Ancient landscape evolution tracked through cosmogenic krypton in detrital zircon

Significance

Reconstructing the evolution of ancient landscapes is essential for understanding how Earth’s surface responds to tectonics, climate, sea-level change, and the biosphere. Cosmogenic nuclide analysis determines how long minerals have resided at and near Earth’s surface. This study demonstrates the application of cosmogenic krypton in detrital zircon, a method suited for older geological systems due to zircon’s high krypton retention and stability during weathering and diagenesis. Quantifying cosmogenic krypton reveals landscape evolution driven by tectonic uplift and sea-level change over millions of years and provides quantitative constraints on denudation and sediment-transport durations. This approach offers insights into landscape and sedimentary deposit formation and highlights the potential for extension of the timescales of cosmogenic nuclide applications.

Abstract

Cosmogenic nuclides have revolutionized our quantitative understanding of landscape evolution via measurements of near-surface erosion, exposure, and burial. Here, we integrate stable cosmogenic krypton in detrital zircon with U–Pb geochronology to extend the temporal limits of cosmogenic nuclide applications and reconstruct Eocene landscape evolution from drill cores of placer deposits in southern Australia. Zircon U–Pb crystallization ages are interpreted to reflect paleodrainage from a deeply weathered ~800,000 km² hinterland and transport via a ~1,000 km littoral drift system. The measured cosmogenic ⁷⁸Kr concentration of detrital zircon samples ranges from ~6.4 × 10⁵ to 1.8 × 10⁷ atoms per gram, suggesting low paleodenudation rates of 0.3 to 0.7 m per My [interquartile range (IQR)]. Such low denudation rates are below those expected by comparison to modern analogs and point to underestimation due to re-exposure during sediment transport and shallow storage. Expressing the concentrations as apparent exposure times, which approximate the near-surface integrated residence time, yields estimates of 0.9 to 2.1 My (IQR). In the stratigraphic and mineralogical context, the dataset records a shift from compositionally mature placers with uniformly high residence times (~1.6 My) to less mature placers with stratigraphically variable residence times (~0.7 to 2.7 My). We infer a shift from prolonged sediment storage, during which the mineral assemblage was modified, to a more dynamic transport regime with higher net transfer rates. The timing suggests eustatic and tectonic forcing, and cosmogenic krypton captures this transition, aiding reconstruction of how ancient landscapes and the sedimentary record coevolve.

Long-term landscape dynamics are closely tied with tectonic and environmental conditions (1, 2). Stratigraphic successions, along with the sediment composition they preserve, represent accessible records of Earth’s past surface processes (3, 4). While the qualitative influence of weathering on sediment composition is well established (5, 6), quantitative approaches that constrain the rates and durations of sediment transport in older sedimentary systems remain limited. This issue hampers the evaluation of how landscapes and the stratigraphic record coevolve.

In situ–produced cosmogenic nuclides provide a means to quantify Earth surface processes such as weathering and erosion (7, 8) and to track sediment transport and burial histories (9–11). Studies on modern sediments show that large portions of Earth’s surface sediment are characterized by extended transport and storage over hundreds of thousands of years (9, 12–14), with associated changes in sediment composition (15–17). For example, the weathering of less resistant minerals during sediment storage increases the relative proportion of stable phases like zircon, thereby contributing to the enrichment of surviving minerals within sediments (18, 19).

A limitation of the most commonly used cosmogenic nuclides (¹⁰Be and ²⁶Al) is their radioactive decay, which restricts their use to late Cenozoic timescales (20). Stable cosmogenic nuclides, such as ³He and ²¹Ne, can extend this range (21, 22). More recently, in situ–produced terrestrial cosmogenic krypton (Kr_cos) in zircon has been introduced as an additional approach for studying landscape evolution (23). Kr_cos offers a distinctive feature among available cosmogenic nuclide systems, as it includes both stable isotopes (e.g., ⁷⁸Kr_cos) and radioactive isotopes (e.g., ⁸¹Kr_cos), providing an internal stable/radionuclide pair allowing for single-element burial dating (23). In older systems, Kr_cos provides an opportunity to explore slowly eroding, zircon-rich settings such as cratons, which remain comparatively understudied in cosmogenic nuclide research (24). By taking advantage of the quantitative retention of Kr in zircon due to its high closure temperature of ~560 to 580 °C (25) and zircon’s stability during weathering and diagenesis (26), this approach can complement ²¹Ne and ³He applications in settings where diffusion (21) or mineral stability may limit temporal resolution. Because of these unique characteristics, Kr_cos in detrital zircon can theoretically be applied to sediments of any depositional age (23) that have not seen medium- or high-grade metamorphic overprinting. Challenges from noncosmogenic Kr components remain, but approaches developed for cosmogenic Ne, such as three-isotope diagrams, provide ways to address these issues (23, 27).

