|
<< Click to Display Table of Contents >> Navigation: Migration > Estimate Q factor |
The Estimate Q factor module measures the seismic quality factor (Q) from pre-stack CMP gathers. The quality factor Q describes how strongly a rock layer absorbs seismic energy: low Q values indicate high attenuation (energy is lost quickly with depth or travel time), while high Q values indicate low attenuation. Accurate Q estimation is essential for inverse-Q filtering (Q compensation), amplitude-versus-offset analysis, and reservoir characterization.
The estimation is based on a Gabor (time-frequency) spectral analysis of the seismic signal. For each analysis location the module collects a super gather — all input traces within a user-defined spatial radius — to improve signal-to-noise ratio before spectral analysis. The amplitude spectrum of the stacked super gather is analyzed in time windows defined either by imported horizon layers or by a regular model grid. The spectral decay between the top and bottom of each window is used to infer the interval Q value for that layer using the chosen Q estimation method (Average, Regression 1, or Regression 2). The results are interpolated spatially and output as both gather-format displays and as pickable Q model items that can be connected to subsequent inverse-Q filtering modules.
The module supports multi-threaded and distributed (remote) execution. Use this module as the first step in a Q-compensation workflow, followed by an inverse-Q filtering module.
The SEG-Y file handle providing the pre-stack seismic data to be analyzed. This should be a CMP-sorted dataset. All traces within the super gather aperture around each analysis bin are read and used together for the Q estimation. Connect this to the same SEG-Y item used in your NMO/stacking workflow.
The sorted trace header index corresponding to the input SEG-Y handle. This provides CMP coordinates, inline/crossline numbers, and offset information used to group traces into super gathers at each analysis bin location.
A bin point vector that defines the spatial locations at which Q values are computed and output. The output Q gathers and Q model items are populated at these bin locations. This is typically the same geometry as your processing grid or a subsampled version of it controlled by the inline and crossline step parameters in the Analysis area group.
The RMS velocity model used to apply NMO correction to the super gather before spectral analysis. Applying NMO flattens reflections, which is necessary to correctly isolate the spectral decay due to attenuation from moveout effects. Connect the same VRMS model you use for standard NMO correction.
An optional mute function that suppresses noisy or unreliable portions of each trace (typically the far-offset stretch zone and shallow refractions) before Q analysis. Providing a consistent mute ensures that only high-quality reflection data contributes to the Q estimation.
A set of time-domain horizons that define the layer boundaries for Q estimation. When the Build type is set to By layers or By model and layers, the module estimates one Q value per layer defined between consecutive horizons. Each layer receives its own interval Q value, which allows the Q model to faithfully represent geologically realistic depth-varying attenuation. This input is not required when Build type is set to By model.
The spatial radius (in metres) around each output bin within which input traces are collected and stacked to form the super gather used for Q analysis. Default: 500 m. Range: 0 to 10000 m.
Increasing this aperture averages more traces together, which improves signal-to-noise but reduces spatial resolution in the Q model. Use larger values in areas with sparse acquisition or noisy data. Use smaller values when you need to resolve rapid lateral changes in attenuation.
The reference velocity (in m/s) used to initialize the spatial interpolation of Q picks across the output bin grid. Default: 2000 m/s. Minimum: 100 m/s.
This value is used as the background fill where Q picks are absent — for example in areas outside the survey or in time zones with no analysis windows. Set it to a value representative of the average velocity in your area.
Enable this option for 3D surveys. Default: off (2D mode). When enabled, the module searches for traces in both the inline and crossline directions when building the super gather, which is appropriate for 3D acquisition geometry. For 2D surveys, leave this option off so that only traces along the 2D line are included.
The mathematical method used to derive the Q value from the frequency-domain spectral ratios computed by the Gabor transform. Default: Average. Options:
Average — estimates Q by averaging the spectral decay across all frequencies in the analysis band. This is the most stable and noise-tolerant option and is recommended as the starting point.
Regression_1 — fits a linear regression to the log spectral ratio as a function of frequency to estimate Q. This can be more accurate when the data has a clean spectral shape but is more sensitive to noise.
Regression_2 — an alternative regression approach that additionally outputs amplitude (A model) and azimuth (Q-azimuth model) attributes alongside the Q factor. Use this option when you need to analyze azimuthal anisotropy of attenuation.
Controls how the time windows for each analysis layer are referenced when the Build type is set to By model or By model and layers. Default: Interval. Options:
Interval — each analysis window spans from the bottom of the previous window to the bottom of the current window (true interval Q). This is the standard mode and directly recovers the layer-by-layer interval Q.
RMS — each analysis window always extends from the top of the model (FromTime) to the current bottom time, accumulating the full travel path. The result is analogous to an RMS (effective) Q rather than a true interval Q.
