Sediments and sedimentary rocks are important targets for paleomagnetic studies because of their widespread occurrence and their ability to provide a continuous record of the geomagnetic field. The ubiquitous distribution of sediments and sedimentary rocks on the earth, ranging from low latitudes to high latitudes in various depositional environments, makes them valuable archives for studying spatial characteristics of the geomagnetic field and permits tests of different models for the behavior of the Earth's magnetic field (e.g., Opdyke and Henry, 1969; Constable and Johnson, 1999; Tauxe and Kent, 2004; Kodama, 2012). The slow, relatively continuous nature of depositional processes gives sedimentary rocks the ability to preserve nearly continuous readings of the geomagnetic field, and thus sedimentary rocks are well-suited to study the long-term temporal behavior of the Earth's magnetic field (e.g., Guyodo and Valet, 1999; Valet, 2003; Valet et al., 2005).

The power of sedimentary rocks for recording the spatial and temporal characteristics of the geomagnetic field hinges on the correspondence of the remanent magnetization to the local geomagnetic field at the time of deposition. Yet acquisition of detrital remanent magnetization (DRM) is complicated because it can involve physical, chemical, and biological processes during deposition and lithification of sedimentary rocks (Verosub et al., 1979; Tauxe and Kent, 1984; Anson and Kodama, 1987; Tauxe et al., 2006; Mitra and Tauxe, 2009; Lund et al., 2010; Scherbakov and Sycheva, 2010; Kodama, 2012). These processes may cause misalignment of magnetization with the ambient field, resulting in an inclination shallower than that of the ambient field, known as “inclination shallowing.” This phenomenon was noticed more than half a century ago (King, 1955), and has been observed in some natural sediments (e.g., Tauxe et al., 1980; Collombat et al., 1993) and repeatedly reported in laboratory redeposition experiments (e.g., Tauxe and Kent, 1984; Deamer and Kodama, 1990). Inclination shallowing has been attributed to depositional effects that are associated with various aspects such as particle size and shape (e.g., King, 1955; Griffiths et al., 1960; Bilardello, 2013; Bilardello et al., 2013), aggregation of sediments, or flocculation (e.g., Lu et al., 1990; Scherbakov and Sycheva, 2008), Brownian motion (e.g., Collinson, 1965), bottom current shear (e.g., Rees, 1961), viscosity, pH, deposition rate and so on (Verosub, 1977; Tauxe et al., 2006; Mitra and Tauxe, 2009; Scherbakov and Sycheva, 2010). Post-depositional processes such as bioturbation and burial compaction can modify the magnetization prior to being fully locked-in (Roberts and Winklhofer, 2004), leading to an acquisition of post-depositional remanent magnetization, or pDRM (Verosub, 1977). While bioturbation may enhance the alignment of magnetization with ambient field, post-depositional compaction can rotate magnetic particles toward the horizontal and thus cause inclination errors (e.g., Anson and Kodama, 1987; Deamer and Kodama, 1990).

Despite the fact that the acquisition mechanism of DRM remains poorly understood and inclination errors are frequently observed in the redeposition experiments, in nature, the inclination shallowing problem does not appear to occur widely in recent sediments. Recent unconsolidated sediments from low through high latitudes show an expected distribution of inclinations as predicted by the GAD model (Opdyke and Henry, 1969; Lund et al., 2010). Also, comparisons of paleomagnetic records of unconsolidated sediments to that of coeval igneous rocks often show that sediments can indeed accurately record the ambient geomagnetic field (e.g., Lund, 1985). In lithified sedimentary rocks, while faithful recording of the geomagnetic field can be demonstrated for many cases (e.g., Li et al., 2013), inclination shallowing does occur sometimes (Kodama, 1997; Tan and Kodama, 1998; Gilder et al., 2001, 2003; Tan et al., 2003, 2010; Vaughn et al., 2005; Tauxe et al., 2008; Bilardello and Kodama, 2010a,b; Kodama, 2012; Huang et al., 2013), and is recognized by comparing sedimentary records either with well-established apparent polar wander path, with coeval igneous paleomagnetic data or by rock magnetic analyses. For some sedimentary rocks such as pelagic carbonates, while many pelagic carbonates preserve high-fidelity records of geomagnetic field, some pelagic carbonates may experience diagnesis and can even be remagnetized (Roberts et al., 2013). Nevertheless, the wide absence of inclination shallowing in recent unconsolidated sediments but its presence in lithified sedimentary rocks point to a universal, common process, that is, post-depositional burial compaction during lithification, as a likely cause to the problem. This notion gains support from the paleomagnetic data of deep sea sediments that underwent compaction. The inclination shallowing of sediments from DSDP site 578 shows a general trend that corresponds to the down-core porosity variations and, overall, the degree of inclination shallowing increases with decreasing porosity down core (Arason and Levi, 1990).

