X-ray CT images on y-z and z-x planes as defined in Fig. 3(a)x are shown in Fig. 3(b), although image reconstructions in any directions are possible after the data acquisition. In Fig. 3(b)x, X-ray CT-generated grayscale images were constructed using histogram equalization with the lower limit value of water, which provides contrastive images for structure observations. Left and right sides of Fig. 3(b) show cross sections on y-z and z-x planes every 15 and 12 units of in-plane resolution (0.313 mm), respectively. The core is not perfectly aligned to the center of stage of X-ray CT scanner; therefore the core appears tilted in the images. Two holes at 36.1 and 74.0 cm, due to shear strengths measurements with a hand held torvane sampler on board, are clearly visible in both y-z and z-x planes. Curved laminar structures possibly due to drilling and sampling processes, are observed near the linear. Significant differences in CT numbers among sediments, IRD (some examples are highlighted by white circles on y = 72 plane in Fig. 3(b)), and pebble at 60 cm are represented by grayscale. Although CT numbers of sediments are similar (clearly shown in Fig. 2), laminae appear in CT number variations by the enhanced images as shown in Fig. 3(b), implying that these structures retain density contrasts. Figure 3(c) shows the CT number variations along the black line on y = 72 plane in Fig. 3(b). CT number difference between peak and trough values are approximately 100–200 as in Fig. 3(c), and it corresponds to 0.1–0.2 g/cm3 changes in density, large enough to discriminate density contrasts. Using the core sedimentation rate of 7.1 cm/ka (Titschack et al., 2009), CT number variations are assessed with temporal resolution of the order often years.
Figure 4 combines (a) X-ray radiograph, (b) X-ray CT image, and (c) line-scan digital core images of the sample with a total length about 50 cm. Note that the “archive-half” was described visually and photographed. X-ray CT and X-ray radiograph was performed using the “working-half”. In Fig. 4(b), y = 72 plane was chosen, which almost corresponds to about 14 mm depth from the core surface, avoiding data gaps at 36.1 and 74.0 cm.
(a) X-ray radiogram, (b) X-ray CT image on y = 72 plane, (c) line-scan digital core image enhanced by histogram equalization, of half-round core sample from Section 307-U1318A-5H4, 22–72 cm. Note that X-ray CT image and X-ray radiogram are mirrored images to correspond to the line-scan digital core image. Scale attached (c) is used for locations mentioned in the text.
Three kinds of images as shown in Fig. 4 are compared with each other to highlight their characteristics, advantages, and limitations. Line-scan digital core images provide detailed visual information on the surface. X-ray radiograph system produces 2-D images of internal 3-D structures, but in a single 2-D shadow projection the depth information is completely mixed. X-ray radiographs reveal laminae on a millimeter scale more clearly but nonquantitatively, and highlight many IRDs as bright spots against dark-colored background sediments, partly because they reflect an integration of information over the slab thickness of 7 mm. On the other hand, X-ray CT system allows us to visualise and measure complete 3-D object structures without sample preparation, and X-ray CT image can be sliced in any sections and reveal sedimentary and post sedimentation structures and materials that cannot be seen on the split core surface. However, because of the lower resolution of X-ray CT images, smaller targets may not be imaged.
Variations in grayscale values of the line-scan digital core image, X-ray CT image and X-ray radiograph reflect laminae. Based on the line-scan digital core image, laminae curved downwards due to drilling disturbance close to liners were frequently observed, which are easily recognized by X-ray CT image. The X-ray CT image (Fig. 4(b)) also captures disturbance of laminae near the pebble at 60 cm. This pebble is clearly captured in the X-ray CT images but cannot be seen on the split core surface. However, disturbances of original sedimentary structures at corresponding range are evidenced in the line-scan digital core image as shown in Fig. 4(c).
Figure 5 shows variations in CT numbers taken along the core axis, with GRA bulk densities on board and laboratorymeasured densities as described in Section 2.4. Densities obtained from GRA measurements and mass/volume methods are not well correlated with each other. This difference may come from the different sample area and volume used for measuring densities: The GRA density measurement provides an estimate integrated over an irradiated cylindrical volume with a diameter of 66 mm and a width of ∼5mm of the whole core sample (Boyce, 1976) and densities were determined by the mass/volume method on discrete cube samples (∼7 cm3) that were cut from split core sections. The GRA bulk densities are about 0.24 g/cm3 higher on average than those of the mass/volume method. One of the plausible reasons is that cores gradually dried after coring, even though the core sample was contained inside plastic liner. The GRA bulk densities show no significant change throughout the 50-cm long core sample, however the densities obtained from mass/volume method show a wider range than those of GRA measurements. It may be related that samplings in the mass/volume method are frequently biased due to uneven distribution of materials and samples do not represent the exact average at the position.
CT numbers (gray line), GRA gamma densities (crosses), mass/volume-derived densities (circles), and MAD density (square), for Section 307-U 1318A-5H4, 22–72 cm. Gray solid, dashed, and dotted lines represent profiles of CT number at x = 150, 160, and 170, respectively. These profiles are obtained using average of 5 profiles from y = 80, 90, 100, 110, and 120. The solid line represents a moving average of 10 CT numbers for the core section depth using 15 profiles.
