Semantic Protocol for Human and Earth Relational Environments
SPHERE is the organising protocol behind this site. It structures how heterogeneous datasets—registries that serve many domains at once—are decomposed into discoverable, purpose-specific units that both humans and AI agents can navigate.
The Problem
A registry like CVR contains 27 object classes and 361 attributes. It serves questions about industry classification, spatial footprint, ownership structure, financial health, and more. No single description captures all of these purposes. An agent asking “where are Copenhagen’s restaurants?” needs different metadata than one asking “who owns this company?”—yet both draw from the same registry.
Flat dataset catalogs force a choice: either one monolithic description per registry (too broad) or one page per entity (too low-level). Neither matches how questions are actually asked.
A Network, Not a Tree
SPHERE was first drawn as a tree crown: spheres as branches, twigs within them, and leaves hanging from one twig each. In practice the data is a network. A Natura 2000 designation is as much a planning instrument as an ecosystem; a drinking water protection zone is planning, groundwater, and resource use at once. Forcing each leaf onto a single branch hid exactly the connections that matter. The names of the elements are kept, but their relations are those of a network:
| Element | Role in the network | Example |
|---|---|---|
| Sphere | A thematic hub—a cluster of related phenomena | Anthroposphere, Hydrosphere, Biosphere |
| Twig | A narrower hub within a sphere | Planning, Groundwater, Ecosystems |
| Leaf | A node: a semantic aspect of a dataset that answers one kind of question | ”Drinking Water Protection” |
| Membership | An edge from a leaf to every twig it is relevant to | Drinking Water Protection → Planning, Groundwater, Resource Utilisation |
| Thread | A membership that carries an explicit rationale | Industry codes → Biosphere Ecosystems, because they signal environmental pressure |
A leaf may name one membership as its primary lens, the preferred entry point for navigation. Leaves that are cross-domain by nature have no primary lens, and no membership is second-class: a twig page lists every leaf that is a member.
The protocol has 5 thematic spheres—Atmosphere, Biosphere, Hydrosphere (including the cryosphere), Geosphere (rock and soil, formerly Lithosphere and Pedosphere), and Anthroposphere (formerly Socio-Technical)—plus the non-thematic Reference Framework (coordinate frames, and the reference units such as municipalities, NUTS regions and addresses that other data refers to by ID). The spheres originate in the geoilum analysis project, where they balanced variables so no domain was over-represented; they were reorganised in 2026 for data organisation.
Two axes run across the spheres:
- What a leaf is about: its twig memberships.
- How the data was produced: the
collection_methodof each realisation (register, in-situ, passive or active remote sensing, derived/modelled, volunteered, cartographic). Remote-sensed data therefore enters the network through its analytical focus, not through a “remote sensing” sphere.
How it works
- Spheres and twigs are the hubs (→ SPHERE)
- Leaves decompose registries into purpose-specific semantic aspects and link to every relevant twig (→ Leaves)
- Each leaf points down to Realisations, which map it onto Dataset documentation (→ Datasets by Collection) and Services (GraphQL, WFS, file download)
- Threads explain the memberships that are not self-evident, enabling lateral discovery
The Architecture Chain
Leaf (semantic aspect, answers a question type)
→ Realisation (implementation mapping, join logic, dataset choice)
→ Dataset (registry info, license, temporal model)
→ Service (GraphQL endpoint, WFS, file download, API key)
An agent navigating SPHERE:
- Identifies the sphere or twig matching its question domain
- Browses the leaves connected to that twig—including those whose primary lens is elsewhere
- Identifies the realisation that best fits its technical or authoritative requirements
- Follows the realisation’s mapping and join logic to the dataset documentation
- Uses service references to access the actual data
Leaf Schema
Every leaf page uses this frontmatter structure:
---
title: <human-readable leaf name>
type: leaf
draft: false
twig_membership:
- <every twig this leaf is relevant to>
primary_lens: <optional; one of twig_membership>
concept: <semantic concept from the collection's semantic organisation>
question: <what kind of question does this leaf help answer?>
threads:
- <thread ids: memberships that carry an explicit rationale>
tags:
- collection/<collection-slug>
---Memberships and Threads: Cross-Domain Connections
Many phenomena sit at the boundary between spheres. Industry codes in CVR classify economic activity (Anthroposphere) but also signal environmental impact (Biosphere) and thematic human partitioning of space (Anthroposphere Perception & Thematics). Rather than duplicating leaves, a leaf simply has a membership in every twig it is relevant to. Where the reason for a membership is not obvious, a thread records it.
An agent exploring the Biosphere finds CVR industry classification listed on the Ecosystems twig, as a member like any other.
For the conventions on memberships, primary lenses, and threads, see SPHERE Thread Conventions.
Backward Compatibility
The existing 338+ pages in Datasets by Collection and SPHERE remain unchanged. They form the Dataset tier of the architecture chain. Leaves add a new semantic layer on top, linking down to those pages. Nothing is removed or reorganised.
Classical Metadata Standards
Three widely-used geospatial classification systems — ISO 19115, INSPIRE, and UN-GGIM — define flat lists of topic categories. SPHERE maps all 67 classical themes into the network, providing a Rosetta Stone between legacy catalog keywords and the sphere/twig/thread structure. This enables systematic harvesting from existing metadata servers.
See Classical Classifications → SPHERE Mapping for the full concordance.
Navigation
| Starting point | Path |
|---|---|
| I have a question about a domain | SPHERE → find sphere → twig → browse member leaves |
| I know the registry | Datasets by Collection → registry page → follow leaf backlinks |
| I want to browse all semantic aspects | All Leaves |
| I have a classical metadata keyword (ISO/INSPIRE/UN-GGIM) | Classical Classifications → find theme → follow to sphere |
| I want the protocol definition | You are here |
Status
SPHERE is in pilot phase. The first leaves cover CVR as a proof of concept, decomposing it into four semantic aspects: industry classification, spatial footprint, ownership structure, and financial profile.
Machine-Readable Discovery
For programmatic agent access, the full sphere/twig/leaf/thread network is available as a JSON index:
- Sphere index: sphere-index.v1.json
This file contains all spheres, leaves with their entities and key attributes, service pointers, and thread cross-references. Agents can fetch this single file to discover everything available.