Preliminary GPS collar write-up

Overview

Context

The following is a first shot at using the GPS collar data pulled off HREC animals in the summer of 2017.

Objectives of the project were two-fold:

• Development
  • Construct a DIY GPS collar system
  • Put it on animals and see if it can handle field conditions
• Application
  • Use preliminary data to compare herding behavior across sheep and cattle
  • Determine spatial and temporal patterns of grazer activity
  • Provide insight into different logging approaches that optimize battery life and location/behavior information

Objectives

Specific research questions addressed here – in anticipation of SRM talk and a methods paper – include:

1. What are the distances between collared animals within the same pasture?
   1. Helps determine how representative collared animals are of the overall herd
   2. Helps determine minimum number of units/herd, and number of units/project
   3. To address this, we deployed three units per herd.
2. How do spatial patterns of use differ between single, regular observations and high-frequency, short-duration bursts of observations?
   1. Battery life benefits greatly from unit shut-down
   2. Less-frequent but dense bursts of positions could gather more behavioral information and extend battery life
   3. However, important to ensure gaps in collection don’t obscure certain patterns of movement
   4. We addressed this by sampling at a high, constant frequency (3 positions/minute for a week) and analysing two ways:
      1. Regular: 1 position every 10 minutes
      2. Burst: 3 positions/minute for 10 minutes, with a 50-minute gap between bursts.

Results

• Overall the system worked well:
  • Very few mechanical issues. The GPS units held up to the conditions.
  • No software issues?
  • Field crew reported pretty quick, low-hassle animal handling
  • Battery life remains an issue but some hardware solutions are promising.

• Collared animals tended to stay within 100m of each other. While a single collar might be sufficient for herds of these sizes, at least two are recommended for data redundancy.

• Burst logging is comparable to real-time logging and appears to be wholly sufficient for our objectives.

Recommendations

• Definitely re-deploy in summer 2018!
• Develop within-case battery-extending hardware.
  • If battery life can be extended to 2-3 weeks, constant sampling throughout the season might be possible.
• Burst logging at 10- or even 5-minute durations, once per hour, will provide robust data and help extend battery life.

Distance between GPS units

Two views on distance between GPS units. Top: Overall summary of mean distance between animals with units within pastures. Bottom: mean distance over hours of the day. Note relatively closer clustering during nighttime periods. Animals generally stayed within 100m of each other.

Patterns of different logging schemes

One day, four pastures

Positions as filtered from high-frequency, continous logging to mimic two logging schemes for one day.

Close-up on one pasture, one day

A focus on the two logging schemes versus real-time data, constant, high-frequency logging

Overall representations of patch distribution

Proportion of positions in burned patches for each of three logging types, by month and grazer type. These data suggest there is little difference between burst and regular logging when compared to the continuous, high-frequency data (truth). The dotted line indicates the expected proportion based on area burned; that all data are below this line suggests livestock are avoiding the burned area but the calculation does not account for watering and loafing time.