Terrestrial cooling in Northern Europe during the Eocene–Oligocene transition

Solent Group Stratigraphy and Paleoclimate.

Terrestrial records of Late Eocene European climate are rare; however, continental sediments of the Solent Group in the Hampshire Basin (Isle of Wight, United Kingdom) provide an age-calibrated (SI Methods) terrestrial EOT record from the mid-North Atlantic region. Sediments of the Solent Group were deposited in a coastal plain setting at ∼45–50°N from the Late Eocene through the Early Oligocene and have never been deeply buried, and nearly all remain unlithified (14). Throughout the sequence, freshwater lacustrine and alluvial sediments are punctuated by a few brief marine incursions that help provide age control (15).

Fossils from the Hampshire Basin indicate a warm Late Eocene climate with abundant marsh vegetation, palms, fruit-eating mammals, and crocodilians (14⇓⇓–17). δ¹⁸O of carbonate from freshwater gastropods, fossil rodent teeth, and charophytes within Solent Group sediments has been interpreted to show peak Late Eocene freshwater summer temperatures of >30 °C, with only minor reduction in growing season temperature across the EOT (10, 15, 18). However, δ¹⁸O-derived temperatures contrast with North Atlantic SST and paleofloral data (6, 7), which record significant temperature decreases across the EOT. These also contradict δ¹⁸O data from North American fossils that show an ∼8 °C cooling during this same time (11). One complication with conventional δ¹⁸O temperature interpretations is that they rely on estimates of the δ¹⁸O_water at the time of formation. Changes in the isotopic composition of precipitation or ambient water (e.g., due to evaporation) can introduce uncertainty in reconstructions of paleotemperature in the Late Eocene to Early Oligocene of the Hampshire Basin (10).

Clumped Isotope Paleothermometry.

The “clumped isotope thermometer” is a temperature proxy based on a measure of the temperature-dependent abundance of doubly substituted ¹³C and ¹⁸O isotopes that are bound to each other within the carbonate lattice (19). The Δ₄₇ of carbonate minerals [a measure of the abundance anomaly of ¹³C-¹⁸O bonds in carbonate-derived CO₂ and defined as Δ₄₇ = (R⁴⁷_actual/R⁴⁷_stochastic − 1) × 1,000] provides a measure of formation temperature, δ¹³C, and δ¹⁸O of carbonate and allows direct calculation of δ¹⁸O of parent water (20).

Δ₄₇ Measurements and Data Reduction.

We collected shells of the freshwater prosobranch gastropod V. lentus from five localities within the Solent Group of the Hampshire Basin, Isle of Wight, for δ¹⁸O and Δ₄₇ temperature measurement. Samples were collected from 10 intervals that span nearly 3 My of the Late Eocene to Early Oligocene (Table S1). Sediments were dried, disagregated, and sieved to separate shell materials. Shells were embedded in epoxy for thin-section analyses (Figs. S2 and S3). For high-resolution δ¹⁸O, complete shells were broken into consecutive pieces, and each was supported internally by epoxy and microsampled. Shell fragments for Δ₄₇ measurements were crushed to a size fraction of <400 µm and stored in a dry chamber before analysis. Splits of this fraction were crushed to <200 µm, using an agate mortar and pestle immediately before Δ₄₇ analysis. Each sample line (Table S1 and Dataset S1) represents a measurement of a split of the <400-µm fraction and two to four splits are used to determine the sample level mean for Δ₄₇. Variability between splits reflects any heterogeneity preserved in the <400-µm fraction.

δ¹³C, δ¹⁸O, and Δ₄₇ of purified CO₂ were measured on a Thermo Finnigan MAT 253 mass spectrometer configured to measure masses 44–49 with sample and reference gas capillaries balanced at a 16-V signal (Methods). Sample gas was measured against a reference of known isotopic composition for a total of 8–10 replicates of 10 cycles. Δ₄₇ values were determined using established methods (19) and normalized through analysis of CO₂ gas with a range of isotopic compositions that was heated for 2 h at 1,000 °C to achieve the “near”-stochastic distribution of isotopologues. Δ₄₈ values provide a check for contaminants with interfering masses that could impact Δ₄₇, and gas with high measured Δ₄₈ values was excluded from our data.

To relate Δ₄₇ measurements to temperature using the empirically determined Δ₄₇-temperature relationship of Ghosh et al. (20), it is critical to correct for scale compression of individual mass spectrometers (21). Our measurements predate establishment of the “Absolute Reference Frame” (ARF) and normalization procedures outlined therein (22). As a result, we normalized Δ₄₇ data to the ARF (Figs. S4 and S5), using measured Δ₄₇ of CO₂ of varied isotopic composition and equilibrated with water at 25 °C, a Carrara marble sample that has been run numerous times and in multiple laboratories, and gases of varied composition heated to 1,000 °C (SI Methods and Table S2).

