https://cpw.cvlcollections.org/files/original/1f144bd9ae38b593c8ab19fb7837a34d.pdf 666a8fbc23bd511b5dd761cd879f8f6b PDF Text Text Dan Prenzlow, Director, Colorado Parks and Wildlife • Parks and Wildlife Commission: Marvin McDaniel, Chair • Carrie Besnette Hauser, Vice-Chair Marie Haskett, Secretary • Taishya Adams • Betsy Blecha • Charles Garcia • Dallas May • Duke Phillips, IV • Luke B. Schafer • James Jay Tutchton • Eden Vardy �Estimating Cougar Predation Rates from GPS Location Clusters Author(s): Charles R. Anderson, Jr. and Frederick G. Lindzey Source: The Journal of Wildlife Management , Apr., 2003, Vol. 67, No. 2 (Apr., 2003), pp. 307-316 Published by: Wiley on behalf of the Wildlife Society Stable URL: https://www.jstor.org/stable/3802772 JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms Wiley and Wildlife Society are collaborating with JSTOR to digitize, preserve and extend access to The Journal of Wildlife Management This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �ESTIMATING COUGAR PREDATION RATES FROM GPS LOCATION CLUSTERS CHARLES R. ANDERSON, JR.,1 Zoology and Physiology Department, University of Wyoming, Box 3166, Unive Laramie, WY 82071, USA FREDERICK G. LINDZEY, U.S. Geological Survey, Wyoming Cooperative Fish and Wildlife Research Unit, Box 316 Station, Laramie, WY 82071, USA Abstract: We examined cougar (Puma concolor) predation from Global Positioning System (GPS) loca (?2 locations within 200 m on the same or consecutive nights) of 11 cougars during September-May Location success of GPS averaged 2.4-5.0 of 6 location attempts/night/cougar. We surveyed potent sites during summer-fall 2000 and summer 2001 to identify prey composition (n = 74; 3-388 days po and record predation-site variables (n = 97; 3-270 days post predation). We developed a model to esti bility that a cougar killed a large mammal from data collected at GPS location clusters where the proba dation increased with number of nights (defined as locations at 2200, 0200, or 0500 hr) of cougar pr a 200-m radius (P< 0.001). Mean estimated cougar predation rates for large mammals were 7.3 days/kill females (1-2.5 yr; n = 3, 90% CI: 6.3 to 9.9), 7.0 days/kill for adult females (n = 2, 90% CI: 5.8 to 10.8), 5 for family groups (females with young; n = 3, 90% CI: 4.5 to 8.4), 9.5 days/kill for a subadult male (1-2 90% CI: 6.9 to 16.4), and 7.8 days/kill for adult males (n = 2, 90% CI: 6.8 to 10.7). We may have slightly cougar predation rates due to our inability to separate scavenging from predation. We detected 45 deer spp.), 15 elk (Cervus elaphus), 6 pronghorn (Antilocapra americana), 2 livestock, 1 moose (Alces alces mammals at cougar predation sites. Comparisons between cougar sexes suggested that females select and males selected elk (P < 0.001). Cougars averaged 3.0 nights on pronghorn carcasses, 3.4 nights o casses, and 6.0 nights on elk carcasses. Most cougar predation (81.7%) occurred between 1901-0500 hr from 2201-0200 hr (31.7%). Applying GPS technology to identify predation rates and prey selection wil agers to efficiently estimate the ability of an area's prey base to sustain or be affected by cougar preda JOURNAL OF WILDLIFE MANAGEMENT 67(2):307-316 Key words: cougar, Global Positioning System, GPS, predation model, predation rate, prey composition, color, Wyoming. 1981, Cunningham et al. 1999) can be important The influences of large carnivores on prey populations and ecosystem processes hinge on in others. preOnly 2 studies found that cougars dation rates that may vary with prey species selected com- 1 ungulate species over another, and in studies, elk were taken over mule deer (0. position, by sex and age of prey, and byboth sex and hemionus;, Hornocker 1970, Murphy 1998). Within age composition of the predator population. Cougar predation rates have been estimated a prey bypopulation, however, young and/or solisnowtracking (Connolly 1949), energeticstary models animals appear most vulnerable to cougar (Hornocker 1970, Ackerman et al. 1986), predation and via (moose: Ross and Jalkotzy 1996; elk: intensive radiotelemetry (Shaw 1977, Harrison Hornocker 1970, Murphy 1998, Nowak 1999; fera 1990, Beier et al. 1995, Murphy 1998, horses Nowak[Equus caballus]: Turner et al. 1992; bighorn sheep: Ross et al. 1997; mule deer: Hornock1999). Cougar predation studies generally have relied on small samples of cougars, because er 1970, of Shaw 1977, Logan and Sweanor 2001) the difficulty of snow- or radiotracking Ackerman these et al's. (1986) predictions based on wide-ranging, solitary animals in rugged terrain. energetic needs for various sex, age, and repro Although deer dominate cougar diets in ductive most classes of cougars suggested increased areas (Spalding and Lesowski 1971, Ackerman predation et rates with cougar body mass and repro al. 1984, Logan and Sweanor 2001), elk (Hornockductive status and were supported by Murphy er 1970, Williams et al. 1995, Murphy 1998,(1998), Nowakalthough the diets in these studies were 1999), bighorn sheep (Ovis canadensis, Williams primarily et mule deer and elk, respectively. Wildlife studies that have utilized GPS technolal. 1995, Wehausen 1996, Ross et al. 1997), moose have primarily addressed GPS collar perfor(Ross and Jalkotzy 1996), and livestock ogy (Shaw mance, reporting large data-storage capacity and high accuracy (Rempel et al. 1995, Moen et al. 1996, Bowman et al. 2000). Application of GPS 1 E-mail: cander@uwyo.edu 307 This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �308 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey J. Wildl. Manage. 