Density‐dependent space use affects interpretation of camera trap detection rates

Camera traps (CTs) are an increasingly popular tool for wildlife survey and monitoring. Estimating relative abundance in unmarked species is often done using detection rate as an index of relative abundance, which assumes that detection rate has a positive linear relationship with true abundance. This assumption may be violated if movement behavior varies with density, but the degree to which movement behavior is density‐dependent across taxa is unclear. The potential confounding of population‐level relative abundance indices by movement would depend on how regularly, and by what magnitude, movement rate and home‐range size vary with density. We conducted a systematic review and meta‐analysis to quantify relationships between movement rate, home‐range size, and density, across terrestrial mammalian taxa. We then simulated animal movements and CT sampling to test the effect of contrasting movement scenarios on CT detection rate indices. Overall, movement rate and home‐range size were negatively correlated with density and positively correlated with one another. The strength of the relationships varied significantly between taxa and populations. In simulations, detection rates were related to true abundance but underestimated change, particularly for slower moving species with small home ranges. In situations where animal space use changes markedly with density, we estimate that up to thirty percent of a true change in relative abundance may be missed due to the confounding effect of movement, making trend estimation more difficult. The common assumption that movement remains constant across densities is therefore violated across a wide range of mammal species. When studying unmarked species using CT detection rates, researchers and managers should explicitly consider that such indices of relative abundance reflect both density and movement. Practitioners interpreting changes in camera detection rates should be aware that observed differences may be biased low relative to true changes in abundance. Further information on animal movement, or methods that do not depend on assumptions of density‐independent movement, may be required to make robust inferences on population trends.     

The Debate About Bait: A Red Herring in Wildlife Research

The use of bait (or attractants) to lure animals to a sampling site is common in wildlife research and important for optimizing species detection rates. The effect of bait on animal movement and space-use, however, is contested, fueled by concerns bait may affect animal movement and increase residency time. If founded, bait may bias parameter estimates from density, species distribution, resource selection, or behavioral models, produce spurious ecological inferences, and skew resulting management recommendations. To test whether animal movement varies with proximity to bait, we used high-resolution global positioning system telemetry data of 10 fishers (Pekania pennanti), temporally paired with 64 baited wildlife camera traps, to quantify the effect of bait on individual and population movement metrics. Although bait appeared to have a significant correlative effect on 1-hour movement segments, landscape characteristics had an effect 1.7 times greater, where the proportion of mixed forest and cultivation explained the majority of variability in animal movements. We contend that maximizing probability of detection and controlling or modeling local-scale landscape variability that could affect the probability of detection is a more important consideration in wildlife research than the effect of bait, which is eclipsed by differences incurred by natural habitat heterogeneity. Failing to maximize the probability of detection may obscure the modest bias potentially presented by the use of bait, or attractants, on ecological inference. © 2019 The Wildlife Society     

Integrating sensory ecology and predator-prey theory to understand animal responses to fire

Fire regimes are changing dramatically worldwide due to climate change, habitat conversion, and the suppression of Indigenous landscape management. Although there has been extensive work on plant responses to fire, including their adaptations to withstand fire and long-term effects of fire on plant communities, less is known about animal responses to fire. Ecologists lack a conceptual framework for understanding behavioural responses to fire, which can hinder wildlife conservation and management. Here, we integrate cue-response sensory ecology and predator-prey theory to predict and explain variation in if, when and how animals react to approaching fire. Inspired by the literature on prey responses to predation risk, this framework considers both fire-naïve and fire-adapted animals and follows three key steps: vigilance, cue detection and response. We draw from theory on vigilance tradeoffs, signal detection, speed-accuracy tradeoffs, fear generalization, neophobia and adaptive dispersal. We discuss how evolutionary history with fire, but also other selective pressures, such as predation risk, should influence animal behavioural responses to fire. We conclude by providing guidance for empiricists and outlining potential conservation applications.     