Zircon is a common accessory mineral found in igneous, metamorphic, and sedimentary rocks and a useful geochronometer due to its unique chemical and physical properties (28). A broad range of analytical techniques (e.g., U–Pb geochronology, trace element geochemistry) have been developed to study zircon. In sedimentary systems, zircon can provide a proxy for sediment provenance (26). Kr_cos complements the analytical toolkit that can be applied to this weathering-resistant mineral that has been critical for understanding our planet’s evolution (29, 30). Importantly, Kr_cos may also help address a knowledge gap in zircon research, that is, this mineral’s surface history, as most techniques commonly applied to zircon (e.g., Lu–Hf, [U-Th]/He) primarily record subsurface geological processes.

Here, we analyze the Kr isotopic composition in detrital zircon from 43 drill core samples from three placer deposits in the Eucla Basin, Southern Australia, which hosts the world’s largest active zircon mine (Jacinth-Ambrosia). To constrain sediment provenance, we integrate the cosmogenic nuclide data with detrital zircon U–Pb geochronology from 4,248 detrital zircon grains. We evaluate the significance of the Kr_cos dataset in regard to the zircon surface history, showing how the near-surface integrated residence time of the zircon grains relates to the sediment composition. Finally, we discuss how changes in past landscapes can be reconstructed through analysis of Kr_cos in a stratigraphic context, demonstrating the potential of Kr_cos to explore Earth’s landscape evolution in early Cenozoic, and potentially older, settings.

Results

Zircon Provenance and Sediment Transport.

The Ooldea barrier system in the eastern Eucla Basin (Fig. 1A) is well studied due to its global significance for zircon resources and relevance in reconstructing the region’s geologic and climatic history (5, 31–33). Within the Ooldea barrier system, the Ambrosia, Tripitaka, and Atacama deposits are the focus of this study (Dataset S1). Ambrosia and Tripitaka are middle Eocene strandline deposits, correlated with the Tortachilla Transgression (41 to 39.5 Ma), and contain low overall heavy mineral concentrations (2.7 and 2.4%) but exceptionally high zircon yields (50 and 65% of heavy mineral concentrates, respectively) (31). Conversely, the younger Atacama deposit is associated with middle-late Eocene strandlines (34) and is mineralogically distinct, with higher total heavy mineral concentration (11.3%) but lower zircon abundance (15% of heavy mineral concentrates) (31).

Fig. 1.

(A) Digital elevation model of the study area (35), showing the locations of the studied placer deposits and the Eocene strandlines associated with mineralization in the Eucla Basin (34). The dashed line indicates the inferred palaeocatchment area, and italicized names refer to geological units and geographic features discussed in this study. The location of the study area within Australia is shown in the Inset in the Bottom Left corner. MDB—Murray-Darling Basin. (B) Results of detrital zircon U–Pb geochronology, presented as cumulative age distributions of the merged concordant U–Pb ages from each deposit compared to potential source areas: Noorina palaeochannel (36), Officer Basin (37), and a putative eastern provenance using the detrital zircon U–Pb ages of the Cretaceous Ceduna Delta (38) as a proxy. Brown vertical bars highlight the main age modes prominent in the central Australian crystalline basement (e.g., refs. 39 and 40). The box plot Inset displays the results of pairwise intradeposit Kolmogorov–Smirnov tests (each sample within a deposit compared against all other samples from the same deposit) as a measure of intradeposit variability in the detrital zircon age spectra. Higher KS-D values correspond to greater dissimilarity (41). Box shows the interquartile range and the central line represents the median. Whiskers show the largest and smallest values above and below the first and third quartiles, respectively, that is within 1.5 times the interquartile range.

To contextualize the cosmogenic nuclide data, it is necessary to first establish the prevailing palaeodrainage system. Detrital zircon U–Pb ages record crystallization associated with high-temperature metamorphic events in their source regions (26, 28, 34, 42), which in turn allows reconstruction of sediment source(s) and thus routing systems (42). We used a rapid ablation U–Pb dating method (43), generating 3,877 concordant U–Pb ages (Fig. 1B and SI Appendix, Fig. S1 and Dataset S2).

