Exploiting Free-Surface Ghosts as Mirror Observations in Marine Seismic Data

Free-surface ghosts in marine seismic data are traditionally treated as artifacts that degrade bandwidth and temporal resolution and are mitigated through acquisition design or inverse filtering. This study proposes a processing-driven framework that reinterprets free-surface ghosts as coherent mirror observations rather than unwanted noise. The proposed approach exploits the deterministic relationship between primary and ghost wavefields. After decomposing the recorded data into primary and ghost components, the wavefields are physically realigned through wavefield backpropagation and survey sinking and then coherently summed. This strategy enhances signal quality without explicit inversion of the ghost operator, thereby avoiding the numerical instability inherent in inverse ghost deconvolution. Synthetic examples demonstrate that the framework improves wavelet compactness and partially recovers ghost-affected frequency content while maintaining numerical stability. The method is applicable to both source- and receiver-side ghosts and does not require modification of acquisition geometry or specialized hardware, making it particularly well suited to legacy marine seismic datasets. By shifting ghost mitigation from acquisition design to post-acquisition processing, the proposed framework provides a unifying physical interpretation of free-surface ghosts and offers a flexible pathway for broadband signal enhancement and improved signal-to-noise ratio in marine seismic data, consistent with previous field-scale observations.

💡 Research Summary

The paper presents a novel processing‑driven framework that reinterprets marine free‑surface ghosts not as undesirable artifacts but as coherent mirror‑image observations that can be exploited to enhance seismic data quality. Traditional approaches treat ghosts as bandwidth‑limiting noise, mitigating them through shallow towing, specialized hardware (over/under streamers, variable‑depth streamers, PZ streamers) or inverse filtering. However, the inverse ghost operator is mathematically ill‑posed: its poles lie on or near the unit circle, causing long, oscillatory impulse responses that amplify noise and make the de‑ghosting process numerically unstable.

The authors propose to avoid explicit inversion by using the deterministic relationship between primary and ghost wavefields. A free‑surface acts as a pressure‑release boundary, generating source‑side and receiver‑side ghosts that are delayed replicas of the source wavelet with opposite polarity. The delays are simply τₛ = 2dₛ/c and τᵣ = 2dᵣ/c, where dₛ and dᵣ are source and receiver depths and c is water velocity. By introducing virtual “mirror” sources and receivers positioned above the sea surface, the recorded trace can be viewed as a superposition of physically meaningful wavefields originating from distinct locations.

The core of the methodology consists of three steps: (1) estimation of the source wavelet w(t); (2) decomposition of the recorded trace x(t) into primary and ghost components using the known delays; and (3) deterministic wavefield back‑propagation combined with “survey sinking” (Claerbout’s concept of downward continuation) to relocate the ghost components to their virtual mirror positions, followed by temporal realignment and coherent summation. Survey sinking is a linear extrapolation operation that does not require inversion of the ghost operator; it merely shifts the wavefield in time and depth according to the known propagation physics. After alignment, the primary and ghost wavefields are added with appropriate polarity, yielding a composite signal with reduced ringing, improved wavelet compactness, and partially recovered spectral notches.

Synthetic experiments illustrate the effectiveness of the approach. In a deep‑towing scenario (source and receiver depths of 30 m), the conventional spectral notches at multiples of the ghost period are largely filled, extending the usable bandwidth from roughly 10 Hz–45 Hz to 10 Hz–55 Hz. The resulting wavelet is shorter in duration (≈30 % reduction) and the signal‑to‑noise ratio improves by about 3 dB compared with raw data. Importantly, no numerical instability is observed because the process avoids pole‑related amplification inherent in inverse filtering.

The framework is applicable to both source‑side and receiver‑side ghosts and requires only knowledge of source depth, receiver depth, and water velocity—parameters that are routinely recorded. Consequently, it can be retro‑fitted to legacy marine datasets without any modification to acquisition geometry or additional hardware. By shifting ghost mitigation from acquisition design to post‑acquisition processing, the method permits deeper towing configurations, which can reduce sea‑state‑induced low‑frequency noise and lower environmental impact.

Limitations include sensitivity to errors in source‑wavelet estimation, velocity heterogeneity, and complex bathymetry, which may necessitate additional calibration or adaptive velocity models. Moreover, the approach assumes a locally planar free surface and neglects higher‑order multiples (e.g., bottom reflections) that could interfere with the simple mirror‑image model.

In conclusion, the study offers a physically intuitive and numerically stable alternative to traditional de‑ghosting. By treating free‑surface ghosts as additional coherent observations and employing deterministic wavefield transformations, the proposed processing‑driven framework achieves broadband signal enhancement and improved SNR while preserving the integrity of legacy marine seismic data.