Corresponding author: A. Schneider-Mor, Department of Geology and Environmental Sciences, Stanford University, Stanford, CA, USA. (firstname.lastname@example.org)
 The flux of sediment and organic carbon from continents to the coastal ocean is an important factor governing organic burial in coastal sediments, and these systems preserve important records of environmental and biogeochemical conditions during past global change events. Burial of organic materials in coastal systems can be promoted by chemical resilience or through protection by association with mineral surfaces, but the role and influence of these processes on organic records from ancient sediments is poorly known. We studied sediment and organic matter burial as particulate organic matter (POM) and mineral-bound organic matter (MOM) in near-shore marine sediments from the Wilson Lake core (New Jersey, USA) that span the Paleocene-Eocene thermal maximum (PETM), a climatic perturbation 55.9 Myr ago. Our results show that distinct POM and MOM fractions can be isolated from sediments. Both fractions appear to be dominated by terrestrial material, but POM consisted primarily of recently synthesized material whereas MOM included a significant fraction of pre-aged organic matter from soils or ancient sediments. Variation in organic burial through the PETM is associated with changes in inorganic nitrogen burial, clay mineralogy, and clastic grain size that we associate with enhanced continental weathering, erosion and redeposition of ancient kaolinites, and eustatic sea level variation, respectively. These results provide a new perspective on factors governing carbon burial and carbon isotope records in ancient marine margin settings and offer information on rate and phasing of late Paleocene/early Eocene Earth system changes that may constrain interpretations of the cause of the PETM climate change event.
 The Paleocene-Eocene thermal maximum (PETM, 55.9 Ma) involved injection of >3000 gigatons (Gt) of carbon into the ocean/atmosphere system [Panchuk et al., 2008; Zeebe et al., 2009] and sea surface warming of 5–8°C over 10–20 thousand years (kyr) [Cui et al., 2011; Zachos et al., 2001]. Various lines of evidence suggest alteration of the hydrological cycle during the PETM, including intensification of seasonal precipitation rates at some midlatitude sites [John et al., 2008; Schmitz and Pujalte, 2007]. The event is marked by a global negative carbon isotope excursion (CIE) of 3–6‰ that affected both marine and continental carbon pools [Bowen et al., 2004; Bowen et al., 2006; Zachos et al., 2006]. The identity of the source or sources of isotopically light carbon contributing to the event remains a matter of debate, with suggestions including methane hydrates, volcanic gas release, or combination of sources [Bowen and Zachos, 2010; Bralower et al., 1997; Dickens, 1995; Secord et al., 2010].
 Marine margin sediments have been a recent focus of efforts to document and understand climatic and carbon cycle perturbation during the PETM. These sedimentary systems can preserve relatively expanded records of the event [Cui et al., 2011; Zachos et al., 2006] and were less affected by the effects of PETM ocean acidification, which led to widespread carbonate dissolution in pelagic sediments [Zachos et al., 2005]. Moreover, sediments deposited on the continental shelf and slope can preserve important records of both marine and terrestrial paleoenvironments [Handley et al., 2008; Storme et al., 2012], facilitating the investigation of both systems and the coupling between them. Important findings from marine margin records to date include constraints on the rate of carbon release at the onset of the PETM [Cui et al., 2011], recognition of oceanic warming preceding the initiation of the CIE [Sluijs et al., 2007], evidence for reduced oxygenation of some ocean water masses during the early stages of the event [Nicolo et al., 2010], shifts in clay mineral assemblages reflective of enhanced continental weathering and/or erosion during the PETM [Robert and Kennett, 1994], and the suggestion of increased organic carbon burial rates on the continental shelves and slopes during the event [John et al., 2008].
 Despite their many positive qualities, marine margin sites have their own set of challenges with respect to developing robust records of the PETM and other global change events. Potential for local environmental change to influence proxy records is high, especially given that the PETM is bounded by eustatic sea level changes that affected local water depths, sedimentary systems, and salinity in coastal environments [Gibson et al., 2000; Sluijs et al., 2008]. For organic proxies, records from marine margin sediments can be particularly complex because they may integrate organic matter with diverse and varying provenance [Blair et al., 2003; Eglinton et al., 1997; Galy et al., 2007; Goñi et al., 2005]. The provenance and nature of organic matter buried in these systems is controlled by a complex of factors including supply, sedimentation rate, organic matter composition, bottom water oxygenation and sediment texture, and mineralogy [Betts and Holland, 1991; Hedges and Keil, 1995; Keil et al., 1994a]. These complexities can lend uncertainty to interpretations of some records [e.g., Cui et al., 2012; Sluijs et al., 2012], but may also provide opportunities to better understand rates and mechanisms of organic matter burial, relationships between organic and sedimentary system dynamics, and coupling between continental and marine systems.
