Anthill: A harvester ant colony built from the published science (1.2.0)
An agent-based model of one colony of the Florida harvester ant, Pogonomyrmex badius, from the day a mated queen lands to the death of the colony. It runs in a web browser and, unchanged, headless in Node for replicate studies.
What the ants do: a claustral queen digs her own shaft and raises the first workers on her reserves; workers dig the nest under local rules, with no ant holding a plan; a one-way, age-based division of labour set by the season of birth, which no shortage reverses; trunk-trail foraging with site fidelity and recruitment; seed storage, opening and germination in the chambers, which is what feeds the larvae; daily weather from the climate normals of the study site, with drought years; nuptial flights after heavy rain; corpse removal; alarm at a disturbance on the foraging ground; and annual nest relocation, in which the store and then the brood are carried along the trail to a new nest the colony digs.
Every value in the model is tagged as measured in this species, borrowed from another ant, or invented, with its source, in a single parameter file. The model is deterministic: one seeded stream per system and a fixed one-minute timestep, so a seed reproduces a run bit for bit in the browser and in Node, which a test pins. A study writes a methods report listing the invented values the results rest on and the acceptance tests the model fails.
Known limits, stated in full in the repository: colonies that pass 700 workers fall back the following year, where real colonies hold about 4300; foraging peaks in July where the field maximum is May to June; the alarm scent is shorter-lived and shorter-ranged than one step of the model.
The biology comes from three decades of field work by Walter R. Tschinkel and Christina L. Kwapich in the Apalachicola National Forest, north Florida. Cite them, not this software, for any claim about the ants.
Release Notes
Version 1.2.0. After 1.0.0 the simulation was shown to researchers who work on this species; 1.1.0 and 1.2.0 are what their comments changed.
1.1.0: chamber height no longer counts the shaft running through a chamber (1.58 cm to 0.97 cm at the architecture gate), and a relocating colony’s new nest starts with a descending, spiralling shaft instead of a straight vertical pipe.
1.2.0: chamber height is read from each chamber’s own floor to its own ceiling. Measured that way, most chambers were half a centimetre high, where the species builds one; a worker now raises a low ceiling before widening the floor, and mean chamber height at the gate is 0.91 cm. Deep chamber spacing and the chamber-series count at 1,200 workers now meet their criteria.
Found and not fixed: by their second or third year grown colonies dig caverns in the upper nest instead of stacked chambers. Colony size also remains short of the field value. Every unmet criterion is recorded in docs/VALIDATION.md, and every design decision in docs/DECISIONS.md.
Runs in a browser at https://adiezc.github.io/ANThill-Sim/ and headless in Node for replicate studies. DOI (all versions): https://doi.org/10.5281/zenodo.22819793
Associated Publications
This is a companion discussion topic for the original entry at https://www.comses.net/codebases/7053dd1c-27e9-42b7-9764-7c4b93833a0a/releases/1.2.0/