| NSF Org: |
EAR Division Of Earth Sciences |
| Recipient: |
|
| Initial Amendment Date: | July 10, 2017 |
| Latest Amendment Date: | August 23, 2017 |
| Award Number: | 1725344 |
| Award Instrument: | Fellowship Award |
| Program Manager: |
Aisha Morris
armorris@nsf.gov (703)292-7081 EAR Division Of Earth Sciences GEO Directorate for Geosciences |
| Start Date: | September 1, 2017 |
| End Date: | August 31, 2019 (Estimated) |
| Total Intended Award Amount: | $87,000.00 |
| Total Awarded Amount to Date: | $174,000.00 |
| Funds Obligated to Date: |
|
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
Berkeley CA US 94702-1480 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
La Jolla CA US 92093-0225 |
| Primary Place of
Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): | Postdoctoral Fellowships |
| Primary Program Source: |
01001718DB NSF RESEARCH & RELATED ACTIVIT 01001819DB NSF RESEARCH & RELATED ACTIVIT |
| Program Reference Code(s): |
|
| Program Element Code(s): |
|
| Award Agency Code: | 4900 |
| Fund Agency Code: | 4900 |
| Assistance Listing Number(s): | 47.050 |
ABSTRACT
Dr. Christopher W. Johnson has been granted an NSF EAR Postdoctoral Fellowship to carry out research and education plans at the University of California San Diego and the University of Southern California. The objective of this project is to develop classifiers for earthquakes, microseisms, and ambient noise using parameters of the waveforms and utilize statistical learning techniques to train a computer program to identify previously undetected microseismic events in upper 1 km of the crust. The properties of the near surface require further exploration to constrain processes that generate earthquakes and characterize the temporal variations in seismic and geodetic records. The expected results from this project will advance the understanding of seismic ground motion, surface deformation, transient deformation signals, and elastic and hydrologic crustal properties for the community at large. This study provides the framework for nascent earthquake detection techniques applicable to regional seismic networks. During the investigation, Dr. Johnson will interact and train undergraduate and graduate students.
The dominant fault system along the Pacific-North American plate boundary is the San Andreas Fault (SAF) that bisects dense urban city centers along the California coast. Southeast of Los Angeles, CA the SAF bifurcates into the San Jacinto Fault Zone (SJFZ), where microseismic events occur daily and six M>5 earthquakes have ruptured in the past 35 years. Fully characterizing the SAF and SJFZ are of great importance to infrastructure planning and hazard mitigation. A recent dense array deployment of >1,100 seismometers recorded ground motions at the SJFZ for 5 weeks. The data set provides a unique opportunity to develop new techniques that characterize the shallow subsurface from 0-1 km depth and fully describe the fault zone environment from the surface to the brittle-ductile transition zone at ~18 km depth. Preliminary results from the deployment indicate >120 small events in one day of the array data, significantly more than the 15 earthquakes detected by the regional seismic network. Labeling as many features of the data as possible will aid in efficiently processing large volumes of data, thereby reducing the computational expense of microseismic earthquake detection. The detailed classification of continuous data should also produce a qualitative description of patterns in the noise and highlight episodes of coherent energy that may represent previously unexplored failure processes in the shallowest portions of the crust. This is a novel study utilizing data from the Earthscope PBO networks to further advance the understanding of plate boundary fault systems.
PUBLICATIONS PRODUCED AS A RESULT OF THIS RESEARCH
Note:
When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external
site maintained by the publisher. Some full text articles may not yet be available without a
charge during the embargo (administrative interval).
Some links on this page may take you to non-federal websites. Their policies may differ from
this site.
PROJECT OUTCOMES REPORT
Disclaimer
This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.
Seismic waveforms are rich in signals arising from multiple processes in the earth, atmosphere and oceans, and human activity. The project utilizes machine learning techniques to extract features from a seismic data set containing unexplained high-frequency emergent and impulsive signals. The data is from a temporary dense seismic array of 1,100 geophones deployed for ~30 days with sensor spacing of 10-30 m, which allows detailed analysis of coherent signals at neighboring sensors. Validation of true ground motions are assessed when signals are recorded at multiple locations and verified with a sensor previously installed at a depth of 150 m. Unsupervised machine learning models extract underlying structure in unlabeled data without any prior information. The project develops a novel training data set by extracting features from the seismic waveforms using unsupervised learning techniques, then clustering the information into groups of similar ground motions. The outcome is millions of labeled examples to train a classification model that has predictive ability for ground motions in the study area. Supervised machine learning classification models are then trained using the large numbers of labeled data to provide learning examples by iteratively modifying millions of model coefficients based on the prediction ability. The fully trained model allows the rapid analysis of large volumes (1.6Tb) of seismic waveforms and assigns a label for the type of waveform identified for each 1 second of continuous data. The results show tectonic events and different classes of non-tectonic weak ground motions as coherent signals across the array. The non-tectonic signals in the continuous seismic records occupy a significant fraction of the day. There is considerable interest in the detection of seismic tremor that represents a bridge between seismic and aseismic motions. The results demonstrate that incorporating machine learning algorithms to extract features, develop data labels, and train models can provide a new perspective on the anatomy of continuous seismic waveforms to better identify tectonic signals. Continuing research in this direction can lead to new discoveries about the full range of sources of weak ground motions, and significant improvements in the ability to detect small genuine microearthquakes and tectonic tremor.
Last Modified: 09/26/2019
Modified by: Christopher W Johnson
Please report errors in award information by writing to: awardsearch@nsf.gov.
