Working with Sentinel-2
Copy this file and pass it to your AI agent to work with Sentinel-2 data.
Sentinel-2 is a constellation of Copernicus satellites that image the whole Earth every few days at up to 10 m per pixel. Fused can find scenes, build cloud-free composites, compute indices such as NDVI and follow them through time, for any area and date range.
This page is the map. Each goal below links to the example that covers it and to the canvas you can open and run.
What do you want to do?
| Goal | Walkthrough | Canvas |
|---|---|---|
| Map an index, follow it through time on H3 hexagons or build a composite, with GeoTIFF, CSV or Parquet download, or the same from one JSON API | Sentinel-2 Explorer | Sentinel2_Explorer |
| Cloud-free composite for any area, downloadable as a COG | Sentinel-2 Composite | Sentinel2_Composite |
| Monthly composites over a long date range, one machine per month | Monthly Median Composite | Monthly_Median_Composite |
| NDVI time series for farm fields | NDVI Time Series for Farm Fields | temporal_pixel_analysis |
| Utility functions to search scenes, compute indices, run statistics, classify, make timelapses and export | See Utility functions | Sentinel2_Toolkit |
| Combine NDVI with other layers on H3 | See Combine with H3 | H3_Parcel_Suitability |
Where the imagery comes from
Sentinel-2 L2A scenes are public on AWS Open Data and are listed in STAC catalogs. The examples use three ways to find them:
| Catalog | Used by |
|---|---|
Fused's partitioned Sentinel-2 L2A STAC index, s3://fused-asset/stac/sentinel-2-l2a/partitioned_2015_2025/ | Monthly Median Composite |
s2-stac-geoparquet on Source Cooperative | Sentinel-2 Composite |
| Earth Search STAC API | Sentinel2_Toolkit |
- The canvases above are built on
sentinel_utils, so you work with this data yourself through its utility functions.
Scale out with .map()
A long date range is many small jobs. Run one job per month or per few days in parallel with .map(). Monthly Median Composite runs one machine per month, and the Explorer reads its H3 table with one job per 10 days. Sentinel-2 Composite splits a large area into blocks that run in parallel. See Scaling out UDFs for the pattern.
Utility functions
Sentinel2_Toolkit is a library of Sentinel-2 helpers, with one node per module and a set of example nodes you can run:
- Search scenes and read their dates and cloud cover.
- Request bands, mosaics and composites, and compute indices (NDVI, NDWI, MNDWI, NDMI, NBR, NDSI, NDRE, EVI, SAVI and BSI).
- Statistics per interval over a polygon, and a Copernicus DEM and AlphaEarth embeddings.
- Examples for classification, change detection, timelapse and export.
The same code is published as public/sentinel_utils in the Fused UDFs repo, as three modules: su_search, su_read and su_analysis. Open the canvas and read its README for the function list.
Combine with H3
NDVI on a hexagon grid joins to any other H3 dataset on the hex column. H3_Parcel_Suitability joins NDVI, elevation, irrigation and weather onto parcels. To do the same with your own data, follow Working with H3. The Explorer also builds an H3 table of an index per day, which you can download as CSV or Parquet.
Reference
udf.map(): run one job per tile or month in parallel- STAC catalogs: query STAC catalogs with
pystacandodc.stac - Canvas gallery: more example canvases
Contains modified Copernicus Sentinel data.