Skip to main content
Verification is the mechanism that lets an agency accept a platform-issued line instead of a physical site visit. This page documents the model. For the operator-facing workflow, see Verification. The model has four parts: source events, the cross-check, the tamper-evident hash, and the submission bundle. Each part has a defined schema so the system is auditable end-to-end.

Source events

A verification is always anchored to a source event that already exists on the platform. The event must carry a minimum set of fields for it to be verifiable. If occurred_at and logged_at differ by more than 72 hours, the event is still eligible but the cross-check is marked with a late-log flag. Assessors decide whether to accept.

The cross-check

Each event type has a defined cross-check that compares the reported event against the independent satellite record. Each cross-check produces a strength grade: The grade, the pass metadata, and the delta values all enter the evidence bundle. Nothing is hidden.

Tamper-evident hash

Each verification produces a SHA-256 hash over the canonical serialization of:
  1. The source event (all fields above).
  2. The satellite pass metadata used for the cross-check (source, tile id, acquisition time, cloud mask).
  3. The measured delta values.
  4. The cross-check grade and any flags.
  5. The verification timestamp and platform version.
The hash is:
  • Deterministic - any party with the raw inputs can recompute it and confirm.
  • Bound to a specific imagery pass - swapping the satellite source or date changes the hash.
  • Recorded on submission - the agency stores the hash; auditors can request the raw inputs and recompute.
The hash is not a blockchain. It does not need to be. It is a portable, reproducible fingerprint that any third party can validate given the raw evidence bundle. Blockchain anchoring is optional and available for agencies that require it.

Evidence bundle

The bundle is what the agency actually receives. It is the same underlying content across all three submission formats; only the presentation differs.

Submission formats

All three preset submission formats are projections over this bundle. They differ in layout and in which optional sections are included. Agency-specific form fields are provided by each agency and mapped once by the platform. New agencies onboard by supplying the field mapping; the underlying bundle does not change.

Confidence, not certainty

A verification confirms consistency between two independent records: what the platform user logged, and what the satellite captured. It does not, and cannot, prove intent or interior quality.
  • What it proves: an operation of the reported type occurred on the reported parcel within the imagery window.
  • What it does not prove: the identity of the operator, the specific product used, or off-parcel activity.
Assessors retain discretion. Verification is designed to move the default from “prove it with a site visit” to “accept unless flagged”, with the flags surfaced honestly in the bundle.

Auditing an existing verification

Any party with the verification id and the raw evidence bundle can:
  1. Retrieve the referenced imagery from the named source (Sentinel-2, Sentinel-1) using the tile id and acquisition timestamp.
  2. Recompute the cross-check metric against the reported geometry.
  3. Recompute the SHA-256 hash over the canonical serialization.
  4. Confirm the recomputed hash matches the one stored by the agency.
If any step fails, the verification is invalid and the agency has grounds to reject the submission even after initial approval. This is what makes it tamper-evident: falsification requires forging an independent satellite record.