The entire Sentinel 2 archive translated into stac-geoparquet enabling scalable bulk access using any tool that understands Parquet. Syncs daily with Earth Search API and uses its COG's as the assets.
Every Sentinel-2 L2A scene that AWS Earth Search indexes, republished as cloud-native GeoParquet you query in place over HTTP. No API and no key stand between a client and the data.
The explorer opens on Collection 1. Its sidebar switch, or
?collection=sentinel-2-l2a,
shows the original index on the same page.
Earth Search, run by Element 84 on the AWS Registry of Open Data, produces the item indexes this catalog republishes, and Copernicus Sentinel-2 is the imagery behind them. This catalog mirrors those indexes without filtering, reclassifying or interpolating anything in them.
No imagery is published here. The Sentinel-2 Cloud-Optimized GeoTIFFs remain in their AWS buckets, and each item row reproduces their URLs verbatim. What publishes here is the item index, one GeoParquet row per scene, with small per-MGRS-tile aggregates that make it practical to plan a query before running it.
Each collection's table:row_count and temporal extent state what it includes
right now, and each year=YYYY/YYYY.json item states its year's. Those fields
are the authority, where this page describes what is mirrored and how to read
it.
Each collection has a README and an agent guide beside it:
sentinel-2-c1-l2a
(agents),
stats-c1 (agents),
sentinel-2-l2a
(agents), and
stats (agents).
A client reads the Parquet over HTTP, with range requests, straight from the bucket:
The parts are sorted by tile, so a tile filter reads a few row groups instead
of the file. To scan many years at once, use the anonymous s3:// door with a
glob, because DuckDB can list a bucket but cannot expand a glob over plain
https://:
The collection READMEs give the full query patterns, including the one that finds cloud-free scenes over a field and the COG URL for each.
Each item comes from the
Earth Search STAC API, fetched and
normalized to one schema per collection. Months the API could not serve during
its outage windows were read instead from the static item JSON in the
sentinel-cogs bucket, which contains the same items. Each archive is one
queryable table, and a row means the same thing in 2019 as it does today.
Column meanings and query patterns are documented per collection, in
sentinel-2-c1-l2a and
sentinel-2-l2a, each with an agent guide beside
it.
Sentinel-2 data and its derivatives fall under the
Copernicus Sentinel Data Terms and Conditions,
which grant free, full and open access for any use (SPDX CC-BY-SA-3.0-IGO).
The citation for the upstream COGs, from their
AWS Registry of Open Data entry:
Sentinel-2 Cloud-Optimized GeoTIFFs, accessed on [DATE] from https://registry.opendata.aws/sentinel-2-l2a-cogs.
Cite this catalog itself as the item index derived from that archive. Contains modified Copernicus Sentinel data.
| Earth Search collection |
stats | The same aggregates over the original index. | browser | derived here |