Gastropod Δ₄₇ Temperatures.

Calibration of the clumped isotope thermometer derives from a variety of biotic and abiotic carbonates from a range of growth temperatures (20). Studies show that molluscs precipitate carbonate in isotopic equilibrium (23), and mollusc Δ₄₇ data fall on the empirical Δ₄₇-temperature calibration curve of Ghosh et al. (20). Recent study of Δ₄₇ of two species of aquatic gastropods in freshwater Pliocene sediments shows no temperature bias related to “vital effects” during gastropod growth and demonstrates that gastropod carbonate precipitates in equilibrium with ambient waters (24). Our examination of a modern unionid mollusc from the Huron River in Michigan produces a Δ₄₇ temperature of ∼19 °C, which compares favorably to mean May to September growing season temperatures that average 19.6 °C (PRISM Climate Group). These data show that Δ₄₇ temperatures for aquatic gastropods reliably match conventional oxygen isotope paleothermometry estimates when water δ¹⁸O is well constrained, unlike terrestrial gastropod Δ₄₇ temperatures (25). However, aquatic gastropod Δ₄₇ data are biased by seasonal temperature variation during the period of shell growth.

Multiple lines of evidence indicate that our Δ₄₇ data record primary values. First, fossil shells are preserved in clay-rich sediments, which potentially reduced fluid-flow that could have altered the primary shell structure during burial. Second, Solent Group sediments have not been deeply buried, heated, or extensively altered, and most sediments remain unlithified (14⇓–16, 26). Although low-temperature diagenetic alteration of aragonite to secondary calcite is commonly observed in the rock record (27), all samples were analyzed by X-ray diffraction and produced patterns consistent with aragonite (28). Third, thin-section analyses of individual shells showed an outer layer with vertical crystals in all shells. The prismatic, alternating vertical crystals and clear growth cessation marks are consistent with modern Viviparus species (Fig. 1 and Figs. S2 and S3), indicating preservation of primary structure. Finally, we conducted high-resolution microsampling of five pristine shells to examine the variability in δ¹⁸O of individual shells across the EOT (Fig. 2). These all preserve fine-scale δ¹⁸O variations that reflect changes in water temperature and composition during the period of shell growth. These data also show a decrease in δ¹⁸O values during the EOT, consistent with changes in local environmental conditions. Preservation of variability on the millimeter scale within the aragonite shell matrix suggests original Δ₄₇ values have also not been altered.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

(A) Thin section of a Late Eocene V. lentus in plane light. The shell structure contains a prismatic layer with alternating vertical crystals similar to those of modern shells and clear cessation marks that show the interruption of shell growth due to food or temperature reductions. (B) Detail of prismatic layer.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

(A–C) Subannual δ¹⁸O record of (A) Early Oligocene, (B) Eocene–Oligocene, and (C) Late Eocene gastropods. Sample ages are shown in upper right corner. Isotopic variability shows preservation of primary δ¹⁸O variations due to temperature or water δ¹⁸O changes. Shells from the EOT are marked by a shift to low δ¹⁸O values, the lowest (sample HAM S4) of which occurs at 33.7 Ma and corresponds to the interval of the coldest Δ₄₇ temperatures during the EOT.

Δ₄₇ Temperature Data and Paleoclimate.

Aquatic gastropods are incapable of regulating body temperature, so shell carbonate is precipitated in thermal equilibrium with ambient water (29). As a result, Δ_{47 shell} of ancient gastropods records seasonal growth temperature. This can be related to MAAT, providing a clear picture of ΔT (°C) across the EOT. Our samples span the EOT reduction in atmospheric CO₂ and Oligocene oxygen isotope event 1 (Oi-1) (Oi-1 onset between samples BOULD S1 and BOULD S5), although no fossil carbonates are available from the peak of the Oi-1 glacial event due to sea-level fall that interrupted deposition (15).

Because of the large sample size required for Δ₄₇ measurements (5–10 mg), we homogenized individual gastropod shell fragments to <400 µm and powdered individual sample splits immediately to <200 µm before analysis. Two to four splits of the coarser fraction were measured for Δ₄₇ to determine “bulk” growth temperature for each sample unit. This approach has the benefit of enabling precise measurement of bulk growth temperature; however, it integrates any seasonal temperature variability as shown by the high-resolution δ¹⁸O data (Fig. 2). This produces a single measure of growing season temperature that is weighted proportionally to periods of higher metabolic and carbonate accumulation rates. Our data show slightly greater bulk isotopic variability between sample splits relative to standard materials, suggesting some heterogeneity at the <400-µm scale.

