Human disturbance compresses the spatiotemporal niche.

Human disturbance may fundamentally alter the way that species interact, a prospect that remains poorly understood. We investigated whether anthropogenic landscape modification increases or decreases co-occurrence-a prerequisite for species interactions-within wildlife communities. Using 4 y of data from >2,000 camera traps across a human disturbance gradient in Wisconsin, USA, we considered 74 species pairs (classifying pairs as low, medium, or high antagonism to account for different interaction types) and used the time between successive detections of pairs as a measure of their co-occurrence probability and to define co-occurrence networks. Pairs averaged 6.1 [95% CI: 5.3, 6.8] d between detections in low-disturbance landscapes (e.g., national forests) but 4.1 [3.5, 4.7] d between detections in high-disturbance landscapes, such as those dominated by urbanization or intensive agriculture. Co-occurrence networks showed higher connectance (i.e., a larger proportion of the possible co-occurrences) and greater proportions of low-antagonism pairs in disturbed landscapes. Human-mediated increases in species abundance (possibly via resource subsidies) appeared more important than behavioral mechanisms (e.g., changes in daily activity timing) in driving these patterns of compressed co-occurrence in disturbed landscapes. The spatiotemporal compression of species co-occurrences in disturbed landscapes likely strengthens interactions like competition, predation, and infection unless species can avoid each other at fine spatiotemporal scales. Regardless, human-mediated increases in co-occurrence with-and hence increased exposure to-predators or competitors might elevate stress levels in individual animals, with possible cascading effects across populations, communities, and ecosystems.

Snapshot Wisconsin: networking community scientists and remote sensing to improve ecological monitoring and management.

Biological data collection is entering a new era. Community science, satellite remote sensing (SRS), and local forms of remote sensing (e.g., camera traps and acoustic recordings) have enabled biological data to be collected at unprecedented spatial and temporal scales and resolution. There is growing interest in developing observation networks to collect and synthesize data to improve broad-scale ecological monitoring, but no examples of such networks have emerged to inform decision-making by agencies. Here, we present the implementation of one such jurisdictional observation network (JON), Snapshot Wisconsin, which links synoptic environmental data derived from SRS to biodiversity observations collected continuously from a trail camera network to support management decision-making. We use several examples to illustrate that Snapshot Wisconsin improves the spatial, temporal, and biological resolution and extent of information available to support management, filling gaps associated with traditional monitoring and enabling consideration of new management strategies. JONs like Snapshot Wisconsin further strengthen monitoring inference by contributing novel lines of evidence useful for corroboration or integration. SRS provides environmental context that facilitates inference, prediction, and forecasting, and ultimately helps managers formulate, test, and refine conceptual models for the monitored systems. Although these approaches pose challenges, Snapshot Wisconsin demonstrates that expansive observation networks can be tractably managed by agencies to support decision making, providing a powerful new tool for agencies to better achieve their missions and reshape the nature of environmental decision-making.

Abundance estimation of unmarked animals based on camera-trap data.

The rapid improvement of camera traps in recent decades has revolutionized biodiversity monitoring. Despite clear applications in conservation science, camera traps have seldom been used to model the abundance of unmarked animal populations. We sought to summarize the challenges facing abundance estimation of unmarked animals, compile an overview of existing analytical frameworks, and provide guidance for practitioners seeking a suitable method. When a camera records multiple detections of an unmarked animal, one cannot determine whether the images represent multiple mobile individuals or a single individual repeatedly entering the camera viewshed. Furthermore, animal movement obfuscates a clear definition of the sampling area and, as a result, the area to which an abundance estimate corresponds. Recognizing these challenges, we identified 6 analytical approaches and reviewed 927 camera-trap studies published from 2014 to 2019 to assess the use and prevalence of each method. Only about 5% of the studies used any of the abundance-estimation methods we identified. Most of these studies estimated local abundance or covariate relationships rather than predicting abundance or density over broader areas. Next, for each analytical approach, we compiled the data requirements, assumptions, advantages, and disadvantages to help practitioners navigate the landscape of abundance estimation methods. When seeking an appropriate method, practitioners should evaluate the life history of the focal taxa, carefully define the area of the sampling frame, and consider what types of data collection are possible. The challenge of estimating abundance of unmarked animal populations persists; although multiple methods exist, no one method is optimal for camera-trap data under all circumstances. As analytical frameworks continue to evolve and abundance estimation of unmarked animals becomes increasingly common, camera traps will become even more important for informing conservation decision-making.

