In addition to volatile degassing, the emplacement of the NAIP may have affected the global climate system through other processes such as increased continental weathering and widespread regional uplift.

Despite the close temporal link between the NAIP emplacement and the PETM, the exact relationship is complicated by multiple concurrent climate forcings, incomplete/imprecise geochronological data (Wilkinson et al., 2017), and uncertainties in the timing and sources of volatile fluxes from the NAIP (Passey and Jolley, 2008; Stoker et al., 2018). These uncertainties are compounded by a lack of expanded sedimentary records that contain both volcanic and climatic proxies in the same sections. The limited dispersal of many geochemical/geological indicators of volcanism means that such proxies are often regionally constrained (Jones, 2015; M. T. Jones et al., 2019). In addition, paleoclimate proxies across the Paleocene–Eocene transition are often complicated by the significant changes in sedimentation, seawater acidification, regional uplift/subsidence, eustatic sea level, and seaway connections. These uncertainties can only be resolved by targeted studies on expanded and continuous sedimentary sequences proximal to the NAIP that contain a multitude of volcanic and climatic proxies.

The eastern North Sea basin is an ideal setting for constraining the uplift history and magmatic activity of the NAIP across the PETM. This region has experienced near-continuous tectonic subsidence since the Late Cretaceous, resulting in high sedimentation rates and expanded sedimentary sequences. The relative proximity of this basin to the NAIP (Fig. 1) resulted in the co-preservation of multiple volcanic and climatic proxies. Additionally, this location was sufficiently distant from LIP activity to avoid subtle climatic indicators being overwhelmed by the volcanic signal and was relatively unaffected by the regional forced regressions caused by thermal uplift. The enclosed nature of the North Sea and the lack of significant thermal or diagenetic overprints have resulted in the exceptional preservation of both inorganic and organic records (Nielsen, 1995). These factors make North Sea sediments ideal for a wide range of integrated geochemical, biological, oceanographic, and volcanological studies. The Limfjord area in Denmark (Fig. 2) offers rare onshore access to North Sea Paleocene and Eocene strata due to glaciotectonic uplift (Pedersen, 2014). These outcrops provide a unique opportunity to attain a high-quality, virtually uninterrupted record of NAIP activity and environmental change spanning the latest Paleocene and early Eocene (Heilmann-Clausen et al., 1985).

This study presents a compilation of new and existing proxy data for volcanism (ash layers, Hg anomalies, Os isotopes) and paleoclimate (lipid biomarkers and C, Li, and Os isotopes) from the Stolleklint Beach section on Fur and other localities around the Limfjord area (Fig. 2). These data are uniquely poised to (1) assess changes in environmental conditions across the PETM and in the earliest Eocene, (2) investigate how the style and magnitude of NAIP activity varied concurrently, and (3) evaluate how these regional signals relate to global changes through comparison to existing PETM localities worldwide.

The succession was logged and sampled at several outcrops and quarries on Fur (M. T. Jones et al., 2019; Stokke et al., 2020a), supplemented with Skarrehage and Ejerslev localities on the island of Mors and at Silstrup in mainland Denmark (Fig. 2a). These sequences were compiled to form a composite stratigraphic succession where the zero of the depth scale is set to the top of the ubiquitous and easily distinguishable felsic Ash -33 (Fig. 2c). The stratigraphic thicknesses vary slightly between localities, so sections within the composite succession are tied to specific localities. The lowermost pre-PETM to post-PETM section (-24.8 to +21.8 m) is based on the Stolleklint Beach locality (M. T. Jones et al., 2019). The Elke and Stendal East quarries (+21.8 to +41.0 m) form the basis of the log between Ash -11 and Ash +98. The uppermost part (+41.0 to +51.0 m) is based on the Silstrup locality (Fig. 2a).

The presence of abundant ash layers in the Danish strata offers insights into the extent and style of NAIP volcanism, and they act as key marker horizons between localities. High-precision radiometric dating of magmatic crystals within ash deposits provides a geochronological framework into which paleoenvironmental records can be placed (Lowe, 2011). Key felsic ash layers in the Danish strata such as Ashes -17 and +19 are important marker horizons across Greenland and Europe (Storey et al., 2007a; Westerhold et al., 2009). Some of the basaltic ash layers are up to 12 cm thick and found 700–1500 km from the known source volcanoes, representing some of the largest explosive basaltic eruptions recorded in the geological record (Egger and Brückl, 2006; Stokke et al., 2020b). The Stolleklint Clay and Fur Formation were systematically logged and sampled, with the thickness of each ash layer recorded and combined into a percentage of the strata (after compaction) per metre of section.

