The geomagnetic excursion dated to ≈190 ka BP, now widely referred to as the Iceland Basin Excursion (IBE), is considered one of the most significant geomagnetic events of the Brunhes Chron (Simon et al., 2016). One of the first attempts to date this excursion with high confidence came from lava flows, notably in the Snake River Plain (Idaho, USA). Champion et al. (1981, 1988) reported transitional paleomagnetic directions in multiple flow units, including Site E in Idaho, and proposed a Potassium-Argon (K-Ar) age of (188±8) ka BP. They named this event the Jamaica excursion, based on terminology used in earlier literature (Ryan, 1972). This naming particularly persisted in the literature related to volcanology.

Subsequent studies began identifying concomitant excursions in marine sediment cores, particularly in the Pacific. Tauxe and Wu (1990) reported excursion-like signals from two cores on the Ontong Java Plateau (ERDC-89P and ERDC-113P), with age estimates around 190 ka BP. Around the same time, Valet and Meynadier (1993) published data from ODP Sites 848 and 851 showing an excursion between 180 and 200 ka BP, which they tentatively linked to the Jamaica event described by Champion et al. (1981, 1988). Records from the western equatorial Pacific also reported an excursion around that age identified in cores NP5, NP7, NGC16, and NGC29 (Yamazaki and Ioka, 1994), later extended to NP35, NGC36, and NGC38 (Yamazaki et al., 1995).

Highly-resolved analyses of North Atlantic sediment cores confirmed the existence of a short, high-amplitude, and global excursion around 190 ka BP (Weeks et al., 1995), that was associated to the Jamaica event, referring to Champion et al. (1988). Other sites further supported these conclusions including a suite of eight cores from the Indian and Western Pacific Oceans (Guyodo and Valet, 1996), three cores from the Acores regions (Lehman et al., 1996), and from Northwest Pacific (Roberts et al., 1997). The name Iceland Basin Excursion was formally introduced in 1997 based on high-resolution magnetic studies of sediment core collected from the Iceland Basin, ODP sites 983 and 984 (Channell, 1999; Channell et al., 1997). Both sites of the Iceland Basin revealed a short full reversal event at ≈186–189 ka BP with magnetization components rotating through 180° and back, in coincidence with a relative paleointensity (RPI) low of approximately 3 ka. These records as well as the other studies from that region (listed below), led to widespread adoption of the term Iceland Basin Excursion.

Since 1997, numerous studies have reported the IBE across a wide range of archives and locations. Here, we provide a non-exhaustive list of these observations. IBE has been robustly documented at multiple ODP sites along the Western and Eastern North Atlantic: Sites 1061, 1062, and 1063 (Channell et al., 2012; Christl et al., 2010; Knudsen et al., 2008; Laj et al., 2006; ODP Leg 172 Scientific Party et al., 1998, p. 172), as well as at Sites 980, 984, and 919 (Channell, 2006, 1999; Channell and Raymo, 2003). Additional detections were reported from the Labrador basin (Stoner et al., 1998), Western and Equatorial Pacific cores (Gee et al., 2000), Lake Baikal (Demory et al., 2005; Oda et al., 2002), Southern Ocean (Stoner et al., 2003), the Portuguese margin (Thouveny et al., 2004), and the Irminger Basin and the Eirik Drift (Evans et al., 2007). IBE was first observed with authigenic 10Be measurements from Portuguese margin marine cores MD95-2040 and MD95-2042 (Carcaillet et al., 2004; Knudsen et al., 2008). IBE was reported in a lava flow at Unzen Volcano (Japan) with a K-Ar age of (197±17) ka BP (Shibuya et al., 2007). Further establishing its global extent, the GDM low linked to IBE has also been observed in an Antarctic ice core with atmospheric 10Be concentrations from Dome Fuji (Horiuchi et al., 2016). Because of its global characteristics, global stack compilations are representative of the collapse of GDM during IBE (Channell, 2014; Simon et al., 2016; Valet et al., 2020, 2024).

