Oct 18

Oct 18th Discussion …

Decision time for NSHMP (& possible predicaments)

Suggested steps forward

Review logic tree options

& evaluation criteria

Decision time for NSHMP

Present Situation PossibilitiesUCERF3 is reviewed in short order and everyone is happy with methodology and resultsUCERF3 is deemed technically flawed and not fixable in time for 2014 NSHM.Given complexity, there is not enough time for adequate review & vetting of UCERF3 for NSHMUCERF3 is found to be technically sound (with inevitable adjustments), but mean hazard changes are up to factors of 10 in many areas (and only understood to some level of specificity)

Probability10%20%30%50%

NSHMP options here?...

Decision time for NSHMP

NSHMP Options if there is no UCERF3:Do not update CA ERF for 2014 NSHMsOnly update non fault-based sources (with alternative regional rates and spatial PDFs)Do (b) and add some of the new faults as type-B sources (bulge will increase) using geologic slip ratesUpdate everything including Type-A faults in a “UCERF2-like” way

Issues:What exactly is “UCERF2-like”?Who would do this (WGCEP is already near burnout & would be demoralized)?This could take as long as fixing UCERF3Given La Quila, would anyone sign off on amodel that lacks multi-fault ruptures?

Decision time for NSHMP

Present Situation PossibilitiesUCERF3 is reviewed in short order and everyone is happy with methodology and resultsUCERF3 is deemed technically flawed and not fixable in time for 2014 NSHM.Given complexity, there is not enough time for adequate review & vetting of UCERF3 for NSHMUCERF3 is found to be technically sound (with inevitable adjustments), but mean hazard changes are up to factors of 10 in many areas (and only understood to some level of specificity)

Probability10%20%30%50%

Given:

El Mayor-Cucapah (Mmax)Darfield-Christchurch (Mmax & triggering)Tahoku (segmentation)M8.6 Sumatra (“weird one”)does anyone believe we know mean hazard within 10%?Are we reluctant to put these out because it implies we’ve misled user communities (or didn’t push the epistemic uncertainties hard enough)?Is looking at hazard implications before finalizing weights cheating?

Oct 18th Discussion …

Decision time for NSHMP (& possible predicaments)

Suggested steps forward

Review logic tree options & evaluation criteria

The WGCEP Path Forward(the only one, in my opinion)

Wrap it up

(finish and publish ASAP)

in part, to satisfy contractual obligation to CEA

Let others decide what to use in 2014 NSHMs

e.g., some weighted average of old and new hazard curves at each grid node

The WGCEP Path Forward(the only one, in my opinion)

Specific steps:

Get feedback here on final branches (& weights)

Decide how to handle convergence and equation set weights

Decide on any a posteriori weighting scheme (

what to add to Morgan’s data-fits table?

)

Finalize calculations & document for review (

by Nov 1

st

if all goes well?

)

Activate review (hands on, aggressive, back and forth in terms of answering questions);

how long will this take? Finish by year end?

Finalize & publish in 2013

Oct 18th Discussion …

Decision time for NSHMP (& possible predicaments)

Suggested steps forward

Review logic tree options & evaluation criteria

Proposed Logic-Tree Branches:

Proposed Logic-Tree Branches:

Two Fault Models

S

ame number as in UCERF2

Weighted equally

Proposed Logic-Tree Branches:

Four Deformation Models

M

uch broader range than in UCERF2

Off-fault moment rates provided (UCERF2 Type-C zones gone!)A priori weights represent an average among those of a special review panel

Proposed Logic-Tree Branches:

Scaling Relationships

Only HB08 &

EllB

used in UCERF2Equal a priori weights

Proposed Logic-Tree Branches:

Slip Along Rupture (

D

sr

)Added Boxcar optionEqual a priori weights

Weldon

et al

. (2007) Average of 13 large events

???

Characteristic Slip?

???

Proposed Logic-Tree Branches:

Total

M

≥

5 Event Rate Per year, inside RELM regionFrom Felzer (Appendix L)UCERF2 had single value of 7.5, which is at the low end here (new best estimate of 8.7 represents a 16% increase)

Proposed Logic-Tree Branches:

Inversion Model

(Fault-Section Nucleation MFD)

Characteristic

UCERF2 Constrained:

* if

Type-A in UCERF2, use UCERF2 nucleation MFD instead.

UCERF2 type MFD

1/3 GR and 2/3 Char

Gutenberg-Richter Constrained:

b

=1

Proposed Logic-Tree Branches:

Inversion Model

(Fault-Section Nucleation MFD)

???

Use only

Characteristic???

???

If

Gutenberg-Richter

requires reducing slip rates by ~40%, how many would give it a relative weight > 10%?

???

Proposed Logic-Tree Branches:

M

max

Off Fault

UCERF2 had “off-fault” values of either 7.0 or 7.6El Mayor-Cucapah exceeded 7.0; so we’ve increased to 7.2 at the low endThe value of 8.0 is newWeights are different for Char vs GR branches

UCERF2

M

max

Proposed Logic-Tree Branches:

Off-Fault Spatial

Seis

PDF

Three options, with two new ones:UCERF3 Smoothed SeisDeformation Model Ave3 more could be added (1 for each deformation model)Weights are different for Char vs GR branchesExactly how these are used shortly…

Proposed Logic-Tree Branches:

Off-Fault Spatial

Seis

PDF

UCERF2 Smoothed Seismicity

UCERF3 Smoothed Seismicity

Average Deformation Model

L

og10(

P

rob

) for Each Grid Cell

(values sum to 1.0)

ABM

NeoKinema

Zeng

???

Use

d

eformation-model-specific off-fault spatial PDFs

???

Proposed Logic-Tree Branches:

Fault Moment-Rate Fixes

This deals with cases where fault moment rates are too high to satisfy all data

Currently only an issues for

GR branches?More on this shortly

Grand Inversion Results

Data

Fits: - Regional MFDs - Slip-rate fits - Paleo event-rate & ave-slip fits - Tabulation of equation-set fits and other metrics (e.g., implied CC)Implications Plots: - Participation rate maps - Parent-section MFDs (also tabulated) - Correlation between paleo sites - Implied segmentation (e.g., on SAF) - Fault-jumping statistics - Slip COVs (e.g., Hecker et al.) - Lots of stuff in SCEC VDOERF-Based Plots: - MFDs in LA and SF Boxes - Hazard curves at sites - Hazard Maps - RTGM at sites - Statewide Losses

We currently have these (and more) implemented, although we haven’t yet had time to examine everything

Evaluation Metrics:

Mean, Min, and Max from all logic-tree branches

UCERF3 Mean

UCERF3 Mean Cumulative

UCERF2