Search

Removal of non-native fish populations can be crucial to the conservation of native species, but often presents a complex challenge for managers. The goal of Trojan Y chromosome (TYC) programs is to skew the non-native sex ratio until only males remain, leading to eradication. We present results from a simulation model used to explore effects of alternative management approaches on an in-progress mechanical removal and TYC program to eradicate non-native brook trout (Salvelinus fontinalis). Simulation results indicated that stocking fingerling YY males (∼137 mm) was more effective than stocking catchable-sized YY males (∼230 mm), although questions about intercohort competition warrant further investigation. Increasing the proportion of mature fingerling YY males reduced treatment time by increasing the number of YY male spawners and increasing density-dependent mortality on young, mature wild brook trout. Maximizing the spatial distribution of YY male releases may be crucial to program success but is also dependent upon immediate dispersal movements. Principles derived from our results can be broadly applied to the management of other aquatic invaded systems using TYC programs to eradicate non-native species.

In long-lived marine top-predators with delayed sexual maturity such as seabirds, adult survival is predicted to drive population dynamics. Major knowledge gaps exist for the cryptic sub-adult stages of the population. Yet as the oceans undergo dramatic change, investigating the trends of top-predator populations and their responses to environmental variability is key for a process-based understanding of climate change. Using integrated population models based on a longitudinal data set spanning nearly 2 decades, we investigated the demographic rates and environmental drivers shaping the dynamics of a longevous marine top-predator population, the black-browed albatross Thalassarche melanophris in the Falklands. We quantified the dynamics of the whole population, including its cryptic sub-adult stages, and examined the effects of candidate environmental covariates on adult survival and productivity parameters. We found that high survival rates of both adults and juveniles are the main contributors to the growth of this albatross population, the only one that is increasing in the South Atlantic. The breeding parameters were affected by environmental fluctuations through bottom-up processes, with a negative effect of higher sea surface temperatures and a positive effect of wind intensity. The relatively constant adult survival was influenced by deeper ecosystem changes, captured by large-scale indices (Southern Annular Mode). Approximately half of the total population is composed of sub-adults, and the population dynamics are highly sensitive to changes in juvenile survival; addressing current knowledge gaps in these cryptic life-history stages is therefore a conservation goal of primary importance.

Informative species abundance estimates are critical for guiding decisions around the conservation and management of ecological systems. There exist many methods for estimating abundance of frequently encountered species and populations with uniquely identifiable individuals. However, for wildlife populations with unmarked individuals that occur at low densities, there exist a variety of behaviors and characteristics that make effectively surveying and sampling challenging or uninformative. Examples of challenging characteristics include the elusive behaviors of low‐density species that occur in complex and rugged terrain. Such characteristics make detection difficult and surveys expensive, dangerous, and potentially biased. To address these challenges, we used a common, non‐invasive field survey method combined with a probability‐based study design and frequently utilized statistical model to estimate abundance of an unmarked mountain goat population in eastern Idaho. We developed a novel data analysis approach using an N‐mixture model that, together with spatially balanced random sampling and a double‐observer field data collection method, directly solves the problem of approximating statistical assumptions, including population closure. We demonstrate that a probability‐based sampling design not only is feasible, but also is important for estimating population parameters for unmarked and low‐density species. With this approach, we present a procedure that offers unbiased abundance estimates, empowering managers to track low‐density species’ population trends across time.

Recently developed methods, including time-to-event and space-to-event models, estimate the abundance of unmarked populations from encounter rates with camera trap arrays, addressing a gap in non-invasive wildlife monitoring. However, estimating abundance from encounter rates relies on assumptions that can be difficult to meet in the field, including random movement, population closure, and an accurate estimate of movement speed. Understanding how these models respond to violation of these assumptions will assist in making them more applicable in real-world settings. We used simulated walk models to test the effects of violating the assumptions of the time-to-event model under four scenarios: (1) incorrectly estimating movement speed, (2) violating closure, (3) individuals moving within simplified territories (i.e., movement restricted to partially overlapping circles), (4) and individuals clustering in preferred habitat. The time-to-event model was robust to closure violations, territoriality, and clustering when cameras were placed randomly. However, the model failed to accurately estimate abundance when movement speed was incorrectly estimated or cameras were placed non-randomly with respect to habitat. We show that the time-to-event model can provide unbiased estimates of abundance when some assumptions that are commonly violated in wildlife studies are not met. Having a robust method for estimating the abundance of unmarked populations with remote cameras will allow practitioners to monitor a more diverse array of populations noninvasively. With the time-to-event model, placing cameras randomly with respect to animal movement and accurately estimating movement speed allows unbiased estimation of abundance. The model is robust to violating the other assumptions we tested.

