GOES-R ABI — download a CONUS granule (live)¶

This notebook fetches a real GOES-East ABI granule from the anonymous noaa-goes19 bucket via the earthlens.goes backend, then decodes one infrared band with pyramids and maps it.

earthlens.goes ships the raw NetCDF and does not decode it — reading and reprojecting the geostationary grid is pyramids' job downstream. This notebook shows that hand-off end to end.

Setup¶

Imports and a notebook-relative output directory. GOES CONUS imagery updates every 5 minutes with a short latency, so we pick a window a few hours in the past to be sure the granules are already published.

In [ ]:

from pathlib import Path

import numpy as np
import pandas as pd

from earthlens.core import EarthLens

out_dir = Path("goes_out")
out_dir.mkdir(exist_ok=True)

from pathlib import Path import numpy as np import pandas as pd from earthlens.core import EarthLens out_dir = Path("goes_out") out_dir.mkdir(exist_ok=True)

In [ ]:

now = pd.Timestamp.utcnow().tz_localize(None).floor("min")
start = now - pd.Timedelta(hours=3)
end = start + pd.Timedelta(minutes=6)  # one CONUS scan (5-min cadence)
start, end

now = pd.Timestamp.utcnow().tz_localize(None).floor("min") start = now - pd.Timedelta(hours=3) end = start + pd.Timedelta(minutes=6) # one CONUS scan (5-min cadence) start, end

Download¶

EarthLens("goes", ...).download() lists the hour prefix, keeps the granules whose scan-start lands in the window, and downloads them whole. It returns the list[pathlib.Path] of raw NetCDF files.

In [ ]:

lens = EarthLens(
    "goes",
    dataset="abi-l2-mcmip",  # Cloud & Moisture Imagery (16 bands)
    satellite="east",  # -> noaa-goes19
    domain="C",  # CONUS
    start=start.strftime("%Y-%m-%d %H:%M"),
    end=end.strftime("%Y-%m-%d %H:%M"),
    fmt="%Y-%m-%d %H:%M",
    lat_lim=[20, 50],
    lon_lim=[-130, -60],
    path=out_dir,
)
paths = lens.download()
[p.name for p in paths]

lens = EarthLens( "goes", dataset="abi-l2-mcmip", # Cloud & Moisture Imagery (16 bands) satellite="east", # -> noaa-goes19 domain="C", # CONUS start=start.strftime("%Y-%m-%d %H:%M"), end=end.strftime("%Y-%m-%d %H:%M"), fmt="%Y-%m-%d %H:%M", lat_lim=[20, 50], lon_lim=[-130, -60], path=out_dir, ) paths = lens.download() [p.name for p in paths]

The filename encodes everything: OR_ABI-L2-MCMIPC-M6_G19_s…_e…_c….nc — the product (MCMIPC), the scan mode (M6), the satellite (G19), and the scan start / end / created timestamps. One CONUS file (~58 MB) carries all 16 ABI bands.

In [ ]:

granule = paths[0]
print("granule :", granule.name)
print("size    :", round(granule.stat().st_size / 1e6, 1), "MB")

granule = paths[0] print("granule :", granule.name) print("size :", round(granule.stat().st_size / 1e6, 1), "MB")

Decode a band with pyramids¶

The granule is raw geostationary NetCDF. pyramids georeferences the ABI scan-angle grid from the CF goes_imager_projection grid-mapping, so we can pull the clean longwave IR window band (CMI_C13, 10.3 µm — the classic IR cloud channel) and warp it to WGS84.

In [ ]:

from pyramids.dataset import Dataset, GeoReference
from pyramids.netcdf import NetCDF

nc = NetCDF.read_file(granule)
band = nc.get_variable("CMI_C13").to_crs(4326)
image = np.asarray(band.read_array()).astype("float32")
image[image >= 65535] = np.nan  # mask the fill value
image.shape

from pyramids.dataset import Dataset, GeoReference from pyramids.netcdf import NetCDF nc = NetCDF.read_file(granule) band = nc.get_variable("CMI_C13").to_crs(4326) image = np.asarray(band.read_array()).astype("float32") image[image >= 65535] = np.nan # mask the fill value image.shape

Map it¶

Colder cloud tops are bright in the inverted IR palette; the warm surface is dark. The values are the granule's packed brightness-temperature counts (lower = warmer), which is why we invert the colormap.

In [ ]:

vmin, vmax = np.nanpercentile(image, [2, 98])

brightness = Dataset.from_array(
    image,
    no_data_value=np.nan,
    geo_ref=GeoReference(geo=band.geotransform, epsg=band.epsg),
)
glyph = brightness.plot(
    cmap="Greys",
    vmin=vmin,
    vmax=vmax,
    title=f"GOES-19 ABI CMI_C13 (10.3 µm) — {granule.name.split('_s')[1][:11]}",
)
glyph.cbar.set_label("packed IR counts (low = warm)")

vmin, vmax = np.nanpercentile(image, [2, 98]) brightness = Dataset.from_array( image, no_data_value=np.nan, geo_ref=GeoReference(geo=band.geotransform, epsg=band.epsg), ) glyph = brightness.plot( cmap="Greys", vmin=vmin, vmax=vmax, title=f"GOES-19 ABI CMI_C13 (10.3 µm) — {granule.name.split('_s')[1][:11]}", ) glyph.cbar.set_label("packed IR counts (low = warm)")

Takeaway¶

• EarthLens("goes", ...).download() returns raw ABI NetCDF granule paths — no decoding, no server-side subset.
• Reading / reprojecting the geostationary grid is a downstream pyramids (or satpy) step — here NetCDF.read_file(...).get_variable("CMI_C13").to_crs(4326).
• Swap domain="F" for Full Disk, domain="M1" for a 1-minute mesoscale sector, or dataset="abi-l1b-rad", variables=["C02"] for a single-channel radiance file. See the catalog explorer.