Deer were introduced to Australia in the 19th century, and today 6 species have been established in the wild. Australians have considered deer a pest and an important economic and hunting resource since their early introduction. We recommend promoting the management of wild deer for the joint outcomes of reducing adverse impacts on agriculture, forestry, and conservation values, and for exploitation and harvest. We also recommend that managers recognize deer as an established component of many Australian ecosystems. If Australians are to have a professional approach to wildlife management, we must move away from the traditional paradigms of protection of native species and eradication of exotic species and embrace wildlife management for all its complexities, challenges, and positive outcomes. We discuss wild deer management as a test case to explore these issues.

Presence–absence data can be useful to wildlife managers in a wide variety of contexts, from monitoring populations at large spatial scales to identifying habitats that are of high value to specific species of conservation concern. However, a key issue is that a species may be declared “absent” from a landscape unit simply as a result of not detecting the species using the prescribed sampling methods. The effect of this imperfect detection is that parameter estimates will be biased, and any modeling of the data provides a description of the surveyors' ability to find the species on the landscape, not where the species is on the landscape. The reliability of so-called “presence–absence” data for making sound management decisions and valid scientific conclusions could therefore be questioned. However, after collecting appropriate data (i.e., repeated surveys of landscape units within a relatively short timeframe), recently developed statistical models can be used to obtain unbiased parameter estimates. Here I provide a nontechnical overview of the issues that pertain to wildlife studies or monitoring programs that seek to make reliable inference about the presence or absence of a target species.

The status of the swift fox (Vulpes velox) became a subject of controversy when a petition was submitted in 1992 to have it listed under the provisions of the Endangered Species Act. Colorado is believed to have one of the largest remaining populations of the species due to significant amounts of short-grass prairie habitat suitable for occupancy. In 1995, we initiated a survey in eastern Colorado to determine swift fox population size and presence or absence on 724.8×6.4 km² grids trapped with 20 live traps. We drew trapping grids from a stratified random sample of grids plotted on 1:100,000 scale Gap vegetation maps of eastern Colorado. Of the grids sampled, 54% were from areas with 75% short-grass habitat. We captured 241 swift foxes (117 male, 121 female, and 3 undetermined) on 51 (71%) of the grids. We estimated a mean capture probability per night of an individual fox, weighted by number of grids and number of trapping occasions for each grid, as 0.234 (SE = 0.022). Highest capture and detection probabilities occurred during the September to March period. We estimated the average number of foxes per grid as 6.96 (SE = 0.794), likely a maximum estimate because of foxes being attracted to the trapping grids. Percentage of short-grass prairie habitat on each grid was an important predictor of both occupancy rate and detection probabilities. We estimated occupancy rate of the 72 grids trapped as 0.821 (SE = 0.0659), using the mean (66.9%) of the short-grass prairie habitat for the 72 grids as a covariate, suggesting that 59.1 (SE = 4.74) of the 72 grids trapped contained foxes. Alternatively, using the percentage of short grass prairie for each grid to estimate a grid-specific occupancy probability, we estimated 56.9 (SE = 4.13) of the 72 grids contained foxes. We estimated nightly detection probabilities of >0.60 for grids trapped in March. We used simulation to compute the power to detect simulated changes in occupancy rates assuming a detection probability of 0.6 and an occupancy rate of 0.8. With only 2 trapping occasions, even with 100 trapping grids, the power was only 80% to detect a change in occupancy from 0.8 to 0.5. With 50 trapping grids and 3 trapping occasions, the power was nearly 80% to detect a 0.3 reduction in occupancy rate and >80% for 4 trapping occasions.

Knowledge of factors influencing animal distribution and abundance is essential in many areas of ecological research, management, and policy-making. Because common methods for modeling and estimating abundance (e.g., capture–recapture, distance sampling) are sometimes not practical for large areas or elusive species, indices are sometimes used as surrogate measures of abundance. We present an extension of the Royle and Nichols (2003) generalization of the MacKenzie et al. (2002) site-occupancy model that incorporates length of the sampling interval into the model for detection probability. As a result, we obtain a modeling framework that shows how useful information can be extracted from a class of index methods we call indirect detection indices (IDIs). Examples of IDIs include scent station, tracking tube, snow track, tracking plate, and hair snare surveys. Our model is maximum likelihood, and it can be used to estimate site occupancy and model factors influencing patterns of occupancy and abundance in space. Under certain circumstances, it can also be used to estimate abundance. We evaluated model properties using Monte Carlo simulations and illustrate the method with tracking tube and scent station data. We believe this model will be a useful tool for determining factors that influence animal distribution and abundance.

The ecological niche is the conceptual foundation for the analysis of species distributions. Habitat variables and resources are therefore thought to determine the distribution of a species. However, metapopulation and source-sink models posit that population processes determine the local distribution of a species; metapopulation theory suggests that many suitable patches could be unoccupied and source-sink theory suggests that unsuitable patches could be occupied. We examined whether the distribution of 2 European anurans (tree frog [Hyla arborea] and natterjack toad [Bufo calamita]) was determined by habitat variables, local population size (as indexed by the number of calling males), or metapopulation characteristics are. We found that the distribution of both species was best predicted by the number of calling males in the previous years. This supports the idea that population processes are more important determinants of distribution than habitat characteristics. We suggest that future distribution models explicitly incorporate population dynamic processes. Our results suggest that there is a threshold number of calling males above which a population is almost certain to persist in the short-term. If more robust population data are not available, such a threshold should be useful for managing threatened anuran species because the number of calling males is easily determined.

Conservation efforts often use habitat models as surrogates for empirical data on the distribution or population size of target species as well as to predict the efficacy of proposed conservation and management actions. Such a model was used to evaluate preserves for the Palm Springs ground squirrel (Spermophilus tereticaudus chlorus) under a proposed Habitat Conservation Plan in the Sonoran Desert of southern California. The habitat model was developed from expert opinion and data from nonprobability-based sample surveys for ground squirrels. We evaluated the model and tested the underlying assumptions about the ground squirrel's habitat affinities, estimated the occupancy rates and detection probabilities of ground squirrels in modeled habitat, quantified the total area of the proposed preserves occupied by ground squirrels, and tested a priori hypotheses about how occupancy differed among vegetation and substrate types. Occupancy rate varied with vegetation and substrate types and, by repeatedly visiting survey points, we found that detection probability was ∼0.22 (SÊ= 0.05). We found ground squirrels were more likely to occupy mesquite (Prosopis glandulosa variety torreyana) on dunes/hummocks (0.99 ± 0.01) than creosote (Larrea tridentata) on dunes/hummocks (0.33 ± 0.09) or other vegetation/substrate types (≤0.08). Our results suggest that this and comparable habitat models should be refined to incorporate detection probabilities and variation in ground squirrel affinity for different vegetation and substrates. In contrast to the existing habitat model, our data suggested that ground squirrel occupancy rate differed by vegetation/substrate types. We suggest that conservation efforts focus on protecting and enhancing the mesquite hummock/dune system that has largely disappeared over the last 50 years due to agricultural and urban development and a lowering of water tables.

