Working Group: Working Group on California Earthquake Probabilities (WGCEP)
GROUP LEADERS

PAST RESEARCH RESULTS

MEETINGS & WORKSHOPS

Item 1
Item 2 Item 3 
RELATED RESEARCH

Item 1
Item 2 Item 3 
Research Objectives
The WGCEP is a collaboration between SCEC, the USGS, and CGS aimed at developing official earthquakeruptureforecast models for California. The project is closely coordinated with the USGS National Seismic Hazard Mapping Program, and has received financial support from the California Earthquake Authority (CEA). The WGCEP has now completed the timeindependent UCERF3 model (UCERF3TI, which relaxes segmentation and includes multifault ruptures) and the longterm, timedependent model (UCERF3TD, which includes elasticrebound effects). We are now working on adding spatiotemporal clustering (UCERF3ETAS) to account for the fact that triggered events can be large and damaging. As the latter will require robust interoperability with realtime seismicity information, UCERF3ETAS will bring us into the realm of operational earthquake forecasting (OEF). We are also starting to plan for UCERF4, which we anticipate will utilize physicsbased simulators to a greater degree (see last bullet below).
Example Research Strategies
 Evaluate fault models in terms of the overall fault connectivity at depth (important for understanding the likelihood of multifault ruptures) and the extent to which faults represent a welldefine surface versus a proxy for a braided deformation zone.
 Evaluate existing deformation models, or develop new ones, in terms of applicability of GPS constraints, categorical sliprate assignments (based on “similar” faults), applicability of backslip methods, and other assumptions. Of particular interest is the extent to which slip rates taper at the ends of faults and at fault connections.
 Evaluate the UCERF3 implication that 30% to 60% of offfault deformation is aseismic.
 Help determine the average alongstrike slip distribution of large earthquakes, especially where multiple faults are involved (e.g., is there reduced slip at fault connections?).
 Help determine the average downdip slip distribution of large earthquakes (the ultimate source of existing discrepancies in magnitudearea relationships). Are surface slip measurements biased with respect to slips at depth?
 Develop a better understanding of the distribution of creeping processes and their influence on both rupture dimension and seismogenic slip rate.
 Contribute to the compilation and interpretation of mean recurrenceinterval constraints from paleoseismic data and/or develop sitespecific models for the probably of events going undetected at a paleosiesmic site.
 Develop ways to constrain the spatial distribution of maximum magnitude for background seismicity (for earthquakes occurring off of the explicitly modeled faults).
 Address the question of whether small volumes of space exhibit a Gutenberg Richter distribution of nucleations (even on faults).
 Develop improved estimates (including uncertainties) of the total longterm rates of observed earthquakes for different sized volumes of space.
 Refine our magnitude completeness estimates (as a function of time, space, and magnitude). Develop such models for realtime applications (as will be needed in operational earthquake forecasting).
 Develop methods for quantifying elasticrebound based probabilities in unsegmented fault models.
 Help quantify the amount of slip in the last event, and/or average slip over multiple events, on any major faults in California (including variations along strike).
 Develop models for faulttofault rupture probabilities, especially given uncertainties in fault endpoints.
 Determine the extent to which seismicity rates vary over the course of historical and instrumental observations (the socalled Empirical Model of previous WGCEPs), and the extent to which this is explained by aftershock statistics.
 Determine the applicability of higherresolution smoothedseismicity maps for predicting the location of larger, more damaging events.
 Explore the UCERF3 “Grand Inversion” with respect to: possible plausibility filters, relaxing the UCERF2 constraints, not overfitting data, alternative equationset weights, applying a characteristicslip model, and applicability of the Gutenberg Richter hypothesis on faults (see report at www.WGCEP.org).
 Develop applicable methods for adding spatiotemporal clustering to forecast model s(e.g., based on empirical models such as ETAS). Are sequencespecific parameters warranted?
 Determine if there is a physical difference between a multifault rupture and a separate event that was triggered quickly.
