The growth and decay of ocean-island volcanoes are intrinsically linked to vertical movements. While the causes for subsidence are better understood, uplift mechanisms remain enigmatic. Santa Maria Island in the Azores Archipelago is an ocean-island volcano resting on top of young lithosphere, barely 480 km away from the Mid-Atlantic Ridge. Like most other Azorean islands, Santa Maria should be experiencing subsidence. Yet, several features indicate an uplift trend instead. In this paper, we reconstruct the evolutionary history of Santa Maria with respect to the timing and magnitude of its vertical movements, using detailed field work and 40Ar/39Ar geochronology. Our investigations revealed a complex evolutionary history spanning ∼6 m.y., with subsidence up to ca. 3.5 Ma followed by uplift extending to the present day. The fact that an island located in young lithosphere experienced a pronounced uplift trend is remarkable and raises important questions concerning possible uplift mechanisms. Localized uplift in response to the tectonic regime affecting the southeastern tip of the Azores Plateau is unlikely, since the area is under transtension. Our analysis shows that the only viable mechanism able to explain the uplift is crustal thickening by basal intrusions, suggesting that intrusive processes play a significant role even on islands standing on young lithosphere, such as in the Azores.

/pdf/emergence-and-evolution-of-santa-maria-island-azores-the-5cgbbxd9ph.pdf

Emergence and evolution of Santa Maria Island (Azores)—The conundrum of uplifted islands revisited

The large low shear-wave velocity provinces (LLSVP) are thermochemical anomalies in the deep Earth's mantle, thousands of km wide and ∼1800 km high. This study explores the hypothesis that the LLSVPs are compositionally subdivided into two domains: a primordial bottom domain near the core-mantle boundary and a basaltic shallow domain that extends from 1100 to 2300 km depth. This hypothesis reconciles published observations in that it predicts that the two domains have different physical properties (bulk-sound versus shear-wave speed versus density anomalies), the transition in seismic velocities separating them is abrupt, and both domains remain seismically distinct from the ambient mantle. We here report underside reflections from the top of the LLSVP shallow domain, supporting a compositional origin. By exploring a suite of two-dimensional geodynamic models, we constrain the conditions under which well-separated “double-layered” piles with realistic geometry can persist for billions of years. Results show that long-term separation requires density differences of ∼100 kg/m3 between LLSVP materials, providing a constraint for origin and composition. The models further predict short-lived “secondary” plumelets to rise from LLSVP roofs and to entrain basaltic material that has evolved in the lower mantle. Long-lived, vigorous “primary” plumes instead rise from LLSVP margins and entrain a mix of materials, including small fractions of primordial material. These predictions are consistent with the locations of hot spots relative to LLSVPs, and address the geochemical and geochronological record of (oceanic) hot spot volcanism. The study of large-scale heterogeneity within LLSVPs has important implications for our understanding of the evolution and composition of the mantle.

https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2016GC006605

Compositional layering within the large low shear-wave velocity provinces in the lower mantle

