in time period 2, and was unchanged in time period 3 (−1.91% ± 1.05%; see Fig. 4F). Visual obstruction transect-level standard deviation displayed a supplementation treatment by time period interaction (P < 0.01), wherein time period 1 both treatments resulted in a net decrease of visual obstruction standard deviation; however, in time period 2 only the cake supplementation treatment resulted in a decrease, with no treatment effects found in time period 3 (see Fig. 5B).

Discussion

We found that supplementation strategy interacted with time period to influence relative use and grazing behavior of heifers during the winter grazing season. The hand-fed cake supplementation strategy in our study resulted in animals selecting for grazing locations near supplement delivery sites for the first two time periods and an overall decrease in the average time spent grazing per day compared with the self-fed protein concentrate treatment. Our results are consistent with previous research demonstrating that supplemented cattle consuming dormant forage graze on average 6–7 h per day (Adams et al., 1986; DelCurto et al., 1990). In addition, our results support previous findings that suggest hand-fed supplementation can disrupt daily grazing activities and/or cause cattle to consume supplement as a substitute to forage, resulting in a decrease in time spent grazing (Adams, 1985).

Our top model evaluating the response of relative grazing use to vegetation characteristics indicates that heifers in both supplementation treatments for all time periods increase relative use with increasing crude protein and standing biomass of perennial grasses. Contrary to our results, previous research suggests cattle typically select for grazing locations higher in forage quality but lower in standing biomass due to the inverse correlation of forage quality and quantity (Wilmshurst et al., 2000; Ganskopp and Bohnert, 2009). Thus, cattle appear to sacrifice short-term intake for the nutritional gains of less abundant but higher-quality forage (Wilmshurst et al., 2000). However, these studies were conducted during the growing season where animals had the ability to select diets from a multitude of plant species at varying stages of phenology. Our study was conducted during the dormant grazing season where most plant species are relatively low in quality. Under relatively uniform foraging conditions, livestock tend to favor grazing locations that maximize intake rate (Black and Kenney, 1984; Laca et al., 1994; Ganskopp and Bohnert, 2009). Therefore, cattle can select grazing locations on the basis of the relative quantity and/or quality of available forage (Senft et al., 1985). Thus, it is likely that the heifers in our study selected for grazing locations with high standing biomass of perennial grasses to maximize intake rate and locations containing plants that maintain quality over the course of the winter (e.g., Krascheninnikovia lanata, common name “winterfat”) but are relatively low in availability.

Most differences in behavioral effects between supplementation treatments tended to diminish in time period 3, which may be due to paddock green up before the end of the grazing period. Paddock green up results in an increase of both forage availability and quality, which has been shown to reduce reliance on supplementation, possibly due to less competition for a limited nutrient supply (Wagnon, 1965; Ducker et al., 1981; Bowman and Sowell, 1997).

Although grazing distribution by cattle has been shown to be constrained by spatial patterns in topography (Gillen et al., 1984; Coughenour, 1991; Van Vuren, 2001; Allred et al., 2011; Kohl et al., 2013) and influenced by extreme weather and temperature conditions (Holechek and Vavra, 1983; Senft and Rittenhouse, 1985; Beaver and Olson, 1997; Parsons et al., 2003), neither slope, aspect, nor temperature were significant drivers in heifer resource use in our study. This may be due to our research paddocks being relatively small ($x¯$ = 65 ha) with little physical constraints due to slope (< 25%) and little variation in temperature within and across paddocks (< 1.5°C). Recent research evaluating cattle grazing distribution on landscapes with subtle topographic variation suggest topographic position indices may be more informative than slope alone (Gersie et al., 2019). Therefore, grazing paddocks that consist of a larger range of available microclimates and using topographic position indices may yield different results.

