Migrating from Google Earth Engine to `earthlens.stac`#

Most workloads that load imagery from a single Earth Engine asset id (ee.ImageCollection("…")) map directly to one of earthlens.stac's eight curated endpoints. The two backends speak different APIs — Earth Engine is a server-side processing language; earthlens.stac is a STAC-API client that hands you Cloud-Optimized GeoTIFFs — but the catalogue overlap is broad enough that the top ten GEE asset ids each have an earthlens.stac equivalent. This page lists them.

When to use which backend #

You want pixels on disk over your AOI/window → earthlens.stac. One COG per (collection, date), anonymous reads on most endpoints (Earth Search, Digital Earth Africa / Australia, NASA VEDA, Brazil Data Cube, Microsoft Planetary Computer), requester-pays on usgs-landsat, S3 keys on cdse.
You want a server-side reduction (e.g. composites, mosaics, indices computed on EE's cluster) → earthlens.openeo (the openEO process-graph backend). See Dynamic World below.
You want GEE-only assets (e.g. Google's proprietary atmospheric correction, internal LST products) → keep earthlens.gee; nothing in earthlens.stac substitutes for it.

Top-10 GEE → `earthlens.stac` mappings #

GEE asset id	`earthlens.stac` endpoint / collection	Notes
`COPERNICUS/S2_SR_HARMONIZED`	`planetary-computer / sentinel-2-l2a` (or `earth-search / sentinel-2-l2a`)	Same MGRS scenes; Earth Search is fully anonymous, MPC mints a SAS token.
`LANDSAT/LC08/C02/T1_L2`	`usgs-landsat / landsat-c2l2-sr`	Authoritative USGS source; requester-pays on `s3://usgs-landsat`. MPC's `landsat-c2-l2` is a mirror.
`LANDSAT/LC09/C02/T1_L2`	`usgs-landsat / landsat-c2l2-sr`	LC08 and LC09 share the SR product; filter by `platform` in the search.
`NASA/HLS/HLSL30/v002`	`planetary-computer / hls2-l30`	Harmonised Landsat (30 m). MPC handles the EDL auth so you don't need an Earthdata account.
`NASA/HLS/HLSS30/v002`	`planetary-computer / hls2-s30`	Harmonised Sentinel-2 (30 m).
`MODIS/061/MOD13Q1`	`bdc / mod13q1-6.1`	16-day NDVI/EVI; BDC reads anonymously today.
`MODIS/061/MYD13Q1`	`bdc / myd13q1-6.1`	16-day NDVI/EVI (Aqua).
`COPERNICUS/DEM/GLO30`	`planetary-computer / cop-dem-glo-30` (or `earth-search / cop-dem-glo-30`)	30 m global DEM, anonymous.
`ESA/WorldCover/v200`	`planetary-computer / esa-worldcover`	2021 global 10 m land cover.
`USGS/3DEP/10m`	`planetary-computer / 3dep-seamless`	USGS 10 m DEM, US-only coverage.

For DEM specifically, earth-search/cop-dem-glo-90 and deafrica/dem_cop_30 / dem_cop_90 are also valid anonymous mirrors with different default extents.

Runnable snippets #

Each snippet replaces the equivalent ee.ImageCollection(...).filterBounds(...).filterDate(...) .first() recipe in Earth Engine and writes a Cloud-Optimized GeoTIFF.

