Equivalent Q Estimation Using a Deep-learning-based Decoupling Method

In seismic exploration, Q model estimation from post-stack seismic data is an important problem since this model is a key prerequisite for reservoir identification and improving the vertical resolution of seismic data. The effects of Q are mainly manifested amplitude reduction and phase distortion of seismic data. In order to obtain high-resolution seismic data to describe oil and gas reservoirs, many Q factor estimation methods have been proposed. These methods can be roughly divided into direct estimation methods and inversion methods. Direct estimation methods, such as logarithmic spectral ratio method, frequency shift method, etc., usually use the attributes of seismic data to estimate Q, but it usually has disadvantages such as poor stability, dependence on source wavelet type, and the need for piecewise estimation of multi-layer Q model (Tonn, 1991). In contrast, the inversion methods regard Q factor as a model parameter and use the inversion algorithm to obtain the dynamic Q curve with traveltime or depth, which improves the calculation efficiency and stability, such as a novel method for Q analysis on reflection seismic data (Wang, 2004). However, the reflectivity and Q factor simultaneously affects the waveform of post-stack seismic data, leading to the fact that the Q model cannot be independently estimated without providing an accurate reflectivity model. The general approach for solving this problem is to simultaneously estimate these two parameters in an alternative iteration way (Wang et al., 2016). However, since neither the Q factor nor the reflectivity has a good initial model, the approach has no convergence guarantee.