Extracting a Detailed Magnetostratigraphy From Weakly Magnetized, Oligocene to Early Miocene Sediment Drifts Recovered at IODP Site U1406 (Newfoundland Margin, Northwest Atlantic Ocean)

Fine‐grained magnetic particles in deep‐sea sediments often statistically align with the ambient magnetic field during (and shortly after) deposition and can therefore record geomagnetic reversals. Correlation of these reversals to a geomagnetic polarity time scale is an important geochronological tool that facilitates precise stratigraphic correlation and dating of geological records globally. Sediments often carry a remanence strong enough for confident identification of polarity reversals, but in some cases a low signal‐to‐noise ratio prevents the construction of a reliable and robust magnetostratigraphy. Here we implement a data‐filtering protocol, which can be integrated with the UPmag software package, to automatically reduce the maximum angular deviation and statistically mask noisy data and outliers deemed unsuitable for magnetostratigraphic interpretation. This protocol thus extracts a clearer signal from weakly magnetized sediments recovered at Integrated Ocean Drilling Program (IODP) Expedition 342 Site U1406 (Newfoundland margin, northwest Atlantic Ocean). The resulting magnetostratigraphy, in combination with shipboard and shore‐based biostratigraphy, provides an age model for the study interval from IODP Site U1406 between Chrons C6Ar and C9n (∼21–27 Ma). We identify rarely observed geomagnetic directional changes within Chrons C6Br, C7r, and C7Ar, and perhaps within Subchron C8n.1n. Our magnetostratigraphy dates three intervals of unusual stratigraphic behavior within the sediment drifts at IODP Site U1406 on the Newfoundland margin. These lithostratigraphic changes are broadly concurrent with the coldest climatic phases of the middle Oligocene to early Miocene and we hypothesize that they reflect changes in bottom water circulation.


Introduction
Since the 1960s, the geomagnetic polarity time scale (GPTS) has become established as a widely applied tool for the development of accurate age models and global stratigraphic correlations. Magnetic reversals observed in marine magnetic anomaly profiles were initially supported and subsequently improved by the construction of magnetostratigraphies from continuous sedimentary sequences recovered during deep-sea drilling. Today most of the geomagnetic field reversals of the Cenozoic era are well dated using astronomical tuning and radiometric dating methods (e.g., Hilgen et al., 2012;Vandenberghe et al., 2012) and therefore provide high-fidelity age control for important geological events. For example, the six reversals within Chron C6Cn (nomenclature cf., Cande & Kent, 1992) occur contemporaneously with the Oligocene-Miocene Transition (OMT) climate event. The precision of the magnetostratigraphic age control is independent from climate proxy records and therefore key to global correlations required for a more complete understanding of the causes and consequences of this climate event.
One major obstacle to develop reliable magnetostratigraphic chronologies of these important geological events is the occurrence of very low natural remanent magnetization (NRM) intensities that is sometimes present in deep-sea sequences. This low intensity can be a primary effect caused by low ferromagnetic mineral supply, result from dilution by nonremanence phases (e.g., Roberts et al., 2013), be authigenic or diagenetic (e.g., Roberts, 2015), or a combination of these. A good example of this problem arises in the 2. Samples and Methods 2.1. Site Description IODP Site U1406 is located near the crest of the J-Anomaly Ridge, on the Newfoundland margin in the northwest Atlantic Ocean (Figure 1). The site is at intermediate paleodepths within the Expedition 342-wide depth transect that aims to document changes in the North Atlantic carbonate compensation depth. The Eocene-Miocene interval at IODP Site U1406 consists of clay-rich nannofossil oozes and chalks (Norris et al., 2014). The sediments are light greenish to grey in color and are well bioturbated with occasional glauconitic horizons (Norris et al., 2014) that suggest prolonged exposure on the sea floor. For example, an exceptionally well-developed, 10 cm-thick glauconite horizon ( Figure 2a) (Billups et al., 2004;Channell et al., 2003), ODP Site 1218 (Lanci et al., 2005;P€ alike et al., 2006b), ODP Sites 1264 through 1266 (Liebrand et al., 2016(Liebrand et al., , 2017, and IODP Site U1334 (Channell et al., 2013).
Geochemistry, Geophysics, Geosystems 10.1002/2017GC007185 van Peer et al., 2017). Additionally, interhole variability is present at 94 m CCSF-M, which consists of a condensed interval in Hole U1406A, a 2 m hiatus in Hole U1406B between 93 and 95 m CCSF-M, and an apparent stratigraphically complete record in Hole U1406C (van Peer et al., 2017). Furthermore, contorted bedding and microfaults ( Figure 2b) appear at 175 m CCSF-M, which has been interpreted as slumping (Norris et al., 2014).

