the investigations are well directed, and the data properly analyzed and archived, then the returns on previous research investments can be compounded as more data are collected. Each observational study within the natural laboratory adds to the database, improving the context for future work. Natural laboratories are well suited to field operations that are logistically complicated and expensive, as in the collection of spatially dense data sets and the monitoring of phenomena over extended time intervals. The committee endorses a recommendation in the recent National Research Council report Basic Research Opportunities in Earth Science, to establish an Earth Science National Laboratory Program within the NSF (2). Many types of earthquake natural laboratories would be able to deliver new data on fault-zone processes. Pore-fluid pressurization in deep boreholes could be used to induce seismic events, which could be recorded by borehole seismometers to characterize nucleation and the early stages of rupture dynamics (a modern-day update of the Rangely oil field experiments conducted in the 1960s; see Chapter 2). Data could be collected from seismic networks in deep mines where small earthquakes are rapidly, and to some extent controllably, generated by the advance of mine faces and other underground developments.
Synoptic studies in natural laboratories could furnish an important observational base for developing theoretical and numerical models of fault systems, and they could yield the essential data by which these models are ultimately validated. Fault zones are complex structures at all scales, and numerical simulations are required to integrate laboratory observations, field data on fault-zone structure and composition, and pore-fluid interactions to obtain constitutive representations of macroscale faulting. Simulations of fault-zone properties must account for many nonlinear phenomena, such as thermal transients during large earthquakes, that may directly alter fault strength (or even result in melting) and may couple with pore-pressure effects. Rigorous physical modeling begins with testable microscale processes and carries out the appropriate analyses that scale up through the geometric complexities of fault networks to understand the implications for natural events. Dynamical simulations at various scales will be needed to assess the discrepancies among laboratory-based friction laws, observed fault-system behaviors (e.g., earthquake productivity, postseismic response), and seismological data on large earthquakes (e.g., fracture energies, particle velocities and accelerations).
Better knowledge of earthquake source physics on the short time scales of fault rupture will improve the understanding of how strong ground motions are generated, as well as the processes that lead up to the fault