Geomagnetic relative paleointensity and direction during the last 40,000 years obtained from a sediment core in the Nankai Trough

Construction of regional geomagnetic secular variation curves for the last several tens of thousands of years is important for understanding the behavior of non-dipole fields and applications to geochronology. Around Japan, secular variation records of older than 10 ka was scarce, in particular for relative paleointensity (RPI). Here, we conducted a paleomagnetic study of a sediment core covering the last ~ 40 kyr taken from a small basin in the Nankai Trough. The core consists of homogenous hemipelagic sediments except for turbidites and volcanic ashes. The age model was constructed based on seven 14 C datings and two volcanic ashes. Turbidites and volcanic ashes were excluded from the construction of secular variation curves because of geologically instantaneous deposition. It was revealed that the magnetization of this core is carried largely by detrital magnetic minerals, although magnetofossils are also contained. Bulk magnetic properties show some temporal changes in magnetic concentration and grain size, but still homogeneous enough for reliable RPI estimations except for turbidites and volcanic ashes. The resultant RPI shows no correlation with the normalizer, anhysteretic remanent magnetization, of the RPI estimations or with a proxy for a magnetic grain size and/or the proportion of magnetofossils to detrital magnetic minerals. The obtained RPI record shows a long-term increasing trend since ~ 40 ka, which coincides with global stack curves. On the other hand, there are some differences in shorter timescale variations, which may reflect non-dipole fields. This study demonstrated that hemipelagic sediments in the Nankai Trough have potential for recovering high-quality RPI records when turbidites and volcanic ashes were excluded and are useful for accumulating records to construct a regional master curve.


Introduction
Geomagnetic field variations have a very wide timescales from shorter than a second to millions of years or longer.Variations of longer than tens of years are considered to be originated from the Earth's interior, and knowledge of secular variations is important for understanding geodynamo processes in the core (Merrill et al. 1998).For the last 10 kyr, a large number of secular variation records were accumulated using archaeological materials, lake and marine sediments, and volcanic rocks, and global field models including non-dipole components were constructed (e.g., CALS10k, Korte et al. 2011).In addition to geomagnetic interests, regional reference curves of secular variations and predictions from global geomagnetic field models are useful also for geochronological applications, for example age estimations and inter-core correlations of sediment cores (Kanamatsu et al. 2017;Korte et al. 2019).
Efforts have been made to extend geomagnetic field models to older times, and a model for the last 100 ka was reported (GGF100k, Panovska et al. 2018).However, the spatial and temporal resolutions of this model is still limited because the number of available records is not enough for older than 10 ka and their site distribution is biased.Thus, further accumulation of regional paleomagnetic secular variation records of high enough resolutions is still necessary.For this purpose, sediments with high enough sedimentation rates are required.Around Japan, which is the target region of this paper, such sediments occur in forearc basins and landward trench slopes.Sediments in these depositional environments were previously imagined to be unsuitable for relative paleointensity (RPI) studies because magnetic properties including magnetic concentration, grain size, and mineralogy potentially have large fluctuations compared with pelagic sediments of slow deposition.Furthermore, in sediments of these regions, turbidites can frequently be intercalated.Turbidite layers need to be detected and excluded when time variations of the geomagnetic field are considered because these layers are of geologically instantaneous deposition.However, it was sometimes not easy to detect them when they are fine grained and difference in appearance compared with sediments below and above is subtle.The potential rock magnetic inhomogeneity and the presence of turbidite layers are probably the main reasons why sediments in forearc basins and landward trench slopes around Japan were seldom used for studies of secular variations, in particular of RPI.However, recent technological development including X-ray CT scan and X-ray fluorescence (XRF) core scan for analyzing sediment cores have made it possible to efficiently identify and characterize turbidite layers (e.g., Rothwell et al. 2006;Van Daele et al. 2014;Okutsu et al. 2018;Hsiung et al. 2021).
In this paper, we present the results of a paleo-and rock magnetic study of a sediment core covering the last ~ 40 kyr, which was taken from the Nankai Trough.Geomagnetic secular variations older than 10 ka were little reported so far around Japan, in particular for variations in RPI.We show that the sediments have magnetic and sediment properties of homogeneous enough for reliable RPI estimations when turbidite and volcanic ash layers are excluded, and that reductive dissolution of magnetic minerals, which can be a problem in hemipelagic sediments with relatively large organic carbon inputs, has not occurred.Then, we present a RPI record as well as directional variations for the last 40 kyr.

