Reciprocated competition between two forest carnivores drives dietary specialization.

Competition shapes animal communities, but the strength of the interaction varies spatially depending on the availability and aggregation of resources and competitors. Among carnivores, competition is particularly pronounced with the strongest interactions between similar species with intermediate differences in body size. While ecologists have emphasized interference competition among carnivores based on dominance hierarchies from body size (smaller = subordinate; larger = dominant), the reciprocity of exploitative competition from subordinate species has been overlooked even though efficient exploitation can limit resource availability and influence foraging. Across North America, fishers Pekania pennanti and martens (Martes spp.) are two phylogenetically related forest carnivores that exhibit a high degree of overlap in habitat use and diet and differ in body size by a factor of 2-5×, eliciting particularly strong interspecific competition. In the Great Lakes region, fishers and martens occur both allopatrically and sympatrically; where they co-occur, the numerically dominant species varies spatially. This natural variation in competitors and environmental conditions enables comparisons to understand how interference and exploitative competition alter dietary niche overlap and foraging strategies. We analysed stable isotopes (δ13 C and δ15 N) from 317 martens and 132 fishers, as well as dietary items (n = 629) from 20 different genera, to compare niche size and overlap. We then quantified individual diet specialization and modelled the response to environmental conditions that were hypothesized to influence individual foraging. Martens and fishers exhibited high overlap in both available and core isotopic δ-space, but no overlap of core dietary proportions. When the competitor was absent or rare, both martens and fishers consumed more smaller-bodied prey. Notably, the dominant fisher switched from being a specialist of larger to smaller prey in the absence of the subordinate marten. Environmental context also influenced dietary specialization: increasing land cover diversity and prey abundance reduced specialization in martens whereas vegetation productivity increased specialization for both martens and fishers. Despite an important dominance hierarchy, fishers adjusted their niche in the face of a subordinate, but superior, exploitative competitor. These findings highlight the underappreciated role of the subordinate competitor in shaping the dietary niche of a dominant competitor.

Seasonality drives the survival landscape of a recovering forest carnivore in a changing world.

Ecological heterogeneity promotes species persistence and diversity. Environmental change has, however, eroded patterns of heterogeneity globally, stifling species recovery. To test the effects of seasonal heterogeneity on a reintroduced carnivore, American martens (Martes americana), we compared metrics of local and season-specific heterogeneity to traditional forest metrics on the survival of 242 individuals across 8 years and predicted a survival landscape for 13 reintroduction sites. We found that heterogeneity-created by forest structure in the growing season and snow in the winter-improved survival and outperformed traditional forest metrics. Spatial variation in heterogeneity created a distinct survival landscape, but seasonal change in heterogeneity generated temporal discordance. All translocation sites possessed high forest heterogeneity but there were greater differences in winter heterogeneity; recovery sites with the poorest snow conditions had the lowest viability. Our work links heterogeneity across seasons to fitness and suggests that management strategies that increase seasonal aspects of heterogeneity may help to recover other sensitive species to continuing environmental change.

Ecological forecasts reveal limitations of common model selection methods: predicting changes in beaver colony densities.

Over the past two decades, there have been numerous calls to make ecology a more predictive science through direct empirical assessments of ecological models and predictions. While the widespread use of model selection using information criteria has pushed ecology toward placing a higher emphasis on prediction, few attempts have been made to validate the ability of information criteria to correctly identify the most parsimonious model with the greatest predictive accuracy. Here, we used an ecological forecasting framework to test the ability of information criteria to accurately predict the relative contribution of density dependence and density-independent factors (forage availability, harvest, weather, wolf [Canis lupus] density) on inter-annual fluctuations in beaver (Castor canadensis) colony densities. We modeled changes in colony densities using a discrete-time Gompertz model, and assessed the performance of four models using information criteria values: density-independent models with (1) and without (2) environmental covariates; and density-dependent models with (3) and without (4) environmental covariates. We then evaluated the forecasting accuracy of each model by withholding the final one-third of observations from each population and compared observed vs. predicted densities. Information criteria and our forecasting accuracy metrics both provided strong evidence of compensatory density dependence in the annual dynamics of beaver colony densities. However, despite strong within-sample performance by the most complex model (density-dependent with covariates) as determined using information criteria, hindcasts of colony densities revealed that the much simpler density-dependent model without covariates performed nearly as well predicting out-of-sample colony densities. The hindcast results indicated that the complex model over-fit our data, suggesting that parameters identified by information criteria as important predictor variables are only marginally valuable for predicting landscape-scale beaver colony dynamics. Our study demonstrates the importance of evaluating ecological models and predictions with long-term data and revealed how a known limitation of information criteria (over-fitting of complex models) can affect our interpretation of ecological dynamics. While incorporating knowledge of the factors that influence animal population dynamics can improve population forecasts, we suggest that comparing forecast performance metrics can likewise improve our knowledge of the factors driving population dynamics.

