Rapid geomagnetic changes inferred from Earth observations and numerical simulations

Modern global geomagnetic field models derived from satellite and observatory data establish features of secular variation, including hemispheric differences and rapid changes at high latitudes linked to core–mantle interactions and core flow. On longer timescales, paleomagnetic studies report very rapid intensity changes (0.75–1.5 μT yr−1) around 1000 BCE in the Levant, substantially exceeding modern maxima, though debated. Directional changes approaching ~1° yr−1 have been reported from Italian sediments near the Matuyama–Brunhes reversal, far exceeding modern rates (~0.1° yr−1), but their reliability remains contested. Kinematic and synthetic field studies have hinted at rapid directional changes during dipole weakening or synthetic reversals, without establishing physical limits, geographic preferences, or dynamical origins. This study asks: how fast can local geomagnetic directions change in general, do such extremes prefer specific locations, and what core-surface processes generate them? Leveraging long-timescale paleofield models (0–100 ka) and geodynamo simulations, the authors quantify extreme directional change rates during stable polarity, excursions, and reversals, and diagnose their physical origins.

Prior work documents: (1) Large secular variation patterns in recent centuries to decades. (2) Archeomagnetic intensity spikes (e.g., Levant, China, Texas) with high dB/dt, whose rates are compatible with core-flow kinematic bounds, and reproduced by simulations via migrating normal-polarity flux patches. (3) Reports of extremely rapid directional change (~1° yr−1) in Italian sediments near the Matuyama–Brunhes reversal, under ongoing debate. (4) Kinematic/synthetic models show site-dependent reversal behaviors and potential ~1° yr−1 dipole-latitude changes due to moving intensity spikes. However, a comprehensive assessment of directional-change speed limits, global spatial preferences, and dynamical compatibility with core physics was lacking.

Data and models: (1) Observational paleofield models: GGF100k covering 0–100 ka (global, time-varying); higher-resolution models spanning the Laschamp excursion (e.g., LSMOD.2) for sensitivity checks. (2) Numerical geodynamo simulations: 16 3D convection-driven dynamo runs spanning magnetic Reynolds numbers Rm ≈ 100–700 (Earth ≈ 1000), including reversing and non-reversing regimes, with input parameters (E, Pm, Ra, Pr = 1, ri/ro = 0.35, boundary conditions with no-slip and insulating/conducting combinations, and fixed heat flux/temperature; some with laterally varying CMB heat flux). Simulations are scaled to physical units using advection time and polar-mean field strength. Computation of directional rates: The magnetic field vector B at Earth’s surface is computed on a 2°×2° latitude–longitude grid over time. Directional rate of change for the local field unit vector B̂ is defined as arccos of the dot product between successive unit vectors at times t and t+Δt; the analogous metric is computed for the virtual dipole vector Pv derived from inclination and declination. Δt and grid spacing are chosen conservatively to avoid overestimating extremes; additional tests with denser sampling quantify sensitivity (potential factors 2–5 higher). Identification of extremes: For each location and time step, the maximum rate of change in direction (∂P/∂t; and analogously using B) is recorded; the global maxima are labeled extremal events. Analyses include time series near extremal events, latitude of occurrences, and statistical distributions (histograms and cumulative distribution functions) of dB/dt by latitude. Simple analogue model: A controlled CMB radial field configuration consisting of a static axial dipole (~80 μT) superimposed with a single moving flux patch constructed with a Fisher–von Mises distribution (half-width ~15°), of normal (A<0) or reversed (A>0) polarity. The patch is translated in latitude and/or longitude in fixed increments (Δt set to 1 for relative rate comparison). Upward continuation provides the surface field to evaluate directional change rates and their spatial dependence. Masking analysis for dynamical attribution: Around each extremal event in simulations, the authors identify four 30°×30° quadrants on the CMB sharing a vertex at the extremal site. They halve the CMB radial field amplitude in one quadrant at a time (f=0.5) in real space, carefully transforming to/from spherical harmonics up to original truncation (ℓ≈128–256) to minimize aliasing. Recomputed surface directional rates reveal which quadrant (and thus which CMB features, notably reversed flux patches) most strongly contributes to the extremal change. Statistical characterization: Rates of change in GGF100k are analyzed via log-normal fits globally and by latitude; ratios of means and standard deviations at higher latitudes relative to the equator summarize latitudinal dependence across both observational and simulated datasets.

