Orbital forcing of terrestrial hydrology , weathering and carbon sequestration during the Palaeocene-Eocene Thermal Maximum

The response of the Earth System to greenhouse-gas driven warming is of critical importance for the future trajectory of our planetary environment. Hypethermal events past climate transients with 25 significant global-scale warming can provide insights into the nature and magnitude of these responses. The largest hyperthermal of the Cenozoic was the Palaeocene-Eocene Thermal Maximum (PETM ~56 Ma). Here we present a new high-resolution cyclostratigraphy for the classic PETM section at Zumaia, Spain. With this new age model we are able to demonstrate that detrital sediment accumulation rates within this continental margin section increased more than four-fold during the 30 PETM, representing a radical change in regional hydrology that drove dramatic increases in terrestrial to marine sediment flux. During the body of the PETM, orbital-scale variations in bulk sediment Si/Fe ratios are evidence for the continued orbital pacing of sediment erosion and transport processes, most likely linked to precession controls on sub-tropical hydroclimates. Most remarkable is that detrital accumulation rates remain high throughout the body of the PETM, and even reach peak values during 35 the recovery phase of the characteristic PETM carbon isotope excursion (CIE). Using a series of Earth System Model inversions, we demonstrate that the silicate weathering feedback alone is insufficient to recover the PETM CIE, and that active organic carbon burial is required to match the observed dynamics of the CIE. Further, that the period of maximum organic carbon sequestration coincides with the peak in detrital accumulation rates observed at Zumaia. Based on these results, we hypothesize that 40 Clim. Past Discuss., https://doi.org/10.5194/cp-2017-131 Manuscript under review for journal Clim. Past Discussion started: 16 November 2017 c © Author(s) 2017. CC BY 4.0 License.


Introduction
The PETM is a marked interval of climate warming (Dunkley Jones et al., 2013), shoaling of oceanic 5 carbonate saturation horizons (Zachos et al., 2005) and a large global negative Carbon Isotope Excursion (CIE) in all rapidly exchangeable marine and terrestrial carbon reservoirs (McInerney and Wing, 2011).The most consistent explanation for these coupled perturbations is the release of carbon from a large shallow lithospheric reservoir, with a depleted carbon isotopic (δ 13 C) signature (~ -20 to -60 ‰), on a multi-millenial timescale (Bowen et al., 2015;Dickens et al., 1995;Gutjahr et al., 2017b;10 Kirtland Turner and Ridgwell, 2016;Zeebe et al., 2016).Although there is still no confidence on the identity of such a large (>4000 Pg C) and unstable carbon reservoir, its release and oxidation within the ocean-atmosphere system caused rising atmospheric CO 2 concentrations, warming and a range of Earth System perturbations associated with pronounced global warming (Sluijs et al., 2007).15 Although considerable attention has been paid to constraining the rates of carbon release, based on deep-ocean carbonate dissolution (Panchuk et al., 2008;Zachos et al., 2005;Zeebe et al., 2009), rates of warming (Meissner et al., 2014;Zeebe et al., 2016), carbon isotope profiles (Bowen et al., 2015;Kirtland Turner and Ridgwell, 2016) and surface ocean pH (Gutjahr et al., 2017a) the mechanisms responsible for both the climatic and isotope recovery at the end of this transient event are still not well 20 constrained (Bowen and Zachos, 2010).The timescales of silicate weathering and carbonate burial (~100-200 ka) are suggested to be too long to drive the main phase of CIE recovery, which the best records available to date indicate is an order of magnitude faster (~10-20 ka) (Bowen and Zachos, 2010).However, observed patterns of enhanced carbonate and possibly biogenic silica deposition in the deep-ocean through the PETM recovery phase are interpreted as the signature of a silicate weathering-25 driven recovery (Penman, 2016;Penman et al., 2016).The alternative mechanism for carbon removal is the substantial burial of organic carbon within either terrestrial or marine sedimentary systems (Bowen and Zachos, 2010).One means of increasing global organic carbon burial rates is through enhanced sedimentation rates on continental margins, which might contribute significantly to C org burial rates with little or no rise in measured total weight percent organic carbon content of sediments (Sluijs et al., 30 2014).There is also direct evidence for substantial increases in the organic carbon content of syn-PETM marine sediments in the eastern Tethys, associated with increased terrestrial to marine weathering fluxes and reduced seawater oxygenation (Carmichael et al., 2017;Gavrilov et al., 1997).
If organic carbon burial does drive recovery from the PETM, there is no explanation for the temporal 35 offset between increased sedimentation rates and organic carbon burial during the body of the PETM (John et al., 2008) -when models indicate no or little carbon removal from the systems -and the major phase of carbon burial required for the CIE recovery at the end of the event.Models that include a continued leakage of carbon to maintain a prolonged CIE provide a mechanism to delay recovery ( Penman et al., 2016;Zeebe et al., 2009), but this delay only exacerbates the mismatch between the observed fast recovery and the timescales of silicate weathering-driven carbon drawdown (Penman et al., 2016).As yet, there is no satisfactory model to explain the timing of the large-scale carbon drawdown required to drive the rapid CIE recovery phase ~100 ka after the onset of the PETM (Bowen and Zachos, 2010).Understanding the ultimate fate of thousands of petagrams of "excess" carbon 5 injected into the ocean-atmosphere system during the PETM has profound implications for understanding the behaviour of the Earth's climate future over the coming millennia.Here, we present new sedimentological records from the Iberian upper continental slope (~1000m paleo-water depth), constrained within a detailed new cyclostratigraphic age model to address the linkage between climatic controls on sedimentation rates and carbon drawdown.10 2 The Zumaia PETM succession

