|
<< Click to Display Table of Contents >> Navigation: Modeling > Finite difference 2D/3D modelling |
The Finite Difference 2D/3D Modelling module generates synthetic seismic data by numerically solving the acoustic wave equation on a velocity model. It propagates a user-supplied wavelet through the model using an explicit finite-difference time-stepping scheme and records the resulting wavefield at prescribed receiver locations. The module supports both 2D and 3D velocity models and automatically detects the dimensionality from the input.
Three modelling modes are available. Exploding reflector initialises the wavefield at every reflector in the model simultaneously and propagates it upward, producing a zero-offset synthetic section without ray tracing. Common-shot gather (CS gather) places a point source at the centre of the model and records the full pre-stack wavefield at all receiver positions, generating an offset-dependent synthetic gather suitable for AVO studies and migration testing. Post-stack RTM back-propagates an existing stacked section through the velocity model using reverse-time migration, converting post-stack time data into a depth-domain image.
Before running the module, use the Suggest grid step based on frequency, Suggest max time step, and Suggest min grid step actions in the Custom Actions section to obtain numerically stable parameter values. Running the modeller with unstable parameters will produce an error and no output.
Connect the depth-domain velocity gather that defines the acoustic properties of the subsurface. This item must contain at least five traces in both inline and crossline directions for a 3D model, or at least five traces in one direction for a 2D model. The module reads interval velocities in metres per second directly from the gather samples. The spatial extent of the model determines the size of the modelling grid. The velocity model must use depth as the vertical axis.
Selects how receiver positions are specified for the output synthetic gather. Currently the only option is Trace headers, which takes receiver coordinates from the connected Output geometry trace vector item. This parameter is visible only in Exploding reflector and CS gather modes; it is hidden in Post-stack RTM mode.
Connect a trace vector item that defines the source and receiver positions for the output synthetic gather. The modeller records the simulated wavefield at each receiver location listed in this item. If no trace vector is connected in Exploding reflector or CS gather mode, receiver positions are derived automatically from the trace geometry embedded in the velocity model gather.
Sets the total recording duration of the output synthetic gather, in seconds. Default is 4 s. Increase this value to capture reflections from deeper targets or wide-angle arrivals. Longer record lengths increase computation time. This parameter is visible in Exploding reflector and CS gather modes only.
Sets the sample interval of the output synthetic gather, in seconds. Default is 0.004 s (4 ms). The output is resampled from the fine internal modelling time step to this coarser sample interval, so this value may be larger than the Modelling step. This parameter is visible in Exploding reflector and CS gather modes only.
Controls where the modelled synthetic gather is delivered. The only active option is Gather, which sends the result directly to the Output gather data connector for display and downstream processing.
Connect a post-stack time-domain seismic section to use as the source wavefield for reverse-time migration. This input is only active in Post-stack RTM mode and is hidden in Exploding reflector and CS gather modes. The section is back-propagated through the velocity model to produce the depth image.
Connect a single-trace gather containing the source wavelet to be injected into the model at the start of the simulation. This wavelet defines the frequency content and shape of the modelled reflections. Use a Ricker or band-limited spike wavelet for standard modelling. The wavelet gather should have a single trace; only the first trace is used. This input is required in Exploding reflector and CS gather modes.
Selects the wave-propagation mode. Default is Exploding reflector.
Exploding reflector — computes a zero-offset synthetic section by treating every impedance contrast in the model as a secondary source that fires simultaneously at time zero. The wavelet is convolved with the reflectivity derived from the velocity model. This mode runs the fastest and is ideal for validating a velocity model or creating a zero-offset synthetic for well tie.
CS gather — places a point explosive source at the horizontal centre of the model and records the full offset-dependent wavefield at all receiver positions. Use this mode to generate pre-stack synthetic data for AVO analysis, offset-dependent amplitude studies, or to test migration algorithms.