Individual animal behavior was a major source of variation in this study, displayed as wide confidence intervals for certain variables in the RUF analysis, and the random effect of individual animal explaining the majority of variation in our top model. Previous research on supplementation and grazing behavior suggests that the nutritional status of the animal can have strong influences on grazing behavior of livestock (Krysl and Hess, 1993; Schauer et al., 2005; Bodine and Purvis, 2003) and grazing behavior may vary with individual supplement intake, weight, and body condition in dormant-season grazing systems (Allison, 1985). Thus, within-herd variation of individual-level factors not measured in our study, such as supplement intake, body weight, and condition, may explain some of the variation in individual grazing behavior and space use.

We found that supplementation strategy had little overall effect on vegetation cover and heterogeneity within paddocks grazed at similar moderate stocking rates, with the exception of bare ground cover being reduced in paddocks managed with the protein concentrate supplementation. Rather, the time period at which grazing occurred influenced the effect of winter grazing on residual vegetation and habitat conditions. As expected, early winter grazing (December–mid-January) resulted in a decrease in visual obstruction and bare ground cover and increased litter and residual grass cover and heterogeneity of litter and residual grass cover. At the initiation of the grazing trial each year, there was little snow or environmental factors limiting forage availability. Forage availability can have major effects on grazing behavior as it forms the bounds from which the animal selects its diet; thus, high forage availability allows cattle to graze selectively (Marten, 1989; Reuter and Moffet, 2016). Selective grazing by livestock at moderate stocking rates, as in our study, often promotes within-paddock heterogeneity resulting in paddocks that have areas of light and heavily grazed vegetation (Coughenour, 1991; Bailey et al., 1998; Fuhlendorf and Engle, 2001; Bailey, 2005). Due to cattle having a strong preference for grass, selective grazing may then result in an increase in heterogeneity of residual grass and litter with an overall decrease in visual obstruction and bare ground cover (Vermeire et al., 2004; Augustine and Derner, 2015; Bailey et al., 2015).

Grazing in midwinter (mid-January–February) reduced visual obstruction, grass and shrub cover, and shrub heterogeneity. Heavy snow accumulations may have limited forage availability for prolonged periods of time during both years of the grazing trial (see Table S2). Limited forage availability likely caused animals to consume a greater proportion of the less preferred forage (Marten, 1989) and focus grazing efforts in areas where less snow had accumulated or when the snow melted from the site. Grazing during midwinter may reduce grazing selectivity, resulting in a reduction in visual obstruction and vegetation cover while reducing shrub heterogeneity but having little effect on other vegetation cover conditions within the paddock. Although bare ground cover was marginally reduced by 3.42% ± 1.49% during the midwinter grazing period, it is likely due to sampling error, as other cover classes either decreased or remained unchanged through the grazing time period.

Grazing in midwinter (mid-January–February) reduced visual obstruction, grass and shrub cover, and shrub heterogeneity. Heavy snow accumulations may have limited forage availability for prolonged periods of time during both years of the grazing trial (see Table S2). Limited forage availability likely caused animals to consume a greater proportion of the less preferred forage (Marten, 1989) and focus grazing efforts in areas where less snow had accumulated or when the snow melted from the site. Grazing during midwinter may reduce grazing selectivity, resulting in a reduction in visual obstruction and vegetation cover while reducing shrub heterogeneity but having little effect on other vegetation cover conditions within the paddock. Although bare ground cover was marginally reduced by 3.42% ± 1.49% during the midwinter grazing period, it is likely due to sampling error, as other cover classes either decreased or remained unchanged through the grazing time period.

We observed a decrease in litter, grass and shrub cover, and grass cover variability during late-winter to early-spring grazing periods. Early March received heavy snow accumulations for short durations of time, temporarily limiting forage availability, which may explain the slight decrease in shrub cover, a more substantial decrease in grass cover, and heterogeneity. Grass is the dominant functional group of vegetation for the experimental paddocks, accounting for approximately 85% of the total vegetative cover. Grass cover is a primary driver of grassland bird use and demography, and residual cover and litter may be the only source of available cover and nesting material available at the beginning of the breeding season (Davis, 2005; Fisher and Davis, 2010). Compared with early winter grazing, we observed that cattle grazing during the late-winter to early-spring time period under both protein supplementation strategies in this study reduced average litter cover, as well as spatial variability of residual grass cover. Management strategies that reduce litter and promote homogenous residual grass cover have previously been associated with declines in nesting success of grassland ground nesting birds (Bowman and Harris, 1980), suggesting that the timing, but not the supplementation strategy, of dormant season grazing may have significant impacts on breeding bird habitat conditions.