Regions near the center of core excluding the pebble at 60 cm and two data gaps at 36.1 cm and 74.0 cm are chosen to show CT number variations in Fig. 5. CT numbers in each voxel of reconstructed CT image are plotted as a function of core section depth. Each average value at x = 150, 160, and 170, using 5 data at y = 80, 90, 100, 110, 120 is used to exclude outliers and to reflect a larger spatial area, as shown in gray lines in Fig. 5. Although the direction of core section depth is not exactly parallel to sediment layer stratification, each CT number variations and three average values show similar trends with peaks in phase. As mentioned before, there might be expected a sort of relation between variations of CT number and densities. Gray lines in Fig. 5 show density variations corresponding to distinct sedimentary sequence, which clearly shown in different colours in Fig. 4. Figure 5 also shows a moving average of 10 CT numbers for the core section depth using 15 profiles at (x, y) = ([150, 160, 170], [80, 90, 100, 110, 120]), to capture long-term trends.
Two GRA density peaks at depths of 35 and 60 cm are shown in Fig. 5, although GRA density changes are relatively small all through this sample. The peak at 35 cm is also recognized at the similar depth in the line-scan digital core image and X-ray radiograph as a bright-colored area in Figs. 4(a) and (c), which corresponds to higher CT number. At this cross section depth, density by mass/volume method is rather high. The GRA density has the second highest value at 60 cm through this sample, where averaged CT number also has a peak at this depth. On the other hand, CT number shows a high peak at 64 cm. This part coincides with a concentration of a white fine-sand layer, only 5–10 mm thick in the line-scan digital core image, X-ray CT image and X-ray radiograph as shown in Fig. 4. However, densities obtained from GRA measurements and mass/volume methods do not correspond to this high CT number. The peak at 64 cm may be out of the target area of GRA density measurements collected at 5 cm intervals with the measurement width of ∼5 mm.
Several attempts have been made to convert CT numbers to equivalent sediment bulk density and have verified the linear correlation between bulk density and X-ray attenuation (e.g., Orsi et al., 1994), although there are many uncontrollable variables, including machine-dependent parameters and statistical parameters of samples such as mean value and standard deviation of CT numbers. Previous calibration experiments on the Technicare Δ-100 CT scanner at 120 kV/25 mA using marine sediments from the northern Gulf of Mexico continental shelf, showed a linear response between bulk density (ρ) and CT number (HU) described by 
(Orsi et al., 1994). They estimated the average density resolution of the system much smaller than 0.01 g/cm3 with the bulk density range from 1.96 to 2.08 g/cm3. A calibration procedure by Orsi and Anderson (1999) using two types of material showed the difference of gradient between the curve for SiO2 and CaCO3 CT number versus bulk density plots, namely 
and 
over a range of 1.0 and 2.2 g/cm3. Inazaki et al. (1995) indicated that the accuracy of bulk density measurements estimated from the CT number is better than ±0.02 g/cm3 by calibration with known standards using medical X-ray CT system (TCT-700S) at 120 kV/55–200 mA. Their suggested linear regression is 
, using water glasses and pressed bentonites with bulk density ranges between 1.10 and 1.95 g/cm3.
Figure 6 shows a relationship between CT numbers and measured densities by the GRA and the mass/volume method of host sediments. Densities obtained by the mass/volume method show smaller values than those from the GRA measurements. The density of a pebble by the water saturation method was also plotted. Small variations in CT number for the pebble of sedimentary origin may reflect that pebble material is rather homogeneous. Densities by the GRA measurements of host sediments and pebble fall roughly on the same trend using silica sediments by Orsi and Anderson (1999). Based on CT number and density data obtained from the GRA measurements and pebble, CT number has a linear relationship with density: 
, over a range of 2.0 and 2.7 g/cm3. Although data are not evenly distributed over this density range, a medical X-ray CT system respond a linear relationship in density up to 2.7 g/cm3. However, this obtained linear relation represents a different offset from that shown by silica sediments (Orsi and Anderson, 1999). Further developments of the relationship between CT number and density are required to conduct more quantitative analysis.
Scatterplot illustrating the relationship between CT number and wet bulk density. Open triangles and squares show all the data points for densities obtained from the GRA measurement and the mass/volume method, respectively. The accompanying horizontal bars represent CT number distribution in the corresponding area to be used for density measurements. Solid gray triangle and square indicate average densities of host sediments by the GRA measurements and the mass/volume method, respectively. The solid circle indicates the density of a pebble by the water saturation method with the measurement error. The horizontal bar of the solid circle shows the full-width at half maximum (FWHM) of CT number distribution as a proxy for estimating data range. Thick gray dashed and dotted lines were determined by Inazaki et al. (1995) and Orsi et al. (1994). Thick gray solid and one-dotted lines were determined by least-squares analysis for silica sediments and carbonate samples by Orsi and Anderson (1999). Dashed line was determined in this study.
Standard conventional techniques for bulk density analysis in sediment cores use raw data from GRA or mass/volume method. These density values obtained from samples represent the gross average. Neither have a spatial resolution better than X-ray CT data-derived density. Moreover, X-ray CT data-derived density offers advantages over these standard conventional techniques as it gives a threedimensional distribution anywhere in the sample. Also, because of the simple process of the conversion from CT number to density, it is expected to reduce the possibility of human error.