67(2):2003 technology to animal movement investigations provides a potential tool for studying predation rates and identifying prey-selection patterns. Our objectives were to determine whether cougar kills of large mammals could be detected from data beaver (Castor canadensis) and yellow-bellied mar- mot (Marmota flaviventris). Other carnivores include coyote (Canis latrans), pine marten (Martes americana), bobcat (Lynx rufus), and black bear (Ursus americanus). records of store-on-board GPS collars and, if METHODS detectable, to develop a model to estimate probWe trailed cougars using hounds and immobiability of a predation event based on characteristics of GPS data records. STUDY AREA lized them upon capture with a mixture of 5 mg/kg Telazol? (Aveco Company, Inc., Cherry Hill, New Jersey, USA) and 1 mg/kg xylazine hydrochloride The Snowy Range, located in southeast delivered in hypodermic dart fired from a CO2 Cougars (>1 yr old and self sufficient) were Wyoming, USA, about 30 km west of Laramie, pistol. is collared with GPS receivers (GPS2000; Lotek, a 2,120 km2 portion of the Medicine Bow NationInc., Newmarket, Ontario, Canada). We proal Forest surrounded by private, Bureau of Land grammed collars to attempt position acquisitions Management, and state-owned lands. The range at 1600, 1900, 2200, 0200, 0500, and 0800 hr each is bounded by state highway 230 on the west, US Interstate 80 on the north, the Laramie River andday, targeting the nocturnal period when cougars Sand Creek drainages on the east, and the are most likely pursuing prey or feeding (AckerWyoming-Colorado border on the south. Eleva- man 1982, Hopkins 1989, Beier et al. 1995). Age tion ranges from about 2,100 m in the valleys to (subadult [1-2.5 yr] or adult [23 yr]) was esti3,652 m at Medicine Bow Peak. Vegetation com- mated from tooth wear, canine ridge eruption, munities are dominated by sagebrush (Artemisia spotting progression, and evidence of lactation tridentata) grasslands in the peripheral valleys, for females (Shaw 1979, Ashman et al. 1983, lodgepole pine (Pinus contorta) with interspersed Lindzey et al. 1989). We estimated age of depenquaking aspen (Populus tremuloides), Rocky Moun- dent cougars (young of collared females) by tain juniper (Juniperus scopulorum), and limber backdating to the time of localized denning activpine (Pinusflexilis) at mid-elevations, and Engel- ity of their mother. mann spruce (Picea engelmannii) /subalpine fir (Abies lasiocarpa) forests with occasional limber Predation Site Surveys pine (Alexander et al. 1986) at higher elevations. We downloaded data from collars upon death Understory dominants in the mid- and high-eleva- of the cougar or when we recaptured the animals tion communities include huckleberry (Vaccinium to remove collars (Apr-May 2000, 2001). Locascoparium), buffaloberry (Shepherdia canadensis), tion data were plotted in ArcView? (Environserviceberry (Amelanchier alnifolia), snowberry mental Systems Research Institute 1999) and (Symphoricarpos spp.), and common juniper (J. sequentially inspected with the Tracking Analyst communis). Riparian areas are composed primarily extension to identify location clusters. We initialof willow (Salix spp.) interspersed with narrowleaf ly defined location clusters as 22 locations within cottonwood (P angustifolia) at low elevations. 200 m during the same or consecutive 16-hr periMule deer and elk are the most common ungu- ods (i.e., 1600-0800 hr). We examined X2 goodlates in the Snowy Range. Post-hunting season com- ness-of-fit to assess potential bias (P < 0.05) in position counts during winter 1999-2000 were time of successful GPS location attempts for each 492 male, 1,704 female, and 1,161 young mule cougar. After we identified potential predation deer, and 416 male, 1,166 female, and 635 young sites (location clusters) from GPS collar records, elk (Wyoming Game and Fish Department 1999). we located them on the ground using a handPronghorn and white-tailed deer (0. virginianus) held GPS receiver and searched for prey remains are abundant in adjacent sagebrush-grassland by walking 5-10-m-wide strip-transects within the habitats and lower-elevation riparian areas, area described by the outer point radius plus 10 respectively. Bighorn sheep and moose are pre-m. The U.S. Department of Defense discontinsent in low numbers. Small mammals common in ued selective availability in spring 2000, between the period when most cougars were monitored the Snowy Range include porcupine (Erethizon dorsatum), pine squirrel (Tamiasciurus hudsoniand when potential predation sites were searched. cus), cottontail (Sylvilagus nuttallii), and snowThus, error associated with cougar locations before spring 2000 was about 43 m (Moen et al. shoe hare (Lepus americanus) and occasionally This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �J. Wildl. Manage. 67(2):2003 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey 309 1996) and that associated with searches was <5 m (Bowman et al. 2000). Variables recorded at To evaluate the model, we applied it to the GPS location record of a cougar (subadult male) not potential predation sites included carcass presincluded in the original analyses and searched ence or absence, portions of carcasses found, sites that the model identified as potential predation sites (i.e., locations at 2200, 0200, or 0500 prey species sex/age (young-of-year/adult), cougar sign, and distance from the cluster center hr). We later included data from this cougar in the overall data set and refit the model to inpoint to the carcass. We used exterior condition crease sample size. of bones (e.g., oily, dry, chalky), presence and condition of bone marrow, hair at cache sites, We summed the predation probabilities associat and vegetation growth under and around bonesed with each cluster to estimate total number of to determine whether remains were of an age predation events for each cougar and estimated consistent with the suspected predation event.individual predation rates by dividing total numPotential predation sites to be searched were ini-ber of days monitored by estimated number of tially selected to maximize sample size by choos-predation events. To construct 90% confidence ing clusters close to one another. Other sites wereintervals around individual cougar predation searched as time permitted. We used the longestrates, we