Implications

We observed high variability in grazing site selection among individual heifers, suggesting individual-level factors could be the dominant drivers in grazing resource use and behavior. Combinations of supplement type, variation in intake, weight and body condition, and winter weather may influence grazing behavior and have significant implications in animal and land management. Monitoring daily grazing behavior without accounting for individual-level factors may not provide meaningful insight about the complex interrelationships that exist between grazing livestock and their environment. Future research examining the effects of supplementation strategies on grazing behavior and resource use should incorporate individual animal measurements in an attempt to account for individual animal variablity. Understanding the effects of supplementation strategies on animal performance, behavior, and paddock use are essential in the development of a cost-effective and sustainable heifer development program.

Our research provides evidence that even in winter dormant forage conditions, heifers appear to still select grazing locations on the basis of the relative quantity and quality of available forage. In addition, supplementation strategy can have an effect on the total time spent grazing per day and grazing resource use during early and midwinter (December–February). However, despite these behavioral differences between supplementation strategies, the time period of when grazing occurred had the largest effects on structural vegetation conditions. The influence of winter grazing on average vegetative conditions, as well as spatial variability (i.e., heterogeneity) appeared to improve in the early winter, with marginal effects in midwinter, and negative effects in late winter to early spring. Thus, grazing in early to midwinter may have marginal effects on vegetation and habitat heterogenity, whereas grazing in late winter to early spring could reduce residual vegetation cover and habitat heterogeneity. Previous studies suggest that a heterogeneous approach to grassland conservation is capable of maintaining biodiversity and agricultural productivity simultaneously (Fuhlendorf et al., 2006). However, local environmental (e.g., edaphic, precipitation) conditions mediate local vegetation and livestock grazing distribution (Pinchak et al., 1991; Smith et al., 1992; Ganskopp and Bohnert, 2009) and their subsquent effects on habitat heterogeneity and grassland wildlife (Lipsey and Naugle, 2017; Vold, 2018). Therefore, future studies evaluating the effects of dormant season grazing management on other grassland ecosystems are recommended.

The following are the supplementary data related to this article.Figure S1Average time spent grazing per day (hrs; ± 95% CI) by A) Supplement treatment and B) time period by heifer's winter grazing rangeland in 2015 – 2016 & 2016 – 2017 at the Fort Keogh Range and Livestock Research Laboratory, Miles City, MT.Figure S1Figure S2Pre-post differences in mean (± 95% CI) bare ground cover classification by A) Supplement treatment and B) time period by heifer's winter grazing rangeland in 2015 – 2016 & 2016 – 2017 at the Fort Keogh Range and Livestock Research Laboratory, Miles City, MT.Figure S2Table S1Average annual production (± SE, kg/ha), Crude Protein (CP ± SE; %), Neutral Detergent Fiber (NDF ± SE; %) and Acid Detergent Fiber (ADF ± SE; %) of the experimental paddocks for the 2 years of grazing (2015 – 2016, 2016 – 2017) at the Fort Keogh Range and Livestock Research Laboratory, Miles City, MTTable S1Table S2Average winter temperature (low, high, mean; °C), total precipitation (cm) and predicted snow fall (based on temperature and precipitation; cm) for the 2 years of grazing (2015 – 2016, 2016 – 2017) at the Fort Keogh Range and Livestock Research Laboratory, Miles City, MTTable S2Table S3Model selection for all models evaluating the effects of vegetation quality and production and supplementation strategy on grazing resource utilization of heifers’ winter grazing rangeland in 2015 – 2016 & 2016 – 2017 at the Fort Keogh Range and Livestock Research Laboratory, Miles City, MTTable S3