Samples and Magnetic Measurements
A total of 159 u-channel samples were taken from the center of archive-half core sections at IODP Site U1406 to avoid sampling of the sheared core-margins that could degrade paleomagnetic measurements (e.g., Acton et al., 2002). Sampling followed the shipboard splice (Norris et al., 2014) and we accounted for stratigraphic complexities and splice revisions (van Peer et al., 2017) by collecting additional off-splice u-channel samples. Working-half core sections were sampled where archive-half core sections were designated as a permanent archive. The NRM of all u-channel samples was measured on a 2G Enterprises Model-760R superconducting rock magnetometer housed in a magnetically shielded room at the National Oceanography Centre Southampton, University of Southampton (UoS), UK. This system was designed for measurement of u-channel samples and has a narrow-bore sample access hole (Weeks et al., 1993), which introduces less convolution (smoothing) during pass-through measurements. We measured the NRM of all u-channel samples repeatedly at 1 cm intervals before and after stepwise alternating field (AF) demagnetization at peak fields of 20-60 mT in 5 mT increments, and at 80 and 100 mT (12 steps in total).
We also collected 21 bulk sediment samples from selected depth levels for room temperature rock magnetic experiments. Each sample weighed 1 g and was encapsulated and immobilized with sodium silicate. After consolidation, we measured hysteresis loops and backfield curves of the samples to estimate hysteresis parameters (i.e., M s saturation magnetization, M r saturation remanence, H cr coercivity of remanence, and H c coercivity). Measurements were carried out on a Princeton Measurements Corporation (now Lake Shore Cryotronics) Model 3900-02 vibrating sample magnetometer (VSM) at UoS, using a 10 s averaging time, 5 mT field-step increments, and a saturating field of 500 mT for both hysteresis loops and backfield curves. We used the Institute for Rock Magnetism Database software (Jackson & Solheid, 2010) to process the measured hysteresis loops.
Fifteen bulk samples were selected for First-Order Reversal Curve (FORC) measurements (Pike et al., 1999;Roberts et al., 2000Roberts et al., , 2014 on a VSM at UoS. FORC diagrams consisted of 80 curves at 3.5 mT field increments and were analyzed using the FORCinel v. 2.03 software (Harrison & Feinberg, 2008), which incorporates VARIFORC smoothing (Egli, 2013) and statistical confidence intervals . All FORC diagrams were mass normalized and drift corrected, with the first point artifact removed and five lower branches subtracted to enhance the FORC signal (Egli, 2013). The generally weak magnetization of the samples required high VARIFORC smoothing factors (i.e., S c,0 5 5-7; S c,1 5 5-8), in addition to a 14 h measurement protocol with a 10 s averaging time.
Magnetic susceptibility of eleven freeze-dried bulk sediment samples was measured either during warming from room temperature (RT) to 6008C followed by cooling to RT in a 200 A/m AC field, on an AGICO multifunction Kappabridge with a CS4 furnace at the Institute for Rock Magnetism (University of Minnesota), or from RT to 7008C followed by cooling to RT in a 300 A/m AC field on an AGICO KLY-4S with a CS3 furnace at UoS. The latter experiments were carried out to test for possible alteration at temperatures >6008C. All samples were heated at 118C/min in an argon atmosphere to minimize potential alteration during heating.

NRM Data Processing and Noise-Masking Protocol
Measurements of the relatively weak NRM (intensity 10 25 to 10 24 A/m after 20 mT peak AF demagnetization) of IODP Site U1406 sediments were prone to noise. Principal component analyses (PCA) (Kirschvink, a) Image Glauconite b) Image Slump Figure 2. Core section photos (Norris et al., 2014) in (a and b) are highlighting the strongly developed glauconitic horizon and slumped interval, respectively Geochemistry, Geophysics, Geosystems 10.1002/2017GC007185 1980) of the u-channel NRM data, using a commonly applied, uniform demagnetization range (i.e., 20-60 mT; e.g., Channell et al., 2013), frequently resulted in maximum angular deviation (MAD) values >158, indicating poorly defined component directions. Recognizing the limitations of applying a uniform demagnetization range for direction assignments, we investigated the application of a set of data selection criteria that automatically masks statistical outliers or measurements significantly affected by environmental or instrument noise, leaving better-resolved data for magnetostratigraphic interpretations. Despite several similarities, our analytical procedure was developed independently from the Zplotit software (Acton, 2011) and was designed specifically for integration with the UPmag software suite (Xuan & Channell, 2009).
