Search for: All records

### Access Files can be accessed and downloaded from the directory via: [https://arcticdata.io/data/10.18739/A2CJ87N8D/](https://arcticdata.io/data/10.18739/A2CJ87N8D/). ### Overview This dataset provides camera trap images from 40 sites on the northern coastal plain of Alaska. Cameras collected motion sensor and time-lapse (5 minute intervals) images from May-September 2019, and 2021-2023.This dataset was used to monitor summer distribution and behavior of wildlife, including land mammals and birds. The images also provide information on seasonal change (snow melt), weather, and plant phenology. The dataset helped to design effective camera-trap protocols for monitoring and comparing wildlife presence in developed (oil fields) and undeveloped (Arctic National Wildlife Refuge) areas. These images create baseline data for future research to assess ecosystem change as human activity increases and temperatures warm in northern Alaska. 

Qualitative studies have suggested that forest changes following a wildfire can challenge a hunter's ability to harvest big game, such as moose (Alces alces). Quantitative effects have not been estimated. Given the increasing prevalence of wildfires, the strong linkages between wildfire and moose habitat, and the importance of moose to the people of the boreal region of North America, our goal was to assess if and how moose harvest patterns changed immediately following a wildfire. To address that goal, we used 36 years (1984-2019) of spatially-explicit wildfire and moose harvest data in Alaska to compare moose harvest variables the year before and year after a wildfire occurred. With a few exceptions, the number of hunters, kills, and success rates were similar (p > 0.05, Effect size < 0.3) between pre- and post-wildfire years. We estimated a weak to moderate effect on change in moose hunter numbers, kills, and success rate in only a small percentage (1.5%) of wildfires that burned a very large proportion (>38%) of a moose harvest reporting unit. Our findings suggest that wildfire has not caused a clear and functional quantitative effect on hunters’ ability to harvest moose in Alaska. 

Abstract Snow conditions are changing rapidly across our planet, which has important implications for wildlife managers. In Alaska, USA, the later arrival of snow is challenging wildlife managers' ability to conduct aerial fall (autumn) moose (Alces alces) surveys. Complete snow cover is required to reliably detect and count moose using visual observation from an aircraft. With inadequate snow to help generate high‐quality moose survey data, it is difficult for managers to determine if they are effectively meeting population goals and optimizing hunting opportunities. We quantified past relationships and projected future trends between snow conditions and moose survey success across 7 different moose management areas in Alaska using 32 years (1987–2019) of moose survey data and modeled snow data. We found that modeled mean snow depth was 15 cm (SD = 11) when moose surveys were initiated, and snow depths were greater in years when surveys were completed compared to years when surveys were canceled. Further, we found that mean snow depth toward the beginning of the survey season (1 November) was the best predictor of whether a survey was completed in any given year. Based on modeled conditions, the trend in mean snow depth on 1 November declined from 1980 to 2020 in 5 out of 7 survey areas. These findings, coupled with future projections, indicated that by 2055, the delayed onset of adequate snow accumulation in the fall will prevent the completion of moose surveys over roughly 60% of Alaska's managed moose areas at this time of the year. Our findings can be used by wildlife managers to guide decisions related to the future reliability of aerial fall moose surveys and help to identify timelines for development of alternate measurement and monitoring methods. 

Abstract Camera traps (CT) have been used to study a wide diversity of wildlife around the world. However, despite their widespread use, standardized protocols are lacking, potentially leading to reduced efficiency and inhibiting study comparisons, generalizability, and repeatability. While there are general guidelines and considerations researchers should be aware of when designing a CT survey, studies have shown the vital importance of selecting sampling schemes and camera settings tailored to specific characteristics of the wildlife system of interest. For many species and regions, optimal sampling protocols have not been thoroughly evaluated, especially in vast open landscapes. We used CT data on barren‐ground caribou (Rangifer tarandus) in the open landscape of arctic Alaska as a case study to evaluate and quantify the influence of camera trigger type (i.e., motion detection vs. time‐lapse) and time‐lapse interval on data generation to inform sampling protocols for future CT research in this system or others like it. Comparing camera trigger types, we found 5 min interval time‐lapse generated seven‐times more images containing caribou compared to motion detection. However, the detection rate of motion detection was over 11‐times greater than time‐lapse resulting in more efficient data collection with respect to camera battery life, data storage, and data processing time. Exploring the effect of time‐lapse interval length, we found detections were highly sensitive to interval length with a 30 min interval producing 33.7% fewer images containing caribou and identifying 22.2% fewer trap days containing caribou compared to a 5 min interval. Our results provide insight into effective CT sampling protocols for open landscapes and highlight the importance of critically evaluating and selecting camera settings that account for characteristics of the study system to ensure adequate data is generated efficiently to address study objectives. 

s 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.