Wildlife surveillance using deep learning methods

相似文献   

Human vs. machine: Detecting wildlife in camera trap images

As the capacity to collect and store large amounts of data expands, identifying and evaluating strategies to efficiently convert raw data into meaningful information is increasingly necessary. Across disciplines, this data processing task has become a significant challenge, delaying progress and actionable insights. In ecology, the growing use of camera traps (i.e., remotely triggered cameras) to collect information on wildlife has led to an enormous volume of raw data (i.e., images) in need of review and annotation. To expedite camera trap image processing, many have turned to the field of artificial intelligence (AI) and use machine learning models to automate tasks such as detecting and classifying wildlife in images. To contribute understanding of the utility of AI tools for processing wildlife camera trap images, we evaluated the performance of a state-of-the-art computer vision model developed by Microsoft AI for Earth named MegaDetector using data from an ongoing camera trap study in Arctic Alaska, USA. Compared to image labels determined by manual human review, we found MegaDetector reliably determined the presence or absence of wildlife in images generated by motion detection camera settings (≥94.6% accuracy), however, performance was substantially poorer for images collected with time-lapse camera settings (≤61.6% accuracy). By examining time-lapse images where MegaDetector failed to detect wildlife, we gained practical insights into animal size and distance detection limits and discuss how those may impact the performance of MegaDetector in other systems. We anticipate our findings will stimulate critical thinking about the tradeoffs of using automated AI tools or manual human review to process camera trap images and help to inform effective implementation of study designs.     

Comparison of Methods of Monitoring Wildlife Crossing-Structures on Highways

ABSTRACT Wildlife crossing-structures (e.g., underpasses and overpasses) are used to mitigate deleterious effects of highways on wildlife populations. Evaluating performance of mitigation measures depends on monitoring structures for wildlife use. We analyzed efficacy of 2 noninvasive methods commonly used to monitor crossing-structure use by large mammals: tracking and motion-activated cameras. We monitored 15 crossing-structures every other day between 29 June and 24 October 2007 along the Trans-Canada Highway in Alberta, Canada. Our objectives were to determine how species-specific detection rates are biased by the detection method used, to determine factors contributing to crossing-event detection, and to evaluate the most cost-effective approach to monitoring. We detected 3,405 crossing events by tracks and 4,430 crossings events by camera for mammals coyote-sized and larger. Coyotes (Canis latrans) and grizzly bears (Ursus arctos) were significantly more likely to be detected by track-pads, whereas elk (Cervus elaphus) and deer (Odocoileus sp.) were more likely to be detected by cameras. Crossing-event detection was affected by species, track-pad length, and number of animals using the crossing structure. At the levels of animal activity observed in our study our economic analysis indicates that cameras are more cost-effective than track-pads for study durations >1 year. Understanding the benefits and limitations of camera and track-pad methods for monitoring large mammal movement at wildlife crossing-structures will help improve the efficiency of studies designed to evaluate the effectiveness of highway mitigation measures.     

物种相对多度指数在红外相机数据分析中的应用及局限

多度是衡量物种种群数量的参数之一, 多度的动态及其影响因素是种群生态学研究的经典问题。物种相对多度指数(relative abundance index, RAI)作为一种简单、便利的指标, 广泛应用于动物本底清查中。但RAI易受物种自身特征、探测率和环境因素的影响, 需要结合其他物种数量分析方法, 以验证其与种群大小的相关性。随着红外相机技术在野生动物调查中的广泛应用, 用红外相机数据估计动物种群数量的研究越来越多。目前, 基于红外相机数据计算RAI的方法有多种, 不同计算方法和应用范围存在差异, 亟需对现有方法和应用进行梳理。本文综述了根据红外相机数据计算物种相对多度的4种主要方法: (1)拍摄一张有效照片所需要的天数; (2)基于单位调查强度的物种拍摄率; (3)每个位点每天的物种拍摄率; (4)某一物种的照片数占所有物种的比例。总结了我国野生动物监测调查中采用红外相机方法计算RAI的应用现状。国内的研究主要采用第2种和第4种计算方法, 其中约72.5%的研究论文应用第2种计算方法, 而第4种方法一般适用于群落中的物种组成比较。我们建议根据红外相机数据计算RAI时尽量使用第2种计算方法, 这有助于研究或管理人员对不同研究中的物种RAI进行比较分析。     

Animal Scanner: Software for classifying humans,animals, and empty frames in camera trap images

