1. Please download animations in the file format of your choice (MPG format is recommended because AVI files are much larger than MPG) August 2010 slow-slip event: ftp://ftp.gps.caltech.edu/pub/minson/chpt_Aug_6-10_2010.avi ftp://ftp.gps.caltech.edu/pub/minson/chpt_Aug_6-10_2010.mpg Full 2010 time series for whole PANGA network: ftp://ftp.gps.caltech.edu/pub/minson/chpt_one_lin_2010.avi ftp://ftp.gps.caltech.edu/pub/minson/chpt_one_lin_2010.mpg ftp://ftp.gps.caltech.edu/pub/minson/chpt_one_lin_2010.avi ftp://ftp.gps.caltech.edu/pub/minson/chpt_one_lin_2010.mpg

2. *Please excuse technical jargon

3. Earthquake monitoring using GPS  Dense real-time 1 Hz GPS network  Utilize GPS with seismic data to identify and analyze events for which GPS data contribute the most Hazardous earthquakes Slow slip events

4. Rapid response rupture models using real-time GPS  Monte Carlo exploration of fault extent, location, and slip Okada model Constraints on fault- finiteness depend greatly on source receiver geometry Moment is well-determined  Need to detect events and estimate offsets in real-time

5. Bayesian changepoint detection  Changepoint: the time that at least one model parameter changes  Can use Bayes’ theorem to compute the probability of a changepoint as a function of time, P(changepoint=t|D)  Can also use Bayes’ theorem to assess significance of potential changepoints

6. Bayesian changepoint detection  It may sound fancy, but it’s not It’s just Bayesian piecewise linear regression What is P(t=t 0 )? ○ The solution is analytical and ridiculously cheap

7. The Nile River & The Aswan Dam

16. Real-time monitoring  That was the retrospective changepoint problem I have a time series. Is there a changepoint or changepoints? If so, when?  For real-time monitoring, need prospective changepoint analysis a.k.a. the quickest detection problem As each new observation arrives, do I now have enough data to identify a changepoint? Solution is to minimize the Bayes’ risk: ○ P(false alarm) + c*[average detection delay] c is tuning parameter

17. Some complications – or – what we really want are real-time offsets  For slow slip events, we have a functional form for earthquakes We get offset as part of change-point analysis  For earthquakes, we could get ramps, waveforms, anything Filter the heck out of the GPS time series until we make it look like a ramp? Window out the earthquake?

18. One more thing  If we want to use GPS stations to detect events (to trigger rapid response or for EEW) and/or we want to retrospectively search for deformation events, we have to discriminate between changepoints due to signals and those due to noise Need to rely on spatio-temporal pattern of stations with possible detections Could treat triggers at GPS stations like seismic data or could do something new

20. A little motivation  Need offshore measurements to understand plate locking and ETS, as well as to determinine coseismic slip Basic science could be done with campaign measurements, but monitoring needs real-time data ○ Real-time might not be more expensive due to the costs of ship time to retrieve instruments and data  The PBO should extend to the plate boundary Even better, should go beyond trench to understand deformation of incoming plate

21. Many kinds of measurements  Mature instrumentation: GPS-acoustic Ocean bottom pressure sensors DART Seafloor fluid flow Campaign relative gravity ○ Japan is deploying 5,100 km of cable to install 154 nodes with OBS/OBP  On the horizon: Seafloor timelapse gravity (absolute gravity) Self-calibrating pressure recorders (decreases instrument drift) Non-electronic seismometer/tiltmeter (decreases instrument drift of tiltmeters) Seafloor interferometric optical fiber strainmeter

22. Q. Who pays? A. No clue  When it comes to monitoring, the USGS is the obvious choice The USGS doesn’t have the money  NOAA might want to expand DART network, install more OBP  FEMA? DHS?

23. Conference Highlights  Best factoid: Turbidite flow from Tohoku hit OBP station P03 and moved it about 1 km to the east  Runner-up best factoid: Largest onshore displacement from Tohoku was 5 m, but largest offshore displacement was 31 m  Best unnecessarily long word: Seismohydrogeodynamic  Best potential pitfall: Will the U.S. Navy stand in the way of possible future offshore network to protect secrecy of their submarine positions?