Ryan K. Brook | Search USGS

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↗

USGS2019

Predicting functional responses in agro-ecosystems from animal movement data to improve management of invasive pests

Functional responses describe how changing resource availability affects consumer resource use, thus providing a mechanistic approach to prediction of the invasibility and potential damage of invasive alien species (IAS). However, functional responses can be context dependent, varying with resource characteristics and availability, consumer attributes, and environmental variables. Identifying context dependencies can allow invasion and damage risk to be predicted across different ecoregions. Understanding how ecological factors shape the functional response in agro‐ecosystems can improve predictions of hotspots of highest impact and inform strategies to mitigate damage across locations with varying crop types and availability. We linked heterogeneous movement data across different agro‐ecosystems to predict ecologically driven variability in the functional responses. We applied our approach to wild pigs ( Sus scrofa ), one of the most successful and detrimental IAS worldwide where agricultural resource depredation is an important driver of spread and establishment. We used continental‐scale movement data within agro‐ecosystems to quantify the functional response of agricultural resources relative to availability of crops and natural forage. We hypothesized that wild pigs would selectively use crops more often when natural forage resources were low. We also examined how individual attributes such as sex, crop type, and resource stimulus such as distance to crops altered the magnitude of the functional response. There was a strong agricultural functional response where crop use was an accelerating function of crop availability at low density (Type III) and was highly context dependent. As hypothesized, there was a reduced response of crop use with increasing crop availability when non‐agricultural resources were more available, emphasizing that crop damage levels are likely to be highly heterogeneous depending on surrounding natural resources and temporal availability of crops. We found significant effects of crop type and sex, with males spending 20% more time and visiting crops 58% more often than females, and both sexes showing different functional responses depending on crop type. Our application demonstrates how commonly collected animal movement data can be used to understand context dependencies in resource use to improve our understanding of pest foraging behavior, with implications for prioritizing spatiotemporal hotspots of potential economic loss in agro‐ecosystems.

California. Florida, Georgia, Louisiana, Missouri,↗