Regional Open Data Cubes — DEAfrica + DEA¶

Two of the new endpoints — deafrica and dea — are Open Data Cube STAC catalogs over their respective continents. Both are anonymous (no credentials needed), and both expose a Water Observations from Space (WOfS) product that flags every clear water pixel detected by Landsat. This notebook pulls a WOfS observation from each over a small bbox to show the endpoint abstraction — same request shape, different region, same anonymous signer.

The two endpoints in the catalog (no network)¶

In [ ]:

from earthlens.stac import Catalog

cat = Catalog()
for ep in ("deafrica", "dea"):
    e = cat.endpoints[ep]
    print(f"{ep:9} url={e.url}")
    print(f"          signer={e.signer}  region={e.region}")
print()
print("deafrica wofs default_assets:", cat.get_collection("deafrica/wofs_ls").default_assets)
print("dea      wofs default_assets:", cat.get_collection("dea/ga_ls_wo_3").default_assets)

from earthlens.stac import Catalog cat = Catalog() for ep in ("deafrica", "dea"): e = cat.endpoints[ep] print(f"{ep:9} url={e.url}") print(f" signer={e.signer} region={e.region}") print() print("deafrica wofs default_assets:", cat.get_collection("deafrica/wofs_ls").default_assets) print("dea wofs default_assets:", cat.get_collection("dea/ga_ls_wo_3").default_assets)

DEAfrica WOfS over Johannesburg (live)¶

WOfS classifies each pixel into a small set of water-state codes; the water asset is the single-band classification we want.

In [ ]:

import tempfile

from earthlens.earthlens import EarthLens

africa = EarthLens(
    data_source="deafrica",
    start="2024-01-01", end="2024-12-31",
    dataset="deafrica/wofs_ls",
    variables=["water"],
    aoi=[28.0, -26.5, 28.5, -26.0],   # Johannesburg
    path=tempfile.mkdtemp(), max_items=1,
)
africa_paths = africa.download()
africa_paths

import tempfile from earthlens.earthlens import EarthLens africa = EarthLens( data_source="deafrica", start="2024-01-01", end="2024-12-31", dataset="deafrica/wofs_ls", variables=["water"], aoi=[28.0, -26.5, 28.5, -26.0], # Johannesburg path=tempfile.mkdtemp(), max_items=1, ) africa_paths = africa.download() africa_paths

DEA WOfS over Canberra (live, same shape)¶

The only thing that changes is data_source, dataset, and the bbox. The signer is still anonymous, the product still emits a single water asset.

In [ ]:

australia = EarthLens(
    data_source="dea",                    # alias 'digital-earth-australia' works too
    start="2024-01-01", end="2024-12-31",
    dataset="dea/ga_ls_wo_3",
    variables=["water"],
    aoi=[149.0, -35.5, 149.5, -35.0],     # Canberra
    path=tempfile.mkdtemp(), max_items=1,
)
aus_paths = australia.download()
aus_paths

australia = EarthLens( data_source="dea", # alias 'digital-earth-australia' works too start="2024-01-01", end="2024-12-31", dataset="dea/ga_ls_wo_3", variables=["water"], aoi=[149.0, -35.5, 149.5, -35.0], # Canberra path=tempfile.mkdtemp(), max_items=1, ) aus_paths = australia.download() aus_paths

Inspect what each endpoint returned¶

Both rasters are single-band uint8 WOfS classifications, projected to each endpoint's native UTM zone. Reading them through pyramids exposes the CRS and shape.

In [ ]:

from pyramids.dataset import Dataset

for label, paths in (("DEAfrica", africa_paths), ("DEA", aus_paths)):
    ds = Dataset.read_file(str(paths[0]))
    print(f"{label:8} bands={ds.band_count} epsg={ds.epsg} shape={ds.read_array().shape}")

from pyramids.dataset import Dataset for label, paths in (("DEAfrica", africa_paths), ("DEA", aus_paths)): ds = Dataset.read_file(str(paths[0])) print(f"{label:8} bands={ds.band_count} epsg={ds.epsg} shape={ds.read_array().shape}")

Where to go next¶

• Both catalogs ship ~10 curated collections — landsat / sentinel-2 ARD, fractional cover, GeoMedian annual composites, Copernicus DEM mirrors. Catalog().get_collection(<key>) lists them.
• Swap to data_source="deafrica" / "dea" with dataset="deafrica/fc_ls" or "dea/ga_ls8c_nbart_gm_cyear_3" for fractional cover / annual NBART GeoMedian.