Post provided by Marina Papadopoulou

Authors

We are three researchers interested in collective animal behaviour. Marina Papadopoulou is a postdoctoral researcher at Tuscia University in Italy, Simon Garnier is a Professor at the New Jersey Institute of Technology (USA), and Andrew King is an Associate Professor at Swansea University (UK). As a Greek-French-Welsh team with empirical, mathematical, and computational backgrounds in different study systems, we put our heads together to provide researchers with the tools needed to quantify and compare collective animal motion.

Animal collectives

The collective motion of animals is often beautiful to watch – whether it’s the pulsing dark shapes created in the open sky by a starling murmuration, or the ant trails meandering through the deep forest. How and why do animals perform such behaviours? Modern technologies, such as drones, high-speed cameras, and bio-logging devices, can now provide data on the movements of the individuals in these large groups, so we can begin to answer these questions.  

With lots of data, one of the best ways to tackle questions about the mechanisms (how?) and function (why?) of collective motion is to identify similarities and differences across different groups, populations, or species: the comparative method. But what sort of unit of comparison should we use, and with so many ways to measure collective movements, what do we compare?

Into the swaRmverse

Unfortunately, there isn’t to date any standardised way to fully quantify and compare collective motion. This gap also roots from the fact that the scientific community interested in collective behaviour is incredibly interdisciplinary, going beyond ecology and evolution to statistical physics, engineering, and complexity science. We thus decided to start building towards a standardised way, and give biologists a tool to have a stronger voice in the field. We thankfully didn’t start from scratch. Simon is the creator of two valuable packages for the analysis of trajectory data, trackdf and swaRm. Standing on their shoulders, we created a new R package for the comparative analysis of collective motion.

Clearly inspired by pop-culture, we envisioned collectives, like spider-men, with their unique characteristics, existing in a multi-dimensional space. Instead of parallel universes, each dimension represents a metric of collective motion. Instead of a character, each unit of comparison is a segment of a group’s trajectory during which individuals are moving in a coordinated manner, that we name an ‘event’ of collective motion. Using dimensionality-reduction techniques, we can create two or three new dimensions to visualise and better understand how events of many collectives relate to each other, from fish schools to ungulate herds, and from the laboratory to the fields of Africa.

Welcome to the Swarm-Verse!

Across the swaRmverse

We have already compared several species using our package: stickleback fish, homing pigeons, sheep, goats, and chacma baboons. When interpreting our first plot comparing events from different species, most of what we saw was obvious: pigeon events clustered together at one end, baboon events at the other, but the placement of some events was unexpected. Goat events took up a central part of the space and overlapped with many other species; are they the GOAT of collective behaviour? Thus, at a broad scale, our package will allow us to identify traditional collective movements (i.e. Peter Parker characters) and more unusual variants (Spider-Ham characters).

Comparisons are also not restricted to empirical data: we can compare simulated collectives from agent-based models for model validation and to create a stronger link between data and theory. We showcase this using the HoPE model of pigeon flocks, previously created by Marina, and we see (with a big relief from her side), that it falls close to the real pigeon flocks and not to the other species.

Beyond the swaRmverse

Our swaRmverse brings together different variants of collective behaviour which, we hope, will inspire more comparative research in the field. Its state reflects just the first steps in a long-term process, we aim for our package to become a living entity of our community. More metrics will be added to capture more variation we see in nature. More functionalities will be added to cover more specialised use-cases and comparisons.

The full article is available here, our (hopefully helpful) step-by-step vignettes here, and our Github repo for official reporting of any issues here. If you are interested in discussing more or adding further functionalities in the package, please also contact us here!

Post edited by Lydia Morley