Materials and age model
The sediment core PC04 used in this study was taken during R/V Hakuho-maru KH-17-2 cruise from the Nankai Trough (Fig. 1).The position and water depth of the coring site are 33° 15.19'ʹ N, 136° 39.22'ʹ E and 2255 m.The site is located at a small hill in an ENE-WSW elongated isolated basin between the accretionary prism and the forearc basin off Kumano.From the topographic features of the basin, it is expected that the basin does not receive sediment supply from the river-submarine canyon system and that it preserves most sedimentary records.Such a basin is called a terminal basin.We consider that the deposition in terminal basins is mostly continuous without significant hiatuses eroded by turbidites because turbidites in the depositional environment of terminal basins like the studied site are thought to be originated by surface sediment remobilization (Okutsu et al. 2018), which may not have capability of significant erosion.
The studied core consists mainly of hemipelagic mud and intercalates turbidites and volcanic ashes (Fig. 2).Sixteen fine-grained turbidite layers were identified based on visual core description, X-ray CT scan images, elemental profiles from an XRF core scanner, and physical properties including magnetic susceptibility (Okutsu et al. 2018;Okutsu 2019).Two volcanic ash layers are intercalated at 1.48-1.53m and 4.19-4.24m in depth, which are identified as the Kikai-Akahoya (K-Ah) ash (median age 7253 cal yr BP, Albert et al. 2019) and the Sambe-Ukinuno (SUk) ash (median age 19,551 cal yr BP, Albert et al. 2018), respectively (Okutsu 2019).These turbidite and volcanic ash layers were excluded in this study for the purpose of constructing a geomagnetic secular variation curve because they represent geologically infinitely short period of time.
To construct the age model of the core, dating with radiocarbon ( 14 C) was conducted using planktonic foraminifers extracted from seven horizons (Table 1 and Fig. 2).The measurements were carried out using a   Ramsey 2009).We assumed zero year for the marine reservoir effect (ΔR = 0).The median value of the calendar age was used to construct the age model shown in Fig. 3.The two ash layers mentioned above were also used as control points.Cumulative depth after excluding turbidite and ash layers were used as the depth scale in Fig. 3 and hereafter in this paper.Constant sedimentation rates were assumed between the dated horizons.Ages above (below) the uppermost (lowermost) dated horizon were extrapolated using the sedimentation rates of the adjacent dated intervals.The average sedimentation rate is approximately 18 cm/kyr, and the sedimentation rate increases upcore.Discrete samples for rock-and paleomagnetic measurements were taken continuously from half-split core surface using 7 cc plastic cubes.In total 306 discrete samples were obtained.

Paleomagnetic direction and relative paleointensity
Natural remanent magnetization (NRM) was measured for all samples using a pass-through cryogenic magnetometer (2G Enterprises Model 760R) in a shielded room at Marine Core Research Institute (MaCRI), Kochi University.Stepwise alternating-field (AF) demagnetization of NRM was conducted at 15 steps, at 5 mT intervals up to 60 mT and at 70 and 80 mT.Then, anhysteretic remanent magnetization (ARM) was imparted at a peak AF of 80 mT superimposed on a DC field of 0.1 mT.The ARM was AF demagnetized at the same steps as the AF demagnetization of NRM.
Characteristic remanent magnetization (ChRM) direction was determined using principal component analysis (PCA, Kirschvink 1980;Fig. 4  maximum angular deviation (MAD) was adopted.Samples with MAD > 15° were excluded from further analyses.Samples facing core-section boundaries were also excluded due to potential physical disturbance.RPI was estimated from NRM intensity normalized by ARM.RPI was calculated as a best-fitting slope in a diagram plotting a pair of NRM and ARM for each AF demagnetization step (Channell et al. 2002;Fig. 4).
The demagnetization interval having at least 5 steps with the largest correlation coefficient was adopted.