Estimación de la Abundancia de Animales No Marcados con Base en Datos de Cámaras Trampa Resumen La rápida mejoría de las cámaras trampa en las décadas recientes ha revolucionado el monitoreo de la biodiversidad. A pesar de su clara aplicación en las ciencias de la conservación, las cámaras trampa han sido utilizadas pocas veces para modelar la abundancia de las poblaciones de animales no marcados. Buscamos resumir los retos que enfrenta la estimación de la abundancia de animales no marcados, compilar una perspectiva general de los marcos analíticos de trabajo existentes y proporcionar una guía para aquellos practicantes que buscan un método adecuado. Cuando una cámara registra múltiples detecciones de animales no marcados, no se puede determinar si las imágenes representan a diferentes individuos en movimiento o a un solo individuo que entra repetidamente a la zona de visión de la cámara. Sumado a esto, el movimiento animal ofusca una definición clara del área de muestreo y, como resultado, del área a la cual corresponde un estimado de abundancia. Después de reconocer estos retos, identificamos seis estrategias analíticas y revisamos 927 estudios con cámaras trampa publicados entre 2014 y 2019 para evaluar el uso y la prevalencia de cada método. Solamente en el 5% de los estudios se usó cualquiera de los métodos de estimación de abundancia que identificamos. La mayoría de estos estudios estimaron la abundancia local o las relaciones de covarianza en lugar de predecir la abundancia o la densidad a lo largo de áreas más amplias. Después, para cada estrategia analítica, recopilamos los requerimientos de datos, suposiciones, ventajas y desventajas para ayudar a los practicantes a navegar el paisaje de los métodos de estimación de abundancia. Cuando los practicantes busquen un método apropiado deberán evaluar la historia de vida del taxón focal, definir cuidadosamente el área del marco de muestreo y considerar cuáles tipos de recolección de datos son posibles. El reto de estimar la abundancia de poblaciones de animales no marcados persiste; aunque existan muchos métodos, no hay método único óptimo para los datos de las cámaras trampa que cumpla con todas las circunstancias. Mientras los marcos analíticos de trabajo sigan evolucionando y la estimación de la abundancia de animales no marcados sea cada vez más común, las cámaras trampa serán todavía más importantes para informar la toma de decisiones de conservación.

Making inference with messy (citizen science) data: when are data accurate enough and how can they be improved?

Measurement or observation error is common in ecological data: as citizen scientists and automated algorithms play larger roles processing growing volumes of data to address problems at large scales, concerns about data quality and strategies for improving it have received greater focus. However, practical guidance pertaining to fundamental data quality questions for data users or managers-how accurate do data need to be and what is the best or most efficient way to improve it?-remains limited. We present a generalizable framework for evaluating data quality and identifying remediation practices, and demonstrate the framework using trail camera images classified using crowdsourcing to determine acceptable rates of misclassification and identify optimal remediation strategies for analysis using occupancy models. We used expert validation to estimate baseline classification accuracy and simulation to determine the sensitivity of two occupancy estimators (standard and false-positive extensions) to different empirical misclassification rates. We used regression techniques to identify important predictors of misclassification and prioritize remediation strategies. More than 93% of images were accurately classified, but simulation results suggested that most species were not identified accurately enough to permit distribution estimation at our predefined threshold for accuracy (<5% absolute bias). A model developed to screen incorrect classifications predicted misclassified images with >97% accuracy: enough to meet our accuracy threshold. Occupancy models that accounted for false-positive error provided even more accurate inference even at high rates of misclassification (30%). As simulation suggested occupancy models were less sensitive to additional false-negative error, screening models or fitting occupancy models accounting for false-positive error emerged as efficient data remediation solutions. Combining simulation-based sensitivity analysis with empirical estimation of baseline error and its variability allows users and managers of potentially error-prone data to identify and fix problematic data more efficiently. It may be particularly helpful for "big data" efforts dependent upon citizen scientists or automated classification algorithms with many downstream users, but given the ubiquity of observation or measurement error, even conventional studies may benefit from focusing more attention upon data quality.

