Comparing Excursions and the Present-Day Field.

It has been suggested that the present-day SAA may expand and deepen, leading to an excursion or reversal (8). Although the mechanisms that initiate these events could be different, the SAA-like intensity structures at 49 and 46 ka do not grow and spread across Earth’s surface to form either excursions or reversals. Rather, the field remains dipole-dominated during and after SAA-like epochs (Figs. 1 and 4). This leads us to infer that SAA-like structures are transitory and not diagnostic of an imminent excursion or reversal.

This finding does not imply that today’s field will not continue its current decrease in intensity. This may persevere for some uncertain amount of time. The current global DM is higher than during both the 49- and 46-ka SAA-like epochs (Fig. 3B) and could drop to at least these values. If the SAA continues to deepen, this will have the greatest consequences for the Americas. It has been noted that even a moderate decrease in dipole strength could have practical implications (25). For example, the current SAA has led to electrical failures on low Earth orbit satellites passing through the SAA (26), and with a continued decrease in field intensity, issues such as this will become more widespread.

Initiation of Excursions.

The generation of the excursions follows a different mechanism. For the Laschamp excursion, as for times before the 49- and 46-ka SAA-like events, the spectra are initially similar to the present day. However, from ∼43 ka onward, the axial dipole component weakens and the directional changes of the excursion are initiated slowly (Fig. 3). Two large-intensity anomalies at the surface grow almost simultaneously (one over central America and one over SE Asia), and these are associated with the growth of reversed flux patches below these locations on the CMB. Through time, the magnitude of the dipole begins to decay more rapidly, and the energy spectrum at the CMB becomes white for all degrees at ∼42 ka (Fig. 4). This does not immediately yield large directional changes associated with the excursion (although some locations now have geomagnetic poles at equatorial latitudes), as even in this state, the dipole energy at the surface exceeds the energy of the non-dipole components. At just before 41 ka, the non-dipole field begins to dominate at the surface as the fall in the axial dipole continues rapidly. Intensity anomalies at the surface grow and new spatially unconnected anomalies develop. Numerous reversed flux patches develop at the CMB, and, immediately before 41 ka, reversed flux covers both poles (Fig. 2). At 41 ka, the dipole reaches its minimum with a magnitude close to zero and then ceases to decay further (Fig. 3). Intensity is now low across the surface, and fully reversed directions are observed at numerous locations; however, there is globally nonuniform directional variability with many locations showing only small directional excursions, as would be expected when the axial dipole reduces close to zero (Fig. 3), but does not reverse (27).

For the initiation of the Mono Lake excursion, the axial dipole begins to decay from a dipole dominated state at ∼36 ka. However, unlike the Laschamp excursion, the decay in the axial dipole is only short-lived, with dipole energy remaining significantly above non-dipole energy at the surface (Fig. 3) and CMB (Fig. 4), with a minimum in the DM occurring just before 34 ka. Multiple surface intensity anomalies appear at mid and low latitudes from 36 ka onward, but they are short-lived and erratic in their spatial evolution and do not resemble SAA-like field structures (Movie S1). Reversed flux patches start to develop after 35 ka, with reversed flux at mid to high latitudes in both hemispheres appearing at ∼35 ka as the energy in the octupole term matches the energy of the dipole term. However, the Mono Lake excursion is not driven into a more substantial excursion, and by 33 ka, the field regains its dominantly dipolar appearance at the surface, although increased energy in some of the higher-order components post-33 ka creates greater spatial complexity in both intensity and Br than for today’s field.

A Mechanism for Excursion and Reversal Generation.