A widely used volcanic proxy is mercury (Hg) content in sedimentary records (Grasby et al., 2019; Percival et al., 2021). Volcanic emissions are a primary source of gaseous Hg0 that is rapidly dispersed through the atmosphere (Pyle and Mather, 2003). Prolonged periods of elevated volcanism, such as from LIPs, directly impact the global Hg cycle (Grasby et al., 2019). Mercury has a strong affinity for organic matter, which provides the main depositional pathway for Hg in aqueous environments (Outridge et al., 2007). Elevated Hg / TOC is interpreted to show enhanced Hg sequestration, either through enrichment of Hg in organic matter or via deposition through other means such as adhesion to clay particles or in sulfides (Sanei et al., 2012). Therefore, anomalously high Hg content or peaks in host-phase normalised ratios (e.g. Hg / TOC, Hg / S) can be indicative of extensive regional or global volcanic activity (Grasby et al., 2019).

High-resolution sampling and analyses were conducted for Hg and TOC through the Stolleklint Clay (M. T. Jones et al., 2019) and the Fur Formation (this study). The lowermost strata (-24.8 to -24.3 m) were continuously sampled, while the mean sample spacing for the rest of the stratigraphy was 8 cm (including 425 new samples). All samples were oven-dried at ≤ 50 ∘C and powdered in either an agate hand mortar or agate disc mill before further analysis. Mercury content was analysed using a Zeeman R-915F (Lumex) high-frequency atomic absorption spectrometer at the University of Lausanne. Samples were heated to > 700 ∘C, and measurements were taken on the direct thermal evaporation of Hg from solid samples. Each aliquot was measured in duplicate, while machine accuracy was confirmed by the analysis of the GSD-11 standard certified reference material (Chinese alluvium: 72.0±3.6 ppb). Total organic carbon (TOC) concentrations were obtained by Rock-Eval pyrolysis (Behar et al., 2001) at the University of Lausanne.

The ratio of lithium isotopes 7Li and 6Li, expressed as δ7Li, is a tracer for silicate weathering in sedimentary records. Lithium is abundant in silicate rocks, largely absent from carbonate rocks, and is not known to be fractionated by organic growth (Kisakürek et al., 2005; Pogge von Strandmann et al., 2016). As such, Li isotopes are dominantly controlled by silicate weathering processes (Pogge von Strandmann et al., 2021). In fluvial waters, Li isotopes reflect the balance between primary rock dissolution and secondary clay formation. Isotopic fractionation through clay formation increases dissolved-δ7Li values, whereas primary rock dissolution does not fractionate Li isotopes and therefore drives dissolved δ7Li to low values (Misra and Froelich, 2012; Pogge von Strandmann et al., 2017b). Low-intensity fluvial weathering regimes (i.e. high erosion relative to weathering rates) are characterised by high primary rock dissolution relative to clay formation, leading to low dissolved-δ7Li values but high dissolved-Li fluxes. Moderate-intensity weathering regimes have greater clay formation and therefore elevated dissolved-δ7Li values but somewhat lower dissolved-Li fluxes. High-intensity weathering regimes induce the dissolution of secondary clays with little primary rock dissolution, which leads to low δ7Li values and low Li fluxes (Dellinger et al., 2015; Pogge von Strandmann et al., 2020). In this study, we analysed the solid (largely clay) material from the sediment sequences, providing a counterpart to the dissolved signal. The advantage compared to marine carbonate sections is that detrital clay sections are not affected by the long (∼ 1 Myr) ocean residence time of Li but instead provide information on local conditions. Selected samples were analysed through the Stolleklint Clay (Pogge von Strandmann et al., 2021) and the Fur Formation (this study). Powdered samples were dissolved first using HF–HNO3–HClO4, then steps of concentrated HNO3 and 6 M HCl. The samples were then purified for Li using a two-column method with AG 50W-X12 resin and 0.2 M HCl as an eluent (Pogge von Strandmann et al., 2017a). The purified samples were analysed at the LOGIC group at University College London using a Nu Plasma 3 multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS) and normalised to IRMM-016 bracketing standards. The measured δ7Li values of secondary standards at this facility are seawater (31.17±0.38 ‰, n=43), USGS BCR-2 (2.57±0.30 ‰, n=11), and USGS SGR-1b (3.82±0.28 ‰, n=9). The total procedural blank is ≤ 0.003 ng Li (Pogge von Strandmann et al., 2019). Long-term external analytical uncertainties using this method are ±0.4 ‰ (2σ) (Pogge von Strandmann et al., 2021).

Samples from the Fur Formation were analysed for clay mineralogy to expand on the existing dataset from Stokke et al. (2021). The clay fraction (< 2 µm) was prepared as oriented aggregate mounds using gravity settling and the Millipore filter transfer method (Moore and Reynolds, 1997). The X-ray diffraction (XRD) clay data were recorded with a step size of 0.01 from 2 to 65 (2θ) at a count time of 0.3 s (2θ) in air-dried samples and with a step size of 0.01 from 2 to 34 (2θ) at a count time of 0.3 s (2θ) on treated samples. Three rounds of treatments were applied: 24 h of ethylene glycol saturation, 1 h heating at 350 ∘C, and 1 h heating at 550 ∘C. The software NewMod II (Reynolds and Reynolds, 2012) was used for semi-quantification of the XRD patterns of inter-stratified clay minerals.

