PosterPDF Available

Individual information access resulting from fine-scale movement decisions

Authors:
  • Max Planck Institute of Animal Behavior
  • Max Planck Institute of Animal Behavior
  • Max Planck Institute of Animal Behavior

Abstract

The patterns captured in animal movement data represent a record of an individual's movement decisions in response to information gleaned from the environment and social partners. Understanding the complex social and environmental factors influencing individual animals is therefore important for interpreting observed movement patterns; however, this contextual information is rarely captured by common bio-logging methodologies. We have developed novel methods for generating high-resolution datasets on the movement and behavior of entire herds of African ungulates and have paired these with high-quality 3D habitat models. Using these data, we will quantify behavioral patterns and microhabitat selection and examine the ways in which an individual's fine-scale decisions affect its access to social and environmental information. How do individuals within a group vary in their access to social and environmental information as a result of their fine scale movement decisions? How do animals adjust their behavioral patterns (particularly vigilance) in response to information availability? How do observed individual movement strategies compare to random strategies or strategies designed to optimize information access?
Individual information access
resulting from fine-scale
movement decisions
The patterns captured in animal movement data represent a record of an
individual’s movement decisions in response to information gleaned from the
environment and social partners. Understanding the complex social and
environmental factors influencing individual animals is therefore important for
interpreting observed movement patterns; however, this contextual
information is rarely captured by common bio-logging methodologies. We
have developed novel methods for generating high-resolution datasets on
the movement and behavior of entire herds of African ungulates and have
paired these with high-quality 3D habitat models. Using these data, we will
quantify behavioral patterns and microhabitat selection and examine the
ways in which an individual’s fine-scale decisions affect its access to social
and environmental information.
How do individuals within a group vary in their
access to social and environmental information as
a result of their fine scale movement decisions?
How do animals adjust their behavioral patterns
(particularly vigilance) in response to information
availability?
How do observed individual movement strategies
compare to random strategies or strategies de-
signed to optimize information access?
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant
agreement No. 748549 and the DFG Centre of Excellence 2117 “Centre for the Advanced Study of Collective Behaviour” (ID: 422037984).
Poster design by Mike Costelloe.
Blair R. Costelloe, Benjamin Koger, Jacob M. Graving, Iain D. Couzin
Max Planck Institute for Ornithology; Centre for the Advanced Study of Collective Behaviour, University of Konstanz
Introduction
To further analyze behavior at the individual level, we use a Bayesian
machine learning model to compress our high-dimensional posture
time-series into a two-dimensional “map” of individual behavior. We then use
this model to estimate the probability distribution of each animal’s full
behavioral repertoire (above left). Density peaks in the distribution (graph
nodes) indicate attractors in the behavioral space, or preferred “modes” of
behavior, which we then cluster into discrete, low-level categories—some-
times called an “unsupervised ethogram”. We then model the dynamics of
individual behavior at longer timescales by analyzing the Markov transition
probabilities (graph edges) between behavioral clusters. The transition
graph provides a snapshot of the temporal dynamics within the behavioral
space and is complex, with weighted and directed (indicated by the
right-hand curvature) connections. The graph exhibits compact, dense-
ly-connected clusters of nodes, indicating higher-order temporal structure,
which we model with graph-theoretic clustering (above right). The structure
of these higher-order clusters indicate that behavior may be organized
hierarchically by the brain over multiple timescales.
Video Data
Collection
We film Grevy’s zebra, plains zebra,
buffalo and impala herds using DJI
Phantom 4 Pro drones. The drones fly
for up to 30 minutes on a single
battery, but we use two drones in relay
to achieve longer observation times.
After collecting baseline footage, two
team members approach the group on
foot, eliciting a detection and escape
response. The resolution of our video
footage is 4K, 60 frames per second.
Visual Field Reconstruction
We use our posture estimates to calculate head and eye locations for each
individual in 3D space. We use a ray-casting algorithm to reconstruct visual
fields for each individual. By examining where these rays intersect with
vegetation structures or conspecifics, we can determine what environmental
and social information each individual has visual access to at any given time.
We can thus explicitly account for this information and examine its effect on
behavioral patterns and decision-making.
Posture Estimation
Behavioral Maps
We use a deep convolutional neural network to estimate the posture of
each animal in each frame. This model is optimized to take in images of
focal individuals and estimate the location of nine keypoints corre-
sponding to different body parts (e.g. head, nose, tail, etc.), as well as
confidence levels for each estimate.
See videos and learn more about the
research at our project website,
herdhover.com
high-frequency
head movements
tail movements
head and tail
movements
low-frequency
head movements
medium frequency
head movements
medium frequency
head movements
10m
1m
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.