• Extreme directional change rates in simulations reach ~10° yr−1, about 10× higher than fastest values reported in individual paleomagnetic records and ~40× higher than modern Holocene model maxima (~0.1° yr−1).
• Observational models show compatible magnitudes when sampled more densely: conservative estimates around the Laschamp excursion in GGF100k and LSMOD.2 are ~2.5–3.5° yr−1; with 10-year sampling, GGF100k reaches ~4.8° yr−1 and LSMOD.2 up to ~22.5° yr−1.
• Extremal directional changes commonly occur during periods of decreased overall field strength but are not confined to polarity reversals; many occur thousands of years away from the axial dipole equatorial crossing and also appear in non-reversing simulations during excursions or regional events.
• Geographic preference: Most extreme events occur at absolute latitudes <40°. The probability of faster directional rates is highest at low latitudes; CDFs and log-normal fits show mean rates decrease with increasing latitude.
• Rarity: Over 100 kyr in GGF100k (~5 million surface-time samples), only one interval exceeds 2° yr−1; 95% of samples are below 0.05° yr−1.
• Physical mechanism: Analogue model and masking experiments identify migration of reversed flux patches across the CMB as the principal cause of extremal directional changes at the surface. Normal-polarity patches can drive changes but generally slower; reversed patches moving toward the observation site create local low-inclination regions where the non-dipole field dominates directional variability.
• Distinct from intensity spikes: Prior work and current analyses indicate intensity spikes tend to arise from rapid migration/intensification of normal-polarity patches at higher latitudes (>50°), whereas directional extremes are linked to reversed flux and occur preferentially at low latitudes, implying different underlying core-surface dynamics for intensity versus directional extremes.

The study demonstrates that rapid directional changes, far exceeding modern observed rates, are dynamically compatible with physically based geodynamo simulations. The alignment between simulations and paleofield models in both amplitude and preferred low-latitude occurrence indicates that such extremes are genuine features of core dynamics, not artifacts of modeling. The identified mechanism—migration of reversed flux patches reducing local inclination and allowing non-dipole field to control direction—directly addresses the research question on the physical origin of rapid directional changes and explains their association with periods of lower field strength and equatorial preference, where reversed flux occurs more frequently. The rarity quantified in long records clarifies why such events are difficult to substantiate in isolated paleomagnetic archives. Comparison with intensity variations underscores that directional and intensity extremes likely probe different aspects of core-surface dynamics, consistent with known spatial sampling biases of inclination/declination versus intensity. Potential influences of a stratified layer at the top of the core and lateral CMB heat flow heterogeneity are discussed: weaker fields can coincide with higher rates of change, while horizontal flux motions responsible for directional extremes may not be impeded by stratification; no clear systematic differences were observed between simulations with and without CMB heat-flow heterogeneity, though the sample size limits robust inference. The results suggest future paleomagnetic searches for rapid directional changes should target low latitudes and times of diminished field strength, especially near excursions, and motivate further dynamical attribution studies.

This work establishes that the geomagnetic field can undergo extremely rapid directional changes—up to ~10° yr−1 in simulations and multi-degree per year in paleofield models—much faster than modern variations. These extremes occur preferentially at low latitudes (<40°), are more likely during intervals of reduced field strength, and are dynamically linked to the migration of reversed flux patches across the core-mantle boundary. The strong agreement between observational models and numerical simulations validates the physical plausibility of such events and provides guidance for future data acquisition: focus on low latitudes and intervals around excursions. Future research should (1) expand global time-dependent paleofield models with improved chronology and resolution, (2) develop automated tracking of normal and reversed flux patches to quantify their statistical contributions across events, (3) explore parameter regimes and boundary conditions (e.g., stratified layers, lateral CMB heat flow) that modulate extremal behavior, and (4) assess whether directional records can yield information about intensity spikes under known sampling biases.

• Numerical simulations do not reach Earth-like extreme parameter values (Ekman and magnetic Prandtl numbers), though they reproduce many Earth-like features; extrapolation uncertainty remains.
• Paleofield model limitations: GGF100k’s spatial/temporal data coverage and chronological uncertainties smooth variations and likely underestimate true extremes; higher-resolution models show larger rates.
• The simple analogue patch model omits interactions among multiple patches and assigns arbitrary time increments, precluding absolute rate calibration.
• Masking attribution assumes static quadrant-based reduction and does not track time-varying patch strengths/directions; global spherical harmonic representation can introduce minor aliasing despite mitigation.
• Sparse and uneven paleomagnetic data, especially outside excursions and at low latitudes, limit direct observational validation of extreme directional events and their geographic distribution.