Uniform California Earthquake Rupture Forecast Version 3 (UCERF3) Building an Operational Earthquake Forecast for California 2014 Working Group on California Earthquake Probabilities Field, E. H., J. R. Arrowsmith, G. P. Biasi, P. Bird, T. E. Dawson, K. R. Felzer, D. D. Jackson, K. M. Johnson, T. H. Jordan, C. Madden, A. J. Michael, K. R. Milner, M. T. Page, T. Parsons, P. M. Powers, B. E. Shaw, W. R. Thatcher, R. J. Weldon, and Y. Zeng
Operational Earthquake Forecasting: (real-time forecast of earthquake likelihoods) Involves (or will involve): 1) A continual updating of authoritative information about the future occurrence of potentially damaging earthquakes (including aftershocks) 2) The officially sanctioned dissemination of this information to enhance earthquake preparedness in threatened communities. e.g., Jordan et al., 2011, Operational Earthquake Forecasting: State of knowledge and guidelines for implementation, final report of the International Commission on Earthquake Forecasting for Civil Protection, Annals Geophys., 54(4), 315-391, doi:10.4401/ag-5350
2010 Haiti Earthquake: USGS PAGER Notifications (e.g., Wald et al., 2008) provide economic loss and fatality estimates following large events After this event, the USGS received requests for information regarding the possibility of triggered earthquakes
2010 Haiti Earthquake: ad hoc report 9 days later Aftershocks: will continue for months if not years events will diminish with time activity during a 30-day period beginning January 21, 2010, is: 3% probability of M 7 events 25% probability of M 6 events 90% probability of M 5 events Expect 2-3 M 5 events
USGS OEF Goal: PAGER-type loss estimates from possibly triggered events?
Uniform California Earthquake Rupture Forecast Version 3 (UCERF3) Building an Operational Earthquake Forecast for California 2014 Working Group on California Earthquake Probabilities Field, E. H., J. R. Arrowsmith, G. P. Biasi, P. Bird, T. E. Dawson, K. R. Felzer, D. D. Jackson, K. M. Johnson, T. H. Jordan, C. Madden, A. J. Michael, K. R. Milner, M. T. Page, T. Parsons, P. M. Powers, B. E. Shaw, W. R. Thatcher, R. J. Weldon, and Y. Zeng
CEA NSF USGS Sources of WGCEP funding Geoscience organizations SCEC USGS Menlo Park USGS Golden WGCEP Organization & Funding Sources Management oversight committee MOC Scientific review panel SRP State of CA CGS Working Group on California Earthquake Probabilities WGCEP ExCom Working group leadership Subcom. A Subcom. B Subcom. C Task-oriented subcommittees
CEA Other core members: Geoscience Morgan Page organizations Karen Felzer NSF Peter Powers SCEC Bruce Shaw USGS USGS Glenn Biasi Menlo Park Dave Jackson Sources of Art Frankel USGS WGCEP funding Jeanne Hardebeck Golden Peter Powers State of CA Kevin Milner CGS Wayne Thatcher Kaj Johnson Yuehua Zeng Peter Bird Chris Madden Others Working Group on California Earthquake Probabilities WGCEP Organization & Funding Sources Management oversight committee MOC Ned Field (USGS, Chair) Thomas Parsons (USGS, Menlo Park) Tim Dawson (CGS) Ray Weldon (U of O) Andy Michael (USGS, Menlo Park) WGCEP ExCom Scientific review panel Working group leadership SRP Subcom. A Subcom. B Subcom. C Task-oriented subcommittees
CEA NSF USGS Sources of WGCEP funding State of CA Geoscience organizations SCEC USGS Menlo Park USGS Golden CGS WGCEP Organization & Funding Sources Management oversight committee MOC Thomas H. Jordan (SCEC, Chair) Thomas Brocher (USGS, Menlo Park ) Jill McCarthy (USGS, Golden ) Chris Wills (CGS) Scientific review panel SRP Working Group on California Earthquake Probabilities WGCEP ExCom Working group leadership Subcom. A Subcom. B Subcom. C Task-oriented subcommittees
