Spatial and Temporal Patterns in Petrogenic Organic Carbon Mobilization During the Paleocene‐Eocene Thermal Maximum

The Paleocene‐Eocene Thermal Maximum (PETM) was a transient global warming event and is recognized in the geologic record by a prolonged negative carbon isotope excursion (CIE). The onset of the CIE was due to a rapid influx of 13C‐depleted carbon into the ocean‐atmosphere system. However, the mechanisms required to sustain the negative CIE remains unclear. Enhanced mobilization and oxidation of petrogenic organic carbon (OCpetro) has been invoked to explain elevated atmospheric carbon dioxide concentrations after the onset of the CIE. However, existing evidence is limited to the mid‐latitudes and subtropics. Here, we determine whether: (a) enhanced mobilization and subsequent burial of OCpetro in marine sediments was a global phenomenon; and (b) whether it occurred throughout the PETM. To achieve this, we utilize a lipid biomarker approach to trace and quantify OCpetro burial in a global compilation of PETM‐aged shallow marine sites (n = 7, including five new sites). Our results confirm that OCpetro mass accumulation rates (MARs) increased within the subtropics and mid‐latitudes during the PETM, consistent with evidence of higher physical erosion rates and intense episodic rainfall events. High‐latitude sites do not exhibit drastic changes in the source of organic carbon during the PETM and OCpetro MARs increase slightly or remain stable, perhaps due a more stable hydrological regime. Crucially, we also demonstrate that OCpetro MARs remained elevated during the recovery phase of the PETM. Although OCpetro oxidation was likely an important positive feedback mechanism throughout the PETM, we show that this feedback was both spatially and temporally variable.

• We assess spatial and temporal patterns in petrogenic organic carbon (OC petro ) mobilization during the Paleocene-Eocene Thermal Maximum • Evidence for enhanced OC petro mobilization in the subtropics and mid-latitudes, likely due to an increase in extreme rainfall events • OC petro mobilization remained elevated during the recovery phase of the Paleocene-Eocene Thermal Maximum Supporting Information: Supporting Information may be found in the online version of this article.
Whether it acts as a positive or negative feedback mechanism largely depends on whether the organic carbon (OC) is "biospheric" (OC bio ), representing relatively recent thermally immature organic carbon (10 2 -10 4 years old; e.g., vegetation and soils), or "petrogenic" (OC petro ), representing ancient rock-derived and thermally mature organic carbon (>10 6 years old; e.g., organic carbon-rich shales).Erosion, mobilization, and the subsequent burial of OC bio in marine sediments helps to sequester CO 2 (Berhe et al., 2007;Stallard, 1998).In contrast, exhumation and oxidation of OC petro during lateral transport from land-to-sea can release CO 2 (Petsch et al., 2000).In modern settings, up to ∼90% of OC petro is oxidized in large catchments such as the Amazon and Himalayan range (e.g., Bouchez et al., 2010;Galy et al., 2008), whereas a lower proportion (∼10%-40%) of OC petro is oxidized in mountain basins with steep rivers (e.g., Hilton et al., 2011Hilton et al., , 2014)).Crucially, regardless of individual catchment dynamics, OC petro has the potential to be oxidized and increase atmospheric CO 2 concentrations.
Several studies have quantified the mobilization and burial of OC petro in modern systems (e.g., Blair et al., 2003;Clark et al., 2017Clark et al., , 2022; T. I. Eglinton et al., 2021 and references therein;Galy et al., 2007, 2015 andreferences therein;Hilton & West, 2020;Hilton et al., 2010, 2011 andreferences therein;Smith et al., 2013) and Holocene sediments (e.g., Hilton et al., 2015;Kao et al., 2008Kao et al., , 2014)).These studies show that erosion and transport of OC petro is largely controlled by a combination of geomorphic and climate processes (e.g., T. I. Eglinton et al., 2021;Hilton, 2017).For example, extreme rainfall events can trigger bedrock landslides (e.g., Hilton et al., 2008) and/or create deeply incised gullies (e.g., Leithold et al., 2006), both of which can expose OC petro to oxidation.However, clastic sediments from hyperpycnal flows and turbidites can act to enhance the preservation of OC petro (e.g., Bouchez et al., 2014;Hilton et al., 2011).As climate model simulations indicate an intensification of the hydrological cycle in response to rising atmospheric CO 2 levels and global temperatures (Lee et al., 2021), the delivery of OC petro to the oceans will likely be enhanced in the future.However, such predictions are based on present-day observations and/or past climate states that span a lower-than-modern atmospheric CO 2 values (e.g., Hilton & West, 2020;Kao et al., 2008).
The geologic record enables investigations into high CO 2 states of the past, providing unique insights into how terrestrial carbon cycle processes may operate in the future.Many studies have focused on the Paleocene-Eocene Thermal Maximum (PETM; ∼56 million years ago) (McInerney & Wing, 2011), a transient carbon cycle perturbation characterized by global warming (∼4-6°C; Inglis et al., 2020;Tierney et al., 2022) and an intensified hydrological cycle (Carmichael et al., 2017 and references therein).The PETM is identified in the geologic record by a negative carbon isotope excursion (CIE) (−4 ± 0.4‰; Elling et al., 2019).The onset of the PETM is on the order-of-millennia (Kirtland Turner, 2018;Zeebe et al., 2014) and is followed by sustained low and stable carbon isotope (δ 13 C) values for ∼94-170 thousand years (Kyrs) (Zeebe & Lourens, 2019), referred to as the "body" of the CIE (Bowen et al., 2006).The body is then followed by a long recovery of ∼50-120 Kyrs (Bowen, 2013;Murphy et al., 2010;Zeebe et al., 2009), which is further divided into Phase I (initial rapid rise in δ 13 C) and Phase II (final gradual rise in δ 13 C) (Röhl et al., 2007).
The onset of the CIE was the result of a rapid influx of 13 C-depleted carbon from one or more reservoirs outside the active global exogenic carbon pool (Dickens et al., 1997).Proposed reservoirs include submarine methane hydrates (Dickens, 2011;Dickens et al., 1995), terrestrial organic carbon (Bowen, 2013;Deconto et al., 2012;Kurtz et al., 2003), and volcanic carbon related to the North Atlantic Igneous Province (Gutjahr et al., 2017;Jones et al., 2019;Storey et al., 2007;Svensen et al., 2004).Less explored are the mechanisms responsible for the prolonged body of the CIE.This feature requires continual input of 13 C-depleted carbon (e.g., Zeebe et al., 2009), thus several feedback mechanisms (either acting individually or in combination) have been proposed.This includes a slow dissociation of oceanic methane hydrates (Zeebe, 2013), pulsed releases of thermogenic methane from vent complexes (e.g., Frieling et al., 2016;Kirtland Turner, 2018), and/or "leaky" terrestrial organic carbon reservoirs (Bowen, 2013).Alternatively, recent work suggests that CO 2 released from OC petro oxidation could explain the extended body of the CIE (Lyons et al., 2019).This theory is based on evidence for an order-of-magnitude increase in the delivery of OC petro to the oceans, ∼10-20 Kyrs after the onset of the PETM.However, this study was limited to the mid-latitudes (Atlantic Coastal Plain) and subtropics (Tanzania), and therefore may not be globally representative.It is also unclear whether enhanced mobilization of OC petro was a persistent feature throughout the PETM or whether it was restricted to the body interval.
Here, we use lipid biomarker thermal maturity ratios to fingerprint OC petro burial in a global compilation of PETMaged shallow marine sites (n = 7, including five new sites).Lipid biomarkers undergo various structural alterations with increasing thermal maturity (e.g., defunctionalization, isomerization, catagenesis, and aromatization; Peters et al., 2005) and thus, can be used to assess the proportion of OC petro in marine sediments (Lyons et al., 2019).We focus on thermally immature, shallow marine sediments as they are "hotspots" for terrestrial organic carbon input (Bianchi et al., 2018).We quantify OC petro burial fluxes before and during the PETM, using a two endmember mixing model.Overall, we aim to determine whether: (a) enhanced mobilization and subsequent burial of OC petro in the ocean was a global phenomenon; and (b) whether it occurred throughout the PETM.