used the model coefficient standard time period between the potential predationerror to obtain a 90% confidence interval for each event and when a carcass was found at a 3-nightestimated predation event, and summed the lower cluster (high probability of predation) to defineand upper bounds to get a confidence interval for the period when large-mammal carcass detectiontotal number of predation events. We then divided may become inconsistent. the total monitoring period by the lower bound estimate and upper bound estimate to get a prePredation Model Development dation-rate confidence interval for each cougar. We used logistic regression analyses (Hosmer We averaged the coefficients of variation across and Lemeshow 1989, SAS Institute 1990, Mehta individuals to obtain predation-rate confidence and Patel 1993) to estimate the probability of aintervals for sex/age and reproductive classes. large-mammal predation event (carcass presence was coded as 1 and absence as 0) as a function ofPrey Composition and Predation Characteristics number of locations in a cluster, number of nights at a cluster (locations from 1600 hr 1 day to 0800 We identified species and sex-age class of prey hr the next), presence/absence of daytime loca-carcasses from hair or skeletal remains following tions distant from the cluster (i.e., daybed),Moore (1974) and Adrian (1996). Remains of sex/age and reproductive class of the cougar, andadults that could not be sexed in the field were all combinations of consecutive, nocturnal loca- sexed via gender polymerase chain reaction tions. We hypothesized that nocturnal presenceamplification of the zinc finger motif found on at location clusters represented cougar predation the x gene (ZFX) and the sex determining region and thus, the latter predictor variable was used to found on the Y gene (SRY), where males exhibit 2 evaluate which combination of nocturnal locabands (ZFX = 425 base pairs, SRY = 225 base pairs) females exhibit 1 band (SRY = 225 base pairs; tions best predicted cougar predationand when examining number of locations and number of Game and Fish Laboratory, Laramie, Wyoming nights at a cluster. We based model selection Wyoming, on USA). We examined diets of male and strength of variable significance (Wald female X2, P < cougars for differences (P< 0.05) using X2 analyses (StatXact-Turbo; Mehta 0.05). We used forward stepwise logistic contingency regresandavoid Patel 1992). We tested for differences in sion for initial model development. To number of nights cougars spent on carcasses of potential inadequacies of stepwise analysis (James and McCulloch 1990), we compared the elk stepwise and mule deer and elk and pronghorn using model to various models containing stepwise 1-tailed varit-tests for unequal variance (P < 0.05); ables selected and each variable not included dur- data were not included if the collar ceased col- ing stepwise analysis (included individually). lecting All 2- GPS data while the cougar was at a kill. examined data for patterns that would sugg way interactions also were investigated. We used were more successful at killing prey the Hosmer-Lemeshow goodness-of-fit testcougars (Hosmer and Lemeshow 1989) to assess adequatecertain fit of times of night by comparing the tim our data to the logistic regression model, cougars where were first located at clusters with expe significant results (P < 0.05) suggest lack-of-fit. ed times using X2 goodness-of-fit analysis. This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �310 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey J. Wildl. Manage. 67(2):2003 ?38 weeks (270 days) after 3 or more nights of cougar presence. We used 66 sites in model development (removing those searched after 38 weeks) 1900 hr) associated with a kill (i.e., daybeds). including GPS data from 2 subadult females (n = Daybeds were defined as locations at 0800, 1600, or 1900 hr that were outside GPS location clusters 21 sites), 2 females with young (n = 19), 1 adult female (n = 14), and I adult male (n = 12). Prey and from which the cougar returned to the conremains were found at 46 of these sites (32 deer, firmed and predicted predation site. Predicted Using ArcView?, we examined distribution of daytime and early evening locations (0800, 1600, predation sites were defined by the predation8 elk, 3 pronghorn, 2 livestock, I porcupine). Because we suspected that small-prey (<15 kg) model as location clusters with probability of predetection was inconsistent at GPS location clusdation >0.5. We included early evening locations ters, and our objective was to identify large-mam(1900 hr) because subsequent analysis suggested mal kills, we considered the single porcupine kill that cougars were not often associated with preas a nonpredation event for model development. dation sites at this time. Differences in daybed Ninety-one percent of location clusters used in distances outside location clusters (P < 0.05) between confirmed and predicted predation sitesmodel development contained 3-dimensional were examined using a 2-tailed t-test for equal GPS positions (signals from ?4 satellites yielded x, variance. For sites with multiple daybeds (1/day)y coordinates and elevation). Mean distance from outside a single cluster, we used the mean dis- kills to cluster centers was 42.6 m (range = 0.0-106.5). Mean number of times a cougar was tance to maintain independence. located at the 45 sites with large mammal kills was RESULTS 9.3 (range = 2-28), and mean number of subseWe collared 11 cougars (2 adult males, 4 adult quent nights spent at the kill was 3.5 (range = females with large young [3 litters 4-8 months 1-8). At clusters where we did not find large prey old, 1 litter 14-17 months old], 2 adult females remains (n = 21), cougars were located an averwithout young, and 3 subadult females) withage GPS of 3.3 times (range = 2-5) and remained 1.9 receivers between September 1999 and January nights (range = 1-6). 