Two ways in
Northmere publishes its index as a static STAC catalogue, which any STAC client can crawl without special support from the server, and as one flat Parquet table that is faster for analysis in DuckDB, pandas or R. Both are regenerated at 02:10 UTC each night.
- Catalogue root
- /files/catalog/stac/catalog.json
- Collections table
- /files/catalog/collections.parquet
- Item footprints
- /files/catalog/footprints/<collection>.geoparquet
- Schema version
- STAC 1.1, Northmere extension 1.2
Example: tiles over a bounding box
import pystac
cat = pystac.Catalog.from_file("https://se2.enotcloud.com/files/catalog/stac/catalog.json")
coll = cat.get_child("sat-2031")
items = [i for i in coll.get_items()
if i.datetime.year == 2025 and i.bbox[0] < 25.5 and i.bbox[2] > 24.0]
print(len(items), "tiles,", sum(i.assets["B04"].extra_fields["file:size"] for i in items) / 1e9, "GB")
Each item lists one asset per band. The asset address is a plain HTTPS URL supporting range requests, so you can open a window with GDAL, rasterio or any COG-aware reader without copying the file.
Example: filter the table with DuckDB
SELECT code, title, size_tb, licence
FROM read_parquet('https://se2.enotcloud.com/files/catalog/collections.parquet')
WHERE domain = 'Climate' AND size_tb > 10
ORDER BY size_tb DESC;
Fields in the collections table
| Column | Type | Meaning |
|---|---|---|
| code | string | Short code, for example CLM-1120 |
| doi | string | Persistent identifier of the current version |
| domain | string | Genomics, Climate or Imagery |
| size_tb | double | Stored size in decimal terabytes |
| n_files | int64 | Number of files in the current version |
| licence | string | SPDX-style licence identifier |
| bbox | list of double | West, south, east, north in degrees; null for non-spatial data |
| time_start, time_end | timestamp | Temporal coverage in UTC |
| embargo_until | date | Null once public |
The STAC catalogue is static, so there is no query endpoint to rate-limit. A full crawl is about 180 MB of JSON; the Parquet table is 41 MB and is usually the better choice.