Field, E. H., Milner, K. R., Hardebeck, J. L., Page, M. T., van der Elst, … · Elst, Thomas H. Jordan, Andrew J. Michael, Bruce E. Shaw, and Maximilian J. Werner Abstract We, the

Field E H Milner K R Hardebeck J L Page M T van der Elst NJordan T H Michael A J Shaw B E amp Werner M (2017) ASpatiotemporal Clustering Model for the Third Uniform CaliforniaEarthquake Rupture Forecast (UCERF3-ETAS) Toward an OperationalEarthquake Forecast Bulletin of the Seismological Society of America107(3) 1049-1081 httpsdoiorg1017850120160173

Peer reviewed version

Link to published version (if available)1017850120160173

Link to publication record in Explore Bristol ResearchPDF-document

This is the author accepted manuscript (AAM) The final published version (version of record) is available onlinevia SSA at httpbssageoscienceworldorgcontentearly201702240120160173 Please refer to anyapplicable terms of use of the publisher

University of Bristol - Explore Bristol ResearchGeneral rights

This document is made available in accordance with publisher policies Please cite only the publishedversion using the reference above Full terms of use are available httpwwwbristolacukpureuser-guidesexplore-bristol-researchebr-terms

A Spatiotemporal Clustering Model for the Third Uniform

California Earthquake Rupture Forecast (UCERF3-ETAS)

Toward an Operational Earthquake Forecast

by Edward H Field Kevin R Milner Jeanne L Hardebeck Morgan T Page Nicholas van derElst Thomas H Jordan Andrew J Michael Bruce E Shaw and Maximilian J Werner

Abstract We the ongoing Working Group on California Earthquake Probabil-ities present a spatiotemporal clustering model for the Third Uniform CaliforniaEarthquake Rupture Forecast (UCERF3) with the goal being to represent after-shocks induced seismicity and otherwise triggered events as a potential basis foroperational earthquake forecasting (OEF) Specifically we add an epidemic-type after-shock sequence (ETAS) component to the previously published time-independent andlong-term time-dependent forecasts This combined model referred to as UCERF3-ETAS collectively represents a relaxation of segmentation assumptions the inclusionof multifault ruptures an elastic-rebound model for fault-based ruptures and a state-of-the-art spatiotemporal clustering component It also represents an attempt to mergefault-based forecasts with statistical seismology models such that information on faultproximity activity rate and time since last event are considered in OEF We describeseveral unanticipated challenges that were encountered including a need for elasticrebound and characteristic magnitudendashfrequency distributions (MFDs) on faults bothof which are required to get realistic triggering behavior UCERF3-ETAS producessynthetic catalogs ofM ge25 events conditioned on any priorM ge25 events that areinput to the model We evaluate results with respect to both long-term (1000 year)simulations as well as for 10-year time periods following a variety of hypotheticalscenario mainshocks Although the results are very plausible they are not always con-sistent with the simple notion that triggering probabilities should be greater if a main-shock is located near a fault Important factors include whether the MFD near faultsincludes a significant characteristic earthquake component as well as whether largetriggered events can nucleate from within the rupture zone of the mainshock BecauseUCERF3-ETAS has many sources of uncertainty as will any subsequent version orcompeting model potential usefulness needs to be considered in the context of actualapplications

Electronic Supplement Figures showing discretization verification of theDistanceDecayCubeSampler average simulated participation rate and average cumu-lative MFD

Introduction

Long-term earthquake forecasts which are applicablefrom decades to centuries represent our first line of defensewith respect to mitigating earthquake risk especially in termsof informing building codes We also know however thataftershocks and otherwise triggered events can be large anddamaging as demonstrated by several earthquakes includingthe 2011M 63 Christchurch New Zealand event (eg Kai-ser et al 2012) The ability to forecast earthquakes on

shorter time scales such as days to decades is known as op-erational earthquake forecasting (OEF) which also involvesthe dissemination of authoritative information to inform riskmitigation decisions (Jordan et al 2011) The history mo-tivation and challenges associated with OEF have been dis-cussed elsewhere (Jordan and Jones 2010 Jordan et al2011 2014 Peresan et al 2012 Wang and Rogers 2014)and significant progress with OEF has been made in both

1

Bulletin of the Seismological Society of America Vol 107 No 1 pp ndash February 2017 doi 1017850120160173

Italy following the 2009 LrsquoAquila earthquake (Marzocchiet al 2014) and in New Zealand following the 2010DarfieldChristchurch events (Gerstenberger et al 2014)Furthermore a workshop was recently held to discussthe potential usefulness of OEF at which there was gener-ally broad support for the capability among a wide varietyof stakeholders (Field et al 2016) With the goal of an OEFsystem for California in mind we present a spatiotemporalclustering component for the Third Uniform CaliforniaEarthquake Rupture Forecast (UCERF3) developed by theongoing Working Group on California Earthquake Proba-bilities (WGCEP)

Modeling Goals

Because our stated goal is to help communities preparefor potentially destructive earthquakes viability of a modelcomes down to its reliability and skill with respect to fore-casting larger triggered events

Table 1 lists a variety of information that one mightconsider when addressing this question with the overallchallenge being how to integrate such constraints into a sys-tem-level model that embodies consilience not only amongthese constraints but also among whatever underlying as-sumptions are being made throughout the model Our modelalso provides forecasts of smaller events which are also use-ful for OEF messaging

An important OEF milestone was the development ofthe short-term earthquake probability (STEP) model (Ger-stenberger et al 2005) which provides real-time aftershockhazard estimates assuming the following as combined byReasenberg and Jones (1989 1994) a GutenbergndashRichter(GR) magnitudendashfrequency distribution (MFD Gutenbergand Richter 1944) the Utsu relationship for the numberof aftershocks as a function of mainshock magnitude (Utsu1971) and the modified Omori law for the temporal behaviorof aftershocks (Utsu 1961) The US Geological Survey(USGS) posted STEP forecasts for California starting in

2005 but the system was taken offline in 2010 due to soft-ware maintenance issues Nevertheless STEP continues tobe used to inform decision making in other parts of the world(eg Gerstenberger et al 2014) In lieu of a detailed reviewof that model here Table 1 indicates which types of con-straints are embodied in STEP together with those utilizedin the UCERF3 epidemic-type aftershock sequence (ETAS)model presented here Many other candidate OEF modelshave also been developed and tested since the introductionof STEP (eg Werner et al 2011 Woessner et al 2011Nanjo et al 2012 and references therein Segou et al2013 Zechar et al 2013 and references therein Helmstetterand Werner 2014 Marzocchi et al 2014 Gerstenbergeret al 2014) What makes UCERF3-ETAS unique to allof these is a more explicit and complete incorporation of geo-logic fault information as well as the inclusion of elastic re-bound (in which rupture probabilities are thought to drop ona fault after experiencing a large event and to grow back withtime as tectonic stresses re-accumulate) In other words ourmodel attempts to merge the point-process-clustering modelsdeveloped by statistical seismologists (eg Ogata 1988 Re-asenberg and Jones 1989 1994 Gerstenberger et al 2005)with the geologically based renewal models typically appliedin official long-term time-dependent forecasts (eg WGCEP1988 1990 1995 2003 2007 Field et al 2009 2015)

All scientifically viable models are ultimately wrongbecause in addition to embodying assumptions and approx-imations we possess only estimates of the constraints listedin Table 1 What we really hope is to have a model that isnevertheless useful for both risk-mitigation efforts as alreadyemphasized and for improving our scientific understandingof earthquakes To both these ends an important goal for ourmodel has been the ability to generate synthetic catalogs alsoknown as stochastic event sets On the practical side gener-ating suites of synthetic catalogs enables constructing a com-plete probability distribution of potential losses as opposedto being limited to just mean loss estimates when utilizingrate-based forecasts such as STEP On the scientific side

Table 1Some of the Informational Constraints (Observed or Model Inferred) that One Might Use to Forecast the Probability

of Triggered Earthquakes

Observation or Model Inference STEP UCERF3-ETAS

Global or regional triggering statistics (generic aftershock parameters)

Sequence-specific deviations from generic aftershock parameters

Spatial and temporal variation in aftershock parameters (within a sequence)

Spatial variation in long-term event rates (background rates)

Location of recent events (ie areas that are active with microseismicity)

Proximity to active faults especially with respect to triggering larger magnitude events

The long-term magnitudendashfrequency distribution inferred for faults

Elastic-rebound implied stress on faults (eg time since last event relative to recurrence interval)

Dynamic or static stress changes imposed by any previous events Implicit Implicit

Also indicated are the constraints used by the short-term earthquake probability (STEP) model (Gerstenberger et al 2005) and by theThird Uniform California Earthquake Rupture Forecast (UCERF3) epidemic-type aftershock sequence (ETAS) A checkmark means thatthe constraint is used explicitly with respect to forecasting nonspontaneous (triggered) events whereas Implicit means that the influencemay be captured by other utilized constraints (eg static or dynamic stress changes being manifested in observed seismicity)

2 E H Field et al

synthetic catalogs provide a powerful and perhaps indispen-sable perspective with respect to testing and evaluating mod-els For example simulations have previously revealed thatexcluding elastic rebound leads to unrealistic aftershock sta-tistics such as accelerating sequences (eg Field 2012)

Another improvement over STEP is to sample triggeredevents from the very same population of ruptures defined inthe underlying long-term model For example we mightwant the likelihood of sampling an M 8 event to dependon the proximity to capable faults (like the San Andreas)rather than assuming the same maximum magnitude and rel-ative likelihood throughout the region In fact the CaliforniaEarthquake Prediction Evaluation Council (CEPEC) isknown to convene when a magnitude sim5 earthquake occursnear the San Andreas fault (SAF Jordan and Jones 2010)but not when such an earthquake occurs well away fromknown faults In addition to fault proximity one might alsowant to consider the activity rate of a fault (eg the meanrecurrence interval) as well as whether the fault has had timeto reload in an elastic-rebound sense

Schoenberg and Bolt (2000) suggested one way to com-bine clustering and elastic-relaxation processes when mod-eling seismicity but did not consider finite faults or non-GR distributions explicitly One of the first attempts to con-sider faults when defining earthquake triggering probabilitieswas the foreshock model of Agnew and Jones (1991) inwhich the a priori likelihood of having a large event basedon fault information in their examples is considered in com-puting the probability of the event being triggered by asmaller prior earthquake This foreshock model was not acomplete OEF model because it did not include aftershocksProbabilities obtained from the Agnew and Jones (1991)methodology are generally higher than those obtained usingOmorindashUtsuGR statistics which Michael (2012a) demon-strated to be a consequence of assuming a characteristic ver-sus GR MFD when sampling aftershocks Michael (2012a)also provided a generalized clustering model that relaxes theGR assumption and noted that having a characteristic distri-bution is the most effective way of getting above-average-triggering probabilities near a fault

The UCERF3-ETAS model presented here essentiallyrepresents an elaborate implementation of the generalizedclustering model of Michael (2012a) because the model ef-fectively samples aftershocks according to the nearby MFDimplied by the UCERF3 long-term model Furthermore andas anticipated by Michael (2012a) we find that the degree ofcharacteristicness (the deviation from GR) is the most impor-tant factor when it comes to the likelihood of triggering largeevents from a given mainshock

Model Description

To achieve all the goals set out above the UCERF3 spa-tiotemporal component utilizes an ETAS model (Ogata1988) which is why we refer to the model as UCERF3-ETAS Elastic-rebound effects are included by building upon

the UCERF3 long-term time-dependent model (UCERF3-TD Field et al 2015) which in turn includes multifault rup-tures and relaxes fault-segmentation assumptions by virtue ofbeing built upon the time-independent model (UCERF3-TIField et al 2014) Details of these previously publishedmodels are reiterated here only to the extent that they areimportant for understanding UCERF3-ETAS Those wantinga more complete understanding may wish to consult theseprevious UCERF3 publications

Model Overview

The ETAS model is used to produce multiple realiza-tions of how an earthquake sequence may progress over timeEach simulation takes a list of previous M ge25 events asinput which could represent an actual catalog one or morehypothetical earthquake scenarios or both For each inputand simulated earthquake the latter of which includes spon-taneous events we randomly generate some number ofM ge25 aftershocks using the ETAS model Each event issampled from the long-term UCERF3 model according tothe current relative likelihood of each possible ruptureincluding both elastic-rebound effects on faults and proxim-ity to the mainshock for triggered events The hope orassumption is that ETAS represents an adequate statisticalproxy for whatever physics is operating in the system Fur-thermore by incorporating observations of smaller earth-quakes to inform the triggering potential of large ones (ieby updating the model with actual earthquake data) we hopeto capture any static or dynamic triggering effects that maybe playing out in the sequence and perhaps the potential im-plications of any induced seismicity

Although conceptually simple the actual UCERF3-ETAS implementation gets complicated with respect toincluding elastic rebound and dealing with uncertainties as-sociated with modeled faults Consequently UCERF3-ETAShas a number of adjustable parameters or variables whichare listed and defined in Table 2 Note that this is not a com-plete list of all possible adjustable parameters which wouldbe much longer but rather a list of the more important onesexplored here For example the TimeSpan specifies the starttime and duration for the desired forecast and ProbModelspecifies if and how elastic-rebound effects are applied inthe model Three of the adjustable parameters (ApplyGRcorrApplyGridSeisCorr and ApplySubSeisForSupraNucl) con-trol the degree of characteristicness allowed throughout themodel which we already noted as having a first-order influ-ence on triggering probabilities The last parameter (Total-RateScaleFactor) allows one to correct for any overall biasesin the total rate of simulated events All of these variableswhich are written in italics throughout this article are de-scribed in more detail below

The rest of this section describes the various componentsand algorithms utilized in UCERF3-ETAS with the goalbeing one or more synthetic earthquake catalogs for a speci-fied TimeSpan The model has been implemented using

A Spatiotemporal Clustering Model for the UCERF3-ETAS 3

OpenSHA which is an open-source object-oriented platformfor conducting seismic-hazard analysis (see Data and Re-sources) Care has been taken to implement generic compo-nents which are not specific to California in order tofacilitate applications elsewhere

The Long-Term Earthquake-Rupture Forecast (ERF)

One of the main model components is a long-term ERFwhich by definition specifies every possible earthquake rup-ture (at some acceptable level of discretization) as well as theassociated occurrence probability of each for a given regionand TimeSpan An ERFmay also have other adjustable param-eters to enable for example setting different levels of accu-racy or for specifying alternative logic-tree branches For thisstudy we utilize one of the UCERF3-TD and although anyone of the 5760 logic-tree-branch options could be chosenresults presented here were obtained using a branch-averagedmodel that is described in the Figure 1 caption

Although an ERF can be thought of as a long list of pos-sible earthquake ruptures the latter are actually bundled insidedifferent earthquake sources In other words an earthquakesource is a list of earthquake ruptures that are somehow re-lated Two types of sources are utilized in UCERF3 fault-based sources and gridded seismicity The first type representssupraseismogenic fault-based sources each of which involvesa unique contiguous set of two or more of the fault subsectionsshown in Figure 1 (in which supraseismogenic means that thealong-strike length of the rupture is greater than or equal to theaverage down-dip width of the faults involved and each sub-section has a length that is half the down-dip width) Eachfault-based source therefore represents a specific rupture area(defined by its unique collection of fault subsections) Rupturemagnitudes are determined from the source area so a fault-based source will have more than one rupture only if multiplemagnitudes are assigned for the given area such aleatory vari-ability is supported as an option in UCERF3 but not utilized inthe results shown here so each fault-based source corresponds

to a single fault-based rupture in this study As such there areabout 260000 fault-based sources including many that re-present multifault ruptures

For the time-dependent model applied here the proba-bility of each rupture depends on how long it has been sinceeach associated fault subsection experienced a supraseismo-genic rupture as defined by the elastic-rebound implemen-tation described in the UCERF3-TD report (Field et al2015) Stated simply these probabilities are computed usinga Brownian passage time (BPT) model after averaging the re-currence interval and time since last event over all the subsec-tions utilized by the given fault-based rupture The calculationis more complicated for faults wherewe only know that the lastevent predated some historical open interval see the UCERF3-TD documentation (Field et al 2015) for full details

Gridded seismicity is the other type of source inUCERF3 in which the region is discretized into about7700 different 01deg-by-01deg cells (Fig 1) and a nucleationMFD is assigned to each The sources represent subseismo-genic ruptures on or near the explicitly modeled faults aswell as all ruptures elsewhere to account for unmodeledfaults Each grid-based source has five ruptures for each dis-crete magnitude in the MFD one for each alternative focalmechanism represented in the model The number of rup-tures also increases if alternative finite rupture surfacesare included (also an option) although we treat gridded seis-micity as point sources for simplicity here The total numberof ruptures from grid-based ruptures is about 1700000

The ETAS Model

Aftershocks are sampled using the ETAS model intro-duced by Ogata (1988) which represents a generalizationof modified Omori aftershock statistics in terms of providinga more detailed accounting of event pedigree That is thesame statistical seismicity laws are used as in Reasenbergand Jones (1989) and STEP (Gerstenberger et al 2005)but in ETAS every earthquake can spawn others so that some

Table 2The UCERF3-ETAS Model Variables or Adjustable Parameters That Are Explored in This Article

Model Variable Name Description

ApplyGRcorr This indicates whether a simulation should force all fault sections to be consistent with a GutenbergndashRichter (GR)distribution in terms of total supraseismogenic rates which is achieved by dividing the nucleation rate of such rupturesby the subsection CharFactor value Default value is False

ApplyGridSeisCorr This tells whether to correct gridded seismicity rates so as to be greater than or equal to the rate of aftershocks implied bythe long-term rate of supraseismogenic ruptures Default value is True

ApplySubSeisForSupraNucl This option makes the location where a given supraseismogenic rupture is likely to nucleate from proportional to the long-term rate of smaller events along the fault surface (inside the polygons of associated fault subsections) Default value isTrue

ProbModel This specifies the type of probability model to apply with the options being POISSON (time independent) FULL_TD(fully time dependent including elastic-rebound triggering [ERT]) and NO_ERT (time dependent but excluding ERT)Default is FULL_TD

TimeSpan The start time and duration of the desired forecastTotalRateScaleFactor This defines the amount by which total long-term rupture rates in the earthquake-rupture forecast (ERF) are scaled to

obtain a better match between total simulated and target M ge25 rates Default value is 114 (a 14 increase)