Seismic waves that are recorded by near-surface sensors are usually disturbed by strong noise. Hence, the recorded seismic data are sometimes of poor quality; this phenomenon can be characterized as a low signal-to-noise ratio (SNR). The low SNR of the seismic data may lower the quality of many subsequent seismological analyses, such as inversion and imaging. Thus, the removal of unwanted seismic noise has significant importance. In this article, we intend to improve the SNR of many seismological datasets by developing new denoising framework that is based on an unsupervised machine-learning technique. We leverage the unsupervised learning philosophy of the autoencoding method to adaptively learn the seismic signals from the noisy observations. This could potentially enable us to better represent the true seismic-wave components. To mitigate the influence of the seismic noise on the learned features and suppress the trivial components associated with low-amplitude neurons in the hidden layer, we introduce a sparsity constraint to the autoencoder neural network. The sparse autoencoder method introduced in this article is effective in attenuating the seismic noise. More importantly, it is capable of preserving subtle features of the data, while removing the spatially incoherent random noise. We apply the proposed denoising framework to a reflection seismic image, depth-domain receiver function gather, and an earthquake stack dataset. The purpose of this study is to demonstrate the framework’s potential in real-world applications. INTRODUCTION Seismic phases from the discontinuities in the Earth’s interior contain significant constraints for high-resolution deep Earth imaging; however, they sometimes arrive as weak-amplitude waveforms (Rost and Weber, 2001; Rost and Thomas, 2002; Deuss, 2009; Saki et al., 2015; Guan and Niu, 2017, 2018; Schneider et al., 2017; Chai et al., 2018). The detection of these weak-amplitude seismic phases is sometimes challenging because of three main reasons: (1) the amplitude of these phases is very small and can be neglected easily when seen next to the amplitudes of neighboring phases that are much larger; (2) the coherency of the weak-amplitude seismic phases is seriously degraded because of insufficient array coverage and spatial sampling; and (3) the strong random background noise that is even stronger than the weak phases in amplitude makes the detection even harder. As an example of the challenges presented, the failure in detecting the weak reflection phases from mantle discontinuities could result in misunderstanding of the mineralogy or temperature properties of the Earth interior. To conquer the challenges in detecting weak seismic phases, we need to develop specific processing techniques. In earthquake seismology, in order to highlight a specific weak phase, recordings in the seismic arrays are often shifted and stacked for different slowness and back-azimuth values (Rost and Thomas, 2002). Stacking serves as one of the most widely used approaches in enhancing the energy of target signals. Shearer (1991a) stacked long-period seismograms of shallow earthquakes that were recorded from the Global Digital Seismograph Network for 5 yr and obtained a gather that shows typical arrivals clearly from the deep Earth. Morozov and Dueker (2003) investigated the effectiveness of stacking in enhancing the signals of the receiver functions. They defined a signal-to-noise ratio (SNR) metric that was based on the multichannel coherency of the signals and the incoherency of the random noise, and they showed that the stacking can significantly improve the SNR of the stacked seismic trace. However, stacking methods have some drawbacks. First, they do not necessarily remove the noise present in the signal. Second, they require a large array of seismometers. Third, they require coherency of arrivals across the array, which are not always about earthquake seismology. From this point of view, a single-channel method seems to be a better substitute for improving the SNR of seismograms (Mousavi and Langston, 2016; 2017). In the reflection seismology community, many noise attenuation methods have been proposed and implemented in field applications over the past several decades. Prediction-based methods utilize the predictive property of the seismic signal to construct a predictive filter that prevents noise. Median filters and their variants use the statistical principle to reject Gaussian white noise or impulsive noise (Mi et al., 2000; Bonar and Sacchi, 2012). The dictionary-learning-based methods adaptively learn the basis from the data to sparsify the noisy seismic data, which will in turn suppress the noise (Zhang, van der Baan, et al., 2018). These methods require experimenters to solve the dictionary-updating and sparse-coding methods and can be very 1552 Seismological Research Letters Volume 90, Number 4 July/August 2019 doi: 10.1785/0220190028 Downloaded from