SemanticGIS — Consolidated Terminology Glossary

Working reference for internal consistency. Compiled from drafting discussion; verify each entry against the manuscript before treating it as final.

1. Top-Level Terms

Term	Definition	Notes
SemanticGIS	The title of the book and name of the methodology/pedagogy it teaches. Full title: SemanticGIS — Designing and Documenting Geospatial Workflows.	Never the grammatical subject of verbs like yields, produces, is executed — it teaches, structures, governs. Geospatial Workflows do the producing/executing.
GIS	Reserved strictly for the operational technology stack — desktop software (QGIS, ArcGIS) or code ecosystems (Python/R libraries).	Deliberately narrowed from Goodchild/Chrisman’s “GI Science.” Framed as a stipulative convention for this book, not a claim about the wider field’s terminology debate.
Stack	Settled. The dedicated term for the computational/technology layer — QGIS, ArcGIS, a Python ecosystem. Interchangeable with GIS in its technology sense.	Takes over the computational sense entirely, freeing “Framework” (below) to revert to generic use.
Geospatial Workflow	The specific, documented sequence of semantic reasoning by which a spatial question (from a Domain Project) is answered. Comprises the five Vantage Points and the continuous practice of Workflow Stewardship.	This is what SemanticGIS teaches practitioners to design. Stewardship is part of the Workflow, not a parallel track.
Domain Project	The overarching real-world endeavour (e.g., urban planning, environmental risk management) that sets the brief and human intent.	Contains one or more Geospatial Workflows. A single Workflow never spans multiple Projects.
Framework	Settled. Generic term, no longer globally reserved — freed once “Stack” took over the computational sense. Anchors Activity 2’s title (“Establishing the Data Framework”), unchanged from the original draft.	Must read unambiguously within its sentence, but no longer needs one fixed meaning book-wide. Manual pass still worthwhile wherever it appears, confirming no stray computational-sense usage remains that should say “Stack” instead.

2. The Five Vantage Points

Reframed from a numbered/chronological list to five Vantage Points: positions the practitioner occupies and revisits, each disclosing an aspect of the workflow no other vantage can. Scoping is always first; Execution is always last. Movement between the three middle vantage points is free and expected.

#	Vantage Point	What it uniquely discloses
1	Scoping the Workflow	Human intent and political/ethical/operational constraint. Always entered first.
2	Establishing the Data Framework	Ontological fit (or gap) between the ideal and the given. See §3 below for internal phase structure.
3	Authoring the Analytical Schema	The tool-agnostic mathematical/logical structure required to answer the question.
4	Designing Spatial Results Communication	Legibility of the result to the Lifeworld it serves. See §4 below.
5	Executing the Analytical Schema	The operational/computational reality of a specific Stack. Always entered last.

Workflow Stewardship — not a sixth Vantage Point (it has no partial view of its own); the continuous documentation practice that records movement between vantages, including every Backward Glance and Epistemic Choice. Visually: not a node, but the annotation layer on the edges connecting the nodes.

Backward Glance — a named phase within Stewardship: the audit performed at a transition between Vantage Points, checking whether the vantage just departed adequately resolved what the next one requires. Not confined to the immediately preceding vantage — can cascade as far back as the actually unresolved question originates (e.g., a Schema-stage gap tracing all the way back to Scoping).

Epistemic Choice — the recurring hard decision, surfaced whenever the idealised model cannot be matched to existing/affordable data: compromise the model, or absorb the cost of harvesting new data. Itself a piece of Design Rationale.

3. Activity 2 — Establishing the Data Framework

Now five tightly iterative phases (revised from four, then five — Recording added), moving from an idealised conception through formalisation and confrontation with observed reality to digital instantiation:

Phase	Activity	Concept-level term	Definition
One	Identifying the Universe of Discourse (Boole, 1854)	Concept	The unconstrained, ideal ontology the workflow would require, independent of what exists or can be observed.
Two	Auditing Existing Data (retitled — “Sensoric Manifold” removed from this phase; see Phase Four)	Feature / Feature Class (GIS-technical terms, deliberately introduced here where the semantic and technical layers confront each other)	Does the data exist, and how faithfully does it correspond to the ideal ontology? Audit is ontological, not merely one of coverage/resolution — existing (secondary) data carries the categories of whoever collected it; it is already-synthesised, not raw manifold.
Three	Formalising the Domain of Discourse	Cognised Existence (mass noun — denotes a kind, not yet a countable instance; grammatically deliberate)	The Universe of Discourse, bounded. Four commitments (thematic specification promoted to first, absorbing discrete/continuous; topological/spatial-relationship constraints retained as fourth): see below. Functions as a validation gate the Analytical Schema must pass through. Triggers a mandatory Backward Glance to Phase Two, re-validating candidate data against the now-rigorous criteria.
Four (new)	Confronting the Sensoric Manifold (Kant: Mannigfaltigkeit der Anschauung — the raw, unconceptualised given, prior to any synthesis; not hardware sensor data, and not existing/secondary datasets, which are already synthesised)	—	Always performed, even where the expected outcome is simple confirmation. Two functions: (1) Validation — does the formalised Domain of Discourse survive direct field contact? (2) Collection — where existing data is insufficient, the manifold is what new data is harvested from; a structured Realisation is extracted per the Domain of Discourse’s criteria. Data collection methodology proper deferred to Chapter [X]; this phase establishes only its epistemic position.
Five (new)	Recording the Domain of Discourse	Realisation (converges here — see below)	The digital instantiation of the formalised model: (1) Schema — a spatial table whose geometry type, fields, and field types faithfully carry the Domain of Discourse’s five commitments (discrete/continuous, thematic specification, MMU, typing, topology), and whose coordinate reference system is additionally fit for the Analytical Schema’s intended operations — CRS fitness is Recording’s own distinct concern, not a restatement of Phase Three’s typing (geometry type itself is settled at Phase Three, under Geometric Type; Recording only checks CRS); (2) Population — entering re-evaluated Accept/Transform Realisations from Phase Two, and digitising Collection Realisations from Phase Four, into that same schema. Corrects a stray earlier claim (still to check for remaining instances) that “creating database tables” belongs to Executing the Analytical Schema (Activity 5) — it does not; see the GIS-as-record vs. GIS-as-analysis distinction below.

Phase Three’s five commitments (re-revised — discrete/continuous promoted to a standalone first commitment)

Superseding the “four commitments, discrete/continuous embedded in thematic specification” decision below this heading in an earlier session. That version kept discrete/continuous as thematic specification’s “second dimension” — coordinate with categorical boundary conditions, inside the same numbered commitment. Reopened because that framing quietly breaks: MMU (the second commitment) presupposes discreteness and is not askable of a continuous Cognised Existence, and burying the distinction that decides this inside commitment one’s prose hid the dependency rather than stating it. The field/object split is judged too central to spatial thinking generally to be a sub-clause of anything else. Now five commitments, not four — every “four commitments” reference across the manuscript has been updated to “five” (1.2, 2.2’s Explanatory material, the Relational Data Model note, the data-specification note); this glossary is the last to catch up.

Discrete or Continuous (NEW — promoted to first, no longer a dimension of thematic specification) — is the Cognised Existence a bounded, individually identifiable phenomenon, or a sampled instance of an unbounded continuous field (e.g., a temperature reading)? Settled before anything else because every other commitment depends on the answer: a continuous Cognised Existence has no boundary for thematic specification’s categorical dimension to bound, no MMU to fall below (its analogue is a minimum sampling/interpolation density), and no discrete extent for a topological containment test. Forward-referenced in full to Chapter [X].
Thematic specification — an intensional definition (necessary and sufficient conditions) for every Cognised Existence, letting a practitioner decide whether a specific phenomenon in the world is a legitimate instance of that class, given the discrete/continuous character already settled at (1). Categorical boundary conditions only now — e.g., does a houseboat, permanently moored and used for public events, count as a building? If venue is elsewhere defined as depending on building, the ambiguity propagates. Worked example: houseboat / building / venue (see below).
MMU (Minimum Mapping Unit) / temporal resolution — spatial and temporal resolution below which distinctions are not honoured. Branches on (1): MMU for discrete Cognised Existences; a minimum sampling density for continuous ones. Temporal resolution applies to both alike.
Typing, in two parts — (a) NOIR attribute typing (Nominal/Ordinal/Interval/Ratio — confirm this is the intended acronym, not “NORI”); and (b) Geometric Type — Point/Line/Polygon/Field, an explicit extension this manuscript adds to Stevens (1946) since spatial representation was no part of his problem. States which spatial operations are legitimate upon a Cognised Existence, exactly as NOIR states which statistical operations are legitimate upon an attribute (e.g., a road typed as Line supports network analysis but not area calculation; typed as Polygon, the reverse). A continuous Cognised Existence is typed as Field automatically, following directly from (1) rather than as a fresh decision here. Geometric Type is an “inadmissible operation” check, same family as NOIR — distinct from coordinate reference system fitness, which is a “computable but silently wrong” failure and remains Phase Five’s (Recording’s) job, not Phase Three’s.
(Retained, now fifth in sequence but stated as part of the same formalisation) Topological/spatial-relationship constraints — e.g., roads may not intersect lakes except at bridges; administrative boundaries must not gap or overlap.