Cyclostratigraphy
Our studied PETM section is located at the Itzurun beach-front of the town of Zumaia, northern Spain (Figure 1; N 043°18´4.5",W 002°15´31.2").Situated within the deep-basin of the Pyrenean Gulf, it is the deep-water end of a sediment distribution system that has been mapped in detail over the past two 15 decades (see references and summary within Pujalte et al., 2015).Logged sections span terrestrial alluvial fans and plains, shallow shelf carbonates, incised channel distribution systems and a deep-water basin (Pujalte et al., 2015 and references therein).Within this, the Zumaia section is regarded as the most complete and representative end member of the deep-water depositional system, recording the distal deposition of fluvial-derived fine grained sediment plumes (Pujalte et al., 2015) at middle-lower 20 bathyal (1000 m) paleodepths (Alegret et al., 2009).The aim of this study is to build on the existing detailed depositional framework developed for the region, and to use the Zumaia section as a key recorder of the temporal patterns of climate-driven sediment flux from terrestrial systems to the deep ocean during the PETM.25 Although the Zumaia section is a classic PETM locality and the subject of a series of bio-, magneto-, chemo-, and lithostratigraphic studies (Alegret et al., 2009;Baceta et al., 2000;Bernaola et al., 2007;Canudo et al., 1995;Dinarès-Turell et al., 2007;Manners et al., 2013;Schmitz et al., 1997;Schmitz et al., 2001;Storme et al., 2012), there is, as yet, no robust cyclostratigraphic framework established for the distinct 'fine-grained siliciclastic unit' (FSU), which spans the majority of the PETM CIE (Figure 30 2).To generate a cyclostratigraphic framework, we took a total of 248 small "thumbnail" 1cc bulk sediment samples at ~3 to 5cm resolution from ~4.3 m below to ~8.2 m above the onset of the characteristic "siliciclastic unit" that marks the body of the PETM at this location.The zero reference point for all sampling is taken at the top of the prominent limestone bed, located ~0.3 m below the base of the SU.In between the limestone and the SU is a 0.3 m thick greenish-grey marl unit, which is 35 distinct from both the limestone / marl units found above and below the SU and the typically red-brown clays and marly clays of the SU itself.The section was logged in detail, and integrated with the existing stratigraphic framework for the site (Figure 2).All samples were analysed for bulk sediment elemental composition by X-ray Fluoresence (XRF) using dried, powered sediment, mounted in wax pellets.Pellets were analysed with a Bruker S8 tiger X-Ray Fluorescence (XRF) Spectrometer with 8-minute analysis time, at the University of Birmingham (UoB).5 CaO concentrations from XRF analyses were converted into high-resolution estimates of weight percent CaCO 3 using a calibration set of 25 discrete calcium carbonate measurements.Weight percent CaCO 3 contents were determined at UoB, based on inorganic carbon contents measured by CO 2 generation with phosphoric acid reaction within a 200ºC furnace using a Shimadzu TOC-V carbon analyser.Detrital MARs were estimated assuming that both detrital and carbonate fractions had an estimated dry bulk 10 density of 1.8 g cm -3 .Organic carbon MARs were calculated following (2009).
A subset of 129 thumbnail samples were also analysed for the stable isotope composition (δ 13 C carb and δ 18 O carb ) of bulk carbonates at the NERC Isotope Geosciences Laboratory (NIGL, Keyworth, Nottingham).Sample material was powdered to yield CO 2 from the equivalent of 10 mg of 100% 15 calcite and reacted with anhydrous phosphoric acid in vacuo overnight at a constant 25°C.Liberated CO 2 was separated from water vapour under vacuum and collected for analysis.Measurements were made on a VG Optima mass spectrometer (standard reproducibility of < ±0.1‰).All data are on the VPDB scale.An additional 67 bulk samples were analysed for total organic carbon content at NIGL using a Carlo Erba 1500 elemental analyser with acetanilide used as the calibration standard.Replicate 20 analyses indicated a precision of ±0.1 wt% in well-mixed samples (1 Standard Deviation, SD).
New analyses resolve a prominent cyclicity in SiO 2 and Si/Fe ratios both below and above the SU (Figure 2), corresponding to the established precession-paced lithological cycles in this setting (Dinarès-Turell et al., 2003;Westerhold et al., 2008).Through the SU itself, 10 Si/Fe cycles are identified 25 between the CIE onset and the δ 13 C inflection point F of Röhl et al. (2007; Figure 2).With a total CIE duration of 90-167 ka (Murphy et al., 2010;Röhl et al., 2007) these are most likely half-precession cycles (10*10.5=105ka) of a similar nature to those in terrestrial palaeosol successions from the Bighorn Basin.These are likely linked to catchment hydrology (e.g.Abdul Aziz et al., 2008), resulting from the interference of both northern and southern hemisphere precession signals within tropical 30 hydroclimates (Short et al., 1991).This is consistent with a strong hydroclimate control on sediment erosion and transport in the Tremp-Graus terrestrial basins during the PETM (Pujalte et al., 2015;Schmitz and Pujalte, 2007).
Above the SU, two prominent thicker carbonate beds are correlated with the 100 ka eccentricity minima 35 of Westerhold et al. (Westerhold et al., 2007;precession cycles 6/7 and 11), which is supported by the placement of the Last Rare Occurrence of the calcareous nannofossil genus Fasciculithus at 8.7 m (Baceta et al., 2006;confirmed by Dunkley Jones, pers. obs.), between cycles 15 and 16, which is consistent with Forada section and ODP Site 690 (Röhl et al., 2007).Significant power is resolved at frequencies of ~2.2 cycles m -1 (~0.45 m period) and 0.7 cycles m -1 (~1.4 m period) (Figure 3), which 40 are taken to be half-precession and precession cycles respectively.This double peak in power spectra is unlikely to be due to a simple step-change in sedimentation rates at the onset of the PETM conditions - with an associated shift in the frequency of precession signals in the depth domain -as power is resolved at both frequencies within the body of the PETM where sedimentation rates are relatively stable (Figure 2).Based on the resolution of these ten cycles within the siliciclastic unit, we have managed to extend the existing cyclostratigraphy for the Zumaia section through the body of the PETM, developing a detailed age model and calculated sedimentation rates (Figure 4). 5