Detrital zircon U–Pb ages across the different deposits can be correlated to crystalline and sedimentary sources within the palaeodrainage of the Eucla Basin (5, 31, 34). These sources include the Archean Yilgarn (44) and Gawler (45) cratons [2.7 to 2.6 billion years ago (Ga)] and the Meso-Neoproterozoic Musgrave, Madura, and Coompana provinces [1.2 to 1.0 and 0.6 to 0.5 Ga (39, 40)]. Age components at 1.9 to 1.5 and 0.45 to 0.35 Ga can be linked to the central Australian Proterozoic Arunta Orogen (46) and the Alice Springs Orogeny (47), respectively, despite crystalline sources of these ages not cropping out in the catchment area. Instead, matching zircon ages are found in sedimentary units of the Officer Basin and Noorina palaeochannel (Fig. 1), which suggests that intermediate sediment storage acted as a source of detrital zircon to the Eucla Basin palaeoshorelines. Contributions from proximal eastern sources appear minor, as indicated by comparatively distinct U–Pb zircon age spectra from the Eocene Garford palaeochannel and larger sediment routing systems that drained the eastern part of the Australian continent (e.g., Cretaceous Ceduna Delta, Fig. 1) (38, 48).

Our zircon U–Pb dating results show limited interdeposit and intradeposit downhole variability. Quantitatively, low variability among all samples is supported by the Kolmogorov–Smirnov D statistic (KS-D), which measures differences in the relative abundance of age modes and ranges from 0 to 1 (41, 49). The median KS-D for all pairwise comparisons is 0.15 ± 0.07 (1SD), corresponding to 15% of the maximum possible variability (a value of 1 would indicate no overlap between any pair of age populations). Individually, Atacama sample U–Pb age populations appear slightly more dispersed (SI Appendix, Fig. S1), with an intradeposit KS-D of 0.14 ± 0.07, compared with 0.12 ± 0.03 for Ambrosia and 0.14 ± 0.04 for Tripitaka (Fig. 1B). These subtle differences are evaluated by a pairwise Fligner–Killeen test (50), which shows that Atacama is significantly more variable than Ambrosia (P = 3 × 10⁻⁵), whereas Tripitaka and Ambrosia do not differ significantly. Regardless, the observed KS-D values are broadly consistent with the typical range of intraformational detrital zircon datasets [e.g., ~0.1 in the Book Cliffs, UT; ref. 51] and suggest stability of the prevailing drainage systems during the time of deposition.

The provenance patterns are broadly consistent with previous studies (32–34, 52) in that sediment appears to be sourced from an extensive radial palaeodrainage system, covering an area greater than 800,000 km². This drainage system consists of multiple smaller catchments, the sediment of which is mixed with preexisting sedimentary material from intermediate storage sites, and subsequently transported in a ~1,000 km long west-to-east longshore drift system (Fig. 1A) that ultimately concentrated heavy minerals in the east of the basin (5, 32–34).

Krypton Isotope Signature in Detrital Zircon.

Kr isotopes were measured on the same drill core samples analyzed for zircon U–Pb geochronology. Because the analyses were performed on bulk zircon separates, the results represent integrated signals of the sampled zircon populations. Under ideal conditions, Kr_cos is expected to follow an air–cosmogenic mixing trend (23), which provides the reference against which deviations are assessed (Fig. 2A).

Fig. 2.

(A) Three-isotope diagram illustrating the processes that produced the Kr in the detrital zircon of the placer deposits (sensu ref. 23). Uncertainties are shown as 1 SD. mfl—mass fractionation line. (B) Downcore pattern of ⁷⁸Kr_cos across the different deposits (using the same symbology as in figure part A), revealing variability among the deposits.

Many of the samples deviate from the expected air–cosmogenic mixing relationship (Fig. 2A and Dataset S3). The scatter toward higher ⁸⁰Kr/⁸²Kr ratios is consistent with the presence of nucleogenic Kr (Kr_nuc), produced via thermal neutron capture by Br (23) over the life-time of the zircon grains, which affects only ⁸⁰Kr and ⁸²Kr but not ⁷⁸Kr due to the absence of suitable natural target isotopes of Br (53). Several data points fall below the air–cosmogenic trend. Fissiogenic ⁸²Kr is a possible but unlikely explanation for this deviation, as the fission yields required would be far higher than the actual yields (54). We therefore consider these data points as experimental outliers and exclude them from further evaluation. In all other cases, Kr_nuc is an interfering Kr component that can be accounted for in the determination of the Kr_cos component.