For NRM data of each 1 cm interval, we first calculated PCA directions for all possible six-step combinations from 10 steps in the 20-80 mT range. We did not anchor the origin to avoid introducing artificially low uncertainty estimates (Heslop & Roberts, 2016). Only PCA directions associated with the six-step combination yielding the minimum MAD value were used for further interpretation. This procedure automatically discards data from noisy measurement steps, which improved the rigor of the resulting PCA directions. In addition, we calculated Fisher's statistics using the data from the six demagnetization steps that produce the optimal PCA directions. The associated a 95 values (error estimate of Fisher's mean) provided an additional check on the measurement data quality. For example, demagnetization data that lied closely on a line in orthogonal vector space would have yielded a small MAD value, but the defined line may not cross near the origin. Such data would be associated with large a 95 values.
To construct the magnetostratigraphy at IODP Site U1406 using only the highest quality NRM data, we removed data from (1) disturbed intervals (e.g., coring-induced or sampling-induced) described during shipboard operations (Norris et al., 2014) and after visual inspection of core-images and u-channel samples; (2) measurements from the top and bottom 5 cm intervals of every u-channel that are prone to convolution effects (e.g., Oda & Xuan, 2014); (3) measurements with spurious ARM acquisition during demagnetization (recognized by substantial increase in NRM intensity during AF demagnetization); and (4) intervals with MAD and/or a 95 values >158 associated with the six-step optimal PCA calculations. The remaining data were then subjected to (5) the Vandamme cutoff procedure (Vandamme, 1994), which iteratively determines the optimal cutoff angle and characteristic angular standard deviation of the directional distribution. This approach removes one outlier (a direction farther from the mean than the calculated optimal cutoff angle) each time until the furthest outlier was within optimal cutoff angle. We note that this procedure can mask transitional directions proximal to reversals of the geomagnetic field. Therefore, we manually examined all Zijderveld diagrams around each reversal and defined the uncertainty (i.e., transitional) interval between the first and last poorly resolved characteristic remanent magnetization direction.
Declination data from every studied core from Holes U1406A and U1406B were rotated using the FlexIT tool data (Norris et al., 2014). We also corrected the declinations of each core by rotating the core so that the circular mean of the declinations, calculated using CircStat (Berens, 2009), was oriented north or south for normal and reversed polarity zones, respectively. FlexIT tool-corrected declinations were generally very consistent with the circular-mean-oriented declinations for Holes U1406A and U1406B. Rotated declinations for Hole U1406C were only calculated using the circular-mean method, because the FlexIT tool was only deployed during the coring of Holes U1406A and U1406B (Norris et al., 2014).

Rock Magnetism
The selected hysteresis parameters yield good quality results (Figure 3a), despite the relatively weak remanence and strong paramagnetic component. Backfield remanence measurements suggest that a lowcoercivity component with H cr of 42-53 mT dominates the samples. Slope-corrected hysteresis loops (calculated using the methods of Jackson & Solheid, 2010), however, generally do not saturate by 500 mT, indicating a potential artifact of slope correction caused by the strong paramagnetic component or at least one additional component with higher coercivity. The FORC diagrams show patterns consistent with a central ridge component, with either small interactions or none, but the ridge is very smoothed due to the 3.5 mT field increment. No interactions may indicate highly anisotropic noninteracting single-domain magnetite, such as intact magnetosome chains (e.g., Egli et al., 2010;Roberts et al., 2013Roberts et al., , 2014. The low-coercivity and large interaction component is consistent with the presence of a coarser-grained (i.e., a multidomain detrital) magnetic component (e.g., Roberts et al., 2012), partially collapsed magnetosome chains Geochemistry, Geophysics, Geosystems 10.1002/2017GC007185 (e.g., Li et al., 2012), or both. At coercivities >60 mT, all FORCs show significant positive values at negative interactions (H u axis), but not at positive interactions (H u axis). This is not consistent with the symmetrical signature of a pseudo-single-domain or multidomain magnetic component (e.g., Roberts et al., 2012), and it may be part of a kidney-shaped imprint that is representative of hematite or monoclinic pyrrhotite (Roberts et al., 2014).