The Zumaia Carbon Isotope Excursion (CIE)
Prior to the onset of the CIE at Zumaia, benthic foraminifera extinctions begin in the greenish-grey marl unit between the top of a characteristic limestone band (0 m) and the onset of the PETM (0.3 m) (Alegret et al., 2009).The main phase of benthic foraminifera extinctions is precisely coupled with the onset of the CIE at 0.3 m (Alegret et al., 2009).The CIE onset is rapid with a ~3.3 ‰ negative shift 10 between 0.3 and 0.38 m (~0 to 4.7 ka after onset) with a further decline to values of ~ -4.0 ‰ or lower above 0.5 m (~12 ka after onset; total CIE of more than ~3.7 ‰; Figure 2).Although the doubling of sedimentation rates at the onset of the PETM limits the effects of any carbonate burndown on the δ 13 C carb record across the onset, δ 13 C carb data appear to diverge from the typical PETM CIE profile between ~0.7 to 2.8m.Here, δ 13 C carb values return to higher values, of ~ -3‰, which lie between those 15 of the peak CIE and pre-excursion values (Figure 2).This is attributed to the increased relative contribution of reworked detrital carbonates, as evidenced by the presence of Cretaceous nannofossils (Dunkley Jones, pers.obs.), during periods of very low autochthonous carbonate flux.This is supported by the return of δ 13 C carb to near peak-excursion values once carbonate contents increase above 2.8m.Above this level, there is a CIE recovery phase similar to other high-resolution records from the deep 20 ocean (e.g.Zachos et al., 2005; Figure 5).
The new age model for Zumaia enables a direct comparison to both existing deep ocean (Rohl et al., 2007) and terrestrial (Bowen et al., 2014) locations that have a comparable cyclostratigraphic framework (Figure 5).In the rate of CIE onset -defined as time to peak negative CIE values -and the 25 magnitude of the CIE excursion, the Zumaia δ 13 C record is clearly more similar to the Big Horn Basin soil carbonate δ 13 C record than to other deep-ocean sites (Figure 5).In both sections the main CIE onset is marked by a ~4‰ negative δ 13 C shift, with a duration of between ~2 and 12 ka.This excursion value is then maintained for ~80 ka, during a relatively stable "plateau" phase (~55.93 to 55.85 Ma), before a relatively quick main recovery phase, which is slightly quicker at Zumaia (~55.85 to 55.82 Ma) than in 30 the Bighorn Basin (55.85 to 55.81 Ma).In the deep ocean, sedimentary records of the PETM onset are typically condensed due to extensive carbonate dissolution (Bralower et al., 2014).Combined with sediment mixing, this tends to smooth δ 13 C carb profiles, reducing the apparent rate of onset of the event, whilst maintaining an appearance of stratigraphic continuity (Bralower et al., 2014;Kirtland Turner and Ridgwell, 2016).The greatly increased flux of fine-grained detrital sediment from nearby terrestrial 35 basins (Pujalte et al., 2015)

