SlideShare a Scribd company logo
1 of 11
Download to read offline
!
!
!
!
!
Summer Studentships 2013-2014
!
Final Project Report
!
!
!
!
!
!!
!
!
!
!!!!
!
!
!
!
!
!
!
!
!
!
Project Details
Student Name: Roydon Nutsford
Name of supervisor(s): Kasper van Wijk
Supervisor department: Physics
Project title:
!Seismic Investigation of the Neal Hot Springs Geothermal Area
Supervisor letter included: Yes ☐ No ☐
Page | !1
!
!
!
!
!
!
Brief statement on how the studentship has contributed to your career development
(1 page limit):
!
!!!Participation in this studentship has allowed me to further improve my skills in the field of geophysics. I have gained
a true appreciation for the purpose of research and how it can make a real impact in the field, and on a career. Part
of the studentship involved learning to write Python code from the ground up. Understanding this still emerging
language in the Geophysics work force today will give me a cutting edge. The ability to use computers to analyse data
will propel me in a career were staying current is important in ever changing fields.
!Through the proceedings of the research I was exposed to real earthquake databases which taught me to read,
comprehend and make use of large data networks from all over the world. This contributes experience to an area of
key skills which will be vital in the work place.
Finally the studentship program has enforced the universal requirement to present scientific data and findings in an
interesting, informative and appropriate way for readers to best understand.
!
!
!
!
!
!
!
Research abstract – not more than 250 words:
!
!!!The aim of the research project is to detect small local earthquakes which are potentially covered or
disguised by background noise. The testing location is Neal Hot Springs, an active geothermal site that
supplies a 23MW geothermal electric power plant near Vale, Oregon.
!In May 2011, 10 temporary seismic stations were set up to record the earth’s surface movements over 20
months. Numerical cross correlation is used to identify similar wave forms. The goal is to locate all of
the local earthquakes, and then differentiate between geothermal seismic activity and natural seismic
activity in the area.
Python code was used to scan seismic data against a nearby ‘template’ earthquake event found from a
previous study (Shaltry, 2013) by Daniel Shaltry, a student at Boise State University. The program
returned coefficients which define the similarity of the waveforms. The method relies on earthquake
pressure, Love and Rayleigh waves to produce similar waveforms, when the earthquakes originate from
similar locations. With three earthquakes already visually identified by Shaltry, the program spotted 9
more. These can be further studied to deduce whether they are a product of surrounding local active
faults, or whether they are the result of the geothermal exploration.
!!
!
!
!
!
!
Page | !2
!
!
!
Summary of research and its significance (1 page limit):
!
!
!
The Geophysics group from Boise State University set up and recorded seismic data in the Neal Hot
Springs area, near Vale, Oregon. Colwell (2012) From their study they identified three local
earthquakes. In this research I used computer analysis to gain higher resolution from the data in order to
locate more earthquakes. Extracting the data from the Incorporated Research Institution for Seismology
(IRIS) database, I used ObSpy toolbox to compare the earthquake waveforms against known seismic
events within the local vicinity, found in the previous study. The hypothesis that earthquakes which
originate in close proximity to one another will produce similar wave forms was tested by scanning the
data for high cross correlation values using Austin Holland xcor code. Holland (2013) If the cross
correlation coefficient was 0.7 or higher, and occurred on at least 5 stations I classified it as an event.
!
With this in mind the goal was to locate more small earthquakes to interpret the activity of local faults
and assess whether the Neal Hot Spring or the power generation plant increases local seismicity. The
correlation code does not match with large earthquakes at large proximity due to a combination of
different earthquake mechanisms and earth lateral variability deforming the waveform to a state where
the coefficient is not high enough. The significance of this study is testing the previous results with new
technology, and trying to increase the ability to distinguish small earthquakes from ambient noise. The
cross correlation process allows the waveforms to be evaluated, giving information about the subsurface
geology in the area, and how this earth material affects earthquake waveforms. Further research
involves studying the distance between stations, and how much the waveform is deformed in that
period. In particular, the effect of high ground water content, in such cases as a geothermal region.
!
!
!
!
!
!
!
Research report - Aims, methods, results & discussion (5 page limit):
!
!
!
Aim:
To test the previous hypothesis with a new technology/methodology and determine the local seismic
activity of the Neal Hot Springs region with respect to earthquakes originating from local faults and to
investigate the waveform variability of earthquakes in, and passing through the area.
!
!
!
!
!
Page | !3
!
!
Method:
For this research project the raw data was produced by students of Boise State University, under the
supervision of Dr Van Wijk, and consisted of 10 stations, each station with a L-22 seismometer, data
acquisition system, GPS, solar panel and battery. The stations ran continuously from May 2011 until
November 2012 and recorded vertical, northward and eastward propagation at a rate of 250 samples per
second. This data was sent to the Incorporated Research Institution for Seismology (IRIS) database. This
data was then sourced and written into a python 2.6.7 code, with the ObSpy 0.9 toolbox handling the
stream data. Boise State University student Daniel Shaltry found three low magnitude earthquakes of
close proximity to the stations. I used these earthquakes as template events. I used a code to extract the
first 5 seconds of each of these earthquakes, capturing the P and S wave arrivals. The next section of the
code looped through the time between May 2011 and November 2012. Xcorr, a cross correlation code
written by Austin Holland was then implemented to compare the template events with the trace data.
Each individual template was correlated with each individual station using parallel computing. An
acceptable correlation coefficient of 0.7 was required from five or more stations to be considered an
earthquake event. The high correlations occurred where the trace data waveforms matched those of
template earthquakes closely, meaning they are of the same origin, and therefore allowing stronger
interpretations to be drawn.
!
Results.
The raw data was retrieved from IRIS and the three earthquakes spotted by Daniel Shaltry were found.
The raw data contained 10 traces which recorded for 20 months. Each trace corresponded to a recording
station. A small script of python code was required to trim the data down to a short template. The
template captures the p-wave and s-wave arrival times of the most representative trace in, along with
the first few seconds of vigorous shaking.
!
!
Original template
!
(figure 1): -Raw data showing first 6 seconds of an earthquake waveform. The wave arrivals are clear but
there is strong noise.
!
Page | !4
!
The Original template has discrete wave arrivals; however, the true waveform is hidden under
considerable noise. The waveform was smoothed by reducing the noise through a series of filters, which
cleaned the raw data as seen in figure 2 below. This made available a manageable template for the
algorithm. The correlation code acting on the cleaned data was able to gain consistent correlation
coefficients of a higher value. To maximise the chance of finding earthquakes I created filtered
templates from all three known waveforms.
!
Filtered template - Earthquake 1
!
(figure 2): Filtered template of waveform 1, ready to be cross-correlated against the data set. Displays a
much smoother representation of the waveform.
!
!
!
!
!
Filtered template – Earthquake 2
!
(figure 3): Filtered template of waveform 2, ready to be cross-correlated against the data set. Using
multiple waveforms from within the area will increase the potential hits with the code.
!
Page | !5
!
!
Filtered template – Earthquake 3
!
(figure 3): Filtered template of waveform 3, ready to be cross-correlated against the data set.
!
!
The above templates were used to scan the data from the 1st May 2011 till the 1st November 2011. In
order to obtain consistency, the same filters applied to the templates were also transferred to the raw
data as the correlation code requires two parameters to control the output it creates. These can be set
to alert when earthquakes of any required similarity are found. In my program I have defined an event as
a correlation value of 0.7, across at least 5 stations. This allows only similar waveforms to be alerted,
and nearly eliminates noise as correlations are required over five stations. The program loops over two
minute intervals. This opens a window were correlations can occur at any time. To minimise the
probability of random correlations meeting the 5 station requirement, I set the required correlation at
0.7, which is a high waveform similarity which does not occur commonly. 

