Source Comparison¶
Source choice begins with the claim, not with whichever layer is easiest to plot. Families differ in evidence unit, geographic reach, temporal support, and publication role. Those differences remain visible when several families share one atlas.
Choose By Question¶
| Question | Primary family | Evidence unit | Role in a combined product | Principal limit |
|---|---|---|---|---|
| What vegetation or pollen setting surrounds a place? | LandClim | site sequence and REVEALS grid | primary pollen context | does not establish a sample-level claim |
| Which pollen sites support cross-place comparison? | Neotoma | palaeoecological site record | primary pollen context | chronology coverage is not uniform |
| What environmental archaeology context surrounds the evidence? | SEAD | environmental-archaeology site | contextual domain | nearby archaeology is not sample proof |
| What dense Swedish archaeology context is available? | RAÄ | registry record and spatial density | contextual domain | intentionally jurisdiction-specific |
| Which lakes and water bodies frame sampling choices? | SVAR | lake, catchment, and water-body record | sampling context | present-day hydrography is not historical evidence |
| Which country or region contains a feature? | Boundaries | governed polygon | geographic framing | boundaries carry no scientific weight by themselves |
| Which human aDNA samples appear in a versioned release? | AADR | release-owned sample metadata | direct human aDNA | metadata publication is not genotype analysis |
| Which animal samples have defensible source, place, and time evidence? | Animal source intake | project, paper, supplement, sample, and evidence relation | direct animal aDNA | recovery and publication maturity differ by record |
PalaeOpen is an interoperability network rather than a captured evidence family. It can strengthen metadata alignment and collaboration without becoming direct support for a published feature.
Comparison Axes¶
flowchart TB
Claim["proposed claim"] --> Unit["evidence unit"]
Unit --> Role["direct, context, sampling, or framing"]
Role --> Space["spatial meaning and precision"]
Space --> Time["temporal meaning and precision"]
Time --> Reach["geographic and source coverage"]
Reach --> Product["eligible product and caveat"]Two layers are comparable only across dimensions they both actually support. Co-location permits a spatial comparison; it does not establish shared time, causal relation, or equal evidence strength.
Publication Presence Is Not Comparison Readiness¶
A retained publication proves that a product descendant exists. It does not prove that every preparation stage required by a new comparison is present in the current snapshot:
| Family posture in the current lifecycle matrix | Safe starting claim | Additional comparison evidence required |
|---|---|---|
| Neotoma, SEAD, animal aDNA: capture, normalization, review, and publication present | inspect the governed reviewed population | verify member-level role, space, time, denominator, and proposed bridge |
| LandClim, RAÄ, boundaries: review artifact absent | use normalized and published material at its declared role | do not claim source-specific review support; add the review contract required by the comparison |
| SVAR, AADR: normalized and review artifacts absent | inspect retained capture and publication surfaces | do not infer missing authorities backward from product fields |
Comparison readiness is therefore evaluated from material evidence, not from navigation depth or map availability. A source can be valuable as framing or context while remaining unavailable for a stronger cross-domain bridge.
Valid Combinations¶
| Combination | Supports | Does not support automatically |
|---|---|---|
| LandClim + Neotoma | complementary pollen and palaeoenvironmental reading | one universal pollen chronology |
| pollen + SEAD or RAÄ | environmental and archaeology context around a place | direct association between a site and a biological sample |
| pollen + SVAR | lake-centered sampling and landscape interpretation | field suitability, permits, or preserved sediment quality |
| direct aDNA + context layers | interpretation of a governed sample within a wider landscape | transfer of context-layer precision to the sample |
| evidence + boundaries | reproducible geographic selection and display | representativeness within the selected area |
Declare The Join¶
Cross-source records do not share a universal key. The bridge itself is a derived claim and must name the rule that relates the members:
| Bridge | What may be joined | Required qualification |
|---|---|---|
| identity | two records proven to describe the same governed sample, site, project, or source member | aliases and labels are not sufficient keys |
| containment | point or geometry inside a governed boundary snapshot | boundary version and endpoint precision remain visible |
| proximity | two defensible geometries under a declared distance method and band | proximity is not association or contemporaneity |
| temporal overlap | comparable point or interval claims in the same time basis | overlap is compatibility, not causation |
| category match | normalized values under one declared vocabulary | normalization cannot erase source-specific meaning |
| aggregate comparison | explicitly partitioned populations with compatible denominators | totals remain typed by observation unit and scope |
Joining by display name, rounded coordinate, or row position is not a stable comparison. Those values can change without the governed identity changing—or remain unchanged while the scientific ownership differs.
Comparison Verdicts¶
Every comparison should end in one of four explicit postures:
| Verdict | Required condition | Defensible use |
|---|---|---|
| aligned | observation units, roles, space, time, and scope support the declared relation | direct comparison within the recorded precision |
| partially aligned | only named dimensions overlap | qualified comparison of those dimensions only |
| contextual | one family informs interpretation but does not measure the target claim | context, prioritization, or hypothesis formation |
| refused | the proposed bridge requires unsupported identity, precision, time, role, or denominator | preserve the gap and state what evidence is missing |
The verdict applies to the proposed claim, not permanently to a source family. SEAD can be valid contextual archaeology for a lake analysis while being refused as same-period evidence when numeric intervals are absent. Boundaries can be aligned for geographic containment while remaining incapable of supporting biological association.