Post-stack RTM — back-propagates the connected input stack through the velocity model using finite-difference reverse-time migration. The output is a depth image. An additional time-stretched version of the depth image is also produced in the Output gather converted connector.
Sets the internal time step used by the finite-difference solver, in seconds. Default is 0.0005 s (0.5 ms). This value must satisfy the Courant-Friedrichs-Lewy (CFL) stability criterion: the product of maximum velocity and time step divided by grid step must be below a threshold that depends on the modelling order. If the chosen value violates this criterion the module will report an error and refuse to run. Use the Suggest max time step action to obtain the largest stable value for the current velocity model and grid step, and then reduce it slightly as a safety margin.
Sets the spatial grid step used internally by the finite-difference solver in all directions (horizontal and vertical), in metres. Default is 10 m. The velocity model is resampled to this grid spacing before modelling begins. Smaller values increase accuracy and resolve higher spatial frequencies but increase both memory usage and computation time. To avoid numerical dispersion the grid step should be at most one-fifth to one-tenth of the shortest wavelength in the model (minimum velocity divided by maximum frequency). Use the Suggest min grid step and Suggest grid step based on frequency actions to obtain recommended values.
An optional group of parameters that applies a trapezoidal bandpass filter to the simulated wavefield before it is recorded. Enable the filter by checking Apply bandpass; the four frequency corner and epsilon fields then become editable.
When checked, a Butterworth bandpass filter is applied to the modelled wavefield using the four frequency corners defined below. Default is off. Enable this option when the modelled data contains unwanted low- or high-frequency numerical noise, or when the output must match the bandwidth of recorded field data.
The low-cut start frequency of the trapezoidal bandpass filter, in Hz. Default is 1 Hz. Energy below this frequency is fully attenuated. Must be less than Frequency 2.
The low-cut end frequency of the trapezoidal bandpass filter, in Hz. Default is 5 Hz. Energy passes at full amplitude above this frequency (up to Frequency 3). Must be greater than Frequency 1 and less than Frequency 3.
The high-cut start frequency of the trapezoidal bandpass filter, in Hz. Default is 70 Hz. Energy begins to roll off above this frequency. Set this value to the highest geologically meaningful frequency in the model.
The high-cut end frequency of the trapezoidal bandpass filter, in Hz. Default is 80 Hz. Energy above this frequency is fully attenuated. Must be greater than Frequency 3. Keep the transition slope (Frequency 4 minus Frequency 3) gentle to avoid ringing.
The error tolerance for the Butterworth bandpass filter design, as a percentage. Default is 1%. This controls the ripple allowed in the filter pass-band. Smaller values produce a sharper filter but may increase numerical artefacts. The default value is appropriate for most applications.
Selects the order of the finite-difference approximation to the Laplacian operator used in the wave equation. Default is Fourth.
Second — uses a 3-point stencil in each direction. It is faster per time step but requires a finer grid and smaller time step to avoid numerical dispersion.
Fourth — uses a 5-point stencil in each direction. It achieves the same accuracy with a coarser grid and larger time step, making it more computationally efficient in practice. This is the recommended choice for most models.
The stability criterion and the suggested time step and grid step values both depend on the chosen order. After changing this setting, re-run the suggestion actions.
When enabled, the top boundary of the model is treated as a free (pressure-release) surface, which generates surface-related multiples and the direct-wave reflection from the top. Default is on. Disable this option when you want a purely absorbing top boundary — for example, to model data that has already had surface-related multiples removed, or to avoid the direct wave contaminating shallow reflections.
Selects the method used to suppress artificial reflections from the edges of the modelling grid. Default is Exponential.
Exponential — applies a sponge layer at all model boundaries where the wavefield amplitude is multiplied by a decaying exponential coefficient on each time step. The thickness of this layer and the decay rate are controlled by the Absorbing bound size in samples and Attenuation coefficient parameters.
Clayton-Engquist — applies first-order absorbing boundary conditions based on the one-way wave equation at each boundary face. This method does not require an extra sponge layer and therefore slightly reduces grid size, but is less effective for waves arriving at oblique angles. When selected, the Absorbing bound size and Attenuation coefficient parameters are hidden.
