Subsalt imaging: Thunder Horse from Kirchhoff to FWI to QI

The field and the geology

• Location: Mississippi Canyon, Gulf of Mexico, ~200 km SE of New Orleans. Water depth 1850 m. Operated by BP (75%) with ExxonMobil (25%) as partner.
• Reservoir: Lower Miocene turbidite sandstone, deposited by deep-water mass-transport / channel-lobe systems before the salt canopy formed above. Net pay 50-100+ m, porosities 20-26%, light oil with associated gas.
• Trap: structural-stratigraphic. The Miocene reservoir is folded into a four-way closure beneath an irregular salt canopy. Salt geometry creates both the trap geometry AND the imaging headache.
• Salt: an allochthonous (laterally-emplaced) salt canopy ~1500-2500 m thick at the field, with irregular base-of-salt topography including pendants, overhangs, and welds that further complicate imaging.
• Pre-FWI imaging (early 2000s): subsalt section was a chaotic mess. Reservoir geometry uncertain by hundreds of meters. Multiple development wells drilled with significant location uncertainty.
• Post-FWI imaging (2010s): clean reservoir image. Reservoir top mapped within tens of meters. Infill drilling guided by high-confidence depth maps. Production extended for decades.

Exercise — the imaging progression in five stages

• Open the widget in Raw Kirchhoff PSDM mode. You see a coherent overburden (top), a high-amplitude SALT TOP (red-yellow band), the salt interior, and BENEATH THE SALT — a noisy, chaotic, low-coherence zone where the reservoir should be. Notice how the salt body has overhangs (western flank) and a complex base. This is exactly what a 2002-era subsalt processing chain delivered: salt is mappable; reservoir is essentially invisible.
• Switch to TTI tomography velocity. The display now shows the velocity model (cool = slow, warm = fast). Sediments grade from ~1500 m/s near the seafloor to ~3500 m/s at depth. The salt is a uniform 4500 m/s body. SUBSALT SEDIMENTS are still poorly constrained — tomography only updated the velocities along reflections it could pick, and there are few clear subsalt reflections to work with. The model is approximately right but UNCERTAIN where it matters most.
• Switch to FWI velocity model. FWI (§8.4) updates the velocity field to MATCH FULL WAVEFORMS, not just picked reflections. Compare carefully: FWI sharpens velocity contrasts at the base of salt, recovers SUBSALT VELOCITY DETAIL that tomography missed, and resolves small-scale features within the reservoir. This is the velocity model that makes everything downstream possible.
• Switch to RTM image (post-FWI). With the FWI velocity in hand, Reverse-Time Migration produces a CLEAN SUBSALT IMAGE. The reservoir now appears as a coherent reflection package beneath the salt. Salt geometry is sharper. The imaging confidence beneath the salt has gone from ~10% (chaos) to ~80%+ (interpretable). This single processing tier converts the prospect from “undrillable” to “developable.”
• Switch to Inverted Ip cube. With a clean RTM image, run model-based inversion (§7.2). The reservoir appears as a LOW-Ip ANOMALY (cool colors) embedded in higher-Ip surrounding shales — the classic Class III turbidite QI signature you saw in §9.1. NOW you can do volumetric net pay, fluid prediction, and well-placement optimization. The full QI capability that took two days to apply at Girassol takes weeks at Thunder Horse — because the imaging tier alone took years to get right.
• Cycle through all five modes again. Notice the progression: each stage UNLOCKS the next. You CANNOT skip steps. RTM without FWI velocity gives slightly-better-positioned chaos. Inversion on a chaotic image gives chaotic Ip. The whole chain is required.

Why subsalt imaging is so hard

Salt bodies have FOUR properties that conspire against conventional imaging:

• High velocity contrast: salt Vp ≈ 4500 m/s vs surrounding sediment Vp ≈ 2200-3500 m/s. This produces strong refraction at salt boundaries and bends ray paths sharply.
• Irregular geometry: salt canopies have overhangs, pendants, and welds. These create caustics (focusing/defocusing zones) that distort the wavefront entering the subsalt section.
• Multiple generation: the high-impedance salt-sediment interface generates strong multiples that contaminate subsalt primary reflections and are hard to remove.
• Low subsalt illumination: rays bent at the salt boundary may not return to the receivers at all, leaving “shadow zones” with no fold beneath certain salt geometries.

Conventional Kirchhoff PSDM uses a single high-frequency ray-bending approximation that CANNOT handle these effects accurately. The result: subsalt reflectors are positioned wrongly, smeared, or missing entirely. Fixing this required two simultaneous advances: WAVE-EQUATION migration (RTM — reverse-time migration) that handles complex wavefronts properly, AND velocity models accurate enough that the wave equation propagates the energy to the right places (FWI).