Bulk magnetic properties
Magnetic susceptibility was measured using an AGICO KappaBridge MKF-1 susceptometer at AORI, The University of Tokyo.Magnetic susceptibility represents the concentration of magnetic minerals in sediments.Isothermal remanent magnetization (IRM) was imparted using a pulse magnetizer (Magnetic Measurements MMPM10) and measured using a spinner magnetometer (Natsuhara-Giken SMD-88) at MaCRI, Kochi University.First, IRM was imparted at 2.5 T, which is regarded as saturation IRM (SIRM) in this paper.Then, IRM of 0.3 T were successively imparted in the direction opposite to the initial IRM.S ratio (S -0.3 T ) was calculated according to the definition of Bloemendal et al. (1992).S -0.3 T represents the proportion of magnetic minerals with coercivity lower than 0.3 T, which provides information on magnetic mineralogy in the sediments.ARM before AF demagnetization mentioned above was converted to ARM susceptibility (k ARM ), and the ratio of k ARM to SIRM was calculated.k ARM /SIRM is used as a magnetic grain-size proxy (Banarjee et al. 1981;King et al. 1982).

First-order reversal curve (FORC) measurements
FORC diagrams have become widely used in rock-and paleomagnetic studies to characterize quantitatively magnetic mineral assemblages from information on the distribution of coercivity (H c ) and local interaction field (H u ) (Pike et al. 1999;Roberts et al. 2000;2014;2022;Egli 2021).We conducted FORC measurements using an alternating-gradient magnetometer (AGM, Princeton Measurements Corporation MicroMag 2900) at AORI, The University of Tokyo.Ten samples were selected from the sediments except for turbidites and volcanic ashes to cover variations of k ARM /SIRM and S ratios.A total of 165 FORCs were measured for each sample, with H c between 0 and 100 mT, H u between − 50 and 50 mT, and a field spacing of approximately 1.3 mT.The maximum applied field was 1.0 T, and the averaging time for each measurement point was 200 ms.FORCinel software (Harrison and Feinberg 2008) was used to produce FORC diagrams, and the VARIFORC algorithm of Egli (2013) was used to smooth the data with smoothing parameters of S c0 = 4, S b0 = 3, and S c1 = S b1 = 7.

Transmission electron microscopy
We conducted observation of magnetic grains using a transmission electron microscope (TEM) in order to directly observe magnetic grains carrying remanent magnetization, in particular to examine the presence of magnetofossils from the characteristic morphology and grain size confined within the single-domain (SD) range (e.g., Kopp and Kirchvink 2008).Magnetic minerals were extracted from two samples at 1.99 and 5.35 m in depth using a magnetic finger (Kirschvink et al. 1992;Yamazaki and Shimono 2013).A TEM (JEOL JEM-1400) at AORI, The University of Tokyo, operated at 120 kV, was used for the observation.

Grain-size analyses
Grain-size analyses of bulk sediments were conducted using a laser diffraction grain size analyzer (Horiba LA-960) at the National Institute of Advanced Industrial Science and Technology.Twenty-eight samples were selected at approximately the same depth intervals excluding turbidites and volcanic ashes.Small amounts of sediments were taken from the paleomagnetic discrete samples, and they were dispersed well using an ultrasonic bath before the measurement.