Given the fact that some sedimentary rocks do suffer from inclination shallowing and over half of the paleomagnetic data in the Global Paleomagnetic Database (http://www.ngu.no/geodynamics/gpmdb/) were derived from sediments and sedimentary rocks (Kodama, 2012), it is crucial to detect and correct for inclination shallowing in sedimentary records in order to achieve accurate interpretation. To this end, it is important to first identify inclination shallowing when a coeval igneous record or a well-established reference apparent polar wander path is not available.

One way of recognizing magnetic inclination shallowing is to examine the distribution of paleomagnetic directions. Distribution of paleomagnetic directions of geomagnetic origin can be predicted from statistical paleosecular variation (PSV) models that are based on paleomagnetic data of lava flows over the past 5 Myrs (e.g., Constable and Parker, 1988; Constable and Johnson, 1999). If the PSV models are applicable to ancient times, distributions of paleomagnetic directions can then be assessed by comparing the directions with those from PSV model predictions. Tauxe and Kent (2004) proposed a PSV model, TK03.GAD, that quantifies the distribution of directions with an elongation parameter E. This model also predicts that E decreases with increasing latitudes, i.e., inclinations, defining a trend of Elongation -Inclination changes (E/I). If observed paleomagnetic directions agree with the E/I trend, the scatter of these directions is consistent with that resulting from geomagnetic secular variation. The PSV model predicts that an elongation direction of paleomagnetic directions should be in the vertical, N-S plane. It can then be deduced that the occurrence of elongation in the horizontal, E-W plane indicates inclination shallowing (Tauxe and Kent, 2004). The inclination shallowing can be corrected by finding the optimum flattening factor f using the E/I method of Tauxe and Kent (2004). The flattening factor f is defined as the ratio of (tan I_o)/(tan I_f) (King, 1955), where I_o is the observed inclination and I_f is that of the applied field. This approach needs a large dataset so that the field distribution can be sufficiently sampled and well represented.

Another approach is to study the magnetic anisotropy of samples. Magnetic anisotropy, also known as magnetic fabrics, measures the preferred orientation of magnetic particles in a rock and is represented graphically by an ellipsoid. It is widely accepted that an oblate fabric is typically produced during deposition in many sedimentary environments, which is characterized by the minimum axes K3 perpendicular or subperpendicular to the bedding (Tarling and Hrouda, 1993). The primary depositional fabrics can be further modified by post-depositional processes. During lithification, post-depositional compaction of sediments would lead to vertical shortening and development of oblate magnetic fabrics, which are characterized by the steepening of the minimum axes and the scatter of the maximum axes K1 and intermediate axes K2 within the bedding plane. Post-depositional compaction would also rotate magnetic particles toward the horizontal plane, giving rise to inclinations shallower than that of the ambient field. Consequently, the occurrence of oblate fabrics is often used as evidence for identifying inclination shallowing. Jackson et al. (1991) developed an elegant theoretical model that relates inclination shallowing to magnetic anisotropy of sedimentary rocks. The model assumes that inclination shallowing is a function of both the bulk magnetic anisotropy of remanence-carrying grains and the magnetic anisotropy of individual magnetic particles. Thus, correction for inclination shallowing can be made after bulk magnetic anisotropy and individual particle anisotropy are determined (e.g., Kodama, 1997; Tan, 2001; Kodama, 2009). This approach has been used in correcting for inclination shallowing of sedimentary rocks of different ages (e.g., Kodama and Davi, 1995; Tan and Kodama, 1998; Tan, 2001; Vaughn et al., 2005; Tan et al., 2010; Bilardello and Kodama, 2010a,b,c; Huang et al., 2013).