Surveys that record the presence or absence of fauna are used widely in wildlife management and research. A false absence occurs when an observer fails to record a resident species. There is a growing appreciation of the importance of false absences in wildlife surveys and its influence on impact assessment, monitoring, habitat analyses, and population modeling. Very few studies explicitly quantify the rate of these errors. Quantifying the rate of false absences provides a basis for estimating the survey effort necessary to assert that a species is absent with a pre-specified degree of confidence and allows uncertainty arising from false absences to be incorporated in inference. We estimated the rate of false absences for 2 species of forest owl and 4 species of arboreal marsupial based on 8 repeat visits to 50 survey locations in south-eastern Australia. We obtained estimates using a generalized zero-inflated binomial model. We presented detectability curves for each species to convey the number of visits required to achieve a specified level of confidence that resident species will be detected. The observation error rates we calculated were substantial but varied between species. For the least detectable species, the powerful owl (Ninox strenua), our standard surveys returned false absences on 87% of visits. However, our surveys of the more detectable sugar glider (Petaurus breviceps) returned a 45% false absence rate. We predict that approximately 18 visits would be required to be 90% sure of detecting resident owls and approximately 5 visits would provide 90% confidence of detecting resident sugar gliders. We fitted hierarchical logistic regression models to the data to describe the variation in detection rates explained by environmental variables. We found that temperature, rainfall, and habitat quality influenced the detectability of most species. Consideration of observation error rates could result in important changes to resource management and conservation planning.

Northern spotted owls (Strix occidentalis caurina) have been studied intensively since their listing as a threatened species by the U.S. Fish and Wildlife Service in 1990. Studies of spotted owl site occupancy have used various binary response measures, but most of these studies have made the assumption that detectability is perfect, or at least high and not variable. Further, previous studies did not consider temporal variation in site occupancy. We used relatively new methods for open population modeling of site occupancy that incorporated imperfect and variable detectability of spotted owls and allowed modeling of temporal variation in site occupancy, extinction, and colonization probabilities. We also examined the effects of barred owl (S. varia) presence on these parameters. We used spotted owl survey data from 1990 to 2002 for 3 study areas in Oregon, USA, and we used program MARK to develop and analyze site occupancy models. We found per visit detection probabilities averaged <0.70 and were highly variable among study years and study areas. Site occupancy probabilities for owl pairs declined greatly on 1 study area and slightly on the other 2 areas. For all owls, including singles and pairs, site occupancy was mostly stable through time. Barred owl presence had a negative effect on spotted owl detection probabilities, and it had either a positive effect on local-extinction probabilities or a negative effect on colonization probabilities. We conclude that further analyses of spotted owls must account for imperfect and variable detectability and barred owl presence to properly interpret results. Further, because barred owl presence is increasing within the range of northern spotted owls, we expect to see further declines in the proportion of sites occupied by spotted owls.

Widespread conversion of deciduous forests to agriculture in the midwestern United States has resulted in landscapes whose remaining native habitats are highly fragmented, with well-documented consequences for wildlife community structure. We analyzed trap data for 5 forest rodent species from 525 sites in 35 study landscapes throughout the upper Wabash River basin, which is dominated by agricultural use and drains >20% of Indiana. We used a recently developed likelihood approach and multi-model inference to obtain unbiased estimates of occurrence probabilities for a species when detection rates are <1, and we used hierarchical generalized linear modeling to evaluate random effects associated with nested data structure. Inclusion of a spatial autocovariate term had important effects on model selection results for 4 of 5 species and improved fit of models. Eastern chipmunks (Tamias striatus), although previously described as sensitive to fragmentation, were nearly ubiquitous in our study region and had greater occupancy rates in smaller woodlots. Fox squirrels (Sciurus niger) are well adapted to patchy landscapes, and their occupancy rates actually increased with patch isolation. Gray squirrels (S. carolinensis) are adversely affected by forest fragmentation; we observed a threshold decline in occupancy rates beyond ∼15 km from the Wabash River, which harbors the only source of contiguous habitat within the entire river basin. Red squirrels (Tamiasciurus hudsonicus) exhibited greater abundance in sites without gray squirrels. Moreover, their occupancy rates were positively related to habitat and landscapes considered suboptimal (e.g., lower basal area of hard-mast bearing trees, more isolated patches); this was further evidence for negative effects of gray squirrels on red squirrel populations. White-footed mice (Peromyscus leucopus) were nearly ubiquitous; although occurrence rates in less forested landscapes increased with distance from edge and greater volume of woody debris.

Declines in populations of many species across the United States are prompting the need for monitoring programs that can detect population change effectively for large suites of species. The proliferation of reliable standardized monitoring techniques enhances the potential that we may meet these burgeoning information needs. However, multiple-species monitoring approaches are still a rarity. In an attempt to meet monitoring information needs on National Forest System (NFS) lands, the U.S. Forest Service developed the Multiple Species Inventory and Monitoring (MSIM) protocol, an approach that employs a variety of vertebrate and habitat survey methods to obtain presence–absence data for a broad spectrum of species at a systematic array of sample points throughout ecoregions. A recent evaluation of the MSIM protocol predicted that its implementation across the Sierra Nevada ecoregion would detect a 20% change in proportion of points occupied for 76% of the species with 80% confidence and power. This evaluation was based on a qualitative estimate of probability of detection and basic habitat association information to estimate probability of presence. We conducted a field test to evaluate the ability of a multi-taxonomic, presence–absence monitoring approach to detect and estimate population change at multiple scales and to meet the monitoring needs of land management agencies. Our field data consisted of bird, mammal, amphibian, and reptile surveys conducted at terrestrial and aquatic sites in the Lake Tahoe basin, located in the central Sierra Nevada. We used a maximum likelihood function to quantitatively estimate the probability of detection and sampling adequacy of species and species groups. We compared these observed estimates to the previously derived expected estimates. We detected 185 species by field testing, including 89% of the expected species and several species that were not expected. Observed probabilities of detection were generally lower than we expected. Sampling adequacy was also lower than we expected but still relatively high, with 66% of the Lake Tahoe species and 47% of all Sierra Nevada species adequately sampled. Species with small geographic ranges, narrow habitat specificity, or small population sizes were considered rare, and they were not sampled as adequately as other species groups. Still, 69 of the rarest species, including 12 species with state or federal designation as endangered, threatened, or sensitive, were estimated to be adequately sampled. Our analysis highlighted the ability of a broad-scale, multi-taxonomic presence–absence monitoring approach, such as the MSIM protocol, to detect population change for a large number and wide variety of vertebrate species. The outcome of our test also suggests that many data yields would be garnered at the scale of National Forests and National Parks, such as distribution and occurrence data for most species, including little-known species, population trend data for many species, habitat relationships and habitat condition data, and data on association strengths of indicator species.