Calcium Carbonate and Detrital Sediment Mass Accumulation Rates
Based on the cyclostratigraphic age model and estimated weight % CaCO 3 , we estimate calcium carbonate, total organic carbon (TOC) and detrital mass accumulation rates (MAR) throughout the 5 study interval (Figures 4 and 5).Carbonate MAR closely follows weight % CaCO 3 , with a marked decline during the body of the SU.The records do, however, diverge in two key intervals: first during the distinct limestone bed below the onset of the PETM, between ~ -0.8 and 0 m (~56.03 to 55.95 Ma), and then in the later stage of the CIE recovery between ~4.4 and 4.7m (55.82 to 55.81 Ma) (Figures 4  and 5).In the first of these, the distinct high carbonate content of the limestone unit is weakly 10 represented in carbonate MAR, indicating that this is the result of clastic sediment starvation, rather than increased carbonate production.The second horizon occurs during the CIE recovery phase, when carbonate MARs nearly treble compared to pre-PETM values before stabilizing at levels ~50% higher than pre-PETM conditions during the post-PETM phase above 4.7 m (~55.81Ma).This persistent increase in carbonate MARs above the PETM is reflected in consistently higher carbonate contents 15 (CaCO 3 wt %) of sediments above the PETM than those in the latest Paleocene.This is clear evidence for significantly increased carbonate accumulation rates post-PETM, even in relatively shallow (~1000m) supra-lysoclinal depths on the continental margins.The rate and timing of this recovery in carbonate MARs is also remarkable, with not only recovery to background levels, but also this switch to significantly elevated MARs all occurring within the main phase of the CIE recovery in less than 20 ka 20 (Figure 5).
The mass accumulation rate of non-carbonate components is here considered equal to the detrital MAR, in the absence of any significant contribution of biogenic silica.Before the CIE, the detrital MARs are relatively low, fluctuating around 2 g cm -2 ka -1 , with a further decline during the precursor limestone 25 unit (-0.8 to 0 m).During the onset of the PETM (0.3 -1.5 m) detrital fluxes increase dramatically, to between three and four times pre-PETM levels.Within the peak-PETM (1.5-1.7 m; ~55.88 to 55.85 Ma), detrital flux declines slightly and shows variability on similar timescales to Si/Fe ratios.It then increases markedly towards the end of the peak-PETM interval, above 3.7 m (~55.85Ma), and reaches peak values for the whole PETM during the main phase of the δ 13 C recovery interval (3.7 -4.2 m; 30 ~55.83 to 55.82 Ma).Both detrital and carbonate MAR increase, to peak values for the whole PETM, during the main interval of CIE recovery, with detrital MARs peaking first, followed by carbonate MARs.We note that these high detrital mass accumulations rates during the body of the PETM do not coincide with a noticeable increase in grain size, with sediments dominated by clay-grade material throughout.This is consistent with the marked absence of coarser grained turbidite deposition during 35 the PETM within the Zumaia section (Clare et al., 2015), although major increases in coarse grained siliciclastic deposits during the PETM are recorded within deep-water channels within the basin (Pujalte et al., 2015).Total organic carbon MARs are relatively stable before the onset of the event, at ~0.01 g cm -2 ka -1 .At the onset of the PETM there is an increase in TOC MARs, but this lags the CIE onset by ~20 ka (Figure 5).During the body of the PETM there are distinct peaks in TOC MARs at ~55.90 Ma 40  and Schmidt, 2010).We also include a 0-D (global average) temperature-dependent terrestrial 10 weathering feedback as described in (Colbourn et al., 2013).We follow a two-part spin-up procedure, first for 20 ka as a 'closed' system (i.e.weathering exactly balancing sedimentation) and with atmospheric CO 2 set to 834 ppm with a δ 13 C of -4.9‰.A further 250 ka 'open