Compensatory mortality in a recovering top carnivore: wolves in Wisconsin, USA (1979-2013).

Populations of large terrestrial carnivores are in various stages of recovery worldwide and the question of whether there is compensation in mortality sources is relevant to conservation. Here, we show variation in Wisconsin wolf survival from 1979 to 2013 by jointly estimating the hazard of wolves' radio-telemetry ending (endpoint) and endpoint cause. In previous analyses, wolves lost to radio-telemetry follow-up (collar loss) were censored from analysis, thereby assuming collar loss was unconfounded with mortality. Our approach allowed us to explicitly estimate hazard due to collar loss and did not require censoring these records from analysis. We found mean annual survival was 76% and mean annual causes of mortality were illegal killing (9.4%), natural and unknown causes (9.5%), and other human-caused mortality such as hunting, vehicle collisions and lethal control (5.1%). Illegal killing and natural mortality were highest during winter, causing wolf survival to decrease relative to summer. Mortality was highest during early recovery and lowest during a period of sustained population growth. Wolves again experienced higher risk of human-caused mortality relative to natural mortality as wolves expanded into areas with more human activity. We detected partial compensation in human- and natural-caused mortality since 2004 as the population saturated more available habitat. Prior to 2004, we detected additivity in mortality sources. Assessments of wolf survival and cause of mortality rates and the finding of partial compensation in mortality sources will inform wolf conservation and management efforts by identifying sources and sinks, finding areas of conservation need, and assessing management zone delineation.

Harvest and group effects on pup survival in a cooperative breeder.

Recruitment in cooperative breeders can be negatively affected by changes in group size and composition. The majority of cooperative breeding studies have not evaluated human harvest; therefore, the effects of recurring annual harvest and group characteristics on survival of young are poorly understood. We evaluated how harvest and groups affect pup survival using genetic sampling and pedigrees for grey wolves in North America. We hypothesized that harvest reduces pup survival because of (i) reduced group size, (ii) increased breeder turnover and/or (iii) reduced number of female helpers. Alternatively, harvest may increase pup survival possibly due to increased per capita food availability or it could be compensatory with other forms of mortality. Harvest appeared to be additive because it reduced both pup survival and group size. In addition to harvest, turnover of breeding males and the presence of older, non-breeding males also reduced pup survival. Large groups and breeder stability increased pup survival when there was harvest, however. Inferences about the effect of harvest on recruitment require knowledge of harvest rate of young as well as the indirect effects associated with changes in group size and composition, as we show. The number of young harvested is a poor measure of the effect of harvest on recruitment in cooperative breeders.

Uncertainty and precaution in hunting wolves twice in a year: Reanalysis of Treves and Louchouarn.

Management of wolves is controversial in many jurisdictions where wolves live, which underscores the importance of rigor, transparency, and reproducibility when evaluating outcomes of management actions. Treves and Louchouarn 2022 (hereafter TL) predicted outcomes for various fall 2021 hunting scenarios following Wisconsin's judicially mandated hunting and trapping season in spring 2021, and concluded that even a zero harvest scenario could result in the wolf population declining below the population goal of 350 wolves specified in the 1999 Wisconsin wolf management plan. TL further concluded that with a fall harvest of > 16 wolves there was a "better than average possibility" that the wolf population size would decline below that 350-wolf threshold. We show that these conclusions are incorrect and that they resulted from mathematical errors and selected parameterizations that were consistently biased in the direction that maximized mortality and minimized reproduction (i.e., positively biased adult mortality, negatively biased pup survival, further halving pup survival to November, negatively biased number of breeding packs, and counting harvested wolves twice among the dead). These errors systematically exaggerated declines in predicted population size and resulted in erroneous conclusions that were not based on the best available or unbiased science. Corrected mathematical calculations and more rigorous parameterization resulted in predicted outcomes for the zero harvest scenario that more closely coincided with the empirical population estimates in 2022 following a judicially prevented fall hunt in 2021. Only in scenarios with simulated harvest of 300 or more wolves did probability of crossing the 350-wolf population threshold exceed zero. TL suggested that proponents of some policy positions bear a greater burden of proof than proponents of other positions to show that "their estimates are accurate, precise, and reproducible". In their analysis, TL failed to meet this standard that they demanded of others.

