Inspect catalogue decisions and coordinate provenance, then open the supporting records directly in the atlas.
Read targeted source reviews, disputed identifications and publication-based dating corrections. Raw catalogue data remain available.
Applies to catalogue results in the atlas. Date, tier and layer filters still apply; rejected records require diagnostic opt-in.
Investigate the three screenshot cases, the coordinate supply chain and class-specific distances. Unknown accuracy stays unknown.
A geometric discrepancy screen with source audits and an optional uncertainty scenario. It does not establish historical road alignment.
Read the feasibility assessment ↗Publication-checked case studies distinguish road work, imperial address, displacement, and later reuse. CIL concordances are labelled by the edition actually consulted.
Dossier filters apply to the publication-checked objects.
This atlas compares reconstructed routes, catalogue localities and publication-checked object histories. Overlapping layers make questions inspectable; they do not generate archaeological validation.
Preparing the evidence layers…
Counts follow the applied atlas filters. Road context segments are not necessarily contemporary with the inscriptions.
Catalogue dots are source locality anchors with unverified precision. Published findspots retain their source limits; repository pins show custody locations. Pin tiers describe locations separately from object associations.
A nearby point and line may depend on the same publication, or represent different periods and location roles. Inspect chronology, original setting and source independence before treating proximity as evidence.
Itiner-e already records confidence classes and segment bibliographies. The useful test is whether a source supports an alignment, a broad corridor, an inscribed distance, a phase of work, or only an imperial dedication. Stones already used in a route reconstruction are not independent confirmation.
A fixed rupestral milestone can constrain its ancient inscription position when original bedrock, Roman milestone function, and a local road relation are publication supported. T2a identifies primary installations beside Roman road remains; T2b identifies documented local co-occurrence with original installation unproved. Reused curbstones or road-fill fragments do not pass that gate. Support form, location precision, date, and source independence remain separate.
These are editorial review categories, not probabilities. A well-documented installed column can provide stronger local evidence than an incompletely published rock text. An object may have a T1 association while its available map pin is only a T5 site locator. Every pin reports its own location tier; exact pavement and dated construction are separate claims.
Original emplacement, archaeological findspot, secondary-use discovery, reported locality, repository, and replica are independent assertions. These are distinct relations, never automatic fallback coordinates. A reused stone’s discovery place is not necessarily its original roadside position. Unlocated cases remain available in the dossier.
Smith’s Shield, Taucheira FS12 and Čatež are mapped through separately documented official map references, published survey maps and bounded historical topography. Approximate map centres carry dashed editorial search envelopes. These show the precision of this interpretation, not a measured accuracy, probability, or claim that the original stone stood exactly at the pin. Click a pin or its envelope, then Locations, for the method, controls, uncertainty and source links. Later reuse, museum institutions and modern replicas retain separate roles and dates. These derived positions are excluded from point-to-road distance tests.
The alignment pilot measures three unchanged screenshot points against the frozen full road snapshot, separately by source certainty. Its offsets are geometric diagnostics. It identifies who provided, extracted and compiled coordinates, while keeping unknown individual georeferencers and accuracy explicitly unknown. The current points lack verified original-position geometry; no low-alignment statistic or independent road verdict is issued. Named observers of a stone and contributors who digitized road lines are separate from the author of a milestone coordinate. Open Alignment test & coordinate provenance under Review tools to read the audited settings, source authors and a clearly labelled uncertainty scenario.
The Catalogue audit control under Review tools lists targeted decisions and opens their evidence directly in the atlas. Supported falsa, interpolated transmissions and genuine monuments misclassified as milestones are separately explained. Excluded records remain searchable and require diagnostic opt-in for mapping. Amber dots flag cautions or review candidates; a broad range, question mark, starred CIL reference or mention of a copy does not establish a forgery. Contested text dates remain conditional claims, available through uncertainty opt-in, and cannot enter road comparisons. The raw Cramond locality for RIB 2313 is withheld following the published Ingliston correction; no replacement coordinate is invented. An incompatible EDH partner record for Beek is retained only in raw provenance. Whole-catalogue screening and targeted publication checks have different coverage, documented in the downloadable audit.
Catalogue date envelopes can contain several inscriptions. The filter tests publication-reviewed date assertions where supplied, otherwise the parsed source dating components. Both remain visible in the inspector; raw metadata is preserved and gaps remain gaps. These are catalogue assignments, not newly publication-verified dates or construction dates for a whole road. Rivers and sea lanes are kept in a separate optional layer and excluded from road-distance comparisons. Zero placeholders are normalized to unknown; raw source values remain inspectable. One-sided dates stay visibly incomplete. The inscription filter uses separate phases and stated date errors; gaps remain gaps. Object-history mode filters discovery, reuse, movement, and custody events, with modern years available up to 2200. Road geometry then supplies geographical context. A report or publication date does not date a move. Bounded-date overlaps appear by default. Undated, incomplete, invalid and unresolved dating components require explicit opt-in and use hollow symbols; they are not dated matches. Rejected catalogue records require a separate diagnostic opt-in. RIB 2392* (Donnington) is excluded on the basis of RIB’s Falsum classification, while its raw source record is preserved. Changing a year or checkbox leaves a visible draft notice until Apply or Enter validates it; year zero and reversed ranges are rejected. Roads and water routes supply geographical context unless their date-metadata filter is enabled. Object-history mode always leaves routes as context. Selection outlines disappear when the selected record is filtered out. A curated pin covering a linked catalogue dot provides a direct catalogue-record control in its inspector.
A named emperor may be an honorand or authority rather than the physical builder. A curator, governor, community, contractor, or workforce requires specific evidence. New construction, repair, paving, bridge work, stone erection, and honorific rededication remain separate actions. Fecit alone need not prove a new road.
EDCS rows are selected by the explicit miliaria category. Six byte-equivalent duplicate records are collapsed. EDH records enrich existing EDCS IDs through explicit partner identifiers; they are not counted as additional stones. Entity resolution across different catalogue IDs remains open.
All catalogue coordinates are labelled as source locality anchors with unverified precision. Six imported coordinate conflicts or superseded locality anchors are withheld from display and proximity tests, while raw source values remain preserved. No automated geocoding, road snapping, or relocation is applied. R1 identifies a separate class of rock-cut roadworks records; these do not count as T1 milestones and their relocated display positions cannot locate ancient roads. Derived endpoints represent exact shared geometry, not ancient settlements or a separately verified road graph.
For a selected stone, the map measures approximate point-to-polyline distance against the loaded road geometry in a local equirectangular projection, searching within 50 km. It reports metadata date overlap and source-dependence status. This is exploratory: location error, source dependence, chronology, and road displacement can dominate the computed distance. No support score or road assignment is produced.
DARE © Johan Åhlfeldt / University of Gothenburg; legacy and tile-announcement licence statements differ (CC BY-SA 3.0 / CC BY 4.0). See the documented provenance below. Esri attribution is displayed on the map. Physical terrain has native coverage through zoom 8; enlargement adds no detail. The verified service retirement schedule lists Physical for March 2028 and National Geographic / Topographic for December 2029.
Research edition · old local catalogue snapshots · selected publication checks · no comprehensive CIL collation, field verification, or complete modern bibliography claimed. Source and licence details are retained in the source audit. Basemaps require an internet connection; data, library, and plain map are bundled.