Evidence Publications¶

Pollenomics publications are derived, versioned views over the governed data state. Reports, maps, evidence tables, ranking outputs, and caveat surfaces are assembled together so a visual result can be inspected alongside its inputs, traceability, and scientific review.

The domain language defines product, manifest, publication member, projection, evidence role, and product scope.

Geographic Lineage¶

flowchart TB
    Data[("governed data and admitted evidence")]
    Data --> World["world evidence surface"]
    World --> Europe["Europe-plus selection"]
    Europe --> Nordic["Nordic selection"]
    Nordic --> Sweden
    Nordic --> Norway
    Nordic --> Finland
    Nordic --> Denmark

The hierarchy is a lineage and subset contract. Regional and country outputs are narrower selections over shared evidence; they are not separately curated databases. Subset validation detects rows that appear in a child without a governed parent or that change meaning between scopes.

Publication As A Governed Query¶

A publication behaves like a recorded database query whose inputs, predicates, members, and non-members remain inspectable:

Query component	Governed publication equivalent
input population	versioned normalized and reviewed evidence identities
relation set	source, sample, site, locality, chronology, coordinate, and geography joins admitted by contract
predicates	product scope, evidence role, identity posture, spatial fitness, temporal fitness, and exclusion rules
projection	declared fields, units, coordinate posture, warnings, and source locators
result identity	bundle manifest, version, member IDs, and companion artifact inventory
non-member accounting	outside-scope, excluded, refused, conflicted, and deferred identities with reasons

flowchart LR
    Population["reviewed evidence population"] --> Relations["governed relations"]
    Relations --> Predicates{"scope and admission predicates"}
    Predicates -->|member| Projection["declared product fields"]
    Predicates -->|non-member| Account["reasoned non-member outcome"]
    Projection --> Manifest["versioned bundle manifest"]
    Account --> Review["exclusion and review surfaces"]

This is why a country map is not produced by clipping arbitrary coordinates. Containment is only one predicate; evidence role, point class, chronology posture, lineage, and product rules continue to govern the selected member.

Projection Invariants¶

A publication projection is valid only when it preserves the semantics of the database state it selects:

Invariant	Required publication behavior
identity preservation	every member resolves to the same typed governed object used during admission
role preservation	direct evidence, context, sampling support, and framing remain distinguishable
precision preservation	projected fields do not imply finer place, time, or identity support than their authorities
state preservation	qualifications and conflicts material to use remain reachable; exclusions remain accounted for
population preservation	numerator, eligible denominator, scope, and observation unit travel with aggregate claims
revision consistency	manifest, members, traceability, warnings, and rendering derive from one database snapshot

Projection may rename columns, select fields, transform coordinate representation, or aggregate declared populations. It may not mint upstream facts, discard a material qualification, or reinterpret a contextual member as direct evidence.

Product Existence And Rebuildability Differ¶

A committed publication can remain available after the current database state reveals that one of its prerequisites is missing. The product then remains a retained, versioned result; it is not silently certified as reproducible from the present inputs.

Question	Governing evidence
Does the product exist?	bundle manifest and committed member inventory
What did it publish?	product contract, evidence rows, exclusions, and review surfaces
Can the current snapshot rebuild it?	materialized source-family stages, exact input artifacts, and passing publication gates
Is it scientifically reusable now?	rebuildability plus the evidence and qualifications required by the proposed reuse

The evidence-stage matrix can therefore report published while also reporting a missing normalized or review stage. That combination is a visible integrity warning, not a contradiction and not permission to infer the missing evidence backward from the product.

flowchart LR
    Product["retained publication"] --> Exists["existence and historical membership"]
    Inputs["current contracted inputs"] --> Ready{"rebuild prerequisites present?"}
    Review["current review and gates"] --> Ready
    Ready -->|yes| Rebuildable["rebuildable current projection"]
    Ready -->|no| Blocked["retained product with explicit blocker"]

Publication Freshness Is A Vector¶

One generation date cannot describe the freshness of a cross-domain bundle. Each family can have a different capture date, evidence revision, and review posture, while the bundle has its own generation identity.