The osmium (187Os / 188Os) isotope system is a powerful paleoceanographic tracer for silicate weathering, specifically identifying changes between evolved and juvenile sources through time (Peucker-Ehrenbrink and Ravizza, 2000; Peucker-Ehrenbrink and Ravizza, 2020). The ratio of rhenium to osmium (Re / Os) is higher in crustal rocks than in the mantle, which means that the in situ beta decay of 187Re to 187Os leads to higher (radiogenic) 187Os / 188Os ratios over time. The crustal 187Os / 188Os average is ∼ 1.40 (Peucker-Ehrenbrink and Jahn, 2001), while the unradiogenic primitive upper-mantle value is ∼ 0.13 (Meisel et al., 2001). The oceanic record reflects the proportional mixing of these two end-members (Peucker-Ehrenbrink and Ravizza, 2000), with the unradiogenic component sourced from the weathering and alteration of subaerial basalts, submarine hydrothermal activity (Dickson et al., 2021), and occasional extraterrestrial bolide impacts (Sato et al., 2013).

Osmium is removed from the water column through the precipitation of Fe–Mn oxides and/or adsorption onto organic material and clays (Racionero-Gómez et al., 2017; Yamashita et al., 2007). The oceanic residence time of Os is 10–55 kyr (Levasseur et al., 1999; Sharma et al., 1997), which is longer than the present-day mixing time of the oceans (1–2 kyr) but considerably shorter than for other weathering tracers such as strontium (> 2 Myr) (Broeker and Peng, 1982). While the global oceans are relatively homogeneous at a given time, 187Os / 188Os ratios can deviate rapidly from the fully mixed global signal in restricted environments such as enclosed basins (Rooney et al., 2016). Enhanced organic matter deposition can reduce Os residence times significantly, and local input and output fluxes become more important in enclosed settings (Martin et al., 2001; Paquay and Ravizza, 2012). Therefore, the 187Os / 188Os proxy has multiple potential uses in this study, including tracking the emplacement and weathering of NAIP lavas and ashes; changes in continental weathering associated with regional uplift or climatic variations; and modifications to the extent of seaway connectivity between the North Sea basin and the Arctic, Tethys, and Atlantic realms.

Individual samples weighing ca. 50 g were selected for Os isotope chemostratigraphy analysis throughout the succession. Consolidated samples were cut using a rock saw and then hand-polished with a diamond-plated polishing pad to remove potential contamination from the saw blade and weathered surfaces. After being dried at room temperature overnight, all samples were crushed to a fine (∼ 30 µm) powder in a SPEX 8500 ShatterBox with a ceramic agate grinding container and puck to homogenise any potential Re and Os heterogeneity. The long half-life of 187Re (ca. 42 Gyr) leads to minimal age correction for Cenozoic samples, but both Re and Os concentrations and isotopic compositions were measured in this study to provide accurate initial Os isotope compositions for chemostratigraphy.

Sample Re and Os isotopic abundance and composition were determined at the Yale Geochemistry and Geochronology Center. Depending on Re concentration, between 0.3 and 1.0 g of sample powder was digested and equilibrated in 8 mL of CrVIO3–H2SO4 with a mixed 190Os–185Re tracer (spike) solution sealed in Carius tubes at 220 µC for 48 h. This dissolution method has been shown to preferentially liberate hydrogenous Re and Os to yield a more accurate and precise depositional age (Kendall et al., 2004). The Re and Os were extracted and purified through solvent extraction (NaOH, (CH3)2CO, and CHCl3), micro-distillation, and anion column chromatography and analysed using negative thermal ionisation mass spectrometry (Selby and Creaser, 2003). Isotopic measurements were performed via static Faraday collection for Re and ion counting using a secondary-electron multiplier in peak-hopping mode for Os on a ThermoElectron TRITON PLUS mass spectrometer (Creaser et al., 1991; Völkening et al., 1991). The Os samples were loaded onto 99.995 % Pt wire (H-Cross, NJ) in 9 N HBr, covered with a saturated solution of Ba(OH)2 in 0.1 N NaOH as an activator, and analysed as oxides of Os. Interference of 187ReO3 on 187OsO3 was corrected using the measured intensity of 185ReO3. Mass fractionation was corrected with 192Os / 188Os = 3.0826 using the exponential fractionation law.