The geological meaning of the origin of Kr_nuc remains cryptic. Zircon and most silicates have negligible Br concentrations and previous work on zircon megacrysts found rare evidence for Kr_nuc (23). Elevated Br concentrations necessary to generate ^80,82Kr_nuc are found in rock types like carbonatites and kimberlites (55), which are volumetrically minor in the rock record and can be excluded as a significant source given the size of the palaeodrainage (Fig. 1A). A more plausible explanation for the observed Kr_nuc, particularly pronounced in the Atacama deposit, is the presence of halogen-bearing inclusions such as apatite, which are common in zircon (56). This interpretation is consistent with experimental results showing that halogen-bearing inclusions can significantly contribute to the nucleogenic noble gas budget of zircon (25).

Among the Kr_cos isotopes, ⁷⁸Kr_cos is particularly relevant for applications to older systems because it is stable and produced solely by spallation reactions (23). The interference-corrected concentration of ⁷⁸Kr_cos ranges from ~6.4 × 10⁵ to 1.8 × 10⁷ atoms g⁻¹ (Fig. 2B) across all deposits. Median values are similar, at 5.7, 6.8, and 6.7 × 10⁶ atoms g⁻¹ for Ambrosia, Tripitaka, and Atacama, respectively, with no significant differences indicated by a Kruskal–Wallis test for equal medians (H = 1.5, P = 0.46). In contrast, the distributions of ⁷⁸Kr_cos concentrations across the deposits differ in their variability (Fig. 2B). Ambrosia and Tripitaka show a unimodal appearance and limited dispersion (1SD of 1.7 and 1.0 × 10⁶ atoms g⁻¹), whereas Atacama displays a broader, more dispersed distribution (1SD of 4.8 × 10⁶ atoms g⁻¹). This difference is statistically supported by a Fligner–Killeen test (χ² =10.10, P = 0.0064), with Atacama’s variance ∼7 times greater than Ambrosia’s and ∼22 times greater than Tripitaka’s, whereas Ambrosia and Tripitaka do not differ significantly (P = 0.35). Given these substantial variance differences, the result is unlikely to arise from the variable sample sizes alone.

Discussion

Does Cosmogenic Krypton Reflect Erosion in the Source Area or Transport Modification?

A central question is whether Kr_cos primarily reflects accumulation during bedrock erosion or modification during sediment transport and storage. In the case of ⁷⁸Kr_cos, muon-induced production of ⁷⁸Kr is negligible since negative-muon capture on Zr has an extremely low reaction probability, equivalent to <0.01% of total near-surface production (23). Thus, spallogenic production dominates and significant accumulation of ⁷⁸Kr_cos begins during bedrock exhumation at depths of ~2 m (7). In cases of rapid transport and deposition that shields sediment from re-exposure, concentrations of cosmogenic nuclides are dominated by those acquired during exhumation prior to erosion from bedrock (57). Conversely, in landscapes characterized by tectonic quiescence, low topographic relief, and large catchments (e.g., the sampled Eocene landscape of central Australia), re-exposure during transport and shallow temporary storage can contribute substantially to the accumulated cosmogenic nuclides (57–59).

One way to contextualize the ⁷⁸Kr_cos concentrations is to calculate apparent catchment-wide palaeodenudation rates, which for the samples analyzed here range from 3.8 ± 1.1 to 0.14 ± 0.03 m My⁻¹ (1SD). Such denudation rates are among the lowest known globally when compared to presently eroding surfaces (SI Appendix, Fig. S2). For comparison, the median of these rates (0.38 m My⁻¹) is only slightly higher than values determined for the McMurdo Dry Valleys in Antarctica (~0.19 m My⁻¹; ref. 60), the Atacama Desert in Chile (~0.1 m My⁻¹; ref. 2), and the Singida diatreme field in Tanzania (<0.1 m My⁻¹; ref. 23). In Australia, similarly low denudation rates occur in the (semi-)arid regions, such as the granite surfaces on the proximal Eyre Peninsula (Fig. 1A) that erode at 0.55 to 1.70 m My⁻¹ (IQR) (61). Such constraints overlap with the Kr_cos-derived denudation rates (IQR = 0.29 to 0.65 m My⁻¹) and could hint at stable long-term denudation of the central-southern Australian landscape. Nonetheless, the considerable offset to lower denudation rates is unexpected given the humid conditions of the Eocene (62, 63), which contrast with today’s aridity, but is consistent with additional accumulation of ⁷⁸Kr_cos through re-exposure of zircon grains after release from bedrock.

Comparison with analogous settings allows for a first-order estimate of how much ⁷⁸Kr_cos was acquired during bedrock erosion versus during transport and storage. An average denudation rate of 1 m My⁻¹ for cratonic shields (7), as an approximation for denudation rates experienced in the headwaters of the Eucla Basin, yields a hypothetical ⁷⁸Kr_cos concentration of ~2.5 × 10⁶ atoms g⁻¹. Compared with the measured median ⁷⁸Kr_cos concentration (~6.6 × 10⁶), this corresponds to an accumulation of about 38% of ⁷⁸Kr_cos during bedrock erosion and the remainder during sedimentary transport and storage.

