Per-dataset output kind¶

Earthdata is the first earthlens backend whose OUTPUT_KIND is resolved per request from the catalog row, not fixed on the class. This notebook shows the catalog side of that (no network / credentials needed). For the live download see Usage.

Catalog: raster vs vector datasets¶

Load the Earthdata catalog and split its datasets by output_kind. Each row already declares whether it resolves to a raster or a vector request, so the partition is a simple comprehension over cat.datasets.

In [ ]:

from earthlens.earthdata import Catalog

cat = Catalog()
raster = [k for k, d in cat.datasets.items() if d.output_kind == 'raster']
vector = [k for k, d in cat.datasets.items() if d.output_kind == 'vector']

from earthlens.earthdata import Catalog cat = Catalog() raster = [k for k, d in cat.datasets.items() if d.output_kind == 'raster'] vector = [k for k, d in cat.datasets.items() if d.output_kind == 'vector']

The raster datasets are the gridded products; the vector ones are the point / footprint products (ICESat-2, GEDI, OCO-2). Print the counts to see the split:

In [ ]:

print('raster datasets:', len(raster))
print('vector datasets:', sorted(vector))

print('raster datasets:', len(raster)) print('vector datasets:', sorted(vector))

What the output kind controls¶

When you build EarthLens(data_source='earthdata', ...), the resolved row's output_kind becomes the instance's OUTPUT_KIND. The facade reads it at download() time to gate aggregate=:

• a raster instance forwards aggregate=AggregationConfig(...) into the backend, which reduces a stack of single-timestep granules per window;
• a vector / tabular instance rejects aggregate= with NotImplementedError.

A single request must not mix output kinds — EarthData rejects that at construction, since one instance carries exactly one kind.

DAAC registry¶

Each CMR provider code maps to a DAAC, its cloud region, and the S3 credentials endpoint used for in-region streaming. Look up the first few provider codes to see the mapping:

In [ ]:

for code in sorted(cat.daacs)[:5]:
    daac = cat.get_daac(code)
    print(f'{code:12s} -> {daac.daac:10s} ({daac.cloud_region})')

for code in sorted(cat.daacs)[:5]: daac = cat.get_daac(code) print(f'{code:12s} -> {daac.daac:10s} ({daac.cloud_region})')