Jason T. Fisher

REVIEW: Wildlife camera trapping: a review and recommendations for linking surveys to ecological processes

Summary Reliable assessment of animal populations is a long-standing challenge in wildlife ecology. Technological advances have led to widespread adoption of camera traps (CTs) to survey wildlife distribution, abundance and behaviour. As for any wildlife survey method, camera trapping must contend with sources of sampling error such as imperfect detection. Early applications focused on density estimation of naturally marked species, but there is growing interest in broad-scale CT surveys of unmarked populations and communities. Nevertheless, inferences based on detection indices are controversial, and the suitability of alternatives such as occupancy estimation is debatable. We reviewed 266 CT studies published between 2008 and 2013. We recorded study objectives and methodologies, evaluating the consistency of CT protocols and sampling designs, the extent to which CT surveys considered sampling error, and the linkages between analytical assumptions and species ecology. Nearly two-thirds of studies surveyed more than one species, and a majority used response variables that ignored imperfect detection (e.g. presence–absence, relative abundance). Many studies used opportunistic sampling and did not explicitly report details of sampling design and camera deployment that could affect conclusions. Most studies estimating density used capture–recapture methods on marked species, with spatially explicit methods becoming more prominent. Few studies estimated density for unmarked species, focusing instead on occupancy modelling or measures of relative abundance. While occupancy studies estimated detectability, most did not explicitly define key components of the modelling framework (e.g. a site) or discuss potential violations of model assumptions (e.g. site closure). Studies using relative abundance relied on assumptions of equal detectability, and most did not explicitly define expected relationships between measured responses and underlying ecological processes (e.g. animal abundance and movement). Synthesis and applications. The rapid adoption of camera traps represents an exciting transition in wildlife survey methodology. We remain optimistic about the technologys promise, but call for more explicit consideration of underlying processes of animal abundance, movement and detection by cameras, including more thorough reporting of methodological details and assumptions. Such transparency will facilitate efforts to evaluate and improve the reliability of camera trap surveys, ultimately leading to stronger inferences and helping to meet modern needs for effective ecological inquiry and biodiversity monitoring.

Body mass explains characteristic scales of habitat selection in terrestrial mammals.

Niche theory in its various forms is based on those environmental factors that permit species persistence, but less work has focused on defining the extent, or size, of a species’ environment: the area that explains a species’ presence at a point in space. We proposed that this habitat extent is identifiable from a characteristic scale of habitat selection, the spatial scale at which habitat best explains species’ occurrence. We hypothesized that this scale is predicted by body size. We tested this hypothesis on 12 sympatric terrestrial mammal species in the Canadian Rocky Mountains. For each species, habitat models varied across the 20 spatial scales tested. For six species, we found a characteristic scale; this scale was explained by species’ body mass in a quadratic relationship. Habitat measured at large scales best-predicted habitat selection in both large and small species, and small scales predict habitat extent in medium-sized species. The relationship between body size and habitat selection scale implies evolutionary adaptation to landscape heterogeneity as the driver of scale-dependent habitat selection.

Spatial patterns of breeding success of grizzly bears derived from hierarchical multistate models.

Conservation programs often manage populations indirectly through the landscapes in which they live. Empirically, linking reproductive success with landscape structure and anthropogenic change is a first step in understanding and managing the spatial mechanisms that affect reproduction, but this link is not sufficiently informed by data. Hierarchical multistate occupancy models can forge these links by estimating spatial patterns of reproductive success across landscapes. To illustrate, we surveyed the occurrence of grizzly bears (Ursus arctos) in the Canadian Rocky Mountains Alberta, Canada. We deployed camera traps for 6 weeks at 54 surveys sites in different types of land cover. We used hierarchical multistate occupancy models to estimate probability of detection, grizzly bear occupancy, and probability of reproductive success at each site. Grizzly bear occupancy varied among cover types and was greater in herbaceous alpine ecotones than in low-elevation wetlands or mid-elevation conifer forests. The conditional probability of reproductive success given grizzly bear occupancy was 30% (SE = 0.14). Grizzly bears with cubs had a higher probability of detection than grizzly bears without cubs, but sites were correctly classified as being occupied by breeding females 49% of the time based on raw data and thus would have been underestimated by half. Repeated surveys and multistate modeling reduced the probability of misclassifying sites occupied by breeders as unoccupied to <2%. The probability of breeding grizzly bear occupancy varied across the landscape. Those patches with highest probabilities of breeding occupancy-herbaceous alpine ecotones-were small and highly dispersed and are projected to shrink as treelines advance due to climate warming. Understanding spatial correlates in breeding distribution is a key requirement for species conservation in the face of climate change and can help identify priorities for landscape management and protection.