In-house Re and Os solutions were continuously analysed during this study to ensure and monitor long-term mass spectrometer reproducibility. The Yale Geochemistry and Geochronology Center Re standard solution measured on the Faraday cups yields an average 185Re / 187Re value of 0.59748±0.0014 (2σ,n=6), which agrees with the accepted value within error (Gramlich et al., 1973). The measured difference between 185Re / 187Re values for the Re solution and the accepted value (0.59738) is used to correct the Re sample data. The Os isotope standard solution used at Yale Metal Isotope Center is the Durham Romil Osmium Standard (DROsS) (Luguet et al., 2008). Over the past 3 years on the Yale Triton, the runs have yielded a 187Os / 188Os ratio of 0.16082±0.000116 (2σ,n=319), which is identical, within uncertainty, to the value reported by other laboratories (e.g. Liu and Pearson, 2014; Luguet et al., 2008). Total procedural blanks during this study were 40±2.0 pg for Re and 0.06±0.11 pg for Os, with an average 187Os / 188Os value of 0.25±0.05 (1σ).

Uncertainties for 187Re / 188Os and 187Os / 188Os are determined by error propagation of uncertainties in Re and Os mass spectrometry measurements, blank abundances and isotopic compositions, spike calibrations, and reproducibility of standard Re and Os isotopic values. The Re–Os isotopic data and 2σ calculated uncertainties for 187Re / 188Os and 187Os / 188Os are used to generate initial Os isotope (187Os / 188Os(i)) compositions with an age of 55.9 Ma.

The extent of North Sea isolation may be tested using existing datasets of Os isotopes (see above) and thaumarchaeotal membrane lipid distributions. The composition of isoprenoid glycerol dialkyl glycerol tetraether lipids (GDGTs) is thought to regulate membrane fluidity of marine Thaumarchaeota (Schouten et al., 2013). As such, GDGT composition is commonly assumed to be largely governed by growth temperature and forms the basis of the TEX86 paleotemperature proxy (Schouten et al., 2002). Typically, the number of cyclo-pentane rings increases with temperature, with the TEX86 ratio presenting the statistically strongest relation (Kim et al., 2010; Schouten et al., 2002). However, other metrics, such as the ring index (Zhang et al., 2016) and fractional abundance of crenarchaeol regio-isomers (fcren') relative to total crenarchaeol (O'Brien et al., 2017), can be employed to detect non-thermal impact on TEX86 (Zhang et al., 2016) and differences in temperature response between communities (e.g. Inglis et al., 2015; O'Brien et al., 2017). Indeed, it has been recognised that certain modern ocean regions have distinct GDGT distributions and responses to temperature, arguably associated with distinct, isolated communities of Thaumarchaeota (e.g. Trommer et al., 2009). Some studies have turned this around to argue for distinct communities based on GDGT distributions in (semi-)restricted settings (Steinig et al., 2020).

High fcren' has been recognised in (warm) saline waters (Steinig et al., 2020; Trommer et al., 2009) and has been proposed to identify such paleoenvironmental conditions (Inglis et al., 2015). In contrast, many freshwater bodies are marked by reduced fcren' (e.g. Blaga et al., 2009; Powers et al., 2010). At times during the latest Paleocene–early Eocene, reduced-salinity water masses occupied the North Sea area (Bujak and Mudge, 1994; Eldrett et al., 2014; Kender et al., 2012; Zacke et al., 2009) and the Arctic Ocean (Pagani et al., 2006; Sluijs et al., 2006). Here we re-examine the published Paleocene–Eocene GDGT data, including those from the North Sea area, focusing on the differences in response to PETM warming between TEX86 and fcren' to assess whether GDGT distributions can be used as a supporting tool to detect basin restriction and/or reduced salinity.

The NAIP ash layers are largely constrained to discreet ash-rich intervals within the Stolleklint Clay and Fur Formation. Four ashes (SK1–SK4) are found just above and below the PETM onset (Fig. 3b), representing the earliest evidence of explosive volcanism in Danish strata (Heilmann-Clausen et al., 2014). There is a distinct lack of visible ash layers for a ∼ 19.3 m interval between the PETM onset and Ash -39, the first of the traditional numbered ash series (Fig. 3a). The second interval of ash-rich strata (-4.9 to +10.9 m) encompasses the recovery of the δ13Corg excursion and includes several felsic ash layers such as Ashes -33 and -17. It is followed by a ∼ 15 m thick ash-poor interval, with only three ash layers > 1 cm thickness. Ash +1 at +26.0 m heralds the main phase of ash deposition in the Fur Formation (Fig. 3a). With the exception of the felsic Ashes +13 and +19, the positive ash series comprises relatively homogenous tholeiitic basalts. Basaltic ash reaches a peak of 31 % of the total sediment (after compaction) between +42.0 and +43.0 m (Ashes +110 to +118; Fig. 3a), representing a considerable portion of the total stratigraphy.

