In spite of the wide use and acceptance of information theoretic approaches in the wildlife sciences, debate continues on the correct use and interpretation of Akaike's Information Criterion as compared to frequentist methods. Misunderstandings as to the fundamental nature of such comparisons continue. Here we agree with Steidl's argument about situation-specific use of each approach. However, Steidl did not make clear the distinction between statistical and biological hypotheses. Certainly model selection is not statistical, or null, hypothesis testing; importantly, it represents a more effective means to test among competing biological, or research, hypotheses. Employed correctly, it leads to superior strength of inference and reduces the risk that favorite hypotheses are uncritically accepted.

Statistical inference is an important element of science, but these inferences are constrained within the framework established by the objectives and design of a study. The choice of approach to data analysis, while important, has far less consequence on scientific inference than claimed by Sleep et al. (2007). Their principal assertion—that when model selection is used as the approach to data analysis, all studies provide a reliable foundation for distinguishing among mechanistic explanatory hypotheses—is incorrect and encourages faulty inferences. Sleep et al. (2007) overlook the critical distinction between inferences that result from studies designed a priori to discriminate among a set of candidate explanations versus inferences that result from exploring data post hoc from studies designed originally to meet pattern-based objectives. No approach to data analysis, including model selection, has the power to overcome fundamental limitations on inferences imposed by study design. The comments by Sleep et al. (2007) reinforce the need for scientists to understand clearly the inferential basis for their scientific claims, including the roles and limitations of data analysis.

The scientific evidence that California condors (Gymnogyps californianus) are frequently sickened and killed by lead poisoning from spent ammunition supports the conclusion that current levels of lead exposure are too high to allow reintroduced condors to develop self-sustaining populations in the wild in Arizona and, by inference, in California. The evidence for lead poisoning and its source comes from the following sorts of data: 1) 18 clinical necropsies revealing high levels of lead in body tissues and (or) presence of lead shotgun pellets and bullet fragments in digestive tracts; 2) moribund condors showing crop paralysis and impending starvation with toxic levels of lead in their blood; 3) widespread lead exposure among free-flying condors, many with clinically exposed or acute levels; 4) temporal and spatial correlations between big game hunting seasons and elevated lead levels in condors; and 5) lead isotope ratios from exposed condors showing close similarity to isotope ratios of ammunition lead but isotope ratios in less exposed condors being similar to environmental background sources, which are different from ammunition lead. Simple population models reveal harmful demographic impacts of unnatural mortality from lead on population trajectories of reintroduced condors. Recent innovations in the manufacture of nonlead shotgun pellets and bullets with superior ballistics now provide for a simple solution to the problem of lead ingestion by condors, many other species of wildlife, and human beings: substitute nontoxic forms of ammunition for traditional lead-based ammunition. The substitution of nontoxic ammunition would be highly efficacious for hunting, economically feasible, and the right thing to do.

In the natural resource and wildlife profession, we face difficulties in the production, diffusion, and transfer of rigorously tested science, especially when facing entrenched management paradigms. We present 3 case studies to illustrate the challenges in changing entrenched management paradigms for endangered species. Here we examine specifically what factors helped or hindered the adoption of management practices through the theoretical framework developed for the dissemination of technologies. An examination of 3 case studies suggests that active communication and advocacy of scientific findings, along with simple, visible results, will aid researchers in the acceptance and adoption of their research. Management agencies that increase openness and communication with outside experts, reduce bureaucratic procedures, and localize decision making increase the likelihood that new scientific ideas will be adopted by the agency. We also suggest adaptive resource management as a strategy for endangered species management may foster many of the characteristics that aid in the adoption of scientific ideas into management activities.

The controversy over the use of null hypothesis statistical testing (NHST) has persisted for decades, yet NHST remains the most widely used statistical approach in wildlife sciences and ecology. A disconnect exists between those opposing NHST and many wildlife scientists and ecologists who conduct and publish research. This disconnect causes confusion and frustration on the part of students. We, as students, offer our perspective on how this issue may be addressed. Our objective is to encourage academic institutions and advisors of undergraduate and graduate students to introduce students to various statistical approaches so we can make well-informed decisions on the appropriate use of statistical tools in wildlife and ecological research projects. We propose an academic course that introduces students to various statistical approaches (e.g., Bayesian, frequentist, Fisherian, information theory) to build a foundation for critical thinking in applying statistics. We encourage academic advisors to become familiar with the statistical approaches available to wildlife scientists and ecologists and thus decrease bias towards one approach. Null hypothesis statistical testing is likely to persist as the most common statistical analysis tool in wildlife science until academic institutions and student advisors change their approach and emphasize a wider range of statistical methods.

This Special Section includes 8 peer-reviewed papers on the wolverine (Gulo gulo) in southern North America. These papers provide new information on current and historical distribution, habitat relations at multiple spatial scales, and interactions with humans. In aggregate, these papers substantially increase our knowledge of wolverine ecology and population dynamics in North America, in many cases replacing previous speculations and informed judgments with empirical information. North American wolverines occur primarily in tundra, taiga, and subalpine environments. These environments become increasingly fragmented at southern latitudes, where wolverine populations occur at low densities and are potentially vulnerable to human-caused mortality. The combination of highly fragmented habitat, demographic sensitivity to adult mortality, and low population densities make local wolverine populations difficult to monitor and easy to overharvest. Where populations are fragmented, persistence is critically dependent on dispersal between habitat islands. Although dispersal dynamics are poorly understood, high levels of genetic structure observed in both current and historical populations indicate that dispersal between mountain ranges is limited. Wolverine biology remains poorly understood, and many fundamental issues need additional research.

Conservation of the wolverine (Gulo gulo) at the southern extent of its North American range requires reliable understandings of past and present distribution patterns and broad-scale habitat relations. We compiled 820 verifiable and documented records of wolverine occurrence (specimens, DNA detections, photos, and accounts of wolverines being killed or captured) in the contiguous United States from museums, the literature, and institutional archives. We spatially referenced 729 records with areal precision ≤1 township (93.2 km²) and temporal precision ≤10 years. Historical records (1827–1960) were located primarily in the western mountains and Great Lakes region. However, our data suggest that the historical distribution of wolverines in the Cascade Range and Sierra Nevada was disjunct, contradicting previous interpretations. Our results indicate that wolverine range in the contiguous United States had contracted substantially by the mid-1900s. Current records (1995–2005) are limited to north-central Washington, northern and central Idaho, western Montana, and northwestern Wyoming. We investigated potential relations between wolverines and alpine vegetation, cold temperatures, and spring snow cover by comparing the distribution of historical wolverine records with Kuchler's potential natural vegetation types, Holdridge's climatic life zones, and EASE snow-cover maps during the latter portion of the wolverine denning period (15 Apr–14 May). In the western mountains, historical wolverine records generally occurred in or near alpine vegetation and climatic conditions, especially at the limits of their distribution in the Cascade Range, Sierra Nevada, and southern Rocky Mountains. However, the only habitat layer that fully accounted for historical distribution patterns was spring snow cover. Causal factors for the extirpation of wolverines from the southern portions of their range in the contiguous United States are unknown, but are likely related to high levels of human-caused mortality and low to nonexistent immigration rates.