This is not an exhaustive list of all possible model variables (eg those in Table 3 could be adjusted as well) but rather only the ones addressed in this study

4 E H Field et al

events in an aftershock sequence are not triggered by themainshock directly but indirectly through a previously trig-gered aftershock Events triggered by the first earthquake arereferred to as primary aftershocks those triggered by oneof the latter are secondary aftershocks and so forth ETASmakes no distinction between aftershocks and any other typeof triggered event

We use the ETAS formulation introduced by Ogata(1998) and used by Hardebeck (2013) in which the nucle-ation rate of M ge Mmin earthquakes as a function of time (t)and space (x) is given as

EQ-TARGETtempintralink-df155154

λt x λ0μx X

fitilttgk10αMiminusMmint minus ti cminusp

times csr dminusq 1

in which λ0 is the total rate of spontaneous events (back-ground events) and μx is the long-term spatial density

of M ge Mmin events The summation is over all events thathave occurred prior to time t in which the first term in brack-ets gives the rate evolution of primary aftershocks with krepresenting overall productivity α representing the magni-tude dependence of triggering p representing the temporaldecay rate and c preventing a singularity at t ti The sec-ond set of brackets in the summation gives the linear densityof aftershock triggering in which r represents distance fromthe rupture surface q represents the linear decay rate d pre-vents a singularity at r 0 and cs is a normalization factorthat ensures a value of 10 when the term is integrated over allspace We adopt the ETAS parameter values also given byHardebeck (2013) listed in Table 3 which were derivedfrom California seismicity Note that the α parameter is fixedto the b-value (10) which ensures that Baringthrsquos law holds in-dependent of mainshock magnitude (Felzer et al 2002)

The number of primary aftershocks expected from aparent of magnitude M is

Figure 1 3D perspective view of the types of earthquake sources utilized in the Third Uniform California Earthquake Rupture Forecast(UCERF3) Each fault-based sourcerupture is represented by a collection of two or more contiguous fault subsections the latter of which aredepicted as black-outlined parallelograms The 01deg-by-01deg grid draped over the entire region defines the gridded seismicity cells describedin the text The colors indicate the average likelihood that each cell or fault subsection will participate in anM ge67 earthquake in a 30-yearperiod following 2014 This study utilizes a branch-averaged UCERF3 long-term time-dependent (UCERF3-TD) model in which only FaultModel 31 is included only the UCERF3 smoothed seismicity branch is used because it performs better for short-term forecasts and only theMid-Aperiodicity branch is used for the elastic-rebound time dependence all other branches contribute to this average model according totheir respective weights (see the UCERF3-TD report including Fig 3 therein for more details on the various branch options) Although theUCERF3 time-independent (UCERF3-TI) and UCERF3-TD models have an option for removing aftershocks according to the Gardner andKnopoff (1974) declustering algorithm this option has not been applied here meaning that all earthquakes are included The minimummagnitude considered by the forecast is also lowered from M 5 to 25


EQ-TARGETtempintralink-df255538NM k10MminusMminZ

t2

t1

c tminuspdt

k10MminusMmin

1 minus pc t21minusp minus c t11minusp 2

in which t1 and t2 represent the start and end times of the Time-Span (relative to the origin time of the parent) Table 4 lists thenumber of primary aftershocks expected in 10 years followingvarious mainshock magnitudes and assuming parameter val-ues adopted here (Table 3) The number of aftershocks for sub-sequent generations (secondary and beyond) depends on theMFD from which one samples and Table 4 includes valuesassuming the total regional MFD for UCERF3-TI (Fig 2)which is GR consistent applies everywhere

One productivity metric we will use extensively in thisarticle is the likelihood that an earthquake will trigger anevent larger than itself which is also listed in Table 4 Theexpected number of primary aftershocks that are larger thanthe mainshock is generally 0053 which represents a 52likelihood of having one or more such events The expectednumber considering all generations is 0165 which repre-sents a 15 probability of one or more such aftershocksThese likelihoods are reduced at the highest mainshock mag-nitudes due to the tapering of the regional MFD (Fig 2)These values are also for a 10-year time period following themainshock for comparison values for 7 days following are33 and 56 for primary and all generations respectivelyand values for 1-year following are 46 and 11 respec-tively This increase in likelihood with forecast duration servesas a reminder that we should never think ldquoitrsquos a bit late for thisto be an aftershockrdquo (Michael 2012b p 630)

A subtle but potentially important issue is that sponta-neous events in the ETAS model represent in part a proxyfor aftershocks of unknown parents The extent to which thisis true can be quantified by the so-called branching ratio(Helmstetter and Sornette 2003) which gives the numberof primary aftershocks expected over infinite time for a pa-rent magnitude sampled randomly from the assumedMFD Itturns out that the branching ratio of the parameters for Har-

debeck (2013) is effectively 10 meaning each event triggerson average one other event implying that all events must betriggered and that the true rate of spontaneous events is zero(λ0 0) Of course we do not have a complete history of allpast events so we still need a nonzero value of λ0 to serve as aproxy for descendants of unknown parents Furthermore λ0must vary with time over the duration of a forecast For exampleif we have no information on past events then λ0 must equal thetotal rate of events at the beginning of the forecast (all events arespontaneous because there are no known parents) but as timeincreases λ0 will go down as the forecast generates events andλ0 will eventually become zero at infinite time

This issue is discussed at length by van der Elst (unpub-lished manuscript 2016 see Data and Resources) and weuse the equations therein to compute time-dependent fractionof spontaneous versus triggered events for our forecastswhich explicitly account for magnitude-dependent dates ofcompleteness in an earthquake catalog We use the UCERF3California catalog compiled by Felzer (2013) together withmagnitude-dependent completeness thresholds defined inher table 9 which thereby provides a list of historicalinstru-mental events between 1850 and 2012 Using this catalog ina forecast that begins in 2012 and given the total model MFDshown in Figure 2 the initial rate of spontaneous events (λ0)is 030 times the total regional rate this fractional valueevolves to 028 at 10 years into the forecast 024 at 100years and 020 at 1000 years Again these values depend onthe branching ratio implied by the ETAS parameters andalthough van der Elst (unpublished manuscript 2016 seeData and Resources) provides additional support for a valuenear 10 for California the practical implications and main

40 50 60 70 80

Magnitude

1 0ndash5

1 0ndash4

1 0ndash3

1 0ndash2

1 0ndash1

1 0 0

1 0 1

1 0 2

Cu

mu

lati

ve R

ate

(per

yr)

TotalOn Fault

Figure 2 The total long-term magnitudendashfrequency distribu-tion (MFD) for the branch-averaged UCERF3 model used here(black) plus on-fault (red) and off-fault (blue) contributions whichsum to equal the total (black) The on-fault contribution includessubseismogenic ruptures inside each fault-section polygon The thindashed lines indicate a b 1 extrapolation in order to show howthe distributions differ from a perfect GutenbergndashRichter (GR) Theplot only goes down toM 40 even though the model goes toM 25

Table 3The ETAS Parameters Used in This Study Representing the

Preferred Values of Hardebeck (2013) Based on anAnalysis of M ge25 California Earthquakes between 1993

and 2011

Parameter Units Value Range

p 107 10ndash14c Years 178 times 10minus5 100 times 10minus6 to 316 times 10minus4

q 196 gt18d km 079 gt063k Yearspminus1 284 times 10minus3 379 times 10minus4 to 497 times 10minus3

α 10 FixedMmin 25 Fixed

The Range gives parameter values with log likelihoods within 5 of themaximum log likelihood over a reasonable range of values of the otherparameters (see Hardebeck 2013 for details)

6 E H Field et al

conclusions presented in this study do not change significantlywith other published values (eg using the ETAS parametersof Hardebeck et al 2008 which have an implied branchingratio of 066 for the regional MFD shown in Fig 2)

Characteristic Versus GutenbergndashRichter MFDs

For each simulated aftershock UCERF3-ETAS effec-tively samples a magnitude from a nucleation MFD or moretechnically from a magnitude density function One of themain findings of the UCERF-TI effort was that in orderto fit all the data constraints many faults require a character-istic MFD which is defined as having elevated rates at highmagnitudes relative to a GR extrapolation from smaller mag-nitudes Figure 3a shows the average time-independent nu-cleation MFD for one of the Mojave subsections of the SAF

for which the supraseismogenic versus subseismogenic con-tributions are plotted separately Note that the rates for supra-seismogenic events are well above the GR extrapolation fromsubseismogenic rates meaning the fault has a characteristicMFD We quantify the degree of characteristicness by whatwe call the CharFactor defined as the cumulative rate at theminimum supraseismogenic magnitude divided by the cu-mulative rate at that magnitude for a perfect GR distributionthe latter is defined as having the same totalM ge25 rate thesame maximum magnitude (highest magnitude with nonzerorate) and a b-value of 10 The CharFactor value is 33 forthe example shown in Figure 3a

Figure 4a shows a map of CharFactor values for eachfault subsection in the UCERF3-TI model together withthe cumulative distribution of values (black curve in Fig 4d)Values range from a low of 00016 (1625) at one end of the

Table 4Expected Number of Aftershocks (Exp N) over 10 Years Following Various Mainshock Magnitudes (Mmain)

Exp N ge M 25 Exp N ge Mmain Exp N ge Mmain minus 10

Mainshock Magnitude (Mmain) Primary Secondary Tertiary Plusdagger Total Primary Total Primary Total

25 00529 00362 00762 01653 0053 016530 01672 01144 02409 05228 0053 016535 05287 03616 07619 16533 0053 016540 16719 11436 24093 52281 0053 016545 52869 36164 76189 16533 0053 0165 0529 16550 16719 11436 24093 52281 0053 0165 0529 16555 52869 36164 76189 16533 0053 0165 0529 16560 16719 11436 24093 52281 0053 0165 0529 16565 52869 36164 76189 16533 0053 0165 0529 16570 16719 11436 24093 52281 0050 0157 0529 16575 52869 36164 76189 16533 0039 0121 0529 16580 16719 11436 24093 52281 0013 0041 0529 165

Based on Monte Carlo simulations and assuming the UCERF3-TI total magnitudendashfrequency distribution shown in Figure 2daggerTertiary Plus includes the third-through-fifteenth generation of aftershocks

3 4 5 6 7 8

Lo

g10

Rat

e (p

er y

r)

3 4 5 6 7 8ndash 8

ndash 6

ndash 4

ndash 2

0

Magnitude Magnitude

ndash 8

ndash 6

ndash 4

ndash 2

0Uncorrected CharFactor Corrected

(a) (b)

Figure 3 Subseismogenic (light gray bins) and supraseismogenic (dark gray bins) time-independent incremental nucleation MFDs for aMojave subsection of the San Andreas fault (SAF) The black line is the combined cumulative MFD the dashed black line is a perfect GRextension of the subseismogenic incremental MFD with the same maximum magnitude and the dotted line is the cumulative distribution forthe latter (a) Original (uncorrected) rates for supraseismogenic ruptures with the CharFactor described in the text being defined as the ratio ofsolid-black to dotted-black lines atM 63 (the minimummagnitude for supraseismogenic ruptures) which gives CharFactor = 33 (b) CorrectMFD in which the supraseismogenic rupture rates have been divided by the CharFactor value of 33 causing the solid-black and dotted-blacklines to agree at M 63


Imperial fault to a high of 161 on the Surprise Valley faultboth of which are labeled in Figure 4a

The CharFactor values have a direct influence on thelikelihood of sampling supraseismogenic ruptures whichare generally the ones we care about from a hazard or riskperspective For example the expected number of M ge63primary aftershocks expected from an M 55 parent is

sim00084 (0053 times 1055minus63) if sampling from a GR distribu-tion (dashed line in Fig 3a) but the expected number be-comes sim0028 if sampling from the characteristic MFD inthat figure with the multiplicative increase being exactlyequal to the CharFactor value of 33 for this fault section

In other words CharFactor values represent the factor bywhich the likelihood of triggering supraseismogenic ruptures

SubectionCharFactorDistribution

CharFactor

Cu

mu

lati

ve F

ract

ion

1 0ndash2 1 0ndash1 1 00

1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)

Log10(CharFactor)ndash2 ndash1 210

UncorrectedValues

Values withApplyGridSeisCorr = True

Values withApplyGridSeisCorr = TrueApplySubSeisForSupraNucl = True

deg511minusdeg021minusdeg521minus

35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km

Surprise Valley faullt

Imperialfault

Robinson Creekfault

San Andreas(Mojave S)

fault

San Andreas(Peninsula)

San Jacinto(Borrego)

fault

Figure 4 Subsection CharFactor values defined as the total rate of supraseismogenic ruptures divided by that implied by a GR extrapo-lation from small magnitudes (see text for details) (a) Uncorrected values (ApplyGridSeisCorr and ApplySubSeisForSupraNucl both set asFalse) (b) Values for case in which ApplyGridSeisCorr = True and ApplySubSeisForSupraNucl = False (c) Values for case when bothApplyGridSeisCorr and ApplySubSeisForSupraNucl are True (d) The cumulative distribution of values in which the black curve corre-sponds to those shown in (a) blue is for those in (b) and red is for those in (c)

8 E H Field et al

differs from that of a perfect GR distribution conditioned onthe occurrence of a nearby parent event This implies that if theCharFactor value for the Mojave section in Figure 3 were afactor of 161 (the value for the Surprise Valley fault) thenthe expected number ofM ge63 primary aftershocks triggeredfrom anM 55 parent would be 135 We need to be somewhatcareful here because the likelihood of having more than one

event will depend on whether elastic-rebound effects are in-cluded if so the occurrence of the first supraseismogenic rup-ture will changereduce the likelihood that others can followFurthermore secondary tertiary and subsequent generationsof aftershocks could also trigger anM ge63 event increasinglikelihoods compared with those cited above Neverthelessthe point stands that a sufficiently high CharFactor value will

˚511minus˚021minus˚521minus˚511minus˚021minus˚521minus

35deg

40deg

˚511minus˚021minus˚521minus

35deg

40deg


35deg

40deg

minus3 minus2 minus1 0 1

Log10(Corrected Nucleation Rate)

0 100 200 300

km

(a) (b)

(c) (d)

0 6 12

Corrected vs Original Ratio

35deg

40deg

0 100 200 300

km


Log10(Aftershock Nucleation Rate)

0 100 200 300

km


Log10(Original Nucleation Rate)

0 100 200 300

km

Figure 5 (a) TheM gt25 nucleation rate (per year) in each grid cell for the branch-averagedUCERF3model used in this study (described inFig 1 caption) (b) The long-term rate of aftershocks expected from the long-term rate of supraseismogenic ruptures only (c) The ratio of (b) to(a) in which values less than 10 are plotted as gray revealing areas where the long-term model (a) must be deficient in M gt25 earthquakes(d) TheM gt25 nucleation rate for the correctedmodel (ApplyGridSeisCorr = True) in which rates in (a) have beenmultiplied by the ratio in (c)


imply near certainty of triggering at least one supraseismo-genic rupture given a smaller nearby event which highlightsthe need to carefully scrutinize the degree of characteristicnessallowed throughout the model

Several factors influence the CharFactor value inferredfor each fault section The subseismogenic MFD reflects anempirical estimate of M ge25 event rates near the fault (ob-tained by integrating smoothed seismicity rates over associ-ated fault-section polygons the latter of which are describedmore below) The fact that the total rate of M ge5 eventsthroughout the entire region is uncertain by about 20 (ac-cording to UCERF3-TI logic-tree branch options) impliesthat we should expect considerable observational uncertaintyat the fault-section polygon level Supraseismogenic nuclea-tion rates for each fault section are determined from theUCERF3-TI grand inversion which incorporates informationon nearby fault connectivity slip-rate estimates paleoseismicevent-rate data where available and other constraints such asstaying as close as possible to the previous model (UCERF2)The high CharFactor value for the Surprise Valley fault couldbe biased from either erroneously low M ge25 event-rate es-timates in that area due to a historical lull or network detectionissues or because a lack of known connectivity with neigh-boring faults produces an erroneously high rate of moderate-sized events when satisfying the slip-rate constraint The pointis that individual CharFactor values are uncertain and we needto be careful not to propagate any poorly constrained attributesof UCERF3 long-term models into UCERF3-ETAS forecastsespecially if they have a first-order influence on consequenttriggering statistics

One reality check is whether the subseismogenic MFDsnear faults are high enough to include the expected long-termrate of aftershocks from supraseismogenic events Figure 5reveals that the answer is no near several faults Figure 5ashows a map of the total long-term event rates implied byUCERF3-TI which mostly reflects the smoothed seismicitymodel used for gridded background seismicity Figure 5bshows the rate of primary aftershock expected from the

long-term rates of fault-based ruptures multiplied by a factorof 2 to approximate the contribution of subsequent genera-tions and Figure 5c shows areas where and the extent towhich the ratio of the latter to the former exceeds 10 (inwhich UCERF3-TI does not satisfy the expected rate of after-shocks from supraseismogenic events) These areas imply alack of self-consistency in that supraseismogenic ruptures inthe model will trigger more M ge25 events than the long-term model exhibits in the first place which represents amodel inconsistency that results in artificially inflated Char-Factor values For UCERF3-ETAS we consequently supportthe correction of gridded seismicity rates so they do not fallbelow the values shown in Figure 5b which amounts tomultiplying the values in Figure 5a by the ratios in Figure 5cleading to the values shown in Figure 5d This correction ismade according to whether the ApplyGridSeisCorr param-eter (Table 2) is set as true or false

The CharFactor values that result from setting Apply-GridSeisCorr = True are shown in Figure 4b along withthe cumulative distribution (blue curve in Fig 4d) revealinga significant reduction of some of the higher CharFactor val-ues The maximum value is now 72 on the Robinson Creekfault (labeled on Fig 4b) and the value for the Surprise Val-ley fault has gone from 161 to 56

