The heterogeneity of rock fabric and fluid distribution induce Wave-induced Fluid Flow (WIFF), a mechanism that significantly affects the velocity dispersion and energy attenuation of seismic waves. We integrate pore structure (intralayer) and fluid saturation (inter-layer) heterogeneity into a unified framework to establish a layered double-porosity patchy-saturated (LDPPS) porous model and obtain the potential, kinetic and dissipation functions via the generalized Biot theory. Based on Hamilton’s principle, we derive the new governing equations for the LDPPS model. Five P-waves and one S-wave are achieved through plane wave analysis. We investigate the properties of fast P-wave and S-wave through changing relevant rock parameters. The results show that there are three attenuation peaks for fast P-wave in the whole frequency band, corresponding to mesoscopic fluid flow (fluid heterogeneity), microscopic fluid flow (fabric heterogeneity) and macroscopic fluid flow (Biot flow). Furthermore, it is found that the model exhibits strong sensitivity to water saturation, permeability, fluid viscosity and heterogeneity scale. Finally, we compare the dispersion and attenuation predicted by our theory with experimental data, and the results show that the theoretical predictions are consistent with reality. Since our proposed theory can account for multiscale fluid flow in the whole frequency band, it is of great significance for fluid evaluation in reservoirs.

Elastic least-squares reverse time migration (EL-SRTM) effectively enhances imaging resolution and interprets multicomponent seismic data. However, traditional ELSRTM approaches, based on first-order gradient optimization methods, do not exploit the Hessian matrix, which is critical for inversion problems. To address this limitation, we propose a truncated Newton ELSRTM method (TN-ELSRTM) utilizing multi-parameter Hessian-vector products to achieve efficient and high-quality seismic imaging. By neglecting the second-order nonlinear terms with minor influence in the Hessian matrix, the Gauss-Newton term is used to represent the Hessian matrix. The Hessian-vector products are computed through single demigration and migration process, enabling the determination of the Newton update direction via a matrix-free conjugate gradient (CG) method. Implemented in a nested iterative framework, the outer loop focuses on gradient computation and model updating, while the inner loop solves for the Newton update direction in a matrix-free manner. This ensures simultaneous minimization of misfit in the data and model spaces. Compared to traditional conjugate gradient ELSRTM (CG-ELSRTM), the proposed TN-ELSRTM method significantly accelerates convergence and requires fewer iterations. It effectively mitigates finite aperture effects, band-limited wavelet, geometric spreading, and multi-parameter coupling. Numerical results demonstrate superior imaging quality, alleviating P- and S-wave reflectivity trade-offs, enhancing deep illumination, expanding horizontal imaging range, increasing vertical high-wavenumber components, and improving resolution. The propsed TN-ELSRTM method produces balanced and high-resolution images for P- and S-wave reflectivity models with greater efficiency.

As an effective time-frequency (TF) analysis method, S transform (ST) has an extensive application in signal processing. However, for broad-band seismic signals, the peaks of the frequency distribution in the TF spectrum of ST biases the actual Fourier spectrum. Besides, the TF resolution of ST is affected by the relatively fixed window function and the Heisenberg uncertainty principle. In order to correct the frequency bias and improve the TF resolution of ST, and at the same time, to increase the flexibility of the window function, we propose a new TF analysis method, called the deconvolutive frequency corrected three-parameter S transform (DFC-TPST). The DFC-TPST includes two steps: modifying the window function to achieve the frequency corrected three-parameter S transform (FC-TPST) and deconvoluting FC-TPST to achieve DFC-TPST. The synthetic example proves the superiority of the method in characterizing seismic signals. Through comparison of field data testing, we find that the proposed method can be well applied to the hydrocarbon detection in tight sandstone reservoir.

Time-frequency (TF) analysis is an important tool for seismic signal analysis, and traditional methods are difficult to achieve high TF resolution and energy aggregation. In this letter, we propose the adaptive time-synchroextracting S transform (ATSEST) for seismic data processing and interpretation. The method first calculates the scale parameter of the window function through the Fourier spectrum of the signal to obtain the adaptive S transform spectrum. After that, the final result is obtained by extracting the TF coefficients at the group delay and removing a large amount of fuzzy energy in the TF spectrum. The results of the synthetic example TF analysis show that the method can improve the localization ability of transient features of seismic signals. We apply the method to the fault identification of field data, and the results show that the coherent attribute slices extracted by using ATSEST can well characterize the fault.

In wave equation modeling, wavefields propagating over time can be regarded as feedforward in a recurrent neural network. Therefore, the seismic inversion problem based on partial differential wave equations can be addressed using automatic differentiation in the state-of-art deep learning frameworks, eliminating the need for explicit backpropagating the residual wavefield. However, one challenge that arises in the context of automatic differentiation is the significant memory usage due to the necessity of storing the hidden states of the RNN (i.e., wavefields in seismic modeling) during forward computation for constructing the computational graph and computing the derivatives during backpropagation. This memory overhead can become a bottleneck, particularly when dealing with large-scale inversion problems. To mitigate this issue, we propose an effective boundary saving strategy that allows for the reconstruction of the computational graph during the backpropagation process. Instead of storing all the intermediate wavefields at each time step, we selectively save the necessary information at the boundaries, thereby significantly reducing the memory footprint. This approach enables us to maintain the convenience and efficiency of automatic differentiation computations while minimizing the memory requirements. Both 2D and 3D numerical experiments validate the accurate reconstruction of wavefields with minimal loss in precision, while the computational graph is simultaneously reconstructed. Consequently, the gradients can also be calculated correctly by automatic differentiation with minimal CPU/GPU memory occupation.

The supervised neural network-based method provides an effective way for seismic data denoising. The noise level of seismic erratic noise, i.e. outlier, varies from traces, time windows and shot gathers. The networks with popular structures may damage reflections because the network cannot learn the exact location of the noises. In order to accommodate the characteristics of erratic noise, we propose using an attention-based network which focuses more on noisy regions. The network outputs an attention map which shows the spatial distribution probability of erratic noise and a raw-noise which may have leakage reflections. To train the network with two kinds of outputs, we use two different loss function to optimize the network. The erratic noises in the specified area can be extracted by multiplying the attention map and the raw-noise, and the reflections in areas without noise are preserved. For generating the training set, we proposed a shuffled noise strategy which starts from the inaccurate denoised data with conventional method. Only the noisy data is available in the whole denoising workflow. The network can learn the noise features effectively with the shuffle noise strategy and can achieve better denoising effect than that based on conventional method. Besides, the denoising capability of the network can be controlled manually by filtering the attention map with probability, and the denoising effect can be further improved through recovering the leaked signals. Synthetic and field data examples show that the proposed method has potential for field erratic noise attenuation.