Cross-Domain Source Roles

Bijux Pollenomics combines evidence families because they answer related questions, not because they are interchangeable. Each family retains its own observation unit, authority, spatial precision, temporal posture, and product role.

An evidence role is an authority boundary, not a display category. It governs which assertions a record may contribute before any map styling, proximity calculation, or ranking is applied. A downstream product can narrow that authority, but cannot promote it.

Role Matrix

Family	Governed unit	Product role	Strongest supported reading
LandClim	pollen site sequence and REVEALS grid cell	environmental context	time-aware pollen and vegetation-reconstruction coverage
Neotoma	pollen site	environmental context	site-level pollen coverage under explicit temporal posture
SEAD	environmental-archaeology site inventory row	archaeology context	mapped Nordic inventory context without current numeric time support
RAÄ	published heritage record aggregated to density cells	archaeology context	Sweden-specific registry density under a declared classification
AADR	release-pinned human aDNA annotation row	direct human evidence	admitted human sample metadata within a versioned product
animal ancient DNA	project-owned sample and its claim records	direct or qualified animal evidence	sample identity, locality, chronology, and coordinate claims that pass the product contract
SVAR	registered lake	geographic and decision-support input	stable lake identity and hydrographic properties
boundaries	country MultiPolygon	geographic framing	reproducible Nordic and country scope selection

Role And Workflow Are Independent

Evidence role describes what a family may support. Intake workflow describes how its governed material reaches the repository. The two dimensions remain separate:

Family boundary	Intake pattern	Why the distinction matters
seven collector-managed families	release, API, or registry capture into a family-owned tree	executable support is visible through source-support, but presence and fitness require separate state evidence
animal ancient-DNA family	paper, archive, supplement, and sample-owned recovery	no synthetic collection adapter can replace project-by-project evidence lineage

This is why the eight-family portfolio has seven collection adapters. The difference is part of the scientific model, not an incomplete count.

Comparison Requires A Bridge

flowchart LR
    Left["evidence family A"] --> Unit["compatible observation unit?"]
    Right["evidence family B"] --> Unit
    Unit --> Place["compatible spatial precision?"]
    Place --> Time["compatible temporal posture?"]
    Time --> Role["roles support the proposed claim?"]
    Role -->|yes| Compare["qualified comparison"]
    Unit -->|no| Refuse["keep separate or aggregate honestly"]
    Place -->|no| Refuse
    Time -->|no| Refuse
    Role -->|no| Refuse

Co-location is only the beginning of a comparison. A defensible cross-domain claim names the bridge between observation units, respects the weakest spatial precision, uses compatible temporal evidence, and does not promote context or framing into direct support.

Examples:

• a LandClim sequence and an animal sample can be compared by overlapping declared BP intervals, but overlap does not prove causation;
• a RAÄ density cell can describe archaeology registry context near a lake, but it cannot supply an event date or site relationship;
• a boundary can select a record into Sweden, but it cannot increase that record's evidence strength; and
• an SVAR lake can anchor a ranking unit, but nearby evidence remains owned by its original family.

Roles Do Not Escalate Through Combination

Combining several contextual layers does not produce direct evidence. A derived product inherits the role of each contribution and may make only the claim supported by the declared bridge:

flowchart LR
    Direct["direct sample evidence"] --> Product["combined product"]
    Primary["primary pollen context"] --> Product
    Context["archaeology context"] --> Product
    Sampling["lake and fieldwork context"] --> Product
    Framing["boundary framing"] --> Product
    Product --> Reading["partitioned, role-preserving interpretation"]

Combination outcome	Defensible statement	Overclaim
direct sample near pollen site	a governed sample and pollen site meet the declared spatial rule	pollen explains the sample
lake near dense registry cells	the lake lies near recorded archaeology context	the lake has high archaeological potential
several context families agree spatially	multiple contextual layers are present under the same scope	independent confirmation of one historical event
country boundary contains features	those features enter the declared geographic product	the country sample is representative

Independence must also be demonstrated rather than assumed. Two records derived from the same upstream source or one copied citation are not independent corroboration merely because they appear in separate layers.