 Develop more objective ways of setting logictree branch weights, especially where there are either known or unknown correlations between branches.
 Develop easily computable hazard or loss metrics that can be used to evaluate and perhaps trim logictree branches.
 Develop techniques for downsampling event sets to enable more efficient hazard and loss calculations.
 Develop novel ways of testing UCERF3, especially ones that can be integrated with CSEP.
 Study and test the behavior of computational earthquakecycle simulators, envisioning that they could become essential ingredients in future UCERF projects and a cornerstone of SCEC5. The goal is to develop the capability of simulators to be able to contribute meaningfully to hazard estimates. Examples of important tasks:
 Study and test, using code verification exercises and more than one code, the sensitivity of simulator results to input details including faultsystem geometry, stressdrop values, tapering of slip, methods of encouraging rupture jumps from fault to fault, cell size, etc.
 Develop physically realistic ways of simulating offfault seismicity.
 Add additional physics into simulators, for example, the inclusion of highspeed frictional weakening and of offfault viscoelastic and heterogeneous elastic properties.
 Develop alternate methods of driving fault slip besides “backslip”.
 Make access to existing simulators easy for new users, including adequate documentation and version numbers, examples of input and output files for initial testing, and access to analysis tools. Publicize availability.
 Develop new approaches to designing simulators and/or of making them more computationally efficient, including the use of better algorithms, point source Greens functions, and GPUs.
 Develop validation tools for simulators, utilize existing UCERF data comparison tools with them, and develop capabilities for simulators to interact with UCERF infrastructure.
 Develop the capability of simulators to deal with UCERF and SCEC CFM fault geometries, both for rectangular and triangular cell representations.
 Create statewide synthetic earthquake catalogs spanning 100 My using as many different simulators as possible, in order to generate statistically significant behavior on even slowslipping faults. Use small timesteps to permit evaluation of shortterm clustering.
 Use these catalogs as synthetic laboratories for CSEP testing as described under CSEP.
 Datamine these catalogs for statistically significant patterns of behavior. Evaluate whether muchshorter observed catalogs are statistically distinguishable from simulated catalogs. Consider and explore what revisions in simulators would make simulated catalogs indistinguishable from observed catalogs.
 Develop and test a variety of statistical methods for determining the predictability of the of earthquakes in these simulated catalogs.
 Compute other data types such as gravity changes, surface deformation, InSAR images, in order to allow additional comparisons between simulated results and observations.
Further suggestions and details can be found at http://www.WGCEP.org, or by contacting the project leader (Ned Field: field@usgs.gov; (626) 6446435).
Recent Results
The Working Group on California Earthquake Probabilities (WGCEP) is charged with developing official, consensus, and timedependent earthquake forecast models for California. The effort builds on a long tradition of previous WGCEPs (e.g., models published in 1988, 1990, 1995, 2003, and 2008), and involves explicit collaboration between SCEC, the USGS, and CGS, with considerable funding from the California Earthquake Authority (http://www.earthquakeauthority.com). The previous WGCEP model was the Uniform California Earthquake Rupture Forecast version 2 (UCERF2, http://www.scec.org/ucerf2/), which was published in 2008. Since that time we have been working on the next model, UCERF3, for which the main goals have been to: 1) relax segmentation and include multifault ruptures; 2) develop an algorithm for computing more selfconsistent longterm elasticreboundbased probabilities; and 3) include spatiotemporal clustering effects in acknowledgement that aftershocks and triggered events can be large and damaging. The latter (spatiotemporal clustering) will bring us into the realm of Operational Earthquake Forecasting (OEF). The need for these enhancements has been exemplified by several recent earthquakes, including the 2011 M9 Tohoku earthquake with respect to segmentation, both the 2010 M7.2 El MayorCucapah and 2012 M8.6 Sumatra earthquakes in regard to multifault ruptures, and the 2011 M6.3 Christchurch earthquake in terms of spatiotemporal clustering. Progress on each of these goals is given below.
