OUP user menu

Interactions between Feral Horses and Desert Bighorn Sheep at Water

Stacey Ostermann-Kelm , Edward R. Atwill , Esther S. Rubin , Mark C. Jorgensen , Walter M. Boyce
DOI: http://dx.doi.org/10.1644/07-MAMM-A-075R1.1 459-466 First published online: 18 April 2008


We studied sympatric populations of native bighorn sheep (Ovis canadensis) and feral horses (Equus caballus) to quantify their spatial and temporal overlap and to determine whether horses interfered with use of water by bighorn sheep. We observed no evidence of direct competition, but our field experiment, which involved placing desert-acclimated domestic horses near watering sites used by bighorn sheep, demonstrated that bighorn sheep avoided sites with horses nearby. The presence of domestic horses near a watering site preferred by bighorn sheep resulted in a 76% reduction in the number of groups of bighorn sheep coming to water at that location and a concomitant increase in the number of bighorn sheep watering at other sites. An experimental approach to studying competition between large mammals has been problematic and to our knowledge this study constitutes the 1st manipulative field experiment to test for competitive interactions between feral horses and native ungulates.

Key words
  • bighorn sheep
  • competition
  • Equus caballus
  • feral horses
  • Ovis canadensis
  • Peninsular Ranges
  • resource partitioning
  • wild horses

Desert bighorn sheep (Ovis canadensis nelsoni) and feral horses (Equus caballus) occur together in many arid regions of western North America where food, water, or both may be limited. Although these sympatric associations create the opportunity for behavioral interactions and potential resource competition, Berger (1985, 1986) questioned whether interactions between feral horses and bighorn sheep would result in niche segregation. A few studies have indicated that there is limited overlap in resource use by these species (Anonymous 1997; Beever and Brussard 2000; Bureau of Land Management 1996; Ganskopp and Vavra 1987), but none have examined the question of whether desert bighorn sheep (hereafter bighorn sheep) may be excluded from watering sites by feral horses.

We conducted a study to determine if interference competition for water existed between sympatric populations of feral horses and federally endangered bighorn sheep (United States Fish and Wildlife Service 2000) in an arid, water-limited region of southern California. We 1st assessed resource use by quantifying spatial and temporal overlap between feral horses and bighorn sheep near water, and then evaluated the responses of bighorn sheep to the presence of domestic horses tethered near a key water source.

Materials and Methods

Study area.—Our study was conducted in Coyote Canyon in Anza-Borrego Desert State Park, a 2,429-km2 natural area located in San Diego and Riverside counties in southern California. Research was focused in Coyote Canyon, a watershed that encompasses approximately 405 km2 in the northwestern portion of the park. Approximate coordinates for the center of the study area are latitude 33.438°N, longitude 116.523°W. Coyote Creek traverses the canyon and forms the longest perennial stream in San Diego County, but there are only 3 short segments of the creek where water is present at the surface and available to animals during the summer months. Together, the Lower, Middle, and Upper Willows segments of Coyote Creek form 3 of the most verdant riparian wetlands in desert regions of southeastern California (Warner and Hendrix 1985). Within Coyote Canyon itself, riparian vegetation is found only in areas in close proximity to surface or subsurface water, extending over approximately 49 ha at Lower Willows, 22 ha at Middle Willows, and 16 ha at Upper Willows. Notably, these 3 areas of Coyote Canyon are included in the federally designated critical habitat for the endangered Peninsular bighorn sheep in southern California (United States Fish and Wildlife Service 2001).

Upland vegetation is dominated by chaparral at elevations above approximately 1,500 m and by piny on pine (Pinus monophylla)-]xmvper (Juniperus californica) associations above approximately 1,200 m. At lower elevations, vegetation is dominated by agave (Agave de serti), ocotillo (Fouquieria splendens), cholla (Opuntia), palo verde (Cercidium floridum), creosote (Larrea tridentata), and palo verde-mesquite (Prosopis) associations (Ryan 1968; Sawyer and Keeler-Wolf 1995). Riparian areas are largely dominated by red willow (Salix laevigata), arroyo willow (Salix lasiolepis), cottonwood (Populus fremontii), and desert fan palm (Washingtonia filifera). Perennial shrub species such as mulefat (Baccharis salicifolia), narrow-leaved willow (Salix exigua), and arrow weed (Pluchea sericea) are mixed with willow-dominated vegetation.