CEA NSF USGS Sources of WGCEP funding State of CA Geoscience organizations SCEC USGS Menlo Park USGS Golden CGS Working Group on California Earthquake Probabilities WGCEP Organization & Funding Sources Management oversight committee MOC WGCEP ExCom Scientific review panel SRP Bill Ellsworth (chair) Art Frankel Duncan Agnew Ramon Arrowsmith Yehuda Ben-Zion Greg Beroza (PC Liason) Mike Blanpied David Schwartz Sue Hough Working Warner group Marzocchi leadership Rick Shoenberg Hamid Haddadi Subcom. A Subcom. B Subcom. C Task-oriented subcommittees This has been a participatory review
CEA Geoscience organizations Process NSF also evaluated SCEC by both NEPEC & Funding & CEPEC Sources Management oversight Scientific review (National and California Earthquake Prediction committee Evaluation Councils) panel USGS USGS Menlo Park Sources of WGCEP funding USGS Golden WGCEP Organization MOC SRP State of CA CGS Working Group on California Earthquake Probabilities WGCEP ExCom Working group leadership Subcom. A Subcom. B Subcom. C Task-oriented subcommittees
CEA NSF USGS Sources of WGCEP funding State of CA ~25% of funding ($2,000,000) Geoscience organizations SCEC USGS Menlo Park USGS Golden CGS WGCEP Organization & Funding Sources Management oversight committee MOC What is CEA and where did it come from? Scientific review panel SRP Working Group on California Earthquake Probabilities WGCEP ExCom Working group leadership Subcom. A Subcom. B Subcom. C Task-oriented subcommittees
Birth of the California Earthquake Authority (CEA) In California, homeowners insurance providers are required by law to offer earthquake coverage Figure from Glenn Pomeroy
Birth of the California Earthquake Authority (CEA) In California, homeowners insurance providers are required by law to offer earthquake coverage The 1994 M 6.7 Northridge earthquake caused a crisis Figure from Glenn Pomeroy
Birth of the California Earthquake Authority (CEA) In California, homeowners insurance providers are required by law to offer earthquake coverage The 1994 M 6.7 Northridge earthquake caused a crisis Almost all insurance companies stopped providing homeowners insurance, which meant banks stopped providing mortgages
Birth of the California Earthquake Authority (CEA) In California, homeowners insurance providers are required by law to offer earthquake coverage The 1994 M 6.7 Northridge earthquake caused a crisis Almost all insurance companies stopped providing homeowners insurance, which meant banks stopped providing mortgages The state stepped in and established the CEA, which is essentially a staterun earthquake reinsurance pool (they insure the insurance companies)
Birth of the California Earthquake Authority (CEA) In California, homeowners insurance providers are required by law to offer earthquake coverage The 1994 M 6.7 Northridge earthquake caused a crisis Almost all insurance companies stopped providing homeowners insurance, which meant banks stopped providing mortgages The state stepped in and established the CEA, which is essentially a staterun earthquake reinsurance pool (they insure the insurance companies) Who decides best available science? CEA is currently the largest earthquake insurance provider in the US (although only 12% of California homeowners have earthquake insurance) They are required by law to use best available science and any regional difference in insurance rates must be backed by science. They can also apply time-dependent probabilities since insurance rates are updated regularly.
