Key Words: Cervus elaphus, competition, distribution, habitat selection, Odocoileus hemionus, resource selection

Authors are wildlife biologist, Ore. Dep. of Fish and Wildl., 1401 Gekeler Lane, La Grande, Ore. 97850, USA; wildlife biologist, Ore. Dep. of Fish and Wildl., 1401 Gekeler Lane, La Grande, Ore. 97850, USA; statistician, Western Ecosystems Technologies, Inc., 2003 Capitol Way, Cheyenne, Wyo. 83001, USA; wildlife biologist, Ore. Dep. of Fish and Wildl., 1401 Gekeler Lane, La Grande, Ore. 97850, USA; wildlife biologist, USDA For. Serv., 1401 Gekeler Lane, La Grande, Ore. 97850, USA; wildlife biologist, USDA For. Serv., 1401 Gekeler Lane, La Grande, Ore. 97850, USA.. We thank J. H. Noyes and T. DelCurto and B. Dick for their comments. Our research was funded under provisions of the Federal Aid in Wildl. Restoration Act (Pittman-Robertson Act), Ore. Dep. of Fish and Wildl., and USDA For. Serv., Pac. Northw. Res. Sta. and Pac. Northw. Region. Our research was approved by the Institutional Animal Care and Use Committee for Starkey and conducted in accordance with approved animal welfare protocols (Wisdom et al. 1993). D. Leckenby, J. W. Thomas, and M. Vavra were instrumental in establishing this research.

Manuscript accepted: April 15, 2000

Cattle graze seasonally on national forests in the Western United States, and mule deer (Odocoileus hemionus) and/or elk (Cervus elaphus) are sympatric with cattle in most of these areas. But the effects of interspecific interactions in terms of both the resources selected and animal distributions across landscapes are poorly understood. At the USDA Forest Service Starkey Experimental Forest and Range (Starkey), located in northeast Oregon, USA, elk and mule deer were free ranging within a 78 km² study area enclosed by a 2.4 m high fence while cattle were moved among pastures in summer on a deferred-rotation schedule. Elk, mule deer, and cattle were located with an automated telemetry system from 1993 - 1996 and locations were linked to a geographic information system (GIS) of Starkey. Our objective was to examine responses of elk and mule deer to cattle at several spatial and temporal levels. We compared elk and mule deer distributions, use of plant communities, and resource selection functions in one cattle pasture (24 km²) during early summer (cattle present in odd-numbered years) and late summer (cattle present in even-numbered years). Elk and deer differed in their spatial and temporal responses to presence of cattle. When cattle were present, the proportion of elk locations within the pasture decreased and use of the ponderosa pine/Douglas fir (Pinus ponderosa/Pseudotsuga menziesii) plant community within the pasture decreased in early summer and increased in late summer. The cattle resource selection function variable for early summer was not a predictor of elk distributions when cattle were present, but it was a predictor on years when cattle were absent. In late summer, the cattle resource selection function variable was a predictor of elk distributions regardless of presence of cattle. For mule deer distributions the cattle resource selection function variable was not a significant predictor in early summer (cattle present or absent), or in late summer when cattle were present, but it was a negative predictor of mule deer distributions when cattle were absent in late summer. Mule deer use increased or decreased in opposite direction from elk use in 3 of 4 season/year combinations for both pasture and ponderosa pine/Douglas fir. Our results suggest that competition for forage could occur between elk and cattle in late summer and that species interactions may be stronger between elk and cattle than mule deer and cattle.

