Source Family Matrix¶

The source-family matrix describes what each governed family contributes and where its authority stops. It is a role and lifecycle inventory, not a ranking of scientific importance.

Governed Families¶

Family	Domain	Evidence role	Normalized unit	Characteristic review	Published use
LandClim	pollen context	primary context	site sequence and grid cell	freshness, coverage, temporal posture	world and regional pollen layers
Neotoma	pollen context	primary context	pollen-site record	site-level temporal comparability	world and regional pollen layers
SEAD	archaeology context	contextual domain	environmental-archaeology site	access, temporal, and normalization legibility	archaeology context layers
RAÄ	archaeology context	contextual domain	Swedish registry record and density surface	coverage and spatial interpretation	Sweden-focused archaeology layers
SVAR	hydrography	sampling context	lake, catchment, and water-body record	registry coverage and candidate linkage	Sweden lake products and overlays
boundaries	geography	framing	country and regional polygon	geometry and scope fitness	world, regional, and country selection
AADR	human ancient DNA	direct evidence	release-owned sample metadata	sample locality and chronology posture	human aDNA country and regional layers
animal aDNA	non-human ancient DNA	direct evidence	sample-owned species record	project, paper, supplement, place, time, coordinate, and archive integrity	admitted atlas and country members

The machine-readable authority for these roles and lifecycle roots is data/source_family_contracts.json. Publication maturity is reported separately because a family can have a complete structural contract while remaining scientifically qualified.

Lifecycle Coverage¶

flowchart LR
    Contract["source-family contract"] --> Capture["captured layer"]
    Capture --> Normalized["normalized layer"]
    Normalized --> Review["reviewed layer"]
    Review --> Published["published layer"]
    Contract --> Metrics["coverage metrics"]
    Metrics --> Review

Every family declares captured, normalized, reviewed, and published surfaces. Each stage is proven by the exact materialized artifacts named in the family contract. A publication can remain committed when an earlier current-stage artifact is absent; the matrix preserves both facts rather than inferring the missing authority backward from its descendant.

Checked-In Stage Metrics¶

The current stage matrix reports this materialized state:

Family	Raw	Normalized	Reviewed	Published	Checked-in metric and interpretation
LandClim	present	present	missing	present	492 sequences and 88 model cells; the source-specific review artifact is absent
Neotoma	present	present	present	present	200 normalized points; temporal capability remains member-specific
SEAD	present	present	present	present	2,172 normalized points; review supports a contextual, not uniformly dated, role
RAÄ	present	present	missing	present	761,917 registry and 318,265 heritage records; source-specific review is absent
boundaries	present	present	missing	present	four country geometries frame membership without scientific weight
SVAR	present	missing	missing	present	a 40,565-lake summary and retained products exist without the contracted normalized registry and review
AADR	present	missing	missing	present	three v66 capture files and retained products exist without governed Homo sapiens normalized and review members
animal aDNA	present	present	present	present	10 species, 40 projects, and 894 species-owned sample-foundation rows are materialized

The animal lifecycle count is the population of species-owned foundation rows, not the 868 recovered project sample-master identities or the 234 admitted point-evidence rows. All three quantities are valid only with their governing unit. The corrected metric no longer reports zero merely because it was reading a field name that the foundation summary does not own.

Read The Matrix As An Evaluated Snapshot¶

The stage matrix is recomputed from the artifacts named by the family contracts. It reports what is materially present in the inspected repository state; it is not a history of every successful collection or publication.

That distinction explains an otherwise surprising pattern: a family can show published: present while normalized or reviewed is missing. The product is a retained checked-in descendant, while the prerequisite named by the current contract is absent. The matrix refuses to infer the prerequisite from the descendant.

Matrix pattern	Defensible interpretation	Invalid inference
raw present; normalized missing	capture exists, but the current normalized contract is not materialized	normalization must be valid because raw bytes exist
normalized present; review missing	represented members exist without the required family review surface	every normalized member is publication-ready
published present; prerequisite missing	a retained product exists but full rebuildability is not demonstrated	the prerequisite is implicitly present
all stages present	every required stage has a materialized artifact	every member supports every scientific use

Historical continuity belongs in capture receipts, product manifests, and revision history. Current stage presence belongs here. Mixing those questions would turn retained outputs into evidence for missing preparation work.

Authority Boundaries¶

Family class	May establish	Cannot establish alone
direct evidence	a bounded claim about its governed observation or sample	collection completeness or representativeness
primary context	source-backed environmental setting central to comparison	identity, place, or time for an unrelated sample
contextual domain	surrounding archaeological or environmental interpretation	direct biological association
sampling context	candidate selection and fieldwork reasoning	feasibility, preservation, permits, or scientific outcome
geographic framing	membership in a declared spatial scope	scientific support for a member record

Observation Units And Safe Joins¶

Cross-family analysis is valid only when the join preserves the unit on both sides. Similar labels and nearby coordinates are candidates for review, not keys.