Mercury concentrations vary from 2 to 303 ppb through the succession, but within each interval Hg content is relatively consistent (Table 1). Mercury peaks are few and limited in amplitude, with only four samples exceeding 100 ppb. Mercury content was also normalised to TOC and plotted as Hg / TOC (Fig. 3). Samples where TOC < 0.2 wt % were excluded from Hg / TOC ratios as the propagated error creates unacceptably high uncertainties (Grasby et al., 2019). The intervals where this occurs are in the glauconite-rich layer at the top of Interval 1 (Fig. 3b) and in part of the diatomite-rich Fur Formation (Intervals 7–9; Fig. 3a). The late Paleocene and pre-PETM strata (Intervals 1 and 2) have relatively low mean Hg content values of 32.5 and 30.0 ppb, respectively (Table 1). When Hg is normalised to TOC, the late Paleocene strata are relatively uniform (Interval 1), while the pre-PETM strata (Interval 2) show more scatter and sporadic Hg / TOC anomalies (Fig. 3b). Mercury content increases slightly across the PETM onset (Interval 3) to an average of 43.8 ppb (Table 1), but a concurrent increase in TOC content leads to lowered Hg / TOC values compared to the interval below.

Within the PETM body, Hg becomes increasingly enriched upwards through the strata, with the largest mean concentrations observed in Interval 6 (67.7 ppb; Table 1). This concentration is comparable to the average shale value (62.4 ppb) from a compilation of global datasets (Grasby et al., 2019). In the lower and middle parts of the PETM (Intervals 4 and 5), the Hg content covaries with TOC enrichments, resulting in an average Hg / TOC ratio of 33.8 ppb wt %-1 (σ= 10.0) for Interval 4 and 26.9 ppb wt %-1 (σ=3.6) for Interval 5. The low standard deviations highlight that Hg / TOC values are remarkably uniform through Intervals 4 and 5 (Table 1). The relative homogeneity ends in the final 5.6 m of the PETM body (Interval 6), where Hg enrichments outpace increased TOC content, coincident with the re-emergence of ash layers in the stratigraphy (Fig. 3a). The PETM recovery (Interval 7) heralds a decrease in Hg and TOC concentrations into the diatomitic Fur Formation. Mercury content rises again slightly into Interval 8, becoming more pronounced in the section between Ashes -17 and +1, with a mean Hg content of 47 ppb (Table 1). Variable TOC enrichments in Interval 8 lead to a large scatter in Hg / TOC ratios (Fig. 3a). Mercury and TOC content gradually declines into Interval 9, reaching a nadir (mean 21.4 ppb Hg) in the strata between Ashes +19 and +118, with many samples < 0.2 wt % TOC. Mercury and TOC content slightly increases again towards the top of the section exposed at Silstrup, but Hg / TOC ratios are comparable to much of the rest of the Fur Formation (Fig. 3a).

There are significant variations in detrital Li isotopes (δ7Li) through the studied section. Within the late Paleocene and pre-PETM strata (Intervals 1 and 2), δ7Li values are typically between -1.1 ‰ and +0.2 ‰. The two samples encompassing the negative δ13C onset (at -24.37 and -24.26 m) show a -3.9 ‰ δ7Li excursion (Fig. 3b), accompanied by a TEX86-based ∼ 10 µC sea surface temperature (SST) warming across the same interval (Schoon et al., 2015; Stokke et al., 2020a). The Li isotopes gradually return to less negative values through the course of the PETM body, returning to -0.7 ‰ by the end of Interval 6 (Fig. 3a). Lithium isotope values then oscillate from -2.4 to +0.2 and back to -2.2 ‰ through the PETM recovery (Interval 7). The post-PETM strata (Interval 8) first show a gradual positive δ7Li excursion, peaking at 0.5 ‰ at ∼ 1.5 m above Ash -17 (Fig. 3a). There is then a progressive negative δ7Li excursion to -4.0 ‰ just above Ash +9. The remaining samples from the upper Fur Formation between Ashes +31 and +98 (Interval 9) have δ7Li values between -1.8 and -1.4 ‰.

The Re and Os abundances and isotopic compositions vary considerably in the studied samples. Elemental Re and Os abundances range from 0.9 to 196.4 ppb and 54 to 1009 ppt, respectively (Table 2). 187Re / 188Os ratios vary between 18 and 2498, while 187Os / 188Os ratios vary from 0.336 to 2.593. Rhenium and Os enrichments appear to covary with TOC content, with the PETM interval showing elevated element abundances compared to pre- and post-PETM strata (Table 2). Using an assigned age of 55.9 Ma, calculated initial 187Os / 188Os values (187Os / 188Os(i)) range from 0.258 to 1.110, with significant and systematic changes through the stratigraphy. The late Paleocene samples (Interval 1) are consistently unradiogenic, with 187Os / 188Os(i) values between 0.26 and 0.40 (Table 2, Fig. 3b). The 187Os / 188Os(i) values are lower in the ash-bearing pre-PETM strata (Interval 2), consistently between 0.27 and 0.29. The PETM onset (Interval 3) is marked by an increase in 187Os / 188Os(i) values from 0.27 to 0.39. Throughout the body of the δ13C excursion (Intervals 4–6), the 187Os / 188Os(i) values remain unradiogenic (0.30–0.39) and remarkably stable for ∼ 24 m of stratigraphy (Table 2). However, the PETM recovery samples (Interval 7) exhibit an extreme change in 187Os / 188Os(i) values from 0.3 to 0.83 across a ∼ 1 m interval containing the felsic Ash -33 (Fig. 3a). This more radiogenic signature persists throughout the Fur Formation, albeit with considerable variations in 187Os / 188Os(i) (0.49–1.11).