Working Groups on California Earthquake Probabilities (WGCEPs) The most official time-dependent earthquake forecasts for California UCERF2 2007 UCERF3 2014 UCERF3 (2014)
All models are wrong, some are useful (Box, 1979) More specifically: All models are an approximation of the system they represent
Working Groups on California Earthquake Probabilities (WGCEPs) The most official time-dependent earthquake forecasts for California A better and better approximation UCERF2 2007 UCERF3 2014 UCERF3 (2014)
Working Groups on California Earthquake Probabilities (WGCEPs) The most official time-dependent earthquake forecasts for California How do we build these models? UCERF2 2007 UCERF3 2014 UCERF3 (2014)
Working Groups on California Earthquake Probabilities (WGCEPs) The most official time-dependent earthquake forecasts for California How do we build these models? UCERF2 2007 UCERF3 2014 UCERF3 (2014)
A tale of two perspectives: Geology & Paleoseismology Statistical Seismology
Working Groups on California Earthquake Probabilities (WGCEPs) The most official time-dependent earthquake forecasts for California Earliest models based mostly on the geologic/paleoseismic perspective UCERF2 2007 UCERF3 2014 UCERF3 (2014)
Geology & Paleoseismology Perspective: Assume fault ruptures fully and only by itself (segmented/characteristic earthquake hypothesis) Compute magnitude from area ( e.g., M=4+log(Area) ) Compute frequency of rupture from slip-rate, or set as paleoseismically inferred event rate Assume Reid s (1910) elastic rebound theory, where probability drops after event and builds with time as tectonic stresses re-accumulate (probability depends on date of last event) Key Assumptions Earthquake recurrence intervals are decades to centuries; aftershocks regarded as small and negligible
Working Groups on California Earthquake Probabilities (WGCEPs) The most official time-dependent earthquake forecasts for California Earliest models based mostly on the geologic/paleoseismic perspective UCERF2 2007 UCERF3 2014 UCERF3 (2014)
A tale of two perspectives: Geology & Paleoseismology Statistical Seismology
Statistical Seismology Perspective: Gutenberg- Richter Distribution in a region Omori-Utsu Law Key Assumption: these small-earthquake statistics apply to large damaging events Distance Decay
Statistical Seismology Perspective: Gutenberg- Richter Distribution Geologic/Paleoseismic rates (on a fault) Omori-Utsu Law Key Assumption: these small-earthquake statistics apply to large damaging events Distance Decay
Statistical Seismology Perspective: Gutenberg- Richter Distribution The USGS has used these statistical models to state aftershock probabilities based on Reasenberg & Jones (1989, 1990, 1994) Omori-Utsu Law Distance Decay
Routinely for M 5 events in California No info on proximity to populated areas
ad hoc assessments elsewhere e.g., 2010 Haiti Earthquake:
USGS s Short-Term Earthquake Probability (STEP) Model Gerstenberger et al. (2005) A real-time aftershock hazard map based on statistical seismology: Gutenberg Richter æ 1 Rate µ10 -M ç è t + c ( ) p Omori Law Distance Decay ö ø 1 r n
USGS s Short-Term Earthquake Probability (STEP) Model Gerstenberger et al. (2005) Reviewed and approved for use by the California Earthquake Prediction Evaluation Council (CEPEC), which advises the state governor & office of emergency services.
USGS s Short-Term Earthquake Probability (STEP) Model Gerstenberger et al. (2005) Limitations: 1) Probability of triggering large damaging aftershock is independent of proximity to large faults
USGS s Short-Term Earthquake Probability (STEP) Model Gerstenberger et al. (2005) Limitations: 1) Probability of triggering large damaging aftershock is independent of proximity to large faults 2) The is no elastic rebound Highest probability for any given rupture will be the moment after it occurs
USGS s Short-Term Earthquake Probability (STEP) Model Gerstenberger et al. (2005) Limitations: 1) Probability of triggering large damaging aftershock is independent of proximity to large faults 2) The is no elastic rebound 3) We d like synthetic catalogs of events rather than a mean forecast rate (e.g., to define complete distribution of possible losses rather than just mean value)
Working Groups on California Earthquake Probabilities (WGCEPs) The most official time-dependent earthquake forecasts for California Earliest models based mostly on the geologic/paleoseismic perspective UCERF3 - An attempt to find consilience: Geology Paleoseismology Statistical Seismology UCERF2 2007 UCERF3 2014 UCERF3 (2014)
