IODP Proceedings    Volume contents     Search



Acquired logging data are presented in a series of data files determined by tool string or sonde, which is standard during all IODP expeditions, and by sampling rate inside Techlog. In the output training data set (Figures F1, F2, F3, F4), data from all sources were compiled and exported in single hole-specific DLIS files in which variables are sorted by a simplified suite of sampling rates. Supplementary data files were separated, and all data were depth matched as best as resolution allowed after down-sampling.

Two primary data sets were created for each hole in a single DLIS file. Both were named “Datafull” to suggest a full suite of log data, one with the suffix “Standard” and another with the suffix “High Res” to indicate the differing sampling rates. As a consequence of this format, some variables were down-sampled from their original sampling rate.

Down-sampled data include the highest resolution outputs from the HLDS (from 0.0254 to 0.0508 m) and magnetic susceptibility data (from 0.0254 to 0.0508 m). A resolution of 0.0508 m was chosen to match the resistivity data resolution. Data set variable content was also modified from the standard IODP processed data set variable suite to aid understanding. A complete list of the variables is in Table T3. Further, data sets were trimmed to new top and bottom depths (Table T4).

Supplementary data files

To make the Expedition 346 training data set a more flexible tool, some supplementary data were separated out from the Datafull DLIS files in a series of file types. Additional data sets comprise

  • FMS image data sets in separate DLIS files,
  • Survey data sets with data for hole inclination, deviation, and azimuth,
  • Physical property track data CSV files containing point data for GRA bulk density, magnetic susceptibility, and NGR,
  • Text files detailing stratigraphy, and,
  • GIF files of core images.

Supplementary data set variables are detailed in Table T5. Training data set file structure was organized by file type to ease the import process (see the “Appendix”).

FMS image data sets were kept separate from the Datafull DLIS files because these images can be considered a less standard data type. In addition, FMS logs contain large volumes of data. Each oriented image consists of 77 variables concatenated, which increases the DLIS file size and reduces usability on underpowered systems. Finally, by separating images from standard data, users can run custom scripts during import.

FMS images were created for full measured intervals in all holes and a duplicated short interval in Hole U1427A. This compressed interval image is designed for operation on underpowered systems and covers the same interval as the GIF images (379.1–430.8 m). In addition to the borehole image array, FMS image DLIS files contain associated tool parameters for dip picking in Techlog. The complete variable list is in Table T5. Images were depth matched to RT_HRLT in the Datafull DLIS data sets.

Further supplementary data sets have been compiled in the form of mimic survey data LAS files. Saved separately from the Datafull DLIS files, survey data are intended to resemble standard industry data format. Survey data comprise azimuth and inclination data exported from the General Purpose Inclinometry Tool on the FMS-sonic tool string. These variables cover the entire interval of the drilled hole and were trimmed to the same depths as the Datafull DLIS files.

Track data CSV files containing GRA bulk density, magnetic susceptibility, and NGR data have been depth matched to Datafull log data sets. Occasionally, these physical property data are not aligned with core recovery flags because the track data originate from a different hole than the core flag. See “Core data” for details of where this may occur.

Intentional errors were incorporated into the data set, which is common in training data sets. One such example is a depth mismatch between the Datafull Standard and the Datafull High Res data sets in Hole U1430B. The high-resolution data set was linearly shifted 4 m downhole. Another example is the repetition of data in “RUN_1, 2, 3, 4” LAS files designed for use in well-renaming exercises.