Camera traps are a popular tool to sample animal populations because they are noninvasive, detect a variety of species, and can record many thousands of animal detections per deployment. Cameras are typically set to take bursts of multiple photographs for each detection and are deployed in arrays of dozens or hundreds of sites, often resulting in millions of photographs per study. The task of converting photographs to animal detection records from such large image collections is daunting, and made worse by situations that generate copious empty pictures from false triggers (e.g., camera malfunction or moving vegetation) or pictures of humans. We developed computer vision algorithms to detect and classify moving objects to aid the first step of camera trap image filtering—separating the animal detections from the empty frames and pictures of humans. Our new work couples foreground object segmentation through background subtraction with deep learning classification to provide a fast and accurate scheme for human–animal detection. We provide these programs as both Matlab GUI and command prompt developed with C++. The software reads folders of camera trap images and outputs images annotated with bounding boxes around moving objects and a text file summary of results. This software maintains high accuracy while reducing the execution time by 14 times. It takes about 6 seconds to process a sequence of ten frames (on a 2.6 GHZ CPU computer). For those cameras with excessive empty frames due to camera malfunction or blowing vegetation automatically removes 54% of the false‐triggers sequences without influencing the human/animal sequences. We achieve 99.58% on image‐level empty versus object classification of Serengeti dataset. We offer the first computer vision tool for processing camera trap images providing substantial time savings for processing large image datasets, thus improving our ability to monitor wildlife across large scales with camera traps.     

Determinants of uncertainty in wildlife responses to human disturbance

Outdoor recreation is increasing in intensity and space. Areas previously inaccessible are now being visited by ever‐growing numbers of people, which increases human–wildlife encounters across habitats. This has raised concern among researchers and conservationists as, even in non‐aggressive encounters, animals often perceive humans as predators and mount physiological and behavioural responses that can have negative consequences. However, despite all the research in recent decades, not many general patterns have emerged, especially at the level of populations, and many studies have yielded seemingly contradictory or inconclusive results. We argue that this is partly due to incomplete knowledge of the number and complexity of factors that may modulate the responses of animals. Thus, we aim to provide a conceptual approach intended to highlight the reasons that make it difficult to detect general patterns. We present a comprehensive compilation of factors modulating animal responses to humans at increasing levels (from sensory detection and immediate behavioural and physiological reactions, to changes in fitness and population trends), which may help understanding the uncertainty in the patterns. We observed that there are many modulating factors, which can be categorized as reflecting characteristics of the recreational activity itself (e.g. intensity of human presence), of the animals concerned (e.g. age or antipredatory strategy), and of the spatio‐temporal context (e.g. habitat or timing of the encounter). Some factors appear to have non‐linear and complex effects, which, if not considered, may lead to erroneous conclusions. Finally, we conclude that the difficulty in finding general patterns will be amplified at higher levels (i.e. at the level of populations), since as we proceed from one level to the next, the number of potential modulating factors accumulates, adding noise and obscuring direct associations between recreation and wildlife. More comprehensive knowledge about which (and how) factors affect animal responses across levels will certainly improve future research design and interpretation, and thus, our understanding of human recreational impacts on wildlife.     

A simple framework for a complex problem? Predicting wildlife–vehicle collisions

Collisions of vehicles with wildlife kill and injure animals and are also a risk to vehicle occupants, but preventing these collisions is challenging. Surveys to identify problem areas are expensive and logistically difficult. Computer modeling has identified correlates of collisions, yet these can be difficult for managers to interpret in a way that will help them reduce collision risk. We introduce a novel method to predict collision risk by modeling hazard (presence and movement of vehicles) and exposure (animal presence) across geographic space. To estimate the hazard, we predict relative traffic volume and speed along road segments across southeastern Australia using regression models based on human demographic variables. We model exposure by predicting suitable habitat for our case study species (Eastern Grey Kangaroo Macropus giganteus) based on existing fauna survey records and geographic and climatic variables. Records of reported kangaroo–vehicle collisions are used to investigate how these factors collectively contribute to collision risk. The species occurrence (exposure) model generated plausible predictions across the study area, reducing the null deviance by 30.4%. The vehicle (hazard) models explained 54.7% variance in the traffic volume data and 58.7% in the traffic speed data. Using these as predictors of collision risk explained 23.7% of the deviance in incidence of collisions. Discrimination ability of the model was good when predicting to an independent dataset. The research demonstrates that collision risks can be modeled across geographic space with a conceptual analytical framework using existing sources of data, reducing the need for expensive or time‐consuming field data collection. The framework is novel because it disentangles natural and anthropogenic effects on the likelihood of wildlife–vehicle collisions by representing hazard and exposure with separate, tunable submodels.     