UCERF2 Problems: 1) Assumes segmentation These inadequacies were recognized in the UCERF2 report, and since exemplified by the Christchurch, Tohoku, and other earthquakes. 2) Excludes multi-fault ruptures 3) Over-predicts M ~6.7 events(?) 4) Elastic rebound not self-consistent Christchurch NZ 5) Lacks spatiotemporal clustering
UCERF2 Problems: UCERF3 Solutions: 1) Assumes segmentation 2) Excludes multi-fault ruptures 3) Over-predicts M ~6.7 events 4) Elastic rebound not self-consistent New method supported by physics-based simulators 5) Lacks spatiotemporal clustering ETAS Operational Eqk Forecasting
Uniform California Earthquake Rupture Forecast Version 3 (UCERF3) Does not assume segmentation or the characteristic earthquake hypothesis (includes multi-fault ruptures) Includes both elasticrebound and spatiotemporal clustering (aftershocks)
Uniform California Earthquake Rupture Forecast Version 3 (UCERF3) Does not assume segmentation or the characteristic earthquake hypothesis (includes multi-fault ruptures) Includes both elasticrebound and spatiotemporal clustering (aftershocks) Uses Epidemic Type Aftershock Sequence model (ETAS; Ogata, 1988) to generate synthetic catalogs of M 2.5 events Main Shock Primary Aftershocks Secondary Aftershocks Tertiary Aftershocks
Epidemic Type Aftershock Sequence (ETAS) Model An empirically based description of triggering statistics Main Shock Key Assumption: l(t,x) =l o m x ( ) å ( t - t i + c) -p c S r + d ( ) + k10 a M i -M min ì í î i : t i < t ü ý þ Statistics inferred from small earthquakes apply to large (damaging) ones. ( ) -q Primary Aftershocks Secondary Aftershocks Hardebeck, Appendix S Tertiary Aftershocks
Example 1-year simulations: Northridge Landers Spontaneous 1 st generation 2 nd generation 3 rd generation
Example 1-year simulations for M7 Mojave SAF event (Tom Jordan s nightmare ):
Example 1-year simulations for M7 Mojave SAF event (Tom Jordan s nightmare ):
Toward operational loss modeling
We pre-compute economic losses and fatalities for every UCERF3 rupture (~500,000) using an OpenSHA implementation of the HAZUS-MH methodology (Porter et al., 2012, SRL): Exposure for California single-family dwellings
We pre-compute economic losses and fatalities for every UCERF3 rupture (~500,000) using an OpenSHA implementation of the HAZUS-MH methodology (Porter et al., 2012, SRL): For a given ETAS simulation, we sum losses for all events that occurred in the synthetic catalog to get a loss estimate.
We pre-compute economic losses and fatalities for every UCERF3 rupture (~500,000) using an OpenSHA implementation of the HAZUS-MH methodology (Porter et al., 2012, SRL): For a given ETAS simulation, we sum losses for all events that occurred in the synthetic catalog to get a loss estimate. Repeat to obtain N different simulated catalogs
We pre-compute economic losses and fatalities for every UCERF3 rupture (~500,000) using an OpenSHA implementation of the HAZUS-MH methodology (Porter et al., 2012, SRL): For a given ETAS simulation, we sum losses for all events that occurred in the synthetic catalog to get a loss estimate. Repeat to obtain N different simulated catalogs Make a histogram of loss values, giving a probability distribution of possible values.
Not just mean expected loss Gains depend on forecast duration For single-family dwellings, but full inventory can also be used Fatalities also available
We now have an operationalizable, end-to-end system to forecast losses in California that: Relaxes segmentation and includes multifault ruptures Includes elastic rebound and spatiotemporal clustering Generates synthetic catalogs (stochastic event sets) Includes very efficient loss calculations
We now have an operationalizable, end-to-end system to forecast losses in California that: Relaxes segmentation and includes multifault ruptures Includes elastic rebound and spatiotemporal clustering Generates synthetic catalogs (stochastic event sets) Includes very efficient loss calculations This is required to get realistic results Otherwise ~85% of triggered large events would re-rupture the same fault (Field, 2012, SSA), which we don t see Leaving it out also produces doomsday sequences, and screws up Båth's Law. This is what has taken us so long
We now have an operationalizable, end-to-end system to forecast losses in California that: Relaxes segmentation and includes multifault ruptures Includes elastic rebound and spatiotemporal clustering Generates synthetic catalogs (stochastic event sets) Includes very efficient loss calculations All models are wrong, some are useful (Box, 1979)
We now have an operationalizable, end-to-end system to forecast losses in California. UCERF3 has assumptions, approximations, and corrections Every component will get a thorough vetting as we operationalize Your work is an important part of this effort
Conclusion: OEF (with losses) is within reach