Several geomagnetic excursions have been identified in the 205–245 ka BP interval, but their amplitude, naming end even number remain debated. Herrero-Bervera et al. (1994) provided one of the most comprehensive syntheses of an RPI low dated to (218±10) ka BP with 40Ar/39Ar across multiple western U.S. sites: Pringle Falls – which gave its name to the event – and Summer Lake (Oregon), Mono Lake (Paoha Island), and Benton Crossing (Long Valley, California), although the first observation of a geomagnetic event around that age was probably observed in lava flows from Albuquerque (Geissman et al., 1990). Given dating uncertainties, some studies associated the Pringle Falls excursion to the Jamaica excursion (Langereis et al., 1997). In the Southern Hemisphere, a prominent geomagnetic reversal was recorded in the Mamaku ignimbrite, erupted from the Taupo volcanic zone, New Zealand. This event has been dated using 40Ar/39Ar methods at (227±8) ka BP (Houghton et al., 1995; Tanaka et al., 1996), with revised estimates at (223±3) ka BP based on improved chronostratigraphy (Mcwilliams, 2001). Updated dating of the excursion recorded in Albuquerque lavas and Pringle Falls ash D – (218±14) ka BP and (211±13) ka BP, respectively – supported that all three sites (Albuquerque, Pringle Falls, Mamaku) likely recorded the same geomagnetic excursion, named Pringle Falls, and dated to (211±13) ka BP (Singer et al., 2008).

However, sediment records complicated this interpretation. For instance, RPI minima were observed around 208 ka BP (Channell, 2006) and 238 ka BP (Channell et al., 2012) in Atlantic sediment cores. Besides, a long minimum is observed in the PISO-1500 stack between 208 and 221 ka BP, while a sharp, decoupled minimum is recorded around 238 ka BP (Channell et al., 2009). This led to question the existence of two, instead of a unique, “Pringle Falls” excursions.

In this sense, the comparison of RPI and cosmogenic nuclide concentrations would solve this ambiguity. Between 210 and 220 ka BP, the RPI low correlates with a cosmogenic nuclide production enhancement, while a marked RPI low and a minor cosmogenic nuclide production enhancement occurs at ca. 236 ka BP (Simon et al., 2016). However, the latter has also been dated around (230±12) ka BP in Mamaku lavas (Shane et al., 1994). Finally, a small amplitude RPI low has been observed around 250–255 ka BP concomitant with a clear enhancement in authigenic 10Be/9Be measured in two West Equatorial Pacific Ocean sediment cores (Simon et al., 2016). This event could correspond to the Calabrian Ridge 0 excursion (Mediterranean Sea, Langereis et al., 1997), the 8α excursion (IODP Leg 172 Sites 1060–1063, Lund et al., 2001), and the Fram strait excursion (Nowaczyk and Frederichs, 1999).

Consequently, a continuous, globally-representative and well-dated record like ice core 10Be from Antarctica would provide valuable information to better understand low-amplitude excursions. However, until now, no study has discussed this double excursion. This literature synthesis supports a cautious approach for dating and identifying the Pringle Falls, Mamaku and Calabrian Ridge 0-like excursion(s?).

To give credit to all these previous studies we will give consensual names to the three excursions we study in this work: the Iceland Basin Excursion (IBE), the Pringle Falls-Excursion (PFE), and the Mamaku Excursion (ME), dated around 190, 210, and 240 ka BP, respectively.

Measurements of 10Be have been performed in the Talos Dome ice core (TALDICE). Talos Dome is situated on the East Antarctic plateau (72°48^′ S, 159°06^′ E; 2316 m a.s.l., Fig. 1), being particularly close to oceans (290 km from the Southern Ocean and 250 km from the Ross Sea) than other drillings like EPICA Dome C (EDC) or Dome Fuji (DF) (Frezzotti et al., 2004). The drilling campaign reached 1620 m, being 175 m above the bedrock (Crotti et al., 2021). It is worthnoting that the drilling was situated in a glacial valley, whose hills reach 1550 m (Crotti et al., 2021).

The ice core was continuously sampled in sections of 20 cm (when possible) between 1470 and 1499 m and between 1505 and 1531 m. The interval between 1499 and 1505 m was not sampled as part of a targeted sampling strategy focusing on depth intervals corresponding to the expected timing of the IBE, PFE, and ME excursions, based on existing age models and marine sediment records. The outermost section of the ice core, dedicated to cosmogenic nuclide measurements, had been stored in clean polyethylene bags at -20 °C in a cooling facility, close to CEREGE. As part of the LGO2i platform at CEREGE research institute, the ice samples were cut with a saw pre-decontaminated with ultrapure ice and the outermost 1 mm of ice was removed with a ceramic knife. Ice samples were always handled with clean nitrile gloves, and left to melt in glass beakers, covered with plastic film, at room temperature.