https://pubs.geoscienceworld.org/ssa/srl/article-pdf/90/4/1552/4790732/srl-2019028.1.pdf by Seismological Society of America, Mattie Adam on 09 July 2019 expensive, computationally speaking. Decomposition-based methods decompose the noisy data into constitutive components, so that one can easily select the components that primarily represent the signal and remove those associated with noise. This category includes singular value decomposition (SVD)-based methods (Bai et al., 2018), empirical-mode decomposition (Chen, 2016), continuous wavelet transform (Mousavi et al., 2016), morphological decomposition (Huang et al., 2017), and so on. Rank-reduction-based methods assume that seismic data have a low-rank structure (Kumar et al., 2015; Zhou et al., 2017). If the data consist of κ complex linear events, the constructed Hankel matrix of the frequency data is a matrix of rank κ (Hua, 1992). Noise will increase the rank of theHankel matrix of the data, which can be attenuated via rank reduction. Such methods include Cadzow filtering (Cadzow, 1988; Zu et al., 2017) and SVD (Vautard et al., 1992). Most of the denoising methods are largely effective in processing reflection seismic images. The applications in more general seismological datasets are seldom reported, partially because of the fact that many seismological datasets have extremely low data quality. That is, they are characterized by low SNR and poor spatial sampling. Besides, most traditional denoising algorithms are based on carefully tuned parameters to obtain satisfactory performance. These parameters are usually data dependent and require a great deal of experiential knowledge. Thus, they are not flexible enough to use in application to many real-world problems. More research efforts have been dedicated to using machine-learning methods for seismological data processing (Chen, 2018a,b; Zhang, Wang, et al., 2018; Bergen et al., 2019; Lomax et al., 2019; McBrearty et al., 2019). Recently, supervised learning (Zhu et al., 2018) has been successfully applied for denoising of the seismic signals. However, supervised methods with deep networks require very large training datasets (sometimes to an order of a billion) of clean signals and their noisy contaminated realizations. In this article, we develop a new automatic denoising framework for improving the SNR of the seismological datasets based on an unsupervised machine-learning (UML) approach; this would be the autoencoder method. We leverage the autoencoder neural network to adaptively learn the features from the raw noisy seismological datasets during the encoding process, and then we optimally represent the data using these learned features during the decoding process. To effectively suppress the random noise, we use the sparsity constraint to regularize the neurons in the hidden layer. We apply the proposed UML-based denoising framework to a group of seismological datasets, including a reflection seismic image, a receiver function gather, and an earthquake stack. We observe a very encouraging performance, which demonstrates its great potential in a wide range of applications. METHOD Unsupervised Autoencoder Method Wewill first introduce the autoencoder neural network that we use for denoising seismological datasets. Autoencoders are specific neural networks that consist of two connected parts (decoder and encoder) that try to copy their input to the output layer. Hence, they can automatically learn the main features of the data in an unsupervised manner. In this article, the network is simply a three-layer architecture with an input layer, a hidden layer, and an output layer. The encoding process in the autoencoder neural network can be expressed as follows: EQ-TARGET;temp:intralink-;df1;323;673 ξ W1x b1 ; 1 in which x is the training sample (x∈Rn), ξ is the activation function. The decoding process can be expressed as follows: EQ-TARGET;temp:intralink-;df2;323;608 x ⌢ ξ W2x b2 : 2 In equations (1) and (2), W1 is the weighting matrix between the input layer and the hidden layer; b1 is the forward bias vector; W2 is the weighting matrix between the hidden layer and output layer; b2 is the backward bias vector; and ξ is the activation function. In this study, we use the softplus function as the activation function: EQ-TARGET;temp:intralink-;df3;323;505 ξ x log 1 e : 3 Sparsity Regularized Autoencoder To mitigate the influence of the seismic noise on the learned features and suppress the trivial components associated with low-amplitude neurons in the hidden layer, we apply a sparsity constraint to the hidden layer; that is, the output or last layer of the encoder. The sparsity constraint can help dropout the extracted nontrivial features that correspond to the noise and a small value in the hidden units. It can thus highlight the most dominant features in the data—the useful signals. The sparse penalty term can be written as follows: EQ-TARGET;temp:intralink-;df4;323;335~ R p ; 4 in which R is the penalty function: EQ-TARGET;temp:

Improving the Signal‐to‐Noise Ratio of Seismological Datasets by Unsupervised Machine Learning

Global seismic data reveal little water in the mantle transition zone

(1992). 22. P. J. Mouginis-Mark et al., in Mars, H. H. Kieffer, B. M. Jakosky, C. W. Snyder, M. S. Matthews, Eds. (Univ. of Arizona Press, Tucson, 1992), p. 424; W. B. Banerdt et al., ibid., p. 249; G. Schubert et al., ibid., p. 147. 23. M. T. Zuber and D. E. Smith, J. Geophys. Res., 102, 28,673 (1997 ). 24. R. J. Phillips et al., ibid. 95, 5089 (1990). 25. J. Phipps Morgan et al., ibid. 100, 8045 (1995). 26. K. L. Tanaka et al., in Mars, H. H. Kieffer, B. M. Jakosky, C. W. Snyder, M. S. Matthews, Eds. (Univ. of Arizona Press, Tucson, 1992), p. 345. 27. S. C. Solomon and J. W. Head, J. Geophys. Res. 82, 9755 (1982). 28. D. H. Scott and K. L. Tanaka, U.S. Geol. Survey Map I-1802-A (1986). 29. R. M. C. Lopes et al., Moon Planets 22, 221 (1980); P. W. Francis and G. Wadge, J. Geophys. Res. 88, 8333 (1983). 30. M. H. Carr, Water on Mars (Oxford Univ. Press, New York, 1996). 31. G. Komatsu and V. R. Baker, J. Geophys. Res. 102, 4151 (1997 ). 32. We use the Chezy equations [R. A. Granger, Fluid Mechanics (Dover, New York, 1995)] to estimate average flow velocity; v, is given by v 5 C(Rs)1/2, where C is the Chezy coefficient, s is local slope, and R is hydraulic radius, defined as the ratio of flow crosssectional area to perimeter of the wetted channel. In SI units, C 5 (1/n)R1/6, where n is the Manning roughness coefficient, here square-root-gravity scaled from an assumed terrestrial value of 0.02. Hydraulic radius is obtained directly from MOLA data, assuming that the water level reached the paired-terraces and attempting to adjust the channel perimeter for post-flow mass wasting. A more difficult problem is estimation of slope. In principle, the apparent slope measured along the groundtrack can be projected into the downstream direction to yield true channel slope. The orbital track makes an angle of 50° with the downstream vector and true slope 5 apparent slope/ cos(50°). The presence of debris flows on the channel floor could locally modify this slope. 33. This velocity and the channel depth of 1300 m yield a Froude number [Fr 5 V/(hg)1/2, where h is channel depth and g is gravity] of 2.0. This value indicates supercritical flow. 34. The filter of the Viking Orbiter Camera nearest to the MOLA wavelength is red [0.5 mm to 0.7 mm; M. H. Carr et al., Icarus 16, 17 (1972)]. We have adopted a scaling factor of 0.8 to convert those images to 1.064 mm to be consistent with spectroscopic observations of the average martian surface in visible and infrared wavelengths [L. A. Soderblom, in Mars, H. H. Kieffer, B. M. Jakosky, C. W. Snyder, M. S. Matthews, Eds. (Univ. of Arizona Press, Tucson, 1992), p. 557]. It is not possible to obtain unbiased values of opacity using the above method, because the Viking images are not themselves corrected for the relevant atmospheric opacity. We used only Viking images that exhibited little or no obscuration. Such opacities in the Viking images may either darken or brighten them, unlike the effect in MOLA reflectivities, which is pure extinction. We estimated that average Viking images were darkened with an opacity of 0.4 and added this to the MOLA opacities. 35. M. C. Malin et al., Science 279, 1681 (1998). 36. We thank MOLA Instrument Manager R. Follas and the rest of the instrument team, and G. Cunningham, B. McAnally, and the MGS spacecraft and operation teams. We also acknowledge helpful reviews from M. Carr and an anonymous reviewer, and contributions from J. Abshire and J. Smith in instrument calibration and performance assessment, G. Neumann, G. Elman, P. Jester, and J. Schott in altimetry processing, F. Lemoine, D. Rowlands, and S. Fricke in orbit determination, and O. Aharonson, D. Brown, J. Frawley, P. Haggerty, S. Hauk, A. Ivanov, P. McGovern, C. Johnson, S. Pratt, and N. Siebert in analysis. The MOLA investigation is supported by the NASA Mars Global Surveyor Project.