Freshness coordinate	Governing evidence	Change that advances it
source freshness	family release, retrieval time, response identity, and digest	a newly identified and accepted capture
preparation freshness	normalized member and transformation revision	accepted member or semantic reconciliation
review freshness	claim decisions, conflicts, and fitness revision	reevaluation against current evidence and rules
product freshness	manifest, contract, member inventory, and generated time	coherent bundle regeneration

flowchart LR
    Source["source freshness by family"] --> Preparation["preparation revision"]
    Preparation --> Review["review revision"]
    Review --> Product["product generation"]
    Product -. "does not advance" .-> Source

Regenerating a map advances product freshness only. It does not make an old capture current or fill a missing review stage. A freshness statement names the coordinate relevant to the claim and exposes older or blocked coordinates that constrain reuse.

Publication Bundle¶

A geographic bundle can include:

a manifest and summary;
an interactive map document and local map assets;
source-family GeoJSON layers;
animal evidence and point-traceability tables;
a map publication contract;
candidate-site rankings and sensitivity results;
a ranking-engine manifest; and
scientific review in machine-readable and narrative forms.

flowchart LR
    Inputs["selected evidence inputs"] --> Manifest["bundle manifest"]
    Manifest --> Map["interactive map"]
    Manifest --> Tables["evidence and traceability tables"]
    Manifest --> Ranking["ranking and sensitivity"]
    Manifest --> Review["scientific review and caveats"]

The map is one consumer of the bundle, not the bundle's authority.

Bundle Identity¶

A publication is identified by the combination of version, geographic scope, member inventory, and governing contracts. Two bundles with the same map title are not interchangeable when any of those fields differ.

Identity field	Question it answers	Why it must travel with reuse
version	which governed data state was published?	later recovery or review can change admission
scope	which geographic selection was applied?	a country result cannot stand for the Nordic or world collection
member inventory	which tables, layers, reviews, and rankings belong together?	prevents artifacts from different runs being combined
contract identity	which rules admitted and qualified records?	makes the visible subset reproducible

The bundle is complete for interpretation only when its manifest, evidence members, review surfaces, and renderings resolve to the same identity. A standalone HTML map can remain useful for exploration, but it is not a complete scientific publication.

Bundle Closure¶

A bundle is closed when every declared member and companion can be reconciled without inferring state from a rendering:

Closure check	Required agreement
inventory	manifest lists every required structured, review, warning, and rendering companion
membership	admitted identities equal manifested identities after declared scope selection
traceability	every visible scientific member resolves to governed evidence and a source locator
qualification	material precision, conflict, context, and refusal posture survives every relevant projection
accounting	counts reconcile by observation unit across eligible, admitted, qualified, and non-member populations
geography	child members are valid subsets of their parent scope and retain invariant evidence meaning
rendering	labels, filters, popups, tables, and narrative do not contradict structured member state

A file-complete directory can still fail closure when its members disagree. A scientifically coherent product can also contain explicit refusals and empty subsets; closure requires those outcomes to be accounted for, not hidden.

Member Decisions Travel With The Bundle¶

Decision field	What it prevents
stable member identifier	confusing the same label across sources, scopes, or versions
source-family role	treating context, framing, and direct evidence as interchangeable
admission or exclusion reason	reading absence as proof that the source has no record
spatial basis and precision	inferring survey precision from marker placement
temporal posture	converting broad, contextual, or unresolved time into a numeric comparison
parent and child membership	double counting one record across nested geographic products

These decisions are part of the publication even when a rendering does not display every field. A map may shorten them for legibility; the evidence and traceability members retain them for audit and reuse.

Bundle Authority Order¶

When publication artifacts disagree, resolve the claim through the narrowest governing surface:

the bundle manifest governs product identity, version, scope, and members;
the publication contract governs required layers, fields, and admission behavior;
point traceability and evidence tables govern visible record identity;
scientific review and warning surfaces govern material qualifications;
maps and narrative reports render and interpret that governed state.

flowchart TD
    Manifest["manifest and product scope"] --> Contract["publication contract"]
    Contract --> Rows["evidence and traceability rows"]
    Rows --> Review["scientific review and warnings"]
    Review --> Map["interactive map"]
    Review --> Narrative["reader narrative"]

A map label or narrative sentence cannot override a manifest member, evidence identifier, precision posture, or material warning. If the rendered surface disagrees, the rendering is stale or defective and must be corrected.