system' spin-up with temperature-dependent weathering feedbacks enabled and input of CO 2 via volcanic outgassing (Colbourn et al., 2013) was used to bring the geological 13 C cycle to steady state.In this, the specified 15 detrital flux to the sediments (non-carbonate) (0.18 g cm -2 ka -1 ), the rain ratio of carbonate to particulate organic carbon (0.2), and the global average concentrations of Mg 2+ and Ca 2+ (29.89 and 18.22 mmol kg -1 , respectively) and alkalinity (1975 µmol kg -1 ) were selected to give an average global mean sedimentary carbonate preservation of 46.75%, comparable with late Paleocene estimates (Panchuk et al., 2008).20 Four inversion experiments were undertaken, all forced with carbon (CO 2 ) inputs to the atmosphere such that the δ 13 C of surface ocean dissolved inorganic carbon (DIC) followed a prescribed inversion target taken from the Zumaia δ 13 C record, adjusted to the initial cGENIE average surface ocean DIC δ 13 C of +2.7‰.Within the very low carbonate interval, where the primary signals appear to be lost due 25 to the disproportionate influence of non-autochthonous carbonates, the δ 13 C target curve was linearly interpolated between the data points at ~19 and 78 ka.Experiments included variation in the isotope composition of the source (-60‰ and -22‰) and CIE recovery driven by silicate weathering only and with active carbon removal to match the CIE profile (Table 1).In order to directly compare modelled and observed carbonate contents, all experiments also include the temporally varying observed changes 30 in detrital fluxes within the sediment model at the model grid-box corresponding to the paleolocation of Zumaia only.
Net input fluxes depend on the δ 13 C composition of the source (  2011;Gutjahr et al., 2017a;Panchuk et al., 2008).In this case the lower bound of carbon input (4154 PgC) is generated by an inversion of the CIE isotopic profile assuming a pure biogenic methane carbon source with isotope composition of -60‰.As an independent test of the reliability of these simulations, in terms of the estimated CIE magnitude and temporal dynamics, the modelled atmospheric δ 13 C excursion is compared to the best available terrestrial record of the CIE, from soil carbonates of the 5Bighorn Basin (Figure 6; Bowen et al., 2014).There is a strong correspondence between the two, in the shape of the onset and recovery, the magnitude of the CIE and the overall duration of the event.
Although these model runs have no independent second constraint on the carbon cycle perturbation through the PETM -for example a measure of surface ocean pH change (Gutjahr et al., 2017c) (Gutjahr et al., 2017c), but the dynamics of this release must be considerably quicker than assumed within the deep-ocean records recovered from Site DSDP Site 401, in order to satisfy the rate of CIE onset observed at Zumaia, which is ~3.3 ‰ over less than 5 ka.We argue that this CIE onset rate is robust, given that it is observed in both terrestrial (Bowen et al., 2015) and marine (this study) sections where the onset of the CIE is expanded and set within a robust intra-PETM cyclostratigraphic age 20 model.The CIE recorded at Zumaia, and within the Big Horn Basin (Bowen et al., 2015), of 4 ‰ or more, is also greater than that assumed in the favoured model simulations of Gutjahr et al. ( 2017) of 2.6 ‰, and is closer to their alternate run assuming a CIE of ~4 ‰, which reconstructs a source carbon source with a δ 13 C composition of ~ -17 ‰.Both these aspects of the Zumaia record -a CIE onset less than 5 ka, and a CIE with magnitude ~4 ‰ or more -requires the rapid release of carbon from a source 25 that is more isotopically-depleted than volcanic CO 2. Whether thermogenic methane generated by sill intrusion during the onset phase of the NAIP (Svensen et al., 2004) can meet the emission rates and carbon isotopic compositions required to generate the early stages of the CIE is yet to be resolved.

