Available Datasets

Datasets are served from the public STAC catalog at:

https://s3-west.nrp-nautilus.io/public-data/stac/catalog.json

The agent discovers dataset paths and schemas dynamically — use the browse_stac_catalog and get_stac_details tools rather than hardcoding any S3 paths.

Current datasets

Dataset	Description
GLWD	Global Lakes and Wetlands Database
Vulnerable Carbon	Conservation International carbon vulnerability data
NCP	Nature Contributions to People biodiversity scores
Countries & Regions	Global administrative boundaries (Overture Maps)
WDPA	World Database on Protected Areas
Ramsar Sites	Wetlands of International Importance
HydroBASINS	Global watershed boundaries (levels 3–6)
iNaturalist	Species occurrence range maps
Corruption Index 2024	Transparency International data

H3 spatial indexing

All datasets are indexed using Uber's H3 hexagonal grid system.

Column	Resolution	Cell area
h8	Resolution 8	~0.74 km²
h4	Resolution 4	~1,771 km²
h0	Resolution 0	~4,357,449 km²

Area calculations

H3 cells are not equal-area — true cell area varies with latitude and icosahedral distortion (res-8 cells range ~0.55–0.82 km²), so a nominal per-resolution constant introduces a systematic error (~6% for California). For a region, feature, or per-group area, sum the exact per-cell area over distinct cells:

-- Exact area in km² (the proper method)
SELECT SUM(h3_cell_area(h8, 'km^2')) AS area_km2
FROM (SELECT DISTINCT h8, h0 FROM read_parquet('s3://...') WHERE ...);

Only for unscoped global aggregates over millions of cells — where materializing every distinct cell would defeat the fast approximate path — fall back to multiplying an approximate count by the nominal constant (APPROX_COUNT_DISTINCT(h8) * 0.737327598, accurate to ~1–2% globally).

See h3-guide.md for the full area guidance.

Cross-dataset joins

Always include h0 in join conditions to enable partition pruning:

SELECT a.h8, a.value, b.other_value
FROM read_parquet('s3://dataset-a/**') a
JOIN read_parquet('s3://dataset-b/**') b
  ON a.h8 = b.h8 AND a.h0 = b.h0   -- h0 required for pruning

Omitting h0 forces a full scan of both datasets and is 5–20× slower.