6.4 Analysing Spatio-Temporal Data – Understanding animal movement
In a dynamic approach to GIS, the way of thinking about spatial features changes. Points become events, which occur at a given location for a specified period of time. Similarly, lines become tracks, representing the movement of entities across a study area. Polygons become dynamic features whose dimensions and boundaries may change over time.
GIS software is not only evolving to store spatio-temporal data like this, but several software packages are now able to analyse temporal variations in spatial data. For example, the software package TerraSeer includes space-time clustering tests such as the Knox test. These tests identify whether events in one time period will occur close to events in the previous time period. For example, space-time clustering statistics might identify how far the risk of foot and mouth disease might increase on cattle farms close to previously reported cases of the disease. Techniques are also becoming available for analysing dynamic polygonal data ( for example, May Yuan’s presentation, examining storm tracks using GIS). However, this object explores analysis tools for analysing tracks and focuses in on tools for investigating patterns of animal movement.
Where do animal track data come from?
A common source of animal track data are satellite telemetry devices, which can be attached to individual animals. Signals from these tags are received by satellites, usually those run by the National Oceanic and Atmospheric Administration (NOAA). The signals received by one or more satellites may then be processed to give a position. This mode of data collection is most often used with marine animals because the positional accuracy of this technology is low (to within hundreds of metres at best). One of the main differences between satellite telemetry and GPS devices is that satellite telemetry devices are generally less accurate.
Another alternative is to use GPS devices, mounted on collars and attached to animals. Whilst this provides more accurate positional data, it may not be possible to ‘fix’ an animal’s position if it moves beneath dense vegetation. Another commonly used technology is radio-tracking, in which a device that emits radio-waves at a specific frequency is attached to an animal. Field ecologists then record these signals from a hand-held or mobile receiver with directional antennae. Generally, the ecologist needs to move from place to place fairly rapidly to fix the animal’s position or to work in pairs with another field team. Together with data on the receiver’s location, this information may be processed using triangulation to calculate the animal’s position.
Some ecologists refer to different types of track:
- tracks without associated time stamps for each positional fix
- irregular tracks, which have associated time stamps for each positional fix, but where the time interval between successive fixes varies
- regular tracks, which have associated time stamps that are a set time interval apart, e.g. every 60 minutes.
Finally, it is also possible to use observation notes from field ecologists, plotted onto a map of the study area. This method is clearly very labour-intensive, since it entails field staff physically following an animal’s movement and either making direct observations or recording its spoors (tracks or other traces of its passage).
Analysing animal movement data
On receiving such data on animal movements, the first task for the GIS analyst is to map the animals’ tracks and search for patterns. Software for displaying animal movements over time enables track data to be animated, showing the changing tracks over time. A second question is how to convert animal tracks into a raster surface that reflects the intensity of habitat use by an animal or group of animals. One approach is to use tools such as ArcGIS’s ‘kernel density’, to ‘smooth’ track waypoints. This approach involves estimating the density of waypoints recorded within a moving window, weighting those waypoints that are closest to the centre of the moving window more highly than those that are further away. However, such an approach ignores the fact that a track represents animal movement. More sophisticated approaches have begun to emerge therefore, notably Brownian Bridge Movement Models (BBMM). Such models estimate the probability of an animal moving over a given distance within a set period, based on successive waypoints from GPS collars and other such devices. These movement probabilities are then used to ‘smooth’ the waypoints that make up recorded animal tracks, so providing a more realistic surface depicting the animals’ intensity of habitat use.
Many animals exhibit territorial behaviour, repeatedly moving around within the same area. For example, a badger will often patrol the same patch of ground every night to avoid coming into contact with animals from neighbouring setts. Spatial ecologists often describe such territorial behaviour as site fidelity. An important preliminary question for the GIS analyst is whether the tracks recorded in the GIS are suggestive of territorial animal behaviour. It is possible to perform a statistical test for site fidelity on track data for a given animal – this is described in more detail in the illustrations below.
If an animal does show territorial behaviour, then it may be appropriate to calculate its home range. The home range would be the area that a badger repeatedly visits every night. There are many methods of measuring an animal’s home range. One of the older methods for measuring home range, the Jennrich-Turner method, is described below.
Test your knowledge of GIS tools for analysing animal movement by answering the following short question and doing the quiz below. Optionally, if you are interested in exploring software other than ArcMap, you may wish to undertake this exercise, which provides a gentle introduction to the analysis of wildlife tracks using the RStudio software.
In GIS terms, how would you describe the difference between a vector line feature and a track?
Tracks represent movement, so each node along will be 'tagged' with a particular time when that point was visited. The arc segments joining the nodes along the line together thus represent the movement of an entity. A line is a more generic concept. As well as representing dynamic features such as storm tracks or animal movement, lines can also represent static features such as a deer fence or a stream. The nodes along a line are not necessarily 'tagged' with a particular time.
References (Essential reading for this learning object indicated by *)
* From early and now somewhat obsolete work on animal tracks, a number of more recent initiatives have started. This site describes work to develop QGIS tools to analyse animal movement: http://www.faunalia.eu/en/animove.html. Similarly, the GeoSpatial Modelling Environment (GME) also provides tools for generating metrics from the movement paths of individual animals: http://www.spatialecology.com/gme/movementpathmetrics.htm There are a number of other, more recent developments that provide for analysis of animal movements. This blog post provides some useful links to other, similar software that is supported in later versions of ArcGIS: http://benmearns.blogspot.co.uk/2010/10/animal-movement-and-home-range-options.html
There is now an international initiative called MoveBank that acts as a repository for animal track data, much of which can be downloaded for free: https://www.movebank.org/
For an example of how GIS may be used to look at the analysis of events, see the TerraSeer web site: www.terraseer.com
For an example of how GIS may be used to look at dynamic polygonal data, see May Yuan’s spatial analysis of storm trajectories: http://ags.ou.edu/~myuan/papers/cagis.pdf
The RStudio practical materials above make use of material from the following resources:
Calenge, C. (2016): Analysis of animal movements in R – the adehabitatLT package. https://mran.microsoft.com/snapshot/2016-08-05/web/packages/adehabitatLT/vignettes/adehabitatLT.pdf
..as well as the R statistical pacakge (https://cran.rstudio.com/) and RStudio (https://www.rstudio.com/products/rstudio/download/). It also draws on some introductory RStudio materials, namely http://stat545.com/block002_hello-r-workspace-wd-project.html and https://www.sitepoint.com/introduction-r-rstudio/.