Population monitoring can take many different forms, and monitoring elusive and endangered species frequently involves a variety of sparse data from different sources. Small populations are often hard to sample precisely and without bias, so when estimates of vital rates like survival or recruitment point to conflicting population trends, it can be hard to determine which is more correct. Furthermore, data can be extremely hard to collect on small populations and it can be helpful to find a way to use all available hard‐won data. To address these issues, we developed an integrated population model (IPM) using all available data to estimate vital rates and abundance for a case study of an endangered woodland caribou (Rangifer tarandus caribou) population. This IPM allowed us to incorporate data from juvenile recruitment surveys, telemetry‐based survival, aerial population counts and mark–resight data, and non‐invasive capture–recapture DNA data to better understand population status and trend. We estimated survival, abundance, and recruitment of four age classes of male and female caribou: young, juveniles, subadults, and adults. The four‐age class structure of the IPM allowed us to estimate recruitment from reproductive‐aged female caribou alone, even though it can be difficult to distinguish age classes—and even sexes—in the field. As part of our IPM, we developed a novel mixture model to break apart data from different age classes when age is unobservable, as it typically is from non‐invasive DNA samples. This helped us decrease bias in juvenile and adult survival estimates from scat data, which was important to our understanding of the population dynamics. Overall, our integrated model provided more precise estimates of population trends than any one method (e.g., telemetry or non‐invasive DNA) alone. This IPM provides a useful, flexible tool for biologists to monitor populations and provides a valuable example of the benefits of integrated population modeling approaches for endangered species management and recovery.

Motion-activated cameras (camera traps) are increasingly used in ecological and management studies for remotely observing wildlife and have been regarded as among the most powerful tools for wildlife research. However, studies involving camera traps result in millions of images that need to be analyzed, typically by visually observing each image, in order to extract data that can be used in ecological analyses. We trained machine learning models using convolutional neural networks with the ResNet-18 architecture and 3,367,383 images to automatically classify wildlife species from camera trap images obtained from five states across the United States. We tested our model on an independent subset of images not seen during training from the United States and on an out-of-sample (or out-of-distribution in the machine learning literature) dataset of ungulate images from Canada. We also tested the ability of our model to distinguish empty images from those with animals in another out-of-sample dataset from Tanzania, containing a faunal community that was novel to the model. The trained model classified approximately 2,000 images per minute on a laptop computer with 16 gigabytes of RAM. The trained model achieved 98% accuracy at identifying species in the United States, the highest accuracy of such a model to date. Out-of-sample validation from Canada achieved 82% accuracy, and correctly identified 94% of images containing an animal in the dataset from Tanzania. We provide an R package (Machine Learning for Wildlife Image Classification; MLWIC) that allows the users to A) implement the trained model presented here and B) train their own model using classified images of wildlife from their studies. The use of machine learning to rapidly and accurately classify wildlife in camera trap images can facilitate non-invasive sampling designs in ecological studies by reducing the burden of manually analyzing images. We present an R package making these methods accessible to ecologists. We discuss the implications of this technology for ecology and considerations that should be addressed in future implementations of these methods.

In many socially monogamous species, divorce is a strategy used to correct for sub-optimal partnerships and is informed by measures of previous breeding performance. The environment affects the productivity and survival of populations, thus indirectly affecting divorce via changes in demographic rates. However, whether environmental fluctuations directly modulate the prevalence of divorce in a population remains poorly understood. Here, using a longitudinal dataset on the long-lived black-browed albatross (Thalassarche melanophris) as a model organism, we test the hypothesis that environmental variability directly affects divorce. We found that divorce rate varied across years (1% to 8%). Individuals were more likely to divorce after breeding failures. However, regardless of previous breeding performance, the probability of divorce was directly affected by the environment, increasing in years with warm sea surface temperature anomalies (SSTA). Furthermore, our state-space models show that warm SSTA increased the probability of switching mates in females in successful relationships. For the first time, to our knowledge, we document the disruptive effects of challenging environmental conditions on the breeding processes of a monogamous population, potentially mediated by higher reproductive costs, changes in phenology and physiological stress. Environmentally driven divorce may therefore represent an overlooked consequence of global change. Fundação para a Ciência e Tecnologia - FCT info:eu-repo/semantics/publishedVersion

Competition is a widespread interaction among carnivores, ultimately manifested through one or more dimensions of the species' ecological niche. One of the most explicit manifestations of competitive interactions regards spatial displacement. Its interpretation under a theoretical context provides an important tool to deepen our understanding of biological systems and communities, but also for wildlife management and conservation. We used Bayesian multispecies occupancy models on camera‐trapping data from multiple sites in Southwestern Europe (SWE) to investigate competitive interactions within a carnivore guild, and to evaluate how species' ecological traits are shaping coexistence patterns. Seventeen out of 26 pairwise interactions departed from a hypothesis of independent occurrence, with spatial association being twice as frequent as avoidance. Association behaviors were only detected among mesocarnivores, while avoidance mainly involved mesocarnivores avoiding the apex predator (n = 4) and mesocarnivore‐only interactions (n = 2). Body mass ratios, defined as the dominant over the subordinate species body mass, revealed an important negative effect ( βˆ=−0.38;CI95=−0.81𝑡𝑜−0.06 ) on co‐occurrence probability, and support that spatially explicit competitive interactions are mostly expressed by larger species able to dominate over smaller ones, with a threshold in body mass ratios of ~4, above which local‐scale intraguild coexistence is unlikely. We found a weak relationship between pairwise trophic niche overlap and the probability of coexistence ( βˆ=−0.19;CI95=−0.58𝑡𝑜0.21 ), suggesting that competition for feeding resources may not be a key driver of competition, at least at the scale of our analysis. Despite local‐scale avoidance, regional‐scale coexistence appears to be maintained by the spatial structuring of the competitive environment. We provide evidence that SWE ecosystems consist of spatially structured competitive environments, and propose that coexistence among near‐sized species is likely achieved through the interplay of “facultative” and “behavioral” character displacements. Factors influencing carnivore coexistence likely include context‐dependent density and trait‐mediated effects, which should be carefully considered for a sound understanding of the mechanisms regulating these communities., This study was funded by research projects CGL2009‐10741, funded by the Spanish Ministry of Science and Innovation and EU‐FEDER, and OAPN 352/2011, funded by the Spanish Organismo Autónomo Parques Nacionales. P. Monterroso enjoyed a Ph.D. grant from the Fundação para a Ciência e a Tecnologia (FCT) (ref: SFRH/BD/37795/2007) and a postdoctoral fellowship by EU‐FEDER funds through the Operational Programme for Competitiveness Factors—COMPETE and by National Funds through FCT—Foundation for Science and Technology (ref: UID/BIA/50027/2013, POCI‐01‐0145‐FEDER‐006821). F. Díaz‐Ruiz enjoyed a research contract “Juan de la Cierva” (ref: FJCI‐2015‐24949) from the Spanish Ministry of Economy, Industry, and Competitiveness.