A more suitable analog for the humid, vegetated Eocene landscape is the Murray–Darling Basin in eastern Australia (Fig. 1A). This catchment is of similar size (~1,000,000 km²) to the inferred Eucla paleodrainage (~800,000 km²) and records catchment-wide denudation rates near its outlet of 3.40 −0.39/+0.53 m My⁻¹ (95% CI; ref. 12). If taken as a meaningful source signal, this would imply only 10 to 13% of the measured ⁷⁸Kr_cos derives from bedrock and hillslope erosion. Despite the arid setting, the smaller Finke, Macumba, and Neales rivers (Fig. 1A) can provide additional insights because they partly erode the same source lithologies as the Eocene Eucla Basin. While the lowest denudation rates match those of Eocene placer deposits (SI Appendix, Fig. S2), the dominant rates of 0.91 to 4.36 m My⁻¹ (IQR) correspond to an interpreted 42 to 9% of the ⁷⁸Kr_cos signal reflecting bedrock erosion, implying that even in these cases, a substantial proportion of the signal may derive from near-surface exposure elsewhere during sediment transport and temporary storage.

The comparison with modern analogs indicates that there is a considerable chance that much of the ⁷⁸Kr_cos was acquired during protracted transport and storage, rather than solely during bedrock erosion. This is consistent with the transport history inferred from the provenance constraints which suggest >100 km of continental sediment routing with potential for fluvial storage, and up to 1,000 km of littoral drift with ample opportunity for re-exposure in paralic and coastal barrier settings. Similar evidence comes from the North Platte River in Nebraska, where cosmogenic ²¹Ne measurements in fluvial sediments are incompatible with a pure bedrock-erosion signal but instead point to substantial cosmogenic production of stable ²¹Ne during shallow storage and transport (59).

An alternative explanation for elevated ⁷⁸Kr_cos concentrations may be inheritance from older sedimentary cycles, for example, through reworking of sandstones. These two scenarios (surface accumulation vs. inherited signals) could be distinguished in modern systems: a progressive increase in downstream ⁷⁸Kr_cos at constant zircon provenance would support accumulation during surface residence, whereas uniform concentrations combined with provenance heterogeneity would point to inheritance. Regardless, both possibilities imply a significant bias in the denudation rates calculated from ⁷⁸Kr_cos and highlight the complex and prolonged surface histories of zircon grains that Kr_cos can help to resolve in older sedimentary systems, similar to how cosmogenic ²⁶Al/¹⁰Be records burial and re-exposure in modern drainage systems (13).

A Million-Year Sediment Transport History.

⁷⁸Kr_cos concentrations can also be considered in terms of calculated apparent exposure times, where the cosmogenic nuclide concentrations inform on the near-surface integrated residence time of zircon grains since exposure near Earth’s surface (57). Framing the results this way also more appropriately accounts for potentially inherited ⁷⁸Kr_cos, which is suggested by provenance constraints that include recycling from the Officer Basin (Fig. 1B). Ambrosia and Tripitaka deposits have consistent apparent exposure times with a main mode of ~1.6 My (Fig. 3A). In contrast, the ⁷⁸Kr_cos concentrations of the Atacama deposit define three modes at ~0.7, 1.9, and 2.7 My (Fig. 3A).

Fig. 3.

Results of the cosmogenic Kr (Kr_cos) determination in the context of the mineralogy of the investigated samples. (A) Kernel density estimates for the combined middle Eocene deposits (Atacama and Tripitaka) and the middle–late Eocene Atacama deposit, illustrating distinct distributions (bandwidth = 0.24, equivalent to the mean uncertainty). (B) Downhole variability of Kr_cos is presented as apparent exposure time and heavy mineral concentration (wt%) for individual samples. Pie charts summarize key mineralogical parameters at the deposit scale: heavy mineral concentration (% HM, pie chart area proportional to % HM) and the ilmenite-to-zircon modal abundance ratio (31). ilm—ilmenite; zrc—zircon. (C) Scatterplot comparing apparent exposure time and heavy mineral concentration. Inset text lists results of a Spearman’s rank correlation test between the two variables, showing Spearman’s rho (ρ) and the associated P-value. Note that only Atacama deposit shows a significant (P < 0.05) correlation.

