Reports¶

Reports assemble a scoped narrative, evidence inventory, map contract, traceability, analytical outputs, and scientific caveats. They are publication bundles over governed evidence—not independent databases and not stronger than the records they cite.

How The Report Tree Is Organized¶

docs/report/index.md is the public entry point
docs/report/world/ is the broadest shared answer
docs/report/regions/europe-plus/ and docs/report/regions/nordic/ are intermediate regional views
docs/report/countries/<country-slug>/ holds the narrowest country bundles

Each child scope is a selection from a governed parent. Country bundles cannot introduce records that lack an upstream identity or silently change the meaning of a shared feature.

The report tree has two complementary axes. Geographic bundles answer where does this product apply? Repository-wide reviews answer how strong is the support, and what remains blocked? A country landing page without its repository review companions can describe the visible subset but cannot establish collection completeness.

Bundle Anatomy¶

Component	Responsibility
landing narrative	state scope, principal findings, and interpretation limits
evidence surface	enumerate admitted source-family content
map contract	declare layers, roles, controls, and publication checks
point traceability	connect visible feature identifiers to governed records
candidate ranking	expose score components and ordered candidates
sensitivity analysis	show whether conclusions survive plausible model changes
scientific review	evaluate evidential strength and unresolved risk
warnings and exclusions	account for records that did not publish

Typed Report Members¶

Files in one report directory are not interchangeable representations of one flat table. Each member has a distinct database role:

Member type	Carries	Cannot govern
manifest	product identity, version, scope, parent, and artifact membership	sample locality or chronology facts
evidence table	projected governed rows with stable identifiers and roles	membership outside its named product
traceability table	relations from visible members to evidence authorities	source interpretation beyond the linked claim
exclusion table	known non-members and product-specific reasons	source absence or global invalidity
review surface	scientific qualifications, coverage, and refusal posture	source-native facts
rendering	visual or narrative presentation of selected state	stronger identity, precision, or readiness than its members

The manifest binds these types into one product without merging their authority. A consumer that imports only the evidence table loses the non-member population; a consumer that imports only the map loses both row identity and decision context.

Bundle Consistency¶

The members are several representations of one governed publication state. They should agree on identity, scope, membership, roles, and qualifications:

Cross-member invariant	Failure signal
manifest names every governed bundle member	an unmanifested file or a manifest path that does not resolve
structured exports agree on stable member identity	CSV, JSON, and GeoJSON contain unexplained member differences
narrative counts reconcile to structured populations	prose total has no matching unit, denominator, or scope
map roles agree with evidence roles	contextual or framing layer is described as direct evidence
warnings and exclusions remain reachable	visible claim loses the caveat or non-member explanation that bounds it
child scopes preserve parent meaning	country or regional output changes the identity, precision, or role of a shared member

A bundle can render successfully while violating one of these invariants. Publication review therefore checks agreement across members, not only whether each file is individually valid.

Read Every Count As A Defined Population¶

A report number is interpretable only when its observation unit and eligible population are explicit. Keep these five parts together:

Part	Question
observation unit	are rows projects, samples, sites, source records, map features, or lakes?
numerator	which governed members satisfy the stated condition?
denominator	which members were eligible to satisfy it?
exclusions	which captured or tracked members were refused, unresolved, or outside scope?
version and geography	which publication state and selection produced the count?

This distinction matters throughout the report tree. A tracked-project count cannot be divided into an admitted-point count without a declared relationship; a country member count cannot stand for source prevalence; and an admitted subset cannot establish recovery completeness when the expected population is unknown.

flowchart TB
    World["world bundle"] --> Europe["Europe-plus bundle"]
    Europe --> Nordic["Nordic bundle"]
    Nordic --> Denmark
    Nordic --> Finland
    Nordic --> Norway
    Nordic --> Sweden
    World -. "identity and subset validation" .-> Sweden

Choose By Question¶

Use geographic publication lineage to compare a world, regional, or country member without losing its parent relation.
Use the report portal to select a geography.
Use the world surface for repository-wide posture and the broadest admitted view.
Use Europe-plus or Nordic for shared regional questions and cross-country comparison.
Use a country bundle such as Sweden for national selection, citations, warnings, and local framing.
Use an analytic sidecar such as Sweden lake evidence richness for a derived ranking and its inputs.
Use a sample table such as Sweden animal aDNA samples for the direct published rows rather than a narrative summary.