Allocating forage and maintaining proper stocking levels of domestic and wild ungulates are fundamental concepts of range management (Stoddart and Smith 1955). Managing rangelands for cattle, elk (Cervus elaphus) and mule deer (Odocoileus hemionus) in montane habitats is complicated because interspecific interactions vary from potential competition (Skovlin et al. 1968, Lindzey et al. 1997, Mackie 1970, Dusek 1975, Knowles and Campbell 1982, Nelson 1982, Austin and Urness 1986, Loft et al. 1991, Wallace and Krausman 1987, Peek and Krausman 1996) to commensalism (Anderson and Scherzinger 1975, Frisina and Morin 1991, Peek and Krausman 1996). Further complicating our understanding of the interactions among these ungulates is that the response of elk or mule deer to cattle may vary seasonally depending on forage availability and quantity (Peek and Krausman 1996, Wisdom and Thomas 1996). Competition occurs when individuals or species use the same resources that are in short supply. Short supply of forage may decrease nutritional planes such that population performance of 1 or more species decreases (Birch 1957, Putnam 1996). In contrast, commensalism occurs when an action of 1 species benefits a second species. Clear understanding of distributions and resource selection of elk and mule deer with and without cattle present is necessary to identify possible interspecific relationships among cattle, elk, and mule deer to better manage the resources.

Elk and mule deer may respond to cattle grazing at several levels or scales. When cattle are present, elk and mule deer may leave a pasture, shift distributions within pastures but not change in how they select resources, stay within a pasture and select resources differently, or not respond in any measurable manner. Indeed, most studies of the interactions of either elk or mule deer with cattle describe changes in distribution when cattle are introduced into pastures (Skovlin et al. 1968, Knowles and Campbell 1982, Wallace and Krausman 1987, Kie et al. 1991, Loft et al. 1991, Clegg et al. 1994), but few studies have contrasted patterns of habitat selection of elk and mule deer with and without cattle present. Our objectives were to examine responses of elk and mule deer to cattle at several spatial and seasonal scales. We tested if elk and mule deer (1) left the pasture, (2) shifted distributions among habitats within a pasture, and (3) changed resource selection when cattle were present compared to when cattle were absent in early summer and late summer.

Study Area

Starkey Experimental Forest and Range (Starkey) is located on the Wallowa-Whitman National Forest, 35 km southwest of La Grande, Ore (45 deg 15’ N, 118 deg 25’ W). An ungulate-proof fence (2.4 m high) of New Zealand woven wire surrounds Starkey enclosing the 101 km² area (Bryant et al. 1993). Ungulate-proof fences also divide Starkey into 3 study areas, and we conducted our study within Main Area (78 km²) (fig. 1). Main Area is divided by barbed-wire fences into several cattle pastures (hereafter "Main Area" refers to the area available to elk and mule deer, while the term "pasture" refers to one of the cattle pastures within Main Area). Cattle grazed Smith-Bally, Halfmoon, and Bear pastures annually. Plant communities characterizing the vegetation include bunchgrass (Festuca idahoensis Elmer, Poa sandbergii Vasey, Agropyron spicatum (Pursh) Scribn. & Smith), ponderosa pine (Pinus ponderosa Dougl.), Douglas fir (Pseudotsuga mensiesii (Mirbel) Franco), grand fir (Abies grandis (Dougl.) Forbes), and lodgepole pine (Pinus contorta Dougl.) (fig. 2). Elevations range between 1100 and 1500 m and from 1993 through 1996 annual precipitation averaged 636 mm. Skovlin (1991), Noyes et al. (1996), and Rowland et al. (1997) provide detailed descriptions of Starkey.

Fig. 1. Main Area at Starkey Exp. Forest and Range, northeast Oregon is summer range for about 300 elk and 200 mule deer. Solid lines depict livestock fences. Pastures are grazed by 500 cow-calf pairs of cattle on a deferred rotation schedule. Cattle grazed sequentially through Smith-Bally (1), Half moon (2), and Bear (3) on odd-numbered years from 15 June to 15 September and through pastures 3, 2, and 1 on even-numbered years from 15 July to 15 October. Meadow Creek pasture was not grazed except for the 3 pastures on the extreme east edge for 20 days each. Livestock did not graze Horse pasture.

