C.1. Problem Statement: The Need for Information Technology Research in Earthquake Science

In the last several years, notable advances have been made in two distinct areas of earthquake science: (1) the dynamics of fault rupture -- what happens on a time scale of seconds to minutes when a single fault breaks during a given earthquake -- and (2) the dynamics of fault systems -- what happens within a network of many faults on a time scale of years to centuries to produce the sequencing of earthquakes in a given region. Combined with the increased availability of terascale computing resources, these advances in geophysics make it possible for the first time to create fully three-dimensional (3D) simulations of fault-rupture and fault-system dynamics. Such simulations are crucial to gaining a fundamental understanding of earthquake phenomena. However, even as the challenges of understanding the geophysics are being met, we are faced with a new problem. Constructing a system-level earthquake simulation from models of constituent phenomena and executing that simulation on suitable computing platforms becomes increasingly complex due to (1) the difficulty of selecting and configuring compatible simulation models that are appropriate for the geophysics problem being studied, and (2) mapping those models onto computing resources for execution. This complexity will limit the ability of non-experts to perform accurate earthquake modeling, restricting the community that can benefit from earthquake simulation and ultimately slowing down progress in earthquake science. We propose to address this problem by developing an integrated modeling framework that automates the process of selecting, configuring, and executing models of earthquake systems. We will achieve this ambitious goal via an innovative integration of knowledge representation, knowledge acquisition, Grids, and digital libraries. The proposed research will be conducted by a collaboration between leading researchers in each of these information technology areas with earthquake scientists associated with the Southern California Earthquake Center.

C.1.a. The Opportunity for Improved Seismic Hazard Analysis

Physics-based simulations can potentially provide enormous practical benefits for assessing and mitigating earthquake risks through seismic hazard analysis. Seismic hazard analysis (SHA) seeks to describe the shaking that can be expected at a given point of the Earth's surface due to earthquakes that are likely to occur over a specified time period [1]. From the perspective of deterministic physics, this calculation requires the coupling of models that represent a series of complex physical processes:

• Unified Structural Representation (USR): a self-consistent, 3D characterization of active faults and material properties (e.g. seismic velocities) needed to describe regional deformation and seismogenic processes.
• Fault System Model (FSM): an evolving representation of the regional stress and deformation fields, capable of predicting when individual fault segments will rupture.
• Rupture Dynamics Model (RDM): a dynamical description of the nonlinear stress/displacement interactions across a rupturing fault as a function of space and time.
• Anelastic Wave Model (AWM): a computation of the propagation, interference, and attenuation of the seismic waves that travel along complex paths from a fault rupture to a target site.
• Site Response Model (SRM): a dynamical description of ground excitation in the near-surface environment at a target site, which, for strong ground motions, often involves significant nonlinearities.

At present, practitioners are forced to make many simplifications and approximations in the application of SHA to earthquake engineering and risk mitigation. Fault system models that include stress interactions are in their early stages of development, and their predictive value has not been demonstrated. Therefore, the FSM is usually replaced by long-term forecasts of the earthquake potential on individual faults derived from historical seismicity catalogs and/or geologic and geodetic slip rates. The RDM is usually replaced by an isotropic source whose strength depends only on earthquake magnitude and thus ignores many important features of real fault ruptures, such as strong azimuthal variations in radiation. The AWM is approximated in terms of an empirical "attenuation relationship" that depends only on source distance, depth, and magnitude, which does not account for wave-interference effects [2]. The SRM is approximated by an empirical function of the rock properties local to the site, rather than a dynamical calculation based on near-surface structure. Recent analyses have attempted to quantify other effects, such as basin depth [3], but a major new study by the Southern California Earthquake Center (SCEC) has concluded that, "any model that attempts to predict ground motion with only a few parameters will have substantial intrinsic variability. Our best hope for reducing such uncertainties is via waveform modeling based on the first principles of physics." [4]

C.1.b. Project Goals and Requirements

SCEC has embarked on an ambitious program to develop physics-based models of earthquake processes and integrate these models into a new scientific framework for seismic hazard analysis and risk management [5]. The success of this program will depend on the construction of a Community Modeling Environment, in which the appropriate simulation models will be developed, documented, and maintained on-line for application by SCEC, earthquake researchers elsewhere around the world, and end-users of earthquake information. This environment will function as a virtual collaboratory for the purposes of knowledge quantification and synthesis, hypothesis formulation and testing, data conciliation and assimilation, and prediction. It will greatly facilitate the system-level understanding of earthquake phenomena, and it has the potential to improve substantially the utilization of SHA in reducing earthquake losses.

