Most geospatial platforms model the world as features: a point, a line, a polygon, and a flat row of attributes attached to each. That model is fast, familiar, and powerful but it quietly omits the things analysts and stewards care about most: relationships and context. Who built this? What happened here, and when? Which report, photograph, or 3D model documents it? What controlled vocabulary defines its type?

Arches is an open-source data management platform, originally built to manage cultural heritage data, that answers contextual questions by design. Instead of bolting attributes onto geometry, Arches models information as semantic graphs grounded in an ontology such as the CIDOC Conceptual Reference Model (CIDOC CRM). Organizations can author graph schemas that describe geospatial resources (e.g.: physical assets and landscapes) alongside the actors (people and groups), temporal entities (activities and events), digital resources (files, images, 3D models, time series), textual resources (reports, books, grey literature), and authority data (thesauri, controlled vocabularies, gazetteers) they connect to. Arches treats geospatial and non-spatial data as first-class, interlinked entities.

Why build a data management platform this way? Arches was designed to solve a genuinely hard problem: how to describe cultural heritage. Judging whether an object or landscape is culturally significant means modeling uncertainty in time, fuzziness in geospatial extent, deep descriptive detail, and the web of relationships between objects, people, events, and documentation. Those relationships provide the context interpreting potential heritage resources demands. By solving for that complexity, Arches is well suited to managing information for other relationship-dense domains: natural resources, engineered works, and beyond.

This talk makes the case that Arches' semantic, relationship-first approach is genuinely uncommon in the data-management world and shows how it does this without abandoning the GIS stack you already use. We'll see how Arches integrates its semantic graphs with the traditional feature-based representations used by QGIS, PostGIS, and OGC services.

There's a payoff for using Arches. Because Arches describes resources semantically rather than through opaque database column names its data is legible to both machines as well as people. Well-typed descriptions make resources easier to query precisely, easier to integrate across datasets, and far more tractable for LLMs and AI agents, which thrive on the explicit class definitions and properties a semantic graph provides.