John A. Litvaitis

Resampling methods for evaluating classification accuracy of wildlife habitat models

Predictive models of wildlife-habitat relationships often have been developed without being tested The apparent classification accuracy of such models can be optimistically biased and misleading. Data resampling methods exist that yield a more realistic estimate of model classification accuracy These methods are simple and require no new sample data. We illustrate these methods (cross-validation, jackknife resampling, and bootstrap resampling) with computer simulation to demonstrate the increase in precision of the estimate. The bootstrap method is then applied to field data as a technique for model comparison We recommend that biologists use some resampling procedure to evaluate wildlife habitat models prior to field evaluation.

Influence of Understory Characteristics on Snowshoe Hare Habitat Use and Density

The influence of forest understory characteristics on snowshoe hare (Lepus americanus) habitat use and density was studied in eastern (Cherryfield) and western (Pierce Pond) Maine during 1981-83. Fecal pellet counts indicated that hares at Cherryfield preferred hardwood and avoided mixedwood and open understories during the leaf-off season (Oct-May) (P < 0.05). At Pierce Pond, hares used softwood more and hardwood and open understories less than expected during leaf off (P < 0.05). Hardwood understories provided the densest cover at Cherryfield, whereas at Pierce Pond softwoods were the densest cover. Hares in both areas used dense understories less during the leaf-on season (Jun-Sep). Spring population densities (0.1-1.7/ha) were correlated with understory density (r = 0.94, P < 0.001). Overwinter survival also was associated with understory density (r = 0.74, P < 0.04). Dense understories provided escape and thermal

Bobcat Habitat Use and Home Range Size in Relation to Prey Density

Bobcat (Felis rufus) diet, habitat use, and home range size were studied in relation to snowshoe hare (Lepus americanus) density and distribution in 2 areas in Maine during 1979-84. Hare remains occurred in 63-76% of bobcat feces collected during all seasons in both areas. Habitat use patterns of 12 transmitter-equipped bobcats in eastern Maine indicated that they used hardwood understories more, and softwood and mixedwood understories less than expected (P 5* (P < 0.05). The average home range of resident male bobcats (95.7 kmn2) was 3 x as large as that of resident females (31.2 km2) (P < 0.05), and home range size was correlated with bobcat weight (r2 = 0.45, P < 0.002). Metabolic home range size (km2/kg075 body wt) of bobcats was inversely correlated with stem cover unit density and estimated hare density (r2 = 0.22, P < 0.05). Estimated hare density and average topographic slope within bobcat home ranges accounted for 50% of the variation in metabolic home range size (P < 0.006). J. WILDL. MANAGE. 50(1):110-117 Many factors influence vertebrate habitat use and home range size, including energy requirements, prey distribution and density, reproductive requirements, and intraand interspecific relations (Stenger 1958, McNab 1963, Schoener 1968, Smith 1968). Among these, prey density has been reported to be inversely correlated with home range size (Smith 1968, Holmes 1970, Mares et al. 1976). Large variation has been observed in bobcat home range sizes. Bailey (1974) reported bobcats in Idaho occupied ranges from 9 to 108 km , and bobcats in Alabama occupied ranges <4 km2 (Miller 1980). Buie (1980) observed a 4-fold increase in home range size in South Carolina 13 years after Marshall and Jenkins (1967) studied bobcats in the same area. Buie suggested that changes in land-use patterns caused declines in prey populations and resulted in decreased bobcat density and increased home range size. Snowshoe hares are the major prey of bobcats in Maine (May 1981, Major 1983). Therefore, bobcat habitat use within home ranges should be influenced by the distribution of hares. Bobcats occurring in areas with relatively dense hare populations may occupy smaller ranges than bobcat in areas with low hare densities. The objective of our study was to determine the relati nship between bobcat habitat use and home range s ze and snowshoe hare distribution and abundance in Maine. This research was supported by the Maine Coop. Wildl. Res. Unit, U.S. Fish and Wildl. Serv., Maine Dep. Inland Fish. and Wildl., Wildl. Manage. Inst., and the Univ. Maine, cooperating. A. G. Clark, J. H. Hunt, and L. Perry made valuable contributions in executing and funding this study. We thank students at the Coll. For. Resour., Univ. Maine, and Unity Coll. that provided field assistance. We also thank J. W. Ault III; J. R. Gilbert; G. L. Jacobson, Jr.; D. M. Leslie, Jr.; R. B. Owen, Jr.; and R. E. Rolley for reviewing this report. This is contribution 1078 from Maine Agric. Exp. Stn.

