This application facilitates the analysis of radar data for polar geophysicists by using Extended Reality (XR) technology to visualize radar returns in an accurate 3D geospatial context.

Project Website: https://pgg.ldeo.columbia.edu/projects/VISER

Note: This project is still in development.

Last Updated 10/31/2024.

Developers:

• Qazi Ashikin, Columbia University
• Anusha Lavanuru, Columbia University
• Leah Kim, Barnard College
• Moises Mata, Columbia University
• Linda Mukarakate, Barnard College
• Greg Ou, Columbia University
• Joel Salzman, Columbia University
• Ben Yang, Columbia University

Additional Contributors:

• Isabel Cordero, Lamont-Doherty Earth Observatory
• Andrew Hoffman, Lamont-Doherty Earth Observatory
• Carmine Elvezio, Columbia University
• Bettina Schlager, Columbia University

Advisors:

• Dr. Robin Bell, Lamont-Doherty Earth Observatory
• Dr. Alexandra Boghosian, Lamont-Doherty Earth Observatory
• Dr. Kirsty Tinto, Lamont-Doherty Earth Observatory
• Professor Steven Feiner, Columbia University

Ice-penetrating radar is a powerful tool to study glaciological processes but the data are often difficult to interpret. These radar measurements are taken by flying a plane equipped with sensors along some path (known as the flightline) and plotting the radar returns. These plots (radargrams) must be interpreted, often manually, in order to distinguish features inside the ice. Glaciologists are intensely interested in identifying the ice surface, bedrock depth, and whatever features can be discerned in the subsurface. An essential step in this process is "picking lines," or identifying contiguous curves in radargrams that correspond to real features in the world (as opposed to noise). Current methods are suboptimal.

VISER aims to improve interpretability of radargrams using XR. First, we place the user inside a digital twin of the relevant ice shelf so that all the data are properly contextualized relative to each other in space. Second, we model the radargrams as Unity GameObjects so that the features visible on the plots (which appear as textures on the GameObjects) also appear in the proper 3D geospatial context. Third, we implement numerous interfaces for analysis and manipulation so that users can explore the data.

The application contains three scenes.

• Home Menu
• Ross Ice Shelf (RIS)
• Petermann Glacier (PG or Petermann)

The Home Menu scene is mostly empty except for an XR menu that allows users to load one of the other scenes.

The RIS scene uses the following assets.

In this scene, radar images are textured on planar GameObjects. Only plots from the portions of the flightline where the plane was flying relatively straight are shown. This scene has the most mature navigation system and includes a minimap.

The flightlines are loaded from CSV files as discrete points.

The Petermann scene uses the following assets.

In this scene, the flightlines and radargrams are generated at runtime as GameObjects rather than living permanently in the scene. The flightlines are rendered as polylines and are broken up into segments during preprocessing. Each radar object is modeled as a triangle mesh, textured with a radargram, and linked to the associated flightline portion. Using meshes enables the entire flightline to be displayed. The flightline coordinates are accurate within the projected coordinate system but the vertical coordinate is snapped to the surface DEM for ease of viewing.

The radargram objects are generated in the following way. In preprocessing, the plots are generated in MATLAB and converted into the three separate files by the greenland_obj.m script.

• .obj (contains the geometry of the mesh)
• .mtl (the material file for the mesh)
• .png (the radargram; this is mapped onto the mesh)

The .obj files then need to be decimated in order to improve performance. This is done in Blender and currently requires manual attention to ensure that the meshes do not deform significantly at the boundaries. These simplified .obj files, along with the .mtl and .png files, are added to the Assets folder. Upon loading the scene, the LoadFlightlines.cs script programmatically generates meshes from the file triples and associates each textured mesh with the corresponding flightline polyline. Users can select a flightline portion to load a mesh.

VISER works on both AR (Hololens) and VR (Quest) headsets. Since the most recent version is optimized for VR, we omit the AR controls in this section.

Here is a list of interactions available everywhere using the Oculus Controllers:

Movement can be accomplished with the joysticks. Tilt forward to shoot a ray; if the ray is white, release the joystick to teleport there. You can also nudge yourself back by tilting back on the joystick.

The scene bounding box can be used to scale everything inside the scene along any of the three axes. Users can grab the bounding box from the corners or the center of any edge and use standard MRTK manipulation to adjust the box's size.

Here is a list of interactions available with the main menu. All of these interactions can be used with the Oculus Controller trigger buttons:

Because the Ross Ice Shelf scene uses a different workflow than the Petermann Glacier scene, some of the controls are different. Here are the menu options that are only available in the RIS scene.

Here is a list of interactions available with the line menu. All of these interactions can be used with the Oculus Controller trigger buttons:

The line menu has a unique sidebar.