2000. Weight of GPS collars was 1.6-2.2% of Cougar Predation Models cougar body mass. We recovered 6 GPS collars between November 1999 and March 2000 via har- Logistic regression analyses suggested that the vest and 4 others via capture between April and number of nights at a cluster best predicted a May 2000; the remaining collar failed just prior to large-mammal predation event, where nights at a retrieval (adult female with 5-8 month-old cluster was defined as cougar presence at 2200, young). We observed no injuries caused by the 0200, or 0500 hr. Cluster duration exhibited the GPS collars, and movement patterns of each strongest relationship (X - = 12.71, P < 0.001) of cougar were comparable to those previously all ob-univariate models and was the only significant tained from VHF radiocollars (C. R. Anderson, variable (P < 0.05) included among multivariate unpublished data). Monitoring periods averaged models compared (Fig. 1). 78 nights (n = 10, range = 28-188), and mean When we applied the initial model (Fig. 1) to number of locations/night ranged from 2.4the to data collected from a single cougar (subadult 5.0/cougar. Time of successful GPS location male) monitored December 2000-May 2001, it attempts did not differ from equality among predicted 38 potential predation sites (locations cougars (X2 < 8.31, P 0.140), suggesting that at suc2200, 0200, or 0500 hr) from GPS records. We cessful locations were unbiased for time of locasearched 36 of the 38 sites for evidence of kills, tion attempt. Nights without a GPS location but precluded 5 sites because they were too occurred only 12 times (0-3/cougar, total monidensely vegetated to facilitate discovery of prey toring = 784 nights for all 10 cougars). We identiremains. We detected 9 large (5 elk, 3 mule deer, fied 188 potential kill sites from GPS data records I moose) and 4 small-mammal predation events and searched 94 sites an average of 201 days after (3 porcupine, 1 coyote) at the 31 sites searched. the potential predation event (mode = 105, range Number of predation events estimated by the model from the GPS records for this subadult = 3-388). We found prey remains at 61 sites. Prey remains were not found at 5 of 17 sites with 3-7 male was 13.4, an overestimate of large prey consecutive nights of cougar presence that werekilled (13.4 vs. 9). searched >38 weeks (273 days) after clustering. Data from this subadult male were then added However, we found remains at all 24 sites searchedto the original data set, and the model was refit to This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �J. Wildl. Manage. 67(2):2003 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey 311 both models (Fig. 1). One notable difference o 0.9 however, was the predicted probability of preda tion with 2 nights recorded at the same location The initial model (Fig. 1, solid line) suggested a predation probability of 0.94 with 2 nights of 0.7 8 0.5 - cougar presence recorded, whereas the latte model (Fig. 1, dashed line) predicted a 0.77 chance of a predation event with 2 nights o 0.3 0.2 0 0 1 2 3 4 cougar presence. 5 Number (GPS of nights Cougar Predation Rates at locations at GP 2200, Monitoring periods for individual cougars Fig. ters 1. Probability ranged 28-188 days (X = 84.0; Table 3). Estimatedo where p = expu large-mammal predation events predation data ranged 3.8-22.8/ from Wyoming, USA, cougar (X = 12.1). Mean estimated large-mammal S 3.9922 (number of predation rate for all cougars was 7.0 days/kill the Hosmer-Lemes (90% CI: 5.8 to 10.4), but ranged from 5.4 days data fit the logisti represents cougar (90% CI: 4.5 to 8.4) for family groups to 9.5 days lected in southeast (90% CI: 6.9 to 16.4) for the subadult male. Per31 observations collected from a single cougar, Dec 2000-May 2001 where u = -3.8467 + 2.5287(number of cent of successful GPS location attempts ranged appeared appropriate for the logistic regression model (Hosmer-Lemeshow 2 = 6.86, P= 0.144). Number of nights from GPS locations at 2200, 0200, or 0500 hr. Prey Composition nights), SEpo = 0.8040, and SE1 = 0.5305; these data 39.3-82.7% (X = 60.5) for individual cougars. We detected prey remains (9 species) at 74 clusters from 11 cougars (Table 1). This included all GPS location clusters searched during initial model development (Fig. 1, solid line; n = 66), those the combined data. Location clusters where small mammals were detected (n = 4) were treated as added for subsequent model development (Fig. 1, dashed line; n = 31), and those searched but nonpredation sites (i.e., coded as 0). Refitting the model with the 31 additional clusters from not included in model development (>38 weeks the subadult male resulted in a similar model. post predation; n = 28). Cougars spent longer periods on elk carcasses (k = 6.0 nights, SD = 3.7, Again, number of nights at GPS location clusters, defined as presence at 2200, 0200, or 0500 nhr, was = 14) than deer (? = 3.4 nights, SD = 1.8, n= 43; the most significant predictor variable = P = 0.010) and pronghorn (X = 3.0 tl5 (X2 = 2.60, 22.72, P< 0.001) among all univariate models and nights, SD = 1.3, n = 6; t18 = 2.70, P = 0.007). cougars killed more mule deer than the only significant variable (P < 0.05) Female included other species, and male cougars killed more elk in multivariate models compared (Fig. 1). other species (X= 20.61, P <0.001; Table 1). Predicted predation probabilities with than number Males killed of nights at GPS location clusters were similar for proportionately more adult male elk Table 1. Prey species (n, [%]) detected at GPS location clusters of cougars by sex, age, and reproductive class in the Snowy Range, southeast Wyoming, USA, Sep-May, 1999-2001. Subadult female Adult female Family group Subadult male Adult male Prey (n = 19) (n = 11) (n = 17) (n = 14) (n = 14) Deera 13 (68) 8 (73) 17 (100) 3 (21) 4 (29) Elk 0 3 (27) 0 5 (36) 7 (50) Pronghorn 4 (21) 0 0 0 2 (14) Porcupine 0 Livestockc Moose Coyote 0 0 0 2 0 4 (29)b (11) 0 0 0 0 0 1 1 a Includes 44 mule deer and 1 white-tailed deer. b Two of 4 porcupines were detected at the same location. C Includes 1 domestic sheep and 1 domestic calf. This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms 0 1 0 (7) 0 (7) 0 (7) 0 �312 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey J. Wildl. Manage. 