Data Compilation
New n-alkane-and/or hopane-based thermal maturity ratios were acquired from the following PETM-aged shallow marine sites: the  1).Additional information (e.g., paleodepth) and a brief description of the lithology for each site can be found within Table S1 and Text S1 in Supporting Information S1, respectively.We also compile n-alkane-and/ or hopane-based thermal maturity ratios from the following published PETM-aged shallow marine sites: TDP Site 14 (Carmichael et al., 2017;Handley et al., 2012); South Dover Bridge (SDB) (Lyons et al., 2019); and Cambridge-Dorchester Airport (CamDor) (Lyons et al., 2019) (Figure 1).Other published biomarker records are available for PETM-aged shallow marine sites, however these sequences are dominated by autochthonous OC petro and show evidence for post-depositional diagenesis (Cui et al., 2021;Handley et al., 2011).

Organic Geochemistry
For this study, samples from ACEX (n = 94), ODP Site 1172 (n = 41), and Ancora (n = 42) were freeze dried, homogenized, and extracted using a MARS5 microwave-assisted extraction system, with: (a) dichloromethane:methanol (DCM:MeOH; 1:1, v:v); (b) DCM:MeOH (9:1, v:v); and (c) DCM, at Harvard University (see Elling et al., 2019).Each solvent mixture was heated for 30 min to 100°C, followed by a hold time of 20 min.The extracts from the three steps were combined into a total lipid extract (TLE) and further divided into five fractions (following Polik et al. (2018)).At the University of Southampton, extracted copper was added to the apolar fractions for 24 hr to remove elemental sulfur.The apolar fractions were then analyzed using a ThermoFisher Trace 1310 gas chromatograph (GC) coupled to a Thermo TSQ8000 Triple Quadrupole mass spectrometer (MS).Helium was used as the carrier gas and separation was achieved with DB-5 column (30 m × 0.25 mm i.d., 0.25 μm film thickness).The GC oven program started at 70°C for 1 min, increased to 130°C at 20°C min −1 , followed by 300°C at 4°C min −1 , which was then held for 20 min.MS scanning occurred between mass-to-charge ratio (m/z) 50 to 650 Da, and an ionization energy of 70 eV.Compound identification was based on retention times, fragmentation patterns, comparison to an in-house standard, and library matches.Kheu River samples (n = 39) were extracted at the University of Bristol by ultrasonicating homogenized samples sequentially with DCM, DCM:MeOH (1:1, v:v), and MeOH.Elemental sulfur was removed from the combined TLE using activated copper turnings.An activated silica column with saturated ammonia in chloroform and chloroform:acetic acid (100:1, v:v) was used to separate the neutral and acid fraction, respectively.The apolar fraction was split from the neutral fraction by eluting with hexane:DCM (9:1, v:v) via separation on an alumina column.The apolar fractions were then analyzed at the University of Bristol on a Thermoquest Finnigan Trace GC interfaced with a Thermoquest Finnigan Trace MS.The GC was fitted with a fused capillary column (50 m × 0.32 mm i.d.) and the carrier gas was helium.The samples were suspended in ethyl acetate and injected at 70°C.The temperature program increased to 130°C (20°C min −1 ), then 300°C (4°C min −1 ), and finally remained isothermal for 20 min.The MS operated with an electron ionization source at 70 eV, scanning over m/z ranges of 50-850 Da.The compounds were integrated on the total ion chromatogram (TIC).
Additional samples (n = 12) from TDP Site 14 were homogenized and extracted at the University of Bristol.
Extractions were achieved via Soxhlet apparatus overnight, using DCM:MeOH (2:1 v:v).The TLE was suspended in hexane:DCM (9:1, v:v) and separated by alumina column chromatography.Co-eluting compounds and/or unresolved complex mixtures were reduced with urea adduction (following Pancost et al., 2008).Elemental sulfur was removed using extracted copper turnings.The apolar fractions were analyzed at the University of Bristol on the same GC-MS as used for Kheu River.The GC was fitted with a CPsil-5CB column (Agilent Technologies, dimethylpolysiloxane stationary phase) and the carrier gas was helium.The samples were injected in ethyl acetate at 70°C.The temperature program increased to 130°C (20°C min −1 ), then 300°C (4°C min −1 ), and finally held for 25 min.The MS operated with an electron ionization source at 70 eV, scanning over m/z ranges of 50-850 Da.
The compounds were integrated on the TIC or using the appropriate mass fragment (e.g., m/z 191).

