Habitat selection by Dall's sheep (Ovis dalli): maternal trade-offs

J. Zool., Lond. (1998) 245, 457±465 # 1998 The Zoological Society of London Printed in the United Kingdom Habitat selection by Dall's sheep (Ovis dalli): maternal trade-offs Janet L. Rachlow 1 and R. Terry Bowyer Institute of Arctic Biology and Department of Biology and Wildlife, University of Alaska Fairbanks, Fairbanks, AK 99775, U.S.A. (Accepted 12 November 1997) Abstract Habitat selection by female Dall's sheep (Ovis dalli dalli) during lambing was studied in interior Alaska, U.S.A., in 1988 and 1989. Selection of habitat changed with the chronology of lambing. During the prelambing period, maternal females selected sites with forage and avoided snow-covered areas. During peak lambing, food and steep slopes continued to be selected. Distance to escape terrain was a critical component of habitat selection by females throughout lambing. Maternal bands that ventured farther from escape terrain were in larger groups. Abundance of forage also increased in areas that were farther from this terrain. Habitat selection by females also differed between years. In 1989, a late spring storm resulted in deep snow and delayed phenology of plants; forage was of lower quality, and the growing season was much shorter in 1989 than in 1988. Births of lambs also occurred later and less synchronously in the second year. Females selected terrain characteristics that were related to avoiding predators in 1988 when forage was more plentiful, but selected sites with forage in 1989 when food was less abundant. Group size of maternal bands was larger in 1988 than in 1989, and females foraged most ef ciently during 1988. Large groups may not have occurred in 1989 because lambs were born asynchronously, preventing cohesive movements of maternal bands with lambs of differing ages. Both the chronology of lambing and a variable environment affected habitat selection by females. Maternal females made trade-offs between the requirements for forage to meet the high energetic costs of lactation and the risk of predation. Key words: Dall's sheep, Ovis dalli dalli, group size, habitat selection, maternal trade-offs INTRODUCTION 1 Present address: Institute of Zoology, Zoological Society of London, Regent's Park, NW1 4RY London, U.K. Behaviour and habitat selection by female ungulates is in uenced by their reproductive state as well as local ecological conditions (Clutton-Brock, Guinness & Albon, 1982; Festa-Bianchet, 1988a; Berger, 1991). Gestation and especially lactation increase nutritional and energetic demands on females (Sadleir, 1980; White & Luick, 1984; Oftedal, 1985). Additionally, maternal females may be constrained in selection of habitat because of vulnerability of young to predation (Berger, 1991; Bleich, Bowyer & Wehausen, 1997). Under such conditions, females likely make trade-offs between foraging in habitats with abundant food and the risk of predation (Berger, 1991; Nicholson, Bowyer & Kie, 1997). Dall's sheep (Ovis dalli dalli), which occur throughout rugged mountain ranges in western Canada and Alaska, USA (Bowyer & Leslie, 1992), are the northernmost species of mountain sheep in North America. Severity of Arctic and sub-arctic environments may pose unique constraints on selection of habitat by Dall's sheep; growth and development of young, and replenishment of female body reserves must occur in a short summer at high latitudes (Bunnell, 1982; Rachlow & Bowyer, 1991, 1994). Considerable information exists about habitat selection by bighorn sheep (Ovis canadensis) (Festa- Bianchet, 1988a, b; Gionfriddo & Krausman, 1986; Krausman & Leopold, 1986), including habitat components selected on summer and winter ranges (Fairbanks, Bailey & Cook, 1987; Wakelyn, 1987). Similarly, habitat use by Dall's sheep outside the lambing period has been studied extensively (Hoefs & Cowan, 1979; Burles & Hoefs, 1984; Hoefs & Bayer, 1983). Female bighorn sheep use high elevations, rugged terrain, and steep slopes during lambing (Geist, 1971;

458 J. L. Rachlow and R. T. Bowyer Bleich et al., 1997). Few studies, however, have addressed the habitat needs of female Dall's sheep around the time of parturition, and none has examined how females respond to changes in the costs of lactation and risk of predation through the lambing period. Likewise, female strategies for coping with the extremely variable climate at high latitudes are largely unstudied. Several factors probably interact to affect selection of habitat by mountain sheep for parturition and rearing of young. Studies of both domestic (Alexander, 1961) and wild sheep have shown that mortality of neonates occurred from exposure and hypothermia as discussed by Wehausen et al. (1987). Predation on neonates also has been implicated as a major cause of death in bighorn sheep (Berger, 1978a; Hass, 1989). Energetic requirements of female mammals shift with onset of lactation (Miller, 1977), and habitat selection often varies in response to such changes in reproductive state (Lott, 1984; Bowyer, 1986; Berger, 1991). Likewise, habitat requirements of young may vary with age as vulnerability to predation changes, and as young begin foraging (Geist, 1971). We tested for differences in habitat selection by female Dall's sheep before lambing, during its peak, and following the season of births. We hypothesize that females should use sites more sheltered from harsh climatic conditions, and closer to escape terrain during the peak period of lambing, re ecting vulnerability of young lambs to hypothermia and predation. Other researchers have suggested that a trade-off between forage quantity and distance to escape terrain is often made by mountain sheep (Risenhoover & Bailey, 1985; Warrick & Krausman, 1987; Berger, 1991; Frid, 1997). If a positive correlation exists between forage abundance and distance to escape terrain, then we predict that females should use areas with lower abundance of forage during peak lambing. During the post-lambing period, however, females should select areas with a greater abundance of forage because of the nutritional demands of lactation (White & Luick, 1984). Further, escape terrain should be less important during the postlambing period when lambs are less vulnerable to predation. Timing of births and subsequent patterns of maternal investment by Dall's sheep in Denali National Park, Alaska, USA, differed markedly between 1988 and 1989; births were later and less synchronous in the second year (Rachlow & Bowyer, 1991). Following parturition in 1989, total time spent suckling per hour was greater than twice as long as in 1988, but weaning occurred more rapidly in the second year (Rachlow & Bowyer, 1994). Berger (1991) documented shifts in foraging behaviour and habitat use relative to reproductive status of female bighorn sheep. We hypothesize that females in our study should compensate for increased energetic demands of lactation in the second year by: (1) foraging in areas of greater forage availability; or (2) increasing total time spent foraging in 1989 relative to 1988. METHODS Study area Igloo Mountain and the surrounding area (approximately 1165 ha) in Denali National Park and Preserve, Alaska, USA (648N, 1508W), was chosen as the site for this study based on its habitat characteristics and traditional use by Dall's sheep for lambing (Murie, 1944). Large adult males seldom occurred on Igloo Mountain during the lambing season. Indeed, spatial separation of the sexes outside the mating season is common amongst ungulates (Bowyer, 1984; Clutton-Brock, Iason & Guinness, 1987; Bleich et al., 1997). Igloo Mountain is separated from nearby mountains by drainages on 3 sides and connected by a low ridge to Sable Mountain to the south. Elevation ranges from 1480 m at the summit to 885 m in river valleys. Topography varies from steep, rocky slopes at higher elevations to gently rolling hills at lower elevations. The climate in the park is typical of mountainous regions at high latitude; summers are short and cool and winters are long and harsh. Average temperature ranges from -17 8C in January to 12 8C in July. Yearly snowfall averages 190 cm. Weather varied markedly between study years; in 1988, patches of vegetation were free of snow by early May; however, in 1989 a late storm deposited 25 cm of fresh snow in mid May. In addition, the growing season in 1988 was nearly twice as long as in 1989; degree days > 5 8C were 576 and 469 in 1988 and 1989, respectively (Rachlow & Bowyer, 1994). White spruce (Picea glauca) forest occurs in valley bottoms, and shrub birch (Betula glandulosa) and willows (Salix sp.) are abundant at lower elevations, especially along drainages. Matted alpine tundra dominated by dryas (Dryas sp.), grasses, lichens, mosses, and forbs occurs above treeline at approximately 950 m. The number of females observed using Igloo Mountain each day was nearly identical in 1988 (29 animals) and 1989 (30 animals), and maternal groups included females with and without lambs (Rachlow & Bowyer, 1991). Predators of Dall's sheep we observed on the study area were grizzly bears (Ursus arctos), wolves (Canis lupus), and golden eagles (Aquilia chrysaetos) (Rachlow & Bowyer, 1991). Other predators of Dall's sheep in Denali Park include wolverines (Gulo gulo), lynxes (Lynx canadensis), and coyotes (Canis latrans). Sampling procedures We observed Dall's sheep from April to July in 1988 and 1989. Field work was conducted 3 days/week in April and early May, and 6 days/week thereafter, for a total of 68 days in 1988 and 74 days in 1989. We initially located groups of females, often accompanied by yearlings, young males (< 3 years old), or lambs, by using binoculars, and then approached on foot to 100 m for observations. We collected data on activity using scan samples of focal groups at 15-min intervals (Altmann,

Maternal trade-offs in Dall's sheep 459 1974), totalling 155 h of observation in 1988 and 235 h in 1989. Per cent of time active (standing versus bedded) and per cent of active time spent feeding (Berger, 1991) were examined using activity data on females chosen at random from focal groups. Females were observed for a minimum of 30 min to be included in this analysis, but most were observed for >1 h. We also recorded the size of the focal groups, which were de ned as animals 100 m from another individual that exhibited similar patterns of activity and movement. Sites used by sheep were identi ed by randomly choosing an adult female (2 years old) in the focal group at 1-h intervals to help minimize dependence among samples. We plotted locations of selected individuals on an infra-red aerial photograph (approximately 12 cm = 1 km), and made sketches to aid in relocating those sites. Care was taken not to disturb sheep during periods of observation and habitat sampling; we discontinued observations if animals exhibited alarm behaviours. We usually sampled sites with sheep immediately after they left the area, but some were sampled on the following day. When possible we located all groups of females and lambs within the study area each day. Randomly chosen sites also were located on the infra-red aerial photograph and sampled in the same manner as sites used by Dall's sheep. These random locations included areas on Igloo Mountain that were at least occasionally used by female Dall's sheep. Measurements of a suite of habitat variables were made