LGND Embeddings

Name: LGND Embeddings
Creator: Clay

This repository contains Clay embedding datasets created and maintained by lgnd.ai. Licensed under CC BY 4.0.

Product Details

Visibility: Public
Owner: Clay
Created: 1 Jun 2026
Last Updated: 2 Jun 2026

Product Contents

root

rasters

Product Details

Product Contents

LGND Embeddings

This repository contains Clay embedding datasets created and maintained by lgnd.ai. Licensed under CC BY 4.0.

Product Details

Visibility: Public
Owner: Clay
Created: 1 Jun 2026
Last Updated: 2 Jun 2026

Product Contents

root

rasters

collection=sentinel-2-l2a

README.md

2.53 KB

README

Rasters

A STAC GeoParquet catalog of every satellite scene processed to produce the embeddings in this repository, organized by collection. Use this dataset to look up source imagery metadata — scene footprints, acquisition times, and COG asset URLs — and join back to the embeddings via stac_item_id.

File

One STAC GeoParquet file per collection:

Schema (sentinel-2-l2a)

Column	Type	Description
`id`	string	Scene identifier (e.g. `S2B_33TVH_20170930_0_L2A`). Join key to `stac_item_id` in the embeddings datasets.
`collection`	string	STAC collection name (e.g. `sentinel-2-l2a`)
`datetime`	timestamp	Acquisition timestamp
`geometry`	binary (WKB)	Scene footprint polygon in EPSG:4326
`bbox`	struct	Bounding box (`xmin`, `ymin`, `xmax`, `ymax`)
`assets`	struct	COG asset URLs per band (see below)
`gsd`	double	Ground sample distance in meters (10.0)
`file_last_modified`	string	Last modified timestamp of source file

Data Format

STAC GeoParquet 1.1.0 with geo and stac-geoparquet metadata
Hilbert curve spatial ordering
256 MB row groups

Usage

Joining rasters to embeddings

License

Licensed under CC BY 4.0.

README

Rasters

File

One STAC GeoParquet file per collection:

Schema (sentinel-2-l2a)

Column	Type	Description
`id`	string	Scene identifier (e.g. `S2B_33TVH_20170930_0_L2A`). Join key to `stac_item_id` in the embeddings datasets.
`collection`	string	STAC collection name (e.g. `sentinel-2-l2a`)
`datetime`	timestamp	Acquisition timestamp
`geometry`	binary (WKB)	Scene footprint polygon in EPSG:4326
`bbox`	struct	Bounding box (`xmin`, `ymin`, `xmax`, `ymax`)
`assets`	struct	COG asset URLs per band (see below)
`gsd`	double	Ground sample distance in meters (10.0)
`file_last_modified`	string	Last modified timestamp of source file

Data Format

STAC GeoParquet 1.1.0 with geo and stac-geoparquet metadata
Hilbert curve spatial ordering
256 MB row groups

Usage

Joining rasters to embeddings

License

Licensed under CC BY 4.0.

1rasters/
2  collection=sentinel-2-l2a/
3    stac.parquet
4  collection=naip/          (future)
5    stac.parquet

1rasters/
2  collection=sentinel-2-l2a/
3    stac.parquet
4  collection=naip/          (future)
5    stac.parquet

1import duckdb
2
3# Find all scenes over Boulder, CO in 2024
4duckdb.sql("""
5    SELECT id, datetime, assets.red.href, assets.nir.href
6    FROM read_parquet(
7        's3://us-west-2.opendata.source.coop/clay/lgnd-embeddings/rasters/collection=sentinel-2-l2a/stac.parquet',
8        hive_partitioning=true
9    )
10    WHERE bbox.xmin > -105.5 AND bbox.xmax < -105.0
11      AND bbox.ymin > 39.9 AND bbox.ymax < 40.1
12      AND datetime >= '2024-01-01'
13      AND datetime < '2025-01-01'
14    LIMIT 10
15""")

1import duckdb
2
3# Find all scenes over Boulder, CO in 2024
4duckdb.sql("""
5    SELECT id, datetime, assets.red.href, assets.nir.href
6    FROM read_parquet(
7        's3://us-west-2.opendata.source.coop/clay/lgnd-embeddings/rasters/collection=sentinel-2-l2a/stac.parquet',
8        hive_partitioning=true
9    )
10    WHERE bbox.xmin > -105.5 AND bbox.xmax < -105.0
11      AND bbox.ymin > 39.9 AND bbox.ymax < 40.1
12      AND datetime >= '2024-01-01'
13      AND datetime < '2025-01-01'
14    LIMIT 10
15""")

1import duckdb
2
3duckdb.sql("""
4    SELECT e.chips_id, e.embedding, r.assets.red.href
5    FROM read_parquet('s3://...monthly-aggregated/**/*.parquet', hive_partitioning=true) e
6    JOIN read_parquet('s3://...rasters/collection=sentinel-2-l2a/stac.parquet', hive_partitioning=true) r
7      ON e.stac_item_id = r.id
8    WHERE e.year = 2024 AND e.month = 6
9    LIMIT 10
10""")

1import duckdb
2
3duckdb.sql("""
4    SELECT e.chips_id, e.embedding, r.assets.red.href
5    FROM read_parquet('s3://...monthly-aggregated/**/*.parquet', hive_partitioning=true) e
6    JOIN read_parquet('s3://...rasters/collection=sentinel-2-l2a/stac.parquet', hive_partitioning=true) r
7      ON e.stac_item_id = r.id
8    WHERE e.year = 2024 AND e.month = 6
9    LIMIT 10
10""")