Skip to main content
Dryad

Data from: Identification of reindeer fine-scale foraging behaviour using tri-axial accelerometer data

Cite this dataset

Rautiainen, Heidi (2024). Data from: Identification of reindeer fine-scale foraging behaviour using tri-axial accelerometer data [Dataset]. Dryad. https://doi.org/10.5061/dryad.8sf7m0cs7

Abstract

Animal behavioural responses to the environment ultimately affect their survival. Monitoring animal fine-scale behaviour may improve understanding of animal functional response to the environment and provide an important indicator of the welfare of both wild and domesticated species. In this study, we illustrate the application of collar-attached acceleration sensors for investigating reindeer fine-scale behaviour. Using data from 19 reindeer, we tested the supervised machine learning algorithms random forests, support vector machines, and hidden Markov models to classify reindeer behaviour into seven classes: grazing, browsing low from shrubs or browsing high from trees, inactivity, walking, trotting, and other behaviours. We implemented leave-one-subject-out cross-validation to assess generalizable results on new individuals. Our main results illustrated that hidden Markov models were able to classify collar-attached accelerometer data into all our pre-defined behaviours of reindeer with reasonable accuracy while random forests and support vector machines were biased towards dominant classes. Random forests using 5-second windows had the highest overall accuracy (85%), while hidden Markov models were able to best predict individual behaviours and handle rare behaviours such as trotting and browsing high. We conclude that hidden Markov models provide a useful tool to remotely monitor reindeer and potentially other large herbivore species behaviour. These methods will allow us to quantify fine-scale behavioural processes in relation to environmental events.

README: Identification of reindeer fine-scale foraging behaviour using tri-axial accelerometer data

 author: "Heidi Rautiainen, with acknowledgements to Måns Karlsson for hidden Markov model"

https://doi.org/10.5061/dryad.8sf7m0cs7

Manuscript: https://doi.org/10.1186/s40462-022-00339-0

Description of data handling, accelerometer data, and behavioural data from video recordings.

Description of files provided

All in .csv format (","-separated columns)

  • Filename: "acceleration.csv" (for step 1)
  • Filename: "annotations.csv" (for step 1)
  • Filename: "df4_2s_allIDs.csv" (for step 2)
Filename: "acceleration.csv"

Description: Accelerometer data in three dimensions (x,y,z), animal id, and timestamp using axy-4 loggers from technosmart. Used for step 1 "Data handling_oneID".

*Column names: *

  • "TagID": Accelerometer ID
  • "Timestamp": H:M:S:OS
  • "X"
  • "Y"
  • "Z"
Filename: "annotations.csv"

Description: Behavioural data from video recordings (labelled and exported csv from BORIS software). Used for step 1 "Data handling_oneID".

N/A's: default columns from BORIS output, where no notes has been made (e.g. "Description", Behavorial.category").

*Column names: *

  • "Observation.id": Chosen observation ID
  • "Observation.date": Date of observations
  • "Description": N/A
  • "Subject": Animal ID (named by accelerometer ID)
  • "Behavior": Observered behaviour
  • "Behavioral.category": Behavioural category
  • "Modifiers": Observered behaviour
  • "Behavior.type": STATE
  • "Start..s.": Starttime of observered behaviour in seconds
  • "Stop..s.": Stoptime of observered behaviour in seconds
  • "Duration..s.": Duration of observered behaviour in seconds
Filename: "df4_2s_allIDs.csv"

Description: Example dataset of processed data and variables ued (following steps in "Data handling_oneID") from 19 individuals. Used for step 2 "Training-and-validation-HMM".

**Description for column names **given for acceleration (X, Y and Z), static acceleration (stX, stY, stZ) and dynamic acceleration (dyX, dyY, dyZ):

  • "samp": Window number (segment)
  • "mrot_": Median (of rotated values) for each axis in each window
  • "drot_": Third quantile (Q3) subtracted by the first quantile (Q1) for each axis
  • "meanrot_": Mean value for each axis in each window
  • "minrot_": Minimum value for each axis in each window
  • "maxrot_": Maximum value for each axis in each window
  • "sdrot_": Standard deviation for each axis in each window
  • "mroll"/"mpitch": Median for roll and pitch in each window
  • "droll"/"dpitch": Third quantile (Q3) subtracted by the first quantile (Q1) for roll and pitch in each window
  • "ID": Animal ID = Accelerometer ID
  • "Timestamp": Y-M-D H:M.S
  • "Modifiers": Behaviour

Description of data handling

Step 1: Data handling_oneID

Files: "annotations.csv" and "acceleration.csv"

Example data given for one individual for data processing.

a. Examples of initial data handling
  • Basic data handling
b. Merging of acceleration and behaviour
  • Fix date and time format
  • Add behaviours (or modifiers) to acceleration file
  • Data merging ends (acceleration for each start-stop of behaviour)
c. Filtering, computation of variables and rotation matrices
  • Filtering using running mean
  • Raw acceleration in bits (X, Y, Z)
  • Static acceleration (stX, stY, stZ)
  • Dynamic acceleration (dX, dY, dZ)
  • Pitch (rotation around Y axis)
  • Roll (rotation around X axis)
  • Rotation matrix around X to account for rotation around the neck
d. Segmentation of data to two second windows with the most common behavior
  • 2, 3 or 5 second windows with most common behavior
  • Example only for using 2 sec window

Step 2: Training-and-validation-HMM

File: "df4_2s_allIDs.csv"

Example code given for the final model (hidden Markov model) for 19 individuals. Random forests, Support vector machines or forward feature selection not included here (see packages used in manuscript). Example data is given for a subset of 200 observations for each individual for demonstration to enable fast processing. 

Code and software

Behaviours labelled from video recordings using BORIS Version 7.9.8. Data handling, model training and validation performed in R (R version 4.0.3 and RStudio version 1.3.1093). Note: Although not part of this manuscript, predictions on new individuals done using Anaconda https://www.anaconda.com/products/distribution to enable fast processing of data. 

Methods

We simultaneously collected video and acceleration data from in total 19 semi-domesticated female reindeer. Three cameras (Axis Communications, 2025-LE Network Camera) were used and placed to cover the whole enclosure and enable video recordings of the animals from different angles. Video recordings were annotated using BORIS Version 7.9.8. All data processing and analyses were performed using R version 4.0.3 and RStudio version 1.3.1093.

 

 

 

Funding

Swedish Research Council for Environment Agricultural Sciences and Spatial Planning