Workflow walkthrough — the modern subsalt chain

• Wide-azimuth (WAZ) acquisition: traditional narrow-azimuth streamer geometry illuminates subsalt from one direction only. WAZ uses multiple boats / multi-azimuth tows to ILLUMINATE FROM MULTIPLE ANGLES, dramatically improving subsalt fold. BP’s Thunder Horse WAZ surveys (2004 onward) were a major industry investment.
• Salt boundary picking: the salt body must be defined geometrically before subsalt imaging can succeed. Top-of-salt is straightforward (strong reflector). Base-of-salt is HARD — often requires multiple passes of imaging and interpretation. Salt body is rebuilt iteratively as imaging improves.
• TTI anisotropic tomography: classical tomography updates the velocity model from picked reflections under a tilted-transverse-isotropy (TTI) assumption (§8.3). Establishes the LARGE-SCALE velocity field. Cannot resolve detail beneath the salt where reflections are sparse or absent.
• FWI velocity update (§8.4): full-waveform inversion uses the entire seismic record (not just picks) to refine the velocity model. Modern subsalt FWI runs at progressively higher frequencies (3 → 5 → 10 Hz) using the multi-scale strategy from §8.4. Each frequency band sharpens the model further. Subsalt velocity detail emerges that tomography missed.
• RTM imaging: with the FWI model, run reverse-time migration. RTM solves the wave equation in the subsurface, naturally handling the complex wavefronts that Kirchhoff cannot. Produces a clean subsalt image suitable for interpretation and QI.
• Pre-stack QI workflow (Parts 5-7): on the RTM image, run AVO analysis, model-based inversion to acoustic and elastic impedance, fluid/lithology classification. This is the same QI workflow that runs on a clean conventional image — the only difference is the years of imaging work needed to get to the start line.
• Drill plan optimization: with high-confidence subsalt structure + reservoir attributes, well placement is optimized. Pre-FWI Thunder Horse wells had ±100 m+ uncertainty in reservoir depth; post-FWI, ±10-30 m. This translates to FEWER DRY HOLES, BETTER WELL PRODUCTIVITY, and EXTENDED FIELD LIFE.

Outcomes and lessons

The Thunder Horse imaging story has had cascading impact across the deepwater industry:

• Validated FWI + RTM as the SUBSALT STANDARD. Pre-2010, FWI was a research topic; post-2010, no major subsalt project ships without it.
• Justified the WAZ acquisition investment ( role="main" aria-label="Lesson content" tabindex="-1"00M+ surveys) by demonstrating concrete recovery uplift from improved imaging.
• Established the IMAGING-LED development model: imaging quality determines economic viability of subsalt prospects, not just well productivity.
• Demonstrated that PROCESSING IS NEVER FINISHED. Thunder Horse has been re-processed multiple times as algorithms improve; each re-processing reveals additional reservoir and adjusts well placement strategy.
• Trained a generation of geophysicists in the multi-tier processing mindset: each tier earns the right to apply the next.

The main lesson that transfers to every subsalt or sub-basalt project worldwide: imaging quality is the LIMITING factor for QI. The best QI workflow on a poor image is worse than a basic QI workflow on a clean image. Investment must flow to imaging FIRST.

Where the Thunder Horse story generalizes

• Strong generalization: every Gulf of Mexico subsalt field (Atlantis, Mad Dog, Tahiti, Stampede, Appomattox, etc.); subsalt prospects in offshore Brazil, West Africa, Mexico; Caspian Sea sub-salt sequences. The processing chain is essentially universal.
• Adapted application: SUB-BASALT imaging in the North Atlantic Margin (Faroe-Shetland, Rockall, NE Atlantic). Basalt has even higher velocity contrast than salt. Same workflow philosophy; harder execution.
• Pre-salt application (§9.5 preview): Brazilian pre-salt carbonates beneath thick salt canopies use the SAME imaging chain. Pre-salt discoveries (Lula/Tupi, Búzios, etc.) were made possible by the maturation of subsalt FWI workflows pioneered at Thunder Horse and analogous fields.
• Onshore application: complex thrust belts (Foothills, Andean Subandino, Papua New Guinea) face similar imaging challenges. Same multi-tier mindset applies.
• Where it does NOT apply: shallow conventional plays with simple velocity structure (most onshore conventional, simple offshore plays). Standard PSDM + tomography is sufficient. FWI is overkill.