Kit fox (Vulpes macrotis Merriam, 1888) populations in the Great Basin Desert have declined and are of increasing concern for managers. Increasing coyote (Canis latrans Say, 1823) abundance and subsequent intraguild interactions may be one cause for this decline. Concurrent monitoring of carnivores is challenging and therefore rarely conducted. One possible solution for monitoring elusive carnivores is using noninvasive genetic sampling. We used noninvasive genetic sampling to collect fecal DNA from kit foxes and coyotes and estimate their densities from 2013–2014 in Utah, USA. We identified individuals based on microsatellite genotypes and estimated density with multisession spatially explicit capture–recapture models. Mean kit fox density was 0.02 foxes·km−2, while coyote densities were up to four times greater (0.07–0.08 coyotes·km−2). Kit fox densities were significantly lower than densities in the 1950s but were comparable with estimates from the late 1990s, suggesting that populations may be stabilizing after a precipitous decline. Our kit fox density estimates were among the lowest documented for the species. Our coyote density estimate was the first reported in our region and revealed that despite seemingly high abundance, densities are low compared with other regions. Our results suggested that kit foxes may be able to coexist with coyotes.

Demographic monitoring is required in threatened species management, yet effective and efficient monitoring is challenging for species that are difficult to capture or susceptible to capture stress. One possible monitoring approach for such species is non-invasive genetic sampling with capture–recapture methods (genetic capture–recapture). We evaluated the performance of genetic capture–recapture in a challenging model system, monitoring the threatened Sonoran pronghorn Antilocapra americana sonoriensis. In an effort to determine the best (i.e. efficient, accurate, precise, cost-effective) method for abundance estimation, we used simulations to examine the optimal genetic capture–recapture faecal sampling design for this population. We simulated encounter histories for 100–300 individuals, with 0.33–3.33 samples/individual/session, in 1–3 sampling sessions. We explored trade-offs between sample size, number of sessions and multi-session (MARK) versus single-session (capwire) closed capture–recapture abundance estimators, and an accurate and precise estimate. We also compared the cost between the genetic capture–recapture approaches and current aerial monitoring methods. Abundance was biased positively in capwire and negatively in MARK. Bias increased and precision decreased with fewer samples/individual/session. Annual genetic capture–recapture monitoring cost was nearly twice the cost of aerial surveys, although genetic capture–recapture methods provided much higher precision. However at the current estimated abundance (c. 200), the same level of precision achieved with aerial methods can be obtained by collecting 0.75 samples/individual in a single session, for an annual cost saving of > USD 4,000. This approach of comparing estimator performance and cost can easily be applied to other systems and is a useful evaluation for managers to implement prior to designing capture–recapture studies.

We studied long-term trends in catches per unit effort (CPUE; fish h-1) of four sport fish taxa in the Flathead River, Montana, following changes in hydroelectric operations at the Slis Ksanka Qisp Project (SKQP; formerly Kerr Dam). Prior to 1997 SKQP operations caused frequent, unnatural flow fluctuations. In 1997 the dam was changed from a power-peaking and load-following facility to a base-load facility. The new operations included seasonal minimum flows and ramping rates that greatly reduced flow fluctuations. Autumn trends in CPUEs of two size classes (substock and stock) of northern pike (Esox lucius), Oncorhynchus spp., brown trout (Salmo trutta), and smallmouth bass (Micropterus dolomieu) were monitored using nighttime electrofishing during 1984–86 before operational changes and again at the onset of operational changes in 1998 and continuing through 2008. We observed little difference in CPUEs between 1984–86 and 1998, the first year after changes, but we documented strong increasing trends in CPUEs of both sizes of all taxa, except stock northern pike, over the long term following operational changes. We also examined long-term patterns in the size structure (total length; TL mm) of fishes following operational changes. All taxa had either initial downward shifts in median TL, decreases in minimum sizes of fish captured, or both, a pattern consistent with enhanced recruitment and survival of smaller fishes. We conclude that modifications in dam operations were associated with increases of four sport fish taxa in the Flathead River and that similar normative flow applications might benefit riverine fish populations elsewhere.