We hypothesized that hunters' use of approximate responses for participation and harvest survey questions was prompted by the cognitive complexity of giving an exact answer. Approximate responses we considered were: (1) rounding days of participation and harvest to numbers ending in zero or 5 (0–5 prototypes) and (2) multiplying days of participation by an integer to approximate harvest. Both types of survey responses result in response heaps (spikes in the data) that can introduce bias. Our analyses were based on the frequency of responses in heaps relative to all responses in a neighborhood (a set of responses close to each other). Data from the Illinois hunter harvest survey (1998–2000) showed that the relative frequency of using approximate responses in a neighborhood was statistically constant across years and species despite differences in day-harvest frequencies. Our results supported accepting the hypothesis that survey responses are influenced by the cognitive complexity of formulating a response. We discuss the following strategies for identifying, reducing, and correcting bias in estimates derived from surveys: (1) basic research to identify the extent to which respondents use prototypes when answering participation and harvest questions; (2) survey methodology modifications to increase the accuracy of answers; and (3) the development of bias correction factors.

Introduced brushtail possums (Trichosurus vulpecula) are a widespread pest of conservation and agriculture in New Zealand, and considerable effort has been expended controlling populations to low densities. A national protocol for monitoring the abundance of possums, termed trap catch index (TCI), was adopted in 1996. The TCI requires that lines of leghold traps set at 20-m spacing are randomly located in a management area. The traps are set for 3 fine nights and checked daily, and possums are killed and traps reset. The TCI is the mean percentage of trap nights that possums were caught, corrected for sprung traps and nontarget captures, with trap line as the sampling unit. We studied 1 forest and 1 farmland area in the North Island, New Zealand, to address concerns that TCI estimates may not be readily comparable because of seasonal changes in the capture probability of possums. We located blocks of 6 trap lines at each area and randomly trapped 1 line in each block in 3 seasons (summer, winter, and spring) in 2000 and 2001. We developed a model to allow for variation in local population size and nightly capture probability, and fitted the model using the Bayesian analysis software BUGS. Capture probability declined with increasing abundance of possums, generating a nonlinear TCI. Capture probability in farmland was lower during spring relative to winter and summer, and to forest during summer. In the absence of a proven and cost-effective alternative, our results support the continued use of the TCI for monitoring the abundance of possums in New Zealand. Seasonal biases in the TCI should be minimized by conducting repeat sampling in the same season.

Highways and railroads have come under increasing scrutiny as potential agents of population and habitat fragmentation for many mammalian species, including grizzly bears (Ursus arctos). Using Global Positioning System (GPS) technology and aerial Very High Frequency (VHF) telemetry, we evaluated the nature and extent of trans-highway movements of 42 grizzly bears along the U.S. Highway 2 (US-2) corridor in northwest Montana, USA, 1998–2001, and we related them to highway and railroad traffic volumes and other corridor attributes. We employed highway and railroad traffic counters to continuously monitor traffic volumes. We found that 52% of the sampled population crossed highways at least once during the study but that crossing frequency was negatively exponentially related to highway traffic volume. We found that grizzly bears strongly avoided areas within 500 m of the highway and that highway crossing locations were clustered at a spatial scale of 1.5 km. Most highway crossings occurred at night when highway traffic volume was lowest but when railroad traffic was highest. Highway crossing locations were flatter, closer to cover in open habitat types, and within grassland or deciduous forest vegetation types. Nighttime traffic volumes were low, averaging about 10 vehicles/hr, allowing bears to cross. However, we project that US-2 may become a significant barrier to bear movement in ∼30 years if the observed trend of increasing traffic volume continues.

The use of Population Viability Analysis has recently been criticized mainly because uncertainties in population predictions are often ignored. We analyzed the viability of Scandinavian wolverine Gulo gulo populations that allow for the inclusion of uncertainties in parameter estimates and stochastic effects on population dynamics. We used the insight gained from analyses of stochastic population models to quantitatively analyze different management options for the wolverine. These management options were based on data from an individual-based demographic study in the Sarek National Park, Sweden, and data from monitoring schemes in Norway and Sweden. Stochastic components in population dynamics of the wolverine were large. Strong density regulation occurred around the carrying capacity. According to the World Conservation Union (2001), the carrying capacity of populations must exceed 46 sexually mature (≥3-yr-old) females to be considered not vulnerable. Continuation of the current levels of offtake in Norway will lead to extinction of the wolverine over large parts of the country. Hence, current rates of mortality of female wolverine make the northern population endangered, whereas the southern population is classified as vulnerable. Management plans allowing harvest of individuals should be based on a proportional threshold harvest strategy. Only 40–60% of the surplus individuals exceeding a certain threshold for harvesting should be removed. Our analyses emphasized the importance of including estimates of precision in parameter estimates, including stochastic factors in recommendations based on Population Viability Analysis, and examining the robustness of the conclusions against variation in crucial parameters by sensitivity analyses.

Leopard seals (Hydrurga leptonyx) are formidable marine predators and require sedation before scientific examination. Mass-specific drug dosage for leopard seals has usually been determined from generic allometric equations or visual estimates. However, the leopard seal is a slender phocid and generic equations are likely to return inaccurate mass estimates, which may lead to fatal overdoses of drugs. We used published and unpublished morphometric data to construct allometric models for estimating leopard seal body mass. The model using volume (Vol), which combined measures of snout-tail length (STL) and the square of girth (G²), provided our best estimate of mass (r² = 0.97). The model using STL alone was sample-site specific (each r² = 0.85), highlighting G as an important measure to obtain where possible. The confidence and prediction intervals associated with each model broadened with increased seal size and decreased sample size, suggesting the use of extra caution when estimating drug dosage for larger seals to avoid over- or under-dosing. We also developed a seal body condition index that can assist wildlife management when deciding if rehabilitation of vagrant seals is warranted. Body condition may also affect the induction, duration and recovery times of anaesthetized leopard seals.