at 2 scales: (1) plots of 1 m 2 (micro-sites) were sampled at locations used by sheep and at randomly located sites; and (2) macro-sites, which all received use by female sheep, by subdividing the study area in 19 zones based on elevation and aspect. Depending upon steepness of terrain and elevation, macro-sites ranged from 15 to 450 ha, with larger sites at lower elevations. Each micro-site was located within one of these zones and assigned the values for macro-habitat measurements associated with that zone. These two scales are relevant to habitat use by Dall's sheep on Igloo Mountain and help avoid potential biases from sampling at a single scale (Bowyer, Kie & Van Ballenberghe, 1996). Habitat characteristics sampled at micro-sites included vegetative, climatic, and physiogeographic variables. We used a point frame (Kershaw, 1964) to sample per cent vegetative cover by species (n = 40 `pin hits' per plot); these data subsequently were summed by forage class (grasses, forbs, browse, dryas, or mosses and lichens) for statistical analyses. Species seldom eaten by Dall's sheep in Denali Park (e.g. Betula, Juniperus, Ledum, etc.) were categorized as `other'. Climatic variables measured in this study included wind speed, wind direction, ambient temperature, and snow cover. We measured wind speed using a wind meter, and determined direction with a compass. Measurements were taken for greatest sustained wind speeds or for maximum gusts when wind speed was variable. We used a thermometer to record ambient temperature to the nearest 1 8C, and values for windchill were calculated from wind speed and temperature. Snow was recorded as present or absent. Terrain features at micro-sites were recorded as follows: slope was measured using an Abney level; aspect was determined with a compass; brokenness was indexed by laying a ne chain diagonally across each plot, measuring this surface distance, and subtracting the distance calculated for a at surface; elevation was determined from a topographic map (scale = 1:63). Directional measurements (aspect and wind direction) were coded as two values (Cartesian coordinates), sine and cosine of the angle, to allow analyses of these circular variables (Zar, 1984). Sine and cosine represent east±west and north±south differentiation, respectively. We estimated distance to escape terrain visually and veri ed this distance on an aerial photograph. We de ned escape terrain as rockland with steep slopes or cliffs on which sheep could outrun predators (Gionfriddo & Krausman, 1986; Wakelyn, 1987). Macro-scale measurements were obtained from a topographic map (scale = 1:63) and an infra-red aerial photograph. Macro-aspect of each zone was categorized as 1 of 8 directions (N, NE, E, SE, S, SW, W, NW), and the central value (degrees) for each category was transformed into sine and cosine. We calculated macro-slope (in degrees) by taking the inverse of the tangent of change in elevation divided by the distance over which the rise occurred (rise/run). Brokenness of terrain was indexed by subtracting straight-line distances across zones from those measured by following contour lines within each macro-site zone. Three values for brokenness were averaged for each zone. Lastly, we rated openness of habitat (Hirth, 1977) in each zone on a scale from 1 to 7 ranging from open rock and scree slopes (1) to closed spruce forest (7); this measure re ects broad habitat types. Thus, each random and used site was characterized on small and large plots to examine habitat selection at different scales (Bowyer, Kie et al., 1996). Statistical analyses Data were tested for normality, and transformations were required for several variables (Zar, 1984). Pearson correlation coef cients were examined to minimize multicollinearity amongst variables; elevation and habitat openness were removed from further analyses because of multicollinearity (r 70 ) with other variables. Additionally, we deleted per cent cover of the `other' vegetative class from analyses due to lack of independence with per cent cover of forage categories. We de ned periods of lambing based on timing of births in each year. Peak lambing included those dates during which 90% of births occurred (Rachlow & Bowyer, 1991); 9 May to 4 June in 1988; and 22 May to 26 June in 1989. Pre-lambing samples were collected before onset of lambing in each year, and post-lambing

460 J. L. Rachlow and R. T. Bowyer samples included those collected from completion of the peak-lambing period through July. We used discriminant function analysis (DFA) with stepwise inclusion of variables (Johnson & Wichern, 1988) to identify habitat components important in separating sites used by sheep from those sampled randomly. Selection of variables was based on maximization of Mahalanobis' distance between random and used sites (Dixon, 1983). DFA was used primarily for model building in these analyses, but only variables with signi cant (P < 0.05) F-values were included in our models. We examined appropriateness of the models by jack-kni ng on a case-by-case basis; prior probabilities were based on sample sizes used to build each DFA model, and posterior probabilities were calculated by eliminating one case, applying DFA, and classifying the case withheld from analysis (Dixon, 1983). In addition, DFA was conducted on rank-transformed data (Conover & Iman, 1981) to assess the effects of violations of the assumption of normality on our analysis; similar models resulted from ranked data, suggesting that the models were apt. DFA models for periods of lambing (pre, peak, and post) within years were overparamatized (Williams, 1983), and hence, omitted from analyses. Hypotheses regarding simultaneous