!
Known earthquake events were picked up by the code during testing. This proved the proficiency of the
code. From the XN data set, Figure 4 displays the 3rd earthquake, including the data from all 10 station
traces, showing distinctive earthquake waveforms. This earthquake was approximately 13km away. This
can be estimated through the wave arrival times at different station locations.
!
!
!
!
!
!
!
!
!
Page | !6
3rd known event across all 10 stations
!
(Figure 4) Two minute loop captured by Python code, showing the arrival times of earthquake 3 at each
station. The magnitude and arrival time vary, however the waveform remains relatively consistent. This
event was picked up on 8 stations. Station 10 recorded a 1, as it is the direct trace of template event.
Two of the remaining stations recorded above 0.9, whilst 5 others recorded above 0.8 in the correlation
code.
!
!
Page | !7
!
Event detection count
!
(Figure 5) displays the times of each event detected by the Python code.
!
!
!
!
!
!
!
!
Discussion:
!
The data produced 9 potential earthquake events over four months. The parameters imposed on the
code were strictly allowing times only with a high probability of containing an earthquake to be
detected. There are no previously sighted events in this period and therefore the nine potential events
show that there could in fact be higher seismic activity in the area than previously expected. As the
events have not been noticed by the previous study, it indicates the earthquakes are small and not
clearly visible to the naked eye.
Month number of events Loop times
June 5 2011-06-06T12:46
2011-06-06T21:20
2011-06-17T18:12
2011-06-17T18:16
2011-06-18T06:16
July 1 2011-08-14T00:00
August 0
September 3 2011-09-25T23:22
2011-09-25T23:24
2011-09-26T00:36
Page | !8
The code has not yet been passed over the second period of data due to various problems. The standard
deviation algorithm was the biggest slowdown in the code as it used up large amounts of computer
memory. This forced the
!
!
data to be feed into the code in short two minute intervals, reducing the efficiency resulting in slow
data processing. Another challenge was caused by the data being sourced from Washington, which
occasionally took too long causing the code to crash. This made running any extended duration of data
infeasible.
Future work is to continue processing the data over the entire recording period. This will give data pre
and post the operation of the hydrothermal power station. Understanding the seismic activity will allow
conclusions to be drawn with respect to the production of artificial seismic due to the activities of the
power station. Further investigation is required to finalise if the found events are in fact earthquakes as
it is possible that random noise has occurred at five stations within a two minute loop. A manual check
will not confirm the event as it is likely to be unreadable by the naked eye. With an entire data set, the
location of individual events can be triangulated to pin point the origin of the event. This will provide
evidence as to where in the region has been most active, and raise the question of why.
Another aspect I would like to carry further is the waveform deformation with distance through the
different surface composition. This can be done through analysing the correlation coefficient at different
linear distances from the source. If distinct changes occur it could even have relevance to fault mapping
or structural geologic mapping.
!
.
!
!
!
!
Conclusion
!
The summer project has allowed me to learn python code, and more about seismology. I have applied a
complex code to real world data gained in a hydrothermal region. The results have been generated after
scanning 4 months of data in order to check for earthquakes of similar origin. The code recorded 9
potential events over this period. Further analysis is required to complete the entire research period.
Once these events have been confirmed as natural seismic and not artificial noise, then further
conclusions can be drawn about the seismic activity in the area.
!
!
!
!
!
!
!
!
Page | !9
!
!
!
!
!
!
!
!
!
!
References (1 page limit):
!
Holland A.
Earthquakes triggered by hydraulic fracturing in South-Central Oklahoma. Bull. seism. Soc. Am.
2013;103:1784-1792.
!
Rodgers, L J. Nicewander, J. Alan, W.
Thirteen ways to look at the correlation coefficient. Am. Stat.
1988;42(1):59-66.
!
!
Shaltry, D. Colwell, C. Liberty, L. van Wijk, k
Seismic Investigation of the Neal Hot Springs Geothermal Area
2013
!
Colwell, C. et al,
Integrated Geophysical Exploration of a Known Geothermal Resource: Neal Hot Springs.
2012
!
Van Wijk, K. Channel, T. Viskupic, K. Smith, M L.
Teaching Geophysics with a Vertical-Component Seismometer
51, 552 (2013); doi: 10.1119/1.4830072
!
!
http://www.iris.edu/hq/
http://www.codecademy.com/
http://docs.obspy.org/
https://github.com/obspy/obspy/wiki
Page | !10
http://stackoverflow.com
!
Page | !11

