Origin of Contessa Road magnetic minerals
The Contessa Road section contains magnetic minerals that have been interpreted to be mainly produced in terrestrial environments32,50,51,56. However, diverse authigenic processes have also been suggested as possible magnetic mineral formation mechanisms within the Scaglia Rossa Formation44,48,57. The Cretaceous sequence of these carbonates has alternating reddish and whitish colours that indicate haematite presence or absence, respectively57. This pattern was used to propose that Scaglia Rossa haematite was a diagenetic product of a goethite precursor57. Haematite and maghemite were interpreted to form during oxic periods, while time intervals with reducing conditions inhibited formation of these minerals57. Abrajevitch et al.48 suggested a similar haematite formation mechanism but at the expense of Fe-rich oxyhydroxides within Scaglia Rossa aeolian dust fractions. A reducing environment implies preferential magnetite and maghemite dissolution over haematite48,57,58, as identified at the Cretaceous/Palaeogene (K/Pg) boundary in whitish Scaglia carbonates48. Our whitish EELMZ carbonates contain goethite and maghemite and lack pigmentary haematite (Figs. 3f, 5c), which is incompatible with oxic and reducing condition alternations as maghemite/haematite formation mechanisms57. Preferential dissolution of low coercivity minerals is also possible due to organic matter degradation and microbial oxidant consumption in organic-rich sediments58. Contessa Road fossil and sedimentary assemblages do not provide evidence of significant organic-rich sediments inputs following the K/Pg boundary44. Furthermore, detailed rock magnetic measurements (e.g., first-order reversal curves) from Palaeogene Contessa Road reddish carbonates do not contain reductive diagenesis signatures (e.g., magnetic Fe sulphide occurrences44,58). Therefore, we interpret LPEE Contessa Road haematite and maghemite to be mainly of detrital origin.
Given the present-day Contessa Road location, post-depositional alteration of Fe-rich minerals could also be a potential maghemite and haematite formation mechanism59. If this were the case, then increased magnetic mineral contents would have resulted in haematite-rich magnetic mineral fractions due to the greater availability of material for oxidation55. Such a pattern contrasts with the long-term compositional change and magnetic mineral content reduction at Contessa Road (Fig. 5), which implies that mineral alteration is not a significant magnetic mineral formation process there and supports our argument for a detrital origin of Contessa Road magnetic minerals.
A detrital origin of Contessa Road haematite and maghemite suggests the occurrence of a widely recognised ferrihydrite→maghemite→haematite transformation in western Tethyan continental areas55,56. Ferrihydrite (Fe5HO8·4H2O) is a common iron-oxide precursor, and its occurrence implies that wet periods with increased physical/chemical weathering generated erodible material56. Gradual ferrihydrite dehydration produces maghemite and ultimately haematite in dry settings55. Thus, magnetic minerals at Contessa Road reveal a dynamic LPEE proto-Mediterranean hydroclimate that consists of a dry setting with intermittent wetter conditions9,28,30,31,32,33,34,35.
A few LPEE Contessa Road samples also contain magnetite. This mineral may have biological and detrital origins in the Scaglia Rossa Formation44,48. Early Palaeogene Contessa Road carbonates have been interpreted to contain mainly biogenic magnetite, while the upper Cretaceous Scaglia Rossa record contains biogenic and detrital magnetite in similar proportions44,48. Contessa Road biogenic magnetite occurrences have been inferred only from low coercivity central ridges in first-order reversal curve diagrams and from isothermal remanent magnetisation acquisition curves with ~20-50 mT coercivity components44,48. However, these features are also compatible with the presence of detrital magnetite and maghemite60. Early Palaeogene isothermal remanent magnetisation acquisition curves differ from those of the LPEE interval, and even if biogenic magnetite is present in the LPEE Contessa Road record, its contribution would be reduced compared to detrital magnetite (Figs. S10–S12; see Supplementary information). Furthermore, strong evidence for biogenic magnetite with chain or partially collapsed chain arrangement (i.e., from transmission electron microscope imaging60,61) is lacking for the Contessa Road section. Therefore, we interpret magnetite within the LPEE Contessa Road record to mainly have a terrigenous origin. This suggests that the Contessa Road rock magnetic record reflects primarily detrital magnetic mineral contents and compositional variations43,44,48,49.