Sample mean Δ₄₇ values range from 0.671 to 0.735 relative to the ARF (22) (Fig. 3A and Table S1). There is debate over temperature calibrations for some types of carbonates (20, 22, 30) and differences between calibrations produce large discrepancies for low temperatures (<15 °C). We use the calibration of Ghosh et al. (20) recalculated to the ARF (22) and assume temperatures reflect the time-integrated water T °C over the period of shell growth. Using this calibration, Δ₄₇ values correspond to water temperatures of 34–20 °C. Before 34 Ma, growing season water temperatures range from 28 to 34 °C, with peak temperatures occurring before the EOT (Fig. 3B). Measured values are similar to June to September mean surface water temperatures from subtropical areas such as the Florida Everglades (SI Methods), an area with a broadly similar climate to that of the Late Eocene Hampshire Basin (16). Growing season temperatures oscillate during initial phases of ice buildup on Antarctica (within EOT, Fig. 3B). Following the major pCO₂ reduction, growing season water temperatures decline significantly from pre-EOT values, ranging from 20 to 24 °C (Fig. 3 A and B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

(A) Δ₄₇ of Hampshire Basin gastropods across the EOT. δ¹⁸O_benthic data (gray x’s) (2) are shown for reference. Error bars reflect SE of the Δ₄₇ measurements. (B) Reconstructed Eocene–Oligocene atmospheric pCO₂ (squares) (5) and Δ₄₇ measured temperature (circles). Error bars reflect 1 SE on multiple analyses. Vertical shaded area covers the period of the Eocene–Oligocene transition (EOT) and the dashed line indicates the Eocene–Oligocene (E-O) boundary.

δ¹⁸O of Eocene–Oligocene Waters.

The δ¹⁸O_gastropod depends on the δ¹⁸O_water and temperature during the formation of shell carbonate. Δ₄₇ temperature data allow calculation of δ¹⁸O water across the EOT, using a temperature-dependent aragonite–water fractionation equation (31):δ¹⁸O_water is highly evaporative before the EOT, coinciding with a period of high pCO₂, high growth temperatures (∼34 °C), and a water δ¹⁸O of ∼ +3‰. Following the decline in pCO₂ (5) and associated changes in marine δ¹⁸O data (8), Hampshire Basin δ¹⁸O_water returns to pre-EOT compositions despite a shift of the ocean δ¹⁸O due to the buildup of Antarctic ice (2). Reconstructed water δ¹⁸O values are more positive than those of seawater, reflecting highly evaporative conditions during the EOT (Table S1). Calculated water δ¹⁸O from Hampshire Basin gastropods shows significant variability before and during the EOT. This variability is similar to that observed in contemporaneous sediments in the Ebro Basin, Spain, which show significant changes to the hydrologic cycle in the Late Eocene and Early Oligocene (32).

During the EOT, shell δ¹⁸O becomes more negative with the lowest δ¹⁸O values evidenced in a shell dating to ∼33.7 Ma (sample HAM S4, Fig. 2) (28). Shells from the EOT suggest a possible increase in within-shell δ¹⁸O variability that may be related to an increase in seasonality or variation in water isotopes. These data are relatively limited at present; however, they show that clear variations in environmental conditions occurred during the EOT.

Δ₄₇ Paleotemperatures and Paleoclimate.

Relating growing season temperature to climatic conditions requires an understanding of the relationship between seasonal surface water temperature and MAAT and between bulk growth temperature and seasonal variability. The gill-breathing aquatic gastropod Viviparus colonizes a range of habitats and requires oxygen and fresh water. It commonly lives in deeper portions of the littoral zone during the cool season (<4 m) and migrates to shallower waters (0–2 m) during the warm growing season to avoid oxygen minimum zones. Viviparus can grow throughout the year but shows a decline in growth rate at cool temperatures (28). Growth rates are limited by temperature and food availability, and warm season growth rates are several times greater than cold season growth rates. In temperate lakes, Viviparus shows minimal growth during the cool months of September to March and rapidly grows during May and June in response to increased temperatures and food availability as photosynthetic rates increase (29). Observations of growth cessation in EOT gastropods (Fig. S3) indicate these shells dominantly record spring to summer growth (29). Collinson (16) suggested that water depths in the plant-bearing units of the Bembridge Marls Member were 0.3–3 m, so EOT snails likely record growth when located in oxygenated surface waters.