Up until this point we assumed that supraseismogenic rup-tures have an equal probability of nucleating from anywhere onthe fault surface An alternative assumption is that the likeli-hood of nucleation correlates with the rate of subseismogenicevents In other words if the rate of little earthquakes is twiceas high at one end of a candidate rupture it might also be twiceas likely to nucleate from that end This assumption serves tofurther reduce the range of CharFactor values throughout themodel although the effect is relatively mild as shown in Fig-ure 4c (and the red curve in Fig 4d) We nonetheless supportthis correction as an option in UCERF3-ETAS and it is set bythe ApplySubSeisForSupraNucl parameter value (Table 2)The final value for the Robinson Creek fault is still 72and the Surprise Valley fault value is now down to 45

Finally UCERF3-ETAS includes the option to imposeGR throughout the model (by setting ApplyGRcorr = True)which effectively divides the nucleation rate of each supraseim-sogenic rupture by the associated CharFactor value The resultof this correction is shown in Figure 3b for the Mojave S ex-ample Note that although the total rate of supraseismogenicruptures now matches GR after this correction (at M ge63 inFig 3b) rates do not necessarily match at higher magnitudesgiven our particular definition of characteristicness

Overall Calculation Sequence

In addition to an ERF as described above the UCERF3-ETAS model takes the following as input (1) the desiredTimeSpan and region for the simulation where the latter isset here as the entire UCERF3 area (Fig 1) (2) a list of allM ge25 earthquakes that have occurred up until the chosensimulation start time plus any fake ruptures that might be

1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion

Seismicity Depth Distribution

Figure 6 The depth distribution of hypocenters assumed forCalifornia derived from the UCERF3 catalog (Felzer 2013)

10 E H Field et al

desired for testing purposes (3) the average depth distributionof earthquake hypocenters for the region (which assumed hereis shown in Fig 6) and (4) yes or no answers to various cal-culation options including whether to include spontaneousevents indirect triggering and whether to apply the correc-tions described above (ApplyGRcorr ApplyGridSeisCorr andApplySubSeisForSupraNucl respectively) Another setting isthe type of time dependence to apply in the model specifiedby the ProbModel parameter (Table 2) the options of whichwill be described below Finally we have a TotalRateScaleFac-tor which simply scales the total rate of spontaneous events inorder to correct any bias between total simulated and targetM ge25 rates

Given these inputs and settings the algorithm used toproduce a simulated catalog of ruptures the desired outcomeis described in Appendix A The particularities of the approachdescribed therein stem from the need for a numerically efficientway of sampling ETAS events while honoring an elastic-re-bound model on faults Appendix A assumes the availabilityof another major model component the ETAS_PrimaryAfter-shockSampler which is described in the next section

Sampling Primary Aftershocks from the ERF

Here we describe how primary aftershocks are randomlysampled from the ERF which is achieved via the ETAS_PrimaryAftershockSampler component Again we want tosample from the population of ERF ruptures according to thecurrent relative probability of each and the distance decay de-fined by equation (2) To do so we discretize the entire regioninto cubes that are about 2 km on a side Specifically each 01deg-by-01deg grid cell (Fig 1) is subdivided into 002deg increments anddepth is discretized at 2 km increments down to 24 km (seeFig S1 available in the electronic supplement of this article)This results in 300 cubes per grid cell and about 2300000cubes for the entire UCERF3 region When sampling an after-shock for a given parent first we randomly select a cube fromwhich the triggered event nucleates and then randomly sampleone of the ruptures that can nucleate from within the cube ac-cording to the current relative probability of each doing so

The next step is to define the rate at which each source orrupture nucleates inside each cube according to the ERF Forgridded seismicity we simply divide each grid-cell rateequally among the 300 associated cubes (the depth depend-ence of nucleation is handled later) Fault-based rupturesmight seem straightforward as well because one could as-sume a uniform distribution of nucleation across the surfaceand then distribute nucleation rates among the cubes crossedby the fault surface according to the rupture area within eachRegrettably this simple approach does not work for two rea-sons elastic-rebound triggering (ERT) and the spatial uncer-tainty of faults both of which are discussed next

Elastic-Rebound Triggering The ETAS model raises aninteresting question with respect to triggering on finite faultswhich is illustrated in Figure 7 and can be described as fol-lows suppose a 50-km-long section of the southern SAF hasjust ruptured According to UCERF3-TD a repeat of this ex-act rupture has a zero probability of occurrence in the nearfuture because of elastic rebound However the probabilityof having an even longer rupture that overlaps the one thatjust occurred is nonzero (although reduced according toUCERF3-TD because the average time since last eventhas gone down) The critical question is this Can a longerrupture nucleate or be triggered from within the 50 kmstretch that just ruptured The answer to this question hasa strong influence on triggering statistics because the vastmajority of aftershocks spawned by the first rupture will oc-cur right along the 50 km section as illustrated in Figure 7so the probability of triggering the longer rupture will bemuch higher if the event can nucleate from this previouslyruptured stretch The alternative is that the longer rupture hasrelatively little likelihood of nucleating along the 50 km sec-tion leaving only aftershocks near the ends of the previousrupture able to do any large event triggering The latterassumption is included as an option in UCERF3-ETAS de-pending on the setting of the ProbModel parameter specifi-cally a setting of FULL_TD (full time dependence) meansthat supraseismogenic ruptures cannot be triggered from a

San Andreas Fault

Just RupturedNo chance of doing so again soon according to UCERF3-TD

Possible RuptureProbability greater than zero according to UCERF3-TD

although reduced due to blue rupture

Can the red rupture be triggered (nucleate) from the blue area (which just ruptured)

Figure 7 This figure poses the question of whether a larger rupture (red) can be trigged from within the area of a smaller supraseismo-genic event (blue area) that has just occurred The question is relevant because most small aftershocks (stars) will be in the blue area asillustrated so the likelihood of triggering the red event will be much higher if the answer is yes


zone that just ruptured and a setting of NO_ERT meansit can

Although it is straightforward to prevent a larger rupturefrom nucleating within the rupture zone of a very recent largeevent the harder question is how this zone transitions backinto nucleation viability with time Regrettably UCERF3-TD tells us nothing about where fault-based ruptures will nu-cleate We therefore make the simple assumption that the rel-ative likelihood that a particular section will nucleate a givenrupture is proportional to


AsηsPAsηs 3

in which ηs is the normalized time-since-last-event of thesection (time-since-last-event divided by the average long-term recurrence interval) As is the area of the sectionand the summation in the denominator is the overall sectionsutilized by the particular rupture (a normalization that en-sures that section-nucleation probabilities sum to 10 forthe rupture so that the overall likelihood of the event remainsunchanged) That the relative triggering probability increaseslinearly with time may seem at odds with typically appliedrenewal models (eg BPT) which have a distinctly nonlin-ear increase However we are only talking about where theevent might nucleate from and we are not modifying theoverall rupture likelihood defined by the ERF (which doeshonor the BPT renewal model) That being said we needto watch for potential biases particularly when simulatingevents over time periods that approach or exceed the recur-rence intervals of faults We also acknowledge that a moreelegant approach is certainly fathomable especially with re-spect to making nucleation likelihood a more integral part ofthe elastic-rebound formulation This will be a significantchallenge for models that relax segmentation and includemultifault ruptures however and any progress will likely re-quire at least inferences from physics-based simulators thatinclude Omori-type triggering (eg Richards-Dinger andDieterich 2012) For UCERF3-ETAS we apply the trigger-ing probabilities defined by equation (3) when the ProbMo-del parameter is set as FULL_TD otherwise the time-since-last-event has no bearing on where a rupture nucleates

There is also observational evidence that smaller sub-seismogenic ruptures exhibit such ERT effects Usinghigh-precision double-difference relocations for observedM 4ndash67 earthquakes van der Elst and Shaw (2015) have shownthat aftershocks that are as large or larger than the parent nu-cleate almost exclusively in the outer regions of the parentaftershock zone which they interpret as evidence for elasticrelaxation Though applying such a recent finding here mayseem premature it does solve another potential problem withrespect to triggering probabilities when transitioning betweensubseismogenic and supraseismogenic mainshocks For ex-ample if an M 63 supraseismogenic mainshock cannot trig-ger larger events from the subsections of its rupture surfacethen the probability of triggering such events could be lowerthan for a slightly smaller subseismogenic mainshock if the

latter can trigger from anywhere In other words we wouldhave a sharp drop in triggering probability as we transitionfrom subseismogenic to supraseismogenic mainshocks (allother things being equal)

To avoid this problem we simply apply the rule thatnon-fault-based ruptures larger than M 4 cannot trigger anysupraseismogenic fault-based ruptures from within a spherecentered on the hypocenter The sphere radius is set equal tothe radius of the source in which the latter is assumed to becircular with an area (A) given by

EQ-TARGETtempintralink-df4313613A 10Mminus4 4

in which M is the magnitude This source area is consistentwith both the Hanks and Bakun (2008) and Shaw (20092013) scaling relationships for M ≲63 events That thesource radius is applied to a spherical volume around the hy-pocenter reflects the fact that we will not generally know theactual rupture plane so we effectively account for any pos-sibility This rule which is only applied when ProbModel isset as FULL_TD is admittedly unsophisticated but never-theless a seemingly reasonable place to start

An interesting implication of ERT is a lowering of trigger-ing probabilities relative to GR because larger events can only betriggered from off the ends of the rupture surface (Fig 7) Morespecifically for supraseismogenic mainshocks both the ex-pected number of aftershocks and rupture lengths scale similarlywith magnitudelength so the overall density of aftershocks overthe rupture surface remains invariant with magnitude Thismeans that for a long and completely isolated fault triggeringlikelihoods will also become invariant with increasing rupturelength or magnitude (again all other things being equal) Toquantify this effect we ran a series of tests for various-sizedevents on the southern SAF (Mojave section and beyond)where other model variables were set so as to meaningfully iso-late the influence of ERT (this included application of CharFac-tor corrections by setting ApplyGRcorr = True and using a time-independent model for overall rupture probabilities) When rup-tures were allowed to nucleate from the mainshock rupturezone the expected number ofM ge63 primary aftershocks wasexactly equal to that predicted by GR for all mainshock mag-nitudes Suppressing such triggering produced the followingM ge63 triggering-likelihood reductions relative to GR

bull 072 for M 50 point-source mainshock on the faultbull 048 for M 55 point-source mainshock on the faultbull 024 for M 63 point-source mainshock on the faultbull 017 for M 63 fault-based mainshockbull 008 for M 70 fault-based mainshockbull 006 for M 74 fault-based mainshock andbull 009 for M 78 fault-based mainshock

As expected the reductions relative to GR becomegreater with increasing mainshock magnitude (because of anincreased nucleation exclusion zone) and become approxi-mately invariant above M 7 although not exactly invariantdue to nearby branching faults in these examples (eg the

12 E H Field et al

M 78 event passes by the Garlock fault but the M 74 doesnot) These test results are for a particular set of ruptures onone particular fault and may not apply to all other possiblemainshocks especially where there are more or less nearbybranching faults Nevertheless the tests do give us a generalidea of the influence of ERT

Note that ERT counters the influence of characteristicMFD In fact if ERT is real one will need a CharFactorof about 60 just to match GR probabilities for the M 63mainshock case above this is just to break even with GRas the CharFactor will have to exceed 60 if we think thetriggerability should be higher (eg because we are on a veryactive fault)

Thus there could be long-term characteristic earthquakebehavior as required by the grand inversion results fromUCERF3 but still could be self-similar foreshockndashmainshockndashaftershock behavior as observed for southern California Italy

New Zealand and global seismicity (as discussed in Michael2012a p 2548 and references therein)

Distributing Nucleation Rates Over Fault Section Poly-gons The other challenge is that fault surfaces in UCERF3actually represent a proxy for all supraseismogenic rupturesoccurring within a fault-zone polygon with the latter beingpredefined in UCERF3-TI for each fault section For exam-ple the polygon for vertically dipping faults extends out to12 km on each side of the trace perpendicular to the strike asshown for a part of Mojave SAF in Figure 8 The intent wasto be more explicit about whether an event like the 1989Loma Prieta earthquake is a hit or miss with respect to theSAF or whether the 2010 El MayorndashCucapah earthquakewas a hit or miss with respect to the Laguna Salada faultA 12 km zone on each side which is about the same asthe seismogenic thickness also seems reasonable in terms

24 km

Latitude

Longitude

Dep

th

Log10 Rate (per year)

M 63 CubeNucleation Rates

(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values

minus8 minus7 minus6 minus5

Figure 8 Illustration of how supraseismogenic on-fault nucleation rates are distributed over the cubes that are within each fault-sectionpolygon in which the latter are depicted with black lines on the upper surface which extend to depth vertically Images at the top showM ge63nucleation rates and those on the bottom depict the CharFactor values described in the main text for cubes near SAF (Mojave S) subsections (a)and (c) correspond to the case in which supraseismogenic fault-based ruptures only nucleate from within cubes that are intersected by the faultsurface (magenta- and red-colored cubes for the San Andreas) leaving other cubes within the subsection polygons to have zero supraseismo-genic rates (white) (b) and (d) show the solution adopted here in which cube nucleation rates decrease linearly out to 10 km and are constantbeyond such that the total nucleation rate in the polygon is preserved and CharFactor values beyond 10 km are 10 (see text for details)


of the spatial reach of supraseismogenic ruptures with re-spect to potential elastic relaxation

To avoid double counting the maximum magnitude forgridded seismicity sources inside a fault-zone polygon is justless than the minimum magnitude of the associated supra-seismogenic ruptures (eg the light- vs dark-gray bins inFig 3) If nucleation of a fault-based rupture can only occurfrom cubes that are intersected by the surface then all othercubes within the polygon will have a zero nucleation rateabove the minimum supraseismogenic magnitude as illus-trated in Figure 8a Even worse because the gridded seismic-ity nucleation rates have been distributed equally among the300 cubes within each grid cell the nucleation MFD forcubes intersected by the rupture surface would have a greatlyexaggerated rate difference between sub- and supraseismo-genic ruptures effectively increasing the CharFactor valuestherein by more than an order of magnitude (Fig 8c)

To avoid these problems we must partition the nucleationrate of each fault-based rupture among all cubes containedwithin the associated fault-zone polygons Partitioning nucle-ation rates uniformly is one option but would imply that thelikelihood of triggering a supraseismogenic event is constantover the polygon area and therefore independent of distancefrom the actual fault plane We add some distance dependenceby assuming a linear nucleation-rate decay with distance fromthe fault surface out to 10 km beyond which the nucleationrates are held constant with an implied CharFactor of 10 andsuch that the total nucleation rate is preserved (ie summingrates over the cubes inside the fault-section polygon equal thetotal nucleation rate for the section) Figure 8b and 8d showsthe consequent nucleation rates and implied CharFactor val-ues respectively for cubes near SAF (Mojave S) subsectionsillustrating that our rate-partitioning algorithm provides bothdecay with distance from the fault surface and a continuoustransition with respect to rates and implied CharFactor valuesfor cubes just outside the polygons (in which CharFactor val-ues are 074 because the off-fault MFD [blue curve in Fig 2] isthat much below the GR extrapolation at M 63)

The linear partitioning of nucleation rates means the im-plied CharFactor values for cubes closest to the fault aregreater than the section average (eg 151 and 103 for Rob-inson Creek and Surprise Valley faults respectively) whichis consequence of applying monotonic decrease with dis-tance from the fault and a continuous transition at the poly-gon boundary This algorithm does not make sense for faultsections that have a CharFactor less than 10 because itwould imply a monotonic increase with distance from thefault We therefore distribute nucleation rates uniformlywhen a fault-section CharFactor is less than 10

Figure 9 shows the implications of this partitioningthroughout the region for cubes at 7 km depth Figure 9aand 9b shows the nucleation rate of M gt25 and M ge63events respectively Figure 9c shows the implied CharFactorof each cube and Figure 9d shows the maximum magnitudein each cube

With the rules for ERT and the distribution of nucleationrates across fault-section polygons hereby established we arenow able to compute the rate at which each ERF rupture nucle-ates inside each cube (ie the nucleation MFD for each cube)

Distance Decay In sampling ruptures we also need to ac-count for the distance decay of aftershocks as given by the sec-ond term in brackets in equation (1) This is the linear distancedecay meaning the relative rate or density of triggered events atdistance r rather than the density at any particular point inspace Two complications arise here First the zone of potentialearthquakes transitions from a sphere when r is less than thedistance to both the free surface and the bottom of the model(24 km here) to something more like the outer surface of acylinder at large distances Second wewant the model to honorthe depth dependence of seismicity given in Figure 6 meaningthat a parent at a depth of 12 km is more likely to trigger anevent 4 km above (at 8 km depth) than at 4 km below (16 kmdepth) Both these complications imply that the triggering den-sity at a point in space depends on the depth of the parent event

We solve this problem by first assuming that the parentevent is located at a cube corner meaning that it has a depthof 0 2 4hellip or 24 km ( Fig S1) For each of these depthswe create a separate DistanceDecayCubeSampler whichcontains the relative likelihood that each neighboring cubewill host a triggered event in which the probability densityfor each cube is calculated in a way that honors both the dis-tance decay (equation 1) and the depth dependence shown inFigure 6 The problem is solved numerically taking care toproperly integrate over the entire volume for the closest cubes(because the nonlinear decaymeans that the value computed atthe center of the cube can be significantly different from thevolume-averaged value) The result is illustrated for two dif-ferent parent depths in Figure 10

The DistanceDecayCubeSampler provides a randomlysampled cube according to the relative probability densityof each and it also gives a random location within the cubewhere the latter accounts for any significant rate-density tran-sitions inside the cube (ie for those in close proximity to theparent which most cubes will be) In lieu of further numeri-cal details on how all this is achieved which can be found inthe computer code Figure S2 shows that random samplesare in good agreement with the target distance decay overdistances from 50 m to 1000 km

Sampling Ruptures Full details on the algorithm used tosample primary aftershocks for a given parent rupture are de-scribed in Appendix B In essence a cube is sampled accord-ing to the distance decay from the parent rupture surface and arupture is sampled from within the cube based on the currentrelative likelihood that each can nucleate from within Thecode does not explicitly sample from a cube MFD but ratherfollows a more numerically efficient procedure in which it firstdecides between a gridded seismicity and fault-based sourceaccording to the current relative nucleation rate of each Fur-thermore fault-section nucleation rates are calculated from

14 E H Field et al

and kept in sync with any changes in the elastic-rebound-implied likelihood of each fault-based rupture Again furtherdetails are given in Appendix B

