How to read an Earth Credit assessment

The six dimensions

An assessment scores a property along six dimensions drawn from the UN System of Environmental-Economic Accounting — Ecosystem Accounting (SEEA EA). Each is normalized to [0, 1.2] against a reference condition for that ecosystem type. Above-reference performance is allowed (rare, mostly seen on restoration sites where natural regeneration outpaces the historical baseline).

A1 Physical: Soil, water, and atmospheric structure. Soil organic carbon, moisture regimes, slope, microclimate.
A2 Chemical: Biogeochemistry. pH, cation-exchange capacity, nutrient cycling, contaminant load.
B1 Compositional: Who is here. Species richness, evenness (Pielou's J, richness-gated), endemism, IUCN red-list presence.
B2 Structural: How the system is built. Canopy height and closure, vegetation cover heterogeneity, dead-wood load, vertical structure.
B3 Functional: What the system does. Net primary productivity, water filtration, fire return interval, trophic completeness.
C1 Landscape: Context. Connectivity (effective mesh size), surrounding land use, fragmentation, climate refugia status.

The penalized geometric mean

The six dimension scores combine into a single ECI using a penalized geometric mean:

ECI = GM(D1', …, Dn') × (1 − (σ/μ)²)

Di' is each dimension score floored at 0.05, so a single bad dimension can't collapse the whole score to zero. This prevents a missing-data dimension from acting like a critical failure.
GM is the geometric mean. It rewards balance — a property that's strong in five dimensions and weak in one is penalized harder than a property that's mediocre across the board.
(σ/μ)² is the squared coefficient of variation. It explicitly penalizes imbalance: an evenly-scored property keeps almost all of its GM; a lopsided one loses a chunk.
If a dimension has zero indicators with data, it's excluded from the GM, not scored as zero. We never punish missing measurements; we just stop claiming we measured them.

Conservative ECI and standard error

Each record carries an ECI and an ECI conservative. The conservative figure applies a small haircut (5 / 10 / 15%, depending on monitoring tier) to leave room for measurement uncertainty. seEci is the standard error of the ECI estimate over the sample window. When pricing or trading credits, use the conservative figure — and verify it against seEci.

Threat decomposition

Earth Credits are not just a number on present condition; they account for what would be lost if the property weren't conserved. The threat multiplier is decomposed into three orthogonal pressures, each scored 0–1, each backed by a public source on the record:

T1 Land-use: Annualized rate of conversion in a 50 km buffer (Hansen / GLAD or equivalent). High when the surrounding landscape is being converted faster than baseline.
T2 Development: The Nature Conservancy's Global Human Modification (GHM) score. High where roads, agriculture, and urbanization compress the property's neighborhood.
T3 Rarity: NatureServe S-rank or IUCN-GET rarity. High where the ecosystem type itself is globally rare or imperiled, regardless of the local pressures.

The composite multiplier is bounded at 0.20–1.80 to keep extreme cases from dominating the credit count.

Earth Credits

The credit formula is simple by design:

Credits = Acres × ECI_conservative × Threat_multiplier

Acres is the monitored area, not the total protected area. ECI is the conservative figure. Threat is the bounded composite. There is no curve fitting, no scaling factor, no "calibration" parameter that can be quietly tuned. A property with bigger acreage, higher condition, and higher threat produces more credits — exactly as intended.

Each annual run produces a new vintage. The first batch issued for property P01 is recorded as P01-2026; the next year's run becomes P01-2027. Past vintages are immutable — methodology revisions or new measurements never rewrite a closed batch. The same id always returns the same record, forever.

Where the data comes from

The registry is built on two primary sources — the Calculator for the scientific score, and Landseed sensor hardware for continuous on-site verification. Everything else (live weather, fire watch, atmospheric context shown on each batch page) is a supplement that fills in only until the registry's own primary infrastructure is fully deployed.