Terminology used within thematic specification:

Phenomenon / phenomena — the vernacular, Kantian-justified term for “a specific encountered particular” (this houseboat, this temperature reading) — replaces the earlier, imprecise “thing.” Grammar: phenomenon singular, phenomena plural — check consistently. Note: this is a different use of “phenomenon” than the phase-naming option considered and rejected earlier (Concept vs. “Phenomena” as a phase-level pairing) — that use was rejected as backwards in Kantian terms; this use (an individual encountered particular) is correctly Kantian and unrelated to that earlier rejected option. Keep the two uses distinct in your own head when drafting.
“Class” replaced by “Cognised Existence” throughout this passage — mechanical fix, stray leftover from earlier TBox/ABox discussion.

Worked example — houseboat / building / venue: Running example tying Phases One through Four together. A houseboat’s status as building is contested (does the Domain require fixed connection to land?); if venue depends on building, the ambiguity inherits downward. Deliberately chosen over the earlier “bike shed” example for alignment with the manuscript’s existing “venue” running example. Open consideration: whether to resurface this same example concretely at Phase Four (field-testing the houseboat at the jetty) to close the loop across all four phases with one continuous case, rather than fresh examples per phase.

Real-world case study — GeoDanmark building specification: Danish national building-object specification (§4, Representation) used as an authoritative, real-world instance of thematic specification — independently confirms houseboats as a genuine edge case (explicitly included: “husbåd og lignende”), gives real MMU thresholds (25 m² outside, 10 m² inside an Area Polygon, both municipally overridable — a real Epistemic Choice made in production), and is used pedagogically to show that formal precision at the structural level (thresholds to the m²) often coexists with unexamined vernacular terms at the level of meaning — e.g., “permanent” is never defined, despite doing the key classificatory work, and is complicated by the same spec’s separate provision for buildings under construction. The verbs skønnes (“is assessed/judged”) and the phrase “og lignende” (“and similar”) appear repeatedly, each silently delegating judgement back to the practitioner. Ties directly to Epistemic Choice / Backward Glance: this is exactly the kind of undocumented judgement Stewardship should surface, not let pass silently. (Full Danish text and English translation on file from drafting session — translation includes translator’s notes on “Områdepolygon” and “forefindes… ikke” worth checking against full GeoDanmark documentation before quoting in print.)

Realisation — a specific, located, data-backed instance (count noun: “a Realisation,” “these Realisations”). First meaningful appearance at Phase Two (is this feature a valid Realisation of the Phase One Concept?), carrying forward into Phase Three/Four (does it satisfy the formalised Cognised Existence, including its thematic specification?). Not used in Phase One — nothing exists yet to realise. Existing-data Realisations (Phase Two) and manifold-derived Realisations (Phase Four) are two different provenances of the same concept, and converge at Phase Five: both are entered into the same physical schema, and the finished spatial table should not itself reveal which provenance a given row came from.

GIS as instrument of record vs. instrument of analysis — resolves an apparent conflict between Phase Five and Activity 5 (Executing the Analytical Schema), both of which involve GIS software. Phase Five uses GIS to build and populate a spatial table — an instrument of record, committing nothing about how the data will later be analysed. Activity 5 uses GIS to execute the Analytical Schema’s mathematical verbs — an instrument of analysis. “GIS enters the workflow only at Activity 5” (Part 1, §1.2) refers to this second sense specifically, not to any use of GIS software whatsoever; both places in the manuscript now say so explicitly.

Open item: Confirm whether your physical copy of Boole is genuinely dated 2005, or if the standard 1958 (Dover) / 2009 (CUP) reprint year applies — verify against your copy before citing.