The hyperbolic decrease in magnetic susceptibility during heating (Figure 3b) suggests the presence of a strong paramagnetic component. This observation is consistent with those from the hysteresis measurements. The hyperbolic decrease cannot be extrapolated beyond 5908C, suggesting the presence of magnetite, which is commonly considered a stable sedimentary remanence carrier. The increase in magnetic susceptibility at 500-5508C in several samples implies the formation of a new strongly magnetic component (i.e., magnetite), the presence of a Hopkinson peak (e.g., Dunlop & € Ozdemir, 2007), or both. Stepwise heating-cooling cycles, however, are nonreversible (Figure 3b3), suggesting that the peak in magnetic susceptibility at 500-5508C is caused by the formation of magnetite during heating, and hence, is not a Hopkinson peak. Destabilization of iron-bearing clay minerals may supply the iron for this magnetite that forms during the measurement process.  Figure 3. Examples of rock magnetic analysis results. (a1-a6) Slope-corrected hysteresis loops (blue), remanent hysteretic magnetization (green), error curve (brown), and backfield curves (thick red line) and First-Order Reversal Curve (FORC) diagrams for six representative samples. Smaller insets highlight the difference between raw (red) and drift and slope-corrected (blue) hysteresis loops using the methods of Jackson and Solheid (2010). FORC diagrams are processed using FORCinel 2.03 (Harrison & Feinberg, 2008) with VARIFORC smoothing (Egli, 2013); 95% confidence intervals   Geochemistry, Geophysics, Geosystems  Table S2) after removing data from disturbed intervals, i.e., steps 1-3 from the aforementioned protocol. The optimized PCA directions calculated on the same measurements using the six-step combination are fairly uniformly distributed between all steps ( Figure 4h, note that 65, 70, and 75 mT demagnetization steps were not used) and reduces the mean MAD to 118 (Figure 4a; Supporting information Table S3). Despite these optimization steps, the PCA directions still show relatively large scatter when plotted on an equal area projection ( Figure 4c). This may be due to the large number of optimized PCA directions that are still associated with MAD values >158 ( Figure 4a). Occasionally, optimized PCA directions with MAD values <158 can be spurious if the PCA lines do not cross near the origin (note that the origin was not anchored for PCA analyses to prevent introducing an artificially low uncertainty estimation, cf., Heslop & Roberts, 2016). In these instances, a 95 calculated using the optimized six-step demagnetization data would be large. Thus, small values for both MAD and a 95 are associated with the most reliable directions. The masking of all characteristic directions with MAD values >158 and/or a 95 > 158 (Figure 4d) reduces the scatter of the remaining data and excludes most of the very low intensity data ( Figure  4i) from which reconstructions of accurate directions is unlikely. This masking results in the retention of 62% of the original data set after removal of stratigraphically disturbed intervals, but a large dispersion persists (Fisher's precision parameter of the mean direction, j 5 6, Figure 4e). An equal area plot of directions ( Figure 4f) and a histogram of the corresponding virtual geomagnetic poles (VGPs) (Figure 4g) show that the Vandamme cutoff procedure (Vandamme, 1994) within our protocol successfully removes the remaining statistical outliers. This procedure retains the main fraction of the data (56% of the original data set with disturbances removed, Supporting information Table S4) close to expected values: IODP Site U1406 was at a paleolatitude of 348 at 25 Ma (van Hinsbergen et al., 2015), corresponding to an expected inclination of 6538, assuming a geocentric axial dipole (GAD) field model.
The mean inclination within normal polarity zones becomes progressively shallower throughout the application of our protocol. The initial inclination of 608 (Figures 4b and 4c) is steeper than the expected 538 value at 25 Ma (van Hinsbergen et al., 2015). We also note that natural and coring-induced compaction commonly causes inclination shallowing (e.g., Kodama, 2012), so in the absence of steep drilling overprints, ARMs, or other directional biases, we predict that we could observe inclinations shallower than 538. Indeed, this is what we observe in our normal polarity data. Mean inclination values within reversed polarity zones, however, are 2368, which are very shallow relative to the expected GAD inclination of 2538 at 25 Ma (van Hinsbergen et al., 2015). We suggest that, in addition to compaction processes, the shallow inclination of reversed polarity zones probably relates to imperfect AF demagnetization of a steep downward-directed component. This is sometimes observed in the Zijderveld diagrams ( Figure 5b) and also supported by the difference in average NRM intensity: the NRM intensity of some reversed polarity intervals is lower than surrounding normal polarity intervals. For example, NRM intensities within polarity zone R6 are usually lower than those in N5 and N6 ( Figure 6) and NRM intensities in R14 are generally lower than those in N13 and N14 (Figure 7). This downward-directed component is too steep to be a primary magnetization but is consistent with a coring-induced magnetization (e.g., Bowles, 2007) or a spurious ARM acquired during NRM demagnetization. We limited the effects of this potential overprint on component directions by not anchoring the NRM component to the origin.