n-Alkane-Based Thermal Maturity Ratios
Modern plants and sediments contain long-chain n-alkanes with an odd-over-even preference (G.Eglinton & Hamilton, 1967), however this is progressively lost during diagenesis.The shift away from a dominance of long-chain n-alkanes with an odd-over-even predominance is captured by the carbon preference index (CPI) (Bush & McInerney, 2013).Modern sediments exhibit high CPI values (>3-30), indicating relatively unaltered thermally immature organic matter (Diefendorf & Freimuth, 2017).In contrast, mature organic matter (e.g., coal and oil) exhibits low CPI values (∼1).CPI values <1 are less common, and typify low-maturity source rocks from carbonates or hypersaline environments.In this study, sites with extensive post-depositional diagenesis were excluded, such that CPI values closer to 1 likely suggests input of allochthonous thermally mature organic matter (e.g., OC petro ).Here, we use the equation as originally defined by Bray and Evans (1961): (1)

Hopane-Based Thermal Maturity Ratios
Hopanes are the diagenetic products of biohopanoids, which are produced by a wide diversity of bacteria and consequently ubiquitous in a range of environments (Kusch & Rush, 2022).The ratios between different hopanes and their various stereoisomers have long been utilized as a thermal maturity proxy in the field of petroleum geochemistry (e.g., Farrimond et al., 1998;Mackenzie et al., 1980).Most of the hopane-based thermal maturity ratios used in this study are normalized (with the exception of Equation 4).Values indicating high thermal maturity likely suggests allochthonous older material (e.g., pre-PETM-aged OC petro ), as sites with post-depositional diagenesis were excluded from this study.We use a multi-ratio approach as each ratio corresponds to different stages of maturity relative to the oil window (i.e., from early diagenesis to the generation of oil), thus enabling insight on the degree of thermal maturation (Figure S1 in Supporting Information S1).However, hopane distributions also vary depending on the lithofacies and/or depositional environment (Peters et al., 2005).Therefore, without knowledge of the source rock at each locality, comparison between the sites should be undertaken with caution.
With the exception of Frankia spp.(Rosa-Putra et al., 2001), all bacteria synthesize hopanoids with a 17β, 21β configuration.However, this changes to a more stable βα and then αβ configuration during early diagenesis and then peak oil generation, respectively (Farrimond et al., 1998;Mackenzie et al., 1980).The shift from ββ to αβ is expressed via the following equation (sometimes referred in literature as "hopanoid isomerization"): Higher thermal maturity is marked by values closer to 1.This equation is applied to the hopanes that contained both isomers (i.e., mostly C 29-31 hopanes).However, caution should be taken when interpreting sediments with input from peats, as C 31 αβ isomers dominate the hopane distribution within acidic wetland environments (Inglis et al., 2018).
The shift from βα (also referred to as moretane; M) to the more stable αβ (also referred to as hopane; H) is assessed via the following equation (sometimes referred in literature as "moretane/hopane ratio"): This equation is applied to the most commonly used C 30 hopane (e.g., French et al., 2012), as well as the less commonly used C 29 hopane (Peters et al., 2005).Values closer to ∼0 indicate higher thermal maturity and oil generation.
The C 29 αβ hopane (also referred to as norhopane; N) is more thermally stable than C 30 αβ hopane.This is assessed via the following equation (sometimes referred in literature as "norhopane/hopane ratio"): As well as a thermal maturity proxy, this ratio has been utilized to differentiate between anoxic carbonate and/or marl source rocks (>1) versus clay-rich source rocks (<1) (Peters et al., 2005).
Toward the early stages of oil generation, there is a change in stereochemistry at the C-22 position, from the biologically favored R configuration to a near equal mix of R and S (Farrimond et al., 1998;Mackenzie et al., 1980;Peters et al., 2005).This is expressed via the following equation (sometimes referred in literature as "homohopane isomerization"): This equation uses C 31-35 hopanes (also referred to as homohopanes) and approaches maximum (equilibrium) values of ∼0.6 as thermal maturity increases and oil is generated.