Elk and mule deer ranged freely throughout Main Area from spring through fall. In winter a portion of the animals were trapped and fed on a feedground where animals were sampled and radio-collars were attached and maintained. Elk and mule deer were also caught in portable traps in winter in Main Area and released after radios were attached. Annual point estimates of population size for elk ranged from 311 to 367 female elk (Noyes et al. 1996, Johnson et al. 2000) and 175 – 218 female mule deer prior to parturition from 1993 – 1996 (B. K. Johnson, unpublished data). Elk and mule deer were exposed to similar kinds and extents of human activities from spring through fall as found on adjacent national forests. Elk and mule deer populations were hunted in late summer or fall, and roads were open to the public from 1 May to mid-December each year. Between mid-December and the end of April each year the Experimental Forest was closed to the public.

From 15 June to 15 October each year, 500 cow-calf pairs were grazed in Main Area on a deferred rotation system. Cattle grazed Smith-Bally pasture from 15 June – 1 August on odd numbered years, and then moved on a set schedule to Halfmoon, and Bear pastures, and left Main Area about 15 September. On even-numbered years, cattle began grazing Bear pasture around 15 July and were then moved through Main Area pastures in reverse order, grazing Smith-Bally pasture last from approximately 1 September – 15 October. Cattle grazed Bear pasture for approximately 6 weeks and Halfmoon pasture for 1 week in this rotation schedule, and the dates of cattle grazing in Bear pasture overlapped between even and odd years.

We limited our analysis of resource selection and plant community uses to Smith-Bally pasture, because cattle were either present in the pasture or concentrated in the southeast portion of, or absent from, Main Area.

Habitat Variables

We developed a geographical information system (GIS) for Starkey (Rowland et al. 1998), and identified 26 variables from the literature (Reynolds 1962, Julander and Jeffery 1964, Cook 1966, Mackie 1970, Perry and Overly 1977, Lyon 1979, Thomas et al. 1979, Lyon 1983, Hershey and Leege 1982, Thomas and Toweill 1982, Irwin and Peek 1983, Senft et al. 1983, Skovlin et al. 1968, Edge et al. 1987, Pedersen et al. 1980, Collins and Urness 1983, Leckenby 1984, Bracken and Musser 1993, Wisdom 1998) that were correlated with elk, mule deer, and cattle distributions. We tested for collinearity (PROC CORR; SAS Institute 1987) among these variables. When 2 variables were correlated (r > 0.37), we deleted the variable we judged to be more difficult to measure accurately or interpret resulting in 15 variables to measure habitat in Smith-Bally pasture (Table 1). Rowland et al. (1998) and Johnson et al. (2000) provide detailed descriptions of derivations of these variables and their representation in the GIS habitat database. Figures 2 and 3 show habitat features from which most variables were derived.

Table 1. Variables used in logistic regression to estimate resource selection of cattle, elk, and mule deer in early and late summer in Smith-Bally pasture at Starkey Exp. Forest and Range, northeast Ore. 1993 - 1996.

Grouping animal locations—We grouped daily totals of elk and mule deer locations that occurred in Smith-Bally, Bear, and Meadow Creek/Horse pastures temporally into Interval, Period and Rotation (Fig. 4.). We identified 2 grazing intervals when cattle were moved among pastures at Starkey, adjusting the dates to be comparable to alternate years. We defined early summer (Interval A) as 20 June – 6 July and late summer (Interval B) as 5 September – 12 October. Early summer dates encompass a shorter interval than the actual cattle grazing dates to match the dates that cattle entered Main Area on alternate years. "Period" identified a day as to whether it was within the first 5 days (0), last 5 days (1), or another day (2) of the interval. "Rotation" denoted whether a year was even or odd.

Fig. 4. Daily elk and mule deer locations were grouped and analyzed temporally by Interval (grazing season), Period (first 5 days, last 5 days), and Rotation (odd or even year) for Smith-Bally pasture and plant community comparisons within Smith-Bally pasture. Resource selection models for elk, mule deer, and cattle were calculated by Interval and Rotation within Smith-Bally pasture, Starkey Exp. Forest and Range, northeast Ore.

To test for elk and mule deer responses to cattle presence within Smith-Bally pasture, we selected the same dates from alternate years when cattle were not in the pasture. We compared locations from 1993 and 1995 with locations from 1994 and 1996. We summarized elk and mule deer use of Bear and Meadow Creek/Horse pastures during these intervals to provide information on general distributions throughout Main Area.