Biogenic versus detrital controls on Contessa Road sediment accumulation
Despite the terrigenous origin of magnetic minerals, biogenic CaCO3 deposition/dissolution have been suggested as the main controlling mechanisms of Contessa Road sediment accumulation21,24,43,44,49,62. Here, we assess CaCO3 dissolution changes at Contessa Road with principal component analysis (PCA; see Methods) of Ca and detrital elements (referred to here as PCAdis). PCAdis reveals a major component that accounts for 82% of data variance (PC1dis) and a second component that only accounts for 8% of data variance (PC2dis). PC1dis-PC2dis relationships confirm major CaCO3 dissolution controls on Contessa Road sediment accumulation with negative correlations between Ca and detrital element loadings (Fig. 6a, S13a, Tab. S2), and with coincidence between PC1dis score increases and CaCO3 decreases21,24,43,44,49,62 (Fig. 6c, d, S13a, Tab. S2). PCA of XRF elements, CFB magnetic concentration parameters, Bcr, and S-ratio (referred to here as PCAall) yields a PC1all (53% of data variance; Fig. 6b). PC1all includes positive detrital element loadings that anti-correlate with Ca and correlate poorly with the negative Bcr loading, and with positive S-ratio and CFB magnetic mineral concentration parameter loadings. S-ratio and CFB magnetic mineral concentration parameter loadings anti-correlate with Bcr and have weak correlations with Ca (Fig. 6b, S13b, Tab. S2). These patterns suggest that positive PC1all scores reveal enhanced CaCO3 dissolution accompanied by high magnetic mineral inputs with increased low coercivity mineral contents (i.e., magnetite, maghemite) but reduced high coercivity fractions (Fig. 6e).
Fig. 6: Principal component analysis (PCA).
a PC1dis- PC2dis plot. b PC1all-PC2all plot. PCAall includes all carbonate free basis (CFB) magnetic mineral concentration parameters (see Supplementary information). Bulk-rock (BR) magnetic mineral concentration parameters were not used. In b, CFB labels were removed for clarity. HCC and LCC correspond to the unmixed high coercivity component and the most significant low coercivity component 1, respectively (see Fig. S10; Supplementary information). c Ca and CaCO3 records. CaCO3 is presented in terms of mean (black) ± 2 standard errors (2SE; grey shaded bands)19. d PC1dis, e PC1all, f PC2all scores. Hyperthermals (Palaeocene-Eocene Thermal Maximum (PETM), Eocene Thermal Maximum (ETM) 2 and ETM 3), smaller carbon cycle perturbations, the early Eocene terrigenous zone (EETZ), and the early Eocene low magnetisation zone (EELMZ) are indicated with orange, purple, green and yellow bands, respectively. The dashed blue line in (d), (e), and (f) indicates 0 in the y-axis. Lithology is presented at the bottom of the figure with reddish (pink) and whitish limestones (grey) and marls (red).
In contrast to PCAdis, PCAall reveals a significant second PCA component (PC2all) that accounts for 23% of data variance (Fig. 6b). PC2all has positive loadings for Bcr and detrital elements, and negative loadings for S-ratio, CFB magnetic mineral concentration parameter, and Ca (Fig. 6b, S13b, Tab. S2). These correlations suggest that PC2all positive scores indicate enhanced CaCO3 dissolution accompanied by reduced magnetic mineral inputs; however, magnetic mineral fractions isolated in PC2all are compositionally dominated by high coercivity minerals (Fig. 6f). More pronounced PC2all peaks with respect to PC1all across short-lived carbon cycle perturbations suggest that CaCO3 dissolution was preferentially coincident with haematite enrichments (Fig. 6e, f), which is confirmed by S-ratio and Bcr values (Fig. 5d, e). In contrast to positive values, PC2all negative scores can be associated with enhanced CaCO3 preservation and high terrigenous material inputs enriched in low coercivity minerals. Pronounced negative PC2all peaks are identified exclusively in intervals of good CaCO3 preservation during the EETZ (Fig. 6e, S14), which contains the highest detrital inputs across the LPEE Contessa Road record according to the CFB magnetic mineral concentration parameters (Fig. 5a–c). This pattern reveals that CaCO3 deposition was not disturbed even when detrital inputs increased substantially (Figs. 5a–c, 6e, S14); therefore, we infer that terrigenous dilution was insignificant and did not play an important role in Contessa Road sediment accumulation19,21. This is consistent with calcareous nannofossil assemblages and sedimentation rate variations, which indicate that, although Contessa Road had terrigenous inputs from western Tethyan continental regions42,49, CaCO3 deposition and dissolution were prevailing drivers of sediment accumulation there21,24,43,44,62.