In order to compare with 10Be fluxes from Dome C over the last century, 11 samples have been measured from TALDICE firn between 4.5 and 10 m with a resolution of 50 cm.

The AICC2023 chronology (Bouchet et al., 2023) was used to date the ice, spanning from (172.3±1.9) to (275.7±1.8) ka BP. Our record covers several transitions from glacial to interglacial periods, from Marine Isotope Stage (MIS) 6.4 to 8.3 (Railsback et al., 2015). Therefore, the annual snow accumulation rate, provided by the AICC2023 chronology, ranges between (4.3±0.8) and (8.1±1.5) cm a⁻¹ with a mean value of 5.5 cm a⁻¹. On average, the ≈20 cm resolution corresponds to ≈300 years (min = 60 a; max = 1375 a).

The flux of 10Be was calculated as 10Be concentration × snow accumulation rate × ice density (i.e., 0.917 g cm⁻³). A half-life corrected flux was also calculated, using the (1.387±0.012) Ma 10Be half-life (Chmeleff et al., 2010; Korschinek et al., 2010). The resulting mean half-life-corrected 10Be flux is (1.36±0.26)×105 at cm-2a-1, considering uncertainty on 10Be concentration, snow accumulation rate, and 10Be half-life. Although this calculation is numerically valid, it likely overestimates the uncertainty associated with accumulation rate. Indeed, in the AICC2023 chronology, uncertainties on snow accumulation rate are inherited from the Paleochrono Bayesian framework (Parrenin et al., 2024) and strongly depend on the input uncertainties. During the interval 170–270 ka BP, the relative age uncertainty is only ±1 %, whereas the relative uncertainty in accumulation rates reaches ±20 % (Bouchet et al., 2023). This large value reflects the prior uncertainty (20 %) assigned to accumulation scenarios, which are derived from a poorly constrained empirical relationship between present-day water isotope in the snow and accumulation (Parrenin et al., 2007). In Paleochrono, such prior scenarios are iteratively adjusted within their uncertainty range to fit the dated horizons. In the AICC2023 simulation, posterior age estimates are constrained, but accumulation variations remain loosely constrained. Consequently, the ±20 % snow accumulation rate uncertainty that would arise from directly propagating the AICC2023 accumulation error is likely overestimated and further work is needed to estimate precisely the uncertainty in accumulation scenarios (Marie Carole Bouchet, personal communication, 2025).

In order to avoid artificial inflation of the uncertainty associated with 10Be flux variations, we propose a more realistic uncertainty propagation to track relative changes through time rather than absolute flux values. Rather than propagating the full absolute uncertainty of the snow accumulation rate, we use the 20 % accumulation rate uncertainty from AICC2023 as a relative uncertainty (0.2) applied to deviations from the mean accumulation rate (Eq. 1). With this formulation, accumulation rates equal to the mean have no associated uncertainty in terms of variability, although they remain subject to absolute uncertainty. This approach therefore does not represent the total uncertainty of the accumulation rate, but rather a “variation uncertainty” that reflects how uncertainty in accumulation affects the amplitude of 10Be flux fluctuations. This corrected uncertainty reflects the fact that variability in 10Be flux is primarily driven by changes in accumulation over time rather than by its absolute range of uncertainty. On average, the corrected uncertainty, used when interpreting 10Be flux variations as in Fig. 2, is 22 % (min 0.2 %; max 53 %) lower than the raw uncertainty. 1Realtive corrected_uncertainty(t)=accumulation(t)-accumulationmean170-270ka×0.2accumulation(t) Prior to comparison with the TALDICE record, the Dome Fuji 10Be fluxes (Horiuchi et al., 2016) were recalculated using the more recent accumulation rate. The DF record spans a period from 170 to 205 ka BP. While the record was published using the DFO-2006 chronology (Kawamura et al., 2007) with snow accumulation rate being retrieved from Parrenin et al. (2007), 10Be fluxes were corrected using the most recent DF chronology, DF2021 (Oyabu et al., 2022).

To identify anomalously low 10Be fluxes, which could bias the interpretation of geomagnetic intensity, we applied an objective statistical criterion. These 10Be minima were identified when the concentration fell below the mean minus one standard deviation, both calculated using a 3 ka rolling window. The corrected 10Be flux was calculated removing the 10Be minima associated with a maximum in major ion concentrations.

