Abstract

The 3-D multiple moveout (MMO) and Iso-stretch radial (ISR) transforms have been used previously to pre-condition data so that multiple events become periodic and the wavelet is stationary in each ISR trace. These transforms make the attenuation of multiples by the autoconvolution method more effective. Auto-convolution is used to predict multiples in the transform space and derive the characteristic wavelet. A least squares optimised filter is derived based on the inverse of this the characteristic wavelet. If the multiple amplitudes form a geometric series, this filter can then be applied to attenuate all surface related multiples in one pass. However, such multiple attenuation can affect the amplitudes of primary events if applied on a trace by trace basis. A further pre-conditioning step has been devised which can reduce the loss of primary energy by subsequent multiple attenuation procedures. This transform makes the wavelet of the same event constant, from trace to trace, across the ISR record. Traces in the transformed ISR record are then stacked into one, to create a Characteristic Multiple Trace. Optimal modelling of the multiple sequence is then achieved on the Characteristic Multiple Trace. Subsequently the modelled multiples are removed from individual traces, preserving the amplitudes of primary events with a dramatic improvement in processing time over trace by trace removal. The performance of this method has been tested on geologically appropriate synthetic models. Surface multiple attenuation is demonstrated on a multi-layered 3D model with a cross-dipping sea floor. This new technique is not computer intensive and does not require velocity analysis prior to its application.