1 / 6

Triggering Models vs. Smoothed Seismicity

Triggering Models vs. Smoothed Seismicity. Testing region: California Target events: M  ≥ 3.95 Testing period : 2008 -2010 Testing method: T-test. PG = probability gain = P / P 0 IG = information gain = log e ( PG ). STEP model. PG = 1.35/eqk. Reference forecast.

vito
Télécharger la présentation

Triggering Models vs. Smoothed Seismicity

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Triggering Models vs. Smoothed Seismicity Testing region: California Target events: M ≥ 3.95 Testing period: 2008-2010 Testing method: T-test PG = probability gain = P/P0 IG = information gain = loge(PG) STEP model PG= 1.35/eqk Reference forecast PG= 10/eqk Information gain per earthquake

  2. Japan and NZ Testing Regions 1 day 3 month 6 month 5 year Total New Zealand 2 8 1 4 15

  3. Darfield Aftershock Forecasting(Gerstenberger & Rhoades) Testing region: New Zealand Target events: M ≥ 4 (ETAS, PPE-1d), M ≥ 5 (PPE-3m, PPE-5y) Testing period: 4 Sept 2010 - 8 Mar 2011 Testing method: N-test Forecast Nobs = 271 (M ≥ 4) 209 are Darfield aftershocks Nobs = 17 (M ≥ 5) Number of earthquakes

  4. Darfield Aftershock Forecasting(Gerstenberger & Rhoades) Testing region: New Zealand Target events: M ≥ 4 (PPE-1d), M ≥ 5 (PPE-3m, PPE-5y) Testing period: 4 Sept 2010 - 8 Mar 2011 Testing method: T-test ETAS model PG= 99/eqk Reference forecast PG= 544/eqk PG= 1480/eqk Information gain per earthquake

  5. CSEP Testing Results • Comparative evaluations have quickly identified errors in model implementation • effective method for model verification (debugging) • 5-yr RELM testing program has been effective in ranking long-term forecasting performance for M ≥ 5 target events in California • RELM paper by Zechar, Schorlemmer, et al. • Aftershock triggering models (e.g., STEP, ETAS) obtain probability gains of 10-1000 relative to seismicity averaging models (e.g. PPE, TripleS) • Substantially more information gain can be obtained by updating forecasts more frequently than at 1-day intervals • Adaptive triggering models out-perform those with time-independent parameters • Gerstenberger’s STEP model currently shows the best short-term performance in California; adaptive models in NZ and Japan are still being evaluated

  6. CSEP Plans • Monitor the performance of 91 CSEP/Japan and 15 CSEP/NZ forecasting models during the active phases of the Tohoku and Darfield sequences • Reduce the updating interval for short-term forecasts from 1 day to 1 hr or less • Improve procedures for adapting forecasts to changes in the seismic environment • Incorporate forecasting models based on physical hypotheses about earthquake generation • e.g., Coulomb stress function, rate/state friction, dynamic vs. static triggering, slow slip events, tidal triggering • Expand prospective testing to models based on non-seismic data • Evaluate hypotheses critical to forecasting large earthquakes • e.g., fault segmentation, maximum magnitude, characteristic earthquakes, strongly coupled seismic gaps • Expand global testing program • Include model classes for legacy methods; e.g., M8/MSc • Develop CSEP capabilities to support operational earthquake forecasting • Prospectively test candidate forecasting procedures • Unify forecasting across temporal and spatial scales (e.g. long-term & short-term) • Reduce testing latency by modeling catalog completeness and accuracy • Expand retrospective testing over the entire history of instrumental catalogs • Initiate model testing using recorded ground motions • Support other prospective testing activities, including earthquake early warning

More Related