67(2):2003 Table 2. Sex-age composition (%) of large prey'detected at GPS location clusters (n = 69) of male (n = 3) and female (n = 8) cougars in the Snowy Range, southeast Wyoming, USA, Sep-May, 1999-2001. Deera Elk Pronghorn Total no. Cougar sex of prey Males Females Young Unkb Males Females Young Unkb Males Females Young Other Male 22 Female Total 5 47 69 9 5 19 14 25 a Includes C Moose d e Includes Includes b Unk = 14 32 27 21 16 44 9 9d 10 14 2 10 mule unknown 2 5 2 4 6 5 0 0 4 1 deer 5 2 4 1 5c 2 4e 3 and 4 1 whit calf. 2 1 adults of domestic unknown sex sheep (femal daybeds to confirmed and probable predaand females from killed proporti female muletiondeer (Table 2). sites did not differ (t63 = -0.25, P= 0.803) and 2 sex-age classes averaged 844 m (SD killed = 612, n = 65, range pron = items detected at GPS location clusters included 90-3380). Multiple daybeds associated with a sina moose calf, 2 livestock (1 Ovis aries, 1 Bosgle taukill occurred at 31 of 65 sites (x = 2.68/site, rus), and 6 small mammals (5 porcupines, SD 1 coy= 1.11, range = 2-6); cougar use of the same ote; Table 1). bed site occurred at only 3 of 31 sites. We pooled data from confirmed (n = 40) and probable predation sites (cougar presence 22 DISCUSSION nights, predation probability 2 0.77, n - 42) Cougar Predation Models because arrival time at clusters was similar (X2 = 3.70, P= 0.594; Fig. 2). Initial locations (presumed Cougar predation rates estimated by the mode time period of kill) were not random across generally time agree with rates estimated from snow tracking, radiotracking, and energetics model intervals (X2 - 37.95, P < 0.001), but did not differ among sex-age and reproductive classes when (Y2 <size of primary prey is considered (Tabl 8.31, P 2 0.140). Kills sharply increased4). from A lower rate estimated by Shaw (1977) may have resulted from undetected kills when a 1901-2200 hr, peaked at 2201-0200 hr, and gradually declined until 0801-1600 hr (Fig. 2). cougar stayed only 1 night. Both Beier et (1995) and our study documented kills where We detected daybeds outside the predation cluster at 28 of 69 (40.6%) confirmed and cougar 37 of was present for a single night. Predat 64 (57.8%) probable predation sites. Distances rates estimated by Hornocker (1970) also ma have been underestimated because the cou energy requirements were derived from capt 0.35 cougars. Why Connolly's (1949; 9.7 days/k 0.3 0.25 S0.2 So.15 S0.1 0.05 1601-1900 1901-2200 2201-0200 0201-0500 0501-0800 0801-1600 Time interval estimate, derived from snowtracking, reflecte wider interval than ours (7 days/kill) is uncle Estimates of predation rate will reflect the and age of cougars in the sample, however. I cougars tracked by Connally (1949) were prim ily subadults or females without young, over predation rate likely would be lower than the e mate derived from our more-inclusive sample Because we used store-on-board GPS collars, the period between recognizing a potential preevent (after the collar was retrieved and Fig. 2. Frequency distribution of time intervals cougars were dation first located at confirmed (n = 40) and probable (n = 42) predation sites combined identified from GPS location clusters data downloaded and analyzed) and searching collected Sep-May, 1999-2001, in the Snowy Range, souththe site was often long, decreasing the chance of east Wyoming, USA. Probable predation sites consist of 22 separating scavenging from predation. Scavengnights of cougar presence (presence at 2200, 0200, or 0500 hr) where cougar predation probability 20.77. ing, documented but infrequent in cougars (Ack- This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �J. Wildl. Manage. 67(2):2003 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey 313 Table 3. Length of monitoring period, percent GPS locations acquired (attempted at 1600, 1900, 2200, 0200, 0500, 0800 hr each day), number of estimated large mammal kills from number of nights (locations at 2200, 0200, or 0500 hr) at GPS location clusters (n = 226), and estimated cougar predation rates of large mammals (days/kill) from 11 cougars of 5 sex-age and reproductive classes Sep-May, 1999-2001 in the Snowy Range, southeast Wyoming, USA. Cougar class Monitoring period % GPS locations No. of estimated Estimated predation Cougar ID (days) acquired large mammal killsa rate (90% CI)b Subadult female 619 64 81.3 10.9 5.9 (4.9 to 8.1) 620 95 61.9 14.1 6.7 (5.7 to 9.4) 632 32 39.3 3.4 9.4 (8.5 to 12.1) Mean 63.7 60.8 9.5 7.3 (6.3 to 9.9) Adult female 628 86 645 67.1 42 Mean 10.4 42.2 64.0 7.4 54.7 8.3 5.7 8.9 (6.9 to (4.7 7.0 (5.8 11.7) to 9.5) to 10.8) Family group 623C 607d 618e Mean 87 112 39.7 82.1 14.7 19.9 5.9 5.6 (5.0 (4.8 to to 9.6) 8.3) 50 82.7 10.9 4.6 (3.7 to 7.3) 83.0 68.2 15.2 5.4 (4.5 to 8.4 Subadult male 635 141 44.0 14.9 9.5 (6.9 to 16.4) Adult male 626 188 59.4 22.8 8.2 (7.0 to 11.3 644 28 66.1 3.8 7.4 (6.5 to 10.1) Mean 108.0 62.8 13.3 7.8 (6.8 to 10.7) Grand mean 84.0 60.5 12.1 7.0 (5.8 to 10.4) a Estimated number of large mammal kills = sum of expu and u = -3.8467 + 2.5287 (number of nights). N b Predation rate = length of monitoring c Litter size = 2, age = 4-7 months. d Litter size = 2, age = 4-8 months. e Litter size = 2, age = 15-17 months. period / esti and food habits and predation rates of this2001 erman et al. 1984, Logan and Sweanor sex-age class may differ from others introdu (Murphy phy 1998, Nowak 1999), may have Model predictions were similar to detectunknow, but likely1998). small amount of bias i ed predation, but only when 1 coyote and 3 porpredation rate estimates. We felt comfort cupine kills were included. The model washad develseparating bones