Shifts in pasture use—To determine if elk and mule deer left the pasture when cattle were introduced, we calculated the number of elk and mule deer locations within the pasture and the total number of elk and mule deer locations in Main Area each day. We modeled daily elk numbers as a temporally correlated Poisson process using the GLIMMIX macro in SAS (Wolfinger and O’Connell 1993). Elk and mule deer counts were modeled using a Poisson distribution to account for the discrete nature of count data, and temporal auto-correlation was modeled using an autoregressive correlation model. The autoregressive parameter [Ar(1)] represents the correlation between consecutive days. We estimated separate autocorrelation coefficients for each year/ treatment combination. In general, Ar(1) ranges between -1.0 and 1.0 with large positive values reducing the effective degrees of freedom for estimation. Consequently, hypothesis tests were adjusted to account for autocorrelation among days. When autocorrelation was positive, effective degrees of freedom and consequently the power of statistical tests were reduced.

To test for cattle effects, we compared the first 5 day period with the last 5 day period (Period) when cattle were present and when cattle were absent (Rotation) within early summer and late summer (Interval). An interaction between Period and Rotation within an Interval was interpreted as an effect due to cattle. For example, if elk counts for the first five day period were similar to elk counts in the last 5 day period when cattle were absent, but elk numbers were disimilar between the first 5 day period and the last 5 day period when cattle were present, this effect was interpreted as an interaction between Period and Rotation (Period*Rotation). Count of locations in the pasture day ^-1 was the response variable. To account for variations in daily counts of locations in Main Area each day, total count of locations from Main Area day ^-1 was included as an offset term in the model. To investigate the effects of this reduction in degrees of freedom, we also conducted analyses under the assumption of independence of days (PROC GENMOD; SAS Institute 1987) and often effects which were significant under the assumption of independence were not detected when we accounted for autocorrelation. In some cases the GLIMMIX macro did not converge (a unique optimal set of parameters was not found). In those cases we assumed that the percentage of elk in each pasture was approximately normally distributed and analyzed the data as a repeated measures design with autoregressive error structure using PROC MIXED.

Shifts in plant community use—To determine if elk or mule deer changed their use of plant communities in response to cattle presence, we used locations within a plant community as the response variable and the total number of locations within the pasture as the offset.

We estimated 2 resource selection functions for cattle (early summer and late summer) and 4 resource selection functions for both elk and mule deer (early summer and late summer, with and without cattle). To examine differences between models within an interval for elk and mule deer, we tested the interaction of Rotation with each variable that was significant in either of the initial models for each species. We used cattle locations obtained within 4 hours after sunrise or 4 hours before sunset when foraging is highest for cattle and all locations obtained for elk and mule deer. We used the peak foraging period for cattle to describe where cattle foraged and where potential interactions between elk and mule deer may be highest. We used all locations for elk and mule deer because if we used only locations from peak foraging periods (i.e., within 2 hours of sunrise or sunset), we would not have a sufficient number of locations from enough animals to estimate resource selection. We used locations from 16 mule deer, 35 elk, and 52 cattle in odd years and 21 mule deer, 45 elk, and 48 cattle in even years, and we had a minimum of 32 locations from each animal in an interval. Leban (1999) recommended using a minimum of 20 animals and 50 locations per animal when estimating resource selection in Starkey.

To evaluate if elk or mule deer were selecting resources similar to cattle, we incorporated the cattle resource selection function score (cattle RSF) from the appropriate Interval as a variable in the logistic regression for the elk or mule deer. We tested for collinearity of the cattle variables with the 15 variables used in the logistic regression and deleted variables that were correlated (r > 0.37) with the cattle or elk resource selection function prior to the analysis.

Shifts among pastures

The proportion of elk locations declined in both early summer (43%) and late summer (21%) when cattle were present, compared to when cattle were absent (6% early, 33% increase late) (Table 2, Fig. 5a., b.). The proportion of elk locations increased in Bear and Meadow/Horse pastures during the same intervals that they were decreasing in Smith-Bally pasture (Table 3).

