Home  /  SCEC Workshops  /  SCEC Community Workshop: Community Stress Model

SCEC Community Workshop: Community Stress Model

Conveners: Jeanne Hardebeck, Brad Aagaard, Thorsten Becker, Bruce Shaw, and John Shaw
Dates: October 15-16, 2012
Location: SCEC Boardroom, University of Southern California, Los Angeles, CA
SCEC Award and Report: 12114

SUMMARY: Crustal stress is a fundamental quantity that is relevant to many aspects of the earthquake problem. SCEC4 has committed to the development of the Community Stress Model (CSM) to provide the SCEC community with better constraints on the stress field, and with a means to formally test physical connections between observations and stress models. The intended CSM end‐product, a model or suite of models for the 4D stress tensor in the California lithosphere, will be useful for numerous SCEC core science issues. These include a better understanding of how faults are loaded, redistribute strain as a result of earthquakes, and evolve over time. There is a range of potential uses for the CSM, including seismic hazard estimates, crustal seismicity studies, dynamic earthquake rupture models and earthquake simulators. Moreover, even the discussions of the issues involved in the construction of a CSM (such as the completeness of physical models, noise, and uncertainties) are expected to lead to scientific progress.

At first CSM workshop, it was decided that the best first step for the CSM would be to compile and compare all of the existing relevant data and models held by the SCEC community. A call went out for contributions of data and models, along with the definition of a common format for the models, and a grid of points covering southern California at which the models should be sampled. Sixteen stress and stressing rate models were submitted, along with some contributed borehole stress measurements.

At the second workshop, held October 15‐16, 2012 at USC, we compared the submitted models, and held focused discussions on topics including: reconciling models, validating models with data, quantifying uncertainty in models, including stochastic stress heterogeneity in models, interaction with user communities, and IT needs. The workshop began with formal presentations of the submitted models and model comparisons, but the majority of time was focused on discussion of important issues. We identified short‐terms needs for the next year of CSM development, and solicited SCEC proposals for 2013 to address those needs.


  1. The orientation of the stress field:

    The orientations of the stress axes of the submitted stress models were impressively similar. Three different classes of models of the stress field were submitted, based on different data and assumptions, and the three types of models agree on the direction of maximum horizontal compressive stress (SHmax) to <15° over almost all of the southern California upper crust (Figure 1). There was mild disagreement in the style of faulting in some regions near the edges of the model.

    Next steps: The model contributors have identified the likely cause of faulting style disagreement and are working to reconcile this aspect of the models.

  2. The deviatoric stress:

    Constraining the amplitude of the deviatoric stress at seismogenic depths is a long‐standing difficult problem, dating from at least the discovery of the “San Andreas heat‐flow paradox.” Unsurprisingly, the submitted models did not agree on the magnitude of the deviatoric stress. There was both an order of magnitude disagreement in the magnitude of the deviatoric stress, and a disagreement as to whether the deviatoric stress increases significantly with depth over the seismogenic zone (Figure 2).

    Next steps: Encourage further research to attempt to place bounds on the magnitude of the deviatoric stress at depth in southern California.

  3. The stressing rate:

    The submitted stressing rate models generally agreed on the scalar stressing rate close to the faults of the San Andreas system. There were significant disagreements along some faults, however, related not only to which faults were included in particular fault‐based models, but also in the stressing rates along those faults and the decay of stressing rate away from the faults (Figure 3). Similar discrepancies in other compilations of deformation models were discussed.

    Next steps: Model contributors should meet to understand how differences in assumptions and data lead to the differences between stressing rate models, and plan strategies to reconcile differences. Additional deformation models have been produced as part of UCERF3: these models need to be translated into stressing rate and folded into our comparisons.

  4. The orientation of the stressing rate tensor:

    Most of the stressing rate models agreed well in the orientations of SHmax and the faulting style near the major faults of the San Andreas system. However, off of the major faults, this agreement breaks down. A particular example is the Ventura area, where a local model produces a more thrust style stressing rate tensor than the regional models. Additionally, the agreement becomes worse with depth, due to differences in assumptions about locking depths.

    Next steps: Same as for (3).

  5. Validating Models with Data:

    Our call for data was not met with much response. So we have identified a high‐priority need to compile additional data and use this data to validate models.

    Next steps: Solicit contributions of borehole stress measurements, geological indicators of stress direction and magnitude, shear‐wave splitting measurements, and other relevant data. Develop metrics for validating models against data, and begin applying these metrics to the submitted models.

  6. Quantifying Model Uncertainty:

    While the call for models included a request for uncertainty estimates, the majority of submitted models did not include any uncertainty. At the workshop we discussed the difficulty of estimating uncertainty for forward models: essentially one must follow a Monte Carlo approach and produce a large suite of models and examine their variability. For most models, this would take too much time computationally.