Partial migration is a common life-history strategy among ungulates living in seasonal environments. The decision to migrate or remain on a seasonal range may be influenced strongly by access to high-quality habitat. We evaluated the influence of access to winter habitat of high quality on the probability of a female white-tailed deer (Odocoileus virginianus) migrating to a separate summer range and the effects of this decision on survival. We hypothesized that deer with home ranges of low quality in winter would have a high probability of migrating, and that survival of an individual in winter would be influenced by the quality of their home range in winter. We radiocollared 67 female white-tailed deer in 2012 and 2013 in eastern Washington, United States. We estimated home range size in winter using a kernel density estimator; we assumed the size of the home range was inversely proportional to its quality and the proportion of crop land within the home range was proportional to its quality. Odds of migrating from winter ranges increased by 3.1 per unit increase in home range size and decreased by 0.29 per unit increase in the proportion of crop land within a home range. Annual survival rate for migrants was 0.85 (SD = 0.05) and 0.84 (SD = 0.09) for residents. Our finding that an individual with a low-quality home range in winter is likely to migrate to a separate summer range accords with the hypothesis that competition for a limited amount of home ranges of high quality should result in residents having home ranges of higher quality than migrants in populations experiencing density dependence. We hypothesize that density-dependent competition for high-quality home ranges in winter may play a leading role in the selection of migration strategy by female white-tailed deer.

Genetic mark–recapture methods are increasingly being used to estimate demographic parameters in species where traditional techniques are problematic or imprecise. The federally endangered Indiana bat Myotis sodalis has declined dramatically and threats such as white-nose syndrome continue to afflict this species. To date, important demographic information for Indiana bats has been difficult to estimate precisely using traditional techniques such as emergence counts. Successful management and protection of Indiana bats requires better methods to estimate population sizes and survival rates throughout the year, particularly during summer when these bats reproduce and are widely dispersed away from their winter hibernacula. In addition, the familial makeup of maternity colonies is unknown, yet important for understanding local and regional population dynamics. We had four objectives in this study. For the first two objectives we investigated the potential use of DNA from fecal samples (fecal DNA) collected at roosts to obtain genetically based mark–recapture estimates of 1) colony size and 2) survival rates, for an Indiana bat maternity colony in Indianapolis, Indiana. The third objective was to compare our genetically based colony-size estimates with emergence counts conducted at the same roost tree to evaluate the genetic mark–recapture method. Our fourth objective was to use fecal DNA to estimate levels of relatedness among individuals sampled at the roost. In the summer of 2008, we collected fecal pellets and conducted emergence counts at a prominent roost tree during three time periods each lasting 7 or 8 d. We genotyped fecal DNA using five highly polymorphic microsatellite loci to identify individuals and used a robust-design mark–recapture approach to estimate survival rates as well as colony size at the roost tree. Emergence count estimates at the roost tree ranged from 100 to 215, whereas genetic mark–recapture estimates were higher, ranging from 122 to 266 and more precise. Apparent survival was 0.994 (SE = 0.04) between sampling periods suggesting that few bats died or permanently emigrated during the course of the study. Relatedness estimates, r, between all pairs of individuals averaged 0.055 ranging from 0 to 0.779, indicating that most individuals were not closely related. We demonstrate here the promise of using fecal DNA to estimate demographic information for Indiana bats and potentially other bat species.

Aim Characterizing animal habitat selection is central to ecology and conservation, and understanding selection across multiple spatial scales is a particular priority for research. However, the ability of selection models to capture multi-scale response has been limited by the dual analytical hurdles of cross-scale collinearity and overlapping landscapes. The aim of this study was to overcome these limitations using a novel, spatially hierarchical approach. Location North America's northern Great Plains (U.S.A. and Canada). Methods We developed a novel adaptation of the occupancy modelling framework that integrates animal response conditionally across scales. We then compared outcomes to those from a traditional multi-scale model. We illustrated our approach using the breeding distribution of two North American grassland songbirds of conservation concern, Sprague's Pipit Anthus spragueii and Chestnut-collared Longspur Calcarius ornatus. Results Our model successfully captured bird response to local habitat within a broader landscape context, even when habitat associations occurred in opposite directions across scale. Probabilities of occupancy were more strongly affected by local conditions when landscape context was favourable than when it was unfavourable. The traditional multi-scale approach extended problems of scale into spatial predictions by over-estimating occurrence where conditions were locally favourable but regionally unsuitable. Main conclusions The spatially hierarchical approach provides an integrated model of habitat selection across scales by allowing broader landscape context to shape local response to conditions. For grassland songbirds, our application enabled targeting that could enhance the expected benefits of conservation when compared to the traditional approach.

To successfully respond to changing habitat, climate or harvest, managers need to identify the most effective strategies to reverse population trends of declining species and/or manage harvest of game species. A classic approach in conservation biology for the last two decades has been the use of matrix population models to determine the most important vital rates affecting population growth rate (λ), that is, sensitivity. Ecologists quickly realized the critical role of environmental variability in vital rates affecting λ by developing approaches such as life-stage simulation analysis (LSA) that account for both sensitivity and variability of a vital rate. These LSA methods used matrix-population modeling and Monte Carlo simulation methods, but faced challenges in integrating data from different sources, disentangling process and sampling variation, and in their flexibility. Here, we developed a Bayesian integrated population model (IPM) for two populations of a large herbivore, elk (Cervus canadensis) in Montana, USA. We then extended the IPM to evaluate sensitivity in a Bayesian framework. We integrated known-fate survival data from radio-marked adults and juveniles, fecundity data, and population counts in a hierarchical population model that explicitly accounted for process and sampling variance. Next, we tested the prevailing paradigm in large herbivore population ecology that juvenile survival of neonates