Determining the period of gastropod growth and carbonate mass accumulation in the gastropod shell is critical to relating Δ₄₇ temperatures to seasonal and mean annual air temperature. The metabolic rate of freshwater fluvial and lacustrine male prosobranch gastropods is dependent upon temperature (33). In turn, shell growth rate is related to oxygen demand. Empirical measures of temperature dependence of oxygen demand for prosobranch gastropods suggest that if temperature is the primary control on shell growth, then peak carbonate mass accumulation should occur during the warm summer months, with the June to September period accounting for nearly 50% of shell growth, and the months of April to October accounting for more than 75% of total mass accumulation (SI Methods, Fig. S6, and Table S3). We assume that Δ₄₇ temperatures of ancient Viviparus integrate spring to autumn water temperatures and are biased toward warm season conditions.

We used empirical lake temperature–air temperature transfer functions to interpret seasonally biased Δ₄₇ temperatures with respect to MAAT (34) for a range of possible shell growing season durations. Modern surface water temperature data indicate that mean growing season water temperature is strongly correlated with seasonal and MAAT, and seasonal water temperatures can exceed air temperatures (34). Growing season water T °C (Δ_{47 gastropod}) can constrain Eocene to Oligocene MAAT (°C) when we consider water–air temperature relationships for the period of gastropod carbonate mass accumulation and growth. We consider three scenarios for shell mass accumulation to relate growing season water T °C (Δ_{47 gastropod}) to MAAT with gastropod carbonate integrating a period of growth from (i) April to October (ii), April to June, or (iii) June to August. Modern relationships are appropriate for an Eocene world because they include environments that are similar to the reconstructed Eocene–Oligocene climate of the Hampshire Basin.

Measured shell Δ₄₇ values reflect water temperature during the growing season. Δ₄₇ temperatures decrease by more than 10 °C across the EOT from peak temperatures of ∼34 °C ± 3 °C to nearly 20 °C ± 2 °C (Fig. 3B) and closely follow patterns of deep ocean temperature change and benthic δ¹⁸O (2). These show that cooling during the EOT produced a large change in water conditions during shell growth. Application of seasonally weighted transfer functions that relate seasonal temperature to MAAT enables calculation of the ΔMAAT °C during the EOT, relative to Late Eocene average MAAT (before 34 Ma). Importantly, uncertainty in the timing of growth and in which transfer function to apply does not significantly impact interpretation of MAAT as long as the period of shell growth dominantly occurs during the warm season (Table 1 and Table S1). The maximum observed ΔMAAT for a single sample level is greater than 10 °C during the EOT and Oi-1, regardless of which air–water temperature transfer function is used. If gastropod carbonate Δ₄₇ is inferred to reflect April to October growth, our data indicate that MAAT decreased from a mean of ∼24 °C before 34 Ma to ∼18 °C in the early Oligocene (Fig. 4A). The assumption of a long growing season is reasonable because Eocene to Oligocene conditions were considerably warmer than present. Reconstructed temperatures closely match high-latitude paleofloral and soil tetraether paleotemperature data and are consistent with North Atlantic SST values in excess of 20 °C before the EOT (6, 7, 35). If, however, cooling during the EOT resulted in a shift in the timing of shell growth to late spring to coincide with peak food inputs, shell carbonate would be biased by water temperatures during this period. Taken together, the ΔMAAT inferred from Δ₄₇ data and the shift from a longer (April to October) to a shorter growing season (April to June) would imply a decrease in MAAT from ∼24 °C before 34 Ma to ∼20 °C after the Oi-1, i.e., a decrease of ∼4 °C from the Late Eocene to Early Oligocene.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

(A) Eocene to Oligocene MAAT from Δ₄₇ water temperature estimates and water–air temperature transfer function (squares). Error bars include propagated error from April to October water–air temperature transfer function calibration and SE on multiple Δ₄₇ measurements. δ¹⁸O_benthic record is shown in x’s with five point running mean and reflects a combination of deep-ocean cooling and changes in ice volume (2). (B) Change in MAAT from Δ₄₇ data (triangles) relative to Late Eocene mean (pre-34 Ma) and the ΔSST relative to Late Eocene mean for ocean drilling program sites 511 (circles), 913 (squares), and 1090 (diamonds) (7). Black line represents the smoothed change in SST °C (7). Error bars include SE of the mean Late Eocene temperature and the error associated with individual annual paleotemperature estimates.

Hampshire Basin and Northern Hemisphere Terrestrial EOT Records.