Wolverine behavior varies spatially with anthropogenic footprint: implications for conservation and inferences about declines

Abstract Understanding a species’ behavioral response to rapid environmental change is an ongoing challenge in modern conservation. Anthropogenic landscape modification, or “human footprint,” is well documented as a central cause of large mammal decline and range contractions where the proximal mechanisms of decline are often contentious. Direct mortality is an obvious cause; alternatively, human‐modified landscapes perceived as unsuitable by some species may contribute to shifts in space use through preferential habitat selection. A useful approach to tease these effects apart is to determine whether behaviors potentially associated with risk vary with human footprint. We hypothesized wolverine (Gulo gulo) behaviors vary with different degrees of human footprint. We quantified metrics of behavior, which we assumed to indicate risk perception, from photographic images from a large existing camera‐trapping dataset collected to understand wolverine distribution in the Rocky Mountains of Alberta, Canada. We systematically deployed 164 camera sites across three study areas covering approximately 24,000 km2, sampled monthly between December and April (2007–2013). Wolverine behavior varied markedly across the study areas. Variation in behavior decreased with increasing human footprint. Increasing human footprint may constrain potential variation in behavior, through either restricting behavioral plasticity or individual variation in areas of high human impact. We hypothesize that behavioral constraints may indicate an increase in perceived risk in human‐modified landscapes. Although survival is obviously a key contributor to species population decline and range loss, behavior may also make a significant contribution.

Grizzly Bear Noninvasive Genetic Tagging Surveys: Estimating the Magnitude of Missed Detections.

Sound wildlife conservation decisions require sound information, and scientists increasingly rely on remotely collected data over large spatial scales, such as noninvasive genetic tagging (NGT). Grizzly bears (Ursus arctos), for example, are difficult to study at population scales except with noninvasive data, and NGT via hair trapping informs management over much of grizzly bears’ range. Considerable statistical effort has gone into estimating sources of heterogeneity, but detection error–arising when a visiting bear fails to leave a hair sample–has not been independently estimated. We used camera traps to survey grizzly bear occurrence at fixed hair traps and multi-method hierarchical occupancy models to estimate the probability that a visiting bear actually leaves a hair sample with viable DNA. We surveyed grizzly bears via hair trapping and camera trapping for 8 monthly surveys at 50 (2012) and 76 (2013) sites in the Rocky Mountains of Alberta, Canada. We used multi-method occupancy models to estimate site occupancy, probability of detection, and conditional occupancy at a hair trap. We tested the prediction that detection error in NGT studies could be induced by temporal variability within season, leading to underestimation of occupancy. NGT via hair trapping consistently underestimated grizzly bear occupancy at a site when compared to camera trapping. At best occupancy was underestimated by 50%; at worst, by 95%. Probability of false absence was reduced through successive surveys, but this mainly accounts for error imparted by movement among repeated surveys, not necessarily missed detections by extant bears. The implications of missed detections and biased occupancy estimates for density estimation–which form the crux of management plans–require consideration. We suggest hair-trap NGT studies should estimate and correct detection error using independent survey methods such as cameras, to ensure the reliability of the data upon which species management and conservation actions are based.

The Mountain Habitats Segmentation and Change Detection Dataset

In this paper, we present a challenging dataset for the purpose of segmentation and change detection in photographic images of mountain habitats. We also propose a baseline algorithm for habitats segmentation to allow for performance comparison. The dataset consists of high resolution image pairs of historic and repeat photographs of mountain habitats acquired in the Canadian Rocky Mountains for ecological surveys. With a time lapse of 70 to 100 years between the acquisition of historic and repeat images, these photographs contain critical information about ecological change in the Rockies. The challenging aspects of analyzing these image pairs come mostly from the perspective (oblique) view of the photographs and the lack of color information in the historic photographs. The baseline algorithm that we propose here is based on texture analysis and machine learning techniques. Classifier training and results validation are made possible by the availability of expert manual ground-truth segmentation for each image. The results obtained with the baseline algorithm are promising and serve as a reference for new and improved segmentation and change detection algorithms.

Spatial Mark-Resight for Categorically Marked Populations with an Application to Genetic Capture-Recapture

The estimation of animal population density is a fundamental goal in wildlife ecology and management, commonly met using mark recapture or spatial mark recapture (SCR) study designs and statistical methods. Mark-recapture methods require the identification of individuals; however, for many species and sampling methods, particularly noninvasive methods, no individuals or only a subset of individuals are individually identifiable. The unmarked SCR model, theoretically, can estimate the density of unmarked populations; however, it produces biased and imprecise density estimates in many sampling scenarios typically encountered. Spatial mark-resight (SMR) models extend the unmarked SCR model in three ways: 1) by introducing a subset of individuals that are marked and individually identifiable, 2) introducing the possibility of individual-linked telemetry data, and 3) introducing the possibility that the capture-recapture data from the survey used to deploy the marks can be used in a joint model, all improving the reliability of density estimates. The categorical spatial partial identity model (SPIM) improves the reliability of density estimates over unmarked SCR along another dimension, by adding categorical identity covariates that improve the probabilistic association of the latent identity samples. Here, we combine these two models into a “categorical SMR” model to exploit the benefits of both models simultaneously. We demonstrate using simulations that SMR alone can produce biased and imprecise density estimates with sparse data and/or when few individuals are marked. Then, using a fisher (Pekania pennanti) genetic capture-recapture data set, we show how categorical identity covariates, marked individuals, telemetry data, and jointly modeling the capture survey used to deploy marks with the resighting survey all combine to improve inference over the unmarked SCR model. As previously seen in an application of the categorical SPIM to a real-world data set, the fisher data set demonstrates that individual heterogeneity in detection function parameters, especially the spatial scale parameter σ, introduces positive bias into latent identity SCR models (e.g., unmarked SCR, SMR), but the categorical SMR model provides more tools to reduce this positive bias than SMR or the categorical SPIM alone. We introduce the possibility of detection functions that vary by identity category level, which will remove individual heterogeneity in detection function parameters than is explained by categorical covariates, such as individual sex. Finally, we provide efficient SMR algorithms that accommodate all SMR sample types, interspersed marking and sighting periods, and any number of identity covariates using the 2-dimensional individual by trap data in conjunction with precomputed constraint matrices, rather than the 3-dimensional individual by trap by occasion data used in SMR algorithms to date.