The abundance and distribution of carnivores and their habitat are key information needed for status assessment, conservation planning, population management, and assessment of the effects of human development on their habitat and populations. We developed a habitat quality rating system, using existing wolverine (Gulo gulo) distribution, wolverine food, ecosystem mapping, and human development data. We used this and empirically derived estimates of wolverine density to predict wolverine distribution and abundance at a provincial scale. Density estimates for wolverines in high-quality habitat averaged 6.2 wolverines/1,000 km² (95% CI = 4.2–9.5). We predicted mean densities ranging from 0.3/1,000 km² in rare-quality habitat to 4.1/1,000 km² in moderate-quality habitat. Our predicted population estimate for wolverines in British Columbia was 3,530 (95% CI = 2,700–4,760). We predicted highest densities of wolverines in interior mountainous regions, moderate densities in interior plateau and boreal forest regions, and low densities in mainland coastal regions and drier interior plateaus. We predicted that wolverines would be rare on Vancouver Island, along the outer mainland coast, and in the dry interior forests, and absent from the Queen Charlotte Islands, interior grassland environments, and areas of intensive urban development.

Delineating a species' geographic range using the spatial distribution of museum specimens or even contemporary detection–non-detection data can be difficult. This is particularly true at the periphery of a species range where species' distributions are often disjunct. Wolverines (Gulo gulo) are wide-ranging mammals with discontinuous and potentially isolated populations at the periphery of their range. One potentially disjunct population occurred in the Sierra Nevada Mountains, California, USA, and appears to have been extirpated by the 1930s. Many early 20th century naturalists believed that this population was connected to other populations occurring in the Cascade Range of northern California, Oregon, and Washington, USA, but a recent analysis of historical records suggests that California wolverines were isolated from other populations in North America. We used DNA extracted from museum specimens to examine whether California wolverines were isolated. Both nuclear and mitochondrial DNA data indicate that California wolverines were genetically distinct from extant populations, suggesting long-term isolation. We identified 2 new control region (mitochondrial DNA) haplotypes located only within California. We used these data and referenced sequences from the Rocky Mountains, USA, to make inferences regarding potential wolverine translocations into California. In addition, we used these genetic data to make inferences about wolverine conservation throughout western North America.

Wolverine (Gulo gulo) distribution in British Columbia, Canada, includes multiple-use lands where human use and resource extraction may influence habitat selection. We evaluated seasonal habitat use by resident adult wolverines using radiotelemetry locations from 2 multiple-use landscapes in British Columbia. Food, predation risk, and human disturbance hypotheses were considered in logistic regression analyses of used and random landscapes. Male wolverine habitat associations were most supported by the food hypothesis in both summer and winter. Moose (Alces alces) winter ranges, valley bottom forests, and avalanche terrain were positively associated with winter male wolverine use. Habitat use by male wolverines in winter was also negatively associated with helicopter skiing areas in the Columbia Mountains. Habitat associations of females were more complex; combinations of variables supporting food, predation risk, or human disturbance hypotheses were included in most supported models from both summer and winter in both study areas. Females were associated with alpine and avalanche environments where hoary marmot (Marmota caligata) and Columbia ground squirrel (Spermophilus columbianus) prey are found in summer. Roaded and recently logged areas were negatively associated with female wolverines in summer. In the Columbia Mountains, where winter recreation was widespread, females were negatively associated with helicopter and backcountry skiing. Moose winter ranges within rugged landscapes were positively associated with females during winter. Our analysis suggests wolverines were negatively responding to human disturbance within occupied habitat. The population consequences of these functional habitat relationships will require additional focused research. Our spatially explicit models can be used to support conservation planning for resource extraction and tourism industries operating in landscapes occupied by wolverines.

Wolverines (Gulo gulo) are distributed across much of northern and western Canada and Alaska, USA, and they extend south into the mountainous western United States. Wolverines occur in most regions of British Columbia, Canada, with the highest population densities occurring in the interior mountainous areas. Wolverine populations in British Columbia have been primarily managed to provide a sustainable harvest for trappers and hunters. We used spatially based population estimates, population vital rate data, and spatially based harvest data to evaluate the sustainability of wolverine harvest (trapping and hunting) from 1985 to 2004. The median annual provincial wolverine harvest from 1985 to 2004 was 172 wolverines per year (𝑥̄ = 174.8), which was less than the median simulated estimate of provincial recruitment (195.9 wolverines/yr; 𝑥̄ = 209.7). Harvests in individual population units ranged from 0 to 280 over the 20-year period. Spatially, wolverine harvest was likely to have been unsustainable in 15 of the 71 population units with wolverines, and it was likely to have been sustainable in the remaining population units. Harvest in 5 of the other 56 population units was marginally sustainable and thus of potential management concern. To improve harvest management of wolverines in British Columbia, wildlife managers should focus on improved data collection and monitoring at a provincial scale, and they should work with trappers and hunters at regional scales to address issues specific to individual population units. Further research is required to improve the reliability of wolverine vital rate and population data.

Although understanding habitat relationships remains fundamental to guiding wildlife management, these basic prerequisites remain vague and largely unstudied for the wolverine. Currently, a study of wolverine ecology conducted in Montana, USA, in the 1970s is the sole source of information on habitat requirements of wolverines in the conterminous United States. The Montana study and studies conducted in Canada and Alaska report varying degrees of seasonal differences in wolverine habitat use. This article provides an empirical assessment of seasonal wolverine habitat use by 15 wolverines (Gulo gulo) radiotracked in central Idaho, USA, in 1992–1996. We controlled for radiotelemetry error by describing the probability of each location being in a habitat cover type, producing a vector of cover type probabilities suited for resource selection analysis within a logistic regression framework. We identified variables that were important to presence of wolverines based on their strength (significance) and consistency (variability in coeff. sign) across all possible logistic regression models containing 9 habitat cover types and 3 topographic variables. We selected seasonal habitat models that incorporated those variables that were strong and consistent, producing a subset of potential models. We then ranked the models in this subset based on Akaike's Information Criterion and goodness-of-fit. Wolverines used modestly higher elevations in summer versus winter, and they shifted use of cover types from whitebark pine (Pinus albicaulis) in summer to lower elevation Douglas fir (Pseudotsuga menziezii) and lodgepole pine (Pinus contorta) communities in winter. Elevation explained use of habitat better than any other variable in both summer and winter. Grass and shrub habitats and slope also had explanatory power. Wolverines preferred northerly aspects, had no attraction to or avoidance of trails during summer, and avoided roads and ungulate winter range. These findings improve our understanding of wolverine presence by demonstrating the importance of high-elevation subalpine habitats to central Idaho wolverines.