from animals that di oped only from winter large mammals, however, and it tho ing the previous fall and from overestimated kills forto this cougar had died earlier and beenlarge-mammal exposed hot a (13 estimated 9 detected). Inclusion of GPS summer conditions. Theandlikelihood of fin of this subadult male with kills of small bones of an animalrecords that had died during or winter of other causes at a location cluster mammals coded as no-kill into the original data likely improves the model by making it more where a cougar coincidentally exhibitedset sedenapplicable to all cougar sex-age and reproductive tary behavior for reasons other than foraging The model will profit from additional (e.g., illness) seems small. Additionally,classes. skeletal data sets remains were found at only 1 of 11 clusters con-to capture variation within and between cougar sex-age and reproductive classes. sisting of primarily daytime locations. Applying GPS collars that can be downloaded on demand Global positioning system locations appeared accurate (which were not available during our study) to enough that circumference of clusters, identify cougar predation should furtherbuffered reduce by 10 m, enclosed predation sites. Global Positioning System error should have been these potential problems. minimal Application of the initial model to predict pre- because all but 9 clusters without prey dation events in a subadult male cougarremains identi- were based on 3-D locations (Moen et al. 1997, fied potential problems. The initial model didBowman et al. 2000). Additionally, accuracy not include GPS records from subadult males, of 2-D locations is similar to 3-D locations if the This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �314 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey J. Wildl. Manage. 67(2):2003 Table 4. Cougar predation rates (days/kill) of ungulates from North American cougar studies. Location Predation rate Cougar sex-agea Primary preyb Technique Source Central Utah 9.7 US MD Snowtracking Connolly (1949) Central Idaho 18-26 AD MD Energetics model Hornocker (1970) 4.5c FG MD Snowtracking West central Arizona 6.8 FG MD Radiotracking Shaw (1977) 10.4 AF MD Southern Utah 8.5 AM MD Energetics model Ackerman et al. (1986) 16.1 AF MD 4.8d FG MD 4.5e FG MD Radiotracking Central British Columbia 2.7-6.4 FG BS, MD Radiotracking Harrison (1990) Southern California 7.6 US MD Radiotracking Beier et al. (1995) Northwest Wyoming 7.5 M Elk Radiotracking Murphy (1998) 11.1 AF Elk, MD 7.2 FG Elk, MD 11.0 SM Elk, MD 10.3 SF MD, Elk Northeast Oregon 7.7 UF MD, Elk Radiotracking Nowak (1999) Southeast Wyoming 7.0 US MD, Elk GPS location clusters This study 7.8 AM Elk, MD 7.0 AF MD, Elk 5.4' FG MD 9.5 SM Elk, MD 7.3 SF MD, PH a AD = unspecified adult, FG = family group, A male, SF = subadult female, UF = adult females w b Primary prey listed if >20% of ungulate diet. P PH = pronghorn. C Observed from snowtracking a female with 3 3 d Estimated from Ackerman et al. (1986:Fig. 2) fo e Observed from radiotracking females with 3 6f Estimated from 3 litters of 2 young each. Ag elevation from the tendency last for adult 3-D males to position kill elk and femaleis that of the 2-D (Moen al. 1997). Fur cougars to kill et mule deer. the mean distance from carcasses to cluster cen- Prey Composition ters (43 m) matched the mean GPS error reported by Moen et al. (1996) for an earlier version of Cougars in the Snowy Range may be partitionthe GPS collar we used. We believe it is unlikely ing prey as suggested by Murphy (1998). Subthat the 9 2-D clusters biased the model because adult female cougars on the Snowy Range killed 6 of them represented only a single night of mule deer and pronghorn, the subadult mostly cougar presence and were probably not largemale killed mostly elk and small mammals, adult mammal kills. females killed predominantly mule deer and elk, females with young preyed on mule deer, and Predation Rates adult males killed more elk than other sex-age The lowest predation rates estimated were from classes (Table 1). Although others have suggested subadults (1 male, 1 female) supporting Murphy's that cougars kill more young and male mule deer (1998) prediction that predation rates should be (Hornocker 1970, Shaw 1977, Murphy 1998), the related to age and experience of cougars. Two sex and age of mule deer killed by female cougars other subadult females, however, killed prey as study area appeared similar to estimated in our frequently as adult females. Predation ratescomposition also in the mule deer population varied between the 2 adult females without young, (Wyoming Game and Fish Department 1999). but this may have been due to 1 being in theOur latersmall sample size may have reduced our stages of pregnancy when captured. The similarichances of detecting selection. Dispersion of colty of adult male predation rates and those ofcougars in relation to nonuniform disperlared subadult and adult females likely reflected theof deer and elk over the range may have sion This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �J. Wildl. Manage. 67(2):2003 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey 315 resulted in the apparent selection of species we observed. Wyoming, for gender assays of cougar prey remains. We thank P. Beier and D. Maher for suggestions on improving the manuscript. Capture MANAGEMENT IMPLICATIONS protocols were reviewed and approved under the Modeling predation rates from GPS location University of Wyoming Animal Care and Use form number A-3216-01. records will significantly reduce the cost Committee and effort of estimating this parameter by precluding LITERATURE CITED the need for intensive ground or aerial monitor- ACKERMAN, B. B. 1982. Cougar predation and ecological ing. In a relatively simple system with a single energetics in southern Utah. Thesis, Utah State Unilarge ungulate prey species, predation rates versity, Logan, Utah, USA. could be estimated simply from GPS records with -, F. G. LINDZEY, AND T. P. HEMKER. 