Photograph a tree, record where it is and what you actually see, and seal the whole evidence set so it cannot be quietly rewritten later. Walk the same route in six months and the record becomes a timeline.
Installs to your home screen. Works with no signal.
These exact photographs, this position, this timestamp and these written observations were fixed together at a moment in time. Change any pixel, any coordinate, any word, and verification fails and says which part broke.
That anyone was standing at those coordinates. A phone reported the position and the record wrote it down. Only dedicated hardware can do better, and until it does, the record says so in plain language.
| Level | What it means |
|---|---|
| L0 | An ordinary survey with no cryptographic seal. |
| L1 | The evidence set is sealed and signed; position measured by the phone.Where TreeKnot is today |
| L2 | A hardware device issues a challenge before capture and binds the sequence of photographs to it. |
| L3 | The device measures time and position itself, independently of the phone. |
The same photographs every time — whole tree, leaf, trunk, root flare, setting — so records stay comparable instead of becoming a pile of arbitrary phone pictures. Position and accuracy are recorded, and the survey waits for the fix to settle.
Every photograph is fingerprinted, the observations are canonicalised, and the whole manifest is signed. The signature covers the evidence, not a conclusion drawn from it.
Walk past the same tree again and the survey recognises it, and the new visit joins that tree's history. The tree changes. The record of what it looked like does not.
You record what is visible — dieback in the crown, a split in the bark at chest height, seedlings coming up in the strip. Not health: 48.
The difference matters when someone disagrees. A score is an opinion presented as a measurement, and once it is signed it looks like a fact. Observations can be argued with while the photographs and the words stay exactly as they were recorded — which is the only thing a signature is actually good for.
Interpretation comes later, from versioned rules applied to fixed evidence. Run a better model against the same photographs in two years and you learn something new without changing what the original survey said.
Pick from a library of Wasatch Front species, or type a name the library does not carry. Either way the survey asks how you worked it out — recognised in the field, a plant ID app, an AI assistant, a field guide, or someone who told you — and names the specific tool.
What it will not do is print a confidence percentage. A model's output is not a calibrated probability of being right, and dressing one up as a measurement is the same mistake as scoring a tree's health out of ten.
Naming the tool is what makes the evidence durable. Run a better model against the same photographs in two years and you learn something new, while the record of what the original survey claimed, and on what basis, stays exactly as it was.
Siberian elm (Ulmus pumila) seeds prolifically, breaks up in storms, and crowds out natives across the Intermountain West. A mature one is also casting shade that a five-gallon replacement will not match for twenty years.
So condition, site suitability and removal priority stay three separate things, and the survey records who has standing to act — city right-of-way, public land, private property — because a recommendation nobody can carry out is just noise.
TreeKnot is in field trial. The survey works, records seal and verify, and repeat visits build a timeline. There is no public tree database yet, and no example records are published here — seeding a verifiable dataset with invented observations would defeat the entire point of it.