The Postspinel Phase Boundary in Mg 2 SiO 4 Determined by in Situ X-ray Diffraction

Topography of upper mantle seismic discontinuities beneath the North Atlantic: The Azores, Canary and Cape Verde plumes

We are mapping the topography of upper mantle seismic discontinuities beneath the North Atlantic and surrounding regions by using precursor arrivals to PP and SS seismic waves that reflect off the seismic discontinuities. Numerous source–receiver combinations have been used in order to collect a large dataset of reflection points beneath our investigation area. We analysed over 1700 seismograms from M W > 5.8 events using array seismic methods to enhance the signal to noise ratio. The measured time lag between PP (SS) arrivals and their corresponding precursors on robust stacks are used to measure the depth of the transition zone boundaries. The reflectors' depths show a correlation between the location of known hotspots and a significantly depressed 410 km discontinuity indicating a temperature increase of 50–300 K compared to the surrounding mantle. For the 660 km discontinuity three distinct behaviours are visible: (i) normal depths beneath Greenland and at a distance of a few hundred kilometres away from known hotspots, (ii) shallower 660 km discontinuity compared with the global average value near hotspots closer to the Mid-Atlantic Ridge, and (iii) very few observations of a 660 km discontinuity at the hotspot locations. We interpret our observations as a large upwelling beneath the southern parts of our study region, possibly due to the South Atlantic convection cell. The thermal anomaly may be ponding beneath the endothermic 660 km phase transformation and likely does not extend through the top of the transition zone as a whole, except for those branches which appear as the thinner upwellings of Azores, Canaries and Cape Verde hotspots at the surface.

Topography of upper mantle seismic discontinuities beneath the North Atlantic: the Azores, Canary and Cape Verde plumes

/pdf/on-the-difficulties-of-detecting-pp-precursors-334cjya3ry.pdf

On the difficulties of detecting PP precursors

We investigate the seismic structure of the upper-mantle and mantle transition zone beneath India and Western China using PP and SS underside reflections off seismic discontinuities, which arrive as precursors to the PP and SS arrival. We use high-resolution array seismic techniques to identify precursory energy and to map lateral variations of discontinuity depths. We find deep reflections off the 410 km discontinuity (P410P and S410S) beneath Tibet, Western China and India at depths of 410–440 km and elevated underside reflections of the 410 km discontinuity at 370–390 km depth beneath the Tien Shan region and Eastern Himalayas. These reflections likely correspond to the olivine to wadsleyite phase transition. The 410 km discontinuity appears to deepen in Central and Northern Tibet. We also find reflections off the 660 km discontinuity beneath Northern China at depths between 660 and 700 km (P660P and S660S) which could be attributed to the mineral transformation of ringwoodite to magnesiowuestite and perovskite. These observations could be consistent with the presence of cold material in the middle and lower part of the mantle transition zone in this region. We also find a deeper reflector between 700 and 740 km depth beneath Tibet which cannot be explained by a depressed 660 km discontinuity. This structure could, however, be explained by the segregation of oceanic crust and the formation of a neutrally buoyant garnet-rich layer beneath the mantle transition zone, due to subduction of oceanic crust of the Tethys Ocean. For several combinations of sources and receivers we do not detect arrivals of P660P and S660S although similar combinations of sources and receivers give well-developed P660P and S660S arrivals. Our thermodynamic modelling of seismic structure for a range of compositions and mantle geotherms shows that non-observations of P660P and S660S arrivals could be caused by the dependence of underside reflection coefficients on the incidence angle of the incoming seismic waves. Apart from reflections off the 410 and 660 km discontinuities, we observe intermittent reflectors at 300 and 520 km depth. The discontinuity structure of the study region likely reflects lateral thermal and chemical variations in the upper-mantle and mantle transition zone connected to past and present subduction and mantle convection processes.

Stephan Lessing

Papers

Topography of upper mantle seismic discontinuities beneath the North Atlantic: The Azores, Canary and Cape Verde plumes

Topography of upper mantle seismic discontinuities beneath the North Atlantic: the Azores, Canary and Cape Verde plumes

On the difficulties of detecting PP precursors

Mantle transition zone structure beneath India and Western China from migration of PP and SS precursors

Evidence of absence of PP precursors