Audit Independence Before Counting Support

Several visible features can share one underlying observation, source, site, or transformation. Count independent support only after resolving lineage:

Apparent multiplicity	Independence question	Safe treatment
one AADR sample appears in two panel selections	do both rows resolve to the same release-owned sample identity?	count one sample and retain both panel memberships
several animal samples share one site	are specimens independently identified while locality is shared?	count samples for sample questions; count one site for site questions
pollen sequence and site summary coexist	is the summary derived from the sequence population?	retain both roles; do not count as independent sources
RAÄ density cell represents many registry rows	is the rendered cell an aggregate rather than an observation?	use the declared member denominator, not one symbol as one site
narrative, CSV, and GeoJSON repeat one member	do formats share the same product-member identity?	count one governed member across presentations

flowchart TD
    Features["visible or exported rows"] --> Lineage["resolve governing identities and sources"]
    Lineage --> Same{"same observation or derived ancestor?"}
    Same -->|yes| Roles["retain presentations and roles; count once at chosen unit"]
    Same -->|no| Independent{"independence supported for this question?"}
    Independent -->|yes| Count["count separately with declared unit"]
    Independent -->|no| Qualify["group or qualify dependence"]

Independence is scoped to the question. Samples from one site may be distinct specimens but not independent spatial observations. Two families can have different upstream owners yet remain dependent through a shared derived boundary or ranking input.

Aggregate Only After Choosing The Scientific Unit

Aggregation can change the question before a cross-domain bridge is applied. A density cell, site total, species count, and sample count are not merely different display resolutions of the same population.

flowchart LR
    Question["declared comparison question"] --> Unit["choose unit in each family"]
    Unit --> Bridge["evaluate typed relation between units"]
    Bridge --> Eligibility["retain supported and refused relations"]
    Eligibility --> Aggregate["aggregate for the declared output"]

Shortcut	Distortion introduced	Governed alternative
join a sample directly to a RAÄ density cell	an aggregate is treated as an archaeological observation	retain the cell's member population and declare the spatial relation as contextual
sum sample and site counts	physical or analytical units are mixed with spatial units	report each denominator separately or select one unit before counting
aggregate time before checking comparability	broad and precise intervals can appear equivalent	classify and gate member intervals before summarizing
count map symbols as independent evidence	repeated or aggregated presentations inflate support	resolve governing identities, then aggregate once at the chosen unit

When only aggregate source material is available, the result remains an aggregate-context relation. Disaggregation from a density surface or summary count into synthetic member evidence is not permitted.

Null And Absence Semantics

Missing values retain family-specific meaning. A missing SEAD interval means the current capture does not support numeric comparison. An excluded animal point can still have a governed sample identity. A lake with no nearby admitted feature is not evidence that the historical phenomenon was absent. A country filter can remove a feature without changing its source record.

For any absence claim, distinguish:

1. absent upstream;
2. not captured;
3. captured but unresolved;
4. resolved but outside product scope;
5. excluded by an admission rule; and
6. admitted but hidden by a reader-controlled filter.

Reuse Contract

A cross-domain extract carries source-family identity, stable record IDs, observation units, evidence roles, geometry and precision, temporal semantics, product membership, and caveats. Counts remain partitioned by unit and role; they are not summed into a synthetic total of “evidence.”

Any derived row should retain the member identities on both sides of the bridge, the rule and parameters, the governing source versions, the result posture, and the reason for refusal when no result was produced. Otherwise the row can be displayed but cannot be independently re-evaluated.

Governing Authority Survives The Join

The product record carries a small authority ledger for every derived claim:

Ledger field	Purpose
member identities	recover the exact source-native records on every side of the relation
evidence roles	prevent a contextual or framing contribution from being presented as direct support
bridge rule and parameters	state whether the relation is containment, distance, time overlap, identifier linkage, or another declared operation
precision posture	cap the result at the weakest relevant place and time precision
source versions	make the result reproducible against the releases actually reviewed
disposition and reason	distinguish admitted, qualified, excluded, deferred, and refused results

A derived score is not this ledger. Scores can order candidates within an already governed decision surface; they cannot erase missing evidence, manufacture independence, or transfer authority between families.

Continue to the source-family matrix, spatiotemporal posture, and cross-domain evidence matrix for family-specific maturity and comparison limits.