Comparing motion capture cameras versus human observer monitoring of mammal movement through fence gaps: a case study from Kenya

Monitoring the movement and distribution of wildlife is a critical tool of an adaptive management framework for wildlife conservation. We installed motion‐triggered cameras to capture the movement of mammals through two purpose‐built migration gaps in an otherwise fenced conservancy in northern Kenya. We compared the results to data gathered over the same time period (1 Jan 2011–31 Dec 2012) by the human observers monitoring mammal tracks left at the same fence gaps in a sandy loam detection strip. The camera traps detected more crossing events, more species and more individuals of each species per crossing event than did the human track observers. We tested for volume detection differences between methods for the five most common species crossing each gap and found that all detection rates were heavily weighted towards the camera‐trap method. We review some of the discrepancies between the methods and conclude that although the camera traps record more data, the management of that data can be time‐consuming and ill‐suited to some time‐sensitive decision‐making. We also discuss the importance of daily track monitoring for adaptive management conservation and community security.     

相机陷阱在野生动物种群生态学中的应用

种群参数估计及空间分布格局是动物生态学和保护生物学领域的重要目标之一.最近十几年来, 相机陷阱(camera trap)作为野外调查的一种非损伤性技术手段,在传统调查方法难以实现的情况下表现出极大优势,被广泛应用于野生动物生态学和保护学研究中.相机陷阱所获取的动物出现数据为野生动物种群提供了极其重要的定量信息.本文从相机陷阱工作原理出发,主要阐述了目前在种群生态学中较为成熟的两类针对具有或不具有天然个体标志物种的模型原理及应用: 1)种群密度和种群数量估计; 2)空间占据率估计.论文特别关注了模型发展的逻辑过程、依赖的假定、使用范围、仍然存在的问题以及未来发展方向.最后, 本文综合分析了相机陷阱在种群参数估计应用中还需注意的问题, 以及其在种群动态和生物多样性研究等方面的发展潜力.     

Of scales and stationarity in animal movements

With recent technological advances in tracking devices, movements of numerous animal species can be recorded with a high resolution over large spatial and temporal ranges. This opens promising perspectives for understanding how an animal perceives and reacts to the multi‐scale structure of its environment. Yet, conceptual issues such as confusion between movement scales and searching modes prevent us from properly inferring the movement processes at different scales. Here, I propose to build on stationarity (i.e. stability of statistical parameters) to develop a consistent theoretical framework in which animal movements are modelled as a generic composite multi‐scale multi‐mode random walk model. This framework makes it possible to highlight scales that are relevant to the studied animal, the nature of the behavioural processes that operate at each of these different scales, and the way in which the processes involved at any given scale can interact with those operating at smaller or larger scales. This explicitly scale‐focused approach should help properly analyse actual movements by relating, for each scale and each mode, the values of the main model parameters (speed, short‐ and long‐term persistences, degree of stochasticity) to the animal's needs and skills and its response to its environment at multiple scales.     

Do Ranger Stations Deter Poaching Activity in National Parks in Thailand?

As protected areas become more accessible via transportation networks, fragmentation, and encroachment from the borders, carnivores in these areas frequently decline. To counter these pressures, patrolling and active wildlife enforcement are widely accepted as fundamental conservation strategies. Using the case example of Khao Yai National Park (KYNP) and data from a camera trap survey, we modeled and evaluated the effectiveness of ranger stations in reducing human access and illegal activities, and in increasing prey and predator presence. This type of data and analysis is needed to monitor and evaluate enforcement effectiveness and develop adaptive management strategies. At KYNP, we used camera‐trapping data as a proxy to evaluate whether or not a positive impact of ranger stations on wildlife distribution could outweigh edge effects from human disturbance. We assessed factors affecting the distribution of poachers and wildlife using Maxent. Our analysis was based on 217 camera trap locations (6260 trap nights) and suggests that ungulates and poachers persist nearby ranger stations. Rangers should increase patrolling efforts of border areas; however, increasing wildlife patrolling in inaccessible areas with mobile range units may be more effective than establishing more ranger stations along park boundaries.     