The goal of the proposed project is to construct an information infrastructure for the SCEC Community Modeling Environment that can satisfy the following four requirements:

R1. Capture and manipulate the knowledge that will permit a variety of users with different levels of sophistication to configure complex computational pathways for (a) rapid prototyping of new SHA algorithms and (b) construction, validation, and dissemination of new SHA products.
R2. Enable execution of physics-based simulations and data inversions that incorporate advances in fault-system dynamics, rupture dynamics, wave propagation, and non-linear site response. These simulations must be capable of resolving the stress-interaction scales, including the inner frictional scales of faulting (< 100 m), as well as the seismic frequencies of engineering interest (> 1 Hz).
R3. Manage large, distributed collections of simulation results, as well as the large sets of geologic, geodetic and seismologic data required to validate the simulations and constrain parameter values.
R4. Provide access to SHA products and methodologies to end-users outside of the SCEC community, including practicing engineers, emergency managers, decision-makers, and the general public. This access must include intelligent handling of queries, workbench environments for user-configured calculations, visualization tools, detailed information about the legacy and pedigree of SCEC products, and mechanisms for educating end-users about SHA methodology.

In the context of the SHA problem and the extended SCEC community, these requirements present an interesting set of challenges for information technology:

• Heterogeneity and multiplicity of the models. Many object types must be manipulated and a number of algorithms must be employed, but the pathway to a computational result is contingent on factors that must be evaluated at each step along the way. Moreover, algorithm inputs and outputs can be very complex and take different meanings for different users, so that this may require not just syntactic translation but also semantic mappings and conversion.
• Distributed development of the models. Models are being developed by different organizations with differing expertise and computational resources, requiring data management and execution environments that span autonomous administration domains.
• Computational requirements of simulation models. A physics-based approach to SHA requires that a large range of potential scenarios be examined, each at much higher spatio-temporal resolution than current simulations. This erects a set of Grand Challenge computational problems that require terascale class computational resources to solve [6].
• Requirements for model management. Evolution of the digital representation is driven by physical events, simulation output results, and remote-sensing input, requiring management of multiple versions. Furthermore, the SCEC approach to SHA uses an integrated analysis of seismic, geodetic and geologic data; thus the modeling environment must manage heterogeneous data collections, including a range of existing collections and archives.
• Diversity of user community. Products of SHA are of interest to a wide range of users, from geophysicists developing the analysis, civil and structural engineers designing buildings and structures, to city planners, news media, and disaster response teams. Thus, unsophisticated users may need to manipulate sophisticated models.

To address these challenges, we propose to develop fundamental new approaches to model generation and simulation management by bringing together four distinct computer science disciplines: (a) knowledge representation and reasoning, (b) interactive knowledge acquisition, (c) digital libraries and information management, and (d) Grid resource-sharing environments. The overall architecture of the proposed SCEC Community Modeling Environment in illustrated in Figure 1 and described in more detail in § C.3. While we focus on the application of this new computer science to the practical problems of earthquake science, the techniques and integrated modeling framework that will be developed in this project will be directly applicable to many other disciplines.