More Related Content

Viewers also liked

On the routing overhead in infrastructureless multihop wireless networks
On the routing overhead in infrastructureless multihop wireless networksOn the routing overhead in infrastructureless multihop wireless networks
On the routing overhead in infrastructureless multihop wireless networksNarendra Singh Yadav
 
Commencement ifhs class of 1995
Commencement ifhs class of 1995Commencement ifhs class of 1995
Commencement ifhs class of 1995bdmhodges
 
Arun BTech Mech 15
Arun BTech Mech 15Arun BTech Mech 15
Arun BTech Mech 15Arun S Kumar
 
Performance Evaluation and Comparison of Ad-Hoc Source Routing Protocols
Performance Evaluation and Comparison of Ad-Hoc Source Routing ProtocolsPerformance Evaluation and Comparison of Ad-Hoc Source Routing Protocols
Performance Evaluation and Comparison of Ad-Hoc Source Routing ProtocolsNarendra Singh Yadav
 
GreenTeamIntroPowerpointATAGUCHI
GreenTeamIntroPowerpointATAGUCHIGreenTeamIntroPowerpointATAGUCHI
GreenTeamIntroPowerpointATAGUCHIJenny Heyden
 
Performance Comparison and Analysis of Table-Driven and On-Demand Routing Pro...
Performance Comparison and Analysis of Table-Driven and On-Demand Routing Pro...Performance Comparison and Analysis of Table-Driven and On-Demand Routing Pro...
Performance Comparison and Analysis of Table-Driven and On-Demand Routing Pro...Narendra Singh Yadav
 
Planning A Sampling Event of Water and Wastewater
Planning A Sampling Event of Water and WastewaterPlanning A Sampling Event of Water and Wastewater
Planning A Sampling Event of Water and WastewaterNandar Nwe (Glory)
 
Modules of Environmental Management System
Modules of Environmental Management SystemModules of Environmental Management System
Modules of Environmental Management SystemNandar Nwe (Glory)
 
cluster based routing protocol for ad hoc networks
cluster based routing protocol for ad hoc networkscluster based routing protocol for ad hoc networks
cluster based routing protocol for ad hoc networksNarendra Singh Yadav
 