The jaguar (Panthera onca) is an endangered species that occasionally visits the southwestern United States from Mexico. The number of jaguar sightings per decade has declined over the last 100 years in Arizona, USA, raising conservation concerns for the species at a local and national level. In 1997, state, federal, and local governments with land-management responsibilities agreed to characterize and identify potential jaguar habitat in Arizona and New Mexico. Specifically, the objectives of our analysis were 2-fold: (1) characterize potential jaguar habitat in Arizona from historic sighting records and (2) create a statewide habitat suitability map. We used a Geographic Information System (GIS) to characterize potential jaguar habitat by overlaying historic jaguar sightings (25) on landscape and habitat features believed important (e.g., vegetation biomes and series, elevation, terrain ruggedness, proximity to perennial or intermittent water sources, human density). The amount of Arizona (%) identified as potential jaguar habitat ranged from 21% to 30% depending on the input variables. Most jaguar sightings were in scrub grasslands between 1,220 and 1,829-m elevation in southeastern Arizona, in intermediately to extremely rugged terrain, and within 10 km of a water source. Conservation efforts should focus on protecting the most suitable jaguar habitat in southeastern Arizona (i.e., Santa Cruz, Pima, Cochise, Pinal, Graham counties), travel corridors within and outside Arizona, and jaguar habitat in the Sierra Madres of Sonora, Mexico.

Eurasian lynx (Lynx lynx) are responsible for significant depredation on domestic lambs in Norway. Recreational hunting of lynx is widely used to limit lynx population growth and to attempt to remove problem individuals. We analyzed the relationship between annual changes in lamb losses and lynx hunting on 2 scales. On the county scale, lamb losses were related to the size of the lynx population that was reduced through the harvest of 294 lynx by hunters during our study, 1995–2001. At the level of individual grazing areas, we documented a significant local effect of lynx harvest (n = 321 lynx); however, the magnitude of this benefit was so small (13 lambs per male lynx, or 2 lambs per female lynx) as to be of little practical benefit. The data indicate that lynx hunting only reduces depredation when it reduces the size of the population.

To overcome costs and logistical constraints associated with capturing and marking a sufficient number of elk (Cervus elaphus) for population estimation, elk are sometimes marked with paintballs fired from a helicopter flying overhead. Although marking elk with paintballs is a cost-effective marking technique, there are some unique aspects of this method that may bias population estimates. When analyzing the mark–recapture data, the assumption that each animal has an independent and identical probability of being marked and resighted may be violated. Instead, the probabilities of marking and resighting are likely to be a function of herd size. We explored the potential biases of paintball, mark–resight surveys of elk and present alternative methods that could be used to generate population estimates from mark–resight data, including Bailey's binomial MLE, cluster sampling, and bootstrap methods. Our results suggest that population estimates will generally be negatively biased; the magnitude of the bias depends upon the degree of herd aggregation and fidelity. To minimize these biases, we recommend (1) marking animals when they occur in many small herds, (2) marking equal proportions of individuals in each herd, and (3) determining the degree of aggregation and fidelity of elk to herds in each area. Although these recommendations do not facilitate the primary goal of the technique (i.e., the ability to mark numerous animals easily), following them will provide more reliable population estimates.

Snowshoe hares (Lepus americanus) undergo remarkable cycles and are the primary prey base of Canada lynx (Lynx canadensis), a carnivore recently listed as threatened in the contiguous United States. Efforts to evaluate hare densities using pellets have traditionally been based on regression equations developed in the Yukon, Canada. In western Montana, we evaluated whether or not local regression equations performed better than the most recent Yukon equation and assessed whether there was concordance between pellet-based predictions and mark–recapture density estimates of hares. We developed local Montana regression equations based on 224 data points consisting of mark-recapture estimates and pellet counts, derived from 38 sites in 2 different areas sampled for 1 to 5 years using 2 different pellet plot shapes. We evaluated concordance between estimated density and predicted density based on pellet counts coupled with regression equations at 436 site-area-season combinations different from those used to develop the regression equations. At densities below 0.3 hares/ha, predicted density based on pellets tended to be greater than for mark–recapture; the difference was usually <1 hare per ha on an absolute scale, but at low densities this translated to proportional differences of 1,000% or greater. At densities above 0.7 hares/ha, pellet regressions tended to predict lower density than mark–recapture. Because local regression equations did not outperform the Yukon equation, we see little merit in further development of local regression equations unless a study is to be conducted in a formal double-sampling framework. We recommend that widespread pellet sampling be used to identify areas with very low hare densities; subsequent surveys using mark–recapture methodology can then focus on higher density areas where density inferences are more reliable.

European wild rabbits (Oryctolagus cuniculus) are frequently translocated for hunting and conservation purposes. Quarantining these animals prior to release reduces the risk of releasing rabbits incubating field infections of myxomatosis or viral haemorrhagic disease (RHD), and it provides a way to vaccinate these animals against both diseases. However the optimal quarantine period needed to achieve these goals is not known. We therefore assessed the effects of quarantine lengths (2, 4, 6, 8 weeks) on rabbit biochemical parameters, immunity induced by vaccination against myxomatosis and RHD, and survival of translocated rabbits. We found that levels of total bilirubin, urea nitrogen, aspartate aminotransferase (AST), and alanine aminotransferase (ALT) enzymatic activities were significantly (P < 0.05) decreased during quarantine, independent of quarantine length, whereas hematocrit levels increased significantly. All pregnant female rabbits aborted or lost litters during quarantine. Seroconversion against myxomatosis and RHD following vaccination was not related to any biochemical parameter at vaccination time, but the proportion of seronegative rabbits that seroconverted was moderate. The heterophil/lymphocyte ratio, conjugated to unconjugated bilirubin ratio, and the serum AST and creatinine levels we measured after capture and transport were directly related to mortality through the quarantine period, whereas we found total serum protein level was negatively related to mortality. Mortality after release was positively related to urea nitrogen concentration and negatively related to hematocrit and the albumin/globulin ratio, but it was independent of quarantine length. Based on our findings, rabbit translocation programs that include a quarantine period could be improved by decreasing the acute stress induced by capture and handling prior to quarantine; facilitating a more rapid access to high quality feed during quarantine; and improving the vaccination protocol. In addition, the release from quarantine of rabbits should be determined by their physical condition, not merely by elapsed time in quarantine.