Orbitally forced LPEE hydroclimate variability in the western Tethys
Orbitally forced LPEE carbon cycle and temperature changes have been suggested as drivers of CaCO3 sedimentation and hydroclimate variability at diverse locations15,16,22. Drier/wetter variations in LPEE western Tethyan continental areas, as inferred from the Contessa Road magnetic mineral assemblage, may be similar to those of well-documented Neogene-Quaternary Mediterranean regions, where orbitally driven mechanisms controlled dry/wet variability47,63,64. Specifically, orbital controls on Contessa Road sedimentation are indicated by enhanced CaCO3 dissolution during long eccentricity, short eccentricity, and precession maxima21,24. We confirm the presence of those orbitally controlled CaCO3 dissolution cycles with statistically significant spectral peaks (>90-95% confidence levels) at short eccentricity and precession periods in PC1dis (Fig. 7, S15; see Methods). Similar spectral peaks in HIRM-CFB also reveal orbital controls on haematite deposition (Figs. 7, 8, S15). Squared coherency spectra of PC1dis and HIRM-CFB, with significant peaks above false-alarm levels (α = 0.05 and 0.10; Fig. 7) at short eccentricity and precession periods, corroborate the presence of these orbital signatures. Some peaks in the PC1dis and HIRM-CFB power spectra may be related to obliquity; however, those peaks do not appear consistently across Contessa Road, which is confirmed by squared coherency analyses (Fig. 7). Hence, obliquity is not further assessed in detail here.
Fig. 7: Spectral analysis.
Power spectra (periodograms in the 1st and 2nd rows), squared coherency spectra (3rd row) and phase spectra (4th row) for PC1dis and hard isothermal remanent magnetisation in a carbonate free basis (HIRM-CFB) for different Contessa Road intervals. a The Palaeocene-Eocene Thermal Maximum (PETM), b the post early Eocene terrigenous zone-pre early Eocene low magnetisation zone (post EETZ-Pre EELMZ) and c the post early Eocene low magnetisation zone (EELMZ) define the intervals in which spectral analyses were carried out. 90% and 95% confidence levels in periodograms are indicated by sky blue and dark blue lines, respectively. False alarm levels α = 0.10 and 0.05 in squared coherency spectra are indicated by sky blue and dark blue dashed lines, respectively. Black shapes in phase spectra represent lag/lead relationships of PC1dis with respect to HIRM-CFB. Maximum and minimum lag/lead relationships are indicated in kyr by numbers next to black shapes in phase spectra. Associated frequencies of short eccentricity and precession are indicated by dark grey and light grey bands, respectively. Short eccentricity (~100 kyr) and precession (~22 kyr) periods are related to spectral peaks over the 90%-95% confidence levels.