We instrumented 36 wolverines (Gulo gulo) on 2 study areas in western Montana and one study area on the Idaho–Montana (USA) border: 14 (9 M, 5 F) on the Pioneer study area, 19 (11 M, 8 F) on the Glacier study area, and 3 (2 M, 1 F) on the Clearwater study area. During 2002–2005, harvest from licensed trapping accounted for 9 (6 M, 3 F) of 14 mortalities, including individuals from all 3 study areas. Based on Akaike's Information Criterion adjusted for small sample sizes (AIC_c) rankings of 8 a priori models, a trapping model and a trapping-by-sex interaction model were equally supported (ΔAIC_c < 2) in explaining wolverine survival. Estimated annual survival was 0.80 when we did not consider harvest, whereas additive mortality from harvest reduced annual survival to 0.57. Glacier National Park in the Glacier study area provided some refuge as evidenced by an annual survival rate of 0.77 compared to 0.51 for the Pioneer Mountain study area. We incorporated these survival rates into a simple Lefkovitch stage-based model to examine rates of population change. The finite rate of population change (λ) for the Glacier study area was 1.1, indicating a stable to slightly increasing population, whereas λ for the Pioneer study area was 0.7, indicating a 30% annual population decrease during our study. Changes in λ for both study areas were most sensitive to adult survival. In 2004, we used a Lincoln Index to estimate that 12.8 ± 2.9 (95% CI) wolverines resided in the 4 mountain ranges comprising the Pioneer study area, suggesting that small, island ranges in western Montana supported few individuals. Our results suggest that if wolverines are harvested, they should be managed within individual mountain ranges or small groupings of mountain ranges to limit mortality to within biologically defined limits in recognition of the increased vulnerabilities owing to low fecundity and low population numbers in small mountain ranges. We found that refugia, such as Glacier National Park, were important by reducing trap mortality and providing immigrants to the surrounding population, but even large parks were inadequate to provide complete protection to wolverines from trapping as they ranged outside park borders.

We designed a novel approach to determining extent of distribution and area of occupancy for wolverines (Gulo gulo) by using aerial surveys of tracks in snow and hierarchical spatial modeling. In 2005 we used a small, fixed-wing aircraft with pilot and one observer to search 575 of 588 survey units for wolverine tracks in approximately 60,000 km² of boreal forest in northwestern Ontario, Canada. We used sinuous flight paths to scan open areas in the forest in the 100-km² survey units. We detected tracks in 138 (24%) of the 575 sampled units. There was strong evidence of occurrence (probability of occurrence >0.80) in 30% of the 588 survey units, weak evidence of occurrence (0.50–0.80) in 12%, weak evidence of absence (0.20–0.50) in 15%, and strong evidence of absence (<0.20) in 43%. Wolverine range comprised 59% of the study area and area of occupancy was 33,400 km². With information on probability of occurrence and core areas of occupation for wolverines in our study area, resource managers and others can examine factors that influence wolverine distribution patterns and use this information to formulate best management practices that will maintain wolverines on the landscape in the face of increasing resource development. Comparing future survey results with those of our 2005 survey will provide an objective way to assess the efficacy of management practices.

We assessed spatial distribution and habitat use by Ross's and lesser snow geese (Chen rossii and C. caerulescens caerulescens) during late brood rearing to begin understanding goose–habitat interactions and monitoring key habitats around a rapidly growing nesting colony located at Karrak Lake, Nunavut, Canada. We conducted aerial surveys to count geese and georeference locations, then used Landsat Thematic Mapper satellite imagery to identify habitats associated with each flock. We observed 435 and 407 flocks and 36,287 and 32,745 birds in 1994 and 1995, respectively. Birds were somewhat uniformly distributed over the 5,000-km² study area, with larger aggregations occurring closer to the coast, about 70 km from the colony. We assessed habitat use using Bonferroni intervals at both the flock and individual scales. At the flock level, birds avoided lichen–heath, used other terrestrial habitats as available, and selected freshwater. At the individual level, geese selected lowland habitats: wet sedge meadow, hummock graminoid tundra, and freshwater, which accounted for about 70% of the birds observed, and avoided upland habitats. Selection of lowland habitats is likely due to greater availability of food and easier predator avoidance compared to drier upland areas. Because most geese in our study used freshwater habitats, our results demonstrate that assessment of carrying capacity, at least in the central Arctic, must be expanded beyond the coastal salt marshes traditionally considered by researchers and managers as primary brood-rearing habitat for mid-continent light geese.

The influence of habitat, waterfowl abundance, and hunting on winter survival of waterfowl is not well understood. We studied late August–March survival of 163 after-hatch-year (AHY) and 128 hatch-year (HY) female mallards (Anas platyrhynchos) radiotagged in Sacramento Valley (SACV) and 885 AHY female northern pintails (A. acuta) radiotagged throughout the Central Valley of California, USA, relative to flooded habitat (HAB), January abundance of each species (JMAL or JPIN), hunter-days (HDY), and a hunting pressure index (HPI) that combined these variables. From EARLY (1987–1994) to LATE (1998–2000), HAB increased 39%, JPIN increased 45%, JMAL increased 53%, HDY increased 21%, duck-hunting season increased from 59 days to 100 days, and the female daily bag limit doubled to 2 for mallards but remained 1 for pintails. Survival (± SE) was greater during LATE versus EARLY for pintails radiotagged in each region (SACV: 93.2 ± 2.1% vs. 87.6 ± 3.0%; Suisun Marsh: 86.6 ± 3.2% vs. 77.0 ± 3.7%; San Joaquin Valley: 86.6 ± 3.1% vs. 76.9 ± 4.1%) but not for SACV mallards (AHY: 70.6 ± 7.2% to 74.4 ± 7.7% vs. 80.1 ± 7.2% to 82.8 ± 5.6%; HY: 48.7 ± 9.1% [1999–2000 only] vs. 63.5 ± 8.8% to 67.6 ± 8.0%). Most pintail (72%) and mallard (91%) deaths were from hunting, and lower HPI and higher JPIN or JMAL were associated with reduced mortality. Increased HAB was associated with reduced winter mortality for pintails but not for SACV mallards. Pintail survival rates that we measured were within the range reported for other North American wintering areas, and during LATE were higher than most, even though our study duration was 68–110 days longer. Winter survival rates of SACV mallards were also within the reported range. However, with higher bag limits and longer seasons, mallard survival during LATE was lower than in most other wintering areas, especially during 1999–2000, when high winds on opening weekend resulted in high hunting mortality. Habitat conservation and favorable agriculture practices helped create a Central Valley wintering environment where natural mortality of mallards and pintails was low and survival varied with hunting mortality. We recommend regulations and habitat management that continue to minimize natural mortality while allowing sustainable harvest at a level that helps maintain strong incentive for management of Central Valley waterfowl habitats, including the large portion that is privately owned.