粪便类固醇激素检测准确性的影响因素

近年来,采用非损伤性取样方法监测野生动物的生理状况越来越受到重视。本文对检测过程中影响动物粪便类固醇激素检测准确性的因素进行了分析总结,包括样品的新鲜程度、样品量和保存方法;激素代谢的日节律和季节性变化;动物的年龄、性别和繁殖状态等,以期为粪便类固醇激素检测技术在野生动物中的准确应用提供借鉴。     

Potentiality and limitations of N-mixture and Royle-Nichols models to estimate animal abundance based on noninstantaneous point surveys

Reliable and accurate information on animal abundance is fundamental for the conservation and management of wildlife. Recently, a number of innovative devices (such as camera traps) have been widely used in field surveys and have largely improved survey efficiency. However, these devices often constitute noninstantaneous point surveys, resulting in the multiple counts of the same animal individuals within a single sampling occasion (i.e., false-positive errors). Many commonly-used statistical models do not explicitly account for the false-positive error, with its effects on estimates being poorly understood. Here, I tested the performance of the commonly-used Poisson-binomial N-mixture and the Royle-Nichols model in the presence of both false-positive and negative errors (i.e., individuals in a population might not be detected). I also implemented the Poisson-Poisson mixture model in the Bayesian framework to evaluate its reliability. The results of the simulation using random walks based on Ornstein-Uhlenbeck processes showed that the Poisson-binomial model was not robust to false-positive errors. In comparison, the Royle-Nichols and Poisson-Poisson models provided reasonable estimates of the number of animals whose home range included the survey point. However, the number of animals whose home range included the survey point is inherently influenced by the size of animal home ranges, and thus cannot be used as a surrogate of animal density. Although the N-mixture and Royle-Nichols models are widely used, their utility might be restricted by this limitation. In conclusion, studies should clearly define the objective of surveys and carefully consider whether the models used are valid.     

基于地形的牲畜空间利用特征及干扰评价——以王朗国家级自然保护区为例

地形是栖息地的基本要素, 从地形评价动物的空间利用特征能够掌握动物的分布规律并进行预测。为掌握保护区内牲畜的空间利用特征, 并评价它们对主要保护动物的潜在影响, 我们于2018年5-11月调查了王朗国家级自然保护区内牛和马的分布, 并结合红外相机监测结果及历史监测数据进行了分析和评价。结果表明: (1)虽然两种牲畜均偏好低海拔、低坡度、光照良好(半阳坡、阳坡)、距水源近的栖息地, 但它们在地形利用上存在显著差异; (2)牲畜活动最频繁的三条沟分别是竹根岔右一支沟、竹根岔正沟和大窝凼洋洞沟, 且呈现不同的干扰特征; (3)基于监测数据, 羚牛(Budorcas taxicolor tibetana)可能是保护区内最易受牛马活动威胁的保护动物。红外相机监测结果显示, 羚牛沿海拔分布现状可能是回避牲畜密集区域的结果。基于本研究, 我们建议: (1)保护区重点关注竹根岔(右一支沟、正沟、白沙沟)、大窝凼(洋洞沟、外侧坡)两个核心区的牲畜活动情况, 并尽快针对放牧采取措施。例如, 持续监测重点干扰区域牲畜的种群数量和空间分布趋势。(2)严格限制牲畜继续向高海拔栖息地入侵。(3)管控放牧投盐等干扰的发生频率。(4)加强执法力度, 防止牲畜对保护区带来的干扰持续和扩大, 威胁物种安全。     

A novel camera trapping method for individually identifying pumas by facial features