Open item: A mechanical search-and-check pass is now owed on “thing” and “class” elsewhere in the manuscript (Phase Two, Phase Four, Epistemic Choice paragraph, and beyond) — same treatment as the “framework”/“SemanticGIS” consistency passes.

4. Activity 3 — Authoring the Analytical Schema

Added this session: the Schema’s only admissible input is the Recorded Domain of Discourse (Phase Five Realisations) — neither Phase One Concepts nor the raw Phase Four Sensoric Manifold qualify, since neither has been typed/geometried/CRS-fitted yet. This makes Authoring strictly downstream of both of Establishing the Data Framework’s gates (the five commitments, and Recording’s CRS-fitness check).

Analytical Schema as a Directed Acyclic Graph (DAG) — nodes are individual spatial operations (verbs: buffer, intersect, spatial join, reclassify); edges are Realisations flowing from one operation’s output into the next operation’s input. Directed: data flows one way, Recorded input → final result, never backward into a completed operation. Acyclic: no operation may depend, even indirectly, on its own output — this is what guarantees the Schema terminates as a finite, well-ordered sequence rather than an undefined circular dependency.

Intermediate/derived Realisations — an operation’s output is not a return to the Domain of Discourse; it is a new Realisation, often of a Cognised Existence the Domain of Discourse never needed to define because it exists only as an artefact internal to the analysis (e.g., a distance-to-nearest-venue value). The Schema must stay as explicit about these intermediate Realisations’ thematic meaning and typing as the Domain of Discourse was about Recorded ones — an operation three nodes downstream can fail silently on a poorly-typed intermediate output exactly as the first operation could on poorly-Recorded input.

Tool-agnosticism secured by the DAG framing — the graph specifies which verb transforms which input into which output, and in what order, but not which software executes it. The same DAG can be translated into more than one tool-bound Analytical Recipe (see §6 below) — translation, not authoring, is where a specific Stack enters.

5. Activity 4 — Designing Spatial Results Communication

“Design” — used throughout in the specific sense of suitability (fitness of a representational choice to its data and its audience), not layout or aesthetic composition. Worth one explicit sentence stating this the first time the activity is introduced.

Two distinct suitability judgements:

What kind of deliverable — governed by MacEachren’s cartography cube (MacEachren and Taylor, 1994, Visualization in Modern Cartography, Pergamon), three axes: public↔private audience, presenting knowns↔revealing unknowns, low↔high interaction.
How the deliverable represents the data — e.g., classification method (equal interval vs. equal count/quantile vs. natural breaks/Jenks), each suitable to different distributions and audiences, none inherently “correct.”

Open item: exact chapter/page citation for the cartography cube within the 1994 volume — you have the physical book; locate the specific chapter and page range before finalising the citation (do not use the placeholder chapter title suggested during drafting without verifying it against the actual table of contents).

6. Activity 5 — Executing the Analytical Schema

Restates the GIS = Stack convention explicitly at the point it first becomes operationally relevant.

Analytical Schema → Analytical Recipe — the Schema (Activity 3) is tool-agnostic; the Recipe (a Semantic Asset) is the tool-bound, executed translation of it (e.g., a QGIS tool chain or a Python script). The same Schema can yield multiple valid Recipes. Translation is not free — software-specific defaults (interpolation methods, topology tolerances, CRS handling) can reintroduce undocumented decisions, which should be returned to Stewardship as a Design Rationale addendum rather than silently absorbed.

X-Y Problem — the informal term (not “overfitting,” not Dennett’s “competence without comprehension” — both considered and rejected as imprecise fits) for a well-formed-looking request that answers the stated question (X) while the real prior question (Y) goes unasked. Used in the Introduction’s AI-tutor passage: a student may receive a correct procedural answer (e.g., how to buffer) without realising the real prior question was different (e.g., how to reproject first). The comprehension gap sits with the human framing the prompt, not the AI.

7. Semantic Assets (output of the whole Workflow)

Design Rationale — permanent record of every “wicked problem” and how it was resolved. The artifact produced by Stewardship (Stewardship = the ongoing verb; Design Rationale = the resulting noun).
[Formalised] Data Model / Domain of Discourse — the text-indexed bridge between raw data and human intent. (Confirm final preferred term here now that “Domain of Discourse” has superseded “Data Model” as the Phase Three output — check no stray “Data Model” references remain.)
Analytical Recipe — the decoupled, reproducible, tool-bound script (DAG) — see §5.