The anisotropy-based inclination correction model is based on a fundamental assumption that the development of magnetic anisotropy is directly related to the degree of inclination flattening. Although this is a reasonable assumption and has gained support from the laboratory compaction experiments with clay-rich sediments (Sun and Kodama, 1992), the assumed relationship between magnetic anisotropy and inclination flattening has rarely been documented in nature. Garcés et al. (1996) showed that laminated redbeds display stronger magnetic susceptibility foliation and shallower inclinations than the bioturbated and massive mass-flow facies in the extensional basin on the Iberia continental margin. Kim and Kodama (2004) reported that shale units carry shallower inclinations and better defined foliations than concretionary intervals in the Northumberland Formation on Hornby Island, British Columbia. Vaughn et al. (2005) found that well-foliated samples carry shallower inclinations than do non-foliated samples from the Cretaceous sedimentary rocks in Baja California. These studies essentially provide semi-quantitative correlation between inclination shallowing and magnetic anisotropy for different lithologies. However, a detailed, correlated progressive development of magnetic anisotropy and inclination shallowing has not been well documented in nature. This is particularly important because oblate fabrics would be gradually developed with depth in a stratigraphic column as overburden loading increases. Inclination flattening would not occur until loading-induced compaction is large enough to rotate magnetic grains toward the horizontal. Therefore, there exists a threshold bulk anisotropy, i.e., the extent to which oblate fabrics develop when compaction-induced inclination shallowing starts to occur. A lack of knowledge about the threshold bulk anisotropy could lead to misidentification and/or over-correction of inclination shallowing using magnetic anisotropy.

In this study, we collected sediment cores from the northern continental slope of the South China Sea (SCS) (Figure 1). The core sites are located ~300 km away from the mouth of Pear River to avoid high sand content that would prevent inclination from shallowing because clay content was found an important controlling factor for inclination shallowing (Sun and Kodama, 1992). Yet the core sites are not from the deepest part of the SCS, but from the continental slope in order to achieve a relatively high or moderate accumulation rate so that details of the gradual development of magnetic fabrics and corresponding changes in inclinations may be recorded.

SCS is one of the largest marginal seas in the western Pacific Ocean. The opening of the SCS resulted from the long-lasting extension along the continental margin of the South China Block due to either extrusion tectonism associated with the collision of India and Eurasia (e.g., Tapponnier et al., 1982) or the subduction of the Pacific plate toward Asia (e.g., Hall, 2002). Continued rifting and subsequent spreading of the SCS led to the formation of a series of predominantly NE-striking grabens/basins along the continental margin of the South China Block, i.e., on the northern slope of the SCS, at least since Eocene-Oligocene (Tapponnier et al., 2001). The basins receive continuous sedimentation all the way to the present day. The sediments deposited since the Late Pleistocene in these basins mainly consist of hemipelagic clays with ~40% of carbonate, and the terrigenous accumulation rates at the middle part of the northern slope reach ~3 g/cm^2*10³ year (Schönfeld and Kudrass, 1993), which is higher than those in the Miocene and early Pliocene (Li et al., 2011). With the relatively high accumulation rate, sediments in these basins on the northern slope of the SCS provide excellent testing grounds for examining the development of magnetic fabrics and changes in inclinations with depth in response to compaction in the course of sedimentary rock formation. It is conceivable that the gradual changes in magnetic fabrics and inclinations with depth could be observed, provided that long enough cores are retrieved, as deposition and post-depositional compaction in these basins are occurring.