Results

UCERF3-ETAS is an admittedly complicated modelwith a lot of calculation options Presenting an exhaustive

set of results and sensitivity tests here is therefore not fea-sible Instead we present a limited set of simulations aimedat highlighting both positive features and more importantlypotential issues and remaining challenges with respect todeploying a potentially useful operational system

We first present results from long-term (eg 1000 year)simulations which not only provide valuable insights intomodel behavior but also suggest some corrections that could

Figure 9 Implications of how nucleation of supraseismogenic fault-based ruptures are distributed among the cubes that are inside eachfault-zone polygon (as described in the text) (ab) Nucleation rates for cubes at 7 km depth for M ge25 and M ge63 events respectively(c) The implied CharFactor value of each cube at 7 km depth (d) The implied maximum magnitude for cubes at 7 km depth


be made We then present 10-year probabilities implied by thehistoricalinstrumental catalog Finally we use hypotheticalearthquake scenarios to quantify the conditional triggeringprobabilities implied by some events of interest Althoughindividual simulations can easily be run on a desktop com-puter we made extensive use of high-performance computingresources in generating the large ensembles presented below(see Acknowledgments)

1000-Year Simulations

Figures 11ndash16 show results aggregated from approxi-mately five hundred 1000-year simulations using the defaultparameter values listed in Table 2 including a ProbModel set-ting of FULL_TD (which includes ERT meaning supraseis-mogenic ruptures are prevented from nucleating from withinrecently ruptured zones) The start year for the simulations is2012 set by when the UCERF3 catalog ends and the latterwas input to the model as potential parents (although this doesnot really significantly influence 1000-year results) Figure 11shows the simulated versus target cumulative MFD for the en-tire region The match is good below M 5 (within 4) butthere is a 25 underprediction near M 7 Figure 12a showssimulated M ge25 nucleation rates throughout the region atthe spatial resolution of the cubes and Figure 12b shows theratio of simulated versus target M ge5 rates in each grid cellin which the mean ratio is 093 the median is 079 the mini-mum is 0062 and the maximum is 54 Areas dominated bygridded seismicity (away from faults) are generally undersi-mulating the rate ofM ge5 events and areas near several faultsare oversimulating such rates This is a natural and expectedmanifestation of the spatial variability of long-term nucleationMFDs throughout the region where areas exhibiting CharFactorvalues less than 10 will undersimulate event rates and vice versa

for areas with values greater than 10 As discussed above thefractional rate of spontaneous versus triggered events dependson the nucleation MFD so to match the rates everywhere inFigure 12b we would really need to vary the fractional rate inspace (in addition to time) in order to compensate for the spatialvariability of long-term MFDs This is also why the defaultvalue of the TotalRateScaleFactor parameter is 114 that is weneed to increase the rate of spontaneous events by 14 tobetter match the total rate of M ge25 events

Figure 12c shows simulated fault-section participationrates in map view and Figure 12d shows the ratio of the latterto target values Figure 13 shows a scatter plot of simulatedversus model target participation rates for each subsectioncolor-coded by the fraction of events on each fault sectionthat were triggered by supraseismogenic fault-based ruptures(as opposed to being spontaneous or triggered by subseismo-genic events) On average 335 of fault-based ruptures arespontaneous 31 are triggered by subseismogenic ruptures(meaning no ancestors were supraseismogenic) and 355of ruptures had at least one supraseismogenic ancestor Thesymbol colors in Figure 13 reveal a strong correlation betweenparticipation-rate discrepancies and the fraction of eventstriggered by supraseismogenic events oversimulated rates gen-erally result from above-average triggering from other fault-based ruptures and vice versa for undersimulated rates

A number of the outliers are labeled in Figures 12d and 13The high ones labeled with orange text represent faults that arebeing triggered by the nearby and relatively active SAF Thelow outliers labeled in blue represent faults where recent earth-quakes suppressed occurrences over the 1000-year simulationduration by virtue of elastic rebound (and long recurrence in-tervals) Also labeled is a point on the SAF (Mojave S) where

minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km

minus9 minus7 minus5 minus3 minus1

minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 18 kmLog10 Prob

(a)

(b)

Figure 10 Illustration of the DistanceDecayCubeSampler de-scribed in the text in which the relative likelihood of sampling anevent in each location is shown by the color (a) and (b) The cases inwhich the depth of the parent event is 6 and 18 km respectively

40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)

Figure 11 The average cumulative MFD obtained from fivehundred 1000-year UCERF3 epidemic-type aftershock sequence(ETAS) simulations (black line) compared with the target MFD (grayline) Default parameter settings were applied in these simulationsand the target represents the total UCERF3-TI distribution (Fig 2)modified according to the influence of having ApplyGridSeisCorr= True (which increases rates below about M 63 by about 46)

16 E H Field et al

the simulated rate is 25 above the target which correspondsto a simulated recurrence interval of 75 years versus the targetvalue of 90 years This too is a manifestation of triggering bysupraseismogenic ruptures UCERF3-TI had to assign a rela-tively high rate for M 63ndash65 events in this area to satisfy

the relatively short recurrence interval implied by the Wright-wood paleoseismic record The occurrence of one such event isapparently overtriggering others in the UCERF3-ETAS simu-lations Figure 14 shows a histogram of simulated recurrenceintervals at this location where the height of the first bin

minus10 minus05 00 05 10

deg511minusdeg021minus

35deg

40deg

King Range2011 CFM

Monte Vista - Shannon 2011 CFM

Morales(East)

HomesteadValley 2011

(1992 Landers Qk)

White Wolf(1952 Kern Co Qk)

Great Valley 13(1983 Coalinga)

SAF Mojave S(Subsection 13)

(a) (b)

(c) (d)Log10(M 50 Nucleation-Rate Ratio)Log10(M 25 Nucleation Rate)

Log10(Subsection Participation Ratio)Log10(Subsection Participation Rate (per yr))

Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km

minus6 minus5 minus4 minus3 minus2 minus1


35deg

40deg

0 100 200 300

km

Figure 12 Results obtained from 1000-year UCERF3-ETAS simulations based on default parameter settings in which results representaverages from 500 different runs (a)M ge25 nucleation rates throughout the region at a resolution of 002deg in latitude and longitude (b) Ratioof simulated to model-target nucleation rates for M ge50 events binned at a 01deg resolution (c) The average rate at which each fault sectionparticipates in a supraseismogenic rupture (d) The ratio of (c) to model-target values (the latter from UCERF3-TI)


implies that 9 of the ruptures are occurring within 7 yearsand the fact that this bin is the mode of the distribution impliesthat such a recurrence interval is the most likely one to be ob-served Elastic rebound prevents a recurrence of an identicalrupture so the triggered events are different but with spatialoverlap Such recurrence intervals are not likely to be observedpaleoseismologically both because they are short and becauseM 63ndash65 events might go undetected If we exclude the firstbin in computing the mean recurrence interval the value is 82years which is still a bit below the 90-year target

On average the match in Figure 13 is relatively good forsections with participation rates greater than 0004 per year(a recurrence interval of less than 250 years) but below thisthreshold simulated rates are 25ndash33 below targets onaverage (red-dashed line) One possible explanation is that a1000-year simulation is not long enough to see the full aver-age-triggering effect on such low-rate faults (ie perhaps weneed more than about four earthquakes in each simulation)We therefore did some 10000-year simulations to see if theagreement transition shifts (eg from 0004 per year to alower value of 00004 because the simulation is 10 timeslonger) The shift was minimal however and certainly notenough to explain the full discrepancy Another benefit ofthese 10000-year simulations is that they allowed us to see

whether any of the discrepancies discussed above grow withtime and fortunately we did not find any such instability

So what else might explain the participation-rate agree-ment transition at 0004 per year in Figure 13 A logndashlog plotof participation-rate discrepancy ratios versus the sectionCharFactor value is shown in Figure 15 together with a linearfit that implies lower rate faults have a lower CharFactor valueon average As explained above faults with lower CharFactorvalues will need to have commensurately more spontaneousevents in order to match long-term rates so perhaps this ex-plains the systematic discrepancy at lower rates in Figure 13

To examine the overall elastic-rebound predictability im-plied by UCERF3-ETAS Figure 16 shows a histogram of nor-malized fault-section recurrence intervals aggregated over allfaults and all simulations The normalization which amounts todividing each subsection recurrence interval by the expectedmean before adding it to the histogram is needed to make thecomparison meaningful The normalized recurrence intervalsimplied by the UCERF3-TD model (without spatiotemporalclustering) and those implied by the RSQSim physics-basedsimulator (Richards-Dinger and Dieterich 2012) are also shownboth adopted from Field (2015) As expected inclusion of spa-tiotemporal clustering in UCERF3-ETAS increases the aperi-odicity (widens the bell curve or increases the coefficient ofvariation) and also increases the relative frequency of the short-est recurrence intervals relative to UCERF3-TD (because ofspatially overlapping triggered ruptures) The UCERF3-ETASdistribution is qualitatively similar to that of RSQSim althoughthe latter has a lower coefficient of variation and a smaller frac-tion in the first bin implying that RSQSim has less spatialoverlap with respect to quickly triggered adjacent ruptures

We now turn to describing the influence of alternativeparameter settings with respect to 1000-year simulations re-stricting the discussion to only noteworthy differences

Figure 13 The average simulated participation rate of supra-seismogenic events on each subsection plotted against the model-target value for the 1000-year UCERF3-ETAS simulations basedon default parameter settings The symbol color indicates the frac-tion of events that were triggered by supraseismogenic fault-basedruptures either directly or indirectly as opposed to being sponta-neous or triggered by subseismogenic events The red dashed linerepresents binned average values in which the bin width is 02 inlog10 space corresponding to 20 bins over the x-axis range

0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity

San Andreas (Mojave S) Subsection 13Recurrence Intervals

Figure 14 Histogram of recurrence intervals for supraseismo-genic events on SAF (Mojave S) subsection 13 obtained from the1000-year UCERF3-ETAS simulations based on default parametersettings Bin widths are 72 years and the mean recurrence intervalis 75 years

18 E H Field et al

Turning Off Corrections In accordance with default val-ues the above simulations used the following setting Apply-GridSeisCorr = True ApplySubSeisForSupraNucl = Trueand TotalRateScaleFactor = 114 Setting the first two toFalse and the last one to 10 (no correction) generally makesthe various discrepancies discussed above worse For exam-ple Figure S3 shows simulated versus target subsectionparticipation rates for this case in which an increase in thescatter is apparent when compared with the default-parametercase (Fig 13) In particular discrepancies are even greater forthe more extreme oversimulated cases with the recurrence in-terval for the Mojave S subsection discussed above now beingdown to 45 years (from a value of 75 years for default param-eters and compared with the 90-year target) The difference ismostly a manifestation of the ApplyGridSeisCorr parameterin which not applying this correction causes the simulation toproduce more earthquakes than exist in the long-term modelnear several faults and these surplus events in turn triggerextra supraseismogenic ruptures due to the artificially highCharFactor values As already mentioned the default value ofthe TotalRateScaleFactor is 114 because a value of 10 causesabout a 14 undersimulation of total M ge25 event rates(not shown)

ProbModel Parameter Figure S4 shows the subsectionparticipation scatter diagram for the case in which ProbModelis set as NO_ERT meaning supraseismogenic ruptures are

free to trigger from anywhere All other parameters were setto default values in this case except TotalRateScaleFactor =10 because the total rates are well matched in fact the entireMFD is fit well as shown in Figure S5 In general simu-lated rates are shifted upward in Figure S4 compared withthe default-parameter case (Fig 13) and there is less averagediscrepancy across a wider range of target values (red-dashed

ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)

ndash50 ndash45 ndash40 ndash35 ndash30 ndash25 ndash20 ndash15

Log10(Participation Rate (per year))

Figure 15 Supraseismogenic participation-rate discrepancy(simulated over target) versus the CharFactor value for each faultsubsection together with the best-fit line (in logndashlog space) Simu-lated values are from the 1000-year UCERF3-ETAS simulationswith default parameter settings and the CharFactor values are thoseshown in Figure 4c accordingly

00

05

10

15

20

Den

sity

Normalized Recurrence Interval

00 10 20 30 40 50

00

05

10

15

20

Den

sity

Normalized Recurrence Interval00 10 20 30 40 50

00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)

Figure 16 (a) Normalized subsection recurrence intervals forsupraseismogenic events from the 1000-year UCERF3-ETAS sim-ulations based on default parameter settings in which the normali-zation involves dividing each recurrence interval by that expectedbefore adding to the histogram (b) The normalized recurrence in-tervals implied by UCERF3-TD (ie without spatiotemporal clus-tering) and (c) the equivalent results obtained from the RSQSimphysics-based simulator (Richards-Dinger and Dieterich 2012)The latter two plots come from Field (2015)


line) However 49 of the fault-based ruptures are both trig-gered and have at least one supraseismogenic ancestor (com-pared with 355 above) which manifests as warmer symbolcolors in Figure S4 compared with Figure 13 Furthermorethe recurrence interval for the Mojave S subsection is 54 yearsin this case

Setting ProbModel = POISSON (no elastic rebound)produces very unrealistic simulations and in fact so manyaftershocks are triggered that the code never terminates Thisis an expected manifestation of CharFactor values that ex-ceed 10 because a characteristic distribution like that shownin Figure 3a has a branching ratio well above 10 meaningthat the number of simulated events will forever increaseElastic rebound alleviates this problem by dropping the rateor likelihood of supraseismogenic ruptures when such anevent has already occurred

One remaining hope for the POISSON option might be tocorrect all fault-section MFDs to be GR consistent which wecan enforce by setting ApplyGRcorr = True However Field(2012) has already shown via simpler models that this too pro-duces unrealistic triggering statistics in that something like80 of large triggered ruptures simply re-rupture the parentsurface which we do not see in nature This problem is alsoexemplified using UCERF3-ETAS in the Discussion section

10-Year Probabilities (with No Scenario Mainshock) The1000-year simulations presented above were intended to pro-vide a better understanding of how the UCERF3-ETASmodelworks Though finding perfect agreement between 1000-yearsimulations and model-target values might be helpful fordeeming UCERF3-ETAS potentially useful such agreementmay not be a necessary condition That is perhaps it is askingtoo much to expect the model to behave well over long timeperiods especially when we know that the fractional rate ofspontaneous events should really vary in both time and spaceFurthermore most practical applications will be interested inwhat the model says about the next few days or years andperhaps the model is more reliable over such timeframes

We therefore examine 10-year probabilities here rather thanlong-term rates based on 10000 simulations for each case andagain starting at the beginning of 2012with the UCERF3 catalogprovided as input (both of which are now more influential) Themain metric of interest here is the probability of one or moresupraseismogenic ruptures occurring on each fault section forwhich historical elastic-rebound effects will also be more influ-ential Figure 17 shows a scatter diagram of simulated versustarget probabilities for each subsection in which the simulatedvalues represent the fraction of runs that had one or more occur-rence within 10 years and for which the target values are com-puted from UCERF3-TD note that the term target here does notimply expected or preferred but rather the value we get whenspatiotemporal clustering is ignored The most striking observa-tion is that simulated values are on average 25 below the tar-get values and 33 below for the higher probability subsections(where target values are greater than 003) The fact that prob-abilities are below target values on most faults is consistent with

the fact that recent history is relatively quiet with respect to largeearthquakes because these low current values will need to aver-age out with elevated probabilities following future large events

The fraction of supraseismogenic ruptures that are bothtriggered and have at least one supraseismogenic ancestor is19 down from 355 in the default 1000-year simulationsabove This implies that 10 years is not enough to see full aver-age effects especially given our recent history This is borneout by the symbol colors in Figure 17 which indicate the frac-tion of supraseismogenic ruptures that were triggered bya UCERF3-catalog earthquake either directly or indirectlyCases with a high rate of historical event triggering (green sym-bols some of which are labeled in Fig 17) also tend to havesimulated probabilities that are well above the target values

Some of the high outliers that are not triggered by his-torical events such as the King Range and Great Valley 08cases labeled in Figure 17 represent triggering from a moreactive nearby fault The reason for some of the other highoutliers is not so obvious such as for the Goldstone Lake andMission Ridge cases labeled in Figure 17 these may be re-lated to clusters of high CharFactor faults producing bursts oflarge events because all such cases have a high fraction oftriggering from other supraseismogenic events (not shown)The lowest outliers subsections of Mohawk Valley representa fault that is relatively isolated and therefore cannot be trig-gered by other fault-based ruptures

Changing ProbModel from FULL_TD to NO_ERT(allowing supraseismogenic triggering from recent rupturezones) produces qualitatively similar results but for which si-mulated probabilities are 9 below target values on averagecompared with 25 below for the FULL_TD case above

Scenario Simulations

An OEF system will presumably be in highest demandfollowing some seismic activity of interest We therefore turnto implied 10-year earthquake probabilities following a va-riety of hypothetical scenarios in which 10000 simulationshave been computed for each case As mentioned above theUCERF3 catalog was provided as input but this does notinfluence the results shown here A variety of plots are pro-vided for the first example to give a feel for the simulationsWe then restrict attention to the likelihood that each scenariowill trigger events that exceed various magnitudes and com-pare these with probabilities obtained assuming that the totalregional GR distribution applies everywhere (as quantifiedin Fig 2 and Table 4) Keep in mind that these scenariosare blind in that they incorporate no actual aftershock datawhereas a real OEF system will have some such data depend-ing on the start time of the forecast

M 50 Scenario on San Andreas (Mojave S) We firstpresent results for an M 50 mainshock located right on theSAF (Mojave S) section and for default model parametersThe location of this scenario can be seen in Figure 18 whichshows the nucleation rate of primary aftershocks averaged

20 E H Field et al

over all 10000 simulations Figure 19 shows the average nu-cleation rate for all generations of aftershocks indicating thatsome large and distant events have been triggered over the10-year time period (with the more distant triggers beingunique to this set of simulations as more realizations wouldbe needed to get true averages)

The average rate that each fault subsection participatesin a supraseismogenic aftershock is shown in Figure 20 andthe average rate at which such aftershocks nucleate fromeach fault section is shown in Figure 21 both of which in-clude all generations of aftershocks As expected mostsupraseismogenic aftershocks occur near the M 5 scenarioTriggering does occur at more distant locations but Figure 21indicates that 10000 simulations is not an adequate numberto establish average nucleation rates on more distant faults