163_eleizalde.ppt
163_eleizalde.ppt163_eleizalde.ppt
163_eleizalde.pptbinovo
 
155_viznoli6proba_txanti.ppt
155_viznoli6proba_txanti.ppt155_viznoli6proba_txanti.ppt
155_viznoli6proba_txanti.pptbinovo
 
Impacts of High-Voltage Power Transmission Lines Project
Impacts of High-Voltage Power Transmission Lines ProjectImpacts of High-Voltage Power Transmission Lines Project
Impacts of High-Voltage Power Transmission Lines ProjectNandar Nwe (Glory)
 
Environmental Management System
Environmental Management SystemEnvironmental Management System
Environmental Management SystemNandar Nwe (Glory)
 

Viewers also liked (17)

Sunu1
Sunu1Sunu1
Sunu1
 
On the routing overhead in infrastructureless multihop wireless networks
On the routing overhead in infrastructureless multihop wireless networksOn the routing overhead in infrastructureless multihop wireless networks
On the routing overhead in infrastructureless multihop wireless networks
 
Commencement ifhs class of 1995
Commencement ifhs class of 1995Commencement ifhs class of 1995
Commencement ifhs class of 1995
 
Sunu1
Sunu1Sunu1
Sunu1
 
Arun BTech Mech 15
Arun BTech Mech 15Arun BTech Mech 15
Arun BTech Mech 15
 
Performance Evaluation and Comparison of Ad-Hoc Source Routing Protocols
Performance Evaluation and Comparison of Ad-Hoc Source Routing ProtocolsPerformance Evaluation and Comparison of Ad-Hoc Source Routing Protocols
Performance Evaluation and Comparison of Ad-Hoc Source Routing Protocols
 
GreenTeamIntroPowerpointATAGUCHI
GreenTeamIntroPowerpointATAGUCHIGreenTeamIntroPowerpointATAGUCHI
GreenTeamIntroPowerpointATAGUCHI
 
Performance Comparison and Analysis of Table-Driven and On-Demand Routing Pro...
Performance Comparison and Analysis of Table-Driven and On-Demand Routing Pro...Performance Comparison and Analysis of Table-Driven and On-Demand Routing Pro...
Performance Comparison and Analysis of Table-Driven and On-Demand Routing Pro...
 
Planning A Sampling Event of Water and Wastewater
Planning A Sampling Event of Water and WastewaterPlanning A Sampling Event of Water and Wastewater
Planning A Sampling Event of Water and Wastewater
 
Modules of Environmental Management System
Modules of Environmental Management SystemModules of Environmental Management System
Modules of Environmental Management System
 
cluster based routing protocol for ad hoc networks
cluster based routing protocol for ad hoc networkscluster based routing protocol for ad hoc networks
cluster based routing protocol for ad hoc networks
 
163_eleizalde.ppt
163_eleizalde.ppt163_eleizalde.ppt
163_eleizalde.ppt
 
155_viznoli6proba_txanti.ppt
155_viznoli6proba_txanti.ppt155_viznoli6proba_txanti.ppt
155_viznoli6proba_txanti.ppt
 
Landslide Safety Tips
Landslide Safety TipsLandslide Safety Tips
Landslide Safety Tips
 
Impacts of High-Voltage Power Transmission Lines Project
Impacts of High-Voltage Power Transmission Lines ProjectImpacts of High-Voltage Power Transmission Lines Project
Impacts of High-Voltage Power Transmission Lines Project
 
Noise Pollution
Noise PollutionNoise Pollution
Noise Pollution
 
Environmental Management System
Environmental Management SystemEnvironmental Management System
Environmental Management System
 

Similar to summer-studentship-report-PDF

Toward Real-Time Analysis of Large Data Volumes for Diffraction Studies by Ma...
Toward Real-Time Analysis of Large Data Volumes for Diffraction Studies by Ma...Toward Real-Time Analysis of Large Data Volumes for Diffraction Studies by Ma...
Toward Real-Time Analysis of Large Data Volumes for Diffraction Studies by Ma...EarthCube
 
Leibniz: A Digital Scientific Notation
Leibniz: A Digital Scientific NotationLeibniz: A Digital Scientific Notation
Leibniz: A Digital Scientific Notationkhinsen
 
One–day wave forecasts based on artificial neural networks
One–day wave forecasts based on artificial neural networksOne–day wave forecasts based on artificial neural networks
One–day wave forecasts based on artificial neural networksJonathan D'Cruz
 
DAQScienceTeamMeetingPosterV1
DAQScienceTeamMeetingPosterV1DAQScienceTeamMeetingPosterV1
DAQScienceTeamMeetingPosterV1Alex Kotsakis
 