UCERF3TI, The TimeIndependent Model
Figure 1. First page of the UCERF3 fact sheet (http://dx.doi.org/10.3133/fs20153009) 
The backbone of UCERF3 is the longterm, timeindependent model (UCERF3TI), which was published as a USGS OpenFile Report on Nov. 5, 2013, and includes a main report, 20 appendices, and various supplements (http://pubs.usgs.gov/of/2013/1165/). The main report and one of the appendices have also been published in a peerreviewed journal (as Field et al. (2014) and Page et al. (2014), respectively). The primary achievement for this model component was relaxing fault segmentation and including multifault ruptures, both limitations of UCERF2. The rates of all earthquakes were solved for simultaneously, and from a broader range of data, using a systemlevel “grand inversion” that is both conceptually simple and extensible. The inverse problem is large and underdetermined, so a range of models was sampled using an efficient simulated annealing algorithm. The approach is more derivative than prescriptive (e.g., magnitudefrequency distributions are no longer assumed), so new analysis tools were developed for exploring solutions. Epistemic uncertainties were also accounted for using 1440 alternative logic tree branches, necessitating access to supercomputers. The most influential uncertainties include alternative deformation models (fault slip rates), a new smoothed seismicity algorithm, alternative values for the total rate of M≥5 events, and different scaling relationships, virtually all of which are new. As a notable first, three deformation models are based on kinematically consistent inversions of geodetic and geologic data, also providing sliprate constraints on faults previously excluded due to lack of geologic data. The grand inversion constitutes a systemlevel framework for testing hypotheses and balancing the influence of different experts. For example, we have demonstrated serious challenges with the GutenbergRichter hypothesis for individual faults. UCERF3TI is still an approximation of the system, however, and the range of models is limited (e.g., constrained to stay close to UCERF2). Nevertheless, UCERF3TI removes the apparent UCERF2 overprediction of M6.57 earthquake rates and also includes types of multifault ruptures seen in nature. Although UCERF3TI fits the data better than UCERF2 overall, there may be areas that warrant further sitespecific investigation. Finally, the supporting products may be of general interest, and we listed key assumptions and avenues for future model improvements in the report.
UCERF3TD, The LongTerm, TimeDependent Model
This model, which builds on UCERFTI, includes longterm, timedependent probabilities based on Reid’s elasticrebound hypothesis, which posits that rupture likelihood drops on a fault after experiencing a large rupture and then builds back up as tectonic stresses reaccumulate with time. A new methodology was developed (Field, 2015) that solves applicability issues in the previous approach for unsegmented models. The new methodology also supports magnitudedependent aperiodicity and accounts for the historic open interval on faults that lack a dateoflastevent constraint (Field and Jordan, 2015). Epistemic uncertainties are represented with a logic tree, producing 5,760 different forecasts. Results for a variety of evaluation metrics have been presented, including logictree sensitivity analyses and comparisons to the previous model (UCERF2). For 30year M≥6.7 probabilities, the most significant changes from UCERF2 are a threefold increase on the Calaveras fault and a threefold decrease on the San Jacinto fault. Such changes are due mostly to differences in the timeindependent models (e.g., fault slip rates), with relaxation of segmentation and inclusion of multifault ruptures being particularly influential. In fact, some UCERF2 segments were simply too long to produce M 6.7 sized events. Probability model differences are also influential, with the implied gains (relative to a Poisson model) being generally higher in UCERF3. Accounting for the historic open interval is one reason. Another is an effective 27% increase in the total elasticreboundmodel weight. The exact factors influencing differences between UCERF2 and UCERF3, as well as the relative importance of logictree branches, vary throughout the region, and they depend on the hazard metric of interest (e.g., M≥6.7 probability changes may not translate to hazard). This sensitivity, coupled with the approximate nature of the model, as well as known limitations, means the applicability of UCERF3 should be evaluated on a casebycase basis. Overall, UCERF3 represents the best model currently available for forecasting California earthquakes. UCERF3TD was been reviewed by our Scientific Review Panel, including the aforementioned supporting papers, and the main report was published in the Bulletin of the Seismological Society of America (Field et al, 2015). A USGS fact sheet was also published (http://pubs.usgs.gov/fs/2015/3009/), and we had a press release on the day the model went public (http://www.usgs.gov/newsroom/article.asp?ID=4146).