    Next steps: Individual model contributors may work on producing Monte Carlo uncertainty estimates. Additionally, the range of the models also gives a measure of the uncertainty.

  7. Stochastic Stress Heterogeneity:

    We discussed how to implement stochastic stress heterogeneity in the CSM. We would like for model users to be able to generate heterogeneity that they can add to any of the “deterministic” CSM models. Because of the stochastic nature of these models, we would like to be able to generate multiple realizations, rather than have a single heterogeneous stress model.

    Next steps: Develop software that users can run to add stochastic stress heterogeneity to any model in the CSM format. This software will be based on existing stochastic stress heterogeneity codes developed by Debbie Smith.

  8. IT Needs:

    Our immediate IT need is for a model and data repository and a web portal and that allows SCEC scientists and other contributors and users to access, compare, and validate CSM models.

    Next steps: Develop a web site that allows the SCEC community to download the submitted CSM data and models, as well as to generate plots of any of the models and/or comparisons of selected models. The plotting tools would be based on GMT scripts developed by Jeanne Hardebeck for the model comparisons shown at the workshop.

In-Person -- Brad Aagaard (USGS), Thorsten Becker (USC), Peter Bird (UCLA), Yuri Fialko (SIO/UCSD), Jeanne Hardebeck (USGS), Egill Hauksson (Caltech), Elizabeth Hearn, Tran Huynh (USC), Dave Jackson (UCLA), Tom Jordan (USC), Elizabeth Madden (Stanford), Jessica Murray (USGS), David Sandwell (UCSD/SIO), Bruce Shaw (Columbia), Debbie Smith (CIW), Joann Stock (Caltech), Anne Strader (UCLA), and Wenzheng Yang (Caltech)
Remote -- Kelly Blake (Navy, COSO), Michele Cooke (UMass Amherst), Attreyee Ghosh (Bangalore Institute of Technology), Kaj Johnson (Indiana), Jack Loveless (Smith), Karen Luttrell (USGS), Scott Marshall (Appalachian State), and John Shaw (Harvard)

Presentation slides may be downloaded by clicking the pdf links following the title. PLEASE NOTE: Slides are the author’s property. They may contain unpublished or preliminary information and should only be used while viewing the talk.

MONDAY, October 15, 2012

09:00 Continental Breakfast Room 169
09:30-10:00 Introduction
  • Overview of workshop
  • Outcomes from 2011 workshop
  • Stress model format
  • Participants
Jeanne Hardebeck
10:00-12:30 Descriptions of Contributed Models
10 min + 5 min Q&A per model
Brad Aagaard, moderator
12:30-13:30 Lunch Room 265
13:30-14:30 Comparison of Models
  • Presentation of comparisons (30 min)
  • Discussion (30 min)
    • Features present in models
    • Consistencies and inconsistencies
Bruce Shaw, moderator

Jeanne Hardebeck

14:30-15:15 Discussion: Model branches
  • Which models can be reconciled and combined?
  • Which classes of models have distinct differences?
Thorsten Becker, moderator
15:15-15:30 Break  
15:30-16:00 Discussion: Validating the Community Stress Model
  • Stress tensor
  • Stress change tensor
Thorsten Becker, moderator
16:00-16:30 Discussion: Quantifying uncertainty in stress models Jeanne Hardebeck, moderator
16:30-17:00 Discussion: Stochastic heterogeneity in stress models
[see also Smith and Heaton (2011)]
Brad Aagaard, moderator
18:00-21:00 Group Dinner University Club

TUESDAY, October 16, 2012

08:30 Continental Breakfast Room 169
09:00-09:30 Introduction
  • Summary of Day 1
  • Community Stress Model objectives
  • Overview of Day 2
Brad Aagaard
09:30-10:45 Moving forward [1.pdf and 2.pdf] Jeanne Hardebeck, moderator
10:45-11:00 Break  
11:00-12:30 Moving forward
  • Discussion: Uncertainty
  • Discussion: Heterogeneity
Bruce Shaw, moderator
12:30-13:30 Lunch Room 265
13:30-14:00 Moving forward
  • Discussion: Model format
    • Ambiguities and limitations
    • Interpolation
    • API versus standard format
  • Discussion: IT infrastructure
Brad Aagaard, moderator
14:00-15:00 Discussion: Community Stress Model
  • What is the SCEC Community Stress Model?
    • Current vision still intact?
    • Model priorities
  • Targeting important isssues with SCEC proposals
    • Improving current models
    • Reconciling discrepancies
    • Consolidating models
Thorsten Becker, moderator
15:00 Adjourn