Maps

Maps are scoped views over admitted evidence, contextual layers, and geographic framing. They reveal spatial relationships quickly, while their contracts and traceability surfaces preserve the slower answer to why each feature appears.

Map Assembly

flowchart LR
    Families["governed source families"] --> Eligibility{"scope and product eligibility"}
    Eligibility -->|admit| Layers["versioned GeoJSON layers"]
    Eligibility -->|refuse| Exclusions["exclusion and gap outputs"]
    Boundaries["geographic framing"] --> Layers
    Layers --> Bundle["manifested map bundle"]
    Bundle --> Browser["interactive map"]
    Bundle --> Trace["point traceability and contract"]

The browser is a consumer. Its filters can hide admitted features but cannot admit new ones. Popup text can summarize a record but cannot replace its evidence row.

The map publication contract makes that boundary inspectable. It declares the scope and parent scope, artifact membership, countries, roles, layer rows, legend sections, filter surfaces, bounds, initial view, basemap, and visible caveats. The HTML map renders the contract; it is not the contract itself.

Three States Behind One View

An interactive map combines three states that must remain distinguishable:

State	Owned by	What can change it
evidence state	source and curated evidence authorities	source capture, normalization, or review decision
product state	manifest, admission, layer, and traceability contracts	governed publication regeneration
view state	browser filters, layers, viewport, selection, and basemap	reader interaction

flowchart LR
    Evidence["governed evidence state"] --> Product["admitted product state"]
    Product --> View["current browser view"]
    View -. cannot mutate .-> Product
    Product -. cannot redefine .-> Evidence

A feature hidden by a country or time filter remains a product member. A feature absent from the product may be excluded, outside scope, unresolved, or unrecovered. A changed viewport has no evidentiary meaning. Reproducible map discussion therefore names all three states instead of treating the screenshot as the dataset.

Reading Layers

Layer role	Safe interpretation	Required follow-up for a stronger claim
direct evidence	admitted sample or observation at declared precision	sample, locality, chronology, coordinate, and citation lineage
environmental context	surrounding palaeoenvironmental signal	source-specific temporal and observation-unit semantics
archaeological context	nearby or scoped archaeology records	dating and registry limits; proximity is not ownership
decision support	ranked or scored candidate geometry	ranking model, inputs, and sensitivity output
framing	boundary or viewport	none as scientific evidence; framing carries no evidence weight

Check Compatibility Before Comparing Layers

Dimension	Compatibility question	If the answer is no
observation unit	are both features sites, samples, modeled cells, registry records, or another declared unit?	describe the layers separately
evidence role	are both direct evidence, or is one contextual or framing information?	do not assign equal evidentiary weight
spatial support	do point, polygon, grid, and approximate-location semantics support the proposed relation?	qualify or refuse the spatial comparison
temporal support	do both records have compatible admitted time postures?	restrict the result to spatial context
population	are capture scope, filters, and exclusions known for both layers?	do not interpret density or emptiness as abundance or absence

The browser can make incompatible features look visually commensurate because all geometry shares one viewport. Compatibility is established from the layer contracts and evidence fields, not from symbol size, color, or proximity.

Build A Comparison Receipt

When two map features motivate a scientific question, record the comparison before interpreting the visual pattern:

Receipt member	Feature A	Feature B	Shared rule
product and member identity	stable scoped identity	stable scoped identity	same bundle or declared cross-product relation
observation unit	sample, site, sequence, cell, lake, or event	corresponding typed unit	no implicit conversion between units
evidence role	direct, context, framing, decision support	corresponding role	interpretation respects the weaker role
spatial posture	geometry owner, basis, precision, CRS	corresponding spatial evidence	distance or containment method plus uncertainty
temporal posture	basis, bounds, precision, comparability	corresponding temporal evidence	explicit overlap or non-comparability rule
source coverage	captured and admitted population	corresponding population	absence and density limits retained

flowchart TD
    Pair["two visible features"] --> Units{"units and roles compatible?"}
    Units -->|no| Separate["describe as separate layers"]
    Units -->|yes| Space{"spatial support compatible?"}
    Space -->|no| Separate
    Space -->|yes| Time{"temporal support compatible?"}
    Time -->|no| Context["spatial context only"]
    Time -->|yes| Compare["bounded spatiotemporal comparison"]

The result remains descriptive unless a separately governed analytical contract supports a stronger inference. A complete receipt can establish co-location and compatible time; it does not establish interaction or cause.

