|Camera-Trap Assessment of Small Mammal Occupancy in a Coastal Virginia Forest|
|Ray Dueser, John Porter|
|Demonstration of Image Recognition Wildlife Feeders|
|Rebecca McPeake, Rachel Lipsey|
This demonstration assessed the effectiveness of six image recognition wildlife feeders for allowing access to target species while excluding non-target species. WiseEye™ game feeders were set at six locations at The Alotian Golf Club near Roland, Arkansas from October 2018 – April 2019. Feeders baited with whole corn were programmed to close when species other than white-tailed deer (Odocoileus virginianus) approached the feeder. A solar panel with a backup marine battery powered the systems. These were set along field edges and dirt roads in a hilly mixed pine and deciduous forest. A moisture detector closed the feeder door to keep bait dry. As a secondary demonstration, three sites were encircled with three Havahart Electronic Deer Repellent™ stakes. Photo images were collected from WiseEye™ cameras and motion-detecting trail cameras. Deer consumed corn from every feeder. However, feeders were unreliable and inconsistent in feed availability. One feeder ceased functioning in February and a new computer box installed. For most systems, battery depletion occurred daily. Doors would automatically close when system failure occurred. Deer would approach the system but the door would not open. Inconsistent image recognition of raccoons (Procyon lotor) and other species resulted in a lag in door closure, or no closure at all. In one system, raccoons shoveled corn onto the ground after prying open the closed door. Feeders were programmed to shock non-target species but proved ineffective for those raccoons. Results indicate these systems which use artificial intelligence for image recognition are not 100% effective at consistently dispensing feed to target species and excluding non-target species. Therefore, ineffective image detection for white-tailed deer and other species is cause of concern for other uses, such as limiting deer access to baits intended for feral hog removal in areas with chronic wasting disease or dispensing poisoned baits for feral hog control.
|Implementation and Maintenance of Remote Live-Streaming Camera Monitoring Systems for Wildlife Management and Other Applications|
|Nadya Seal Faith|
Remote streaming camera monitoring systems (RSCMS) are networks of cameras that can transmit live-streaming feeds wirelessly through radio antennas in remote locations. RSCMS are useful for gathering data for research, monitoring health and behavior of various wildlife, and are a source for public outreach and education, with minimal human presence. The Hopper Mountain National Wildlife Refuge Complex (HMNWRC) uses a RSCMS to monitor California condor nests for all of these reasons and recognizes that these systems may have a wide variety of applications for other organizations. Several of these wireless connections are in difficult to access terrain with connections that span between hundreds of feet to over twenty miles in montane chaparral with thousands of feet of elevation gain and loss. Here we present a poster that will highlight our manual detailing the creation, implementation, and maintenance of such a RSCMS, using the HMNWRC California Condor Nest Monitoring program as an example. The manual outlines broad applications of these camera systems, discusses advantages and limitations, outlines step-by-step instructions for setting up a camera monitoring system with sections on software, hardware, wireless communications, and troubleshooting, as well as a cost estimate breakdown for those who wish to implement a similar RSCMS.
|Camera Deployment for Optimal Detection of Critical Life Events in Fisher|
|Stephanie Cunningham, Timothy Pyszczynski, Timothy Watson, Rachel Bakerian, Paul Jensen, Jacqui Frair|
Trail cameras are often used to monitor detection/non-detection of wildlife and, increasingly, to record activity and critical life events such as recruitment. For example, fisher (Pekania pennanti) use tree cavities for reproduction, and trail cameras may be used to document females moving kits between den trees. During spring 2020, we set 2-7 trail cameras at dens of GPS-collared fishers in northern New York and compared photos of the fishers leaving or returning to the den to activity periods inferred from GPS and accelerometer data. Our camera arrays missed several departure and return events, as well as entire active periods. Generally, the probability of detection should increase with the number of cameras deployed at a survey site, but how many cameras and what arrangement will provide optimal coverage? To answer this question, we deployed three baited camera arrays, with each having 15 cameras set approximately 5 meters away from a center tree and facing inwards. Any photos captured within 1-minute grouping defined a detection event. We thinned detection events to only include the first and last event when consecutive events were less than 25 minutes apart. We virtually rarefied cameras and recorded whether or not a fisher was detected by the remaining cameras, restricting our rarefied combinations to a maximum of 6 equally spaced cameras. We predicted detection rate as a function of the number of cameras using a binomial generalized linear mixed model. With an average of 4-5 cameras deployed on fisher reproductive studies to date, detection rates may be as low as 0.60-0.73. A detection rate of ≥0.90 required more than 6 cameras, indicating a more intensive design than in typically considered for capturing ephemeral life history events. Greater understanding of detection probabilities will inform more effective use of camera traps for estimating demographic rates of cryptic species.