Juvenile survival is highly variable in ungulate populations and often influences their dynamics. However, this vital rate is difficult to estimate with common wildlife management methods. Yet managers would benefit from being able to predict juvenile survival to reliably assess population dynamics for harvest management. In the case of mule deer (Odocoileus hemionus), previous studies reported that overwinter survival is the demographic parameter that influences population dynamics. We predicted winter survival of mule deer fawns under a range of habitat quality, weather, and predation regimes. We modeled overwinter survival of 2,529 fawns within 11 Population Management Units (PMU) in Idaho, 2003–2013. We used remotely sensed Normalized Difference Vegetation Index (NDVI) as a measure of summer plant productivity and both Moderate Resolution Infrared Spectroscopy Snow Data (MODIS SNOW) and the modeled Snow Data Assimilation System (SNODAS) as measures of winter snow conditions to capture spatiotemporal variation in winter survival. We used Bayesian hierarchical models to estimate survival, including covariates at the appropriate spatial and temporal resolution for each level: individual, capture site, PMU, and ecotype scales. We evaluated the predictive capacity of models using internal validation and external (out-of-sample) validation procedures comparing non-parametric Kaplan-Meier (KM) survival estimates with estimates from Bayesian hierarchical models. Statewide survival of fawns from 16 December to 1 June ranged from 0.32 to 0.71 during 2003 to 2013, with relatively low survival in 2006, 2008, and 2011 in most Game Management Units. Survival for individual PMUs ranged from 0.09 in the Weiser-McCall PMU in 2011 to 0.95 in the same PMU in 2005. Internal validation revealed models predicted KM survival well, over a range of R2 from 0.78 to 0.82, with the most complex model explaining the most variance as expected. However, because our goal was to predict winter survival of mule deer in the future, we evaluated our candidate models by withholding 2 years of data and then predicted those years with each model. The best-supported predictive model was our simplest model with 3 covariates, accounting for 71% of the variance in withheld years. Forage quality in late summer-fall increased winter mule deer survival, whereas early and late winter snow cover decreased survival. At finer-spatial scales within ecotypes, our internal validation was slightly better in aspen (0.86), similar in conifer (0.80), and poorer in shrub-steppe (0.60) ecotypes than our best statewide overall survival models, which accounted for 82% of the variance. Our analyses demonstrate the generality versus precision tradeoff across ecotypes and spatial scales to understand the extent that our survival models may be applied to different landscapes with varied predator communities, climate, and plant nutrition. © 2016 The Wildlife Society.

Abundance estimators are often evaluated with simulations, or by comparing estimates to populations of known size. Advances in noninvasive genetic sampling have fueled an increase in the use of noninvasive genetic sampling-based capture–recapture. However, when working with free-ranging populations of unknown size, managers often lack data necessary to select the appropriate estimator. This leads to uncertainty regarding how choice of estimator or sampling design influence estimates, and managers may select estimators based on funding or logistical constraints. Alternatively, comparing estimates from multiple estimators can provide managers with greater confidence in estimates, or highlight potential differences. We used noninvasive genetic sampling to estimate the abundance of free-ranging kit foxes Vulpes macrotis and coyotes Canis latrans. We generated estimates of abundance with two non-spatial likelihood-based estimators: 1) robust design Huggins capture–recapture models and 2) single-occasion capture with replacement (CAPWIRE) models. We compared these with recently published estimates derived from spatially explicit capture–recapture (SECR) models. For both species, estimates from Huggins models were generally lower than those from SECR. Abundance estimates from CAPWIRE, which was developed specifically for noninvasive genetic sampling and generates estimates from a single sampling occasion, tended to be biased low with high precision. Our results suggest that choice of estimator and sampling design can significantly influence estimates, and that the relationship between estimators varied between species. Our results further suggest that single-occasion sampling often employed with CAPWIRE abundance estimation may produce biased results and be inappropriate for species requiring dispersed sampling strategies.

Connectivity is important for population persistence and can reduce the potential for inbreeding depression. Connectivity between populations can also facilitate disease transmission; respiratory diseases are one of the most important factors affecting populations of bighorn sheep (Ovis canadensis). The mechanisms of connectivity in populations of bighorn sheep likely have implications for spread of disease, but the behaviors leading to connectivity between bighorn sheep groups are not well understood. From 2007–2012, we radio-collared and monitored 56 bighorn sheep in the Salmon River canyon in central Idaho. We used cluster analysis to define social groups of bighorn sheep and then estimated connectivity between these groups using a multi-state mark-recapture model. Social groups of bighorn sheep were spatially segregated and linearly distributed along the Salmon River canyon. Monthly probabilities of movement between adjacent male and female groups ranged from 0.08 (±0.004 SE) to 0.76 (±0.068) for males and 0.05 (±0.132) to 0.24 (±0.034) for females. Movements of males were extensive and probabilities of movement were considerably higher during the rut. Probabilities of movement for females were typically smaller than those of males and did not change seasonally. Whereas adjacent groups of bighorn sheep along the Salmon River canyon were well connected, connectivity between groups north and south of the Salmon River was limited. The novel application of a multi-state model to a population of bighorn sheep allowed us to estimate the probability of movement between adjacent social groups and approximate the level of connectivity across the population. Our results suggest high movement rates of males during the rut are the most likely to result in transmission of pathogens among both male and female groups. Potential for disease spread among female groups was smaller but non-trivial. Land managers can plan grazing of domestic sheep for spring and summer months when males are relatively inactive. Removal or quarantine of social groups may reduce probability of disease transmission in populations of bighorn sheep consisting of linearly distributed social groups. © 2016 The Wildlife Society.

