A few important issues for performing nonlinear refraction traveltime tomography have been identified. They include the accuracy of the traveltime and raypath calculations for refraction, the physical information in the refraction traveltime curves, and the characteristics of the refraction traveltime errors. Consequently, we develop a shortest path ray-tracing method with an optimized node distribution that can calculate refraction traveltimes and raypaths accurately in any velocity model. We find that structure ambiguity caused by short and long rays in the seismic refraction method may influence the inversion solution significantly. Therefore, we pose a nonlinear inverse problem that explicitly minimizes the misfits of the average slownesses (ratios of traveltimes to the corresponding ray lengths) and the apparent slownesses (derivatives of traveltimes with respect to distance). As a result, we enhance the resolution as well as the convergence speed. To regularize our inverse problem, we use the Tikhonov method to avoid solving an ill-posed inverse problem. Errors in refraction traveltimes are characterized in terms of a common-shot error, a constant deviation for recordings from the same shot, and a relative traveltime-gradient error with zero mean with respect to the true gradient of the traveltime curve. Therefore, we measure the uncertainty of our tomography solution using a nonlinear Monte Carlo approach and compute the posterior model covariance associated with two different types of random data vectors and one random model vector. The nonlinear uncertainty analysis indicates that the resolution of a tomography solution may not correspond to the ray coverage. We apply this tomography technique to image the shallow velocity structure at a coastal site near Boston, Massachusetts. The results are consistent with a subsequent drilling survey.

Intratumoral immunotherapy is an emerging modality for the treatment of solid tumors. Toll-like receptor (TLR) agonists have shown promise for eliciting immune responses, but systemic administration often results in the development of adverse side effects. Herein, we investigate whether localized delivery of the TLR agonist, resiquimod (R848), via platelet membrane-coated nanoparticles (PNP-R848) elicits antitumor responses. The membrane coating provides a means of enhancing interactions with the tumor microenvironment, thereby maximizing the activity of R848. Intratumoral administration of PNP-R848 strongly enhances local immune activation and leads to complete tumor regression in a colorectal tumor model, while providing protection against repeated tumor re-challenges. Moreover, treatment of an aggressive breast cancer model with intratumoral PNP-R848 delays tumor growth and inhibits lung metastasis. Our findings highlight the promise of locally delivering immunostimulatory payloads using biomimetic nanocarriers, which possess advantages such as enhanced biocompatibility and natural targeting affinities.

We have developed rapid 3-D dc resistivity forward modeling and inversion algorithms that use conjugate gradient relaxation techniques. In the forward network modeling calculation, an incomplete Cholesky decomposition for preconditioning and sparse matrix routines combine to produce a fast and efficient algorithm (approximately 2 minutes CPU time on a Sun SPARCstation 2 for 50 X 50 X 20 blocks). The side and bottom boundary conditions are scaled impedance conditions that take into account the local current flow at the boundaries as a result of any configuration of current sources.For the inversion, conjugate gradient relaxation is used to solve the maximum likelihood inverse equations. Since conjugate gradient techniques only require the results of the sensitivity matrix A or its transpose A T multiplying a vector, we are able to bypass the actual computation of the sensitivity matrix and the inversion of A T A, thus greatly decreasing the time needed to do 3-D inversions.We demonstrate 3-D resistivity tomographic imaging using pole-pole resistivity data collected during an experiment for a leakage monitoring system near evaporation ponds at the Mojave Generating Station in Laughlin, Nevada.

An immediate report of the source focal mechanism with full automation after a destructive earthquake is crucial for timely characterizing the faulting geometry, evaluating the stress perturbation, and assessing the aftershock patterns. Advanced technologies such as Artificial Intelligence (AI) has been introduced to solve various problems in real-time seismology, but the real-time source focal mechanism is still a challenge. Here we propose a novel deep learning method namely Focal Mechanism Network (FMNet) to address this problem. The FMNet trained with 787,320 synthetic samples successfully estimates the focal mechanisms of four 2019 Ridgecrest earthquakes with magnitude larger than Mw 5.4. The network learns the global waveform characteristics from theoretical data, thereby allowing the extensive applications of the proposed method to regions of potential seismic hazards with or without historical earthquake data. After receiving data, the network takes less than two hundred milliseconds for predicting the source focal mechanism reliably on a single CPU.

One of the key objectives in geophysics is to characterize the subsurface through the process of analyzing and interpreting geophysical field data that are typically acquired at the surface. Data-driven deep learning methods have enormous potential for accelerating and simplifying the process but also face many challenges, including poor generalizability, weak interpretability, and physical inconsistency. We present three strategies for imposing domain knowledge constraints on deep neural networks (DNNs) to help address these challenges. The first strategy is to integrate constraints into data by generating synthetic training datasets through geological and geophysical forward modeling and properly encoding prior knowledge as part of the input fed into the DNNs. The second strategy is to design nontrainable custom layers of physical operators and preconditioners in the DNN architecture to modify or shape feature maps calculated within the network to make them consistent with the prior knowledge. The final strategy is to implement prior geological information and geophysical laws as regularization terms in loss functions for training the DNNs. We discuss the implementation of these strategies in detail and demonstrate their effectiveness by applying them to geophysical data processing, imaging, interpretation, and subsurface model building.

The accurate and automated determination of small earthquake (M_L < 3.0) locations is still a challenging endeavor due to low signal-to-noise ratio in data. However, such information is critical for monitoring seismic activity and assessing potential hazards. In particular, earthquakes caused by industrial injection have become a public concern, and regulators need a solid capability for estimating small earthquakes that may trigger the action requirements for operators to follow in real time. In this study, we develop a fully convolutional network and locate earthquakes induced during oil and gas operations in Oklahoma with data from 30 network stations. The network is trained by 1,013 cataloged events (M_L ≥ 3.0) as base data along with augmented data accounting for smaller events (3.0 > M_L ≥ 0.5), and the output is a 3D volume of the event location probability in the Earth. The prediction results suggest that the mean epicenter errors of the testing events (M_L ≥ 1.5) vary from 3.7 to 6.4 km, meeting the need of the traffic light system in Oklahoma, but smaller events (M_L = 1.0, 0.5) show errors larger than 11 km. Synthetic tests suggest that the accuracy of ground truth from catalog affects the prediction results. Correct ground truth leads to a mean epicenter error of 2.0 km in predictions, but adding a mean location error of 6.3 km to ground truth causes a mean epicenter error of 4.9 km. The automated system is able to distinguish certain interfered events or events out of the monitoring zone based on the output probability estimate. It requires approximately one hundredth of a second to locate an event without the need for any velocity model or human interference.