Multiple suppression on single-channel marine data used to be a problem
Several techniques of multiple suppression are well known to work with 'big' full-offset marine seismic data, including Radon demultiple, SRME, Tau-P deconvolution, etc. However, efficient suppression of multiple reflections on high-resolution offshore seismic data acquired with one or just a few channels at zero or small offsets until now have been a deep problem. Predictive deconvolution, the mainstay of single-channel marine demultiple processing, in most cases is nearly useless. It either does not affect the strongest multiples or erases a blank gap around the unwanted event killing all adjacent reflections.
New efficient demultiple algorithm for shallow seismics
The RadExPro software now provides a new efficient demultiple algorithm for near-offset single-channel or stacked offshore seismic data. The algorithm is implemented in a brand-new processing module called Zero-Offset DeMultiple. It brings really amazing results being capable to significantly suppress multiples without disturbing criss-crossing useful reflections.
Examples are shown below. Principal multiple reflections are indicated with arrows. Note the primary reflections interfering with the multiples. When the multiples are significantly suppressed, the interfering primaries are still there and not disturbed.
Click on the images below for a full-size view:
How does it work?
The algorithm is based on adaptive subtraction of a model of multiples from the original wave field. The model is obtained from the data itself either by static shift of the original traces of by auto-convolution. However, what makes the algorithm that effective is the adaptive subtraction approach used.
In general terms, a dedicated shaping filter is calculated for each trace basing on both the original data traces and the model traces. This filter, when applied to the trace, is trying to minimize the RMS amplitudes of whatever is found similar between the trace and the model. When the filter is calculated it is accounted for non-stationarity to make it adaptive to the events that are quite similar but not exactly the same. This makes the subtraction impressively efficient even for rather approximate models when the arrival time or/and the amplitude of the modeled multiple differ from the actual observation.
The algorithm is implemented as a single processing routine with simple user-friendly interface:
The parameters of the routine are discussed in the RadExPro Plus User Manual.