Improving the accessibility and transferability of machine learning algorithms for identification of animals in camera trap images: MLWIC2.

Motion-activated wildlife cameras (or "camera traps") are frequently used to remotely and noninvasively observe animals. The vast number of images collected from camera trap projects has prompted some biologists to employ machine learning algorithms to automatically recognize species in these images, or at least filter-out images that do not contain animals. These approaches are often limited by model transferability, as a model trained to recognize species from one location might not work as well for the same species in different locations. Furthermore, these methods often require advanced computational skills, making them inaccessible to many biologists. We used 3 million camera trap images from 18 studies in 10 states across the United States of America to train two deep neural networks, one that recognizes 58 species, the "species model," and one that determines if an image is empty or if it contains an animal, the "empty-animal model." Our species model and empty-animal model had accuracies of 96.8% and 97.3%, respectively. Furthermore, the models performed well on some out-of-sample datasets, as the species model had 91% accuracy on species from Canada (accuracy range 36%-91% across all out-of-sample datasets) and the empty-animal model achieved an accuracy of 91%-94% on out-of-sample datasets from different continents. Our software addresses some of the limitations of using machine learning to classify images from camera traps. By including many species from several locations, our species model is potentially applicable to many camera trap studies in North America. We also found that our empty-animal model can facilitate removal of images without animals globally. We provide the trained models in an R package (MLWIC2: Machine Learning for Wildlife Image Classification in R), which contains Shiny Applications that allow scientists with minimal programming experience to use trained models and train new models in six neural network architectures with varying depths.

A SEROSURVEY OF DISEASES OF FREE-RANGING GRAY WOLVES (CANIS LUPUS) IN MINNESOTA, USA.

We tested serum samples from 387 free-ranging wolves ( Canis lupus ) from 2007 to 2013 for exposure to eight canid pathogens to establish baseline data on disease prevalence and spatial distribution in Minnesota's wolf population. We found high exposure to canine adenoviruses 1 and 2 (88% adults, 45% pups), canine parvovirus (82% adults, 24% pups), and Lyme disease (76% adults, 39% pups). Sixty-six percent of adults and 36% of pups exhibited exposure to the protozoan parasite Neospora caninum . Exposure to arboviruses was confirmed, including West Nile virus (37% adults, 18% pups) and eastern equine encephalitis (3% adults). Exposure rates were lower for canine distemper (19% adults, 5% pups) and heartworm (7% adults, 3% pups). Significant spatial trends were observed in wolves exposed to canine parvovirus and Lyme disease. Serologic data do not confirm clinical disease, but better understanding of disease ecology of wolves can provide valuable insight into wildlife population dynamics and improve management of these species.

Wolf (Canis lupus) Generation Time and Proportion of Current Breeding Females by Age.

Information is sparse about aspects of female wolf (Canis lupus) breeding in the wild, including age of first reproduction, mean age of primiparity, generation time, and proportion of each age that breeds in any given year. We studied these subjects in 86 wolves (113 captures) in the Superior National Forest (SNF), Minnesota (MN), during 1972-2013 where wolves were legally protected for most of the period, and in 159 harvested wolves from throughout MN wolf range during 2012-2014. Breeding status of SNF wolves were assessed via nipple measurements, and wolves from throughout MN wolf range, by placental scars. In the SNF, proportions of currently breeding females (those breeding in the year sampled) ranged from 19% at age 2 to 80% at age 5, and from throughout wolf range, from 33% at age 2 to 100% at age 7. Excluding pups and yearlings, only 33% to 36% of SNF females and 58% of females from throughout MN wolf range bred in any given year. Generation time for SNF wolves was 4.3 years and for MN wolf range, 4.7 years. These findings will be useful in modeling wolf population dynamics and in wolf genetic and dog-domestication studies.