Read A Bundle In Order¶

flowchart LR
    Identity["identify version and scope"] --> Inventory["verify bundle members"]
    Inventory --> Contract["read admission contract"]
    Contract --> Evidence["inspect evidence rows"]
    Evidence --> Qualification["apply review and caveats"]
    Qualification --> Rendering["interpret map or report"]

This order separates product discovery from claim evaluation. The manifest locates the right product; the contract explains why records are present; the evidence rows support the claim; review surfaces preserve qualifications; and the rendering helps the reader see the result. Reversing that order risks treating visual prominence as scientific authority.

Worked Bundle: World v66¶

The checked-in world manifest makes the reading order concrete:

Surface	Governed fact	Reader decision
`world_bundle.json`	`scope_key: world`, version `v66`, generated `2026-06-22`	select the product identity before opening a rendering
`world_map_publication_contract.json`	five declared layer rows and their filter behavior	determine which roles are eligible at world scope
`world_point_traceability.json`	visible feature-to-evidence relations	resolve a marker to its governing row
`world_scientific_review.json`	qualifications and comparison limits	constrain interpretation
`world_map.html`	interactive rendering of the governed members	explore after scope and role are known

The map contract currently declares 1,231 AADR features, one dromedary context feature, 26 goat features, 207 horse features, and four boundary features. Those counts are not one population: they mix human samples, differently qualified animal evidence features, and geographic framing. The manifest binds them into one product without making their observation units equivalent.

flowchart LR
    Bundle["world_bundle.json / v66"] --> Contract["world map contract"]
    Contract --> AADR["1,231 human sample features"]
    Contract --> Animal["234 animal evidence features"]
    Contract --> Boundaries["4 framing features"]
    AADR --> Trace["point traceability"]
    Animal --> Trace
    Trace --> Review["scientific review"]
    Review --> Map["world_map.html"]

A valid world citation names world, v66, the relevant member identity, and its governing evidence role. “Visible on the world map” is not enough to distinguish a human sample, animal context feature, or boundary polygon.

Publication Gates¶

Animal publication checks currently enforce that:

published points retain the identity support declared by their point class: final sample lineage or visibly provisional project context;
provisional project context does not become recovered sample evidence;
sample-site disagreement is not flattened into one project locality;
blocked sample-site rows do not publish as exact sites;
unresolved or conflicting chronology does not enter country or atlas output;
project-level locality substitution does not masquerade as sample evidence;
broad chronology does not publish as a numeric time window; and
contextual temporal rows remain distinguishable from numeric comparisons.

Passing these safeguards permits the qualified publication surface. It does not authorize unrestricted precision or final-release maturity; those stronger claims have separate evidence requirements.

Choose A Publication¶

Geographic publication lineage explains how world, regional, and country products preserve identity and subset meaning.
Reports provide scope-oriented narrative and inventories.
Maps provide visual exploration across evidence families.
Map inputs identify the source files behind visible layers.
Point rules explain eligibility and coordinate posture.
Filters and popups explain interactive selection and displayed metadata.
Publication types distinguish scientific evidence, context, framing, and review surfaces.
Collection summary binds publications to the collected source state.
Limits records the boundaries that remain after a bundle passes its structural checks.

The generated report portal provides direct access to world, regional, country, review, and caveat outputs.

Reuse Rule¶

Reuse the smallest surface that supports the claim. A map screenshot is enough to illustrate layout, but not to support a sample-level scientific assertion. For that, retain the evidence row, source lineage, locality and chronology posture, coordinate basis, bundle version, and applicable caveats.

A reusable citation package therefore contains the product identity and the narrowest evidence member that supports the statement. If that member cannot be named, the statement is not yet traceable enough for scientific reuse.

For a count, also retain the observation unit, numerator, eligible population, exclusions, and scope. “234 points” is a product-membership statement; it is not a recovery rate until a defensible denominator and recovery rule are named.

The revision and state model defines the database snapshot from which a projection receives its authority.