skip to main content


Title: Can machine learning reveal precursors of reversals of the geomagnetic axial dipole field?
SUMMARY

It is well known that the axial dipole part of Earth’s magnetic field reverses polarity, so that the magnetic North Pole becomes the South Pole and vice versa. The timing of reversals is well documented for the past 160 Myr, but the conditions that lead to a reversal are still not well understood. It is not known if there are reliable ‘precursors’ of reversals (events that indicate that a reversal is upcoming) or what they might be. We investigate if machine learning (ML) techniques can reliably identify precursors of reversals based on time-series of the axial magnetic dipole field. The basic idea is to train a classifier using segments of time-series of the axial magnetic dipole. This training step requires modification of standard ML techniques to account for the fact that we are interested in rare events—a reversal is unusual, while a non-reversing field is the norm. Without our tweak, the ML classifiers lead to useless predictions. Perhaps even more importantly, the usable observational record is limited to 0–2 Ma and contains only five reversals, necessitating that we determine if the data are even sufficient to reliably train and validate an ML algorithm. To answer these questions we use several ML classifiers (linear/non-linear support vector machines and long short-term memory networks), invoke a hierarchy of numerical models (from simplified models to 3-D geodynamo simulations), and two palaeomagnetic reconstructions (PADM2M and Sint-2000). The performance of the ML classifiers varies across the models and the observational record and we provide evidence that this is not an artefact of the numerics, but rather reflects how ‘predictable’ a model or observational record is. Studying models of Earth’s magnetic field via ML classifiers thus can help with identifying shortcomings or advantages of the various models. For Earth’s magnetic field, we conclude that the ability of ML to identify precursors of reversals is limited, largely due to the small amount and low frequency resolution of data, which makes training and subsequent validation nearly impossible. Put simply: the ML techniques we tried are not currently capable of reliably identifying an axial dipole moment (ADM) precursor for geomagnetic reversals. This does not necessarily imply that such a precursor does not exist, and improvements in temporal resolution and length of ADM records may well offer better prospects in the future.

 
more » « less
Award ID(s):
1953778
NSF-PAR ID:
10370738
Author(s) / Creator(s):
; ; ; ; ;
Publisher / Repository:
Oxford University Press
Date Published:
Journal Name:
Geophysical Journal International
Volume:
231
Issue:
1
ISSN:
0956-540X
Page Range / eLocation ID:
p. 520-535
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. SUMMARY Earth’s internal magnetic field is dominated by the contribution of the axial dipole whose temporal variations are wide ranging and reflect characteristic timescales associated with geomagnetic reversals and large scale palaeosecular variation, ranging down to decadal and subannual field changes inferred from direct observations. We present a new empirical power spectrum for the axial dipole moment based on composite magnetic records of temporal variations in the axial dipole field that span the frequency range 0.1 to 5 × 105  Myr–1 (periods from 10 million to 2 yr). The new spectrum is used to build a stochastic representation for these time variations, based on an order 3 autoregressive (AR) process and placed in the context of earlier stochastic modelling studies. The AR parameter estimates depend on the frequency of transitions in the spectral regime and may be influenced by Ohmic diffusion, advection and torsional oscillations in Earth’s core. In several frequency ranges across the interval 200–5000 Myr–1(5000 to 200 yr periods) the empirical power spectrum lies above the AR3 model and may be influenced by Magneto–Coriolis (MC) waves in Earth’s core. The spectral shape and parameter estimates provide a potentially useful guide for developing assessments of whether numerical dynamo simulations meet criteria for being considered Earth like. 
    more » « less
  2. We consider a stochastic differential equation model for Earth's axial magnetic dipole field. The model's parameters are estimated using diverse and independent data sources that had previously been treated separately. The result is a numerical model that is informed by the full paleomagnetic record on kyr to Myr time scales and whose outputs match data of Earth's dipole in a precisely defined feature-based sense. Specifically, we compute model parameters and associated uncertainties that lead to model outputs that match spectral data of Earth's axial magnetic dipole field but our approach also reveals difficulties with simultaneously matching spectral data and reversal rates. This could be due to model deficiencies or inaccuracies in the limited amount of data. More generally, the approach we describe can be seen as an example of an effective strategy for combining diverse data sets that is particularly useful when the amount of data is limited. 
    more » « less
  3. Abstract

    Observations of relative paleointensity reveal several forms of asymmetry in the time dependence of the virtual axial dipole moment (VADM). Slow decline of the VADM into a reversal is often followed by a more rapid rise back to a quasi‐steady state. Asymmetry is also observed in trends of VADM during times of stable polarity. Trends of increasing VADM over time intervals of a few 10s of kyr are more intense and less frequent than decreasing trends. We examine the origin of this behavior using stochastic models. The usual (Langevin) model can account for asymmetries during reversals, but it cannot reproduce the observed asymmetry in trends during stable polarity. Better agreement is achieved with a different class of stochastic models in which the dipole is generated by a series of impulsive events in time. The timing of each event occurs randomly as a Poisson process and the amplitude is also randomly distributed. Predicted trends replicate the observed asymmetry when the generation events are large and the recurrence time is long (typically longer than 3 kyr). Large and infrequent generation events argue against dipole generation by small‐scale turbulent flow. Instead, the observations favor a mechanism that relies on expulsion of poloidal magnetic field from the core.

     
    more » « less
  4. Abstract

    The historical trend in the axial dipole is sufficient to reverse the field in less than 2 kyr. Assessing the prospect of an imminent polarity reversal depends on the probability of sustaining the historical trend for long enough to produce a reversal. We use a stochastic model to predict the variability of trends for arbitrary time windows. Our predictions agree well with the trends computed from paleomagnetic models. Applying these predictions to the historical record shows that the current trend is likely due to natural variability. Furthermore, an extrapolation of the current trend for the next 1 to 2 kyr is highly unlikely. Instead, we compute the trend and time window needed to reverse the field with a specified probability. We find that the dipole could reverse in the next 20 kyr with a probability of 2%.

     
    more » « less
  5. Abstract

    Fluctuations in the geomagnetic field occur over a broad range of timescales. Short‐period fluctuations are called secular variation, whereas excursions and reversals are viewed as anomalous transient events. An open question is whether distinct mechanisms are required to account for these different forms of variability. Clues are sought in trendsbof the axial dipole moment from six time‐dependent geomagnetic field models. Variability inbhas a well‐defined dependence on the time interval (or window) for the trend. The variance ofbreveals a simple relationship to trends during excursions and reversals. This connection hints at a link between reversals, excursions and secular variation. Stochastic models exhibit a similar behavior in response to random fluctuations in dipole generation. We find that excursions, reversals and secular variation can be distinguished on the basis of trend durations rather than differences in the underlying physical process. While this analysis does not rule out distinct physical mechanisms, the paleomagnetic observations suggest that such distinctions are not required.

     
    more » « less