Monitoring carnivores is critical for conservation, yet challenging because they are rare and elusive. Few methods exist for monitoring wide-ranging species over large spatial and sufficiently long temporal scales to detect trends. Remote cameras are an emerging technology for monitoring large carnivores around the world because of their low cost, non-invasive methodology, and their ability to capture pictures of species of concern that are difficult to monitor. For species without uniquely identifiable spots, stripes, or other markings, cameras collect detection/non-detection data that are well suited for monitoring trends in occupancy as its own independent useful metric of species distribution, as well as an index for abundance. As with any new monitoring method, prospective power analysis is essential to ensure meaningful trends can be detected. Here we test camera-based occupancy models as a method to monitor changes in occupancy of a threatened species, grizzly bears (Ursus arctos), at large landscape scales, across 5 Canadian national parks (~ 21,000 km2). With n = 183 cameras, the top occupancy model estimated regional occupancy to be 0.79 across all 5 parks. We evaluate the statistical power to detect simulated 5–40% declines in occupancy between two sampling years and test applied questions of how power is affected by the spatial scale of interest (park level vs. regional level), the number of cameras deployed, and duration of camera deployment. We also explore several ecological mechanisms (i.e., spatial patterns) of decline in occupancy, and examine how power changes when focusing only on grizzly bears family groups. As hypothesized, statistical power increased with the number of cameras and with the number of days deployed. Power was unaffected, however, by the ecological mechanisms of decline, indicating that our systematic sampling design can detect a decline regardless of whether occupancy declined due to range edge attrition, ecological trap or other mechanisms. Despite their lower occupancy, power was similarly high for grizzly bear family groups compared to grizzly bears in general. We highlight which study design attributes contributed to high power and we provide advice for establishing cost-effective camera-based programs for monitoring large carnivore occupancy at large spatial scales.

In response to growing concern over lethal shipwhale collisions, a number of efforts have been developed intended to enhance the ability of ships to avoid whales. However, the effec- tiveness of avoidance by large ships depends upon the ships detecting whales at a distance suffi- cient to allow for an appropriate avoidance measure. Here we explore the issue of whale detection using over 3000 unique detections of humpback whales recorded by observers stationed aboard large cruise ships in Alaska, USA. We used point transect distance sampling methods to generate detection functions necessary to understand the probability of whale detection and how it varies with distance under different environmental and biological characteristics. Detection probability of surfacing whales decreased markedly with increasing distance from the ship. We found visibility and group size to be the most important variables influencing detection. The worst visibility condi- tions reduced detection probability to near 0 at 1000 m. Compared to detecting a single whale, a group of 2 or 3 whales almost doubled detection probability at 1000 m. Surface active behavior increased detection compared to spouting while showing no flukes. In southeastern Alaska, single whales that spouted during excellent visibility conditions were most commonly encountered and had a detection probability of 0.569 at 1000 m. Understanding the ability of mariners to detect whales at distances sufficient to invoke avoidance measures is a key component in the effective- ness of 'ships avoiding whales' and is germane to efforts to reduce lethal shipwhale collisions.

The greater sage-grouse (Centrocercus urophasianus) is a species of conservation concern in western North America that is experiencing ongoing population declines due to habitat loss, energy development, disease, and other factors. It is therefore imperative to have robust estimates of population size and trends in this species across its range as part of monitoring, management, and conservation efforts. Greater sage-grouse are typically monitored by conducting counts of males at breeding leks, but the relationship between this index and true population size is unknown. In an attempt to improve the analyses of this population index, we examined the potential of N-mixture models to evaluate population size, detection probability, and trend in greater sage-grouse using lek count data collected over space and time. We used simulations to test how well the models recovered abundance and growth rate parameters with increasingly sparse count data. We found that the models correctly recovered parameters for scenarios with both constant and variable detection probability, even with up to 90% of the data missing, where 95% of simulations contained the true population growth rate parameter value within the 95% credible interval. We then applied the model to 13 years of lek count data from Montana, USA, collected at 2 spatial scales. Statewide, we determined that the population was decreasing by 7% per year on average over this time period, and that mean annual detection probability ranged from 0.20 to 0.76. In contrast, regressions of naive counts over time showed a 9% annual decrease in population size with a confidence interval spanning 0. High interannual variability in detection probability demonstrates that naive counts do not accurately measure interannual variability in population size, and may lead to misleading trends in population size over time. Although N-mixture models still have limitations, they are a promising approach for conducting robust analyses of population trends for species that aggregate at discrete breeding sites, even when datasets are sparse or uneven. © 2016 The Wildlife Society.

Population abundance estimates are important for management but can be challenging to determine in low-density, wide-ranging, and endangered species, such as Sonoran pronghorn (Antilocapra americana sonoriensis). The Sonoran pronghorn population has been increasing; however, population estimates are currently derived from a biennial aerial count that does not provide survival or recruitment estimates. We identified individuals through noninvasively collected fecal DNA and used robust-design capture-recapture to estimate abundance and survival for Sonoran pronghorn in the United States from 2013 to 2014. In 2014 we generated separate population estimates for pronghorn gathered near 13 different artificial water holes and for pronghorn not near water holes. The population using artificial water holes had 116 (95% CI 102-131) and 121 individuals (95% CI 112-132) in 2013 and 2014, respectively. For all locations, we estimated there were 144 individuals (95% CI 132-157). Adults had higher annual survival probabilities (0.83, 95% CI 0.69-0.92) than fawns (0.41, 95% CI 0.21-0.65). Our use of targeted noninvasive genetic sampling and capture-recapture with Sonoran pronghorn fecal DNA was an effective method for monitoring a large proportion of the population. Our results provided the first survival estimates for this population in over 2 decades and precise estimates of the population using artificial water holes. Our method could be used for targeted sampling of broadly distributed species in other systems, such as in African savanna ecosystems, where many species congregate at watering sites.

