Sentinel-2 COG Downloader/Loader

Search Element 84 Earth Search for Sentinel-2 L2A Cloud-Optimized GeoTIFFs, stitch every tile needed for full bounding-box coverage, keep only the bands you ask for, and return an xarray array. I know this is a bit niche but I've found myself writing this code multiple times over the years so I'm putting it in a package.

Install

pip install sentinel-2-cog-downloader

With uv:

uv add sentinel-2-cog-downloader

Usage

from s2_cog_dl import Downloader, NoViableDataError

downloader = Downloader(
    bbox=[-122.86, 37.61, -122.15, 38.11],  # west, south, east, north
    datetime="2026-01-01",                  # target time; ±14 days by default
    bands=["red", "green", "blue", "nir"],  # or Sentinel-2 aliases like "B04"
    max_cloud_coverage=25,
)

try:
    result = downloader.download()
except NoViableDataError as exc:
    raise

array = result.to_array()   # DataArray with a "band" dimension
dataset = result.to_dataset()

A single date is expanded by search_window_days (default 14). Pass an explicit range to search only that window:

Downloader(bbox=bbox, datetime=["2026-01-01", "2026-01-15"])

Output CRS defaults to the UTM zone of the bbox. Pixel size is metres (resolution=10) and is converted to degrees when crs="EPSG:4326".

Publish (maintainers)

Build and upload with the stdlib-adjacent PyPA tools:

pip install -e ".[publish]"
make publish-test   # TestPyPI
make publish        # PyPI

Or with uv:

uv build
uv publish --publish-url https://test.pypi.org/legacy/
uv publish

twine / uv publish expect TWINE_USERNAME / UV_PUBLISH_TOKEN (or an interactive login) with a PyPI API token.