Between 170 and 270 ka BP, TALDICE 10Be concentrations vary between (0.25±0.03)×104 at g⁻¹ and (7.1±0.2)×104 at g⁻¹, with mean concentration and uncertainty values of 2.5×104 at g⁻¹ and 0.08×104 at g⁻¹, respectively (Fig. 2A). This period spans MIS 8 to 6, including Termination III and the MIS 7, a time marked by pronounced climatic fluctuations (Railsback et al., 2015) (Fig. 2C). After accounting for variations in snow accumulation rates using the AICC2023 chronology (Bouchet et al., 2023), 10Be fluxes show values between (1.9±0.4)×104 at cm-2a-1 and (31.0±6.1)×104 at cm⁻² a⁻¹ (Fig. 2B). These flux variations highlight notable production changes independent of deposition variations. Concurrently, VADM values fluctuated indicating intervals of geomagnetic instability (Channell et al., 2009; Simon et al., 2016). In particular, 4 geomagnetic excursions are reported with little to strong depletions of the VADM, around 190, 215, 240 ka BP, and 258 ka BP (Fig. 2E). An increase in the 10Be flux related to a background production is expected during such minima due to reduced geomagnetic shielding, discussed in Sect. 5.2 and 5.3. One geomagnetic excursion is already clearly recorded in TALDICE: a pronounced 10Be flux peak associated with the IBE excursion around 190 ka BP. Another event, previously identified in two sediment cores from the South-West Pacific (Simon et al., 2016), likely occurred around 258 ka BP but is not recorded in TALDICE due to the absence of measurements caused by a missing section of the ice core.

A total of 40 minima in 10Be concentration were identified across the TALDICE record studied in this work. Because minima are defined at the 10Be sampling resolution, a low-concentration interval extending over 40 cm is counted as two distinct minima. The 10Be minima appear to coincide with maxima in the concentrations of major ions (16 coincidences out of the 23 maxima observed in major ions, Fig. 3). This association is statistically significant (permutation test's p value <0.0001), though no direct quantitative relationship can be established. The major ions involved originate from a variety of sources, including oceanic sea spray (Na⁺, Cl⁻, Mg²⁺, MSA, SO42-), crustal dust (Ca²⁺, Mg²⁺), and volcanic (SO42-) sources. In addition to ion concentration peaks, many of the 10Be minima are also associated with decreases in the Cl-/Na+ ratio which is typically used to study changes in the relative contributions of marine aerosols or alterations in transport processes (Legrand et al., 2017).

To test whether the identified 10Be concentration minima preferentially occur under particular conditions, we compared their distribution against δD and [Ca²⁺]. Of the 40 minima identified, 31 occurred during glacial intervals (δD <-300 ‰), which is proportional to the fraction of the record spent in glacial conditions (77 %). A χ2 test confirms no significant increase in the number of minima during glacials (p=1.00), and δD values at 10Be minima are statistically indistinguishable from non-minima levels (Mann-Whitney U test, p=0.889). In contrast, [Ca²⁺] concentrations are systematically higher at 10Be minima. While median [Ca²⁺] is only slightly elevated at minima compared to the background (Mann–Whitney U p=0.53), contingency tests using thresholds show strong enrichment: for example, 32 % of 10Be minima exceed 15 ppb Ca²⁺ compared to 13 % of the background (χ2=8.0, p=0.005), and 7 10Be minima exceed 30 ppb compared to only 9 out of 221 non-minima samples (χ2=8.4, p=0.005). This suggests that short-lived 10Be minima preferentially coincide with dust-rich conditions.