The climate in the study area is characterized by hot summers, cool, wet winters, and a bimodal precipitation pattern (Dammitt 2000). Precipitation occurs mainly in November through February and July through September, with most rain falling during the winter. Between 1948 and 2003, annual precipitation averaged 149.1 mm, but was 29.7 mm and 118.6 mm in 2002 and 2003, respectively (Western Regional Climate Center, Borrego Desert Park Station). During the 2003 field season, major flash flooding in Coyote Canyon occurred on 1 and 21 August and substantially increased the water flow in Coyote Creek for several days after both flood events.

Human use of Coyote Canyon is highest at Lower Willows and considerably less at Middle and Upper Willows. Access to Middle and Upper Willows is limited by rough, 4-wheel-drive-only road access. A recreational equestrian trail passes through the canyon, but the entire canyon from Lower Willows through Upper Willows is closed to the public from 30 June to 1 September to minimize disturbance to bighorn sheep using the area for water. Our study was conducted during the summer months of 2002 and 2003, when the canyon was closed to public use.

Study populations.—Bighorn sheep in the Peninsular Ranges of California have been protected under state law since 1971, and were listed as a federally endangered species in 1998 (United States Fish and Wildlife Service 2000). During the period of this study, the estimated number of adult and yearling bighorn sheep in Coyote Canyon was approximately 35 (Torres 2003). This population estimate was obtained by using mark-recapture methodology and resighting radiocollared bighorn sheep from a helicopter (Torres 2003). Predators of bighorn sheep in the study area include mountain lions (Puma concolor), bobcats (Lynx rufus), and coyotes (Canis latrans).

Park records and anecdotal evidence indicate that feral horses were 1st introduced to Coyote Canyon around 1920. Data from periodic ground-based and helicopter surveys for bighorn sheep indicate that the population of feral horses in Coyote Canyon has ranged from 20 to 40 individuals from the 1970s through 2000 (Anza-Borrego Desert State Park, in litt.). Twenty-three feral horses were counted in a census of horses conducted via helicopter in May 2000, and 34 were individually identified during field studies in the summer of 2002. Because of drought and poor forage conditions in fall 2002 and winter 2002–2003, all 29 horses that remained in Coyote Canyon were captured in March 2003 and translocated to a wild horse sanctuary (foals, subadults, and adult females), or a Bureau of Land Management holding facility (adult males).

Evaluation of habitat overlap and interactions.—Ground reconnaissance trips were conducted during June and July 2002 to identify watering sites and to establish belt transects for estimating overlap in habitat use. We recorded surface water and bighorn sheep and horse sign (tracks and feces) via Trimble ProXR and Geo3 global positioning system units (Trimble, Sunnyvale, California), and estimated that there was 765, 490, and 360 linear m of suitable watering habitat for bighorn sheep and feral horses along the Lower, Middle, and Upper Willows areas of Coyote Creek, respectively. Fifty-six 1 x 100-m transects were then positioned along Coyote Creek at Lower Willows (n= 28), Middle Willows (n = 21), and Upper Willows (n = 7). In each of the 3 areas, transects were placed perpendicular to and centered on Coyote Creek at 20-m intervals. We recorded the presence or absence of feral horse and bighorn sheep feces and tracks in each of the one hundred 1-m2 quadrats located along each transect. We excluded from analysis quadrats near the recreational equestrian trail, and where loose sand or extensive disturbance interfered with our ability to clearly distinguish sign.