 Here we report the results of a multiproxy study of marginal marine sediments and organic matter spanning the Paleocene-Eocene boundary in the Wilson Lake core, New Jersey, USA. Previous work on sediments from this and nearby cores has contributed to many of the advances in our understanding of the PETM noted above [Gibson et al., 2000; John et al., 2008; Sluijs et al., 2008] and has provided evidence for regional warming of surface waters [Zachos et al., 2006] and enhanced marine productivity [Gibbs et al., 2006] during the event. We focus on characterizing two distinct organic matter fractions—particulate organic matter (POM, free organic particles, commonly preserved as a result of their resilience to chemical and biological degradation [Hedges et al., 1986]) and mineral-bound organic matter (MOM, organic compounds derived from soils, freshwater, or marine systems and adsorbed to and protected by their association with mineral surfaces, especially those of clays [Keil et al., 1994a, 1994b; Meyers, 1994]). Our results demonstrate for the first time that free particulate and mineral-bound organic fractions with distinct provenance and chemistry can be isolated from organic matter in Paleocene-Eocene-age marine margin sediments. We compare records from these fractions with data on sediment mineralogy, grain size, and inorganic nitrogen content to elucidate changes in the land/ocean organic-sedimentary system through the PETM. These results have implications for the interpretation of isotopic and clay mineral records from marine margin sites and allow us to demonstrate and characterize distinct patterns of change in continental and marine systems at this site through the PETM carbon cycle and climate perturbation.
 The Wilson Lake core (39°39′N, 75°03′W) was recovered in eastern New Jersey by the United States Geological Survey in 1993 [Miller, 1997] (Figure 1). We obtained samples from 28 intervals spanning the PETM and physically separated two organic matter fractions—a free particulate fraction and a fraction strongly bound to mineral grains—that have been shown to have different provenance, composition, and reactivity in modern systems [Sohi et al., 2001]. Samples were disaggregated lightly by hand, and approximately 2–3 g was transferred to a 50 mL centrifuge tube. This material was reacted with 1 N hydrochloric acid to remove calcium carbonate and 0.1 M NaOH to remove highly labile organics and contaminants and assist in the dispersion of clays and organic colloids allowing separation of the different fractions [Oades, 1988]. Bulk carbonate concentrations were determined by weighting the dried samples before and after the acid treatment, with an approximate uncertainty of 1% based on the precision of the analytical balance. The total uncertainty may be larger if some small particles were lost during the acid treatment, but these loses were minimized by using centrifugation and aspiration to remove solutions from the sample tubes, rather than decantation. Although acid-base treatment has been shown to lead to the loss of soluble organic matter in some sediments, this is less of a concern for ancient rock samples such as those studied here, where most preserved organic matter is relatively nonreactive [Brodie et al., 2011].
 The samples were then immersed in NaI solution (density = 1.8 g/cm3) and agitated using a Branson 450 ultrasonic probe (15 min, 30–40 W) [Sohi et al., 2001]. After sonication, samples were centrifuged for 30 min at 5000 rpm to isolate free organic particulates in suspension from mineral material, and the supernate and particles in suspension were aspirated onto a precombusted, 1.5 µm glass fiber filter (GFF). The agitation/centrifugation/aspiration procedure was repeated until no new particulates were collected on the filter. The POM for each sample was defined as the aggregate of the material collected on the filters, and POM concentration was measured as the net dry weight of particles collected during filtration (Figure 2). Visual inspection of filters under a binocular microscope confirmed that little to no contamination by mineral material was present. Any organics remaining in the residual mineral fraction were considered MOM.