Paleomagnetic direction and intensity
ChRM directions are well resolved on the Zijderveld diagrams presenting stepwise AF demagnetization of NRM (Fig. 4).MADs of most samples are less than 15° (Fig. 5).
A soft secondary component of probably viscous remanent magnetization origin was erased in the first few steps of the AF demagnetization.The PCA fitting was not anchored to the origin, but deviations from the origin were small in general.Figure 5 shows the resultant variations of inclination and declination with age.Declinations are relative because the core was not azimuthally oriented.Here, the mean declination was aligned to zero.
Directional variations in the depth scale before excluding turbidites and tephra layers are presented in Additional file 1: Fig. S1.The SUk tephra and some turbidites showed anomalous directions inconsistent with those above and below them.
The directional records of Core PC04 are compared with the records from off Shikoku, Southwest Japan (Ohno et al. 1993) and Lake Biwa, Central Japan (Hayashida et al. 2007) (Fig. 6).All three records show inclinations shallower than that expected from the geocentric axial dipole field at each site.The inclination anomaly around Japan for the last 100 kyr is estimated to be − 2° to − 4° from the global geomagnetic field model for the last 100 kyr (Panovska et al. 2019), which explains only part of the observed shallow inclinations.Hence these inclination records may partly be influenced by inclination shallowing in detrital remanent magnetization (DRM) acquisition processes (e.g., Griffiths et al. 1960).The variation patterns of declination and inclination agree in some time intervals, but there are differences in other intervals.Various factors may be responsible for the differences.The age control of Core PC04 is not good for older than ~ 20 kyr cal BP.In the Off Shikoku (Ohno et al., 1993) Lake Biwa (Hayashida et al., 2007)  Off Shikoku (Ohno et al., 1993) Lake Biwa (Hayashida et al., 2007) (b) Inclination core off Shikoku, turbidite layers might have been overlooked because modern techniques for detecting turbidites such as X-ray CT scan were not available at that time.DRM acquisition processes and lock-in depth in the freshwater sediments of Lake Biwa may be different from those of marine sediments (Tauxe et al. 2006).We consider that the absence of the AT tephra in our core does not necessarily indicate a hiatus.Instead, we infer that the tephra may be diluted by a turbidite and hence not easy to be detected.The horizon expected from the age model corresponds to the interval of frequent thin turbidites.
NRM-ARM demagnetization plots of Core PC04 for calculating RPIs show linear relationship in general except for the first few demagnetization steps (Fig. 4).RPI variations with age are presented in Fig. 5. See Additional file 1: Fig. S1 for RPI in the depth scale including tephras and turbidites.The age of the bottom of the core is estimated to be ~ 40 kyr cal BP, although it is not well constraint.It is inferred that the core does not reach the Laschamp excursion (41.4 ka, Channell et al. 2017) because no directional flip is observed.Alternatively, shallow inclination near the bottom of the core with low RPI might be a manifestation of the Laschamp excursion.
Reliability of the obtained RPI record and comparison with other records will be discussed later.