Magnetostratigraphy
Our noise-masking procedure applied to optimized PCA directions produces clearly defined reversed or normal polarity zones (designated R1-N14 in Figures 6 and 7). Data from overlapping intervals of the three holes show very consistent results. The stratigraphic uncertainties in reversal positions (Table 1) range between 4.5 and 65.5 cm (median 19 cm), compared to 4-67 cm (median 18 cm) in the shipboard data (Norris et al., 2014), excluding core gaps. The only exception is a 2 m uncertainty in the position of the N11/R12 reversal correlation. The magnetostratigraphic directions based on shore-based u-channel measurements, however, are of much higher precision and quality relative to the shipboard interpretation (cf., Norris et al., 2014, Figures F17-F19), notably after the masking of noisy data by the protocol. Thus, in this Geochemistry, Geophysics, Geosystems 10.1002/2017GC007185 (e, f) Component directions on equal area projections for the optimal range PCA results after cutoff of MAD or a 95 > 158, and after the Vandamme cutoff (Vandamme, 1994), respectively. (g) Histograms of virtual geomagnetic pole (VGP) latitudes after applying each of the methods highlight the removal of low-latitude directions after application of the protocol. (h) Histograms of peak AF steps used in the optimal six-step PCA analysis, prior to and after MAD and a 95 cutoff. (i) Histograms of intensity after 20 mT peak AF demagnetization in all data (including disturbances), and those excluding data masked by the after MAD and a 95 cutoff, and by the Vandamme-cutoff. Note that the protocol typically removes data with low intensity and high intensity (related to various disturbances).
Geochemistry, Geophysics, Geosystems 10.1002/2017GC007185 application to sediments recovered from IODP Site U1406, this protocol provides well-resolved magnetostratigraphic directions. This provides the statistical background to confidently identify normal and reversed polarity zones suitable for high precision and detailed correlation to timescales and other stratigraphic records.
The magnetostratigraphic framework also provides the high-fidelity age control required for environmental and paleoclimatic studies of the OMT interval. This framework resolves inconsistencies in the shipboard identification of the C6Cn reversals (Norris et al., 2014) through the correlation of zones N6 to N8 to the three normal subchrons of Chron C6Cn. In addition, we identify putative normal polarity intervals (Table 1)  We calculate sedimentation rates (Figure 8) based on the correlation of the Site U1406 magnetostratigraphy (Table 1) to the Geological Time Scale 2012 (GTS2012, Hilgen et al., 2012;Vandenberghe et al., 2012). The Miocene interval of this record (35-95 m CCSF-M; 21-23 Ma) has the highest linear sedimentation rate (LSR) of the studied interval, with values up to 3 cm/kyr, apart from Subchron C6AAr.3r when LSR is 0.9 cm/kyr. Within the Oligocene part of the record (95-175 m CCSF-M; 23-27 Ma) LSR gradually decreases from 2.5 cm/kyr above the slump to 1 cm/kyr in Subchron C6Cn.2r (latest Oligocene), except during Subchrons C7n.1n (LSR 2.9 cm/kyr) and C8n.1r (LSR 0.8 cm/kyr).

Occurrence of Rarely Recorded Geomagnetic Events
The moderate LSR (average 2.3 cm/kyr) calculated at IODP Site U1406 makes it adequate for recording chrons, subchrons, and possibly some short-duration geomagnetic events (e.g., Roberts & Winklhofer, 2004). Short-duration events were originally labeled cryptochrons after identification in marine magnetic anomaly (MMA) profiles (e.g., Cande & Kent, 1992LaBrecque et al., 1977) and may represent either brief periods of low geomagnetic field intensity (Cande & LaBrecque, 1974), or short polarity intervals (Blakely, 1974). These events can be classed as subchrons (cf., Cande & Kent, 1992) once such short-duration directional changes are globally identified and include two well-characterized geomagnetic polarity reversals. A more detailed description of the frequency of the occurrence of short-duration events may help with a more fundamental understanding of geodynamo processes; for example, why does the geomagnetic field reverse at irregular or on very short time intervals?
Several short-duration, rarely recorded geomagnetic events have been previously reported during Subchrons C6AAr.2r and C8n.1n, and Chrons C6Br and C7Ar at South Atlantic Ocean ODP Site 1090 and at equatorial Pacific Ocean ODP Site 1218 and IODP Site U1334 (Channell et al., 2003(Channell et al., , 2013Lanci et al., 2005; Figures 6 and 7). However, the relatively stratigraphically condensed character of these sites (average sedimentation rates 1-1.5 cm/kyr) makes it difficult to obtain a complete inventory of short-duration geomagnetic events during the Oligocene to Miocene epochs, and at present, these potential geomagnetic events are observed only in these records. Site U1406 has a higher LSR (up to 3 cm/kyr over the studied interval) and may therefore be better suited to record short-duration geomagnetic events than the records from the Pacific and South Atlantic Ocean. We can thus use the detailed magnetostratigraphy from Site U1406 to test the fidelity of previously published magnetostratigraphic records, including these rarely observed geomagnetic events.

