IODP Proceedings    Volume contents     Search


Methods and materials

All data incorporated into the data set originate from the offshore phase of Expedition 346 (shipboard data). Expedition 346 logged four sites in the marginal sea:

  • Site U1423B: 249 m deep hole at 41°41.95′N, 139°4.98′E and 1785 meters below sea level (mbsl);
  • Site U1425B: 407 m deep hole at 39°29.44′N, 134°26.55′E and 1909 mbsl;
  • Site U1427A: 548 m deep hole at 35°57.92′N, 134°26.06′E and 330 mbsl; and
  • Site U1430B; 275 m deep hole at 37°54.16′N, 131°32.25′E and 1072 mbsl.

Data originate from multiple sources, which are included in the IODP standard measurements list (

Downhole log data

As with most IODP expeditions, Expedition 346 shipboard data include a suite of downhole logs generated by two tool strings. The first tool string used in each hole was the Paleo combination (paleo combo), a variation of the triple combination (triple combo) tool string on which the porosity tool (Accelerator Porosity Sonde [APS]) was replaced with the Magnetic Susceptibility Sonde (MSS). The second tool string was the Formation MicroScanner (FMS)-sonic.

Together, these tool strings generate a suite of downhole logs that include information on formation resistivity, natural gamma radiation (NGR), density, magnetic susceptibility, sonic velocity, and borehole diameter (Table T1). The FMS also produces electrical resistivity images of the borehole wall at a 5 mm resolution. For more information, see “Downhole measurements” in the “Methods” chapter (Tada et al., 2015b).

Expedition logging data processed by the Lamont-Doherty Earth Observatory Borehole Research Group were downloaded from the Scientific Ocean Drilling log database ( Processed data are already principally depth matched between logging runs and corrected to seafloor depth.

Only a single instance of each variable was kept, and many accessory variables were removed to simplify the data set and enhance accessibility. This gives a streamlined data set that contains only the most commonly used log variables. The chosen variables universally originate from primary logging runs for both the FMS-sonic and paleo combo tool strings.

FMS images were recreated and saved as borehole image arrays in the Techlog format with the following image processing parameters:

  • 20 m window size for dynamic images,
  • Global button processing method; and
  • Depth-shifted FMS images to match calculated true formation resistivity.

A bit-size variable was also interpolated from unprocessed original data and modified to match info in the Proceedings volume (11716 inches [~11.4375 cm] instead of 9¾ and 11¾ inches recorded in the original DLIS files).

Core data

Even with high-quality downhole log data, recovering core material for groundtruthing and sampling is standard for all IODP expeditions. In the case of Expedition 346, recovering core from the marginal sea was also an essential target to complete to achieve the scientific objectives. Expedition 346 is an exception, however, in volume of core; at 6135.3 m, more core was recovered than during any other single IODP expedition at the time. This significant recovery enables the core and discrete sample physical property data sets to cover almost the complete depth range of each hole.

Downhole log and core data are complementary because core samples provide material for classic sedimentary, petrological, and structural analyses and log data provide measurements that are continuous with depth and under in situ conditions. Further, core and log analyses measure over different scales and capture different structural and petrological controls. For example, because of its length, the Dipole Sonic Imager acoustic velocity sonde on the FMS-sonic tool string (Table T1) has an approximate vertical resolution of 107 cm when sampling at 15 cm intervals. Core-based acoustic velocity measurements can be conducted at the centimeter scale to prioritize matrix characterization over structure. Just as unique data can be acquired downhole, core measurements can also include data that are not always possible to obtain in a downhole environment, such as color spectroscopy. This integration of multiple data types is key to understanding the power of the interdisciplinary data sets recovered by IODP.

The full suite of IODP standard measurements was gathered for the total depth cored at each site. However, for the purposes of this training data set, only a few specific discrete sample data types were included: bulk density, grain density, and porosity. All three data types are recovered from moisture and density (MAD) measurements collected using the helium pycnometry method outlined by Blum (1997). Here, wet mass, dry mass, and dry volume are measured on ~10 cm3 push-core samples to calculate water content and porosity.

When importing IODP core data into a log data set, depth values can be taken as either core depth below seafloor, Method A (CSF-A) or Method B (CSF-B). CSF-A represents distance below seafloor calculated from drilling depth, core depth, and measurement location in the core. CSF-B is identical except that where core recovery is >100%, a compression algorithm is applied to account for core expansion caused by decompression.