Figure 22 shows that the distance decay implied byaftershocks of this scenario are in good agreement with thatimposedassumed in which the increasing discrepancy be-yond about 300 km is due to the fact that we are not simu-lating events outside the California region (Fig 1) Likewisethe average temporal decay for primary aftershocks shownin Figure 23 is in good agreement with that imposedassumed and the temporal decay for all aftershocks (alsoshown) has a lower rate of decay as expected by the temporaldelay associated with indirect triggering

A histogram of the maximum magnitude trigged in eachsimulation is shown in Figure 24 for both primary and all gen-erations of aftershocks The modal value is 38 consistent withBaringthrsquos law that the magnitude of the largest aftershock tends to

be 12 magnitude units below that of the mainshock (Richter1958) Figure 24 also shows a propensity to trigger largerevents with M 63ndash65 aftershocks being particularly likelyin this location (for reasons discussed above)

Finally Figure 25 shows the cumulative MFD of trig-gered events or more accurately the average number of eventstriggered above each magnitude (not normalized to rates andnote that we nevertheless continue referring to such plots asMFDs to avoid introducing another acronym) Several curvesare shown in this figure including the mean mode medianand 25th and 975th percentiles (the latter implying that 5 ofsimulations had values above or below these lines) The meanfor primary aftershocks only is also shown (green curve)which has values that are about half the total mean at highermagnitudes The most striking result is that the total mean iswell above the median and mode the latter two of which aresimilar and imply a maximum magnitude of 38ndash40 (againconsistent with Baringthrsquos law) The total mean is clearly beingpulled high by large event triggering whereby such eventsproduce additional aftershocks that significantly skew theaverages In fact the mean curve implies that we should ex-

Figure 17 Simulated versus target (UCERF3-TD implied)10-year probability of one or more supraseismogenic events on eachsubsection based on 10000 UCERF3-ETAS simulations using de-fault parameter settings Symbol colors indicate the faction ofevents that were triggered by an event in the historicalinstrumentalUCERF3 catalog (either directly or indirectly) minus5 minus4 minus3 minus2 minus1 0

Log10 M 25 Nucleation Rate


35deg

40deg

0 100 200 300

km

Figure 18 Average M ge25 nucleation rate of primary after-shocks for the M 50 scenario on the SAF (Mojave S) section (be-tween subsections 10 and 11) obtained from 10000 10-yearsimulations using default parameter settings (Table 2) The regionwas discretized into 002deg latitude and longitude bins for this plotand rates should be multiplied by the 10-year duration to get theaverage number of nucleations in each bin


pect one M ge5 aftershock on average compared with a 15likelihood of one or more assuming the total regional GRThe mean likelihood of triggering an M ge63 event is about10 and the likelihood of an M ge78 aftershock is about05 For comparison the dashed line in Figure 25 shows thetotal mean number of aftershocks assuming that the regionalGR distribution in Figure 2 applies everywhere

One way to avoid this mean-value distortion is to insteadquantify the fraction of simulations that had one or more after-shocks exceeding each magnitude the result of which is alsoshown with the red line in Figure 25 (and labeled as ldquoFractWith 1 Or Morerdquo) This fraction is 10 (100) at low magni-tudes meaning virtually all simulations included an M ge25aftershock with fractions being about 13 7 and 05 atM ge50 M ge67 and M ge75 respectively In terms of anappropriate evaluation metric it is not yet clear that this frac-tion is superior to the mean value so we have quantified bothbut utilize mean values in the discussions that follow the dif-ference is relatively small at the more hazardous magnitudesanyway

For convenience the mean number of aftershocks ex-pected above various magnitudes as implied by our simula-tions is summarized in Table 5 for all the scenarios examinedin this article The ratio of this value to that expected from aGR distribution or more specifically from the assumption

that the total regional GR in Figure 2 applies everywhere isalso listed in the table We refer to the latter as the RM factormeaning the ratio of the expected number to the regional GRvalue at the given magnitude (or the ratio of the solid blackline to the dashed black line in Fig 25) For example R50 625 for the scenario discussed above meaning that we shouldexpect 625 times more M ge50 aftershocks than impliedby the regional GR average The RM factors for the othermagnitude thresholds listed are R55 664 R63 125R70 778 R74 148 andR78 435 all of which seemplausible given that the scenario is located right on one of themost active faults in the region

RM factors reflect the effective MFD being sampled inthe vicinity of the mainshock which in turn is influenced by(1) the degree of characteristicness in the long-term MFDs(2) time-dependent probability gains implied by elastic re-bound which are presently about a factor of 15 on MojaveS subsections (Field et al 2015) and (3) the influence ofERT which influences where supraseismogenic ruptures nu-cleate from Tracking the exact influence of each of these is achallenge given the spatial variability in the long-term MFDs(eg as characterized by CharFactor values Fig 9c) and the

minus4 minus2 0



35deg

40deg

0 100 200 300

km

Figure 19 Same as Figure 20 but including all generations ofaftershocks for the M 50 SAF (Mojave S) scenario


Log10 Participation Rate for All Aftershocks


35deg

40deg

0 100 200 300

km

Figure 20 The average rate at which each fault section partic-ipates in any aftershock for the M 50 SAF (Mojave S) scenarioobtained from 10000 10-year simulations using default parametersettings (Table 2) Multiply rates by the 10-year duration to get theaverage number of participations

22 E H Field et al

fact that the occurrence of one event can change the likelihoodthat others will follow (by virtue of elastic rebound)

Other San Andreas (Mojave S) Scenarios Simulationswere conducted for other scenarios involving SAF (Mojave S)subsections Specifically mainshock magnitudes of 55 6370 74 and 78 were applied and both FULL_TD and NO_ERT values for the ProbModel parameter were consideredScenarios with M gt63 are modeled as supraseismogenicfault-based ruptures and those with M lt63 are treated aspoint sources One of the M 63 scenarios was treated as asupraseismogenic fault-based rupture and another was treatedas a point source for comparison Plots equivalent to those pre-sented for the M 5 scenario above are also available in Dataand Resources for the other scenarios discussed in this articlein particular the maps showing the rate of M ge25 primaryaftershocks are a good way to see the locationrupture surfaceof each scenario and the cumulative MFD plots are a morecomplete representation of the data summarized in Table 5

Results for theM 55 scenario are qualitatively similar tothose for the M 50 case except that RM factors are reduced

by about 35 due to the increased effect of ERT (a largervolume from which supraseismogenic ruptures cannot rup-ture) For the M 63 scenario treated as a point source (non-fault-based rupture) RM factors range from a value of 155 to812 reflecting even further reductions due to ERT Valuesfor the M 63 scenario treated as a fault-based rupture aresimilar ranging from 147 to 626

Perhaps the most striking result is that RM factors are be-low 10 for the M 7 scenario ranging from 013 to 069 de-pending on the magnitude the mean cumulative MFD for thisscenario is also shown in Figure 26 The increasing influenceof ERT at greater magnitude is influential here but apparentlyany remaining elastic-rebound probability gains and character-isticness are not making up for this RM factors are even lowerfor the higher magnitude scenarios ranging from 005 to 058for the M 74 mainshock and from 001 to 044 for the M 78event As we go to higher magnitudes elastic-rebound effectsare further suppressing the possibility of triggered events

Table 5 also lists the results for the NO_ERT ProbModeloption in which supraseismogenic aftershocks are allowed tobe triggered from within the rupture zone of the mainshocksAs expected RM values are larger typically by a factor of 5compared with the FULL_TD simulations The smallestchange is for the M 78 scenario in which R50 went from044 to 047 and the biggest change is for theM 74 scenarioin which R78 went from 005 to 172 For the M 7 scenarioexpected numbers generally go from being less than GR forthe FULL_TD case to being greater than GR for the NO_ERTcase (see also Fig 26) However numbers generally stay be-low GR for the M 74 and 78 scenarios

San Andreas (Peninsula) Scenarios Scenarios involvingSAF (Peninsula) sections near San Francisco were also


Log10 Trigger Rate for All Aftershocks


35deg

40deg

0 100 200 300

km

Figure 21 The average rate at which supraseismogenic after-shocks of any generation nucleate from each fault section for theM 50 SAF (Mojave S) scenario obtained from 10000 10-year sim-ulations using default parameter settings (Table 2) Multiply rates bythe 10-year duration to get the average number of nucleations

01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)

Simulation LowerUpper 95 of Mean

Simulation Mean

ExpectedTarget Distance Decay

Figure 22 The distance decay for aftershocks of theM 50 SAF(Mojave S) scenario obtained from 10000 10-year simulations usingdefault parameter settings (Table 2) The plot includes all generationsof aftershocks but distance is measured relative to the immediate pa-rent of each aftershock The black and gray solid lines represent themean and 95 confidence of the mean respectively and the dashedgray line represents that imposedassumed in the model


investigated including anM 55 point source and M 63 and70 fault-based ruptures In addition to being of practical con-cern this fault is also of interest by virtue of having a Char-Factor of sim055 (it is anticharacteristic with respect toM ge63 rates Fig 4c) and by having a time-dependent prob-ability gain of sim064 meaning it is statistically underdue (seefig 4b of Field et al 2015) These facts coupled with theinfluence of ERT predict RM factors below 10 Indeed forthe FULL_TD ProbModel option the range of values in Ta-ble 5 is a low of R63 004 for the M 63 scenario and ahigh of R50 071 for the M 55 scenario Switching toNO_ERT increases expected numbers by up to a factor of 6

(eg the likelihood of theM 63 scenario triggeringM ge78events) with a maximum of R74 094 for the M 55 sce-nario That is none of the RM factors currently exceeds 10not even for the NO_ERT case meaning that conditionaltriggering probabilities are currently below regional GR val-ues In the future the elastic-rebound probability gain maygo from the current value of 064 to something like 20 buteven then only some RM factors will exceed 10 and only upto a value of about 30

Surprise Valley M 55 Scenario As noted above the Sur-prise Valley fault has a CharFactor value of 45 (Fig 4c)which is second only to a subsection of the Robinson Creekfault Simulations of the latter turned out to be less dramaticthan for Surprise Valley due to other influences so we dem-onstrate the influence of extreme CharFactor values here us-ing an M 55 scenario located right on the Surprise Valleyfault RM factors range from 256 to 474 for the FULL_TDcase and from 102 to 205 for the NO_ERT case Perhaps themost extreme consequence is that theM 55 mainshock has a053 probability of triggering an M ge63 event for theNO_ERT case (or a 01 chance for the FULL_TD case) Thisresult alleviates any concerns that extreme CharFactor valuesmight lead to unphysical conditional triggering probabilitiesbecause even for this most extreme case we are not sayingthat an M 63 event is certain to happen This conclusiondepends on setting ApplyGridSeisCorr = True because un-reasonable conditional probabilities do result without thiscorrection as discussed above Note also that the probability

25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat

ions All Aftershocks

Primary Aftershocks

Scenario M 50

Maximum-Magnitude Histogram

Figure 24 Histogram of the maximum aftershock magnitudefor M 50 SAF (Mojave S) scenario obtained from 10000 10-yearsimulations using default parameter settings (Table 2)

3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean

Fract With 1 Or More

Regional GR Mean

Figure 25 Cumulative number of aftershocks greater thaneach magnitude for the M 50 SAF (Mojave S) scenario obtainedfrom 10000 10-year simulations using default parameter settings(Table 2) See the legend for a description of the various curvesThe Mean Mode Median and fractile curves correspond to allaftershocks (all generations) Primary Mean is the mean for primaryaftershocks only and the Regional GR Mean is the total expectednumber assuming the total regional GR distribution in Figure 2 ap-plies everywhere Fract With 1 or More represents the fraction ofsimulations that had one or more aftershocks greater than or equal tothe given magnitude

ndash40 ndash30 ndash20 ndash10 00 10 20 30

Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean

ExpectedTarget Temporal Decay

Primary Aftershocks

All Aftershocks

Figure 23 The temporal decay for aftershocks of the M 50SAF (Mojave S) scenario obtained from 10000 10-year simula-tions using default parameter settings (Table 2) Results are plottedseparately for primary aftershocks and all aftershocks (as labeled)The black and gray solid lines represent the mean and 95 confi-dence of the mean respectively and the dashed gray line representsthat imposedassumed for primary aftershocks

24 E H Field et al

Table 5Average Expected Number (N ge M) and the Ratio with Respect to Assuming a GR Distribution (RM) Implied by the Various Mainshock


Scenario ProbModel N ge 50 R50 N ge 55 R55 N ge 63 R63 N ge 70 R70 N ge 74 R74 N ge 78 R78

SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128

SAF Mojave SM 63 PtSrc

FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658

SAF Mojave SM 63 FSS

FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000

San JacintoBorregoM 55

FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703

PtSrc means treated as point source (gridded seismicity) and FSS means treated as a fault-system-solution rupture (fault based) SAF San Andreas fault


of having an M 55 mainshock in this area is very low in thefirst place which is why it has a high CharFactor value

San Jacinto (Borrego) M 55 Scenario This case exempli-fies the consequence of having a particularly low CharFactorbecause the value is 012 at the hypocenter of this scenario (andthe time-dependent probability gain is near 10) TheR63 valueis 009 and 018 for the FULL_TD and NO_ERT cases respec-tively In other words the conditional probability of triggeringa supraseismogenic rupture given something smaller nearby iswell below the GR average This result reflects a relatively highrate of microseismicity near this fault with the low CharFactorvalue effectively preventing these numerous events from over-triggering supraseismogenic ones (relative to the rate of the lat-ter implied by the long-term model)

Central Valley M 50 Scenario This case exemplifies amainshock that is far away from any modeled faults Herethere should be minimal difference between the FULL_TDand NO_ERT cases because these parameters only influencesupraseismogenic on-fault triggering Simulations imply thatR50 072 with values decreasing at high magnitudeswhich is consistent with the fact that the off-fault MFD (bluecurve in Fig 2) tapers more quickly than the total MFD(black curve) at higher magnitudes The number of M ge5primary aftershocks is consistent with that in GR but thenumber of M ge5 events triggered in subsequent generationsis deficient due to fewer larger events being triggered byvirtue of the greater tapering This is precisely why thelong-term simulations underpredict the target rate of eventsin off-fault areas (Fig 12b) because the fractional rate of

spontaneous versus triggered events should really be ad-justed according to the off-fault MFD

Bombay Beach M 48 Scenario This scenario correspondsto one studied extensively by Michael (2012a) in terms of itslikelihood of triggering anM ge7 earthquake on the Coachellasection of the SAF The M 48 event which actually occurredon 24March 2009 is located about 4 km from the fault-surfacerepresentation in UCERF3 The probability gain in this area is19 according to UCERF3-TD and the CharFactor value is065 for the closest subsection (and note that the CharFactorwent from above 10 to below it when we set ApplyGridSeis-Corr = True Fig 4a vs Fig 5bc) Here the metric of interestis the likelihood of triggering anM ge7 event within 3 days forwhich Michael (2012a) found a 000035ndash0013 range depend-ing on whether a GR or characteristic MFD is assumed Thevalue we obtained is 00012 meaning that this fraction of thesimulations exhibited one or more triggedM ge7 events withinthe first 3 days Both the FULL_TD and NO_ERT ProbModeloptions produced similar values for this relatively small main-shock The RM factors based on 10-year forecasts range from alow of R70 162 to a high of R78 83 (Table 5)

Discussion

Main Challenges

The goal of this study has been to bring fault-based in-formation into OEF which has traditionally been basedsolely on GR and OmorindashUtsu statistics (Jordan et al 2011)As stated in the Introduction UCERF3-ETAS essentially rep-resents an elaborate implementation of the Michael (2012a)generalized clustering model in that it effectively samplesearthquakes according to the spatial distribution of MFDs im-plied by UCERF3-TD In so doing we explicitly include in-formation on fault proximity activity rate and susceptibilityfrom an elastic-rebound perspective

A number of significant challenges have arisen in con-structing the model including the following

1 How fault-based nucleation rates transition into the sur-rounding region especially given uncertainties in thefault surfaces themselves

2 How and if elastic rebound suppresses the likelihood ofother ruptures nucleating from within the rupture zone ofprevious events

3 That the degree of characteristicness implied on faultsvaries widely and that much of this may represent uncer-tainty rather than reality

4 That the previously inferred long-term rate of smallerevents near some faults is not consistent with the numberof aftershocks expected from larger ruptures on that faultimplying that CharFactors must be biased high

5 That if one wants to match the long-term model then thefractional rate of spontaneous versus triggered events inthe ETAS model must not only vary in time but also inspace because the nucleation MFDs vary spatially

3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r

Figure 26 Mean cumulative number of aftershocks greaterthan each magnitude (solid lines) for the M 70 SAF (Mojave S)scenario obtained from 10000 10-year simulations using defaultparameter settings (Table 2) Results are shown separately for Prob-Model set as full-time dependence (FULL_TD) versus NO_ERT (inwhich ERT means elastic-rebound triggering as labeled) Thedashed line represents the result assuming the total regional GRMFD in Figure 2 applies everywhere and the dotted lines representmodal values from the simulations

26 E H Field et al

6 That one must consider the depth distribution of seismic-ity as well as the finite thickness of the seismogeniclayer when matching the distance decay of aftershocks

Although most of us thought that UCERF3-ETASwould be relatively straightforward to construct in retrospectit is surprising that the model works as well as it does givenall the challenges encountered and the approximate ap-proaches that were used to solve them The question is how-ever whether the model is good enough to be useful oruseful enough to be worth operationalizing

Potential Model Usefulness

UCERF3-ETAS simulations look realistic in terms ofthe overall productivity and spatiotemporal decay of after-shocks and the results maintain a plausible degree of elas-tic-rebound predictability (Fig 16) However the 1000-yearsimulations do exhibit differences from long-term modelrates some of which can be attributed to a need for a spatiallydependent fractional rate of spontaneous earthquakes whichis not presently supported in the model

Supraseismogenic ruptures can only occur in one of threeways (1) spontaneously (2) triggered by a smaller nearbyevent that is not itself a descendant of any supraseismogenicrupture or (3) triggered by a supraseismogenic rupture eitherdirectly or indirectly Discrepancies between simulated andmodel-expected fault-section participation rates correlate moststrongly with the degree of triggering from other supraseismo-genic fault-based ruptures with the highest outliers represent-ing lesser faults adjacent to and consequently being triggeredby the much more active SAF (eg Figs 12 and 13) Like-wise a well-isolated fault will exhibit undersimulated ratesby virtue of only rarely being triggered by a neighboring faultAlthough these more extreme outliers are easily explained wecannot yet claim to understand every discrepancy that themodel produces