From July through October 2002, we conducted observations of bighorn sheep and feral horses at Lower, Middle, and Upper Willows from 3 fixed locations for 30 days (10 days per site). Observations were conducted by teams of 2 or 3 people from dawn (approximately 0600 h) to dusk (approximately 1830 h). Ambient temperature was recorded every 30 min. Locations and behaviors of individual feral horses and groups of bighorn sheep were recorded upon initial sighting, and then via scan sampling (Altmann 1974) at 30-min intervals for as long as the animals were in sight. When bighorn sheep were within 100 m of riparian vegetation, their locations and behaviors were recorded at 5-min intervals in order to record their path of movement near water.

We also classified locations of each individual horse and group of bighorn sheep as being within riparian, wash, or upland habitat. We defined riparian habitat to be only those areas within 5 m of water or riparian vegetation (e.g., Salix). Areas near the creek having loose sand and sparse vegetation outside of riparian areas were considered wash habitat, and all other areas were considered upland. Locations of animals were later digitized into Arc View 3.3 (ESRI, Redlands, California). Overlap between bighorn sheep and feral horses was quantified by using Arc View 3.3 to overlay a 50 × 50-m grid over the study area (the viewshed at each riparian area) and counting the number of grid cells containing locations of bighorn sheep or feral horses at each of the Willow areas.

Horse experiment.—We implemented an experiment in summer (16 June-14 September) 2003 using tethered domestic horses to test the null hypothesis that horses had no effect on use of suitable watering areas at Lower Willows by bighorn sheep. This area was chosen because observational data collected in 2002 indicated that Lower Willows area was a highuse watering area for bighorn sheep. Because all feral horses had been removed from the canyon in March 2003, and the canyon was closed to the public, we were able to control for the presence and absence of horses. The experimental design included 10 days of baseline data collection with no horses present, followed by 25 days with horses present, and an additional 20 days with horses absent (nonhorse days). Each of the 25 days with horses (horse days) was randomly assigned a priori, and interspersed among the nonhorse days.

Each day, 1 or 2 observers used binoculars and spotting scopes to detect bighorn sheep coming to water. Upon initial sighting of bighorn sheep, we recorded their location (on a 10-m-resolution aerial photograph map), group size, group composition, and the date, time, ambient temperature, and wind speed. With the aid of a stopwatch timer, tape recorder, and binoculars, we recorded the location of the geographic center of the group at 30-s intervals. Animals separated by more than 100 m were considered separate groups. If multiple groups of bighorn sheep were in view at the same time, scan sampling alternated among groups so that locations were recorded at approximately 60-s intervals. The number and time of bighorn sheep visits to watering sites was recorded through a combination of observations and remote-triggered passive infrared Trailmaster cameras (Goodson and Associates, Inc., Lenexa, Kansas).

During the initial 10-day period of baseline data collection we determined that bighorn sheep drank at 4 discrete locations from Coyote Creek near Lower Willows that we termed watering sites 1–4. We chose to place the horses near site 1 between 1100 and 1500 h because bighorn sheep watered most frequently during this time period and at this location. On the 25 horse days, 2 or 3 desert-acclimated domestic horses were walked to the area and tethered to camouflaged metal posts for a 4-h period (from 1100 to 1500 h). Horses were able to move freely within an approximately 2-m-diameter circle, had constant access to a bucket of water, and were under close observation at all times. A horse handler, who had radio contact with the bighorn sheep observers, was stationed in a shade blind within 100 m of the horses to assist with any horse-related problems and to remove the horses at 1500 h. The horse handler's blind was out of view from the horses and from the bighorn sheep watering sites. The horses were individually tied at posts approximately 25 m apart, within 70 m of the water at site 1 and 150 m from site 2. From site 1, bighorn sheep had unobstructed views of the horses, but the rugged topography prevented bighorn sheep at sites 2, 3, and 4 from being able to see the horses and vice versa. Site 3 consisted of surface water 200–400 m downstream from the horses, whereas site 4 was >500 m away.