Family	Stable member identity	Safe cross-domain relation	Unsafe shortcut
LandClim	governed site or grid-cell identity	explicit site/grid relation with stated temporal basis	treating a model cell as an observed pollen sequence
Neotoma	database and site/deposit identifiers	retained database identity plus reviewed site relation	joining on site name alone
SEAD	SEAD site identifier	declared distance or containment from its reported point	interpreting proximity as sample association
RAÄ	national registry identity	declared spatial relation within Swedish registry scope	replacing SEAD or specimen provenance with a nearby heritage record
SVAR	registered water-body identity	reviewed candidate-to-lake relation	assuming a same-named water body is the intended sampling basin
boundaries	governed geometry identity and version	point-in-polygon membership	using country membership as coordinate validation
AADR	release-owned sample identifier	explicit reviewed locality or publication relation	matching a human sample by place label alone
animal aDNA	project, accession, sample, and evidence locators	sample-owned locality, chronology, and source relations	collapsing all samples in one project to one place or date

Derived relations retain both member identifiers, the rule that connected them, input versions, and the evidence posture of each input. This allows a distance, containment, or comparison to be recomputed without turning it into source truth.

Product Admission Matrix¶

The same family can be eligible for one product and ineligible for another.

Product use	Minimum source posture	Refusal condition
spatial inventory	stable member identity and supported geometry	unresolved identity or geometry
country membership	supported geometry and governed boundary version	missing geometry or undeclared boundary basis
time-aware comparison	comparable numeric intervals with retained basis	absent, contextual-only, or incomparable chronology
contextual overlay	declared contextual role and reproducible spatial relation	context presented as direct association
specimen claim	sample-owned identity and claim-specific evidence	project-, paper-, or locality-level evidence substituted for the sample
fieldwork prioritization	governed lake identity plus explicit decision inputs and caveats	registry presence or proximity presented as feasibility

Publication is therefore a claim-specific decision, not the final stage of a uniform source pipeline. Eligibility must be recomputed when the source, normalization, boundary, comparison rule, or product contract changes.

Read Across A Family, Then Down A Claim¶

The matrix supports two different readings:

flowchart LR
    Family["one source family"] --> Lifecycle["capture through publication"]
    Claim["one proposed claim"] --> Dimensions["identity, semantics, space, time, and role"]
    Lifecycle --> Fitness["family readiness for this claim"]
    Dimensions --> Fitness

Reading across one family shows which lifecycle evidence is materialized. Reading down one claim compares only the dimensions needed for that claim. Neither reading authorizes a global family ranking. LandClim can be mature for pollen context while AADR has a governed release capture but lacks its contracted normalized and review members; their record counts and roles are not competing measures of quality.

This distinction also prevents lifecycle completion from being mistaken for scientific readiness. A family may have captured, normalized, reviewed, and published artifacts while still carrying a material temporal or geographic qualification.

Maturity Is Multidimensional¶

Maturity cannot be reduced to one color or score. Review source identity, acquisition reproducibility, normalized semantics, spatial support, temporal support, evidence-role clarity, product admission, and visible limits independently.

A family can be:

structurally complete but temporally uneven;
geographically broad but locally weak;
scientifically valuable but context-only;
well captured but not publication-ready;
admitted to one product while excluded from another.

Evaluate Maturity By Claim¶

Use the matrix as a set of independent questions rather than a ladder:

Axis	Question	Failure must remain visible as
identity	Can each governed member be traced to one source-native object and release?	unresolved identity, collision, or missing locator
acquisition	Were expected assets captured under recorded access and use conditions?	blocked asset, partial capture, or unknown denominator
semantics	Does normalization preserve the family's observation unit and field meaning?	unsupported mapping, ambiguous value, or schema drift
space	Is location represented at the precision supported by the source?	approximate, regional, substituted, withheld, or unresolved geometry
time	Is chronology numeric, contextual, broad, absent, or inapplicable?	explicit temporal class and uncertainty
role	Is the family direct evidence, primary context, contextual, sampling context, or framing?	refusal of role substitution
publication	Which named product admits the record and under what qualification?	exclusion, deferral, warning, or empty membership

A family can be strong on one axis and weak on another without contradiction. For example, a stable national registry can have excellent identity and spatial coverage while carrying no uniform chronology suitable for same-period comparison.

flowchart LR
    Family["source family"] --> Identity["identity and acquisition"]
    Family --> Meaning["unit and semantics"]
    Family --> Space["spatial support"]
    Family --> Time["temporal support"]
    Family --> Role["evidence role"]
    Identity --> Fitness["claim-specific fitness"]
    Meaning --> Fitness
    Space --> Fitness
    Time --> Fitness
    Role --> Fitness
    Fitness --> Product["qualified product decision"]

Current Review Surfaces¶

These reports describe the checked-in state. They do not replace the captured and normalized records that govern individual facts.

The revision and state model explains why lifecycle presence, record fitness, and publication membership remain separate database states.