Models and Analysis of Animal Movements: From Individual Tracks to Mass Dispersal

1. What they did

Almost everything in an animal’s life, from finding food and mates to avoiding predators and migrating, depends on how they move. Tracking technology now produces huge amounts of movement data, but those data are hard to analyse: each location depends on the one before, and positions are often recorded at irregular intervals. This dissertation develops mathematical models and statistical methods for describing animal movement, starting with single animals’ paths and ending with the spread of whole populations. The author describes the work as mainly about methods, not testing ecological ideas.

It has two parts:

• Models of movement (Chapters 2 and 3). The author re-examined the widely used correlated random walk, adapted it to locations recorded at irregular intervals, and tested it on the GPS track of a dugong in Shark Bay, Western Australia. The author then presented a more flexible continuous model that can describe drifting, spiralling and turning paths, tested on laboratory videos of a swimming alga.
• Ecological applications (Chapters 4 to 6). The author used movement scales to predict how often moving animals meet (predators and prey, foragers and food), and applied this to the survival of young salmon migrating down the Snake and Columbia rivers. Next came a way to find sudden changes in behaviour along a track, tested it on simulated data and a Magellanic penguin track, and applied it to five of seven foraging trips by a nursing female northern fur seal in the Kuril Islands, Russia. Finally, the author modelled how differences between individuals shape the spread of a population, using published mark-recapture data on bluehead chub in a Tennessee creek, salmon travel times between two dams, and the 2006 Seattle marathon.

A closing chapter reflects on movement models in ecology and on movement studies in conservation.

2. What they found

Movement scales

• Almost all movement is correlated over short times and random over long times, so the useful question is at what time and distance a path loses its “memory”.
• The dugong’s direction stayed consistent for only tens of minutes: the half-time of persistence was 9.55 minutes. With locations every hour, any pattern in the dugong’s turning would be undetectable. Of 5040 possible 15-minute positions over 52.5 days, only 1471 were recorded.
• The penguin kept moving in one direction for several hours at a time.

Encounters and salmon survival

• In earlier data on 287 release groups of juvenile spring Chinook salmon (1993 to 2003), survival was linked much more strongly to distance travelled than to travel time.
• The models explain why: when predators sit and wait, survival is like running a “gauntlet” that depends only on distance, and faster river flows would help little. Speed only matters once predators move around and forage actively.
• In the mid-Columbia hydroacoustic data (Appendix C), survival and travel times differed strongly between river sections; nearly 10 times more sockeye died in the tailraces or passing two of the dams than in any mid-reservoir section.
• For two-dimensional searching, the author suggests that a forager’s movement scale should be about the distance between targets or more, so a forager’s movement may give clues about where their prey is.

Fur seal foraging trips

• The method detected sudden, clear changes in behaviour, even in gappy data.
• Travelling out from and back to the rookery showed strongly autocorrelated (steady, directed) movement, and changes were often found around dusk and dawn, with more activity between sunset and sunrise.
• On the first trip (5.1 days, up to 96 km from the rookery), speed in the main direction of travel dropped from about 4 to about 2 km per hour at the first change point during the first night. During the day, speed dropped to 0 to 0.5 km per hour.
• The direction of travel stayed consistent for 30 to 40 minutes while travelling, but only 4 to 15 minutes during the rest of the trip. The change in this “steadiness” often separated travel from other behaviour better than speed did.

Differences between individuals

• A model that lets movement vary continuously between individuals matched the chub’s dispersal (a sharp peak with long tails) with few parameters, but could not explain its skew.
• For salmon, the new travel-time model fitted better than the standard model in almost all cases. Steelhead appeared to migrate with each fish choosing its own steady speed downstream and closely following river flow. Chinook moved in a more random, diffusion-like way, perhaps feeding and holding station, and did not respond to higher flows.
• In the marathon, both differences between runners and variable pacing shaped finishing times. Differences between women runners rose sharply in the harder second half of the race.

3. What they concluded

• Researchers should embrace rather than avoid the dependence between successive locations, because the time and length scales at which movement becomes independent tell us a lot about the movement itself.
• These scales help predict dispersal, migration and encounter rates.
• The change-point method can describe complex behaviour from gappy tracks and could be used with Argos, archival or radio telemetry data, and on land. Its real value will come from comparing the results with other data, such as diving, stomach temperature (prey capture) and environmental conditions.
• Differences between individuals should be built into dispersal models; for migrating salmon they reveal a behavioural signal that standard models miss.
• The author is frank that the work does not test ecological hypotheses, and that inferring how animals respond to their surroundings and internal state “remains largely unfulfilled”.

6. Other technologies, skills or innovations

• GPS tags towed by dugongs; Fastloc GPS data loggers with dive, temperature and light sensors on the fur seal, plus a stomach temperature sensor; satellite tags (ARGOS) on penguins.
• Passive integrated transponder (PIT) tags and hydroacoustic tags with detector arrays for juvenile salmon; video tracking of algae in the laboratory; timing chips in a marathon.
• Annotated R code in the appendices so readers can apply the methods.

7. Implications for conservation (from the paper)

• For juvenile salmon, the author suggests managers study what happens within reservoirs, where predators are and how all the animals move, rather than relying on reservoir-wide averages. Raising river flows is only likely to help survival when predators forage actively.
• The most direct use of movement studies in conservation has been identifying critical habitat.
• The author also cautions that movement information alone does not save species: conservation is driven first by political will and economics, with examples from Steller sea lion and Columbia River salmon research. Understanding movement and behaviour can still deepen understanding of how ecological systems work.