Joel Sotelo Flores
Active

Ijen Pyroclast Micro-CT and Pore-Network Analysis

Deep-learning segmentation and stereometric analysis of pyroclastic materials from the 1817 Kawah Ijen eruption.

Joel conducts computational volcanology research on pyroclastic micro-CT volumes from the 1817 Kawah Ijen eruption. The work uses U-Net-based deep-learning segmentation, stereometric analysis, Dragonfly, OpenPNM, electron microprobe data, and scanning electron microscopy to investigate vesicularity, pore connectivity, permeability, anisotropy, mineral chemistry, fractionation trends, and pressure-temperature constraints.

ImagingSegmentationStereometryPhysical Interpretation
Grayscale scanning electron microscope image showing vesicles and solid material in a pyroclast sample.
Binary PyRO-FOAMS segmentation mask corresponding to the original SEM image.
Original SEM image and aligned binary segmentation produced with PyRO-FOAMS.

Scientific question

How can pyroclast textures, pore structure, and mineral chemistry be quantified to help reconstruct the eruptive history of the 1817 Kawah Ijen eruption?

Why it matters

The measured textural and chemical characteristics connect to reconstruction of eruptive history for the 1817 Kawah Ijen eruption.

Data and materials

  • Pyroclastic micro-CT volumes
  • Volcanic clasts
  • 2D thin-section imagery
  • Electron microprobe data
  • Scanning electron microscope data
  • Materials associated with the 1817 Kawah Ijen eruption

Methods

  • U-Net-based deep-learning segmentation for 3D pore-space identification
  • Stereometric analysis and automated vesicle analysis
  • Dragonfly for volumetric imaging workflows
  • OpenPNM for pore-network analysis
  • Porosity, permeability, pore-connectivity, and anisotropy analysis
  • Mineral-chemistry integration and fractionation interpretation
  • Pressure-temperature constraint analysis

My contribution

  • Conduct computational volcanology research on pyroclastic micro-CT volumes
  • Train and evaluate U-Net-based deep-learning models for 3D pore-space segmentation
  • Built PyRO-FOAMS, a Python-based stereometric analysis program inspired by Shea et al. (2010)
  • Used Dragonfly and OpenPNM to quantify porosity, permeability, pore connectivity, and anisotropy
  • Integrated electron microprobe and scanning electron microscope data with textural observations

Current results

The work enables quantitative analysis of connected pore networks, porosity, permeability, pore connectivity, anisotropy, vesicularity, and eruption-related textural characteristics.

Research outputs

Associated manuscripts and software are in preparation.