The ORCA framework

One reconstruction method. Several classes of subsurface problem.

ORCA is the reconstruction framework we develop and apply. Every problem class below is a variant of the same underlying problem — sparse or missing observations, a complete volume to reconstruct, calibrated uncertainty on the result. The mask geometry differs. The method does not.

The unifying frame

Recognizing the unification matters for how we approach the work. Every improvement to the core method — better training data, more robust architecture, improved uncertainty quantification — carries across to every problem class on the list.

Problem class Observations available What is missing
2D-to-3D reconstruction Sparse 2D seismic lines The entire inter-line volume
Infill 3D volume with holes Platform shadows, drilling obstructions, permit gaps
Obscured-zone imaging Surrounding good data Below gas clouds, beneath salt canopies, through shallow chaos
Footprint removal Mid/far offset data Near offsets at shallow levels

Where we focus today

We don't work on all four classes at the same time. Current development centers on 2D-to-3D reconstruction; the others are research directions we'd take up alongside concrete POC engagements as they arise.

Active focus

2D-to-3D reconstruction

Where current development is concentrated. Turns sparse 2D seismic lines into probabilistic 3D volumes — P10, P50, P90. We've validated the framework on the open-source Penobscot dataset (real 2D + real 3D over the same area) and are looking for POC engagements on comparable basins.

│  ║  │  ║  │  ║  │  ║  │  ║  │
│  ║  │  ║  │  ║  │  ║  │  ║  │
│  ║  │  ║  │  ║  │  ║  │  ║  │
│  ║  │  ║  │  ║  │  ║  │  ║  │
   ↑                          ↑
   2D line             unobserved
Mask geometry: parallel sparse lines.
Read the Penobscot case study →
Active research

Infill

Reconstructs missing zones in existing 3D surveys — platform shadows, drilling obstructions, permit boundaries. Dense surrounding context makes the inversion better-conditioned than 2D-to-3D, with matching constraints at the gap boundaries. No published case study yet; a good fit for a first POC.

█████████████████████████████
█████████░░░░░░░░░░██████████
█████████░  hole  ░██████████
█████████░░░░░░░░░░██████████
█████████████████████████████
Mask geometry: bounded hole in dense data.
Concept →
Future direction

Obscured-zone imaging

Recovers structural geometry distorted by gas clouds, salt canopies, or shallow chaotic zones. The hardest variant — obscuration affects both the masked zone and the surrounding data through wavefield distortion. Research direction, not yet active development.

█████████████████████████████
█████████▓▓▓▓▓▓▓▓▓▓██████████
████▓▓▓▓▓▓▓ gas ▓▓▓▓▓▓███████
█████████▓▓▓▓▓▓▓▓▓▓██████████
░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
Mask geometry: distortion + shadow below.
Concept →
Future direction

Footprint removal

Reconstructs the near-offset gap at shallow depths in some land and marine acquisitions. Uses the NMO equation as a physics-based constraint relating recorded mid/far offsets to the missing near offsets. Research direction, not yet active development.

      near     mid     far
        ╲     │      ╱
         ╲    │     ╱
   shallow╲   │    ╱
          ╲╲╲ │   ╱
   missing→╳╲╲│  ╱  ← available
            ╲╲│ ╱
deep     ────╲│╱──── available
Mask geometry: offset cone, depth-shallow.
Concept →

What we hand back

The deliverable from every ORCA engagement has the same shape: a reconstruction with explicit, quantified uncertainty. RBF interpolation gives one answer. Sparse inversion gives one answer. The ORCA framework gives a distribution of answers — each geologically plausible — with the spread quantifying what the data constrains versus what the prior assumes.

That changes what we put on the table at the end of a POC. Instead of "here is our best estimate of the structure," the deliverable is "here is the range of structures consistent with your data, and here is how additional acquisition would reduce that range."

Regulators care about this too. CCUS site approvals increasingly require explicit uncertainty quantification on structural closure. A single deterministic surface does not answer the regulator's question. A P10–P90 envelope does.