06/ FOUNDATIONSB / VIRTUAL FENCINGestablished elsewhere → untested here

The boundary is coordinates.Nothing is built, and nothing blocks anything that isn’t wearing a collar.

A physical fence excludes everything that meets it. A virtual fence excludes only the animals carrying collars — everything else in the landscape moves through the line as though it were not there. In a landscape whose conservation target includes a migrating reptile, that is the whole reason to prefer coordinates over wire.

REGISTER

The film

Nineteen stops in four moves: the wire, the collar, the learning, and the two properties that follow from it. Click a stop — where a transition has a clip it plays and lands on the next frame, and the labels appear once the picture is still.

01

Open ground

STOPPED · 01 / 19

NO IMAGE IN THIS FILM IS A PHOTOGRAPH. Every frame is generated. The pasture is generic — no Galápagos location or vegetation is depicted. The tortoise stands for what an uncollared animal is; it is not a filmed Galápagos scene. The hardware is drawn generically and no manufacturer is named or shown.

What the collar actually does

A boundary is drawn in software as a set of coordinates. Each animal wears a GPS collar that knows where it is relative to that line. There is nothing in the ground and nothing in the air.

As the animal approaches, the collar plays a tone that rises in pitch. If the animal keeps going past the line, a brief electrical pulse follows. The pairing is the whole technology. The animal is not learning a map — it is learning that the tone predicts the pulse. Once that association forms, it turns at the sound, and the pulse becomes rare.

Because it is learned rather than built, three things follow. It needs a training period. Animals that have met an electric fence before learn faster. And it is not restraint: an animal that is determined, or frightened, can walk through.

Two further things follow from the boundary being software rather than structure. It can be moved without dismantling or building anything, so a grazing cell is relocated on the cadence the vegetation requires rather than the cadence fence-building allows. And the management goal it serves is suppression of regrowth by grazing — without herbicide.

What this does not do

It does not remove established plants; grazing manages regrowth. It does not work on an untrained animal, and it does not restrain one. And it does not target Rubus niveus in the first phase. The first-phase target vegetation is Pennisetum purpureum — elephant grass — chosen because its palatability is not in question, which takes one behavioural assumption off the critical path. Suppression of R. niveus is the objective of the programme, not of the first cells.

Not GHRI data: no pulse rate, boundary interaction or containment outcome has been measured by this project.

ESTABLISHED ELSEWHERE

  • Virtual fencing for invasive-species control is already published as a concept — Bennett et al. (2026), Biological Conservation 318:111866.
  • Rangeland pilots exist in North America. Application to a woody invasive on a tropical oceanic island has not been reported, and the return interval for Rubus niveus under equatorial conditions is unknown.
  • Collar activity-index measurements correlate strongly with directly observed grazing intensity — Frontiers in Veterinary Science, 2025.
  • Heifers with previous electric-fence experience acquire the association faster.
  • Manufacturer specification: cattle collar battery up to about a year, goat collar about eight months, a fourteen-day solar reserve, nested exclusion zones at 10–15 m radius. None of it field-verified here.

UNTESTED HERE

  • Zero collars deployed. No animals collared, no site instrumented, no field season.
  • No telemetry, no boundary-interaction counts, no containment data, no behavioural observation, no vegetation baseline.
  • No supply agreement of any kind. No equipment ordered or received.
  • GPS performance under closed Scalesia canopy is untested.
  • No equatorial or tropical deployment experience is confirmed for the hardware.

The planned pilot

ANIMALS
Approximately 10–15 borrowed cattle, as filed. The internal record also carries a recommended pilot size of 15–20 and a minimum of 8–10, and instructs that herd size be treated as a range rather than a settled number.
SITE
Private agricultural land in the Santa Cruz highlands, outside Park territory.
DESIGN
Three intensively measured cells of about 0.3 ha in a paired-block design blocked by property, plus additional lightly monitored grazing area.
SEASONS
Spanning both the garúa and the dry season.
FIELDWORK
Early 2027.
WHAT IT IS FOR
Estimating the size of an effect, not testing a hypothesis definitively. The number of animals is small and the design is honest about what that allows.

Where this stands

PILOT IN PREPARATION

Pilot project in preparation, on private agricultural land in the Santa Cruz highlands. Regulatory consultations with the competent agencies have not yet been submitted. No equipment has been deployed in the field.

Who does what

The collars are a commercial system built by a third party. GHRI would buy or borrow hardware and run a pilot on private land; it does not speak for the manufacturer, and the manufacturer does not speak for it.

The grazing is a management tool, not a product. Nothing in this track produces anything to sell.