Two-Endmember Mixing Model
The fraction of OC petro (f petro ) was calculated for each hopane-based thermal maturity ratio (X mix ; Table 1), following the two endmember mixing model from Lyons et al. (2019): where X background and X petro is the defined immature and mature endmembers, respectively.The endmembers for C 31-35 S/(S + R) ratio follow the definitions in Lyons et al. (2019), where X background is the contemporaneous carbon value of 0 and X petro is the most thermally mature value of 0.6.The endmembers for C 29-30 βα/(βα + αβ) ratio also follow the definitions in Lyons et al. (2019), where X background is 1 and X petro is 0. For this study, the endmembers of the αβ/(αβ + ββ) ratio was defined as 0 for X background and 1 for X petro .Note that C 29 αβ/C 30 αβ and T s /(T s + T m ) ratios were excluded due to their strong dependence on the source rock and/or depositional environment (Peters et al., 2005).

Mass Accumulation Rates
The mass accumulation rate (MAR; in gC cm 2 Kyr −1 ) of OC petro was calculated for all the new and published f petro data, following Lyons et al. (2019): where LSR is the linear sedimentation rate (cm Kyr −1 ), ρ is the dry bulk density (g cm −3 ), and TOC is the total organic carbon (TOC) or C org (%) (see Table 1).As published bulk density values are only available for one site (ODP Site 1172), a constant ρ value of 1.8 g cm −3 was assumed across all the sites (following Dunkley Jones et al., 2018).However, we acknowledge that changes in dry bulk density may influence absolute MARs, especially in sites with major lithological changes (see Text S1 in Supporting Information S1).The TOC values and LSR were acquired for each location from published studies (Table 1).C org records from ODP Site 1172 (Papadomanolaki et al., 2022) and TDP Site 14 (Aze et al., 2014) were linearly interpolated to match the depths of the biomarker data, using R Package Astrochron (Meyers, 2014).LSR estimates were obtained (where possible) for three key time intervals: (i) pre-PETM (Paleocene); (ii) the "core" (onset and body of the CIE) of the PETM; (iii) and the recovery of the PETM (see Text S1 in Supporting Information S1).This was available for all the sites with the exception of ODP Site 1172, which lacks the recovery interval.Note that the recovery at Ancora and SDB were further divided into: (iiia) Phase I; and (iiib) Phase II.Kheu River does not have LSR data, thus estimates were taken from the nearby Aktumsuk section (Uzbekistan; John et al., 2008).Both Kheu River and Aktumsuk comprises shallow marine deposits that exhibits TOC values from ∼0.1% pre-PETM to a maximum of ∼8.5% during the PETM (Bolle et al., 2000;Dickson et al., 2014).Similarly, LSRs from within the core interval of SDB was assumed to be the same for the entire PETM section at CamDor (following  2b) and hopane-based thermal maturity ratios exhibit relatively stable trends throughout the sequence, suggesting that the organic carbon source did not distinctly change.Note that potential information may be missing due poor core recovery between ∼388 and 384.5 mcd (Sluijs et al., 2006).However, C 30 αβ/(αβ + ββ) (Figure 2c), C 31 S/(S + R) (Figure 2d), and T s / (T s + T m ) (Figure 2f) values slightly increase (i.e., higher thermal maturity) between pre-PETM and the core of the PETM, by an average of 0.01, 0.01, and 0.08, respectively.These indices then decline during the recovery interval.C 31 αβ/(αβ + ββ) and C 30 βα/(βα + αβ) ratios (Figure 2c) exhibit the opposite trend, with lower thermal maturity during the core interval and the C 30 βα/(βα + αβ) ratio (Figure 2e) continuing to decline into the recovery of the PETM.