Coyote Movements, Habitat Use, and Food Habits in Southwestern Oklahoma

Eighteen radio-equipped coyotes (Canis latrans) were found to have home ranges averaging 68.7 km2 for adult females, 31.3 km2 for adult males, 39.9 km2 for yearling females, and 1.0 km2 for pups. Coyotes involved in pup rearing occupied smaller home ranges than did unmated coyotes. Home ranges of adult females overlapped, as did adjacent male-female home ranges. Coyotes were located more in savanna and less in prairie than expected (P < 0.05) by the availability of these habitats. Seventy-eight percent of the observations of coyotes were of lone animals, 17% were of pairs, and less than 6% were of more than 2 coyotes. Analysis of coyote scats indicated rodents, fruits and seeds, and fawn deer (Odocoileus virginianus) were important foods. J. WILDL. MANAGE. 44(1):62-68 This study was initiated to investigate patterns of coyote movements and food habits on the southern Great Plains. Specific objectives were to (1) determine coyote home ranges, (2) determine coyote group size, (3) describe habitat use, and (4) estimate food habits of coyotes. The study was conducted on the 23,917ha Wichita Mountains National Wildlife Refuge (WMNWR) in southwestern Oklahoma. The WMNWR supports bison (Bison bison), longhorn cattle, elk (Cervus canadensis), and white-tailed deer. The refuge contains 12,505 ha of woodland, 8,547 ha of grassland, and 405 ha of marsh and water. Detailed descriptions of the floral communities were provided by Buck (1964) and Crockett (1964). We thank J. A. Bissonette, F. L. Knopf, and P. A. Vohs, Jr. for comments during preparation of this manuscript. Assistance in the field by D. Martin, D. L. Peters, and S. Pratt is acknowledged. Special thanks are due L. Crabb, D. Snay, C. Harrison, and K. Podborney of the Division of Animal Damage Control, U.S. Fish and Wildlife Service, for their assistance in trapping coyotes. G. Stout, Fish a d Wildlife Biologist, Fort Sill, is also acknowl dged for his assistance throughout the study. METHODS AND MATERIALS Coyotes were captured with steel leghold traps that had offset, padded jaws, 2-pronged drags, and a tranquilizer tab (Basler 1965) containing 80 mg of Tranvet (Diamond Laboratories, Des Moines, Iowa). (Mention of a manufacturer does not indicate endorsement by the authors or cooperating agencies.) Traps were set where coyote scats and tracks were most evident. Captured coyotes were weighed, sexed, and aged (yearling or adult). Coyotes were then ear-tagged and fitted with radio-transmitter collars (Wildlife Materials, Inc., Carbondale, Ill.) and released at capture sites. Coyote pups were captured by hand and age was determined by weight and development (Gier 1968, Bekoff and Jamieson 1975). Those estimated to be 6 weeks of age or older were ear-tagged and fitted with radio-transmitters having expandable collars. Most relocations were obtained through 1 This study was supported by Pittman-Robertson Project W-122-R; the Oklahoma Cooperative Wildlife Research Unit, U.S. Fish and Wildlife Service, Oklahoma Department of Wildlife Conservation, Wildlife Management Institute and Oklahoma State University, cooperating. 2 Present address: U.S. Fish and Wildlife Service, 333 W. 4th St., Tulsa, OK 74103. 62 J. Wildl. Manage. 44(1):1980 This content downloaded from 157.55.39.246 on Tue, 27 Sep 2016 05:18:40 UTC All use subject to http://about.jstor.org/terms COYOTES IN SOUTHWESTERN OKLAHOMA * Litvaitis and Shaw 63 triangulation using portable receivers (AVM Corp., Champaign, Ill.) and handheld 4-element Yagi antennas. At least 2 signal azimuths were taken from known locations and later plotted on aerial photographs (scale = 1:7,920). The accuracy of this technique was estimated to be +-50 through placement and subsequent relocation of transmitters in the field. Once a visual estimation was made of the coyotes location, an effort was made to take signal azimuths within 1 km of the coyote to minimize error. Most relocations were taken from May through August, 1976 and 1977, from 0600-2000 CST. Daily movements were evaluated by plotting straight-line distances between consecutive daily locations and monitoring at 2.5-hour intervals for 24hour periods. The capture site, all relocations, and the carcass or transmitter recovery site were plotted for each coyote, and home range boundaries were determined by the minimum area method (Mohr 1947). Habitat use was measured by comparing the number of relocations in each habitat to the proportion of that habitat found within coyote home ranges. Habitats included prairie, prairie creek, savanna, and woodland (closed-canopy forest). The percentage of each habitat within a home range was determined using a modified acreage grid (Mosby 1971) and aerial photographs (scale = 1:20,000). Habitat preference was expressed as a percentage of the expected value: [(Obs. Exp.)/Exp.] x 100, with positive (+) and negative (-) values indicating departures of observed from expected values (Follmann 1973). Coyote group sizes were determined through direct observations from hilltops during morning and evening, and from chance observations during travel through the study area. Coyote food habits were determined through analysis of scats collected biweekly and dried in a desiccating oven prior to storage. Food items were identified through comparison with a reference collection and were reported by percentage occurrence. Relative availability of various prey groups was estimated by several methods. Small rodent abundance was sampled bimonthly using the North American Standard method (Calhoun 1948) consisting of 3 snap-traps placed within a 1.5-m radius at 15.4-m intervals along a 292-m transect. Traps were baited with a peanut butter, raisin, and oat mixture and rebaited for 3 consecutive evenings per trapping period. One transect passed through savanna and woodland and another was located exclusively in prairie. Lagomorph population indices were derived from roadside counts (Lord 1961) begun at approximately V2 hour before sunrise or 1 hour after sunset along a 32km route. The route was driven biweekly at 32 km/hour. Availability of whitetailed deer fawns was estimated from a concurrent study of fawn mortality (Bartush 1978). Data were evaluated using t-tests and significance was assigned at P < 0.05.

