Low-dimensional analysis of geomagnetic reversals

Abstract:

Low-dimensional models for Earth's magnetic dipole have attracted attention recently because they may be a powerful tool to study the dominant dynamics over geological time-scales, where direct numerical simulation remains challenging. We investigate the extent to which several low-dimensional models can explain the Earth's dipole dynamics by comparing them to the signed relative paleointensity over the past 2 million years. Our comparisons of models and data are done by Bayesian statistics, which allows us to incorporate nonlinearity and uncertainty into the computations. The comparison, or data assimilation, reveals the strengths and weaknesses of each low-dimensional model and suggests improvements to the low-dimensional models. We also investigate if low-dimensional models can predict dipole reversals by performing extensive numerical experiments, and by hind-casting the Laschamp event, the Bruhnes-Matuyama reversal, as well as four other reversals documented over the past two million years. Our analysis stresses the need for models of geomagnetic reversals to faithfully account for the full spectrum of variability of paleomagnetic intensity.