After removing the 10Be minima associated with major ion maxima, rolling mean averages can be calculated to smooth the record and obtain the first-order variations, which are likely to result. Testing 1, 3, and 5 ka rolling mean averages (Fig. S1 in the Supplement) illustrate the trade-off between noise reduction and signal preservation. Given the mean resolution of our record (300 a) we selected a 3 ka rolling mean average as a practical compromise. This choice provides stable background trends while preserving the amplitude of GDM-related variations. The resulting 3 ka averaged 10Be flux record can be compared to geomagnetic reconstructions, including the Dome Fuji ice core data (Horiuchi et al., 2016) in Fig. 4, and authigenic 10Be/9Be records from marine sediment cores (Simon et al., 2016) and the 10Be production (Poluianov et al., 2016) calculated from RPI-based VADM (Channell et al., 2009) in Fig. 5. Flux enhancement during geomagnetic excursions or events depends strongly on the choice of the background value. When discussing paleomagnetism, the background should be defined as the 10Be profile without minima. Here, we consider either the full 170–270 ka BP interval 10Be flux mean average, after removing 10Be minima (see Sect. 5.1) but includes the geomagnetic events, of (1.44±0.26)×105 at cm-2a-1 or a fixed value of 1.1×105 at cm-2a-1 as background. The latter could be considered as representative of the long-term background 10Be flux, which is close to the minimum values the 3 ka rolling mean average around 178 and 223 ka BP. Depending on the chosen background, the flux enhancement factors during specific events are as follows: 1.55 or 2.02 for the 190 ka BP event (IBE); 1.23 or 1.62 for the 205–215 ka BP event (PFE); and 1.17 or 1.53 for the 240 ka BP event (ME). As a general idea, the flux enhancement factor for IBE represents 1.4 times the modern value of ≈1.64×105 at cm⁻² a⁻¹ at Dome C (Jouzel et al., 2026). It is evident that defining a clear background value would have been preferable; however, this was not possible due to the limitations inherent in our profile and our understanding of the polar bias. We therefore advocate the necessity of having this dual scenario, mean or fixed values, with further work being necessary to refine this question.

Post-depositional artifacts in the geomagnetically-induced 10Be signal in ice cores have been widely discussed, particularly in Greenland, where dust contributions have been shown to significantly affect the total 10Be budget especially during glacial periods during which the dust level are higher than during interglacials (Baumgartner et al., 1997). Among known confounding factors, snow accumulation plays a central role: when poorly constrained, 10Be concentrations tend to mirror accumulation variability (Yiou et al., 1985), complicating interpretations in terms of solar or geomagnetic modulation (Delaygue and Bard, 2011). Additional pre-depositional processes – such as stratospheric transport, volcanic injections (Baroni et al., 2011), wind-induced grain metamorphism (Lal et al., 2001), and general atmospheric aerosol dynamics (Zheng et al., 2024) – can all influence the 10Be deposition. However, in the case presented here, we report distinct 10Be minima that cannot be explained by variations in accumulation, deposition processes, or atmospheric circulation.

These 10Be minima co-occur with sharp maxima in major ion concentrations (e.g., Na⁺, Cl⁻, Ca²⁺), yet the ions originate from diverse sources. This rules out scenarios similar to single-source volcanic fallout (Baroni et al., 2011), terrestrial dust input (Baumgartner et al., 1997), or irregular snow redistribution (Poizat et al., 2024) and extreme atmospheric events such as atmospheric rivers (Wille et al., 2021). Instead, the inverse relationship between 10Be and major ions suggests a post-depositional control. Spikes in the concentration of major ions have been observed in the deep section of EDC that had been linked to impurities migration in the ice crystal boundaries (Traversi et al., 2009). Anomalous 10Be signals were also reported in the deepest parts of the EDC core (>700 ka; >3100 m) (Raisbeck et al., 2006), documenting 10Be maxima in EDC rather than minima as in TALDICE. This apparent contrast with TALDICE likely reflects differences in analytical protocols. In particular, the EDC study did not involve ion-exchange resin, unlike the present study. As reported in Kappelt et al. (2025), horizontal migration of Be was proposed to explain these maxima, as “smoothing over several thousands of years does not yield a distribution resembling the expected production signal smoothed by a vertical migration” (Kappelt et al., 2025). Under deep-ice conditions characterized by large grain sizes and enhanced impurity relocation, 10Be may become incorporated into dust-rich aggregates or mineral phases at grain boundaries. Such particle-bound Be may not be quantitatively recovered by the ion-exchange protocol, leading to apparent 10Be minima despite elevated concentrations of major ions. However, it is worth noting that some of the EDC 10Be maxima were also concomitant with spikes in other species, including dust (Raisbeck et al., 2006). Similarly, major ion spikes in EDC were associated with low acidity (Raisbeck et al., 2006), which is different from the high acidity observed in TALDICE.

