Publication Limits¶

The checked-in publications are credible, inspectable evidence surfaces, but they do not support every possible pollenomics claim. Limits are separated by cause so missing source material, weak comparability, conservative admission, and absent runtime capability are not mistaken for one generic disclaimer.

Current Release Posture¶

The repository's generated release assessment refuses final-release language. Its current blocking dimensions are:

animal data recovery: sample extraction is not yet sufficiently complete and region-agnostic for the strongest coverage claims;
SEAD treatment: access and temporal comparability remain weaker than the mature context surfaces; and
geographic extensibility: the publication lineage and onboarding contract exist, but broader geographic proof is not yet established.

The animal publication gate passes its current anti-overclaim checks while still prohibiting language that implies unrestricted precision. These two results are compatible: a conservative subset can be safely published before the underlying recovery program is complete.

Release Posture And Member Fitness Have Different Scope¶

A release-level blocker does not invalidate every governed member, and a valid member does not clear a release-level blocker. The decision scopes remain separate:

Decision scope	Question	Example valid outcome
claim	does this evidence support this place, time, identity, or relation?	exact coordinate accepted; numeric chronology refused
member	may this governed object enter this named product?	qualified spatial member admitted
family	is this source family prepared and reviewed for its declared role?	context publication allowed with missing stronger temporal capability
bundle	are manifest, members, review, caveats, and rendering coherent?	bundle closed with explicit exclusions
release	may the repository use final or unrestricted maturity language?	refused while recovery and SEAD treatment remain blocking

flowchart LR
    Claim["claim decisions"] --> Member["member fitness"]
    Member --> Bundle["bundle closure"]
    Family["family readiness"] --> Bundle
    Bundle --> Release["release posture"]
    Member -. "does not clear" .-> Release
    Release -. "does not erase" .-> Member

This is why the 170 sample-backed animal locality features remain usable under their product contracts while the repository refuses final-release language. They represent 331 distinct admitted samples. Reuse cites the narrowest passing decision and carries every broader blocker that limits the proposed interpretation.

Quantified Boundaries¶

The current governed state makes several limits measurable:

Surface	Current signal	Consequence
Animal foundation preparation	1,450 final rows: 557 fully grounded, 335 partially grounded, 11 blocked by metadata, 391 by location detail, and 156 by chronology	preparation depth is measurable, but grounding posture is not point eligibility
Animal sample recovery	1,455 recovered raw rows and 1,450 final rows across a 40-project registry; only four projects have a trustworthy expected count	recovered rows are auditable, but collection completeness is usually unknown
Animal locality	1,368 direct sample-site assignments and 87 unresolved among 1,455 raw review rows	unresolved samples cannot be described as exact sample sites; a place-name assignment alone does not admit map coordinates
Animal publication points	170 sample-backed locality features representing 331 samples; 117 domesticated-core and 53 wild/progenitor-context	the point subset is traceable but not a complete recovery census
Neotoma time	23,281 captured age claims: 11,526 comparable, 10,622 context-only, 294 refused, and 839 unresolved	sample-level temporal admission is separate from site-span context and accepted taxonomic classification
SEAD time	25,109 claims: 14,264 comparable, 10,144 context-only, 60 explicitly refused, and 641 unresolved	use comparable interval overlap as archaeology context; never infer same-period support from ineligible claims
RAÄ scope	Sweden-specific density source	do not generalize its coverage to the Nordic region

These numbers are a snapshot of the governed artifacts, not permanent project constants. The applicable manifests and review surfaces remain authoritative when the data state changes.

The animal counts cannot be placed into one percentage without changing their meaning. The 1,450-row foundation classifies final sample preparation, the 1,455-row raw sample master governs recovery before final admission, and the 170-feature point surface governs locality-level product membership for 331 samples. A publication limit must name which population is blocked and which stronger claim the blocker prevents.

Limits By Dimension¶

Dimension	Supported use	Unsupported inference
Coverage	inspect tracked and admitted records	assume absence means no relevant evidence exists
Locality	distinguish exact, qualified, broad, and blocked place claims	treat a project or region label as a sample site
Chronology	compare records carrying compatible temporal semantics	convert broad or contextual dates into precise numeric overlap
Coordinates	inspect direct or explicitly derived mapping posture	read marker precision as source precision
Cross-domain maps	explore co-located evidence families	infer causal, biological, or temporal association from proximity
Lake ranking	prioritize candidates under declared scoring and sensitivity	replace bathymetry, access, permitting, coring design, or field verification
AADR	use versioned public metadata	infer genotype processing or population-genetic analysis
Country views	inspect filtered descendants of the shared atlas state	treat a country bundle as an independent complete database

How Limits Propagate¶

A derived surface cannot be more certain than the evidence and rules that produced it. Filtering, ranking, and rendering may narrow a question; they do not repair missing recovery, ambiguous identity, broad chronology, or weak coordinate support.

flowchart LR
    Recovery["source recovery"] --> Identity["record identity"]
    Identity --> Qualification["locality and chronology"]
    Qualification --> Admission["publication admission"]
    Admission --> Derived["map, report, or ranking"]
    Derived --> Claim["reader claim"]

At each arrow, the downstream surface inherits the strongest applicable limit from upstream. A precise marker cannot improve an approximate coordinate; a country filter cannot make recovery complete; and a ranking cannot turn contextual proximity into causal evidence.