Absolute age estimates allow for the quantification of sediment deposition rates for environmental proxies such as TOC and Hg. A robust geochronological framework is achievable for the Stolleklint Clay due to the clear onset, body, and recovery of the PETM δ13C excursion and for parts of the Fur Formation due to the prevalence of ash layers that act as key marker horizons (Westerhold et al., 2009). In contrast, the chronology of the pre-PETM strata is poorly constrained due to intense bioturbation and a potential unconformity below the glauconite-rich horizon.

Sedimentation rates during the PETM can be estimated by two methods. The first method uses the estimated durations of the PETM onset and body from astronomically calibrated localities and modelling. If the 11 cm of preserved strata is the entire PETM onset (Fig. 3b), and assuming a maximum duration of 5 kyr for this interval (Kirtland Turner, 2018), the minimum average sedimentation rates at the start of the hyperthermal were ∼ 2.2 cm kyr-1 (Table 1). For the PETM body, van der Meulen et al. (2020) predict a CIE duration of 101±9 kyr. Including propagated uncertainties, this duration gives an average sedimentation rate of 23.8±4.5 cm kyr-1, up to an order of magnitude greater than the sedimentation rate during the onset interval (Table 1). The alternative method is to use marker horizons of known ages. The PETM onset is constrained by astronomical tuning to 55.93 Ma, based on the best-fit orbital solution of Westerhold et al. (2017). A bentonite horizon within the PETM CIE in Svalbard has a 206Pb / 238U age of 55.785±0.034 Ma (Charles et al., 2011), estimated to be equivalent to ∼ 2 m above Ash -33 in the Danish strata (Stokke et al., 2020a). Using these two ages gives a sedimentation rate of 17.9±2.8 cm kyr-1, including uncertainties in the thickness of the Stolleklint Clay and the position of the Svalbard marker horizon in the Fur strata. The slight disparity between these two sedimentation rate estimates may arise because the latter value includes part of the PETM recovery, where deposition rates were reduced at Fur. Overall, the two methods are in good agreement, and we base our age model on the former estimate.

Estimating Fur Formation deposition rates is aided by the presence of numerous ash layers, which have been correlated with tuff horizons within the East Greenland flood basalts at the Gronau Nunatak (Heister et al., 2001; Storey et al., 2007a), offshore along the north-western Europe continental margin of Goban Spur (DSDP Site 550), and the Norwegian continental margin (ODP Site 642; Fig. 1) (Knox, 1984, 1985). In particular, two prominent felsic ash layers (Ashes -17 and +19) are used to constrain the geochronology of the early Eocene. Ash -17 was Ar–Ar-dated to 55.48±0.12 Ma once corrected using the 28.201 Ma Fish Canyon Tuff calibration (Kuiper et al., 2008; Storey et al., 2007a). Astronomical tuning using records from Site 550 and the Walvis Ridge (ODP Sites 1262 and 1263) predicts a ∼ 200 kyr interval between the two ashes (Röhl et al., 2007; Westerhold et al., 2009), which gives an estimated age of ∼ 55.28 Ma for Ash +19 (Fig. 3). The measured thickness between Ash -17 and Ash +19 based on the Stolleklint and Elke outcrops is 18.67 m, or 18.02 m excluding ash intervals, which gives an estimated deposition rate of 9.0 cm kyr-1 for this section of the Fur Formation (Table 1).