1984. Cougar no need to investigate kill sites. In more complex food habitats in southern Utah. Journal of Wildlife systems, where cougars can switch among prey Management 48:147-155. species, kill-site investigations may be required. I, _ , AND - . 1986. Predictive energetics model of cougars. Pages 333-352 in S. D. Miller and In either case, identifying predation rates and D. Everetts, editors. Proceedings of the International prey selection will allow managers to estimate the Cat Symposium, Kingsville, Texas, USA. effect of cougar predation on the area's preyADRIAN, popW. J., editor. 1996. Wildlife forensic field manual. Second edition. Colorado Division of Wildlife, ulation (s) if prey abundance can be estimated. Fort Collins, Colorado, USA. The model we constructed to estimate predaALEXANDER, R. R., G. R. HOFFMAN, AND J. M. WIRSING. tion events from GPS records of cougar move1986. Forest vegetation of the Medicine Bow Nationments may improve as more GPS records with al Forest in southeastern Wyoming: a habitat type verified kills by various sex and age cougars classification. are U.S. Forest Service Research Paper RMadded. As we collected 125 observations in 46 271. U.S. Forest Service, Rocky Mountain Range Experiment Station, Fort Collins, Colorado, USA. days, additional records could be generated ASHMAN, D. L., G. C. CHRISTENSEN, M. L. HESS, G. K. quickly. Continuing advances in GPS collar tech- AND M. S. WICKERSHAM. 1983. The mounTSUKAMOTO, tain lion in Nevada. Federal Aid in Wildlife Restoranology will further enhance a model-based tion Project W-48-15, Final Report. Nevada Departapproach for estimating predation rates. For of Wildlife. example, GPS data accuracy increased ment when BEIER, P., D. CHOATE, AND R. H. BARRETT. 1995. Move- selective availability was discontinued in 2000. The cost of GPS collars continues to decrease, ment patterns of mountain lions during different behaviors. Journal of Mammalogy 76:1056-1070. potentially allowing purchase of more collars to BOWMAN, J. L., C. O. KOCHANNY, S. DEMARAIS, AND B. D. increase sample size. Collars are becoming LEOPOLD. 2000. Evaluation of a GPS collar for white- tailed deer. Wildlife Society Bulletin 28:141-145. lighter and capable of collecting an increasing CONNOLLY, E.J. 1949. Food habits and life history of the number of locations. Timed breakaway devices mountain lion. Thesis, University of Utah, Salt Lake preclude the need to recapture collared cougars. City, Utah, USA. And, improvements in the reliability of remote CUNNINGHAM, S. C., C. R. GUSTAVSON, AND W. B. BALLARD. downloading systems will allow biologists more 1999. Diet selection of mountain lions in southeast Arizona. Journal of Range Management 52:202-207. timely retrieval of data, improving estimates ofENVIRONMENTAL SYSTEMS RESEARCH INSTITUTE. 1999. prey species and their sex and age. The advance- Introduction to ArcView? GIS, version 3.2. ESRI, ment of GPS technology provides opportunities Redlands, California, USA. to investigate more detailed hypotheses of preda-HARRISON, S. 1990. Cougar predation on bighorn sheep in the Junction Wildlife Management Area, British tor-prey relationships and ecosystem processes. Columbia. Thesis, University of British Columbia, Vancouver, British Columbia, Canada. HOPKINS, R. A. 1989. Ecology of the puma in the Diablo Range, California. Dissertation, University of CaliforThe Wyoming Game and Fish Department, nia, Berkeley, California, USA. Wyoming Animal Damage Management Board, HORNOCKER, M. G. 1970. An analysis of mountain lion and the Pope and Young Club funded this project. predation upon mule deer and elk in the Idaho primWe are indebted to D. Wroe for the intense effort itive area. Wildlife Monographs 21. required to capture and handle cougars. Field HOSMER, D. W., AND S. LEMESHOW. 1989. Applied logistic John Wiley & Sons, New York, New York, assistance from H. Cruickshank, L. Johnson, regression. and ACKNOWLEDGMENTS USA. T. Chapman was appreciated. We thank D. France JAMES, C. F., AND C. E. MCCULLOCH. 1990. Multivariate of France Flying Service, Saratoga, Wyoming, for analysis in ecology and systematics: panacea or panaerial telemetry, and K. Sargent and D. Hawkdora's of box? Annual Review of Ecology and Systematics 21:129-166. the Wyoming Game and Fish Laboratory, Laramie, This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms �316 ESTIMATING COUGAR PREDATION FROM GPS DATA * Anderson and Lindzey J. Wildl. Manage. 67(2):2003 LINDZEY, F. G., B. B. ACKERMAN, D. BARNHURST, T. BECKER, T. P. HEMKER, S. P. LAING, C. MECHAM, W. D. VAN SICKLE. 1989. Boulder-Escalante cougar project. Final Report. Utah Division of Wildlife Research, Salt Lake City, Utah, USA. LOGAN, K. A, AND L. L. SWEANOR. 2001. Desert puma: evolutionary ecology and conservation of an enduring carnivore. Island Press, Washington, D.C., USA. MEHTA, C., AND N. PATEL. 1992. StatXact-Turbo: statistical software for exact nonparametric inference. Cytel Software, Cambridge, Massachusetts, USA. , AND . 1993. LogXact-Turbo: logistic ing winter. Canadian Journal of Zoology 74:771-775. SAS INSTITUTE. 1990. SAS technical report P-200. Release 6.04 edition. SAS Institute, Cary, North Car- olina, USA. SHAW, H. G. 1977. Impact of mountain lion on mule deer and cattle in northwestern Arizona. Pages 17-32 in R. L. Phillips and C.Jonkel, editors. Proceedings of the 1975 Predator Symposium. University of Mon- tana, Missoula, Montana, USA. . 1979. A mountain lion field guide. Special Report No. 9. Arizona Game and Fish Department Phoenix, Arizona, USA. regression software featuring exact methods. Cytel . 1981. Comparison of mountain lion predation Software, Cambridge, Massachusetts, USA. on cattle on two study areas in Arizona. Pages MOEN, R., J. PASTOR, AND Y. COHEN. 1997. Accuracy of 306-316 in L. Nelson and J. M. Peek, editors. ProGPS telemetry collar locations with differential corceedings of the Wildlife Livestock Relationships Symrection. Journal of Wildlife Management 61:530-539. posium. University of Idaho, Forest, Wildlife, and S, - , AND C. C. SCHWARTz. 