Two Strategies for Attenuating Internal Multiples in Land Data

Abstract

Abstract Two new internal multiple attenuation methodologies are presented. The first approach uses a 3D wave equation based multiple modeling technique followed by 3D adaptive subtraction. The recorded data are back propagated/propagated through a reflectivity model (obtained from a preliminary migration) of the overburden. All possible multiple raypaths are modeled by using different combinations of two sub-windows in this reflectivity model. Each combination results in a specific multiple model. The generated multiple models are then simultaneously adapted and subtracted from the input data. The second approach is a 3D dip extraction and filtering technique, taking advantage of any dip discrimination between the primaries and multiples. It works on post stack and pre stack volumes. We show how the two methods are complementary and how they can be combined. The first approach can be applied to any seismic data plagued with strong internal multiples and is perfectly adapted to modern dense, wide azimuth surveys. The second approach works on any seismic volume in which multiples and primaries have different dips in a given domain. Some real data examples from South Oman 3D seismic datasets are shown. All are characterized by a heavy multiple contamination, generated by strong shallow reflectivity sequences in a flattish overburden, and overlying deeper, weaker and heavily structured primaries. By revealing structures previously invisible, these techniques add a lot of value to existing seismic data sets. The generation mechanism of internal multiples is often quite complex making them difficult to predict. The main benefits of the methods described here are that they work in absence of move out discrimination and without the need of a precise identification of the internal multiple generators. Introduction In some areas of Middle East, it is common to have shallow carbonate sequences with strong reflectivity, high velocity contrasts, overlying lower reflectivity clastic series with weak impedance contrasts and lower velocities. This geologic sequence yields a heavy contamination of the lower unit by energetic interbed multiples generated in the upper unit ("curtain of multiples"). These multiples often have the same, or higher apparent velocity, than the primaries, offering no opportunity to remove them by move-out discrimination. Similarly, prediction methods (gapped deconvolution) have proved unsatisfactory. And in addition, it is mostly impossible to identify clearly the generators, given the large number of possible candidates. Historically these unwanted multiples have been removed by surgical 3D FK filtering after flattening of the suspected multiple generators. This method has been used for many years but has several strong limitations, in addition to the well known FK artifacts. Hence new methodologies were required to overcome the failings of this 3D FK approach.