Science is the process by which we observe the world, infer general principles that systematize those observations, and then deduce the observable consequences of those principles. These activities involve the collection and analysis of information.
Computational science is an emerging discipline and, as such, one whose definition is itself evolving. The traditional definition of computational science is that it undertakes an investigative approach to the understanding of systems (physical, biological, mechanical, etc.) through the use of mathematical models that are solved on high performance computers. Computational science grew out of necessity; the development of the modern computer was driven, in large part, by the need to solve complex equations in science and engineering. Today computational science includes not only simulation and numerical modeling, but also the powerful new ways in which data generated from experiments and observation can be manipulated and probed to gain scientific insight. These activities are often referred to as data mining and analysis. The broadest definition of computational science would encompass the development and application of problem-solving methodologies that rely on advanced computational resources, including processor, storage, network, and display technologies.
It is helpful to state what computational science is not. First, it is not computer science. Computer science is a specific discipline: the science of the computer itself. Computer science is often involved in important and ground-breaking computational science research, but there no requirement that all computational science directly involve computer science. Second, all uses of a computer in service of science are not necessarily computational science. The computer has undoubtedly become an indispensable tool in such activities as scholarly writing, literature research, production and manipulation of graphics and images, and experimental control and data acquisition. These activities are not normally regarded as computational science, even though they may require a significant amount of technical expertise. Third, computational science is not confined to supercomputing, i.e., the use of the most advanced and largest processors.
The purpose of computational science is the science, and its procedures follow the usual course of scientific inquiry. Research begins with a question, a conjecture, or a hypothesis to be tested and ends with a moment of discovery that constitutes at least a partial answer. The intermediate steps generally include framing the question in terms of a model; this most often means a determination of the equations governing the model. The scientist then looks for previously unrealized implications of this model. In computational science the governing equations must be put in terms amenable to solution by a numerical algorithm. This algorithm must then be converted to a machine-executable form through programming. This application must then be placed on a suitable hardware platform and, ideally, implemented in an appropriate and efficient manner for this platform. After testing and validation, the resulting program is used to execute a series of numerical experiments that have the potential to probe the question. The outcomes of these experiments are often themselves rather complex, requiring a comparable computational effort to analyze. Finally, out of this analysis one aims for new insights into the system under study. Computational science can be so effective in facilitating the exploration of data, that the outcomes often close the loop by assisting in the formulation of new questions for further research.
Many of these research steps are familiar; they are the same steps taken in any theoretical or experimental science. But computational science requires additional procedures and new areas of expertise. Discipline-trained scientists will not necessarily possess all of the requisite knowledge and, given the pace at which the technology evolves, may find it difficult even to maintain an appropriate level of expertise. Continual advanced training and research collaborations are essential.