Detection of Radio Emission from Fireballs
Detection of Radio Emission from FireballsDetection of Radio Emission from Fireballs
Detection of Radio Emission from FireballsCarlos Bella
 
Analysis Of Three-Prong Events In The Heavy Ion Interaction Of 17.0 MeV U 132...
Analysis Of Three-Prong Events In The Heavy Ion Interaction Of 17.0 MeV U 132...Analysis Of Three-Prong Events In The Heavy Ion Interaction Of 17.0 MeV U 132...
Analysis Of Three-Prong Events In The Heavy Ion Interaction Of 17.0 MeV U 132...Arlene Smith
 
Abstracts Of The Emerging Scholars Program Research Projects Fall 2010 Suppor...
Abstracts Of The Emerging Scholars Program Research Projects Fall 2010 Suppor...Abstracts Of The Emerging Scholars Program Research Projects Fall 2010 Suppor...
Abstracts Of The Emerging Scholars Program Research Projects Fall 2010 Suppor...Claire Webber
 
Recovery of aftershock sequences using waveform cross correlation: from catas...
Recovery of aftershock sequences using waveform cross correlation: from catas...Recovery of aftershock sequences using waveform cross correlation: from catas...
Recovery of aftershock sequences using waveform cross correlation: from catas...Ivan Kitov
 
NNBAR SESAPS PRESENTATION FINAL
NNBAR SESAPS PRESENTATION FINALNNBAR SESAPS PRESENTATION FINAL
NNBAR SESAPS PRESENTATION FINALJoshua Barrow
 
INCIDENCE OF ABSORPTION AT THE INTERFACE OF GALAXIES AND IGM
INCIDENCE OF ABSORPTION AT THE INTERFACE OF GALAXIES AND IGM INCIDENCE OF ABSORPTION AT THE INTERFACE OF GALAXIES AND IGM
INCIDENCE OF ABSORPTION AT THE INTERFACE OF GALAXIES AND IGM University of Delaware
 
Aip pg book of abstracts
Aip pg book of abstractsAip pg book of abstracts
Aip pg book of abstractsSiddartha Verma
 
Self-organzing maps in Earth Observation Data Cube Analysis
Self-organzing maps in Earth Observation Data Cube AnalysisSelf-organzing maps in Earth Observation Data Cube Analysis
Self-organzing maps in Earth Observation Data Cube AnalysisLorena Santos
 
The benefit of hindsight in observational science - Retrospective seismologica...
The benefit of hindsight in observational science - Retrospective seismologica...The benefit of hindsight in observational science - Retrospective seismologica...
The benefit of hindsight in observational science - Retrospective seismologica...Elizabeth Entwistle
 

Similar to summer-studentship-report-PDF (20)

Intel Report
Intel ReportIntel Report
Intel Report
 
Intel Report
Intel ReportIntel Report
Intel Report
 
Toward Real-Time Analysis of Large Data Volumes for Diffraction Studies by Ma...
Toward Real-Time Analysis of Large Data Volumes for Diffraction Studies by Ma...Toward Real-Time Analysis of Large Data Volumes for Diffraction Studies by Ma...
Toward Real-Time Analysis of Large Data Volumes for Diffraction Studies by Ma...
 
Zhang weglein-2008
Zhang weglein-2008Zhang weglein-2008
Zhang weglein-2008
 
Climate Extremes Workshop - Networks and Extremes: Review and Further Studies...
Climate Extremes Workshop - Networks and Extremes: Review and Further Studies...Climate Extremes Workshop - Networks and Extremes: Review and Further Studies...
Climate Extremes Workshop - Networks and Extremes: Review and Further Studies...
 
Leibniz: A Digital Scientific Notation
Leibniz: A Digital Scientific NotationLeibniz: A Digital Scientific Notation
Leibniz: A Digital Scientific Notation
 
Terra Seismic : Global Earthquake Prediction Systems
Terra Seismic : Global Earthquake Prediction SystemsTerra Seismic : Global Earthquake Prediction Systems
Terra Seismic : Global Earthquake Prediction Systems
 
mclaren2012
mclaren2012mclaren2012
mclaren2012
 
One–day wave forecasts based on artificial neural networks
One–day wave forecasts based on artificial neural networksOne–day wave forecasts based on artificial neural networks
One–day wave forecasts based on artificial neural networks
 
DAQScienceTeamMeetingPosterV1
DAQScienceTeamMeetingPosterV1DAQScienceTeamMeetingPosterV1
DAQScienceTeamMeetingPosterV1
 
American Journal of Biometrics & Biostatistics
American Journal of Biometrics & BiostatisticsAmerican Journal of Biometrics & Biostatistics
American Journal of Biometrics & Biostatistics
 
Detection of Radio Emission from Fireballs
Detection of Radio Emission from FireballsDetection of Radio Emission from Fireballs
Detection of Radio Emission from Fireballs
 
Analysis Of Three-Prong Events In The Heavy Ion Interaction Of 17.0 MeV U 132...
Analysis Of Three-Prong Events In The Heavy Ion Interaction Of 17.0 MeV U 132...Analysis Of Three-Prong Events In The Heavy Ion Interaction Of 17.0 MeV U 132...
Analysis Of Three-Prong Events In The Heavy Ion Interaction Of 17.0 MeV U 132...
 
