Data System¶
Pollenomics is a multi-domain evidence system with a single publication lineage. Pollen, archaeology, hydrography, boundaries, human ancient DNA, animal ancient DNA, and field observations retain separate scientific roles while contributing to governed geographic publications.
Evidence Lifecycle¶
flowchart LR
Select["select and identify source"] --> Capture["capture versioned material"]
Capture --> Normalize["normalize without strengthening"]
Normalize --> Review["review fitness and uncertainty"]
Review --> Publish["publish admitted evidence"]
Publish --> Audit["audit lineage and limits"]The lifecycle is represented in machine-readable contracts and checked-in artifacts. It supports forward questions—what can be published from this source?—and reverse questions—what evidence supports this visible result?
Current Governed Snapshot¶
The checked-in v66 collection summary identifies seven collector-managed
families: AADR, boundaries, LandClim, Neotoma, RAÄ, SEAD, and SVAR. Animal
ancient DNA is the eighth contracted family, but it follows a literature,
archive-project, supplement, and sample curation path rather than the uniform
collector path.
The Nordic publication illustrates why these counts must remain separate:
| Published surface | Current members | Meaning |
|---|---|---|
| AADR human aDNA | 1,231 samples | release-pinned human sample metadata selected for four Nordic countries |
| LandClim | 492 site sequences | primary pollen and vegetation context |
| Neotoma | 200 sites | palaeoecological site context with explicit time postures |
| SEAD | 2,172 mapped sites | environmental-archaeology context; 23 additional reviewed rows lack country assignment |
| animal aDNA | 2 Nordic localities | admitted animal points in this regional product, not the size of the global animal evidence base |
| fieldwork | 1 documented location | a checked-in observation at Lyngsjön, not a regional sampling census |
| REVEALS | 88 grid cells | modelled pollen context |
| RAÄ | 106 density cells | Sweden-specific archaeology density context |
| boundaries | 4 country polygons | geographic framing only |
The rows are not additive measures of “total evidence.” They have different observation units, denominators, geographic reach, and evidential roles. The bundle manifest records membership; the source and evidence records determine what each member can support.
Snapshot Labels Do Not Collapse State¶
v66 identifies the collector snapshot used by the current source capture and
many retained products. It does not assert that every normalized, reviewed,
and published descendant was rebuilt in one atomic transaction. Three state
coordinates must travel together:
| Coordinate | Establishes | Does not establish |
|---|---|---|
| source snapshot | upstream selection, retrieval context, and captured content | complete normalization or review |
| database revision | the coherent governed objects, claims, and decisions under inspection | membership in every publication |
| product identity | the manifest, scope, members, caveats, and renderings released together | freshness of every upstream family |
flowchart LR
Snapshot["source snapshot"] --> Database["governed database revision"]
Database --> Product["product manifest"]
Snapshot -. "not interchangeable" .-> ProductReproduction therefore names all three when they differ. A retained v66
product remains inspectable even when the current evidence-stage matrix marks
one prerequisite as missing; the correct conclusion is “retained but not
currently rebuildable from a complete chain,” not that the missing stage must
exist because the product does.
Boundary Outputs¶
| Boundary | Input question | Durable output |
|---|---|---|
| selection | Is the source relevant, identifiable, accessible, and licensable for its intended role? | source identity and selection rationale |
| capture | Which exact material was acquired, from where, and when? | raw artifact, retrieval metadata, and content identity |
| normalization | Which stable repository fields preserve the source record? | normalized record plus transformation and null semantics |
| review | What can this record support at its actual precision? | coverage, conflict, fitness, and recovery surfaces |
| publication | Which reviewed records belong to this product and geography? | manifest, layers, tables, contract, and caveats |
| audit | Can a visible or missing member be explained? | traceability, subset validation, exclusions, and refusal |
A successful boundary does not imply success at the next boundary. Capture can complete while extraction remains incomplete; normalization can complete while locality remains unresolved; review can complete with a defensible refusal.
One Evidence Transaction¶
An evidence transaction begins with an identifiable upstream object and ends with either a qualified product member or an explicit non-publication outcome. The intermediate records are not disposable preparation files; together they preserve the scientific reasoning behind the outcome.
| Transaction state | Durable fact | Reader-safe conclusion |
|---|---|---|
| selected | source identity, intended role, access, and license are known | the source is relevant enough to pursue |
| captured | exact material and retrieval context are recorded | this is the material the repository received |
| normalized | source meaning is represented under declared fields | records can be inspected consistently without claiming fitness |
| reviewed | identity, coverage, precision, conflict, and comparability are evaluated | the record has a claim-specific evidence posture |
| admitted | a named product contract accepts or qualifies the record | the record may appear in that product at the declared strength |
| refused or deferred | the failed rule or missing evidence is recorded | non-publication is explainable and revisable |
flowchart LR
Upstream["identifiable upstream object"] --> Capture["captured material"]
Capture --> Record["normalized record"]
Record --> Review{"claim-specific review"}
Review -->|supported| Member["qualified product member"]
Review -->|unsupported| Refusal["reasoned refusal"]
Review -->|evidence unavailable| Deferred["recovery state"]The transaction never infers success from pipeline completion. A technically successful capture can still end in refusal when scientific ownership, precision, or comparability is inadequate.