selection of habitat components were tested using two-way multivariate analysis of variance (MANOVA) (Johnson & Wichern, 1988). Class variables were `use' and either `period' of lambing (pre, peak, or post) or `year' (1988 or 1989). Selection for or against habitat components was indicated by signi cant variation between sites used by sheep and random sites (use versus availability). Further, interactions between selection and period effects in the MANOVA were used to test hypotheses about variation in selection of habitat components relative to timing of births and nursing of lambs. Likewise, interactions between selection and year effects were used to test hypotheses regarding yearly differences in habitat selection. Differences in habitat selection by time interval were inferred from a signi cant interaction between use and either period or year. The Hotelling± Lawley trace was used to test for overall differences of multivariate models (SAS Institute, 1985). We applied 2- way ANOVAs with a sequential Bonferroni test (Rice, 1989) to univariate statistics to determine which variables contributed most to overall differences in selection. We used multiple linear regression with stepwise inclusion of variables to identify habitat variables related to group size (Neter, Wasserman & Kutner, 1985). Yearly trends in group size, per cent of time that sheep were active, and per cent of active time spent feeding were analysed with the Mann±Whitney U-test (Zar, 1984). We examined variation amongst periods using a Kruskal±Wallis test (Zar, 1984). In uence of group size on feeding behaviours was tested using simple linear regression. We used a Z-test to compare per cent time feeding by females without and with lambs (Zar, 1984). RESULTS Use and selection of lambing habitat Lambing habitat for Dall's sheep was characterized by steep, rugged terrain intermixed with forage, including grasses and dryas (Table 1). Such habitat typically was at high elevation and free of snow. Discriminant function analysis (DFA) with years and lambing periods combined (full model) indicated that nine variables were useful in discriminating between random sites and those used by Dall's sheep (Table 2). Standardized canonical correlations indicated that macro-brokenness, per cent cover of grasses and dryas, and distance to escape terrain were in uential variables. Discriminant function analysis within each period of lambing (pre, peak, and post) indicated that similar variables were selected during each period with a few notable exceptions (Fig. 1). Windchill, macro-brokenness, and browse were useful in discriminating between random sites and those used by sheep only during peak lambing, whereas snow was important only during the pre-lambing period. Two additional variables, macroaspect (E±W) (canonical correlation, peak = 0.40, post = 0.43) and micro-slope (canonical correlation, post = 70.65) also were selected. These variables re ect use of an east-facing meadow on Igloo Mountain and use of relatively level microsites on steep macro-slopes. A two-way MANOVA, including 11 variables that entered either the full model (Table 2) or one of the within-period models (Fig. 1), indicated that random sites and those used by Dall's sheep differed overall when these variables were considered simultaneously (F = 30.23, d.f. = 11,428, P < 0.001). A signi cant interaction between selection of habitat and period of lambing in the two-way MANOVA (F = 5.68, d.f. = 11, 428, P < 0.001) indicated that changes in selection of habitat were associated with the chronology of lambing. Variables that differed most among periods of lambing were macro-brokenness, per cent cover of browse, and snow (Fig. 1); these variables were signi cant at joint con dence of 95% using a sequential Bonferroni test. Distance to escape terrain and cover of important forage (grass or dryas) were selected in all periods of lambing (Fig. 1). Signi cant positive correlations were observed between distance to escape terrain and per cent cover of grasses (r = 0.21, P<0.01), browse (r = 0.44, P<0.01), but not dryas (r = 0.02, P>0.50). Additionally, stepwise multiple regression of habitat variables against distance to escape terrain (dependent variable) indicated that per cent cover of browse, dryas, and grasses all were related positively to distance to escape terrain (R 2 = 0.37, F = 42.54, P > 0.001; YÃ = 0.63 + 0.91 browse + 0.30 dryas + 0.43 grasses ± 0.02 macro-slope). These data indicate that females made a trade-off between forage abundance and distance to escape terrain in selection of habitat because areas with more forage tended to be further from escape terrain.

Maternal trade-offs in Dall's sheep 461 Table 1. Means ( SD) of components for habitat of Dall's sheep measured in interior Alaska, U.S.A. Total sample sizes for sites used by sheep and for randomly located sites are given for 1988 and 1989 combined Sites used by sheep Random sites n = 316 n = 231 Variable XÅ sd XÅ sd Distance to escape (m) a 20.50 46.07 319.11 358.67 Elevation (m) 1187.49 101.06 066.54 135.03 Micro-brokenness (cm) a 17.67 13.09 21.87 19.83 Micro-slope (degrees) b 19.39 10.15 20.60 12.28 Micro-aspect (N±S) 70.20 0.68 70.15 0.67 Micro-aspect (E±W) 0.34 0.62 0.29 0.67 Wind direction (N±S) 70.09 0.73 70.31 0.67 Wind direction (E±W) 0.17 0.66 0.07 0.68 Wind speed (km/h) 16.91 12.44 14.40 14.22 Temperature (8C) 12.75 6.72 12.01 7.10 Windchill (8C) 12.06 7.72 14.34 7.75 Snow presence 0.02 0.15 12.25 8.60 Percentage cover Grasses c 20.53 23.79 15.14 20.98 Browse c 9.45 17.43 18.30 25.22 Forbs a 2.77 6.35 1.96 6.28 Dryas a 15.12 21.65 5.60 12.94 Moss-lichen a 5.23 9.63 8.94 14.46 Other 4.93 11.58 14.33 24.01 Macro-brokenness 1.97 