Carbon Removal During the PETM Recovery
Two fundamental unknowns remain about the recovery phase of the PETM: first, what is the ultimate 30 sink of thousands of petagrams of carbon when added to the ocean-atmopshere system as CO 2 ?; and, second, what controls the temporal dynamics of carbon flux into this sink, and thence controls the timing of the relatively rapid CIE and PETM climate recovery?The combination of the age-constrained CIE records from Zumaia, the inverse Earth System modelling of this CIE and the general excellent agreement between the atmospheric CIE generated in these simulations with independent records from 35 the terrestrial realm (Figure 6; Bowen et al., 2015), provide new insights into these key questions.In simulations without active carbon removal, the only process restoring atmospheric pCO 2 to pre-CIE levels is the silicate weathering feedback, where net removal depends on the size of the carbon input and associated temperature change.This "weathering only" experiment fails to match the observed CIE recovery, with minimal recovery of atmospheric CO 2 δ 13 C even when a significant mass of carbon has been removed from the atmosphere (Figure 6).This is due to the isotopic imbalance between the ultimate carbon sink in this case -marine carbonates (-1 to 4‰) -compared to the carbon source (-22 or -60 ‰).With the weathering feedback only, the rate of removal of carbon from the atmosphere (Figure 6), and associated climate recovery, also occurs on timescales significantly longer (> 100 ka) 5 than indicated by climate proxy data, where temperatures are closely coupled to the recovery of the CIE (Dunkley Jones et al., 2013).The configuration of the silicate weathering feedback within the model is thus an insufficient driver of PETM recovery on two counts: first its inability to drive recovery in the carbon isotopic composition of the exogenic carbon reservoir, and second, in its inability to match the rate of recovery of the climate system at the end of the PETM.The Zumaia record also provides 10 evidence for a rapid climate transition at the end of the PETM, independent of carbon-cycle system parameters (e.g.δ 13 C and CaCO 3 accumulation), in the observed rapid switch back to pre-PETM detrital fluxes during the recovery phase.This implies a rapid climate recovery -back to a hydrological and erosive system similar to pre-event conditions -that is very closely coupled to the CIE recovery (Figure 5).15 To correctly simulate both the CIE and the climate system (temperature, hydrology) recovery rates, the active removal of large masses of isotopically light carbon is required during the PETM recovery phase.Simple isotope mass balance requirements require a greater mass of carbon to be removed than was first released, if the isotopic composition of the sink is more positive (organic carbon -22 %) than that of the 20 source (biogenic methane -60 %).In these methane input / organic carbon burial simulations (Experiment 4), the required excess carbon removal (removal mass minus input mass) causes a remnant post-event climate forcing, with an ~60% decline in post-PETM atmospheric CO 2 levels relative to pre-PETM conditions (Figure 6), and a global mean surface temperature reduction of ~ 4ºC.At present there is no evidence for such a post-PETM cooling, which would suggest that the source and sink terms 25 are relatively close in their isotopic compositions and precludes a biogenic methane input / organic carbon burial scenario.
The simulations that best match the constraints of the CIE and PETM recovery are those with the active removal of carbon, focused on the main phase of the CIE recovery, and with an isotopic composition 30 close to that of the original carbon source.The failure of the silicate weathering feedback alone to match the rate and magnitude of CIE recovery is similar to the recent inverse modelling by Gutjahr et al. (2017), where 2,500 Pg of carbon removal, with isotopic composition of -30.5 ‰, was required to match the CIE.The larger masses of carbon removal required in the simulations presented here are due to the smaller magnitude CIE recorded at DSDP Site 401, the target CIE curve for Gutjahr et al. (2017), 35 and the isotopically lighter composition of the carbon sink used (-30.5 ‰) by Gutjahr et al. (2017) compared to this study (-22 ‰).
The question remains as to what is the driver of enhanced carbon burial during the CIE recovery interval?In the Zumaia records it is notable that exactly this interval -the CIE recovery phase -is 40 directly coupled with maxima in total sedimentation rates and detrital mass accumulation rates (Figure 5).There is also a close coupling between these peak detrital fluxes and a transient peak in carbonate MAR, which is exactly the coupling predicted if increased detrital accumulation is accompanied by the widespread burial of organic carbon, with a resultant rapid increase in ocean pH and carbonate saturation state (Bowen and Zachos, 2010).Inverse Earth System modelling of the CIE at this location demonstrates that this interval of peak MARs (~80 to 120 ka after onset) is coincident with ~8700 Pg of carbon drawdown from the ocean-atmosphere system in the -22 ‰ δ 13 C input and removal scenario, 5 equivalent to ~87 % of total removal fluxes.The magnitude of increase in detrital accumulation rates at the end of the event (from <6 to >8 g cm -2 ka -1 ) is also of the same order as the ~40% increase in organic carbon burial rates estimated for the CIE recovery (Bowen and Zachos, 2010).Although Zumaia is a single PETM record from a single location, it has striking similarities to other Tethyan deep-water successions, such as Forada in Italy (Giusberti et al., 2016), and is likely representative of 10 regional-scale dynamics of climate and the associated response of continental margin sedimentary systems.Within the main phase of the PETM, variability within the Forada section is interpreted to be precession-paced wet and arid phases associated with a monsoonal-type subtropical climate (Giusberti et al., 2016).Measured TOC MARs through the Zumaia section are elevated during the recovery interval, ~55.83 to 55.81 Ma (Figure 5), although these data are not able to discriminate between the 15 burial fluxes of old (pre-PETM) reworked carbon and those for newly produced (intra-PETM) organic carbon -only the later will contribute to the removal of carbon from the exogenic carbon cycle.The Zumaia records cannot yet be used to trace the rates of burial of new-formed organic carbon through the PETM recovery phase, however they do provide a feasible mechanism for the timing of the recovery phase of the PETM, namely a strong precession-forcing of tropical and sub-tropical hydrological 20 regimes and associated sediment erosion, flux and accumulation.Alongside the observed precessionforced variations in sediment accumulation, there is likely a complex influence of temperature and CO 2 on the balance between rates of net primary production and organic carbon remineralization during the PETM (Cotton et al., 2015).This production / remineralization balance will control the availability of newly formed sedimentary organic carbon for long-term burial in the sedimentary carbon sink (Cotton 25 et al., 2015).During conditions of peak PETM warmth, enhanced organic carbon remineralization in terrestrial systems may have drastically limited the flux to continental margin systems, but with even a gradual recovery from peak-warmth conditions -forced, for example, by known enhanced carbon burial in the anoxic marine basins of the eastern Tethys during the peak-PETM (Carmichael et al., 2017) -a subtle shift in the production to respiration balance towards the availability of organic carbon for long-30 term burial, combined with a precession-forced peak in sediment flux and accumulation rates across terrestrial and continental margin systems, could account for the magnitude and timing of the sudden switch to widespread carbon sequestration required during the CIE recovery phase.The greatly enhanced detrital accumulation rates observed in continental margin settings such as Zumaia could also effectively mask patterns of enhanced burial of newly formed organic matter (Sluijs et al., 2014), unless 35 sedimentation rates within the PETM can be accurately constrained.Future studies should also focus on discriminating the relative contributions of newly formed (intra-PETM) and reworked "fossil" (pre-PETM) organic carbon within these continental margin successions, as only the former will contribute to carbon sequestration.

