Application of 3-D FKK Filtering in 3-D High-density Onshore Seismic Field Data

For linear noise such as seismic ground roll, 3-D frequency-wavenumber-wavenumber (3-D FKK) domain filtering suppression is better than 2-D frequency-wavenumber (F-K) domain filtering. In recent years, with the continuous development of computer processing speed and memory capacity, high-density data acquisition in seismic exploration has been widely applied in the hydrocarbon industry, opening up the application of 3-D FKK filtering methods. We applied the 3-D FKK filtering software developed locally by ourselves to a 3-D high-density onshore seismic field dataset from a coal mines in western China. The case study demonstrates that the linearity of the noise in the field data is better represented by constructing the single shot records as a minimum dataset. Both theoretical synthetic models and the 3-D high-density onshore seismic field data numerical filtering experiments demonstrate that the feasibility of our 3-D FKK filtering software. The filtering effects are good. Introduction Currently, 3-D seismic exploration is a popular technique (e.g., Yan et al., 2017), but the 3-D filtering has not developed significantly. The common approach is to first filter data inline and then crossline (two-pass filtering method, Stewart, 1989; Peardon and Bacon, 1992; Li and Larner, 1993) to replace the true 3-D filtering, which is low efficiency because of switching data between the horizontal and vertical directions. Separating linear coherent noise, such as ground roll from reflections, remains a key challenge in seismic processing. The coherent noise, especially the ground roll, are difficult to grasp in the form of 3-D shot gather datasets in 2-D. The apparent velocity of the coherent noise identified in 2-D is different from its true apparent velocity, and the apparent velocity of each shot gather is also different. Therefore, it is not ideal to filter out the coherent noise with a two-pass filtering method. The distribution of signal and ground roll are different in the frequency-wavenumber domain, so we can identify the noise energy in frequency-wavenumber relationship, and we can design the factors removed, we cut off the energy that is higher (or lower) than a certain dip angle to allow for filtering. The specific implementation of the cut-off can be carried out in the time-space domain (Treitel et al., 1967; Hale and Claerbout, 1983; Claerbout, 1985; Li and Larner, 1993; Liao et al., 2018); or in the frequency-wavenumber domain (Wiggins, 1966; Peardon and Bacon, 1992). As seismic studies undertake investigation of more complex geology, more innovative methods were needed. Examples include derivative filters (Melo et al., 2009; Nie et al., 2014), empirical model decomposition (Bekara and van der Baan, 2009), time-frequency transforms (Askari and Siahkoohi, 2008; Liu and Fomel, 2013), and compressed sensing and sparsity promotion (Wang et al., 2008; Gholami, 2014). A generic shortcoming of these methods is the operational loss of the lower end of the body wave frequency spectrum. By adapting the redundant lifting scheme, a wavelet transform method, to seismic data, Aghayan et al., (2016) determined how the wavelet domain can be used to suppress coherent and random noise. They proposed the new method that was wavelet-based ground roll noise suppression using a synthetic shot gather and two real gathers. High-density seismic exploration has been developed in recent years in order to improve the quality of seismic data acquisition and processing techniques, which use a single-sensor recording, vibroseis alternating scanning and digital detector technologies (Rajab et al., 2006). High-density seismic exploration data with small temporal and spatial sampling intervals, and high folding numbers, eliminates the influence of the false frequency in processing, which improves the separation of the signal and noise in the 3-D FKK domain greatly. With different ways to realize filtering the two-pass filtering method first filters in the FKx domain and then in the FKy domain. While one-pass filtering method removes the ground roll energy in the FKK domain directly. Peardon and Bacon (1992) used a marine seismic section that was a single line from a 3-D survey for examples of 3-D FKK filtering application. In this paper, we extended their research and applied 3-D FKK filtering method (Stewart, 1989) on a 3-D high-density onshore seismic field data from a coal mines in western China, which was typically noisier than marine seismic field data. Both theoretical synthetic models and the 3-D high-density onshore seismic field data numerical filtering experiments demonstrate that the feasibility of our 3-D FKK filtering software. The filtering results show that the denoising effects are good.