Habitat management for ducks has significantly influenced prairies and wetlands used by other species. However, the effects of management on other species have not been clearly assessed. We present the first study to compare the nesting success of ducks with the productivity of coexisting passerines. We evaluated effects of cattle grazing, subdivision of fields, habitat edges, year, and vegetation structure on duck and songbird nesting success in 32 mixed-grass prairie fields in southern Alberta, Canada. Duck and songbird nesting success were not correlated. Duck nesting success was influenced by timing of grazing and vegetation structure, and tended to be higher in wetter years, whereas nesting success of most songbirds was not influenced by vegetation structure or grazing, and was sometimes higher in drier years. Local habitat management for ducks cannot be assumed to benefit songbirds. However, some management strategies, such as those that promote tall grass and short litter, might benefit both taxa.

The North American Breeding Bird Survey (BBS) is used extensively to make inferences about populations of many North American bird species and is increasingly being used for avian conservation planning. How well BBS routes represent the landscape is poorly known, even though accuracy of representation could significantly affect inferences made from BBS data. We used digital landcover data to examine how well landcover within 400-m buffers around BBS routes represented the surrounding landscape (the route neighborhood) for 52 routes in the Prairie Pothole Region of North Dakota and South Dakota. Differences in composition between landcover along BBS routes and the route neighborhood were not statistically significant for upland cover classes. The area of temporary and seasonal wetland basins was accurately represented by BBS routes in our study area, but the area of semipermanent and permanent wetland basins was significantly underrepresented along BBS routes. Number of wetland basins and upland patches was higher along routes. Area of urban, forest, and hay landcover classes was higher along routes, although differences were not statistically significant. Amount of bias in landcover representation was negatively correlated with the proportion of each landcover type in the study area, but bias was not correlated with area of the route neighborhoods. Differences between landcover along BBS routes and the route neighborhood were primarily attributable to increased anthropogenic activity along roads and siting of roads away from relatively large, deep water bodies. Our results suggest that inferences made from BBS data in our study region are likely biased for species that are associated with deeper-water habitats or are strongly influenced by landscape fragmentation. Inferences made from BBS data for species associated with uplands or shallow wetlands are less likely to be biased because of differences in landcover composition.

Roadside survey data have been used frequently to assess species occurrence and population trends and to establish conservation priorities. However, most studies using such data assume that samples are representative of either the amount of habitat or its rate of change at larger spatial scales. We tested both of these assumptions for the Breeding Bird Survey (BBS) from 1974 to 2001 in New Brunswick, Canada. Our study focused on mature forest—a cover type that we predicted would be characterized by rapid change due to human activities and that is of high ecological importance. We also sought to determine whether land cover changes adjacent to BBS routes were related to bird population trends detected in BBS data. Within all 3 time periods examined (1970s, 1980s, and 1990s), the amount of mature forest adjacent to BBS routes was significantly lower than in surrounding 1° blocks of latitude and longitude. This could be problematic for studies that use roadside data to compare the relative abundance of species. On average, mature forest declined at a rate of −1.5% per year over the 28-year study period. We detected no significant difference in the rate of change between degree blocks and BBS routes over this time span. However, in the 1970s and 1980s, mature forest declined more rapidly in degree blocks (−2.7%/yr) than adjacent to BBS routes (−0.5/yr). We also found that the BBS trend for a mature forest-associated species, blackburnian warbler (Dendroica fusca), was correlated with the trend in mature forest along BBS routes. This, combined with slower rates of mature forest change along routes in the 1970s and 1980s, suggests that BBS data may have underestimated population declines during this period. It is important that research be conducted to test for potential biases in roadside surveys caused by uneven rates of landscape change, particularly in regions characterized by rapid habitat alteration.

Developing comprehensive conservation strategies requires knowledge of factors influencing population growth and persistence. Although variable prey resources are often associated with fluctuations in raptor demographic parameters, the mechanisms of food limitation are poorly understood, especially for a generalist predator like the northern goshawk (Accipiter gentilis). To determine the reproductive responses of goshawks to variable prey populations, we evaluated 823 goshawk breeding opportunities on the Kaibab Plateau, Arizona, USA, during 1994–2002. Concurrently, density was estimated for 4 prey species (2 avian, 2 mammalian). We explored the relationship between goshawk reproduction and prey density at one temporal scale (year) and 2 spatial scales (study area, forest type). Prey density for all 4 species combined accounted for 89% of the variation in goshawk reproduction within the entire study area (P < 0.001), 74% in mixed conifer forest (P = 0.003) and 85% in ponderosa pine (Pinus ponderosa) forest (P < 0.001). We found that an incremental increase in prey density resulted in a greater increase in goshawk reproduction in ponderosa pine forest than in mixed conifer forest, suggesting that the denser structural conditions of mixed conifer forest may have reduced prey availability. Red squirrel (Tamiasciurus hudsonicus) density explained more annual variation in goshawk reproduction within the study area (r² = 0.87, P < 0.001), mixed conifer forest (r² = 0.80, P = 0.001), and ponderosa pine forest (r² = 0.85, P < 0.001) than did any other individual species. Although certain prey species were more strongly correlated with fluctuations in goshawk reproduction than were others, the high model selection uncertainty and the strong relationship between total prey density and number of goshawk fledglings produced indicated that alternate prey species were readily substituted for one another. Therefore, conservation strategies concerned with the status of goshawk populations should incorporate forest management practices that increase the abundance, diversity, and availability of prey resources.