Kheu River
C 16 -C 35 n-alkanes were identified in the apolar fraction, in addition to C 27 -C 31 hopanes (including αβ, βα, and ββ isomers).Prior to the PETM and during the recovery, the CPI drops below 1 (Figure 4b), which may suggest input of low-maturity source rocks from carbonates or hypersaline environments.The CPI also oscillate drastically between ∼1 and ∼3 within the lower depths of the core of the PETM (∼0-50 cm; Figure 4b).This section of high variability is also reflected in the C 29 αβ/C 30 αβ (Figure 4d) and C 29 βα/(βα + αβ) (Figure 4e) ratios, suggesting rapid changes in the organic carbon source.However, it may also represent greater sampling resolution within the PETM.Overall, the average of all the thermal maturity ratios exhibit lower thermal maturity during the core interval.In addition, the C 29 αβ/C 30 αβ ratio present values >1 during the PETM (Figure 4d), potentially indicating input from a clay-rich source rock.With the exception of T s /(T s + T m ) (Figure 4f), all of the ratios increase in higher thermal maturity during the recovery to either higher than pre-PETM (i.e., C 29 αβ/(C 29 αβ + C 30 αβ) and C 29-30 βα/(βα + αβ) ratios) or near pre-PETM values (i.e., C 29-31 αβ/(αβ + ββ) ratio).

Ancora
The apolar fraction contains C 15 -C 34 n-alkanes and C 27 to C 31 hopanes (including αβ, βα, and ββ isomers).CPI ranges from 1 to 2.2 and is similar during the pre-PETM and PETM interval (Figure 5b).Similarly, C 30-31 αβ/(αβ + ββ) values remain relatively constant, albeit exhibiting a very slight decline by an average of 0.01-0.03(i.e., decreasing thermal maturity; Figure 5c).On the other hand, C 31 S/(S + R) (Figure 5d) and C 30 βα/(βα + αβ) (Figure 5e) values peak toward higher thermal maturity during the core of the PETM, reaching a maximum of 0.38 and 0.04, respectively.C 31 S/(S + R) values exhibit a drastic shift during the PETM (Figure 5d) and there is near equal mix of 22S and 22R isomers, suggesting potential transient input of thermally mature organic carbon.Changes in the C 31 S/(S + R) ratio and C 30 βα/(βα + αβ) ratio do not occur synchronously, instead C 31 S/(S + R) values lag behind by ∼1.5 mcd.The CPI remains >3 (i.e., low thermal maturity), with the exception of five data points which occur during the core of the PETM (Figure 6b).Most noticeable is the large variability in the hopane-based thermal maturity ratios pre-PETM and for the first ∼4 m of the core of the PETM.In the upper ∼5 m of the core of the PETM, the ratios are more stable and in general agreement.This interval mostly exhibits more thermally mature values than during the pre-PETM section (e.g., C 31 αβ/(αβ + ββ) increases by an average of 0.5; Figure 6c), suggesting a potential shift to an input of thermally mature organic carbon.For example, C 29-31 αβ/(αβ + ββ) values are close to its mature endmember of 1 (Figure 6c).