During Expedition 346, MAD samples were taken at regular intervals of 1 or 2 per core, usually from Sections 2 and 5 (see “Physical properties” in the “Methods” chapter [Tada et al., 2015b]), providing an average spatial resolution downhole of 1 point per 4.75 m. Discrete samples were not taken for MAD analysis from all holes at a site because of prior sampling for a particular depth/stratigraphic range in a previous hole at that site, assuming that stratigraphy should not vary significantly with 15 m of lateral shift. Where core data logged in a hole at a site were unavailable for this reason, data were loaded into the training data set from the nearest adjacent hole at the same site. On occasion, discrete sample data from two holes were combined to generate a complete data set for the depth interval. This was done for Hole U1423B, where samples for the uppermost 205 m were taken from Hole U1423A (15 m northward), and Hole U1430B, where all samples were taken from Hole U1430A (15 m northward).

Expedition 346 core data were also gathered through the use of “track” core logging systems, which nondestructively acquire data from both whole-round and split-core sections. Gamma ray attenuation (GRA) bulk density, magnetic susceptibility, and NGR data were acquired in this way and are included in the training data set.

Spatial resolution for GRA density and magnetic susceptibility is one measurement every 2, 2.5, or 5 cm along whole-round cores (section dependent), whereas NGR data were acquired at a resolution of 8 measurements per 150 cm core section, generating a mean downhole measurement resolution of 18.75 cm. Because of the measurement technique, track data have a consistent sampling interval downcore. However, data still need to be imported as point data with an inconsistent sampling rate because cleaning the data removes bad measurement points, inconsistent section lengths produce odd numbers, and incomplete core recovery and expanded recovery shifts the true measurement point depths.

The final form of core data included in the training data set comes from core images. Line scan images were taken of all archive-half core sections, which is standard procedure during IODP expeditions. Because of the high resolution of these images and associated high-RAM requirements, only a 50 m interval was taken in Hole U1427A (379.1–430.8 m CSF-A; Cores 58H through 68H). In addition, images were compressed to aid data set functionality on underpowered systems. Images have been renamed with top and bottom depths (CSF-B) in the file name to allow Techlog to assign them to the correct interval.

Additional data sources

Expedition data are extensive but not exhaustive. Additional data were created to supplement the shipboard data and ease the interpretation process. Data were created in the form of additional log variables written into the DLIS files, including bad hole flags calculated from bit size and bulk density correction, a core recovery flag from the IODP section recovery reports (CSF-B), and a shale volume calculation derived from the NGR log. Additionally, an interval.txt variable detailing basin-wide lithostratigraphy identified from shipboard data and duplicate logging runs for training with Techlog-specific tools were created.

Bad hole flags were created to help identify areas where data quality and interpretation may be weakened. Two bad hole variables were created, one from caliper tool diameter and one from bulk density correction. Caliper-based borehole quality flags are generated based on the principle that data quality is reduced where the Hostile Environment Litho-Density Sonde (HLDS) caliper arm reads a hole diameter > 14 inches (35.5 cm) because of an increased proportion of drilling mud (as opposed to formation) being measured. This is key for the density tool (HLDS), which is an eccentralized tool that requires good contact of its pad with the borehole wall (Schlumberger, 2015). Bulk density correction borehole quality flags are generated using density variance between the near and far detectors on the HLDS. A correction is applied to the signal to calculate true formation density. This correction is automated in modern tools and is calculated empirically for different drilling-mud types. The following bulk density corrections are used to create bad hole flags:

  • 0 = bulk density correction_min ≤ bulk density correction ≤ bulk density correction_max,
  • 0 = –0.1 ≤ bulk density correction ≤ 0.1, and
  • 1 = any other outcome.

Core recovery flags written into the DLIS files were generated using section summaries from the IODP Laboratory Information Management Information (LIMS) database. Section reports were collated, and where the section top depth (CSF-B) did not match the bottom depth of the preceding section, a gap in recovery was assumed. Section depth values were collated into simplified CSV files and imported into the DLIS as a flag variable with the same sampling rate as the depth reference: 1 = core, 0 = no core recovery. Section reports provided higher resolution flag variables than core reports (1.5 m instead of ~9 m) and more accurately account for gaps in discrete sample and physical property track data sets.

Shale volume calculations (VSH variables) were generated from NGR logs acquired by the Enhanced Digital Telemetry Cartridge using the following calculation:

GRindex = (GR – GRmatrix)/(GRshale – GRmatrix),


  • GR = gamma ray log (API),
  • GRmatrix = gamma ray log reading 100% matrix rock, and
  • GRshale = gamma ray log reading 100% shale.

The linear method was applied (VSH = GRindex), and VSH was calculated using the same default parameters for all holes: GRmatrix = 10 API and GRshale = 100 API.

Additional stratigraphy data comprises a TXT file that plots expedition lithostratigraphy identified from core as Techlog intervals (Table T2) with top depths and bottom depths for each zone in the borehole.

Additional logging runs were duplicated for training with Techlog’s well-identification solver program. The duplicated logging runs were intentionally misnamed in the LAS header so that they can be assigned correctly as a showcase for the software.