GOES-R ABI — catalog explorer (no network)¶

The earthlens.goes backend fetches raw NOAA GOES-R ABI NetCDF granules from the anonymous noaa-goes* AWS buckets. Before downloading anything, it helps to understand the three axes of a GOES request — satellite, product, and domain — and the spectral bands the imager carries.

This notebook is fully offline: it only reads the bundled catalog (earthlens.goes.Catalog). By the end you will know which products, domains, satellites, and ABI channels are available, and how they map to an S3 key.

Setup¶

Load the catalog and pandas for tidy tables.

In [1]:

import pandas as pd

from earthlens.goes import Catalog

cat = Catalog()
cat  # a compact repr: how many products / available datasets

import pandas as pd from earthlens.goes import Catalog cat = Catalog() cat # a compact repr: how many products / available datasets

Out[1]:

Catalog(datasets=17, available_datasets=17)

Products¶

A GOES request names a product family with dataset=. Each row maps a friendly key to its ABI product_group (the S3 prefix stem), the processing level, and the domains it publishes. band_split marks the products that store one file per ABI channel (so variables= selects which channels to fetch).

In [2]:

products = pd.DataFrame(
    [
        {
            "dataset": key,
            "product_group": p.product_group,
            "level": p.level,
            "domains": ", ".join(p.domains),
            "band_split": p.band_split,
            "description": p.description.split(" - ")[0],
        }
        for key, p in sorted(cat.datasets.items())
    ]
)
products

products = pd.DataFrame( [ { "dataset": key, "product_group": p.product_group, "level": p.level, "domains": ", ".join(p.domains), "band_split": p.band_split, "description": p.description.split(" - ")[0], } for key, p in sorted(cat.datasets.items()) ] ) products

Out[2]:

	dataset	product_group	level	domains	band_split	description
0	abi-l1b-rad	ABI-L1b-Rad	L1b	C, F, M1, M2	True	ABI radiances (Level-1b)
1	abi-l2-acha	ABI-L2-ACHA	L2	C, F, M1, M2	False	Cloud Top Height.
2	abi-l2-acm	ABI-L2-ACM	L2	C, F, M1, M2	False	Clear Sky Mask.
3	abi-l2-actp	ABI-L2-ACTP	L2	C, F, M1, M2	False	Cloud Top Phase.
4	abi-l2-adp	ABI-L2-ADP	L2	C, F, M1, M2	False	Aerosol Detection (smoke / dust) product.
5	abi-l2-aod	ABI-L2-AOD	L2	C, F	False	Aerosol Optical Depth at 550 nm.
6	abi-l2-cmip	ABI-L2-CMIP	L2	C, F, M1, M2	True	Cloud & Moisture Imagery (single-band)
7	abi-l2-cod	ABI-L2-COD	L2	C, F	False	Cloud Optical Depth.
8	abi-l2-cps	ABI-L2-CPS	L2	C, F, M1, M2	False	Cloud Particle Size distribution.
9	abi-l2-dmw	ABI-L2-DMW	L2	C, F, M1, M2	False	Derived Motion Winds.
10	abi-l2-dsi	ABI-L2-DSI	L2	C, F, M1, M2	False	Derived Stability Indices (CAPE, lifted index,...
11	abi-l2-fdc	ABI-L2-FDC	L2	C, F, M1, M2	False	Fire / Hot Spot Characterization.
12	abi-l2-lst	ABI-L2-LST	L2	C, F, M1, M2	False	Land Surface (skin) Temperature.
13	abi-l2-mcmip	ABI-L2-MCMIP	L2	C, F, M1, M2	False	Cloud & Moisture Imagery (multi-band)
14	abi-l2-rrqpe	ABI-L2-RRQPE	L2	F	False	Rainfall Rate / Quantitative Precipitation Est...
15	abi-l2-sst	ABI-L2-SST	L2	F	False	Sea Surface (skin) Temperature (Full Disk only).
16	abi-l2-tpw	ABI-L2-TPW	L2	C, F, M1, M2	False	Total Precipitable Water.

Domains¶

ABI scans four domains. CONUS (C) and Full Disk (F) each have their own S3 prefix suffix; the two Mesoscale sub-sectors (M1, M2) share one …M prefix and are told apart by a filename token. Cadence is informational — the backend lists the actual hour prefix rather than computing an expected count.

