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.

Mixed ancestry from wild and domestic lineages contributes to the rapid expansion of invasive feral swine.

Invasive alien species are a significant threat to both economic and ecological systems. Identifying the processes that give rise to invasive populations is essential for implementing effective control strategies. We conducted an ancestry analysis of invasive feral swine (Sus scrofa, Linnaeus, 1758), a highly destructive ungulate that is widely distributed throughout the contiguous United States, to describe introduction pathways, sources of newly emergent populations and processes contributing to an ongoing invasion. Comparisons of high-density single nucleotide polymorphism genotypes for 6,566 invasive feral swine to a comprehensive reference set of S. scrofa revealed that the vast majority of feral swine were of mixed ancestry, with dominant genetic associations to Western heritage breeds of domestic pig and European populations of wild boar. Further, the rapid expansion of invasive feral swine over the past 30 years was attributable to secondary introductions from established populations of admixed ancestry as opposed to direct introductions of domestic breeds or wild boar. Spatially widespread genetic associations of invasive feral swine to European wild boar deviated strongly from historical S. scrofa introduction pressure, which was largely restricted to domestic pigs with infrequent, localized wild boar releases. The deviation between historical introduction pressure and contemporary genetic ancestry suggests wild boar-hybridization may contribute to differential fitness in the environment and heightened invasive potential for individuals of admixed domestic pig-wild boar ancestry.

Detection error influences both temporal seroprevalence predictions and risk factors associations in wildlife disease models.

Understanding the prevalence of pathogens in invasive species is essential to guide efforts to prevent transmission to agricultural animals, wildlife, and humans. Pathogen prevalence can be difficult to estimate for wild species due to imperfect sampling and testing (pathogens may not be detected in infected individuals and erroneously detected in individuals that are not infected). The invasive wild pig (Sus scrofa, also referred to as wild boar and feral swine) is one of the most widespread hosts of domestic animal and human pathogens in North America.We developed hierarchical Bayesian models that account for imperfect detection to estimate the seroprevalence of five pathogens (porcine reproductive and respiratory syndrome virus, pseudorabies virus, Influenza A virus in swine, Hepatitis E virus, and Brucella spp.) in wild pigs in the United States using a dataset of over 50,000 samples across nine years. To assess the effect of incorporating detection error in models, we also evaluated models that ignored detection error. Both sets of models included effects of demographic parameters on seroprevalence. We compared our predictions of seroprevalence to 40 published studies, only one of which accounted for imperfect detection.We found a range of seroprevalence among the pathogens with a high seroprevalence of pseudorabies virus, indicating significant risk to livestock and wildlife. Demographics had mostly weak effects, indicating that other variables may have greater effects in predicting seroprevalence.Models that ignored detection error led to different predictions of seroprevalence as well as different inferences on the effects of demographic parameters.Our results highlight the importance of incorporating detection error in models of seroprevalence and demonstrate that ignoring such error may lead to erroneous conclusions about the risk associated with pathogen transmission. When using opportunistic sampling data to model seroprevalence and evaluate risk factors, detection error should be included.

Propagule size and structure, life history, and environmental conditions affect establishment success of an invasive species.

Population dynamics of species that are recently introduced into a new area, e.g., invasive species and species of conservation concern that are translocated to support global populations, are likely to be dominated by short-term, transient effects. Wild pigs (Sus scrofa, or wild boar) are pulsed-resource consumers of mast nuts that are commonly introduced into new areas. We used vital rate data (i.e., survival and fecundity) for wild pigs in Germany under varying forage conditions to simulate transient population dynamics in the 10-years following introduction into a new environment. In a low forage environment (i.e., conditions similar to their native range), simulated wild pig populations maintained a stable population size with low probability of establishment, while in environments with better quality forage (i.e., conditions similar to parts of their invasive range), high juvenile fecundity and survival facilitated rapid population growth and establishment probability was high. We identified a strategy for simulating population dynamics of species whose reproduction and survival depend on environmental conditions that fluctuate and for predicting establishment success of species introduced into a new environment. Our approach can also be useful in projecting near-term transient population dynamics for many conservation and management applications.

Rat eradication and the resistance and resilience of passerine bird assemblages in the Falkland Islands.

Norway rats (Rattus norvegicus) were introduced to the Falkland Islands and are detrimental to native passerines. Rat eradication programmes are being used to help protect the avifauna. This study assesses the effectiveness of eradication programmes while using this conservation practice as a natural experiment to explore the ecological resistance, resilience and homeostasis of bird communities. We conducted bird surveys on 230 islands: 85 in the presence of rats, 108 that were historically free of rats and 37 from which rats had been eradicated. Bird detection data were used to build occupancy models for each species and estimate species-area relationships. Count data were used to estimate relative abundance and community structure. Islands with invasive rats had reduced species richness of passerines and a different community structure than islands on which rats were historically absent. Although the species richness of native passerines was remarkably similar on eradicated and historically rat-free islands, community structure on eradicated islands was more similar to that of rat-infested islands than to historically rat-free islands. The results suggest that in the Falkland Islands, species richness of passerines is not resistant to invasive rats, but seems to be resilient following their removal. In contrast, community structure seems to be neither resistant nor resilient. From a conservation perspective, rat eradication programmes in the Falkland Islands appear to be effective at restoring native species richness, but they are not necessarily beneficial for species of conservation concern. For species that do not recolonize, translocations following eradications may be necessary.