Description

The Local Matched Filter module adaptively reshapes one seismic gather (the Model) so that it locally resembles a second, reference gather (the Input) as closely as possible. It does this by designing a space- and time-varying shaping filter independently in a series of overlapping analysis windows: at each window position the module studies a local group of neighboring traces, derives the least-squares filter that best transforms the Model data into the Input data, applies that filter to the Model trace, and blends the overlapping results back together into a single output gather. The output is also rescaled locally so its energy matches that of the Input gather, preventing the filter design from introducing artificial gain changes.

Use this module whenever two seismic datasets that should represent the same subsurface response differ in wavelet shape, phase, or amplitude because of differences in acquisition or processing history - for example when matching a monitor survey to a base survey in 4D repeatability work, when shaping a modeled or predicted noise or multiple component so it can be subtracted from the recorded data, or when generally equalizing one volume to another prior to comparison or merging.

Input data

Input Gather

The reference seismic gather that the Model gather is matched against. This is the data set whose waveform character, amplitude, and timing the module tries to reproduce when it reshapes the Model gather. It must have the same number of traces and samples as the Model Matched to Gather input.

Model Matched to Gather

The second gather (the Model) that is locally reshaped by the adaptively derived filter so that it best matches the Input Gather. This is typically a repeat/monitor survey that must be conformed to a base survey, or a modeled/predicted component (such as a noise or multiple model) that needs to be shaped to resemble the corresponding energy in the real data before it is subtracted downstream. It must share the same trace and sample geometry as the Input Gather.

Parameters

Horizontal sliding window

Sets the width, in traces, of the spatial averaging window used to build the local matching filter at each output trace. For every trace, the module accumulates cross-correlation and autocorrelation statistics from this many neighboring traces, centered on the trace being processed, before solving for the shaping filter, which makes the filter estimate more stable in the presence of trace-to-trace noise. Narrower windows adapt more quickly to lateral changes in wavelet character but are more sensitive to noise; wider windows give smoother, more robust filters but blur lateral variations, for example across a fault or a strong change in reflectivity.

Default: 40 traces (approximately ±20 traces around the current trace). Minimum: 1 trace.

Time window

Sets the length, in time, of the analysis window used at each position of the sliding window to estimate the matching filter statistics and to which the resulting filter is applied. Larger windows average the filter estimate over a longer time interval, producing smoother and more stable results when the mismatch between the two gathers changes gradually with time; shorter windows track faster time-varying changes in wavelet character but need a wider Horizontal sliding window or a shorter Operator Length to remain stable.

Default: 500 ms. Minimum: 15 ms.

Taper

Length of the taper applied to each end of the analysis window before the correlation and filter design step. The taper is added on top of the Time window (the working window used internally is the Time window plus twice this taper length), and it suppresses edge artifacts that would otherwise appear where the window is cut off abruptly. Increase this value if the output shows instability or ringing near window boundaries.

Default: 20 ms. Minimum: 0 ms.

Time Sliding Step

Sets the distance, in time, between successive positions of the sliding analysis window as it steps down each trace. A smaller step recomputes the local matching filter more often, so the result tracks time-varying differences between the Model and Input gathers more precisely, at the cost of longer processing time. A larger step reduces computation but assumes the ideal matching filter changes only slowly with time.

Default: 250 ms. Minimum: 4 ms.

Operator Length

Sets the length, in time, of the shaping filter that is solved for and applied to the Model gather within each analysis window. A longer operator can compensate for larger phase and amplitude differences between the Model and Input gathers, but requires more data to estimate reliably and can introduce ringing artifacts if pushed too long. A shorter operator is more robust and stable but can only correct simple gain and phase differences between the two gathers.

Default: 200 ms. Minimum: 4 ms.

Max Time Shift

Sets the largest static time misalignment, in time, expected between the Model gather and the Input gather. Set this to a value at least as large as the largest timing difference you expect between the two data sets being matched, for example residual timing drift between a monitor survey and a base survey in 4D processing.

Default: 200 ms. Minimum: 40 ms.

Noise

Pre-whitening (stabilization) factor applied to the zero-lag autocorrelation before solving for the matching filter, expressed as a fraction of the Model trace's own energy (0.01 = 1%). This regularizes the filter design when the Model data has weak or unstable energy within an analysis window, preventing the solution from becoming an unstable, overly "spiky" filter. Increase this value, for example to 0.05-0.1, if the output shows noisy, unstable, or ringy artifacts; keep it low when the Model and Input gathers are already a close match and a sharper, more precise filter is desired.

Default: 0.01 (1%). Minimum: 0.

Output data

Output Gather

The Model gather after being locally reshaped by the adaptive matching filter so that it best matches the Input Gather trace by trace and window by window. Its amplitude is rescaled within each analysis window to preserve the Input gather's local energy level, so the output is not just phase- and waveform-matched but also amplitude-consistent with the reference data. Use this output directly for 4D-matched comparisons, or feed it into a subsequent difference step to isolate and subtract a matched noise, multiple, or repeatability component from the original data.

References

Verschuur, D. J., Berkhout, A. J. and Wapenaar, C. P. A., 1992, Adaptive surface-related multiple elimination: Geophysics, 57, 1166-1177.