In [3]:

domains = pd.DataFrame(
    [
        {
            "domain": key,
            "name": d.name,
            "cadence_minutes": d.cadence_minutes,
            "prefix_suffix": d.prefix_suffix,
            "subsector": d.subsector or "-",
        }
        for key, d in cat.domains.items()
    ]
)
domains

domains = pd.DataFrame( [ { "domain": key, "name": d.name, "cadence_minutes": d.cadence_minutes, "prefix_suffix": d.prefix_suffix, "subsector": d.subsector or "-", } for key, d in cat.domains.items() ] ) domains

Out[3]:

	domain	name	cadence_minutes	prefix_suffix	subsector
0	C	CONUS	5.0	C	-
1	F	Full Disk	10.0	F	-
2	M1	Mesoscale 1	1.0	M	M1
3	M2	Mesoscale 2	1.0	M	M2

Satellites → buckets¶

satellite= accepts an operational role (east / west) or an explicit satellite number (16 / 18 / 19). The role→bucket map lives in the catalog because East/West rotate as new GOES satellites commission.

In [4]:

pd.DataFrame(
    sorted(cat.satellites.items()), columns=["satellite", "bucket"]
)

pd.DataFrame( sorted(cat.satellites.items()), columns=["satellite", "bucket"] )

Out[4]:

	satellite	bucket
0	16	noaa-goes16
1	18	noaa-goes18
2	19	noaa-goes19
3	east	noaa-goes19
4	west	noaa-goes18

ABI spectral channels¶

The imager carries 16 spectral bands. For the band-split products (abi-l1b-rad, abi-l2-cmip) you pick channels with variables=["C02", "C13", ...]. The visible / near-IR bands sit at short wavelengths; the water-vapour and window bands are thermal-infrared.

In [5]:

channels = pd.DataFrame(
    [
        {"channel": k, "wavelength_um": c.wavelength_um, "name": c.name}
        for k, c in cat.channels.items()
    ]
)
channels

channels = pd.DataFrame( [ {"channel": k, "wavelength_um": c.wavelength_um, "name": c.name} for k, c in cat.channels.items() ] ) channels

Out[5]:

	channel	wavelength_um	name
0	C01	0.47	Blue (visible)
1	C02	0.64	Red (visible)
2	C03	0.86	Veggie (near-IR)
3	C04	1.37	Cirrus (near-IR)
4	C05	1.60	Snow/Ice (near-IR)
5	C06	2.20	Cloud particle size (near-IR)
6	C07	3.90	Shortwave window (IR)
7	C08	6.20	Upper-level water vapour (IR)
8	C09	6.90	Mid-level water vapour (IR)
9	C10	7.30	Lower-level water vapour (IR)
10	C11	8.40	Cloud-top phase (IR)
11	C12	9.60	Ozone (IR)
12	C13	10.30	Clean longwave window (IR)
13	C14	11.20	Longwave window (IR)
14	C15	12.30	Dirty longwave window (IR)
15	C16	13.30	CO2 longwave (IR)

A quick look at where each band sits on the spectrum:

In [6]:

import matplotlib.pyplot as plt

fig, ax = plt.subplots(figsize=(8, 3))
ax.bar(channels["channel"], channels["wavelength_um"], color="#3b7dd8")
ax.set_ylabel("wavelength (µm)")
ax.set_title("ABI spectral channels")
ax.tick_params(axis="x", rotation=90)
plt.tight_layout()
plt.show()

import matplotlib.pyplot as plt fig, ax = plt.subplots(figsize=(8, 3)) ax.bar(channels["channel"], channels["wavelength_um"], color="#3b7dd8") ax.set_ylabel("wavelength (µm)") ax.set_title("ABI spectral channels") ax.tick_params(axis="x", rotation=90) plt.tight_layout() plt.show()

From a request to an S3 key (still no network)¶

Constructing the backend resolves the satellite to a bucket and the product + domain to an S3 prefix — without touching the network. This is the prefix the backend will list per hour of the requested window.

In [7]:

from earthlens.goes import GOES

goes = GOES(
    start="2026-07-03 12:00", end="2026-07-03 12:10",
    lat_lim=[20, 50], lon_lim=[-130, -60],
    dataset="abi-l2-mcmip", satellite="east", domain="C",
    fmt="%Y-%m-%d %H:%M",
)
print("bucket:", goes._bucket)
print("prefix:", goes._prefix())
print("hours :", [str(h) for h in goes.time.dates])

from earthlens.goes import GOES goes = GOES( start="2026-07-03 12:00", end="2026-07-03 12:10", lat_lim=[20, 50], lon_lim=[-130, -60], dataset="abi-l2-mcmip", satellite="east", domain="C", fmt="%Y-%m-%d %H:%M", ) print("bucket:", goes._bucket) print("prefix:", goes._prefix()) print("hours :", [str(h) for h in goes.time.dates])

bucket: noaa-goes19
prefix: ABI-L2-MCMIPC
hours : ['2026-07-03 12:00:00']

Takeaway¶

• A GOES request is satellite × product × domain × time window.
• dataset= picks the product family; variables= picks ABI channels for the band-split products; satellite= and domain= resolve to a bucket + S3 prefix, all from the catalog.
• Next: the CONUS granule download notebook fetches a real granule and decodes a band with pyramids.