RMS2 — an extended RMS mode that additionally computes amplitude (A) and azimuthal Q attributes in one pass. Available only when Q estimation type is Regression_2.
The spatial interpolation method used to spread the Q picks computed at analysis bins across the full output geometry grid. Default: Triangulation. Options:
Triangulation — Delaunay triangulation with linear interpolation within triangles. Produces smooth results and is well-suited to regularly sampled analysis grids.
Voronoi — nearest-neighbour interpolation using Voronoi cells. Each output bin inherits the Q value of the nearest analysis bin. Use this option when you want sharp boundaries between analysis zones rather than smooth gradients.
This parameter group controls the spatial extent and temporal range of the Q analysis. The sub-parameters are described below.
Determines how the time windows for Q analysis are defined. Default: By layers. Options:
By layers — the Input horizons define the top and bottom of each Q estimation window. One Q value is computed per layer. Requires the Input horizons parameter to be connected.
By model — the time range and step (From time, To time, Step time) define a regular grid of analysis windows. No horizons are required. This mode is useful for a first-pass Q scan without an interpreted horizon set.
By model and layers — combines both approaches: the regular model time grid is used as the base, but window boundaries are snapped to horizon times when a horizon falls within a time step. This prevents windows from crossing geological layer boundaries and produces the most geologically consistent Q model.
Controls which area of each horizon layer is filled with estimated Q values when Build type is By layers. Default: Whole layer surface. Options:
Whole layer surface — Q is estimated and populated everywhere across the full spatial extent of each layer.
Stay inside layer — Q is only estimated within the area where both the top and bottom horizon of that layer exist (their intersection). Use this when horizons have limited lateral extent to avoid extrapolating Q into undefined areas.
When enabled, Q is also estimated for the interval between the surface (time zero / topography datum) and the first horizon or the model start time. Default: off. When disabled, the near-surface interval above the first horizon is filled with the Maximum Q factor value (treated as transparent).
Enable this option when the near-surface weathering layer or low-velocity zone causes significant attenuation that should be included in the Q model.
The first inline number in the analysis area. Default: -1 (no limit — process from the first available inline). Set this to a positive inline number to restrict computation to a sub-area of the survey.
The last inline number in the analysis area. Default: -1 (no limit — process to the last available inline). Use together with the First inline number to define a rectangular sub-area for testing or partial processing.
The first crossline number in the analysis area. Default: -1 (no limit). For 2D data this parameter is not applicable.
The last crossline number in the analysis area. Default: -1 (no limit). For 2D data this parameter is not applicable.
The number of inlines to skip between successive Q analysis locations. Default: 45. Minimum: 1. Increase this value to perform a coarser, faster scan of the survey. Decrease it to obtain a denser Q model at the cost of longer computation time. The spatial interpolation step will fill in Q values between analysis points, so a step of 10 to 50 inlines typically provides sufficient density for smooth Q models.
The number of crosslines to skip between successive Q analysis locations. Default: 45. Minimum: 1. For 2D surveys this parameter is not applicable. Set to the same order of magnitude as the inline step to produce a roughly isotropic Q sampling density.
The start time (in seconds) of the model-based Q analysis window. Default: 0 s. Active when Build type is By model or By model and layers. Set this to the shallowest time from which you want to estimate Q. If Calculate Q from topo to first layer is enabled, a single additional window from time zero to this value is also analyzed.
The end time (in seconds) of the model-based Q analysis window. Default: 15 s. Minimum: 0.001 s. Active when Build type is By model or By model and layers. Set this to the deepest time of interest. Values beyond the length of the input traces are automatically clamped to the trace length.
The time interval (in seconds) between successive analysis windows when using a model-based grid. Default: 0.25 s. Minimum: 0.001 s. Smaller steps produce a finer vertical Q resolution but increase computation time. The step is automatically clamped to be no smaller than the sample interval of the data. For typical seismic bandwidths, steps of 0.1 to 0.5 s are appropriate.
This parameter group controls the time-frequency decomposition used to extract spectral information from the super gather. The Gabor transform uses Gaussian-windowed short-time Fourier transforms to produce a smoothly time-varying spectrum, which is then analyzed for frequency-dependent amplitude decay to estimate Q.
The lower boundary (in Hz) of the frequency band used for Q estimation. Default: 0 Hz. Set this above the lowest reliable signal frequency of your data (typically 5 to 15 Hz) to avoid contaminating the spectral analysis with low-frequency noise. Very low frequencies often have poor signal-to-noise ratio and can distort the Q estimate.
The upper boundary (in Hz) of the frequency band used for Q estimation. Default: 125 Hz. Set this below the Nyquist frequency of your data and below the frequency at which the signal-to-noise ratio drops significantly. Using frequencies where noise dominates will produce unreliable Q estimates.