 Mineral-bound organic carbon (MOC) concentration, the C/N value of the particulate fraction, and δ13C values for both fractions were determined by using a Carlo-Erba elemental analyzer coupled to a PDZ/Europa 20-20 isotope ratio mass spectrometer in the Purdue Stable Isotope Facility. Carbon isotope ratios and wt% carbon values were measured simultaneously on 25–50 mg aliquots of the mineral fraction with a mean standard deviation, based on replicate analysis of samples, of 0.3‰ and 0.017% for δ13C and wt% C, respectively. POM samples were scraped from the GFF before analysis. During this process, pieces of the filter were inevitably added to the sample, meaning that the wt% C and N could not be measured directly. We therefore estimated POC amounts from the measured POM weights by assuming that the POM was 45% carbon [Craft et al., 1991; Hedges et al., 1986; Yu et al., 2010]. We then estimated particulate organic nitrogen (PON) amounts using the measured sample C/N.
 Clay minerals in marginal marine sediments can carry a significant load of inorganic N (usually as NH4+) adsorbed to their interlayers [Hinman, 1966; Muller, 1977; Raju and Mukhopadhyay, 1976; Stevenson and Dhariwal, 1959]. We isolated and measured the concentration of total N (MTN) and inorganic N (MIN) separately within the mineral fraction. To isolate inorganic N, we removed the organic nitrogen by the method of Silva and Bremner . Briefly, aliquots of the samples were reacted with 20 mL KOBr-KOH for 2 h, then 60 mL of double-distilled water was added and the solution was brought to a boil for minimum for 5 min. The solution was cooled overnight and washed three times with 0.5 M KCl, then rinsed three more times with distilled deionized water. Nitrogen content was measured using a PDZ Europa ANCA-GSL elemental analyzer interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer, with mean standard deviation based on replicate analyses of samples of 0.4‰ and 0.008%.
 The <2 µm silicate fraction of the sediments was isolated by centrifugation [Jackson, 1969]. Semiquantitative estimates of clay mineral abundance in the <2 µm silicate fraction were made based on XRD measurements of glycolated splits run from 2° to 80° 2θ (PanAlytical X'Pert PRO MPD using a cobalt X-ray source). Mineral abundance was calculated based on peak heights and areas [Biscaye, 1965; Johns and Jonas, 1954; Petschick et al., 1996; Soller and Owens, 1991], with an approximate uncertainty of 5–10% based on the peak area calculation. Grain size distributions were measured on the siliciclastic fraction using a Malvern Sediment Analyzer at Indiana University-Purdue University Indianapolis. The samples were digested and heated with 30% hydrogen peroxide to remove organic matter and then wet sieved (500 µm) to remove coarse detritus. Samples were allowed to settle for 24–36 h before being analyzed in triplicate on the Malvern. We present data aggregated into clay, silt, and sand size classes, with cutoffs of 0.39 µm and 63 µm based on Wentworth size classifications [Wentworth, 1922].
 Carbon isotope and C concentration data show distinct values and patterns of change for POC and MOC fractions (Figure 3). Particulate fraction δ13C values (δ13CPOC) closely match the pattern of reference curves from known, autochthonous substrates (dynocysts and bulk carbonate [Sluijs et al., 2007; Zachos et al., 2006], which themselves show similar patterns with subtle differences as discussed in Sluijs et al. ), showing a sharp decrease of 5‰ at the onset of the PETM. The majority of this change in our record is documented by a discrete shift of 3.5‰ between adjacent samples at 110.06 and 109.5 mcd. This shift is followed by an interval of relatively stable values between 109.5 and 99.49 mcd, then a gradual increase toward pre-event values that is broken by a more abrupt increase across a previously documented disconformity at 96 mcd (Figure 3a–c). In contrast, the δ13CMOC record shows a much smaller decline of ~3‰ across the 110 mcd level, followed by a gradual and continuous decrease to a minimum at 97.5 mcd and a relatively rapid recovery across the disconformity (Figure 3d).
 Particulate organic carbon concentrations are stable at ~0.2 wt% from the base of our record to 108.51 mcd, through the onset of the CIE in all of the substrates measured here (Figure 3e). Above this level, POC concentrations exhibit higher variability, with a number of transient increases above the 0.2 wt% background to values as high as 0.9 wt%. Mineral-bound organic carbon concentrations average 0.21 wt% but exhibit substantial variability throughout the study interval (Figure 3f). There is some suggestion that MOC concentrations are lower in the pre-PETM samples; but given the range of values and limited number of samples within the pre-PETM interval, this conclusion is somewhat speculative. With the exception of discrete intervals within the PETM and post-PETM sediments where POC concentrations are high, POC and MOC each comprise approximately half of the total organic C recovered from the samples.