10.1002/2017GC007185
At IODP Site U1406, we observe similar short-duration intervals of normal polarity within Chrons C6Br and C7Ar (Figures 5i-5l and 5q-5t, yellow shaded bars in Figures 6 and 7), similar to Channell et al. (2003Channell et al. ( , 2013 and Lanci et al. (2005), and also, for the first time in a sedimentary record, in Chron C7r (Figures 5m-5p, yellow shaded bars in Figures 6 and 7). However, evidence for a potential reversed polarity interval during Subchron C8n.1n is less clear (Figures 5u-5x), and Subchron C6AAr.2r does not contain an interval of normal polarity, in contrast to previous reports from Site U1334 that identified a new polarity interval in Subchron C6AAr.2r (Channell et al., 2013). Preliminary rock magnetic experiments indicate that these short intervals of normal polarity do not correspond to anomalous rock magnetic or lithostratigraphic properties (e.g., Figure  3, Sample U1406B-13H-3A, 22 cm is in the middle of the normal polarity interval within Chron C7Ar, Zijderveld diagram in Figure 5r). Therefore, we conclude that these short-duration changes in magnetic polarity are not caused by a change in magnetic or sediment characteristics (e.g., mineralogy or grain-size) but reflect genuine features of the geomagnetic field. In total, we observe three new short-duration polarity intervals with high confidence and tentatively identify a fourth interval at Site U1406.
The short polarity change in Chron C6Br occurs at 73.5 m CCSF-M in both Holes U1406A (Figures 5i-5l) and U1406C ( Figure 6). This interval is present between 22.331 and 22.354 Ma (Table 1), based on linear interpolation between the reversal boundaries of Chron C6Br using the GTS2012 age model . Another short-duration polarity change occurs at 12% from the top of Chron C7r (116.4 m CCSF-M, Figures 5m-5p and 6), with its top and bottom reversals at 24.509 and 24.525 Ma (Table 1). A cryptochron is present in the middle of Chron C7r in MMA profiles from the North Pacific (Cande & Kent, 1992), but this has never been unequivocally demonstrated to reflect directional variability. Thus, the interval of normal polarity within Chron C7r at Site U1406 may be the same geomagnetic feature as observed in MMA profiles if the sedimentation rate at Site U1406, the spreading rate in the North Pacific, or both, varied substantially throughout Chron C7r. Testing this correlation will require further work on the integration of the  (Billups et al., 2004;Liebrand et al., 2016;P€ alike et al., 2006a, 2006b. Timing of the Oligocene-Miocene Transition (OMT) and Mid-Oligocene Glacial Interval (MOGI) intervals follow Liebrand et al. (2017). (b) The magnetostratigraphy for IODP Site U1406 from this study. Magnetobiostratigraphic tie points are plotted in (c), and our correlation to the geomagnetic polarity time scale Vandenberghe et al., 2012) in (e). Calculated linear sedimentation rates, with uncertainties propagating from the depth domain, are shown in (d).
Geochemistry, Geophysics, Geosystems 10.1002/2017GC007185 MMA profiles and the sedimentary record of Site U1406, such as the use of relative paleointensity to exclude variability in field intensity or astronomical tuning to provide alternative age control.
The third anomalous short-duration polarity interval observed at Site U1406 occurs within Chron C7Ar (Table 1 and Figures 5q-5t and 7). A complex sequence of paleomagnetic directions, present in Core U1406B-13H, hampers straightforward interpretation. Most of Chron C7Ar consists of an interval of uncertain polarity with southward-directed declination and positive inclination (125-127 m CCSF-M). The last step of our protocol, the Vandamme cutoff (Vandamme, 1994), masks almost all directions in this uncertain interval. The top of Chron C7Ar, however, consists of a 1.3 m thick interval of reversed polarity in which a 0.7 m thick interval of normal polarity is present at 124.5 m CCSF-M. Zijderveld diagrams of Core U1406B-13H demonstrate that this normal polarity event within C7Ar is not part of the interval of uncertain polarity but instead represents a genuine feature of the geomagnetic field (Figures 5q-5t). We estimate the age of the normal polarity event boundaries in Chron C7Ar to 24.997 and 25.023 Ma (Table 1). Another possible, rarely observed polarity event may be present in Chron C8n.1n at 128.65 m CCSF-M, but it is not clearly expressed at IODP Site U1406 (Figures 5u-5x and 7): the declination rotates by only 908 and inclination does not decrease below 08. It is unclear whether this is a true geomagnetic feature or a coring or diagenetic artifact, so we refrain from assigning an age or duration to this feature.