The thickness of the exponential sponge layer added around the model boundaries, measured in grid samples. Default is 45 samples. At the default 10 m grid step this corresponds to a 450 m sponge zone. Increase this value if you observe residual edge reflections in the output. A thicker sponge absorbs incident energy more gradually and is more effective at oblique angles, but it increases memory consumption. This parameter is only active when Absorbing bound type is set to Exponential.
The per-step amplitude decay coefficient applied within the exponential sponge layer. Default is 0.0053. Larger values damp the wavefield more aggressively within the sponge zone. If the value is too large it can cause instability near the inner boundary of the sponge layer, which shows as low-frequency noise in the output. If it is too small the wave is not fully absorbed before it reaches the grid edge and reflects back. The default value works well with the default sponge thickness of 45 samples. This parameter is only active when Absorbing bound type is set to Exponential.
Selects whether the finite-difference computation runs on the CPU or GPU. GPU acceleration can substantially reduce modelling time for large models.
Enables distribution of the computation across multiple processing nodes in a cluster environment.
Controls the number of gathers sent to each distributed node per work chunk. Larger values reduce communication overhead but may lead to uneven load balancing when nodes have different processing speeds.
When enabled, restricts the number of CPU threads used on each distributed processing node.
An optional text label appended to the distributed job name for identification in the cluster scheduler.
When enabled, allows pinning threads to specific CPU cores via the Affinity field.
Specifies the CPU core affinity mask when Set custom affinity is enabled.
Sets the number of CPU threads used for the finite-difference computation on the local machine. For large velocity models, using all available cores significantly reduces wall-clock time.
When enabled, the module is bypassed and passes input data through to the output unchanged. Use this setting to temporarily disable modelling within a workflow without disconnecting the module.
The primary modelling result. In Exploding reflector and CS gather modes this contains the time-domain synthetic gather with the sample interval and length specified by Output sample ratio and Output data length. In Post-stack RTM mode this contains the depth-domain migrated image.
Available in Post-stack RTM mode only. Contains the RTM depth image stretched back to the time domain using the velocity model for the depth-to-time conversion. This time-domain version facilitates direct comparison with the input post-stack section and assists in interpreting the depth migration result.
Read-only display field. After the module runs it shows the minimum interval velocity (m/s) found in the velocity model. Together with the maximum velocity, this value is used to compute the recommended grid step and time step.
Read-only display field. After the module runs it shows the maximum interval velocity (m/s) found in the velocity model. The stability of the finite-difference scheme is governed by the maximum velocity: the higher the maximum velocity, the smaller the time step required for a given grid step.
Read-only display field. Shows the grid step in metres recommended by the Suggest grid step based on frequency or Suggest min grid step action, based on the velocity model and the chosen modelling order. Use this value as a guide when setting the Model grid step parameter.
Read-only display field. Shows how many grid samples fit within one wavelength of the highest-frequency component of the wavelet, given the current grid step and velocity model. A value of at least 5 to 10 samples per wavelength is required to avoid numerical dispersion; values below 5 indicate that the Model grid step is too coarse.
Calculates and displays the recommended Model grid step based on the dominant frequency of the input wavelet and the minimum velocity in the connected velocity model. The result accounts for the chosen modelling order and is written to the Recommended grid step information field. Run this action first when setting up the modelling parameters.
Reads the maximum velocity from the velocity model and the current Model grid step, then computes the largest time step that satisfies the CFL stability criterion for the chosen modelling order. The suggested value is written directly into the Modelling step parameter. Run this action after fixing the Model grid step to ensure numerical stability.
Reads the maximum velocity from the velocity model and the current Modelling step, then computes the minimum grid step required to keep the scheme stable for the chosen modelling order. The suggested value is written directly into the Model grid step parameter. Use this action when you have a fixed time step and need to determine the coarsest grid that remains stable.