Baumgartner et al. (1997) proposed another mechanism to explain the covariations between 10Be and dust. Indeed, in the deepest and warmest ice of the GRIP core (Greenland), up to 40 %–50 % of 10Be become dust-bound – higher than the <5 % seen in Holocene ice – due to ice metamorphism. The migration of 10Be and dusts at the ice grain boundaries would result in higher local concentrations and thus favour the adsorption of 10Be onto dust particles. Although such effects are expected to be less pronounced in low-dust Antarctic settings, deeper sections of ice cores, where ice crystals are large, may promote localized enrichment of major ions and 10Be adsorption onto grain-boundary dust, muting the dissolved-phase signal. This mechanism is also in agreement with higher relocations and reactions of dusts in deep TALDICE (Baccolo et al., 2021). The significant connection between 10Be minima occurrence and elevated Ca²⁺ concentrations, particularly above 30 ppb, suggests that extreme, non-atmospheric conditions are the main drivers of these 10Be minima. The absence of a significant relationship with the glacial/interglacial state, as defined by δD, reinforces the idea that these anomalies are not controlled by large-scale atmospheric changes in production or transport, but instead reflect in-ice processes. In this way, while further experimental work is needed to directly test this hypothesis, including stronger leaching protocols, ion remobilisation associated with ice grain metamorphism, leading to 10Be incorporation into larger mineral aggregates that are not released by our extraction protocol, provides a plausible mechanism for the observed 10Be minima.

These short-scale 10Be minima are typically confined to one or two consecutive samples representing 20 to 40 cm in depth, with only two events exceeding 60 cm. Their limited extent implies that the broader 10Be flux record remains intact, preserving the long-term geomagnetic signal. In the TALDICE core, once these short-lived minima are accounted for and removed, the 10Be record robustly mirror geomagnetic variability. Past studies had suggested to account for these outliers calculating a rolling median (Raisbeck et al., 2006). However, the median method results in similar 10Be fluxes to those based on a rolling mean average (Fig. S2). On average, the method based on minima identification results in +10 % 10Be flux compared to the median method (Fig. S2).

Reconstructing 10Be flux requires reliable estimates of both concentration and snow accumulation rate, as both parameters can vary by up to a factor of two during geomagnetic excursions or between glacial/interglacial variations, respectively. In particular, assessing snow accumulation rates across Termination III and MIS 7 is challenging, as this interval is marked by rapid and complex climatic variability (e.g., Caillon et al., 2003; Pérez-Mejías et al., 2017). This makes it essential to confirm the accuracy and precision of the accumulation reconstruction.

For the period 170–270 ka BP, the mean 3 ka rolling 10Be flux in TALDICE, including the background and the geomagnetic events, is 1.32×105 at cm-2a-1, which slightly increases to 1.40×105 at cm-2a-1 when the identified 10Be minima are excluded. For the overlapping period 170–190 ka BP, the TALDICE 10Be flux relative variations are in good agreement with the Dome Fuji (DF) record (Horiuchi et al., 2016) (Fig. 4). This coherence not only highlights the homogeneous 10Be deposition over East Antarctica, but also indicates that accumulation changes are well captured in both age models, AICC2023 for TALDICE (Bouchet et al., 2023) and the Dome Fuji DFO-2006 chronology (Kawamura et al., 2007). This agreement is further improved when the recent Dome Fuji chronology from Oyabu et al. (2022) is applied (R2=0.44, with DF values calculated on TALDICE timestep), compared to the older DFO-2006 chronology (R2=0.37). The consistent temporal evolution of 10Be fluxes across these independent ice cores supports the reliability of these datasets for investigating variations in the GDM.

Despite this coherence in temporal evolution, DF exhibits systematically higher absolute 10Be fluxes than TALDICE (≈180 %, Fig. 4). While the mean 10Be flux in TALDICE is 1.51×105 at cm⁻² a⁻¹ over 170–190 ka BP (accounting for the minima), the DF ice core shows significantly higher values, reaching 2.74×105 at cm-2a-1 (Horiuchi et al., 2016), or 2.68×105 at cm-2a-1 when recalculated with the revised Dome Fuji chronology (Oyabu et al., 2022). The TALDICE values, however, are consistent with those from other Antarctic sites, such as EDC, which reports mean fluxes of approximately 1.44×105 at cm-2a-1 for the period 200–300 ka BP (Cauquoin, 2013).

