Dynamic deconvolution in the time-frequency domain is an effective method to improve seismic resolution. However, the traditional methods usually suffer from the limitation of window width and shape, resulting in insufficient time-frequency resolution and phase distortion of seismic signatures, which significantly limits their applicability in complex geological conditions. To overcome this problem, we propose an adaptive local frequency modulation transform (ALFMT), which is directly embedded in the dynamic deconvolution method. ALFMT can dynamically adjust the shape of the analysis window based on the adaptive local frequency information so that the time-frequency energy is effectively focused near the local frequency, thereby improving the focusing and resolution of the time-frequency representation. The synthetic examples and actual seismic application demonstrate that the ALFMT-based dynamic deconvolution can effectively compensate for the seismic amplitude energy, and the compensated data has a higher resolution and clearer reflection characteristics.

Time-frequency analysis (TFA) serves as a critical tool in seismic signal processing and interpretation, particularly for characterizing non-stationary signals. However, conventional TFA methods, such as the short-time Fourier transform (STFT) and continuous wavelet transform (CWT), suffer from inherent limitations, including energy smearing and insufficient time-frequency resolution, which hinder their ability to meet the demands of high-precision seismic interpretation. By detecting local maxima along the frequency direction, LMSST significantly improves energy concentration in time-frequency representations (TFRs). Meanwhile, the W transform is known for high resolution in low-frequency regions and a flexible windowing function, which surpasses conventional TFA methods. However, both methods face limitations when applied independently to complex seismic signals, particularly in scenarios demanding high precision and resolution. To overcome these challenges, the local maximum synchrosqueezing W transform (LMSSWT) is proposed. This approach combines the adaptive windowing of the W transform with the precise frequency reallocation of the LMSST, resulting in a more centralized and energy-concentrated time-frequency representation. Furthermore, an inverse LMSSWT is also derived to ensure completeness and accurate signal reconstruction. By synergizing the W transform’s adaptive windowing with the energy concentration of LMSST, LMSSWT overcomes key limitations in time-frequency analysis of complex seismic signals, offering a powerful tool for high-resolution reservoir prediction and hydrocarbon detection. The effectiveness and applicability of the proposed method are validated through synthetic tests and practical data applications.

The wave-induced fluid flow (WIFF) occurring in the ubiquitous layered porous media (e.g., shales) usually causes the appreciable seismic energy dissipation, which further leads to the frequency dependence of wave velocity (i.e., dispersion) and elastic anisotropy parameters. The relevant knowledge is of great importance for geofluid discrimination and hydrocarbon exploration in the porous shale reservoirs. We derive the wave equations for a periodic layered transversely isotropy medium with a vertical axis of symmetry (VTI) concurrently with the annular cracks (PLPC medium) based on the periodic-layered model and anisotropic Biot's theory, which simultaneously incorporate the effects of microscopic squirt fluid flow, mesoscopic interlayer fluid flow and macroscopic global fluid flow. Notably, the microscopic squirt shorten fluid flow emerges between the annular-shaped cracks and stiff pores, which generates one attenuation peak. Specifically, we first establish the stress-strain relationship and pore fluid pressure in a PLPC medium, and then use them to derive the wave equations by means of the Newton's second law. The plane analysis is implemented on the wave equations to yield the analytic solutions for phase velocities and attenuation factors of four waves, namely, fast P-wave, slow P-wave, SV-wave and SH-wave, and the anisotropy parameters can be therefore computed. Simulation results show that P-wave velocity have three attenuation peaks throughout the full frequency band, which respectively correspond to the influences of interlayer flow, the squirt flow and the Biot flow. Through the results of seismic velocity dispersion and attenuation at different incident angles, we find that the WIFF mechanism also has a significant impact on the dispersion characteristics of elastic anisotropy parameters within the low-mid frequency band. Moreover, it is shown that several poroelastic parameters, such as layer thickness ratio, crack aspect ratio and crack density have notable influence on seismic dispersion and attenuation. We compare the proposed modeled velocities with that given by the existing theory to confirm its validity. Our formulas and result can provide a better understanding of wave propagation in PLPC medium by considering the unified impacts of micro-, meso- and macro-scale WIFF mechanisms, which potentially lays a theoretical basis of rock physics for seismic interpretation.

Deterministic inversion based on deep learning has been widely utilized in model parameters estimation. Constrained by logging data, seismic data, wavelet and modeling operator, deterministic inversion based on deep learning can establish nonlinear relationships between seismic data and model parameters. However, seismic data lacks low-frequency and contains noise, which increases the non-uniqueness of the solutions. The conventional inversion method based on deep learning can only establish the deterministic relationship between seismic data and parameters, and cannot quantify the uncertainty of inversion. In order to quickly quantify the uncertainty, a physics-guided deep mixture density network (PG-DMDN) is established by combining the mixture density network (MDN) with the deep neural network (DNN). Compared with Bayesian neural network (BNN) and network dropout, PG-DMDN has lower computing cost and shorter training time. A low-frequency model is introduced in the training process of the network to help the network learn the nonlinear relationship between narrowband seismic data and low-frequency impedance. In addition, the block constraints are added to the PG-DMDN framework to improve the horizontal continuity of the inversion results. To illustrate the benefits of proposed method, the PG-DMDN is compared with existing semi-supervised inversion method. Four synthetic data examples of Marmousi Ⅱ model are utilized to quantify the influence of forward modeling part, low-frequency model, noise and the pseudo-wells number on inversion results, and prove the feasibility and stability of the proposed method. In addition, the robustness and generality of the proposed method are verified by the field seismic data.

Oil and gas are important energy resources and industry materials. They are stored in pores and fractures of subsurface rocks over thousands of meters in depth, making the finding and distinguishing them to be a significant challenge. The geophysical methods, especially the seismic and well-logging methods, are the effective ways to identify the oil and gas reservoirs and are widely used in industry. Due to the complexity of near surface and subsurface structures of new exploration targets, the geophysical methods based on advanced computation methods and physical principles are continuously proposed to cope with the emerging challenges. Thus, some new advances in theory and technology of oil and gas geophysics are summarized in this editorial material, especially focusing on the geophysical data processing, numerical simulation technology, rock physics modeling, and reservoir characterization.