Magnetic and sediment properties
Changes in bulk magnetic properties with age after excluding tephras and turbidites are shown in Fig. 7. See Additional file 1: Fig. S2 for those in the depth scale including tephras and turbidites.Magnetic susceptibility of Core PC04 ranges from ~ 2 to ~ 6 × 10 -4 SI excluding tephras and turbidites (Fig. 7a).The two tephras and some turbidites have much higher magnetic susceptibility (Additional file 1: Fig. S2).k ARM /SIRM changes between ~ 1.5 to 5 × 10 -4 m/A (Fig. 7c).A lower ratio suggests a larger average magnetic grain size (Banerjee et al. 1981;King et al. 1982).The influence of tephras and turbidites on k ARM /SIRM is small in general, although some of them produce minor peaks or lows (Additional file 1: Fig. S2).This suggests that their magnetic grain sizes do not differ significantly from hemipelagic sediments above and below.Nonetheless, some turbidites accompany anomalously high RPIs (Additional file 1: Fig. S1).S -0.3 T of 0.95 or higher throughout the core indicates that the magnetization of this core is mainly carried by low-coercivity magnetic minerals like magnetite (Fig. 7d).These magnetic-property variations indicate that magnetic mineral dissolution by reduction diagenesis has not occurred in the studied sediments.It is well documented that reductive dissolution of magnetic minerals results in one order of magnitude or more drop in magnetic concentration, a coeval increase in average magnetic grain size, and a decrease in S -0.3 T to ~ 0.8 (Yamazaki et al. 2003;Yamamoto et al. 2007;Korff et al. 2016).The increase in average magnetic grain size is due to earlier loss of finer magnetic grains, and the decrease in S -0.3 T is due to that high-coercivity magnetic minerals like hematite is more resistive to reduction diagenesis than magnetite.The ratio of SIRM/k less than ~ 25 indicates that greigite formation in anoxic environments has not occurred in the core (Fig. 7e).Sediments containing greigite is known to have high SIRM/k, ~ 30 × 10 3 A/m or higher (Oldfield et al. 2003;Liu et al. 2018;Chen et al. 2021).In early diagenesis processes, the growth of secondary greigite occurs when reduction further proceeds after magnetite dissolution at a certain depth below the iron-redox boundary (e.g., Roberts 2015).Thus, the absence of greigite estimated from SIRM/k is consistent with the observation that reductive dissolution of magnetite has not started in the studied core.
The magnetic properties presented in Fig. 7 show coeval changes; they are lower between ~ 13 to ~ 27 kyr cal BP except for an increase in magnetic susceptibility between ~ 17 to ~ 20 kyr cal BP.The low k ARM /SIRM and S -0.3 T indicate that the average magnetic grain size and the proportion of high-coercivity magnetic component are larger in this time interval.The time interval roughly corresponds to the Marine Isotope Stage (MIS) 2, suggesting a connection to global climatic changes.Then, the interval of higher values in the magnetic properties below it might correspond to MIS 3.
The FORC diagrams in the studied sediments show a ridge-like feature along H u ≈ 0, which is superimposed on a broad component with a wide H u distribution (Fig. 8).The former is called the central ridge, which indicates the presence of non-interacting SD grains and is interpreted to be carried by intact chains of magnetofossils (Chen et al. 2007;Yamazaki 2008;Egli et al. 2010;Li Jinhua et al. 2012;Roberts et al. 2012;Chang et al. 2014).The TEM observation supports the existence of magnetofossils (Fig. 9); magnetofossils are identified from the characteristic morphologies, these are hexagonal prism, bullet-shape, and equant octahedron, and SD sizes (tens of nanometers in magnetite).The broad component with a wide H u distribution indicates the presence of significant magnetostatic interactions, which is interpreted to be carried by interacting SD, vortex, and multi-domain (MD) grains of mainly detrital origin.Magnetofossils of collapsed chains and those originated from magnetotactic bacteria with multistranded magnetosome chains should have significant magnetostatic interactions, and hence also contribute to the broad component.The variations of the FORC distributions among the measured samples are consistent with magnetic grain-size changes estimated from k ARM /SIRM.The peaks in the FORC distributions of samples with lower k ARM /SIRM have lower H c values around 10 mT, as exemplified by the sample at 17.5 kyr shown in Fig. 8b.The broad component of such samples shows a more MD-like feature; contours of the FORC distributions diverge toward the H u axis.The extracted central-ridge component is relatively small compared with the ridge-free broad component (Fig. 8df ).We semi-quantitatively evaluated the proportion of the two components by integrating separately the FORC distributions of the two in the profiles along H c = 20 mT (Fig. 8g).The result indicates that the FORC proportion of the central-ridge component is about 5% or less, and this does not change significantly within the core.This suggests that the contribution of magnetofossils to the magnetization is minor compared with the detrital component in the studied sediments.
The results of sediment grain-size analyses show that the observed grain-size distribution can be explained by two components assuming a log-normal distribution (Fig. 10a-c); the finer component has the median size of approximately 10 μm, and the median size of the coercer component ranges from ~ 50 to ~ 70 μm.The median size of the former is uniform throughout the core, whereas the size of the latter slightly increases with age (Fig. 10d).The finer component occupies ~ 70% in volume when all grains are assumed to have a spherical shape, and the proportion of the finer component slightly decreases with age (Fig. 10e).Sediment grain-size changes are decoupled from magnetic grain-size variations.Despite the average magnetic grain-size increase between ~ 13 to ~ 27 kyr cal BP, sediment grain size does not show any remarkable change in this time interval.

