The interevent time fingerprint of t riggering for induced seismicity Mark Naylor
Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent...
Transcript of Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent...
![Page 1: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/1.jpg)
Background seismicity rates frominterevent-time statistics:
Spatial patterns appear stationary through time
Jeanne HardebeckUnited States Geological Survey
![Page 2: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/2.jpg)
Background earthquake rate
Assumes two earthquake classes:
1) Background events: occur as direct response toloading (e.g. tectonic, magmatic.)
2) Triggered events: caused by other earthquakes.
![Page 3: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/3.jpg)
Background earthquake rate
Important for:• Time-independent seismic hazard assessment.• Background rate changes related to loadingrate changes => fault physics; time-dependentseismic hazard.• Background rate changes related to stressshadows => static versus dynamic stresstriggering debate.
![Page 4: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/4.jpg)
Background earthquake rate
Difficult to measure:• Background rate obscured by triggered events.• Standard declustering methods (Reasenberg,Gardner & Knopoff) rely on two unknown,subjectively adjustable parameters: the length-scale and time-scale that define a “triggered”event.• A poster today: Van Stiphout “How far can wetrust declustering algorithms?”
![Page 5: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/5.jpg)
Earthquake interevent-time distribution
- Time between successive events i and i+1: Δti = ti+1 - ti.- N earthquakes, catalog duration T, given area and magnitude range.- Normalize by average interevent time, τi=Δti*(N/T).
![Page 6: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/6.jpg)
Short interevent times: clustered => power law.Long interevent times: Poissonian => exponential.
Gamma distribution combinespower law and exponential:
Gamma distribution formulationfrom Hainzl et al., BSSA 2006,following Corral, PRL, 2004.
Earthquake interevent-time distribution
![Page 7: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/7.jpg)
Method of Hainzl et al., BSSA 2006:Background fraction found objectivelyand easily from mean and variance ofinterevent-time distribution:
Follows from gamma distribution:
Earthquake interevent-time distribution
![Page 8: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/8.jpg)
Follows from gamma distribution:
Gamma distribution is apoor fit to our data!
Earthquake interevent-time distribution
Method of Hainzl et al., BSSA 2006:Background fraction found objectivelyand easily from mean and variance ofinterevent-time distribution:
![Page 9: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/9.jpg)
Earthquake interevent-time distributionTheoretical distribution:
Another form of the theoreticaldistribution from Gutenberg-Richter and Omori laws wasderived by Saichev and Sornette,JGR 2007.
![Page 10: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/10.jpg)
![Page 11: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/11.jpg)
2000 ETAS simulations: Hainzl et al method works well, but requires acorrection based on direct p-value (not usually known for real data.)
![Page 12: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/12.jpg)
Northern California
1980-2007
![Page 13: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/13.jpg)
Method used for USseismic hazard maps.
ObjectiveSubjective, using recommended parameters: Reasenberg Gardner & Knopoff
![Page 14: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/14.jpg)
![Page 15: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/15.jpg)
Calaveras
Parkfield
High rate
![Page 16: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/16.jpg)
![Page 17: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/17.jpg)
Calaveras Fault
![Page 18: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/18.jpg)
San Andreas Fault: Parkfield
![Page 19: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/19.jpg)
![Page 20: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/20.jpg)
Schorlemmer et al., JGR 2004
![Page 21: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/21.jpg)
Conclusions:
• The Hainzl et al method can accurately recover thebackground fraction, for ETAS simulations, but requires acorrection based on the direct p-value.
• Gamma distribution poorly fits the observed and ETASinterevent-time distributions, especially for high direct p-value.
• The background rate in Northern California is spatiallyvariable, but appears generally stable through time.
• Parkfield seismicity is remarkably stationary: earthquakelocations, background rate, and b-value.
![Page 22: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/22.jpg)
Future Work - Methodology:
• Develop a method to use the theoretical interevent-timedistribution (rather than gamma distribution) to findbackground fraction and direct p-value.
• Develop a quantitative measure of uncertainty for theestimated background rate. Explore sources of error such ascatalog incompleteness.
![Page 23: Background seismicity rates from interevent-time statistics · 2017. 4. 3. · Short interevent times: clustered => power law. Long interevent times: Poissonian => exponential. Gamma](https://reader035.fdocuments.net/reader035/viewer/2022071416/6111c45d06e07a669300f5f6/html5/thumbnails/23.jpg)
Future Work - Applications:
• Quantify the stability of background rate through time.• New background rates for seismic hazard estimates.• Search for temporal changes in background rate, especiallyin areas of changing stressing rate:
- Volcanic areas: does background rate change due toloading from magma movement?- Subduction zones: does background rate change due toloading from episodic slow-slip events?
- Stress shadows: do they exist?• Etc…