Such million-year near-surface residence times of zircon are consistent with burial and exposure estimates from other cosmogenic systems. Although the radioactive ²⁶Al/¹⁰Be system records subsurface burial (i.e., decay during shielding from cosmic rays) rather than surface exposure traced by stable cosmogenic nuclides, it provides complementary constraints that ultimately relate to the residence time of sediment within the landscape (12–14). About one-third of global fluvial systems show >500 ky of sediment subsurface storage based on the discordance of ²⁶Al/¹⁰Be denudation rates (9, 13). In the aforementioned Murray–Darling Basin, considered a suitable modern analog for the Eocene Eucla system, ²⁶Al/¹⁰Be discordance near the basin outlet indicates storage of 0.8 to 1.5 My (95% CI; ref. 12). Higher minimum burial estimates are reported from rivers such as the Tiber and Lena with ~2.5 and 3.2 My, respectively (9), whereas stable cosmogenic ³He analyses of alluvial metal and alloy grains yield surface residence times of up to ~7 My (22).

The long apparent exposure times inferred from ⁷⁸Kr_cos imply complex grain histories and increased potential for modification of sediment composition through surface weathering, which is consistent with independent geological evidence. Comparing deposit-scale mineral composition shows that Ambrosia and Tripitaka are compositionally mature with high zircon concentrations, while Atacama is less mature (less zircon, but more labile ilmenite) (Fig. 3B). Moreover, in the Atacama deposit, apparent exposure times are negatively correlated with heavy mineral concentration (Spearman’s ρ = –0.60, P = 0.002), whereas no significant correlations are observed in Ambrosia or Tripitaka (P > 0.05; Fig. 3 B and C).

Elevated Kr_nuc in the Atacama deposit also indicates less intense weathering. This is because Kr_nuc is consistent with a higher abundance of apatite inclusions. Mineral grains with inclusions are more susceptible to chemical weathering and attrition during storage and transport (64), as inclusions can induce anisotropic stress (65) and thus promote fracturing, as confirmed by leaching experiments of zircon (66). This interpretation does not imply a different source, as the U–Pb age spectra are similar across all deposits, but instead reflects differential preservation controlled by the zircons’ surface history. The preferential survival of such inclusion-bearing zircon grains in the Atacama sediment is therefore consistent with lower degrees of weathering during sediment storage if the relationship of Kr_nuc and the inclusion inventory of zircon is correct. Greater intradeposit homogeneity of the zircon U–Pb age populations of samples within Ambrosia and Tripitaka (Fig. 1B) is also consistent with more protracted reworking, and ultimate concentration from a well-mixed heavy mineral reservoir.

The combined observations reveal differences between the older (Ambrosia and Tripitaka) and younger (Atacama) shoreline deposits in maturity and Kr_cos composition. These differences suggest that enrichment of resistant minerals such as zircon is linked to their predepositional surface history. Modification of sediment composition can occur during sediment transport and storage at near-surface conditions (17), for example, through chemical weathering in floodplains, paleochannels, and coastal barriers and can be extensive with removal of less weathering-resistant minerals on the 10⁵ to 10⁶ years timescales (15). To understand the conditions of sediment generation, transport, and storage that formed the deposits, it is necessary to place them into their broader geological context.

A Window Into Ancient Landscapes.

The tectonic and climatic history of southern Australia provides the framework for understanding the Kr_cos data. Rifting between Australia and Antarctica began around 90 Ma, transforming southern Australia into a passive margin and influencing the region’s subsequent landscape evolution (67, 68). During the early Cenozoic hothouse climate, the region experienced tectonic quiescence, leading to intense chemical weathering and the formation of a deep regolith mantle (69) (Fig. 4A). In the middle Eocene, basin rejuvenation initiated erosion of this deeply weathered landscape (5, 31). This landscape would have represented a substrate enriched in ⁷⁸Kr_cos and stable minerals due to the breakdown of less resistant minerals in response to extensive weathering and sediment recycling (19).

Fig. 4.

(A) Eocene history of the Eucla Basin in the context of regional geological events. The deposition of the Ambrosia and Tripitaka deposits (indicated by the positions of their respective symbols, with black lines denoting the uncertainties in depositional age) occurred following a protracted period of deep weathering in their inferred palaeocatchment areas during the prolonged Tortachilla transgression, which is coeval with the onset of accelerated seafloor spreading and increased sedimentation in the study area (5). This spreading was followed by a period of more dynamic sea-level fluctuations, including the Tuketja-Tuit transgression, during which the Atacama deposit formed. Tilting/uplift and offshore hiatuses from ref. 70. (B) Schematic model of signal propagation and modification during sediment generation, transport, and storage. The sediment composition is progressively modified toward a more stable composition (less labile phases), with rates dependent on the climatic and tectonic boundary conditions of the drainage system. This modification is captured by the concentration and distribution of ⁷⁸Kr_cos in detrital zircon, where high concentration reflects high apparent exposure times (and low denudation rates) and vice versa.

