Report on the CIG/IRIS Workshop on Computational Seismology

Based upon discussions during this meeting we make the following structural recommendations:

1. Establishment of a Seismology Working Group. The initial members of this Working Group are Alan Levander, Mike Ritzwoller, Jeroen Tromp, and Michael Wyssession. This working group works closely with the seismologist (initially Ritzwoller or Tromp) on the CIG Scientific Steering Committee to set and carry out seismology priorities for CIG. The Working Group will consider expanding its membership to meet disciplinary breadth needs as appropriate to changing priorities and needs of CIG, and will discuss its composition and the duration of service with the CIG Scientific Steering Committee.
2. A principal job of the Seismology Working Group is to establish and oversee a number of Seismology Task Groups whose role it is to implement specific seismology priorities within the context of CIG.

   We established the following short-term (i.e., 2005-2006) goals:

   • A number of open-source 1D codes (e.g., Generalized Ray Theory, reflectivity, WKBJ, and normal mode) and 3D codes (e.g., finite-difference, finite-element, spectral-element, and eikonal solvers) will be brought within the CIG framework. The Seismology Working Group will set priorities shortly following the Workshop and will define the needed Seismology Task Groups to implement the codes within the framework.
   • Codes in the framework will be open source, well documented, and come with a number of sample benchmarks. We will encourage contributions of all seismic modeling and analysis codes with the understanding that CIG will provide different levels of support to develop each for the framework, depending on priority of the project.
   • The goal is to provide a set of modular codes to foster maximum flexibility and inter-operability. There is much interest in hybrid method development.
   • Particular attention needs to be focused on defining model parameterization, which includes source, receivers, and Earth model parameterization. There will likely be a need for development of translators to move from one Earth model definition to another. Again, this will be the focus of one or more Seismology Task Groups.
   • Whenever possible, CIG seismology efforts will be coordinated with SPICE, GEON, IRIS, and SCEC. Examples include model parameterization, the organization of workshops, visualization, and grid computing. SPICE, in particular, provides a natural partner for CIG with its emphasis on code development and training.
   • In the summer/fall of 2006, a theoretical & computational seismology workshop will be organized. In addition to scientific presentations, the meeting will offer hands-on experience with the CIG framework and the seismology codes then available through the CIG framework.
   • Conversations will be held with the IRIS/DMS and IRIS leadership to establish a highly benchmarked global 1D code at the IRIS/DMC to produce synthetic seismograms that will be available to accompany data orders from the DMC. This is seen as a first step toward an increasingly sophisticated and powerful system of synthetic seismogram generation available through the IRIS DMS.
   • The Seismology Working Group is tasked with working toward the completion of these goals. The list of short-term goals will be evaluated and updated at the next workshop in the summer/fall of 2006. All efforts will be coordinated with the CIG Science Steering Committee.
   We identified the following long-term (i.e., 3-4 year) goals:
   • Automated/On-Demand simulations; i.e., the establishment of a Seismology Science Portal. o Automated simulations: Near real-time 1D and/or 3D synthetics to accompany IRIS data for all events over a certain magnitude threshold using past and emerging events in the Harvard CMT catalog. Synthetics would be made optionally available with the data through CIG/IRIS web-servers.
   • On-Demand simulations: At a higher level of user control, on-demand simulations for user-specified Earth models, source descriptions, receiver configurations, and run parameters is also seen as a useful future research tool.
   • TeraGrid resources are useful for both the automated and user-controlled on-demand simulations.
   • GEON and SCEC serve as models for aspects of this functionality, and conduits for advice, guidance, and support.
   • There is the need for a database of seismic models, including structural models of the crust and mantle together with databases of topography and bathymetry. Various resolutions are needed to match the capabilities of codes being developed under CIG. Mechanisms for the contribution of models must be established.
   • Investigate the feasibility to facilitate the development of data processing tools for field and laboratory use. These could include both low-level routines for standard data manipulation (e.g., filtering, simple array analysis), higher-level functionality such as earthquake location, traveltime picking, moment tensor analysis, and high-level functionality such as tomography and receiver functions (perhaps with migration), shear wave splitting, etc. This may be accomplished through a Task Group.
   • 2D and 3D visualization is increasingly important in seismology and presents an area of overlap with other CIG efforts that requires coordination.
   • Imaging/tomographic tools may also be included productively within the CIG framework.
   • There is concern about the long-term viability of the Harvard CMT project. CIG may be able to help in ensuring its continued operation.
   • The European SPICE group has an interesting funding model that might be successfully tried here: SPICE provides training fellowships for post-docs and graduate students. For CIG such awardees would develop existing codes for the CIG framework as part of their responsibilities for funding. The awardees could be institutionally based, but be funded with the understanding that they would spend a substantial amount of time at Caltech learning framework coding.