We were concerned that the presence of the horse handler and other horse equipment might influence behavior of bighorn sheep. The initial 10 baseline days of the study provided data on bighorn sheep in the absence of horses, equipment, or the horse handler. To test for potential effects of equipment and handler, equipment (shade blind, horse hitching posts, and water buckets) and no handler were placed at the site on 10 of 30 nonhorse days, and equipment and handler were placed at the site on 10 additional nonhorse days. The various treatments used in this experiment are referred to as baseline, equipment only, equipment plus handler, nonhorse (which includes no-horse days, equipment only, and equipment plus handler days), and horse days.

This study was performed in a humane manner and conformed to guidelines sanctioned by the American Society of Mammalogists (Gannon et al. 2007). Animal procedures were approved by and conducted according to University of California Protocol for Animal Use and Care no. 10670.

Statistical analyses.—We compared the number of transect quadrats containing sign from horses or bighorn sheep using Pearson's chi-square tests or Wilcoxon rank-sum tests (Sokal and Rohlf 1995) in S-PLUS 2000 Professional Release 2 (MathSoft, Inc., Cambridge, Massachusetts). We compared the total number of bighorn sheep and horses seen in 2002 at Upper, Lower, and Middle Willows using Pearson's chi-square tests, Fisher's exact test, or Wilcoxon rank-sum tests (Sokal and Rohlf 1995) in S-PLUS 2000 Professional Release 2 (MathSoft, Inc.).

We used Poisson regression or, when necessary, negative binomial regression in STATA 7.0 (Stata Corporation, College Station, Texas) to assess potential changes in watering patterns by bighorn sheep caused by the presence of horse equipment, the handler, or the horses. Poisson regression is an appropriate statistical model for discrete, count-based response variables (e.g., visits of groups of bighorn sheep); negative binomial regression is an alternative to Poisson when the count data exhibit excessive variation around the mean. We began our analysis by testing whether the rate of visitation by bighorn sheep to watering areas differed by treatment condition (baseline, equipment, or equipment plus handler) using mixed-effects negative binomial regression (StataCorp; Stata Corporation, College Station, Texas). In this regression model, site was treated as a random effect, and treatment as a categorical fixed-effect.

We 1st used fixed-effects Poisson regression (StataCorp 2003) to test whether the presence of horses caused a reduction in the number of bighorn sheep watering at Lower Willows per day. Next, we used fixed-effects Poisson regression to determine if the presence of horses was associated with temporal (periods 1, 2, or 3) or spatial (sites 1–4) shifts in the hourly visitation rate by bighorn sheep. For this analysis, we used the cluster option in STATA to group data by watering site (sites 1–4) and by time period (period 1 [0800–1100 h], period 2 [1101–1500 h], and period 3 [1501–1700 h]). Categorical covariates included treatment, whether horses were present the previous day (a lagged effect), whether precipitation occurred within the previous 24 h, and whether precipitation occurred within the previous 3 days. Continuous covariates included maximum air temperature per time period, maximum wind speed per time period, and day of data collection (1–55). A forward-stepping algorithm was used to build the model, with a P-value ≤ 0.10 based on a conditional t-test used as the criterion for inclusion in the model. All biologically plausible 2-way interactions and polynomial continuous variables were screened for inclusion in the model.

Because bighorn sheep are gregarious animals, animals within a group cannot be considered independent. Therefore, we conducted our analysis using the outcome variable “number of groups of sheep” rather than the number of individual bighorn sheep visiting a watering site per time period.


Overlap in habitat use by bighorn sheep and feral horses.— We found that use of the 3 riparian areas by horses differed, as indicated by the presence of horse hoofprints (Pearson's χ2 = 117.6, d.f. = 66, P < 0.001) and horse feces (Pearson's χ2 = 131.1, df.= 36, P < 0.001; Table 1). Transects at Upper and Middle Willows contained more horse hoofprints than transects at Lower Willows (Z = −5.29, P < 0.001; Z = −7.58, P < 0.001, respectively). Horse hoofprints were found in 42%, 36%, and 5% of the quadrats examined at Upper, Middle, and Lower Willows, respectively, whereas horse feces were found in 34%, 6%, and 1% of the quadrats examined at Upper, Middle, and Lower Willows, respectively. Horse feces were significantly more common at Upper Willows than at Middle Willows (Z = −4.29, P < 0.001), and at Middle Willows than at Lower Willows (Z = −5.19, P < 0.001).

