Introduction

The earth sciences differ from other scientific disciplines in their focus on processes one cannot easily repeat or control. Examples are the nucleation of earthquakes as brittle failure along faults, or the creeping flow processes in the Earth's interior driving plate tectonics and the geologic activity of our planet. The inherent experimental limitations and the indirect nature of our observations explain the need for sophisticated modeling approaches. And with continued growth of computer hardware performance, the crossing of some long-standing thresholds in capacity and capability computing is finally underway. For instance, it is now feasible to implement earth models having in excess of 1–10 billion grid points, a number that matters because it allows us to overcome in three-dimensional (3-D) models the disparate length scales characteristic of key geologic phenomena: an earthquake rupturing a fault segment over a distance of some 100 km while emanating seismic energy throughout the planet (10 000 km), or the peculiar nature of plate tectonics with deformation concentrated along narrow plate boundaries of 10–100 km width separated by plates of dimension 1 000–10 000 km. But capable hardware and raw compute power are not sufficient by themselves to advance demanding earth system simulations. Equally critical are sophisticated software, visualization tools, data portals, and shared middleware and software libraries, collectively known as cyberinfrastructure.

Cyberinfrastructure forms a comprehensive modeling environment to integrate computing hardware, data, networks, digitally enabled sensors, observatories, and experimental facilities.