1996.Range Experiment Station, Moscow, Idaho, USA. Effects of moose movement and habitat use on GPS SPALDING, D. J., AND J. LESOWSKI. 1971. Winter food of the cougar in south-central British Columbia. Journal collar performance. Journal of Wildlife Management 60:659-668. of Wildlife Management 35:378-381. MOORE, T. D., L. E. SPENCE, AND C. E. DUGNOLLE.TURNER, 1974. J. W., M. L. WOLFE, AND J. F. KIRKPATRICK. 1992. Identification of the dorsal guard hairs of some mamSeasonal mountain lion predation on a feral horse mals of Wyoming. Bulletin No. 14. Wyoming Game population. CanadianJournal of Zoology 70:929-934. and Fish Department, Cheyenne, Wyoming, USA. WEHAUSEN, J. D. 1996. Effects of mountain lion predaMURPHY, K. M. 1998. The ecology of the cougar tion (Puma on bighorn sheep in the Sierra Nevada and concolor) in the northern Yellowstone Ecosystem: Granite Mountains of California. Wildlife Society Bulletin 24:471-479. interactions with prey, bears, and humans. Dissertation, University of Idaho, Moicow, Idaho, USA. WILLIAMS, J. S., J. J. MCCARTHY, AND H. D. PICTON. 1995. NOWAK, M. C. 1999. Predation rates and foraging ecoloCougar habitat use and food habits on the Montana gy of adult female mountain lions in northeastRocky Ore- Mountain Front. Intermountain Journal of Sciences 1:16-28. gon. Thesis, Washington State University, Pullman, Washington, USA. WYOMING GAME AND FISH DEPARTMENT. 1999. Annual big REMPEL, R. S., A. R. RODGERS, AND K. F. ABRAHAM.game 1995. herd unit reports: Laramie Region. Wyoming Performance of a GPS animal location system under Game and Fish Department, Cheyenne, Wyoming, USA. boreal forest canopy. Journal of Wildlife Management 59:543-551. Ross, I. P., AND M. G. JALKOTZY. 1996. Cougar predation on moose in southwestern Alberta. Alces 32:1-8. Received 31 December 2001. Accepted 8 December 2002. , , AND M. FESTA-BIANCHET. 1997. Cougar predation on bighorn sheep in southwestern AlbertaAssociate dur- Editor: White, Jr. This content downloaded from 174.198.140.53 on Tue, 06 Jul 2021 20:26:22 UTC All use subject to https://about.jstor.org/terms � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal Articles Description An account of the resource CPW peer-reviewed journal publications Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Estimating cougar predation rates from GPS location clusters Rights Information about rights held in and over the resource <div class="element"> <div class="element-text"><a href="http://rightsstatements.org/vocab/InC-NC/1.0/" target="_blank" rel="noreferrer noopener">In Copyright - Non-Commercial Use Permitted</a></div> </div> Date Created Date of creation of the resource. 2003-04 Subject The topic of the resource Cougar Global Positioning System (GPS) Predation model Predation rate Prey composition Color Wyoming Description An account of the resource <span>We examined cougar (Puma concolor) predation from Global Positioning System (GPS) location clusters (≥2 locations within 200 m on the same or consecutive nights) of 11 cougars during September-May, 1999-2001. Location success of GPS averaged 2.4-5.0 of 6 location attempts/night/cougar. We surveyed potential predation sites during summer-fall 2000 and summer 2001 to identify prey composition (n = 74; 3-388 days post predation) and record predation-site variables (n = 97; 3-270 days post predation). We developed a model to estimate probability that a cougar killed a large mammal from data collected at GPS location clusters where the probability of predation increased with number of nights (defined as locations at 2200, 0200, or 0500 hr) of cougar presence within a 200-m radius (P&lt;0.001). Mean estimated cougar predation rates for large mammals were 7.3 days/kill for subadult females (1-2.5 yr; n = 3, 90% CI: 6.3 to 9.9), 7.0 days/kill for adult females (n = 2, 90% CI: 5.8 to 10.8), 5.4 days/kill for family groups (females with young; n = 3, 90% CI: 4.5 to 8.4), 9.5 days/kill for a subadult male (1-2.5 yr; n = 1, 90% CI: 6.9 to 16.4), and 7.8 days/kill for adult males (n = 2, 90% CI: 6.8 to 10.7). We may have slightly overestimated cougar predation rates due to our inability to separate scavenging from predation. We detected 45 deer (Odocoileus spp.), 15 elk (Cervus elaphus), 6 pronghorn (Antilocapra americana), 2 livestock, 1 moose (Alces alces), and 6 small mammals at cougar predation sites. Comparisons between cougar sexes suggested that females selected mule deer and males selected elk (P &lt; 0.001). Cougars averaged 3.0 nights on pronghorn carcasses, 3.4 nights on deer carcasses, and 6.0 nights on elk carcasses. Most cougar predation (81.7%) occurred between 1901-0500 hr and peaked from 2201-0200 hr (31.7%). Applying GPS technology to identify predation rates and prey selection will allow managers to efficiently estimate the ability of an area's prey base to sustain or be affected by cougar predation.</span> Creator An entity primarily responsible for making the resource Anderson Jr, Charles R. Lindzey, Frederick G. Language A language of the resource English Is Part Of A related resource in which the described resource is physically or logically included. The Journal of Wildlife Management Format The file format, physical medium, or dimensions of the resource application/pdf Extent The size or duration of the resource. 10 pages Bibliographic Citation A bibliographic reference for the resource. Recommended practice is to include sufficient bibliographic detail to identify the resource as unambiguously as possible. Anderson, C. R., Jr., and F. G. Lindzey. 2003. Estimating cougar predation rates from GPS location clusters. The Journal of Wildlife Management 67:307-316. <a href="https://www.jstor.org/stable/3802772" target="_blank" rel="noreferrer noopener">https://www.jstor.org/stable/3802772</a> Type The nature or genre of the resource Article