View this table:
Table 1

Percentage (number) of 1-m2 quadrats containing feral horse or bighorn sheep sign (hoofprints or feces) in transects at Lower (LW), Middle (MW), and Upper Willows (UW) in Coyote Canyon, Anza-Borrego Desert State Park, California, during summer 2002.

Quadrats examinedQuadrats with horse printsQuadrats with horse fecesQuadrats with sheep printsQuadrats with sheep fecesQuadrats containing both sheep and horse sign
LW28004.6 (130)1.0 (28)3.6 (102)3.3 (93)0.1 (4)
MW210036.1 (758)6.0 (125)21.1 (444)7.8 (163)10.9 (228)
uw70041.7 (292)33.9 (237)2.1 (15)1.7 (12)1.9 (13)

Tracks and fecal pellets from bighorn sheep were significantly more common at Middle Willows (21% and 8% of quadrats) than at Lower Willows (4% and 3% of quadrats; Z = −5.90, P < 0.001; Z = −2.75, P < 0.001, respectively) or Upper Willows (2% and 2% of quadrats; Z = 3.96, P < 0.001; Z = 2.75, P < 0.001, respectively). We found no difference in the number of fecal pellet groups from bighorn sheep at Lower Willows and Upper Willows. Quadrats containing both horse and bighorn sheep hoofprints or feces or both were found at all 3 riparian areas (Table 1). Evidence of overlap between the 2 species was more prevalent at Middle Willows (11%; Pearson's X2 = 69.0, d.f. = 26, P < 0.001), than at both Upper and Lower Willows, where <2% of the quadrats contained sign of both species.

Observational data revealed that overlap in use of habitats near surface water by feral horses and bighorn sheep was highest at Middle and Upper Willows. Horses were detected in 24% of 50 × 50-m grid cells at Middle Willows and 27% of grid cells in the Upper Willows areas where bighorn sheep were also observed. Overlap in use of grid cells by the 2 species was much less (3%) in the Lower Willows area.

During the 30 observation days in 2002, we recorded 34 groups of bighorn sheep, for which we recorded 342 locations and behavioral observations. At least 1 group of bighorn sheep was observed on 19 of 30 days (Appendix I). We documented 25 (74%) of the 34 groups of bighorn sheep at water between 0800 and 1700 h. Bighorn sheep were seen more frequently at Middle Willows than Lower Willows (Z = −2.63, P= 0.009) or Upper Willows (Z = −2.71, P = 0.007). During the 30 observations days, we observed 31 groups of horses over 19 days (Appendix I) and recorded 1,468 locations of horses and behavioral observations. A larger number of individual horses was seen at Middle Willows than at Lower Willows (Z = −2.31, P = 0.02) or Upper Willows (Z = −2.06, P= 0.04), and at Upper Willows than at Lower Willows (Z = −3.33, P < 0.001). Horses commonly used riparian areas. We recorded horses in wash or riparian areas on 588 occasions between 0600 and 2000 h, and for an average of 4.9, 2.3, and 0.3 h per day at Upper, Middle, and Lower Willows, respectively.