8. Verification / Citation Checklist

Citation	Status
Tomlinson (1969), Journal of Geography (Chigaku Zasshi), 78(1), 45–48	✅ Verified directly via J-STAGE.
Tomlinson (1968), in Stewart (ed.), Land Evaluation — CSIRO Symposium, Macmillan of Australia, pp. 200–210	✅ Verified via the Chigaku Zasshi reprint’s own footnote (secondary-source chain — legitimate, not directly inspected). Recommended citation form: cite 1968 original as primary, note reprint in brackets.
Boole, G. (1854), An Investigation of the Laws of Thought	✅ Concept and near-exact phrasing verified against primary source. Confirm your copy’s actual reprint year (2005 vs. 1958/2009).
MacEachren, A.M. and Taylor, D.R.F. (eds.) (1994), Visualization in Modern Cartography, Pergamon	⚠️ Volume-level attribution verified via multiple secondary sources. Chapter/page-level detail not yet verified — locate in your physical copy.
NOIR vs. NORI (Stevens, 1946 levels of measurement)	⚠️ Confirm intended acronym — standard term is NOIR. Note: manuscript now extends Stevens with a companion Geometric Type dimension (Point/Line/Polygon/Field) not present in the original 1946 framework — flag this extension explicitly wherever Stevens is cited, so it reads as a stated addition rather than a misattribution.
“Den Erkendbare Eksistens”	Noted as Danish rather than German orthography — confirm intended source language before citing linguistically in-text.
GeoDanmark building specification (§4, Representation)	Danish source text on file from drafting session, with English working translation. ⚠️ Confirm precise defined meaning of “Områdepolygon” and “forefindes… ikke” against full GeoDanmark documentation before quoting the translation in print; cite the specific document/version and access date.

Compiled as a working document — treat as a living checklist to run against the full draft once assembled, not a final, print-ready appendix.

9. Task, Operator, Operation (added this session — not yet reconciled against §§2–3’s older Vantage-Point/five-commitment language above; this section reflects current thinking, the rest of the document does not)

Task — a tool-agnostic, reusable unit of work smaller than an Activity/Phase and possibly smaller than a single Analytical Schema node (e.g., georeferencing, creating a table, correcting a hydrologically-sound DTM, print layout design). A Task decomposes into:

one or more human decision points — judgement calls that are not themselves mechanically executable (e.g., deciding which points on an image are usable control points; deciding column names and data types for a new table); and
one or more Operations, each carried out by an Operator.

Operation — the formal, executable step (unchanged from its existing use at §4/DAG-node level: the verb — buffer, intersect, CREATE TABLE ..., apply-transformation). Operator — the agent that carries out an Operation (a person, a tool, a function). Kept as two distinct terms rather than synonyms: “Operation” names the thing done, “Operator” names who or what does it. A Task’s write-up should keep the decision points and the Operations distinct — conflating them (e.g., writing “create the table” as if it were a single atomic step) hides exactly the judgement Stewardship is meant to surface.

Task composition is a graph, not a hierarchy. A larger Task (e.g., building a hydrological model) can be composed of several smaller Tasks and/or Operations (e.g., correct the DTM → flow direction → flow accumulation), but a smaller Task is not “owned” by any one larger Task — the same correct the DTM Task can be referenced from several different larger Tasks (a hydrological model, a viewshed workflow, a landslide-susceptibility workflow) without duplicating its definition. This mirrors the Analytical Schema’s own DAG structure, and mirrors the reuse pattern wanted for the teaching site (one node, linked into multiple context views — see the Canvas/Kumu discussion).

Task write-up template — mirrors Schema/Recipe:

Task Schema (tool-agnostic): the decision points + the Operations, in the abstract, with no reference to any specific Stack.
Task Recipe(s) (tool-bound, one per Stack): “This Task is implemented in QGIS by …”, “in ArcGIS by …“. Multiple valid Recipes can exist per Task Schema, exactly as for the Analytical Schema/Recipe pair at §6.

Open item: decide whether the book represents Tasks as a numbered glossary/appendix or as an actual linked graph (network diagram / Canvas), given the graph (not tree) composition structure above.

SemanticGIS

Explorer

concepts