Seismic Modeling
Seismic modeling is the computational process of generating synthetic seismic data from an assumed or known subsurface earth model, used to predict what recorded seismic data should look like in order to calibrate well-to-seismic ties, validate amplitude interpretations, support AVO analysis, optimize survey acquisition design, and quantify uncertainty in exploration and development decisions; methods range from the simple 1D convolutional model (reflectivity series convolved with a source wavelet) to full 3D finite-difference solutions of the elastic wave equation.
Key Takeaways
- The 1D convolutional model generates a synthetic seismogram by convolving an acoustic impedance-derived reflectivity series with an estimated source wavelet, providing a fast well-to-seismic tie calibration tool.
- Full-waveform modeling (finite-difference or finite-element methods) solves the elastic or acoustic wave equation numerically, capturing multiples, diffractions, mode conversions, and geometrically complex propagation effects that the convolutional model ignores.
- Ray-tracing modeling offers a computationally efficient intermediate approach, tracking wavefront geometry through a velocity model to compute traveltimes and approximate amplitudes for complex structural interpretation.
- Seismic modeling underpins AVO analysis by predicting how reflection amplitude should vary with offset for different fluid and lithology combinations, allowing interpreters to distinguish gas sands from brine sands or shales.
- Survey design modeling (also called seismic survey simulation) uses synthetic data generation to evaluate whether a proposed acquisition geometry will image target reflectors with sufficient resolution and signal-to-noise ratio before committing to field acquisition costs.
Fast Facts
A single 1D synthetic seismogram can be generated in milliseconds on a desktop workstation. A full 3D elastic finite-difference model of a typical exploration block at 20 Hz dominant frequency requires hundreds of CPU-hours or GPU-accelerated computing. The convolutional model was formalized by Robinson and Treitel in the 1950s-1960s and remains the most widely used well-tie tool in the industry. Full-waveform inversion (FWI) is the inverse application of seismic modeling: observed data are iteratively matched to model predictions to derive velocity and density models.
Tip: When calibrating a synthetic seismogram to real seismic data, use a statistical wavelet extracted directly from the seismic data near the well rather than a theoretical zero-phase Ricker wavelet; the extracted wavelet captures phase rotations and bandwidth limitations specific to the acquisition and processing sequence, producing a much tighter tie between synthetic and real data.
What Is Seismic Modeling?
Seismic modeling is fundamentally the forward problem of geophysics: given a description of the earth (velocity, density, and elastic properties as a function of position), compute what seismic waves would look like after traveling through that earth and being recorded at surface receivers. This synthetic data can then be compared to real recorded seismic data to assess how well the earth model matches reality, diagnose interpretation ambiguities, or plan new data acquisition.
The simplest and most common form, the 1D convolutional model, assumes that seismic waves travel vertically and that each subsurface layer produces a reflection proportional to the acoustic impedance contrast at its boundaries. The resulting reflectivity series is convolved mathematically with the seismic source wavelet to produce a synthetic trace that looks visually like a real seismic section and can be compared side-by-side with the recorded data at a well location. This process, called well-to-seismic tie or calibration, is the first step in virtually every seismic interpretation project.
How Seismic Modeling Works
For the 1D convolutional approach, the earth model is built from well log data: compressional velocity (Vp) and bulk density (RHOB) logs are multiplied to compute acoustic impedance as a function of depth. Depth is converted to two-way time using a velocity-depth relationship derived from check-shot surveys or VSP data. The impedance contrasts at layer boundaries are computed as reflection coefficients: RC = (Z2 - Z1) / (Z2 + Z1). This series of reflection coefficients, convolved with an estimated or extracted wavelet in the time domain, produces the synthetic seismogram. Phase and polarity conventions must match the recorded data processing sequence, which is where systematic calibration errors most often arise.
More physically complete modeling uses numerical solutions to the wave equation. Finite-difference time-domain (FDTD) methods discretize the subsurface into a regular grid and propagate wavefields forward in time step by step, applying the elastic wave equations at every grid point. These methods capture all wave types including P-waves, S-waves, surface waves, and multiples, making them the standard tool for full-waveform inversion and complex structural illumination studies. The computational cost scales as the fourth power of the dominant frequency-to-grid-spacing ratio, which is why these methods are expensive at high frequencies.
Ray-tracing modeling provides an efficient alternative for computing traveltimes and approximate amplitudes in 2D and 3D velocity models. Ray tracing tracks the path of individual ray bundles through the velocity model using Snell's law at each interface, computing two-way traveltime, geometrical spreading, and reflector illumination for each source-receiver pair. It is commonly used for migration velocity analysis, normal moveout correction assessment, and survey planning.