Survey data improve population management, yet those data often have associated bias. We quantified one source of bias in moose survey data (observer detection probability, p), by using repeated ground-observations of calves-at-heel of radio-collared moose in Colorado, USA. Detection probabilities, which varied both spatially and temporally, were estimated using an occupancy-modelling framework. We provide an efficient offset for modelled calf-at-heel occupancy (ψ) estimates that accommodates summer calf mortality. Detection probabilities were most efficiently modelled with seasonal variation, with the lowest probability of detecting calves-at-heel occurring during parturition (i.e. May) and later autumn periods (after August). Our most efficiently modelled detection probability estimate for summer was 0.80 (SE = 0.05). During the four years of this study, ψ estimates ranged from 0.54–0.84 (SE = 0.08–0.11). Accounting for 91.7% monthly calf survival corrected ψ estimates downward (ψ = 0.42–0.65). Our results suggest that repeated ground-based observations of individual cow moose, during summer months, can be can a cost-effective strategy for estimating a productivity parameter for moose. Ground survey results can be further improved by accounting for calf mortality.

1. Motion-activated cameras (“camera traps”) are increasingly used in ecological and management studies for remotely observing wildlife and have been regarded as among the most powerful tools for wildlife research. However, studies involving camera traps result in millions of images that need to be analyzed, typically by visually observing each image, in order to extract data that can be used in ecological analyses.2. We trained machine learning models using convolutional neural networks with the ResNet-18 architecture and 3,367,383 images to automatically classify wildlife species from camera trap images obtained from five states across the United States. We tested our model on an independent subset of images not seen during training from the United States and on an out-of-sample (or “out-of-distribution” in the machine learning literature) dataset of ungulate images from Canada. We also tested the ability of our model to distinguish empty images from those with animals in another out-of-sample dataset from Tanzania, containing a faunal community that was novel to the model.3. The trained model classified approximately 2,000 images per minute on a laptop computer with 16 gigabytes of RAM. The trained model achieved 98% accuracy at identifying species in the United States, the highest accuracy of such a model to date. Out-of-sample validation from Canada achieved 82% accuracy, and correctly identified 94% of images containing an animal in the dataset from Tanzania. We provide an R package (Machine Learning for Wildlife Image Classification; MLWIC) that allows the users to A) implement the trained model presented here and B) train their own model using classified images of wildlife from their studies.4. The use of machine learning to rapidly and accurately classify wildlife in camera trap images can facilitate non-invasive sampling designs in ecological studies by reducing the burden of manually analyzing images. We present an R package making these methods accessible to ecologists. We discuss the implications of this technology for ecology and considerations that should be addressed in future implementations of these methods.

Broad scale population estimates of declining species are desired for conservation efforts. However, for many secretive species including large carnivores, such estimates are often difficult. Based on published density estimates obtained through camera trapping, presence/absence data, and globally available predictive variables derived from satellite imagery, we modelled density and occurrence of a large carnivore, the jaguar, across the species’ entire range. We then combined these models in a hierarchical framework to estimate the total population. Our models indicate that potential jaguar density is best predicted by measures of primary productivity, with the highest densities in the most productive tropical habitats and a clear declining gradient with distance from the equator. Jaguar distribution, in contrast, is determined by the combined effects of human impacts and environmental factors: probability of jaguar occurrence increased with forest cover, mean temperature, and annual precipitation and declined with increases in human foot print index and human density. Probability of occurrence was also significantly higher for protected areas than outside of them. We estimated the world’s jaguar population at 173,000 (95% CI: 138,000–208,000) individuals, mostly concentrated in the Amazon Basin; elsewhere, populations tend to be small and fragmented. The high number of jaguars results from the large total area still occupied (almost 9 million km2) and low human densities (< 1 person/km2) coinciding with high primary productivity in the core area of jaguar range. Our results show the importance of protected areas for jaguar persistence. We conclude that combining modelling of density and distribution can reveal ecological patterns and processes at global scales, can provide robust estimates for use in species assessments, and can guide broad-scale conservation actions.

Bias in vital rate estimation may come from failing to meet a variety of assumptions during the stages of sampling, monitoring, and analysis, though most are not commonly addressed in published studies. Here, we pay specific attention to various forms of censoring and truncation that present challenges for telemetry-based monitoring of survival. We use simulations to assess how uncertainty about times of death and imperfect detection probabilities affect Kaplan–Meier survival estimates. We then treat monitoring of threatened woodland caribou (Rangifer tarandus caribou) in west-central Alberta as a case study to test for potential effects of non-random right censoring and interval censoring on survival estimates. We report that monitoring frequency (e.g., daily vs. monthly) and associated uncertainty about the exact time of death do not inherently induce bias on Kaplan–Meier point estimates of survival nor affect estimates of variance. Removing individuals from the at-risk pool during intervals for which they were not detected did induce a negative bias on resulting survival estimates when the probability of detection was independent of the animals' fates. We recommend using subsequent detections to impute animals' fates during missed intervals in such cases when all animals' fates are eventually known or permanently right censored and when missed detections are not related to changes in mortality risk. Although some assumptions remain difficult to test and of continued concern, we find no evidence of biases in the methodology of Alberta's woodland caribou monitoring program. Our results lend credence to recent evidence of widespread declines in woodland caribou populations across the province. © 2015 The Wildlife Society.