The main proxies for the eruption style and intensity of NAIP activity are ash deposition, Os isotope chemostratigraphy, clay mineralogy, and Hg anomalies, and each of these proxies has its strengths and weaknesses. Ash layers are conclusive evidence of explosive eruptions. However, the eastern North Sea is 700–1500 km from the known NAIP source volcanoes (Fig. 1), which means that only the largest explosive eruptions are preserved in Danish strata. Several ashes show evidence of magma–water interactions in their formation (Stokke et al., 2020b), suggesting that phreatomagmatic processes led to the explosive nature of eruptions. Therefore, while the distribution of ashes is a clear indicator of extreme explosive volcanism, it may reflect changes in eruption style and may not be indicative of overall volcanic activity. Osmium isotopes can be used as a passive tracer for the weathering and erosion of LIP basalts and ashes affecting ocean chemistry, providing the signal is discernible from other factors such as changing continental weathering and other sources of unradiogenic Os such as extraterrestrial inputs and mid-ocean-ridge spreading (Dickson et al., 2021). Another indicator of the weathering and erosion of LIP basalts and ash is the presence of the minerals smectite and zeolite (Nielsen et al., 2015; Stefánsson and Gíslason, 2001). While clay mineralogy is affected by changes in climate and hydrology, the presence of both smectite and zeolite minerals in variable abundance throughout the Fur stratigraphy has been linked mainly to weathering of volcanic products of NAIP origin (Heilmann-Clausen et al., 1985; Stokke et al., 2021) rather than to reworking of smectite-rich sediments (Li et al., 2020) or post-deposition flocculation (Deconinck and Chamley, 1995).

Sedimentary mercury is generally accepted as a viable proxy for large-scale volcanism in the geological record, with Hg and Hg / TOC anomalies coeval with periods of (subaerial) LIP emplacement (Grasby et al., 2019). However, other factors can impact Hg and TOC content non-uniformly (Grasby et al., 2019; Percival et al., 2015; Frieling et al., 2023), so it is important to rigorously assess individual localities taking these factors into account. Directly comparing (normalised) Hg records between localities is complicated by differences in depositional environments, such as organic matter source and content, lithology, sediment accumulation rates, and redox state (Grasby et al., 2019). It is also currently challenging to differentiate between thermogenic and volcanic sources for Hg in the sedimentary record, since both are large potential Hg sources (Svensen et al., 2023). The regional distribution of Hg emissions would be heavily affected by whether Hg degassing is subaerial or submarine (M. T. Jones et al., 2019; Percival et al., 2018). Passive submarine degassing around modern volcanic systems can limit Hg deposition to just tens of kilometres from the source in enclosed, shallow marine environments (e.g. Tomiyasu et al., 2007), so the depth and explosivity of submarine emissions will have a major impact on subsequent Hg dispersal. Thermogenic degassing from the NAIP was most likely dominated by explosive shallow marine venting (Svensen et al., 2004), which may suggest that Hg anomalies proximal to the NAIP are more likely to be of contact metamorphic origin. However, given that widespread sill intrusions and continental flood basalts are both chronologically constrained to somewhere in this time interval and that the ratio of submarine versus subaerial Hg emissions for each source is poorly constrained, the Hg record can only give an overall qualitative indicator of NAIP activity across this interval.

Existing data across the PETM suggest larger Hg anomalies closer to the NAIP (M. T. Jones et al., 2019; Keller et al., 2018; Kender et al., 2021; Liu et al., 2019; Tremblin et al., 2022), indicating that subaqueous emissions from volcanic and/or thermogenic sources may well have limited Hg deposition to more proximal settings than from atmospheric distribution. In fact, even within the North Sea basin there appears to be a substantial gradient in the magnitude and frequency of Hg anomalies, decreasing in intensity from north-west to south-east (M. T. Jones et al., 2019; Kender et al., 2021). Therefore, the Danish strata do not experience significant Hg anomalies compared to more proximal localities. However, what the Fur succession does have is an expanded sedimentary sequence and a well-constrained geochronology, which allows us to estimate Hg mass accumulation rates (Hga) using Eq. (1): 1Hga=Hgc(SR)ρHgangcm-2yr-1=Hgcngg×SRcmyr×ρgcm3, where Hgc is the mean measured Hg concentration, SR is the estimated sedimentation rate, and ρ is the density of the host sediments. The density of the Stolleklint Clay is measured as 1.4 g cm-3, while the Fur Formation diatomite has a measured density of 0.8 g cm-3 (Pedersen et al., 2004). The calculated Hga values for each interval are shown in Table 1.

The Paleocene–Eocene transition is marked by several regional and eustatic changes in sea levels. A global sea-level rise on the order of a few metres during the PETM is attributed to the thermal expansion of seawater (Sluijs et al., 2008). While this effect would be important for shallow slope/margin environments, it is insufficient to affect seaway connectivity. In contrast, the transient thermal uplift in the centre of the NAIP closed the Atlantic connection via the Faroe–Shetland Basin until ∼ 54 Ma (Hartley et al., 2011; Shaw Champion et al., 2008; White and Lovell, 1997). Evidence from the δ18O record of shark-tooth apatite indicates a North Sea surface-water freshening in the early Eocene (Zacke et al., 2009), suggesting that the Atlantic connection via the English Channel was also temporarily restricted. This North Sea freshening is corroborated by an influx of low-salinity tolerant dinocyst taxa into the North Sea prior to the PETM CIE (Kender et al., 2012). There is also evidence of a transient closure of the shallow marine connection to the Peri-Tethys (Radionova et al., 2003) and a restriction of the Norwegian–Greenland Seaway (Hovikoski et al., 2021) around the late Paleocene to early Eocene. Although the exact timings of these events are not well constrained, these observations indicate the potential for restriction of the pre-breakup Nordic and North Sea basins. The multiple datasets from the Danish Paleogene strata in this study and prior work can be compared with other global high-resolution records to explore changes to the oceanic connections between the North Sea and the Atlantic, Tethys, and Arctic oceans.