Marine records of the EOT show significant cooling during Antarctic ice buildup associated with changes in atmospheric pCO₂ (1, 7, 8). These show larger temperature decreases at high latitudes, whereas tropical regions show much smaller declines. Much of the cooling may occur in the lead-up to the Oi-1 onset (9) and cooling is associated with dramatically increased seasonality (9). Terrestrial records, which generally have coarser temporal resolution, show the full spectrum of climatic change during the EOT. Fossil records show a significant turnover event (Grande Coupure) coinciding with the Oi-1 or earliest Oligocene (15, 36). This turnover has been attributed to a local shift to drier conditions (36), competition from new arrivals from Asia (37), and increased seasonality (38) and lags the Eocene–Oligocene (E-O) boundary by ∼100–300 ky. However, European climate records indicate significant regional climatic heterogeneity. Paleosol geochemical records from Spain show a significant decline in chemical weathering (∼30%) across the EOT coincident with falling pCO₂, but no change in precipitation or temperature (39). Paleosol records from the Hampshire Basin suggest minimal temperature change but an increase in precipitation across the EOT (39). Combined isotopic records of fossil teeth and carbonate from the Hampshire Basin were interpreted to show no major cooling in summer temperatures during the EOT (10). These all contrast with North Atlantic marine data that show significant cooling of the surface ocean (7).

Like European records, North American terrestrial records show a range of responses to Antarctic glaciation and declining CO₂. Isotopic proxies based on bone and teeth show cooling during the EOT (∼8 °C) that lags global marine records by several hundred thousand years (11). North American paleosols show cooler and drier conditions across the EOT, with some sites indicating significant aridification during this transition (e.g., ref. 40). However, these records do not indicate cooling of the magnitude suggested by isotopic data from fossils (11). Paleosol and sedimentologic data would suggest that at least part of the drop in North American continental temperature during the EOT, as determined from fossil horse tooth δ¹⁸O, may be an artifact of changes in precipitation δ¹⁸O related to changing hydrologic regimes (39).

Our clumped isotope data from the Hampshire Basin show that coastal northern Europe land temperatures decreased by an average of ∼4–6 °C from the Late Eocene to the Early Oligocene. This cooling is somewhat less than the inferred cooling in continental North America (11); however, temperature declines are associated with a shift in the regional isotopic composition of ambient water. This change in water δ¹⁸O may result from alteration of the regional hydrologic cycle related to global cooling and could reflect increased seasonality during and following the EOT. Importantly, Δ₄₇ data from the Hampshire Basin contradict previous conventional stable isotope temperature estimates from the same localities that suggest minor or no change in growth temperatures across the EOT (10, 18). The new Δ₄₇ data show a larger change in surface water δ¹⁸O than was indicated by conventional fossil δ¹⁸O data. Such a result suggests that reconstructed climate records based on oxygen isotopes of terrestrial carbonates are significantly improved when paired with temperature estimates that are independent of assumptions of water δ¹⁸O.

Δ₄₇ MAAT Estimates and Atmospheric CO₂.

pCO₂ decreased from >1,000 ppm to <600 ppm between 35 and 32 Ma, punctuated by a brief peak in pCO₂ before the marine oxygen isotope shift (5). Paleoclimate models show that a reduction of this magnitude (from 1,120 ppm to 560 ppm) should be accompanied by significant decreases in high-latitude MAAT, with much of the temperature decrease due to reductions in cold month mean temperature (CMMT) and increased seasonality (6).

Hampshire Basin Δ₄₇ data show a decline in growing season temperature and MAAT (Figs. 3 and 4) during a period of declining atmospheric pCO₂ and the onset of major Antarctic glaciation. Pre-EOT peak CO₂ concentrations coincide with the highest temperatures and most isotopically enriched surface waters, which suggests that pre-EOT increases in atmospheric pCO₂ may have resulted in an interval of high temperatures and increased evaporation before the onset of major Antarctic glaciation. Initial phases of the EOT are characterized by an increase in local climatic variability relative to pre-EOT conditions and precede the major drop in temperature that coincides with the Oi-1 glacial event. Δ₄₇ values record growing season and MAAT, but do not directly constrain seasonality.

Our data show that coastal northern Europe temperatures at ∼45–50°N decreased, on average, by ∼4–6 °C from the Late Eocene to Early Oligocene. This cooling was synchronous with a decrease in the concentration of pCO₂ and the buildup of Antarctic ice and is similar to the amount of cooling observed in North Atlantic SST (Fig. 4B) (7). The timescale and magnitude of terrestrial temperature decreases during the EOT indicate coupling of terrestrial climate to atmospheric pCO₂ and ocean temperatures and suggest that rapid North Atlantic SST cooling during the EOT and Oi-1 may have resulted from a strong pCO₂–climate feedback during this critical climate transition.