 C/N values for both fractions are similar after correction for the subtraction of inorganic N from the mineral fraction (average for MOM = 29.2, POM = 36.8). Values for both fractions exhibit similar ranges of variation, although variation is higher for MOM near the base of the PETM and for POM within and above the CIE (Figure 3g and h). Spikes in C/N values for MOM at 110.54 and 107.61 mcd are associated with very low values of mineral-bound organic N (≤0.003 wt%) and are likely artifacts. MOM C/N values are relatively high (~40) throughout the first two thirds of the PETM interval, then drop rapidly between 101.65 and 101.5 mcd to values of ~25 (Figure 3h).
 Grain size data show a clear shift from sand-silt-dominated sediments in pre-PETM samples to silt-clay-dominated sediments during the PETM (Figure 4). The shift occurs abruptly between 111.5 and 109.5 mcd, with decreasing sand and increasing clay values slightly preceding an increase in the silt size fraction starting at 110.54 mcd. The siliciclastic fraction in the pre-PETM sediments contains on average 33% sand, 38% silt, and 29% clay, whereas sediments within the body of the CIE (as defined by carbonate and dinocyst reference curves) are on average 0.5% sand, 60.5% silt, and 39% clay (Figure 4b–d). Grain size distributions are more variable in the post-PETM interval, showing transient coarsening associated with previously identified unconformities in the section. Carbonate concentration averages ~10% with no significant difference between pre-PETM and PETM samples (t test, p = 0.124), and increases abruptly to values as high as 24% during the post-PETM period (Figure 4a).
 Mineralogical data from the siliciclastic fine clay (<2 µm) fraction show substantial changes (Figure 5) consistent with previous work on nearby cores [Gibson et al., 2000]. Pre-PETM sediments contain 13% kaolinite, 50% illite, and 37% smectite on average. Kaolinite abundance increases in a stepwise fashion from 13% to 38% of the clay fraction between the adjacent samples at 110.06 and 109.5 mcd, and continues to increase gradually until it reaches relatively constant values of 45% between 103.51 and 98.51 mcd (Figure 5). The kaolinite increase within this fraction is largely at the expense of smectite, which averages about 15% of the clay fraction during the PETM. When taking into account the changes in grain size distributions during the PETM, the abundance of each clay mineral group as a fraction of the total sediment increases during this interval. Clay mineral distributions begin to recover above 98.51 mcd and reach values that are similar to those of the pre-PETM distribution, although more variable and characterized by slightly lower kaolinite abundance, above 94.49 mcd (Figure 5).
 Inorganic nitrogen comprises about 70% of the total nitrogen in the mineral fraction of pre-PETM sediments and ~70–94% of mineral fraction nitrogen during the PETM. Concentrations of mineral-fraction organic N (MON, expressed relative to total sample weight) are very low (<0.01%) throughout most of the PETM and pre-PETM intervals, with a transient increase to values just above 0.01 wt% between 110.06 and 108.51 mcd and somewhat higher values above 101.5 mcd. MON concentrations in post-PETM samples from above the disconformity are between 0.01 and 0.02 wt%. Inorganic N concentrations (MIN) show a strong increase beginning at 101.6 mcd, from values of ~0.01 to ~0.03 wt%. The MIN values remain high throughout the PETM, dropping abruptly across the 96 mcd disconformity.
Preservation of Discrete Organic Fractions through the PETM
 Multisubstrate data from the Wilson Lake core clearly show distinct stratigraphic patterns of δ13C change in the POC and MOC fractions through the PETM, with changes in the MOC fraction being generally less abrupt than and lagging those in POC (Figure 3). The global shifts in the δ13C value of atmospheric and oceanic carbon pools at the onset and termination of the PETM were abrupt and essentially synchronous [Koch et al., 1992; McInerney and Wing, 2011]. This suggests that the two organic matter fractions we measured are either differently modulated by changes in the mixing ratios of C from isotopically distinct sources or they time-average the global CIE signal in different ways.