Population decline of the American alligator (Alligator mississippiensis) was observed in Lake Apopka in central Florida, USA, in the early 1980s. This decline was thought to result from adult mortality and nest failure caused by anthropogenic increases in sediment loads, nutrients, and contaminants. Reproductive impairment also was reported. Extensive restoration of marshes associated with Lake Apopka has been conducted, as well as some limited restoration measures on the lake. Monitoring by the Florida Fish and Wildlife Conservation Commission (FFWCC) has indicated that the adult alligator population began increasing in the early 1990s. We expected that the previously reported high proportion of complete nest failure (θ₀) during the 1980s may have decreased. We collected clutches from alligator nests in Lake Apopka from 1983 to 2003 and from 5 reference areas from 1988 to 1991, and we artificially incubated them. We used a Bayesian framework with Gibbs sampler of Markov chain Monte Carlo simulation to analyze θ₀. Estimated θ₀ was consistently higher in Lake Apopka compared with reference areas, and the difference in θ₀ ranged from 0.19 to 0.56. We conducted change point analysis to identify and test the significance of the change point in θ₀ in Lake Apopka between 1983 and 2003, indicating the point of reproductive recovery. The estimated Bayes factor strongly supported the single change point hypothesis against the no change point hypothesis. The major downward shift in θ₀ probably occurred in the mid-1990s, approximately a generation after the major population decline in the 1980s. Furthermore, estimated θ₀ values after the change point (0.21) were comparable with those of reference areas (0.07–0.31). These results combined with the monitoring by FFWCC seem to suggest that anthropogenic habitat degradation caused reproductive impairment of adult females and decreases in θ₀ occurred with the sexual maturity of a new generation of breeding females. Long-term monitoring is essential to understand population changes due to habitat restoration. Such information can be used as an input in planning and evaluating restoration activities.

Sea turtle embryo mortality in natural nests due to environmental and anthropogenic factors can be very high. To increase hatching success of these endangered species, nest translocation to hatcheries immediately after egg-laying is a common management tool. To test the viability of delayed translocation, we moved 50 loggerhead  sea turtle (Caretta caretta) nests to a beach hatchery after various times (0–96 hr) after egg-laying at Boavista Island (Republic of Cabo Verde, western Africa). We transported eggs in a rigid plastic container, being careful to maintain their original vertical orientation. Delayed translocation times of 0 hours, 12 hours, 24 hours, 84 hours, or 96 hours after egg-laying did not have any effect on hatching success, incubation period, or hatchling size and mass. Delayed translocation slightly increased the duration of the translocation process because of extra precautions taken (e.g., maintaining axial orientation, protecting eggs from mechanical shocks). We conclude that delayed nest translocation can be done in a safe and effective way, thereby increasing the efficiency of the whole monitoring program. Finally, delayed translocation, accompanied by an evaluation of fertility, would seem to permit the removal of undeveloped eggs and to facilitate their subsequent exploitation by local communities without affecting turtle nesting success.

Downing population reconstruction uses harvest-by-age data and backward addition of cohorts to estimate minimum population size over time. Although this technique is currently being used for management of black bear (Ursus americanus) and white-tailed deer (Odocoileus virginianus) populations, it had not undergone a rigorous evaluation of accuracy. We used computer simulations to evaluate the impacts of collapsing age classes and violating the assumptions of this technique on population reconstruction estimates and trends. Changes in harvest rate or survival over time affected accuracy of reconstructed population estimates and trends. The technique was quite robust to collapsing age classes as far as 3 for bears and deer. This method would be suitable for estimating population growth rate (λ) for populations experiencing no trend in harvest rate or natural mortality rate over time. Our evaluation showed Downing population reconstruction to be a potentially valuable tool for managing harvested species with high harvest rates and low natural mortality, with possible application to black bear and white-tailed deer populations.

Wildlife–vehicle collisions cause numerous human fatalities and injuries, and generate considerable expenses in property damage each year. Certain characteristics of the road and its surroundings are known to have an impact on collision probability. Roadside salt pools increase the risk of collision by attracting moose (Alces alces) to the side of the road. In the Laurentides Wildlife Reserve of Québec, Canada, roadside salt pools were drained and filled with rocks to deter moose from drinking. We surveyed 12 roadside salt pools during 3 consecutive summers (2003–2005) from mid-May to mid-August. Seven salt pools were managed in autumn 2004, and 5 pools were left untreated. We equipped all 12 sites with electronic apparatus that allowed us to detect moose attendance and study their behavior. We also measured physical, chemical, and environmental characteristics of these pools and other unvisited pools in order to correlate moose attendance with specific habitat criteria. We found that moose mostly attended roadside salt pools from mid-June to mid-July, with a decrease in August. Moose attendance was significantly correlated with visual obstruction toward the road and water availability. Management of the pools caused a decrease in mean length of time moose spent at them. Number of visits decreased significantly at night (by 90%), which was when most visits occurred, but not during the day. The proposed management practice prevented all visiting moose from drinking brackish water. These results suggest that moose should eventually lose interest in treated salt pools, therefore decreasing the risk of moose–vehicle collisions on the road.

Trapping brown treesnakes (Boiga irregularis; BTS) with live-mouse (Mus domesticus) lures is the principal control technique for this invasive species on Guam. Lure-based trapping is also used on other islands as a precaution against undetected arrivals and in response to verified BTS sightings. However, the effectiveness of lure-based trapping on other islands is questionable, as it has yielded no BTS despite other evidence of their presence. Some evidence suggests that high rodent numbers may interfere with BTS control. To test the relationship between rodent abundance and snake trappability, we conducted a controlled, replicated field experiment incorporating a rodenticide treatment during a BTS mark–recapture study. Using open population modeling in Program MARK, we estimated BTS apparent survival and recapture probabilities. Rodent reduction increased BTS recapture probabilities by 52–65% in 2002 and 22–36% in 2003, and it decreased apparent survival by <1% both years. This appears to be the first published instance of manipulating wild prey to influence snake behavior. Rodent reduction may enhance detection and control of BTS with traps on Guam and other islands. It may also amplify the effectiveness of oral toxicants against BTS.

