Can evidence travel without losing its validity?

Its scientific problem is narrower and more demanding: can evidence remain interpretable as it moves from physical reality, to observation, to model output, to scientific judgment, to verification, to an institutional decision?

Those stages are often compressed by software into a single status such as “verified.” AgEvidence treats that compression as a risk. A record can be cryptographically intact and scientifically invalid. A model can be reproducible and out of domain. A scientifically credible result can still be insufficient for a particular method or buyer.

This lens explains the scientific boundaries that keep the evidence portable without pretending that portability creates universal scientific authority.

The chain begins outside software. A physical event occurs. Someone or something observes it. A source record represents that observation. A model may transform it. A scientist may interpret the result. An authorized practitioner may verify against scoped criteria. An institution may decide whether to rely on it.

AgEvidence’s scientific discipline is to preserve the transitions so confidence cannot silently jump from one stage to another.

The scientific risk is validity laundering: a true statement in one layer is treated as evidence for a different statement in another.

A hash can show that bytes did not change. It cannot show that the original measurement was accurate. A model can execute reproducibly. That does not show it was scientifically applicable to the target population. A practitioner can verify a scoped method. That does not create universal buyer acceptance.

There is no single universal answer to “is there enough evidence?”

A research use, a carbon methodology, a buyer claim, an assurance engagement and a financing decision can require different evidence over the same history. The architecture should preserve the historical evidence and then evaluate it against an explicit purpose rather than declaring the evidence globally sufficient or insufficient.

Uncertainty is part of the evidence, not a formatting problem.

Measurement error, sampling design, calibration state, model domain, temporal coverage, spatial support, imputation, conflicting records and missingness should survive the evidence pipeline. A number should not become more precise because it entered a database or a PDF.

A measured physical outcome does not automatically become an attributable contribution, a legal right, an allocation, a claim or an accepted statement.

Those steps can contain scientific, contractual, methodological and governance judgments. Keeping them separate is particularly important in climate and natural-capital workflows where an observed change may be real but the desired commercial wording may still be unsupported.

AgEvidence can preserve and test evidence lineage. It cannot become the source of scientific authority simply because it stores the evidence.

It may establish deterministic integrity, declared model identity, profile conformance, evidence completeness relative to stated requirements and attributable review history. It does not by itself establish truth of the original observation, scientific validity, methodology applicability, regulatory eligibility, legal title, accredited assurance or claim authorization.