 POC δ13C values most closely track the carbonate δ13C curve for the Wilson Lake core both in amplitude and pattern. The two curves exhibit subtle differences—for example, the excursion onset is somewhat more prolonged in the carbonate record, which may reflect the lower sampling resolution of the POC record but would also be consistent with atmosphere-down propagation of light carbon during the event's onset [Thomas et al., 2002]. Overall, these records exhibit a pattern similar to that seen at many other continental and marine sites worldwide where highly resolved records are available from well-constrained substrates [e.g., Bowen and Zachos, 2010; Diefendorf et al., 2010; McInerney and Wing, 2011]. For instance, δ13C values of n-C31 alkanes and pedogenic carbonates from a thick section of paleosols in the Bighorn Basin (WY) show a very similarly shaped excursion with abrupt onset and recovery [McInerney and Wing, 2011]. Bulk carbonates and taxonomically specific foramineferal records from multiple pelagic ocean sites also record a similarly shaped excursion once corrected for changes in sediment accumulation rates [Murphy et al., 2010]. Although we cannot definitively rule out other possibilities, the most parsimonious explanation for the strong similarities seen among these disparate records is that they each sample a common, well-mixed ocean/atmosphere C source without being modulated by substantial shifts in substrate composition, C source, or time averaging. If this is the case, the abrupt 3.5‰ drop in δ13CPOC values we observe at the onset of the CIE would have occurred over no more than 6.7 to 8.2 kyr (using sediment accumulation rates of 6.8 to 8.4 cm/kyr [John et al., 2008]), an estimate consistent with the shorter end of the range of existing estimates for the duration of the CIE onset [Cui et al., 2011; Zachos et al., 2001] and favoring a relatively rapid release of C from highly volatile reduced C reservoirs at the beginning of the PETM.
 In contrast, the pattern of change in MOC δ13C values differs from that of other high-fidelity PETM records, particularly in its gradual and continual δ13C decrease throughout the CIE and the position of the δ13C minimum. These differences could reflect systematic changes in MOC sources throughout the PETM, with a transient increase in carbon derived from relatively 13C-enriched (e.g., marine) sources early in the CIE. Alternatively, the pattern could result from relatively long residence times for MOC (e.g., in soils and sedimentary systems) before burial, which could cause the delivery of light, PETM-derived MOC to the coastal margin to lag that of other organic and inorganic fractions. In either case, our results demonstrate that these physically separable organic matter pools are characterized by distinct dynamics and reflect two different pathways of organic C burial in the marginal marine sediments at Wilson Lake.
 Carbon/nitrogen values provide one metric of organic matter source. Ratios for terrestrial plant biomass range widely, from values of ~20 (grasses) to ~200 (trees) [Hedges et al., 1986; Hedges et al., 1997]. Within soils, C/N values vary among pools dominated by different compound classes, and relatively high values (e.g., >100) tend to be preserved in resilient particulates (e.g., black carbon; [Schmidt and Noack, 2000]), whereas mineral-bound compounds tend to exhibit lower values (~5–20) [Sollins et al., 2006]. Marine organic matter commonly has C/N values lower than 10 [Hedges et al., 1986]. C/N values for both the particulate and mineral-bound fractions are consistently >20 throughout the study interval (Figure 3g–h). Due to the high abundance of inorganic N, mineral-bound organic C/N values are much higher than bulk C/N measured on non-KOBr-treated samples; but with the exception of two discrete intervals close to the onset of the CIE, C/NMOM values are lower than C/NPOM. Together, these observations are consistent with a terrestrially dominated source for the organic matter in both fractions, with some variation in sources throughout the event.