We evaluated elk (Cervus elaphus) reproductive success following removal of human disturbance during calving season, by comparing data from 2 segregated groups of free-ranging elk (control and treatment) from 1 pre-disturbance year, 2 disturbance years, and 2 post-disturbance years. Treatment-group elk were subjected to simulated recreational activity during calving season in disturbance years but not in pre- or post-disturbance years. Control animals experienced only ambient levels of disturbance throughout the study, and their calf/cow proportions (proportion of marked adult female elk maintaining a calf in Jul and Aug) were similar throughout the 5 years. We observed reduced productivity of treatment-group elk compared to controls during disturbance years after adjusting for nontreatment year differences. We hypothesized that productivity would return to, or potentially exceed, pre-disturbance levels following removal of disturbance. Productivity rebounded following release from disturbance, and full recovery was achieved by the second post-disturbance year. However, we did not observe productivity in excess of pre-disturbance years, as might be expected if release from energetic demands of maintaining a calf in 1 year increases probability of maintaining a calf in the following year. Our results are consistent with hypotheses that human-induced disturbance during parturition periods can reduce reproductive success and that removal of disturbance can allow productivity to recover to pre-disturbance levels. Managers of wildlife and wildlife habitat should consider potential impacts of human-induced disturbance on wildlife populations. Wildlife populations, depressed by human-induced disturbance during the neonatal period, may have the ability to rebound if the disturbance is removed.

We studied arboreal lichen production and availability for Rocky Mountain elk (Cervus elaphus nelsoni), white-tailed deer (Odocoileus virginianus), and mule deer (O. hemionus) during winter in western Montana, USA. In May and June of 1997 and 1998, we collected lichen litterfall inside and outside exclosures to assess deer and elk use of lichens and recorded associated tree stand characteristics. Bryoria spp. and Nodobryoria spp. composed >99% of the lichen litterfall. Lichen litterfall use by deer and elk on 2-yr sites averaged 7.91 kg/ha for the severe winter of 1996–1997, and 6.02 kg/ha for the milder winter of 1997–1998. The greater use of lichens in winter 1996–1997 was probably due to increased ungulate densities in forested habitats during a severe winter. A strong linear relationship between lichen availability and lichens consumed suggested that lichen use was driven by availability rather than opportunistic foraging and that ungulates may have selected habitat partly because of lichen availability. Tree stand variables were partially correlated with lichen availability and consumption by ungulates but were different between winters, indicating that variables in addition to those we measured contributed to availability and consumption.

We studied the impact of red fox (Vulpes vulpes) predation on free-ranging roe deer (Capreolus capreolus) neonates during 14 years in a mixed forest/agricultural landscape in south central Sweden. A large-scale natural predator removal experiment occurred when an outbreak of sarcoptic mange (Sarcoptes scabiei) reduced the red fox population initially and caused subsequent variations in red fox abundance. We estimated relative red fox abundance by dividing number of fox observations with number of person-days in field. Red fox predation accounted for 88% of known mortality in roe deer fawns. Predation was closely correlated to red fox abundance, and there was a strong negative correlation between fox abundance and overall fawn survival. Fox abundance was the only factor with a significant effect on between year variation in fawn survival. Annual predation varied between zero and 90%. Our study recorded the highest predation rates for roe deer juveniles established to date. High fawn survival in years of low fox abundance suggested that predation mortality was additive during summer. Our results supported the conclusion that the large national increase of roe deer in Sweden during the 1980s and 1990s was related to lower fox predation on fawns and also indicated that the roe deer population density was well below habitat carrying capacity at the onset of the mange epidemic. In Sweden, in general, a thorough fox control will most likely result in increased fawn survival and higher potential roe deer harvest.

Movements (e.g., migration, dispersal) of white-tailed deer (Odocoileus virginianus) vary greatly over the geographic range of the species. Therefore, region-specific, empirical information is needed to effectively manage deer populations. Movements of white-tailed deer have been well documented in forest dominated habitats; however, little information related to white-tailed deer movements exists in intensively (>80%) cultivated areas. From January 2001 to August 2002, we monitored movements of 77 (61 adult, 16 young) female white-tailed deer in southwest Minnesota. We collected 6,867 locations, calculated 130 home ranges, and documented 149 seasonal movements. Fifteen percent of deer were nonmigratory, whereas 35% were facultative migrators, and 42.5% were obligate migrators. Mean distance between summer and winter home range was 10.1 km. Temperature and snow depth had the greatest influence on initiation of seasonal migration, whereas crop emergence and harvest had minimal effects. Four deer (8%) dispersed a mean distance of 71.3 km with 1 adult female moving a straight-line distance of 205 km. All dispersing deer occupied a temporary staging area for approximately 1 month between previous winter and new summer ranges. Mean home range (95% use area) in winter (5.2 km²) was over twice as large as home range in summer (2.3 km²). Movements exhibited by white-tailed deer in southwest Minnesota were influenced by large annual fluctuations in climate and a highly fragmented landscape dominated by row-crop agriculture. We provide data beneficial to biologists managing northern populations of white-tailed deer in fragmented environments by detailing the relationship between climate, intensive agriculture, and deer movements.

The relationship between hunters and their environment is a key component in managing wildlife populations. Identifying hunter's characteristics, motivations, and efforts is crucial to understanding if a hunt will be successful. We predicted that landscape characteristics and moose (Alces alces) densities would affect success of hunts. As in wildlife management programs elsewhere, moose hunters in interior Alaska, USA, must return harvest tickets to the Alaska Department of Fish and Game. These tickets provide location of hunts (Uniform Coding Units) and other details. Our modeling of responses (1997–2001) from harvest tickets indicated that location of hunts, mode of transportation, hunting regulations, use of commercial services, year, density of roads, hunter-to-moose ratio, moose density, and residency of hunters were important predictors of success. In addition, we documented that the linear-regression approach to measuring catch per unit effort (CPUE) was inappropriate because it produced an inverse, but not significant, relationship between hunting effort and success. This outcome occurred because most hunts, particularly for large mammals, ended with the harvesting of an animal. Likewise, modeling of hunter success with logistic regression was similarly biased by measures of hunter effort. We established that a time-to-event Weibull regression provided substantial improvement over standard models of CPUE. Weibull regression accurately represented the positive relationship between effort and success, and it can be used to model length of hunt and other covariates related to hunters and landscape characteristics for predicting success.

