Summary It has been shown that seismic interferometry can be used to predict and subtract direct and/or scattered ground from land seismic data. This method of removal is limited as it requires each survey source to be located next to a receiver, such that “virtual estimates of the ground roll can be subtracted from the real source data. Here, I show how one of the inputs for interferometry can be replaced by a simple model of surface-wave propagation, allowing ground predicted between a source and receiver. The new approac is illustrated using conventional 3D point seismic data.
Introduction Interferometric ground-roll removal is an approach to ground-roll attenuation based on the cross crossconvolution) and summation of wavefields observed at a pair of receiver locations (Curtis et al., al., 2006; Halliday et al., 2007). The resu correlation and summation process is an estimate of the wavefield at one receiver as if a source had been placed at the other (van Manen et al., 2006; Wapenaar and Fokkema, 2006). This process can be configured such that the result is dominated by direct and/or scattered surface waves (Halliday et al., 2010). Provided each source in a survey is located close to a receiver, the resulting estimates can be adaptively subtracted from source records. One potential drawback of this approach is that, in order to be able to attenuate problematic crossline scattered ground roll, the wavefields observed at each receiver must be illuminated from a sufficient range of azimuths interferometry requires illumination of receivers from all azimuths (e.g., sources enclosing the receivers; Figure 1a), but lines of sources may also suffice (as illustrated in Figure 2a). Thus interferometry requires many source positions, potentially in addition to the regular positions used in a survey. This can considerably increase acquisition costs. For example, in Halliday et al. ( nine closely spaced source lines were used to attenuate scattered ground roll from a single source line. drawback is that each source position in the survey must have a receiver located nearby, such that a “virtual estimate of the ground roll can be made. This is unlikely to be the case in a conventional 3D land seismic is this issue that we address here. We consider a new approach where one input to interferometry is replaced by a model of the direct ground roll. This allows consideration of where the ground roll can be estimated between a source and a receiver, allowing the method conventional 3D acquisition geometries. Example 2: Point-receiver cross-spread data We now consider examples using point-source, pointreceiver data recorded as part of a 3D land seismic survey using a cross-spread acquisition geometry. The ground roll recorded in this data set shows some scattering within the noise cone and, therefore, allows testing of the method on both direct and scattered ground roll. The same preprocessing steps are applied to all data used in this abstract: vibroseis correlation, airwave removal, and anomalous amplitude attenuation (spike and noise burst removal).