A phenology of fear: Investigating scale and seasonality in predator-prey games between wolves and white-tailed deer.

Predators and prey engage in games where each player must counter the moves of the other, and these games include multiple phases operating at different spatiotemporal scales. Recent work has highlighted potential issues related to scale-sensitive inferences in predator-prey interactions, and there is growing appreciation that these may exhibit pronounced but predictable dynamics. Motivated by previous assertions about effects arising from foraging games between white-tailed deer and canid predators (coyotes and wolves), we used a large and year-round network of trail cameras to characterize deer and predator foraging games, with a particular focus on clarifying its temporal scale and seasonal variation. Linear features were strongly associated with predator detection rates, suggesting these play a central role in canid foraging tactics by expediting movement. Consistent with expectations for prey contending with highly mobile predators, deer responses were more sensitive to proximal risk metrics at finer spatiotemporal scales, suggesting that coarser but more commonly used scales of analysis may miss useful insights into prey risk-response. Time allocation appears to be a key tactic for deer risk management and was more strongly moderated by factors associated with forage or evasion heterogeneity (forest cover, snow and plant phenology) than factors associated with the likelihood of predator encounter (linear features). Trade-offs between food and safety appeared to vary as much seasonally as spatially, with snow and vegetation phenology giving rise to a "phenology of fear." Deer appear free to counter predators during milder times of year, but a combination of poor foraging state, reduced forage availability, greater movements costs, and reproductive state dampen responsiveness during winter. Pronounced intra-annual variation in predator-prey interactions may be common in seasonal environments.

Correction: Demographic and Component Allee Effects in Southern Lake Superior Gray Wolves.

[This corrects the article DOI: 10.1371/journal.pone.0150535.].

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.

Demographic and Component Allee Effects in Southern Lake Superior Gray Wolves.

Recovering populations of carnivores suffering Allee effects risk extinction because positive population growth requires a minimum number of cooperating individuals. Conservationists seldom consider these issues in planning for carnivore recovery because of data limitations, but ignoring Allee effects could lead to overly optimistic predictions for growth and underestimates of extinction risk. We used Bayesian splines to document a demographic Allee effect in the time series of gray wolf (Canis lupus) population counts (1980-2011) in the southern Lake Superior region (SLS, Wisconsin and the upper peninsula of Michigan, USA) in each of four measures of population growth. We estimated that the population crossed the Allee threshold at roughly 20 wolves in four to five packs. Maximum per-capita population growth occurred in the mid-1990s when there were approximately 135 wolves in the SLS population. To infer mechanisms behind the demographic Allee effect, we evaluated a potential component Allee effect using an individual-based spatially explicit model for gray wolves in the SLS region. Our simulations varied the perception neighborhoods for mate-finding and the mean dispersal distances of wolves. Simulation of wolves with long-distance dispersals and reduced perception neighborhoods were most likely to go extinct or experience Allee effects. These phenomena likely restricted population growth in early years of SLS wolf population recovery.

Are the numbers adding up? Exploiting discrepancies among complementary population models.