 Previous work has demonstrated that palynomorphs in the Wilson Lake Paleocene-Eocene boundary sediments are dominated by marine dinoflagellate cysts, with terrestrial plant pollen being essentially absent [Sluijs and Brinkhuis, 2009]. Similarly, measurements of the Branching Index of Tetraether lipids (BIT; [Hopmans et al., 2004]) suggest that the distribution of these lipids is dominated by marine sources throughout the Paleocene-Eocene interval, despite the high clastic content of the sediments and inferred increases in terrigenous sediment influx to the site during the PETM [Zachos et al., 2006]. Together with our results, these observations imply that organic preservation at Wilson Lake must have been subject to relatively strong depositional and taphonomic filtering. Visual inspection indicates that the POM recovered in our study is dominated by abundant sub-millimeter-sized, black, angular, and subangular fragments (Figure A1). We interpret these particles to be burned biomass, and abundant graphitic black carbon has been previously documented in contemporaneous sediments from the nearby Bass Lake core [Moore and Kurtz, 2008]. We suggest that degradation of less chemically resilient terrestrial POM during transport to and across the marine margin must have contributed to a terrestrial POM assemblage dominated by nonreactive black carbon at this site. Neither the black carbon nor the mineral-bound organic matter [Poirier et al., 2005] would be expected to preserve an abundance of the a polar lipids comprising the BIT index, perhaps explaining why this index might not be reflective of the sources of the organic matter isolated here.
 Although the C/N data for MOM suggest the potential for some changes in source through the PETM, the pattern and nature of the changes is not coherent with the pattern observed for δ13CMOC and is not consistent with variation in marine versus terrestrial organic matter source as the dominant control on δ13C change in this organic fraction. Discrete intervals of high C/N values near the onset of the event (Figure 3) may suggest short-lived fluctuations in the transport of mineral-bound organic material across the continental margin, and the abrupt drop in C/N values at 102 mcd may represent a modest shift to a more marine-dominated MOM assemblage, but none of these features are accompanied by substantial changes in δ13CMOC. Minimum δ13CMOC values between 97 and 96 m depth coincide with some of the lowest C/NMOM values in the study interval, in contrast to expectations if a shift from relatively δ13C-enriched marine organic carbon to δ13C-depeleted terrestrial carbon were driving change in the δ13CMOC record. Instead, we suggest that the distinct pattern of δ13C change in this fraction, as compared with the POM and reference records, could reflect time averaging of the MOM record due to a prolonged residence and transit time within soils, fluvial, and shelf sedimentary systems before deposition and burial at Wilson Lake. Although it is likely that MOM burial varied in response to transient fluctuations in sedimentary system dynamics throughout the event, the overall pattern of δ13C variation implies that a pool of material with relatively long (104 year) residence time continued to contribute pre-PETM photosynthate to the MOM fraction throughout most of the CIE interval. The residence time for this faction is consistent with values estimated for soil-derived MOM deposited in marginal marine systems in the modern ocean [Goñi et al., 2005; Mollenhauer et al., 2007].
Organic and Sedimentary Provenance
 Another potential control on MOM provenance and burial through the PETM is variation in the provenance of the mineral material with which this organic matter is associated. The source of the abundant kaolinite buried along the New Jersey Margin during the PETM remains uncertain, and some authors have suggested that the dramatic change in clay mineral assemblages reflects enhanced erosion of preexisting continental kaolinite deposits rather than a change in weathering regimes within earliest Eocene soils [John et al., 2012; John et al., 2008; Thiry, 2000]. In this case, concentrations and δ13C values of the MOM fraction during the PETM could reflect an increase in the exhumation and reburial of ancient (pre-Paleogene) kerogen, a process common today particularly along continental margins characterized by steep topography [Blair et al., 2004; Blair et al., 2003]. One implication, if correct, would be that a fraction of the excess organic carbon burial previously inferred during the PETM at sites like Wilson Lake [John et al., 2012; John et al., 2008] would actually have been derived from the lithosphere as opposed to the active ocean/atmosphere/biosphere (“exogenic”) carbon pools and would have had little effect on the exogenic system.
 Terrigenous clays exported across the continental margin also carry and sequester inorganic nitrogen, primarily in form of ammonium formed in soils during organic matter decomposition and fixed to interlayer clay minerals where it replaces potassium in the lattice [Muller, 1977; Schubert and Calvert, 2001; Stevenson and Dhariwal, 1959]. The ammonium load on soil-derived clays depends on a variety of factors, including clay mineralogy (e.g., only 2:1 clays such as smectite and illite fix appreciable amounts of NH4+) and soil chemistry (pH, concentrations of other solutes such as Na+ and Ca2+ that promote interlayer cation exchange) [Raju and Mukhopadhyay, 1976; Vandermarel, 1954]. As a result, patterns of MIN variation in the Wilson Lake sediments may reflect variation in clay mineral source conditions and/or provenance.
