How does seismic inversion improve exploration accuracy?

I. High-level Purpose and Where It Fits in the Value Chain

Seismic inversion converts seismic amplitudes (reflectivity) into quantitative rock-property volumes (acoustic/elastic impedance, Vp/Vs, density, facies probabilities). This directly improves exploration accuracy by turning qualitative images into measurable reservoir attributes.

• I.1 Purpose: Derive subsurface properties that correlate with lithology, porosity, and fluids to reduce interpretation ambiguity and improve prospect risking.
• I.2 Placement: Occurs after seismic processing and before prospect evaluation/volumetrics; integrates tightly with petrophysics, rock physics, and static modeling.
• I.3 Accuracy gains:
  • Transforms amplitudes into calibrated property volumes with uncertainty, enabling objective cutoffs (net pay, facies, fluids).
  • Uses AVO/elastic response to separate lithology vs. fluid effects, limiting false positives from bright spots alone.
  • Delivers spatial continuity of properties between sparse wells, sharpening structural–stratigraphic interpretation and well placement.

II. Step-by-Step Process Flow

• II.1 Seismic preconditioning
  • Denoise, deghost, deconvolve, multiple/peg-leg attenuation, Q-compensation, angle-domain migration.
  • Generate angle/offset stacks (far–near) for AVO integrity; preserve amplitudes via true-amplitude processing.
• II.2 Wavelet and well tie
  • Estimate time-variant wavelet per sector/azimuth; tie synthetic seismograms to seismic; minimize phase/time shift.
  • QC: correlation coefficient, residual phase, stretch/squeeze within tolerances.
• II.3 Low-frequency model (LFM)
  • Build from well logs (Vp, Vs, ?), trends vs. depth/facies, structural horizons; optionally FWI low-wavenumber update.
  • Control non-uniqueness below seismic band; quantify prior uncertainty.
• II.4 Rock physics calibration
  • Establish Vp–Vs–?–porosity–saturation relations; e.g., Gardner: $$\rho=aV_p^b\quad\text{(estimated)}$$
  • Map elastic attributes to facies/fluid probabilities using petrophysical cutoffs.
• II.5 Inversion execution
  • Post-stack impedance inversion: fast AI volumes for initial property maps.
  • Simultaneous pre-stack inversion: angle stacks to recover AI, SI (shear impedance) and density; derive Vp/Vs, ??, µ?.
  • Facies- or geostatistics-constrained inversion: integrate trends and variograms; output property ensembles.
• II.6 QC and uncertainty
  • Blind-well validation, seismic-reconvolution misfit, crossplot separability, variogram consistency.
  • Quantify uncertainty via Monte Carlo/Bayesian posteriors; deliver P10–P90 property volumes.
• II.7 Interpretation and integration
  • Convert property volumes to net-to-gross, porosity, fluid probability; map sweet spots and risks.
  • Feed static models and prospect volumetrics; iterate with updated logs/VSPs as new data acquired.

III. Major Equipment/Components and Their Functions

• III.1 Acquisition systems: to deliver amplitude-preserved, wide-azimuth, broad-band data
  • Marine sources/streamers, ocean-bottom nodes, land nodal arrays; positioning and timing units.
• III.2 Well calibration data:
  • Open/cased-hole logs (sonic monopole/dipole, density, resistivity), checkshots/VSPs, core lab measurements.
• III.3 Processing and inversion stack:
  • HPC clusters/GPUs for migrations, FWI, and simultaneous inversion; seismic/petrophysical software suites.
• III.4 QA/QC toolset:
  • Attribute engines (AVO, spectral decomposition), rock-physics toolkits, uncertainty and geostatistics modules.

IV. Key Performance Drivers (Efficiency, Cost, Safety, Emissions)

• IV.1 Signal quality and bandwidth: Broad frequency (e.g., 2–90 Hz marine), clean multiples, accurate amplitudes; directly controls vertical resolution (˜ ?/4) and property fidelity.
• IV.2 Angle/azimuth coverage: Adequate near–far offsets and multi-azimuth sampling improve inversion stability and fluid/lithology discrimination.
• IV.3 Wavelet and ties: Robust, possibly time-variant wavelets and high R² well ties reduce bias; mis-tied wavelets propagate systematic errors.
• IV.4 Low-frequency model quality: Proper priors with uncertainty bounds mitigate non-uniqueness; FWI can supply low-wavenumber velocity/density trends.
• IV.5 Rock physics realism: Facies-dependent trends, anisotropy and fluid-substitution where justified; prevents over-interpretation of AVO anomalies.
• IV.6 Method selection and regularization: Deterministic vs. Bayesian vs. sparse-spike; tuning smoothness and model covariance to preserve stratigraphic detail without amplifying noise.
• IV.7 Cycle time and cost: Efficient HPC utilization and targeted areas of interest cut runtime; early screening via post-stack before pre-stack reduces spend.
• IV.8 Safety/emissions: Better prospect risking reduces unnecessary appraisal wells (lower HSE exposure and emissions per discovery).

V. Typical Challenges/Bottlenecks and Mitigation

• V.1 Non-uniqueness (band-limited data):
  • Mitigate with strong priors (LFM), Bayesian ensembles, and blind-well validation; incorporate facies constraints.
• V.2 Low-frequency gap:
  • Use checkshots/VSPs and regional trends; leverage FWI for low-k background; propagate LFM uncertainty into results.
• V.3 Wavelet non-stationarity:
  • Estimate sectoral/time-variant wavelets; apply Q-comp and deghosting; re-QC ties by stratigraphic interval.
• V.4 Anisotropy and azimuthal effects:
  • Account for VTI/HTI in processing and inversion; use multi-azimuth data to avoid biased AVO gradients.
• V.5 Illumination and multiples:
  • Adopt OBN/WAZ where needed; advanced multiple prediction; target-oriented migration and angle-gathers for QC.
• V.6 Scale mismatch (logs vs. seismic):
  • Upscale logs with stratigraphic consistency; use thin-bed aware inversion and spectral broadening where justified.
• V.7 Time–depth conversion risk:
  • Constrain velocity models with checkshots/VSP/FWI; propagate depth uncertainty to volumetrics.

VI. Why This Activity Matters Economically or Operationally

• VI.1 Prospect risking uplift: Elastic/facies probabilities cut false positives and high-side bias; typical exploration chance-of-success improvement (estimated): 10–30%.
• VI.2 Better volumetrics and net pay: Property-driven NTG/porosity maps reduce P-uncertainty; P10–P90 narrowing by 15–40% (estimated).
• VI.3 Optimized well placement: Targeting high-probability sweet spots reduces appraisal wells and sidetracks; lateral landing accuracy improvement 20–40% (estimated).
• VI.4 Cycle time and cost avoidance: Early elimination of uneconomic leads saves seismic re-shoots and rig days; fewer dry holes lower capital at risk.
• VI.5 Decision quality: Quantified uncertainty (posteriors) supports disciplined investment gates and reserves classification.