A possible interpretation for excursions is that field contributions of all scales vary with time, but with a weaker base state (a weaker axial dipole), variations in the axial dipole are sufficient to produce an excursion (27). Excursions can therefore be considered a part of normal secular variation, but with a weaker axial dipole contribution and a greater chance of producing a zero or reversed dipole field. We argue that the field has two possible mean states, one in which the axial dipole is dominant at the CMB, which is broadly stable, and one in which the axial dipole matches the higher degrees, when fluctuations in the axial dipole can produce an excursion. The question remains: Are random fluctuations sufficient for a full reversal? To answer this question, a reexamination of the most recent reversal, the Matuyama–Brunhes, building on previous modeling (30), but with an expanded dataset, will be necessary.

We infer that for excursions to occur, a weakening of the field across much of the globe spreading from multiple sources is required, and not just localized weakening expanding from an SAA-like feature. They also require the growth of reversed flux patches in both hemispheres, with reversed flux transiting the poles. However, the amount and duration of reversed flux across the poles differs for the Laschamp and Mono Lake excursions. Reversed flux occurs simultaneously across both poles during the Laschamp (Fig. 2) and is present for ∼2 ka, whereas across the Mono Lake, reversed flux is located primarily over the North Pole (Fig. 2) for only a few hundred years (Movie S1).

Although today’s dipole is weakening, it is still substantially stronger than the higher-degree components of the field and exceeds the dipole moment from our model through the majority of 40–50 ka. Today’s secular variation is instead comparable to the SAA-like states at 49 and 46 ka, which did not lead to an excursion. Similar arguments apply to the Mono Lake excursion; although it does not reach the magnitude and extent of the Laschamp excursion, it still starts from a more geographically spread weak state than from a single SAA-like feature.

Relationship Between SAA-Like Structures and Cosmogenic Nuclide Production.

The above inferences depend on the robustness of our model; however, there is evidence from a complementary source: variations in cosmogenic nuclide production. Peaks in nuclide production (¹⁰Be and ³⁶Cl) coincide with the intensity lows of the Laschamp (29, 31) and Mono Lake (31, 32) excursions (Fig. 3C). However, there are also peaks in production that do not match known excursions. One production peak of particular note occurs at 46 ka and is coeval with one of our SAA-like structures, supporting our observation of a weakened field at this time. This peak is evident as a low in the 1/800 y filtered ¹⁰Be- and ³⁶Cl-derived dipole variations obtained from Greenland ice core records (7, 28) and is hinted at in the stacks of sediment ¹⁰Be/9Be records from the Portuguese margin and west equatorial Pacific (29) (Fig. 3C). There are two other times when there are decreases in both the ¹⁰Be- and ³⁶Cl-derived dipole data. The low at ∼31 ka is approximately coeval with a decrease in DM in our model. The low at ∼37 ka does not appear to be correlated with any field behavior in our model; however, this low appears negligible in the same records filtered at 1/3,000 y (31). The origin of the anticorrelation of the ¹⁰Be- and ³⁶Cl-derived dipole ice core records between 49 and 50 ka is currently unknown (31), and we do not consider cosmogenic nuclide data before 48 ka in our analysis.

Cosmogenic nuclide records reflect globally averaged production through postproduction atmospheric mixing processes (33). Numerous factors can influence cosmogenic nuclide production and concentration (31). On time scales longer than hundreds of years, production results primarily from variations in the geomagnetic field, although possible millennial-scale solar activity and climate influences may complicate estimates of dipole change (31). It is commonly assumed that production is related to changes in the dipole component of the field. This is based on the altitude of modulation of cosmic rays and the rapid fall-off of non-dipole components with distance from the CMB (31). However, we observe only a minor reduction in DM at 46 ka in our model (Fig. 3) and suggest instead that localized areas of weak intensity, such as SAA-like structures, might additionally allow cosmogenic nuclides to penetrate the geomagnetic field. We do not claim that the apparently coveal timing of one our SAA-like structures with a period of increased cosmogenic nuclide production unequivocally links the two events; however, this observation would be worth pursuing by researchers in the future. This conjecture could be tested by incorporating the geomagnetic model presented here into models of cosmogenic nuclide production for this time.