skip to main content

Explore Research Products in the NSF-PAR

  • NSF Public Access
  • Search Results
  • Probabilistic inversion of seafloor compliance for oceanic crustal shear velocity structure using mixture density neural networks
Title: Probabilistic inversion of seafloor compliance for oceanic crustal shear velocity structure using mixture density neural networks
SUMMARY Measurements of various physical properties of oceanic sediment and crustal structures provide insight into a number of geological and geophysical processes. In particular, knowledge of the shear wave velocity (VS) structure of marine sediments and oceanic crust has wide ranging implications from geotechnical engineering projects to seismic mantle tomography studies. In this study, we propose a novel approach to nonlinearly invert compliance signals recorded by colocated ocean-bottom seismometers and high-sample-rate pressure gauges for shallow oceanic shear wave velocity structure. The inversion method is based on a type of machine learning neural network known as a mixture density neural network (MDN). We demonstrate the effectiveness of the MDN method on synthetic models with a fixed deployment depth of 2015 m and show that among 30 000 test models, the inverted shear wave velocity profiles achieve an average error of 0.025 km s−1. We then apply the method to observed data recorded by a broad-band ocean-bottom station in the Lau basin, for which a VS profile was estimated using Monte Carlo sampling methods. Using the mixture density network approach, we validate the method by showing that our VS profile is in excellent agreement with the previous result. Finally, we argue that the mixture density network approach to compliance inversion is advantageous over other compliance inversion methods because it is faster and allows for standardized measurements.  more » « less
Award ID(s):
1658214
NSF-PAR ID:
10333930
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
Geophysical Journal International
Volume:
227
Issue:
3
ISSN:
0956-540X
Page Range / eLocation ID:
1879 to 1892
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ( , Journal of Geophysical Research: Solid Earth)
    Abstract

    We present models of compressional and shear velocity structure of the oceanic sediments and upper crust surrounding the Hawaiian islands. The models were derived from analysis of seafloor compliance data and measurements of Ps converted phases originating at the sediment‐bedrock interface. These data were estimated from continuous broadband ocean bottom seismometer acceleration and pressure records collected during the Plume‐Lithosphere Undersea Mantle Experiment, an amphibious array of wideband and broadband instruments with an aperture of over 1,000 km. Our images result from a joint inversion of compliance and Ps delay data using a nonlinear inversion scheme whereby deviation from a priori constraints is minimized. In our final model, sediment thickness increases from 50 m at distal sites to over 1.5 km immediately adjacent to the islands. The sedimentary shear velocity profiles exhibit large regional variations. While sedimentary structure accounts for the majority of the compliance signal, we infer variations in shear velocity in the uppermost bedrock on the order of ±5%. We also require relatively high values of Poisson's ratio in the uppermost crust. Lower crustal velocities are generally seen to the north and west of the islands but do not appear well correlated with the Hawaiian Swell bathymetry. A region of strong low velocity anomalies to the northeast of Hawaii may be associated with the Molokai fracture zone.

     
    more » « less
  2. ; ; ; ( , Geophysical Journal International)
    SUMMARY

    The southern boundary of the Cayman Trough in the Caribbean is marked by the Swan Islands transform fault (SITF), which also represents the ocean–continent transition of the Honduras continental margin. This is one of the few places globally where a transform continental margin is currently active. The CAYSEIS experiment acquired an ∼165-km-long seismic refraction and gravity profile (P01) running across this transform margin, and along the ridge-axis of the Mid-Cayman Spreading Centre (MCSC) to the north. This profile reveals not only the crustal structure of an actively evolving transform continental margin, that juxtaposes Mesozoic-age continental crust to the south against zero-age ultraslow spread oceanic crust to the north, but also the nature of the crust and uppermost mantle beneath the ridge-transform intersection (RTI). The traveltimes of arrivals recorded by ocean-bottom seismographs (OBSs) deployed along-profile have been inverse and forward modelled, in combination with gravity modelling, to reveal an ∼25-km-thick continental crust that has been continuously thinned over a distance of ∼65 km to ∼10 km adjacent to the SITF, where it is juxtaposed against ∼3–4-km-thick oceanic crust. This thinning is primarily accommodated within the lower crust. Since Moho reflections are only sparsely observed, and, even then, only by a few OBSs located on the continental margin, the 7.5 km s–1 velocity contour is used as a proxy to locate the crust–mantle boundary along-profile. Along the MCSC, the crust–mantle boundary appears to be a transition zone, at least at the seismic wavelengths used for CAYSEIS data acquisition. Although the traveltime inversion only directly constrains the upper crust at the SITF, gravity modelling suggests that it is underlain by a higher density (>3000 kg m–3) region spanning the width (∼15 km) of its bathymetric expression, that may reflect a broad region of metasomatism, mantle hydration or melt-depleted lithospheric mantle. At the MCSC ridge-axis to the north, the oceanic crust appears to be forming in zones, where each zone is defined by the volume of its magma supply. The ridge tip adjacent to the SITF is currently in a magma rich phase of accretion. However, there is no evidence for melt leakage into the transform zone. The width and crustal structure of the SITF suggests its motion is currently predominantly orthogonal to spreading. Comparison to CAYSEIS Profile P04, located to the west and running across-margin and through 10 Ma MCSC oceanic crust, suggests that, at about this time, motion along the SITF had a left-lateral transtensional component, that accounts for its apparently broad seabed appearance westwards.