Two gravity cores GHE24 L and GHE27L were retrieved from the northern slope of the SCS (Figure 1) in 2011. GHE24L is 5.98 m long and was collected from the water depth of 1387 m. GHE27L is 5.73 m long and was taken from the water depth of 1533 m (Figure 1).

Two cores were split into halves and paleomagnetic specimens were collected at an ~2 cm interval using non-magnetic 2.0 × 2.0 × 2.0 cm³ plastic cubes. Each plastic cube was gently pushed into the soft sediments with the arrow at the bottom of the plastic cube pointing upcore and was then carefully removed to avoid disturbance. A total of 286 and 298 oriented cubic specimens were collected from GHE24L and GHE27L, respectively.

Anisotropy of magnetic susceptibility (AMS) of all specimens was measured with a kappabridge KLY-3 susceptibility meter to estimate changes in magnetic fabrics with depth. Anisotropy of anhysteretic remanence (AAR) of selected samples from both GHE24L and GHE27L were determined following McCabe et al. (1985)'s approach in which an ARM was applied along nine different orientations in a coercivity window, typically 20–80 mT, in a 0.023 mT DC field and between the AAR acquisition of each orientation is the demagnetization of the samples at 100 mT. The AAR experiments were conducted in the paleomagnetism laboratory of the Institute of SCS Oceanography in the Chinese Academy of Science. Thermomagnetic properties of representative samples were measured by heating the samples up to 700°C while their susceptibilities were monitored in order to constrain the magnetic mineralogy. In addition, magnetic hysteresis of selected samples was measured in the Institute of Geology and Geophysics in the Chinese Academy of Science to estimate the grain size of the magnetic particles.

To obtain a record of changes in magnetic inclination with depth, a total of 118 selected specimens, on an average of ~10 cm spacing, were subjected to stepwise, typically 12 steps, alternating field demagnetization with a Molspin demagnetizer. The remanence of these samples was measured with a 2G Enterprise Inc. 755 rock magnetometer that is housed in a magnetically shielded room (residual field is less than 300 nT) in the paleomagnetism laboratory of Nanjing University. The demagnetization data are presented graphically with vector-endpoint diagrams (Zijderveld, 1967) and were analyzed using principal component analysis (Kirschvink, 1980). Fisher statistics (Fisher, 1953) was utilized to calculate mean directions. The paleomagnetic software packages PaleoMac v6.2 (Cogné, 2003) and Anisoft 4.2 (by Martin Chadima and Vit Jelinek) were used to perform data analyses. In addition, regression analyses were performed to quantify the down-core changes in the degree of magnetic anisotropy and Jackson et al. (1991)'s model was employed to correct for inclination shallowing and to estimate particle anisotropy “a”.

In order to determine whether inclination shallowing occurred in these two cores and how much the inclination is shallowed, the expected inclinations at the core sites must be calculated. According to the geocentric axial dipole model (GAD), the expected inclinations of GHE24L and GHE27L are 35.9 and 35.8°, respectively (Figures 8A,B).

Down-core variations of AMS and AAR of both GHE24L and GHE27L exhibit different characteristics above and below 2 m (Figures 2–4), i.e., non-foliated fabrics above 2 m and predominantly foliated fabrics with a general trend of linear or quasilinear increase in intensity with depth below 2 m (Figures 2A, 3A, 4E). These disparate features of AMS and AAR fabrics between the above 2 m and the below 2 m segments of the two cores encourage us to believe that variations of inclination with depth should be treated separately to examine the relationship between development of magnetic fabrics and recording accuracy of inclinations with increasing depth.

Paleomagnetic and rock magnetic analyses of two gravity cores GHE24L and GHE27L from the northern slope of the SCS reveal that these two cores document the detailed changes in inclinations and the gradual development of magnetic anisotropy in a stratigraphic column where compaction is occurring. The results allow us to draw the following conclusions.