The half-width (in seconds) of the Gaussian window used in the Gabor transform. Default: 0.1 s. Minimum: 0.008 s. A larger window provides better frequency resolution but poorer time resolution — it averages spectral characteristics over a longer time interval. A smaller window improves time localization but reduces the accuracy of low-frequency spectral estimation. For most datasets a value of 0.05 to 0.15 s is appropriate.
The time step (in seconds) at which the Gaussian analysis window is advanced through the trace during the Gabor transform. Default: 0.01 s. Minimum: 0.002 s. Smaller steps produce a more densely sampled time-frequency representation and are more accurate but increase computation time. The step should generally be smaller than the Gauss half window to ensure adequate overlap between successive analysis positions.
This parameter group contains advanced settings that control data reading performance, parallelism, and output quality bounds. These settings are rarely changed from their defaults but can be adjusted for performance tuning or to control the range of output Q values.
The number of bins processed in each distributed execution chunk sent to a remote worker node. Default: 100 bins. Minimum: 1. Larger values reduce communication overhead in distributed runs but increase memory use per chunk. Leave at the default unless you are tuning performance on a cluster.
Enables multi-threading within each distributed execution chunk. Default: off. When enabled, the Gabor transform computations across bins within a single chunk are parallelized. This can improve throughput on multi-core worker nodes when running in distributed mode. For local single-machine runs, the standard multi-threaded execution controlled by the main thread count setting is usually sufficient.
The upper limit applied to estimated Q values. Default: 500 (dimensionless). Any Q estimate above this value is clamped to the maximum. Very high Q values (above several hundred) are physically unrealistic for typical sedimentary sequences and usually indicate areas of poor signal quality or insufficient spectral bandwidth. This value is also used as the fill value for intervals where Q is not computed (such as the near-surface zone when Calculate Q from topo to first layer is disabled).
The lower limit applied to estimated Q values. Default: 20 (dimensionless). Any Q estimate below this value is replaced by this minimum. Very low Q values can arise from noise, cycle-skipping, or thin layers with ambiguous spectral content. This floor prevents unphysically strong attenuation from contaminating the Q model and the downstream inverse-Q filter.
A gather displaying the estimated amplitude factor (A) as a function of bin and time. This output is populated when the Engine type is set to RMS2 together with Q estimation type Regression_2. Each trace represents how the amplitude term of the spectral model varies with time at one spatial location. Use this output for QC of the spectral fitting quality and for azimuthal amplitude analysis.
A gather displaying the estimated interval Q as a function of bin location and two-way time. Each sample value is the interval Q for the analysis window containing that time sample. This is the primary output for reviewing the layer-by-layer attenuation structure. Connect this output to a velocity analysis display to inspect Q laterally or to a seismic section display to overlay Q on the stacked section.
A gather displaying the effective (average) Q computed by integrating the interval Q model from the surface downward. At each time sample t, the average Q is calculated as Q_avg(t) = t / integral(1/Q_int dt from 0 to t). This quantity is the appropriate input for a constant-Q inverse-Q filter applied in the time domain, as it correctly accounts for the accumulated attenuation over the full travel path.
A gather displaying a dimensionless Q ratio attribute (clamped between 1.0 and 2.0) within the analysis time range (From time to To time), and zero outside. This output is produced when Q estimation type is Regression_2 with Engine type RMS or RMS2. It can be used as a normalized attenuation indicator for relative comparison between locations.
A gather displaying the azimuth of the principal attenuation axis (in degrees) as a function of bin and time, within the analysis time range. Values are output in degrees (converted from radians internally). This output reflects azimuthal anisotropy of attenuation, which can be related to fracture orientation or stress fields. Populated only when Engine type is RMS2.
The primary output Q model stored as a vertical velocity picking item (GVerticalVelocityPickingItem). This model contains the spatially interpolated interval Q values across the full survey area and can be connected directly to inverse-Q filtering modules. It is the recommended output to pass to subsequent Q-compensation processing.
The spatially interpolated amplitude factor model stored as a vertical velocity picking item. Populated only when Engine type is RMS2. This model captures the spatial variation of the amplitude scaling term in the spectral attenuation model and can be used for amplitude balancing or further geophysical interpretation.
The spatially interpolated Q-ratio model stored as a vertical velocity picking item. Populated only when Q estimation type is Regression_2. This model provides a normalized measure of attenuation variability across the survey area and is useful for identifying zones of anomalously high or low attenuation relative to the regional trend.
The spatially interpolated azimuthal Q anisotropy model stored as a vertical velocity picking item. Populated only when Engine type is RMS2. The stored values represent the azimuth angle (in degrees) of the minimum-Q direction, which typically aligns with the dominant fracture orientation. Use this model together with the Q model for azimuthal Q compensation or fracture characterization.