High-Resolution Seismic Tomography of Long Beach, CA, Using Machine Learning

Date: 4/24/2019

Time: 02:30 PM

Room: Elliott Bay

We use a machine learning-based tomography method to obtain high-resolution subsurface geophysical structure in Long Beach, CA, from seismic noise recorded on a “large-N” array with 5204 geophones (~13.5 million travel times). This method, called locally sparse travel time tomography (LST), exploits the dense sampling obtained by ambient noise processing on large arrays by learning dictionaries of local, or small-scale, geophysical features from the data. These features are obtained in LST using dictionary learning, an unsupervised machine learning method. Without the need for smoothing constraints often required in conventional tomography, dictionary learning in LST models both discontinuous and smooth slowness features and helps provide high-resolution slowness estimates as permitted by the data. Using LST, we obtain a high-resolution 1 Hz Rayleigh wave phase speed map of Long Beach. Among the geophysical features shown in the map, the important Silverado aquifer is well isolated relative to previous surface wave tomography studies. The 1 Hz Rayleigh wave sensitivity depth range (~100-500 m) is occupied by Pleistocene and Holocene deposits, which contain most of the ground water resources in the area. Specifically, the Silverado formation (Lower Pleistocene age, 300-580 ka), accounts for nearly 90% of the total ground water extraction in the region considered here. The Silverado unit is characterized by relatively high-density and high-velocity coarse-grained sediments, which result in high phase-speed anomalies detectable by the method. Our results show promise for LST in obtaining detailed geophysical structure in travel time tomography studies. In this particular application, we have potentially further characterized the structure of the important Silverado aquifer in Long Beach, CA.

Presenting Author: Kim B. Olsen

Authors