AVO (amplitude variation with offset) modeling extends the 1D synthetic to the pre-stack domain. The Zoeppritz equations or their linear Aki-Richards approximations predict how reflection amplitude varies with the angle of incidence for a given contrast in Vp, Vs, and density. Modeling AVO responses for different fluid scenarios (gas, brine, oil) in a target reservoir allows the interpreter to define an expected AVO class and amplitude signature for the prospect, which is then tested against the actual pre-stack seismic data to evaluate DHI (direct hydrocarbon indicator) credibility.
Seismic Modeling Across International Jurisdictions
In Canada, seismic modeling is central to WCSB exploration and development workflows. The AER requires that well completion and stimulation programs in tight gas and tight oil plays be supported by reservoir characterization studies that routinely include synthetic seismogram generation and well-to-seismic tie documentation. Geoscience teams at operators including Canadian Natural Resources, Cenovus, and ConocoPhillips Canada use 3D full-waveform modeling for detailed Montney and Duvernay reservoir characterization, particularly for detecting fracture-prone zones and optimizing horizontal well placement relative to seismic amplitude anomalies.
In the United States, seismic modeling supports exploration decisions from the Permian Basin to the Gulf of Mexico deepwater. The BOEM requires environmental impact assessments for offshore seismic surveys, and modeling is used to design surveys that achieve imaging objectives with minimum air gun array output. Deepwater GOM exploration relies heavily on full-waveform inversion and finite-difference modeling to handle complex salt geometry, sub-salt illumination, and anisotropy effects that make conventional ray-based processing inadequate for high-confidence amplitude interpretation.
In Norway, the Norwegian Petroleum Directorate (NPD) and Sodir maintain the DISKOS national data repository, which includes processed seismic surveys and well synthetic seismograms for all NCS wells with formal data submission requirements. Equinor, TotalEnergies Norway, and Aker BP routinely use rock physics-constrained seismic modeling for 4D (time-lapse) seismic interpretation at producing fields such as Sleipner, Gullfaks, and Troll, where repeated seismic surveys track fluid substitution and pressure changes during production for reservoir management decisions.
In the Middle East, Saudi Aramco employs some of the world's largest seismic modeling computing infrastructures for detailed characterization of the complex Arab-D carbonate reservoir system in the Ghawar and Safaniya fields. Anisotropic full-waveform modeling is used for fracture characterization in tight gas fields and for planning advanced acquisition geometries in structurally complex areas of the Rub' al Khali basin. ADNOC similarly uses seismic modeling at scale for development planning of offshore Abu Dhabi carbonate and clastic reservoirs, integrating seismic with reservoir simulation models in a coupled workflow.
Synonyms and Related Terminology
Seismic modeling is also called forward seismic modeling, synthetic seismogram generation, or seismic simulation. The inverse process is called seismic inversion or full-waveform inversion (FWI). Related terms include synthetic seismogram, AVO analysis, reflectivity series, acoustic impedance, seismic wavelet, and well-to-seismic tie.
FAQ
Q: What is the difference between seismic modeling and seismic inversion?
A: Seismic modeling (the forward problem) computes synthetic seismic data from a known earth model. Seismic inversion (the inverse problem) works in the opposite direction: it uses recorded seismic data to estimate the earth model, most commonly the acoustic or elastic impedance distribution. Modeling is used to validate and calibrate interpretations; inversion is used to extract quantitative rock and fluid properties from the seismic data.
Q: How accurate is a 1D synthetic seismogram compared to real seismic data?
A: A well-calibrated 1D synthetic typically achieves a cross-correlation of 0.7-0.9 with the real seismic data in a structurally simple, laterally homogeneous setting. In structurally complex areas with strong lateral velocity variation, anisotropy, or significant multiples, the 1D model breaks down and 2D or 3D finite-difference modeling is required to explain the observed data accurately.
Why Seismic Modeling Matters
Seismic modeling is indispensable at every stage of the exploration and development lifecycle. During exploration, it determines whether a proposed acquisition program can resolve the target structure and detect the expected amplitude anomaly at the risk-reducing cost-per-well economics demanded by modern petroleum geoscience. During appraisal, well-to-seismic tie and AVO modeling translate wireline log rock physics into seismic observables, providing the evidentiary link between well data and the broader 3D seismic volume. In development and production, 4D modeling monitors reservoir fluid changes and pressure support, directly informing well placement and injection optimization decisions worth hundreds of millions of dollars annually.