Pollenomics Publication Model

Publications combine unlike evidence without declaring it equivalent. Every layer has a role, every product has a scope, and every visible feature remains downstream of an admission decision.

The model deliberately separates three questions:

1. Is the source represented faithfully? — acquisition, identifiers, and normalization answer this.
2. Is the record fit for a declared use? — evidence review and admission answer this.
3. Is the product internally coherent? — manifests, traceability, release checks, and rendered outputs answer this.

A positive answer at one boundary does not imply a positive answer at the next. This is why a normalized row can remain excluded, and why a coherent map can be internally valid without claiming complete recovery, uniform precision, or fitness for an undeclared analysis.

Evidence Roles

Role	Examples	Supports	Does not support alone
direct evidence	admitted human and animal aDNA samples, direct field observation	a source-backed record at its declared place and time posture	regional completeness or causation
scientific context	LandClim, Neotoma, SEAD, RAÄ	environmental or archaeology setting	sample identity or automatic temporal overlap
geographic framing	boundaries and SVAR lake registry	scope, filtering, lake identity, and map extent	scientific association
decision support	lake rankings and sensitivity scenarios	reproducible prioritization under declared inputs	sampling readiness or optimal coring location
accountability	caveats, exclusions, drift reviews, traceability reports	why evidence is present, qualified, or absent	a stronger claim than the governing record

Scope Inheritance

flowchart TB
    Evidence["governed evidence state"] --> World["world publication"]
    World --> Europe["Europe-plus scope"]
    Europe --> Nordic["Nordic scope"]
    Nordic --> Countries["country bundles"]
    Nordic --> Lakes["Sweden lake decision support"]

The hierarchy selects records; it does not clone truth. A country bundle is a filtered descendant of broader evidence state and cannot legitimately carry a stronger locality, chronology, coordinate, or source claim. Specialized lake products retain their scenario inputs and remain attached to the same Nordic publication family.

Publication Admission

Normalization makes a record consistent enough to review. It does not make the record universally publishable. Admission is evaluated against a named product:

1. confirm source and record identity;
2. evaluate the evidence dimensions required by the product;
3. apply geographic and domain scope;
4. admit, qualify, or exclude with a recorded reason;
5. emit membership and traceability alongside presentation.

Filtering in a browser changes visibility after publication. It cannot admit an excluded record or alter its evidence role.

Product Contract

Every publication resolves six elements:

Element	Reader-facing meaning
scope	geography, source families, species, and scenarios considered
membership	stable identifiers actually included in the product
roles	direct evidence, context, framing, decision support, or accountability
semantics	spatial precision, temporal posture, units, and null meaning
traceability	joins from visible objects back to governed evidence and sources
release posture	checks passed, claims refused, and known limitations

The world bundle is the broad product assembly. Europe-plus, Nordic, country, and lake-decision products inherit its evidence semantics while applying narrower membership or a specialized decision contract. A descendant product can be more selective; it cannot silently strengthen its ancestors' facts.

Product Classes and Claim Ceilings

Product class	May claim	Must not imply
evidence register	governed source, identifier, and review state for each listed record	that every relevant source or record has been recovered
point layer	admitted spatial membership with row-level provenance and visible coordinate class	uniform precision, absence of refused evidence, or analytical independence
contextual layer	source-specific environmental, archaeological, or geographic context	direct sample support or causation from proximity
regional bundle	reproducible selection from the parent evidence state	stronger evidence semantics than the parent product
decision-support output	ranking under named inputs, transformations, and scenarios	field readiness, optimality, or a scientific conclusion independent of assumptions

The claim ceiling travels with membership and semantics. A country subset does not turn contextual chronology into a numeric interval; a filtered map does not turn an approximate coordinate into a source-supplied one; a high lake rank does not become a coring recommendation without field evidence.

Membership Accounting

A product is accountable when every candidate considered by its contract has one recoverable outcome. Use mutually exclusive states for accounting even when the rendering groups them differently:

Candidate outcome	Visible in the product?	Required record
admitted	yes	membership and traceability
admitted with qualification	yes, with the qualification	membership, traceability, and visible caveat
excluded	no	named failed rule and governing evidence
unresolved or recovery-bound	no	ambiguity or recovery record describing missing support
outside scope	no	scope predicate and candidate identity

flowchart LR
    Population["declared candidate population"] --> Outcome{"one governed outcome"}
    Outcome --> Admit["admitted"]
    Outcome --> Qualified["admitted with qualification"]
    Outcome --> Excluded["excluded"]
    Outcome --> Recovery["unresolved or recovery-bound"]
    Outcome --> Outside["outside scope"]
    Admit --> Account["product accounting"]
    Qualified --> Account
    Excluded --> Account
    Recovery --> Account
    Outside --> Account

The accounting identity is conceptual but strict: the declared candidate population must equal the sum of those disjoint outcomes. Browser filters do not alter that population; they only change which already-admitted members are displayed. A record missing from both membership and accountability surfaces is not a harmless omission—it is an unexplained product-integrity gap.

Release Gates Preserve Honest Claims

Release checks test the relationship between evidence and publication rather than declaring the underlying science complete. For animal evidence, the current gate confirms that published points retain required traceability, blocked sample-site and chronology rows do not leak into exact outputs, and contextual chronology is not rendered as numeric time. The same gate refuses collection-completeness and uniform-precision claims because project recovery remains incomplete and admitted rows retain different evidence classes.

That combination is intentional: a product can be valid for its declared membership and still decline a broader completeness claim. Passing integrity checks means the output says no more than its governed evidence supports.

Reading Mixed-Domain Outputs

Interpret the layer before interpreting proximity. Two nearby points can have different chronology postures, spatial precision, source maturity, and scientific roles. The map supports comparison by keeping those differences visible; it does not erase them.

Continue to the cross-domain evidence matrix for maturity by family, published reports for the output tree, and geographic evidence surfaces for map behavior.