During the Tortachilla transgression, a prolonged sea-level highstand facilitated the development of the Ooldea barrier system and the formation of the Ambrosia and Tripitaka deposits (Fig. 4). The longevity of the highstand allowed sediment storage and mixing across the coastal plain and the shelf while reducing offshore transport (71). Combined with the invariant long apparent exposure times (~1.6 My) in the mature deposits and the small intradeposit variation in the detrital zircon U–Pb ages (KS-D Inset in Fig. 1B and SI Appendix, Fig. S1), these findings suggest extended in situ reworking in coastal environments is a probable mechanism to explain both the mineralogical maturity and the well-mixed, high-concentration Kr_cos signal (Figs. 2B and 4).

In contrast to the downcore invariance in the Ambrosia and Tripitaka deposits, the more variable downcore pattern of apparent exposure times in the Atacama deposit (Fig. 3B) suggests temporal variation in the surface history of its sediment. This interpretation is supported by the Fligner–Killeen test on the ⁷⁸Kr_cos concentrations, which confirms a significant difference in variance between the deposits (Results). Such observations are consistent with increased sea level fluctuation during the subsequent Tuketja-Tuit transgression (i.e., during Atacama deposition, Fig. 4A) and increased river incision (72), together implying an allogenic control (i.e., sea level and uplift) on the denudational activity in the hinterland of the deposits. This interpretation is further supported through the highest intradeposit variability in the detrital zircon U–Pb age distributions of the Atacama deposit (KS-D Inset in Fig. 1B).

The Atacama deposit, with its increased downhole variability in zircon U–Pb ages and the first appearance of lower apparent exposure times of these zircon (Fig. 3), captures an inflection point in landscape dynamics (Fig. 4B). This shift is characterized by episodic pulses of relatively more direct sediment supply, with less weathered sediment composition (i.e., lower apparent exposure times) and higher proportions of labile phases (i.e., higher heavy mineral concentration). These sediment pulses appear to interrupt the homogenized and mature sediment supply that dominated in the middle Eocene (Fig. 4B). This interpretation is consistent with the statistically significant correlation of heavy mineral concentration and apparent exposure times in the middle–late Eocene (Fig. 3C), which resembles a typical maturity trend, i.e., decreasing heavy mineral concentration with increasing weathering (15, 73). In the context of the landscape evolution, we interpret this change as reflecting reduced coastal storage and reworking during the middle-late Eocene and a more effective propagation of source signals in the transfer zone.

Overall, the analysis of Kr_cos in a stratigraphic context, along with detrital geochronology and mineralogy, suggests that modification of sediment composition and thus the alteration of the source (hinterland) denudation signal can be attributed to extended sediment transport and storage (>1.5 My), primarily driven by environmental conditions. Surface weathering, sea level changes, and uplift are recognized as key controls on sediment transport and signal propagation, and Kr_cos provides a stable cosmogenic nuclide approach that can quantify the magnitude of such effects in periods where short-lived cosmogenic radionuclides such as ¹⁰Be and ²⁶Al are inapplicable. Ultimately, Kr_cos in detrital zircon enhances this mineral’s versatility to address geological questions by filling a knowledge gap in our understanding of zircon’s surface history, offering complementary opportunities to evaluate Earth’s long-term landscape evolution.

Materials and Methods

Sample Preparation.

Detrital zircon grains were concentrated using a combination of heavy liquid separation at 2.9 g cm⁻³ and a magnetic separator (S.G. Frantz, United States). A representative split of each detrital zircon concentrate was taken using an ASC Scientific Sample Microsplitter (United States). The split of grains was utilized for U–Pb geochronology and purity testing. For U–Pb analysis, the grains were bulk-mounted on double-sided tape and subsequently embedded in epoxy resin. The mounted grains were ground to approximate half-grain thickness and then polished. Grain mounts were carbon coated and underwent automated mineral identification using energy-dispersive X-ray spectrometry (EDS) and backscattered electron imaging using a TESCAN Integrated Mineral Analyzer (TESCAN, Czech Republic). Results suggest a purity of zircon concentrations of ~96 to 99 wt%. Nonzircon grains are mainly quartz and aluminosilicates, i.e., minerals that do not contain a significant abundance of target elements for production of Kr_cos (Rb, Sr, Y, Zr, and Nb; ref. 23). As such, contaminants diluted the Kr signal but did not contribute to the Kr budget.

