species · agnt.eco

A graph reads what an ecosystem holds dear.

Curated Species nodes from the nine agnt eco agents, ranked by a six-component centrality score derived from the knowledge graph itself — not from how often the curator typed “important”.

→ biosemiotic layer — what each species reads as meaning (sourced)

network · 01

Nine systems, their curated species

Each agent's curated Species_<Agent> set, with the top three by centrality. Click an agent to see the full ranked list and per-species component breakdowns. Or see which species are shared across multiple agents →

Loading…
methodology · 02

How the centrality score is built

Importance in an ecosystem is multi-dimensional. A single number is a compromise — but a careful compromise is more useful than narrative pull alone. Each species gets six component scores, each normalised inside its agent's set, then combined into one 0–100 figure.

25 %
structural degree

Web of connections

A weighted sum of all the species' edges in the graph. Trophic edges (FEEDS, MYCORRHIZAL_WITH, HOSTS) count more than indicator or mention edges. Species that hold the web together rank higher.

20 %
cross-agent recognition

Shared across systems

How many of the nine agents also list this species in their own central set. A species that surfaces in three ecosystems is probably doing real work in each of them.

15 %
trophic position

Where it sits in the food web

Apex predators, keystone mutualists, and ecosystem engineers get a boost. The metric uses the species' trophic_role string + the topology of its in/out edges.

15 %
curation depth

Evidence on the page

Number of verified facts + cited publications + ingested events tied to this species. The more an agent has actually read and recorded about a species, the more central it is to that agent's reasoning.

15 %
event load

Frequency of agency

How often this species shows up in dated events — collapses, migrations, ingestions, deaths, returns. A species that drives the agent's timeline is central by virtue of mattering through time.

10 %
habitat breadth

Reach across habitats

How many distinct sub-habitats / sub-locations of the agent's ecosystem this species occupies. Rewards species that knit the ecosystem's micro-geographies together.

# Final score (0–100) score = 100 × ( 0.25 × structural_degree + 0.20 × cross_agent_recognition + 0.15 × trophic_position + 0.15 × curation_depth + 0.15 × event_load + 0.10 × habitat_breadth )
A score is a compromise. Centrality numbers are useful for navigation and for spotting where an agent's attention is concentrated — but they are not the species' importance to the ecosystem in any final sense. Read the rationale and the dependencies below each species before trusting the ranking.
network analysis · 03

What the network shares

Eleven systems — nine agent-curated ecosystems plus the bulk inventories of Rondane and the Amsterdamse Bos — read against one another as a single list. The visualisations below count how often a scientific name appears in more than one place, and which pairs of systems share the most species. Hover any ribbon, marker or arc to read the actual names.

—
unique species across the network
—
species shared by two or more systems
—
of 55 system pairs share at least one species

Chord — pairwise sharing

Arc thickness = number of species shared between two systems. Hover a ribbon to read them.

Map — top shared species

Each marker is a system; each arc is one of the 15 most-shared species drawn between the systems that hold it.

Shared species — full list

All species recorded in two or more systems, sorted by count. Click a header to sort.
scientific name common family count systems
methods & caveats

Sources. Nine envai agents and Rondane: live read of the Species_<Agent> labels in Neo4j. Amsterdamse Bos: cached GBIF fungi + lichen inventory (lichen_fungi_bos.json), because the Bos Neo4j stack currently holds zero species nodes. Generated —.

Sampling is asymmetric. The envai agents are curated indicator-species lists of roughly 25–44 names each; Rondane and the Amsterdamse Bos are bulk inventories of 713 and 894 names respectively. Co-occurrence counts therefore lean toward pairs that include one of those two bulk nodes — read the chord ring with that in mind.

Amsterdamse Bos is fungi-only here. Non-fungal overlaps with that node will read as zero until a fuller inventory is wired in.

Matching is case-insensitive on scientific_name; subspecies and authorship strings are not normalised, so a small number of genuine overlaps will be missed.