We analyzed how the application of scent-marking, forest clearing, and supplemental feeding correlated with the number of moose (Alces alces)–train collisions along the most vulnerable railroad stretch in Norway. Data on 1,045 collisions has been compiled for 18 years since 1985, and remedial actions have occurred during various periods since 1990. We used sections of the rail line where remedies had never been applied as control sections to estimate the expected number of collisions per year and per km. In this way, we took into account the yearly variation in the number of accidents by using the difference between the actual number of accidents and the expected number of accidents as our response variable. We compared the difference between periods when remedies were applied to periods without any remedy. We found a general 46% decrease in the number of accidents during years with a remedy compared to what would have been expected the same years without any remedy. Forest clearing and supplemental feeding seem to be reliable ways of reducing the number of collisions. Scent was only applied for short distances in a few years, and the beneficial effects we observed were questionable. We conclude that mitigative efforts may substantially reduce accidental mortality in moose populations if applied for long distances. We discuss the economics of game-vehicle collisions by performing a simple calculation to visualize the need for a bio-economic approach to the problem.

The Fortymile Caribou Herd (FCH) is the most prominent caribou herd in interior Alaska. A large portion of the FCH calving and summer range lies beneath heavily used Military Operations Areas (MOA) that are important for flight training. We observed the behavior of Grant's cow caribou (Rangifer tarandus granti) and their calves before, during, and immediately following low-level military jet overflights. We also monitored movements of radiocollared cow caribou and survival of their calves. We conducted fieldwork from mid May through early June 2002. We concluded that military jet overflights did not cause deaths of caribou calves in the FCH during the calving period nor result in increased movements of cow–calf pairs over the 24-hour period following exposure to overflights. Short-term responses to overflights were generally mild in comparison to caribou reactions to predators or perceived predators. Caribou responses to overflights were variable, but responses were generally greater as slant distances decreased and jet speeds increased. A-10 jets caused less reaction than F-15s and F-16s. Although we found that short-term reactions of caribou to jet overflights were mild, we advise against assuming there are no long-term effects on calving caribou from jet overflights.

A large barren-ground caribou (Rangifer tarandus granti) population (the Bering Seacoast Herd) historically ranged across southwest Alaska. The size of this herd peaked in the early 1860s but declined by the late 1880s. Caribou numbers remained low in southwest Alaska for the next 100 years. Biologists have argued that periodic dispersal has been an important factor in caribou population dynamics. However, others conclude there was no credible evidence that significant interchange between herds has ever occurred in Alaska. Since 1981, we monitored 318 radiocollared caribou and documented dramatic population growth, erratic movements, shifts from traditional ranges, and changes in migratory behavior. We also documented shifts in calving distribution that may contrast with conventional concepts of calving tradition and herd identity. Some biologists have concluded caribou herds can be considered closed populations for management purposes because the number of dispersing caribou is so small that it has no influence on population dynamics. We propose that the current definition of a herd may be appropriate for short-term management; however, over long time frames and large spatial scales, metapopulations may better describe caribou ecology and be more useful in long-term caribou conservation.

Reindeer (Rangifer tarandus) is considered one of the most polygynous ungulates, so high variance in male mating success is expected. We assessed the variance in male mating success in a wild reindeer population with natural adult sex ratio in order to test this assumption. We obtained samples from 321 calves from the same year class and from 206 adult males. These samples represented an estimated 84% of the potential sires in the population. By analyses of 23 microsatellite loci and use of a likelihood ratio approach, we assigned 49% and 90% of the calves a sire at 95% and 80% confidence interval, respectively. Approximately 50% of the assigned calves had 1 or no paternal half-sibs, and approximately two thirds of the males contributed genetically to the cohort. At 80% confidence interval, 2.9% of the males sired >7 offspring, and 20 calves were assigned to the most successful male. Mean and variance in the number of calves assigned by different male age classes increased with age, although several of the yearling males were more successful than some of the adults. Our findings did not support the traditional assumption that most reindeer calves were fathered by a small group of highly successful males, suggesting that alternative mating strategies may play a more prominent role in reindeer than previously expected. Within a sustainable management concept, our findings demonstrate the importance of maintaining a reasonable proportion of young and adult males in the population.

Collecting skin biopsies from large whales for genetic analysis is often subject to national permit, and in the case of cow-calf pairs, it may be prohibited. We present results of 906 biopsy attempts on southern right whales (Eubalaena australis) in South African waters between 1995 and 1997, including 147 cow-calf pairs. Our sampling success was higher for biopsy darts with a bore of 4 mm compared to 4.6 mm. Contact periods averaged 17.7 min for cow-calf pairs and 25.4 min for whales unaccompanied by calves. There were no significant differences in the short-term reactions of males and females to biopsying, but the reaction of single animals of either sex was greater than for larger groups. Cows accompanied by calves had the strongest reactions, which were significantly greater than even single females. We found evidence of sensitization to repeat biopsying (over periods of hours to 65 days) for cows but not calves (n = 20). We compared the subsequent reproductive history of 117 biopsied cows with that of 163 unbiopsied cows from the same years, and we compared the distribution of calving intervals for biopsied animals with 829 intervals recorded from 1985 to 1995. We did not detect any adverse effects on the proportion of successful reproductive cycles, and hence calf survival, or the proportion of longer-than-normal cycles, although the power of all the statistical tests was low. We concluded that any prohibition on the biopsy sampling of cow-calf pairs should be carefully reconsidered in the light of the valuable genetic insights such sampling could achieve.

Early hatched waterfowl are more likely to enter the breeding population. Managers' primary tool to increase nesting success in the Prairie Pothole Region (PPR) of North America is managing upland vegetation for duck nesting cover. To determine whether managed covertypes affect early-season nesting success, we modeled seasonal variation in nesting success using >17,000 duck nests found in managed and unmanaged covertypes in prairie Canada from 1993 to 2000. Nesting success was higher in most managed covertypes than in unmanaged covertypes early in the nesting season. Planted cover appeared to be the best managed covertype for increasing early-season nesting success as it had high early-season nesting success, and was selected by nesting ducks in greater proportion than its availability; however, nesting success in planted cover declined later in the nesting season while nesting success in most unmanaged covertypes increased. Nevertheless, even with reduced nesting success late in the season, planted cover was more productive than surrounding unmanaged covertypes. Future waterfowl management efforts should focus on providing safe nesting cover early in the nesting season.

