Launch Watch · Analysis

Published 4 October 2026, 08:54 HST (UTC−10)

The September 2017 flare: comparing NOAA’s X9.3 and X14.6 records

Compare the 2017 flare bulletin, GOES-15 operational readings and a GOES-16 science-report peak, with a reproducible scaling-only test.

NOAA’s September 2017 bulletin calls the 6 September flare X9.3. A later NOAA science-report product lists the peak at the same 12:02 UTC minute label as X14.6. That difference needs more than a renamed flare class: the historical operational data and the science report use different processing, and the numerical records compared here come from different spacecraft.

We placed the 2017 SWPC bulletin, archived GOES-15 one-minute data, and the version 1-0-1 composite science flare report side by side. We then calculated what happens when one documented scale factor is removed from the operational series. The exercise reproduces part of the numerical difference and shows exactly where a scaling-only explanation stops.

Keep the product beside the number

The two numerical records describe the XRS-B long band, nominally 0.1–0.8 nanometers. NOAA’s flare-report readme explains that the report’s peak irradiances derive from one-minute averages and that the product can select data from different satellites. The source-spacecraft field is therefore essential. The GOES-16 value is not a newly labeled GOES-15 row.

Method: remove one documented scale factor

NOAA’s operational XRS readme instructs users to divide the long-band operational irradiance by 0.7 to remove the historical SWPC scaling factor. Applied to the GOES-15 peak, that gives 0.00093293 ÷ 0.7 = 0.001332757 watts per square meter, rounded here after calculation.

We applied the same division to every long-band value in the two-hour window. All 120 retained rows have a zero B-band quality flag and positive irradiance. The original series and its rescaled version peak at the same row because multiplying every positive value by the same constant preserves their order. The dashed curve is consequently a useful arithmetic comparison, not an independent observation.

The GOES-16 science-report peak remains higher: 0.001464 minus 0.001332757 is approximately 0.000131243 watts per square meter. Relative to the scaling-only GOES-15 result, the difference is about 9.85%. Removing the factor explains a large part of the gap but does not reconcile these two products.

Why the remaining 9.85% is not a calibration-error estimate

We changed one scale factor and held the operational values otherwise fixed. NOAA’s science-quality readme describes additional processing differences. This article does not reconstruct that pipeline, compare the two spacecraft on a common calibrated time series, or isolate the contribution of any one adjustment.

Even the matching 12:02 minute labels need care. The operational readme states that GOES-13 through GOES-15 averaged timestamps fall 1.024–3.072 seconds after the start of their included data. Matching the printed minute therefore does not establish identical acquisition or integration intervals across products. The comparison joins a dated event; it does not prove sample-for-sample temporal equivalence.

Preserve the history when citing the flare

For an account of what SWPC reported in September 2017, retain the bulletin’s X9.3 label and cite that record. For a calculation using the retrieved version 1-0-1 science report, retain its X14.6 value, GOES-16 source and product version. Silently replacing every historical X9.3 reference with X14.6 would erase the record being cited.

The reproducible result is that dividing this GOES-15 operational peak by 0.7 does not equal the separately sourced GOES-16 science-report peak. Keep spacecraft, passband, cadence, time convention and processing version with the number before comparing historical flare magnitudes. These data do not establish a new flare-energy total, a spacecraft failure or an effect on a particular launch.