This pattern persists over the last millennium, between 965 and 1713 CE, the mean 10Be fluxes were 1.69×105 at cm-2a-1 and 2.79×105 at cm-2a-1 at Dome C and Dome F, respectively (Jouzel et al., 2026). These values may be the results of the occurrence of solar minima during this interval corresponding to enhanced 10Be production (Wolf, 1280–1350 CE; Spörer, 1420–1570 CE; Maunder 1645–1715 CE; Dalton 1790–1830 CE, Bard et al., 2000; Berggren et al., 2009; Horiuchi et al., 2008). Although no 10Be measurements are available for the last millennium at Talos Dome, comparison of recent century data shows similar mean fluxes between Talos Dome (1.49×105 at cm-2a-1, Table S1) and Dome C (1.69×105 at cm-2a-1, Jouzel et al., 2026). Independent evidence from the last millennium further indicates that Dome Fuji generally exhibits higher 10Be fluxes than other East Antarctic sites. A recent multi-site compilation and modelling study (Jouzel et al., 2026) reports that measured 10Be fluxes at Dome Fuji exceed those at EPICA Dome C by ≈70 %, consistent with earlier observations. These results suggest that persistent differences in 10Be flux over Antarctica likely reflect regional atmospheric and depositional processes, rather than artefacts related to accumulation estimates derived from age models or methodological biases.

Jouzel et al. (2026) further suggest that this persistent contrast may reflects regional atmospheric and depositional processes specific to the high-elevation interior of East Antarctica, including a transition from predominantly wet deposition north of 75° S to dry-dominated deposition south of this boundary (Horiuchi et al., 2022), as well as enhanced stratosphere-troposphere exchanges over the highest Antarctic domes. While current global aerosol-climate models remain limited in their ability to resolve such sharp spatial gradients over Antarctica (Golubenko et al., 2024; Jouzel et al., 2026; Zheng et al., 2024), these mechanisms provide a physically grounded explanation for the systematically elevated 10Be fluxes observed at Dome Fuji.

Importantly, we find no significant relationship between the Dome Fuji / TALDICE 10Be flux ratio and δD (Spearman test, p=0.46), indicating that differences in 10Be flux over Antarctica is not modulated by glacial-interglacial climate variability. This result implies that the atmospheric and depositional mechanisms responsible for the DF / TALDICE ratio either remain stable through time or compensate one another across climate states. Consequently, while absolute 10Be fluxes may differ between Antarctic sites, their relative temporal variations primarily reflect changes in cosmogenic production. This reinforces the use of Antarctic ice cores as robust records of global 10Be production (modulo hemispheric polar bias; Adolphi et al., 2023) and supports their application to reconstruct variations in the geomagnetic dipole moment.

The comparison between atmospheric 10Be fluxes from ice cores and authigenic 10Be/9Be ratios from marine sediments reveals overall agreement (Fig. S4A), supporting the reliability of both archives in recording geomagnetic events. Previous studies have highlighted the close correspondence between atmospheric and authigenic 10Be records (e.g., Czymzik et al., 2020; Horiuchi et al., 2016). For IBE, high-resolution alignment is observed between ice cores (TALDICE and Dome Fuji, see Sect. 5.3.1) and marine sediment cores (Fig. 5). In particular, MD05-2920 (Simon et al., 2016) and KR0515-PC2 (Horiuchi et al., 2016) captures the same three 10Be peaks at 174, 182, and 187 ka BP. PFE is also identified in both MD05-2920 and MD05-2930, while ME is not detected in MD05-2930, likely due to the core's lower temporal resolution (ca. 4 ka around 240 ka BP), which may not resolve the short 2–3 ka duration of the event.

A systematic time offset between the oceanic and the ice core records is observed (Figs. 5, S4A). A similar 3 to 4.5 ka offset between oceanic and Dome F ice core records was previously reported by Horiuchi et al. (2016), without explicitly identifying the dominant cause of the age offset. When comparing authigenic 10Be/9Be from core MD05-2920 to the TALDICE 10Be flux, the best correlation is obtained when the oceanic record is shifted 3 ka younger (R2=0.37, Fig. S4), which remains within the uncertainties of the marine core age model (Tachikawa et al., 2014). Because this comparison relies solely on 10Be-based proxies, it is not directly affected by magnetic lock-in processes that influence RPI records. Potential physical causes would instead involve atmospheric or oceanic transport and mixing processes. Nevertheless, does this 3 ka offset result from age model uncertainties or reflect a physical lag in the system, thereby limiting the possibility to use paleomagnetic events as chronostratigraphic horizons?

A phase shift linked to mixing processes would result in a delayed and attenuated 10Be signal. Although excursion amplitudes are consistent between the two archives (Fig. 5), this depends on complex processes that can be involved such as polar bias linked to incomplete atmospheric mixing (Adolphi et al., 2023), oceanic circulation and transport (Jeromson et al., 2025; Savranskaia et al., 2021) and bioturbation in sediments (Raisbeck et al., 1985). Nevertheless, if a physical phase shift occurred, we would expect the ice core signal to lead the oceanic record due to various oceanic mixing effects, thereby resulting in a delayed (and attenuated) marine authigenic 10Be/9Be signal.