Enhanced OC petro Mass Accumulation Rates in the Subtropics and Mid-Latitudes During the PETM
A previous study from Tanzania (TDP Site 14) reported a relative increase in the thermally mature αβ hopanes during the PETM (Carmichael et al., 2017;Handley et al., 2012).Here, we present new hopane-based thermal maturity data that reveals rapidly fluctuating values within the first ∼4 m of the core of the PETM (Figure 6).Similar patterns were observed in the bulk δ 13 C of organic carbon (δ 13 C org ; Figure 6a), the n-alkane δ 13 C record, the chain-length distributions of n-alkanes, and the branched and isoprenoid tetraether (BIT) index (Figure S2 in Supporting Information S1; Aze et al., 2014;Carmichael et al., 2017;Handley et al., 2008Handley et al., , 2012)).The δ 13 C org and n-alkane δ 13 C records were previously suggested to reflect episodic reworking of older (pre-PETM) material rather than changes in the atmospheric carbon reservoir (Aze et al., 2014;Handley et al., 2008).The hopane-based thermal maturity ratios within this study confirms this variable delivery of OC petro .In contrast, the upper ∼5 m of the core of the PETM exhibits more stability in the hopane-based thermal maturity ratios (Carmichael et al., 2017;Handley et al., 2012), δ 13 C org values, and n-alkane δ 13 C values (Aze et al., 2014;Handley et al., 2008), indicating a switch from an episodic to persistent delivery of OC petro (Carmichael et al., 2017;Handley et al., 2012).
The hopane-based thermal maturity ratios also indicate that the OC petro within this interval is of higher thermal maturity.During the PETM, a rise in thermally mature hopanes and LSRs increases OC petro MARs by an average of 8 × 10 −3 gC cm 2 Kyr −1 (Figure 7).Enhanced OC petro MAR is consistent with a shift from predominantly marine organic carbon to a terrestrial organic carbon source (e.g., an increase in the abundance of long-chain n-alkanes produced by vascular plants and brGDGTs produced by soil bacteria; Carmichael et al., 2017;Handley et al., 2008Handley et al., , 2012)).Whilst there is greater LSR and terrigenous sediment during the PETM, C org values declined.This drop was attributed to the larger contribution of clay (Handley et al., 2012).Evidence includes an abundance of kaolinite, suggestive of intensified physical erosion (John et al., 2012), and high Li/Al combined with low Na/Al, suggestive of exhumation of older weathered clay.These additional proxies also indicate processes that support an increase in the mobilization and accumulation of OC petro during the PETM.
OC petro MAR is largely driven by a shift in LSR from 0.8 cm Kyr −1 (pre-PETM) to 11.28 cm Kyr −1 (PETM) (Table 1; Stassen et al., 2012), and thus any uncertainty in the LSRs will also be reflected in the MAR estimates.The higher OC petro MAR is consistent with evidence of terrestrial input to the Atlantic Coastal Plain during the PETM, including a higher abundance of kaolinite (Gibson et al., 2000), detrital magnetic minerals (Kopp et al., 2009), charcoal, seed pods, and terrestrial spores (Self-Trail et al., 2017).In addition, there is an increase in the terrestrial aquatic ratio (TAR; Bourbonniere & Meyers, 1996;Lyons et al., 2019).Indirect evidence includes changes in the marine microfossil assemblage toward benthic foraminifera (Self-Trail et al., 2017) and dinoflagellates (Sluijs & Brinkhuis, 2009) that can tolerate brackish water with high sediment input (Self-Trail et al., 2017).However, with the exception of the abrupt peaks of C 31 S/(S + R) at ∼169-171 mcd (Figure 5d) and C 30 βα/(βα + αβ) at ∼171-173 mcd (Figure 5e), the thermal maturity ratios at Ancora are relatively stable compared to SDB and CamDor (Lyons et al., 2019).Unlike Ancora, SDB and CamDor are characterized by a 6‰ increase in δ 13 C org values during the PETM (Lyons et al., 2019), which was argued to represent reworking of older (pre-PETM) material and not an increase in primary production (Lyons et al., 2019).This 13 C enrichment is not observed at Ancora (Figure 5a; Elling et al., 2019) and is consistent with the relatively stable thermal maturity ratios during the PETM.
The average OC petro MARs at Kheu River exhibits an increase (by 3 × 10 −2 gC cm 2 Kyr −1 ) during the PETM (Figure 7), driven by an order-of-magnitude rise in C org values from an average background level of ∼0.1 wt.% (preand post-PETM) to ∼4.4 wt.% (Dickson et al., 2014).However, in contrast to the sites discussed thus far, thermal maturity ratios at Kheu River shift to immature values during the core of the PETM (Figure 4).During the PETM, the n-alkane distribution is dominated by long-chain homologs characteristic of vascular plants (Dickson et al., 2014).It can therefore be argued that the shift observed in the thermal maturity ratios is mostly due to enhanced input of the OC bio (i.e., immature hopanes such as ββ isomers) transported from land, although in situ production cannot be dismissed.An increase in the Chemical Index of Alteration (CIA) and spike in Ti/Al during the PETM not only corroborates evidence for terrestrial input but possibly erosion of older (pre-PETM) material (Dickson et al., 2014).As such, both OC petro and (to a larger extent) OC bio were likely delivered to this site.OC bio burial may negate CO 2 released via enhanced OC petro oxidation (e.g., Bowen & Zachos, 2010;John et al., 2008;Kaya et al., 2022;Papadomanolaki et al., 2022;Sluijs, Röhl, et al., 2008).Therefore, understanding whether the Kheu River region was a net carbon source or sink requires further investigations, and this study highlights the need to quantify both OC bio and OC petro in marine sediments.Regardless, the subtropical and mid-latitude sites all exhibit an increase in OC petro MAR during the PETM, and thus suggest that OC petro oxidation may provide an additional source of CO 2 .

Limited Change in Organic Carbon Sources in the High-Latitudes During the PETM
In the subtropics and mid-latitudes, average OC petro MAR increased between 8 × 10 −3 and 6 × 10 −2 gC cm 2 Kyr −1 during the PETM for a given site (see Section 4.1).In the high-latitudes, OC petro MARs in the Arctic (ACEX) and the southwest Pacific Ocean (ODP Site 1172) either increase (by 7 × 10 −2 gC cm 2 Kyr −1 ) or decrease (by 3 × 10 −4 gC cm 2 Kyr −1 ), respectively (Figure 7).The decline observed at ODP Site 1172 is due to a small drop in C org values and LSRs.The marked rise at ACEX is mostly driven by a peak in TOC values, from a minimum of 1.3% (pre-PETM) to a maximum of 4.9% (core PETM) (Elling et al., 2019).Absolute abundances of palynomorphs from ACEX suggest that TOC is a mixture of marine and terrestrial organic matter (Sluijs, Röhl, et al., 2008).However, both sites, with the exception of the C 31 S/(S + R) ratio at ODP Site 1172 (Figure 3d), have stable thermal maturity ratios throughout the record.This indicates that although the supply of organic carbon increased during the PETM, the organic carbon source did not distinctly change.Intriguingly, there is an antiphase relationship between C 30 αβ/(αβ + ββ) and C 31 αβ/(αβ + ββ) at ACEX (Figure 2c), perhaps suggesting subtle changes in the organic carbon source during the PETM.Decoupling between the C 30 and C 31 indices could be due to a greater input of acidic peats, which are dominated by C 31 αβ hopanes but lack abundant C 30 αβ isomers (Inglis et al., 2018).The contribution of OC bio from acidic peats at ACEX has also been inferred from brGDGTs (Sluijs et al., 2020).