We saw both bighorn sheep and feral horses on 11 days, but both species occurred simultaneously on only 4 occasions (Appendix I). Feral horses and bighorn sheep were never observed using water at the same time, nor were bighorn sheep ever observed to water when horses were in wash or riparian habitat within sight of the watering source. We did not detect evidence of direct interference competition (i.e., aggression) between feral horses and bighorn sheep, and therefore behavioral data are not presented. We observed 4 situations where horses and bighorn sheep were obviously in view of each other. In 2 cases, the horses did not appear to be aware of the bighorn sheep and the bighorn sheep were displaced (flushed) by the horses. In 1 instance, 2 females and a male walked within 35 m of 4 browsing horses in order to join another group of bighorn sheep nearby. In the 2nd instance, 2 females that were within 300 m of a group of horses fled a short distance uphill (<100 m) to rejoin a group of 14 other bighorn sheep and then continued browsing while the 12 horses came to water approximately 200 m below the bighorn sheep.

Horse experiment.—During 55 days of observation, we documented a total of 372 bighorn sheep coming to water in 82 groups between 0800 and 1700 h. Most of these bighorn sheep (318 bighorns in 64 groups) watered between 1100 and 1500 h, the period during which horses were positioned near site 1 (Table 2). On 4 occasions we documented bighorn sheep running in apparent response to movement (e.g., shaking or turning their head or body) of the domestic horses tied near site 1.

View this table:
Table 2

Percentage (number) of bighorn sheep groups observed at each watering site when tethered horses were present or absent near site 1. Data are from observations made between 1100 and 1500 h at Lower Willows of Coyote Canyon, Anza-Borrego Desert State Park, California, during summer 2003.

Watering siteHorses absent near site 1 (n = 30)aHorses present near site 1 (n = 25)a
Site 153 (18)3(1)
Site 26(2)24 (7)
Site 36(2)13(4)
Site 435 (12)60 (18)
  • a Number of days of observation.

Our Poisson regression models showed no effect of equipment or equipment plus the handler (P > 0.10) on the rate of bighorn sheep visits to water. Therefore, we pooled the 30 nonhorse days for comparisons with the 25 horse days. We found that the presence of horses near site 1 did not result in an overall reduction in the number of groups of bighorn sheep watering at Lower Willows that day. However, our final Poisson regression model (Table 3) indicated that when horses were present, there was a significant decrease in the rate of watering by bighorn sheep at site 1 and a corresponding significant increase in the rate of watering by bighorn sheep at sites 2–4, and particularly at site 4 (horses present × site 4, P= 0.005; Fig. 1). When horses were not present, there was no difference in watering rates by bighorn sheep at sites 1 and 4 (site 4, P = 0.27; Table 3). Coefficients for interactions between horses and time period were not significant, indicating there was no evidence of temporal partitioning of watering sites. Compared to referent conditions (i.e., site 1, during period 1, with no horses), significantly fewer bighorn sheep watered at site 2 (P= 0.007) and at site 3 (P = 0.01). More bighorn sheep watered during period 2 than during period 1 (P < 0.001; Table 3).

View this table:
Table 3

Estimated maximum-likelihood coefficients for a Poisson regression model fitted to data on rates of watering (hourly visits to water with and without horses present) by bighorn sheep at 4 different sites in Coyote Canyon, Anza-Borrego Desert State Park, California, during summer 2003. The referent condition to which all other factors are compared is horses absent at site 1 during time period 1 (0800-1100 h).a

ParameterCoefficient95% CI for variable xP -valueIncident rate ratio (IRR)IRR 95% CI
Present−1.44−2.47, −0.400.0070.240.08, 0.67
Watering site
Site lb0
Site 2−1.50−2.59, −0.420.0070.220.08, 0.66
Site 3−1.28−2.27, −, 0.75
Site 4−0.41−1.14, 0.320.270.670.32, 1.38
Time period
Time period 1 (0800-1100 h)b0
Time period 2 (1101-1500 h)2.071.01, 3.13<().()() 17.942.74, 23.0
Time period 3 (1501-1700 h)−1.22−3.43, 0.990.280.290.03, 2.7
Maximum air temperature (°F)0.04−0.003,, 1.08
Day0.07−0.007,, 1.16
Day2−0.002−0.003, −0.00030.020.9980.996, 0.999
Horse × watering site
Horses present × site lb0
Horses present × site 21.970.42, 3.530.017.21.5, 34.1
Horses present × site 31.620.10,, 23.0
Horses present × site 41.740.51, 2.970.0055.71.7, 19.4
Constant−8.52−12.5, −4.54<0.001
  • a Incidence rate ratios (IRRs) are the ratios of the rate of sheep visits per hour for the indicated factor, over the rate for the referent condition (no horses, site 1, period 1). For example, the IRR for horses present is 0.24, meaning that all other things being equal, having horses present results in a significant reduction to just 24% of the hourly visit rate at site 1 compared to when horses were absent (or a 76% reduction in the watering rate). For continuous variables the interpretation is for a single unit change; for example, as maximum air temperature increases by 1°F, the sheep visitation rate at site 1 increased by about 4%. 95% CI − 95% confidence interval.