The brief normal polarity interval in Subchron C6AAr.2r, previously identified at IODP Site U1334 (Channell et al., 2013), is not recorded at IODP Site U1406. Potential explanations for this observation include (1) a transient interval of low sedimentation rates at IODP Site U1406 during this geomagnetic event, (2) nonuniform geomagnetic field behavior during reversals (e.g., Glatzmaier & Roberts, 1995), or (3) poor preservation of the magnetic signal (e.g., Kodama, 2012). Despite generally high LSRs at IODP Site U1406, sedimentation rates in sediment drifts are commonly more variable than at pelagic sites (e.g., Rebesco et al., 2014). A transient interval with low LSR could have occurred in Subchron C6AAr.2r, and a low LSR reduces the possibility that short-duration geomagnetic features can be accurately reconstructed, due to lock-in and smoothing effects during the recording of the paleomagnetic signal (e.g., Roberts & Winklhofer, 2004).
Nonuniform geomagnetic field behavior during reversals can also influence the recording of a shortduration polarity event. Both magnetohydrodynamic models of reversals (e.g., Glatzmaier & Roberts, 1995) and paleointensity records of geomagnetic reversals preserved in flood basalts (e.g., Pr evot et al., 1985) reveal that the axial dipole moment deteriorates to zero and then increases in strength in the opposite direction. During the weak dipole reversal phase, nondipole components may dominate and determine where and with what characteristics the magnetic field is recorded. The influence of such nondipole components varies with latitude, and may therefore have a different effect on equatorial sites, e.g., IODP Site U1334 (Channell et al., 2013), than on midlatitude sites, e.g., IODP Site U1406 studied here. It is important to note that it is currently uncertain if sediments are able to accurately record weak transitional directions, with arguments in favor of (e.g., Laj & Channell, 2015) and against (e.g., Valet et al., 2016) this hypothesis. A latitudinally dispersed stack of VGPs of the last four reversals (e.g., Clement, 2004) and numerical geodynamo models (Wicht, 2005;Wicht et al., 2009) also suggest that the duration of a polarity transition increases with the latitude at which the transition is observed or recorded. This means that equatorial records (e.g., ODP Site 1218 and IODP Site U1334) are more likely to record the discrete polarity boundaries and stable polarity phase of short-duration subchrons than midlatitude to high-latitude records (e.g., ODP Site 1090 and IODP Site U1406), assuming similar sedimentation rates at all sites. We also note that quasistable, transitional directions of short-duration polarity events identified at Site U1406 may be filtered out by our noise-reducing protocol. The noise-reducing protocol helps identifying well-expressed features in the magnetostratigraphic data, but it may not be suited to aid the identification of weakly expressed features such as transitional directions with weaker moments and outlier directions.
When evaluating the occurrence and fidelity of short-duration polarity events at Site U1406, we must also consider chemical processes potentially affecting the magnetic signal. NRM intensity (average 10 24 to 10 25 A/m after 20 mT peak AF demagnetization) of Site U1406 sediments is about 1-2 orders of magnitude lower than those of ODP Sites 1090, 1218, and IODP Site U1334 sediments (average 10 23 A/m after 20 mT peak AF demagnetization). The long-term, down-core increase in NRM intensity at IODP Site U1406 probably reflects sedimentary compaction and changes in magnetic concentration, which is also observed in the shipboard dry bulk density (Norris et al., 2014, Figure 24). Partial reductive dissolution of iron oxides is one explanation for the weak NRM intensities encountered at Site U1406, supported by the generally grey-green Geochemistry, Geophysics, Geosystems 10.1002/2017GC007185 sediment color that is characteristic of reduced sediments (e.g., Channell et al., 2013;Giosan et al., 2002;Roberts, 2015). Further evidence comes from the presence of Mn 21 but little Fe 21 in the pore waters (Norris et al., 2014) that are also characteristic of reduced environments (Canfield & Thamdrup, 2009;Roberts, 2015). Diagenetic sulfate reduction, however, is unlikely, because the concentration of pore water sulfate is similar to seawater values (Norris et al., 2014) and because the modest level of organic-matter burial (total organic carbon <0.5 wt %, Norris et al., 2014) limits the activity of sulfate-reducers in these sediments. Therefore, it is less likely that remanence-carrying iron sulfides could have formed. We argue that the supply and degradation of organic matter drives changes in redox potential that reduces some of the magnetite but does not progress to sulfate reduction allowing the authigenic formation of remanence-carrying iron sulfides.

Dynamic Behavior in Sediment Drifts
Three stratigraphic complexities are present in the Oligocene to lower Miocene sediments from IODP Site U1406. First, contorted bedding attributable to slumping distorts bedding in all three holes in the middle of the Chron C9n interval (Figures 2b and 8). Second, the stratigraphic sequence recovered at IODP Site U1406 shows large interhole variability in the interval of Chron C6Cn (latest Oligocene to earliest Miocene; Figures 6 and 8; van Peer et al., 2017). Third, the well-developed glauconite horizon identified in the three holes at 34.5 m CCSF-M marks a 2 Myr hiatus in the lower Miocene (Figures 2a and 8), estimated from the revised Base of S. belemnos (Table 2) and the C6Ar/C6AAn reversal (Table 1).