 Our data show that concentrations of mineral-bound inorganic nitrogen vary by factor of 1.5 through the PETM (Figure 6). We compare measured MIN values with the abundance of illite and smectite, which we calculate here as the percentage of each clay mineral group measured in the <2 µm fraction times the percent abundance of the <4 µm grain size fraction to approximate changes in the total population of clays (i.e., we assume that the mineralogy of the <2 µm and 2–4 µm fractions are the same). The data segregate into two regimes (Figure 6a). Values from the intervals below and above those characterized by CIE δ13CMOC values (samples below 110 mcd and above 95.5 mcd, respectively) have lower concentrations of MIN, and MIN values are tightly correlated with illite + smectite abundance (R2 = 0.73, p < 0.001; Figure 6a). Samples from between 110 and 95.5 mcd have higher MIN values relative to their illite + smectite concentrations, and MIN concentrations in these samples are not significantly correlated with illite + smectite. Identical patterns exist if the clay mineral abundances are calculated for the <2 µm only.
 These results suggest that clay ammonium loads on pre-PETM and post-PETM 2:1 clays are relatively constant, as might be expected if the clays were derived from a uniform, relatively invariant system. In contrast, the PETM-influenced samples have a separate, distinct pattern of MIN loading consistent with a change in clay provenance and/or chemical environment of ammonium fixation for these samples. The shift in MIN regimes during the PETM could reflect a shift in clay source associated with increased erosion and transport of ancient clay-bearing deposits [John et al., 2008; Thiry, 2000]. In this scenario, increased contributions of 2:1 clays from Cretaceous and older deposits that formed under different chemical conditions and/or were affected by subsurface processes promoting NH4+ substitution [Sterne et al., 1982] could have increased the MIN concentration in the Wilson Lake PETM sediments. However, it is also possible that the increase in MIN values was associated with a change in soil chemistry during the PETM (e.g., higher concentrations of Na+ and Ca2+ in solution within more rapidly weathering PETM soils) or increases in ammonium fixation within secondary environments (e.g., during clay mineral residence within anoxic estuarine sediments) without substantial changes in clay provenance. Although we cannot conclusively distinguish between these scenarios, we note that the observed lack of correlation between illite + smectite abundance and MIN concentrations for the PETM samples suggests that within this interval the 2:1 clay assemblage likely contains a variable mixture clays from the two different MIN regimes (i.e., pre-PETM soil-derived clays and either eroded ‘ancient’ clays or PETM soil-derived clays).
 In contrast to the observations for MIN, concentrations of mineral-bound organic carbon are not significantly correlated with illite + smectite abundance for any interval or combination of intervals of the record (Figure 6b). MOC concentrations are weakly but significantly correlated with MIN concentrations in the aggregated pre-PETM and post-PETM intervals (R2 = 0.40, p = 0.03), but not within the PETM interval, where ratios of MIN to MOC are approximately two times higher than pre-event and post-event. As described above, the MOM fraction preserves a carbon isotope excursion that is damped early in the event but eventually achieves an amplitude equal to that recorded by particulate organic carbon. This implies that the MOM fraction may be diluted by ancient organic matter throughout part of the PETM, but is dominated by PETM-derived material late in the body of the PETM and during the early recovery phase, near the δ13CMOC minimum. The samples at 97.51 and 96.41 mcd, comprising the δ13CMOC minimum, correspond closely with a rapid decline in kaolinite content but have illite + smectite and MIN contents indistinguishable from other samples earlier in the body of the PETM (Figure 6). This pattern suggests that the damped expression of the PETM CIE in the early part of the δ13CMOC record may in part reflect dilution of PETM organic matter in the mineral fraction by the addition of ancient organic matter associated with eroded kaolinites. We suggest that as the flux of kaolinite to the marine margin abated in the initial stages of the PETM recovery, authigenic, PETM-formed 2:1 clays carrying high MIN loads and low-δ13C organic matter continued to be eroded from soils and ponded sediments and deposited on the shelf, leading to the full expression of CIE minimum values in δ13CMOC values within this core interval.