     
    more » « less
  3. ; ; ( , Journal of Geophysical Research: Solid Earth)
    Abstract

    We use Eikonal tomography to derive phase and group velocities of surface waves for the plate boundary region in Southern California. Seismic noise data in the period range 2 and 20 s recorded in year 2014 by 346 stations with ~1‐ to 30‐km station spacing are analyzed. Rayleigh and Love wave phase travel times are measured using vertical‐vertical and transverse‐transverse noise cross correlations, and group travel times are derived from the phase measurements. Using the Eikonal equation for each location and period, isotropic phase and group velocities and 2‐psi azimuthal anisotropy are determined statistically with measurements from different virtual sources. Starting with the SCEC Community Velocity Model, the observed 2.5‐ to 16‐s isotropic phase and group dispersion curves are jointly inverted on a 0.05° × 0.05° grid to obtain local 1‐D piecewise shear wave velocity (Vs) models. Compared to the starting model, the final results have generally lowerVsin the shallow crust (top 3–10 km), particularly in areas such as basins and fault zones. The results also show clear velocity contrasts across the San Andreas, San Jacinto, Elsinore, and Garlock Faults and suggest that the San Andreas Fault southeast of San Gorgonio Pass is dipping to the northeast. Investigation of the nonuniqueness of the 1‐DVsinversion suggests that imaging the top 3‐kmVsstructure requires either shorter period (≤2 s) surface wave dispersion measurements or other types of data set such as Rayleigh wave ellipticity.

     
    more » « less
  4. ( , Geochemistry, Geophysics, Geosystems)
    Abstract

    I present a 3‐D isotropic shear wave velocity model of the crust and uppermost mantle beneath the Alaska‐Aleutian subduction zone offshore of the Alaska Peninsula, based on seismic data recorded by the Alaska Amphibious Community Seismic Experiment (AACSE) array and some other networks. The model derives from Rayleigh wave phase speed measurements extracted from ambient seismic noise. A new three‐station interferometry (Zhang et al., 2020) approach is applied to improve the data coverage of ambient noise surface waves. Based on the ambient noise Rayleigh wave dispersion data, a Bayesian Monte Carlo inversion is performed to produce the shear wave velocity model. There are several prominent structures captured by the model, including: (1) The major sedimentary basins across the study region are identified by model. (2) Crustal thickness estimates are related with the geological structures. (3) The imaged slab edge is consistent with both the Slab 2.0 model (Hayes et al., 2018) and earthquake locations. And lots of geological and tectonic features related to subduction zone are captured, including the serpentinized forearc and partial melting zone beneath the Aleutian arc volcanoes. (4) Near the Shumagin gap, reduction in Vs is observed at the uppermost part of the incoming Pacific plate, consistent with the active source study of Shillington et al. (2015). The Vs reduction reflects hydration of the oceanic plate which could be related to local seismicity variation.

     
    more » « less
  5. ; ; ; ; ; ( , Geophysical Journal International)
    SUMMARY Ocean bottom distributed acoustic sensing (OBDAS) is emerging as a new measurement method providing dense, high-fidelity and broad-band seismic observations from fibre-optic cables deployed offshore. In this study, we focus on 35.7 km of a linear telecommunication cable located offshore the Sanriku region, Japan, and apply seismic interferometry to obtain a high-resolution 2-D shear wave velocity (VS) model below the cable. We first show that the processing steps applied to 13 d of continuous data prior to computing cross-correlation functions (CCFs) impact the modal content of surface waves. Continuous data pre-processed with 1-bit normalization allow us to retrieve dispersion images with high Scholte-wave energy between 0.5 and 5 Hz, whereas spatial aliasing dominates dispersion images above 3 Hz for non-1-bit CCFs. Moreover, the number of receiver channels considered to compute dispersion images also greatly affects the resolution of extracted surface-wave modes. To better understand the remarkably rich modal nature of OBDAS data (i.e. up to 30 higher modes in some regions), we simulate Scholte-wave dispersion curves for stepwise linear VS gradient media. For soft marine sediments, simulations confirm that a large number of modes can be generated in gradient media. Based on pre-processing and theoretical considerations, we extract surface wave dispersion curves from 1-bit CCFs spanning over 400 channels (i.e. ∼2 km) along the array and invert them to image the subsurface. The 2-D velocity profile generally exhibits slow shear wave velocities near the ocean floor that gradually increase with depth. Lateral variations are also observed. Flat bathymetry regions, where sediments tend to accumulate, reveal a larger number of Scholte-wave modes and lower shallow velocity layers than regions with steeper bathymetry. We also compare and discuss the velocity model with that from a previous study and finally discuss the combined effect of bathymetry and shallow VS layers on earthquake wavefields. Our results provide new constraints on the shallow submarine structure in the area and further demonstrate the potential of OBDAS for high-resolution offshore geophysical prospecting. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email