We used a mark–recapture analysis combining data from live recaptures of paradise shelduck (Tadorna variegata) at multiple banding sites with reported recoveries from birds shot by hunters to analyze the movement of paradise shelduck between molting sites. Our model allows survival, recapture, band recovery, and movement probabilities to depend on the possibly unknown location of the birds at the time live recaptures are made. Temporary emigration can be included in the model using unobserved states in which capture cannot occur. Survival probabilities varied through time and were different in the first year following banding compared to other years. At 2 of the sites, an estimated 30% of birds emigrated each year with evidence that the emigration was permanent. At the third site, an estimated 92% emigrated, but there was evidence that the emigration was temporary.

We banded northern pintails (Anas acuta; n = 13,645) at a single site on the Yukon-Kuskokwim Delta (YKD), Alaska, USA, from 1990 to 2001. We used recaptures from our site in combination with hunter recoveries to model annual survival, recovery rates, and fidelity to our capture location. Most recoveries (>90%) occurred in the Pacific Flyway with 64% reported from California's Central Valley. Our top candidate models allowed survival to vary by sex but not by age or year. Estimated annual survival was 77.6% (95% CI: 73.9–81.0%) for males and 60.2% (95% CI: 53.2–67.0%) for females. Reporting rates varied by age, sex, and year; estimates for adult males exceeded those for adult females by 3.5 times. Within sexes, reporting rates of hatch-year pintails exceeded those of adults. Estimated recovery rates were considerably lower than those estimated during the 1950s–1970s for winter banded pintails (Hestbeck 1993b), but there were no differences in survival rates. This suggests that changes in harvest regulations may not have influenced annual survival in this population. The propensity of banded pintails to return to our capture site (fidelity rate) varied between sexes and was positively correlated with water conditions in prairie Canada. Our estimates of fidelity rates varied from 77.4% to 87.2% for males and 89.8% to 94.3% for females. Our fidelity estimates suggest that some level of subpopulation structuring may exist for northern pintails. Additionally, our estimates of fidelity support previous observations of northern pintails overflying poor wetland habitat conditions on the Canadian prairies.

We investigated the effects of prescribed understory fire and shrub density on a population of dusky-footed woodrats (Neotoma fuscipes) live-trapped each spring and fall from October 1993 to 2001. We performed model selection with program MARK on capture–recapture data using Pollock's robust design to estimate survival and temporary emigration. Abundance increased from 1993 to a peak in 1997 and then decreased steadily from 1997 to 2001, accompanied by reduced reproduction. Survival was age- and time-dependent. Differential temporary emigration probabilities were explained by home-range size differences. Survival and abundance were higher in plots with high shrub cover vs. plots with low shrub cover. A low-medium intensity, prescribed understory fire in the fall of 1997 had no negative effect on survival or temporary emigration. The only fire-related negative demographic effect was a reduction in the number of juveniles per adult female on burned plots in the spring immediately after the fire. Prescribed understory fire in oak woodland is unlikely to alter woodrat populations significantly if patches of well-structured habitat are maintained.

Naturally occurring stable isotopes in foodwebs can be used to determine the relative contributions of endogenous and exogenous nutrients to avian eggs in cases where birds move between isotopically distinct biomes or habitats to breed. We measured δ¹³C and δ¹⁵N values in somatic muscle tissues and eggs of Barrow's goldeneye (Bucephala islandica) together with those isotope values in amphipods from wetlands used by birds breeding on the Chilcotin Plateau in central British Columbia, Canada. Females that had recently arrived on the breeding grounds had muscle tissue isotope values similar to those found in coastal wintering birds and were considerably more enriched in ¹³C than were samples from local foodwebs. However, δ¹⁵N values of amphipods were highly variable among wetlands, resulting in a nondistinct exogenous δ¹⁵N endpoint for our dual-isotope mixing model. Therefore, we only used the model based on δ¹³C values to estimate nutrient sources to eggs. In 2000, first-laid eggs were more enriched in both isotopes than fourth- or eighth-laid eggs. Considerable endogenous protein input to egg yolk and albumen was detected for the first laid egg (yolk: range = 0–92.7%, median = 23.7%; albumen: range = 0–78.6%, median = 28.7%) with less endogenous contribution of somatic lipids (first egg: range = 0–100%, median = 4.9%). Using archived tissue samples of muscle and developing ovarian follicles from birds collected in 1993–1994, we found no δ¹³C isotopic evidence for endogenous protein contribution to egg yolk. Our results demonstrate the utility of the stable isotope approach in cases where isotopic endpoints are well established. Barrow's goldeneye showed a mixed strategy of endogenous vs. exogenous nutrient allocation to reproduction that varied by individual females, laying order, and year. We encourage managers to use this approach to quantify nutrient allocations from various biomes to reproduction in waterfowl to better understand the importance of wintering sites to reproduction.

Large raptors such as northern goshawks (Accipiter gentilis) frequently have been the focus of conservation efforts because of their large area requirements and habitat sensitivity. We examined landscape-scale habitat selection of breeding goshawks and created a habitat model of goshawk landscape use in an area (1,600 km²) that contained forested, agricultural, and urban landscapes in northern Japan. From a systematic nest survey and extensive radio telemetry, we delineated 36 male home ranges (18 of them were approximated with similar-sized plots around nests) and 44 similar-sized unoccupied plots over the study area. The proportions of forest cover in each home range varied widely (between 2 and 88%). Despite this variation, goshawk landscape use was estimated accurately (classification accuracy = 84%) with a stepwise logistic regression model; the probability of occupancy was positively associated with the proportion of forest interior (>200 m from the forest edge) and that of open land <200 m from the forest edge, but it was negatively associated with the proportion of water and the mode of the slope. In home ranges, the proportions of forest interior and open land <200 m from forest edge complemented each other with the change in forest cover in such a way that the sum of the 2 variables always occupied >30% of the home ranges. These results suggest that a preference for the 2 different habitats, both of which may have a high prey availability, enables goshawks to nest in landscapes with a wide variety of forest fragmentation. The adverse effect of water on goshawk occupancy was due to the tendency of goshawks not to include a large river (>800 m wide) in their home ranges. This suggests that goshawks need connectivity between forest patches in their home ranges. The proportion of urban land was 19% lower in home ranges than in unoccupied plots. These results indicate that management efforts should focus on maintaining landscapes with high prey availability and connectivity between forest patches and on preventing urbanization over entire home ranges.

