Basic Earth Science and Society - Basic Research Opportunities in Earth Science - page 23

ting their behaviors, can be of immense practical value; thus, achieving a
predictive understanding has become a major research goal. Because most
geosystems involve many components that interact nonlinearly over a wide
spectrum of spatial and temporal scales, the behaviors they display are not
amenable to classical theoretical analysis and manual calculation. Indeed, because
of the rapidly expanding capabilities for observing terrestrial activity, the data
volumes on geosystems will soon be measured in petabytes. The interpretation of
such vast quantities of data lies beyond the expertise and ability of the lone
scientist, requiring collaborations among large groups of investigators from a
variety of disciplines. The primary integrative mechanism for this
multidisciplinary activity is the system-level model.
Only in the past few years have computational capabilities permitted
numerical simulations of an interesting spectrum of geosystem behaviors in three
spatial dimensions. For instance, although the first attempts to model solid-state
convection in the Earth's mantle date from the early 1970s, the numerical
resolution required to represent mantle convection properly in three dimensions
was attained only in the 1990s. The first simulation of a self-sustaining core
dynamo based on a realistic set of governing equations was not achieved until
1995. For some problems, such as simulations of active fault systems, full three-
dimensional calculations over the appropriate scale range exceed the capabilities
of even the largest available computers.
Continuing progress in geosystem modeling will depend heavily on
improvements to the computational infrastructure of Earth science, including
computational algorithms for exploiting parallel computers and other hardware,
access to distributed computing and collaborative environments, advanced
methods for code development and sharing, software libraries, visualization
tools, and data management capabilities. The need for community models that can
function as “virtual laboratories” for the study of particular geosystems presents a
major challenge because new organizational structures will have to be set up to
develop, verify, and maintain the requisite software components. The strategies
and tools for this type of collaborative research are being developed by computer
scientists in partnership with other research communities, and Earth scientists can
learn and profit from participating in these efforts.
Most geosystems are so complex that the ability to extrapolate the observed
behaviors into new regimes and confirm them with additional data becomes an
essential measure of how well a system is understood. Predictions made from
system-level models thus play an integral role in an iterated cycle of data
gathering and analysis, hypothesis testing, and model improvement. The reliance
on this type of model-based empiricism has significant implications for the
organizational structure of geoscience, in addition to its epistemology, because it
offers a framework for integrating