Bridging the usability gap in open source animal movement algorithms
Brock, Alison M Ashbury, Rainer Hirk, Florian Schwendinger, Margaret C Crofoot
whoami
I help scientists with their data analysis at Max Planck Institute of Animal Behavior.
travelpaths package provides a consistent interface that simplifies the process of running multiple methods on an animal movement dataset.
Target End User
Researchers who collect movement data using both traditional and automated methods.
Typically R users with a passion for science not for coding. Much of the analysis is carried out by early career researchers.
Scientific Scope
Animal Movement Data typically consists of x, y, time, and individual id.
Movement Classification is a diverse group of methodologies that break a longer movement path into distinct sections, and optionally group those section.
This step is performed to understand a variety of biologically significant characteristics including but not limited to
Key locations
Activity state
Overall direction of travel
Gurarie et al. 2017
The Opportunity
Lots of peer-reviewed methods to classify an animal path into segments of interest and group those segments.
The Problem
Takes time to learn how to implement each method. Each user has to dig into each method one by one.
Method 1
flowchart LR
A1[Look for Code] --> A2[Install from CRAN] --> A3[Reshape Data] --> A4[Run Model] --> A5[Reshape Results] --> A6[Interpret]
Method 2
flowchart LR
B1[Look for Code] --> B2[Install from Github] --> B3[Determine Data Format] --> B4[Run Model] --> B5[Reshape Results] --> B6[Interpret]
Method 3
flowchart LR
C1[Look for Code] --> C2[Email Author] -->|Wait| C3[Generalize Code] --> C4[Run Model] --> C5[Reshape Results] --> C6[Interpret]
The Solution
travelpaths contributer integrates each method one time
Two (of many) ways to identify change points in animal movement data
“Correlated Velocity Models” aka CVM Explicitly models speed and direction changes. Can be used to identify change points ie points where the best model changes. Gurarie et al. 2017
“Change Point Test” aka CPT Uses permutation test based approach to identify changes in direction. Byrne et al. 2009
Finding code (without travelpaths)
Best Case Scenario (CVM)
Direct (working!) link from paper to code
Mid Case Scenario (CPT)
Broken link to in paper
Author was responsive via email
Reshaping for Input (without trackframe)
CVM
Z location data. Can be: a complex vector, a two-column matrix or data frame, an ltraj object from adehabitatLT or a move object from the move package.