Discussion
In this study, we have examined magnetic and sedimentological properties of Core PC04 to assess the reliability of the RPI record.Currently, possible influence on RPI of sediment magnetic-property variations associated with environmental changes (so-called "lithological contamination") is a serious problem of RPI estimations (Xuan and Channell 2008;Valet et al. 2011;Yamazaki et al. 2013).Inverse correlation between RPI and k ARM / SIRM is frequently reported in previous RPI studies using marine sediments (Hofmann and Fabian 2009;Sakuramoto et al. 2017;Yamazaki et al. 2013;Inoue et al. 2021;Li Jiaxi et al. 2022).Recent studies indicate that magnetofossils have lower RPI recording efficiency than the detrital component (Gai et al. 2021;Inoue et al. 2021;Li Jiaxi et al. 2022;Yamazaki et al. 2023).On the other hand, k ARM /SIRM is a proxy for the proportion of magnetofossils to detrital magnetic minerals (Egli 2004;Yamazaki 2008;Zhang et al. 2022) as well as for magnetic grain size; k ARM /SIRM increases with increasing proportion of magnetofossils.Thus, changing proportion of magnetofossils to detrital magnetic minerals can cause the inverse correlation between RPI and k ARM /SIRM.In the studied core, no correlation between RPI and k ARM / SIRM is observed (Fig. 11a).Our rock-magnetic analyses show that the contribution of magnetofossils to the magnetization of the sediments is very small.It is thus expected that RPI acquisition efficiency of the sediments does not change throughout the core.In this core, k ARM / SIRM is considered to represent magnetic grain-size changes rather than the proportion of magnetofossils to detrital magnetic minerals because of the minor contribution of magnetofossils.In summary, it is considered that the lithological contamination to RPI is minimal in the studied core.In addition, obtained RPIs do not correlate with the normalizer of the RPI estimations, that is ARM, either (Fig. 11b), which also supports the reliability of our RPI record (Tauxe 1993).This implies that differences in the magnetizability of sediments due to the differences in magnetic mineral concentration are appropriately corrected by ARM.In Fig. 11, the data distribution can be divided into three groups, which correspond to different age intervals.k ARM /SIRM and k ARM values younger than ~ 12 ka are high in general, whereas those between ~ 12 to 26 ka are low.They form clusters in Fig. 11 because the upcore transition from low to high are rapid (Fig. 7).The age of the transition roughly corresponds to the MIS 2/1 boundary.The RPI variation ranges in the two age intervals are similar.The low RPI interval older than ~ 26 ka tends to have medium k ARM / SIRM and k ARM values.
In the studied core, there is no sign of magnetic mineral dissolution or secondary greigite formation associated with reduction diagenesis, which may erase and/ or overprint original paleomagnetic records.Sediments with high sedimentation rates are preferable for obtaining high-resolution paleomagnetic secular variation records.However, such sediments often have high organic-carbon contents and hence in anoxic environments, where reduction diagenesis prevails (e.g., Yamazaki et al. 2003;Roberts 2015;Korff et al. 2016).It was revealed that sediment grain size is uniform throughout the core when turbidite and volcanic ash layers are excluded.This suggests that hydrodynamic conditions, which may influence DRM acquisition (Jezek and Gilder 2006), would not have changed significantly.Thus, hemipelagic sediments in terminal basins in the Nankai Trough region, which have relatively high sedimentation rates without reduction diagenesis, can potentially be suitable materials for geomagnetic secular variation studies, although previously they were not used widely.Fluctuations of magnetic mineral concentration, magnetic grain size, and mineralogy in this core are relatively small (Fig. 7), which meets the empirical criteria for reliable RPI estimation (Tauxe 1993).
The RPI record of Core PC04 is compared with the GLOPIS-75 stack covering the last 75 kyr (Laj et al. 2014) (Fig. 12).Both records show an increasing trend since ~ 40 ka.The apparent RPI decease after ~ 3 kyr cal BP in Core PC04 is most likely due to physical disturbance of the very surface sediments at the coring.The RPI increase since the Laschamp Excursion is commonly recognized also in the lower resolution global PRI stacks, Sint-2000 for the last 2 Myr.(Valet et al. 2005) and PISO-1500 for the last 1.5 Myr.(Channell et al. 2009).On the other hand, some shorter time-scale changes do not agree between the two.A high from ~ 25 to ~ 22 kyr cal BP occurs only in Core PC04.However, the PISO-1500 stack also shows a rather strong paleointensity around 22 ka comparable to that during the last 10 ka, although the age resolution is limited.A high from ~ 33 to ~ 30 kyr cal BP exists only in the GLOPIS-75.These differences may suggest existence of a non-dipole field in the Asian region.Alternatively, the GLOPIS-75 stack might be influenced by a non-dipole component; the GLOPIS-75 stack was established by stacking 24 RPI records from the North and South Atlantic Oceans, the Mediterranean, and the Indian Ocean, but the site distribution is still biased in the Atlantic region (Laj et al. 2004).Checking consistency of the PC04 RPI record with other records from nearby sites is hindered by the scarcity of RPI records for the last 40 kyr near Japan except for Holocene.Hayashida et al. (2007) reported normalized intensity, NRM/ARM, from a sediment core in Lake Biwa, Central Japan.Although Hayashida et al. (2007) mentioned that the magnetic properties of the core may not be suitable for RPI reconstruction and avoided to use the term RPI, their record also shows a high between ~ 25 and ~ 22 kyr cal BP when their age model was recalculated using the methodology used for Core PC04.
Further accumulation of reliable regional RPI records with high enough resolution is required to discuss growth and decay of non-dipole components.This study demonstrated that for this purpose hemipelagic sediments with relatively high sedimentation rates deposited in terminal basins between accretionary prisms and forearc basins have potential for recovering high-quality RPI records.