Conclusions
New high-resolution bulk sediment chemistry records across the PETM interval of the Zumaia section allow, for the first time, the extension of the well-established precession-based cyclostratigraphy through the characteristic low carbonate siliciclastic unit that marks peak-PETM conditions.Ten intra-PETM cycles in bulk sediment Si/Fe ratios are consistent with the newly established cyclostratigraphic 5 framework for the Big Horn Basin and its correlation to classic PETM deep-ocean sites (Westerhold et al., 2017).This new cyclostratigraphic framework for Zumaia, together with new bulk carbonate isotopic data and Earth System modelling of the Zumaia PETM CIE, constrains some key components of the PETM event: • CIE onset rates and magnitude recorded at Zumaia are in close agreement with soil carbonate nodule 10 data from the Big Horn Basin, and imply PETM onset times of less than 5000 years, with a CIE magnitude of at least ~4 ‰.
• During the deposition of the characteristic PETM siliciclastic unit of Zumaia, sedimentation rates double and the mass accumulation rates of detrital sediments increase by up to four times relative to pre-event conditions.15 • Dramatic increases in sedimentation rates during the main phase of the PETM are interpreted to represent significant changes in regional hydrology, most likely an increase in the magnitude and frequency of extreme rainfall and runoff events, potentially within an overall lower mean annual precipitation regime (Carmichael et al., 2017;Carmichael et al., 2016).These changes in sedimentation accumulation rates -and inferred perturbations to hydrology -persist throughout the PETM interval, 20 only returning towards pre-event levels during the rapid phase of CIE recovery.This supports a close coupling between the CIE and the climate system.
• During the PETM bulk sediment composition and flux vary on timescales consistent with a component of precession forcing of regional subtropical hydroclimates.This infers a significant role for orbital-paced insolation variability in pacing global climate throughout the extreme climate state of the 25 PETM.
• Earth System Model (cGENIE) inversion experiments, designed to replicate the observed CIE dynamics at Zumaia, reproduce an atmospheric CIE that is close to the best available terrestrial records of the CIE (Bowen et al., 2015), in timing and magnitudes.Depending on the δ 13 C composition of the source of PETM carbon, estimated total carbon inputs during the onset of the PETM are ~3,400 Pg C 30 (methane source, -60 ‰) or ~11,300 Pg C (organic carbon source, -22 ‰).
• Simulations with a silicate weathering only feedback and mechanism for carbon cycle recovery fail to reproduce the observed record in two key ways: first, they are unable to recover the coupled CIEclimate system at a rate fast enough to match the observations (<20 ka); and, second, the final sink of carbon within this feedback -marine carbonates -is isotopically too heavy to provide any significant 35 recovery of the CIE.
• Simulations that match the patterns of CIE recovery require active organic or methane carbon removal, of a magnitude similar to the original carbon input.Simulations with an isotopic mismatch between source carbon (methane -60 ‰) and sink carbon (organic matter -22 ‰) require significantly more carbon removal at the end of the event than originally released during the onset.This causes a 40 marked post event cooling of ~4 ºC relative to pre-event conditions that is inconsistent with current climate records.• Peak sedimentation rates and detrital accumulation rates are observed through the recovery phase of the CIE and not the onset.We suggest that this peak in sedimentation rates is coupled to the precession forcing of subtropical terrestrial hydrology and that this intra-PETM variability provides a key to 5 understanding the timing of the recovery phase of the CIE.A widespread, precession-forced peak in sediment flux within tropical to sub-tropical environments, potentially coupled to increased rainfall, runoff and sediment transport, provides a mechanism for increased organic carbon burial across a range of terrestrial and continental margin sediment sinks.If this is coupled with a gradual change in the temperature / CO 2 -controlled balance between net primary production to organic carbon 10 remineralization rates (Cotton et al., 2015), such that organic carbon becomes more available for burial as temperatures begin to recover from peak-PETM conditions, this may have the potential to provide a rapid recovery feedback through a cycle of cooling and enhanced organic carbon burial.
The results presented here demonstrate that the Earth System was capable of the sequestration of several 15 thousand petagrams of carbon at the end of the PETM, but only after nearly one hundred thousand years within a profoundly altered climate state.The mechanism of carbon burial also appears to rely on a radical perturbation to global hydroclimates, sufficient to cause erosion and dramatically increased sediment flux rates to continental margins between three and four times pre-perturbation levels.
into the Zumaia section creates an expanded rather than condensed PETM onset with far greater stratigraphic and temporal fidelity.Once plotted on a common age model the dissolution-controlled lag in time to peak CIE conditions in the deep-ocean bulk δ 13 C is clearly visible, and of the order of 70 ka even in the shallowest and most complete sections (ODP Sites 690 & 1263; Clim.Past Discuss., https://doi.org/10.5194/cp-2017-131Manuscript under review for journal Clim.Past Discussion started: 16 November 2017 c Author(s) 2017.CC BY 4.0 License.peak CIE ~55.86 Ma; Figure 5).Once peak values are reached, the deep-ocean sites show a similar quick CIE recovery phase to Zumaia between 55.85 and 55.82 Ma.
Clim.Past Discuss., https://doi.org/10.5194/cp-2017-131Manuscript under review for journal Clim.Past Discussion started: 16 November 2017 c Author(s) 2017.CC BY 4.0 License.and ~55.85 Ma, but in general values remain between 0.015 and 0.02 g cm -2 ka -1 .Coincident with recovery the CIE recovery phase, TOC MARs increase again to over ~0.02 g cm -2 ka -1 between 55.83 and 55.81 Ma. 3 Inverse Modelling the Zumaia CIE and the PETM carbon source To constrain the rates of carbon input and sequestration through the onset and termination of the PETM, 5 based on the CIE profile and cyclostratigraphy from Zumaia, we undertook a series of PETM simulations using the late Palaeocene-early Eocene configuration of the cGENIE Earth System Model.This includes a 3-D dynamic ocean model, 2D energy balance atmospheric model, and a representation of marine biogeochemical cycling plus the preservation of carbonates in deep-sea sediments (Ridgwell or deep ocean carbonate saturation state(Panchuk et al., 2008;Zeebe et al., 2009) -model simulations with a 10 mixed source input with δ 13 C of -22 ‰ are close in total mass input across the PETM onset (11,316 Pg C) to the preferred simulations of Gutjahr et al. (2017; ~10,200 Pg C), which are constrained by surface ocean pH changes.The combination of these two records -pH from DSDP Site 401 and the CIE dynamics from Zumaia -is, however, informative.Total carbon input masses, constrained by the most detailed surface ocean pH records of the PETM available to date, are modelled to be >10,000 Pg C 15