An Approach Toward Understanding Wildlife-Vehicle Collisions

Among the most conspicuous environmental effects of roads are vehicle-related mortalities of wildlife. Research to understand the factors that contribute to wildlife-vehicle collisions can be partitioned into several major themes, including (i) characteristics associated with roadkill hot spots, (ii) identification of road-density thresholds that limit wildlife populations, and (iii) species-specific models of vehicle collision rates that incorporate information on roads (e.g., proximity, width, and traffic volume) and animal movements. We suggest that collision models offer substantial opportunities to understand the effects of roads on a diverse suite of species. We conducted simulations using collision models and information on Blanding’s turtles (Emydoidea blandingii), bobcats (Lynx rufus), and moose (Alces alces), species endemic to the northeastern United States that are of particular concern relative to collisions with vehicles. Results revealed important species-specific differences, with traffic volume and rate of movement by candidate species having the greatest influence on collision rates. We recommend that future efforts to reduce wildlife-vehicle collisions be more proactive and suggest the following protocol. For species that pose hazards to drivers (e.g., ungulates), identify collision hot spots and implement suitable mitigation to redirect animal movements (e.g., underpasses, fencing, and habitat modification), reduce populations of problematic game species via hunting, or modify driver behavior (e.g., dynamic signage that warns drivers when animals are near roads). Next, identify those species that are likely to experience additive (as opposed to compensatory) mortality from vehicle collisions and rank them according to vulnerability to extirpation. Then combine information on the distribution of at-risk species with information on existing road networks to identify areas where immediate actions are warranted.

Niche dimensions of New England cottontails in relation to habitat patch size

We examined physical condition, niche dimensions, and survival of New England cottontails (Sylvilagus transitionalis) that occupied 21 habitat patches of different sizes during winter. Rabbits on small patches (≤2.5 ha) were predominantly males, and both sexes had lower body mass than individuals on large patches (≥5.0 ha). Niche indices (β, where β ranges from 0 to 1. and values approaching 1 indicate generalized resource use) of habitat use revealed that rabbits on small patches used a greater variety of microhabitats (based on understory stem density: βs, and proximity to cover: βc) than rabbits occupying large patches (βs=0.65, βc=0.66). Rabbits on small patches also consumed low quality forage more often and fed at sites farther from escape cover than rabbits on large patches. There were no significant correlations between rabbit densities and niche dimensions. Niche expansion was not a result of compertitive release or relaxation of predator pressure. Rabbits on small patches apparently modified their niche dimensions in response to resource limitations. This response included occupying sites with limited understory cover that apparently resulted in rabbits on small patches having a lower survival rate (0.35) than rabbits on large patches (0.69) during a 10-week monitoring period. Skewed sex ratios and low survival rates among rabbits on small patches suggest that these habitats act as sinks to dispersing, juveniles from large (source) patches. As a result, local populations of New England cottontails may become vulnerable to extinction if larte patches of habitat are not maintained.