Broadcast surveys using conspecific calls are currently the most effective method for detecting northern goshawks (Accipiter gentilis) during the breeding season. These surveys typically use alarm calls during the nestling phase and juvenile food-begging calls during the fledgling-dependency phase. Because goshawks are most vocal during the courtship phase, we hypothesized that this phase would be an effective time to detect goshawks. Our objective was to improve current survey methodology by evaluating the probability of detecting goshawks at active nests in northern Minnesota in 3 breeding phases and at 4 broadcast distances and to determine the effective area surveyed per broadcast station. Unlike previous studies, we broadcast calls at only 1 distance per trial. This approach better quantifies (1) the relationship between distance and probability of detection, and (2) the effective area surveyed (EAS) per broadcast station. We conducted 99 broadcast trials at 14 active breeding areas. When pooled over all distances, detection rates were highest during the courtship (70%) and fledgling-dependency phases (68%). Detection rates were lowest during the nestling phase (28%), when there appeared to be higher variation in likelihood of detecting individuals. EAS per broadcast station was 39.8 ha during courtship and 24.8 ha during fledgling-dependency. Consequently, in northern Minnesota, broadcast stations may be spaced 712 m and 562 m apart when conducting systematic surveys during courtship and fledgling-dependency, respectively. We could not calculate EAS for the nestling phase because probability of detection was not a simple function of distance from nest. Calculation of EAS could be applied to other areas where the probability of detection is a known function of distance.

A captive Attwater's prairie-chicken (Tympanuchus cupido attwateri; APC) breeding program was initiated in 1992 to avoid extinction of the APC, currently one of the most endangered birds in the United States. We addressed the efficacy of using pen-reared birds to supplement wild populations or to establish new populations. From 19 July 1996 to 30 October 1997, we released 119 pen-reared APC on Attwater Prairie Chicken National Wildlife Refuge (APCNWR) and Galveston Bay Prairie Preserve (GBPP), in an effort to supplement wild populations. We used radiotelemetry to determine survival, movements, and reproduction of pen-raised APC. We evaluated the effects of acclimation period, date of release, transmitter style, and release habitat on post-release survival. Birds held in acclimation pens for 14 days experienced higher 6-month survival than those held for 3 days (47.4% and 19.4%, respectively). We did not observe a difference in survival for different date of release, release habitats, or transmitter styles. Survival 2 weeks post-release was similar for birds released in 1996 and 1997 on APCNWR (61.7% and 50.8%, respectively) and on GBPP (75.0% and 81.8%, respectively). Movements and monthly ranges for juvenile males and females were similar to those of wild APC. Overall site and year nest success was 44%. Recruitment on APCNWR was zero, while recruitment on GBPP was unknown. Future research is needed to determine why broods did not survive.

Lesser prairie-chicken (Tympanuchus pallidicinctus) populations have declined rangewide, and one of the principal causes is thought to be low nest success. Little is known about the relationship of vegetation structure and human intrusion to lesser prairie-chicken nest location and success. We conducted our study from 1997 to 2002 in southwestern Kansas, USA, on 2 sand-sagebrush (Artemisia filifolia) prairie areas managed for livestock production. We determined apparent nest success (26%) for 200 of 209 lesser prairie-chicken nests located. Nest sites had taller grass, greater sand-sagebrush density, and higher visual obstruction than random locations in the surrounding prairie. We recorded the distances from nests to 6 anthropogenic features (wellheads, buildings, improved roads, unimproved roads, transmission lines, center-pivot irrigation fields) to determine whether the features were related to nest location and success. Sand-sagebrush habitat around 5 of 6 features (all except unimproved roads) was avoided for 80 m (wellheads) to >1,000 m (buildings) by nesting lesser prairie-chickens, but distances to the features were not substantial predictors of apparent nest success. Grass height, sagebrush plant density, and sagebrush height were the most important vegetation characteristics influencing nest success.

Populations of the lesser prairie-chicken (Tympanuchus pallidicinctus), an endemic grouse of the south-central United States, have declined precipitously. This species occurs in short- and mixed-grass prairies with sandy soils. Apart from perennial grasses of short stature, prairie-chicken habitat is characterized by dryland shrubs of the sand shinnery community, particularly the shinnery oak (Quercus havardii) and sand sagebrush (Artemisia filifolia). We measured microhabitat and microclimate characteristics at bird-centered and random points at the southwestern (New Mexico) and northeastern (Oklahoma) edges of the species' range. We estimated survival by locating radio-tagged prairie-chickens (n = 544) from April 1999 to June 2003. We found that lesser prairie-chickens used sites within the sand shinnery community that had a higher cover and greater density of shrubs (ANOVA: P < 0.0001). Microclimate differed substantially between occupied and random sites (MANOVA: P < 0.0001), and prairie-chicken survival was higher in microhabitat that was cooler, more humid, and less exposed to the wind. Survivorship was higher for adults that chose microhabitat with a higher cover of shrubs and grasses and a higher density of vegetation. Survivorship was higher for prairie-chickens that used sites with >20% cover of shrubs than for those choosing 10–20% cover; in turn, survivorship was higher for prairie-chickens choosing 10–20% cover than for those choosing <10% cover (Cox regression: P < 0.05). Whereas vegetation may recover following moderate habitat disturbance, land managers applying herbicides or otherwise removing shrubs should understand the potentially negative effects of reduced shrub cover on adult survivorship of lesser prairie-chickens.

It is difficult to survey crevice-nesting seabirds because nest-sites are hard to identify and count, and the number of adult birds attending a colony can be extremely variable within and between days. There is no standardized method for surveying crevice-nesting horned puffins (Fratercula corniculata), and consequently little is known about abundance or changes in their numbers. We examined the variability in colony attendance of horned puffins at 5 breeding colonies in the North Pacific to assess whether variation in count data can be reduced to a level that would allow us to detect changes in the number of birds attending a colony. We used within-year measures of variation in attendance to examine the power to detect a change in numbers between 2 years, and we used measures of among-year variation to examine the power to detect trends over multiple years. Diurnal patterns of attendance differed among colonies, and among-day variation in attendance was generally lowest from mid- to late-incubation to early chick rearing. Within-year variation in water counts was lower than in land counts, and variation was lower using a daily index based on 5 counts per day than it was using 1 count per day. Measures of among-year variation in attendance also were higher for land-based than water-based counts, and they were higher when we used a 10-day survey period than when we used a 30-day period. The use of either 1 or 5 counts a day during the colony-specific diurnal peak of attendance had little influence on levels of among-year variation. Overall, our study suggests that variation in count data may be reduced to a level that allows detection of trends in numbers. However, more studies of interannual variability in horned puffin attendance are needed. Further, the relationship between count data and breeding population size needs more study before the number of birds present at the colony can be used with confidence as an index of population trend.