An Overview of Computational Science
Definition of Computational Science. Computational
Science is the field of study that integrates natural science, computer
science, and applied mathematics. The
problems typically come from one of the natural sciences, the (simulation
and optimization) models and solution algorithms are usually mathematical
in nature, and the algorithm implementation requires computer science
knowledge for accurate, efficient, and reliable results.
Computational Science is an interdisciplinary subject within
an emerging curriculum (it is not just computer science).
There is no doubt of its importance.
Computational Science is an interdisciplinary subject, which has proved to be an effective way to generate new knowledge. It is increasingly being used by natural and social scientists, engineers, mathematicians, and computer scientists, often with high performance computing tools, to produce new scientific results.
The employment of the computer in research and development to simulate physical events and process large quantities of data is now established as a legitimate third paradigm of scientific methodology, along with the traditional use of scientific theory and observation or experiment, for understanding the physical world.
Some large universities have graduate Computational Science programs. A few undergraduate institutions now have programs as well. This discipline is now even working its way down to the K-12 levels.
Reasons for Studying Computational Science. Computational science is now the “glue” that binds experiment to theory. The experimentalist and the computational scientist must work closely and interactively together to perfect the simulation model that best describes the current theory. This is an ongoing learning process for all parties. Frequently an interdisciplinary team is needed.
Currently in industry, unlike research institutions, there is frequently only time to simulate, and not enough time to actually experiment. Many important decisions at all levels are made based upon simulation and optimization models. These models are now being used in a more quantitative way than before. Numerical simulation is also completely safe, unlike dangerous but necessary experiments such as weapons testing. It can be much more economical than performing an expensive investigation. Some activities can only be simulated. In fact, the greatest percentage of the computational science software is written in legacy FORTRAN or C with the newer software being written in C++ (for object-oriented code) or F90/F95 (which permits the implementation of parallel processing).
Many scientific problems are now so complex that it is no longer sufficient to just know how to use modeling software. The use of modeling software alone is not considered to be computational science. One must now know enough about programming, mathematics and the problem itself to be able to modify the source code in order to tailor it to solve the physical problem at hand.
One must also know both numerical algorithms (continuous and discrete) and the limitations of floating-point arithmetic as well as algorithms to effectively manipulate data structures. This is necessary for the increasingly complex problems. This is also where the computer scientist can make the most significant contribution. Many of these are of order of magnitude O(n7) where n is the number of variables, so it is crucial to be able to reduce this order to something more computationally tractable (e.g., reducing a job that could take 4 years to run to one that could run in 6 hours). Computational scientists normally have this type of training. For example, a computational scientist would never use a matrix inverter to solve a system of linear equations because of wasted computation time and the fact that the additional roundoff errors could corrupt the solution.
Much more accurate and plentiful data is now obtained from a variety of sensors (including satellites) in all scientific disciplines. Much of this data is being placed in web-accessible repositories. Models are needed to help interpret this data, and this may now be done through remote computing.
Activities of Computational Science. The activities involved in Computational Science may be conveniently placed in three categories:
Filtering and Enhancing Field and Lab Data
Integrating Sensor Data
Accessing Worldwide Scientific Databases
Employing Appropriate Solution Methods
Understanding Algorithm Parameter Selection
Utilizing Software from Research and Industry
Selecting Symbolic and Numerical Software Tools
Visualizing Physical Representations
Interpreting Computational Results
Problems Solved by Computational Science. The methods of Computational Science have been applied to a wide range of problems such as aeronautical design, environmental improvement, neuroscience, pharmaceutical design, and weather forecasting. More recently, high performance computation, traditionally used in physics and chemistry, has been applied to biology (e.g., bioinformatics), geology, environmental studies, and some social sciences. The development of Computational Science has had a profound effect on the way that science and engineering is done.
Due to the increasing complexity of real-world problems, it is absolutely essential that interdisciplinary collaboration be an integral part of the educational curriculum and in scientific research.
Computational scientists are needed in all branches of science, but the most significant need is for computational biologists. Many agree that “this is the century of the biologist.”
Wittenberg’s Computational Science Minor. Wittenberg’s Computational Science minor began in 2003. Three years prior to this, Wittenberg University’s Mathematics and Computer Science Department pioneered an introductory computational course for science students - Computational Models and Methods, COMP/MATH 260. Several components of this course were developed collaboratively with faculty members from each of the science departments. This course was very popular, because of its “applied” nature, and became a permanent part of the curriculum. This led to a proposal by the Wittenberg Science Department Chairs to create the new minor. At Wittenberg, this course forms the core of this new minor whose other courses come from selected courses in the disciplines of biology, chemistry, computer science, economics, geology, mathematics, and physics.
While some of the activities learned in a Computational Science program
are of the cutting-edge variety, by the time some of Wittenberg’s graduates
become science-oriented practitioners and researchers, many of the models
and methods they have learned will be commonplace.
This minor enables Wittenberg students to learn the essential concepts, principles, and skills of computational science, and to gain experience in the laboratory applying them successfully to problems from their own discipline. They will be better prepared to conduct increasingly computational, cutting edge research in graduate school or industry. The participating faculty, similarly, are becoming much better versed in applying computational science methodologies to their discipline, constituting a core of faculty with sufficient expertise to extend computational methods into other disciplines such as psychology, management, and economics.
