High-frequency accelerometer monitoring of foraging and movement behavior in a secretive predator (Central Rat Snakes, Pantherophis alleghaniensis)
Faculty Mentor(s) Name(s)
Dr. Dominic DeSantis
Abstract
Advancements in bio-logging technology have transformed the study of animals in nature. Accelerometer (ACTs) dataloggers remotely and continuously log three-dimensional acceleration (upward, downward, and side-to-side) derived from subject-motion. When paired with machine learning techniques, automated classification of distinct behavioral states from these high-resolution data can quantify real-time activity budgets in wild-ranging subjects. The goal of this project is to expand upon a recently validated framework for ACT monitoring of snake behavior. Using Rat Snakes (Pantherophis alleghaniensis) as a model for large-bodied constrictors, we conducted a series of captive behavioral trials with ACT-equipped snakes that stimulated a series of key behaviors: full-body locomotion, immobility, predatory strikes, constriction, and prey swallowing. We then produced an extensive validation dataset for supervised model development containing 38 observations of each behavior class. Model training and testing procedures were conducted using an open-source web application, demonstrating the increasing accessibility of animal-borne ACT studies. The top performing model indicated a combined class accuracy of 82.69%, with class subset accuracies of 95%. Model development is a fundamental step toward field-recording of predatory behaviors in snakes. Translation to wild-ranging individuals require overcoming additional hurdles, including refinement of a method for long-term ACT attachment at 25% of snake snout-to-vent lengths, and assessing the generalizability of our model to field data. We envision this technique transforming field studies of snake behavioral ecology, as unlocking real-time monitoring of foraging efficiency facilitates improved interpretation of the causes and consequences of variation in individual behavior, and its effects on population trajectories.
Start Date
27-3-2024 9:00 AM
End Date
27-3-2024 9:50 AM
Location
Magnolia Ballroom
High-frequency accelerometer monitoring of foraging and movement behavior in a secretive predator (Central Rat Snakes, Pantherophis alleghaniensis)
Magnolia Ballroom
Advancements in bio-logging technology have transformed the study of animals in nature. Accelerometer (ACTs) dataloggers remotely and continuously log three-dimensional acceleration (upward, downward, and side-to-side) derived from subject-motion. When paired with machine learning techniques, automated classification of distinct behavioral states from these high-resolution data can quantify real-time activity budgets in wild-ranging subjects. The goal of this project is to expand upon a recently validated framework for ACT monitoring of snake behavior. Using Rat Snakes (Pantherophis alleghaniensis) as a model for large-bodied constrictors, we conducted a series of captive behavioral trials with ACT-equipped snakes that stimulated a series of key behaviors: full-body locomotion, immobility, predatory strikes, constriction, and prey swallowing. We then produced an extensive validation dataset for supervised model development containing 38 observations of each behavior class. Model training and testing procedures were conducted using an open-source web application, demonstrating the increasing accessibility of animal-borne ACT studies. The top performing model indicated a combined class accuracy of 82.69%, with class subset accuracies of 95%. Model development is a fundamental step toward field-recording of predatory behaviors in snakes. Translation to wild-ranging individuals require overcoming additional hurdles, including refinement of a method for long-term ACT attachment at 25% of snake snout-to-vent lengths, and assessing the generalizability of our model to field data. We envision this technique transforming field studies of snake behavioral ecology, as unlocking real-time monitoring of foraging efficiency facilitates improved interpretation of the causes and consequences of variation in individual behavior, and its effects on population trajectories.