0.43 1.72 0.38 Macro-slope (degrees) 3.43 6.44 17.23 7.33 Macro-aspect (N±S) 70.12 0.58 70.01 0.73 Macro-aspect (E±W) 0.41 0.70 0.32 0.62 Macro-site openness 2.70 1.35 4.45 2.13 a a Data were log transformed for analyses. b Data were square root transformed. c Data were arcsine transformed. Table 2. Standardized canonical correlation coef cients for habitat variables useful in discriminating between sites used by Dall's sheep and randomly located sites on Igloo Mountain in interior Alaska, U.S.A., when all data were considered (full model). Order of variables indicates the order in which each was added in stepwise D.F.A. Per cent of sites classi ed correctly using this model was evaluated with a jack-kni ng procedure Variable Habitat selection between years Full mode Distance to escape terrain 70.742 Percentage cover Grasses 0.310 Dryas 0.361 Macro-aspect (E±W) 0.380 Macro-brokenness 0.459 Micro-slope 70.363 Macro-slope 0.303 Macro-aspect (N±S) 70.286 Snow presence 70.178 Percentage classi ed correctly 84.3 Selection of habitat also varied between 1988 and 1989. DFA models were similar between years (Table 3); however, per cent cover of forage classes (grass, dryas, and browse) was useful in discriminating between random sites and those used by sheep only in 1989. Forage variables were strongly selected in 1989, whereas terrain features (macro-brokenness and macro-slope) entered the DFA model only in 1988 (Table 3). An interaction between use and year in the two-way MANOVA indicated a signi cant difference in selection between years (F = 3.12, d.f. = 11, 429, P < 0.001). Although Dall's sheep used sites with slightly more grass and dryas in 1988 than in 1989, lower availability of these forage classes on random sites in the second year of study (Fig. 2) resulted in them being more strongly selected by females in 1989 (Table 3). Effects of group size on habitat selection Group size of maternal bands also was related to selection of habitat by Dall's sheep. Although the predictive nature of this relationship was not strong, stepwise multiple regression with group size as the dependent variable indicated that distance to escape terrain and per cent cover of dryas were signi cantly related to group size (R 2 = 0.06, F = 11.18, P<0.001; YÃ = 5.30 + 2.01 escape + 1.43 dryas). Likewise, group size varied signi cantly with distance to escape terrain in multiple regressions performed independently for 1988 (R 2 = 0.08, F = 10.99, P < 0.002; YÃ = 7.04 + 3.71 escape) and 1989 (R 2 = 0.07, F = 5.74, P = 0.004; YÃ = 0.93 + 1.42 escape

462 J. L. Rachlow and R. T. Bowyer Selection (canonical correlation coefficient) 0.9 0.7 0.5 0.3 0.1 0.1 0.3 0.5 0.7 0.9 0.9 0.7 0.5 0.3 0.1 0.1 0.3 0.5 0.7 0.9 0.9 0.7 0.5 0.3 0.1 0.1 0.3 0.5 0.7 0.9 Pre-lambing period Peak-lambing period Post-lambing period Escape Grass Dryas Broken Slope Snow Browse Windchill Habitat variables Fig. 1. Selection of habitat by female Dall's sheep during pre (above), peak (middle), and post (below) lambing in interior Alaska, U.S.A., 1988±1989. Escape = distance to escape terrain, grass = % cover of grasses, dryas = % cover of dryas, broken = macro-brokenness of terrain, slope = macro-slope, snow = presence or absence of snow, browse = % cover of browse, and windchill. Canonical correlations indicate the strength and direction of selection. For instance, the negative correlation for distance to escape terrain indicates females used sites closer to such terrain than randomly located sites. The positive correlative for grass indicates there was more of this forage class at sites used by females than at random sites. + 0.16 macro-slope). Dryas did not enter the models for either year independently, and macro-slope was signi cantly related to group size only in 1989. Mean size of maternal bands of Dall's sheep differed between years, but was relatively constant within each year (Fig. 3). Per cent of time females were active (not bedded) did not differ between years (1988 = 49.5%, 1989 = 51.4%; Mann±Whitney U= 0.10, d.f. = 1, P=0.76). Per cent of active time spent feeding, however, was greater in 1988 than in 1989 (Fig. 4). Because females were active for comparable proportions of time in both years, but spent a greater per cent of active time foraging in 1988, females spent more total time foraging in 1988. Group size was related to activity and feeding behaviours. Regressions of per cent time active and per cent Per cent cover 30 25 20 15 10 5 0 30 25 20 15 10 5 0 active time spent feeding against group size with years combined indicated that per cent of time active increased signi cantly with group size (b=0.84, P = 0.026). Per cent of active time spent feeding, however, did not show a signi cant trend with group size (b=0.20, P = 0.440). Females in larger groups spent a greater total time feeding (% time active 6 % active time spent foraging). This conclusion is supported by the results of yearly comparisons of foraging behaviour; females spent more total time foraging in 1988 when groups were larger. DISCUSSION Random sites Sites with sheep 1988 1989 Grass Dryas Browse Fig. 2. Per cent cover (XÅ + se) of forage classes at random sites and those used by female Dall's sheep in interior Alaska, U.S.A. Periods of lambing are combined within years (n =72 in 1988 and n = 117 in 1989 for random sites; n = 157 in 1988 and n = 194 in 1989 for sites used by sheep). We hypothesized that selection of habitat by female Dall's sheep would vary through the lambing period. The con icting needs of the female to acquire forage necessary to support the high costs of lactation and to minimize risk of predation for the neonate change through this period and re ect such maternal con icts.