Read A Bundle In Authority Order¶

flowchart LR
    Landing["landing narrative"] --> Manifest["bundle manifest"]
    Manifest --> Members["structured member files"]
    Members --> Trace["point and row traceability"]
    Trace --> Evidence["governing evidence"]
    Manifest --> Contract["map and product contract"]
    Contract --> Caveats["warnings, exclusions, and reviews"]

The landing page states the question, scope, and principal results.
The bundle manifest establishes product identity, parent scope, version, countries, and artifact membership.
JSON, CSV, and GeoJSON members carry the reusable records.
Traceability surfaces connect visible identifiers to governed evidence.
Contracts define roles, controls, bounds, and comparison semantics.
Reviews and exclusions qualify both visible and absent evidence.

Narrative is the orientation layer, not the final authority for a row-level claim. When two surfaces disagree, the stable member identity, governing evidence, and narrower qualification control; the disagreement itself becomes a publication-integrity finding.

Cite A Reproducible Claim¶

A report citation should identify enough state to recover the statement after the publication changes:

Citation component	Purpose
product and geographic scope	identifies which governed population was published
version or repository revision	fixes the publication state
stable member or aggregate identity	locates the exact observation or defined population
source family and observation unit	prevents unlike rows from being conflated
governing evidence or traceability surface	supports the row-level fact
material caveat or exclusion context	preserves the boundary of the claim

For a count, retain the numerator definition and eligible denominator. For a map observation, retain the product member IDs rather than only coordinates or a screenshot. For a narrative conclusion, retain the analytical or review surface that supplies the reasoning.

Review A Report Change¶

Totals are the last comparison, not the first. Review a regenerated report in causal order:

compare product identity, parent scope, input versions, and contract;
compare artifact membership and stable member identifiers;
classify added and removed members by admission, exclusion, scope, or recovery cause;
compare changed fields by evidence owner, unit, precision, and null state;
compare warnings, refusals, and traceability coverage;
only then reconcile summary counts and narrative claims.

flowchart LR
    Before["prior bundle"] --> Identity["product and member identity diff"]
    After["current bundle"] --> Identity
    Identity --> Cause["admission, scope, evidence, or recovery cause"]
    Cause --> Semantics["field and qualification diff"]
    Semantics --> Totals["count reconciliation"]
    Totals --> Narrative["public claim review"]

A zero count delta does not prove semantic stability: one member can replace another, a coordinate can become approximate, or a chronology can lose numeric comparability. Conversely, a changed total is expected when a declared scope or admission rule changes and every affected identity is accounted for.

Compare Scopes Without Double Counting¶

World, Europe-plus, Nordic, and country products are nested selections, not independent observations. The same feature identifier can legitimately appear in several bundles. Combining those bundles by concatenation would duplicate evidence.

For cross-scope comparison, compare member identifiers against the parent, confirm each child's subset relationship, and then group by evidence role and temporal posture. A country count describes membership in that country product; summing country counts is valid only when the manifests establish disjoint membership for the observation unit being counted.

When scopes overlap, use a set operation over stable member identifiers and report the intersection. Do not infer independence from separate report files or separate map layers.

Review And Refusal Companions¶

Use these beside the main bundle when the question is about strength rather than description:

Absence from a report is not evidence of absence. Consult the animal atlas exclusion report and animal intake recovery review to distinguish a publication refusal from incomplete recovery. When a narrative sentence and its linked evidence disagree, the narrower governed evidence and visible caveat control the interpretation.

See publication projections for the identity, role, precision, state, population, and revision constraints shared by every report bundle.