  • b Referent condition for Poisson regression model.

Maximum air temperature per time period and day of the study influenced the rate of bighorn sheep watering (Table 3). Incident rate ratios indicated that for each degree Fahrenheit the temperature increased, the visitation rates by bighorn sheep increased 4%. The negative and positive coefficients for the variables day and day2, respectively, indicate that the rate of bighorn sheep visits to water by bighorn sheep initially increased as the study proceeded, but then leveled and eventually declined toward the end of the study. The covariates maximum wind speed per time period, precipitation in the previous 1 or 3 days, and presence of horses the previous day were not associated with spatial or temporal partitioning of watering sites, nor the overall rate of visits by bighorn sheep to Lower Willows.


In contrast to the study by Berger (1986) and other studies (Anonymous 1997; Beever and Brussard 2000; Bureau of Land Management 1996; Ganskopp and Vavra 1987), we found clear spatial overlap in habitats used by desert bighorn sheep and feral horses, particularly near water. Data from field observations and transect surveys showed that feral horses and bighorn sheep both used the same watering sources and occasionally interacted with each other near water (Table 1). Where use by feral horses was highest in terms of hours of use per day (Upper Willows), use by bighorn sheep was lowest.

Our finding that feral horses frequently spent several hours per day in or near riparian areas corroborates the findings of Crane et al. (1997) and Menard et al. (2002), who reported that feral horses preferred streamside or marsh-bog-meadow habitats during warm seasons. In contrast, Ganskopp and Vavra (1987) reported that feral horses in Oregon typically moved rapidly to and from water, with few feeding or loafing activities in the immediate vicinity of drinking areas. Ganskopp and Vavra (1987) suggested their finding may be explained by the plentiful water sources in their study area.

Although feral horses and bighorn sheep used all 3 Willow areas, both species rarely used these areas at the same time. On the 4 occasions when both species were present simultaneously, we saw no evidence of aggressive interactions and therefore we concluded that feral horses did not directly interfere or compete with bighorn sheep at water. This conclusion also was supported by the observations of Berger (1986) and Ganskopp and Vavra (1987).

Our field experiment demonstrated that bighorn sheep avoided water sources when horses were present nearby (Table 2; Fig. 1). Controls built into the experiment showed that this response was due to the presence of horses and was not influenced by the presence of horse handlers or equipment. This spatial partitioning of water resources is evidence of indirect interference competition, and warrants further investigation.

Fig. 1

Comparison of watering rates by bighorn sheep (Ovis canadensis; groups of sheep per hour) per watering site when tethered horses (Equus caballus) were a) absent and b) present at site 1. Curves were plotted on the fitted relationship (Poisson regression model; Table 3) between watering rate and temperature (°F) during time period 2 (midday, 1100–1500 h) at Lower Willows in Coyote Canyon, Anza-Borrego Desert State Park, California, during summer 2003.