The Li isotope record for the Stolleklint Clay and Fur Formation shows large variations both during and after the PETM (Fig. 3). A major excursion in δ7Li values of -4 ‰ in under 10 kyr, which is coeval with the PETM onset and matches other global siliciclastic and carbonate Li isotope records, is interpreted to be a direct weathering response to an enhanced hydrological cycle during global warming (Pogge von Strandmann et al., 2021). However, the post-PETM record continues to show considerable variability, with three negative δ7Li excursions broadly coinciding with the start of Intervals 7, 8, and 9, each followed by positive excursions (Fig. 3). Notably, the excursion at the start of Interval 9 is of a comparable magnitude (-4 ‰) to the excursion during the PETM onset, albeit a gradual change over 15 m (∼ 160 kyr) rather than a sharp response over a 12 cm interval (Fig. 6).

Numerous factors could have influenced the detrital δ7Li signal to generate the observed fluctuations in the Fur Formation. Increased surface temperatures elevate both chemical weathering and erosion and may increase erosion more than chemical weathering, leading to a lower-weathering-intensity regime, as proposed to explain the negative δ7Li excursion during the PETM onset (Pogge von Strandmann et al., 2021). However, the lack of correlation between Li isotopes and paleotemperatures in post-PETM sediments suggests that climate was not the main driving factor of later δ7Li excursions (Fig. 6). Alternatively, uplift associated with the emplacement of the NAIP could affect detrital δ7Li values (Dellinger et al., 2017, 2015), since the increase in topography would steepen slopes and increase erosion rates, which could also lower the weathering intensity. This effect could account for the negative δ7Li excursion observed at the start of the PETM recovery (Interval 7), which was coeval with the NAIP uplift that isolated the North Sea from the North Atlantic Ocean. However, this NAIP uplift phase was the last major regression that is recognised in proximal settings (Hartley et al., 2011; Shaw Champion et al., 2008; White and Lovell, 1997), so continued uplift events were unlikely to be responsible for the negative δ7Li excursions during Intervals 8 and 9 (Fig. 6).

Variations in clay mineralogy and/or clay content could also cause δ7Li excursions in the bulk sediments, but there is little correlation between Li isotopes and clay mineralogy through the Fur Formation (Fig. 6). Peaks in kaolinite content, possibly indicating enhanced runoff transporting this denser clay further into the catchment (Nielsen et al., 2015; Stokke et al., 2021), do not correlate with δ7Li values (Fig. 6). The δ7Li excursions also do not appear to be linked to changes in the bulk lithology. The heavily laminated sections in Intervals 4 (Stolleklint Clay) and 8 (Fur Formation diatomite) show opposing δ7Li trends, while the upper sections of Interval 9 (unlaminated diatomite), where the clay component comprises 100 % smectite, continue to show some isotopic variability (Fig. 6). In summary, variations in clay mineralogy or lithology do not appear to be driving the observed δ7Li fluctuations.

One proxy that does broadly correlate with δ7Li variations in the post-PETM sediments is the 187Os / 188Os record (Fig. 6). Osmium isotopes reflect a basin-wide signal, while the detrital δ7Li record in Danish strata is likely representative of siliciclastic input from the Fennoscandian shield on the north-eastern North Sea margin (Anell et al., 2012). Assuming that the North Sea remained largely isolated from the North Atlantic Ocean during the deposition of the Fur Formation, the 187Os / 188Os(i) record suggests an increase in basalt-derived fluxes towards the start of Interval 9 (Fig. 6). The correlation with the decreasing δ7Li values suggests that volcanic ash was likely being deposited on the Fennoscandian platform in increasing quantities towards the end of Interval 8 and that these ash deposits were rapidly weathered and eroded, forming isotopically light clays that were transported into the basin. During the 2010 eruption of Eyjafjallajökull in Iceland, clay formation occurred within just a few months of ash deposition in local catchments (Olsson et al., 2014; Paque et al., 2016; Pogge von Strandmann et al., 2019), and this process was likely to have been faster in the warmer Eocene climate. The gradual δ7Li excursion of -4 ‰ from Interval 8 to Interval 9 also coincides with an increase in Hg content (Fig. 3a) and estimated Hg accumulation rates (Fig. 5), which is also consistent with a volcanic origin. In combination, these proxies suggest that the post-PETM variations in δ7Li values were largely governed by the rapid weathering and erosion of terrestrial ash deposits from the NAIP, which reached a post-PETM peak around the start of Interval 9.