Earthquake Fingerprint Scanning to Discern Induced Seismicity

Transcription

1 Earthquake Fingerprint Scanning to Discern Induced Seismicity Mike Brudzinski Rob Skoumal Brian Currie Steve Holtkamp Running a Background Check on Suspicious Earthquakes

2 Template Poten,al Match Match Poten,al Not a Match Youngstown Earthquake Magnitude Original Detec,on Threshold Year Skoumal et al., EPSL, 2014

3 Skoumal et al., submi7ed

4 Template Families Parents = Ini=al earthquake templates Children = Matched earthquakes Look for Cousins: Empirical subspace detector Try other templates: Deriva=ve of template waveforms Stack of matched waveforms Deriva=ve of the stack Skoumal et al., SRL, in press

5 Op3mized Rou3ne Computa=onal efficiency + cluster use > 1 billion correla3ons (>1 year of data) per sec Op=mized parameters for regional network Bandpass frequency Template length 5-15 Hz ~35 sec Sta3on configura3on TA N54A, M54A, O56A Skoumal et al., EPSL, 2014

6 Op3mized Rou3ne Local Data Op=mized parameters for local network Bandpass frequency Template length 1-45 Hz 6 sec Sta3on configura3on ODNR AWM1-4 Skoumal et al., SRL, in press

7 Correla3ons Improve Rela3ve Loca3ons and Magnitudes Double Difference Reloca=on Correla=on- based Magnitude Es=mates Holtkamp et al., AAPG, in press

8 Fingerprint Scanning vs. Tradi3onal a) Tradi=onal detec=on with local deployment in Azle, TX b) 3 sta=on regional template matching Local catalog from Hornbach et al., Nature Comm., 2015

9 All Ohio Earthquakes Since swarms triggered by hydraulic fracturing 3 swarms triggered by wastewater injec=on 17 natural earthquakes Skoumal et al., JGR, revised

10 Youngstown: Wastewater Injec3on Skoumal et al., EPSL, 2014; Holtkamp et al., AAPGB, 2015

11 Poland, OH: Hydraulic Fracturing Skoumal et al., BSSA, 2015

12 Poland Magnitudes over Time Nearly all earthquakes during 6 stages Screen- out occurred immediately before earthquakes (observed in UK and Canada too) Max magnitude ~3 Skoumal et al., BSSA, 2015

13 Poland Cross- Sec3on Skoumal et al., BSSA, 2015; Davies et al., MPG, 2013

14 Harrison County Friberg et al., SRL, 2014 S=mula=on Skoumal et al., JGR, submi7ed

15 Harrison County Part 2 Skoumal et al., in prep

16 Oklahoma Skoumal et al., submi7ed

17 Annual Distribu3on of Matched Events Skoumal et al., submi7ed

18 Compared with Injected Volumes Skoumal et al., submi7ed

19 Comparison with Hydraulic Fracturing

20 Now it s Your Turn to try Large- E My student compiled a Matlab script for you to try running mul=- sta=on waveform cross- correla=on Download and unpack tarball from the web page Run: /bin/sh./rcorr.sh This will use the input parameters in file.txt It obtains waveforms using IRIS DMC webservices wget calls Matlab does the cross- correla3ons (~200,000 per sec) Produces an output file in corr_results/ directory

21 cat file.txt #Folder for output file corr_results #Corr Start #Corr End #Template Length, sec 30 #Bandpass Filter, Hz 5 15 #MAD Coefficient 8 Input file

22 Input file (con3nued) #Downsample Rate 40 #Deriva3ve 0 #Data loca3on IRIS_data #Template loca3on #Num Cores 1 #Download data / Save data 1 1

23 Input file (con3nued) #Template 3mes somich_ _14_ N4 J47A - - BHE N4 J47A - - BHN N4 J47A - - BHZ N4 L46A - - BHE N4 L46A - - BHN N4 L46A - - BHZ N4 L48A - - BHE N4 L48A - - BHN N4 L48A - - BHZ 1

24 Output file cat corr_results/somich txt Format: epoch- =me NNCC MAD #channels NNCC: Network Normalize Correla3on Coefficient MAD: Median Absolute Devia3on awk $2/$3>15 corr_results/somich txt

25 Let s Try Some Examples I made some example template files for you Where: in examples/ directory of where you are Events are listed in events.txt When you pick an event, copy the examples/file.[id].txt to file.txt and re- run rcorr.sh Use the awk command to find matches above 15 MAD While wai3ng: Look at where the event occurred Look at where the template sta=ons are rela=ve to the event

26 events.txt #EventID Time La=tude Longitude Depth/km Author Catalog Contributor ContributorID MagType Magnitude MagAuthor EventLoca=onName T17:08: CERI NEIC PDE NEIC PDE- M MB 4.3 NEIC KENTUCKY T22:06: CERI NEIC PDE NEIC PDE- M MD 2.3 CERI KENTUCKY T02:23: OH NEIC PDE NEIC COMCAT product.xml ML 3.2 US OHIO T07:12: SE NEIC PDE NEIC ALERT MD 2.1 SE MISSISSIPPI T09:37: US NEIC PDE NEIC COMCAT product.xml MB_LG 2.4 US CENTRAL TEXAS T09:10: US NEIC PDE NEIC COMCAT product.xml MB_LG 3.3 US NEBRASKA T15:42: US NEIC PDE NEIC COMCAT product.xml MB_LG 3.3 US MICHIGAN T03:10: SE NEIC PDE NEIC ALERT MD 2.7 SE TENNESSEE

27 What Did You Find? How many matches above 15 MAD? What are the NNCC for those matches? What are the =mes rela=ve to the template? Use epoch command to convert epoch 3me to day- 3me /opt/antelope/5.5/bin/epoch What kind of sta=ons were used in the template? Try adding a day or two before and awer in input file Do you think the events are natural or induced? What other informa3on would you want to know?

28