Maternal trade-offs in Dall's sheep 463 Table 3. Standardized canonical correlation coef cients for DFA models of habitat selection by Dall's sheep in interior Alaska, U.S.A., for 1988 and 1989 with lambing periods combined within each year. Percent of sites classi ed correctly was evaluated using a jack-kni ng procedure Variable 1988 model 1989 model Distance to escape terrain 70.421 70.1224 Percentage cover Grasses 0.334 Dryas 0.466 Browse 0.242 Macro-aspect (E±W) 0.701 0.277 Macro-brokenness 0.664 Micro-slope 70.413 70.312 Macro-slope 0.571 Macro-aspect (N±S) Snow presence Percentage classi ed correctly 84.3 84.6 Indeed, Dall's sheep signi cantly altered important components of habitat they selected through the three periods of lambing (Fig. 1). Distance to escape terrain, however, was useful in discriminating between randomly located sites and those used by Dall's sheep during all periods of lambing and in both years. Habitat features constituting escape terrain have been noted consistently as essential components of habitat for mountain sheep (Geist, 1971; Krausman & Leopold, 1986; Bleich et al., 1997). Moreover, Festa-Bianchet (1988b) reported that bighorn sheep made trade-offs between forage and risk of predation. Berger (1991) also associated relative levels of predation risk with habitat types for bighorn sheep. He reported that lambs in open or at terrain were killed by predators more frequently than lambs occurring in steep or rugged terrain (71% vs 22%, respectively). Considering the relatively high densities of wolves (Miquelle, Peek & Van Ballenberghe, 1992) and grizzly bears (Albert & Bowyer, 1991) in Denali Park, similar relative risks of predation probably were associated with these types of habitat on Igloo Mountain. Selection of other habitat features by female Dall's sheep varied with the chronology of lambing. Greater cover of forage (grass and dryas), and lower snow cover were selected by sheep prior to lambing. During peak lambing, however, terrain features associated with predator avoidance (macro-brokenness and macro-slope) were strongly selected for use by maternal females. Lambs occurred in areas with more browse during peak lambing, perhaps because these shrubs provided shelter from the wind (Fig. 1), or concealment cover from predators. Concealment cover is an important component of habitat for young ungulates (Bowyer, 1986; Bowyer, Kie & Van Ballenberghe, 1998). Foraging ef ciency (per cent active time spent feeding) declines when females have young at heel or use habitats with an increased risk of predation (Berger, 1978b; Molvar & Bowyer, 1994; Frid, 1997). Indeed, Dall's sheep females with lambs followed this pattern Group size 14 12 10 8 6 4 2 0 8 1988 1989 14 32 Grass Dryas Browse Lambing chronology Fig. 3. Group size (XÅ + sd) of maternal bands of Dall's sheep during lambing in 1988 and 1989 in interior Alaska, U.S.A. Sample sizes are indicated above bars. Group size differed signi cantly between years (Mann±Whitney U = 27.3, d.f. = 1, P < 0.001), but not within years (1988: Kruskal±Wallis test, w 2 = 2.84, d.f. = 2, P= 0.24; 1989: w 2 = 2.22, d.f. = 2, P = 0.33). Active time spent feeding (%) 100 90 80 70 60 50 1988 1989 n = 8 15 37 46 26 32 PRE PEAK POST Lambing chronology Fig. 4. Per cent of active time spent foraging by female Dall's sheep during lambing in interior Alaska, U.S.A., 1988±1989. Sample size (n) is the number of observation periods. Females spent more active time foraging in 1988 than in 1989 (Mann± Whitney U = 5.10, d.f. = 1, P = 0.024). Females spent less active time foraging when they had lambs (peak and post, combined) than prior to parturition in 1988 (Z = 2.19, P < 0.05) and in 1989 (Z = 2.79, P < 0.02). (Fig. 4). Differences in foraging ef ciency between years (Fig. 4) were most likely related to group size, which was larger in 1988 (Fig. 3). Foraging ef ciency also tends to increase with increasing group size in bighorn sheep (Berger, 1978b, 1991). Morphological traits that enhance foraging ef ciency have been related to overwinter survival in Soay sheep (Ovis aries) (Ilius et al., 1995). Likewise, behaviours that increase an individual's 38 22 25

464 J. L. Rachlow and R. T. Bowyer feeding ef ciency also may contribute to survival in a severe, Arctic climate. We hypothesize that differences in synchrony of parturition were an important factor in uencing the disparate patterns of habitat selection and sociality that occurred between years. Ninety per cent of births occurred in a 27-day period in 1988 and in a 36-day period in 1989 (Rachlow & Bowyer, 1991). Thus, the differing ages of young likely were an impediment to the formation and cohesive movement of maternal bands of Dall's sheep in 1989. Spatial associations in bighorn sheep change with the age of lambs (Berger, 1979). Timing of births and patterns of maternal investment by female Dall's sheep differed markedly between 1988 and 1989 (Rachlow & Bowyer, 1991, 1994). In the second year of study, a late snowstorm substantially delayed the growth of plants, and a relatively cool growing season resulted in lower quality of forage for sheep; the