Density and distribution of a brown bear (Ursus arctos) population within the Caucasus biodiversity hotspot

Population declines and extirpations of large mammalian carnivores are major concerns for global biodiversity conservation. Many large carnivores are vulnerable to conflict with humans and attract conservation attention for their flagship appeal and ecological importance. Coexisting with carnivores requires an understanding of carnivore distribution and abundance relative to human activities and disturbances. Such knowledge is often hindered by the rare and elusive nature of carnivores and the lack of systematic ecological surveys in biodiverse regions facing high levels of threat. The Caucasus Ecoregion is one such biodiversity hotspot harboring several threatened mammal species for which there is a paucity of reliable data, including brown bears (Ursus arctos). Caucasus brown bear populations have declined significantly from historical times and may be isolated and vulnerable to disturbance from development activities such as mining, as well as increasing hunting pressure. To inform land-use planning and bear conservation in the Caucasus Ecoregion, we conducted systematic surveys in May–October 2015 in the foothills of the Caucasus Mountains within the Vayots Dzor region of Armenia. We used noninvasive genetic sampling, camera trapping, and statistical models that account for imperfect detection to estimate density and distribution of the bear population in the 1,000-km2 study area. Across 34 sampling sites, we obtained 3,163 camera-trap photos of brown bears and genotyped 28 individual bears (7 males and 21 females). Spatially explicit capture-recapture models revealed an unexpectedly high density of bears (59.4/1,000 km2; females = 44.6, 95% confidence interval, CI = 25.4–78.4; males = 14.8, 95% CI = 6.6–34.0), and multi-method occupancy models indicated that bears were distributed across most of the study area (ψ = 0.85; SE = 0.07). These results provide robust evidence that a significant population of brown bears persists in Armenias Vayots Dzor region, despite a history of hunting and habitat loss that have driven declines in brown bear populations throughout much of the Caucasus Ecoregion. Continued persistence of this flagship species may be threatened by mining, poaching, and other anthropogenic pressures in the region, underscoring the urgent need for strategic conservation planning, impact mitigation, and expanded ecological monitoring within this biodiversity hotspot.

Cumulative effects of climate and landscape change drive spatial distribution of Rocky Mountain wolverine (Gulo gulo L.)

Abstract Contemporary landscapes are subject to a multitude of human‐derived stressors. Effects of such stressors are increasingly realized by population declines and large‐scale extirpation of taxa worldwide. Most notably, cumulative effects of climate and landscape change can limit species’ local adaptation and dispersal capabilities, thereby reducing realized niche space and range extent. Resolving the cumulative effects of multiple stressors on species persistence is a pressing challenge in ecology, especially for declining species. For example, wolverines (Gulo gulo L.) persist on only 40% of their historic North American range. While climate change has been shown to be a mechanism of range retractions, anthropogenic landscape disturbance has been recently implicated. We hypothesized these two interact to effect declines. We surveyed wolverine occurrence using camera trapping and genetic tagging at 104 sites at the wolverine range edge, spanning a 15,000 km2 gradient of climate, topographic, anthropogenic, and biotic variables. We used occupancy and generalized linear models to disentangle the factors explaining wolverine distribution. Persistent spring snow pack—expected to decrease with climate change—was a significant predictor, but so was anthropogenic landscape change. Canid mesocarnivores, which we hypothesize are competitors supported by anthropogenic landscape change, had comparatively weaker effect. Wolverine population declines and range shifts likely result from climate change and landscape change operating in tandem. We contend that similar results are likely for many species and that research that simultaneously examines climate change, landscape change, and the biotic landscape is warranted. Ecology research and species conservation plans that address these interactions are more likely to meet their objectives.

Visualizing Category-Specific Changes in Oblique Photographs of Mountain Landscapes

Our paper proposes a method for visualizing the spatial distribution of classes for a multi-class image segmentation problem. We apply this method for the case of mountain landscape images, where classes are defined by landscape categories. The proposed method builds class-specific distribution maps. Our contribution is two-fold. First, the class-specific distribution maps allow for the visualization of class-specific changes computed from pairs of images depicting the same landscape at different moments in time. Second, these maps enable us to calculate prior class probabilities for statistical scene segmentation purposes.