The challenges described above become particularly acute when "all hell breaks loose" during a big earthquake. Modern seismic information systems, such as TriNet in Southern California, can, within a few minutes, locate regional earthquakes and produce preliminary maps of ground shaking to guide emergency response [7]. A long-term goal is to have the capability for automatically configuring an expanded set of modeling resources to assimilate seismic, geodetic, and geologic data as they are acquired in real time, to create new products such as ground-motion and damage predictions, and to distribute the output to multidisciplinary teams scattered across the region. These teams will need the means for jointly visualizing, manipulating, and modifying the products and for communicating the results to non-specialists -- engineers, emergency managers, government officials, and the media. All of these operations will have to be done under potentially stressful conditions using distributed, multiply-connected computational systems that are robust to major regional disruptions in power, communications, and transportation. The proposed project does not directly address the issues of real-time operations and robustness. However, the SCEC Community Modeling Environment, as envisaged here, should greatly facilitate the development of these capabilities by USGS scientists and others responsible for real-time seismic information systems in Southern California and elsewhere.

Figure 1. Proposed architecture for SCEC's Community Modeling Environment. Knowledge-rich community models will be developed and curated jointly by geophysicists and knowledge engineers to produce a comprehensive knowledge base of earthquake science. Knowledge representation and reasoning techniques will enable semantic translations and term mappings as well as complex inference. The Grid will provide access to computing and storage resources to execute complex computational pathways. The community models will be grounded in data collections and software repositories through digital libraries technology. This knowledge-rich environment will enable interactive knowledge acquisition tools to guide unsophisticated users in constructing complex computational pathways that result in sophisticated simulations and thus more accurate seismic hazard analysis.

C.1.c. The Need for a Large, Integrated Project

SCEC is a consortium of 40 universities and research organizations funded by the NSF, USGS, and other government agencies to (1) gather new information about earthquakes in Southern California, (2) integrate the available knowledge into a comprehensive and predictive understanding of earthquake phenomena, and (3) communicate this understanding to engineers, emergency managers, government officials, and the general public. Recent estimates by the Federal Emergency Management Agency (FEMA) ascribe nearly half of the national earthquake risk to Southern California, with one-quarter concentrated in Los Angeles county alone [8]. SCEC thus serves a high-risk population of more than 20 million people as its regional center for earthquake information and coordinated earthquake studies. This coordination is essential to the development of the comprehensive data sets, consensus models, and consistent scientific judgements needed for public policy in earthquake risk management and mitigation.

Southern California is a superb natural laboratory for understanding the fundamentals of earthquake processes, endowed with many active faults and diverse tectonic regimes astride the rapidly deforming Pacific-North America plate boundary. Data on earthquakes in this part of the world are outstanding, and the integration of this information into a comprehensive and predictive understanding of earthquake behavior requires the resources of a multidisciplinary consortium capable of system-level research. SCEC coordinates its activities across a scientific community that includes over 150 professional scientists, as well as many postdocs and graduate students. In addition to a national distribution of research universities, the organizations participating in SCEC include the U. S. Geological Survey (USGS) and the California Division of Mines and Geology (CDMG), which have statutory responsibilities for characterizing earthquake hazards on the national and state level, respectively.

We have formed a SCEC/IT Partnership to develop an advanced information infrastructure for system-level earthquake science in Southern California. Our partnership comprises SCEC, USC's Information Sciences Institute (ISI), the San Diego Supercomputer Center (SDSC), the Incorporated Institutions for Research in Seismology (IRIS, a 97-institution consortium), and the U.S. Geological Survey. The funding requested in this proposal will support four project elements:

• Fundamental IT research by ISI and SDSC on how to integrate knowledge representation and reasoning, knowledge acquisition, digital libraries and information management, and Grid resource sharing into a methodology that will support the SCEC Community Modeling Environment.
• Application of this new methodology by SCEC, USGS, and IRIS scientists to SHA for the purpose of reducing earthquake losses in Southern California.
• Transfer of the methodology to other regions and extension to other Earth science problems by IRIS and (at no cost to NSF) the USGS.
• Use of products from the SCEC Community Modeling Environment to educate students at all levels and inform the general public about earthquake hazards.

The scope of this effort clearly requires a large, integrated project involving an extended collaboration among many disciplines and research organizations that spans both earthquake and computer sciences.

Section C.2: Geoscience Approach and Research

Proposal: Table of Contents

Back to ITR Proposal

Back to the CME Home Page

Phone 213/740-5843

Fax 213/740-0011

e-mail: SCECinfo@usc.edu