Population Before Result¶
A comparison declares its population before it calculates matches. This keeps coverage and missingness from disappearing behind a result count.
flowchart LR
Scope["governed source population"] --> Eligible["eligible for declared bridge"]
Eligible --> Tested["members actually evaluated"]
Tested --> Matched["members satisfying the rule"]
Scope --> Excluded["excluded with reason"]
Eligible --> Deferred["not testable with current evidence"]| Denominator | Meaning |
|---|---|
| governed population | all members in the named source release and scope |
| eligible population | members whose observation unit, role, place, and time posture permit the proposed bridge |
| tested population | eligible members for which the operation completed |
| matched population | tested members satisfying the declared rule |
| excluded or deferred population | members outside policy or lacking the evidence needed for evaluation |
Percentages name their denominator. “Five nearby sites” is incomplete when the reader cannot tell whether five of six eligible members or five of six hundred captured members were tested.
Distinguish Absence Before Comparing Counts¶
An empty cell or missing match can encode different evidence states:
| Absence class | Meaning | Treatment in a comparison |
|---|---|---|
| absent upstream | the named release does not contain the object under the searched identity and scope | report source absence only after identity and scope are fixed |
| not captured | the repository has not acquired the necessary upstream member | exclude from tested population and retain recovery requirement |
| rejected during preparation | a source row was observed but could not satisfy parsing or normalization contracts | retain rejected-row reason and source denominator |
| unresolved evidence | the governed object exists but identity, place, time, or relation is insufficient | defer the affected bridge; do not encode as zero |
| outside product scope | the evidence is governed but not a member of the selected product | retain as a non-member, not missing source data |
| valid non-match | eligible members were tested and did not satisfy the declared rule | count in tested population and report the negative result |
Only the last row is a comparison result. The others describe source, preparation, evidence, or product state and must remain visible separately.
Comparison Packet¶
A reusable comparison preserves the inputs and bridge that produced its verdict. At minimum, record:
| Packet member | Why it is necessary |
|---|---|
| family and member identities | prevents display labels from becoming join keys |
| source releases and product scope | fixes the populations being compared |
| observation units and evidence roles | states what each member can establish |
| spatial basis and precision | bounds distance, containment, and co-location claims |
| temporal class, basis, and interval eligibility | separates numeric overlap from contextual time |
| denominator and missingness posture | prevents visible counts from implying completeness |
| comparison rule | makes the spatial, temporal, or categorical bridge explicit |
| verdict and qualifications | preserves aligned, partial, contextual, or refused meaning |
flowchart LR
Members["identified source members"] --> Bridge["declared comparison rule"]
Space["spatial support"] --> Bridge
Time["temporal support"] --> Bridge
Roles["observation units and roles"] --> Bridge
Bridge --> Verdict["qualified verdict"]
Verdict --> Packet["portable comparison packet"]Without the packet, a downstream table can retain the result while losing why the comparison was permitted. That is especially dangerous for derived counts or distance bands, which appear precise even when their members carry unlike coverage or time semantics.
The packet also records the lifecycle authority used for every input. If the comparison depends on a normalized interval or review posture, the named artifact must exist and resolve to the member; a field copied into a retained publication is not a substitute.
Example: Proximity Without Temporal Equivalence¶
Suppose a pollen site, a heritage record, and an ancient-DNA sample fall within the same distance band of a registered lake. The geometry supports a declared proximity calculation. It does not establish that the observations are contemporaneous, that the heritage record relates to the specimen, or that the lake is a feasible coring target.
A defensible result retains each member ID and evidence role, the distance rule, each spatial precision, the available temporal classes, and a contextual verdict. A single combined “evidence count” would discard the differences that make the comparison interpretable.
The non-members matter as well. If 20 sites are eligible for a proximity test, three lack point-quality geometry, and two are outside the selected product, the result population is 15 tested sites—not an unexplained count of 15 nearby features. Eligibility, tested membership, matches, and exclusions are separate denominators.
Reuse Outside Current Products¶
A source family is portable when its identity, acquisition, normalized semantics, evidence role, and review criteria remain meaningful in the new scope. Jurisdiction-specific sources may still be reusable as a pattern, but their records and authority do not travel beyond their governed coverage.
Before reuse, establish:
- the denominator and geographic scope;
- whether the normalized record retains source-native meaning;
- which temporal and spatial comparisons remain valid;
- whether license and retrieval context permit the intended use;
- which new review or publication gate is required.
Reuse must also preserve negative evidence. Exclusions, unresolved identities, missing intervals, and jurisdictional limits cannot be dropped merely because the destination schema has no field for them. A schema that cannot retain the qualification cannot inherit the original claim strength.
Comparison Changes Are Evidence Changes¶
A changed verdict can originate in four places: a source member changed, a normalization rule changed, the bridge or its parameters changed, or the product population changed. A reusable result identifies which cause applies. Recomputing a distance or overlap under new code without retaining the prior rule identity makes source change and method change indistinguishable.
When a verdict strengthens—from refused to contextual, contextual to partially aligned, or partially aligned to aligned—the packet names the newly recovered evidence. When it weakens, the earlier result remains traceable and affected products are reassessed. Neither direction is reduced to a changed count.
The source-family matrix records the repository-wide roles, while shared normalization explains how different families become structurally comparable without becoming scientifically interchangeable.