Orbital signals in PC1dis represent coupled CaCO3/δ13C variations that were also used for Contessa Road age model development19 (see Methods). Thus, peaks and troughs of the filtered PC1dis short eccentricity signal coincide with those of the ZB18a astronomical solution23 (Fig. 8a; see Methods). There are no available LPEE astronomical solutions against which we can compare the filtered PC1dis precession signal. However, multiple studies have identified precession in CaCO3 records17,18,21,22,24. Although these precession signals seem to contrast across different records, CaCO3 variability has been widely related to orbitally driven lysocline depth variations that resulted from precession controls on temperature and the carbon cycle17,18. At diverse localities, LPEE enhanced CaCO3 dissolution coincided with lighter δ13C and δ18O values on precession timescales17,65. These variations were amplified during eccentricity maxima17,18,21,65, which is also visible in the form of amplitude modulation patterns of precession in our PC1dis record (Fig. S16). This suggests that the PC1dis precession signal can be a useful indicator of the impacts of precession-related light carbon concentration/temperature variability on CaCO3 dissolution. Accordingly, enhanced CaCO3 dissolution, associated with higher temperatures and increased light carbon concentrations in the ocean/atmosphere, coincides with PC1dis precession maxima18,21,22,24,62 (Fig. 8c). The filtered HIRM-CFB precession signal is typically out of phase with that of the PC1dis record (Fig. 8c), which is confirmed by phase spectra (see Methods) that consistently indicate lag/lead relationships between ~3 kyr and ~10 kyr (Fig. 7). Out-of-phase HIRM-CFB precession with respect to PC1dis reveals that enhanced CaCO3 dissolution, and its associated temperature and light carbon concentration increases, coincide with haematite formation/ transportation reductions on precession timescales, which is a similar pattern to that identified for orbitally controlled Neogene-Quaternary sedimentation in Mediterranean zones47,63,64.
Fig. 8: Orbital signals.
Short (blue) and long (grey) eccentricity signals of the ZB18a astronomical solution23, and filtered short eccentricity signals of (a) PC1dis and (b) hard isothermal remanent magnetisation in a carbonate free basis (HIRM-CFB). Intervals in which the filtered HIRM-CFB short eccentricity signal does not coincide well with the ZB18a astronomical solution23 are indicated by rectangles with dashed black lines. c Filtered precession signals of PC1dis (black) and HIRM-CFB (brown). Hyperthermals (Palaeocene-Eocene Thermal Maximum (PETM), Eocene Thermal Maximum (ETM) 2 and ETM 3), smaller carbon cycle perturbations, the early Eocene terrigenous zone (EETZ), and the early Eocene low magnetisation zone (EELMZ) are indicated with orange, purple, green and yellow bands, respectively. Lithology is presented at the bottom of the figure with reddish (pink) and whitish limestones (grey) and marls (red).
Precession controls on CaCO3 dissolution and temperature identified at Contessa Road21,24 suggest that precession-driven seasonality played a role in western Tethyan hydroclimate variability. Northern hemisphere summer insolation maxima and winter insolation minima during precession minima (perihelion in the northern hemisphere summer) are expected to induce summer temperature increases and winter temperature decreases in sub-tropical regions such as the proto-Mediterranean and Mediterranean zones63. These precession-driven insolation maxima conditions have been indicated to promote enhanced CaCO3 dissolution17,21 and intensify seasonal precipitation in Mediterranean zones47,64. Hence, our finding of reduced haematite production during such times can be related to moisture increases and chemical weathering amplification, which promoted haematite precursor formation56. During the opposite precession phase, reduced precession-driven seasonality coincides with PC1dis minima and HIRM-CFB maxima (Fig. 8c). This suggests that precession maxima conditions enhanced ferrihydrite transformation into haematite, which reveals strengthened aridity and likely expanded proto-Mediterranean semi-arid and arid zones in western Tethyan continental regions34,47,64. This pattern explains the origins of the HIRM-CFB precession signal via aeolian dust deposition, which has been recognised as the dominant sedimentation mechanism that controlled haematite accumulation at Contessa Road44,48,49,57.
Precession-related HIRM-CFB variability suggests that long and short eccentricity also impacted western Tethyan hydroclimates through amplitude modulation of precession. These eccentricity controls are indicated by the filtered HIRM-CFB short eccentricity signal, which consistently has high haematite contents that coincide with short eccentricity maxima in the ZB18a astronomical solution23 and with short-lived LPEE carbon cycle perturbations (Fig. 8b). This pattern can be a partial result of high-amplitude precession cycles associated with short eccentricity maxima conditions, which strengthened insolation controls on ferrihydrite/haematite production with intensified wet phases that accelerated chemical weathering during precession-driven insolation maxima conditions, and subsequent periods with expanded arid zones that enhanced the ferrihydrite→haematite transformation. This orbitally forced mechanism, in conjunction with short-lived carbon cycle perturbations, systematically strengthened western Tethyan hydroclimate variability. Enhanced orbital controls on the hydrological cycle and/or non-linear hydroclimate responses to short-lived global warming events15,16 should have caused seasonal precipitation reductions that promoted predominantly dry conditions during HIRM-CFB short eccentricity maxima periods9,16,28,37. Our finding of orbitally controlled dry/wet variations suggests considerable similarity between western Tethys LPEE hydroclimate variability and the well-documented Quaternary Mediterranean-type climate. Thus, we propose that LPEE hydroclimates of proto-Mediterranean regions were driven by winter storm track activity14,47 and/or by enhanced monsoon fluctuations16,33,66 that penetrated western Tethyan areas.
