USGS LandsatLook — Landsat C2 requester-pays¶

usgs-landsat is the authoritative USGS STAC for Landsat Collection 2. Earth Search mirrors the same data anonymously, but USGS keeps SR / ST / L1 as separate collections and serves them from a requester-pays bucket (s3://usgs-landsat) — so reads need an AWS identity and the bucket owner stays uncharged.

Earthlens wires this in via a per-collection signer: aws-requester-pays override on every USGS LandsatLook row (the same pattern Earth Search uses for its landsat-c2-l2 mirror). The notebook below is a recipe: it constructs the request, but only runs download() when AWS_ACCESS_KEY_ID + AWS_SECRET_ACCESS_KEY are set in the environment.

The endpoint + curated collections (no network)¶

In [ ]:

from earthlens.stac import Catalog

cat = Catalog()
e = cat.endpoints["usgs-landsat"]
print(f"url   : {e.url}")
print(f"signer: {e.signer}")
print(f"region: {e.region}")
for key in ("usgs-landsat/landsat-c2l2-sr", "usgs-landsat/landsat-c2l2-st", "usgs-landsat/landsat-c2l1"):
    col = cat.get_collection(key)
    print(f"  {key}  signer={col.signer}  default={col.default_assets}")

from earthlens.stac import Catalog cat = Catalog() e = cat.endpoints["usgs-landsat"] print(f"url : {e.url}") print(f"signer: {e.signer}") print(f"region: {e.region}") for key in ("usgs-landsat/landsat-c2l2-sr", "usgs-landsat/landsat-c2l2-st", "usgs-landsat/landsat-c2l1"): col = cat.get_collection(key) print(f" {key} signer={col.signer} default={col.default_assets}")

Surface Reflectance over SF Bay (gated on AWS creds)¶

Same request shape as Earth Search — only data_source and dataset change. If your environment has no AWS keys, the recipe prints the request and stops; with keys it pulls a single Landsat-9 surface-reflectance scene as a 4-band COG.

In [ ]:

import os
import tempfile

from earthlens.earthlens import EarthLens

request = dict(
    data_source="usgs-landsat",          # alias 'landsat' works too
    start="2024-06-01", end="2024-08-31",
    dataset="usgs-landsat/landsat-c2l2-sr",
    variables=["red", "green", "blue", "nir08"],
    aoi=[-122.5, 37.5, -122.0, 38.0],     # SF Bay
    path=tempfile.mkdtemp(),
    max_items=1,
)
have_aws = bool(os.environ.get("AWS_ACCESS_KEY_ID")) and bool(os.environ.get("AWS_SECRET_ACCESS_KEY"))
print("AWS creds present:", have_aws)
if have_aws:
    paths = EarthLens(**request).download()
    print("wrote:", paths)
else:
    paths = []
    print("set AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY to run; recipe shown above.")

import os import tempfile from earthlens.earthlens import EarthLens request = dict( data_source="usgs-landsat", # alias 'landsat' works too start="2024-06-01", end="2024-08-31", dataset="usgs-landsat/landsat-c2l2-sr", variables=["red", "green", "blue", "nir08"], aoi=[-122.5, 37.5, -122.0, 38.0], # SF Bay path=tempfile.mkdtemp(), max_items=1, ) have_aws = bool(os.environ.get("AWS_ACCESS_KEY_ID")) and bool(os.environ.get("AWS_SECRET_ACCESS_KEY")) print("AWS creds present:", have_aws) if have_aws: paths = EarthLens(**request).download() print("wrote:", paths) else: paths = [] print("set AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY to run; recipe shown above.")

Inspect (only when the read ran with AWS creds)¶

In [ ]:

if paths:
    from pyramids.dataset import Dataset

    ds = Dataset.read_file(str(paths[0]))
    print(f"bands={ds.band_count} epsg={ds.epsg} shape={ds.read_array().shape}")

if paths: from pyramids.dataset import Dataset ds = Dataset.read_file(str(paths[0])) print(f"bands={ds.band_count} epsg={ds.epsg} shape={ds.read_array().shape}")

Variants¶

• Surface Temperature: switch dataset to usgs-landsat/landsat-c2l2-st and variables to ["lwir"] (LST band).
• Level-1 TOA: dataset="usgs-landsat/landsat-c2l1" exposes pan, cirrus, and lwir11/lwir12 alongside the spectral bands.
• Earth Search mirror: data_source="earth-search", dataset="earth-search/landsat-c2-l2" reads the same data — also requester-pays, since the Earth Search row carries the same per-collection signer override.