Abstracts Of The Emerging Scholars Program Research Projects Fall 2010 Suppor...
Abstracts Of The Emerging Scholars Program Research Projects Fall 2010 Suppor...Abstracts Of The Emerging Scholars Program Research Projects Fall 2010 Suppor...
Abstracts Of The Emerging Scholars Program Research Projects Fall 2010 Suppor...
 
Recovery of aftershock sequences using waveform cross correlation: from catas...
Recovery of aftershock sequences using waveform cross correlation: from catas...Recovery of aftershock sequences using waveform cross correlation: from catas...
Recovery of aftershock sequences using waveform cross correlation: from catas...
 
NNBAR SESAPS PRESENTATION FINAL
NNBAR SESAPS PRESENTATION FINALNNBAR SESAPS PRESENTATION FINAL
NNBAR SESAPS PRESENTATION FINAL
 
INCIDENCE OF ABSORPTION AT THE INTERFACE OF GALAXIES AND IGM
INCIDENCE OF ABSORPTION AT THE INTERFACE OF GALAXIES AND IGM INCIDENCE OF ABSORPTION AT THE INTERFACE OF GALAXIES AND IGM
INCIDENCE OF ABSORPTION AT THE INTERFACE OF GALAXIES AND IGM
 
Aip pg book of abstracts
Aip pg book of abstractsAip pg book of abstracts
Aip pg book of abstracts
 
Self-organzing maps in Earth Observation Data Cube Analysis
Self-organzing maps in Earth Observation Data Cube AnalysisSelf-organzing maps in Earth Observation Data Cube Analysis
Self-organzing maps in Earth Observation Data Cube Analysis
 
The benefit of hindsight in observational science - Retrospective seismologica...
The benefit of hindsight in observational science - Retrospective seismologica...The benefit of hindsight in observational science - Retrospective seismologica...
The benefit of hindsight in observational science - Retrospective seismologica...
 