Cosmogenic Krypton Determination.

Kr_cos measurements were performed using the facilities and Kr measurement methodology in place at the University of Cologne as detailed by Dunai et al. (23). We use a sea-level high latitude (SLHL) production rate of ⁷⁸Kr_cos of 3.26 atoms g⁻¹ y⁻¹ derived from the reported ⁸¹Kr_cos production rate and the ⁸¹Kr/⁷⁸Kr ratio (23).⁷⁸Kr_cos concentrations are calculated from the horizontal distance between samples and the tie-line between atmospheric and nucleogenic endmembers (23) in ⁷⁸Kr/⁸²Kr vs. ⁸⁰Kr/⁸²Kr space (Fig. 2A), and ⁸²Kr abundances (Dataset S3). For each deposit, duplicates for one sample were measured, and results were averaged for statistical purposes. Production rates were scaled to a palaeolocation equivalent to a present-day latitude and longitude of 28 °S and 128 °E, respectively, and an elevation of 400 m. Latitude and longitude approximations correspond to the present-day central Eucla Basin, and the elevation corresponds to the median elevation of the palaeocatchment (Fig. 1A) when transposed onto a palaeoDEM at 40 Ma (74).

Scaling of production rates used the SPRITE model (75) that has been designed to obtain more meaningful cosmogenic nuclide production rates for pre-Quaternary samples. Using the aforementioned location and elevation constraints, a sample density of 2.65 g cm⁻³ and the atmospheric conversion of Valdes et al. (76), SPRITE returns scaling factors per 0.25 My bin, from 0 to 70 Ma. SPRITE scaling factors are influenced by atmospheric depth and cutoff rigidity (75) and are independent of the nuclide-host mineral production rate, allowing the resultant scaling factors to be applied to the ⁷⁸Kr_cos SLHL production rate. The apparent exposure times of each sample was determined by summing the amount of ⁷⁸Kr_cos produced per 0.25 Myr bin, until the measured cosmogenic nuclide composition of the sample was reached. For this purpose, 41 Ma (for Ambrosia and Tripitaka) and 39 Ma (for Atacama) were used as the time of sediment burial and the cessation of nuclide accumulation, thus scaling factors from 39 or 41 to 70 Ma, at intervals of 0.25 My, were used to calculate the apparent exposure times of each sample.

The production rate scaling factor was calculated using the simplifying assumptions of no self-shielding and no topographic shielding. While less accurate, these assumptions are necessitated by unknowns of topography and depth of intermediate burial episodes over the time periods considered here. The implication of making such an assumption is that the calculated apparent exposure times represent minimum constraints, as they do not account for factors such as topographic shielding or deep burial, which would increase these estimates or lower the denudation rates. Computation of kernel density estimates and determination of the most abundant modes of the apparent exposure times (Fig. 3A) were performed using the R package IsoplotR (77).

Zircon U–Pb Geochronology.

Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) was performed at the John de Laeter Centre, Curtin University, Australia. We employed a “rapid ablation” approach (e.g., refs. 43 and 78) and applied a “blind-dating strategy” (79), i.e., the center of randomly selected zircon grains, which were identified using automated phase identification (Sample Preparation), were targeted. Laser ablation used a RESOlution LR 193 nm ArF excimer laser with a laser fluence of 3 J cm⁻² and repetition rate of 20 Hz for ~6 s of analysis time and 20 s of background capture. No cleaning pulses preceded the analyses. The sample cell was flushed by ultrahigh purity He (350 mL min⁻¹) and N₂ (3.8 mL min⁻¹) and ablations were 38 μm circular spots. U–Pb isotopic analysis was conducted using an Agilent 7900 Quadrupole ICP-MS with high-purity Ar as the carrier gas (flow rate 0.98 L min⁻¹). Data reduction was performed using the U–Pb geochronology data reduction scheme in iolite4 (80) and a linear model to correct for downhole fractionation. GJ1 zircon (81) was used as the primary reference material. Secondary reference materials 91500 (82), Plešovice (83), and Maniitsoq (84) were measured as unknowns to monitor precision and accuracy. No common Pb correction was performed. Age calculations were performed using the R package IsoplotR (77). Data were filtered for discordance using a −10 to 10% concordia distance threshold. Results of LA-ICP-MS analysis including unknowns and reference materials are tabulated in Dataset S2.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Data, Materials, and Software Availability

Code data have been deposited in Zenodo (85). All other data are included in the manuscript and/or supporting information.