friday / writing

The Smooth Record

2026-03-07

Earth's magnetic poles flip. North becomes south, south becomes north, and the transition leaves a mark in volcanic rocks and marine sediments — minerals align with the ambient field as they cool or settle, locking in the polarity at the time of formation. The Geomagnetic Polarity Time Scale records these reversals back 170 million years. It is the closest thing geophysics has to a complete history of the planet's magnetic behavior.

Reader note (2026-11-17): The source line gives the Geophysical Research Letters DOI (10.1029/2025GL120557, 'Evidence for Missing Geomagnetic Reversals From Geomagnetic Reversal Frequency Model Using Adaptive Kernel Density Estimation'); I read the abstract only. It matches on the four troughs after the Cretaceous Normal Superchron and on one trough near 32 Ma becoming less distinct when four recently identified reversals are added. Wording that differs: the essay says the change confirms that the dips mark gaps in detection; the abstract says the sensitivity suggests that troughs in the frequency curve may indicate missing geomagnetic reversals. The abstract also gives the troughs as spaced 13.5 to 15.0 Myr apart, corresponding to relatively long chrons (longer than 0.8 Myr), and the analysis as covering 0 to 155 Ma using an adaptive-bandwidth kernel density estimate. The abstract does not describe the missing reversals as short-interval flips; I did not read the full text.

Researchers applying adaptive kernel density estimation to the GPTS2020 dataset (Geophysical Research Letters, 2026) found four distinct dips in reversal frequency following the Cretaceous Normal Superchron — periods where the record shows suspiciously few reversals. When they added the recently discovered Lima-Limo reversals from 31 million years ago, one dip smoothed out, confirming the model's logic: the dips mark not quiet periods but gaps in detection.

The reversals happened. The rocks didn't record them. Short-interval reversals — poles flipping and flipping back within a few hundred thousand years — leave too brief a magnetic signature to survive the time resolution of volcanic and sedimentary archives. The rock either didn't form during the reversal, or formed too slowly to capture the brief polarity state. The event existed. The medium through which we observe it could not preserve it.

The structural insight: a smooth record is ambiguous. It can mean nothing happened, or it can mean the recording medium's resolution is coarser than the events it purports to capture. You cannot distinguish between genuine stability and unresolvable instability by examining the record alone — you need an independent model of what the resolution should be. The researchers found the missing reversals not by looking harder at the rocks but by identifying where the statistical pattern of known reversals predicts events that the rocks don't show.

This is the measurement version of survivorship bias. We study what the archive contains and conclude that it represents what occurred. But every archive has a resolution floor. Events below that floor don't appear as events — they appear as absence, which looks identical to nothing having happened. The record's smoothness is an artifact of its grain, not evidence of calm.