UCERF3ETAS, Spatiotemporal Clustering for OEF
With the timeindependent and timedependent models published (described above), we have now turned our attention to including spatiotemporal clustering. In recognition that triggered events can be large and damaging, the ultimate goal is to deploy an Operational Earthquake Forecast (OEF) for California, now listed as one of the USGS‘s strategicaction priorities (http://pubs.usgs.gov/of/2012/1088; page 32). In short, OEF aims to provide realtime forecasts to help communities prepare for earthquakes. To this end, we have added an Epidemic Type Aftershock Sequence (ETAS) component to UCERF3 (UCERF3ETAS). Most notably, our model represents a merging of ETAS with finitefault based forecasts, as well as the inclusion of elastic rebound (both firsts, as far as we are aware). In fact, inclusion of elasticrebound turns out to be critical in terms of getting spatiotemporal clustering statistic correct (otherwise ~85% of large triggered events simply rerupture the same fault, which we don’t see in nature). UCERF3ETAS is currently being “testdriven”. Our intent is to continue documenting the model and subjecting it to more rigorous testing (e.g., via CSEP) over the next year. Toward operationalization, the USGS and SCEC are cofunding a series of OEFrelated workshops at the USGS Powell Center in Fort Collins, CO (https://powellcenter.usgs.gov). The first workshop, held in March 2015, addressed the "Potential Uses of OEF", for which a report has been written and submitted to Seismological Research Letters (Field et al., 2016). Forthcoming workshop topics include "Best Available Science for OEF", "Operationalization Challenges for OEF", and "Verification and Validation of OEF", where the latter includes testing effectiveness of product messaging.
Select Publications
 UCERF3TI OpenFile Report, including 20 appendices: http://pubs.usgs.gov/of/2013/1165/
 Field, E. H., R. J. 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 (2014). Uniform California Earthquake Rupture Forecast, version 3 (UCERF3)—The timeindependent model, Bull. Seism. Soc. Am, Vol. 104, No. 3, pp. 1122–1180, June 2014, doi: 10.1785/0120130164
 Page, M. T., E. H. Field, K. R. Milner, and P. M. Powers (2014). The UCERF3 Grand Inversion: Solving for the Long‐Term Rate of Ruptures in a Fault System, Bull. Seism. Soc. Am., v. 104, p. 11811204, doi:10.1785/0120130180. SCEC Contribution 1999
 Field, E. H., R. J. 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 (2015). LongTerm, TimeDependent Probabilities for the Third Uniform California Earthquake Rupture Forecast (UCERF3), Bull. Seism. Soc. Am. 105, 511–543, doi: 10.1785/0120140093. SCEC Contribution 2099
 Field, E. H. (2015). Computing elasticreboundmotivated earthquake probabilities in unsegemented fault models – A new methodology supported by physicsbased simulators, Bull. Seismol. Soc. Am. 105, doi: 10.1785/0120140094. SCEC Contribution 2065
 Field, E. H., and T. H. Jordan (2015). TimeDependent renewalmodel probabilities when date of last earthquake is unknown, Bull. Seismol. Soc. Am. 105, 459–463, doi: 10.1785/0120140096. SCEC Contribution 1991
 Field, E.H., and 2014 Working Group on California Earthquake Probabilities, 2015, UCERF3: A new earthquake forecast for California’s complex fault system: U.S. Geological Survey Fact Sheet 2015–3009, 6 p., http://dx.doi.org/10.3133/fs20153009.
 Field, E. H., T. H. Jordan, L. M. Jones, A. J. Michael, M. L. Blanpied, and the other workshop participants (2016). The Potential Uses of Operational Earthquake Forecasting (A Workshop Report), submitted to Seismological Research Letters. SCEC Contribution 6029