[1] We describe and present a new model of global subduction zone geometries, called Slab1.0. An extension of previous efforts to constrain the two-dimensional non-planar geometry of subduction zones around the focus of large earthquakes, Slab1.0 describes the detailed, non-planar, three-dimensional geometry of approximately 85% of subduction zones worldwide. While the model focuses on the detailed form of each slab from their trenches through the seismogenic zone, where it combines data sets from active source and passive seismology, it also continues to the limits of their seismic extent in the upper-mid mantle, providing a uniform approach to the definition of the entire seismically active slab geometry. Examples are shown for two well-constrained global locations; models for many other regions are available and can be freely downloaded in several formats from our new Slab1.0 website, http://on.doi.gov/d9ARbS. We describe improvements in our two-dimensional geometry constraint inversion, including the use of ‘average’ active source seismic data profiles in the shallow trench regions where data are otherwise lacking, derived from the interpolation between other active source seismic data along-strike in the same subduction zone. We include several analyses of the uncertainty and robustness of our three-dimensional interpolation methods. In addition, we use the filtered, subduction-related earthquake data sets compiled to build Slab1.0 in a reassessment of previous analyses of the deep limit of the thrust interface seismogenic zone for all subduction zones included in our global model thus far, concluding that the width of these seismogenic zones is on average 30% larger than previous studies have suggested.

https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1029/2011JB008524

Slab1.0: A three‐dimensional model of global subduction zone geometries

The main difficulty in extending seismic processing to anisotropic media is the recovery of anisotropic velocity fields from surface reflection data. We suggest carrying out velocity analysis for transversely isotropic (TI) media by inverting the dependence of P‐wave moveout velocities on the ray parameter. The inversion technique is based on the exact analytic equation for the normal‐moveout (NMO) velocity for dipping reflectors in anisotropic media. We show that P‐wave NMO velocity for dipping reflectors in homogeneous TI media with a vertical symmetry axis depends just on the zero‐dip value Vnmo(0) and a new effective parameter η that reduces to the difference between Thomsen parameters e and δ in the limit of weak anisotropy. Our inversion procedure makes it possible to obtain η and reconstruct the NMO velocity as a function of ray parameter using moveout velocities for two different dips. Moreover, Vnmo(0) and η determine not only the NMO velocity, but also long‐spread (nonhyperbolic) P‐wave moveout ...

/pdf/velocity-analysis-for-transversely-isotropic-media-45d3sqg5nv.pdf

Velocity analysis for transversely isotropic media

Subduction zones are home to the most seismically active faults on the planet. The shallow megathrust interfaces of subduction zones host Earth’s largest earthquakes and are likely the only faults capable of magnitude 9+ ruptures. Despite these facts, our knowledge of subduction zone geometry—which likely plays a key role in determining the spatial extent and ultimately the size of subduction zone earthquakes—is incomplete. We calculated the three-dimensional geometries of all seismically active global subduction zones. The resulting model, called Slab2, provides a uniform geometrical analysis of all currently subducting slabs.

https://science.sciencemag.org/content/sci/362/6410/58.full.pdf

Slab2, a comprehensive subduction zone geometry model

Terra infirma: Understanding salt tectonics

BACKGROUND The solid Earth, oceans, and atmosphere together form a complex interacting geosystem. Processes relevant to understanding Earth’s geosystem behavior range in spatial scale from the atomic to the planetary, and in temporal scale from milliseconds to billions of years. Physical, chemical, and biological processes interact and have substantial influence on this complex geosystem, and humans interact with it in ways that are increasingly consequential to the future of both the natural world and civilization as the finiteness of Earth becomes increasingly apparent and limits on available energy, mineral resources, and fresh water increasingly affect the human condition. Earth is subject to a variety of geohazards that are poorly understood, yet increasingly impactful as our exposure grows through increasing urbanization, particularly in hazard-prone areas. We have a fundamental need to develop the best possible predictive understanding of how the geosystem works, and that understanding must be informed by both the present and the deep past. This understanding will come through the analysis of increasingly large geo-datasets and from computationally intensive simulations, often connected through inverse problems. Geoscientists are faced with the challenge of extracting as much useful information as possible and gaining new insights from these data, simulations, and the interplay between the two. Techniques from the rapidly evolving field of machine learning (ML) will play a key role in this effort. ADVANCES The confluence of ultrafast computers with large memory, rapid progress in ML algorithms, and the ready availability of large datasets place geoscience at the threshold of dramatic progress. We anticipate that this progress will come from the application of ML across three categories of research effort: (i) automation to perform a complex prediction task that cannot easily be described by a set of explicit commands; (ii) modeling and inverse problems to create a representation that approximates numerical simulations or captures relationships; and (iii) discovery to reveal new and often unanticipated patterns, structures, or relationships. Examples of automation include geologic mapping using remote-sensing data, characterizing the topology of fracture systems to model subsurface transport, and classifying volcanic ash particles to infer eruptive mechanism. Examples of modeling include approximating the viscoelastic response for complex rheology, determining wave speed models directly from tomographic data, and classifying diverse seismic events. Examples of discovery include predicting laboratory slip events using observations of acoustic emissions, detecting weak earthquake signals using similarity search, and determining the connectivity of subsurface reservoirs using groundwater tracer observations. OUTLOOK The use of ML in solid Earth geosciences is growing rapidly, but is still in its early stages and making uneven progress. Much remains to be done with existing datasets from long-standing data sources, which in many cases are largely unexplored. Newer, unconventional data sources such as light detection and ranging (LiDAR), fiber-optic sensing, and crowd-sourced measurements may demand new approaches through both the volume and the character of information that they present. Practical steps could accelerate and broaden the use of ML in the geosciences. Wider adoption of open-science principles such as open source code, open data, and open access will better position the solid Earth community to take advantage of rapid developments in ML and artificial intelligence. Benchmark datasets and challenge problems have played an important role in driving progress in artificial intelligence research by enabling rigorous performance comparison and could play a similar role in the geosciences. Testing on high-quality datasets produces better models, and benchmark datasets make these data widely available to the research community. They also help recruit expertise from allied disciplines. Close collaboration between geoscientists and ML researchers will aid in making quick progress in ML geoscience applications. Extracting maximum value from geoscientific data will require new approaches for combining data-driven methods, physical modeling, and algorithms capable of learning with limited, weak, or biased labels. Funding opportunities that target the intersection of these disciplines, as well as a greater component of data science and ML education in the geosciences, could help bring this effort to fruition. The list of author affiliations is available in the full article online.