We used 38,709 fixes collected from December 2003 through June 2006 from 44 elk (Cervus elaphus) fitted with Global Positioning System collars and hourly traffic data recorded along 27 km of highway in central Arizona, USA, to determine how traffic volume affected elk distribution and highway crossings. The probability of elk occurring near the highway decreased with increasing traffic volume, indicating that elk used habitat near the highway primarily when traffic volumes were low (<100 vehicles/hr). We used multiple logistic regression followed by model selection using Akaike's Information Criterion to identify factors influencing probability of elk crossings. We found that increasing traffic rates reduced the overall probability of highway crossing, but this effect depended on both season and the proximity of riparian meadow habitat. Elk crossed highways at higher traffic volumes when accessing high quality foraging areas. Our results indicate that 1) managers assessing habitat quality for elk in areas with high traffic-volume highways should consider that habitat near highways may be utilized at low traffic volumes, 2) in areas where highways potentially act as barriers to elk movement, increasing traffic volume decreases the probability of highway crossings, but the magnitude of this effect depends on both season and proximity of important resources, and 3) because some highway crossings still occurred at the high traffic volumes we recorded, increasing traffic alone will not prevent elk–vehicle collisions. Managers concerned with elk–vehicle collisions could increase the effectiveness of wildlife crossing structures by placing them near important resources, such as riparian meadow habitat.

We used video surveillance at 4 wildlife underpasses along 27 km of Arizona State Route 260, USA, to monitor elk (Cervus elaphus) responses to traffic volume and traffic type during underpass use. Passage rates at the highest traffic category (>10–27 vehicles/min) were not lower than passage rates when no vehicles were present, whereas passage rates at low, intermittent traffic volume (>0–1 vehicles/min) were 15% lower. Once elk entered an underpass, semi-trailer trucks were 4 times more likely than passenger vehicles to cause flight behavior when traffic levels were intermittent versus when traffic was continuous. Overall, traffic volumes of >10–27 vehicles per minute did not decrease the effectiveness of wildlife underpasses as a means of mitigating elk population subdivision. However, if flight away from underpasses at intermediate traffic levels causes elk to cross the highway at other points and thereby increases the potential for costly elk–vehicle collisions, we recommend that managers consider measures to reduce traffic noise and visual stimuli, especially those caused by semi-trailer trucks.

Bighorn sheep (Ovis canadensis) use of Sheep Lakes mineral site, Rocky Mountain National Park, Colorado, USA, has decreased since 1996. Officials were concerned that human disturbance may have been contributing to this decline in use. We evaluated effects of vehicular traffic and other road-related disturbance on bighorn use of Sheep Lakes in the summers of 2002 and 2003. We found that the time and number of attempts required by bighorn to reach Sheep Lakes was positively related to the number of vehicles and people present at Sheep Lakes. Further, the number of bighorn individuals and groups attempting to visit Sheep Lakes were negatively affected by disturbance associated with the site. The number of vehicles recorded the hour before bighorn tried to access Sheep Lakes best predicted an animal's failure to cross Fall River Road and reach Sheep Lakes. We conclude that human and road-related disturbance at Sheep Lakes negatively affected bighorn use of the mineral site. Because Sheep Lakes may be important for bighorn sheep, especially for lamb production and survival, the negative influence of disturbance may compromise health and productivity of the Mummy Range bighorn sheep.

We developed and validated a density-adjusted spatial model to predict moose (Alces alces) highway-crossing probability to see if the model could be used as an index of moose–vehicle collision risk. We installed Global Positioning System telemetry collars on 47 moose in the north of the Laurentides Wildlife Reserve, Québec, for 2–36 months. We recorded only 84 highway crossings in spring (0.29% of 28,967 2-hr steps) and 122 crossings in summer (0.18% of 68,337 2-hr steps), despite a high sampling effort and having captured moose close to highways. Moose movement rates during movement steps crossing a highway were on average 3 times higher than during the steps preceding or following highway crossing. Paths used by moose when crossing a highway were characterized by a high proportion of food stands, low proportion of lakes and rivers, and topography typical of a valley. Highway-crossing sites were located in valleys with brackish pools and forest stands providing coniferous cover but a low proportion of lakes and rivers. We adjusted moose crossing probability for local variation in moose density using aerial survey data and assessed crossing probability along the highways in the entire Laurentides Wildlife Reserve. We tested the model using moose–vehicle accident data from 1990 to 2002. The relationship between the density-adjusted crossing probability and number of accidents was relatively loose at the 1-km scale but improved markedly when using longer highway sections (5–15 km; r > 0.80). Our results demonstrate that roads and their surroundings are perceived as low-quality habitat by moose. We also conclude that road segments installed along secondary valleys could be a highly strategic site to deploy mitigation measures such as fences and that it could be desirable to increase the width of road shoulders to reduce forest cover and to eliminate brackish pools to reduce cervid–vehicle collisions. We suggest using empirical data such as location of vehicle–wildlife collisions to plan mitigation measures at a fine scale.

Overabundant elk (Cervus elaphus) populations have become a significant problem in many areas of North America. This is particularly true for protected areas where high densities of elk can cause long-term ecological degradation. When lethal control is not acceptable in these environments, resource managers must often consider alternative methods for reducing the size of resident elk populations. A potential management alternative is controlling the fertility of female elk. A promising new approach to wildlife contraception involves the use of biodegradable implants containing the gonadotropin-releasing hormone (GnRH) agonist leuprolide. During fall 2002–spring 2004, we compared pregnancy rates, reproductive behavior, daily activity patterns, and body condition of 17 free-ranging female elk treated with a leuprolide formulation with 17 untreated females, in Rocky Mountain National Park, Colorado, USA. After treatment, the pregnancy rate of treated elk was 0%, whereas 79% of control elk became pregnant. The effects of treatment were reversed the subsequent year with the pregnancy rate of treated females 100% compared with 91% for controls. Reproductive behaviors were similar for treated and control elk during the breeding and postbreeding seasons; treated elk did not exhibit postrut reproductive behaviors. Moreover, we found no differences in daily activity patterns of experimental groups during the breeding or postbreeding seasons. Treated elk did not show improved body condition over pregnant females. Instead, treated females catabolized proportionately more body fat during winter after treatment and at a higher rate than pregnant control elk. However, this effect was reversed the next spring with no difference in body fat between treated and control elk. We conclude from this experiment that leuprolide, administered as a controlled release formulation, offers a safe and effective method of controlling fertility in free-ranging female elk. However, practical application is limited by treatment duration and the need to treat females before the breeding season.