Most abundance estimation methods assume that all sampled individuals are identified correctly. In practice, this assumption may be difficult to meet and can bias abundance estimates, especially when morphologically similar species overlap in range. Over the past 2 decades, Kittlitz's Murrelet (Brachyramphus brevirostris) populations appear to have declined across parts of their Alaskan range, where they co-occur with the Marbled Murrelet (B. marmoratus). Recently, the reliability of Kittlitz's Murrelet declines has been questioned due to variability and uncertainty in species identification between the 2 species. We conducted a field experiment to quantify misidentification and partial identification (identification to genus [Brachyramphus] level only) of Kittlitz's and Marbled murrelets during abundance surveys, and to evaluate the relative impacts of environmental and observational factors on misidentification and partial identification. We applied these results to previously collected survey ...

Recent major reductions in summer arctic sea ice extent could be expected to be affecting the distributions and life histories of arctic marine biota adapted to living adjacent to sea ice. Of major concern are the effects of ice reductions, and associated increasing SST, on the most abundant forage fish in the Arctic, Arctic cod (Boreogadus saida), the primary prey for the region’s upper trophic level marine predators. The black guillemot (Cepphus grylle mandtii) is an ice-obligate diving seabird specializing in feeding on Arctic cod and has been studied annually since 1975 at a breeding colony in the western Beaufort Sea. The data set is one of the few allowing assessment of the response of an upper trophic marine predator to recent decadal changes in the region’s cryosphere. Analysis of oceanographic conditions north of the colony from 1975 to 2012 for the annual period when parents provision young (mid-July to early September), found no major regime shifts in ice extent or SST until the late 1990s with major decreases in ice and increases in SST in the first decade of the 21st Century. We examined decadal variation in late summer oceanographic conditions, nestling diet and success, and overwinter adult survival, comparing a historical period (1975–1984) with a recent (2003–2012) one. In the historical period sea ice retreated an average of 1.8 km per day from 15 July to 1 September to an average distance of 95.8 km from the colony, while in the recent period ice retreat averaged 9.8 km per day to an average distance of 506.9 km for the same time period. SST adjacent to the island increased an average of 2.9 C between the two periods. While Arctic cod comprised over 95% of the prey provided to nestlings in the historical period, in the recent period 80% of the years had seasonal decreases, with Arctic cod decreasing to

Kittlitz's Murrelet (Brachyramphus brevirostris) is a dispersed-nesting seabird endemic to Alaska and eastern Russia that may have experienced considerable population declines in some parts of its range in the past few decades. Poor reproduction has been suggested as the demographic bottleneck, yet there are no direct estimates of reproduction in a glaciated area where this species reaches its highest densities at sea during the breeding season. The lack of demographic information in glacial habitats has limited our ability to interpret population trends and to clarify whether the presence of glaciers affects reproductive performance. Between 2007 and 2012, we radio-tagged Kittlitz's Murrelets to measure breeding propensity, nesting success, and fecundity in the heavily glaciated landscape of Icy Bay, Alaska, USA. Of 156 radio-tagged birds, 20% were breeders, 68% were potential breeders, and 12% were nonbreeders. Radio-tagged males (29%) were more likely to be breeders compared to females (11%). ...

Increasing elk (Cervus elaphus nelsoni) populations across the West in response to increased demand for recreational and hunting opportunities may have negative, unintended consequences for disease transmission risk. Historically, free-ranging elk populations were not thought to sustain brucellosis (Brucella abortus), but recent studies suggest increasing elk densities may result in free-ranging elk serving as maintenance hosts for the disease. We evaluated spatial variation in elk density, group sizes, and adult female brucellosis seroprevalence in 39 elk management districts in the Greater Yellowstone Ecosystem using a Bayesian approach. We used modeled relationships to estimate the effects of reducing elk density by 10–90% on grouping patterns and seroprevalence rates. Reducing the density of the 3 highest density elk herds by 10%, 50%, and 90% was predicted to result in a 9%, 39%, and 59% decrease in mean group size, whereas reducing the density of the 3 lowest density elk herds was predicted to result in only a 0%, 0.7%, and 1.3% decrease in mean group size. We estimated seroprevalence rates of 0.01–0.27 across management districts, and seroprevalence increased as elk density increased. For the 7 of 39 management districts with >10% estimated seroprevalence, 10%, 50%, and 90% reductions in elk density resulted in predicted mean seroprevalence reductions of 2%, 7%, and 9%, respectively. For the 14 management districts with ≤1% estimated seroprevalence, 10%, 50%, and 90% reductions in elk density resulted in no measurable change in predicted mean seroprevalence. Our results suggest that elk density has an important effect on elk group sizes, which may influence the risk of brucellosis transmission and resultant exposure rates. Manipulating elk density may in turn affect brucellosis seroprevalence rates. However, debate among the diverse stakeholders involved in elk management on the effectiveness of reducing density, group sizes, and brucellosis exposure rates in elk, relative to other interests and objectives, is necessary prior to manipulation of elk density for this purpose. © 2015 The Wildlife Society.