Camera traps (CTs), used in conjunction with capture–mark–recapture analyses (CMR; photo‐CMR), are a valuable tool for estimating abundances of rare and elusive wildlife. However, a critical requirement of photo‐CMR is that individuals are identifiable in CT images (photo‐ID). Thus, photo‐CMR is generally limited to species with conspicuous pelage patterns (e.g., stripes or spots) using lateral‐view images from CTs stationed along travel paths. Pumas (Puma concolor) are an elusive species for which CTs are highly effective at collecting image data, but their suitability to photo‐ID is controversial due to their lack of pelage markings. For a wide range of taxa, facial features are useful for photo‐ID, but this method has generally been limited to images collected with traditional handheld cameras. Here, we evaluate the feasibility of using puma facial features for photo‐ID in a CT framework. We consider two issues: (1) the ability to capture puma facial images using CTs, and (2) whether facial images improve human ability to photo‐ID pumas. We tested a novel CT accessory that used light and sound to attract the attention of pumas, thereby collecting face images for use in photo‐ID. Face captures rates increased at CTs that included the accessory (n = 208, χ ² = 43.23, p ≤ .001). To evaluate if puma faces improve photo‐ID, we measured the inter‐rater agreement of 5 independent assessments of photo‐ID for 16 of our puma face capture events. Agreement was moderate to good (Fleiss’ kappa = 0.54, 95% CI = 0.48–0.60), and was 92.90% greater than a previously published kappa using conventional CT methods. This study is the first time that such a technique has been used for photo‐ID, and we believe a promising demonstration of how photo‐ID may be feasible for an elusive but unmarked species.     

The role of wildlife science in wetland ecosystem restoration: Lessons from the Everglades

There has been little discussion of how and when to integrate wildlife science into ecological restoration projects. The recent emergence of wetland ecosystem restoration offers an opportunity to use wildlife science to increase the probability of a project being successful. This paper traces the evolution of wetland ecosystem restoration in North America and proposes three roles for wildlife science in wetland ecosystem restoration: (1) contribute to conceptual ecosystem models, (2) develop quantitative performance measures and restoration targets that track the progress of restoration, and (3) achieve social feasibility by sustaining long-term public support for a project. The extensive knowledge base for many species of wildlife makes them especially useful for contributing to conceptual ecosystem models. Wildlife species are often the subject of long-term monitoring and research because they have commercial value, are conspicuous, or have aesthetic appeal. Wildlife parameters can be good performance measures for large-scale restoration projects because some species integrate information over large spatial scales and are long-lived. Parameters associated with threatened or endangered wildlife species should get special consideration as performance measures because the information will meet multiple needs rather than just those of the conceptual ecosystem model. Finally, wetland ecosystem restoration projects need to sustain funding over decades to ensure the restored system is self-sustaining. Wildlife are a valued resource that can help achieve the social feasibility of a project by providing a way to communicate complex science in terms that society understands and values.     

Global spread of helminth parasites at the human–domestic animal–wildlife interface

Changes in species distributions open novel parasite transmission routes at the human–wildlife interface, yet the strength of biotic and biogeographical factors that prevent or facilitate parasite host shifting are not well understood. We investigated global patterns of helminth parasite (Nematoda, Cestoda, Trematoda) sharing between mammalian wildlife species and domestic mammal hosts (including humans) using >24,000 unique country‐level records of host–parasite associations. We used hierarchical modelling and species trait data to determine possible drivers of the level of parasite sharing between wildlife species and either humans or domestic animal hosts. We found the diet of wildlife species to be a strong predictor of levels of helminth parasite sharing with humans and domestic animals, followed by a moderate effect of zoogeographical region and minor effects of species’ habitat and climatic niches. Combining model predictions with the distribution and ecological profile data of wildlife species, we projected global risk maps that uncovered strikingly similar patterns of wildlife parasite sharing across geographical areas for the different domestic host species (including humans). These similarities are largely explained by the fact that widespread parasites are commonly recorded infecting several domestic species. If the dietary profile and position in the trophic chain of a wildlife species largely drives its level of helminth parasite sharing with humans/domestic animals, future range shifts of host species that result in novel trophic interactions may likely increase parasite host shifting and have important ramifications for human and animal health.