To investigate the delay, we can examine the cross-correlation between the 10Be/9Be from core MD05-2920 and the 10Be flux from TALDICE, similar to the approach of Horiuchi et al. (2016). Such analysis (Fig. S4B) shows the highest correlation when MD05-2920 is shifted by 3 to 3.7 ka earlier. Another approach consists in examining the evolution of the delay between well-defined paleomagnetic events identified in both records (marked with stars in Fig. 5). If major changes in oceanic circulation influenced the sedimentation of oceanic 10Be, the lag would vary between glacial and interglacial periods (Savranskaia et al., 2024). However, we observe no variation in the lag across glacial–interglacial variations (Fig. S4C), indicating a consistent phase relationship between atmospheric and authigenic 10Be signals. This approach yields a mean offset of 2.3 ka (Fig. S4C). Taken together, these results suggest that the observed delay is the result of age model uncertainty in the marine core, and that 10Be production events are likely recorded synchronously in both oceanic and ice core archives. This conclusion is consistent with previous findings that suggest a limited reservoir effect and minimal climatic bias in 10Be-based chronologies (Ménabréaz et al., 2012). These findings reinforce the value of 10Be as a reliable synchronizing tool across sediment and ice archives. In the context of the Beyond EPICA project, they emphasize the role of 10Be in refining chronologies and investigating climatic transitions, particularly across complex intervals such as the Mid-Pleistocene Transition (1.2–0.9 Ma, Fischer et al., 2013; Parrenin et al., 2017; Wolff et al., 2022).

During IBE, the TALDICE 10Be flux record closely follows variations in oceanic authigenic 10Be/9Be from a global compilation (Frank et al., 1997) (Fig. 5). In particular, the timing, duration, and stepped structure of the collapse and subsequent recovery of the GDM are consistent between the ice cores and marine records (Fig. 5). For instance, the ≈7 ka plateau of elevated 10Be flux observed in TALDICE has also been observed in Dome F ice core (Fig. 5 in Horiuchi et al., 2016) and is in agreement with oceanic records (Knudsen et al., 2008).

A distinct 10Be maximum is observed at ≈232 ka in the MD05-2930 record, for which no clear counterpart is identified in other archives. To further investigate this feature, we examined the δ18O record of planktonic calcite from the same core (Regoli et al., 2015), using its recently revised chronology (Bowman et al., 2024). Interestingly, the 10Be maximum coincides with a local δ18O peak. This raises two possible interpretations: either an incorrect identification of the MIS 7.5 interval in this portion of the record, or the influence of local or regional oceanographic processes affecting both authigenic 10Be and δ18O signals, such as changes in circulation, sedimentation dynamics, or bioturbation. This observation highlights that small-amplitude or isolated features in single marine records should be interpreted with caution and cannot be systematically integrated into global stacks without independent validation. Nevertheless, the overall agreement between ice core and marine records at multi-millennial timescales indicates that such discrepancies remain limited and do not affect the main conclusions of this study. More broadly, this emphasizes the need for cautious use of stacked records as chronological constraints, particularly when relying on 10Be alone.

Because of this broad global agreement, a comparison of 10Be production calculated from RPI-based VADM and authigenic 10Be/9Be with ice core 10Be fluxes is possible. Using the VADM reconstruction PISO-1500 (Channell et al., 2009) and assuming a constant solar modulation potential of 650 MV, 10Be production was calculated following Poluianov et al. (2016). Between 170 and 270 ka BP, 10Be production was at a mean of 10.5×105 at cm-2a-1 (min = 7.4; max = 15.9). Compared with the 1.44×105 at cm-2a-1 mean flux in TALDICE (removing 10Be minima), this reveals a mean scaling factor of 7.3 (min = 5.1; max = 11.0). This scaling factor reflects the combined influence of geographic production patterns, including hemispheric asymmetries in production (Panovska et al., 2023) and atmospheric transport and deposition (Delaygue and Bard, 2011; Golubenko et al., 2024; Heikkilä et al., 2009). Consequently, quantitative interpretation of these ratios remains difficult, which highlights the need for continued model/data integration to constrain transport and deposition processes.