Habitat destruction and degradation are major factors in reducing abundance, placing populations and species in jeopardy. Monitoring changes to habitat and identifying locations of habitat for a species, after disturbance, can assist mitigation of the effects of human-caused or -amplified habitat disturbance. Like many areas in the western United States, the Pinaleño Mountains of southeastern Arizona, USA, have suffered catastrophic fire and large-scale insect outbreaks in the last decade. The federally endangered Mt. Graham red squirrel (Tamiasciurus hudsonicus grahamensis) is only found in the Pinaleño Mountains, and to assess effects of forest disturbance on habitat we modeled their potential habitat by identifying characteristics of cover surrounding their centrally defended middens. We classified high-spatial resolution satellite imagery into ground cover classes, and we used logistic regression to determine areas used by squirrels. We also used known midden locations in conjunction with slope, elevation, and aspect to create a predictive habitat map. Squirrels selected areas of denser forest with higher seedfall for midden sites. Among active middens, those in the densest and least damaged forests were occupied in more seasons than those in more fragmented and damaged areas. The future conservation of red squirrels and the return of healthy mature forests to the Pinaleño Mountains will rely on preservation of mixed conifer zones of the mountain and active restoration of spruce–fir forests to return them to squirrel habitat. Our ability to evaluate the spectrum of fine- to coarse-scale disturbance effects (individual tree mortality to area wide boundaries of a disturbance) with high-resolution satellite imagery shows the utility of this technique for monitoring future disturbances to habitat of imperiled species.

The conservation of bison (Bison bison) from near extinction to >4,000 animals in Yellowstone National Park has led to conflict regarding overabundance and potential transmission of brucellosis (Brucella abortus) to cattle. We estimated survival and birth rates from 53 radiocollared adult female bison during 1995–2001, and we used calf:adult (C:A) ratios to estimate reproduction with the combined effects of pregnancy, fetal loss, and neonatal mortality during 1970–1997. Annual survival of adult females was high (0.92; 95% CI = 0.87–0.95) and constant. Birth rates differed by brucellosis status and age. Birth rates were 0.40 calves per female (95% CI = 0.15–0.65) for brucellosis-positive 3 year olds, 0.63 (95% CI = 0.39–0.87) for individuals testing negative, and 0.10 (95% CI = 0.00–0.24) for individuals contracting brucellosis that birth year (sero-converters). Birth rates were 0.64 (95% CI = 0.52–0.76) for brucellosis-positive individuals ≥4 years old, 0.81 (95% CI = 0.73–0.89) for brucellosis-negative individuals, and 0.22 (95% CI = 0.00–0.46) for sero-converters. Spring C:A ratios were negatively correlated with snow pack (β = −0.01 to −0.03, R² = 0.26–0.60, P < 0.05). Growth rate was highly elastic to adult survival (0.51), and juvenile survival (0.36) was 3 times more elastic than fecundity (0.12). Simulations suggested brucellosis eradication via vaccination would result in increased birth rates and a 29% increase in population growth (λ = 1.09), possibly leading to more bison movements outside the park. Our results will help park managers evaluate bison population dynamics and explore consequences of management actions and disease control programs.

Population reduction (PR) and trap–vaccinate–release (TVR) were used to control an outbreak of the raccoon (Procyon lotor) variant of rabies in Ontario, Canada, during 1999. Raccoon density declined significantly following PR; however, density increased significantly 9–10 months post–PR. Recolonization of PR areas with raccoons marked in TVR areas was very low with only 1.8% (32/1,759) of those raccoons dispersing into PR areas during 1999–2000. In addition, the number of raccoons dispersing from TVR areas into PR areas was not significantly greater than the number of raccoons dispersing into treatment areas where population reduction had not been implemented during 1999–2000. Raccoon movements in this study, determined by mark–recapture, averaged <4.0 km during 1999–2000 and did not differ among treatments (with or without PR). In summary, there was not a mass exodus of raccoons from TVR areas to PR areas during 1999–2000, suggesting there was not an immediate, dramatic, vacuum effect drawing raccoons into areas devoid of raccoons due to PR. However, one year post–PR, raccoon populations were near or above pre–PR densities (due to a high reproductive rate of resident raccoons), suggesting that PR would have to be an annual event if rabies is not eliminated from an area.

Depredation of livestock by large carnivores is an important but poorly understood source of human–carnivore conflict. We examined patterns of livestock depredation by jaguars (Panthera onca) and pumas (Puma concolor) on a ranch–wildlife reserve in western Brazil to assess factors contributing to prey mortality. We predicted jaguars would kill a greater proportion of calves than yearling and adult cattle and that proximity to suitable habitat would increase mortality risk. We further speculated that exposure to predation risk would promote livestock grouping and increased movement distance. We recorded 169 cattle mortality incidents during 2003–2004, of which 19% were due to predation by jaguars and pumas. This level of mortality represented 0.2–0.3% of the total livestock holdings on the ranch. Jaguars caused most (69%) cattle predation events, and survival in allotments was lower for calves than for other age classes. Forest proximity was the only variable we found to explain patterns of livestock mortality, with predation risk increasing as distance to forest cover declined. Due to low predation risk, cattle movement patterns and grouping behavior did not vary relative to level of spatial overlap with radiocollared jaguars. The overall effect of predation on cattle was low and livestock likely constituted an alternative prey for large cats in our study area. However, selection of calves over other age cohorts and higher predation risk among cattle in proximity to forest cover is suggestive of selection of substandard individuals. Cattle ranchers in the Pantanal region may reduce cattle mortality rates by concentrating on losses due to nonpredation causes that could be more easily controlled.

In a study of predation on ground-nesting birds at South San Francisco Bay (South Bay), California, USA, we analyzed stomach contents and stable isotopes of carbon and nitrogen to identify commonly consumed prey. We obtained the stomach contents from 206 nonnative red foxes (Vulpes vulpes regalis) collected in the South Bay area and Monterey County during 1995–2001 and from 68 feral cats (Felis silvestris) from the South Bay area during 2001–2002. We determined prey identity, biomass, and frequency, described seasonal diet trends, and derived an Index of Relative Importance. Avian species were the most frequent prey we found in the stomachs of red foxes from South Bay (61%), whereas small rodents were most frequent for red foxes from Monterey County (62%). Small rodents were the most frequent prey we found in feral cats (63%). Carbon and nitrogen isotopic signatures for foxes supported stomach content findings. However, isotope results indicated that cats received a majority of their energy from a source other than rodents and outside the natural system, which differed from the stomach content analysis. We demonstrated the utility of both stable isotope and stomach content analyses to establish a more complete understanding of predators' diets. This information aids natural resource managers in planning and evaluating future predator-removal programs and increases our understanding of the impacts of nonnative foxes and cats on native species.