Community structure of snakes in a human-dominated landscape

Abstract We studied occupancy, species richness, abundance, and size distributions of snakes on habitat patches that ranged from 0.2 to 120 ha within a landscape undergoing substantial land-use changes. Additionally, we examined the role of intraguild predation in structuring snake communities by equipping five racers ( Coluber constrictor , the largest snake in our study area) with transmitters, and then compared the abundance of smaller snakes within areas of intensive (75% isopleth home range) and limited activity (outside of 75% isopleth) by racers. As expected, small patches ( Thamnophis sirtalis ). Species richness was greatest on large patches, and snakes also tended to be more abundant on large patches. Counter to our expectations, the proportion of large-bodied individuals on small patches was greater. The distribution of small-bodied snakes on one large patch may have been affected by racers because the proportion of small snakes was less in areas intensively used by racers. Additionally, the abundance of the most generalist species in our study (garter snakes) was less in areas of intensive use by racers. In the northeastern United States, populations from a variety of taxonomic groups (including insects, migratory songbirds, mammals, and snakes) are rapidly declining in response to the loss of early-successional habitats. Our study lends support to current initiatives that recommend restoration and maintenance of parcels of early-successional habitats to sustain regional biological diversity.

Are pre-Columbian conditions relevant baselines for managed forests in the northeastern United States?

Populations of a number of taxa associated with shrublands, early-successional forests, and other disturbance-generated habitats (collectively referred to as thickets) are declining in the northeastern United States. To assure that species dependent on thicket habitats persist, intervention is warranted. However, conservationists concerned with the status of thicket-dependent species are confronted with two important questions. How much habitat is needed? And how should these habitats be distributed? Natural disturbance regimes have been recommended as a baseline that managers should consider while providing thicket habitats. Within the Northeast, historic disturbance regimes varied substantially among forest types. Coastal regions were characterized by extensive barrens where regular and often times large-scale disturbances that resulted in >15% of the area being covered by regenerating forest stands. Among inland forests, natural disturbances were usually small and resulted in seedling-sapling stands and beaver (Castor canadensis) impoundments covering <6% of the area. Under these conditions, thicket-affiliated species were probably distributed in small, disjunct populations that shifted in space and time. Current efforts to maintain thicket habitats must deal with a range of current land-uses and a legacy of historic uses. Additionally, the effectiveness of management protocols that mimic natural disturbances is limited among many forests. Increasing ownership parcelization, a relatively young forest, and landscape fragmentation substantially reduce the practicality and suitability of small-scale disturbances for generating thicket habitats. Large, clustered patches may be more practical and beneficial, especially in urbanized landscapes. In rural areas, silvicultural manipulations should be applied on a “sliding scale” relative to forest age. Timber harvests that emulate the range of variability of natural disturbances may become appropriate in these areas as forest stands mature. Addressing the needs of thicket-dependent species in the northeastern United States will require creativity, a willingness to explore a variety of solutions, and public support.

TERRESTRIAL HABITAT USE BY NESTING PAINTED TURTLES IN LANDSCAPES WITH DIFFERENT LEVELS OF FRAGMENTATION

Abstract As part of a comprehensive study of the effects of landscape change on the demography of aquatic turtles, we investigated the use of upland habitats by adult female painted turtles (Chrysemys picta) during the nesting season. We intentionally selected three ponds that were surrounded by a range of land uses and development (e.g., road density ranged from 0.9 to 7.3 km/km2). Terrestrial locations of transmitter-equipped turtles (n = 8) and unmarked females that were incidentally encountered (n = 10) were recorded to: i) determine the distance traveled to nesting habitat, ii) estimate the frequency that female turtles crossed roads to reach nesting habitat, and iii) examine how the abundance of potential nesting habitat affected distance traveled. Average straight-line distance from shoreline to the site a female turtle was encountered ranged from 54 to 115 m for the three ponds. We estimated that marked turtles made five road crossings and two turtles were killed by vehicles. Distance traveled was negatively correlated with the abundance of nesting habitat near a pond. Although adult females are probably less vulnerable when nesting occurs near pond shores, nests created near pond edges may be more vulnerable to predation. As a result, providing additional nesting habitat away from roads and other potential sources of mortality may enhance recruitment among local turtle populations.

FOOD AND ENERGY USE BY CAPTIVE COYOTES

Four eastern coyotes (Canis latrans var.) were fed white-tailed deer (Odocoileus oirginianus), showshoe hare (Lepus americanus), and laboratory mice (Mus musculus) to determine their digestion of dry matter, nutrients, and partitioning of dietary gross energy. Dry matter digestibility of the deer diet (96.8%) was higher (P < 0.05) than of the hare (81.5%) or mouse (83.2%) diets. The digestible energy value of deer (5.69 kcal/g dry matter) differed (P < 0.05) from the other diets, and metabolizable energy values of the deer and mouse diets (4.99,5.07 kcal/g dry matter) were greater (P < 0.05) than that of the hare diet (4.01 kcallg dry matter). The prey required to fulfill th e minimum energy demands at the metabolizable level of a 12.9-kg coyote was estimated to be 8 deer, 105 hares, or 4,800 mice per year.