length of the growing season also was much shorter in 1989 than in 1988 (Rachlow & Bowyer, 1991, 1994). Such differences between years are re ected in the availability of grass and dryas on random sites (Fig. 2). A nutritious diet is an essential component for lactation in ungulates (White & Luick, 1984; Loudon, 1985). Thus, female Dall's sheep selected forage classes more strongly in 1989 when food was less available (Table 3, Fig. 2); they also foraged less ef ciently in 1989 (Fig. 4), ostensibly because forage was more widely distributed. Because of increased maternal investment following parturition in the second year (Rachlow & Bowyer, 1994), we predicted that females would compensate by feeding in areas with more forage or spend more time foraging in 1989. Both of these hypotheses, however, were rejected. Forage availability was likely too low in the second year for females to compensate, although they did select classes of forage more strongly in 1989 than in 1988. Low availability of forage also may have resulted in females weaning lambs more rapidly in 1989 (Rachlow & Bowyer, 1994). Finally, the strong in uence of group size on use of terrain features also affected habitat use by female Dall's sheep. Sheep occurred in larger groups in 1988 than in 1989 (Fig. 2) when they ventured farther from escape terrain into areas with more forage; females also foraged more ef ciently in 1988 than in 1989 (Fig. 4). Also, females selected other terrain features in 1988 when forage was more available (Table 3). Risk of predation clearly effects selection of habitats by females with young (Kohlman, Muller, & Alkon, 1996; Bleich et al., 1997), and females in our study likely traded off the need to acquire suf cient forage to support lactation with the risk of predation. Hence, both forage availability and predation risk constrained habitat selection in female Dall's sheep. Acknowledgements Funding was provided by the Institute of Arctic Biology and the Department of Biology and Wildlife at the University of Alaska Fairbanks, the National Park Service, Sigma Xi, the J. Stelmock Memorial Fund, and C. Ott. We appreciate the help of many National Park personnel especially L. Adams and J. Dolle-Molle. We thank A. Means, B. Platt, and M. Bohac for their assistance in the eld. We thank P. Quang for statistical guidance. We also thank M. Petrula, H. Maier, S. Hartz, and J. Keener for their help. M. Smith and W. Testa provided computer assistance, and D. Klein, E. Murphy, and J. Sedinger helped to improve the manuscript. REFERENCES Albert, D. M. & Bowyer, R. T. (1991). Factors related to grizzly bear±human interactions in Denali National Park. Wildl. Soc. Bull. 19: 339±349. Alexander, G. (1961). Energy expenditure and mortality in newborn lambs. Proc. 4th int. Cong Anim. Repro., 3: 630±637. Altmann, J. (1974). Observational study of behaviour: sampling methods. Behaviour 49: 227±267. Berger, J. (1978a). Maternal defensive behavior in bighorn sheep. J. Mamm. 59: 620±621. Berger, J. (1978b). Group size, foraging, and antipredator ploys: an analysis of bighorn sheep decisions. Behav. Ecol. Sociobiol. 4: 91±99. Berger, J. (1979). Weaning, social environments and ontogeny of spatial associations in bighorn sheep. Biol. Behav. 4: 363±372. Berger, J. (1991). Pregnancy incentives and predation constraints in habitat shifts: experimental and eld evidence for wild bighorn sheep. Anim. Behav. 41: 66±77. Bleich, V. C., Bowyer, R. T. & Wehausen, J. W. (1997). Sexual segregation in mountain sheep: resources or predation? Wildl. Monogr. 134: 1±50. Bowyer, R. T. (1984). Sexual segregation in southern mule deer. J. Mamm. 65: 410±417. Bowyer, R. T. (1986). Habitat selection by southern mule deer. Calif. Fish Game 72: 153±169. Bowyer, R. T., Kie, J. G. & Van Ballenberghe, V. (1996). Sexual segregation in black-tailed deer: effects of scale. J. Wildl. Manage. 60: 10±17. Bowyer, R. T., Kie, J. G. & Van Ballenberghe, V. (1998). Habitat selection by neonatal black-tailed deer: climate, forage, or risk of predation? J. Mamm. 79: 415±425. Bowyer, R. T. & Leslie, D. M., Jr. (1992). Ovis dalli. Mammal. Spec. 393: 1±7. Bunnell, F. L. (1982). The lambing period of mountain sheep: synthesis, hypothesis, and tests. Can. J. Zool. 60: 1±14. Burles, D. W. & Hoefs, M. (1984). Winter mortality of Dall sheep, Ovis dalli dalli, in Kluane National Park Reserve, Yukon Territory. Can Field-Nat. 98: 479±484. Clutton-Brock, T. H., Guinness, F. E. & Albon, S. D. (1982). Red deer: behavior and ecology of two sexes. Chicago: University of Chicago Press. Clutton±Brock, T. H., Iason, G. R. & Guinness, F. E. (1987). Sexual segregation and density-related changes in habitat use in male and female red deer (Cervus elaphus). J. Zool. (Lond.) 211: 275±289. Conover, W. J. & Iman, R. L. (1981). Rank transformations as a bridge between parametric and non±parametric statistics. Am. Stat. 35: 124±130. Dixon, W. J. (1983). BMDP Statistical software. Berkeley: University of California Press. Fairbanks, W. S., Bailey, J. A. & Cook, R. S. (1987). Habitat use by a low-elevation, semi-captive bighorn population. J. Wildl. Mgmt 51: 912±91 5.