Machine learning for data-driven discovery in solid Earth geoscience

The problem of estimating relative time (or depth) shifts between two seismic images is ubiquitous in seismic data processing. This problem is especially difficult where shifts are large and vary rapidly with time and space, and where images are contaminated with noise or for other reasons are not shifted versions of one another. A new solution to this problem requires only simple extensions of a classic dynamic time warping algorithm for speech recognition. A key component of that classic algorithm is a nonlinear accumulation of alignment errors. By applying the same nonlinear accumulator repeatedly in all directions along all sampled axes of a multidimensional image, I obtain a new and effective method for dynamic image warping (DIW). In tests where known shifts vary rapidly, this new method is more accurate than methods based on crosscorrelations of windowed images. DIW also aligns seismic reflectors well in examples where shifts are unknown, for images with differences not limited to time shifts.

Dynamic warping of seismic images

In reflection seismology, full-waveform inversion (FWI) can generate high-wavenumber subsurface velocity models but often suffers from an objective function with local minima caused mainly by the absence of low frequencies in seismograms. These local minima cause cycle skipping when the low-wavenumber component in the initial velocity model for FWI is far from the true model. To avoid cycle skipping, we discovered a new wave-equation reflection traveltime inversion (WERTI) to update the low-wavenumber component of the velocity model, while using FWI to only update high-wavenumber details of the model. We implemented the low- and high-wavenumber inversions in an alternating way. In WERTI, we used dynamic image warping (DIW) to estimate the time shifts between recorded data and synthetic data. When compared with correlation-based techniques often used in traveltime estimation, DIW can avoid cycle skipping and estimate the time shifts accurately, even when shifts vary rapidly. Hence, by minimizing tr...

Wave-equation reflection traveltime inversion with dynamic warping and full-waveform inversion

Fault interpretation enhances our understanding of complex geologic structures and stratigraphy apparent in 3D seismic images. Common steps in this interpretation include image processing to highlight faults, the construction of fault surfaces, and estimation of fault throws. Although all three of these steps have been automated to some extent by others, fault interpretation today typically requires significant manual effort, suggesting that further improvements in automatic methods are feasible and worthwhile. I first used an efficient algorithm to compute images of fault likelihoods, strikes, and dips from a 3D seismic image. From these three fault images, I then automatically extracted fault surfaces as meshes of quadrilaterals that coincide with ridges of fault likelihood. A quadrilateral mesh is a simple data structure alongside which one can easily gather samples of the 3D seismic image. I automatically estimated fault throws by minimizing differences in values of samples gathered from oppos...

Methods to compute fault images, extract fault surfaces, and estimate fault throws from 3D seismic images

Stable explicit filters for depth extrapolation of seismic wavefields may be derived through a modification of the conventional Taylor series method. The modified Taylor series method described here yields extrapolators with maximally-flat amplitude spectra in their passband, while ensuring that no spectral components in the wavefield are amplified by any factor greater than one. The price for stability is increased phase error. For low normalized frequencies, implicit extrapolators are more accurate than the 39-coefficient stable explicit extrapolator described here. However, the small spatial sampling intervals required to obtain high phase accuracy in implicit extrapolation imply that this accuracy is rarely achieved in practice. Over the wide range of normalized frequencies likely to be encountered in practice, stable explicit extrapolators outperform implicit ones. The method presented here for deriving stable explicit extrapolators is in no formal sense optimal. It is only guaranteed to yield stable extrapolators. In my limited experience with alternative methods for designing stable extrapolators, the method presented here produced the least phase error while ensuring stability. Nevertheless, a simple method for designing optimal (in some sense) stable explicit extrapolators would be preferred over the method presented here. 7 refs., 15 figs.

/pdf/stable-explicit-depth-extrapolation-of-seismic-wavefields-50r3c3fy9p.pdf

Stable explicit depth extrapolation of seismic wavefields

Numerous methods have been proposed to automatically extract fault surfaces from 3D seismic images, and those surfaces are often represented by meshes of triangles or quadrilaterals. However, extraction of intersecting faults is still a difficult problem that is not well addressed. Moreover, mesh data structures are more complex than the arrays used to represent seismic images, and they are more complex than necessary for subsequent processing tasks, such as that of automatically estimating fault slip vectors. We have represented a fault surface using a simpler linked data structure, in which each sample of a fault corresponded to exactly one seismic image sample, and the fault samples were linked above and below in the fault dip directions, and left and right in the fault strike directions. This linked data structure was easy to exchange between computers and facilitated subsequent image processing for faults. We then developed a method to construct complete fault surfaces without holes using thi...

Dave Hale

Papers

Dynamic warping of seismic images

Wave-equation reflection traveltime inversion with dynamic warping and full-waveform inversion

Methods to compute fault images, extract fault surfaces, and estimate fault throws from 3D seismic images

Stable explicit depth extrapolation of seismic wavefields

3D seismic image processing for faults