`COPERNICUS/S2_SR_HARMONIZED` → MPC sentinel-2-l2a (RGB)#

from earthlens import EarthLens

paths = EarthLens(
    data_source="planetary-computer",
    variables={"sentinel-2-l2a": ["B04", "B03", "B02"]},
    lat_lim=[40.40, 40.45], lon_lim=[-3.72, -3.67],
    start="2024-06-01", end="2024-06-20",
    path="./out", max_items=1,
).download()

`LANDSAT/LC08/C02/T1_L2` → USGS LandsatLook (requester-pays)#

from earthlens import EarthLens

paths = EarthLens(
    data_source="usgs-landsat",        # alias 'landsat' works too
    variables={"usgs-landsat/landsat-c2l2-sr": ["red", "green", "blue", "nir08"]},
    lat_lim=[37.5, 38.0], lon_lim=[-122.5, -122.0],
    start="2024-06-01", end="2024-08-31",
    path="./out", max_items=1,
).download()

USGS LandsatLook reads from s3://usgs-landsat (requester-pays) — make sure AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY are set.

`NASA/HLS/HLSL30/v002` → MPC hls2-l30 #

paths = EarthLens(
    data_source="planetary-computer",
    variables={"hls2-l30": ["B04", "B05"]},   # red + nir for NDVI
    lat_lim=[40.40, 40.45], lon_lim=[-3.72, -3.67],
    start="2024-06-01", end="2024-08-31",
    path="./out", max_items=1,
).download()

`MODIS/061/MOD13Q1` → BDC mod13q1-6.1 (16-day NDVI, anonymous)#

paths = EarthLens(
    data_source="bdc",                  # alias 'brazil-data-cube' works too
    variables={"bdc/mod13q1-6.1": ["NDVI"]},
    lat_lim=[-23.7, -23.2], lon_lim=[-46.8, -46.3],
    start="2024-01-01", end="2024-12-31",
    path="./out", max_items=1,
).download()

`COPERNICUS/DEM/GLO30` → MPC cop-dem-glo-30 (DEM, anonymous)#

paths = EarthLens(
    data_source="planetary-computer",
    variables={"cop-dem-glo-30": ["data"]},
    lat_lim=[40.40, 40.45], lon_lim=[-3.72, -3.67],
    start="2020-01-01", end="2024-12-31",
    path="./out", max_items=1,
).download()

`ESA/WorldCover/v200` → MPC esa-worldcover (10 m land cover)#

paths = EarthLens(
    data_source="planetary-computer",
    variables={"esa-worldcover": ["map"]},
    lat_lim=[40.40, 40.45], lon_lim=[-3.72, -3.67],
    start="2021-01-01", end="2021-12-31",
    path="./out", max_items=1,
).download()

`USGS/3DEP/10m` → MPC 3dep-seamless (10 m DEM, US only)#

paths = EarthLens(
    data_source="planetary-computer",
    variables={"3dep-seamless": ["data"]},
    lat_lim=[37.5, 38.0], lon_lim=[-122.5, -122.0],
    start="2020-01-01", end="2024-12-31",
    path="./out", max_items=1,
).download()

Server-side reductions: `GOOGLE/DYNAMICWORLD/V1`#

There is no STAC equivalent for Dynamic World — it is a server-side ML reduction (run on Google's cluster) rather than a downloadable raster catalogue. The equivalent in earthlens is the process-graph backend:

paths = EarthLens(
    data_source="openeo",
    variables={"DYNAMIC_WORLD": [...]},   # see earthlens.openeo
    ...
).download()

Same applies to other GEE-resident reducers — composite-on-the-fly, cloudless mosaics, server-side spectral indices. Use earthlens.openeo (or stay on earthlens.gee) when the workload is the reduction, not the raster.

What does not migrate #

The following GEE assets have no earthlens.stac equivalent today:

Google's proprietary cloud-masking / atmospheric-correction layers (e.g. COPERNICUS/S2_CLOUD_PROBABILITY). Use the QA bands shipped with the STAC products (e.g. Sentinel-2 SCL, Landsat qa_pixel).
Region-of-interest reducers (ee.Reducer.mean() etc.). Use earthlens.openeo for those, or pull the COGs and reduce locally with pyramids.
GOOGLE/* (Dynamic World, Open Buildings, Floods). See the openEO note above; Open Buildings is also available as a flat-S3 release that an earthlens.s3-style dataset row can serve.