Full-waveform inversion (FWI) is a powerful method for providing a high-resolution description of the subsurface. However, the misfit function of the conventional FWI method (metric l2-norm) is usually dominated by spurious local minima owing to its nonlinearity and ill-posedness. In addition, FWI requires intensive wavefield computation to evaluate the gradient and step length. We have considered a general inversion method using a deep neural network (DNN) for the FWI problem. This deep-learning inversion method reparameterizes physical parameters using the weights of a DNN, such that the inversion amounts to reconstructing these weights. One advantage of this deep-learning inversion method is that it can serve as an iterative regularization method, benefiting from the representation of the network. Thus, it is suitable to solve ill-posed nonlinear inverse problems. Furthermore, this method possesses good computational efficiency because it only requires first-order derivatives. In addition, it can easily be accelerated by using multiple graphics processing units and central processing units, for weight updating and forward modeling. Synthetic experiments, based on the Marmousi2, 2004 BP, and a metal ore model, are used to show the numerical performance of the deep-learning inversion method. Comprehensive comparisons with a total-variation regularized FWI are presented to show the ability of our method to recover sharp boundaries. Our numerical results indicate that this deep-learning inversion approach is effective, efficient, and can capture salient features of the model.

Summary. Due to the influence of variations in landform, geophysical data acquisition is usually subsampled. Reconstruction of the seismic wavefield from subsa

We consider numerical solution methods for seismic impedance inversion problems in this paper. The inversion process is ill-posed. To tackle the ill-posedness of the problem and take the sparsity of the reflectivity function into consideration, an l ₁ norm regularization model is established. In computation, a nonmonotone gradient descent method based on Rayleigh quotient for solving the minimization model is developed. Theoretical simulations and field data applications are performed to verify the feasibility of our methods.

Image restoration refers to minimizing the degradation caused by the sensor environment, such as misfocused CCD cameras, nonuniform motions, atmospheric aerosols, and atmospheric turbulences. In image restoration problems, it is reasonable to add nonnegative constraints because of the physical meaning of the image. Then, the problem can be expressed as a quadratic programming problem with nonnegative constraints. However, in previous research, a parameterization technique is needed to reduce the problem into an unconstrained optimization problem. To avoid parameterization techniques, we apply the recently developed projected Barzilai–Borwein (PBB) method to solve this quadratic programming problem. Also, a novel approach for reducing the cost of matrix-vector multiplication is proposed when applying BB and PBB methods for atmospheric image restoration. The numerical experiments show that this method is promising for large-scale image restoration problems.

Geoscientists mainly identify subsurface geologic features using exploration-derived seismic data. Classification or segmentation of 2D/3D seismic images commonly relies on conventional deep learning methods for image recognition. However, complex reflections of seismic waves tend to form high-dimensional and multi-scale signals, making traditional convolutional neural networks (CNNs) computationally costly. Here we propose a highly efficient and resource-saving CNN architecture (SeismicPatchNet) with topological modules and multi-scale-feature fusion units for classifying seismic data, which was discovered by an automated data-driven search strategy. The storage volume of the architecture parameters (0.73 M) is only ~2.7 MB, ~0.5% of the well-known VGG-16 architecture. SeismicPatchNet predicts nearly 18 times faster than ResNet-50 and shows an overwhelming advantage in identifying Bottom Simulating Reflection (BSR), an indicator of marine gas-hydrate resources. Saliency mapping demonstrated that our architecture captured key features well. These results suggest the prospect of end-to-end interpretation of multiple seismic datasets at extremely low computational cost.

Distributed acoustic sensing (DAS) technology is, increasingly, the seismic acquisition mode of choice for its high spatial sampling rate, low cost, and nonintrusive deployability. It is being widely evaluated as an enabler of seismic monitoring for CO2 sequestration in building subsurface time-lapse images and in characterizing near-surface environments. To advance this evaluation, field seismic surveys with optical fibers have been conducted at the Containment and Monitoring Institute’s Field Research Station (CaMI.FRS) in Newell County, Alberta, Canada. In comparison to the standard geophones, optical fibers deployed in surface trenches at CaMI.FRS have recorded high-quality surface waves, rich in low frequencies and exhibiting limited spatial aliasing. These benefits have motivated us to apply the full-waveform inversion (FWI) approach to image the S-wave velocity (VS) and attenuation (quality factor QS) models at shallow site using the surface waves recorded by optical fibers. Compared to the conventional surface-wave dispersion approach, FWI can intrinsically incorporate fundamental and high-order modes and produce VS model with high spatial resolution that resolves horizontal variations. The low-frequency components below 10 Hz measured in the DAS recordings are helpful to overcome the cycle-skipping problem of FWI. Following the adjoint-state method, QS sensitivity kernel can be calculated efficiently with memory strain variables. The QS model is iteratively estimated with a new misfit function measuring root-mean-square amplitude differences, which helps to reduce the trade-off artifacts. The synthetic data obtained from the inverted models are consistent with the observed data in amplitude and phase. The inversion results provide valuable information to characterize the near-surface environments at CaMI.FRS and are expected to support seismic imaging in deeper CO2 injection zones.