summer-studentship-report-PDF

  • 1. ! ! ! ! ! Summer Studentships 2013-2014 ! Final Project Report ! ! ! ! ! !! ! ! ! !!!! ! ! ! ! ! ! ! ! ! ! Project Details Student Name: Roydon Nutsford Name of supervisor(s): Kasper van Wijk Supervisor department: Physics Project title: !Seismic Investigation of the Neal Hot Springs Geothermal Area Supervisor letter included: Yes ☐ No ☐ Page | !1
  • 2. ! ! ! ! ! ! Brief statement on how the studentship has contributed to your career development (1 page limit): ! !!!Participation in this studentship has allowed me to further improve my skills in the field of geophysics. I have gained a true appreciation for the purpose of research and how it can make a real impact in the field, and on a career. Part of the studentship involved learning to write Python code from the ground up. Understanding this still emerging language in the Geophysics work force today will give me a cutting edge. The ability to use computers to analyse data will propel me in a career were staying current is important in ever changing fields. !Through the proceedings of the research I was exposed to real earthquake databases which taught me to read, comprehend and make use of large data networks from all over the world. This contributes experience to an area of key skills which will be vital in the work place. Finally the studentship program has enforced the universal requirement to present scientific data and findings in an interesting, informative and appropriate way for readers to best understand. ! ! ! ! ! ! ! Research abstract – not more than 250 words: ! !!!The aim of the research project is to detect small local earthquakes which are potentially covered or disguised by background noise. The testing location is Neal Hot Springs, an active geothermal site that supplies a 23MW geothermal electric power plant near Vale, Oregon. !In May 2011, 10 temporary seismic stations were set up to record the earth’s surface movements over 20 months. Numerical cross correlation is used to identify similar wave forms. The goal is to locate all of the local earthquakes, and then differentiate between geothermal seismic activity and natural seismic activity in the area. Python code was used to scan seismic data against a nearby ‘template’ earthquake event found from a previous study (Shaltry, 2013) by Daniel Shaltry, a student at Boise State University. The program returned coefficients which define the similarity of the waveforms. The method relies on earthquake pressure, Love and Rayleigh waves to produce similar waveforms, when the earthquakes originate from similar locations. With three earthquakes already visually identified by Shaltry, the program spotted 9 more. These can be further studied to deduce whether they are a product of surrounding local active faults, or whether they are the result of the geothermal exploration. !! ! ! ! ! ! Page | !2
  • 3. ! ! ! Summary of research and its significance (1 page limit): ! ! ! The Geophysics group from Boise State University set up and recorded seismic data in the Neal Hot Springs area, near Vale, Oregon. Colwell (2012) From their study they identified three local earthquakes. In this research I used computer analysis to gain higher resolution from the data in order to locate more earthquakes. Extracting the data from the Incorporated Research Institution for Seismology (IRIS) database, I used ObSpy toolbox to compare the earthquake waveforms against known seismic events within the local vicinity, found in the previous study. The hypothesis that earthquakes which originate in close proximity to one another will produce similar wave forms was tested by scanning the data for high cross correlation values using Austin Holland xcor code. Holland (2013) If the cross correlation coefficient was 0.7 or higher, and occurred on at least 5 stations I classified it as an event. ! With this in mind the goal was to locate more small earthquakes to interpret the activity of local faults and assess whether the Neal Hot Spring or the power generation plant increases local seismicity. The correlation code does not match with large earthquakes at large proximity due to a combination of different earthquake mechanisms and earth lateral variability deforming the waveform to a state where the coefficient is not high enough. The significance of this study is testing the previous results with new technology, and trying to increase the ability to distinguish small earthquakes from ambient noise. The cross correlation process allows the waveforms to be evaluated, giving information about the subsurface geology in the area, and how this earth material affects earthquake waveforms. Further research involves studying the distance between stations, and how much the waveform is deformed in that period. In particular, the effect of high ground water content, in such cases as a geothermal region. ! ! ! ! ! ! ! Research report - Aims, methods, results & discussion (5 page limit): ! ! ! Aim: To test the previous hypothesis with a new technology/methodology and determine the local seismic activity of the Neal Hot Springs region with respect to earthquakes originating from local faults and to investigate the waveform variability of earthquakes in, and passing through the area. ! ! ! ! ! Page | !3
  • 4. ! ! Method: For this research project the raw data was produced by students of Boise State University, under the supervision of Dr Van Wijk, and consisted of 10 stations, each station with a L-22 seismometer, data acquisition system, GPS, solar panel and battery. The stations ran continuously from May 2011 until November 2012 and recorded vertical, northward and eastward propagation at a rate of 250 samples per second. This data was sent to the Incorporated Research Institution for Seismology (IRIS) database. This data was then sourced and written into a python 2.6.7 code, with the ObSpy 0.9 toolbox handling the stream data. Boise State University student Daniel Shaltry found three low magnitude earthquakes of close proximity to the stations. I used these earthquakes as template events. I used a code to extract the first 5 seconds of each of these earthquakes, capturing the P and S wave arrivals. The next section of the code looped through the time between May 2011 and November 2012. Xcorr, a cross correlation code written by Austin Holland was then implemented to compare the template events with the trace data. Each individual template was correlated with each individual station using parallel computing. An acceptable correlation coefficient of 0.7 was required from five or more stations to be considered an earthquake event. The high correlations occurred where the trace data waveforms matched those of template earthquakes closely, meaning they are of the same origin, and therefore allowing stronger interpretations to be drawn. ! Results. The raw data was retrieved from IRIS and the three earthquakes spotted by Daniel Shaltry were found. The raw data contained 10 traces which recorded for 20 months. Each trace corresponded to a recording station. A small script of python code was required to trim the data down to a short template. The template captures the p-wave and s-wave arrival times of the most representative trace in, along with the first few seconds of vigorous shaking. ! ! Original template ! (figure 1): -Raw data showing first 6 seconds of an earthquake waveform. The wave arrivals are clear but there is strong noise. ! Page | !4