Table 2. Elk and mule deer use of Smith-Bally pasture and ponderosa pine/Douglas fir (PP/DF), lodgepole pine (LP), grand fir (GF), and grassland, shrubland, and meadow (BG/MD) plant communities in response to cattle grazing rotation. Mean proportions (Mean) and standard errors (SE) are of daily elk or mule deer observations in pasture or plant community during the first 5 days (0) and last 5 days (1) of the interval. Pasture proportions are for Main Area and plant community proportions are for within Smith-Bally pasture. Cattle grazed Smith-Bally pasture in early summer in 1993 and 1995 (Odd) and in late summer in 1994 and 1996 (Even). Significance levels were determined from PROC MIXED and adjusted for autocorrelation among days within a 5 day period within an interval within a year.

Fig. 5. Proportion of elk locations in Smith-Bally pasture and in the ponderosa pine/Douglas fir plant community during early summer and late summer, when cattle were present (open circles, n = 20 days) compared to when cattle were absent (closed circles, n = 20 days), Starkey Exp. Forest and Range, northeast Ore., 1993 - 1996.

The proportion of mule deer locations within the pasture increased 35% in early summer, but decreased 23% in late summer when cattle were present, compared to when cattle were absent (2% increase early, 8% increase late)(Table 2, Fig. 6a., b.). The proportion of mule deer locations in Bear and Meadow/Horse pastures concurrently stayed level or changed in opposite directions to mule deer proportions in Smith-Bally pasture (Table 3).

Table 3. Elk and mule deer use of Bear and Meadow Creek pastures in response to cattle grazing rotation in Smith-Bally pasture. Mean proportions (Mean) and standard errors (SE) are of daily elk or mule deer observations in pasture during the first 5 days (0) and last 5 days (1) of the interval. Cattle grazed Smith-Bally pasture in early summer in 1993 and 1995 (Odd) and in late summer in 1994 and 1996 (Even). Significance levels were determined from PROC MIXED and adjusted for autocorrelation among days within a 5 day period within an interval within a year.

Fig. 6. Proportion of mule deer locations in Smith-Bally pasture and in the ponderosa pine/Douglas fir plant community during early summer and late summer, when cattle were present (open circles, n = 20 days) compared to when cattle were absent (closed circles, n = 20 days), Starkey Exp. Forest and Range, northeast Ore., 1993 - 1996.

Shifts among plant communities

Only in the ponderosa pine/Douglas fir plant community were consistent changes evident in amount of use during intervals for elk and mule deer. However, mule deer response was generally opposite from that of elk (Table 2). In early summer, the mean daily proportion of elk locations in the ponderosa pine/Douglas fir community decreased 81% when cattle were present but increased 21% when cattle were absent (Table 2, Fig. 5c.). In late summer the proportion of elk locations in the ponderosa pine/Douglas fir community increased 36% when cattle were present and decreased 44% when cattle were absent (Table 2, Fig. 5d.). In early summer the proportion of mule deer locations increased 33% when cattle were present and increased 54% when cattle were absent. In late summer the proportion of mule deer locations decreased 33% when cattle were present and increased 133% when cattle were absent (Table 2, Fig 6c.,d.). In all other plant communities, significant shifts in use were not evident. Proportion of cattle use in the 4 plant communities declined in the bunchgrass community and increased in lodgepole pine and grand fir communities from early summer to late summer (Table 4).

Table 4. Cattle locations (%) within 4 plant communities in Smith-Bally pasture in early and late summer, 1993-1996, Starkey Exp. Forest and Range, northeast Ore. Locations were obtained within 4 hours after sunrise or 4 hours before sunset. Cattle grazed this pasture in early summer in 1993 and 1995 and late summer in 1994 and 1996.

Resource selection models

Table 5. Variables in resource selection functions of cattle in early summer and late summer in Smith-Bally pasture, using locations obtained within 4 hours after sunrise or 4 hours before sunset with a LORAN-C automated telemetry system, Starkey Exp. Forest and Range, northeast Ore. 1993 – 1996.