The triggering of supraseismogenic ruptures is influ-enced by the interaction of several factors including the pop-ulation of nearby faults the degree of characteristicness onthose faults the elastic-rebound-implied probability of eachrupture and whether elastic rebound further restricts inwhich triggered ruptures can nucleate At this point it is dif-ficult to say whether long-term simulation discrepancies are aconcern with respect to potential usefulness Either way 10-year probabilities may be more reliable because the latterseem to be reasonable when based on the historicalinstru-mental catalog as input (above average triggering probabil-ities on faults that are near recent large events and belowaverage elsewhere) Applying a spatially variable fractionalrate of spontaneous versus triggered events would certainlyhelp to reduce the remaining discrepancies how much soremains to be seen and it begs the question of how close isclose enough which should probably be answered beforeany such complexity is added to the model especially whenthe revised model will still be an approximation

Perhaps the more troubling aspect of UCERF3-ETAS isthe conditional triggering probabilities implied by scenariosAt first blush one might expect the number of expected after-shocks to always be greater than regional GR-implied valueswhen the mainshock is near a more active fault and belowaverage when not near any known fault (eg reflecting thefact that CEPEC convenes when a magnitude sim5 earthquakeoccurs near the SAF as noted in the Introduction) This is notborne out by UCERF3-ETAS as many of the fault-basedscenarios imply probabilities below the GR value includingan M 7 event on the Mojave S section of the SAF (a faultconsidered to be locked and loaded)

As noted by Michael (2012a) the degree of character-isticness has a first-order influence on conditional triggeringprobabilities near a fault For example the San Jacinto faulthas CharFactor values well below 10 thereby implying con-ditional probabilities that are lower than GR CharFactor val-ues vary widely throughout the region even along a givenfault (eg Fig 9c) which means that we will have commen-surate variability in the conditional triggering probabilitiesimplied by an event

If all the CharFactor values are real (not due to uncer-tainties) then we need to retain them in order for the modelto correctly reproduce both long-term rates and short-termprobabilities For example a high CharFactor is needed pre-cisely because small events are rare in that area and the highvalue compensates accordingly to produce the correct rate oflarge events (or the correct portion triggered by smallerevents) The same goes for low CharFactor faults For exam-ple if we ignored the low value on the San Jacinto fault andinstead set CharFactor ge10 then the conditional probabilityfor a large event being triggered from the next small earth-quake might not look unreasonable However the aggregateconditional probability implied by several such eventswhich would surely occur given the high rate of microseis-micity might approach an unreasonably high value near 10which could be a basis for rejecting the model

The greater concern is that CharFactor values are highlyuncertain and that we are consequently propagating noise Ifwe assume that supraseismogenic rates are correct but thatsubseismogenic rates represent noise then high CharFactorvalues will overstate conditional probabilities (because thereare really more small events than we thought) and vice versafor low CharFactor values The problem with this interpre-tation is that our small event rates are based on a model that isone of the most skillful with respect to forecasting the loca-tion of future small events (Helmstetter et al 2007 Zecharet al 2013) so we know the rates cannot represent com-plete noise

The more likely situation is that we have some uncer-tainty in the nucleation rate of both sub- and supraseismo-genic ruptures and that these are getting propagated intoour conditional probabilities However resolving these un-certainties is not likely to convert the conditional triggeringprobabilities near all faults to a value greater than the GRequivalent especially if ERT is a reality The latter remains


an open question because we do not yet have a basis forrejecting either the FULL_TD or NO_ERT ProbModel op-tions Furthermore even if ruptures cannot nucleate from thezone of a recent rupture allowing them to do so may stillprovide a better statistical proxy overall We also do not re-ally know what large event-triggering statistics really aregiven the paucity of observations so expecting faults to havegreater conditional probabilities than GR may be naiumlve

This brings us to what may be the biggest conceptualproblem with UCERF3-ETAS which applies to the Michael(2012a) generalized clustering model as well That is we areusing a long-termMFD to define conditional triggering prob-abilities when in reality that MFD is extremely time depen-dent For example a fault will exhibit its highest rate ofseismicity immediately following a large event at whichtime the rate or likelihood of another large event on the faultis lowest due to elastic rebound As time goes on the rate orprobability of another big one steadily increases while therate of smaller events decays back down indefinitely perhapsto some background rate Of course the occurrence ofnearby events will modulate these rates somewhat but theoverall MFD time evolution should remain on average If thisbehavior represents reality which UCERF3-ETAS assumesit does then why would conditional probabilities depend onsome hypothetical long-term MFD that may never apply atany point in time Trying to use the fully time-dependentMFD in defining conditional triggering probabilities wouldnot help because it would significantly lower the likelihoodof one fault-based rupture triggering another one fartherdown the fault (because aftershocks from the first spillingonto the surface of the second would temporarily lowerthe CharFactor in that area making the triggering less likely)

Perhaps UCERF3-ETAS can nevertheless be a usefulproxy especially if properly tuned in terms of for examplechanging CharFactor values to produce acceptable conditionalprobabilities However we do not presently know what valuesto tune the model to not even the acceptable range which isagain due to the paucity of large event-triggering data This isa general problem that will plague the evaluation of any can-didate OEF model We also want to avoid having to customizeUCERF3-ETAS for every earthquake that occurs especially ifdoing so requires expert judgment

Another question is whether UCERF3-ETAS might beuseful with respect to OEF for induced seismicity The an-swer would seem to be yes in that the model makes no dis-tinction between natural and induced earthquakes hence ifthe rates of the latter are changed then so too will consequenttriggering probabilities The question is whether UCERF3-ETAS is the best approach in terms of an OEF for inducedevents For example Llenos and Michael (2013) found thathaving a temporally varying rate of spontaneous eventsprovides a clear improvement when using ETAS to modelinduced events especially when swarm-type behavior ispresent UCERF3-ETAS does not presently support such ratevariation However if one wants to consider proximity to

faults then something like UCERF3-ETAS will be neededto model induced seismicity

Finally potential usefulness will need to be ascertained inthe context of specific applications such as public prepared-ness emergency response building inspection and taggingbuilding-code adjustment setting earthquake-insurance ratesand pricing insurance-linked securities such as catastrophicbonds and reinsurance UCERF3-ETAS may be an adequateapproximation for some of these uses but perhaps not forothers An interesting question is what tests will be necessaryand sufficient in terms of deeming usefulness Running themodel for all significant historical California ruptures is anobvious task that should be pursued but it is not yet clear ex-actly how we would deem success or failure Furthermore wehave so far struggled to produce a single viable UCERF3-ETAS logic-tree branch whereas multiple sources of uncer-tainty will need to be explored and quantified Operationali-zation will also present unique challenges perhaps including aneed for real-time access to high-performance computing Inthe interest of brevity we do not discuss these issues furtherhere but suffice it to say considerable work remains withrespect to ascertaining usefulness

Scientific Implications

One primary contribution of this study is in providingfurther evidence that one cannot combine statistical trigger-ing models such as ETAS with finite-fault models withoutalso including elastic rebound This assertion which wasfirst made by Field (2012) using relatively simple modelsis supported here with an example from UCERF3-ETASWe already explained how characteristic MFDs combinedwith the POISSON ProbModel option imply a branchingratio much greater than 10 and therefore producing runawaysequences What remains to be exemplified here is that evenGR-consistent faults also suffer problems Consider the M 7Mojave S scenario discussed above but in which ProbMo-del = POISSON and ApplyGRcorr = True (making all faultsGR-consistent as exemplified in Fig 3b) Figure 27 showsthat according to this model if a supraseismogenic ruptureis triggered as a primary aftershock it has a 77 chance ofhaving more than half of its rupture surface overlap withthat of the parent event We just do not see this kind ofre-rupture occurring in nature at least not anywhere near77 of the time Furthermore and as discussed abovevan der Elst and Shaw (2015) have presented observationalevidence that aftershocks as large or larger than the parentnucleate almost exclusively in the outer regions of the pa-rent aftershock zone giving further evidence of some elas-tic-relaxation process

Though we are very confident that this need for elasticrebound will stand the test of time there are legitimate ques-tions about whether UCERF3-ETAS has gotten it right Therecurrence interval distributions shown in Figures 14 and 16exhibit a plausible degree of elastic-rebound predictabilityOne potentially important metric is the relative height of

28 E H Field et al

the first bin (the fraction of very short recurrence intervals)which is to some degree a measure of the spatial overlap ofruptures triggered on adjacent sections of a fault Is this thecorrect amount of potential overlap Should it be more orless We do not really know and it seems that our best hopefor addressing such questions is to try to collect relevant ob-servational data and to study the behavior implied by phys-ics-based simulators

Another scientific conclusion is that some degree of char-acteristicness is required on at least some faults That is if youbelieve that the conditional probability of triggering a largeevent should be significantly higher if you are near a seeminglycapable fault then you are also saying that the fault has a rel-atively characteristic MFD at least in the context of the gen-eralized clustering model (Michael 2012a) What this studyadds is that you will need even more characteristicness if ERTis a reality because the suppression of triggering from the rup-ture zone of the parent significantly lowers the consequentlarge event-triggering probabilities (illustrated in Fig 7)

Although these scientific inferences seem robust wealso acknowledge that the approach we have taken here maybe fundamentally flawed Perhaps once you exceed the scaleof the seismogenic layer and thereby enter the range of mag-nitudes that cause the most damage maybe ETAS statisticsare no longer reliable or perhaps maybe not at the levelof specificity represented in UCERF3-ETAS Is our modelthereby misleading One key to addressing this question willbe a comparison with more physics-based approaches Forexample one improvement might be to use Coulomb stress-change calculations (eg King et al 1994) or some otherreasoning to condition which aftershocks can nucleate asopposed to the isotropic distribution assumed here We em-phasize that the viability of any such enhancement should betested by evaluating implied synthetic earthquake catalogsbecause we hope to have demonstrated that such analysesare a powerful if not indispensable way to identify modelinconsistencies If we want both more physics and syntheticcatalogs then physics-based simulators that include bothelastic rebound and OmorindashUtsu-type triggering seem thelogical way to go (eg RSQSim Richards-Dinger and Diet-erich 2012) The challenge will be including off-fault seis-micity in such models as well as inferring the conditionaltriggering probabilities implied by smaller events (eg forthe M 48 Bombay Beach scenario)

This study also raises some questions with respect tomodel testing As mentioned the model that specifies the long-term rate of smaller earthquakes in UCERF3-ETAS hasdemonstrated skill in the formal prospective tests conductedby the Collaboratory for the Study of Earthquake Predictability(CSEP Helmstetter et al 2007 Zechar et al 2013 Helmstet-ter and Werner 2014 Rhoades et al 2014) However we hadto change these rates near several faults (via the ApplyGridSeis-Corr parameter Fig 5) in order to prevent artificially highCharFactor values from giving unreasonable conditional trig-gering probabilities in those areasWemight also be tempted toadjust the CharFactor values elsewhere to produce more rea-sonable results which would likely involve changing the long-term rate of smaller events in those areas too The cost of thesemodel adjustments aimed at improving performance at thelarger more hazardous magnitudes might very well result inrelatively poor performance in CSEP tests because the latterare dominated by smaller earthquakes Clearly one key to im-proved testing andor tuning of any OEF model will be a betterempirical quantification of large event-triggering statisticsespecially on a global scale

Conclusions

UCERF3-ETAS can be viewed as an attempt to eitheradd spatiotemporal clustering to a modern fault-based model(UCERF3-TD) or to fold fault-based information into a state-of-the-art statistical seismology model (ETAS) Perhaps ourmost significant finding is that both elastic-rebound effectsand characteristic MFDs near faults are apparently requiredto produce realistic behavior indicating a potential resolution


Log10(Conditional Probability)

35deg

0 100 200 300

km

Total ConditionalProbability = 077

Figure 27 For the SAF Mojave S M 7 scenario (thick blackline) this figure shows all possible supraseismogenic aftershocks thathave at least half of their rupture surface overlapping that of the parent(offset to the northeast for visibility) For ProbModel = POISSON thecolor indicates the conditional probability of each being sampled as aprimary aftershock given some supraseismogenic rupture has beentriggered The total conditional probability that one of these ruptureswill be sampled as a primary aftershock is 77


with respect to debates on whether either one is actuallyneeded (eg Kagan et al 2012 Tormann et al 2015 2016Buumlrgmann et al 2016)

The UCERF3-ETAS simulations evaluated thus far pro-duce very plausible results or at least we cannot yet rejectthose based on preferred parameter settings Some of the re-sults are somewhat surprising however at least in comparisonwith the naiumlve assumption that conditional triggering proba-bilities should always be greater than regional averages whena mainshock is near an active fault One first-order influenceon conditional triggering probabilities is the degree of charac-teristicness in MFDs near faults which varies widely through-out California (at least according to UCERF3) Anotherimportant question is whether larger aftershocks can be trig-gered from within the rupture zone of the parent event be-cause currently viable answers to this question can changetriggering likelihoods by an order of magnitude (eg Fig 26)A third challenge is rectifying large uncertainties in the physi-cal manifestation of faults with the fact that most triggeringoccurs within a few kilometers of the mainshock

The bottom line is that there are many uncertainties as-sociated with UCERF3-ETAS and there are many aspects ofthe model that we could likely improve upon However thiswill always be the case with any such forecast hence the rel-evant question is whether the model is good enough to be po-tentially useful or useful enough to be worth operationalizingThis question can only be answered in the context of actualuses because it may be good enough for one application but notanother Any further development of UCERF3-ETAS shouldtherefore at least be informed by anticipated uses

Going forward one key to improving OEF will be thecollection of more global large event-triggering data both interms of tuning and testing models Another will be the useof physics-based simulators to help for example narrow theviable range of elastic-rebound behavior Finally our resultsdo not imply that simply ignoring faults in OEF will producea more scientifically plausible or useful model

Data and Resources

Plots for all the scenario simulations presented in thisarticle are available at httpwwwWGCEPorgUCERF3‑ETAS (last accessed November 2016) All calculations weremade using OpenSHA (httpwww OpenSHAorg last ac-cessed April 2016 Field et al 2003) which in turn utilizesGeneric Mapping Tools (httpgmtsoesthawaiiedu last ac-cessed January 2012) and JFree-Chart (httpwwwjfreeorgjfreechart last accessed March 2012) for making plots Theunpublished manuscript by N J van der Elst (2016) ldquoIn-cluding orphaned aftershocks of long-past mainshocks inthe ETAS modelrdquo

Acknowledgments

The Southern California Earthquake Center (SCEC) Community Mod-eling Environment University of Southern California Center for High-Per-formance Computing and Communications and the Texas Advanced

Computing Center provided access to the high-performance computing Thisstudy was sponsored by the California Earthquake Authority the US Geo-logical Survey (USGS) the USGS John Wesley Powell Center for Analysisand Synthesis and the SCEC SCEC is supported in part by the NationalScience Foundation under Cooperative Agreement EAR-1033462 and bythe USGS under Cooperative Agreement G12AC20038 We are also gratefulto Andrea Llenos and two anonymous Bulletin of the Seismological Societyof America reviewers for their very thoughtful and helpful comments TheSCEC Contribution Number for this article is 7163 Any use of trade prod-uct or firm names is for descriptive purposes only and does not implyendorsement by the US Government

References

Agnew D C and L M Jones (1991) Prediction probabilities from fore-shocks J Geophys Res 96 11959ndash11971

Buumlrgmann R N Uchide Y Hu and T Matsuzawa (2016) Tohoku rupturereloaded Nat Geosci 9 183ndash184

Felzer K R (2013) Appendix K The UCERF3 earthquake catalog USGeol Surv Open-File Rept 2013-1165-K and California Geol SurvSpecial Rept 228-K

Felzer K T T W Becker R E Abercrombie G Ekstroumlm and J R Rice(2002) Triggering of the 1999 Mw 71 Hector Mine earthquake byaftershocks of the 1992 Mw 73 Landers earthquake J GeophysRes 107 2190 doi 1010292001JB000911

Field E H (2012) Aftershock statistics constitute the strongest evidence forelastic relaxation in large earthquakesmdashTake 2 presented at the 2012Meeting of the SSA San Diego California 17ndash19 April

Field E H (2015) Computing elastic-rebound-motivated earthquake prob-abilities in unsegmented fault models A new methodology supportedby physics-based simulators Bull Seismol Soc Am 105 544ndash559doi 1017850120140094

Field E H R J Arrowsmith G P Biasi P Bird T E Dawson K RFelzer D D Jackson K M Johnson T H Jordan C Madden et al(2014) Uniform California Earthquake Rupture Forecast version 3(UCERF3)mdashThe time-independent model Bull Seismol Soc Am104 1122ndash1180 doi 1017850120130164

Field E H R J Arrowsmith G P Biasi P Bird T E Dawson K RFelzer D D Jackson K M Johnson T H Jordan C Madden et al(2015) Long-term time-dependent probabilities for the Third UniformCalifornia Earthquake Rupture Forecast (UCERF3) Bull SeismolSoc Am 105 511ndash543 doi 1017850120140093

Field E H T E Dawson K R Felzer A D Frankel V Gupta T HJordan T Parsons M D Petersen R S Stein R J Weldon IIand C J Wills (2009) Uniform California Earthquake Rupture Fore-cast Version 2 (UCERF 2) Bull Seismol Soc Am 99 2053ndash2107doi 1017850120080049

Field E H T H Jordan and C A Cornell (2003) OpenSHA A devel-oping community-modeling environment for seismic hazard analysisSeismol Res Lett 74 406ndash419

Field E H T H Jordan L M Jones A J Michael M L Blanpied andOther Workshop Participants (2016) The potential uses of operationalearthquake forecasting Seismol Res Lett 80 1ndash10 doi 1017850220150174

Gardner J K and L Knopoff (1974) Is the sequence of earthquakes insouthern California with aftershocks removed Poissonian Bull Seis-mol Soc Am 64 1363ndash1367