  • 5. ! The Original template has discrete wave arrivals; however, the true waveform is hidden under considerable noise. The waveform was smoothed by reducing the noise through a series of filters, which cleaned the raw data as seen in figure 2 below. This made available a manageable template for the algorithm. The correlation code acting on the cleaned data was able to gain consistent correlation coefficients of a higher value. To maximise the chance of finding earthquakes I created filtered templates from all three known waveforms. ! Filtered template - Earthquake 1 ! (figure 2): Filtered template of waveform 1, ready to be cross-correlated against the data set. Displays a much smoother representation of the waveform. ! ! ! ! ! Filtered template – Earthquake 2 ! (figure 3): Filtered template of waveform 2, ready to be cross-correlated against the data set. Using multiple waveforms from within the area will increase the potential hits with the code. ! Page | !5
  • 6. ! ! Filtered template – Earthquake 3 ! (figure 3): Filtered template of waveform 3, ready to be cross-correlated against the data set. ! ! The above templates were used to scan the data from the 1st May 2011 till the 1st November 2011. In order to obtain consistency, the same filters applied to the templates were also transferred to the raw data as the correlation code requires two parameters to control the output it creates. These can be set to alert when earthquakes of any required similarity are found. In my program I have defined an event as a correlation value of 0.7, across at least 5 stations. This allows only similar waveforms to be alerted, and nearly eliminates noise as correlations are required over five stations. The program loops over two minute intervals. This opens a window were correlations can occur at any time. To minimise the probability of random correlations meeting the 5 station requirement, I set the required correlation at 0.7, which is a high waveform similarity which does not occur commonly. 
 ! Known earthquake events were picked up by the code during testing. This proved the proficiency of the code. From the XN data set, Figure 4 displays the 3rd earthquake, including the data from all 10 station traces, showing distinctive earthquake waveforms. This earthquake was approximately 13km away. This can be estimated through the wave arrival times at different station locations. ! ! ! ! ! ! ! ! ! Page | !6
  • 7. 3rd known event across all 10 stations ! (Figure 4) Two minute loop captured by Python code, showing the arrival times of earthquake 3 at each station. The magnitude and arrival time vary, however the waveform remains relatively consistent. This event was picked up on 8 stations. Station 10 recorded a 1, as it is the direct trace of template event. Two of the remaining stations recorded above 0.9, whilst 5 others recorded above 0.8 in the correlation code. ! ! Page | !7
  • 8. ! Event detection count ! (Figure 5) displays the times of each event detected by the Python code. ! ! ! ! ! ! ! ! Discussion: ! The data produced 9 potential earthquake events over four months. The parameters imposed on the code were strictly allowing times only with a high probability of containing an earthquake to be detected. There are no previously sighted events in this period and therefore the nine potential events show that there could in fact be higher seismic activity in the area than previously expected. As the events have not been noticed by the previous study, it indicates the earthquakes are small and not clearly visible to the naked eye. Month number of events Loop times June 5 2011-06-06T12:46 2011-06-06T21:20 2011-06-17T18:12 2011-06-17T18:16 2011-06-18T06:16 July 1 2011-08-14T00:00 August 0 September 3 2011-09-25T23:22 2011-09-25T23:24 2011-09-26T00:36 Page | !8
  • 9. The code has not yet been passed over the second period of data due to various problems. The standard deviation algorithm was the biggest slowdown in the code as it used up large amounts of computer memory. This forced the ! ! data to be feed into the code in short two minute intervals, reducing the efficiency resulting in slow data processing. Another challenge was caused by the data being sourced from Washington, which occasionally took too long causing the code to crash. This made running any extended duration of data infeasible. Future work is to continue processing the data over the entire recording period. This will give data pre and post the operation of the hydrothermal power station. Understanding the seismic activity will allow conclusions to be drawn with respect to the production of artificial seismic due to the activities of the power station. Further investigation is required to finalise if the found events are in fact earthquakes as it is possible that random noise has occurred at five stations within a two minute loop. A manual check will not confirm the event as it is likely to be unreadable by the naked eye. With an entire data set, the location of individual events can be triangulated to pin point the origin of the event. This will provide evidence as to where in the region has been most active, and raise the question of why. Another aspect I would like to carry further is the waveform deformation with distance through the different surface composition. This can be done through analysing the correlation coefficient at different linear distances from the source. If distinct changes occur it could even have relevance to fault mapping or structural geologic mapping. ! . ! ! ! ! Conclusion ! The summer project has allowed me to learn python code, and more about seismology. I have applied a complex code to real world data gained in a hydrothermal region. The results have been generated after scanning 4 months of data in order to check for earthquakes of similar origin. The code recorded 9 potential events over this period. Further analysis is required to complete the entire research period. Once these events have been confirmed as natural seismic and not artificial noise, then further conclusions can be drawn about the seismic activity in the area. ! ! ! ! ! ! ! ! Page | !9
  • 10. ! ! ! ! ! ! ! ! ! ! References (1 page limit): ! Holland A. Earthquakes triggered by hydraulic fracturing in South-Central Oklahoma. Bull. seism. Soc. Am. 2013;103:1784-1792. ! Rodgers, L J. Nicewander, J. Alan, W. Thirteen ways to look at the correlation coefficient. Am. Stat. 1988;42(1):59-66. ! ! Shaltry, D. Colwell, C. Liberty, L. van Wijk, k Seismic Investigation of the Neal Hot Springs Geothermal Area 2013 ! Colwell, C. et al, Integrated Geophysical Exploration of a Known Geothermal Resource: Neal Hot Springs. 2012 ! Van Wijk, K. Channel, T. Viskupic, K. Smith, M L. Teaching Geophysics with a Vertical-Component Seismometer 51, 552 (2013); doi: 10.1119/1.4830072 ! ! http://www.iris.edu/hq/ http://www.codecademy.com/ http://docs.obspy.org/ https://github.com/obspy/obspy/wiki Page | !10