Coverage And Naming

Coverage is multidimensional. A source can be geographically broad but temporally sparse, contain many project records but few recovered samples, or support normalization without supporting publication. One count cannot represent all of those conditions.

Coverage Dimensions

Dimension	Question	Example measure
acquisition	Was the intended source material captured?	expected and recovered files, projects, papers, or supplements
extraction	Were usable records recovered from the capture?	extracted samples or sites versus supported expectations
evidence	Are identity, locality, chronology, and coordinates supported?	resolved, qualified, unresolved, and conflicting records
geographic	Which places are represented?	countries, regions, sites, or registered lakes
temporal	Which time spans and precision classes are represented?	numeric intervals, contextual labels, unresolved chronology
publication	Which records satisfy a named product contract?	admitted, qualified, excluded, and deferred members

Coverage claims name both the dimension and denominator. “Twelve projects tracked” is not equivalent to “all samples recovered,” and “four countries published” is not equivalent to balanced source coverage across those countries.

The resulting coverage description is a vector rather than a single score:

(acquired, extracted, evidenced, geolocated, dated, admitted, published)

Each element has its own observation unit and denominator. Keeping those elements separate exposes where apparent abundance becomes an evidence bottleneck—for example, many captured rows but few supported numeric chronologies, or many governed samples but few exact-point admissions.

Coverage Rule

Every coverage statement names five things: the observation unit, governing scope, source version, lifecycle stage, and denominator basis. If any of the five is unknown, report the known counts and the unresolved denominator rather than converting uncertainty into a percentage.

Coverage is evaluated at the boundary claimed. Capture coverage cannot stand in for extraction coverage; extraction cannot stand in for supported localities or chronologies; and product membership cannot stand in for source completeness. Cross-family totals remain separate unless a comparison contract defines a common observation unit and eligible population.

Denominator Contract

Every coverage statement should resolve this tuple:

observation unit + governing scope + source version + lifecycle stage + denominator basis

For example, “admitted sample points from AADR v66 within the Nordic product scope” is interpretable only when the bundle names those members and the denominator explains whether excluded, unresolved, and out-of-scope samples are counted separately. Without that basis, a percentage can appear precise while comparing different populations.

Use explicit states for the denominator:

• known and admitted — satisfies the named product contract;
• known and qualified — visible under a narrower claim;
• known and excluded — evaluated but outside the contract;
• known and unresolved — tracked but missing a required decision or fact;
• not assessed — belongs to the declared population but has not been evaluated;
• outside scope — not part of the denominator for this question.

Unknown is not zero, and absence from a product is not evidence of absence in the source or historical record.

Coverage At A Glance

Statement	Valid denominator	Misleading substitute
source acquisition coverage	declared upstream objects expected for the named release	all objects that might exist upstream
sample recovery coverage	source samples expected from captured material	number of downloaded files
chronology coverage	governed records assessed for chronology	only records already carrying numeric bounds
point-publication coverage	candidates evaluated under the named point contract	every record with a place name
country publication coverage	product members evaluated inside the declared geography	all repository records mentioning the country

The denominator includes known exclusions and unresolved records when they belong to the evaluated population. Reporting only admitted rows turns a fitness result into a circular success rate.

Naming Rules

Names identify ownership and scope; they must not imply evidence that the artifact does not contain. Apply these rules to governed records and published descendants:

• use stable repository identifiers for joins and retain source-native keys;
• include family, geography, species, version, scenario, or evidence posture when it changes the population a reader would infer;
• keep display names and aliases as properties rather than identity;
• reserve lifecycle terms for stages that are materially present;
• treat final as final for one declared contract, never as universal scientific finality;
• name refusals, gaps, conflicts, and exclusions by the decision they record, not as undifferentiated errors.

Lifecycle Names Encode Authority

Name component	Meaning
raw or source capture	acquired material before repository interpretation
normalized	repository-owned representation with source linkage
review, audit, ledger, or queue	evaluation, conflict, exclusion, or unresolved work
candidate	record evaluated for a use but not automatically published
final	admitted input for a declared downstream contract, not universal scientific finality
report, map, bundle, or summary	derived presentation or publication surface

A broad name must not conceal a narrow scope. Country, region, version, species, scenario, and evidence posture belong in the artifact identity when they materially constrain interpretation.

Identity And Display Names

Stable identifiers join evidence; display names help readers. A site, sample, project, source record, and product feature may share words while remaining different objects. Preserve their identifiers independently and record any alias or normalized label as a property rather than using it as the join key.

Coordinates are not names. Two records at the same rounded point need not be the same object, and one record with revised coordinates need not be a new object. Likewise, a species label, country name, or paper title can change without changing the governed identity it describes.

Reading Counts Safely

Before comparing two counts, confirm that they share the same unit, scope, version, and evidence stage. Project counts cannot be compared directly with sample counts; source sites cannot be compared directly with published features; and an excluded record remains evidence even though it is absent from a map.

When a count changes, compare member identifiers first. Then classify each addition, removal, and modification by source, normalization, curation, admission, or scope. Aggregate arithmetic can confirm the result but cannot explain it.

Compare Releases By Membership

A release comparison separates four changes:

Change	Meaning
added identity	a governed object entered the declared population
removed identity	an object left the population, with an exclusion or source reason
changed evidence	the object remained, but an owned fact or qualification changed
changed admission	the evidence remained, but product membership or role changed

This separation prevents a corrected coordinate from appearing as a new sample, a stricter admission rule from appearing as source loss, or a display name change from appearing as scientific turnover. Counts summarize the membership diff; stable identifiers and decision records explain it.