The Effects of the Temperature Interval Selection on Paleointensity Estimates Over the Past 5 Million Years

Abstract:

Over the past 5 million years, where data are plentiful and plate motions small, the time-averaged geomagnetic field is well approximated by a GAD model in paleomagnetic field directions. However, site-level data from geographical dispersed studies for this time interval obtained from variants of well-established Thellier-Thellier series experimental procedures fail to coincide with the GAD prediction of latitudinal dependence in field intensities (Fig. 1A). Although advanced data selection protocols are typically used in modern studies, residual biases due to chemical alteration, multidomain grains, and/or viscous remanent magnetization (VRM) can still be responsible for unreliable estimations of the paleointensity. In particular, the common use of auto-selected temperature segments may not adequately exclude the effects of VRM, which are likely to be common but difficult to detect on the Arai diagrams.

Combining the PINT2015.05 database and several studies published since 2015, we analyze 1052 specimen-level Thellier-Thellier series paleointensity data with their temperature intervals (Fig. 1B) from 323 site-level data of 29 studies for normal and reverse polarity intervals over the past 5 million years. Site-level data using standard selection criteria show a virtual axial dipole moment (VADM) estimate of 6.77×10²² Am², comparable to the present-day geomagnetic field strength of 8 ×10²² Am², but no discernable latitude variation. Moreover, the overall mean VADM for site-level data for normal polarity (7.34×10²² Am²) is ~20% higher than for reverse polarity (6.12×10²² Am²). This polarity asymmetry is most likely due to contamination by VRM along the normal (present-day) field direction. Indeed, data from the maximum temperature interval below 350 ℃ (n=20) provide a mean VADM of 8.44×10²² Am² even closer to the present-day VADM; data from the temperature interval higher than 350 ℃ (n=144) provide a much lower mean VADM of 5.97×10²² Am², and reveal a latitudinal variation more consistent with the GAD model (Fig. 1C) while the mean VADMs for normal and reverse polarity data are more equal.

Our results suggest that less than 30% of the reported specimen-level paleointensity data avoid the effect of VRM overprints, which may even dominate in some samples, because of a low-temperature interval selected.