The filtered HIRM-CFB short eccentricity signal also contains intervals that are distinctly out of phase with the ZB18a astronomical solution (rectangles with dashed black lines in Fig. 8b). This visual observation is confirmed by lag/lead relationships of up to ~77 kyr in PC1dis/HIRM-CFB phase spectra (Fig. 7) and indicates a contrasting HIRM-CFB pattern with respect to short-lived carbon cycle perturbations. Out-of-phase HIRM-CFB short eccentricity cycles occur mainly in coincidence with long eccentricity minima in the ZB18a astronomical solution23. This may suggest that only a combination of high light carbon concentrations and enhanced precession-driven haematite production, which is expected to be driven partially by long eccentricity maxima, can generate hydroclimate variations recorded in the short eccentricity band15. However, short eccentricity maxima also coincide well with the ZB18a astronomical solution23 in some Contessa Road intervals without short-lived carbon cycle perturbations (Fig. 8b). This might suggest that the HIRM-CFB record was not sufficiently sensitive to short eccentricity variations during intervals in which long eccentricity forcing reduced precession cycle amplitudes. Nevertheless, this interpretation would conflict with the out-of-phase short eccentricity cycles following the I1 and I2 events (Fig. 8b). Another hypothesis to explain the variable phase behaviour of the filtered HIRM-CFB short eccentricity signal is that Contessa Road haematite originated at different latitudes with contrasting orbitally forced hydroclimate conditions. However, this interpretation is not consistent with sedimentary provenance studies that indicate proto-Mediterranean continental zones as detrital material sources for Scaglia Rossa carbonates42.44. Post-depositional authigenic pigmentary haematite formation could have disturbed orbital signals, but this does not seem to apply to Contessa Road, where magnetic minerals seem to have a mainly detrital origin. Finally, emergence of short eccentricity signals in other early Eocene hydroclimate records has been related partially to power transfer from precession to short eccentricity due to asymmetric climate system responses to insolation forcing (e.g., clipping15). We have no evidence to assess conclusively the significance of the HIRM-CFB short eccentricity cycles that are out of phase with the ZB18a astronomical solution23. Hence, we leave this as an open question and instead focus on the amplified effects of short-lived carbon cycle perturbations and enhanced orbitally-driven seasonal contrast15,16 on western Tethyan aridification.
Carbon cycle controls on proto-Mediterranean hydroclimates
The impact of short-lived carbon cycle perturbations on LPEE hydroclimates indicates that the hydrological cycle changed over longer periods than those associated with orbital frequencies9,15,16,28. The Contessa Road magnetic mineral record has an unusual long-term pattern in which a gradual haematite content drop (HIRM-CFB reductions) coincided with its compositional enrichment in magnetic mineral fractions (higher Bcr values) (Fig. 5). This pattern can be explained by a gradual transition to drier western Tethys conditions, which inhibited chemical weathering and ferrihydrite production, but enhanced transformation of available ferrihydrite into haematite. This interpretation is consistent with modelling results that indicate that atmospheric circulation changes under high temperatures produced LPEE dry-drier sub-tropical conditions9. However, this interpretation also implies that contrasting wetter conditions would have been characterised by increased magnetic mineral contents and compositional prevalence of magnetite/maghemite over haematite. These relationships between magnetic mineral composition and content variability with respect to hydroclimate conditions are confirmed by PCA of Mrs-CFB, ARM-CFB, HIRM-CFB, S-ratio, and Bcr (referred to here as PCAarid; Fig. 9a, S13c), which has a PC1arid (69% of data variance) with positive ARM-CFB, HIRM-CFB, Mrs-CFB, and S-ratio loadings, and a negative Bcr loading.