Based on the magnetostratigraphic age model, the three stratigraphic complexities correlate broadly to intervals characterized by the highest benthic d 18 O values observed in records from the equatorial Pacific (P€ alike et al., 2006b), equatorial Atlantic (P€ alike et al., 2006a), and South Atlantic Oceans (Billups et al., 2004;Liebrand et al., 2016Liebrand et al., , 2017; Figure 8). The slump in the middle of the Chron C9n interval (27 Ma) coincides with the coldest phase of the Mid-Oligocene Glacial Interval (MOGI; Liebrand et al., 2017). It is also associated with a change in the package architecture of sediment drifts on the Newfoundland margin, indicated by seismic reflector H4 (Boyle et al., 2017). Reflector H4 is interpreted to mark the transition between an underlying seismically transparent unit and an overlying wavy, mounded seismic unit (Boyle et al., 2017). Due to the similarities in the depocenter of these units, Boyle et al. (2017) hypothesize that it is unlikely that the depth and pathway of deep-water circulation changed substantially across H4. Seismic reflector H4 may therefore represent a change in the flow volume of the bottom water mass over the Newfoundland margin.
The interval with large interhole variability (Chron C6Cn, 23 Ma) and the well-developed glauconite horizon (Chron C6Ar, 21 Ma) also coincide with cold phases near the OMT climate event and in the early Miocene, respectively. In contrast to the slump in Chron C9n, however, the large interhole variability and the glauconite horizon intervals do not coincide with well-defined seismic reflectors on the Newfoundland margin. Pronounced lateral variability over short length scales between the three holes around the OMT climate event suggests that deposition of the sediment drifts was more dynamic during the OMT climate event than before and after it. The glauconite horizon highlights a period of prolonged exposure at the seabed that we interpret as a period of nondeposition. We hypothesize that bottom water currents and sediment supply influence sediment drift settings in similar fashion as observed in the Quaternary (e.g., Stow et al., 2008). In this context, the stratigraphic complexities reflect a change in sediment supply and bottom water current strength, for example, their combined ability to erode or deposit sediment at IODP Site U1406. As in the Quaternary period, these processes are likely modulated by global climate. Future work on water mass and sediment provenance tracers may help unravel the complex combined effects of tectonic events (e.g., opening or closing of gateways) and long-term paleoclimatic trends on bottom water circulation in northwest Atlantic Ocean during the Oligocene to early Miocene.

Conclusions
A noise-detecting protocol implemented to cull noisy or erroneous paleomagnetic directions from large, continuous paleomagnetic data sets substantially improves the signal-to-noise ratio of the magnetostratigraphy in weakly magnetized sediments of IODP Site U1406 and is ready for integration with the UPmag software suite (Xuan & Channell, 2009). We correlate our shore-based u-channel sample magnetostratigraphy for IODP Site U1406 to Chrons C6Ar through C9n of the GPTS (21-27 Ma; Hilgen et al., 2012;Vandenberghe et al., 2012) using shipboard and shore-based biostratigraphic datums. We identify short-duration geomagnetic events in Chrons C6Br, C7Ar, and C7r, and maybe in Subchron C8n.1n, of which three were previously Geochemistry, Geophysics, Geosystems 10.1002/2017GC007185 identified by Channell et al. (2003Channell et al. ( , 2013 and Lanci et al. (2005). Rock magnetic properties of the sediments suggest these events are not recording artifacts but are genuine features of the geomagnetic field.
Two hiatuses (each 2 Myr long, occurring during Chron C6Ar and C9n, 21 and 27 Ma, respectively) punctuate an otherwise continuous sediment drift record at Site U1406. Additionally, the magnetostratigraphy dates substantial interhole variability at Site U1406 to the Oligocene-Miocene Transition climate event. This observation highlights the dynamic character of sediment drifts and that the construction of a robust splice and age model for such depositional environments requires the integration of multiple independent stratigraphic data sets. The three intervals of stratigraphic complexity at Site U1406 broadly correspond to global cold periods inferred from correlation to benthic d 18 O records from ODP Sites 926, 1090, 1218, and 1264. We hypothesize that this stratigraphic complexity is caused by a change in sediment supply or bottom water current strength during colder climates. Future work on sediment and water mass provenance and their links to global climate may help test and refine our understanding of pre-Pliocene ocean circulation in the North Atlantic Ocean.