Claims The Release Does Not Support¶

The current publications must not be used to claim that:

the collection is an exhaustive inventory of pollenomics evidence;
absence from a map establishes biological, archaeological, or historical absence;
all visible points have equivalent locality or chronological precision;
every animal point represents a recovered source-native sample;
proximity between evidence families establishes association, contemporaneity, or causation;
a ranked lake has been field-validated or is ready for sampling; or
passing a publication gate establishes final-release completeness.

These are claim boundaries, not defects hidden behind a generic disclaimer. Each boundary names the additional evidence or validation that a stronger interpretation would require.

Known Not-Materialized Context¶

The Site 1040 context near Wadi Halfa is not published on the animal point surface. Its paper-backed place statement and approximate named-place resolution remain traceable in readiness accounting, where the governing reason is no_admitted_sample_backed_locality_candidate.

Proposed use	Current decision	Evidence needed for stronger use
show dromedary context as a published point	not supported	recover and admit a sample-backed locality candidate under the point contract
count recovered animal samples	exclude this context from the recovered-sample population	recover and resolve the source-native sample row
compare sample chronology numerically	not supported	sample-owned chronology with eligible numeric semantics
claim exact excavation location	not supported	source-supplied or otherwise defensible site-level coordinate provenance
infer an independent biological observation	not supported	final sample identity and evidence separating the observation from project context

This explicit exclusion is not a reason to discard the 170 admitted locality features. It is also not a license to average the missing evidence away. Downstream work must keep the context outside the published feature population until its admission reason changes.

Visible Absence¶

flowchart LR
    Missing["not visible in publication"] --> Cause{"why?"}
    Cause --> NotCaptured["source not captured"]
    Cause --> Deferred["recovery or curation deferred"]
    Cause --> Blocked["known evidence failure"]
    Cause --> Outside["outside selected scope"]
    Cause --> Unsupported["publication type does not support it"]

These causes have different meanings. The exclusion and recovery surfaces are necessary companions to the visible atlas because a clean map alone cannot distinguish them.

Absence is therefore not one claim. “Not captured,” “captured but unresolved,” “reviewed and refused,” “outside geographic scope,” and “unsupported by this publication type” require different language and different next evidence. A responsible downstream analysis should retain the reason rather than replace all five with a single missing-value code.

Absence class	Defensible statement	Evidence needed to go further
not captured	the governed collection has no captured record	source discovery and recovery
captured but unresolved	a candidate record exists but lacks publishable qualification	identity, locality, chronology, or citation resolution
reviewed and refused	the record failed a named publication rule	corrected evidence and a new review decision
outside scope	the record was not selected for this product	inspect the parent or applicable geographic bundle
unsupported role	this publication type cannot answer the question	use a surface with the required evidence role

Map Boundaries¶

Basemap tiles may still depend on external providers even though publication assets and Leaflet code are bundled locally.
RAÄ is a Sweden-specific context source.
Country and regional filters do not repair weak source geography.
Popups summarize governed fields; they are not complete provenance records.
Candidate rankings are sensitive to available layers, scoring assumptions, geographic scope, and missing evidence.

Responsible Use¶

Before relying on a publication:

identify the bundle version and geographic scope;
determine the source family and its evidence role;
inspect point traceability and scientific review for consequential claims;
preserve precision and temporal posture when quoting or transforming data;
check exclusions and recovery gaps before interpreting absence; and
avoid language stronger than the current release posture.

For lake prioritization, treat the ranking as a reproducible screening model, not a fieldwork decision. A high score identifies a candidate supported by the available and weighted layers. It does not establish sediment preservation, basin geometry, access, legal permission, sampling feasibility, or the absence of uncaptured evidence around lower-ranked lakes.

For cross-domain interpretation, report proximity and temporal compatibility as separate findings. Co-location is not contemporaneity; contemporaneity is not causation; contextual archaeology or pollen is not direct evidence about a particular human or animal sample.

Match A Limit To The Intended Reuse¶

Intended reuse	Limit that must travel with it
recovered-sample analysis	project denominator coverage, identity posture, and exclusion of project-context members
point or distance analysis	locality ownership, coordinate basis, precision, withheld geography, and boundary sensitivity
chronology-aware comparison	source wording, numeric eligibility, dating basis, precision, and incomparable records
source-coverage statement	discovery and capture population, access failures, expected denominator, and recovery state
map screenshot or extract	product version, scope, active filters, member identities, roles, caveats, and exclusion context

A generic citation to “the atlas” cannot carry these limits. The reusable unit is the bounded claim plus its evidence and qualification packet.

The generated repository_final_release_refusal and repository_credibility_dashboard record the current machine-derived assessment. They should be read with the specific evidence review relevant to the claim, not as substitutes for it.