Relationship to Sedimentology and Sea Level
 Previous work has documented the regional impact of sea level change on sedimentary systems along the New Jersey margin, including transient fining of sediments associated with a minor transgression at the PETM onset, shifts in clay mineral assemblages during the PETM body, and coarsening and disconformities associated with end-PETM regression [Gibson et al., 2000; Gibson et al., 1993; Sluijs et al., 2007]. Our new data indicate that initial changes in grain size in the Wilson Lake core are coincidental with “precursor” shifts in dinoflagellate cyst assemblages and sea surface temperatures previously reported at Wilson Lake [Sluijs et al., 2007] and slightly precede shifts in clay mineral abundance and δ13C of both the POM and MOM fractions. At the end of the event, clay mineral assemblages show substantial recovery toward the “normal” non-PETM composition almost coincidently with initiation of the δ13CPOC recovery, whereas grain size distributions remain relatively stable until the disconformity marking the end-PETM regression, some 2.51 m above (~12.9 kyr later assuming a PETM sediment accumulation rate of 19.5 cm/kyr [John et al., 2008]). MOM carbon isotope ratios shift discretely to their minimum values at the same time that clay mineral assemblages are transitioning away from PETM conditions.
 These results suggest temporary decoupling of changes in the marine and continental environments during the intervals of rapid environmental change marking the beginning and end of the PETM. Grain size appears to be closely linked to eustatic sea level change, and with it perturbations of marine fauna and temperature. Coupled changes in these systems appear not to have been entirely coincidental with changes in the carbon cycle recorded by the δ13C records. Taking the shifts in clay mineral assemblage as indicative of changes in weathering and/or erosion regimes on the continents, these changes seem to closely match inflections in our high-fidelity record of carbon cycle change (δ13CPOC). Together, the results may suggest that carbon cycle perturbation and CO2 greenhouse forcing exerted particularly strong and direct forcing on continental systems during the PETM, whereas changes in the oceans were modulated by other environmental and/or oceanographic shifts that occurred independently of, and perhaps contributed to triggering, the carbon cycle changes.
Summary and Conclusion
 Our work offers a new perspective on the evolution of a continental margin system through a major geological climate change event, and suggests several implications for the study of such systems in general and for understanding the PETM specifically. First, we clearly demonstrate that Paleogene marine-margin sediments contain physically separable organic fractions of contrasting provenance that can be associated with different mechanisms of organic matter stabilization and burial. The recognition of discrete POM and MOM fractions in these sediments creates opportunities for more widespread analysis of these physical fractions and development of models for their relative roles in carbon cycle response during ancient climate change events. It also suggests that variation in mixing ratios or provenance of these fractions could be a strong control on bulk organic carbon isotope records from marginal marine settings (and perhaps other depositional systems), and should be considered in the interpretation of such records [Cui et al., 2011; Dupuis et al., 2003; Steurbaut et al., 2003; Storme et al., 2012].
 Our results also show that the coupling of sedimentary and organic systems across the land/ocean interface was complex and dynamic through the PETM. These systems responded to both climatic and eustatic forcing, with different timing and responses to these forcings. Decoupling between sea-level driven changes in grain size and climate-related changes in continental erosion and weathering regimes suggests that the sequence of events at the PETM onset may have initiated with changes in the oceanic system, with the propagation of extreme environmental changes to land occurring through perturbation of the carbon cycle as recorded in carbon isotope records. During the body of the PETM, both the composition and provenance of detrital clays shifted in response to changing continental weathering and erosional regimes, and discrete intervals of higher POM accumulation may have been associated with secular fluctuations in continental climate conditions [Aziz et al., 2008; Kraus and Riggins, 2007]. During the recovery phase of the event, the global recovery of ocean/atmosphere δ13C values coincided with reduced erosion and deposition of kaolinite, but clay-rich PETM soils and sediments continued to be eroded and deposited at the Wilson Lake site until the deposystem was perturbed by sea-level regression and the end-PETM disconformity.
 We thank D. Bessler and J. Greene for all their devoted help at the lab, S. Oleynik for his support with isotope measurement, and S.L. Story for her help with the XRD analysis. In addition, we thank K.J. Licht for the grain size analysis and the UC Davis Stable Isotope Facility for nitrogen isotope analysis. This work was supported by the Donors of the American Chemical Society Petroleum Research Fund and US National Science Foundation grant EAR-0628302 to GJB.
Appendix Table 1
Table A1. Isotopic and concentration data for mineral-associated and particulate organic fractions