Gerstenberger M C G H McVerry D A Rhoades and M W Stirling(2014) Seismic hazard modeling for the recovery of ChristchurchNew Zealand Earthq Spectra 30 17ndash29

Gerstenberger M C S Wiemer L M Jones and P A Reasenberg (2005)Real-time forecasts of tomorrowrsquos earthquakes in California Nature435 328ndash331

Gutenberg B and C F Richter (1944) Frequency of earthquakes in Cal-ifornia Bull Seismol Soc Am 34 185ndash188

Hanks T C and W H Bakun (2008) M-log A observations of recent largeearthquakes Bull Seismol Soc Am 98 490

30 E H Field et al

Hardebeck J L (2013) Appendix S Constraining epidemic type aftershocksequence (ETAS) parameters from the Uniform California EarthquakeRupture Forecast Version 3 catalog and validating the ETASmodel formagnitude 65 or greater earthquakesUS Geol Surv Open-File Rept2013-1165-S and California Geol Surv Special Rept 228-S

Hardebeck J L K R Felzer and A J Michael (2008) Improved tests revealthat the accelerating moment release hypothesis is statistically insignifi-cant J Geophys Res 113 no B08310 doi 1010292007JB005410

Helmstetter A and D Sornette (2003) Importance of direct and indirecttriggered seismicity in the ETAS model of seismicity Geophys ResLett 30 no 11 1576 doi 1010292003gl017670

Helmstetter A and M J Werner (2014) Adaptive smoothing of seismicityin time space and magnitude for time-dependent earthquake forecastsfor California Bull Seismol Soc Am 104 809ndash822

Helmstetter A Y Y Kagan and D D Jackson (2007) High-resolutiontime-independent grid-based forecast forM ge5 earthquakes in Califor-nia Seismol Res Lett 78 78ndash86

Jordan T H and L M Jones (2010) Operational earthquake forecastingsome thoughts on why and how Seismol Res Lett 81 no 4 571ndash574

Jordan T H Y-T Chen P Gasparini R Madariaga I Main W Marzoc-chi G Papadopoulos G Sobolev K Yamaoka and J Zschau (2011)Operational earthquake forecasting state of knowledge and guidelinesfor implementation final report of the International Commission onEarthquake Forecasting for Civil Protection Ann Geophys 54no 4 315ndash391 doi 104401ag-5350

Jordan T H W Marzocchi A J Michael and M C Gerstenberger (2014)Operational earthquake forecasting can enhance earthquake prepared-ness Seismol Res Lett 85 955ndash959

Kagan Y Y D D Jackson and R J Geller (2012) Characteristic earth-quake model 1884-2011 RIP Seismol Res Lett 83 951ndash953 doi1017850220120107

Kaiser A E C Holden R J Beavan R D Beetham R A Benites ACelentano D Collet W J Cousins M Cubrinovski G D Dellowet al (2012) The Mw 62 Christchurch earthquake of February 2011Preliminary report New Zeal J Geol Geophys 55 no 1 67ndash90 doi101080002883062011641182

King G C P R S Stein and J Lin (1994) Static stress changes and thetriggering of earthquakes Bull Seismol Soc Am 84 935ndash953

Llenos A L and A J Michael (2013) Modeling earthquake rate changesin Oklahoma and Arkansas Possible signatures of induced seismicityBull Seismol Soc Am 103 2850ndash2861 doi 1017850120130017

Michael A J (2012a) Fundamental questions of earthquake statistics andestimation of earthquake probabilities from possible foreshocks BullSeismol Soc Am 102 2547ndash2562 doi 1017850120090184

Michael A J (2012b) Do aftershock probabilities decay with time Seis-mol Res Lett 83 630ndash632 doi 1017850220120061

Marzocchi W A M Lombardi and E Casarotti (2014) The establishmentof an operational earthquake forecasting system in Italy Seismol ResLett 85 961ndash969 doi 1017850220130219

Nanjo K Z H Tsuruoka S Yokoi Y Ogata G Falcone N Hirata YIshigaki T H Jordan K Kasahara K Obara et al (2012) Predict-ability study on the aftershock sequence following the 2011 Tohoku-Oki Japan earthquake First results Geophys J Int 191 653ndash658

Ogata Y (1988) Statistical models of point occurrences and residual analy-sis for point processes J Am Stat Assoc 83 9ndash27

Ogata Y (1998) Space-time point-process models for earthquake occur-rences Ann Inst Stat Math 50 no 2 379ndash402

Peresan A V G Kossobokov and G F Panza (2012) Operational earth-quake forecastprediction Rend Fis Acc Lincei 23 131ndash138

Reasenberg P A and L M Jones (1989) Earthquake hazard after a main-shock in California Science 243 1173ndash1176

Reasenberg P A and L M Jones (1994) Earthquake aftershocks UpdateScience 265 1251ndash1252

Rhoades D A M C Gerstenberger A Christophersen J D Zechar DSchorlemmer M J Werner and T H Jordan (2014) Regional earth-quake likelihood models II Information gains of multiplicative hy-brids Bull Seismol Soc Am 104 3072ndash3083

Richards-Dinger K and J H Dieterich (2012) RSQSim earthquake sim-ulator Seismol Res Lett 83 983ndash990

Richter C F (1958) Elementary Seismology W F Freeman San FranciscoCalifornia

Schoenberg F and B Bolt (2000) Short-term exciting long-term correctingmodels for earthquake catalogs Bull Seismol Soc Am 90 849ndash858

Segou M T Parsons and W Ellsworth (2013) Comparative evaluation ofphysics-based and statistical forecasts in northern California J Geo-phys Res 118 6219ndash6240 doi 1010022013JB010313

Shaw B E (2009) Constant stress drop from small to great earthquakes inmagnitude-area scaling Bull Seismol Soc Am 99 871 doi 1017850120080006

Shaw B E (2013) Appendix E Evaluation of magnitude-scaling relation-ships and depth of rupture Recommendation for UCERF3 US GeolSurv Open-File Rept 2013-1165-E and California Geol Surv Spe-cial Rept 228-E

Tormann T B Enescu J Woessner and S Wiemer (2015) Randomness ofmegathrust earthquakes implied by rapid stress recovery after the Ja-pan earthquake Nat Geosci 8 152ndash158

Tormann T B Enescu J Woessner and S Wiemer (2016) Reply to ldquoTo-hoku rupture reloadedrdquo Nat Geosci 9 183ndash185

Utsu T (1961) A statistical study on the occurrence of aftershocks Geo-phys Mag 30 521ndash605

Utsu T (1971) Aftershocks and earthquake statistics (III)mdashAnalyses of thedistribution of earthquakes in magnitude time and space with specialconsideration to clustering characteristics of earthquake occurrence(1) J Facul Sci 3 379ndash441

van der Elst N J and B E Shaw (2015) Larger aftershocks happen fartheraway Nonseparability of magnitude and spatial distributions of after-shocksGeophys Res Lett 42 5771ndash5778 doi 1010022015GL064734

Wang K and G C Rogers (2014) Earthquake preparedness should notfluctuate on a daily or weekly basis Seismol Res Lett 85 569ndash571

Werner M J A Helmstetter D D Jackson and Y Y Kagan (2011) High-resolution long-term and short-term earthquake forecasts for Califor-nia Bull Seismol Soc Am 101 1630ndash1648

Woessner J S Hainzl W Marzocchi M J Werner A M Lombardi F Cat-alli B Enescu M Cocco M C Gerstenberger and SWiemer (2011) Aretrospective comparative forecast test on the 1992 Landers sequence JGeophys Res 116 no B05305 doi 1010292010JB007846

Working Group on California Earthquake Probabilities (WGCEP) (1988)Probabilities of large earthquakes occurring in California on theSan Andreas fault US Geol Surv Open-File Rept p 62

WGCEP (1990) Probabilities of large earthquakes in the San Francisco BayRegion California US Geol Surv Circular p 51

WGCEP (1995) Seismic hazards in southern California probable earth-quakes 1994-2024 Bull Seismol Soc Am 85 379ndash439

WGCEP (2003) Earthquake probabilities in the San Francisco Bay region2002ndash2031 US Geol Surv Open-File Rept 2003-214

WGCEP (2007) The UniformCalifornia Earthquake Rupture Forecast Version2 (UCERF 2) US Geol Surv Open-File Rept 2007-1437

Zechar J D D Schorlemmer M J Werner M C Gerstenberger D ARhoades and T H Jordan (2013) Regional earthquake likelihoodmodels I First-order results Bull Seismol Soc Am 103 787ndash798

Appendix A


The UCERF3-ETAS calculation sequence is describedhere with the goal of producing a synthetic catalog of eventsgiven the following model inputs an earthquake-ruptureforecast (ERF) the desired simulation TimeSpan the regionof interest (see Fig 1 for the results presented here) a list ofall M ge25 earthquakes that have occurred up until the


chosen simulation start time (ObsEqkList) plus any fake rup-tures that are desired for testing purposes the average depthdistribution of earthquake hypocenters for the region (Seis-DepthDistribution for which that assumed here is shown inFig 6) and values for the other model parameters (Apply-GRcorr ApplyGridSeisCorr ApplySubSeisForSupraNuclTotalRateScaleFactor and ProbModel see Table 2)

Given these inputs we take the following steps to producethe desired list of simulated ruptures (SimulatedRupsList)

1 We compute the total expected regional rate of M ge25events for the specified TimeSpan by summing the equiv-alent Poisson rate of each rupture in the ERF (the latterprovides rupture probabilities Pr that can be converted torate as minus ln1 minus Pr=ΔT in which ΔT is the forecast du-ration) This total rate is dominated by gridded seismicitybecause the minimum magnitude of fault-based rupturesis generally around M 63 almost four magnitude unitsabove the minimum magnitude of gridded seismicity(M 25 here)

2 Then we reset the forecast duration in the ERF to 1 or 2years (regardless of the specified simulation duration) andcompute the total equivalent Poisson rate for each griddedseismicity and fault-based source The results are storedin an array (SourceRatesArray) and the values for fault-based sources will get updated during the simulation toaccount for any elastic-rebound probability changes Inwhat follows we assume that the relative probability offault-based ruptures remains constant (unchanged by elas-tic rebound) between the occurrence times of such eventswhich is a safe assumption given the average interval be-tween UCERF3 fault-based ruptures is about 2 years Theresults are therefore not sensitive to the exact durationspecified here as long as it is less than a few years

3 Next we generate a SpontaneousEventSampler whichholds the current yearly rate of each rupture and providesrandom sampling among them based on their relative rates

4 For each event in the ObsEqkList we then compute theexpected number of primary aftershocks N from equa-tion (2) and randomly sample an actual number assuminga Poisson process For each primary aftershock we thenrandomly sample an event time according to the temporaldecay represented in equation (1) Then we create anEqkRupture for each event set the origin time and parentinformation accordingly and put it into a chronologicallyordered list for further processing (EventsToProcessList)because other attributes such as magnitude and locationwill be determined later Again this process is repeatedfor every earthquake in the ObsEqkListThen we randomly sample the origin time of spontaneousevents considering the time dependence in λ0 over theTimeSpan as described in the main text For each we cre-ate an associated EqkRupture and put it into the EventsTo-ProcessList (again other attributes will be filled in later)

5 Next we create an ETAS_PrimaryAftershockSamplerwhich randomly selects a primary rupture from the ERF

for a given parent event taking into account both the rel-ative rates of events throughout the model and the dis-tance decay of aftershocks This component which isdescribed extensively in the main text has access to theSourceRatesArray defined here in order to stay informedabout any elastic-rebound probability changes that occurduring the simulation

6 Now we loop over the EventsToProcessList which is inchronological order to fill in the other information abouteach event (because they currently contain only origin timeand parent information) For each event we do the following

bull If the event is spontaneous (no parent)Sample an ERF rupture from SpontaneousEventSam-pler and fill in event attributes for the triggered earth-quake accordingly (eg magnitude rake and rupturesurface) If it is a gridded seismicity rupture randomlychoose an epicenter assuming a uniform distributionwithin the grid cell and then randomly sample a hypo-central depth from the SeisDepthDistribution (Fig 6)

bull Otherwise it is a triggered event so do the followingSample an ERF rupture from the ETAS_PrimaryAfter-shockSampler and fill in the event attributes accordingly(eg magnitude rake and rupture surface) Again de-tails on how this is done are given in the main text

bull Move the event from EventsToProcessList to Simulate-dRupsList

bull Sample a number of primary aftershocks and their ori-gin times for the event as outlined in step (4) above (butonly over what remains of the simulation time span) andadd each to the EventsToProcessList (which againmaintains chronological order within the list)

bull If the sampled event is a fault-system rupture (supraseis-mogenic) and the ProbModel parameter is not set asPOISSON apply elastic-rebound probability changesin the ERF by setting the date of last event on all sub-sections of the rupture to the origin time and change thestart time of the forecast to the origin time of thesampled event as well Then update the SourceRatesArray (the pointer held by ETAS_PrimaryAftershock-Sampler) and the SpontaneousEventSampler accordinglywhich will reduce probabilities for events that utilize thesections just ruptured and will update the probabilities ofother events due to time having marched on

bull Continue looping through EventsToProcessList until itis empty

Appendix B

Algorithm for Sampling Primary Aftershocksfrom the ERF

This section provides further details on how the ETAS_PrimaryAftershockSampler component samples aftershocksfrom the ERF For a given parent event the algorithm pro-ceeds as follows

32 E H Field et al

1 Define the point on the parent rupture surface from whichthe triggering occurs If the mainshock is a fault-basedrupture randomly sample a point from that surface (depthdependence of seismicity is not applied in this samplingbut it could be) For non-fault-based ruptures use the hy-pocenter ifM le4 or randomly sample a point from withina sphere centered on the hypocenter for which the radiusof the sphere is given by equation (4) and the samplingdistribution is uniform within This is the ParentLocationor again the point from which the triggering occurs

2 Translate this location to the nearest cube corner (essentiallymoving the ParentLocation by no more than sim17 kmwhich is about half the distance between opposite cube cor-ners) This is referred to as the TranslatedParentLocation

3 According to the depth of the TranslatedParentLocationuse the appropriate DistanceDecayCubeSampler to sam-ple a neighboring cube from which the triggered eventnucleates (see main text for details)

4 From the chosen cube randomly sample whether the rup-ture is from a gridded seismicity source or from one ofany fault subsections therein (according to current rela-tive nucleation rates for each and accounting for any elas-tic-rebound effects on the faults)

5 If the sampled event is a gridded seismicity rupture usethe DistanceDecayCubeSampler to randomly sample ahypocentral location inside the cube and translate thislocation by the exact opposite amount applied in step(2) For now gridded seismicity events are treated aspoint sources but finite surfaces could be assigned if de-sired Also sample a magnitude and focal mechanism ac-cording to their relative likelihoodsOtherwise the sampled event is from a fault subsectionso randomly sample one of the ruptures that can nucleatefrom the subsection according to the current relative like-lihood of each doing so

6 Return information about the sampled event

US Geological SurveyDenver Federal CenterPO Box 25046 MS 966Denver Colorado 80225-0046fieldusgsgov

(EHF)

Southern California Earthquake CenterUniversity of Southern California3651 Trousdale Parkway 169Los Angeles California 90089-0742

(KRM THJ)

US Geological Survey345 Middlefield Road MS 977Menlo Park California 94025

(JLH AJM)

US Geological Survey525 S Wilson AvenuePasadena California 91106

(MTP Nv)

Lamont Doherty Earth ObservatoryColumbia UniversityPalisades New York 10964

(BES)

School of Earth Sciences and Cabot InstituteUniversity of BristolWills Memorial Building Queenrsquos RoadBristol BS8 1RJ UK

(MJW)

Manuscript received 2 June 2016


A Spatiotemporal Clustering Model for the Third Uniform

California Earthquake Rupture Forecast (UCERF3-ETAS)

Toward an Operational Earthquake Forecast

by Edward H Field Kevin R Milner Jeanne L Hardebeck Morgan T Page Nicholas van derElst Thomas H Jordan Andrew J Michael Bruce E Shaw and Maximilian J Werner

Abstract We the ongoing Working Group on California Earthquake Probabil-ities present a spatiotemporal clustering model for the Third Uniform CaliforniaEarthquake Rupture Forecast (UCERF3) with the goal being to represent after-shocks induced seismicity and otherwise triggered events as a potential basis foroperational earthquake forecasting (OEF) Specifically we add an epidemic-type after-shock sequence (ETAS) component to the previously published time-independent andlong-term time-dependent forecasts This combined model referred to as UCERF3-ETAS collectively represents a relaxation of segmentation assumptions the inclusionof multifault ruptures an elastic-rebound model for fault-based ruptures and a state-of-the-art spatiotemporal clustering component It also represents an attempt to mergefault-based forecasts with statistical seismology models such that information on faultproximity activity rate and time since last event are considered in OEF We describeseveral unanticipated challenges that were encountered including a need for elasticrebound and characteristic magnitudendashfrequency distributions (MFDs) on faults bothof which are required to get realistic triggering behavior UCERF3-ETAS producessynthetic catalogs ofM ge25 events conditioned on any priorM ge25 events that areinput to the model We evaluate results with respect to both long-term (1000 year)simulations as well as for 10-year time periods following a variety of hypotheticalscenario mainshocks Although the results are very plausible they are not always con-sistent with the simple notion that triggering probabilities should be greater if a main-shock is located near a fault Important factors include whether the MFD near faultsincludes a significant characteristic earthquake component as well as whether largetriggered events can nucleate from within the rupture zone of the mainshock BecauseUCERF3-ETAS has many sources of uncertainty as will any subsequent version orcompeting model potential usefulness needs to be considered in the context of actualapplications

Electronic Supplement Figures showing discretization verification of theDistanceDecayCubeSampler average simulated participation rate and average cumu-lative MFD

Introduction

Long-term earthquake forecasts which are applicablefrom decades to centuries represent our first line of defensewith respect to mitigating earthquake risk especially in termsof informing building codes We also know however thataftershocks and otherwise triggered events can be large anddamaging as demonstrated by several earthquakes includingthe 2011M 63 Christchurch New Zealand event (eg Kai-ser et al 2012) The ability to forecast earthquakes on

shorter time scales such as days to decades is known as op-erational earthquake forecasting (OEF) which also involvesthe dissemination of authoritative information to inform riskmitigation decisions (Jordan et al 2011) The history mo-tivation and challenges associated with OEF have been dis-cussed elsewhere (Jordan and Jones 2010 Jordan et al2011 2014 Peresan et al 2012 Wang and Rogers 2014)and significant progress with OEF has been made in both