Climate Exerts Primary Control on OC petro Mobilization During the PETM
Various factors may explain why some shallow marine sediments are characterized by enhanced delivery of OC petro during the PETM.Modern observations have identified a strong link between rainfall and efficient erosion/transfer of organic carbon from land-to-sea (e.g., T. I. Eglinton et al., 2021;Hilton, 2017).In the subtropics, evidence for changes in the hydrological cycle during the PETM are scarce.Previous work at TDP Site 14 revealed that the hydrogen isotope of n-alkanes (δ 2 H n-alkanes ) increased during the PETM, which was inferred to represent a shift toward more arid climate conditions (Carmichael et al., 2017;Handley et al., 2008).Enhanced aridity could lead to minimal vegetation cover, hindering soil development, and maximizing the potential for erosion and mobilization of OC petro (e.g., Hilton et al., 2008;Leithold et al., 2006).Furthermore, there are large fluctuations in δ 2 H n-alkanes values, which may indicate oscillations between dry and wet climate states and/or an increase in extreme precipitation events (Carmichael et al., 2017;Handley et al., 2008).Modeling studies over subtropical Africa during the PETM support the latter (Carmichael et al., 2018).Episodic and intense rainfall on a landscape prone to erosion would explain the highly variable delivery of different organic carbon sources, as shown by the hopane-based thermal maturity data (this study), δ 13 C org values, and n-alkane δ 13 C values (Aze et al., 2014;Handley et al., 2008).
Analogous to TDP Site 14, Kheu River also exhibits high variability in the thermal maturity ratios (e.g., CPI, C 29 αβ/C 30 αβ, and C 29 βα/(βα + αβ); Figure 4), chain-length distributions of n-alkanes, BIT index, grain-size, and CIA during the PETM (Dickson et al., 2014).Although two brief intervals of marine transgression have been noted in this region (Shcherbinina et al., 2016), the biomarker records are more variable and thus appear to be more consistent with episodic changes in precipitation.There are multiple lines of evidences associating other mid-latitude sites with increased transient and extreme rainfall events during the PETM.For example, the deposition of conglomerates in the Pyrenees (Chen et al., 2018;Schmitz & Pujalte, 2003, 2007), changes in paleosol weathering indices, and the abundance and composition of nodules in the Bighorn Basin (e.g., Kraus & Riggins, 2007;Kraus et al., 2013).There is also evidence for greater freshwater runoff in the Atlantic Coastal Plain (i.e., Ancora, SDB, and CamDor) during the PETM, with the development of a river-dominated shelf referred to as the "Appalachian Amazon" (Doubrawa et al., 2022;Kopp et al., 2009;Self-Trail et al., 2017).This is consistent with high-resolution climate models that suggest the western Atlantic region was dominated by an increase in extratropical cyclones and more extreme rainfall events (Kiehl et al., 2021;Rush et al., 2021;Shields et al., 2021).Although the hydrological cycle likely exerted a first-order control on the mobilization of terrestrial organic carbon, other ecological and/or geologic controls could have also been important.For example, the dominance of OC bio at Kheu River may reflect abundant vegetation cover (e.g., Goñi et al., 2013).On the other hand, the dominance of OC petro at TDP Site 14 may reflect greater availability of OC petro -rich rock and/or exacerbated erosion of OC petro caused by limited soil and vegetation (e.g., Hilton et al., 2011).
Model simulations indicate an increase in precipitation in the high-latitudes for a PETM-type warming event (e.g., Carmichael et al., 2016;Cramwinckel et al., 2023;Winguth et al., 2010).Proxies also reconstruct northern and southern high-latitudes to be wetter at the onset of the PETM (e.g., evidence from palynomorphs (Korasidis et al., 2022;Sluijs et al., 2006;Willard et al., 2019), fossilized plants (Harding et al., 2011), the hydrogen isotopic composition of n-alkanes (δ 2 H n-alkanes ;Pagani et al., 2006), and clay-mineralogy (Dypvik et al., 2011;Kaiho et al., 1996;Robert & Kennett, 1994)).Yet, biomarker evidence from high-latitude sites (i.e., ACEX and ODP Site 1172) indicates limited changes in the source of organic carbon during the PETM.This suggests that in order to exhume and mobilize OC petro , changes in rainfall seasonality and frequency of extreme precipitation events may be required (see Section 4.1).Alternatively, there may be other feedback mechanisms and/or more regional controls beyond the hydrological cycle.In modern systems, local geomorphic processes play a strong role in regulating OC petro transport from land-to-sea (e.g., Hilton & West, 2020).Variability in OC petro MARs could also be attributed to changes in sea level during the PETM.Indeed, various studies have suggested marine transgression during the PETM, including: ACEX (Sluijs et al., 2006); ODP Site 1172 (Sluijs et al., 2011); Kheu River (Shcherbinina et al., 2016); the Atlantic Coastal Plain (John et al., 2008); and elsewhere (Jiang et al., 2023;Li et al., 2020 and references therein;Sluijs, Brinkhuis, et al., 2008 and references therein).However, although sea level rise is expected to reduce the supply of terrestrial organic carbon into the marine realm, this is rarely observed (e.g., Sluijs et al., 2014) and most PETM sites are characterized by enhanced terrigenous material during the PETM (Carmichael et al., 2017 and references therein).

