Geology topics

Eric Odell

Publications and source records attributed to Eric Odell.

3 recordsLinked to original sources

USGS2026

Monitoring wolverine (Gulo gulo) distribution across the western United States for regional and national assessment

Population monitoring of at-risk species is fundamental to directing effective and timely conservation actions. For wide-ranging threatened species, such as the wolverine ( Gulo gulo ), monitoring across multiple jurisdictions is necessary to track their distribution to inform collaborative regional and national management and policy. We repeated a landscape-scale, multi-jurisdictional remote camera survey first completed in 2017 to evaluate wolverine occupancy throughout the western United States in 2022. Our objectives were to 1) expand the sampling frame to include areas of potential recolonization (Colorado, Oregon, Utah, USA), 2) evaluate whether overall and state-specific occurrence has changed from 2017 in the common sampling frame of Idaho, Montana, Washington, and Wyoming, USA, and 3) evaluate spatial drivers of wolverine occurrence in 2022. From a sampling frame of 770 sites (15 × 15-km cells) identified as potential wolverine habitat, 229 sites were surveyed in 2022; no wolverines were detected in Colorado, Oregon, or Utah. We predicted wolverine occupancy and 95% highest posterior density intervals (HPDI) in states with detections to be 0.47 (0.37–0.58) in Idaho, 0.35 (0.21–0.52) in Montana, 0.29 (0.09–0.54) in Washington, and 0.35 (0.27–0.47) in Wyoming. We found weak to moderate evidence of a small decline in overall occupancy ( 𝜓 ) from the common sampling frame (2017: 0.44 [0.37–0.52]; 2022: 0.39 [0.30–0.49]; P( 𝜓 2022 <𝜓 2017 ) = 0.81), driven by lower occurrences in Montana and Washington. However, this finding is complicated by a change in detection methodology that may have contributed to lower occurrences in Montana and Washington. Lastly, we found strong support (probability of an effect >0.95) that within areas of predicted wolverine habitat, increasing habitat connectivity and the proportion of a site that was previously burned (2001–2015) were associated with increased wolverine occupancy, while increasing normalized difference vegetation index (NDVI) was associated with decreased wolverine occupancy. Our expanded sampling frame provides the means to track wolverine occupancy and range extent—including recolonizations and contractions—across the western United States, providing insights at both the state and national scales.

Colorado, Idaho, Montana, Oregon, Utah, Washington↗

USGS2020

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.

Ecology and Evolution↗

USGS2017

Crossing boundaries in a collaborative modeling workspace

There is substantial literature on the importance of bridging across disciplinary and science–management boundaries. One of the ways commonly suggested to cross boundaries is for participants from both sides of the boundary to jointly produce information (i.e., knowledge co-production). But simply providing tools or bringing people together in the same room is not sufficient. Here we present a case study documenting the mechanisms by which managers and scientists collaborated to incorporate climate change projections into Colorado’s State Wildlife Action Plan. A critical component of the project was the use of a collaborative modeling and visualization workspace: the U.S. Geological Survey’s Resource for Advanced Modeling (RAM). Using video analysis and pre/post surveys from this case study, we examine how the RAM facilitated cognitive and social processes that co-produced a more salient and credible end product. This case provides practical suggestions to scientists and practitioners who want to implement actionable science.

Society and Natural Resources↗