1

Bulletin of the Seismological Society of America Vol 107 No 1 pp ndash February 2017 doi 1017850120160173


Modeling Goals



















2 E H Field et al





Model Description



Model Overview












The ETAS Model











4 E H Field et al




λt x λ0μx X


times csr dminusq 1







t2

t1

c tminuspdt

k10MminusMmin







40 50 60 70 80

Magnitude

1 0ndash5

1 0ndash4

1 0ndash3

1 0ndash2

1 0ndash1

1 0 0

1 0 1

1 0 2

Cu

mu

lati

ve R

ate

(per

yr)

TotalOn Fault




and 2011






6 E H Field et al









25 00529 00362 00762 01653 0053 016530 01672 01144 02409 05228 0053 016535 05287 03616 07619 16533 0053 016540 16719 11436 24093 52281 0053 016545 52869 36164 76189 16533 0053 0165 0529 16550 16719 11436 24093 52281 0053 0165 0529 16555 52869 36164 76189 16533 0053 0165 0529 16560 16719 11436 24093 52281 0053 0165 0529 16565 52869 36164 76189 16533 0053 0165 0529 16570 16719 11436 24093 52281 0050 0157 0529 16575 52869 36164 76189 16533 0039 0121 0529 16580 16719 11436 24093 52281 0013 0041 0529 165


3 4 5 6 7 8

Lo

g10

Rat

e (p

er y

r)

3 4 5 6 7 8ndash 8

ndash 6

ndash 4

ndash 2

0

Magnitude Magnitude

ndash 8

ndash 6

ndash 4

ndash 2


(a) (b)








CharFactor

Cu

mu

lati

ve F

ract

ion


1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)


UncorrectedValues




35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


Imperialfault

Robinson Creekfault


fault



fault


8 E H Field et al




35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)




Modeling Goals



















2 E H Field et al





Model Description



Model Overview












The ETAS Model











4 E H Field et al




λt x λ0μx X


times csr dminusq 1







t2

t1

c tminuspdt

k10MminusMmin







40 50 60 70 80

Magnitude

1 0ndash5

1 0ndash4

1 0ndash3

1 0ndash2

1 0ndash1

1 0 0

1 0 1

1 0 2

Cu

mu

lati

ve R

ate

(per

yr)

TotalOn Fault




and 2011






6 E H Field et al









25 00529 00362 00762 01653 0053 016530 01672 01144 02409 05228 0053 016535 05287 03616 07619 16533 0053 016540 16719 11436 24093 52281 0053 016545 52869 36164 76189 16533 0053 0165 0529 16550 16719 11436 24093 52281 0053 0165 0529 16555 52869 36164 76189 16533 0053 0165 0529 16560 16719 11436 24093 52281 0053 0165 0529 16565 52869 36164 76189 16533 0053 0165 0529 16570 16719 11436 24093 52281 0050 0157 0529 16575 52869 36164 76189 16533 0039 0121 0529 16580 16719 11436 24093 52281 0013 0041 0529 165


3 4 5 6 7 8

Lo

g10

Rat

e (p

er y

r)

3 4 5 6 7 8ndash 8

ndash 6

ndash 4

ndash 2

0

Magnitude Magnitude

ndash 8

ndash 6

ndash 4

ndash 2


(a) (b)








CharFactor

Cu

mu

lati

ve F

ract

ion


1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)


UncorrectedValues




35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


Imperialfault

Robinson Creekfault


fault



fault


8 E H Field et al




35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)







Model Description



Model Overview












The ETAS Model











4 E H Field et al




λt x λ0μx X


times csr dminusq 1







t2

t1

c tminuspdt

k10MminusMmin







40 50 60 70 80

Magnitude

1 0ndash5

1 0ndash4

1 0ndash3

1 0ndash2

1 0ndash1

1 0 0

1 0 1

1 0 2

Cu

mu

lati

ve R

ate

(per

yr)

TotalOn Fault




and 2011






6 E H Field et al









25 00529 00362 00762 01653 0053 016530 01672 01144 02409 05228 0053 016535 05287 03616 07619 16533 0053 016540 16719 11436 24093 52281 0053 016545 52869 36164 76189 16533 0053 0165 0529 16550 16719 11436 24093 52281 0053 0165 0529 16555 52869 36164 76189 16533 0053 0165 0529 16560 16719 11436 24093 52281 0053 0165 0529 16565 52869 36164 76189 16533 0053 0165 0529 16570 16719 11436 24093 52281 0050 0157 0529 16575 52869 36164 76189 16533 0039 0121 0529 16580 16719 11436 24093 52281 0013 0041 0529 165


3 4 5 6 7 8

Lo

g10

Rat

e (p

er y

r)

3 4 5 6 7 8ndash 8

ndash 6

ndash 4

ndash 2

0

Magnitude Magnitude

ndash 8

ndash 6

ndash 4

ndash 2


(a) (b)








CharFactor

Cu

mu

lati

ve F

ract

ion


1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)


UncorrectedValues




35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


Imperialfault

Robinson Creekfault


fault



fault


8 E H Field et al




35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)










The ETAS Model











4 E H Field et al




λt x λ0μx X


times csr dminusq 1







t2

t1

c tminuspdt

k10MminusMmin







40 50 60 70 80

Magnitude

1 0ndash5

1 0ndash4

1 0ndash3

1 0ndash2

1 0ndash1

1 0 0

1 0 1

1 0 2

Cu

mu

lati

ve R

ate

(per

yr)

TotalOn Fault




and 2011






6 E H Field et al









25 00529 00362 00762 01653 0053 016530 01672 01144 02409 05228 0053 016535 05287 03616 07619 16533 0053 016540 16719 11436 24093 52281 0053 016545 52869 36164 76189 16533 0053 0165 0529 16550 16719 11436 24093 52281 0053 0165 0529 16555 52869 36164 76189 16533 0053 0165 0529 16560 16719 11436 24093 52281 0053 0165 0529 16565 52869 36164 76189 16533 0053 0165 0529 16570 16719 11436 24093 52281 0050 0157 0529 16575 52869 36164 76189 16533 0039 0121 0529 16580 16719 11436 24093 52281 0013 0041 0529 165


3 4 5 6 7 8

Lo

g10

Rat

e (p

er y

r)

3 4 5 6 7 8ndash 8

ndash 6

ndash 4

ndash 2

0

Magnitude Magnitude

ndash 8

ndash 6

ndash 4

ndash 2


(a) (b)








CharFactor

Cu

mu

lati

ve F

ract

ion


1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)


UncorrectedValues




35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


Imperialfault

Robinson Creekfault


fault



fault


8 E H Field et al




35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)






λt x λ0μx X


times csr dminusq 1







t2

t1

c tminuspdt

k10MminusMmin







40 50 60 70 80

Magnitude

1 0ndash5

1 0ndash4

1 0ndash3

1 0ndash2

1 0ndash1

1 0 0

1 0 1

1 0 2

Cu

mu

lati

ve R

ate

(per

yr)

TotalOn Fault




and 2011






6 E H Field et al









25 00529 00362 00762 01653 0053 016530 01672 01144 02409 05228 0053 016535 05287 03616 07619 16533 0053 016540 16719 11436 24093 52281 0053 016545 52869 36164 76189 16533 0053 0165 0529 16550 16719 11436 24093 52281 0053 0165 0529 16555 52869 36164 76189 16533 0053 0165 0529 16560 16719 11436 24093 52281 0053 0165 0529 16565 52869 36164 76189 16533 0053 0165 0529 16570 16719 11436 24093 52281 0050 0157 0529 16575 52869 36164 76189 16533 0039 0121 0529 16580 16719 11436 24093 52281 0013 0041 0529 165


3 4 5 6 7 8

Lo

g10

Rat

e (p

er y

r)

3 4 5 6 7 8ndash 8

ndash 6

ndash 4

ndash 2

0

Magnitude Magnitude

ndash 8

ndash 6

ndash 4

ndash 2


(a) (b)








CharFactor

Cu

mu

lati

ve F

ract

ion


1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)


UncorrectedValues




35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


Imperialfault

Robinson Creekfault


fault



fault


8 E H Field et al




35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)




t2

t1

c tminuspdt

k10MminusMmin







40 50 60 70 80

Magnitude

1 0ndash5

1 0ndash4

1 0ndash3

1 0ndash2

1 0ndash1

1 0 0

1 0 1

1 0 2

Cu

mu

lati

ve R

ate

(per

yr)

TotalOn Fault




and 2011






6 E H Field et al









25 00529 00362 00762 01653 0053 016530 01672 01144 02409 05228 0053 016535 05287 03616 07619 16533 0053 016540 16719 11436 24093 52281 0053 016545 52869 36164 76189 16533 0053 0165 0529 16550 16719 11436 24093 52281 0053 0165 0529 16555 52869 36164 76189 16533 0053 0165 0529 16560 16719 11436 24093 52281 0053 0165 0529 16565 52869 36164 76189 16533 0053 0165 0529 16570 16719 11436 24093 52281 0050 0157 0529 16575 52869 36164 76189 16533 0039 0121 0529 16580 16719 11436 24093 52281 0013 0041 0529 165


3 4 5 6 7 8

Lo

g10

Rat

e (p

er y

r)

3 4 5 6 7 8ndash 8

ndash 6

ndash 4

ndash 2

0

Magnitude Magnitude

ndash 8

ndash 6

ndash 4

ndash 2


(a) (b)








CharFactor

Cu

mu

lati

ve F

ract

ion


1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)


UncorrectedValues




35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


Imperialfault

Robinson Creekfault


fault



fault


8 E H Field et al




35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)











25 00529 00362 00762 01653 0053 016530 01672 01144 02409 05228 0053 016535 05287 03616 07619 16533 0053 016540 16719 11436 24093 52281 0053 016545 52869 36164 76189 16533 0053 0165 0529 16550 16719 11436 24093 52281 0053 0165 0529 16555 52869 36164 76189 16533 0053 0165 0529 16560 16719 11436 24093 52281 0053 0165 0529 16565 52869 36164 76189 16533 0053 0165 0529 16570 16719 11436 24093 52281 0050 0157 0529 16575 52869 36164 76189 16533 0039 0121 0529 16580 16719 11436 24093 52281 0013 0041 0529 165


3 4 5 6 7 8

Lo

g10

Rat

e (p

er y

r)

3 4 5 6 7 8ndash 8

ndash 6

ndash 4

ndash 2

0

Magnitude Magnitude

ndash 8

ndash 6

ndash 4

ndash 2


(a) (b)








CharFactor

Cu

mu

lati

ve F

ract

ion


1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)


UncorrectedValues




35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


Imperialfault

Robinson Creekfault


fault



fault


8 E H Field et al




35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)








CharFactor

Cu

mu

lati

ve F

ract

ion


1 01

1 02

00

10

02

04

06

08

(a) (b)

(c) (d)


UncorrectedValues




35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


35deg

40deg

0 100 200 300

km


Imperialfault

Robinson Creekfault


fault



fault


8 E H Field et al




35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)






35deg

40deg


35deg

40deg


35deg

40deg



0 100 200 300

km

(a) (b)

(c) (d)

0 6 12


35deg

40deg

0 100 200 300

km



0 100 200 300

km



0 100 200 300

km












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)












1 3 5 7 9 11 13 15 17 19 21 23

Depth (km)

000

005

010

015

020

025

030F

ract

ion



10 E H Field et al







San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)









San Andreas Fault










AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)






AsηsPAsηs 3










12 E H Field et al






24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)








24 km

Latitude

Longitude

Dep

th



(a) (b)

(c) (d)

minus2 minus1 0 1 2

Log10 (CharFactor )

CubeCharFactor

Values














14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)













14 E H Field et al


Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)




Results












minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)









minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0

Parent Depth 6 km


minus118˚

minus116˚

34˚

36˚ 24

LongitudeLatitude

Depth

0


(a)

(b)


40 50 60 70 80 90

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

10 0

10 1

10 2

Cum

ulat

ive

Rat

e (p

er y

r)


16 E H Field et al





35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)







35deg

40deg

King Range2011 CFM


Morales(East)


(1992 Landers Qk)




(a) (b)



Simulatedvs

Target Rates



35deg

40deg

0 100 200 300

km

minus2 minus1 0 1 2


35deg

40deg

0 100 200 300

km



35deg

40deg

0 100 200 300

km










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)










0 5 0 100 150 200 250 300 350 400 450

Years

0E0

1E-3

2E-3

3E-3

4E-3

5E-3

6E-3

7E-3

8E-3

9E-3

1E-2

11E-2

12E-2

13E-2

14E-2

15E-2

16E-2

Den

sity



18 E H Field et al




ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)






ndash2

ndash1

0

1

Log1

0(C

harF

acto

r)




00

05

10

15

20

Den

sity


00 10 20 30 40 50

00

05

10

15

20

Den

sity


00

05

10

15

20

Den

sity


00 10 20 30 40 50

UCERF3-ETAS

UCERF3-TD

RSQSim

(a)

(b)

(c)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)















20 E H Field et al










35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)












35deg

40deg

0 100 200 300

km








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)








minus4 minus2 0



35deg

40deg

0 100 200 300

km





35deg

40deg

0 100 200 300

km


22 E H Field et al











35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)













35deg

40deg

0 100 200 300

km


01 10 10 100 1000

Distance (km)

10ndash7

10ndash6

10ndash5

10ndash4

10ndash3

10ndash2

10ndash1

100

Afte

rsho

ck D

ensi

ty (

per

km)


Simulation Mean







25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)






25 30 35 40 45 50 55 60 65 70 75 80 85 90Magnitude

0

100

200

300

400

500

600

700

800

Num

Sim

ulat


Primary Aftershocks

Scenario M 50



3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104

Cum

ulat

ive

Num

ber

Mean

Mode

Median

25975 Fractiles

Primary Mean


Regional GR Mean



Log10 Time (Days)

10ndash3

10ndash2

10ndash1

100

101

102

103

Rat

e (p

er d

ay)


Simulation Mean


Primary Aftershocks

All Aftershocks


24 E H Field et al




SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)






SAF Mojave SM 50

FULL_TD 10317 625 03464 664 01034 125 00122 778 00080 1480 00054 435

SAF Mojave SM 50

NO_ERT 28180 171 09421 181 02626 317 00423 270 00241 446 00148 119

SAF Mojave SM 55

FULL_TD 21791 418 07261 440 02066 790 00248 500 00151 883 00110 280

SAF Mojave SM 55

NO_ERT 86379 165 28433 172 07338 281 01210 244 00760 445 00502 128


FULL_TD 55615 169 18208 175 04359 264 00484 155 00286 265 00201 812


NO_ERT 29031 882 95457 917 24509 148 03980 127 02434 226 01630 658


FULL_TD 61376 186 21504 207 06988 424 00461 147 00236 219 00155 626


NO_ERT 26093 793 84805 815 20705 125 03741 120 02241 208 01461 590

SAF Mojave SM 70

FULL_TD 11423 069 31868 061 01510 018 00211 013 00096 018 00050 040

SAF Mojave SM 70

NO_ERT 24218 147 69864 134 06886 083 02082 133 01539 285 01096 883

SAF Mojave SM 74

FULL_TD 23956 058 64953 050 01508 007 00197 005 00079 006 00016 005

SAF Mojave SM 74

NO_ERT 31508 076 87965 067 04674 023 01420 036 00840 062 00537 172

SAF Mojave SM 78

FULL_TD 46032 044 12841 039 04892 009 00400 004 00112 003 00009 001

SAF Mojave SM 78

NO_ERT 48481 047 13708 042 07399 014 01377 014 00410 012 00060 008

SAF PeninsulaM 55

FULL_TD 03693 071 01008 061 00050 019 00016 032 00009 053 00001 025

SAF PeninsulaM 55

NO_ERT 03865 074 01082 066 00079 030 00026 052 00016 094 00000 000

SAF PeninsulaM 63

FULL_TD 21673 066 05889 057 00339 021 00123 039 00072 067 00001 004

SAF PeninsulaM 63

NO_ERT 22387 068 06015 058 00391 024 00139 044 00079 073 00006 024

SAF PeninsulaM 70

FULL_TD 10607 064 28391 054 01513 018 00473 030 00283 052 00028 023

SAF PeninsulaM 70

NO_ERT 11129 067 29995 057 01818 022 00562 036 00296 055 00040 032

Surprise Valley55

FULL_TD 16443 315 07817 474 01009 386 00127 256 00000 000 00000 000

Surprise Valley55

NO_ERT 53264 102 23494 142 05334 204 01018 205 00002 012 00000 000


FULL_TD 03056 059 00853 052 00023 009 00007 014 00002 012 00000 000


NO_ERT 03655 070 01052 064 00048 018 00012 024 00007 041 00002 051

Central ValleyM 50

FULL_TD 01268 077 00383 073 00062 075 00002 013 00000 000 00000 000

Central ValleyM 50

NO_ERT 01126 068 00321 062 00039 047 00003 019 00000 000 00000 000

Bombay BeachM 48

FULL_TD 02091 201 00581 176 00085 162 00034 344 00020 586 00007 830

Bombay BeachM 48

NO_ERT 02407 231 00716 217 00126 241 00044 445 00023 674 00006 703








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)








Discussion

Main Challenges








3 4 5 6 7 8 9

Magnitude

10ndash4

10ndash3

10ndash2

10ndash1

100

101

102

103

104C

umul

ativ

e N

umbe

r


26 E H Field et al






















28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)
























28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)












28 E H Field et al





Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)







Conclusions




35deg

0 100 200 300

km








Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)







Data and Resources


Acknowledgments



References

















30 E H Field et al











































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)













































Appendix A



















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)


















Appendix B



32 E H Field et al








(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)










(EHF)


(KRM THJ)


(JLH AJM)


(MTP Nv)


(BES)


(MJW)



Documents

Field, E. H., Milner, K. R., Hardebeck, J. L., Page, M. T., van der Elst, … · Elst, Thomas H. Jordan, Andrew J. Michael, Bruce E. Shaw, and Maximilian J. Werner Abstract We, the