1 · The Calculator — per-coordinate scoring

The Landseed Calculator is the EC-M-1.1 scoring engine. Given a latitude/longitude/radius, it fires fourteen parallel fetchers and returns the complete record that drives credit issuance:

Sentinel-2 L2A (10 m NDVI / NDWI / NDMI / EVI), with cloud filtering
ORNL DAAC MODIS time series (NDVI, LAI, FPAR, NPP, LST day/night)
ISRIC SoilGrids v2.0 — SOC, pH, CEC, clay, bulk density at 0–30 cm and 0–200 cm depth profiles
GBIF + iNaturalist (species richness, family/order spread, threatened-species presence, Chao1 completeness)
NASA FIRMS (recent fire detections within the buffer)
NASA POWER + Open-Meteo CAMS (precipitation, temperature, PAR, aridity, air quality, NO₂, SO₂)
Global Forest Watch / Hansen-UMD (tree cover, recent loss within the buffer)
WDPA (protected-area overlap and IUCN category)
RESOLVE Ecoregions 2017 (biome classification, NNH status)
OpenStreetMap Overpass (road density)
GloFAS / ECMWF (river discharge)
ESA WorldCover 2021 (land-cover heuristic)

Every dimension score in the registry is the output of those fetchers run through the EC-M-1.1 scoring engine. The Calculator is the only place credits get computed. Any other "scoring" surface is presentation only. Each batch detail page shows the actual Calculator response under "Calculator output · raw measurements" — the same numbers that fed the dimension scores, presented as the engine emitted them.

2 · Landseed sensor hardware — continuous verification

The Calculator establishes baseline scores from remote sensing. Landseed sensor infrastructure closes the loop with continuous on-site measurement. Two hardware lines feed into each batch:

Earth Pulse Nodes — modular sensor clusters deployed across each property. Soil probes, water sondes, microclimate sensors, particulate counters; data delivered continuously via cellular, satellite, or LoRa depending on site.
Earth Scanner — periodic deployments providing camera traps, acoustic recorders (ARU), LiDAR canopy sweeps, and environmental DNA sampling. Higher-fidelity, lower-frequency than the Pulse Nodes.

Together these produce the live camera feeds, bioacoustic streams, soil-chemistry readings, fire-detection events, and species-detection counters that should appear on each batch page. As of the current vintage these feeds are simulated placeholders across the validation portfolio. As deployment progresses, simulated streams are replaced with live streams property by property; the record id is preserved.

What's currently a supplement

The site climate readout and the 50 km fire watch on each batch page draw from public weather and FIRMS APIs as a bootstrap signal. Those panels are clearly marked live and exist to keep the registry useful while sensor deployment is in progress; they do not contribute to credit issuance and will be replaced by primary Earth Pulse + Scanner streams as those come online.

Reading the registry

Click any row on the registry to open the full record. You'll see the six dimensions as bars, the threat decomposition with each source cited, the species and structural data that fed the dimension scores, the monitoring infrastructure on site, and the financial picture. Every number shown is reproducible from the underlying dataset (linked at the bottom of every record).

Provenance

Every record stamps its methodology version (currently EC-M-1.1) and its vintage (year of assessment). Methodology updates produce a new version string and don't retroactively rewrite existing records. The registry is append-only.

Source code for the scoring engine and the methodology specification lives in the Landseed forge. The JSON dataset behind this registry is publicly downloadable.

What's not here yet

The current entries are simulated — the ten-property pre-minting portfolio used to validate EC-M-1.1 across grassland, old-growth forest, estuary, desert, montane, marine, savanna, mangrove, and high-altitude ecosystems. They are real ecosystems modeled at realistic ranges, but the actual sensors are not yet streaming. As monitoring goes online, live assessments will replace the simulated ones one by one. The id space (P01–P10) will be preserved.

External anchoring of the ledger root to durable public infrastructure begins at the first formal vintage close. The architecture and verification recipe are documented at /anchor/; specific public-chain selections are deliberately deferred. Until external anchoring is in place, the registry's tamper-evidence is the same as any append-only public ledger: the audit trail is the commit history of the source repository, and every record's commitment is independently verifiable against the published ledger root.

A short guide to a record.