Studies of reintroduced animals are beneficial to evaluate the success of reintroduction programs and to understand factors influencing fitness of reintroduced individuals. The geographic distribution of the federally threatened Louisiana black bear (Ursus americanus luteolus) has been reduced to 3 isolated populations due to habitat loss and excessive harvest. We reintroduced 23 adult female Louisiana black bears and their cubs to east-central Louisiana, USA, and documented postrelease space use, survival, movements, and reproduction. Individual females used larger home ranges after reintroduction than they had in the source population (P = 0.037). Spring ranges of reintroduced females were smaller than summer, autumn, and annual ranges (all P < 0.09), which did not differ from each other (all P > 0.60). Survival of reintroduced females did not differ between their first (S = 0.933) and second (S = 1.00) year after release or from annual survival of females in the source population (S = 0.964–1.00). Mean straight-line distance traveled by females from their release sites to the center of established home ranges or last recorded locations was 22.7 km. Six females reproduced after reintroduction and produced 15 cubs. Mean postrelease litter size of parturient reintroduced females (2.5) was similar to reported mean litter size of females in the source population (2.4). Our results suggest that the Louisiana reintroduction program is proceeding favorably; however, future studies should continue to monitor survival and reproduction of reintroduced females in Louisiana. Additional demographic parameters (i.e., cub survival) should be estimated to allow for population viability analysis to determine if the new population is self-sustaining.

Sheep predation by coyotes (Canis latrans) is a major problem for sheep producers in North America. Solutions are facilitated by a basic understanding of the trophic dynamic context of this problem, one that likely varies geographically in important qualitative ways. Little is known about vertebrate trophic dynamics in Mediterranean ecosystems, where prey are diverse and their biomass is strongly influenced multi-annually by variable rainfall. We used long-term data sets from north-coastal California, USA, to investigate whether wild prey fluctuations caused immediate negative effects on sheep predation via a reduction in the coyote functional response or delayed positive effects on sheep predation via a numerical response by coyote predators. Because we could not measure prey biomass directly, we used variables associated with lower trophic levels (e.g., annual plant productivity, vole abundance, rainfall) as proxies for wild prey biomass. Coyote population growth rate was positively correlated with lower-trophic-level variables of the previous year, suggesting a numerical response, and sheep (ad F lambs) predation was positively correlated with coyote abundance in the current year. Sheep predation also was negatively correlated with lower-trophic-level variables of the current year, suggesting an immediate buffering effect of wild prey on sheep predation. Together, coyote abundance and lower-trophic-level variables explained 47% of the multi-annual variation in sheep kills. The negative pathway between lower-trophic-level variables and sheep predation was stronger than the positive pathway, possibly due to the erratic nature of multi-annual fluctuations in lower-trophic-level variables, which could prevent the numerical response from reaching its full potential. Monthly analyses revealed a type III functional response of coyotes to lambs, which is expected to enhance buffering effects of wild prey on sheep predation. Our findings suggest the dominant effect of wild prey biomass on sheep predation by coyotes in this Mediterranean-type community is as a buffer.

Assessing the effectiveness of monitoring techniques designed to determine presence of forest carnivores, such as American marten (Martes americana), is crucial for validation of survey results. Although comparisons between techniques have been made, little attention has been paid to the issue of detection probabilities (p). Thus, the underlying assumption has been that detection probabilities equal 1.0. We used presence–absence data obtained from a track-plate survey in conjunction with results from a saturation-trapping study to derive detection probabilities when marten occurred at high (>2 marten/10.2 km²) and low (≤1 marten/10.2 km²) densities within 8 10.2-km² quadrats. Estimated probability of detecting marten in high-density quadrats was p = 0.952 (SE = 0.047), whereas the detection probability for low-density quadrats was considerably lower (p = 0.333, SE = 0.136). Our results indicated that failure to account for imperfect detection could lead to an underestimation of marten presence in 15–52% of low-density quadrats in the Black Hills, South Dakota, USA. We recommend that repeated site-survey data be analyzed to assess detection probabilities when documenting carnivore survey results.

Abundance of Rio Grande wild turkeys (Meleagris gallopavo intermedia) has declined in the southeastern Edwards Plateau (EP) of Texas, USA, whereas abundance has remained stable in the northwestern EP. Invertebrates are a critical protein source for poults <6 weeks posthatch. We collected invertebrates at brood and paired locations in both the stable and declining regions. Our objective was to determine if differences in invertebrate abundance existed in regions typified by declining versus stable Rio Grande wild turkey abundance. We found no difference in invertebrate abundance between brood or paired locations within regions, but invertebrate abundance, whether measured as dry mass or frequency, was greater in the stable region. Decreased invertebrate abundance may have contributed to the decline in Rio Grande wild turkey abundance in the southeastern Edwards Plateau.

Fixed-kernel density estimates using radiotelemetry locations are frequently used to quantify home ranges of animals, interactions, and resource selection. However, all telemetry data have location error and no studies have reported the effects of error on utilization distribution and area estimates using fixed-kernel density estimators. We simulated different home range sizes and shapes by mixing bivariate-normal distributions and then drawing random samples of various sizes from these distributions. We compared fixed-kernel density estimates with and without error to the true underlying distributions. The effects of telemetry error on fixed-kernel density estimates were related to sample size, distribution complexity, and ratio of median Circular Error Probable to home range size. We suggest a metric to assess the adequacy of the telemetry system being used to estimate an animal's space use before a study is undertaken. Telemetry location error is unlikely to significantly affect fixed-kernel density estimates for most wildlife telemetry studies with adequate sample sizes.

Most ecologists use statistical methods as their main analytical tools when analyzing data to identify relationships between a response and a set of predictors; thus, they treat all analyses as hypothesis tests or exercises in parameter estimation. However, little or no prior knowledge about a system can lead to creation of a statistical model or models that do not accurately describe major sources of variation in the response variable. We suggest that under such circumstances data mining is more appropriate for analysis. In this paper we 1) present the distinctions between data-mining (usually exploratory) analyses and parametric statistical (confirmatory) analyses, 2) illustrate 3 strengths of data-mining tools for generating hypotheses from data, and 3) suggest useful ways in which data mining and statistical analyses can be integrated into a thorough analysis of data to facilitate rapid creation of accurate models and to guide further research.

Ecologists often develop complex regression models that include multiple categorical and continuous variables, interactions among predictors, and nonlinear relationships between the response and predictor variables. Nomograms, which are graphical devices for presenting mathematical functions and calculating output values, can aid biologists in interpreting and presenting these complex models. To illustrate benefits of nomograms, we developed a logistic regression model of elk (Cervus elaphus) resource selection. With this model, we demonstrated how a nomogram helps scientists and managers interpret interactions among variables, compare the relative biological importance of variables, and examine predicted shapes of relationships (e.g., linear vs. nonlinear) between response and predictor variables. Although our example focused on logistic regression, nomograms are equally useful for other linear and nonlinear models. Regardless of the approach used for model development, nomograms and other graphical summaries can help scientists and managers develop, interpret, and apply statistical models.