Temperature and day of the study also significantly influenced watering rates of bighorn sheep. The significant negative polynomial variable for day suggests that visitation rates by bighorn sheep initially increased, but then gradually decreased during the study period. This decrease may reflect the heavy rain and flash flooding that occurred within the study area beginning on day 28 of the 55-day study. The variables related to rain (precipitation within 24 h and precipitation within the last 3 days) were not significantly related to use of watering sites by bighorn sheep. Categorical variables for rain 1 and 3 days previously may not have been sufficiently sensitive to indicate when rain was an important influence on watering by bighorn sheep. A continuous variable that quantified precipitation would have more accurately accounted for any variation in watering by bighorn sheep due to rain; however, these data were not available. Rainfall in the desert is highly variable and patchy, making accurate quantification of rain in the study area difficult. The variable day2 appeared to best account for the effect of rainfall.

Temporal partitioning of water resources by populations of desert bighorn sheep in Nevada and Arizona has been documented in response to human disturbance (Campbell and Remington 1981; Leslie and Douglas 1980). Perhaps we detected only spatial partitioning because there was an abundance of nearby watering sites available where horses were not present. In our study area, spatial partitioning was apparently less costly in terms of energy and predation risk than temporal resource partitioning. The avoidance behavior we documented likely underestimated the potential impact of feral horses because the 2 or 3 horses in our experiment were restrained and band sizes of feral horses in Coyote Canyon ranged from 4 to 16 horses. Larger numbers of free-ranging horses in and near water sources would be expected to have a greater effect on behavior of bighorn sheep.

An experimental approach to studying competition between large mammals has been problematic (Stewart et al. 2002) and to our knowledge our study constitutes the 1st manipulative field experiment to study potential competitive interactions between feral horses and bighorn sheep. Our study clearly showed that bighorn sheep avoided watering sites when horses were nearby, which constitutes evidence of indirect interference competition (Diamond 1978; Loft et al. 1991). Inference from our study would have been strengthened by having replicates of the experiment involving tethering domestic horses near watering sites used by bighorn sheep. Although further research is clearly needed, our study demonstrates that the presence of horses has the potential to negatively impact bighorn sheep by causing them to avoid watering sites during the hot summer months.


We thank L. Hendrickson, K. Keim, and S. Martin for assistance with fieldwork and logistics. S. Tedeschi and C. Bohn assisted with data collection and entry. We thank P. Jorgensen, J. Dice, and M. Fuzie for their assistance with funding, equipment, and logistical support. L. Jee provided assistance with global positioning system units and the geographic information system. Domestic horses were provided and handled by C. Green. G. Lee provided assistance with logistics, data entry, and editing. Funding was provided by Anza-Borrego Desert State Park and the Wildlife Health Center at the University of California, Davis.

Appendix I

View this table:
Appendix I

Numbers of feral horses and bighorn sheep groups observed at Lower, Middle, and Upper Willows in Coyote Canyon, Anza-Borrego Desert State Park, California; summer 2002. Numbers in parentheses indicate the number of groups seen. NR = not recorded.

DateDay no.High temperatureLow temperatureNo. horsesNo. sheep
Lower Willows
23 July11017607(1)
24 July2102815(1)4(2)
19 September3957300
21 September41007907(1)
23 September510681014 (3)
24 September61068908(3)
26 September71058906(1)
27 September894745(1)0
1 October9806300
2 October10795800
2/10 days6/10 days
Middle Willows
30 July1NRNR17 (2)18(1)
11 August21028609(1)
12 August3103NR5(1)15 (3)
13 Augusta4103845 (1)13 (2)
14 Augusta5102835 (1)21 (4)
19 August6928913 (1)0
20 Augusta7967416 (2)16 (2)
21 August89268025 (2)
22 August99868016 (2)
26 August10104885(1)5 (1)
7/10 days9/10 days
Upper Willows
5 Julya19486514 (1)
6 July29582130
17 July398809(2)12 (1)
18 July4977611 (3)9(1)
4 September5100808(3)0
5 September689817(1)8(1)
9 September7906711 (3)0
10 September889728(2)0
11 September990716(2)0
12 September1092728(2)0
10/10 days4/10 days
  • a Days on which horses and sheep were in view at the same time.


  • Associate Editor was Rick A. Sweitzer.

Literature Cited

View Abstract