Geographic Subsets

World, Europe-plus, Nordic, and country maps form a subset lineage. Narrowing scope must preserve feature identity and meaning. It may apply a stricter product rule, but it must not invent a child feature, strengthen coordinate precision, or reclassify context as direct evidence.

Subset validation answers product lineage, not scientific coverage. A valid Nordic subset can still contain uneven source recovery by country, species, or period. The manifest establishes which governed members were selected; family coverage and exclusion surfaces explain the population from which selection occurred.

Spatial Relationships The Map Does Not Prove

Visual pattern	What is visible	What remains unproven
two nearby points	co-location at the displayed precision	contemporaneity, interaction, or shared cause
dense cluster	many admitted features in the rendered area	historical abundance or uniform sampling effort
point inside a boundary	coordinate falls within current framing geometry	historical political affiliation
lake near archaeology context	proximity under the current map projection	depositional linkage or field suitability
empty area	no visible admitted feature under current filters	absence in the source, past landscape, or recovery queue

Temporal posture, coordinate confidence, source role, and product membership must be compatible before proximity becomes a defensible comparison. The map helps locate candidates for that comparison; it does not perform the evidential join by appearance.

Visual Encoding Is A Transformation

Marker size, color, opacity, clustering, layer order, and basemap style change what attracts attention. They do not change product membership or evidence strength unless the contract explicitly assigns them a governed meaning.

Visual effect	What to verify before interpreting it
larger or darker symbol	the field, scale, units, aggregation rule, and missing-value treatment
cluster count	observation unit, clustering radius, zoom level, and whether members share identity
heat or density surface	source population, cell or kernel method, normalization, and coverage bias
overlapping markers	whether records share coordinates, represent one site, or are merely co-located
layer hidden beneath another	membership in the manifest and current layer order
apparent empty area	active filters, source coverage, exclusions, and geometry support

A basemap supplies orientation, not corroboration. Labels, roads, relief, and modern land use belong to the selected basemap state and must not be read as governed historical or palaeoenvironmental evidence.

Filter Order And Empty Results

Filters operate over admitted product members. A time filter may exclude rows with numeric non-overlap, preserve or separately classify rows with unavailable chronology, and leave non-temporal framing untouched according to the contract. It must not silently interpret missing time as outside the selected period.

When a filtered view is empty, distinguish four cases: no product members in scope, no members matching the declared condition, members without comparable fields, and members hidden by another active control. Those states support different conclusions and should not share one “no data” interpretation.

Auditing A Feature

1. record the map scope, version, layer, and stable feature identifier;
2. locate the feature in the corresponding layer or evidence-row export;
3. inspect the geography's publication contract and point-traceability table;
4. follow source, sample, site, chronology, and coordinate identifiers to the governed data surface;
5. check the exclusion output when an expected feature is absent.

If a visible symbol represents an aggregate, continue from the symbol to its member or denominator surface before making a record-level statement. A cell with 27,450 registry records is one rendered aggregate and 27,450 selected source members; it is neither one archaeological site nor 27,450 independent historical events.

A cluster supports a statement about the visible product, not necessarily sampling intensity or historical abundance. An empty area may mean no admitted record, incomplete source recovery, incompatible temporal support, or true source absence; the map alone cannot choose among those explanations.

Preserve Meaning When Exporting A View

A screenshot preserves pixels but loses membership and provenance. A GeoJSON file preserves features but can still lose the parent scope, evidence roles, filter state, caveats, and exclusion context. A reusable map extract therefore travels with its publication contract, product manifest, selected feature identifiers, traceability, temporal semantics, coordinate-confidence fields, and visible warnings.

Record filters separately from product membership. “Hidden by the reader” and “excluded by the publication contract” are different states and lead to different scientific interpretations.

An export receipt should also record whether filtered-out members had incomparable or missing values. A time-filtered count that silently drops text-only chronology answers “members with comparable numeric time in this window,” not “all evidence in this period.”

For a reproducible view, record the bundle version, scope, active layers, filter values, selected identifiers, basemap, viewport, and export time. Those details reproduce presentation. The product manifest and evidence rows remain necessary to reproduce membership and scientific meaning.

Audit anchors include the world map publication contract, Nordic point traceability, atlas input audit, and animal atlas exclusion report. Candidate admission and refusal are also recorded in data/adna/final/atlas/animal_atlas_candidate_accountability.md. Continue with map inputs, point rules, and filters and popups.