|Estimating Habitat-Specific Densities of Spotted Hyena Using Camera Traps and Acoustic Detectors in Northern Botswana – SRIP|
|Worldwide, large carnivore populations have declined over the past century, mainly because of increased conflict with humans. According to the IUCN Red list, the current conservation status of spotted hyena (Crocuta Crocuta) is “least concern” however, in many African countries spotted hyenas are under threat when residing outside protected areas. This is due to direct persecution by humans, given its perceived responsibility for most livestock attacks. The culturally developed prejudices projected on to the spotted hyena make its conservation difficult. Conventional census techniques are inappropriate for estimating population densities of nocturnal species such as the spotted hyena. In this study, we will compare the performance of two non-invasive methods, camera trapping and passive acoustic monitoring to estimate habitat-specific densities of spotted hyenas in Northern Botswana. Camera trapping is commonly used in ecological studies because it is cost effective, non-invasive, can be deployed for long periods and facilitates collection of data without the researcher being in the field most of the time. Camera trapping is continuously gaining popularity in mammal research due to statistical advancements that allow researchers to calculate population densities from camera trap data. However, like many other techniques, camera trapping has its own shortcomings, such as the sampling being restricted to a small window in front of each camera trap. Incomparison, passive acoustic detectors record animal sounds and can cover a much wider detection range. Passive acoustic detectors, however, rely on animals producing sounds that are loud enough to be detected. This study will determine whether camera traps and passive acoustic monitoring devices provide comparable estimates of spotted hyena densities.|
|Effects of Forest Management on the Distribution of Weasels in Maine and Consequences of Analyzing Similar-Looking Species Data Collected By Camera Traps – SRIP|
|Bryn Evans, Alessio Mortelliti|
|Three species of weasel occur in the state of Maine (northeastern United States). Primarily short-tailed weasel (Mustela erminea), most common in the north, and long-tailed weasel (Mustela frenata), most common in the south, and potentially sporadic occurrences of least weasel (Mustela nivalis) along the northwestern border. Although these generalist predators seem to be resilient to habitat modification, they have not received much study, and to our knowledge there has never been an investigation of weasel responses to varying degrees of timber harvest in working forests. Our objective is to help fill this knowledge gap with a rigorous camera trapping study, designed to cover a gradient of forest disturbance intensity and configuration (i.e. large continuous blocks of harvest, smaller areas of harvest surrounded by less modified forests, etc). We collected data from 197 locations over both summer and winter survey seasons from 2017 to 2020, and obtained over 70,000 images of weasels. Preliminary occupancy modeling reveals that, when pooled together, weasel species do not appear negatively impacted by commercial timber harvest activities. At a local scale, forest disturbance seems positively correlated with initial occupancy probability, and at a broader scale disturbance encourages colonization of areas not already occupied by weasels. While in line with the literature to date, these preliminary findings may obfuscate differences in habitat preference and vulnerability between weasel species. The next phase of our analyses will be differentiating images between species. We anticipate short-tailed and long-tailed weasels will provide sufficient detections for their own data sets, and we will assign confidence levels to species identities to test the role that mistaken identities could play in interpreting models. We hope the results of false-positive models, and comparing summer and winter detection probabilities, will provide valuable guidance to managers and researchers interested in better understanding weasels in working forests.|
|Factors Affecting Cost-Efficiency of Game Cameras for Monitoring Wildlife Populations – SRIP|
|Savannah Fournier, Cady Sartini, Lauren Welvaert, Gretchen Puls, Wesley Dixon, Guha Dharmarajan, Amy Davis|
|Camera traps are often utilized for monitoring wildlife populations and are assumed to offer benefits as a cost-effective and noninvasive alternative to traditional trapping methods. However, few attempts have been made to date to quantify the full extent of labor required to interpret camera trap information. Factors that could affect the efficiency of using cameras include specific camera settings, including the number of photos taken in a trigger, the sensitivity of the camera, and placement of the camera, all of which impact the number of photos which must be processed by trained personnel. Other factors that could affect efficiency include how much training personnel has received or specific methods used for photo processing. This study sought to understand factors affecting the efficiency of using cameras to monitor wildlife. Reconyx game cameras were used to monitor raccoon populations for the National Rabies Monitoring Program from January-April of 2019 and 2020. We will compare the number of photos generated from different camera sites and settings along with how long it takes individual observers to maximize their efficiency to determine specific recommendations for future projects to maximize cost-benefits.|
|Testing New Technology for Wildlife-Livestock Conflict Mitigation: Evaluating AI-Enabled Camera Traps – SRIP|
|Taylor Bayne, Jared Beaver, Jeffrey Mosley, Lance McNew|
|Private working lands make up 65% of Montana’s 93 million acres. They provide invaluable ecosystem services, help address climate change, and are essential in facilitating the connected, open landscapes necessary for wildlife survival and successful conservation efforts. However, expanding wildlife populations and increasing landscape fragmentation concentrate livestock and wildlife in a smaller area, making conflict inevitable. A semi-paradoxical situation has arisen, in which the burden of losses caused by wildlife-livestock conflicts places owners of working lands in opposition to wildlife and wildland protection. Therefore, there is a need to continue developing effective and sustainable conflict prevention tools. As part of the solution, we will evaluate “smart” game cameras equipped with edge artificial intelligence (edge AI) and internet of things (IoT), that reduces the number of false positive images and enables real-time identification and notification of wildlife captured in camera images. The reliability and effectiveness of the smart cameras will be compared to existing game cameras used for wildlife conservation in both field and laboratory settings. Initial camera deployment to establish study protocols will begin summer of 2021, with full deployment and data collection in spring of 2022. Field evaluations will be conducted in the Paradise Valley of Montana, an area that is experiencing an increased prevalence of wildlife-livestock conflict and that is at high risk of land turnover and fragmentation due to the aesthetic quality of the landscape. Elk (Cervus elaphus) and grizzly bears (Ursus arctos) are the focal wildlife species. Development of an effective means of non-invasive, real-time wildlife monitoring could be used as a conflict prevention tool in and of itself as well as a means of evaluating existing and future conflict prevention methods.|