Launch Watch · Evidence study

Published 3 October 2026, 14:31 HST (UTC−10)

GOES-16 in 2019: an audited calibration review calendar

Audit 29 GOES-16 calibration-log rows from 2019, including a repeated gap notice, distinct same-day intervals and band-specific review points.

Two GOES-16 Band 2 images from opposite sides of 23 April 2019 may differ partly because a calibration correction occurred that day. A November gap can have a different explanation, and one missing-scan description appears twice in the log. Before interpreting image changes, it helps to put the operational record beside the observations.

We audited the NOAA STAR GOES-16 calibration table and selected all 29 rows whose start date falls in 2019. The resulting calendar preserves duplicate descriptions, missing end dates and distinct same-day intervals. It is a review aid for historical imagery, not an automatic list of images to discard.

What changed after reviewing the rows

Keep both November 13 entries visible, but flag the repetition. Rows 23 and 26 have the same dates and the same description after whitespace normalization: a missing 18:10 full-disk image and incomplete 18:20 image. Their event titles differ. Removing the duplicate description leaves 28 date-and-description combinations, which still must not be called 28 independent incidents.

Keep the two August 17 entries separate. One synchronization interval runs from 06:00 to 16:50 UTC on August 17. The other runs from 21:20 that evening to 03:40 UTC on August 18. Deduplicating by date or event title alone would erase a distinct interval.

Do not stretch blank end dates into outages. Three selected rows have no end date: the April 2 Mode 6A transition, April 23 Band 2 correction and July 25 ground-system upgrade. They record changes with continuing relevance, but a blank cell does not establish continuous degraded operation.

April 23 contains two different review points

The log gives 11:00 UTC for a Band 2 calibration change with an approximately 6.2% reduction in radiance. It gives 20:15 UTC for a Band 7 resampler update addressing cold pixels around fire. Those are different bands, different interventions and different times.

For a before-and-after image comparison, that distinction changes the next step. A Band 2 brightness discontinuity calls for checking which calibration applies. A Band 7 fire-region artifact calls for checking the resampling correction and product quality information. Neither notice establishes that a particular pixel in your scene was affected; that requires the scene itself.

Use the calendar as a triage sequence

1. Identify the product. Record satellite, band, scan sector, observation time and processing provenance before matching it to a notice.
2. Match dates and explicit scope. A date match alone is too broad. Check the band, affected image format and the notice’s exact or approximate time.
3. Separate change dates from impairment intervals. A calibration implementation, a missing scan and intermittent detector striping should not become interchangeable duration bars.
4. Inspect the actual data. Check the relevant image and quality information, then document whether it is usable for your particular comparison.

A start-year filter misses carryover context

The 29-row selection deliberately asks which notices begin in 2019. A check for everything that could affect a 2019 scene needs an interval-overlap search as well. In the retained table, three earlier-starting rows have finite listed end dates in 2019: Channel 8 striping beginning 31 December 2018, Band 5 striping beginning in 2017, and the Band 7 cold-pixel issue also beginning in 2017.

The first extends through 31 January 2019 in the source dates. The latter two end on 24 June and 23 April 2019 respectively. Earlier rows with blank ends require additional interpretation rather than an automatic assumption that they remain active. One older source cell also contains concatenated material, another reason to preserve the original log instead of treating the extracted table as a flawless incident database.

Keep the remaining limits in view

NOAA describes this as a discretionary collection of events with known or potential Level 1b data-quality effects. It is not a complete availability record. Without a completeness guarantee and a suitable observation denominator, counts of its rows cannot establish a satellite failure rate.

The dates also have different meanings. December’s Band 12 row spans repeated striping and temporary release, not necessarily uninterrupted impairment. The November 12 description gives an outage interval and missing full-disk labels that do not line up exactly; both are preserved. The July 25 row sits in the GOES-16 table but its description explicitly discusses a GOES-17 calibration improvement, which should not be reassigned to GOES-16.

We have not retrieved affected calibrated scenes or measured a new pixel bias. The useful output is the audited calendar and its exceptions. Pair it with your exact imagery, record why a notice matters, and make the include-or-exclude decision at the product level.