Large carnivores are difficult to monitor because they tend to be sparsely distributed, sensitive to human activity, and associated with complex life histories. Consequently, understanding population trend and viability requires conservationists to cope with uncertainty and bias in population data. Joint analysis of combined data sets using multiple models (i.e., integrated population model) can improve inference about mechanisms (e.g., habitat heterogeneity and food distribution) affecting population dynamics. However, unobserved or unobservable processes can also introduce bias and can be difficult to quantify. We developed a Bayesian hierarchical modeling approach for inference on an integrated population model that reconciles annual population counts with recruitment and survival data (i.e., demographic processes). Our modeling framework is flexible and enables a realistic form of population dynamics by fitting separate density-dependent responses for each demographic process. Discrepancies estimated from shared parameters among different model components represent unobserved additions (i.e., recruitment or immigration) or removals (i.e., death or emigration) when annual population counts are reliable. In a case study of gray wolves in Wisconsin (1980-2011), concordant with policy changes, we estimated that a discrepancy of 0% (1980-1995), -2% (1996-2002), and 4% (2003-2011) in the annual mortality rate was needed to explain annual growth rate. Additional mortality in 2003-2011 may reflect density-dependent mechanisms, changes in illegal killing with shifts in wolf management, and nonindependent censoring in survival data. Integrated population models provide insights into unobserved or unobservable processes by quantifying discrepancies among data sets. Our modeling approach is generalizable to many population analysis needs and allows for identifying dynamic differences due to external drivers, such as management or policy changes.

A metapopulation approach to African lion (Panthera leo) conservation.

Due to anthropogenic pressures, African lion (Panthera leo) populations in Kenya and Tanzania are increasingly limited to fragmented populations. Lions living on isolated habitat patches exist in a matrix of less-preferred habitat. A framework of habitat patches within a less-suitable matrix describes a metapopulation. Metapopulation analysis can provide insight into the dynamics of each population patch in reference to the system as a whole, and these analyses often guide conservation planning. We present the first metapopulation analysis of African lions. We use a spatially-realistic model to investigate how sex-biased dispersal abilities of lions affect patch occupancy and also examine whether human densities surrounding the remaining lion populations affect the metapopulation as a whole. Our results indicate that male lion dispersal ability strongly contributes to population connectivity while the lesser dispersal ability of females could be a limiting factor. When populations go extinct, recolonization will not occur if distances between patches exceed female dispersal ability or if females are not able to survive moving across the matrix. This has profound implications for the overall metapopulation; the female models showed an intrinsic extinction rate from five-fold to a hundred-fold higher than the male models. Patch isolation is a consideration for even the largest lion populations. As lion populations continue to decline and with local extinctions occurring, female dispersal ability and the proximity to the nearest lion population are serious considerations for the recolonization of individual populations and for broader conservation efforts.

Osseous-associated cervical spondylomyelopathy in dogs: 27 cases (2000-2012).

OBJECTIVE: To evaluate the signalment, neurologic examination and imaging findings, and outcome in dogs treated medically or surgically for osseous-associated cervical spondylomyelopathy (OACSM). DESIGN: Retrospective case series. ANIMALS: 27 client-owned dogs. PROCEDURES: Medical records for dogs with OACSM (diagnosis made in 2000 through 2012) were reviewed. Collected data included signalment, neurologic examination findings (graded from 0 [normal] to 5 [tetraplegia]), imaging findings, treatment, and outcome. From MRI and CT images, measurements were obtained for subjective grading of spinal cord compression. RESULTS: Among the 27 dogs, the median age was 2 years; there were 15 Great Danes, 3 Mastiffs, 3 Newfoundlands, and 6 other large-breed dogs. For medically treated dogs (n = 7), the median initial neurologic grade was 2; for surgically treated dogs (20), the median initial neurologic grade was 3. Magnetic resonance imaging revealed dorsolateral spinal cord compression in 22 dogs and lateral spinal cord compression in 5 dogs. Dogs with more severe compressions were slightly more likely to undergo surgical than medical treatment. Median survival time of medically treated dogs was 43 months, and that of surgically treated dogs was 60 months. Fifteen of 19 dogs treated surgically had improved neurologic grades at 4 to 8 weeks after surgery and had a good to excellent long-term outcome. CONCLUSIONS AND CLINICAL RELEVANCE: Surgical treatment of dogs with OACSM resulted in neurologic improvement and was associated with a good long-term outcome. For dogs that received medical treatment, neurologic deterioration continued but some patients did well for several years.