Conclusions
We have presented a RPI record since ~ 40 kyr cal BP obtained from a sediment core taken from the Nankai Trough.RPI in this period was little reported previously from the NE Asian region.Rock magnetic and sediment grain-size analyses revealed that when turbidite and volcanic ash layers are excluded the sediments are homogeneous enough for reliable RPI estimations.Reductive dissolution of magnetic minerals or growth of secondary greigite have not occurred in the studied core.The magnetic minerals in the sediments are mainly of detrital origin, and the contribution of magnetofossils to the magnetization is small.The obtained RPI record shows no sign of lithological contamination.When the RPI record of Core PC04 is compared with that of the GLO-PIS-75 stack, an increasing trend since ~ 40 kyr cal BP is commonly observed.On the other hand, there are some differences in shorter timescale variations; a high from ~ 25 to ~ 22 kyr cal BP occurs only in Core PC04, and a high from ~ 33 to ~ 30 kyr cal BP exists only in the GLOPIS-75.Non-dipole fields in the Asian and/or North Atlantic regions may be responsible for the differences.

Fig. 1 a
Fig. 1 a Shaded relief map of the Nankai Trough and surrounding areas.Red box indicates the area of panel b.b Topographic contour map in the vicinity of the site of Core KH-17-2 PC04.Contour intervals are 200 m (bold) and 50 m (thin)

Fig. 2
Fig. 2 Simplified columnar section of Core PC04.Median values of calendar ages are presented at the horizons where dating with 14 C was conducted

Fig. 3 Fig. 4 (
Fig.3The age model of Core PC04.Blue dots are based on dating with 14 C, and red dots are based on the identification of volcanic ashes: K-Ah: Kikai-Akahoya ash, SUk: Sanbe-Ukinuno ash.The depth scale is the cumulative depth after turbidite and volcanic ash layers were excluded

Fig. 5
Fig. 5 Maximum angular deviation (MAD), declination, inclination, and relative paleointensity (from left to right) obtained from Core PC04.Blue and red diamonds along the age axis indicate age-control points based on 14 C dating and volcanic ashes, respectively.Black triangles show the locations of core-section boundaries.Error bars in declination and inclination are based on MAD values, and those in relative paleointensity are from the uncertainties of the best-fitting slopes in NRM-ARM demagnetization diagrams(Kono and Tanaka 1984)

(
See legend on previous page.)

Fig. 7
Fig. 7 Bulk magnetic properties of Core PC04. a Magnetic susceptibility (k), b anhysteretic remanent magnetization (ARM), c the ratio of ARM susceptibility (k ARM ) to saturation iso-thermal remanent magnetization (SIRM), d S ratio (S -0.3 T ), and e the ratio of SIRM to k. Blue and red diamonds along the age axis indicate age-control points based on 14 C dating and volcanic ashes, respectively.Black triangles show the locations of core-section boundaries

Fig. 8 Fig. 10
Fig. 8 Results of first-order reversal curve (FORC) measurements.a-c Examples of FORC diagrams, d-f profiles of FORC distribution along H c = 20 mT (dashed lines in panels a-c): the central-ridge component (blue), the ridge-free background component (black), and the total of the two (red), and g temporal variations of the relative proportion of the central-ridge component (blue) and the ridge-free background component (red) in FORC distribution based on the integrals of the individual components along H c = 20 mT

Fig. 11 aFig. 12
Fig. 11 a Relation between relative paleointensity and k ARM /SIRM, and b relation between relative paleointensity and k ARM