Maternal trade-offs in Dall's sheep 465 Festa-Bianchet, M. (1988a). Birthdate and survival in bighorn lambs (Ovis canadensis). J. Zool. (Lond.) 214: 653±661. Festa-Bianchet, M. (1988b). Seasonal range selection in bighorn sheep: con icts of forage quality, forage quantity, and predator avoidance. Oecologia 75: 580±586. Frid, A. (1997). Vigilance by female Dall's sheep: interactions between predation risk factors. Anim. Behav. 53: 799±808. Geist, V. (1971). Mountain sheep: a study in behavior and evolution. Chicago: University of Chicago Press. Gionfriddo, J. B. & Krausman, P. R. (1986). Summer habitat use by mountain sheep. J.Wildl. Mgmt 50: 331±336. Hass, C. C. (1989). Bighorn lamb mortality: predation, inbreeding and population effects. Can. J. Zool. 67: 699±705. Hirth, D. H. (1977). Social behavior of white-tailed deer in relation to habitat. Wildl. Monogr. 53: 1±55. Hoefs, M. & Bayer, M. ( 1983). Demographic characteristics of an unhunted Dall sheep (Ovis dalli dalli) population in southwest Yukon, Canada. Can. J. Zool. 61: 1346±1357. Hoefs, M. & Cowan, I. McT. (1979). Ecological investigation of a population of Dall sheep (Ovis dalli dalli Nelson). Syesis 12: 1±81. Illius, A. W., Albon, S. D., Pemberton, J. M., Gordon, I. J. & Clutton-Brock, T. H. (1995). Selection for foraging ef ciency during a population crash in Soay sheep. J. Anim. Ecol. 64: 481±492. Johnson, R. A. & Wichern, D. W. (1988). Applied multivariate statistics. Englewood Cliffs, NJ: Prentice Hall. Kershaw, K. K. (1964). Quantitative and dynamic ecology. London: Edward Arnold. Kohlmann, S. G., Muller, D. M. & Alkon, P. U. (1996). Antipredator constraints on lactating Nubian ibexes. J. Mamm. 77: 1122±1131. Krausman, P. R. & Leopold, B. D. (1986). Habitat components for desert bighorn sheep in the Harquahala Mountains, Arizona. J. Wildl. Mgmt 50: 504±508. Lott, D. F. (1984). Intraspeci c variation in the social systems of wild vertebrates. Behaviour 88: 266±325. Loudon, A. S. I. (1985). Neonatal mortality and lactation in mammals. Symp. zool. Soc. Lond. 54: 183±207. Millar, J. S. (1977). Adaptive features of mammalian reproduction. Evolution 31: 370±386. Miquelle, D. G., Peek, J. M. & Van Ballenberghe V. (1992). Sexual segregation in Alaskan moose. Wildl. Monogr. 122: 1±57. Molvar, E. M. & Bowyer, R. T. (1994). Costs and bene ts of group living in a recently social ungulate: the Alaskan moose. J. Mamm. 75: 621±630. Murie, A. (1944). The wolves of Mount McKinley. Washington: U.S. Government Printing Of ce. Neter, J., Wasserman, W. & Kutner, M. H. (1985). Applied linear statistical models. 2nd ed. Homewood: Irwin. Nicholson, M. C., Bowyer, R. T. & Kie, J. G. (1997). Habitat selection and survival of mule deer: trade-offs associated with migration. J. Mamm. 78: 483±504. Oftedal, O. T. (1985). Pregnancy and lactation. In Bioenergetics of wild herbivores: 215±238. Hudson, J. R. & White, R. G., (Eds). Boca Raton, FL: CRC Press. Rachlow, J. L. & Bowyer, R. T. (1991). Interannual variation in timing and synchrony of parturition in Dall's sheep. J. Mamm. 72: 487±492. Rachlow, J. L. & Bowyer, R. T. (1994). Variability in maternal behavior by Dall's sheep: environmental tracking or adaptive strategy? J. Mamm. 75: 328±337. Rice, W. R. (1989). Analyzing tables of statistical tests. Evolution 43: 223±225. Risenhoover, L. L. & Bailey, J. A. (1985). Foraging ecology of mountain sheep: implications for habitat management. J. Wildl. Mgmt 49: 797±804. Sadleir, R. M. F. S. (1980). Energy and protein intake in relation to growth of suckling black-tailed deer fawns. Can. J. Zool. 58: 1347±1354. Wakelyn, L. (1987). Changing habitat conditions on bighorn sheep ranges in Colorado. J. Wildl. Mgmt 51: 904±912. Warrick, G. D. & Krausman, P. R. (1987). Foraging behavior of female mountain sheep in western Arizona. J. Wildl. Mgmt 51: 99±104. Wehausen, J. D., Bleich, V. C., Thong, B. & Russi, T. L. (1987). Recruitment dynamics in a southern California mountain sheep population. J. Wildl. Mgmt 51: 86±98. White, R. G. & Luick, J. R. (1984). Plasticity and constraints in the lactational strategy of reindeer and caribou. Symp. zool. Soc. Lond. 51: 215±232. Williams, B. K. (1983). Some observations on the use of discriminant analysis in ecology. Ecology 64: 1233±1291. Zar, J. H. (1984). Biostatistical analysis. Englewood Cliffs, NJ: Prentice-Hall.