Figure 1 :
Figure 1: Location of Zumaia section and relationship to the terrestrial and shallow shelf carbonate depositional environments; after Pujalte et al. (2015).

Figure 2 .
Figure 2. New sediment geochemistry data for the Zumaia section.From left to right: lithological log produced during sample collection for this study (2011); bulk carbonate δ 13 C; SiO 2 weight percent and Si/Fe ratio based on weight percents of oxides (SiO 2 /Fe 2 O 3 ) from bulk sediment XRF analyses, pre-PETM and post-PETM precession cycles numbered in black, intra-PETM half-precession cycles numbered in red; Si/Fe ratios with Gaussian filters of the dominant cycles 0.45m (red) and 1m (blue);5

Figure 3 .
Figure 3. MTM power spectra for Si/Fe ratios; highlighted are spectral peaks equivalent to cycles with period >1 m and ~0.45 m.

Figure 4 .
Figure 4. New carbonate, detrital and organic carbon mass accumulation and sedimentation rates for the Zumaia section.From left to right: lithological log produced during sample collection for this study (2011); bulk carbonate δ 13 C; weight percent CaCO 3 derived from XRF CaCO calibrated to discrete CaCO 3 analyses (see methods); CaCO 3 mass accumulation rate (light blue);5

Figure 5 . 5 from 10 15
Figure 5.Comparison of the Zumaia records against terrestrial (Big Horn Basin) and deep-ocean sites (Walvis Ridge, ODP Sites 1262, 1263, 1265, 1266, 1267; Maud Rise ODP Site 690) on the same cyclostratigraphic age model.From left to right: bulk carbonate δ 13 C for deep ocean sites; bulk carbonate δ 13 C for Zumaia; soil nodule δ 13 C from the Big Horn Basin; CaCO 3 records 5

Figure 6 .
Figure 6.Results of inverse modelling with cGENIE Earth System Model.From left to right: the bulk carbonate δ 13 C for Zumaia which drives model inversions (black); soil nodule δ 13 C record of the Big Horn Basin (yellow), as representative of atmospheric CO 2 δ 13 C for comparison to model results; modelled δ 13 C composition of atmospheric carbon dioxide; carbon emission rate in PgC5

Table 1 )
, with smaller carbon inputs required for a source with a more negative isotope composition.Total carbon inputs are also slightly 35 larger where no removal fluxes are allowed (Experiment 1).In all experiments, the majority of C input occurs in the first 20 ka with a low rate of C input (<0.05 Pg C yr-1) required to maintain low isotope values until 78 ka.Total modelled C input is broadly comparable to previous model-based estimates constrained by deep-ocean carbonate dissolution that exceed ~4000 Pg C (Bowen et al., 2015; Cui et al., Clim.Past Discuss., https://doi.org/10.5194/cp-2017-131Manuscript under review for journal Clim.Past Discussion started: 16 November 2017 c Author(s) 2017.CC BY 4.0 License.

Table 1 .
20 ACKNOWLEDGEMENTS T.DJ acknowledges funding from a Royal Society Dorothy Hodgkin Fellowship, L.A. project CGL2014-58794-P (Spanish Ministry of Economy and Competitiveness and FEDER funds), and R.D.P. acknowledges the Royal Society Wolfson Research Merit Award.This research was supported by the Deutsche Forschungsgemeinschaft (DFG) to T.W.We would also like to thank Kernow Analytical 25 Technology for use of the T-SEP TM apparatus, Victoriano Pujalte for guidance in accessing sampling sites, Enrique Cantero and Georgia Hole for assistance in sampling, and Chris Kendrick for assistance with TOC and d 13 C TOC analysis.Finally, we thank Dr Ian Bull, Alison Kuhl and James Williams of the Natural Environment Research Council Life Sciences Mass Spectrometry Facility for technical assistance.30 Summary details of cGENIE simulations detailing isotopic compositions of source carboninput during the onset of the PETM to match Zumaia CIE profile -and the nature of modelled CIE recovery, whether by silicate weathering only (Experiment 1), or by active carbon removal.The isotopic composition of this carbon removal ("Sink") is also listed.Net C fluxes are calculated by the model to 5 match the CIE profile and do not include carbon removed through silicate weathering.