Timing and Implications for CO 2 Release During the PETM
Enhanced OC petro delivery was suggested to have occurred ∼10-20 Kyrs after the onset of the PETM (i.e., within the body of the CIE) by Lyons et al. (2019).Here, we confirm that elevated OC petro MARs occurred within the core of the PETM at several other sites (i.e., ACEX, Kheu River, Ancora; Figure 8).However, the exact timing within the core (i.e., onset or body) cannot be determined due to the lack of robust age constraints.The sites where the recovery phases were defined (i.e., ACEX, Kheu River, Ancora, and SDB), enables insight into whether enhanced OC petro MARs continued after the core interval or recovered to pre-PETM values.Interestingly, at both Ancora and SDB, median OC petro MARs are higher than the core of the PETM in Phase II and I, respectively (Figure 8).There is a decrease in OC petro MAR during Phase I of the recovery at Ancora, however this interval consists of a single data point.Although an increase in OC petro MAR during the recovery is not observed at ACEX and Kheu River, values do not return to pre-PETM levels.This suggests that at certain localities, terrestrial organic carbon cycle perturbations continued into the recovery phase.If OC petro was oxidized, it may have provided an additional source of CO 2 during the recovery.In this scenario, other negative feedback mechanisms are required to negate the additional carbon released and help assist in the recovery of the PETM.Several processes have been proposed, such as silicate weathering (Penman et al., 2014) and/or enhanced OC bio burial, either on land (Bowen, 2013;Bowen & Zachos, 2010) or within the ocean (John et al., 2008;Ma et al., 2014).For example, exacerbated weathering and erosion during the PETM (Pogge von Strandmann et al., 2021) may increase nutrient delivery from land-to-sea, stimulating primary productivity and therefore OC bio burial (Kaya et al., 2022;Papadomanolaki et al., 2022).However, the source of sequestered organic carbon in ocean sediments (i.e., terrestrial vs. marine) remains a major source of uncertainty.
Overall, Lyons et al. (2019) inferred between 10 2 and 10 4 PgC was released as CO 2 globally due to oxidation of OC petro during the PETM.This assumed that the three sites (i.e., SDB, CamDor, and TDP Site 14) are globally representative.However, this study demonstrates that enhanced OC petro MARs was mostly restricted to the subtropics and mid-latitudes, suggesting that global estimates may be lower than previously inferred.In addition, the maximum value of 10 4 PgC assumed that 85% of OC petro was oxidized.However, increased erosion of clastic sediments can aid the preservation of OC petro (e.g., Bouchez et al., 2014;Burdige, 2007).Furthermore, intense precipitation events (characteristic of the subtropics and mid-latitudes; e.g., Carmichael et al., 2017;  Handley et al., 2008;Kiehl et al., 2021;Kraus & Riggins, 2007;Kraus et al., 2013;Rush et al., 2021;Schmitz & Pujalte, 2003, 2007;Shields et al., 2021) may reduce the transfer time of OC petro from source to sink, thereby reducing the possibility for oxidation (e.g., Hilton et al., 2011).However, it is important to consider that shallow marine sites will likely integrate an expansive catchment area, which incorporate slow meandering rivers as well as steep mountainous rivers.In the former system, the extent of OC petro oxidized could be as high as ∼90% (e.g., Bouchez et al., 2010;Galy et al., 2008).This is especially likely at sites where large freshwater input was evident, such as the Atlantic Coastal Plain (Doubrawa et al., 2022;Kopp et al., 2009;Self-Trail et al., 2017).We also demonstrate that CO 2 release may have continued into the recovery of the PETM, suggesting that other feedback mechanisms (e.g., OC bio burial) were necessary to aid in the recovery of the Earth's climate system.To constrain global estimates of CO 2 emitted from OC petro oxidation, future work is required to elucidate these uncertainties.For example, Raman spectroscopy could help identify the oxidation efficiency based on the degree of highly degradable versus recalcitrant organic carbon (Sparkes et al., 2018), whilst paleo-digital elevation models may provide further insight on sediment routing systems and transit time during the PETM (Lyster et al., 2020).

Conclusion
This study uses a multi-biomarker approach to reconstruct the mobilization of petrogenic organic carbon (OC petro ) during the PETM.We find widespread evidence for enhanced OC petro mass accumulation rates (MARs) in the subtropics and mid-latitudes during the PETM.In this region, we argue that extreme rainfall events exacerbated erosion, mobilization, and burial of OC petro in the marine realm.In addition, we demonstrate that high OC petro MARs persisted into the recovery phase of the PETM.However, the high-latitude sites do not exhibit a distinct change in the source of organic carbon during the PETM.This may be due to a more stable hydrological regime and/or additional controls, such as geomorphic processes or sea level change.Overall, OC petro oxidation likely acted as an additional source of CO 2 during the PETM.However, further work is needed to determine the exact contributions of OC petro as a positive feedback mechanism during the PETM and other transient warming events.

Figure 8 .
Figure 8. Violin plots of OC petro MARs (gC cm 2 Kyr −1 ) for the defined time intervals of sites with the recovery phase (a) ACEX, (b) Kheu River, (c) Ancora, and (d) SDB.The thick dashed line represents the median and the thin dashed line extends from the 25th to 75th percentiles.Note the discontinuous y-axis and two different scales of (b) Kheu River.