Transaction Identity Remains Split¶
An evidence transaction is reproducible only when its identities are recorded at their own scopes:
| Identity | Why it is separate |
|---|---|
| capture identity | names the upstream version, acquired bytes, retrieval context, and hash |
| record identity | names the normalized source-native object without implying publication fitness |
| claim identity | names one assertion about the object and its evidence locator |
| decision identity | names the claim, intended use, rule, outcome, and qualification |
| member identity | names the admitted object within a specific product manifest |
The same object can therefore retain an accepted spatial claim, an unresolved
numeric chronology claim, and different membership decisions across world,
regional, and country products. A single status column could not express
that state without discarding which claim and product the status qualifies.
Cardinality Is Scientific Meaning¶
Project PRJEB90141 contributes four recovered goat sample identities in the
current curated species view. They resolve to four localities—Direkli Cave,
Blagotin-Poljna, Sang-e Chakhmaq, and Acemhöyük—and retain four sample-owned
chronology intervals. The project accession connects the records; it does not
collapse them into one project point or one project date.
flowchart TD
Project["PRJEB90141"] --> Direkli["SAMEA4453841<br/>Direkli Cave"]
Project --> Blagotin["SAMEA4588077<br/>Blagotin-Poljna"]
Project --> Sange["SAMEA4591758<br/>Sang-e Chakhmaq"]
Project --> Acem["SAMEA4591761<br/>Acemhöyük"]
Direkli --> DTime["13169–13316 BP"]
Blagotin --> BTime["7840–8030 BP"]
Sange --> STime["7939–8170 BP"]
Acem --> ATime["3833–3984 BP"]This one-to-many relation is not implementation detail. A project-level join would assign the wrong place or time to at least three samples; a locality aggregation that discarded sample membership would make the published point impossible to trace back to the physical or analytical unit.
Evidence Foundation Topics¶
The database foundation is the set of rules that keeps a source fact recoverable while it moves toward a public claim. These topics govern that foundation:
| Topic | Governing question | Continue with |
|---|---|---|
| provenance | Which acquired object, version, native key, and transformation support this record? | Provenance and publication linkage |
| source selection | Why is this source suitable for its declared evidence role, and what would disqualify it? | Source selection and refresh |
| refresh lifecycle | Which stages are materially rebuilt, validated, and reviewed when upstream state changes? | Source selection and refresh |
| coverage | Which observation unit, lifecycle stage, scope, and denominator does a completeness statement describe? | Coverage and naming |
| naming | Which durable identity owns the object, and which labels are only presentation or aliases? | Coverage and naming |
| fact ownership | Which record governs a value repeated across normalized, reviewed, and published surfaces? | Data architecture handbook |
| publication lineage | Which decision and manifest connect governed evidence to a visible member? | Provenance and publication linkage |
These rules keep database preparation visible as evidence work. They prevent a normalized export from being mistaken for a capture, a map coordinate from becoming locality authority, and a successful rebuild from being described as scientific completeness.
System References¶
| Question | Reference |
|---|---|
| How do all evidence families and tracked roots fit together? | Data system overview |
| Which file governs a fact repeated across outputs? | Data architecture handbook |
| How do unlike domains appear together without becoming equivalent? | Pollenomics publication model |
| How does a publication resolve to provenance? | Provenance and publication linkage |
| Why was a source selected and what happens during refresh? | Source selection and refresh |
| How are coverage and durable identifiers represented? | Coverage and naming |
| Which evidence dimensions are strong, contextual, or incomplete? | Cross-domain evidence matrix |
| Why does animal aDNA require project- and sample-level recovery? | Animal ancient-DNA evidence |
| Where do the governing artifacts live? | Data directory layout |
| How are claims admitted, qualified, or refused? | Evidence curation |
| How are conflicts and missing support recovered? | Conflicts and recovery |
Read By Question¶
| Reader question | Begin with | Then inspect |
|---|---|---|
| Why is a marker visible? | product manifest and point traceability | governing evidence row, admission posture, and source locator |
| Why is a known record absent? | exclusions, refusals, and product scope | missing or conflicted identity, place, time, coordinate, or geography evidence |
| What does a count measure? | artifact schema and denominator | source coverage, normalized population, and product membership separately |
| Can two layers be compared? | evidence roles and observation units | spatial precision, temporal posture, scope, and comparison rule |
| Can an export be reused elsewhere? | portable record identifiers and provenance | license, source version, qualifications, and destination assumptions |
Authority Boundaries¶
- Source captures govern acquired identity and provenance.
- Normalized records govern repository-owned representation.
- Review surfaces govern scientific fitness, uncertainty, and refusal.
- Publication manifests govern the selected output and its members.
- A downstream report never becomes the authority for an upstream sample, locality, chronology, coordinate, or source claim.
Cross-Domain Interpretation¶
The system does not reduce all evidence to one measure. A temporal interval, pollen sequence, archaeology record, lake polygon, administrative boundary, and ancient-DNA sample carry different units and uncertainty. Publication preserves those differences through source roles, temporal semantics, coordinate posture, layer labeling, and visible caveats.
For claim-level inspection, continue to Sources, Curation, Evidence, or Publications.