Fig. 9: Hydroclimate variability of western Tethyan continental areas.
a PC1arid-PC2arid plot and PC1arid scores. The early Eocene low magnetisation zone (EELMZ) was excluded because of its anomalous magnetic mineral assemblage (see Supplementary information). b Contessa Road stable carbon isotope (δ13C) record presented in terms of mean (black) ± 2 standard errors (SE) (grey shaded bands)19. Hyperthermals (Palaeocene-Eocene Thermal Maximum (PETM), Eocene Thermal Maximum (ETM) 2 and ETM 3), smaller carbon cycle perturbations, the early Eocene terrigenous zone (EETZ), and the early Eocene low magnetisation zone (EELMZ) are indicated with orange, purple, green and yellow bands, respectively.
From PC1arid the wettest LPPE western Tethys conditions occurred during the EETZ (Fig. 9a). This interval with high detrital inputs (Fig. 5) indicates possible fluvial transport of terrigenous sediment to Contessa Road, which contrasts with the rest of the section in which aeolian dust deposition controlled detrital sedimentation44,48,49,57. PC1arid has a gradual drop that mirrors δ13C, which suggests that the protracted transition to drier proto-Mediterranean zones was a response to the LPEE long-term carbon cycle perturbation and its associated higher temperatures12,20 (Fig. 9). Dry hydroclimates in these areas were especially pronounced during the PETM (Fig. 9a) and L1-L2 events. The ETM 2 and ETM 3 were also characterised by dry settings; however, these events have similar PC1arid values to other intervals that do not coincide with short-lived carbon cycle perturbations. These variable dry hydroclimate responses indicate that long-term warming exerted larger controls on western Tethyan hydroclimate variability compared to the short-lived global warming events, and suggests that seasonal precipitation systems (e.g., monsoons, storm tracks) may have had non-linear responses to LPEE light carbon injections15.16.
Possible non-linear hydroclimate responses of proto-Mediterranean regions to LPEE light carbon injections are tested via estimation of hydroclimate recovery timescales. Irreversible δ13C shifts from CIE conditions to more positive values across hyperthermal recovery phases coincide with PC1arid transitions from negative to more positive values (Fig. 9). These PC1arid changes allow us to constrain hydroclimate recovery timescales with ~27 kyr and ~24 kyr durations for the PETM and ETM 2, respectively. Hydroclimate recovery following these events is well defined by the beginning of wetter conditions for the PETM and by a less dry setting for the ETM 2. The ETM 3 may have had a similar ~25 kyr-long hydroclimate recovery period that coincided with δ13C recovery; nevertheless, the interrupted PC1arid record during the EELMZ (see Supplementary results) does not allow verification of the reliability of this estimate. The narrow range of hydroclimate recovery estimates for events with contrasting magnitudes13,22,25,26 allow us to suggest that optimised carbon removal following hyperthermal peak conditions19,27 also re-established the hydrological cycle to pre-event-like conditions, which confirms the non-linear response of western Tethyan hydroclimate drivers to carbon cycle perturbations. The ~24-27-kyr-long hydroclimate recovery estimates also reveal that hydrological cycle changes associated with massive carbon cycle perturbations may last even longer than other detrimental impacts of global warming19 (e.g., ocean acidification), and indicate that increased temperatures, such as those induced by anthropogenic global warming, can disrupt hydroclimate variability for thousands of years.
Overall, we infer from the Contessa Road record that LPEE long-term and short-lived carbon cycle perturbations induced, in general, dry-drier proto-Mediterranean hydroclimate conditions that disturbed orbitally driven dry-wet variations. Our estimates of hydroclimate recovery following peak hyperthermal conditions support model projections, which indicate that global warming can adversely impact the climate system for thousands of years. Along with model results9,28,35, our documented dry-drier response of proto-Mediterranean zones to increased temperatures indicates future dry conditions in Mediterranean zones and suggests that an anthropogenic SSP 8.5 hothouse world can potentially cause extensive dryland aridification in sub-tropical latitudes29,67.
